mirror of
https://github.com/rui314/mold.git
synced 2026-11-09 04:49:33 -10:00
Compare commits
2
Commits
| Author | SHA1 | Date | |
|---|---|---|---|
 Rui Ueyama
|
e390a70f6c | ||
|
Rui Ueyama
|
d7412d60cc |
No files matched your search
@@ -1,13 +0,0 @@
|
||||
# These are supported funding model platforms
|
||||
|
||||
github: [rui314] # Replace with up to 4 GitHub Sponsors-enabled usernames e.g., [user1, user2]
|
||||
patreon: # Replace with a single Patreon username
|
||||
open_collective: # Replace with a single Open Collective username
|
||||
ko_fi: # Replace with a single Ko-fi username
|
||||
tidelift: # Replace with a single Tidelift platform-name/package-name e.g., npm/babel
|
||||
community_bridge: # Replace with a single Community Bridge project-name e.g., cloud-foundry
|
||||
liberapay: # Replace with a single Liberapay username
|
||||
issuehunt: # Replace with a single IssueHunt username
|
||||
otechie: # Replace with a single Otechie username
|
||||
lfx_crowdfunding: # Replace with a single LFX Crowdfunding project-name e.g., cloud-foundry
|
||||
custom: # Replace with up to 4 custom sponsorship URLs e.g., ['link1', 'link2']
|
||||
@@ -0,0 +1,10 @@
|
||||
---
|
||||
name: Bug report
|
||||
about: Create a report to help us improve
|
||||
title: ''
|
||||
labels: ''
|
||||
assignees: ''
|
||||
|
||||
---
|
||||
|
||||
Delete this message and describe the bug here. Note that we only fix bugs in the most recent release of mold, so please make sure that you are using the most recent version.
|
||||
@@ -1,93 +0,0 @@
|
||||
name: Build all tarballs
|
||||
|
||||
on:
|
||||
schedule:
|
||||
- cron: '0 0 * * *'
|
||||
workflow_dispatch:
|
||||
|
||||
jobs:
|
||||
check-changes:
|
||||
runs-on: ubuntu-24.04
|
||||
outputs:
|
||||
has_changes: ${{ github.event_name == 'workflow_dispatch' || steps.check.outputs.has_changes }}
|
||||
steps:
|
||||
- name: Check for new commits since last successful run
|
||||
id: check
|
||||
env:
|
||||
GH_TOKEN: ${{ github.token }}
|
||||
run: |
|
||||
last=$(gh run list -R ${{ github.repository }} -w build-all.yml -b main -s success -L1 --json createdAt -q '.[0].createdAt')
|
||||
if [ -z "$last" ] || [ "$last" = "null" ] || [ "$(gh api repos/${{ github.repository }}/commits?sha=main\&since=$last\&per_page=1 | jq length)" -gt 0 ]; then
|
||||
echo "has_changes=true" >> "$GITHUB_OUTPUT"
|
||||
fi
|
||||
|
||||
build-linux:
|
||||
needs: check-changes
|
||||
if: needs.check-changes.outputs.has_changes == 'true'
|
||||
strategy:
|
||||
matrix:
|
||||
include:
|
||||
- { target: x86_64, os: ubuntu-24.04 }
|
||||
- { target: aarch64, os: ubuntu-24.04-arm }
|
||||
- { target: arm, os: ubuntu-24.04-arm }
|
||||
- { target: riscv64, os: ubuntu-24.04 }
|
||||
- { target: ppc64le, os: ubuntu-24.04 }
|
||||
- { target: s390x, os: ubuntu-24.04 }
|
||||
- { target: loongarch64, os: ubuntu-24.04 }
|
||||
|
||||
runs-on: ${{ matrix.os }}
|
||||
|
||||
permissions:
|
||||
contents: read
|
||||
packages: write
|
||||
|
||||
steps:
|
||||
- name: Checkout Repository
|
||||
uses: actions/checkout@v6
|
||||
|
||||
- name: Install Podman
|
||||
run: sudo apt-get update && sudo apt-get install -y podman qemu-user-static
|
||||
|
||||
- name: Login to GitHub Container Registry
|
||||
uses: redhat-actions/podman-login@v1
|
||||
with:
|
||||
registry: ghcr.io
|
||||
username: ${{ github.repository_owner }}
|
||||
password: ${{ secrets.GITHUB_TOKEN }}
|
||||
|
||||
- name: Build a tarball
|
||||
run: ./dist.sh ${{ matrix.target }}
|
||||
|
||||
- name: Upload artifact
|
||||
uses: actions/upload-artifact@v7
|
||||
with:
|
||||
name: ${{ matrix.target }}
|
||||
path: dist/mold-*.tar.gz
|
||||
compression-level: 0
|
||||
|
||||
build-windows:
|
||||
needs: check-changes
|
||||
if: needs.check-changes.outputs.has_changes == 'true'
|
||||
runs-on: windows-latest
|
||||
steps:
|
||||
- name: Checkout Repository
|
||||
uses: actions/checkout@v6
|
||||
|
||||
- name: Build and Archive
|
||||
shell: pwsh
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake -T clangcl ..
|
||||
cmake --build . --config Release -j $Env:NUMBER_OF_PROCESSORS
|
||||
cmake --install . --config Release --prefix ../mold-install
|
||||
cd ..
|
||||
New-Item -ItemType Directory -Force dist | Out-Null
|
||||
$version = $Env:TAG -replace '^v', ''
|
||||
Compress-Archive -Path mold-install\* -DestinationPath dist\mold-$version-windows-x86_64.zip
|
||||
|
||||
- name: Upload artifact
|
||||
uses: actions/upload-artifact@v7
|
||||
with:
|
||||
name: win-x86_64
|
||||
path: dist/mold-*.*
|
||||
@@ -1,44 +0,0 @@
|
||||
name: Build native tarballs
|
||||
|
||||
on:
|
||||
push:
|
||||
workflow_dispatch:
|
||||
|
||||
jobs:
|
||||
build-tarballs:
|
||||
strategy:
|
||||
matrix:
|
||||
include:
|
||||
- { target: x86_64, os: ubuntu-24.04 }
|
||||
- { target: aarch64, os: ubuntu-24.04-arm }
|
||||
- { target: arm, os: ubuntu-24.04-arm }
|
||||
|
||||
runs-on: ${{ matrix.os }}
|
||||
|
||||
permissions:
|
||||
contents: read
|
||||
packages: write
|
||||
|
||||
steps:
|
||||
- name: Checkout Repository
|
||||
uses: actions/checkout@v6
|
||||
|
||||
- name: Install Podman
|
||||
run: sudo apt-get update && sudo apt-get install -y podman qemu-user-static
|
||||
|
||||
- name: Login to GitHub Container Registry
|
||||
uses: redhat-actions/podman-login@v1
|
||||
with:
|
||||
registry: ghcr.io
|
||||
username: ${{ github.repository_owner }}
|
||||
password: ${{ secrets.GITHUB_TOKEN }}
|
||||
|
||||
- name: Build a tarball
|
||||
run: ./dist.sh ${{ matrix.target }}
|
||||
|
||||
- name: Upload artifact
|
||||
uses: actions/upload-artifact@v7
|
||||
with:
|
||||
name: ${{ matrix.target }}
|
||||
path: dist/mold-*.tar.gz
|
||||
compression-level: 0
|
||||
+29
-161
@@ -1,172 +1,40 @@
|
||||
name: CI
|
||||
on:
|
||||
push:
|
||||
branches: [ main ]
|
||||
pull_request:
|
||||
env:
|
||||
UBSAN_OPTIONS: print_stacktrace=1:halt_on_error=1
|
||||
branches: [ main ]
|
||||
jobs:
|
||||
build-asan-tsan:
|
||||
strategy:
|
||||
matrix:
|
||||
target:
|
||||
- ''
|
||||
- '-DMOLD_USE_ASAN=On'
|
||||
- '-DMOLD_USE_TSAN=On'
|
||||
runs-on: ubuntu-24.04
|
||||
timeout-minutes: 60
|
||||
build-clang:
|
||||
runs-on: ubuntu-20.04
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- uses: rui314/setup-mold@staging
|
||||
- run: sudo ./install-build-deps.sh
|
||||
- name: build
|
||||
- uses: actions/checkout@v2
|
||||
- name: apt-get
|
||||
run: sudo dpkg --add-architecture i386 && sudo apt-get update && sudo apt-get install build-essential libstdc++-10-dev cmake clang libssl-dev zlib1g-dev libxxhash-dev libtbb-dev git bsdmainutils dwarfdump libc6-dev:i386 lib32gcc-10-dev libstdc++-10-dev-arm64-cross gcc-aarch64-linux-gnu g++-aarch64-linux-gnu qemu-user
|
||||
- name: ccache
|
||||
uses: hendrikmuhs/ccache-action@v1
|
||||
- name: make
|
||||
run: |
|
||||
sudo apt-get install -y clang-18 clang gcc-multilib gdb dwarfdump zstd
|
||||
mkdir build
|
||||
cd build
|
||||
cmake -DCMAKE_BUILD_TYPE=Debug -DCMAKE_C_COMPILER=clang-18 -DCMAKE_CXX_COMPILER=clang++-18 ${{ matrix.target }} ..
|
||||
cmake --build . -j$(nproc)
|
||||
- run: cd build; ctest --output-on-failure -j$(nproc)
|
||||
- name: archive test results
|
||||
uses: actions/upload-artifact@v7
|
||||
if: failure()
|
||||
with:
|
||||
name: test-results-clang
|
||||
path: |
|
||||
build
|
||||
!build/CMakeFiles
|
||||
|
||||
build-msan:
|
||||
runs-on: ubuntu-24.04
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- uses: redhat-actions/podman-login@v1
|
||||
with:
|
||||
registry: ghcr.io
|
||||
username: ${{ github.repository_owner }}
|
||||
password: ${{ secrets.GITHUB_TOKEN }}
|
||||
- run: .github/workflows/run-msan.sh
|
||||
|
||||
build-multi-archs:
|
||||
runs-on: ubuntu-24.04
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- name: install-build-deps
|
||||
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
|
||||
make -j$(nproc)
|
||||
- name: make test
|
||||
run: |
|
||||
sudo ./install-build-deps.sh
|
||||
sudo ./install-cross-tools.sh
|
||||
sudo .github/workflows/install-extras.sh
|
||||
- name: build
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake ..
|
||||
cmake --build . -j$(nproc)
|
||||
- run: cd build; ctest --output-on-failure -j$(nproc)
|
||||
- name: archive test results
|
||||
uses: actions/upload-artifact@v7
|
||||
if: failure()
|
||||
with:
|
||||
name: test-results-multi-archs
|
||||
path: |
|
||||
build
|
||||
!build/CMakeFiles
|
||||
|
||||
build-distros:
|
||||
strategy:
|
||||
matrix:
|
||||
distro:
|
||||
- alpine
|
||||
- archlinux
|
||||
- debian:11 # GCC 10 and CMake 3.18 - the minimum supported by mold
|
||||
- fedora
|
||||
- gentoo/stage3
|
||||
- opensuse/tumbleweed
|
||||
- ubuntu:22.04
|
||||
- ubuntu:25.04
|
||||
runs-on: ubuntu-latest
|
||||
timeout-minutes: 60
|
||||
container: ${{ matrix.distro }}
|
||||
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
|
||||
make test -j$(nproc)
|
||||
build-gcc11:
|
||||
runs-on: ubuntu-20.04
|
||||
container: gcc:11.1.0
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- run: ./install-build-deps.sh
|
||||
- name: build
|
||||
- uses: actions/checkout@v2
|
||||
- name: apt-get
|
||||
run: dpkg --add-architecture i386 && apt-get update && apt-get install -y sudo libssl-dev zlib1g-dev libxxhash-dev cmake clang dwarfdump bsdextrautils libgcc-9-dev gcc-aarch64-linux-gnu g++-aarch64-linux-gnu libc6-dev:i386 lib32gcc-10-dev qemu-user
|
||||
- name: ccache
|
||||
uses: hendrikmuhs/ccache-action@v1
|
||||
- name: make
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake ..
|
||||
cmake --build . -j$(nproc)
|
||||
- run: cd build; ctest --output-on-failure -j$(nproc)
|
||||
- name: sanitize distro name
|
||||
if: failure()
|
||||
id: sanitize
|
||||
run: echo "name=$(echo '${{ matrix.distro }}' | tr '/:' '--')" >> "$GITHUB_OUTPUT"
|
||||
- name: archive test results
|
||||
uses: actions/upload-artifact@v7
|
||||
if: failure()
|
||||
with:
|
||||
name: test-results-${{ steps.sanitize.outputs.name }}
|
||||
path: |
|
||||
build
|
||||
!build/CMakeFiles
|
||||
|
||||
build-macos:
|
||||
runs-on: macos-latest
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- name: build
|
||||
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
|
||||
make CC=gcc CXX=g++ -j$(nproc)
|
||||
- name: make test
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake ..
|
||||
cmake --build . -j$(sysctl -n hw.physicalcpu)
|
||||
|
||||
build-windows:
|
||||
runs-on: windows-latest
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- name: build
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake -T clangcl ..
|
||||
cmake --build . -j $Env:NUMBER_OF_PROCESSORS
|
||||
|
||||
build-msys:
|
||||
runs-on: windows-latest
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- name: Setup MSYS2
|
||||
uses: msys2/setup-msys2@v2
|
||||
with:
|
||||
msystem: UCRT64
|
||||
update: true
|
||||
pacboy: gcc-libs:p libwinpthread-git:p tbb:p zlib:p zstd:p dlfcn:p cc:p cmake:p ninja:p
|
||||
- name: build
|
||||
shell: msys2 {0}
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake -GNinja -DMOLD_USE_MIMALLOC=OFF -DMOLD_USE_SYSTEM_TBB=ON ..
|
||||
cmake --build . -j $(nproc)
|
||||
|
||||
build-freebsd:
|
||||
runs-on: ubuntu-latest
|
||||
timeout-minutes: 60
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
- name: Build and test
|
||||
uses: vmactions/freebsd-vm@v1
|
||||
with:
|
||||
usesh: true
|
||||
run: |
|
||||
./install-build-deps.sh
|
||||
mkdir build
|
||||
cd build
|
||||
cmake ..
|
||||
cmake --build . -j$(nproc)
|
||||
ctest --output-on-failure -j$(nproc)
|
||||
export PATH="/usr/lib/ccache:/usr/local/opt/ccache/libexec:$PATH"
|
||||
make test -j$(nproc)
|
||||
@@ -1,75 +0,0 @@
|
||||
#!/bin/bash -x
|
||||
|
||||
apt-get update
|
||||
apt-get install -y wget xz-utils
|
||||
|
||||
# Install a 32-bit RISC-V toolchain
|
||||
mkdir /rv32
|
||||
wget -O- --progress=dot:mega https://github.com/riscv-collab/riscv-gnu-toolchain/releases/download/2025.06.13/riscv32-glibc-ubuntu-24.04-gcc-nightly-2025.06.13-nightly.tar.xz | tar -C /rv32 --strip-components=1 --xz -xf -
|
||||
|
||||
ln -sf /rv32/sysroot /usr/riscv32-linux-gnu
|
||||
echo '/rv32/bin/riscv32-unknown-linux-gnu-gcc -L/usr/riscv32-linux-gnu "$@"' > /usr/bin/riscv32-linux-gnu-gcc
|
||||
echo '/rv32/bin/riscv32-unknown-linux-gnu-g++ -L/usr/riscv32-linux-gnu "$@"' > /usr/bin/riscv32-linux-gnu-g++
|
||||
chmod 755 /usr/bin/riscv32-linux-gnu-{gcc,g++}
|
||||
|
||||
for i in objdump objcopy strip; do
|
||||
ln -sf /rv32/bin/riscv32-unknown-linux-gnu-$i /usr/bin/riscv32-linux-gnu-$i
|
||||
done
|
||||
|
||||
# Install a LoongArch toolchain
|
||||
mkdir /larch
|
||||
wget -O- --progress=dot:mega https://github.com/loongson/build-tools/releases/download/2025.06.06/x86_64-cross-tools-loongarch64-binutils_2.44-gcc_15.1.0-glibc_2.41.tar.xz | tar -C /larch --strip-components=1 --xz -xf -
|
||||
|
||||
cp -r /larch/loongarch64-unknown-linux-gnu/lib/* /larch/target/lib64
|
||||
ln -sf /larch/target /usr/loongarch64-linux-gnu
|
||||
|
||||
for i in gcc g++ objdump objcopy strip; do
|
||||
ln -sf /larch/bin/loongarch64-unknown-linux-gnu-$i /usr/bin/loongarch64-linux-gnu-$i
|
||||
done
|
||||
|
||||
wget -O /usr/local/bin/qemu-loongarch64 --progress=dot:mega https://github.com/loongson/build-tools/releases/download/2025.06.06/qemu-loongarch64
|
||||
chmod 755 /usr/local/bin/qemu-loongarch64
|
||||
|
||||
# Install ARM32 big-endian toolchain
|
||||
mkdir /armeb
|
||||
wget -O- --progress=dot:mega https://toolchains.bootlin.com/downloads/releases/toolchains/armebv7-eabihf/tarballs/armebv7-eabihf--glibc--stable-2024.05-1.tar.xz | tar -C /armeb --strip-components=1 --xz -xf -
|
||||
|
||||
ln -sf /armeb/armeb-buildroot-linux-gnueabihf/sysroot /usr/armeb-linux-gnueabihf
|
||||
echo '/armeb/bin/armeb-linux-gcc -L/usr/armeb-linux-gnu "$@"' > /usr/bin/armeb-linux-gnueabihf-gcc
|
||||
echo '/armeb/bin/armeb-linux-g++ -L/usr/armeb-linux-gnu "$@"' > /usr/bin/armeb-linux-gnueabihf-g++
|
||||
chmod 755 /usr/bin/armeb-linux-gnueabihf-{gcc,g++}
|
||||
|
||||
for i in objdump objcopy strip; do
|
||||
ln -sf /armeb/bin/armeb-linux-$i /usr/bin/armeb-linux-gnueabihf-$i
|
||||
done
|
||||
|
||||
# Install ARM64 big-endian toolchain
|
||||
mkdir /aarch64be
|
||||
wget -O- --progress=dot:mega https://toolchains.bootlin.com/downloads/releases/toolchains/aarch64be/tarballs/aarch64be--glibc--stable-2024.05-1.tar.xz | tar -C /aarch64be --strip-components=1 --xz -xf -
|
||||
|
||||
ln -sf /aarch64be/aarch64_be-buildroot-linux-gnu/sysroot /usr/aarch64_be-linux-gnu
|
||||
echo '/aarch64be/bin/aarch64_be-linux-gcc -L/usr/aarch64_be-linux-gnu "$@"' > /usr/bin/aarch64_be-linux-gnu-gcc
|
||||
echo '/aarch64be/bin/aarch64_be-linux-g++ -L/usr/aarch64_be-linux-gnu "$@"' > /usr/bin/aarch64_be-linux-gnu-g++
|
||||
chmod 755 /usr/bin/aarch64_be-linux-gnu-{gcc,g++}
|
||||
|
||||
for i in objdump objcopy strip; do
|
||||
ln -sf /aarch64be/bin/aarch64_be-linux-$i /usr/bin/aarch64_be-linux-gnu-$i
|
||||
done
|
||||
|
||||
# Install SH4 big-endian toolchain
|
||||
mkdir /sh4aeb
|
||||
wget -O- --progress=dot:mega https://toolchains.bootlin.com/downloads/releases/toolchains/sh-sh4aeb/tarballs/sh-sh4aeb--glibc--stable-2024.05-1.tar.xz | tar -C /sh4aeb --strip-components=1 --xz -xf -
|
||||
|
||||
ln -sf /sh4aeb/sh4aeb-buildroot-linux-gnu/sysroot /usr/sh4aeb-linux-gnu
|
||||
echo '/sh4aeb/bin/sh4aeb-linux-gcc -L/usr/sh4aeb-linux-gnu "$@"' > /usr/bin/sh4aeb-linux-gnu-gcc
|
||||
echo '/sh4aeb/bin/sh4aeb-linux-g++ -L/usr/sh4aeb-linux-gnu "$@"' > /usr/bin/sh4aeb-linux-gnu-g++
|
||||
chmod 755 /usr/bin/sh4aeb-linux-gnu-{gcc,g++}
|
||||
|
||||
for i in objdump objcopy strip; do
|
||||
ln -sf /sh4aeb/bin/sh4aeb-linux-$i /usr/bin/sh4aeb-linux-gnu-$i
|
||||
done
|
||||
|
||||
# Install Intel SDE CPU emulator for CET-related tests
|
||||
mkdir /sde
|
||||
wget -O- --progress=dot:mega https://downloadmirror.intel.com/850782/sde-external-9.53.0-2025-03-16-lin.tar.xz | tar -C /sde --strip-components=1 --xz -xf -
|
||||
ln -s /sde/sde64 /usr/bin
|
||||
@@ -1,104 +0,0 @@
|
||||
name: Build & attach tarballs on tag push
|
||||
|
||||
on:
|
||||
push:
|
||||
tags: ['v[0-9]*']
|
||||
|
||||
permissions:
|
||||
contents: write
|
||||
packages: write
|
||||
id-token: write
|
||||
|
||||
env:
|
||||
TAG: ${{ github.ref_name }}
|
||||
|
||||
jobs:
|
||||
build-linux:
|
||||
strategy:
|
||||
matrix:
|
||||
include:
|
||||
- { target: x86_64, os: ubuntu-24.04 }
|
||||
- { target: aarch64, os: ubuntu-24.04-arm }
|
||||
- { target: arm, os: ubuntu-24.04-arm }
|
||||
- { target: riscv64, os: ubuntu-24.04 }
|
||||
- { target: ppc64le, os: ubuntu-24.04 }
|
||||
- { target: s390x, os: ubuntu-24.04 }
|
||||
- { target: loongarch64, os: ubuntu-24.04 }
|
||||
|
||||
runs-on: ${{ matrix.os }}
|
||||
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
|
||||
- name: Install Podman
|
||||
run: sudo apt-get update && sudo apt-get install -y podman qemu-user-static
|
||||
|
||||
- name: Login to GitHub Container Registry
|
||||
uses: redhat-actions/podman-login@v1
|
||||
with:
|
||||
registry: ghcr.io
|
||||
username: ${{ github.repository_owner }}
|
||||
password: ${{ secrets.GITHUB_TOKEN }}
|
||||
|
||||
- run: ./dist.sh ${{ matrix.target }}
|
||||
|
||||
- uses: actions/upload-artifact@v7
|
||||
with:
|
||||
name: ${{ matrix.target }}
|
||||
path: dist/mold-*.tar.gz
|
||||
|
||||
build-windows:
|
||||
runs-on: windows-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
|
||||
- name: Build and Archive
|
||||
shell: pwsh
|
||||
run: |
|
||||
mkdir build
|
||||
cd build
|
||||
cmake -T clangcl ..
|
||||
cmake --build . --config Release -j $Env:NUMBER_OF_PROCESSORS
|
||||
cmake --install . --config Release --prefix ../mold-install
|
||||
cd ..
|
||||
New-Item -ItemType Directory -Force dist | Out-Null
|
||||
$version = $Env:TAG -replace '^v', ''
|
||||
Compress-Archive -Path mold-install\* -DestinationPath dist\mold-$version-x86_64-windows.zip
|
||||
|
||||
- uses: actions/upload-artifact@v7
|
||||
with:
|
||||
name: win-x86_64
|
||||
path: dist/mold-*.*
|
||||
|
||||
publish:
|
||||
needs: [build-linux, build-windows]
|
||||
runs-on: ubuntu-latest
|
||||
steps:
|
||||
- uses: actions/checkout@v6
|
||||
|
||||
- uses: actions/download-artifact@v8
|
||||
with:
|
||||
path: dist
|
||||
pattern: '*'
|
||||
merge-multiple: true
|
||||
|
||||
- name: Ensure release exists (create if missing)
|
||||
env:
|
||||
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
run: |
|
||||
set -e
|
||||
if gh release view "$TAG" >& /dev/null; then
|
||||
echo "Existing release found for $TAG"
|
||||
else
|
||||
echo "Creating draft release for $TAG"
|
||||
gh release create "$TAG" --draft \
|
||||
--title "$TAG" \
|
||||
--notes "Automated draft – fill the changelog later."
|
||||
fi
|
||||
|
||||
- name: Upload tarballs
|
||||
env:
|
||||
GH_TOKEN: ${{ secrets.GITHUB_TOKEN }}
|
||||
run: |
|
||||
gh release upload "$TAG" dist/* --clobber
|
||||
gh release view "$TAG" --json assets --jq '.assets[].name'
|
||||
@@ -1,37 +0,0 @@
|
||||
#!/bin/bash
|
||||
set -e -x
|
||||
cd "$(dirname $0)"/../..
|
||||
|
||||
if [ "$GITHUB_REPOSITORY" = '' ]; then
|
||||
image=mold-msan
|
||||
image_build="podman build -t $image -"
|
||||
else
|
||||
# If this script is running on GitHub Actions, we want to cache
|
||||
# the created container image in GitHub's container repostiory.
|
||||
image=ghcr.io/$GITHUB_REPOSITORY/mold-msan
|
||||
image_build="podman build -t $image --output=type=registry --layers --cache-to $image --cache-from $image -"
|
||||
fi
|
||||
|
||||
cat <<EOF | $image_build
|
||||
FROM mirror.gcr.io/library/ubuntu:24.04
|
||||
RUN apt-get update && \
|
||||
apt-get install -y --no-install-recommends build-essential gcc g++ wget ca-certificates make cmake ninja-build python3 clang-20 llvm-20-dev libclang-20-dev libclang-rt-20-dev libtbb-dev mold && \
|
||||
rm -rf /var/lib/apt/lists
|
||||
RUN mkdir /llvm && \
|
||||
cd /llvm && \
|
||||
wget -O- --progress=dot:mega https://github.com/llvm/llvm-project/archive/refs/tags/llvmorg-21.1.1.tar.gz | tar xzf - --strip-components=1 && \
|
||||
cmake -S ./runtimes -B build -G Ninja -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_RUNTIMES='libcxx;libcxxabi;libunwind' -DCMAKE_C_COMPILER=clang-20 -DCMAKE_CXX_COMPILER=clang++-20 -DLLVM_USE_SANITIZER=MemoryWithOrigins && \
|
||||
cmake --build build -j\$(nproc) && \
|
||||
cmake --install build --prefix /msan && \
|
||||
rm -rf /llvm
|
||||
EOF
|
||||
|
||||
# Build mold and run its tests
|
||||
podman run -it --rm --pull=never -v "$(pwd):/mold:ro" --security-opt seccomp=unconfined $image bash -c "
|
||||
set -e
|
||||
mkdir /build
|
||||
cd /build
|
||||
cmake -GNinja -DCMAKE_BUILD_TYPE=Debug -DCMAKE_C_COMPILER=clang-20 -DCMAKE_CXX_COMPILER=clang++-20 -DMOLD_USE_MSAN=ON -DMOLD_STDLIB_PREFIX=/msan -DMOLD_USE_MIMALLOC=0 -DMOLD_USE_SYSTEM_TBB=1 -DMOLD_USE_MOLD=1 /mold
|
||||
cmake --build . -j\$(nproc)
|
||||
setarch -R ctest --output-on-failure -j\$(nproc)
|
||||
"
|
||||
@@ -1,33 +0,0 @@
|
||||
name: Update manpage
|
||||
|
||||
on:
|
||||
push:
|
||||
paths:
|
||||
- 'docs/mold.md'
|
||||
branches:
|
||||
- main
|
||||
workflow_dispatch:
|
||||
|
||||
jobs:
|
||||
update-manpage:
|
||||
runs-on: ubuntu-latest
|
||||
|
||||
steps:
|
||||
- name: Check out repository
|
||||
uses: actions/checkout@v6
|
||||
|
||||
- name: Install ronn
|
||||
run: sudo apt-get update && sudo apt-get install -y ronn
|
||||
|
||||
- name: Generate mold.1 from mold.md
|
||||
run: ronn --roff docs/mold.md
|
||||
|
||||
- name: Configure Git
|
||||
run: |
|
||||
git config --global user.name 'Rui Ueyama'
|
||||
git config --global user.email '[email protected]'
|
||||
|
||||
- name: Commit and push if mold.1 is updated
|
||||
run: |
|
||||
git add docs/mold.1
|
||||
git diff --staged --quiet || (git commit -m "Update mold.1 (automated commit)" && git push)
|
||||
+7
-6
@@ -1,11 +1,12 @@
|
||||
*.o
|
||||
*.so
|
||||
*~
|
||||
**/.DS_Store
|
||||
.cache
|
||||
core
|
||||
gentoo
|
||||
/mold-*.tar.gz
|
||||
/build*
|
||||
compile_commands.json
|
||||
/mold
|
||||
/dist
|
||||
/test/tmp
|
||||
options.inc
|
||||
core
|
||||
ld
|
||||
gentoo
|
||||
out
|
||||
-493
@@ -1,493 +0,0 @@
|
||||
# You can customize a build by specifying CMake options. An option may be
|
||||
# given in the -Dvariable=value form. For a boolean variable, `ON` or `1`
|
||||
# means true, while `OFF` or `0` means false.
|
||||
#
|
||||
# Here are a couple of common cmake options:
|
||||
#
|
||||
# -DCMAKE_C_COMPILER=<command-name>
|
||||
#
|
||||
# Specifies the C compiler name to use. The default value is `cc`.
|
||||
#
|
||||
# -DCMAKE_CXX_COMPILER=<command-name>
|
||||
#
|
||||
# Specifies the C++ compiler name to use. The default value is `c++`.
|
||||
#
|
||||
# -DCMAKE_INSTALL_PREFIX=<directory>
|
||||
#
|
||||
# Specifies the install target directory. The default value is `/usr/local`.
|
||||
#
|
||||
# -DCMAKE_BUILD_TYPE=[Debug | Release | RelWithDebInfo | MinSizeRel]
|
||||
#
|
||||
# Specifies the build type. The default is `Release`, which is the right
|
||||
# option unless you are debugging mold.
|
||||
#
|
||||
# An example of a cmake command line is shown below:
|
||||
#
|
||||
# $ cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_CXX_COMPILER=g++-12 ..
|
||||
#
|
||||
# where `..` refers to this directory.
|
||||
#
|
||||
# With cmake, you may run `cmake --install .` instead of `make install` to
|
||||
# install build artifacts to system directories. If you want to install
|
||||
# artifacts to a temporary target directory, run `cmake --install . --prefix
|
||||
# <dir>`.
|
||||
#
|
||||
# You can see the current cmake variables and their values by running
|
||||
# `cmake -N -L .` in a build directory.
|
||||
#
|
||||
# Note that in this file, we provide various dials and knobs to configure
|
||||
# how to build mold. However, as a policy, we do not provide a way to
|
||||
# enable/disable any individual mold's feature. In other words, we do not
|
||||
# provide options like `--enable-foo` or `--disable-foo`. The motivation
|
||||
# behind this is build reproducibility. We want to guarantee that all builds
|
||||
# of the mold linker of the same version will have the exact same set of
|
||||
# features and behave exactly the same.
|
||||
|
||||
cmake_minimum_required(VERSION 3.14)
|
||||
project(mold VERSION 2.41.0)
|
||||
|
||||
include(CMakeDependentOption)
|
||||
include(CheckSymbolExists)
|
||||
include(GNUInstallDirs)
|
||||
|
||||
add_executable(mold)
|
||||
target_compile_features(mold PRIVATE cxx_std_20)
|
||||
|
||||
if(MINGW)
|
||||
target_link_libraries(mold PRIVATE dl)
|
||||
else()
|
||||
target_link_libraries(mold PRIVATE ${CMAKE_DL_LIBS})
|
||||
endif()
|
||||
|
||||
# Build mold itself using mold if -DMOLD_USE_MOLD=ON
|
||||
option(MOLD_USE_MOLD "Use mold to build mold" OFF)
|
||||
if(MOLD_USE_MOLD)
|
||||
target_link_options(mold PRIVATE -fuse-ld=mold)
|
||||
|
||||
if(CMAKE_BUILD_TYPE MATCHES "Deb")
|
||||
target_link_options(mold PRIVATE -Wl,--gdb-index)
|
||||
endif()
|
||||
|
||||
if(CMAKE_BUILD_TYPE MATCHES "^Rel")
|
||||
target_link_options(mold PRIVATE -Wl,--gc-sections,--icf=safe)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
if(NOT "${CMAKE_CXX_COMPILER_FRONTEND_VARIANT}" STREQUAL "MSVC")
|
||||
target_compile_options(mold PRIVATE
|
||||
-fno-exceptions
|
||||
-fno-unwind-tables
|
||||
-fno-asynchronous-unwind-tables
|
||||
-ffunction-sections
|
||||
-fdata-sections
|
||||
-Wall
|
||||
-Wextra
|
||||
-Wno-sign-compare
|
||||
-Wno-unused-function
|
||||
-Wno-unused-parameter
|
||||
-Wno-missing-field-initializers
|
||||
-ggnu-pubnames)
|
||||
endif()
|
||||
|
||||
if(CMAKE_BUILD_TYPE STREQUAL "Debug")
|
||||
target_compile_options(mold PRIVATE -D_GLIBCXX_ASSERTIONS)
|
||||
endif()
|
||||
|
||||
if(CMAKE_SYSTEM_NAME STREQUAL "OpenBSD")
|
||||
set(OPENBSD ON)
|
||||
endif()
|
||||
|
||||
# Build mold with -flto if -DMOLD_LTO=ON
|
||||
option(MOLD_LTO "Build mold with link-time optimization enabled")
|
||||
if(MOLD_LTO)
|
||||
set_property(TARGET mold PROPERTY INTERPROCEDURAL_OPTIMIZATION ON)
|
||||
endif()
|
||||
|
||||
# Enable AddressSanitizer if -DMOLD_USE_ASAN=ON
|
||||
option(MOLD_USE_ASAN "Build mold with AddressSanitizer" OFF)
|
||||
if(MOLD_USE_ASAN)
|
||||
target_compile_options(mold PRIVATE -fsanitize=address -fsanitize=undefined)
|
||||
target_link_options(mold PRIVATE -fsanitize=address -fsanitize=undefined)
|
||||
endif()
|
||||
|
||||
# Enabled ThreadSanitizer if -DMOLD_USE_TSAN=ON
|
||||
option(MOLD_USE_TSAN "Build mold with ThreadSanitizer" OFF)
|
||||
if(MOLD_USE_TSAN)
|
||||
target_compile_options(mold PRIVATE -fsanitize=thread)
|
||||
target_link_options(mold PRIVATE -fsanitize=thread)
|
||||
endif()
|
||||
|
||||
# Enable MemorySanitizer if -DMOLD_USE_MSAN=ON
|
||||
option(MOLD_USE_MSAN "Build mold with MemorySanitizer" OFF)
|
||||
if(MOLD_USE_MSAN)
|
||||
set(MOLD_STDLIB_PREFIX /usr CACHE FILEPATH "prefix of instrumented stdlib")
|
||||
target_compile_options(mold PRIVATE
|
||||
-g -Og -fsanitize=memory -fsanitize-memory-track-origins
|
||||
-nostdinc++ -isystem ${MOLD_STDLIB_PREFIX}/include/c++/v1
|
||||
-DENABLE_MSAN_UNPOISON
|
||||
)
|
||||
target_link_options(mold PRIVATE
|
||||
-fsanitize=memory
|
||||
-nostdlib++ -L ${MOLD_STDLIB_PREFIX}/lib -lc++
|
||||
-Wl,-rpath,${MOLD_STDLIB_PREFIX}/lib
|
||||
)
|
||||
endif()
|
||||
|
||||
# Statically-link libstdc++ if -DMOLD_MOSTLY_STATIC=ON.
|
||||
#
|
||||
# This option is intended to be used by `./dist.sh` script to create a
|
||||
# mold binary that works on various Linux distros. You probably don't
|
||||
# need nor want to set this to ON.
|
||||
option(MOLD_MOSTLY_STATIC "Statically link libstdc++ and some other libraries" OFF)
|
||||
if(MOLD_MOSTLY_STATIC)
|
||||
target_link_options(mold PRIVATE -static-libstdc++)
|
||||
endif()
|
||||
|
||||
# Find zlib. If libz.so is not found, we compile a bundled one and
|
||||
# statically-link it to mold.
|
||||
find_package(ZLIB QUIET)
|
||||
if(ZLIB_FOUND AND NOT MOLD_MOSTLY_STATIC)
|
||||
target_link_libraries(mold PRIVATE ZLIB::ZLIB)
|
||||
else()
|
||||
set(ZLIB_BUILD_EXAMPLES OFF CACHE INTERNAL "")
|
||||
add_subdirectory(third-party/zlib EXCLUDE_FROM_ALL)
|
||||
target_include_directories(zlibstatic INTERFACE third-party/zlib
|
||||
$<TARGET_PROPERTY:zlibstatic,BINARY_DIR>)
|
||||
target_link_libraries(mold PRIVATE zlibstatic)
|
||||
endif()
|
||||
|
||||
# Find BLAKE3 cryptographic hash library. Just like zlib, if libblkae3.so
|
||||
# is not found, we compile a bundled one and statically-link it to mold.
|
||||
find_package(BLAKE3 QUIET)
|
||||
if(BLAKE3_FOUND AND NOT MOLD_MOSTLY_STATIC)
|
||||
target_link_libraries(mold PRIVATE BLAKE3::blake3)
|
||||
else()
|
||||
function(mold_add_blake3)
|
||||
set(BUILD_SHARED_LIBS OFF)
|
||||
add_subdirectory(third-party/blake3/c EXCLUDE_FROM_ALL)
|
||||
target_link_libraries(mold PRIVATE blake3)
|
||||
target_include_directories(mold PUBLIC third-party/blake3/c)
|
||||
endfunction()
|
||||
|
||||
mold_add_blake3()
|
||||
endif()
|
||||
|
||||
# Find zstd compression library. If zstd.h is not found, we compile a
|
||||
# bundled one and statically-link it to mold.
|
||||
include(CheckIncludeFile)
|
||||
check_include_file(zstd.h HAVE_ZSTD_H)
|
||||
|
||||
if(HAVE_ZSTD_H AND NOT MOLD_MOSTLY_STATIC)
|
||||
target_link_libraries(mold PRIVATE zstd)
|
||||
else()
|
||||
set(ZSTD_BUILD_PROGRAMS OFF)
|
||||
set(ZSTD_BUILD_CONTRIB OFF)
|
||||
set(ZSTD_BUILD_TESTS OFF)
|
||||
set(ZSTD_MULTITHREAD_SUPPORT OFF)
|
||||
set(ZSTD_BUILD_SHARED OFF)
|
||||
set(ZSTD_BUILD_STATIC ON)
|
||||
add_subdirectory(third-party/zstd/build/cmake EXCLUDE_FROM_ALL)
|
||||
target_include_directories(mold PUBLIC third-party/zstd/lib)
|
||||
target_link_libraries(mold PRIVATE libzstd_static)
|
||||
endif()
|
||||
|
||||
# Find mimalloc. mimalloc is an alternative malloc implementation
|
||||
# optimized for multi-threaded applications.
|
||||
#
|
||||
# If you want to use the usual libc's malloc, pass -DMOLD_USE_MIMALLOC=OFF.
|
||||
#
|
||||
# We enable mimalloc by default for 64-bit targets. It doesn't seem to
|
||||
# be stable on 32-bit targets.
|
||||
cmake_dependent_option(
|
||||
MOLD_USE_MIMALLOC "Use mimalloc" ON
|
||||
"CMAKE_SIZEOF_VOID_P EQUAL 8; NOT APPLE; NOT ANDROID; NOT OPENBSD; NOT MOLD_USE_ASAN; NOT MOLD_USE_TSAN" OFF)
|
||||
|
||||
cmake_dependent_option(
|
||||
MOLD_USE_SYSTEM_MIMALLOC "Use system or vendored mimalloc" OFF
|
||||
MOLD_USE_MIMALLOC OFF)
|
||||
|
||||
# By default, we build a bundled mimalloc and statically-link it to
|
||||
# mold. If you want to dynamically link to the system's
|
||||
# libmimalloc.so, pass -DMOLD_USE_SYSTEM_MIMALLOC=ON.
|
||||
if(MOLD_USE_MIMALLOC)
|
||||
if(MOLD_USE_SYSTEM_MIMALLOC)
|
||||
find_package(mimalloc REQUIRED)
|
||||
target_link_libraries(mold PRIVATE mimalloc)
|
||||
else()
|
||||
function(mold_add_mimalloc)
|
||||
set(MI_BUILD_STATIC ON CACHE INTERNAL "")
|
||||
set(MI_BUILD_TESTS OFF CACHE INTERNAL "")
|
||||
set(MI_NO_OPT_ARCH ON CACHE INTERNAL "")
|
||||
add_subdirectory(third-party/mimalloc EXCLUDE_FROM_ALL)
|
||||
target_compile_definitions(mimalloc-static PRIVATE MI_USE_ENVIRON=0)
|
||||
target_link_libraries(mold PRIVATE mimalloc-static)
|
||||
endfunction()
|
||||
|
||||
mold_add_mimalloc()
|
||||
endif()
|
||||
endif()
|
||||
|
||||
# Find TBB. TBB (OneTBB or Intel TBB) is a high-level threading library.
|
||||
# Use of this library is mandatory.
|
||||
#
|
||||
# By default, we build a bundled one and statically-link the library
|
||||
# to mold. If you want to link to the system's libtbb2.so, pass
|
||||
# -DMOLD_USE_SYSTEM_TBB=ON.
|
||||
option(MOLD_USE_SYSTEM_TBB "Use system or vendored TBB" OFF)
|
||||
if(MOLD_USE_SYSTEM_TBB OR BLAKE3_USE_TBB)
|
||||
find_package(TBB REQUIRED)
|
||||
target_link_libraries(mold PRIVATE TBB::tbb)
|
||||
else()
|
||||
function(mold_add_tbb)
|
||||
set(BUILD_SHARED_LIBS OFF)
|
||||
set(TBB_TEST OFF CACHE INTERNAL "")
|
||||
set(TBB_STRICT OFF CACHE INTERNAL "")
|
||||
add_subdirectory(third-party/tbb EXCLUDE_FROM_ALL)
|
||||
target_compile_definitions(tbb PRIVATE __TBB_DYNAMIC_LOAD_ENABLED=0)
|
||||
target_link_libraries(mold PRIVATE TBB::tbb)
|
||||
endfunction()
|
||||
|
||||
mold_add_tbb()
|
||||
endif()
|
||||
|
||||
# We always use Clang to build mold on Windows. MSVC can't compile mold.
|
||||
if(WIN32)
|
||||
if(MSVC AND NOT CMAKE_CXX_COMPILER_ID STREQUAL "Clang")
|
||||
message(FATAL_ERROR
|
||||
"Your compiler is not supported; install Clang from Visual Studio Installer and re-run cmake with '-T clangcl'")
|
||||
endif()
|
||||
|
||||
target_compile_definitions(mold PRIVATE NOGDI NOMINMAX)
|
||||
if(MINGW)
|
||||
target_compile_definitions(mold PRIVATE _WIN32_WINNT=0xA00)
|
||||
target_link_libraries(mold PRIVATE bcrypt)
|
||||
endif()
|
||||
else()
|
||||
include(CheckLibraryExists)
|
||||
check_library_exists(m pow "" LIBM_FOUND)
|
||||
if(LIBM_FOUND)
|
||||
target_link_libraries(mold PRIVATE m)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
# Build mold-wrapper.so
|
||||
if(NOT APPLE AND NOT WIN32)
|
||||
add_library(mold-wrapper SHARED)
|
||||
install(TARGETS mold-wrapper DESTINATION ${CMAKE_INSTALL_LIBDIR}/mold)
|
||||
|
||||
# Remove the default `lib` prefix
|
||||
set_target_properties(mold-wrapper PROPERTIES PREFIX "")
|
||||
target_link_libraries(mold-wrapper PRIVATE ${CMAKE_DL_LIBS})
|
||||
target_sources(mold-wrapper PRIVATE src/mold-wrapper.c)
|
||||
endif()
|
||||
|
||||
# If atomics doesn't work by default, add -latomic.
|
||||
# We need the flag on riscv, armv6 and m68k.
|
||||
include(CheckCXXSourceCompiles)
|
||||
check_cxx_source_compiles("#include <atomic>
|
||||
int main() {
|
||||
std::atomic_uint8_t a;
|
||||
std::atomic_uint16_t b;
|
||||
std::atomic_uint32_t c;
|
||||
std::atomic_uint64_t d;
|
||||
return ++a + ++b + ++c + ++d;
|
||||
}" HAVE_FULL_ATOMIC_SUPPORT)
|
||||
|
||||
if(NOT HAVE_FULL_ATOMIC_SUPPORT)
|
||||
target_link_libraries(mold PRIVATE atomic)
|
||||
endif()
|
||||
|
||||
# Add -pthread
|
||||
if(NOT APPLE AND NOT MSVC)
|
||||
target_compile_options(mold PRIVATE -pthread)
|
||||
target_link_options(mold PRIVATE -pthread)
|
||||
endif()
|
||||
|
||||
check_symbol_exists(madvise sys/mman.h HAVE_MADVISE)
|
||||
check_symbol_exists(uname sys/utsname.h HAVE_UNAME)
|
||||
|
||||
# Create a .cc file containing the current git hash for `mold --version`.
|
||||
add_custom_target(git_hash
|
||||
COMMAND ${CMAKE_COMMAND}
|
||||
-DSOURCE_DIR=${CMAKE_SOURCE_DIR}
|
||||
-DOUTPUT_FILE=${CMAKE_BINARY_DIR}/mold-git-hash.h
|
||||
-P ${CMAKE_SOURCE_DIR}/lib/update-git-hash.cmake
|
||||
DEPENDS lib/update-git-hash.cmake
|
||||
BYPRODUCTS mold-git-hash.h
|
||||
VERBATIM)
|
||||
|
||||
add_dependencies(mold git_hash)
|
||||
|
||||
# Almost all functions are template in mold which take a target type
|
||||
# (e.g. X86_64) as its type parameter. Since we suport more than 10
|
||||
# targets, compiling a single source file for all the targets is very
|
||||
# slow.
|
||||
#
|
||||
# As a workaround, we create a .cc file for each target and spawn many
|
||||
# compiler instances. This is hacky but greatly reduces compile time
|
||||
# on a multicore machine.
|
||||
#
|
||||
# You can build mold for a specific set of targets by passing, e.g.,
|
||||
# -DMOLD_TARGETS='X86_64;ARM64LE', though this is strongly discouraged
|
||||
# for build reproducibility. Use this option only if you build mold
|
||||
# frequently for your personal use; otherwise, always build mold with all
|
||||
# targets enabled. We provide this flag in the trust that you will not
|
||||
# abuse it, so please don’t betray that trust.
|
||||
set(MOLD_TARGETS
|
||||
X86_64 I386 ARM64LE ARM64BE ARM32LE ARM32BE RV32LE RV32BE RV64LE
|
||||
RV64BE PPC32 PPC64V1 PPC64V2 S390X SPARC64 M68K SH4LE SH4BE
|
||||
LOONGARCH32 LOONGARCH64
|
||||
CACHE STRING "List of supported targets")
|
||||
|
||||
list(GET MOLD_TARGETS 0 MOLD_FIRST_TARGET)
|
||||
|
||||
list(APPEND MOLD_TEMPLATE_FILES
|
||||
src/arch-arm32.cc
|
||||
src/arch-arm64.cc
|
||||
src/arch-i386.cc
|
||||
src/arch-loongarch.cc
|
||||
src/arch-m68k.cc
|
||||
src/arch-ppc32.cc
|
||||
src/arch-ppc64v1.cc
|
||||
src/arch-ppc64v2.cc
|
||||
src/arch-riscv.cc
|
||||
src/arch-s390x.cc
|
||||
src/arch-sh4.cc
|
||||
src/arch-sparc64.cc
|
||||
src/arch-x86-64.cc
|
||||
src/archive-file.cc
|
||||
src/cmdline.cc
|
||||
src/error.cc
|
||||
src/filetype.cc
|
||||
src/gc-sections.cc
|
||||
src/gdb-index.cc
|
||||
src/icf.cc
|
||||
src/input-files.cc
|
||||
src/input-sections.cc
|
||||
src/linker-script.cc
|
||||
src/main.cc
|
||||
src/mapfile.cc
|
||||
src/output-chunks.cc
|
||||
src/passes.cc
|
||||
src/relocatable.cc
|
||||
src/shrink-sections.cc
|
||||
src/thunks.cc
|
||||
src/tls.cc
|
||||
)
|
||||
|
||||
if(WIN32 AND NOT MINGW)
|
||||
list(APPEND MOLD_TEMPLATE_FILES src/lto-win32.cc)
|
||||
else()
|
||||
list(APPEND MOLD_TEMPLATE_FILES src/lto-unix.cc)
|
||||
endif()
|
||||
|
||||
if(WIN32)
|
||||
list(APPEND MOLD_TEMPLATE_FILES
|
||||
src/output-file-win32.cc
|
||||
src/subprocess-win32.cc
|
||||
)
|
||||
else()
|
||||
list(APPEND MOLD_TEMPLATE_FILES
|
||||
src/output-file-unix.cc
|
||||
src/subprocess-unix.cc
|
||||
)
|
||||
endif()
|
||||
|
||||
function(mold_instantiate_templates SOURCE TARGET)
|
||||
set(PATH ${CMAKE_BINARY_DIR}/${SOURCE}.${TARGET}.cc)
|
||||
if(NOT EXISTS ${PATH})
|
||||
file(WRITE ${PATH} "#define MOLD_${TARGET} 1
|
||||
#define MOLD_TARGET ${TARGET}
|
||||
#include \"${CMAKE_SOURCE_DIR}/${SOURCE}\"
|
||||
")
|
||||
endif()
|
||||
target_sources(mold PRIVATE ${PATH})
|
||||
endfunction()
|
||||
|
||||
foreach (SOURCE IN LISTS MOLD_TEMPLATE_FILES)
|
||||
foreach(TARGET IN LISTS MOLD_TARGETS)
|
||||
mold_instantiate_templates(${SOURCE} ${TARGET})
|
||||
set(HAVE_TARGET_${TARGET} 1)
|
||||
endforeach()
|
||||
endforeach()
|
||||
|
||||
# Add other non-template source files.
|
||||
target_sources(mold PRIVATE
|
||||
lib/aho-corasick.cc
|
||||
lib/compress.cc
|
||||
lib/crc32.cc
|
||||
lib/demangle.cc
|
||||
lib/filepath.cc
|
||||
lib/glob.cc
|
||||
lib/hyperloglog.cc
|
||||
lib/perf.cc
|
||||
lib/random.cc
|
||||
lib/tar.cc
|
||||
src/elf.cc
|
||||
src/entry.cc
|
||||
third-party/rust-demangle/rust-demangle.c
|
||||
)
|
||||
|
||||
if(WIN32)
|
||||
target_sources(mold PRIVATE
|
||||
src/jobs-win32.cc
|
||||
src/mapped-file-win32.cc
|
||||
src/signal-win32.cc
|
||||
)
|
||||
else()
|
||||
target_sources(mold PRIVATE
|
||||
src/jobs-unix.cc
|
||||
src/mapped-file-unix.cc
|
||||
src/signal-unix.cc
|
||||
)
|
||||
endif()
|
||||
|
||||
# Create config.h file
|
||||
configure_file(lib/config.h.in config.h)
|
||||
include_directories(${CMAKE_CURRENT_BINARY_DIR})
|
||||
|
||||
# Test configs
|
||||
include(CTest)
|
||||
|
||||
if(BUILD_TESTING)
|
||||
# Create the ld symlinks required for testing
|
||||
if(NOT WIN32)
|
||||
add_custom_command(
|
||||
TARGET mold POST_BUILD
|
||||
COMMAND ${CMAKE_COMMAND} -E create_symlink mold ld
|
||||
BYPRODUCTS ld
|
||||
WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
|
||||
VERBATIM)
|
||||
endif()
|
||||
|
||||
if(${UNIX})
|
||||
add_subdirectory(test)
|
||||
endif()
|
||||
endif()
|
||||
|
||||
if(NOT CMAKE_SKIP_INSTALL_RULES)
|
||||
install(TARGETS mold RUNTIME DESTINATION ${CMAKE_INSTALL_BINDIR})
|
||||
install(FILES docs/mold.1 DESTINATION ${CMAKE_INSTALL_MANDIR}/man1/)
|
||||
install(FILES LICENSE DESTINATION ${CMAKE_INSTALL_DOCDIR})
|
||||
|
||||
function(mold_install_relative_symlink OLD NEW)
|
||||
install(CODE "
|
||||
get_filename_component(PREFIX_ABS \${CMAKE_INSTALL_PREFIX}/ ABSOLUTE)
|
||||
get_filename_component(OLD_ABS ${OLD} ABSOLUTE BASE_DIR \${PREFIX_ABS})
|
||||
get_filename_component(NEW_ABS ${NEW} ABSOLUTE BASE_DIR \${PREFIX_ABS})
|
||||
get_filename_component(NEW_DIR \${NEW_ABS} DIRECTORY)
|
||||
file(RELATIVE_PATH OLD_REL \${NEW_DIR} \${OLD_ABS})
|
||||
message(STATUS \"Installing symlink: \$ENV{DESTDIR}\${NEW_ABS} -> \${OLD_REL}\")
|
||||
file(MAKE_DIRECTORY \$ENV{DESTDIR}\${NEW_DIR})
|
||||
file(CREATE_LINK \${OLD_REL} \$ENV{DESTDIR}\${NEW_ABS} SYMBOLIC)")
|
||||
endfunction()
|
||||
|
||||
if(NOT WIN32)
|
||||
mold_install_relative_symlink(${CMAKE_INSTALL_BINDIR}/mold${CMAKE_EXECUTABLE_SUFFIX}
|
||||
${CMAKE_INSTALL_LIBEXECDIR}/mold/ld${CMAKE_EXECUTABLE_SUFFIX})
|
||||
mold_install_relative_symlink(${CMAKE_INSTALL_BINDIR}/mold${CMAKE_EXECUTABLE_SUFFIX}
|
||||
${CMAKE_INSTALL_BINDIR}/ld.mold${CMAKE_EXECUTABLE_SUFFIX})
|
||||
mold_install_relative_symlink(${CMAKE_INSTALL_MANDIR}/man1/mold.1
|
||||
${CMAKE_INSTALL_MANDIR}/man1/ld.mold.1)
|
||||
endif()
|
||||
endif()
|
||||
@@ -0,0 +1,54 @@
|
||||
# Contributing to mold
|
||||
|
||||
Summary: When you send a patch, you must agree that the patch is
|
||||
released under the dual license of the GNU AGPLv3 and the MIT license.
|
||||
As an indication of agreement, please add a `Signed-off-by` line to
|
||||
your commit message using `git commit -s`.
|
||||
|
||||
Here is the long official version:
|
||||
|
||||
mold is currently licensed under the GNU AGPLv3, but I want to relicense
|
||||
it under a more liberal license in the future. To make this possible,
|
||||
I want all patches to be submitted under the dual license of the GNU
|
||||
AGPLv3 and the MIT license (the "Dual License") from the beginning.
|
||||
|
||||
To make it clear that contributors have agreed to release their
|
||||
patches under the Dual License, I introduced a sign-off procedure just
|
||||
like the Linux kernel project did. The sign-off is a simple line at
|
||||
the end of the git commit message, which certifies that you wrote it
|
||||
or otherwise have the right to pass it on as an open-source patch. The
|
||||
rules are pretty simple: if you can certify the below:
|
||||
|
||||
> ## Developer's Certificate of Origin
|
||||
>
|
||||
> By making a contribution to this project, I certify that:
|
||||
>
|
||||
> 1. The contribution was created in whole or in part by me and I
|
||||
> have the right to submit it under the Dual License; or
|
||||
>
|
||||
> 2. The contribution is based upon previous work that, to the best
|
||||
> of my knowledge, is covered under the the Dual License and I
|
||||
> have the right under the Dual License to submit that work with
|
||||
> modifications, whether created in whole or in part by me, under
|
||||
> the Dual License (unless I am permitted to submit under
|
||||
> different licenses), as indicated in the file; or
|
||||
>
|
||||
> 3. The contribution was provided directly to me by some other
|
||||
> person who certified (1), (2) or (3) and I have not modified it.
|
||||
>
|
||||
> 4. I understand and agree that this project and the contribution
|
||||
> are public and that a record of the contribution (including all
|
||||
> personal information I submit with it, including my sign-off) is
|
||||
> maintained indefinitely and may be redistributed consistent with
|
||||
> this project or the open source license(s) involved.
|
||||
|
||||
then you just add a line saying:
|
||||
|
||||
```
|
||||
Signed-off-by: Random J Developer <[email protected]>
|
||||
```
|
||||
|
||||
using your real name (sorry, no pseudonyms or anonymous contributions.)
|
||||
This will be done for you automatically if you use ``git commit -s``.
|
||||
Reverts should also include "Signed-off-by". ``git revert -s`` does that
|
||||
for you.
|
||||
@@ -1,21 +1,671 @@
|
||||
MIT License
|
||||
Note: I'm looking for a sponsor who wants to purchase the copyright
|
||||
of this work and relicense it under a more liberal license such as
|
||||
the MIT license. For now, mold is released under the GNU AGPL v3.
|
||||
|
||||
Copyright (c) 2023 Rui Ueyama
|
||||
If you just want to use mold but do not want to use AGPL software
|
||||
in your organization, a commercial license is available. Please
|
||||
contact [email protected].
|
||||
|
||||
Permission is hereby granted, free of charge, to any person obtaining a copy
|
||||
of this software and associated documentation files (the "Software"), to deal
|
||||
in the Software without restriction, including without limitation the rights
|
||||
to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
|
||||
copies of the Software, and to permit persons to whom the Software is
|
||||
furnished to do so, subject to the following conditions:
|
||||
--
|
||||
|
||||
The above copyright notice and this permission notice shall be included in all
|
||||
copies or substantial portions of the Software.
|
||||
GNU AFFERO GENERAL PUBLIC LICENSE
|
||||
Version 3, 19 November 2007
|
||||
|
||||
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
||||
IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
||||
FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
|
||||
AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
||||
LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
|
||||
OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
|
||||
SOFTWARE.
|
||||
Copyright (C) 2007 Free Software Foundation, Inc. <https://fsf.org/>
|
||||
Everyone is permitted to copy and distribute verbatim copies
|
||||
of this license document, but changing it is not allowed.
|
||||
|
||||
Preamble
|
||||
|
||||
The GNU Affero General Public License is a free, copyleft license for
|
||||
software and other kinds of works, specifically designed to ensure
|
||||
cooperation with the community in the case of network server software.
|
||||
|
||||
The licenses for most software and other practical works are designed
|
||||
to take away your freedom to share and change the works. By contrast,
|
||||
our General Public Licenses are intended to guarantee your freedom to
|
||||
share and change all versions of a program--to make sure it remains free
|
||||
software for all its users.
|
||||
|
||||
When we speak of free software, we are referring to freedom, not
|
||||
price. Our General Public Licenses are designed to make sure that you
|
||||
have the freedom to distribute copies of free software (and charge for
|
||||
them if you wish), that you receive source code or can get it if you
|
||||
want it, that you can change the software or use pieces of it in new
|
||||
free programs, and that you know you can do these things.
|
||||
|
||||
Developers that use our General Public Licenses protect your rights
|
||||
with two steps: (1) assert copyright on the software, and (2) offer
|
||||
you this License which gives you legal permission to copy, distribute
|
||||
and/or modify the software.
|
||||
|
||||
A secondary benefit of defending all users' freedom is that
|
||||
improvements made in alternate versions of the program, if they
|
||||
receive widespread use, become available for other developers to
|
||||
incorporate. Many developers of free software are heartened and
|
||||
encouraged by the resulting cooperation. However, in the case of
|
||||
software used on network servers, this result may fail to come about.
|
||||
The GNU General Public License permits making a modified version and
|
||||
letting the public access it on a server without ever releasing its
|
||||
source code to the public.
|
||||
|
||||
The GNU Affero General Public License is designed specifically to
|
||||
ensure that, in such cases, the modified source code becomes available
|
||||
to the community. It requires the operator of a network server to
|
||||
provide the source code of the modified version running there to the
|
||||
users of that server. Therefore, public use of a modified version, on
|
||||
a publicly accessible server, gives the public access to the source
|
||||
code of the modified version.
|
||||
|
||||
An older license, called the Affero General Public License and
|
||||
published by Affero, was designed to accomplish similar goals. This is
|
||||
a different license, not a version of the Affero GPL, but Affero has
|
||||
released a new version of the Affero GPL which permits relicensing under
|
||||
this license.
|
||||
|
||||
The precise terms and conditions for copying, distribution and
|
||||
modification follow.
|
||||
|
||||
TERMS AND CONDITIONS
|
||||
|
||||
0. Definitions.
|
||||
|
||||
"This License" refers to version 3 of the GNU Affero General Public License.
|
||||
|
||||
"Copyright" also means copyright-like laws that apply to other kinds of
|
||||
works, such as semiconductor masks.
|
||||
|
||||
"The Program" refers to any copyrightable work licensed under this
|
||||
License. Each licensee is addressed as "you". "Licensees" and
|
||||
"recipients" may be individuals or organizations.
|
||||
|
||||
To "modify" a work means to copy from or adapt all or part of the work
|
||||
in a fashion requiring copyright permission, other than the making of an
|
||||
exact copy. The resulting work is called a "modified version" of the
|
||||
earlier work or a work "based on" the earlier work.
|
||||
|
||||
A "covered work" means either the unmodified Program or a work based
|
||||
on the Program.
|
||||
|
||||
To "propagate" a work means to do anything with it that, without
|
||||
permission, would make you directly or secondarily liable for
|
||||
infringement under applicable copyright law, except executing it on a
|
||||
computer or modifying a private copy. Propagation includes copying,
|
||||
distribution (with or without modification), making available to the
|
||||
public, and in some countries other activities as well.
|
||||
|
||||
To "convey" a work means any kind of propagation that enables other
|
||||
parties to make or receive copies. Mere interaction with a user through
|
||||
a computer network, with no transfer of a copy, is not conveying.
|
||||
|
||||
An interactive user interface displays "Appropriate Legal Notices"
|
||||
to the extent that it includes a convenient and prominently visible
|
||||
feature that (1) displays an appropriate copyright notice, and (2)
|
||||
tells the user that there is no warranty for the work (except to the
|
||||
extent that warranties are provided), that licensees may convey the
|
||||
work under this License, and how to view a copy of this License. If
|
||||
the interface presents a list of user commands or options, such as a
|
||||
menu, a prominent item in the list meets this criterion.
|
||||
|
||||
1. Source Code.
|
||||
|
||||
The "source code" for a work means the preferred form of the work
|
||||
for making modifications to it. "Object code" means any non-source
|
||||
form of a work.
|
||||
|
||||
A "Standard Interface" means an interface that either is an official
|
||||
standard defined by a recognized standards body, or, in the case of
|
||||
interfaces specified for a particular programming language, one that
|
||||
is widely used among developers working in that language.
|
||||
|
||||
The "System Libraries" of an executable work include anything, other
|
||||
than the work as a whole, that (a) is included in the normal form of
|
||||
packaging a Major Component, but which is not part of that Major
|
||||
Component, and (b) serves only to enable use of the work with that
|
||||
Major Component, or to implement a Standard Interface for which an
|
||||
implementation is available to the public in source code form. A
|
||||
"Major Component", in this context, means a major essential component
|
||||
(kernel, window system, and so on) of the specific operating system
|
||||
(if any) on which the executable work runs, or a compiler used to
|
||||
produce the work, or an object code interpreter used to run it.
|
||||
|
||||
The "Corresponding Source" for a work in object code form means all
|
||||
the source code needed to generate, install, and (for an executable
|
||||
work) run the object code and to modify the work, including scripts to
|
||||
control those activities. However, it does not include the work's
|
||||
System Libraries, or general-purpose tools or generally available free
|
||||
programs which are used unmodified in performing those activities but
|
||||
which are not part of the work. For example, Corresponding Source
|
||||
includes interface definition files associated with source files for
|
||||
the work, and the source code for shared libraries and dynamically
|
||||
linked subprograms that the work is specifically designed to require,
|
||||
such as by intimate data communication or control flow between those
|
||||
subprograms and other parts of the work.
|
||||
|
||||
The Corresponding Source need not include anything that users
|
||||
can regenerate automatically from other parts of the Corresponding
|
||||
Source.
|
||||
|
||||
The Corresponding Source for a work in source code form is that
|
||||
same work.
|
||||
|
||||
2. Basic Permissions.
|
||||
|
||||
All rights granted under this License are granted for the term of
|
||||
copyright on the Program, and are irrevocable provided the stated
|
||||
conditions are met. This License explicitly affirms your unlimited
|
||||
permission to run the unmodified Program. The output from running a
|
||||
covered work is covered by this License only if the output, given its
|
||||
content, constitutes a covered work. This License acknowledges your
|
||||
rights of fair use or other equivalent, as provided by copyright law.
|
||||
|
||||
You may make, run and propagate covered works that you do not
|
||||
convey, without conditions so long as your license otherwise remains
|
||||
in force. You may convey covered works to others for the sole purpose
|
||||
of having them make modifications exclusively for you, or provide you
|
||||
with facilities for running those works, provided that you comply with
|
||||
the terms of this License in conveying all material for which you do
|
||||
not control copyright. Those thus making or running the covered works
|
||||
for you must do so exclusively on your behalf, under your direction
|
||||
and control, on terms that prohibit them from making any copies of
|
||||
your copyrighted material outside their relationship with you.
|
||||
|
||||
Conveying under any other circumstances is permitted solely under
|
||||
the conditions stated below. Sublicensing is not allowed; section 10
|
||||
makes it unnecessary.
|
||||
|
||||
3. Protecting Users' Legal Rights From Anti-Circumvention Law.
|
||||
|
||||
No covered work shall be deemed part of an effective technological
|
||||
measure under any applicable law fulfilling obligations under article
|
||||
11 of the WIPO copyright treaty adopted on 20 December 1996, or
|
||||
similar laws prohibiting or restricting circumvention of such
|
||||
measures.
|
||||
|
||||
When you convey a covered work, you waive any legal power to forbid
|
||||
circumvention of technological measures to the extent such circumvention
|
||||
is effected by exercising rights under this License with respect to
|
||||
the covered work, and you disclaim any intention to limit operation or
|
||||
modification of the work as a means of enforcing, against the work's
|
||||
users, your or third parties' legal rights to forbid circumvention of
|
||||
technological measures.
|
||||
|
||||
4. Conveying Verbatim Copies.
|
||||
|
||||
You may convey verbatim copies of the Program's source code as you
|
||||
receive it, in any medium, provided that you conspicuously and
|
||||
appropriately publish on each copy an appropriate copyright notice;
|
||||
keep intact all notices stating that this License and any
|
||||
non-permissive terms added in accord with section 7 apply to the code;
|
||||
keep intact all notices of the absence of any warranty; and give all
|
||||
recipients a copy of this License along with the Program.
|
||||
|
||||
You may charge any price or no price for each copy that you convey,
|
||||
and you may offer support or warranty protection for a fee.
|
||||
|
||||
5. Conveying Modified Source Versions.
|
||||
|
||||
You may convey a work based on the Program, or the modifications to
|
||||
produce it from the Program, in the form of source code under the
|
||||
terms of section 4, provided that you also meet all of these conditions:
|
||||
|
||||
a) The work must carry prominent notices stating that you modified
|
||||
it, and giving a relevant date.
|
||||
|
||||
b) The work must carry prominent notices stating that it is
|
||||
released under this License and any conditions added under section
|
||||
7. This requirement modifies the requirement in section 4 to
|
||||
"keep intact all notices".
|
||||
|
||||
c) You must license the entire work, as a whole, under this
|
||||
License to anyone who comes into possession of a copy. This
|
||||
License will therefore apply, along with any applicable section 7
|
||||
additional terms, to the whole of the work, and all its parts,
|
||||
regardless of how they are packaged. This License gives no
|
||||
permission to license the work in any other way, but it does not
|
||||
invalidate such permission if you have separately received it.
|
||||
|
||||
d) If the work has interactive user interfaces, each must display
|
||||
Appropriate Legal Notices; however, if the Program has interactive
|
||||
interfaces that do not display Appropriate Legal Notices, your
|
||||
work need not make them do so.
|
||||
|
||||
A compilation of a covered work with other separate and independent
|
||||
works, which are not by their nature extensions of the covered work,
|
||||
and which are not combined with it such as to form a larger program,
|
||||
in or on a volume of a storage or distribution medium, is called an
|
||||
"aggregate" if the compilation and its resulting copyright are not
|
||||
used to limit the access or legal rights of the compilation's users
|
||||
beyond what the individual works permit. Inclusion of a covered work
|
||||
in an aggregate does not cause this License to apply to the other
|
||||
parts of the aggregate.
|
||||
|
||||
6. Conveying Non-Source Forms.
|
||||
|
||||
You may convey a covered work in object code form under the terms
|
||||
of sections 4 and 5, provided that you also convey the
|
||||
machine-readable Corresponding Source under the terms of this License,
|
||||
in one of these ways:
|
||||
|
||||
a) Convey the object code in, or embodied in, a physical product
|
||||
(including a physical distribution medium), accompanied by the
|
||||
Corresponding Source fixed on a durable physical medium
|
||||
customarily used for software interchange.
|
||||
|
||||
b) Convey the object code in, or embodied in, a physical product
|
||||
(including a physical distribution medium), accompanied by a
|
||||
written offer, valid for at least three years and valid for as
|
||||
long as you offer spare parts or customer support for that product
|
||||
model, to give anyone who possesses the object code either (1) a
|
||||
copy of the Corresponding Source for all the software in the
|
||||
product that is covered by this License, on a durable physical
|
||||
medium customarily used for software interchange, for a price no
|
||||
more than your reasonable cost of physically performing this
|
||||
conveying of source, or (2) access to copy the
|
||||
Corresponding Source from a network server at no charge.
|
||||
|
||||
c) Convey individual copies of the object code with a copy of the
|
||||
written offer to provide the Corresponding Source. This
|
||||
alternative is allowed only occasionally and noncommercially, and
|
||||
only if you received the object code with such an offer, in accord
|
||||
with subsection 6b.
|
||||
|
||||
d) Convey the object code by offering access from a designated
|
||||
place (gratis or for a charge), and offer equivalent access to the
|
||||
Corresponding Source in the same way through the same place at no
|
||||
further charge. You need not require recipients to copy the
|
||||
Corresponding Source along with the object code. If the place to
|
||||
copy the object code is a network server, the Corresponding Source
|
||||
may be on a different server (operated by you or a third party)
|
||||
that supports equivalent copying facilities, provided you maintain
|
||||
clear directions next to the object code saying where to find the
|
||||
Corresponding Source. Regardless of what server hosts the
|
||||
Corresponding Source, you remain obligated to ensure that it is
|
||||
available for as long as needed to satisfy these requirements.
|
||||
|
||||
e) Convey the object code using peer-to-peer transmission, provided
|
||||
you inform other peers where the object code and Corresponding
|
||||
Source of the work are being offered to the general public at no
|
||||
charge under subsection 6d.
|
||||
|
||||
A separable portion of the object code, whose source code is excluded
|
||||
from the Corresponding Source as a System Library, need not be
|
||||
included in conveying the object code work.
|
||||
|
||||
A "User Product" is either (1) a "consumer product", which means any
|
||||
tangible personal property which is normally used for personal, family,
|
||||
or household purposes, or (2) anything designed or sold for incorporation
|
||||
into a dwelling. In determining whether a product is a consumer product,
|
||||
doubtful cases shall be resolved in favor of coverage. For a particular
|
||||
product received by a particular user, "normally used" refers to a
|
||||
typical or common use of that class of product, regardless of the status
|
||||
of the particular user or of the way in which the particular user
|
||||
actually uses, or expects or is expected to use, the product. A product
|
||||
is a consumer product regardless of whether the product has substantial
|
||||
commercial, industrial or non-consumer uses, unless such uses represent
|
||||
the only significant mode of use of the product.
|
||||
|
||||
"Installation Information" for a User Product means any methods,
|
||||
procedures, authorization keys, or other information required to install
|
||||
and execute modified versions of a covered work in that User Product from
|
||||
a modified version of its Corresponding Source. The information must
|
||||
suffice to ensure that the continued functioning of the modified object
|
||||
code is in no case prevented or interfered with solely because
|
||||
modification has been made.
|
||||
|
||||
If you convey an object code work under this section in, or with, or
|
||||
specifically for use in, a User Product, and the conveying occurs as
|
||||
part of a transaction in which the right of possession and use of the
|
||||
User Product is transferred to the recipient in perpetuity or for a
|
||||
fixed term (regardless of how the transaction is characterized), the
|
||||
Corresponding Source conveyed under this section must be accompanied
|
||||
by the Installation Information. But this requirement does not apply
|
||||
if neither you nor any third party retains the ability to install
|
||||
modified object code on the User Product (for example, the work has
|
||||
been installed in ROM).
|
||||
|
||||
The requirement to provide Installation Information does not include a
|
||||
requirement to continue to provide support service, warranty, or updates
|
||||
for a work that has been modified or installed by the recipient, or for
|
||||
the User Product in which it has been modified or installed. Access to a
|
||||
network may be denied when the modification itself materially and
|
||||
adversely affects the operation of the network or violates the rules and
|
||||
protocols for communication across the network.
|
||||
|
||||
Corresponding Source conveyed, and Installation Information provided,
|
||||
in accord with this section must be in a format that is publicly
|
||||
documented (and with an implementation available to the public in
|
||||
source code form), and must require no special password or key for
|
||||
unpacking, reading or copying.
|
||||
|
||||
7. Additional Terms.
|
||||
|
||||
"Additional permissions" are terms that supplement the terms of this
|
||||
License by making exceptions from one or more of its conditions.
|
||||
Additional permissions that are applicable to the entire Program shall
|
||||
be treated as though they were included in this License, to the extent
|
||||
that they are valid under applicable law. If additional permissions
|
||||
apply only to part of the Program, that part may be used separately
|
||||
under those permissions, but the entire Program remains governed by
|
||||
this License without regard to the additional permissions.
|
||||
|
||||
When you convey a copy of a covered work, you may at your option
|
||||
remove any additional permissions from that copy, or from any part of
|
||||
it. (Additional permissions may be written to require their own
|
||||
removal in certain cases when you modify the work.) You may place
|
||||
additional permissions on material, added by you to a covered work,
|
||||
for which you have or can give appropriate copyright permission.
|
||||
|
||||
Notwithstanding any other provision of this License, for material you
|
||||
add to a covered work, you may (if authorized by the copyright holders of
|
||||
that material) supplement the terms of this License with terms:
|
||||
|
||||
a) Disclaiming warranty or limiting liability differently from the
|
||||
terms of sections 15 and 16 of this License; or
|
||||
|
||||
b) Requiring preservation of specified reasonable legal notices or
|
||||
author attributions in that material or in the Appropriate Legal
|
||||
Notices displayed by works containing it; or
|
||||
|
||||
c) Prohibiting misrepresentation of the origin of that material, or
|
||||
requiring that modified versions of such material be marked in
|
||||
reasonable ways as different from the original version; or
|
||||
|
||||
d) Limiting the use for publicity purposes of names of licensors or
|
||||
authors of the material; or
|
||||
|
||||
e) Declining to grant rights under trademark law for use of some
|
||||
trade names, trademarks, or service marks; or
|
||||
|
||||
f) Requiring indemnification of licensors and authors of that
|
||||
material by anyone who conveys the material (or modified versions of
|
||||
it) with contractual assumptions of liability to the recipient, for
|
||||
any liability that these contractual assumptions directly impose on
|
||||
those licensors and authors.
|
||||
|
||||
All other non-permissive additional terms are considered "further
|
||||
restrictions" within the meaning of section 10. If the Program as you
|
||||
received it, or any part of it, contains a notice stating that it is
|
||||
governed by this License along with a term that is a further
|
||||
restriction, you may remove that term. If a license document contains
|
||||
a further restriction but permits relicensing or conveying under this
|
||||
License, you may add to a covered work material governed by the terms
|
||||
of that license document, provided that the further restriction does
|
||||
not survive such relicensing or conveying.
|
||||
|
||||
If you add terms to a covered work in accord with this section, you
|
||||
must place, in the relevant source files, a statement of the
|
||||
additional terms that apply to those files, or a notice indicating
|
||||
where to find the applicable terms.
|
||||
|
||||
Additional terms, permissive or non-permissive, may be stated in the
|
||||
form of a separately written license, or stated as exceptions;
|
||||
the above requirements apply either way.
|
||||
|
||||
8. Termination.
|
||||
|
||||
You may not propagate or modify a covered work except as expressly
|
||||
provided under this License. Any attempt otherwise to propagate or
|
||||
modify it is void, and will automatically terminate your rights under
|
||||
this License (including any patent licenses granted under the third
|
||||
paragraph of section 11).
|
||||
|
||||
However, if you cease all violation of this License, then your
|
||||
license from a particular copyright holder is reinstated (a)
|
||||
provisionally, unless and until the copyright holder explicitly and
|
||||
finally terminates your license, and (b) permanently, if the copyright
|
||||
holder fails to notify you of the violation by some reasonable means
|
||||
prior to 60 days after the cessation.
|
||||
|
||||
Moreover, your license from a particular copyright holder is
|
||||
reinstated permanently if the copyright holder notifies you of the
|
||||
violation by some reasonable means, this is the first time you have
|
||||
received notice of violation of this License (for any work) from that
|
||||
copyright holder, and you cure the violation prior to 30 days after
|
||||
your receipt of the notice.
|
||||
|
||||
Termination of your rights under this section does not terminate the
|
||||
licenses of parties who have received copies or rights from you under
|
||||
this License. If your rights have been terminated and not permanently
|
||||
reinstated, you do not qualify to receive new licenses for the same
|
||||
material under section 10.
|
||||
|
||||
9. Acceptance Not Required for Having Copies.
|
||||
|
||||
You are not required to accept this License in order to receive or
|
||||
run a copy of the Program. Ancillary propagation of a covered work
|
||||
occurring solely as a consequence of using peer-to-peer transmission
|
||||
to receive a copy likewise does not require acceptance. However,
|
||||
nothing other than this License grants you permission to propagate or
|
||||
modify any covered work. These actions infringe copyright if you do
|
||||
not accept this License. Therefore, by modifying or propagating a
|
||||
covered work, you indicate your acceptance of this License to do so.
|
||||
|
||||
10. Automatic Licensing of Downstream Recipients.
|
||||
|
||||
Each time you convey a covered work, the recipient automatically
|
||||
receives a license from the original licensors, to run, modify and
|
||||
propagate that work, subject to this License. You are not responsible
|
||||
for enforcing compliance by third parties with this License.
|
||||
|
||||
An "entity transaction" is a transaction transferring control of an
|
||||
organization, or substantially all assets of one, or subdividing an
|
||||
organization, or merging organizations. If propagation of a covered
|
||||
work results from an entity transaction, each party to that
|
||||
transaction who receives a copy of the work also receives whatever
|
||||
licenses to the work the party's predecessor in interest had or could
|
||||
give under the previous paragraph, plus a right to possession of the
|
||||
Corresponding Source of the work from the predecessor in interest, if
|
||||
the predecessor has it or can get it with reasonable efforts.
|
||||
|
||||
You may not impose any further restrictions on the exercise of the
|
||||
rights granted or affirmed under this License. For example, you may
|
||||
not impose a license fee, royalty, or other charge for exercise of
|
||||
rights granted under this License, and you may not initiate litigation
|
||||
(including a cross-claim or counterclaim in a lawsuit) alleging that
|
||||
any patent claim is infringed by making, using, selling, offering for
|
||||
sale, or importing the Program or any portion of it.
|
||||
|
||||
11. Patents.
|
||||
|
||||
A "contributor" is a copyright holder who authorizes use under this
|
||||
License of the Program or a work on which the Program is based. The
|
||||
work thus licensed is called the contributor's "contributor version".
|
||||
|
||||
A contributor's "essential patent claims" are all patent claims
|
||||
owned or controlled by the contributor, whether already acquired or
|
||||
hereafter acquired, that would be infringed by some manner, permitted
|
||||
by this License, of making, using, or selling its contributor version,
|
||||
but do not include claims that would be infringed only as a
|
||||
consequence of further modification of the contributor version. For
|
||||
purposes of this definition, "control" includes the right to grant
|
||||
patent sublicenses in a manner consistent with the requirements of
|
||||
this License.
|
||||
|
||||
Each contributor grants you a non-exclusive, worldwide, royalty-free
|
||||
patent license under the contributor's essential patent claims, to
|
||||
make, use, sell, offer for sale, import and otherwise run, modify and
|
||||
propagate the contents of its contributor version.
|
||||
|
||||
In the following three paragraphs, a "patent license" is any express
|
||||
agreement or commitment, however denominated, not to enforce a patent
|
||||
(such as an express permission to practice a patent or covenant not to
|
||||
sue for patent infringement). To "grant" such a patent license to a
|
||||
party means to make such an agreement or commitment not to enforce a
|
||||
patent against the party.
|
||||
|
||||
If you convey a covered work, knowingly relying on a patent license,
|
||||
and the Corresponding Source of the work is not available for anyone
|
||||
to copy, free of charge and under the terms of this License, through a
|
||||
publicly available network server or other readily accessible means,
|
||||
then you must either (1) cause the Corresponding Source to be so
|
||||
available, or (2) arrange to deprive yourself of the benefit of the
|
||||
patent license for this particular work, or (3) arrange, in a manner
|
||||
consistent with the requirements of this License, to extend the patent
|
||||
license to downstream recipients. "Knowingly relying" means you have
|
||||
actual knowledge that, but for the patent license, your conveying the
|
||||
covered work in a country, or your recipient's use of the covered work
|
||||
in a country, would infringe one or more identifiable patents in that
|
||||
country that you have reason to believe are valid.
|
||||
|
||||
If, pursuant to or in connection with a single transaction or
|
||||
arrangement, you convey, or propagate by procuring conveyance of, a
|
||||
covered work, and grant a patent license to some of the parties
|
||||
receiving the covered work authorizing them to use, propagate, modify
|
||||
or convey a specific copy of the covered work, then the patent license
|
||||
you grant is automatically extended to all recipients of the covered
|
||||
work and works based on it.
|
||||
|
||||
A patent license is "discriminatory" if it does not include within
|
||||
the scope of its coverage, prohibits the exercise of, or is
|
||||
conditioned on the non-exercise of one or more of the rights that are
|
||||
specifically granted under this License. You may not convey a covered
|
||||
work if you are a party to an arrangement with a third party that is
|
||||
in the business of distributing software, under which you make payment
|
||||
to the third party based on the extent of your activity of conveying
|
||||
the work, and under which the third party grants, to any of the
|
||||
parties who would receive the covered work from you, a discriminatory
|
||||
patent license (a) in connection with copies of the covered work
|
||||
conveyed by you (or copies made from those copies), or (b) primarily
|
||||
for and in connection with specific products or compilations that
|
||||
contain the covered work, unless you entered into that arrangement,
|
||||
or that patent license was granted, prior to 28 March 2007.
|
||||
|
||||
Nothing in this License shall be construed as excluding or limiting
|
||||
any implied license or other defenses to infringement that may
|
||||
otherwise be available to you under applicable patent law.
|
||||
|
||||
12. No Surrender of Others' Freedom.
|
||||
|
||||
If conditions are imposed on you (whether by court order, agreement or
|
||||
otherwise) that contradict the conditions of this License, they do not
|
||||
excuse you from the conditions of this License. If you cannot convey a
|
||||
covered work so as to satisfy simultaneously your obligations under this
|
||||
License and any other pertinent obligations, then as a consequence you may
|
||||
not convey it at all. For example, if you agree to terms that obligate you
|
||||
to collect a royalty for further conveying from those to whom you convey
|
||||
the Program, the only way you could satisfy both those terms and this
|
||||
License would be to refrain entirely from conveying the Program.
|
||||
|
||||
13. Remote Network Interaction; Use with the GNU General Public License.
|
||||
|
||||
Notwithstanding any other provision of this License, if you modify the
|
||||
Program, your modified version must prominently offer all users
|
||||
interacting with it remotely through a computer network (if your version
|
||||
supports such interaction) an opportunity to receive the Corresponding
|
||||
Source of your version by providing access to the Corresponding Source
|
||||
from a network server at no charge, through some standard or customary
|
||||
means of facilitating copying of software. This Corresponding Source
|
||||
shall include the Corresponding Source for any work covered by version 3
|
||||
of the GNU General Public License that is incorporated pursuant to the
|
||||
following paragraph.
|
||||
|
||||
Notwithstanding any other provision of this License, you have
|
||||
permission to link or combine any covered work with a work licensed
|
||||
under version 3 of the GNU General Public License into a single
|
||||
combined work, and to convey the resulting work. The terms of this
|
||||
License will continue to apply to the part which is the covered work,
|
||||
but the work with which it is combined will remain governed by version
|
||||
3 of the GNU General Public License.
|
||||
|
||||
14. Revised Versions of this License.
|
||||
|
||||
The Free Software Foundation may publish revised and/or new versions of
|
||||
the GNU Affero General Public License from time to time. Such new versions
|
||||
will be similar in spirit to the present version, but may differ in detail to
|
||||
address new problems or concerns.
|
||||
|
||||
Each version is given a distinguishing version number. If the
|
||||
Program specifies that a certain numbered version of the GNU Affero General
|
||||
Public License "or any later version" applies to it, you have the
|
||||
option of following the terms and conditions either of that numbered
|
||||
version or of any later version published by the Free Software
|
||||
Foundation. If the Program does not specify a version number of the
|
||||
GNU Affero General Public License, you may choose any version ever published
|
||||
by the Free Software Foundation.
|
||||
|
||||
If the Program specifies that a proxy can decide which future
|
||||
versions of the GNU Affero General Public License can be used, that proxy's
|
||||
public statement of acceptance of a version permanently authorizes you
|
||||
to choose that version for the Program.
|
||||
|
||||
Later license versions may give you additional or different
|
||||
permissions. However, no additional obligations are imposed on any
|
||||
author or copyright holder as a result of your choosing to follow a
|
||||
later version.
|
||||
|
||||
15. Disclaimer of Warranty.
|
||||
|
||||
THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY
|
||||
APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT
|
||||
HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY
|
||||
OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO,
|
||||
THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
|
||||
PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM
|
||||
IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF
|
||||
ALL NECESSARY SERVICING, REPAIR OR CORRECTION.
|
||||
|
||||
16. Limitation of Liability.
|
||||
|
||||
IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
|
||||
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS
|
||||
THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY
|
||||
GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE
|
||||
USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF
|
||||
DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD
|
||||
PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS),
|
||||
EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF
|
||||
SUCH DAMAGES.
|
||||
|
||||
17. Interpretation of Sections 15 and 16.
|
||||
|
||||
If the disclaimer of warranty and limitation of liability provided
|
||||
above cannot be given local legal effect according to their terms,
|
||||
reviewing courts shall apply local law that most closely approximates
|
||||
an absolute waiver of all civil liability in connection with the
|
||||
Program, unless a warranty or assumption of liability accompanies a
|
||||
copy of the Program in return for a fee.
|
||||
|
||||
END OF TERMS AND CONDITIONS
|
||||
|
||||
How to Apply These Terms to Your New Programs
|
||||
|
||||
If you develop a new program, and you want it to be of the greatest
|
||||
possible use to the public, the best way to achieve this is to make it
|
||||
free software which everyone can redistribute and change under these terms.
|
||||
|
||||
To do so, attach the following notices to the program. It is safest
|
||||
to attach them to the start of each source file to most effectively
|
||||
state the exclusion of warranty; and each file should have at least
|
||||
the "copyright" line and a pointer to where the full notice is found.
|
||||
|
||||
<one line to give the program's name and a brief idea of what it does.>
|
||||
Copyright (C) <year> <name of author>
|
||||
|
||||
This program is free software: you can redistribute it and/or modify
|
||||
it under the terms of the GNU Affero General Public License as published
|
||||
by the Free Software Foundation, either version 3 of the License, or
|
||||
(at your option) any later version.
|
||||
|
||||
This program is distributed in the hope that it will be useful,
|
||||
but WITHOUT ANY WARRANTY; without even the implied warranty of
|
||||
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
|
||||
GNU Affero General Public License for more details.
|
||||
|
||||
You should have received a copy of the GNU Affero General Public License
|
||||
along with this program. If not, see <https://www.gnu.org/licenses/>.
|
||||
|
||||
Also add information on how to contact you by electronic and paper mail.
|
||||
|
||||
If your software can interact with users remotely through a computer
|
||||
network, you should also make sure that it provides a way for users to
|
||||
get its source. For example, if your program is a web application, its
|
||||
interface could display a "Source" link that leads users to an archive
|
||||
of the code. There are many ways you could offer source, and different
|
||||
solutions will be better for different programs; see section 13 for the
|
||||
specific requirements.
|
||||
|
||||
You should also get your employer (if you work as a programmer) or school,
|
||||
if any, to sign a "copyright disclaimer" for the program, if necessary.
|
||||
For more information on this, and how to apply and follow the GNU AGPL, see
|
||||
<https://www.gnu.org/licenses/>.
|
||||
@@ -0,0 +1,120 @@
|
||||
ifeq ($(origin CC), default)
|
||||
CC = clang
|
||||
endif
|
||||
|
||||
ifeq ($(origin CXX), default)
|
||||
CXX = clang++
|
||||
endif
|
||||
|
||||
GIT_HASH ?= $(shell [ -d .git ] && git rev-parse HEAD)
|
||||
|
||||
OS ?= $(shell uname -s)
|
||||
|
||||
CPPFLAGS = -g -pthread -std=c++20 -fPIE \
|
||||
-DMOLD_VERSION=\"0.9.6\" \
|
||||
-DGIT_HASH=\"$(GIT_HASH)\" \
|
||||
$(EXTRA_CPPFLAGS)
|
||||
LDFLAGS += $(EXTRA_LDFLAGS)
|
||||
LIBS = -pthread -lz -lxxhash -ldl -lm
|
||||
|
||||
SRCS=$(wildcard elf/*.cc)
|
||||
HEADERS=$(wildcard elf/*.h)
|
||||
OBJS=$(SRCS:.cc=.o)
|
||||
|
||||
PREFIX ?= /usr
|
||||
DEST = $(DESTDIR)$(PREFIX)
|
||||
DEBUG ?= 0
|
||||
LTO ?= 0
|
||||
ASAN ?= 0
|
||||
TSAN ?= 0
|
||||
|
||||
ifeq ($(DEBUG), 1)
|
||||
CPPFLAGS += -O0
|
||||
else
|
||||
CPPFLAGS += -O2
|
||||
endif
|
||||
|
||||
ifeq ($(LTO), 1)
|
||||
CPPFLAGS += -flto -O3
|
||||
LDFLAGS += -flto
|
||||
endif
|
||||
|
||||
ifeq ($(ASAN), 1)
|
||||
CPPFLAGS += -fsanitize=address
|
||||
LDFLAGS += -fsanitize=address
|
||||
else ifeq ($(TSAN), 1)
|
||||
CPPFLAGS += -fsanitize=thread
|
||||
LDFLAGS += -fsanitize=thread
|
||||
else ifneq ($(OS), Darwin)
|
||||
# By default, we want to use mimalloc as a memory allocator.
|
||||
# Since replacing the standard malloc is not compatible with ASAN,
|
||||
# we do that only when ASAN is not enabled.
|
||||
ifdef SYSTEM_MIMALLOC
|
||||
LIBS += -lmimalloc
|
||||
else
|
||||
MIMALLOC_LIB = out/mimalloc/libmimalloc.a
|
||||
CPPFLAGS += -Ithird-party/mimalloc/include
|
||||
LIBS += -Wl,-whole-archive $(MIMALLOC_LIB) -Wl,-no-whole-archive
|
||||
endif
|
||||
endif
|
||||
|
||||
ifdef SYSTEM_TBB
|
||||
LIBS += -ltbb
|
||||
else
|
||||
TBB_LIB = out/tbb/libs/libtbb.a
|
||||
LIBS += $(TBB_LIB)
|
||||
CPPFLAGS += -Ithird-party/tbb/include
|
||||
endif
|
||||
|
||||
ifneq ($(OS), Darwin)
|
||||
LIBS += -lcrypto
|
||||
endif
|
||||
|
||||
all: mold mold-wrapper.so
|
||||
|
||||
mold: $(OBJS) $(MIMALLOC_LIB) $(TBB_LIB)
|
||||
$(CXX) $(CXXFLAGS) $(OBJS) -o $@ $(LDFLAGS) $(LIBS)
|
||||
ln -sf mold ld
|
||||
|
||||
mold-wrapper.so: elf/mold-wrapper.c Makefile
|
||||
$(CC) -fPIC -shared -o $@ $< -ldl
|
||||
|
||||
$(OBJS): $(HEADERS) Makefile
|
||||
|
||||
$(MIMALLOC_LIB):
|
||||
mkdir -p out/mimalloc
|
||||
(cd out/mimalloc; CFLAGS=-DMI_USE_ENVIRON=0 cmake -G'Unix Makefiles' ../../third-party/mimalloc)
|
||||
$(MAKE) -C out/mimalloc mimalloc-static
|
||||
|
||||
$(TBB_LIB):
|
||||
mkdir -p out/tbb
|
||||
(cd out/tbb; cmake -G'Unix Makefiles' -DBUILD_SHARED_LIBS=OFF -DTBB_TEST=OFF -DCMAKE_CXX_FLAGS=-D__TBB_DYNAMIC_LOAD_ENABLED=0 -DTBB_STRICT=OFF ../../third-party/tbb)
|
||||
$(MAKE) -C out/tbb tbb
|
||||
(cd out/tbb; ln -sf *_relwithdebinfo libs)
|
||||
|
||||
test tests check: all
|
||||
$(MAKE) -C test --output-sync --no-print-directory
|
||||
|
||||
install: all
|
||||
install -m 755 -d $(DEST)/bin
|
||||
install -m 755 mold $(DEST)/bin
|
||||
strip $(DEST)/bin/mold
|
||||
|
||||
install -m 755 -d $(DEST)/lib/mold
|
||||
install -m 644 mold-wrapper.so $(DEST)/lib/mold
|
||||
strip $(DEST)/lib/mold/mold-wrapper.so
|
||||
|
||||
install -m 755 -d $(DEST)/share/man/man1
|
||||
install -m 644 docs/mold.1 $(DEST)/share/man/man1
|
||||
rm -f $(DEST)/share/man/man1/mold.1.gz
|
||||
gzip -9 $(DEST)/share/man/man1/mold.1
|
||||
|
||||
uninstall:
|
||||
rm -f $(DEST)/bin/mold
|
||||
rm -f $(DEST)/share/man/man1/mold.1.gz
|
||||
rm -rf $(DEST)/lib/mold
|
||||
|
||||
clean:
|
||||
rm -rf *.o elf/*.o *~ mold mold-wrapper.so test/tmp out ld
|
||||
|
||||
.PHONY: all test tests check clean $(TESTS)
|
||||
@@ -1,205 +1,125 @@
|
||||
# mold: A Modern Linker
|
||||
|
||||
mold is a faster drop-in replacement for existing Unix linkers. It is several
|
||||
times quicker than the LLVM lld linker, the second-fastest open-source linker,
|
||||
which I initially developed a few years ago. mold aims to enhance developer
|
||||
productivity by minimizing build time, particularly in rapid
|
||||
debug-edit-rebuild cycles.
|
||||
|
||||
|
||||
Here is a performance comparison of GNU ld, GNU gold, LLVM lld, and
|
||||
mold when linking final debuginfo-enabled executables for major large
|
||||
programs on a simulated 16-core, 32-thread machine.
|
||||
mold is a high-performance drop-in replacement for existing Unix
|
||||
linkers. It is several times faster than LLVM lld linker, the (then-)
|
||||
fastest open-source linker which I originally created a few years ago.
|
||||
Here is a performance comparison of GNU gold, LLVM lld, and mold for
|
||||
linking final executables of major large programs.
|
||||
|
||||
|
||||
| Program (linker output size) | GNU gold | LLVM lld | mold | mold w/ preloading
|
||||
|-------------------------------|----------|----------|-------|-------------------
|
||||
| Firefox 87 (1.6 GiB) | 29.2s | 6.16s | 1.69s | 0.79s
|
||||
| Chrome 86 (1.9 GiB) | 54.5s | 11.7s | 1.85s | 0.97s
|
||||
| Clang 13 (3.1 GiB) | 59.4s | 5.68s | 2.76s | 0.86s
|
||||
|
||||
| Program (linker output size) | GNU ld | GNU gold | LLVM lld | mold
|
||||
|-------------------------------|--------|----------|----------|------
|
||||
| MySQL 8.3 (0.47 GiB) | 10.84s | 7.47s | 1.64s | 0.46s
|
||||
| Clang 19 (1.56 GiB) | 42.07s | 33.13s | 5.20s | 1.35s
|
||||
| Chromium 124 (1.35 GiB) | N/A | 27.40s | 6.10s | 1.52s
|
||||
(These numbers are measured on an AMD Threadripper 3990X 64-core
|
||||
machine with 32 threads enabled. All programs are built with debug
|
||||
info enabled.)
|
||||
|
||||
mold is so fast that it is only 2x _slower_ than the `cp` command on the same
|
||||
machine. If you find that mold is not faster than other linkers, feel
|
||||
free to [file a bug report](https://github.com/rui314/mold/issues).
|
||||
Let me explain the "w/ preloading" column. mold supports the file
|
||||
preloading feature. That is, if you run mold with `-preload` flag
|
||||
along with other command-line flags, it becomes a daemon and halts
|
||||
after parsing input files. Then, if you invoke mold with the same
|
||||
command-line options (except `-preload` flag), it tells the daemon to
|
||||
reload only updated files and proceed. With this feature enabled, and
|
||||
if most of the input files haven't been updated, mold achieves a
|
||||
near-`cp` performance or even exceeds it, as the throughput of file
|
||||
copy using the `cp` command is about 2 GiB/s on my machine.
|
||||
|
||||
mold supports x86-64, i386, ARM64, ARM32, 64-bit/32-bit little/big-endian
|
||||
RISC-V, 32-bit PowerPC, 64-bit big-endian PowerPC ELFv1, 64-bit little-endian
|
||||
PowerPC ELFv2, s390x, 64-bit/32-bit LoongArch, SPARC64, m68k, and SH-4.
|
||||
So, mold is extremely fast per se and even faster with a bit of cheating.
|
||||
|
||||
## Why does linking speed matter?
|
||||
Why is mold so fast? One reason is that it simply uses faster
|
||||
algorithms and efficient data structures than other linkers do.
|
||||
The other reason is that the new linker is highly parallelized.
|
||||
|
||||
If you are using a compiled language such as C, C++, or Rust, a build consists
|
||||
of two phases. In the first phase, a compiler compiles source files into
|
||||
object files (`.o` files). In the second phase, a linker takes all object
|
||||
files and combines them into a single executable or shared library file.
|
||||
Here is a side-by-side comparison of per-core CPU usage of lld (left)
|
||||
and mold (right). They are linking the same program, Chromium
|
||||
executable.
|
||||
|
||||
The second phase can be time-consuming if your build output is large. mold can
|
||||
speed up this process, saving you time and preventing distractions while
|
||||
waiting for a lengthy build to finish. The difference is most noticeable
|
||||
during rapid debug-edit-rebuild cycles.
|
||||
|
||||
|
||||
## Installation
|
||||
As you can see, mold uses all available cores throughout its execution
|
||||
and finishes quickly. On the other hand, lld failed to use available
|
||||
cores most of the time. In this demo, the maximum parallelism is
|
||||
artificially capped to 16 so that the bars fit in the GIF.
|
||||
|
||||
Binary packages for the following systems are currently available:
|
||||
Currently, mold is being developed with Linux/x86-64 as the primary
|
||||
target platform. mold can link many user-land programs including large
|
||||
ones such as web browsers for that target. It also has preliminary
|
||||
Linux/i386 support. Supporting other OSes and ISAs are planned after
|
||||
Linux/x86-64 support is complete.
|
||||
|
||||
[](https://repology.org/project/mold/versions)
|
||||
## How to build
|
||||
|
||||
## How to Build
|
||||
mold is written in C++20, so you need a very recent version of GCC or
|
||||
Clang. I'm using Ubuntu 20.04 as a development platform. In that
|
||||
environment, you can build mold by the following commands.
|
||||
|
||||
mold is written in C++20, so if you build mold yourself, you will need a
|
||||
recent version of a C++ compiler and a C++ standard library. We recommend GCC
|
||||
10.2 or Clang 16.0.0 (or later) and libstdc++ 10 or libc++ 7 (or later).
|
||||
|
||||
### Install Dependencies
|
||||
|
||||
To install build dependencies, run `./install-build-deps.sh` in this
|
||||
directory. It will detect your Linux distribution and attempt to install the
|
||||
necessary packages. You may need to run it as root.
|
||||
|
||||
### Compile mold
|
||||
|
||||
```shell
|
||||
git clone --branch stable https://github.com/rui314/mold.git
|
||||
cd mold
|
||||
./install-build-deps.sh
|
||||
cmake -DCMAKE_BUILD_TYPE=Release -DCMAKE_CXX_COMPILER=c++ -B build
|
||||
cmake --build build -j$(nproc)
|
||||
sudo cmake --build build --target install
|
||||
```
|
||||
$ sudo apt-get install build-essential libstdc++-10-dev cmake clang libssl-dev zlib1g-dev libxxhash-dev git
|
||||
$ git clone https://github.com/rui314/mold.git
|
||||
$ cd mold
|
||||
$ git checkout v0.9.6
|
||||
$ make
|
||||
```
|
||||
|
||||
You might need to pass a C++20 compiler command name to `cmake`. In the
|
||||
example above, `c++` is passed. If that doesn't work for you, try a specific
|
||||
version of a compiler, such as `g++-10` or `clang++-12`.
|
||||
The last `make` command creates `mold` executable.
|
||||
|
||||
By default, `mold` is installed to `/usr/local/bin`. You can change the
|
||||
installation location by passing `-DCMAKE_INSTALL_PREFIX=<directory>`.
|
||||
For other cmake options, see the comments in `CMakeLists.txt`.
|
||||
If you don't have Ubuntu 20.04, or if for any reason `make` in the
|
||||
above commands doesn't work for you, you can use Docker to build it in
|
||||
a Docker environment. To do so, just run `./build-static.sh` in this
|
||||
directory. The script creates a Ubuntu 20.04 Docker image, installs
|
||||
necessary tools and libraries to it, and builds mold as a static binary.
|
||||
|
||||
If you are not using a recent enough Linux distribution, or if `cmake` does
|
||||
not work for you for any reason, you can use Podman to build mold in a
|
||||
container. To do so, run `./dist.sh` in this directory instead of using
|
||||
`cmake`. The shell script will pull a container image, build mold and auxiliary
|
||||
files inside it, and package them into a single tar file named
|
||||
`dist/mold-$version-$arch-linux.tar.gz`. You can extract the tar file anywhere
|
||||
and use the mold executable in it.
|
||||
`make test` depends on a few more packages. To install, run the following commands:
|
||||
|
||||
```
|
||||
$ sudo dpkg --add-architecture i386
|
||||
$ sudo apt update
|
||||
$ sudo apt-get install bsdmainutils dwarfdump libc6-dev:i386 lib32gcc-10-dev libstdc++-10-dev-arm64-cross gcc-10-aarch64-linux-gnu g++-10-aarch64-linux-gnu
|
||||
```
|
||||
|
||||
## How to use
|
||||
|
||||
<details><summary>A classic way to use mold</summary>
|
||||
On Unix, the linker command (which is usually `/usr/bin/ld`) is
|
||||
invoked indirectly by `cc` (or `gcc` or `clang`), which is typically
|
||||
in turn indirectly invoked by `make` or some other build system command.
|
||||
|
||||
On Unix, the linker command (usually `/usr/bin/ld`) is indirectly invoked by
|
||||
the compiler driver (typically `cc`, `gcc`, or `clang`), which is in turn
|
||||
indirectly invoked by `make` or other build system commands.
|
||||
A classic way to use `mold`:
|
||||
- `clang` before 12.0: pass `-fuse-ld=<absolute-path-to-mold-executable>`;
|
||||
- clang after 12.0: pass `--ld-path=<absolute-path-to-mold-executable>`;
|
||||
- gcc: `--ld-path` patch [has been declined by GCC maintainers](https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573833.html), instead they advise to use a [workaround](https://gcc.gnu.org/pipermail/gcc-patches/2021-June/573823.html): create directory `<dirname>`, then `ln -s <path-to-mold> <dirname>/ld`, and then pass `-B<dirname>` (`-B` tells GCC to look for `ld` in specified location).
|
||||
|
||||
If you can specify an additional command line option for your compiler driver
|
||||
by modifying the build system's config files, add one of the following flags
|
||||
to use mold instead of `/usr/bin/ld`:
|
||||
It is sometimes very hard to pass an appropriate command line
|
||||
option to `cc` to specify an alternative linker.
|
||||
To deal with the situation, mold has a feature to intercept all
|
||||
invocations of `/usr/bin/ld`, `/usr/bin/ld.lld` or `/usr/bin/ld.gold`
|
||||
and redirect it to itself. To use the feature, run `make` (or another
|
||||
build command) as a subcommand of mold as follows:
|
||||
|
||||
- For Clang: pass `-fuse-ld=mold`
|
||||
|
||||
- For GCC 12.1.0 or later: pass `-fuse-ld=mold`
|
||||
|
||||
- For GCC before 12.1.0: the `-fuse-ld` option does not accept `mold` as a
|
||||
valid argument, so you need to use the `-B` option instead. The `-B` option
|
||||
tells GCC where to look for external commands like `ld`.
|
||||
|
||||
If you have installed mold with `make install`, there should be a directory
|
||||
named `/usr/libexec/mold` (or `/usr/local/libexec/mold`, depending on your
|
||||
`$PREFIX`), and the `ld` command should be there. The `ld` is actually a
|
||||
symlink to `mold`. So, all you need is to pass `-B/usr/libexec/mold` (or
|
||||
`-B/usr/local/libexec/mold`) to GCC.
|
||||
|
||||
If you haven't installed `ld.mold` to any `$PATH`, you can still pass
|
||||
`-fuse-ld=/absolute/path/to/mold` to clang to use mold. However, GCC does not
|
||||
accept an absolute path as an argument for `-fuse-ld`.
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>If you are using Rust</summary>
|
||||
|
||||
Create `.cargo/config.toml` in your project directory with the following:
|
||||
|
||||
```toml
|
||||
[target.'cfg(target_os = "linux")']
|
||||
linker = "clang"
|
||||
rustflags = ["-C", "link-arg=-fuse-ld=/path/to/mold"]
|
||||
```
|
||||
$ path/to/mold -run make <make-options-if-any>
|
||||
```
|
||||
|
||||
where `/path/to/mold` is an absolute path to the mold executable. In the
|
||||
example above, we use `clang` as a linker driver since it always accepts the
|
||||
`-fuse-ld` option. If your GCC is recent enough to recognize the option, you
|
||||
may be able to remove the `linker = "clang"` line.
|
||||
Here's an example showing how to link Rust code when using the
|
||||
cargo package manager.
|
||||
|
||||
```toml
|
||||
[target.'cfg(target_os = "linux")']
|
||||
rustflags = ["-C", "link-arg=-fuse-ld=mold"]
|
||||
```
|
||||
$ path/to/mold -run cargo build
|
||||
```
|
||||
|
||||
If you want to use mold for all projects, add the above snippet to
|
||||
`~/.cargo/config.toml`.
|
||||
Internally, mold invokes a given command with `LD_PRELOAD` environment
|
||||
variable set to its companion shared object file. The shared object
|
||||
file intercepts all function calls to exec-family functions to replace
|
||||
`argv[0]` with `mold` if it is `/usr/bin/ld`, `/usr/bin/ld.gold` or
|
||||
`/usr/bin/ld.lld`.
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>If you are using Nim</summary>
|
||||
|
||||
Create `config.nims` in your project directory with the following:
|
||||
|
||||
```nim
|
||||
when findExe("mold").len > 0 and defined(linux):
|
||||
switch("passL", "-fuse-ld=mold")
|
||||
```
|
||||
|
||||
where `mold` must be included in the `PATH` environment variable. In this
|
||||
example, `gcc` is used as the linker driver. Use the `-fuse-ld` option if your
|
||||
GCC is recent enough to recognize this option.
|
||||
|
||||
If you want to use mold for all projects, add the above snippet to
|
||||
`~/.config/config.nims`.
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>If you are using Conan package manager</summary>
|
||||
|
||||
You can configure [Conan](https://github.com/conan-io) to download the latest
|
||||
version of `mold` and use it as the linker when building your dependencies and
|
||||
projects from source. Please see the instructions [here](https://conan.io/center/recipes/mold).
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>mold -run</summary>
|
||||
|
||||
It is sometimes very hard to pass an appropriate command line option to `cc`
|
||||
to specify an alternative linker. To address this situation, mold has a
|
||||
feature to intercept all invocations of `ld`, `ld.bfd`, `ld.lld`, or `ld.gold`
|
||||
and redirect them to itself. To use this feature, run `make` (or another build
|
||||
command) as a subcommand of mold as follows:
|
||||
|
||||
```shell
|
||||
mold -run make <make-options-if-any>
|
||||
```
|
||||
|
||||
Internally, mold invokes a given command with the `LD_PRELOAD` environment
|
||||
variable set to its companion shared object file. The shared object file
|
||||
intercepts all function calls to `exec(3)`-family functions to replace
|
||||
`argv[0]` with `mold` if it is `ld`, `ld.bf`, `ld.gold`, or `ld.lld`.
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>GitHub Actions</summary>
|
||||
|
||||
You can use our [setup-mold](https://github.com/rui314/setup-mold) GitHub
|
||||
Action to speed up GitHub-hosted continuous builds. Although GitHub Actions
|
||||
run on a 4 core machine, mold is still significantly faster than the default
|
||||
GNU linker, especially when linking large programs.
|
||||
|
||||
</details>
|
||||
|
||||
<details><summary>Verify that you are using mold</summary>
|
||||
|
||||
mold leaves its identification string in the `.comment` section of an output
|
||||
mold leaves its identification string in `.comment` section in an output
|
||||
file. You can print it out to verify that you are actually using mold.
|
||||
|
||||
```shell
|
||||
```
|
||||
$ readelf -p .comment <executable-file>
|
||||
|
||||
String dump of section '.comment':
|
||||
@@ -207,69 +127,505 @@ String dump of section '.comment':
|
||||
[ 2b] mold 9a1679b47d9b22012ec7dfbda97c8983956716f7
|
||||
```
|
||||
|
||||
If `mold` is present in the `.comment` section, the file was created by mold.
|
||||
If `mold` is in `.comment`, the file is created by mold.
|
||||
|
||||
</details>
|
||||
# Design and implementation of mold
|
||||
|
||||
<details><summary>Online manual</summary>
|
||||
For the rest of this documentation, I'll explain the design and the
|
||||
implementation of mold. If you are only interested in using mold, you
|
||||
don't need to read the below.
|
||||
|
||||
Since mold is a drop-in replacement, you should be able to use it without
|
||||
reading its manual. However, if you need it, [mold's man page](docs/mold.md)
|
||||
is available online. You can read the same manual by running `man mold`.
|
||||
## Motivation
|
||||
|
||||
</details>
|
||||
Here is why I'm writing a new linker:
|
||||
|
||||
## Why is mold so fast?
|
||||
- Even though lld has significantly improved the situation, linking is
|
||||
still one of the slowest steps in a build. It is especially
|
||||
annoying when I changed one line of code and had to wait for a few
|
||||
seconds or even more for a linker to complete. It should be
|
||||
instantaneous. There's a need for a faster linker.
|
||||
|
||||
One reason is that it utilizes faster algorithms and more efficient data
|
||||
structures compared to other linkers. Another reason is that mold is highly
|
||||
parallelized.
|
||||
- The number of cores on a PC has increased a lot lately, and this
|
||||
trend is expected to continue. However, the existing linkers can't
|
||||
take the advantage of the trend because they don't scale well for more
|
||||
cores. I have a 64-core/128-thread machine, so my goal is to create
|
||||
a linker that uses the CPU nicely. mold should be much faster than
|
||||
other linkers on 4 or 8-core machines too, though.
|
||||
|
||||
Here is a side-by-side comparison of per-core CPU usage for lld (left) and
|
||||
mold (right), linking the same program, a Chromium executable.
|
||||
- It looks to me that the designs of the existing linkers are somewhat
|
||||
too similar, and I believe there are a lot of drastically different
|
||||
designs that haven't been explored yet. Developers generally don't
|
||||
care about linkers as long as they work correctly, and they don't
|
||||
even think about creating a new one. So there may be lots of low
|
||||
hanging fruits there in this area.
|
||||
|
||||
|
||||
## Basic design
|
||||
|
||||
As you can see, mold uses all available cores throughout its execution and
|
||||
finishes quickly. In contrast, lld fails to utilize available cores most of
|
||||
the time. In this demo, the maximum parallelism is artificially capped at 16,
|
||||
so that the bars fit in the GIF.
|
||||
- In order to achieve a `cp`-like performance, the most important
|
||||
thing is to fix the layout of an output file as quickly as possible, so
|
||||
that we can start copying actual data from input object files to an
|
||||
output file as soon as possible.
|
||||
|
||||
For details, please see the [design notes](docs/design.md).
|
||||
- Copying data from input files to an output file is I/O-bounded, so
|
||||
there should be room for doing computationally-intensive tasks while
|
||||
copying data from one file to another.
|
||||
|
||||
## Sponsors
|
||||
- We should allow the linker to preload object files from disk and
|
||||
parse them in memory before a complete set of input object files
|
||||
is ready. To do so, we need
|
||||
to split the linker into two in such a way that the latter half of
|
||||
the process finishes as quickly as possible by speculatively parsing
|
||||
and preprocessing input files in the first half of the process.
|
||||
|
||||
It is taken for granted nowadays that compiler toolchains can be easily
|
||||
installed and used for free, and people may not think too much about the
|
||||
individuals behind these "free tools". mold supports many projects, but it
|
||||
is essentially a one-person project. This situation is similar to the one
|
||||
depicted in the following xkcd illustration.
|
||||
- One of the most computationally-intensive stage among linker stages
|
||||
is symbol resolution. To resolve symbols, we basically have to throw
|
||||
all symbol strings into a hash table to match undefined symbols with
|
||||
defined symbols. But this can be done in the preloading stage using
|
||||
[string interning](https://en.wikipedia.org/wiki/String_interning).
|
||||
|
||||
[](https://xkcd.com/2347)
|
||||
- Object files may contain a special section called a mergeable string
|
||||
section. The section contains lots of null-terminated strings, and
|
||||
the linker is expected to gather all mergeable string sections and
|
||||
merge their contents. So, if two object files contain the same
|
||||
string literal, for example, the resulting output will contain a
|
||||
single merged string. This step is computationally intensive, but string
|
||||
merging can be done in the preloading stage using string interning.
|
||||
|
||||
If you think that the "Nebraska guy" should be rewarded, please consider
|
||||
becoming our [GitHub sponsor](https://github.com/sponsors/rui314)!
|
||||
- Static archives (.a files) contain object files, but the static
|
||||
archive's string table contains only defined symbols of member
|
||||
object files and lacks other types of symbols. That makes static
|
||||
archives unsuitable for speculative parsing. Therefore, the linker
|
||||
should ignore the symbol table of static archive and directly read
|
||||
static archive members.
|
||||
|
||||
We thank everyone who sponsors our project. In particular, we'd like to acknowledge
|
||||
the following people and organizations who have sponsored $128/month or more:
|
||||
- If there's a relocation that uses a GOT of a symbol, then we have to
|
||||
create a GOT entry for that symbol. Otherwise, we shouldn't. That
|
||||
means we need to scan all relocation tables to fix the length and
|
||||
the contents of a .got section. This is computationally intensive,
|
||||
but this step is parallelizable.
|
||||
|
||||
### Corporate sponsors
|
||||
## Compatibility
|
||||
|
||||
<a href="https://mercury.com"><img src="docs/mercury-logo.png" align=center height=120 width=400 alt=Mercury></a>
|
||||
- GNU ld, GNU gold and LLVM lld support essentially the same set of
|
||||
command line options and features. mold doesn't have to be
|
||||
completely compatible with them. As long as it can be used for
|
||||
linking large user-land programs, I'm fine with that. It is OK to
|
||||
leave some command line options unimplemented; if mold is blazingly
|
||||
fast, other projects would still be happy to adopt it by modifying
|
||||
their projects' build files.
|
||||
|
||||
<a href="https://cybozu-global.com"><img src="docs/cyboze-logo.png" align=center height=120 width=133 alt=Cybozu></a>
|
||||
- mold emits Linux executables and runs only on Linux. I won't avoid
|
||||
Unix-ism when writing code. I don't want to think about portability
|
||||
until mold becomes a thing that's worth being ported.
|
||||
|
||||
<a href="https://www.emergetools.com"><img src="docs/emerge-tools-logo.png" align=center height=120 width=240 alt="Emerge Tools"></a><br>
|
||||
## Linker Script
|
||||
|
||||
- [G-Research](https://www.gresearch.co.uk)
|
||||
- [Signal Slot Inc.](https://github.com/signal-slot)
|
||||
- [GlareDB](https://github.com/GlareDB)
|
||||
Linker script is an embedded language for the linker. It is mainly
|
||||
used to control how input sections are mapped to output sections and
|
||||
the layout of the output, but it can also do a lot of tricky stuff.
|
||||
Its feature is useful especially for embedded programming, but it's
|
||||
also an awfully underdocumented and complex language.
|
||||
|
||||
### Individual sponsors
|
||||
We have to implement a subset of the linker script language anwyay,
|
||||
because on Linux, /usr/lib/x86_64-linux-gnu/libc.so is (despite its
|
||||
name) not a shared object file but actually an ASCII file containing
|
||||
linker script code to load the _actual_ libc.so file. But the feature
|
||||
set for this purpose is very limited, and it is okay to implement them
|
||||
to mold.
|
||||
|
||||
- [Wei Wu](https://github.com/lazyparser)
|
||||
- [kyle-elliott](https://github.com/kyle-elliott)
|
||||
- [Bryant Biggs](https://github.com/bryantbiggs)
|
||||
- [kraptor23](https://github.com/kraptor23)
|
||||
- [Jinkyu Yi](https://github.com/jincreator)
|
||||
- [Pedro Navarro](https://github.com/pedronavf)
|
||||
Besides that, we really don't want to implement the linker script
|
||||
language. But at the same time, we want to satisfy the user needs that
|
||||
are currently satisfied with the linker script language. So, what
|
||||
should we do? Here is my observation:
|
||||
|
||||
- Linker script allows doing a lot of tricky stuff, such as specifying
|
||||
the exact layout of a file, inserting arbitrary bytes between
|
||||
sections, etc. But most of them can be done with a post-link binary
|
||||
editing tool (such as `objcopy`).
|
||||
|
||||
- It looks like there are two things that truly cannot be done by a
|
||||
post-link editing tool: (a) mapping input sections to output
|
||||
sections, and (b) applying relocations.
|
||||
|
||||
From the above observation, I believe we need to provide only the
|
||||
following features instead of the entire linker script language:
|
||||
|
||||
- A method to specify how input sections are mapped to output
|
||||
sections, and
|
||||
|
||||
- a method to set addresses to output sections, so that relocations
|
||||
are applied based on desired addresses.
|
||||
|
||||
I believe everything else can be done with a post-link binary editing
|
||||
tool.
|
||||
|
||||
## Details
|
||||
|
||||
- As we aim to the 1-second goal for Chromium, every millisecond
|
||||
counts. We can't ignore the latency of process exit. If we mmap a
|
||||
lot of files, \_exit(2) is not instantaneous but takes a few hundred
|
||||
milliseconds because the kernel has to clean up a lot of
|
||||
resources. As a workaround, we should organize the linker command as
|
||||
two processes; the first process forks the second process, and the
|
||||
second process does the actual work. As soon as the second process
|
||||
writes a result file to a filesystem, it notifies the first process,
|
||||
and the first process exits. The second process can take time to
|
||||
exit, because it is not an interactive process.
|
||||
|
||||
- At least on Linux, it looks like the filesystem's performance to
|
||||
allocate new blocks to a new file is the limiting factor when
|
||||
creating a new large file and filling its contents using mmap.
|
||||
If you already have a large file in the buffer cache, writing to it is
|
||||
much faster than creating a new fresh file and writing to it.
|
||||
Based on this observation, mold overwrites to an existing
|
||||
executable file if exists. My quick benchmark showed that I could
|
||||
save 300 milliseconds when creating a 2 GiB output file.
|
||||
Linux doesn't allow to open an executable for writing if it is
|
||||
running (you'll get a "text busy" error if you attempt). mold
|
||||
falls back to the usual way if it fails to open an output file.
|
||||
|
||||
- The output from the linker should be deterministic for the sake of
|
||||
[build reproducibility](https://en.wikipedia.org/wiki/Reproducible_builds)
|
||||
and ease of debugging. This might add a little bit of overhead to
|
||||
the linker, but that shouldn't be too much.
|
||||
|
||||
- A .build-id, a unique ID embedded to an output file, is usually
|
||||
computed by applying a cryptographic hash function (e.g. SHA-1) to
|
||||
an output file. This is a slow step, but we can speed it up by
|
||||
splitting a file into small chunks, computing SHA-1 for each chunk,
|
||||
and then computing SHA-1 of the concatenated SHA-1 hashes
|
||||
(i.e. constructing a [Merkle
|
||||
Tree](https://en.wikipedia.org/wiki/Merkle_tree) of height 2).
|
||||
Modern x86 processors have purpose-built instructions for SHA-1 and
|
||||
can compute SHA-1 pretty quickly at about 2 GiB/s. Using 16
|
||||
cores, a build-id for a 2 GiB executable can be computed in 60 to 70
|
||||
milliseconds.
|
||||
|
||||
- BFD, gold, and lld support section garbage collection. That is, a
|
||||
linker runs a mark-sweep garbage collection on an input graph, where
|
||||
sections are vertices and relocations are edges, to discard all
|
||||
sections that are not reachable from the entry point symbol
|
||||
(i.e. `_start`) or a few other root sections. In mold, we are using
|
||||
multiple threads to mark sections concurrently.
|
||||
|
||||
- Similarly, BFD, gold an lld support Identical Comdat Folding (ICF)
|
||||
as yet another size optimization. ICF merges two or more read-only
|
||||
sections that happen to have the same contents and relocations.
|
||||
To do that, we have to find isomorphic subgraphs from larger graphs.
|
||||
I implemented a new algorithm for mold, which is 5x faster than lld
|
||||
to do ICF for Chromium (from 5 seconds to 1 second).
|
||||
|
||||
- [Intel Threading Building
|
||||
Blocks](https://github.com/oneapi-src/oneTBB) (TBB) is a good
|
||||
library for parallel execution and has several concurrent
|
||||
containers. We are particularly interested in using
|
||||
`parallel_for_each` and `concurrent_hash_map`.
|
||||
|
||||
- TBB provides `tbbmalloc` which works better for multi-threaded
|
||||
applications than the glib'c malloc, but it looks like
|
||||
[jemalloc](https://github.com/jemalloc/jemalloc) and
|
||||
[mimalloc](https://github.com/microsoft/mimalloc) are a little bit
|
||||
more scalable than `tbbmalloc`.
|
||||
|
||||
## Size of the problem
|
||||
|
||||
When linking Chrome, a linker reads 3,430,966,844 bytes of data in
|
||||
total. The data contains the following items:
|
||||
|
||||
| Data item | Number
|
||||
| ------------------------ | ------
|
||||
| Object files | 30,723
|
||||
| Public undefined symbols | 1,428,149
|
||||
| Mergeable strings | 1,579,996
|
||||
| Comdat groups | 9,914,510
|
||||
| Regular sections¹ | 10,345,314
|
||||
| Public defined symbols | 10,512,135
|
||||
| Symbols | 23,953,607
|
||||
| Sections | 27,543,225
|
||||
| Relocations against SHF_ALLOC sections | 39,496,375
|
||||
| Relocations | 62,024,719
|
||||
|
||||
¹ Sections that have to be copied from input object files to an
|
||||
output file. Sections that contain relocations or symbols are for
|
||||
example excluded.
|
||||
|
||||
## Internals
|
||||
|
||||
In this section, I'll explain the internals of mold linker.
|
||||
|
||||
### A brief history of Unix and the Unix linker
|
||||
|
||||
Conceptually, what a linker does is pretty simple. A compiler compiles
|
||||
a fragment of a program (a single source file) into a fragment of
|
||||
machine code and data (an object file, which typically has the .o
|
||||
extension), and a linker stitches them together into a single
|
||||
executable or a shared library image.
|
||||
|
||||
In reality, modern linkers for Unix-like systems are much more
|
||||
complicated than the naive understanding because they have gradually
|
||||
gained one feature at a time over the 50 years history of Unix, and
|
||||
they are now something like a bag of lots of miscellaneous features in
|
||||
which none of the features is more important than the others. It is
|
||||
very easy to miss the forest for the trees, since for those who don't
|
||||
know the details of the Unix linker, it is not clear which feature is
|
||||
essential and which is not.
|
||||
|
||||
That being said, one thing is clear that at any point of Unix history,
|
||||
a Unix linker has a coherent feature set for the Unix of that age. So,
|
||||
let me entangle the history to see how the operating system, runtime,
|
||||
and linker have gained features that we see today. That should give
|
||||
you an idea of why a particular feature has been added to a linker in the
|
||||
first place.
|
||||
|
||||
1. Original Unix didn't support shared libraries, and a program was
|
||||
always loaded to a fixed address. An executable was something like
|
||||
a memory dump that was just loaded to a particular address by the
|
||||
kernel. After loading, the kernel started executing the program by
|
||||
setting the instruction pointer to a particular address.
|
||||
|
||||
The most essential feature for any linker is relocation processing.
|
||||
The original Unix linker of course supported that. Let me explain
|
||||
what that is.
|
||||
|
||||
Individual object files are inevitably incomplete as a program,
|
||||
because when a compiler created them, it only see a part of an
|
||||
entire program. For example, if an object file contains a function
|
||||
call that refers to another object file, the `call` instruction in the
|
||||
object cannot be complete, as the compiler has no idea as to what
|
||||
is the called function's address. To deal with this, the compiler
|
||||
emits a placeholder value (typically just zero) instead of a real
|
||||
address and leaves metadata in an object file saying "fix offset X
|
||||
of this file with an address of Y". That metadata is called
|
||||
"relocation". Relocations are typically processed by the linker.
|
||||
|
||||
It is easy for a linker to apply relocations for the original Unix
|
||||
because a program is always loaded to a fixed address. It exactly
|
||||
knows the addresses of all functions and data when linking a
|
||||
program.
|
||||
|
||||
Static library support, which is still an important feature of Unix
|
||||
linker, also dates back to this early period of Unix history.
|
||||
To understand what it is, imagine that you are trying to compile
|
||||
a program for the early Unix. You don't want to waste time to
|
||||
compile libc functions every time you compile your program (the
|
||||
computers of the era were incredibly slow), so you have already
|
||||
placed each libc function into a separate source file and compiled
|
||||
them individually. That means you have object files for each libc
|
||||
function, e.g., printf.o, scanf.o, atoi.o, write.o, etc.
|
||||
|
||||
Given this configuration, all you have to do to link your program
|
||||
against libc functions is to pick up the right set of libc object
|
||||
files and give them to the linker along with the object files of your
|
||||
program. But, keeping the linker command line in sync with the
|
||||
libc functions you are using in your program is bothersome. You can
|
||||
be conservative; you can specify all libc object files to the
|
||||
command line, but that leads to program bloat because the linker
|
||||
unconditionally link all object files given to it no matter whether
|
||||
they are used or not. So, a new feature was added to the linker to
|
||||
fix the problem. That is the static library, which is also called
|
||||
the archive file.
|
||||
|
||||
An archive file is just a bundle of object files, just like zip
|
||||
file but in an uncompressed form. An archive file typically has the
|
||||
.a file extension and named after its contents. For example, the
|
||||
archive file containing all libc objects is named `libc.a`.
|
||||
|
||||
If you pass an archive file along with other object files to the
|
||||
linker, the linker pulls out an object file from the archive _only
|
||||
when_ it is referenced by other object files. In other words,
|
||||
unlike object files directly given to a linker, object files
|
||||
wrapped in an archive are not linked to the output by default.
|
||||
An archive works as a supplement to complete your program.
|
||||
|
||||
Even today, you can still find a libc archive file. Run `ar t
|
||||
/usr/lib/x86_64-linux-gnu/libc.a` on Linux should give you a list
|
||||
of object files in the libc archive.
|
||||
|
||||
2. In the '80s, Sun Microsystems, a leading commercial Unix vendor at the
|
||||
time, added shared library support to their Unix variant, SunOS.
|
||||
|
||||
(This section is incomplete.)
|
||||
|
||||
## Concurrency strategy
|
||||
|
||||
In this section, I'll explain the high-level concurrency strategy of
|
||||
mold.
|
||||
|
||||
In most places, mold adopts data parallelism. That is, we have a huge
|
||||
number of pieces of data of the same kind, and we process each of them
|
||||
individually using parallel for-loop. For example, after identifying
|
||||
the exact set of input object files, we need to scan all relocation
|
||||
tables to determine the sizes of .got and .plt sections. We do that
|
||||
using a parallel for-loop. The granularity of parallel processing in
|
||||
this case is the relocation table.
|
||||
|
||||
Data parallelism is very efficient and scalable because there's no
|
||||
need for threads to communicate with each other while working on each
|
||||
element of data. In addition to that, data parallelism is easy to
|
||||
understand, as it is just a for-loop in which multiple iterations may
|
||||
be executed in parallel. We don't use high-level communication or
|
||||
synchronization mechanisms such as channels, futures, promises,
|
||||
latches or something like that in mold.
|
||||
|
||||
In some cases, we need to share a little bit of data between threads
|
||||
while executing a parallel for-loop. For example, the loop to scan
|
||||
relocations turns on "requires GOT" or "requires PLT" flags in a
|
||||
symbol. Symbol is a shared resource, and writing to them from multiple
|
||||
threads without synchronization is unsafe. To deal with it, we made
|
||||
the flag an atomic variable.
|
||||
|
||||
The other common pattern you can find in mold which is build on top of
|
||||
the parallel for-loop is the map-reduce pattern. That is, we run a
|
||||
parallel for-loop on a large data set to produce a small data set and
|
||||
process the small data set with a single thread. Let me take a
|
||||
build-id computation as an example. Build-id is typically computed by
|
||||
applying a cryptographic hash function such as SHA-1 on a linker's
|
||||
output file. To compute it, we first consider an output as a sequence
|
||||
of 1 MiB blocks and compute a SHA-1 hash for each block in parallel.
|
||||
Then, we concatenate the SHA-1 hashes and compute a SHA-1 hash on the
|
||||
hashes to get a final build-id.
|
||||
|
||||
Finally, we use concurrent hashmap at a few places in mold. Concurrent
|
||||
hashmap is a hashmap to which multiple threads can safely insert items
|
||||
in parallel. We use it in the symbol resolution stage, for example.
|
||||
To resolve symbols, we basically have to throw in all defined symbols
|
||||
into a hash table, so that we can find a matching defined symbol for
|
||||
an undefined symbol by name. We do the hash table insertion from a
|
||||
parallel for-loop which iterates over a list of input files.
|
||||
|
||||
Overall, even though mold is highly scalable, it succeeded to avoid
|
||||
complexties you often find in complex parallel programs. From high
|
||||
level, mold just serially executes the linker's internal passes one by
|
||||
one. Each pass is parallelized using parallel for-loops.
|
||||
|
||||
## Rejected ideas
|
||||
|
||||
In this section, I'll explain the alternative designs I currently do
|
||||
not plan to implement and why I turned them down.
|
||||
|
||||
- Placing variable-length sections at end of an output file and start
|
||||
copying file contents before fixing the output file layout
|
||||
|
||||
Idea: Fixing the layout of regular sections seems easy, and if we
|
||||
place them at beginning of a file, we can start copying their
|
||||
contents from their input files to an output file. While copying
|
||||
file contents, we can compute the sizes of variable-length sections
|
||||
such as .got or .plt and place them at end of the file.
|
||||
|
||||
Reason for rejection: I did not choose this design because I doubt
|
||||
if it could actually shorten link time and I think I don't need it
|
||||
anyway.
|
||||
|
||||
The linker has to de-duplicate comdat sections (i.e. inline
|
||||
functions that are included in multiple object files), so we
|
||||
cannot compute the layout of regular sections until we resolve all
|
||||
symbols and de-duplicate comdats. That takes a few hundred
|
||||
milliseconds. After that, we can compute the sizes of
|
||||
variable-length sections in less than 100 milliseconds. It's quite
|
||||
fast, so it doesn't seem to make much sense to proceed without
|
||||
fixing the final file layout.
|
||||
|
||||
The other reason to reject this idea is because there's good a
|
||||
chance for this idea to have a negative impact on linker's overall
|
||||
performance. If we copy file contents before fixing the layout, we
|
||||
can't apply relocations to them while copying because symbol
|
||||
addresses are not available yet. If we fix the file layout first, we
|
||||
can apply relocations while copying, which is effectively zero-cost
|
||||
due to a very good data locality. On the other hand, if we apply
|
||||
relocations long after we copy file contents, it's pretty expensive
|
||||
because section contents are very likely to have been evicted from
|
||||
CPU cache.
|
||||
|
||||
- Incremental linking
|
||||
|
||||
Idea: Incremental linking is a technique to patch a previous linker's
|
||||
output file so that only functions or data that are updated from the
|
||||
previous build are written to it. It is expected to significantly
|
||||
reduce the amount of data copied from input files to an output file
|
||||
and thus speed up linking. GNU BFD and gold linkers support it.
|
||||
|
||||
Reason for rejection: I turned it down because it (1) is
|
||||
complicated, (2) doesn't seem to speed it up that much and (3) has
|
||||
several practical issues. Let me explain each of them.
|
||||
|
||||
First, incremental linking for real C/C++ programs is not as easy as
|
||||
one might think. Let me take malloc as an example. malloc is usually
|
||||
defined by libc, but you can implement it in your program, and if
|
||||
that's the case, the symbol `malloc` will be resolved to your
|
||||
function instead of the one in libc. If you include a library that
|
||||
defines malloc (such as libjemalloc or libtbbmallc) before libc,
|
||||
their malloc will override libc's malloc.
|
||||
|
||||
Assume that you are using a nonstandard malloc. What if you remove
|
||||
your malloc from your code, or remove `-ljemalloc` from your
|
||||
Makefile? The linker has to include a malloc from libc, which may
|
||||
include more object files to satisfy its dependencies. Such code
|
||||
change can affect the entire program rather than just replacing one
|
||||
function. The same is true for adding malloc to your program. Making
|
||||
a local change doesn't necessarily result in a local change in the
|
||||
binary level. It can easily have cascading effects.
|
||||
|
||||
Some ELF fancy features make incremental linking even harder to
|
||||
implement. Take the weak symbol as an example. If you define `atoi`
|
||||
as a weak symbol in your program, and if you are not using `atoi`
|
||||
at all in your program, that symbol will be resolved to address
|
||||
0. But if you start using some libc function that indirectly calls
|
||||
`atoi`, then `atoi` will be included in your program, and your weak
|
||||
symbol will be resolved to that function. I don't know how to
|
||||
efficiently fix up a binary for this case.
|
||||
|
||||
This is a hard problem, so existing linkers don't try too hard to
|
||||
solve it. For example, IIRC, gold falls back to full link if any
|
||||
function is removed from a previous build. If you want to not annoy
|
||||
users in the fallback case, you need to make full link fast anyway.
|
||||
|
||||
Second, incremental linking itself has an overhead. It has to detect
|
||||
updated files, patch an existing output file and write additional
|
||||
data to an output file for future incremental linking. GNU gold, for
|
||||
instance, takes almost 30 seconds on my machine to do a null
|
||||
incremental link (i.e. no object files are updated from a previous
|
||||
build) for chrome. It's just too slow.
|
||||
|
||||
Third, there are other practical issues in incremental linking. It's
|
||||
not reproducible, so your binary isn't going to be the same as other
|
||||
binaries even if you are compiling the same source tree using the
|
||||
same compiler toolchain. Or, it is complex and there might be a bug
|
||||
in it. If something doesn't work correctly, "remove --incremental
|
||||
from your Makefile and try again" could be a piece of advice, but
|
||||
that isn't ideal.
|
||||
|
||||
So, all in all, incremental linking is tricky. I wanted to make full
|
||||
link as fast as possible, so that we don't have to think about how
|
||||
to work around the slowness of full link.
|
||||
|
||||
- Defining a completely new file format and use it
|
||||
|
||||
Idea: Sometimes, the ELF file format itself seems to be a limiting
|
||||
factor in improving the linker's performance. We might be able to make a
|
||||
far better one if we create a new file format.
|
||||
|
||||
Reason for rejection: I rejected the idea because it apparently has
|
||||
a practical issue (backward compatibility issue) and also doesn't
|
||||
seem to improve the performance of linkers that much. As clearly
|
||||
demonstrated by mold, we can create a fast linker for ELF. I believe
|
||||
ELF isn't that bad, after all. The semantics of the existing Unix
|
||||
linkers, such as the name resolution algorithm or the linker script,
|
||||
have slowed the linkers down, but that's not a problem of the file
|
||||
format itself.
|
||||
|
||||
- Watching object files using inotify(2)
|
||||
|
||||
Idea: When mold is running as a daemon for preloading, use
|
||||
inotify(2) to watch file system updates so that it can reload files
|
||||
as soon as they are updated.
|
||||
|
||||
Reason for rejection: Just like the maximum number of files you can
|
||||
simultaneously open, the maximum number of files you can watch using
|
||||
inotify(2) isn't that large. Maybe just a single instance of mold is
|
||||
fine with inotify(2), but it may fail if you run multiple of it.
|
||||
|
||||
The other reason for not doing it is because mold is quite fast
|
||||
without it anyway. Invoking stat(2) on each file for file update
|
||||
check takes less than 100 milliseconds for Chrome, and if most of
|
||||
the input files are not updated, parsing updated files takes almost
|
||||
no time.
|
||||
Executable
+37
@@ -0,0 +1,37 @@
|
||||
#!/bin/bash -x
|
||||
# This is a shell script to build a statically-linked mold
|
||||
# executable using Docker, so that it is easy to build mold
|
||||
# on non-Ubuntu 20.04 machines.
|
||||
#
|
||||
# A docker image used for building mold is persistent.
|
||||
# Run `docker image rm mold-build-ubuntu20` to remove the image
|
||||
# from disk.
|
||||
|
||||
set -e
|
||||
|
||||
# If the existing file is not statically-linked, remove it.
|
||||
[ -f mold ] && ! ldd mold 2>&1 | grep -q 'not a dynamic executable' && \
|
||||
rm mold
|
||||
|
||||
cat <<EOF | docker build -t mold-build-ubuntu20 -
|
||||
FROM ubuntu:20.04
|
||||
RUN apt-get update && \
|
||||
TZ=Europe/London apt-get install -y tzdata && \
|
||||
apt-get install -y build-essential git clang lld cmake \
|
||||
libstdc++-10-dev libxxhash-dev zlib1g-dev libssl-dev && \
|
||||
rm -rf /var/lib/apt/lists/*
|
||||
EOF
|
||||
|
||||
LDFLAGS='-fuse-ld=lld -static'
|
||||
|
||||
# libstdc++'s `std::__glibcxx_rwlock_rdlock` refers these symbols
|
||||
# as weak symbols, although they need to be defined. Otherwise,
|
||||
# the program crashes after juping to address 0.
|
||||
# So, we force loading symbols as a workaround.
|
||||
LDFLAGS="$LDFLAGS -Wl,-u,pthread_rwlock_rdlock"
|
||||
LDFLAGS="$LDFLAGS -Wl,-u,pthread_rwlock_unlock"
|
||||
LDFLAGS="$LDFLAGS -Wl,-u,pthread_rwlock_wrlock"
|
||||
|
||||
docker run -it --rm -v `pwd`:/mold -u $(id -u):$(id -g) \
|
||||
mold-build-ubuntu20 \
|
||||
make -C /mold -j$(nproc) EXTRA_LDFLAGS="$LDFLAGS"
|
||||
@@ -1,247 +0,0 @@
|
||||
#!/bin/bash
|
||||
#
|
||||
# This script creates a mold binary distribution. The output is written to
|
||||
# the `dist` directory as `mold-$version-$arch-linux.tar.gz` (e.g.
|
||||
# `mold-2.40.0-x86_64-linux.tar.gz`).
|
||||
#
|
||||
# This script aims to produce reproducible outputs. That means each time
|
||||
# it's run on the same git commit, it generates a bit-for-bit identical
|
||||
# binary file regardless of when or where it's executed. This property
|
||||
# serves as a strong safeguard against supply chain attacks. With a
|
||||
# reproducible build, anyone can independently verify that the binary
|
||||
# files published on our GitHub release page were built from the git
|
||||
# commit tagged for release by rebuilding the binaries themselves.
|
||||
#
|
||||
# Debian provides snapshot.debian.org to host all historical binary
|
||||
# packages. We use it to construct a container image pinned to a
|
||||
# particular timestamp. snapshot.debian.org is known to be very slow,
|
||||
# but that shouldn't be a big problem for us because we only need that
|
||||
# site the first time.
|
||||
#
|
||||
# The mold executable created by this script is statically linked to
|
||||
# libstdc++, but dynamically linked to glibc, libm and a few other
|
||||
# libraries, as these libraries are almost always available on any Linux
|
||||
# system. We can't statically link glibc because doing so would disable
|
||||
# dlopen(), which is required to load the LTO linker plugin.
|
||||
#
|
||||
# We use a reasonably old Debian version for the build environment because
|
||||
# a binary dynamically linked against a newer version of glibc won't work
|
||||
# on a system with an older version of glibc.
|
||||
#
|
||||
# We prefer to build mold with Clang rather than GCC because mold's
|
||||
# Identical Code Folding works best with the LLVM address significance
|
||||
# table (.llvm_addrsig). Building a release binary with GCC produces a
|
||||
# slightly larger binary than with Clang.
|
||||
#
|
||||
# We need a recent version of Clang to build mold. If it's not available
|
||||
# via apt-get, we'll build it ourselves.
|
||||
#
|
||||
# This script can be used to create non-native binaries (e.g., building
|
||||
# aarch64 binary on x86-64) because Podman automatically runs everything
|
||||
# under QEMU if the container image is not native. To use this script for
|
||||
# non-native builds, you may need to install the qemu-user-static package.
|
||||
|
||||
set -e -x
|
||||
cd "$(dirname $0)"
|
||||
|
||||
usage() {
|
||||
echo "Usage: $0 [ x86_64 | aarch64 | arm | riscv64 | ppc64le | s390x | loongarch64 ]"
|
||||
exit 1
|
||||
}
|
||||
|
||||
case $# in
|
||||
0)
|
||||
arch=$(uname -m)
|
||||
if [ $arch = arm64 ]; then
|
||||
arch=aarch64
|
||||
elif [[ $arch = arm* ]]; then
|
||||
arch=arm
|
||||
fi
|
||||
;;
|
||||
1)
|
||||
arch="$1"
|
||||
;;
|
||||
*)
|
||||
usage
|
||||
esac
|
||||
|
||||
# Create a Podman image.
|
||||
if [ "$GITHUB_REPOSITORY" = '' ]; then
|
||||
image=mold-builder-$arch
|
||||
image_build="podman build --arch $arch -t $image -"
|
||||
else
|
||||
# If this script is running on GitHub Actions, we want to cache
|
||||
# the created container image in GitHub's container repostiory.
|
||||
image=ghcr.io/$GITHUB_REPOSITORY/mold-builder-$arch
|
||||
image_build="podman build --arch $arch -t $image --output=type=registry --layers --cache-to $image --cache-from $image -"
|
||||
fi
|
||||
|
||||
case $arch in
|
||||
x86_64)
|
||||
# Debian 9 (Stretch) released in June 2017.
|
||||
#
|
||||
# We use a Google-provided mirror (gcr.io) instead of the official Docker
|
||||
# Hub (docker.io) because docker.io has a strict rate limit policy.
|
||||
#
|
||||
# The toolchain in Debian 9 is too old to build mold, so we rebuild it
|
||||
# from source. We download source archives from official sites and build
|
||||
# them locally, rather than downloading pre-built binaries from somewhere
|
||||
# else, to avoid relying on unverifiable third-party binary blobs. Podman
|
||||
# caches the result of each RUN command, so rebuilding is done only once
|
||||
# per host.
|
||||
cat <<EOF | $image_build
|
||||
FROM mirror.gcr.io/library/debian:stretch@sha256:c5c5200ff1e9c73ffbf188b4a67eb1c91531b644856b4aefe86a58d2f0cb05be
|
||||
ENV DEBIAN_FRONTEND=noninteractive TZ=UTC
|
||||
RUN sed -i -e '/^deb/d' -e 's/^# deb /deb /g' /etc/apt/sources.list && \
|
||||
echo 'Acquire::Retries "10"; Acquire::http::timeout "10"; Acquire::Check-Valid-Until "false";' > /etc/apt/apt.conf.d/80-retries && \
|
||||
apt-get update && \
|
||||
apt-get install -y --no-install-recommends wget file make gcc g++ zlib1g-dev libssl-dev ca-certificates && \
|
||||
rm -rf /var/lib/apt/lists
|
||||
|
||||
# Build CMake 3.27
|
||||
RUN mkdir /build && \
|
||||
cd /build && \
|
||||
wget -O- --progress=dot:mega https://cmake.org/files/v3.27/cmake-3.27.7.tar.gz | tar xzf - --strip-components=1 && \
|
||||
./bootstrap --parallel=\$(nproc) && \
|
||||
make -j\$(nproc) && \
|
||||
make install && \
|
||||
rm -rf /build
|
||||
|
||||
# Build GCC 14
|
||||
RUN mkdir /build && \
|
||||
cd /build && \
|
||||
wget -O- --progress=dot:mega https://ftpmirror.gnu.org/gcc/gcc-14.2.0/gcc-14.2.0.tar.gz | tar xzf - --strip-components=1 && \
|
||||
mkdir gmp mpc mpfr && \
|
||||
wget -O- --progress=dot:mega https://ftpmirror.gnu.org/gmp/gmp-6.3.0.tar.gz | tar xzf - --strip-components=1 -C gmp && \
|
||||
wget -O- --progress=dot:mega https://ftpmirror.gnu.org/mpc/mpc-1.3.1.tar.gz | tar xzf - --strip-components=1 -C mpc && \
|
||||
wget -O- --progress=dot:mega https://ftpmirror.gnu.org/mpfr/mpfr-4.2.1.tar.gz | tar xzf - --strip-components=1 -C mpfr && \
|
||||
./configure --prefix=/usr --enable-languages=c,c++ --disable-bootstrap --disable-multilib && \
|
||||
make -j\$(nproc) && \
|
||||
make install && \
|
||||
ln -sf /usr/lib64/libstdc++.so.6 /usr/lib/x86_64-linux-gnu/libstdc++.so.6 && \
|
||||
rm -rf /build
|
||||
|
||||
# Build GNU binutils 2.43
|
||||
RUN mkdir /build && \
|
||||
cd /build && \
|
||||
wget -O- --progress=dot:mega https://ftpmirror.gnu.org/binutils/binutils-2.43.tar.gz | tar xzf - --strip-components=1 && \
|
||||
./configure --prefix=/usr && \
|
||||
make -j\$(nproc) && \
|
||||
make install && \
|
||||
rm -fr /build
|
||||
|
||||
# Build Python 3.12.7
|
||||
RUN mkdir /build && \
|
||||
cd /build && \
|
||||
wget -O- --progress=dot:mega https://www.python.org/ftp/python/3.12.7/Python-3.12.7.tgz | tar xzf - --strip-components=1 && \
|
||||
./configure && \
|
||||
make -j\$(nproc) && \
|
||||
make install && \
|
||||
rm -rf /build
|
||||
|
||||
# Build LLVM 20
|
||||
RUN mkdir /build && \
|
||||
cd /build && \
|
||||
wget -O- --progress=dot:mega https://github.com/llvm/llvm-project/archive/refs/tags/llvmorg-20.1.3.tar.gz | tar xzf - --strip-components=1 && \
|
||||
mkdir b && \
|
||||
cd b && \
|
||||
cmake -DCMAKE_BUILD_TYPE=Release -DLLVM_ENABLE_PROJECTS=clang ../llvm && \
|
||||
cmake --build . -j\$(nproc) && \
|
||||
cmake --install . --strip && \
|
||||
rm -rf /build
|
||||
EOF
|
||||
;;
|
||||
aarch64 | arm | ppc64le | s390x)
|
||||
# Debian 11 (Bullseye) released in August 2021
|
||||
#
|
||||
# We don't want to build Clang for these targets on QEMU becuase it
|
||||
# would take an extremely long time. Also, I believe old Linux boxes
|
||||
# are typically x86-64.
|
||||
cat <<EOF | $image_build
|
||||
FROM mirror.gcr.io/library/debian:bullseye-20240904@sha256:8ccc486c29a3ad02ad5af7f1156e2152dff3ba5634eec9be375269ef123457d8
|
||||
ENV DEBIAN_FRONTEND=noninteractive TZ=UTC
|
||||
RUN sed -i -e '/^deb/d' -e 's/^# deb /deb /g' /etc/apt/sources.list && \
|
||||
echo 'Acquire::Retries "10"; Acquire::http::timeout "10"; Acquire::Check-Valid-Until "false";' > /etc/apt/apt.conf.d/80-retries && \
|
||||
apt-get update && \
|
||||
apt-get install -y --no-install-recommends build-essential gcc-10 g++-10 clang-16 cmake && \
|
||||
ln -sf /usr/bin/clang-16 /usr/bin/clang && \
|
||||
ln -sf /usr/bin/clang++-16 /usr/bin/clang++ && \
|
||||
rm -rf /var/lib/apt/lists
|
||||
EOF
|
||||
;;
|
||||
riscv64)
|
||||
cat <<EOF | $image_build
|
||||
FROM mirror.gcr.io/riscv64/debian:unstable-20240926@sha256:25654919c2926f38952cdd14b3300d83d13f2d820715f78c9f4b7a1d9399bf48
|
||||
ENV DEBIAN_FRONTEND=noninteractive TZ=UTC
|
||||
RUN sed -i -e '/^URIs/d' -e 's/^# http/URIs: http/' /etc/apt/sources.list.d/debian.sources && \
|
||||
echo 'Acquire::Retries "10"; Acquire::http::timeout "10"; Acquire::Check-Valid-Until "false";' > /etc/apt/apt.conf.d/80-retries && \
|
||||
apt-get update && \
|
||||
apt-get install -y --no-install-recommends build-essential gcc-14 g++-14 clang-18 cmake && \
|
||||
ln -sf /usr/bin/clang-18 /usr/bin/clang && \
|
||||
ln -sf /usr/bin/clang++-18 /usr/bin/clang++ && \
|
||||
rm -rf /var/lib/apt/lists
|
||||
EOF
|
||||
;;
|
||||
loongarch64)
|
||||
cat <<EOF | $image_build
|
||||
FROM mirror.gcr.io/loongarch64/debian:sid@sha256:0356df4e494bbb86bb469377a00789a5b42bbf67d5ff649a3f9721b745cbef77
|
||||
ENV DEBIAN_FRONTEND=noninteractive TZ=UTC
|
||||
RUN sed -i -e 's!http[^ ]*!http://snapshot.debian.org/archive/debian-ports/20250620T014755Z!g' /etc/apt/sources.list && \
|
||||
echo 'Acquire::Retries "10"; Acquire::http::timeout "10"; Acquire::Check-Valid-Until "false";' > /etc/apt/apt.conf.d/80-retries && \
|
||||
apt-get update && \
|
||||
apt-get install -y --no-install-recommends build-essential gcc-14 g++-14 clang-19 cmake && \
|
||||
ln -sf /usr/bin/clang-19 /usr/bin/clang && \
|
||||
ln -sf /usr/bin/clang++-19 /usr/bin/clang++ && \
|
||||
rm -rf /var/lib/apt/lists
|
||||
EOF
|
||||
;;
|
||||
*)
|
||||
usage
|
||||
;;
|
||||
esac
|
||||
|
||||
version=$(sed -n 's/^project(mold VERSION \(.*\))/\1/p' CMakeLists.txt)
|
||||
dest=mold-$version-$arch-linux
|
||||
|
||||
# Source tarballs available on GitHub don't contain .git directory.
|
||||
# Clone the repo if missing.
|
||||
[ -d .git ] || git clone --branch v$version --depth 1 --bare https://github.com/rui314/mold .git
|
||||
|
||||
# We use the timestamp of the last Git commit as the file timestamp
|
||||
# for build artifacts.
|
||||
timestamp=$(git log -1 --format=%ct)
|
||||
|
||||
# `uname -m` in an ARM32 container running on an ARM64 host reports it
|
||||
# not as ARM32 but as ARM64. That confuses BLAKE3's cmake script and
|
||||
# erroneously enables NEON SIMD instructions. `setarch` can be used to
|
||||
# change the output of `uname -m`.
|
||||
setarch=
|
||||
[ $arch = arm ] && setarch='setarch linux32'
|
||||
|
||||
mkdir -p dist
|
||||
|
||||
# Build mold in a container.
|
||||
#
|
||||
# SOURCE_DATE_EPOCH is a standardized environment variable that allows
|
||||
# build artifacts to appear as if they were built at a specific time.
|
||||
# We use it to control how the compiler expands the C/C++ __DATE__ and
|
||||
# __TIME__ macros.
|
||||
podman run --arch $arch -it --rm --userns=host --pids-limit=-1 --network=none \
|
||||
--pull=never -v "$(pwd):/mold:ro" -v "$(pwd)/dist:/dist" $image \
|
||||
$setarch bash -c "
|
||||
set -e
|
||||
export SOURCE_DATE_EPOCH=$timestamp
|
||||
mkdir /build
|
||||
cd /build
|
||||
cmake -DCMAKE_BUILD_TYPE=Release -DMOLD_MOSTLY_STATIC=1 -DCMAKE_C_COMPILER=clang -DCMAKE_CXX_COMPILER=clang++ /mold
|
||||
cmake --build . -j\$(nproc)
|
||||
cmake --install .
|
||||
cmake -DMOLD_USE_MOLD=1 .
|
||||
cmake --build . -j\$(nproc)
|
||||
ctest --output-on-failure -j\$(nproc)
|
||||
cmake --install . --prefix $dest --strip
|
||||
find $dest -print | xargs touch --no-dereference --date=@$timestamp
|
||||
find $dest -print | sort | tar -cf - --no-recursion --files-from=- | gzip -9nc > /dist/$dest.tar.gz
|
||||
cp mold /dist
|
||||
sha256sum /dist/$dest.tar.gz
|
||||
"
|
||||
+2
-2
@@ -31,7 +31,7 @@ If a program is statically-linked, there's no dynamic loader that
|
||||
rewrites the GOT entries. Therefore, if a program is
|
||||
statically-linked, a libc's startup routine does that on behalf of the
|
||||
dynamic loader. Concretely, a startup routine interprets all dynamic
|
||||
relocations between `__rela_iplt_start` and `__rela_iplt_stop`
|
||||
relocations between `__rela_iplt_start` and `__rela_iplt_start`
|
||||
symbols. It is linker's responsibility to emit dynamic relocations
|
||||
for IFUNC symbols even if it is linking a statically-linked program
|
||||
and mark the beginning and the ending of a `.rela.dyn` section with
|
||||
@@ -41,7 +41,7 @@ The bug was my linker didn't define `__rela_iplt_start` and
|
||||
`__rela_iplt_stop` symbols. Since these symbols are weak, they are
|
||||
initialized to zero. From the point of the initializer function,
|
||||
there's no dynamic relocations between `__rela_iplt_start` and
|
||||
`__rela_iplt_stop` symbols. That left GOT entries for IFUNC symbols
|
||||
`__rela_iplt_start` symbols. That left GOT entries for IFUNC symbols
|
||||
untouched.
|
||||
|
||||
The proper fix was to emit dynamic relocations for IFUNC symbols and
|
||||
|
||||
@@ -1 +0,0 @@
|
||||
<svg version="1.1" viewBox="0.0 0.0 600.0 371.0" fill="none" stroke="none" stroke-linecap="square" stroke-miterlimit="10" width="600" height="371" xmlns:xlink="http://www.w3.org/1999/xlink" xmlns="http://www.w3.org/2000/svg"><path fill="#ffffff" d="M0 0L600.0 0L600.0 371.0L0 371.0L0 0Z" fill-rule="nonzero"/><path stroke="#333333" stroke-width="1.0" stroke-linecap="butt" d="M82.5 275.5L581.5 275.5" fill-rule="nonzero"/><path stroke="#cccccc" stroke-width="1.0" stroke-linecap="butt" d="M82.5 231.5L581.5 231.5" fill-rule="nonzero"/><path stroke="#cccccc" stroke-width="1.0" stroke-linecap="butt" d="M82.5 186.5L581.5 186.5" fill-rule="nonzero"/><path stroke="#cccccc" stroke-width="1.0" stroke-linecap="butt" d="M82.5 142.5L581.5 142.5" fill-rule="nonzero"/><path stroke="#cccccc" stroke-width="1.0" stroke-linecap="butt" d="M82.5 98.5L581.5 98.5" fill-rule="nonzero"/><path stroke="#cccccc" stroke-width="1.0" stroke-linecap="butt" d="M82.5 53.5L581.5 53.5" fill-rule="nonzero"/><clipPath id="id_0"><path d="M82.55 53.916668L581.45 53.916668L581.45 275.45L82.55 275.45L82.55 53.916668Z" clip-rule="nonzero"/></clipPath><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M158.0 275.0L136.0 275.0L136.0 229.0C136.0 227.89543 136.89543 227.0 138.0 227.0L156.0 227.0C157.10457 227.0 158.0 227.89543 158.0 229.0Z" fill-rule="nonzero"/><path fill="#4285f4" clip-path="url(#id_0)" d="M158.0 275.0L136.0 275.0L136.0 229.0C136.0 227.89543 136.89543 227.0 138.0 227.0L156.0 227.0C157.10457 227.0 158.0 227.89543 158.0 229.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M307.0 275.0L285.0 275.0L285.0 91.0C285.0 89.89543 285.89542 89.0 287.0 89.0L305.0 89.0C306.10458 89.0 307.0 89.89543 307.0 91.0Z" fill-rule="nonzero"/><path fill="#4285f4" clip-path="url(#id_0)" d="M307.0 275.0L285.0 275.0L285.0 91.0C285.0 89.89543 285.89542 89.0 287.0 89.0L305.0 89.0C306.10458 89.0 307.0 89.89543 307.0 91.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M182.0 275.0L160.0 275.0L160.0 244.0C160.0 242.89543 160.89543 242.0 162.0 242.0L180.0 242.0C181.10457 242.0 182.0 242.89543 182.0 244.0Z" fill-rule="nonzero"/><path fill="#ea4335" clip-path="url(#id_0)" d="M182.0 275.0L160.0 275.0L160.0 244.0C160.0 242.89543 160.89543 242.0 162.0 242.0L180.0 242.0C181.10457 242.0 182.0 242.89543 182.0 244.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M331.0 275.0L309.0 275.0L309.0 131.0C309.0 129.89543 309.89542 129.0 311.0 129.0L329.0 129.0C330.10458 129.0 331.0 129.89543 331.0 131.0Z" fill-rule="nonzero"/><path fill="#ea4335" clip-path="url(#id_0)" d="M331.0 275.0L309.0 275.0L309.0 131.0C309.0 129.89543 309.89542 129.0 311.0 129.0L329.0 129.0C330.10458 129.0 331.0 129.89543 331.0 131.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M480.0 275.0L458.0 275.0L458.0 156.0C458.0 154.89543 458.89542 154.0 460.0 154.0L478.0 154.0C479.10458 154.0 480.0 154.89543 480.0 156.0Z" fill-rule="nonzero"/><path fill="#ea4335" clip-path="url(#id_0)" d="M480.0 275.0L458.0 275.0L458.0 156.0C458.0 154.89543 458.89542 154.0 460.0 154.0L478.0 154.0C479.10458 154.0 480.0 154.89543 480.0 156.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M206.0 275.0L184.0 275.0L184.0 270.0C184.0 268.89542 184.89543 268.0 186.0 268.0L204.0 268.0C205.10457 268.0 206.0 268.89542 206.0 270.0Z" fill-rule="nonzero"/><path fill="#fbbc04" clip-path="url(#id_0)" d="M206.0 275.0L184.0 275.0L184.0 270.0C184.0 268.89542 184.89543 268.0 186.0 268.0L204.0 268.0C205.10457 268.0 206.0 268.89542 206.0 270.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M355.0 275.0L333.0 275.0L333.0 254.0C333.0 252.89543 333.89542 252.0 335.0 252.0L353.0 252.0C354.10458 252.0 355.0 252.89543 355.0 254.0Z" fill-rule="nonzero"/><path fill="#fbbc04" clip-path="url(#id_0)" d="M355.0 275.0L333.0 275.0L333.0 254.0C333.0 252.89543 333.89542 252.0 335.0 252.0L353.0 252.0C354.10458 252.0 355.0 252.89543 355.0 254.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroke-opacity="0.0" clip-path="url(#id_0)" d="M504.0 275.0L482.0 275.0L482.0 250.0C482.0 248.89543 482.89542 248.0 484.0 248.0L502.0 248.0C503.10458 248.0 504.0 248.89543 504.0 250.0Z" fill-rule="nonzero"/><path fill="#fbbc04" clip-path="url(#id_0)" d="M504.0 275.0L482.0 275.0L482.0 250.0C482.0 248.89543 482.89542 248.0 484.0 248.0L502.0 248.0C503.10458 248.0 504.0 248.89543 504.0 250.0Z" fill-rule="nonzero"/><path stroke="#000000" stroke-width="2.0" stroke-linecap="butt" stroLine truncated
|
||||
|
Before Width: | Height: | Size: 99 KiB |
@@ -1,46 +0,0 @@
|
||||
# Coding Guidelines
|
||||
|
||||
mold is written in C++20, but as is the case with every C++ project,
|
||||
it has local coding rules. In this document, I'll explain some of them
|
||||
and try to give justifications for why I chose such rules.
|
||||
|
||||
## DOs
|
||||
|
||||
- Always use `i64` (which is a type alias for `int64_t` in mold) for
|
||||
integers unless you have a reason not to. For example, even if you know
|
||||
that a loop counter won't exceed 100, you should stop thinking about it
|
||||
and just use `i64`.
|
||||
|
||||
Justification: Local variables are usually on CPU registers, so on
|
||||
64-bit CPUs, there's no performance peanlty on choosing `i64` over
|
||||
`i32`. Even if a compiler has to spill register values to the stack,
|
||||
I don't think there's an observable difference between `i32` and
|
||||
`i64`. Therefore, extra 32 bits are essentially free. On 32-bit CPUs,
|
||||
they are not free, but that's OK because we are writing mold for modern
|
||||
computers. mold will still run on 32-bit computers but a bit slowly.
|
||||
|
||||
By always using `i64`, we can eliminate the need to think about the
|
||||
"right" size for each variable. It also reduces the risk of integer
|
||||
overflow.
|
||||
|
||||
Exceptions: If you have to allocate a very large number (e.g. millions)
|
||||
of the same object, its size matters. In that case, use a smaller type.
|
||||
|
||||
## DON'Ts
|
||||
|
||||
- Don't use `auto` unless its actual type is obvious in the very narrow
|
||||
context. Currently, we use `auto` only for lambdas.
|
||||
|
||||
Justification: I think `auto` makes code writing easier but code reading
|
||||
harder, because readers have to make a guess as to what is the actual
|
||||
type of `auto`. If you are already familiar with the existing codebase,
|
||||
you may be able to guess it easily, but that's not always the case.
|
||||
I want to keep the mold codebase friendly to first-time readers.
|
||||
|
||||
- Don't over-use inheritance. In mold, most classes don't have parents,
|
||||
and even if they do, their class hierarchy is very shallow. Currently
|
||||
its height is just two (i.e. abstract classes and their implementations).
|
||||
|
||||
Justification: Designing class hierarchies is fun as it feels like
|
||||
taxonomy, but I don't think that always help writing code. It looks like
|
||||
simpler class hierarchy makes its code simpler.
|
||||
Binary file not shown.
|
Before Width: | Height: | Size: 23 KiB |
-491
@@ -1,491 +0,0 @@
|
||||
[This document was written in 2020, and the contents are outdated.
|
||||
Specifically, we no longer believe that object preloading is a good
|
||||
idea. That being said, most of the points in this document still hold
|
||||
even today. Therefore, I'll keep this document as-is.]
|
||||
|
||||
## Design and implementation of mold
|
||||
|
||||
For the rest of this documentation, I'll explain the design and the
|
||||
implementation of mold. If you are only interested in using mold, you
|
||||
don't need to read the below.
|
||||
|
||||
### Motivation
|
||||
|
||||
Here is why I'm writing a new linker:
|
||||
|
||||
- Even though lld has significantly improved the situation, linking is
|
||||
still one of the slowest steps in a build. It is especially
|
||||
annoying when I changed one line of code and had to wait for a few
|
||||
seconds or even more for a linker to complete. It should be
|
||||
instantaneous. There's a need for a faster linker.
|
||||
|
||||
- The number of cores on a PC has increased a lot lately, and this
|
||||
trend is expected to continue. However, the existing linkers can't
|
||||
take the advantage of the trend because they don't scale well for more
|
||||
cores. I have a 64-core/128-thread machine, so my goal is to create
|
||||
a linker that uses the CPU nicely. mold should be much faster than
|
||||
other linkers on 4 or 8-core machines too, though.
|
||||
|
||||
- It looks to me that the designs of the existing linkers are somewhat
|
||||
too similar, and I believe there are a lot of drastically different
|
||||
designs that haven't been explored yet. Developers generally don't
|
||||
care about linkers as long as they work correctly, and they don't
|
||||
even think about creating a new one. So there may be lots of low
|
||||
hanging fruits there in this area.
|
||||
|
||||
### Basic design
|
||||
|
||||
- In order to achieve a `cp`-like performance, the most important
|
||||
thing is to fix the layout of an output file as quickly as possible, so
|
||||
that we can start copying actual data from input object files to an
|
||||
output file as soon as possible.
|
||||
|
||||
- Copying data from input files to an output file is I/O-bounded, so
|
||||
there should be room for doing computationally-intensive tasks while
|
||||
copying data from one file to another.
|
||||
|
||||
- We should allow the linker to preload object files from disk and
|
||||
parse them in memory before a complete set of input object files
|
||||
is ready. To do so, we need
|
||||
to split the linker into two in such a way that the latter half of
|
||||
the process finishes as quickly as possible by speculatively parsing
|
||||
and preprocessing input files in the first half of the process.
|
||||
|
||||
- One of the most computationally-intensive stage among linker stages
|
||||
is symbol resolution. To resolve symbols, we basically have to throw
|
||||
all symbol strings into a hash table to match undefined symbols with
|
||||
defined symbols. But this can be done in the preloading stage using
|
||||
[string interning](https://en.wikipedia.org/wiki/String_interning).
|
||||
|
||||
- Object files may contain a special section called a mergeable string
|
||||
section. The section contains lots of null-terminated strings, and
|
||||
the linker is expected to gather all mergeable string sections and
|
||||
merge their contents. So, if two object files contain the same
|
||||
string literal, for example, the resulting output will contain a
|
||||
single merged string. This step is computationally intensive, but string
|
||||
merging can be done in the preloading stage using string interning.
|
||||
|
||||
- Static archives (.a files) contain object files, but the static
|
||||
archive's string table contains only defined symbols of member
|
||||
object files and lacks other types of symbols. That makes static
|
||||
archives unsuitable for speculative parsing. Therefore, the linker
|
||||
should ignore the symbol table of static archive and directly read
|
||||
static archive members.
|
||||
|
||||
- If there's a relocation that uses a GOT of a symbol, then we have to
|
||||
create a GOT entry for that symbol. Otherwise, we shouldn't. That
|
||||
means we need to scan all relocation tables to fix the length and
|
||||
the contents of a .got section. This is computationally intensive,
|
||||
but this step is parallelizable.
|
||||
|
||||
### Linker Script
|
||||
|
||||
Linker script is an embedded language for the linker. It is mainly
|
||||
used to control how input sections are mapped to output sections and
|
||||
the layout of the output, but it can also do a lot of tricky stuff.
|
||||
Its feature is useful especially for embedded programming, but it's
|
||||
also an awfully underdocumented and complex language.
|
||||
|
||||
We have to implement a subset of the linker script language anwyay,
|
||||
because on Linux, /usr/lib/x86_64-linux-gnu/libc.so is (despite its
|
||||
name) not a shared object file but actually an ASCII file containing
|
||||
linker script code to load the _actual_ libc.so file. But the feature
|
||||
set for this purpose is very limited, and it is okay to implement them
|
||||
to mold.
|
||||
|
||||
Besides that, we really don't want to implement the linker script
|
||||
language. But at the same time, we want to satisfy the user needs that
|
||||
are currently satisfied with the linker script language. So, what
|
||||
should we do? Here is my observation:
|
||||
|
||||
- Linker script allows doing a lot of tricky stuff, such as specifying
|
||||
the exact layout of a file, inserting arbitrary bytes between
|
||||
sections, etc. But most of them can be done with a post-link binary
|
||||
editing tool (such as `objcopy`).
|
||||
|
||||
- It looks like there are two things that truly cannot be done by a
|
||||
post-link editing tool: (a) mapping input sections to output
|
||||
sections, and (b) applying relocations.
|
||||
|
||||
From the above observation, I believe we need to provide only the
|
||||
following features instead of the entire linker script language:
|
||||
|
||||
- A method to specify how input sections are mapped to output
|
||||
sections, and
|
||||
|
||||
- a method to set addresses to output sections, so that relocations
|
||||
are applied based on desired addresses.
|
||||
|
||||
I believe everything else can be done with a post-link binary editing
|
||||
tool.
|
||||
|
||||
### Details
|
||||
|
||||
- As we aim to the 1-second goal for Chromium, every millisecond
|
||||
counts. We can't ignore the latency of process exit. If we mmap a
|
||||
lot of files, \_exit(2) is not instantaneous but takes a few hundred
|
||||
milliseconds because the kernel has to clean up a lot of
|
||||
resources. As a workaround, we should organize the linker command as
|
||||
two processes; the first process forks the second process, and the
|
||||
second process does the actual work. As soon as the second process
|
||||
writes a result file to a filesystem, it notifies the first process,
|
||||
and the first process exits. The second process can take time to
|
||||
exit, because it is not an interactive process.
|
||||
|
||||
- At least on Linux, it looks like the filesystem's performance to
|
||||
allocate new blocks to a new file is the limiting factor when
|
||||
creating a new large file and filling its contents using mmap.
|
||||
If you already have a large file in the buffer cache, writing to it is
|
||||
much faster than creating a new fresh file and writing to it.
|
||||
Based on this observation, mold overwrites to an existing
|
||||
executable file if exists. My quick benchmark showed that I could
|
||||
save 300 milliseconds when creating a 2 GiB output file.
|
||||
Linux doesn't allow to open an executable for writing if it is
|
||||
running (you'll get a "text busy" error if you attempt). mold
|
||||
falls back to the usual way if it fails to open an output file.
|
||||
|
||||
- The output from the linker should be deterministic for the sake of
|
||||
[build reproducibility](https://en.wikipedia.org/wiki/Reproducible_builds)
|
||||
and ease of debugging. This might add a little bit of overhead to
|
||||
the linker, but that shouldn't be too much.
|
||||
|
||||
- A .build-id, a unique ID embedded to an output file, is usually
|
||||
computed by applying a cryptographic hash function (e.g. SHA-1) to
|
||||
an output file. This is a slow step, but we can speed it up by
|
||||
splitting a file into small chunks, computing SHA-1 for each chunk,
|
||||
and then computing SHA-1 of the concatenated SHA-1 hashes
|
||||
(i.e. constructing a [Merkle
|
||||
Tree](https://en.wikipedia.org/wiki/Merkle_tree) of height 2).
|
||||
Modern x86 processors have purpose-built instructions for SHA-1 and
|
||||
can compute SHA-1 pretty quickly at about 2 GiB/s. Using 16
|
||||
cores, a build-id for a 2 GiB executable can be computed in 60 to 70
|
||||
milliseconds.
|
||||
|
||||
- BFD, gold, and lld support section garbage collection. That is, a
|
||||
linker runs a mark-sweep garbage collection on an input graph, where
|
||||
sections are vertices and relocations are edges, to discard all
|
||||
sections that are not reachable from the entry point symbol
|
||||
(i.e. `_start`) or a few other root sections. In mold, we are using
|
||||
multiple threads to mark sections concurrently.
|
||||
|
||||
- Similarly, BFD, gold an lld support Identical Comdat Folding (ICF)
|
||||
as yet another size optimization. ICF merges two or more read-only
|
||||
sections that happen to have the same contents and relocations.
|
||||
To do that, we have to find isomorphic subgraphs from larger graphs.
|
||||
I implemented a new algorithm for mold, which is 5x faster than lld
|
||||
to do ICF for Chromium (from 5 seconds to 1 second).
|
||||
|
||||
- [Intel Threading Building
|
||||
Blocks](https://github.com/oneapi-src/oneTBB) (TBB) is a good
|
||||
library for parallel execution and has several concurrent
|
||||
containers. We are particularly interested in using
|
||||
`parallel_for_each` and `concurrent_hash_map`.
|
||||
|
||||
- TBB provides `tbbmalloc` which works better for multi-threaded
|
||||
applications than the glib'c malloc, but it looks like
|
||||
[jemalloc](https://github.com/jemalloc/jemalloc) and
|
||||
[mimalloc](https://github.com/microsoft/mimalloc) are a little bit
|
||||
more scalable than `tbbmalloc`.
|
||||
|
||||
### Size of the problem
|
||||
|
||||
When linking Chrome, a linker reads 3,430,966,844 bytes of data in
|
||||
total. The data contains the following items:
|
||||
|
||||
| Data item | Number
|
||||
| ------------------------ | ------
|
||||
| Object files | 30,723
|
||||
| Public undefined symbols | 1,428,149
|
||||
| Mergeable strings | 1,579,996
|
||||
| Comdat groups | 9,914,510
|
||||
| Regular sections¹ | 10,345,314
|
||||
| Public defined symbols | 10,512,135
|
||||
| Symbols | 23,953,607
|
||||
| Sections | 27,543,225
|
||||
| Relocations against SHF_ALLOC sections | 39,496,375
|
||||
| Relocations | 62,024,719
|
||||
|
||||
¹ Sections that have to be copied from input object files to an
|
||||
output file. Sections that contain relocations or symbols are for
|
||||
example excluded.
|
||||
|
||||
### Internals
|
||||
|
||||
In this section, I'll explain the internals of mold linker.
|
||||
|
||||
#### A brief history of Unix and the Unix linker
|
||||
|
||||
Conceptually, what a linker does is pretty simple. A compiler compiles
|
||||
a fragment of a program (a single source file) into a fragment of
|
||||
machine code and data (an object file, which typically has the .o
|
||||
extension), and a linker stitches them together into a single
|
||||
executable or a shared library image.
|
||||
|
||||
In reality, modern linkers for Unix-like systems are much more
|
||||
complicated than the naive understanding because they have gradually
|
||||
gained one feature at a time over the 50 years history of Unix, and
|
||||
they are now something like a bag of lots of miscellaneous features in
|
||||
which none of the features is more important than the others. It is
|
||||
very easy to miss the forest for the trees, since for those who don't
|
||||
know the details of the Unix linker, it is not clear which feature is
|
||||
essential and which is not.
|
||||
|
||||
That being said, one thing is clear that at any point of Unix history,
|
||||
a Unix linker has a coherent feature set for the Unix of that age. So,
|
||||
let me entangle the history to see how the operating system, runtime,
|
||||
and linker have gained features that we see today. That should give
|
||||
you an idea of why a particular feature has been added to a linker in the
|
||||
first place.
|
||||
|
||||
1. Original Unix didn't support shared libraries, and a program was
|
||||
always loaded to a fixed address. An executable was something like
|
||||
a memory dump that was just loaded to a particular address by the
|
||||
kernel. After loading, the kernel started executing the program by
|
||||
setting the instruction pointer to a particular address.
|
||||
|
||||
The most essential feature for any linker is relocation processing.
|
||||
The original Unix linker of course supported that. Let me explain
|
||||
what that is.
|
||||
|
||||
Individual object files are inevitably incomplete as a program,
|
||||
because when a compiler created them, it only see a part of an
|
||||
entire program. For example, if an object file contains a function
|
||||
call that refers to another object file, the `call` instruction in the
|
||||
object cannot be complete, as the compiler has no idea as to what
|
||||
is the called function's address. To deal with this, the compiler
|
||||
emits a placeholder value (typically just zero) instead of a real
|
||||
address and leaves metadata in an object file saying "fix offset X
|
||||
of this file with an address of Y". That metadata is called
|
||||
"relocation". Relocations are typically processed by the linker.
|
||||
|
||||
It is easy for a linker to apply relocations for the original Unix
|
||||
because a program is always loaded to a fixed address. It exactly
|
||||
knows the addresses of all functions and data when linking a
|
||||
program.
|
||||
|
||||
Static library support, which is still an important feature of Unix
|
||||
linker, also dates back to this early period of Unix history.
|
||||
To understand what it is, imagine that you are trying to compile
|
||||
a program for the early Unix. You don't want to waste time to
|
||||
compile libc functions every time you compile your program (the
|
||||
computers of the era were incredibly slow), so you have already
|
||||
placed each libc function into a separate source file and compiled
|
||||
them individually. That means you have object files for each libc
|
||||
function, e.g., printf.o, scanf.o, atoi.o, write.o, etc.
|
||||
|
||||
Given this configuration, all you have to do to link your program
|
||||
against libc functions is to pick up the right set of libc object
|
||||
files and give them to the linker along with the object files of your
|
||||
program. But, keeping the linker command line in sync with the
|
||||
libc functions you are using in your program is bothersome. You can
|
||||
be conservative; you can specify all libc object files to the
|
||||
command line, but that leads to program bloat because the linker
|
||||
unconditionally link all object files given to it no matter whether
|
||||
they are used or not. So, a new feature was added to the linker to
|
||||
fix the problem. That is the static library, which is also called
|
||||
the archive file.
|
||||
|
||||
An archive file is just a bundle of object files, just like zip
|
||||
file but in an uncompressed form. An archive file typically has the
|
||||
.a file extension and named after its contents. For example, the
|
||||
archive file containing all libc objects is named `libc.a`.
|
||||
|
||||
If you pass an archive file along with other object files to the
|
||||
linker, the linker pulls out an object file from the archive _only
|
||||
when_ it is referenced by other object files. In other words,
|
||||
unlike object files directly given to a linker, object files
|
||||
wrapped in an archive are not linked to the output by default.
|
||||
An archive works as a supplement to complete your program.
|
||||
|
||||
Even today, you can still find a libc archive file. Run `ar t
|
||||
/usr/lib/x86_64-linux-gnu/libc.a` on Linux should give you a list
|
||||
of object files in the libc archive.
|
||||
|
||||
2. In the '80s, Sun Microsystems, a leading commercial Unix vendor at the
|
||||
time, added shared library support to their Unix variant, SunOS.
|
||||
|
||||
(This section is incomplete.)
|
||||
|
||||
### Concurrency strategy
|
||||
|
||||
In this section, I'll explain the high-level concurrency strategy of
|
||||
mold.
|
||||
|
||||
In most places, mold adopts data parallelism. That is, we have a huge
|
||||
number of pieces of data of the same kind, and we process each of them
|
||||
individually using parallel for-loop. For example, after identifying
|
||||
the exact set of input object files, we need to scan all relocation
|
||||
tables to determine the sizes of .got and .plt sections. We do that
|
||||
using a parallel for-loop. The granularity of parallel processing in
|
||||
this case is the relocation table.
|
||||
|
||||
Data parallelism is very efficient and scalable because there's no
|
||||
need for threads to communicate with each other while working on each
|
||||
element of data. In addition to that, data parallelism is easy to
|
||||
understand, as it is just a for-loop in which multiple iterations may
|
||||
be executed in parallel. We don't use high-level communication or
|
||||
synchronization mechanisms such as channels, futures, promises,
|
||||
latches or something like that in mold.
|
||||
|
||||
In some cases, we need to share a little bit of data between threads
|
||||
while executing a parallel for-loop. For example, the loop to scan
|
||||
relocations turns on "requires GOT" or "requires PLT" flags in a
|
||||
symbol. Symbol is a shared resource, and writing to them from multiple
|
||||
threads without synchronization is unsafe. To deal with it, we made
|
||||
the flag an atomic variable.
|
||||
|
||||
The other common pattern you can find in mold which is build on top of
|
||||
the parallel for-loop is the map-reduce pattern. That is, we run a
|
||||
parallel for-loop on a large data set to produce a small data set and
|
||||
process the small data set with a single thread. Let me take a
|
||||
build-id computation as an example. Build-id is typically computed by
|
||||
applying a cryptographic hash function such as SHA-1 on a linker's
|
||||
output file. To compute it, we first consider an output as a sequence
|
||||
of 1 MiB blocks and compute a SHA-1 hash for each block in parallel.
|
||||
Then, we concatenate the SHA-1 hashes and compute a SHA-1 hash on the
|
||||
hashes to get a final build-id.
|
||||
|
||||
Finally, we use concurrent hashmap at a few places in mold. Concurrent
|
||||
hashmap is a hashmap to which multiple threads can safely insert items
|
||||
in parallel. We use it in the symbol resolution stage, for example.
|
||||
To resolve symbols, we basically have to throw in all defined symbols
|
||||
into a hash table, so that we can find a matching defined symbol for
|
||||
an undefined symbol by name. We do the hash table insertion from a
|
||||
parallel for-loop which iterates over a list of input files.
|
||||
|
||||
Overall, even though mold is highly scalable, it succeeded to avoid
|
||||
complexties you often find in complex parallel programs. From high
|
||||
level, mold just serially executes the linker's internal passes one by
|
||||
one. Each pass is parallelized using parallel for-loops.
|
||||
|
||||
### Rejected ideas
|
||||
|
||||
In this section, I'll explain the alternative designs I currently do
|
||||
not plan to implement and why I turned them down.
|
||||
|
||||
- Placing variable-length sections at end of an output file and start
|
||||
copying file contents before fixing the output file layout
|
||||
|
||||
Idea: Fixing the layout of regular sections seems easy, and if we
|
||||
place them at beginning of a file, we can start copying their
|
||||
contents from their input files to an output file. While copying
|
||||
file contents, we can compute the sizes of variable-length sections
|
||||
such as .got or .plt and place them at end of the file.
|
||||
|
||||
Reason for rejection: I did not choose this design because I doubt
|
||||
if it could actually shorten link time and I think I don't need it
|
||||
anyway.
|
||||
|
||||
The linker has to de-duplicate comdat sections (i.e. inline
|
||||
functions that are included in multiple object files), so we
|
||||
cannot compute the layout of regular sections until we resolve all
|
||||
symbols and de-duplicate comdats. That takes a few hundred
|
||||
milliseconds. After that, we can compute the sizes of
|
||||
variable-length sections in less than 100 milliseconds. It's quite
|
||||
fast, so it doesn't seem to make much sense to proceed without
|
||||
fixing the final file layout.
|
||||
|
||||
The other reason to reject this idea is because there's good a
|
||||
chance for this idea to have a negative impact on linker's overall
|
||||
performance. If we copy file contents before fixing the layout, we
|
||||
can't apply relocations to them while copying because symbol
|
||||
addresses are not available yet. If we fix the file layout first, we
|
||||
can apply relocations while copying, which is effectively zero-cost
|
||||
due to a very good data locality. On the other hand, if we apply
|
||||
relocations long after we copy file contents, it's pretty expensive
|
||||
because section contents are very likely to have been evicted from
|
||||
CPU cache.
|
||||
|
||||
- Incremental linking
|
||||
|
||||
Idea: Incremental linking is a technique to patch a previous linker's
|
||||
output file so that only functions or data that are updated from the
|
||||
previous build are written to it. It is expected to significantly
|
||||
reduce the amount of data copied from input files to an output file
|
||||
and thus speed up linking. GNU BFD and gold linkers support it.
|
||||
|
||||
Reason for rejection: I turned it down because it (1) is
|
||||
complicated, (2) doesn't seem to speed it up that much and (3) has
|
||||
several practical issues. Let me explain each of them.
|
||||
|
||||
First, incremental linking for real C/C++ programs is not as easy as
|
||||
one might think. Let me take malloc as an example. malloc is usually
|
||||
defined by libc, but you can implement it in your program, and if
|
||||
that's the case, the symbol `malloc` will be resolved to your
|
||||
function instead of the one in libc. If you include a library that
|
||||
defines malloc (such as libjemalloc or libtbbmallc) before libc,
|
||||
their malloc will override libc's malloc.
|
||||
|
||||
Assume that you are using a nonstandard malloc. What if you remove
|
||||
your malloc from your code, or remove `-ljemalloc` from your
|
||||
Makefile? The linker has to include a malloc from libc, which may
|
||||
include more object files to satisfy its dependencies. Such code
|
||||
change can affect the entire program rather than just replacing one
|
||||
function. The same is true for adding malloc to your program. Making
|
||||
a local change doesn't necessarily result in a local change in the
|
||||
binary level. It can easily have cascading effects.
|
||||
|
||||
Some ELF fancy features make incremental linking even harder to
|
||||
implement. Take the weak symbol as an example. If you define `atoi`
|
||||
as a weak symbol in your program, and if you are not using `atoi`
|
||||
at all in your program, that symbol will be resolved to address
|
||||
0. But if you start using some libc function that indirectly calls
|
||||
`atoi`, then `atoi` will be included in your program, and your weak
|
||||
symbol will be resolved to that function. I don't know how to
|
||||
efficiently fix up a binary for this case.
|
||||
|
||||
This is a hard problem, so existing linkers don't try too hard to
|
||||
solve it. For example, IIRC, gold falls back to full link if any
|
||||
function is removed from a previous build. If you want to not annoy
|
||||
users in the fallback case, you need to make full link fast anyway.
|
||||
|
||||
Second, incremental linking itself has an overhead. It has to detect
|
||||
updated files, patch an existing output file and write additional
|
||||
data to an output file for future incremental linking. GNU gold, for
|
||||
instance, takes almost 30 seconds on my machine to do a null
|
||||
incremental link (i.e. no object files are updated from a previous
|
||||
build) for chrome. It's just too slow.
|
||||
|
||||
Third, there are other practical issues in incremental linking. It's
|
||||
not reproducible, so your binary isn't going to be the same as other
|
||||
binaries even if you are compiling the same source tree using the
|
||||
same compiler toolchain. Or, it is complex and there might be a bug
|
||||
in it. If something doesn't work correctly, "remove --incremental
|
||||
from your Makefile and try again" could be a piece of advice, but
|
||||
that isn't ideal.
|
||||
|
||||
So, all in all, incremental linking is tricky. I wanted to make full
|
||||
link as fast as possible, so that we don't have to think about how
|
||||
to work around the slowness of full link.
|
||||
|
||||
- Defining a completely new file format and use it
|
||||
|
||||
Idea: Sometimes, the ELF file format itself seems to be a limiting
|
||||
factor in improving the linker's performance. We might be able to make a
|
||||
far better one if we create a new file format.
|
||||
|
||||
Reason for rejection: I rejected the idea because it apparently has
|
||||
a practical issue (backward compatibility issue) and also doesn't
|
||||
seem to improve the performance of linkers that much. As clearly
|
||||
demonstrated by mold, we can create a fast linker for ELF. I believe
|
||||
ELF isn't that bad, after all. The semantics of the existing Unix
|
||||
linkers, such as the name resolution algorithm or the linker script,
|
||||
have slowed the linkers down, but that's not a problem of the file
|
||||
format itself.
|
||||
|
||||
- Watching object files using inotify(2)
|
||||
|
||||
Idea: When mold is running as a daemon for preloading, use
|
||||
inotify(2) to watch file system updates so that it can reload files
|
||||
as soon as they are updated.
|
||||
|
||||
Reason for rejection: Just like the maximum number of files you can
|
||||
simultaneously open, the maximum number of files you can watch using
|
||||
inotify(2) isn't that large. Maybe just a single instance of mold is
|
||||
fine with inotify(2), but it may fail if you run multiple of it.
|
||||
|
||||
The other reason for not doing it is because mold is quite fast
|
||||
without it anyway. Invoking stat(2) on each file for file update
|
||||
check takes less than 100 milliseconds for Chrome, and if most of
|
||||
the input files are not updated, parsing updated files takes almost
|
||||
no time.
|
||||
Binary file not shown.
|
Before Width: | Height: | Size: 51 KiB |
@@ -1,55 +0,0 @@
|
||||
This page explains the following warning message and how to fix it.
|
||||
mold emits this message when it sees an object file that may not be
|
||||
compatible with mold.
|
||||
|
||||
|
||||
```
|
||||
mold: warning: foo.o: this file may cause a segmentation fault because it requires an executable stack. See https://github.com/rui314/mold/tree/main/docs/execstack.md for more info.
|
||||
```
|
||||
|
||||
# Background
|
||||
|
||||
On modern computers, the stack area (to which local variables are
|
||||
stored) cannot contain executable code. If the control reaches the
|
||||
stack area, the CPU refuses to execute any code there and the program
|
||||
is usually terminated due to segmentation fault.
|
||||
|
||||
This is a security measure. The stack area used to be executable (old
|
||||
CPUs generally execute any code as long as it is in a readable memory
|
||||
region), but that provided an easy attack vector to a malicious user.
|
||||
They wrote executable code to the stack area using some buffer
|
||||
overflow bug and jumped there to run arbitrary code in a remote server
|
||||
process.
|
||||
|
||||
To prevent this type of attack, the stack area is no longer executable
|
||||
since the early 2000s. On Linux, the stack's executable-ness is
|
||||
controlled by a bit in an executable, and the loader respects that
|
||||
bit. The bit is set by the linker.
|
||||
|
||||
GCC had (and still has) a feature that depends on the executable
|
||||
stack, so they invented a way to tell the linker to mark the stack
|
||||
executable. Specifically, if an object file contains a
|
||||
`.note.GNU-stack` section with the `SHF_EXECSTACK` bit, GNU linker
|
||||
silently makes the stack of an output file executable.
|
||||
|
||||
But the GNU linker's behavior is dangerous. If you accidentally link
|
||||
an object file that has that marker section, the entire stack area
|
||||
silently becomes executable, disabling the security mechanism.
|
||||
|
||||
Therefore, mold simply ignores that marker section. If you are using
|
||||
mold, you need to explicitly pass `-z execstack` to the linker to make
|
||||
the stack executable.
|
||||
|
||||
# What caused this issue?
|
||||
|
||||
You are likely to use GCC's [Nested
|
||||
Functions](https://gcc.gnu.org/onlinedocs/gcc/Nested-Functions.html)
|
||||
feature which still depends on the executable stack.
|
||||
|
||||
# How to fix it?
|
||||
|
||||
If you know what you are doing, pass `-z execstack` to mold. Beware
|
||||
that this will significantly weaken your program's security.
|
||||
|
||||
If you don't want to pass `-z execstack`, rewrite your code so that
|
||||
your code does not depend on the executable stack.
|
||||
@@ -1,158 +0,0 @@
|
||||
The very concept of linking is simple: a compiler compiles a piece of
|
||||
source code into an object file (a file containing machine code), and
|
||||
a linker combines object files into a single executable or a shared
|
||||
library file. However, the actual implementation of the linker for
|
||||
modern systems is much more complicated because hardware, operating
|
||||
system, compiler and linker all have many more features.
|
||||
|
||||
In this file, I'll explain random topics in the glossary format that
|
||||
you need to understand to read mold code.
|
||||
|
||||
## DSO
|
||||
|
||||
A .so file. Short for Dynamic Shared Object. Often called as a
|
||||
shared library, a dynamic libray or a shared object as well.
|
||||
|
||||
An DSO contains common functions and data that are used by multiple
|
||||
executables and/or other DSOs. At runtime, a DSO is loaded to a
|
||||
contiguous region in the virtual address.
|
||||
|
||||
## Object file
|
||||
|
||||
A .o file. An object file contains machine code and data, but it
|
||||
cannot be executed because it's not self-contained. For example,
|
||||
if you compile a C source file containing a call of `printf`,
|
||||
the actual function code of `printf` is not included in the resulting
|
||||
object file. You include `stdio.h`, but that teaches the compiler
|
||||
only about `printf`'s type, and the compiler still don't know what
|
||||
`printf` actually does. Therefore, it cannot emit code for `printf`.
|
||||
|
||||
You need to link an object file with other object file or a shared
|
||||
library to make it exectuable.
|
||||
|
||||
## Virtual address space
|
||||
|
||||
A pointer has a value like 0x803020 which is an address of the
|
||||
pointee. But it doesn't mean that the pointee resides at the
|
||||
physical memory address 0x803020 on the computer. Modern CPUs
|
||||
contains so-called Mmeory Management Unit (MMU), and all access to
|
||||
the memory are first translated by MMU to the physical address.
|
||||
The address before translation is called the "virtual address".
|
||||
Unless you are doing the kernel programming, all addresses you
|
||||
handle are virtual addresses.
|
||||
|
||||
The OS kernel controls the MMU so that each process owns the entire
|
||||
virtual address space. So, even if two process uses the same virtual
|
||||
address, they don't conflict. They are mapped to different physical
|
||||
addresses.
|
||||
|
||||
The existence of MMU has several implications to the linker. First,
|
||||
we can link the main executable to a specific address. On process
|
||||
startup, there's no code or data in the virtual address space, so
|
||||
the mapping of the main executable always succeed. However, it's not
|
||||
true to DSOs because they are loaded after the main executable and
|
||||
possibly other DSOs. Therefore, shared libraries must be linked in a
|
||||
way that they can be loaded to any address in the virtual address
|
||||
space.
|
||||
|
||||
## Relocation
|
||||
|
||||
A piece of information for the linker as to how to link object files
|
||||
or a dynamic objects.
|
||||
|
||||
Object files can refer functions or data in other object files. For
|
||||
example, if you compile a function which calls a non-local function
|
||||
`foo`, the resulting code contains something like this:
|
||||
|
||||
```
|
||||
26: e8 00 00 00 00 callq 2b <bar+0xb>
|
||||
27: R_X86_64_PLT32 foo-0x4
|
||||
```
|
||||
|
||||
The above `callq` is the instruction to call a function at the
|
||||
machine code level. It's opcode is `0xe8` in x86-64, so the
|
||||
instruction begins with `0xe8`. The following four bytes are
|
||||
displacement; that is, the address of the branch target relative to
|
||||
the end of this `callq` instruction. Notice that the displacement is
|
||||
0. The compiler couldn't fill the displacement because it has no
|
||||
idea as to where `foo` will be at runtime. So, the compiler write 0
|
||||
as a placeholder and instead write a relocation `R_X86_64_PLT32`
|
||||
with `foo` as its associated symbol. The linker reads this
|
||||
relocation, computes the offsets between this call instruction and
|
||||
function `foo` and overwrite the placeholder value 0 with an actual
|
||||
displacement.
|
||||
|
||||
There are many different types of relocations. For example, if you
|
||||
want to fix up not with a displacement but with an absolute address
|
||||
of a symbol, you need to use `R_X86_64_ABS64` instead.
|
||||
|
||||
## Static library
|
||||
|
||||
A .a file. Often called as an archive file or just archive as well.
|
||||
|
||||
A static library is a container just like tar or zip. Actually,
|
||||
there's no technical reason to not use tar or (uncompressed) zip,
|
||||
but traditionally the .a file format is used by the linker.
|
||||
|
||||
A static library contains object files and can be passed to the
|
||||
linker along with other object files and/or archives.
|
||||
|
||||
A linker pulls out object files from an archive only if it is needed
|
||||
to resolve undefined symbols. In other words, object files in an
|
||||
archive are not linked by default and used as a complement to supply
|
||||
missing definitions. This is ideal for a library because you don't
|
||||
want to link library functions unless you are actually using them.
|
||||
|
||||
Contrary to archive files, object files directly given to a linker
|
||||
are always linked to the output.
|
||||
|
||||
To maximize the benefit of archive files, a library often used as a
|
||||
static library is broken down to small files to separate each
|
||||
function individually (for example, look at
|
||||
https://git.musl-libc.org/cgit/musl/tree/src/stdio). By doing this,
|
||||
you import only used functions.
|
||||
|
||||
A static file is created by `ar`, whose command line arguments are
|
||||
similar to `tar`. A static library contains the symbol table which
|
||||
offers a quick way to look up an object file for a defined symbol,
|
||||
but mold does not use the static library's symbol table. mold
|
||||
doesdn't need a symbol table to exist in an archive, and if exists,
|
||||
mold just ignores it.
|
||||
|
||||
See also: DSO (dynamic library)
|
||||
|
||||
## Symbol
|
||||
|
||||
A symbol is a label assigned to a specific location in an input file
|
||||
or an output file. For example, if you define function `foo` and
|
||||
compile it, the resulting object file contains a symbol `foo`
|
||||
pointing to the beginning of the machine code for `foo`.
|
||||
|
||||
Usually, a symbol name is a function or a variable name. If an
|
||||
object is anonymous (such the one for a string literal), a compiler
|
||||
generated a unique symbol, which often starts with `.` to avoid
|
||||
conflict with user-defined symbols.
|
||||
|
||||
For C++, symbol name is a complex "mangled" name. We need to mangle
|
||||
identifiers because a simple name such as `foo` cannot be uniquely
|
||||
identify a function or a data in C++, because for example `foo` may
|
||||
be in a namespace or defined as a static member in some class. If
|
||||
`foo` is an overloaded function, we need to distinguish different
|
||||
`foo`s by its type. Therefore, C++ compiler mangles an identifier by
|
||||
appending nmaepsace names, type information and such so that
|
||||
different things get different names.
|
||||
|
||||
For example, a function `int foo(int)` in a namespace `bar` is
|
||||
mangled as `_ZN3bar3fooEi`.
|
||||
|
||||
A symbol can be either defined or undefined. A defined symbol points
|
||||
to some location in a file which may contain the function's machine
|
||||
code or the variable's initial value. An undefined symbol does not
|
||||
point to anywhere. It needs to be merged with a defined symbol with
|
||||
the same name at link-time. This merging process is called "name
|
||||
resolution".
|
||||
|
||||
For example, if your program is using `printf`, it usually contains
|
||||
`printf` as an undefined symbol. You need to link it with `libc.a`
|
||||
or `libc.so`, which contain a defined symbol of `printf`, to make a
|
||||
complete program.
|
||||
@@ -1,179 +0,0 @@
|
||||
# Instrumenting mold with MemorySanitizer
|
||||
|
||||
## Introduction
|
||||
|
||||
Per <https://github.com/google/sanitizers/wiki/MemorySanitizerLibcxxHowTo>:
|
||||
|
||||
> If you want MemorySanitizer to work properly and not produce any false
|
||||
> positives, you must ensure that all the code in your program and in
|
||||
> libraries it uses is instrumented (i.e. built with `-fsanitize=memory`).
|
||||
> In particular, you would need to link against an MSan-instrumented C++
|
||||
> standard library. We recommend to use [libc++](https://libcxx.llvm.org/)
|
||||
> for that purpose.
|
||||
|
||||
## Building instrumented libc++
|
||||
|
||||
Build an MSan-instrumented libc++ from source:
|
||||
|
||||
```sh
|
||||
cd ~
|
||||
git clone https://github.com/llvm/llvm-project
|
||||
cd llvm-project
|
||||
cmake -S ./runtimes -B build-libcxx -G Ninja \
|
||||
-DCMAKE_BUILD_TYPE=Release \
|
||||
-DLLVM_ENABLE_RUNTIMES="libcxx;libcxxabi" \
|
||||
-DCMAKE_C_COMPILER=clang \
|
||||
-DCMAKE_CXX_COMPILER=clang++ \
|
||||
-DLLVM_USE_SANITIZER=MemoryWithOrigins
|
||||
cmake --build build-libcxx -- cxx cxxabi
|
||||
export LIBCXX="$HOME/llvm-project/build-libcxx" # for subsequent build steps
|
||||
```
|
||||
|
||||
Upon success, `./build-libcxx/{include,lib}` will contain the resulting
|
||||
headers and shared libraries.
|
||||
|
||||
## Linking mold against instrumented libc++
|
||||
|
||||
Use the `MOLD_USE_MSAN` and `MOLD_STDLIB_PREFIX` cmake variables to link
|
||||
mold against the instrumented build of libc++:
|
||||
|
||||
```sh
|
||||
cd ~
|
||||
git clone https://github.com/rui314/mold.git
|
||||
cd mold
|
||||
cmake -B build -G Ninja \
|
||||
-DCMAKE_BUILD_TYPE=Debug \
|
||||
-DCMAKE_C_COMPILER=clang \
|
||||
-DCMAKE_CXX_COMPILER=clang++ \
|
||||
-DMOLD_USE_MSAN=ON \
|
||||
-DMOLD_STDLIB_PREFIX="$LIBCXX" \
|
||||
-DMOLD_USE_MIMALLOC=OFF \
|
||||
-DMOLD_USE_SYSTEM_TBB=ON
|
||||
cmake --build build
|
||||
```
|
||||
|
||||
Most of mold's tests (except those for `-flto`) should work at this
|
||||
point. Run them like normal:
|
||||
|
||||
```sh
|
||||
ctest --test-dir build
|
||||
```
|
||||
|
||||
Any resulting MemorySanitizer errors should be visible in
|
||||
`./build/Temporary/Testing/LastTest.log`.
|
||||
|
||||
## Building instrumented LTO plugin (experimental)
|
||||
|
||||
Exercising `-flto` with MemorySanitizer (and without false positives)
|
||||
requires instrumenting the transitive dependencies of `lto-unix.cc`:
|
||||
|
||||
- `libiberty.a`
|
||||
- `liblto_plugin.so`
|
||||
- `LLVMgold.so`
|
||||
|
||||
> [!NOTE]
|
||||
> This is more involved and time-consuming than building only libc++ from
|
||||
> source. The cost-benefit ratio of this additional instrumentation and
|
||||
> test coverage may be unfavorable in many cases.
|
||||
|
||||
> [!IMPORTANT]
|
||||
> The following steps are experimental and unlikely to work exactly as-is
|
||||
> under CI runners and individual developer environments. Consider the
|
||||
> following a starting point rather than a complete HOWTO.
|
||||
|
||||
Build an MSan-instrumented GNU libiberty:
|
||||
|
||||
```sh
|
||||
cd ~
|
||||
git clone [email protected]:gcc-mirror/gcc.git
|
||||
cd gcc/libiberty
|
||||
export CC=clang
|
||||
export CFLAGS="-g -Og -fsanitize=memory -fsanitize-memory-track-origins"
|
||||
export LDFLAGS="-fsanitize=memory"
|
||||
./configure
|
||||
make -j$(nproc)
|
||||
unset CC CFLAGS LDFLAGS
|
||||
```
|
||||
|
||||
Install the resulting `./libiberty.a` into the build toolchain being
|
||||
tested with mold. If the system image is ephemeral or disposable (e.g. a
|
||||
short-lived VM or container), a quick-and-dirty install could look like:
|
||||
|
||||
```sh
|
||||
sudo cp /usr/lib/libiberty.a{,.bak}
|
||||
sudo cp ./libiberty.a /usr/lib/
|
||||
```
|
||||
|
||||
Build an MSan-instrumented GCC LTO plugin library:
|
||||
|
||||
```sh
|
||||
cd ~/gcc/lto-plugin # use already-cloned repo from earlier step
|
||||
export CC=clang
|
||||
export CFLAGS="-g -Og -fsanitize=memory -fsanitize-memory-track-origins"
|
||||
export LDFLAGS="-fsanitize=memory"
|
||||
./configure --with-libiberty=/usr/lib
|
||||
make -j$(nproc)
|
||||
unset CC CFLAGS LDFLAGS
|
||||
```
|
||||
|
||||
Install the resulting `liblto_plugin.so`. With the same caveats discussed
|
||||
above, a simple install into an ephemeral environment could look like:
|
||||
|
||||
```sh
|
||||
export GCC_VERSION="$(gcc -dumpversion)"
|
||||
sudo cp /usr/lib/gcc/x86_64-pc-linux-gnu/$GCC_VERSION/liblto_plugin.so{,.bak}
|
||||
sudo cp .libs/liblto_plugin.so /usr/lib/gcc/x86_64-pc-linux-gnu/$GCC_VERSION/
|
||||
```
|
||||
|
||||
Build an MSan-instrumented LLVM LTO plugin library:
|
||||
|
||||
```sh
|
||||
cd ~/llvm-project # use already-cloned repo from earlier step
|
||||
|
||||
CF="-nostdinc++ -isystem $LIBCXX/include/c++/v1"
|
||||
INCDIR="$(find /usr -name plugin-api.h -type f | head -1 | xargs dirname)"
|
||||
|
||||
function configure
|
||||
{
|
||||
cmake -S ./llvm -B build-plugin -G Ninja \
|
||||
-DLLVM_ENABLE_PROJECTS=clang \
|
||||
-DLLVM_TARGETS_TO_BUILD=X86 \
|
||||
-DCMAKE_BUILD_TYPE=Release \
|
||||
-DCMAKE_C_COMPILER=clang \
|
||||
-DCMAKE_CXX_COMPILER=clang++ \
|
||||
-DCMAKE_C_FLAGS="$CF" \
|
||||
-DCMAKE_CXX_FLAGS="$CF" \
|
||||
-DCMAKE_EXE_LINKER_FLAGS="$1" \
|
||||
-DCMAKE_POSITION_INDEPENDENT_CODE=ON \
|
||||
-DLLVM_BINUTILS_INCDIR="$INCDIR" \
|
||||
-DLLVM_USE_SANITIZER=MemoryWithOrigins
|
||||
}
|
||||
|
||||
# workaround for linker issues: configure twice with different LDFLAGS
|
||||
configure "-nostdlib++ -L $LIBCXX/lib -Wl,--rpath=$LIBCXX/lib"
|
||||
configure "-nostdlib++ -L $LIBCXX/lib -lc++ -Wl,--rpath=$LIBCXX/lib"
|
||||
|
||||
cmake --build build-plugin -- LLVMgold.so
|
||||
```
|
||||
|
||||
Install the resulting `LLVMgold.so`. With the same caveats discussed
|
||||
above, a simple install into an ephemeral environment could look like:
|
||||
|
||||
```sh
|
||||
sudo cp /usr/lib/LLVMgold.so{,.bak}
|
||||
sudo cp ./build-plugin/lib/LLVMgold.so /usr/lib/
|
||||
```
|
||||
|
||||
mold's tests for `-flto` should now (mostly) work, though there do seem
|
||||
to be issues around `gcc` inferring `-flto` when not explicitly specified.
|
||||
One workaround is to use `clang` as `TEST_CC`:
|
||||
|
||||
```sh
|
||||
cd ~/mold # use already-cloned and built repo from earlier step
|
||||
TEST_CC=clang TEST_CXX=clang++ ctest --test-dir build
|
||||
```
|
||||
|
||||
## References
|
||||
|
||||
- <https://github.com/google/sanitizers/wiki/MemorySanitizerBootstrappingClang>
|
||||
- <https://llvm.org/docs/GoldPlugin.html#lto-how-to-build>
|
||||
Binary file not shown.
|
Before Width: | Height: | Size: 13 KiB |
+683
-571
File diff suppressed because it is too large.
Load diff
Binary file not shown.
|
After Width: | Height: | Size: 95 KiB |
-985
@@ -1,985 +0,0 @@
|
||||
mold(1) -- a modern linker
|
||||
==========================
|
||||
|
||||
## SYNOPSIS
|
||||
|
||||
`mold` [_option_...] _file_...
|
||||
|
||||
## DESCRIPTION
|
||||
`mold` is a faster drop-in replacement for the default GNU ld(1).
|
||||
|
||||
### How to use
|
||||
See <https://github.com/rui314/mold#how-to-use>.
|
||||
|
||||
### Compatibility
|
||||
|
||||
`mold` is designed to be a drop-in replacement for the GNU linkers for
|
||||
linking user-land programs. If your user-land program cannot be built due to
|
||||
missing command-line options, please file a bug at
|
||||
<https://github.com/rui314/mold/issues>.
|
||||
|
||||
`mold` supports a very limited set of linker script features, which is just
|
||||
sufficient to read `/usr/lib/x86_64-linux-gnu/libc.so` on Linux systems (on
|
||||
Linux, that file is contrary to its name not a shared library but an ASCII
|
||||
linker script that loads a real `libc.so` file.)
|
||||
|
||||
Beyond that, we have no plan to support any additional linker script features.
|
||||
The linker script is an ad-hoc, over-designed, complex language which we
|
||||
believe needs to be replaced by a simpler mechanism. We have a plan to add a
|
||||
replacement for the linker script to `mold` instead.
|
||||
|
||||
### Archive symbol resolution
|
||||
|
||||
Traditionally, Unix linkers are sensitive to the order in which input files
|
||||
appear on the command line. They process input files from the first (leftmost)
|
||||
file to the last (rightmost) file one-by-one. While reading input files, they
|
||||
maintain sets of defined and undefined symbols. When visiting an archive file
|
||||
(`.a` files), they pull out object files to resolve as many undefined symbols
|
||||
as possible and move on to the next input file. Object files that weren't
|
||||
pulled out will never have a chance for a second look.
|
||||
|
||||
Due to this behavior, you usually have to add archive files at the end of a
|
||||
command line, so that when a linker reaches archive files, it knows what
|
||||
symbols remain as undefined.
|
||||
|
||||
If you put archive files at the beginning of a command line, a linker doesn't
|
||||
have any undefined symbols, and thus no object files will be pulled out from
|
||||
archives. You can change the processing order by using the `--start-group` and
|
||||
`--end-group` options, though they make a linker slower.
|
||||
|
||||
`mold`, as well as the LLVM lld(1) linker, takes a different approach. They
|
||||
remember which symbols can be resolved from archive files instead of
|
||||
forgetting them after processing each archive. Therefore, `mold` and lld(1)
|
||||
can "go back" in a command line to pull out object files from archives if they
|
||||
are needed to resolve remaining undefined symbols. They are not sensitive to
|
||||
the input file order.
|
||||
|
||||
`--start-group` and `--end-group` are still accepted by `mold` and lld(1)
|
||||
for compatibility with traditional linkers, but they are silently ignored.
|
||||
|
||||
### Dynamic symbol resolution
|
||||
|
||||
Some Unix linker features are difficult to understand without comprehending
|
||||
the semantics of dynamic symbol resolution. Therefore, even though it's not
|
||||
specific to `mold`, we'll explain it here.
|
||||
|
||||
We use "ELF module" or just "module" as a collective term to refer to an
|
||||
executable or a shared library file in the ELF format.
|
||||
|
||||
An ELF module may have lists of imported symbols and exported symbols, as well
|
||||
as a list of shared library names from which imported symbols should be
|
||||
imported. The point is that imported symbols are not bound to any specific
|
||||
shared library until runtime.
|
||||
|
||||
Here is how the Unix dynamic linker resolves dynamic symbols. Upon the start
|
||||
of an ELF program, the dynamic linker constructs a list of ELF modules which,
|
||||
as a whole, consist of a complete program. The executable file is always at
|
||||
the beginning of the list followed by its dependent shared libraries. An
|
||||
imported symbol is searched from the beginning of the list to the end. If two
|
||||
or more modules define the same symbol, the one that appears first in the list
|
||||
takes precedence over the others.
|
||||
|
||||
This Unix semantics are contrary to systems such as Windows that have a
|
||||
two-level namespace for dynamic symbols. On Windows, for example, dynamic
|
||||
symbols are represented as a tuple of (`symbol-name`, `shared-library-name`),
|
||||
so that each dynamic symbol is guaranteed to be resolved from some specific
|
||||
library.
|
||||
|
||||
Typically, an ELF module that exports a symbol also imports the same symbol.
|
||||
Such a symbol is usually resolved to itself, but that's not the case if a
|
||||
module that appears before it in the symbol search list provides another
|
||||
definition of the same symbol.
|
||||
|
||||
Let's take `malloc` as an example. Assume that you define your version of
|
||||
`malloc` in your main executable file. Then, all `malloc` calls from any
|
||||
module are resolved to your function instead of the one in libc, because the
|
||||
executable is always at the beginning of the dynamic symbol search list. Note
|
||||
that even `malloc` calls within libc are resolved to your definition since
|
||||
libc exports and imports `malloc`. Therefore, by defining `malloc` yourself,
|
||||
you can overwrite a library function, and the `malloc` in libc becomes dead
|
||||
code.
|
||||
|
||||
These Unix semantics are tricky and sometimes considered harmful. For example,
|
||||
assume that you accidentally define `atoi` as a global function in your
|
||||
executable that behaves completely differently from the one in the C standard.
|
||||
Then, all `atoi` function calls from any modules (even function calls within
|
||||
libc) are redirected to your function instead of the one in libc, which will
|
||||
very likely cause a problem. That is a somewhat surprising consequence for an
|
||||
accidental name conflict. On the other hand, this semantic is sometimes useful
|
||||
because it allows users to override library functions without rebuilding
|
||||
modules containing them.
|
||||
|
||||
Whether good or bad, you should keep these semantics in mind to understand
|
||||
Unix linkers' behaviors.
|
||||
|
||||
### Build reproducibility
|
||||
|
||||
`mold`'s output is deterministic. That is, if you pass the same object files
|
||||
and the same command-line options to the same version of `mold`, it is
|
||||
guaranteed that `mold` produces the bit-for-bit identical output. The linker's
|
||||
internal randomness, such as the timing of thread scheduling or iteration
|
||||
orders of hash tables, doesn't affect the output.
|
||||
|
||||
`mold` does not have any host-specific default settings. This is contrary to the
|
||||
GNU linkers, for which some configurable values, such as system-dependent
|
||||
library search paths, are hard-coded. `mold` depends only on its command-line
|
||||
arguments.
|
||||
|
||||
## OPTION NOTATIONS
|
||||
|
||||
Multi-letter long options may precede either a single dash or double dashes,
|
||||
except for those starting with the letter "o". For historical reasons, long
|
||||
options beginning with "o" must precede double dashes.
|
||||
|
||||
For example, you can spell `--as-needed` as `-as-needed`, but `--omagic` must
|
||||
not be spelled as `-omagic`. `-omagic` will be interpreted not as `--omagic`
|
||||
but as `-o magic`.
|
||||
|
||||
## MOLD-SPECIFIC OPTIONS
|
||||
|
||||
* `--chroot`=_dir_:
|
||||
Set _dir_ as the root directory.
|
||||
|
||||
* `--color-diagnostics`=[ _auto_ | _always_ | _never_ ]:
|
||||
Show diagnostic messages in color using ANSI escape sequences. `auto` means
|
||||
that `mold` prints out messages in color only if the standard output is
|
||||
connected to a TTY. Default is `auto`.
|
||||
|
||||
* `--color-diagnostics`:
|
||||
Synonym for `--color-diagnostics=auto`.
|
||||
|
||||
* `--no-color-diagnostics`:
|
||||
Synonym for `--color-diagnostics=never`.
|
||||
|
||||
* `--detach`, `--no-detach`:
|
||||
Permit or do not permit mold to create a debug info file in the background.
|
||||
|
||||
* `--fork`, `--no-fork`:
|
||||
Spawn a child process and let it do the actual linking. When linking a large
|
||||
program, the OS kernel can take a few hundred milliseconds to terminate a
|
||||
`mold` process. `--fork` hides that latency. By default, it does fork.
|
||||
|
||||
Note that `--fork` also hides the resource usage statistics reported by
|
||||
time(2), since it doesn't call waitpid(2) on the child process. If you
|
||||
need those statistics, pass `--no-fork`.
|
||||
|
||||
* `--perf`:
|
||||
Print performance statistics.
|
||||
|
||||
* `--print-dependencies`:
|
||||
Print out dependency information for input files.
|
||||
|
||||
Each line of the output for this option shows which file depends on which
|
||||
file to use a specific symbol. This option is useful for debugging why some
|
||||
object file in a static archive got linked or why some shared library is
|
||||
kept in an output file's dependency list even with `--as-needed`.
|
||||
|
||||
* `--relocatable-merge-sections`:
|
||||
By default, `mold` doesn't merge input sections by name when merging input
|
||||
object files into a single output object file for `-r`. For example,
|
||||
`.text.foo` and `.text.bar` aren't merged for `-r` even though they are
|
||||
merged into `.text` based on the default section merging rules.
|
||||
|
||||
This option changes the behavior so that `mold` merges input sections by
|
||||
name by the default section merging rules.
|
||||
|
||||
* `--repro`:
|
||||
Archive input files, as well as a text file containing command line options,
|
||||
in a tar file so that you can run `mold` with the exact same inputs again.
|
||||
This is useful for reporting a bug with a reproducer. The output filename is
|
||||
`path/to/output.tar`, where `path/to/output` is an output filename specified
|
||||
by `-o`.
|
||||
|
||||
* `--reverse-sections`:
|
||||
Reverse the order of input sections before assigning them the offsets in the
|
||||
output file.
|
||||
|
||||
This option is useful for finding bugs that depend on the initialization
|
||||
order of global objects. In C++, constructors of global objects in a single
|
||||
source file are guaranteed to be executed in the source order, but there's
|
||||
no such guarantee across compilation units. Usually, constructors are
|
||||
executed in the order given to the linker, but depending on it is a mistake.
|
||||
|
||||
By reversing the order of input sections using `--reverse-sections`, you can
|
||||
easily test that your program works in the reversed initialization order.
|
||||
|
||||
* `--run` _command_ _arg_...:
|
||||
Run _command_ with `mold` as `/usr/bin/ld`. Specifically, `mold` runs a
|
||||
given command with the `LD_PRELOAD` environment set to intercept exec(3)
|
||||
family functions and replaces `argv[0]` with itself if it is `ld`,
|
||||
`ld.gold`, or `ld.lld`.
|
||||
|
||||
* `--separate-debug-file`, `--separate-debug-file`=_file_:
|
||||
Bundle debug info sections into a separate file instead of embedding them in
|
||||
an output executable or a shared library. mold creates a debug info file in
|
||||
the background by default, so that you can start running your executable as
|
||||
soon as possible.
|
||||
|
||||
By default, the debug info file is created in the same directory as is the
|
||||
output file, with the `.dbg` file extension. That filename is embedded into
|
||||
the output file so that `gdb` can automatically find the debug info file for
|
||||
the output file. For more info about gdb features related to separate debug
|
||||
files, see
|
||||
<https://sourceware.org/gdb/current/onlinedocs/gdb.html/Separate-Debug-Files.html>.
|
||||
|
||||
mold holds a file lock with flock(2) while creating a debug info file in the
|
||||
background.
|
||||
|
||||
If you don't want to create a debug info file in the background, pass the
|
||||
`--no-detach` option.
|
||||
|
||||
* `--shuffle-sections`, `--shuffle-sections`=_number_:
|
||||
Randomize the output by shuffling the order of input sections before
|
||||
assigning them the offsets in the output file. If a _number_ is given, it's
|
||||
used as a seed for the random number generator, so that the linker produces
|
||||
the same output for the same seed. If no seed is given, a random number is
|
||||
used as a seed.
|
||||
|
||||
This option is useful for benchmarking. Modern CPUs are sensitive to a
|
||||
program's memory layout. A seemingly benign change in program layout, such
|
||||
as a small size increase of a function in the middle of a program, can
|
||||
affect the program's performance. Therefore, even if you write new code and
|
||||
get a good benchmark result, it is hard to say whether the new code improves
|
||||
the program's performance; it is possible that the new memory layout happens
|
||||
to perform better.
|
||||
|
||||
By running a benchmark multiple times with randomized memory layouts using
|
||||
`--shuffle-sections`, you can isolate your program's real performance number
|
||||
from the randomness caused by memory layout changes.
|
||||
|
||||
* `--spare-program-headers`=_number_:
|
||||
Append the given number of `PT_NULL` entries to the end of the program
|
||||
header, so that post-link processing tools can easily add new segments by
|
||||
overwriting the null entries.
|
||||
|
||||
Note that ELF requires all `PT_LOAD` segments to be sorted by `p_vaddr`.
|
||||
Therefore, if you add a new LOAD segment, you may need to sort the entire
|
||||
program header.
|
||||
|
||||
* `--stats`:
|
||||
Print input statistics.
|
||||
|
||||
* `--thread-count`=_count_:
|
||||
Use _count_ number of threads.
|
||||
|
||||
* `--threads`, `--no-threads`:
|
||||
Use multiple threads. By default, `mold` uses as many threads as the number of
|
||||
cores or 32, whichever is smaller. The reason it is capped at 32 is because
|
||||
`mold` doesn't scale well beyond that point. To use only one thread, pass
|
||||
`--no-threads` or `--thread-count=1`.
|
||||
|
||||
* `--quick-exit`, `--no-quick-exit`:
|
||||
Use or do not use `quick_exit` to exit.
|
||||
|
||||
* `--zero-to-bss`:
|
||||
Convert all-zero data sections into BSS.
|
||||
|
||||
When this option is enabled, `mold` scans input data sections that are not
|
||||
of type `SHT_NOBITS` and checks whether their contents consist solely of
|
||||
zero bytes. Such sections are then converted into BSS (`SHT_NOBITS`)
|
||||
sections. This reduces the size of the output file, since BSS sections
|
||||
occupy no space in the file image.
|
||||
|
||||
This behavior is especially useful for user-defined sections created with
|
||||
`__attribute__((section(".sectname")))` that contain uninitialized global
|
||||
variables. GCC and Clang do not automatically mark such sections as BSS
|
||||
even if their contents are entirely zero, and instead emit them as regular
|
||||
data sections.
|
||||
|
||||
For example, consider `__attribute__((section(".sectname"))) int vec[256];`.
|
||||
|
||||
By default, this results in a `.sectname` section of type `SHT_PROGBITS`
|
||||
filled with zeros. With `--zero-to-bss`, the linker will recognize it as
|
||||
empty data and convert it to a `SHT_NOBITS` section, reducing the output
|
||||
file size without changing runtime semantics.
|
||||
|
||||
* `-z rewrite-endbr`, `-z norewrite-endbr`:
|
||||
As a security measure, some CPU instruction sets have recently gained a
|
||||
feature to protect control flow integrity by disallowing indirect branches
|
||||
by default. If the feature is enabled, the instruction that is executed
|
||||
immediately after an indirect branch must be an branch target marker
|
||||
instruction, or a CPU-level fault will raise. The marker instruction is also
|
||||
known as "landing pad" instruction, to which indirect branches can land.
|
||||
This feature makes ROP attacks harder to conduct.
|
||||
|
||||
To use the feature, a function whose pointer is taken needs to begin with a
|
||||
landing pad because a function call via a function pointer is compiled to an
|
||||
indirect branch. On the other hand, if a function is called only directly
|
||||
(i.e. referred to only by _direct_ branch instructions), it doesn't have to
|
||||
begin with it.
|
||||
|
||||
By default, the compiler always emits a landing pad at the beginning of each
|
||||
global function because it doesn't know whether or not the function's
|
||||
pointer is taken in another translation unit. As a result, the resulting
|
||||
binary has more attack surface than necessary.
|
||||
|
||||
If `--rewrite-endbr` is given, mold conducts a whole program analysis
|
||||
to identify functions whose addresses are actually taken and rewrites
|
||||
landing pads with no-ops for non-address-taken functions, reducing the
|
||||
attack surface.
|
||||
|
||||
This feature is currently available only on x86-64.
|
||||
|
||||
## GNU-COMPATIBLE OPTIONS
|
||||
|
||||
* `--help`:
|
||||
Report usage information to stdout and exit.
|
||||
|
||||
* `-v`, `--version`:
|
||||
Report version information to stdout.
|
||||
|
||||
* `-V`:
|
||||
Report version and target information to stdout.
|
||||
|
||||
* `-E`, `--export-dynamic`, `--no-export-dynamic`:
|
||||
When creating an executable, using the `-E` option causes all global symbols
|
||||
to be put into the dynamic symbol table, so that the symbols are visible
|
||||
from other ELF modules at runtime.
|
||||
|
||||
By default, or if `--no-export-dynamic` is given, only symbols that are
|
||||
referenced by DSOs at link-time are exported from an executable.
|
||||
|
||||
* `-F` _libname_, `--filter`=_libname_:
|
||||
Set the `DT_FILTER` dynamic section field to _libname_.
|
||||
|
||||
* `-I` _file_, `--dynamic-linker`=_file_, `--no-dynamic-linker`:
|
||||
Set the dynamic linker path to _file_. If no `-I` option is given, or if
|
||||
`--no-dynamic-linker` is given, no dynamic linker path is set to an output
|
||||
file. This is contrary to the GNU linkers which set a default dynamic linker
|
||||
path in that case. This difference doesn't usually make any difference
|
||||
because the compiler driver always passes `-I` to the linker.
|
||||
|
||||
* `-L` _dir_, `--library-path`=_dir_:
|
||||
Add _dir_ to the list of library search paths from which `mold` searches
|
||||
libraries for the `-l` option.
|
||||
|
||||
Unlike the GNU linkers, `mold` does not have default search paths. This
|
||||
difference doesn't usually make any difference because the compiler driver
|
||||
always passes all necessary search paths to the linker.
|
||||
|
||||
* `-M`, `--print-map`:
|
||||
Write a map file to stdout.
|
||||
|
||||
* `-N`, `--omagic`, `--no-omagic`:
|
||||
Force `mold` to emit an output file with an old-fashioned memory layout.
|
||||
First, it makes the first data segment not aligned to a page boundary.
|
||||
Second, text segments are marked as writable if the option is given.
|
||||
|
||||
* `-S`, `--strip-debug`:
|
||||
Omit `.debug_*` sections from the output file.
|
||||
|
||||
* `-T` _file_, `--script`=_file_:
|
||||
Read linker script from _file_.
|
||||
|
||||
* `-X`, `--discard-locals`:
|
||||
Discard temporary local symbols to reduce the sizes of the symbol table and
|
||||
the string table. Temporary local symbols are local symbols starting with
|
||||
`.L`. Compilers usually generate such symbols for unnamed program elements
|
||||
such as string literals or floating-point literals.
|
||||
|
||||
* `-e` _symbol_, `--entry`=_symbol_:
|
||||
|
||||
Use _symbol_ as the entry point symbol instead of the default entry
|
||||
point symbol _start.
|
||||
|
||||
* `-f` _shlib_, `--auxiliary`=_shlib_:
|
||||
Set the `DT_AUXILIARY` dynamic section field to _shlib_.
|
||||
|
||||
* `-h` _libname_, `--soname`=_libname_:
|
||||
Set the `DT_SONAME` dynamic section field to _libname_. This option is used
|
||||
when creating a shared object file. Typically, when you create `libfoo.so`,
|
||||
you want to pass `--soname=foo` to a linker.
|
||||
|
||||
* `-l` _libname_:
|
||||
Search for `lib`_libname_`.so` or `lib`_libname_`.a` from library search
|
||||
paths.
|
||||
|
||||
* `-m` _target_:
|
||||
Choose a _target_.
|
||||
|
||||
* `-o` _file_, `--output`=_file_:
|
||||
Use _file_ as the output file name instead of the default name `a.out`.
|
||||
|
||||
* `-r`, `--relocatable`:
|
||||
Instead of generating an executable or a shared object file, combine input
|
||||
object files to generate another object file that can be used as an input to
|
||||
a linker.
|
||||
|
||||
* `-s`, `--strip-all`:
|
||||
Omit `.symtab` section from the output file.
|
||||
|
||||
* `-u` _symbol_, `--undefined`=_symbol_:
|
||||
If _symbol_ remains as an undefined symbol after reading all object files,
|
||||
and if there is a static archive that contains an object file defining
|
||||
_symbol_, pull out the object file and link it so that the output file
|
||||
contains a definition of _symbol_.
|
||||
|
||||
* `-y` _symbol_, `--trace-symbol`=_symbol_:
|
||||
Trace references to _symbol_.
|
||||
|
||||
* `--Bdynamic`:
|
||||
Link against shared libraries.
|
||||
|
||||
* `--Bstatic`:
|
||||
Do not link against shared libraries.
|
||||
|
||||
* `--Bsymbolic`:
|
||||
When creating a shared library, make global symbols export-only (i.e. do not
|
||||
import the same symbol). As a result, references within a shared library are
|
||||
always resolved locally, negating symbol override at runtime. See "Dynamic
|
||||
symbol resolution" for more information about symbol imports and exports.
|
||||
|
||||
* `--Bsymbolic-functions`:
|
||||
This option has the same effect as `--Bsymbolic` but works only for function
|
||||
symbols. Data symbols remain being both imported and exported.
|
||||
|
||||
* `--Bsymbolic-non-weak`:
|
||||
This option has the same effect as `--Bsymbolic` but works only for non-weak
|
||||
symbols. Weak symbols remain being both imported and exported.
|
||||
|
||||
* `--Bsymbolic-non-weak-functions`:
|
||||
This option has the same effect as `--Bsymbolic` but works only for non-weak
|
||||
function symbols. Data symbols and weak function symbols remain being both
|
||||
imported and exported.
|
||||
|
||||
* `--Bno-symbolic`:
|
||||
Cancel `--Bsymbolic`, `--Bsymbolic-functions`, `--Bsymbolic-non-weak` and
|
||||
`--Bsymbolic-non-weak-functions`.
|
||||
|
||||
* `--Map`=_file_:
|
||||
Write map file to _file_.
|
||||
|
||||
* `--Tbss`=_address_:
|
||||
Alias for `--section-start=.bss=`_address_.
|
||||
|
||||
* `--Tdata`=_address_:
|
||||
Alias for `--section-start=.data=`_address_.
|
||||
|
||||
* `--Ttext`=_address_:
|
||||
Alias for `--section-start=.text=`_address_.
|
||||
|
||||
* `--allow-multiple-definition`:
|
||||
Normally, the linker reports an error if there are more than one definition
|
||||
of a symbol. This option changes the default behavior so that it doesn't
|
||||
report an error for duplicate definitions and instead use the first
|
||||
definition.
|
||||
|
||||
* `--allow-shlib-undefined`, `--no-allow-shlib-undefined`:
|
||||
Even if mold succeeds in linking a main executable without undefined symbol
|
||||
errors, you may still encounter symbol lookup errors at runtime because the
|
||||
dynamic linker cannot find some symbols in shared libraries in any ELF
|
||||
module. This occurs because mold ignores undefined symbols in shared
|
||||
libraries by default.
|
||||
|
||||
If you pass `--no-allow-shlib-undefined`, mold verifies that undefined
|
||||
symbols in shared libraries given to the linker can be resolved at
|
||||
link-time. In other words, this converts the runtime error to a link-time
|
||||
error.
|
||||
|
||||
Note that you need to pass all shared libraries, including indirectly
|
||||
dependent ones, to the linker as arguments for `-l`. If a shared library
|
||||
depends on a library that's not passed to the linker, the verification will
|
||||
be skipped for that file.
|
||||
|
||||
* `--as-needed`, `--no-as-needed`:
|
||||
By default, shared libraries given to the linker are unconditionally added
|
||||
to the list of required libraries in an output file. However, shared
|
||||
libraries after `--as-needed` are added to the list only when at least one
|
||||
symbol is actually used by the output file. In other words, shared libraries
|
||||
after `--as-needed` are not added to the list of needed libraries if they
|
||||
are not needed by a program.
|
||||
|
||||
The `--no-as-needed` option restores the default behavior for subsequent
|
||||
files.
|
||||
|
||||
* `--build-id`=[ `md5` | `sha1` | `sha256` | `fast` | `uuid` | `0x`_hexstring_ | `none` ]:
|
||||
Create a `.note.gnu.build-id` section containing a byte string to uniquely
|
||||
identify an output file. `sha256` compute a 256-bit cryptographic hash of an
|
||||
output file and set it to build-id. `md5` and `sha1` compute the same hash
|
||||
but truncate it to 128 and 160 bits, respectively, before setting it to
|
||||
build-id. `uuid` sets a random 128-bit UUID. `0x`_hexstring_ sets
|
||||
_hexstring_. `fast` is a synonym for `sha256`.
|
||||
|
||||
* `--build-id`:
|
||||
Synonym for `--build-id=sha256`.
|
||||
|
||||
* `--no-build-id`:
|
||||
Synonym for `--build-id=none`.
|
||||
|
||||
* `--compress-debug-sections`=[ `zlib` | `zlib-gabi` | `zlib:`_N_ | `zstd` | `zstd:`_N_ | `none` ]:
|
||||
Compress DWARF debug info (`.debug_*` sections) using the zlib or zstd
|
||||
compression algorithm. `zlib-gabi` is an alias for `zlib`. `zlib:`_N_
|
||||
selects the zlib compression level, where _N_ is between 0 and 9 (0
|
||||
means no compression), and `zstd:`_N_ selects the zstd compression
|
||||
level, where _N_ is between 1 and 22. Higher levels achieve better
|
||||
compression ratios but are slower. `zlib` is equivalent to `zlib:1`
|
||||
and `zstd` is equivalent to `zstd:3`.
|
||||
|
||||
* `--defsym`=_symbol_=_value_:
|
||||
Define _symbol_ as an alias for _value_.
|
||||
|
||||
_value_ is either an integer (in decimal or hexadecimal with `0x` prefix) or
|
||||
a symbol name. If an integer is given as a value, _symbol_ is defined as an
|
||||
absolute symbol with the given value.
|
||||
|
||||
* `--default-symver`:
|
||||
Use soname as a symbol version and append that version to all symbols.
|
||||
|
||||
* `--demangle`, `--no-demangle`:
|
||||
Demangle C++ and Rust symbols in log messages.
|
||||
|
||||
* `--dependency-file`=_file_:
|
||||
Write a dependency file to _file_. The contents of the written file is
|
||||
readable by make(1), which defines only one rule with the linker's output
|
||||
file as a target and all input files as its prerequisites. Users are
|
||||
expected to include the generated dependency file into a Makefile to
|
||||
automate the dependency management. This option is analogous to the
|
||||
compiler's `-MM -MF` options.
|
||||
|
||||
* `--dynamic-list`=_file_:
|
||||
Read a list of dynamic symbols from _file_. Same as
|
||||
`--export-dynamic-symbol-list`, except that it implies `--Bsymbolic`. If
|
||||
_file_ does not exist in the current directory, it is searched from library
|
||||
search paths for the sake of compatibility with GNU ld.
|
||||
|
||||
* `--eh-frame-hdr`, `--no-eh-frame-hdr`:
|
||||
Create `.eh_frame_hdr` section.
|
||||
|
||||
* `--emit-relocs`:
|
||||
The linker usually "consumes" relocation sections. That is, the linker
|
||||
applies relocations to other sections, and relocation sections themselves
|
||||
are discarded.
|
||||
|
||||
The `--emit-relocs` instructs the linker to leave relocation sections in the
|
||||
output file. Some post-link binary analysis or optimization tools such as
|
||||
LLVM Bolt need them.
|
||||
|
||||
* `--enable-new-dtags`, `--disable-new-dtags`:
|
||||
By default, `mold` emits `DT_RUNPATH` for `--rpath`. If you pass
|
||||
`--disable-new-dtags`, `mold` emits `DT_RPATH` for `--rpath` instead.
|
||||
|
||||
* `--execute-only`:
|
||||
|
||||
Traditionally, setting the executable bit to 1 for a memory page implies
|
||||
that the page also become readable, which allows machine code to be read
|
||||
as data at runtime. That is actually what an attacker often does after
|
||||
gaining a limited control of a process to find pieces of machine code
|
||||
they can use to gain the full control of the process. As a mitigation,
|
||||
recent processors including some ARM64 ones allows "execute-only" pages.
|
||||
If a page is execute-only, you can call a function there as long as you
|
||||
know its address but can't read it as data.
|
||||
|
||||
This option marks text segments as execute-only by setting just the "X"
|
||||
bit instead of "RX". Note that on most systems, the absence of the "R"
|
||||
bit in the text segment serves just as a hint. If you run a program
|
||||
linked with `--execute-only` on a processor that doesn't support
|
||||
execute-only pages, your executable will likely still function normally,
|
||||
but the text segment will remain readable.
|
||||
|
||||
* `--exclude-libs`=_libraries_ ...:
|
||||
Mark all symbols in the given _libraries_ hidden.
|
||||
|
||||
* `--export-dynamic-symbol`=_symbol_:
|
||||
Put symbols matching _symbol_ in the dynamic symbol table. _symbol_ may be a
|
||||
glob pattern in the same syntax as for the `--export-dynamic-symbol-list` or
|
||||
`--version-script` options.
|
||||
|
||||
* `--export-dynamic-symbol-list`=_file_:
|
||||
Read a list of dynamic symbols from _file_.
|
||||
|
||||
* `--fatal-warnings`, `--no-fatal-warnings`:
|
||||
Treat warnings as errors.
|
||||
|
||||
* `--fini`=_symbol_:
|
||||
Call _symbol_ at unload-time.
|
||||
|
||||
* `--gc-sections`, `--no-gc-sections`:
|
||||
Remove unreferenced sections.
|
||||
|
||||
* `--gdb-index`:
|
||||
Create a `.gdb_index` section to speed up GNU debugger. To use this, you
|
||||
need to compile source files with the `-ggnu-pubnames` compiler flag.
|
||||
|
||||
* `--hash-style`=[ `sysv` | `gnu` | `both` | `none` ]:
|
||||
Set hash style.
|
||||
|
||||
* `--icf`=[ `safe` | `all` | `none` ], `--no-icf`:
|
||||
It is not uncommon for a program to contain many identical functions that
|
||||
differ only in name. For example, a C++ template `std::vector` is very
|
||||
likely to be instantiated to the identical code for `std::vector<int>` and
|
||||
`std::vector<unsigned>` because the container cares only about the size of
|
||||
the parameter type. Identical Code Folding (ICF) is a size optimization to
|
||||
identify and merge such identical functions.
|
||||
|
||||
If `--icf=all` is given, `mold` tries to merge all identical functions. This
|
||||
reduces the size of the output most, but it is not a "safe" optimization. It
|
||||
is guaranteed in C and C++ that two pointers pointing two different
|
||||
functions will never be equal, but `--icf=all` breaks that assumption as two
|
||||
identical functions have the same address after merging. So a care must be
|
||||
taken when you use this flag that your program does not depend on the
|
||||
function pointer uniqueness.
|
||||
|
||||
`--icf=safe` is a flag to merge functions only when it is safe to do so.
|
||||
That is, if a program does not take an address of a function, it is safe to
|
||||
merge that function with other function, as you cannot compare a function
|
||||
pointer with something else without taking an address of a function.
|
||||
|
||||
`--icf=safe` needs to be used with a compiler that supports `.llvm_addrsig`
|
||||
section which contains the information as to what symbols are address-taken.
|
||||
LLVM/Clang supports that section by default. Since GCC does not support it
|
||||
yet, you cannot use `--icf=safe` with GCC (it doesn't do any harm but can't
|
||||
optimize at all.)
|
||||
|
||||
`--icf=none` and `--no-icf` disables ICF.
|
||||
|
||||
* `--ignore-data-address-equality`:
|
||||
Make ICF to merge not only functions but also data. This option should be
|
||||
used in combination with `--icf=all`.
|
||||
|
||||
* `--image-base`=_addr_:
|
||||
Set the base address to _addr_.
|
||||
|
||||
* `--init`=_symbol_:
|
||||
Call _symbol_ at load-time.
|
||||
|
||||
* `--no-undefined`:
|
||||
Report undefined symbols (even with `--shared`).
|
||||
|
||||
* `--noinhibit-exec`:
|
||||
Create an output file even if errors occur.
|
||||
|
||||
* `--package-metadata`=_percent-encoded-string_:
|
||||
Embed a specified string into the `.note.package` section. This option
|
||||
is designed for build scripts that generate binary packages, such as
|
||||
`.rpm` or `.deb`, to include package metadata in each executable. It
|
||||
simplifies the process of identifying the corresponding package for a
|
||||
given executable or core file.
|
||||
|
||||
An argument to this option is treated as percent-encoded and decoded
|
||||
before being inserted into the section, allowing you to avoid the use of
|
||||
the comma (`,`) character in the argument. This is useful because the
|
||||
compiler replaces all occurrences of commas in `-Wl,` with spaces before
|
||||
forwarding them to the linker. Note that `mold` always interprets the
|
||||
argument as percent-encoded, so you also need to escape all occurrences
|
||||
of `%` as `%25`.
|
||||
|
||||
* `--pack-dyn-relocs`=[ `relr` | `android` | `android+relr` | `none` ]:
|
||||
Compress dynamic relocations in the output file.
|
||||
|
||||
If `relr` is specified, all `R_*_RELATIVE` relocations are put into
|
||||
`.relr.dyn` section instead of `.rel.dyn` or `.rela.dyn` section. Since
|
||||
`.relr.dyn` section uses a space-efficient encoding scheme, specifying this
|
||||
flag can reduce the size of the output. This is typically most effective for
|
||||
position-independent executable.
|
||||
|
||||
If `android` is specified, all dynamic relocations are encoded in Android's
|
||||
packed relocation format (APS2) and stored in the `.rel.dyn` or `.rela.dyn`
|
||||
section, whose section type is changed to `SHT_ANDROID_REL` or
|
||||
`SHT_ANDROID_RELA`. The dynamic table is updated to use `DT_ANDROID_REL`
|
||||
/ `DT_ANDROID_RELA` and their `SZ` variants in place of the standard tags.
|
||||
|
||||
`android+relr` combines both: `R_*_RELATIVE` relocations go to `.relr.dyn`
|
||||
and the rest are APS2-encoded. This is the configuration used by most
|
||||
Android system binaries.
|
||||
|
||||
Note that a runtime loader has to support `.relr.dyn` to run executables or
|
||||
shared libraries linked with `--pack-dyn-relocs=relr`. As of 2026, glibc
|
||||
(since 2.36), musl (since 1.2.4), Android, ChromeOS, and Fuchsia all
|
||||
support it. The APS2 format, on the other hand, is only handled by
|
||||
Android's bionic loader; the encoder itself is target-agnostic, but the
|
||||
produced binary can only be loaded on platforms where bionic (or a
|
||||
compatible loader) handles `DT_ANDROID_REL(A)`.
|
||||
|
||||
* `--pie`, `--pic-executable`, `--no-pie`, `--no-pic-executable`:
|
||||
Create a position-independent executable.
|
||||
|
||||
* `--print-gc-sections`, `--no-print-gc-sections`:
|
||||
Print removed unreferenced sections.
|
||||
|
||||
* `--print-icf-sections`, `--no-print-icf-sections`:
|
||||
Print folded identical sections.
|
||||
|
||||
* `--push-state`, `--pop-state`:
|
||||
`--push-state` saves the current values of `--as-needed`, `--whole-archive`,
|
||||
`--static`, and `--start-lib`. The saved values can be restored by
|
||||
pop-state.
|
||||
|
||||
`--push-state` and `--pop-state` pairs can nest.
|
||||
|
||||
These options are useful when you want to construct linker command line
|
||||
options programmatically. For example, if you want to link `libfoo.so` by
|
||||
as-needed basis but don't want to change the global state of `--as-needed`,
|
||||
you can append `--push-state --as-needed -lfoo --pop-state` to the linker
|
||||
command line options.
|
||||
|
||||
* `--relax, --no-relax`:
|
||||
Rewrite machine instructions with more efficient ones for some relocations.
|
||||
The feature is enabled by default.
|
||||
|
||||
* `--require-defined`=_symbol_:
|
||||
Like `--undefined`, except the new symbol must be defined by the end of the
|
||||
link.
|
||||
|
||||
* `--retain-symbols-file`=_file_:
|
||||
Keep only symbols listed in _file_. _file_ is a text file containing a
|
||||
symbol name on each line. `mold` discards all local symbols as well as
|
||||
global symbol that are not in _file_. Note that this option removes symbols
|
||||
only from `.symtab` section and does not affect `.dynsym` section, which is
|
||||
used for dynamic linking.
|
||||
|
||||
* `--rpath`=_dir_:
|
||||
Add _dir_ to runtime search paths.
|
||||
|
||||
* `--section-start`=_section_=_address_:
|
||||
Set _address_ to section. _address_ is a hexadecimal number that may start
|
||||
with an optional `0x`.
|
||||
|
||||
* `--shared`, `--Bshareable`:
|
||||
Create a share library.
|
||||
|
||||
* `--spare-dynamic-tags`=_number_:
|
||||
Append the given number of `DT_NULL` entries to the end of the `.dynamic`
|
||||
section, so that post-link processing tools can easily add new dynamic tags
|
||||
by overwriting the null entries.
|
||||
|
||||
* `--start-lib`, `--end-lib`:
|
||||
Handle object files between `--start-lib` and `--end-lib` as if they were in
|
||||
an archive file. That means object files between them are linked only when
|
||||
they are needed to resolve undefined symbols. The options are useful if you
|
||||
want to link object files only when they are needed but want to avoid the
|
||||
overhead of running ar(3).
|
||||
|
||||
* `--static`:
|
||||
Do not link against shared libraries.
|
||||
|
||||
* `--sysroot`=_dir_:
|
||||
Set target system root directory to _dir_.
|
||||
|
||||
* `--trace`:
|
||||
Print name of each input file.
|
||||
|
||||
* `--undefined-glob`=_pattern_:
|
||||
Synonym for `--undefined`, except that `--undefined-glob` takes a glob
|
||||
pattern instead of just a single symbol name.
|
||||
|
||||
* `--undefined-version`, `--no-undefined-version`:
|
||||
By default, `mold` warns on a symbol specified by a version script or by
|
||||
`--export-dynamic-symbol` if it is not defined. You can silence the warning
|
||||
by `--undefined-version`.
|
||||
|
||||
* `--unique`=_pattern_:
|
||||
Don't merge input sections that match the given glob pattern _pattern_.
|
||||
|
||||
* `--unresolved-symbols`=[ `report-all` | `ignore-all` | `ignore-in-object-files` | `ignore-in-shared-libs` ]:
|
||||
How to handle undefined symbols.
|
||||
|
||||
* `--version-script`=_file_:
|
||||
Read version script from _file_. If _file_ does not exist in the current
|
||||
directory, it is searched from library search paths for the sake of
|
||||
compatibility with GNU ld.
|
||||
|
||||
* `-w`, `--no-warnings`:
|
||||
Suppress warnings and cancel `--fatal-warnings`.
|
||||
|
||||
* `--warn-common`, `--no-warn-common`:
|
||||
Warn about common symbols.
|
||||
|
||||
* `--warn-once`:
|
||||
Only warn once for each undefined symbol instead of warn for each relocation
|
||||
referring an undefined symbol.
|
||||
|
||||
* `--warn-unresolved-symbols`, `--error-unresolved-symbols`:
|
||||
Normally, the linker reports an error for unresolved symbols.
|
||||
`--warn-unresolved-symbols` option turns it into a warning.
|
||||
`--error-unresolved-symbols` option restores the default behavior.
|
||||
|
||||
* `--whole-archive`, `--no-whole-archive`:
|
||||
When archive files (`.a` files) are given to the linker, only object files
|
||||
that are needed to resolve undefined symbols are extracted from them and
|
||||
linked to an output file. `--whole-archive` changes that behavior for
|
||||
subsequent archives so that the linker extracts all object files and links
|
||||
them to an output. For example, if you are creating a shared object file and
|
||||
you want to include all archive members to the output, you should pass
|
||||
`--whole-archive`. `--no-whole-archive` restores the default behavior for
|
||||
subsequent archives.
|
||||
|
||||
* `--wrap`=_symbol_:
|
||||
Make _symbol_ be resolved to `__wrap_`_symbol_. The original symbol can be
|
||||
resolved as `__real_`_symbol_. This option is typically used for wrapping an
|
||||
existing function.
|
||||
|
||||
* `-z cet-report`=[ `warning` | `error` | `none` ]:
|
||||
Intel Control-flow Enforcement Technology (CET) is a new x86 feature
|
||||
available since Tiger Lake which is released in 2020. It defines new
|
||||
instructions to harden security to protect programs from control hijacking
|
||||
attacks. You can tell the compiler to use the feature by specifying the
|
||||
`-fcf-protection` flag.
|
||||
|
||||
`-z cet-report` flag is used to make sure that all object files were
|
||||
compiled with a correct `-fcf-protection` flag. If `warning` or `error` are
|
||||
given, `mold` prints out a warning or an error message if an object file was
|
||||
not compiled with the compiler flag.
|
||||
|
||||
`mold` looks for `GNU_PROPERTY_X86_FEATURE_1_IBT` bit and
|
||||
`GNU_PROPERTY_X86_FEATURE_1_SHSTK` bit in `.note.gnu.property` section to
|
||||
determine whether or not an object file was compiled with `-fcf-protection`.
|
||||
|
||||
* `-z now`, `-z lazy`:
|
||||
By default, functions referring to other ELF modules are resolved by the
|
||||
dynamic linker when they are called for the first time. `-z now` marks an
|
||||
executable or a shared library file so that all dynamic symbols are resolved
|
||||
when a file is loaded to memory. `-z lazy` restores the default behavior.
|
||||
|
||||
* `-z origin`:
|
||||
Mark object requiring immediate `$ORIGIN` processing at runtime.
|
||||
|
||||
* `-z ibt`:
|
||||
Turn on `GNU_PROPERTY_X86_FEATURE_1_IBT` bit in `.note.gnu.property` section
|
||||
to indicate that the output uses IBT-enabled PLT. This option implies `-z
|
||||
ibtplt`.
|
||||
|
||||
* `-z ibtplt`:
|
||||
Generate Intel Branch Tracking (IBT)-enabled PLT which is the default on
|
||||
x86-64. This is the default.
|
||||
|
||||
* `-z execstack`, `-z noexecstack`:
|
||||
By default, the pages for the stack area (i.e. the pages where local
|
||||
variables reside) are not executable for security reasons. `-z execstack`
|
||||
makes it executable. `-z noexecstack` restores the default behavior.
|
||||
|
||||
* `-z keep-text-section-prefix`, `-z nokeep-text-section-prefix`:
|
||||
Keep `.text.hot`, `.text.unknown`, `.text.unlikely`, `.text.startup`, and
|
||||
`.text.exit` as separate sections in the final binary instead of merging
|
||||
them as `.text`.
|
||||
|
||||
* `-z rodynamic`:
|
||||
Make the `.dynamic` section read-only.
|
||||
|
||||
* `-z relro`, `-z norelro`:
|
||||
Some sections such as `.dynamic` have to be writable only during a module is
|
||||
being loaded to memory. Once the dynamic linker finishes its job, such
|
||||
sections won't be mutated by anyone. As a security mitigation, it is
|
||||
preferred to make such segments read-only during program execution.
|
||||
|
||||
`-z relro` puts such sections into a special segment called `relro`. The
|
||||
dynamic linker makes a relro segment read-only after it finishes its job.
|
||||
|
||||
By default, `mold` generates a relro segment. `-z norelro` disables the
|
||||
feature.
|
||||
|
||||
* `-z sectionheader`, `-z nosectionheader`:
|
||||
`-z nosectionheader` tell the linker to omit the section header.
|
||||
By default, the linker does not omit the section header.
|
||||
|
||||
* `-z separate-loadable-segments`, `-z separate-code`, `-z noseparate-code`:
|
||||
If one memory page contains multiple segments, the page protection bits are
|
||||
set in such a way that the needed attributes (writable or executable) are
|
||||
satisfied for all segments. This usually happens at a boundary of two
|
||||
segments with two different attributes.
|
||||
|
||||
`separate-loadable-segments` adds paddings between segments with different
|
||||
attributes so that they do not share the same page.
|
||||
|
||||
`separate-code` adds paddings only between executable and non-executable
|
||||
segments. This is the default.
|
||||
|
||||
`noseparate-code` does not add any paddings between segments.
|
||||
|
||||
* `-z defs`, `-z nodefs`:
|
||||
Report undefined symbols (even with `--shared`).
|
||||
|
||||
* `-z shstk`:
|
||||
Enforce shadow stack by turning `GNU_PROPERTY_X86_FEATURE_1_SHSTK` bit in
|
||||
`.note.gnu.property` output section. Shadow stack is part of Intel
|
||||
Control-flow Enforcement Technology (CET), which is available since Tiger
|
||||
Lake (2020).
|
||||
|
||||
* `-z start_stop_visibility`=[ `hidden` | `protected` ]:
|
||||
If a section name is valid as a C identifier (i.e., it matches
|
||||
`/^[_a-zA-Z][_a-zA-Z0-9]*$/`), mold creates `__start_SECNAME` and
|
||||
`__stop_SECNAME` symbols to mark the beginning and end of the section,
|
||||
where `SECNAME` is the section name.
|
||||
|
||||
You can make these marker symbols visible from other ELF modules by passing
|
||||
`-z start_stop_visibility=protected`. Default is `hidden`.
|
||||
|
||||
* `-z text`, `-z notext`, `-z textoff`:
|
||||
`mold` by default reports an error if dynamic relocations are created in
|
||||
read-only sections. If `-z notext` or `-z textoff` are given, `mold` creates
|
||||
such dynamic relocations without reporting an error. `-z text` restores the
|
||||
default behavior.
|
||||
|
||||
* `-z max-page-size`=_number_:
|
||||
Some CPU ISAs support multiple memory page sizes. This option specifies the
|
||||
maximum page size that an output binary can run on. In general, binaries
|
||||
built for a larger page size can run on a system with a smaller page size,
|
||||
but not vice versa. The default value is 4 KiB for i386, x86-64, and RISC-V,
|
||||
and 64 KiB for ARM64.
|
||||
|
||||
* `-z nodefaultlib`:
|
||||
Make the dynamic loader ignore default search paths.
|
||||
|
||||
* `-z nodelete`:
|
||||
Mark DSO non-deletable at runtime.
|
||||
|
||||
* `-z nodlopen`:
|
||||
Mark DSO not available to dlopen(3). This option makes it possible for the
|
||||
linker to optimize thread-local variable accesses by rewriting instructions
|
||||
for some targets.
|
||||
|
||||
* `-z nodump`:
|
||||
Mark DSO not available to dldump(3).
|
||||
|
||||
* `-z nocopyreloc`:
|
||||
Do not create copy relocations.
|
||||
|
||||
* `-z initfirst`:
|
||||
Mark DSO to be initialized first at runtime.
|
||||
|
||||
* `-z interpose`:
|
||||
Mark object to interpose all DSOs but executable.
|
||||
|
||||
* `-(`, `-)`, `-EL`, `-O`_number_, `--dc`, `--dp`, `--end-group`, `--no-add-needed`, `--no-copy-dt-needed-entries`, `--nostdlib`, `--rpath-link=Ar dir`, `--sort-common`, `--sort-section`, `--start-group`, `--warn-constructors`, `--warn-once`, `--fix-cortex-a53-835769`, `--fix-cortex-a53-843419`, `-z combreloc`, `-z common-page-size`, `-z nocombreloc`:
|
||||
Ignored
|
||||
|
||||
## ENVIRONMENT VARIABLES
|
||||
|
||||
* `MOLD_JOBS`:
|
||||
If this variable is set to `1`, only one `mold` process will run at a time.
|
||||
If a new mold process is initiated while another is already active, the new
|
||||
process will wait until the active one completes before starting.
|
||||
|
||||
The primary reason for this environment variable is to minimize peak memory
|
||||
usage. Since mold is designed to operate with high parallelism, running
|
||||
multiple mold instances simultaneously may not be beneficial. If you execute
|
||||
N instances of mold concurrently, it could require N times the time and N
|
||||
times the memory. On the other hand, running them one after the other might
|
||||
still take N times longer, but the peak memory usage would be the same as
|
||||
running just a single instance.
|
||||
|
||||
If your build system invokes multiple linker processes simultaneously and
|
||||
some of them often get killed due to out-of-memory errors, you might
|
||||
consider setting this environment variable to `1` to see if it addresses the
|
||||
OOM issue.
|
||||
|
||||
Currently, any value other than `1` is silently ignored.
|
||||
|
||||
* `MOLD_DEBUG`:
|
||||
If this variable is set to a non-empty string, `mold` embeds its
|
||||
command-line options in the output file's `.comment` section.
|
||||
|
||||
* `MOLD_REPRO`:
|
||||
Setting this variable to a non-empty string has the same effect as passing
|
||||
the `--repro` option.
|
||||
|
||||
## SEE ALSO
|
||||
|
||||
gold(1), ld(1), elf(5), ld.so(8)
|
||||
|
||||
## AUTHOR
|
||||
|
||||
Rui Ueyama <ruiu@cs.stanford.edu>
|
||||
|
||||
## BUGS
|
||||
|
||||
Report bugs to <https://github.com/rui314/mold/issues>.
|
||||
@@ -0,0 +1,461 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static void write_adr(u8 *buf, u64 val) {
|
||||
u32 hi = (val & 0x1ffffc) << 3;
|
||||
u32 lo = (val & 3) << 29;
|
||||
*(u32 *)buf = (*(u32 *)buf & 0x9f00001f) | hi | lo;
|
||||
}
|
||||
|
||||
// Returns [hi:lo] bits of val.
|
||||
static u64 bits(u64 val, u64 hi, u64 lo) {
|
||||
return (val >> lo) & (((u64)1 << (hi - lo + 1)) - 1);
|
||||
}
|
||||
|
||||
static u64 page(u64 val) {
|
||||
return val & ~(u64)0xfff;
|
||||
}
|
||||
|
||||
template <>
|
||||
void GotPltSection<AARCH64>::copy_buf(Context<AARCH64> &ctx) {
|
||||
u64 *buf = (u64 *)(ctx.buf + this->shdr.sh_offset);
|
||||
|
||||
// The first slot of .got.plt points to _DYNAMIC.
|
||||
buf[0] = ctx.dynamic ? ctx.dynamic->shdr.sh_addr : 0;
|
||||
buf[1] = 0;
|
||||
buf[2] = 0;
|
||||
|
||||
for (Symbol<AARCH64> *sym : ctx.plt->symbols)
|
||||
buf[sym->get_gotplt_idx(ctx)] = ctx.plt->shdr.sh_addr;
|
||||
}
|
||||
|
||||
static void write_plt_header(Context<AARCH64> &ctx, u8 *buf) {
|
||||
// Write PLT header
|
||||
static const u8 plt0[] = {
|
||||
0xf0, 0x7b, 0xbf, 0xa9, // stp x16, x30, [sp,#-16]!
|
||||
0x10, 0x00, 0x00, 0x90, // adrp x16, .got.plt[2]
|
||||
0x11, 0x02, 0x40, 0xf9, // ldr x17, [x16, .got.plt[2]]
|
||||
0x10, 0x02, 0x00, 0x91, // add x16, x16, .got.plt[2]
|
||||
0x20, 0x02, 0x1f, 0xd6, // br x17
|
||||
0x1f, 0x20, 0x03, 0xd5, // nop
|
||||
0x1f, 0x20, 0x03, 0xd5, // nop
|
||||
0x1f, 0x20, 0x03, 0xd5, // nop
|
||||
};
|
||||
|
||||
u64 gotplt = ctx.gotplt->shdr.sh_addr + 16;
|
||||
u64 plt = ctx.plt->shdr.sh_addr;
|
||||
|
||||
memcpy(buf, plt0, sizeof(plt0));
|
||||
write_adr(buf + 4, bits(page(gotplt) - page(plt + 4), 32, 12));
|
||||
*(u32 *)(buf + 8) |= bits(gotplt, 11, 3) << 10;
|
||||
*(u32 *)(buf + 12) |= ((gotplt) & 0xfff) << 10;
|
||||
}
|
||||
|
||||
static void write_plt_entry(Context<AARCH64> &ctx, u8 *buf, Symbol<AARCH64> &sym) {
|
||||
u8 *ent = buf + sym.get_plt_idx(ctx) * AARCH64::plt_size;
|
||||
|
||||
static const u8 data[] = {
|
||||
0x10, 0x00, 0x00, 0x90, // adrp x16, .got.plt[n]
|
||||
0x11, 0x02, 0x40, 0xf9, // ldr x17, [x16, .got.plt[n]]
|
||||
0x10, 0x02, 0x00, 0x91, // add x16, x16, .got.plt[n]
|
||||
0x20, 0x02, 0x1f, 0xd6, // br x17
|
||||
};
|
||||
|
||||
u64 gotplt = sym.get_gotplt_addr(ctx);
|
||||
u64 plt = sym.get_plt_addr(ctx);
|
||||
|
||||
memcpy(ent, data, sizeof(data));
|
||||
write_adr(ent, bits(page(gotplt) - page(plt), 32, 12));
|
||||
*(u32 *)(ent + 4) |= bits(gotplt, 11, 3) << 10;
|
||||
*(u32 *)(ent + 8) |= (gotplt & 0xfff) << 10;
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltSection<AARCH64>::copy_buf(Context<AARCH64> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
write_plt_header(ctx, buf);
|
||||
for (Symbol<AARCH64> *sym : symbols)
|
||||
write_plt_entry(ctx, buf, *sym);
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltGotSection<AARCH64>::copy_buf(Context<AARCH64> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
for (Symbol<AARCH64> *sym : symbols) {
|
||||
u8 *ent = buf + sym->get_pltgot_idx(ctx) * AARCH64::pltgot_size;
|
||||
|
||||
static const u8 data[] = {
|
||||
0x10, 0x00, 0x00, 0x90, // adrp x16, GOT[n]
|
||||
0x11, 0x02, 0x40, 0xf9, // ldr x17, [x16, GOT[n]]
|
||||
0x20, 0x02, 0x1f, 0xd6, // br x17
|
||||
0x1f, 0x20, 0x03, 0xd5, // nop
|
||||
};
|
||||
|
||||
u64 got = sym->get_got_addr(ctx);
|
||||
u64 plt = sym->get_plt_addr(ctx);
|
||||
|
||||
memcpy(ent, data, sizeof(data));
|
||||
write_adr(ent, bits(page(got) - page(plt), 32, 12));
|
||||
*(u32 *)(ent + 4) |= bits(got, 11, 3) << 10;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<AARCH64>::apply_reloc(Context<AARCH64> &ctx,
|
||||
ElfRel<AARCH64> &rel,
|
||||
u64 loc, u64 val) {
|
||||
u8 *base = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64:
|
||||
*(u64 *)(base + loc) = val;
|
||||
return;
|
||||
case R_AARCH64_PREL32:
|
||||
*(u32 *)(base + loc) = val - this->shdr.sh_addr - loc;
|
||||
return;
|
||||
case R_AARCH64_PREL64:
|
||||
*(u64 *)(base + loc) = val - this->shdr.sh_addr - loc;
|
||||
return;
|
||||
}
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<AARCH64>::apply_reloc_alloc(Context<AARCH64> &ctx, u8 *base) {
|
||||
ElfRel<AARCH64> *dynrel = nullptr;
|
||||
std::span<ElfRel<AARCH64>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
if (ctx.reldyn)
|
||||
dynrel = (ElfRel<AARCH64> *)(ctx.buf + ctx.reldyn->shdr.sh_offset +
|
||||
file.reldyn_offset + this->reldyn_offset);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<AARCH64> &rel = rels[i];
|
||||
if (rel.r_type == R_AARCH64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<AARCH64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
const SectionFragmentRef<AARCH64> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
auto overflow_check = [&](i64 val, i64 lo, i64 hi) {
|
||||
if (val < lo || hi <= val)
|
||||
Error(ctx) << *this << ": relocation " << rel << " against "
|
||||
<< sym << " out of range: " << val << " is not in ["
|
||||
<< lo << ", " << hi << ")";
|
||||
};
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : rel.r_addend)
|
||||
#define P (output_section->shdr.sh_addr + offset + rel.r_offset)
|
||||
#define G (sym.get_got_addr(ctx) - ctx.got->shdr.sh_addr)
|
||||
#define GOT ctx.got->shdr.sh_addr
|
||||
|
||||
switch (rel_exprs[i]) {
|
||||
case R_BASEREL:
|
||||
*dynrel++ = {P, R_AARCH64_RELATIVE, 0, (i64)(S + A)};
|
||||
*(u64 *)loc = S + A;
|
||||
continue;
|
||||
case R_DYN:
|
||||
*dynrel++ = {P, R_AARCH64_ABS64, (u32)sym.get_dynsym_idx(ctx), A};
|
||||
*(u64 *)loc = A;
|
||||
continue;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64:
|
||||
*(u64 *)loc = S + A;
|
||||
continue;
|
||||
case R_AARCH64_LDST8_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 0) << 10;
|
||||
continue;
|
||||
case R_AARCH64_LDST16_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 1) << 10;
|
||||
continue;
|
||||
case R_AARCH64_LDST32_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 2) << 10;
|
||||
continue;
|
||||
case R_AARCH64_LDST64_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 3) << 10;
|
||||
continue;
|
||||
case R_AARCH64_LDST128_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 4) << 10;
|
||||
continue;
|
||||
case R_AARCH64_ADD_ABS_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A, 11, 0) << 10;
|
||||
continue;
|
||||
case R_AARCH64_MOVW_UABS_G0_NC:
|
||||
*(u32 *)loc |= bits(S + A, 15, 0) << 5;
|
||||
continue;
|
||||
case R_AARCH64_MOVW_UABS_G1_NC:
|
||||
*(u32 *)loc |= bits(S + A, 31, 16) << 5;
|
||||
continue;
|
||||
case R_AARCH64_MOVW_UABS_G2_NC:
|
||||
*(u32 *)loc |= bits(S + A, 47, 32) << 5;
|
||||
continue;
|
||||
case R_AARCH64_MOVW_UABS_G3:
|
||||
*(u32 *)loc |= bits(S + A, 63, 48) << 5;
|
||||
continue;
|
||||
case R_AARCH64_ADR_GOT_PAGE: {
|
||||
i64 val = page(G + GOT + A) - page(P);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
write_adr(loc, bits(val, 32, 12));
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_ADR_PREL_PG_HI21: {
|
||||
i64 val = page(S + A) - page(P);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
write_adr(loc, bits(val, 32, 12));
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_CALL26:
|
||||
case R_AARCH64_JUMP26:
|
||||
if (sym.file) {
|
||||
i64 val = S + A - P;
|
||||
overflow_check(val, -((i64)1 << 26), (i64)1 << 26);
|
||||
*(u32 *)loc |= (val >> 2) & 0x3ffffff;
|
||||
} else {
|
||||
// On ARM, calling an weak undefined symbol jumps to the
|
||||
// next instruction.
|
||||
*(u32 *)loc |= 1;
|
||||
}
|
||||
continue;
|
||||
case R_AARCH64_PREL32: {
|
||||
i64 val = S + A - P;
|
||||
overflow_check(val, -((i64)1 << 31), (i64)1 << 32);
|
||||
*(u32 *)loc = val;
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_LD64_GOT_LO12_NC:
|
||||
*(u32 *)loc |= bits(G + GOT + A, 11, 3) << 10;
|
||||
continue;
|
||||
case R_AARCH64_LD64_GOTPAGE_LO15: {
|
||||
i64 val = G + GOT + A - page(GOT);
|
||||
overflow_check(val, 0, 1 << 15);
|
||||
*(u32 *)loc |= bits(val, 14, 3) << 10;
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21: {
|
||||
i64 val = page(sym.get_gottp_addr(ctx) + A) - page(P);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
write_adr(loc, bits(val, 32, 12));
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC:
|
||||
*(u32 *)loc |= bits(sym.get_gottp_addr(ctx) + A, 11, 3) << 10;
|
||||
continue;
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_HI12: {
|
||||
i64 val = S + A - ctx.tls_begin + 16;
|
||||
overflow_check(val, 0, (i64)1 << 24);
|
||||
*(u32 *)loc |= bits(val, 23, 12) << 10;
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12_NC:
|
||||
*(u32 *)loc |= bits(S + A - ctx.tls_begin + 16, 11, 0) << 10;
|
||||
continue;
|
||||
case R_AARCH64_TLSGD_ADR_PAGE21: {
|
||||
i64 val = page(sym.get_tlsgd_addr(ctx) + A) - page(P);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
write_adr(loc, bits(val, 32, 12));
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_TLSGD_ADD_LO12_NC:
|
||||
*(u32 *)loc |= bits(sym.get_tlsgd_addr(ctx) + A, 11, 0) << 10;
|
||||
continue;
|
||||
case R_AARCH64_TLSDESC_ADR_PAGE21: {
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
// adrp x0, 0 -> movz x0, #tls_ofset_hi, lsl #16
|
||||
i64 val = (S + A - ctx.tls_begin + 16);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
*(u32 *)loc = 0xd2a00000 | (bits(val, 32, 16) << 5);
|
||||
} else {
|
||||
i64 val = page(sym.get_tlsdesc_addr(ctx) + A) - page(P);
|
||||
overflow_check(val, -((i64)1 << 32), (i64)1 << 32);
|
||||
write_adr(loc, bits(val, 32, 12));
|
||||
}
|
||||
continue;
|
||||
}
|
||||
case R_AARCH64_TLSDESC_LD64_LO12:
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
// ldr x2, [x0] -> movk x0, #tls_ofset_lo
|
||||
u32 offset_lo = (S + A - ctx.tls_begin + 16) & 0xffff;
|
||||
*(u32 *)loc = 0xf2800000 | (offset_lo << 5);
|
||||
} else {
|
||||
*(u32 *)loc |= bits(sym.get_tlsdesc_addr(ctx) + A, 11, 3) << 10;
|
||||
}
|
||||
continue;
|
||||
case R_AARCH64_TLSDESC_ADD_LO12:
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
// add x0, x0, #0 -> nop
|
||||
*(u32 *)loc = 0xd503201f;
|
||||
} else {
|
||||
*(u32 *)loc |= bits(sym.get_tlsdesc_addr(ctx) + A, 11, 0) << 10;
|
||||
}
|
||||
continue;
|
||||
case R_AARCH64_TLSDESC_CALL:
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
// blr x2 -> nop
|
||||
*(u32 *)loc = 0xd503201f;
|
||||
}
|
||||
continue;
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
#undef P
|
||||
#undef G
|
||||
#undef GOT
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<AARCH64>::apply_reloc_nonalloc(Context<AARCH64> &ctx, u8 *base) {
|
||||
std::span<ElfRel<AARCH64>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<AARCH64> &rel = rels[i];
|
||||
if (rel.r_type == R_AARCH64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<AARCH64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
const SectionFragmentRef<AARCH64> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : rel.r_addend)
|
||||
#define P (output_section->shdr.sh_addr + offset + rel.r_offset)
|
||||
#define G (sym.get_got_addr(ctx) - ctx.got->shdr.sh_addr)
|
||||
#define GOT ctx.got->shdr.sh_addr
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64:
|
||||
*(u64 *)loc = S + A;
|
||||
continue;
|
||||
case R_AARCH64_ABS32:
|
||||
*(u32 *)loc = S + A;
|
||||
continue;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
break;
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
#undef P
|
||||
#undef G
|
||||
#undef GOT
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<AARCH64>::scan_relocations(Context<AARCH64> &ctx) {
|
||||
ASSERT(shdr.sh_flags & SHF_ALLOC);
|
||||
|
||||
this->reldyn_offset = file.num_dynrel * sizeof(ElfRel<AARCH64>);
|
||||
std::span<ElfRel<AARCH64>> rels = get_rels(ctx);
|
||||
bool is_writable = (shdr.sh_flags & SHF_WRITE);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<AARCH64> &rel = rels[i];
|
||||
if (rel.r_type == R_AARCH64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<AARCH64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (sym.get_type() == STT_GNU_IFUNC) {
|
||||
sym.flags |= NEEDS_GOT;
|
||||
sym.flags |= NEEDS_PLT;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // DSO
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_ADR_GOT_PAGE:
|
||||
case R_AARCH64_LD64_GOT_LO12_NC:
|
||||
case R_AARCH64_LD64_GOTPAGE_LO15:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_AARCH64_CALL26:
|
||||
case R_AARCH64_JUMP26:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21:
|
||||
case R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_AARCH64_ADR_PREL_PG_HI21: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, NONE, ERROR, ERROR }, // DSO
|
||||
{ NONE, NONE, ERROR, PLT }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSGD_ADR_PAGE21:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_ADR_PAGE21:
|
||||
case R_AARCH64_TLSDESC_LD64_LO12:
|
||||
case R_AARCH64_TLSDESC_ADD_LO12:
|
||||
if (!ctx.relax_tlsdesc || sym.is_imported)
|
||||
sym.flags |= NEEDS_TLSDESC;
|
||||
break;
|
||||
case R_AARCH64_ADD_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST16_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST32_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST64_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST128_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST8_ABS_LO12_NC:
|
||||
case R_AARCH64_MOVW_UABS_G0_NC:
|
||||
case R_AARCH64_MOVW_UABS_G1_NC:
|
||||
case R_AARCH64_MOVW_UABS_G2_NC:
|
||||
case R_AARCH64_MOVW_UABS_G3:
|
||||
case R_AARCH64_PREL32:
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_HI12:
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12_NC:
|
||||
case R_AARCH64_TLSGD_ADD_LO12_NC:
|
||||
case R_AARCH64_TLSDESC_CALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,468 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <>
|
||||
void GotPltSection<I386>::copy_buf(Context<I386> &ctx) {
|
||||
u32 *buf = (u32 *)(ctx.buf + this->shdr.sh_offset);
|
||||
|
||||
// The first slot of .got.plt points to _DYNAMIC.
|
||||
buf[0] = ctx.dynamic ? ctx.dynamic->shdr.sh_addr : 0;
|
||||
buf[1] = 0;
|
||||
buf[2] = 0;
|
||||
|
||||
for (Symbol<I386> *sym : ctx.plt->symbols)
|
||||
buf[sym->get_gotplt_idx(ctx)] = sym->get_plt_addr(ctx) + 6;
|
||||
}
|
||||
|
||||
static void write_plt_header(Context<I386> &ctx, u8 *buf) {
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 plt0[] = {
|
||||
0xff, 0xb3, 0x04, 0, 0, 0, // pushl 4(%ebx)
|
||||
0xff, 0xa3, 0x08, 0, 0, 0, // jmp *8(%ebx)
|
||||
0x90, 0x90, 0x90, 0x90, // nop
|
||||
};
|
||||
memcpy(buf, plt0, sizeof(plt0));
|
||||
} else {
|
||||
static const u8 plt0[] = {
|
||||
0xff, 0x35, 0, 0, 0, 0, // pushl (GOTPLT+4)
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *(GOTPLT+8)
|
||||
0x90, 0x90, 0x90, 0x90, // nop
|
||||
};
|
||||
memcpy(buf, plt0, sizeof(plt0));
|
||||
*(u32 *)(buf + 2) = ctx.gotplt->shdr.sh_addr + 4;
|
||||
*(u32 *)(buf + 8) = ctx.gotplt->shdr.sh_addr + 8;
|
||||
}
|
||||
}
|
||||
|
||||
static void write_plt_entry(Context<I386> &ctx, u8 *buf, Symbol<I386> &sym,
|
||||
i64 idx) {
|
||||
u8 *ent = buf + sym.get_plt_idx(ctx) * I386::plt_size;
|
||||
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 data[] = {
|
||||
0xff, 0xa3, 0, 0, 0, 0, // jmp *foo@GOT(%ebx)
|
||||
0x68, 0, 0, 0, 0, // pushl $reloc_offset
|
||||
0xe9, 0, 0, 0, 0, // jmp .PLT0@PC
|
||||
};
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = sym.get_gotplt_addr(ctx) - ctx.gotplt->shdr.sh_addr;
|
||||
} else {
|
||||
static const u8 data[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0x68, 0, 0, 0, 0, // pushl $reloc_offset
|
||||
0xe9, 0, 0, 0, 0, // jmp .PLT0@PC
|
||||
};
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = sym.get_gotplt_addr(ctx);
|
||||
}
|
||||
|
||||
*(u32 *)(ent + 7) = idx * sizeof(ElfRel<I386>);
|
||||
*(u32 *)(ent + 12) = ctx.plt->shdr.sh_addr - sym.get_plt_addr(ctx) - 16;
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltSection<I386>::copy_buf(Context<I386> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
write_plt_header(ctx, buf);
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++)
|
||||
write_plt_entry(ctx, buf, *symbols[i], i);
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltGotSection<I386>::copy_buf(Context<I386> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 data[] = {
|
||||
0xff, 0xa3, 0, 0, 0, 0, // jmp *foo@GOT(%ebx)
|
||||
0x66, 0x90, // nop
|
||||
};
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
u8 *ent = buf + i * sizeof(data);
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = symbols[i]->get_got_addr(ctx) - ctx.gotplt->shdr.sh_addr;
|
||||
}
|
||||
} else {
|
||||
static const u8 data[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0x66, 0x90, // nop
|
||||
};
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
u8 *ent = buf + i * sizeof(data);
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = symbols[i]->get_got_addr(ctx);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<I386>::apply_reloc(Context<I386> &ctx,
|
||||
ElfRel<I386> &rel,
|
||||
u64 loc, u64 val) {
|
||||
u8 *base = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_NONE:
|
||||
return;
|
||||
case R_386_32:
|
||||
*(u32 *)(base + loc) = val;
|
||||
return;
|
||||
case R_386_PC32:
|
||||
*(u32 *)(base + loc) = val - this->shdr.sh_addr - loc;
|
||||
return;
|
||||
}
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<I386>::apply_reloc_alloc(Context<I386> &ctx, u8 *base) {
|
||||
ElfRel<I386> *dynrel = nullptr;
|
||||
std::span<ElfRel<I386>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
if (ctx.reldyn)
|
||||
dynrel = (ElfRel<I386> *)(ctx.buf + ctx.reldyn->shdr.sh_offset +
|
||||
file.reldyn_offset + this->reldyn_offset);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<I386> &rel = rels[i];
|
||||
if (rel.r_type == R_386_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<I386> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
const SectionFragmentRef<I386> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
auto overflow_check = [&](i64 val, i64 lo, i64 hi) {
|
||||
if (val < lo || hi <= val)
|
||||
Error(ctx) << *this << ": relocation " << rel << " against "
|
||||
<< sym << " out of range: " << val << " is not in ["
|
||||
<< lo << ", " << hi << ")";
|
||||
};
|
||||
|
||||
auto write8 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 8);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write8s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 7), 1 << 7);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write16 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 16);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write16s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 15), 1 << 15);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : this->get_addend(rel))
|
||||
#define P (output_section->shdr.sh_addr + offset + rel.r_offset)
|
||||
#define G (sym.get_got_addr(ctx) - ctx.got->shdr.sh_addr)
|
||||
#define GOTPLT ctx.gotplt->shdr.sh_addr
|
||||
|
||||
switch (rel_exprs[i]) {
|
||||
case R_BASEREL:
|
||||
*dynrel++ = {P, R_386_RELATIVE, 0};
|
||||
*(u32 *)loc = S + A;
|
||||
continue;
|
||||
case R_DYN:
|
||||
*dynrel++ = {P, R_386_32, (u32)sym.get_dynsym_idx(ctx)};
|
||||
*(u32 *)loc = A;
|
||||
continue;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
write8(S + A);
|
||||
continue;
|
||||
case R_386_16:
|
||||
write16(S + A);
|
||||
continue;
|
||||
case R_386_32:
|
||||
*(u32 *)loc = S + A;
|
||||
continue;
|
||||
case R_386_PC8:
|
||||
write8s(S + A);
|
||||
continue;
|
||||
case R_386_PC16:
|
||||
write16s(S + A);
|
||||
continue;
|
||||
case R_386_PC32:
|
||||
case R_386_PLT32:
|
||||
*(u32 *)loc = S + A - P;
|
||||
continue;
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X:
|
||||
*(u32 *)loc = sym.get_got_addr(ctx) + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_GOTOFF:
|
||||
*(u32 *)loc = S + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_GOTPC:
|
||||
*(u32 *)loc = GOTPLT + A - P;
|
||||
continue;
|
||||
case R_386_TLS_GOTIE:
|
||||
*(u32 *)loc = sym.get_gottp_addr(ctx) + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_TLS_LE:
|
||||
*(u32 *)loc = S + A - ctx.tls_end;
|
||||
continue;
|
||||
case R_386_TLS_IE:
|
||||
*(u32 *)loc = sym.get_gottp_addr(ctx) + A;
|
||||
continue;
|
||||
case R_386_TLS_GD:
|
||||
*(u32 *)loc = sym.get_tlsgd_addr(ctx) + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_TLS_LDM:
|
||||
*(u32 *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_TLS_LDO_32:
|
||||
*(u32 *)loc = S + A - ctx.tls_begin;
|
||||
continue;
|
||||
case R_386_SIZE32:
|
||||
*(u32 *)loc = sym.esym().st_size + A;
|
||||
continue;
|
||||
case R_386_TLS_GOTDESC:
|
||||
if (sym.get_tlsdesc_idx(ctx) == -1) {
|
||||
static const u8 insn[] = {
|
||||
0x8d, 0x05, 0, 0, 0, 0, // lea 0, %eax
|
||||
};
|
||||
memcpy(loc - 2, insn, sizeof(insn));
|
||||
*(u32 *)loc = S + A - ctx.tls_end;
|
||||
} else {
|
||||
*(u32 *)loc = sym.get_tlsdesc_addr(ctx) + A - GOTPLT;
|
||||
}
|
||||
continue;
|
||||
case R_386_TLS_DESC_CALL:
|
||||
if (ctx.arg.relax && !ctx.arg.shared) {
|
||||
// call *(%rax) -> nop
|
||||
loc[0] = 0x66;
|
||||
loc[1] = 0x90;
|
||||
}
|
||||
continue;
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
#undef P
|
||||
#undef G
|
||||
#undef GOTPLT
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<I386>::apply_reloc_nonalloc(Context<I386> &ctx, u8 *base) {
|
||||
std::span<ElfRel<I386>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<I386> &rel = rels[i];
|
||||
if (rel.r_type == R_386_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<I386> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
const SectionFragmentRef<I386> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
auto overflow_check = [&](i64 val, i64 lo, i64 hi) {
|
||||
if (val < lo || hi <= val)
|
||||
Error(ctx) << *this << ": relocation " << rel << " against "
|
||||
<< sym << " out of range: " << val << " is not in ["
|
||||
<< lo << ", " << hi << ")";
|
||||
};
|
||||
|
||||
auto write8 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 8);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write8s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 7), 1 << 7);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write16 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 16);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write16s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 15), 1 << 15);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : this->get_addend(rel))
|
||||
#define G (sym.get_got_addr(ctx) - ctx.got->shdr.sh_addr)
|
||||
#define GOTPLT ctx.gotplt->shdr.sh_addr
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
write8(S + A);
|
||||
continue;
|
||||
case R_386_16:
|
||||
write16(S + A);
|
||||
continue;
|
||||
case R_386_32:
|
||||
*(u32 *)loc = S + A;
|
||||
continue;
|
||||
case R_386_PC8:
|
||||
write8s(S + A);
|
||||
continue;
|
||||
case R_386_PC16:
|
||||
write16s(S + A);
|
||||
continue;
|
||||
case R_386_PC32:
|
||||
*(u32 *)loc = S + A;
|
||||
continue;
|
||||
case R_386_GOTPC:
|
||||
*(u32 *)loc = GOTPLT + A;
|
||||
continue;
|
||||
case R_386_GOTOFF:
|
||||
*(u32 *)loc = S + A - GOTPLT;
|
||||
continue;
|
||||
case R_386_TLS_LDO_32:
|
||||
*(u32 *)loc = S + A - ctx.tls_begin;
|
||||
continue;
|
||||
case R_386_SIZE32:
|
||||
*(u32 *)loc = sym.esym().st_size + A;
|
||||
continue;
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
#undef GOTPLT
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<I386>::scan_relocations(Context<I386> &ctx) {
|
||||
ASSERT(shdr.sh_flags & SHF_ALLOC);
|
||||
|
||||
this->reldyn_offset = file.num_dynrel * sizeof(ElfRel<I386>);
|
||||
std::span<ElfRel<I386>> rels = get_rels(ctx);
|
||||
bool is_writable = (shdr.sh_flags & SHF_WRITE);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<I386> &rel = rels[i];
|
||||
if (rel.r_type == R_386_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<I386> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (sym.get_type() == STT_GNU_IFUNC) {
|
||||
sym.flags |= NEEDS_GOT;
|
||||
sym.flags |= NEEDS_PLT;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
case R_386_16: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, ERROR, ERROR, ERROR }, // DSO
|
||||
{ NONE, ERROR, ERROR, ERROR }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_386_32: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // DSO
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_386_PC8:
|
||||
case R_386_PC16: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ ERROR, NONE, ERROR, ERROR }, // DSO
|
||||
{ ERROR, NONE, COPYREL, PLT }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_386_PC32: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ BASEREL, NONE, ERROR, ERROR }, // DSO
|
||||
{ BASEREL, NONE, COPYREL, PLT }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X:
|
||||
case R_386_GOTPC:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_386_PLT32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_386_TLS_GOTIE:
|
||||
case R_386_TLS_LE:
|
||||
case R_386_TLS_IE:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_386_TLS_GD:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_386_TLS_LDM:
|
||||
sym.flags |= NEEDS_TLSLD;
|
||||
break;
|
||||
case R_386_TLS_GOTDESC:
|
||||
if (!ctx.arg.relax || ctx.arg.shared)
|
||||
sym.flags |= NEEDS_TLSDESC;
|
||||
break;
|
||||
case R_386_GOTOFF:
|
||||
case R_386_TLS_LDO_32:
|
||||
case R_386_SIZE32:
|
||||
case R_386_TLS_DESC_CALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,676 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <>
|
||||
void GotPltSection<X86_64>::copy_buf(Context<X86_64> &ctx) {
|
||||
u64 *buf = (u64 *)(ctx.buf + this->shdr.sh_offset);
|
||||
|
||||
// The first slot of .got.plt points to _DYNAMIC, as requested by
|
||||
// the x86-64 psABI. The second and the third slots are reserved by
|
||||
// the psABI.
|
||||
buf[0] = ctx.dynamic ? ctx.dynamic->shdr.sh_addr : 0;
|
||||
buf[1] = 0;
|
||||
buf[2] = 0;
|
||||
|
||||
for (Symbol<X86_64> *sym : ctx.plt->symbols)
|
||||
buf[sym->get_gotplt_idx(ctx)] = sym->get_plt_addr(ctx) + 6;
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltSection<X86_64>::copy_buf(Context<X86_64> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
// Write PLT header
|
||||
static const u8 plt0[] = {
|
||||
0xff, 0x35, 0, 0, 0, 0, // pushq GOTPLT+8(%rip)
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *GOTPLT+16(%rip)
|
||||
0x0f, 0x1f, 0x40, 0x00, // nop
|
||||
};
|
||||
|
||||
memcpy(buf, plt0, sizeof(plt0));
|
||||
*(u32 *)(buf + 2) = ctx.gotplt->shdr.sh_addr - this->shdr.sh_addr + 2;
|
||||
*(u32 *)(buf + 8) = ctx.gotplt->shdr.sh_addr - this->shdr.sh_addr + 4;
|
||||
|
||||
// Write PLT entries
|
||||
i64 relplt_idx = 0;
|
||||
|
||||
static const u8 data[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOTPLT
|
||||
0x68, 0, 0, 0, 0, // push $index_in_relplt
|
||||
0xe9, 0, 0, 0, 0, // jmp PLT[0]
|
||||
};
|
||||
|
||||
for (Symbol<X86_64> *sym : symbols) {
|
||||
u8 *ent = buf + sym->get_plt_idx(ctx) * X86_64::plt_size;
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = sym->get_gotplt_addr(ctx) - sym->get_plt_addr(ctx) - 6;
|
||||
*(u32 *)(ent + 7) = relplt_idx++;
|
||||
*(u32 *)(ent + 12) = this->shdr.sh_addr - sym->get_plt_addr(ctx) - 16;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void PltGotSection<X86_64>::copy_buf(Context<X86_64> &ctx) {
|
||||
u8 *buf = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
static const u8 data[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0x66, 0x90, // nop
|
||||
};
|
||||
|
||||
for (Symbol<X86_64> *sym : symbols) {
|
||||
u8 *ent = buf + sym->get_pltgot_idx(ctx) * X86_64::pltgot_size;
|
||||
memcpy(ent, data, sizeof(data));
|
||||
*(u32 *)(ent + 2) = sym->get_got_addr(ctx) - sym->get_plt_addr(ctx) - 6;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<X86_64>::apply_reloc(Context<X86_64> &ctx,
|
||||
ElfRel<X86_64> &rel,
|
||||
u64 loc, u64 val) {
|
||||
u8 *base = ctx.buf + this->shdr.sh_offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_NONE:
|
||||
return;
|
||||
case R_X86_64_32:
|
||||
*(u32 *)(base + loc) = val;
|
||||
return;
|
||||
case R_X86_64_64:
|
||||
*(u64 *)(base + loc) = val;
|
||||
return;
|
||||
case R_X86_64_PC32:
|
||||
*(u32 *)(base + loc) = val - this->shdr.sh_addr - loc;
|
||||
return;
|
||||
case R_X86_64_PC64:
|
||||
*(u64 *)(base + loc) = val - this->shdr.sh_addr - loc;
|
||||
return;
|
||||
}
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
static u32 relax_gotpcrelx(u8 *loc) {
|
||||
switch ((loc[0] << 8) | loc[1]) {
|
||||
case 0xff15: return 0x90e8; // call *0(%rip) -> call 0
|
||||
case 0xff25: return 0x90e9; // jmp *0(%rip) -> jmp 0
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_rex_gotpcrelx(u8 *loc) {
|
||||
switch ((loc[0] << 16) | (loc[1] << 8) | loc[2]) {
|
||||
case 0x488b05: return 0x488d05; // mov 0(%rip), %rax -> lea 0(%rip), %rax
|
||||
case 0x488b0d: return 0x488d0d; // mov 0(%rip), %rcx -> lea 0(%rip), %rcx
|
||||
case 0x488b15: return 0x488d15; // mov 0(%rip), %rdx -> lea 0(%rip), %rdx
|
||||
case 0x488b1d: return 0x488d1d; // mov 0(%rip), %rbx -> lea 0(%rip), %rbx
|
||||
case 0x488b25: return 0x488d25; // mov 0(%rip), %rsp -> lea 0(%rip), %rsp
|
||||
case 0x488b2d: return 0x488d2d; // mov 0(%rip), %rbp -> lea 0(%rip), %rbp
|
||||
case 0x488b35: return 0x488d35; // mov 0(%rip), %rsi -> lea 0(%rip), %rsi
|
||||
case 0x488b3d: return 0x488d3d; // mov 0(%rip), %rdi -> lea 0(%rip), %rdi
|
||||
case 0x4c8b05: return 0x4c8d05; // mov 0(%rip), %r8 -> lea 0(%rip), %r8
|
||||
case 0x4c8b0d: return 0x4c8d0d; // mov 0(%rip), %r9 -> lea 0(%rip), %r9
|
||||
case 0x4c8b15: return 0x4c8d15; // mov 0(%rip), %r10 -> lea 0(%rip), %r10
|
||||
case 0x4c8b1d: return 0x4c8d1d; // mov 0(%rip), %r11 -> lea 0(%rip), %r11
|
||||
case 0x4c8b25: return 0x4c8d25; // mov 0(%rip), %r12 -> lea 0(%rip), %r12
|
||||
case 0x4c8b2d: return 0x4c8d2d; // mov 0(%rip), %r13 -> lea 0(%rip), %r13
|
||||
case 0x4c8b35: return 0x4c8d35; // mov 0(%rip), %r14 -> lea 0(%rip), %r14
|
||||
case 0x4c8b3d: return 0x4c8d3d; // mov 0(%rip), %r15 -> lea 0(%rip), %r15
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_gottpoff(u8 *loc) {
|
||||
switch ((loc[0] << 16) | (loc[1] << 8) | loc[2]) {
|
||||
case 0x488b05: return 0x48c7c0; // mov 0(%rip), %rax -> mov $0, %rax
|
||||
case 0x488b0d: return 0x48c7c1; // mov 0(%rip), %rcx -> mov $0, %rcx
|
||||
case 0x488b15: return 0x48c7c2; // mov 0(%rip), %rdx -> mov $0, %rdx
|
||||
case 0x488b1d: return 0x48c7c3; // mov 0(%rip), %rbx -> mov $0, %rbx
|
||||
case 0x488b25: return 0x48c7c4; // mov 0(%rip), %rsp -> mov $0, %rsp
|
||||
case 0x488b2d: return 0x48c7c5; // mov 0(%rip), %rbp -> mov $0, %rbp
|
||||
case 0x488b35: return 0x48c7c6; // mov 0(%rip), %rsi -> mov $0, %rsi
|
||||
case 0x488b3d: return 0x48c7c7; // mov 0(%rip), %rdi -> mov $0, %rdi
|
||||
case 0x4c8b05: return 0x49c7c0; // mov 0(%rip), %r8 -> mov $0, %r8
|
||||
case 0x4c8b0d: return 0x49c7c1; // mov 0(%rip), %r9 -> mov $0, %r9
|
||||
case 0x4c8b15: return 0x49c7c2; // mov 0(%rip), %r10 -> mov $0, %r10
|
||||
case 0x4c8b1d: return 0x49c7c3; // mov 0(%rip), %r11 -> mov $0, %r11
|
||||
case 0x4c8b25: return 0x49c7c4; // mov 0(%rip), %r12 -> mov $0, %r12
|
||||
case 0x4c8b2d: return 0x49c7c5; // mov 0(%rip), %r13 -> mov $0, %r13
|
||||
case 0x4c8b35: return 0x49c7c6; // mov 0(%rip), %r14 -> mov $0, %r14
|
||||
case 0x4c8b3d: return 0x49c7c7; // mov 0(%rip), %r15 -> mov $0, %r15
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
// Apply relocations to SHF_ALLOC sections (i.e. sections that are
|
||||
// mapped to memory at runtime) based on the result of
|
||||
// scan_relocations().
|
||||
template <>
|
||||
void InputSection<X86_64>::apply_reloc_alloc(Context<X86_64> &ctx, u8 *base) {
|
||||
ElfRel<X86_64> *dynrel = nullptr;
|
||||
std::span<ElfRel<X86_64>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
if (ctx.reldyn)
|
||||
dynrel = (ElfRel<X86_64> *)(ctx.buf + ctx.reldyn->shdr.sh_offset +
|
||||
file.reldyn_offset + this->reldyn_offset);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<X86_64> &rel = rels[i];
|
||||
if (rel.r_type == R_X86_64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<X86_64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
const SectionFragmentRef<X86_64> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
auto overflow_check = [&](i64 val, i64 lo, i64 hi) {
|
||||
if (val < lo || hi <= val)
|
||||
Error(ctx) << *this << ": relocation " << rel << " against "
|
||||
<< sym << " out of range: " << val << " is not in ["
|
||||
<< lo << ", " << hi << ")";
|
||||
};
|
||||
|
||||
auto write8 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 8);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write8s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 7), 1 << 7);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write16 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 16);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write16s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 15), 1 << 15);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32 = [&](u64 val) {
|
||||
overflow_check(val, 0, (i64)1 << 32);
|
||||
*(u32 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32s = [&](u64 val) {
|
||||
overflow_check(val, -((i64)1 << 31), (i64)1 << 31);
|
||||
*(u32 *)loc = val;
|
||||
};
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : rel.r_addend)
|
||||
#define P (output_section->shdr.sh_addr + offset + rel.r_offset)
|
||||
#define G (sym.get_got_addr(ctx) - ctx.got->shdr.sh_addr)
|
||||
#define GOT ctx.got->shdr.sh_addr
|
||||
|
||||
switch (rel_exprs[i]) {
|
||||
case R_BASEREL:
|
||||
*dynrel++ = {P, R_X86_64_RELATIVE, 0, (i64)(S + A)};
|
||||
*(u64 *)loc = S + A;
|
||||
continue;
|
||||
case R_DYN:
|
||||
*dynrel++ = {P, R_X86_64_64, (u32)sym.get_dynsym_idx(ctx), A};
|
||||
*(u64 *)loc = A;
|
||||
continue;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
write8(S + A);
|
||||
continue;
|
||||
case R_X86_64_16:
|
||||
write16(S + A);
|
||||
continue;
|
||||
case R_X86_64_32:
|
||||
write32(S + A);
|
||||
continue;
|
||||
case R_X86_64_32S:
|
||||
write32s(S + A);
|
||||
continue;
|
||||
case R_X86_64_64:
|
||||
*(u64 *)loc = S + A;
|
||||
continue;
|
||||
case R_X86_64_PC8:
|
||||
write8s(S + A - P);
|
||||
continue;
|
||||
case R_X86_64_PC16:
|
||||
write16s(S + A - P);
|
||||
continue;
|
||||
case R_X86_64_PC32:
|
||||
write32s(S + A - P);
|
||||
continue;
|
||||
case R_X86_64_PC64:
|
||||
*(u64 *)loc = S + A - P;
|
||||
continue;
|
||||
case R_X86_64_PLT32:
|
||||
write32s(S + A - P);
|
||||
continue;
|
||||
case R_X86_64_GOT32:
|
||||
write32s(G + A);
|
||||
continue;
|
||||
case R_X86_64_GOT64:
|
||||
*(u64 *)loc = G + A;
|
||||
continue;
|
||||
case R_X86_64_GOTPC32:
|
||||
write32s(GOT + A - P);
|
||||
continue;
|
||||
case R_X86_64_GOTPC64:
|
||||
*(u64 *)loc = GOT + A - P;
|
||||
continue;
|
||||
case R_X86_64_GOTPCREL:
|
||||
write32s(G + GOT + A - P);
|
||||
continue;
|
||||
case R_X86_64_GOTPCREL64:
|
||||
*(u64 *)loc = G + GOT + A - P;
|
||||
continue;
|
||||
case R_X86_64_GOTPCRELX:
|
||||
if (sym.get_got_idx(ctx) == -1) {
|
||||
u32 insn = relax_gotpcrelx(loc - 2);
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(S + A - P);
|
||||
} else {
|
||||
write32s(G + GOT + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_REX_GOTPCRELX:
|
||||
if (sym.get_got_idx(ctx) == -1) {
|
||||
u32 insn = relax_rex_gotpcrelx(loc - 3);
|
||||
loc[-3] = insn >> 16;
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(S + A - P);
|
||||
} else {
|
||||
write32s(G + GOT + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_TLSGD:
|
||||
if (sym.get_tlsgd_idx(ctx) == -1) {
|
||||
// Relax GD to LE
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x8d, 0x80, 0, 0, 0, 0, // lea 0(%rax), %rax
|
||||
};
|
||||
memcpy(loc - 4, insn, sizeof(insn));
|
||||
|
||||
i64 val = S - ctx.tls_end + A + 4;
|
||||
overflow_check(val, -((i64)1 << 31), (i64)1 << 31);
|
||||
*(u32 *)(loc + 8) = val;
|
||||
i++;
|
||||
} else {
|
||||
write32s(sym.get_tlsgd_addr(ctx) + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_TLSLD:
|
||||
if (ctx.got->tlsld_idx == -1) {
|
||||
// Relax LD to LE
|
||||
static const u8 insn[] = {
|
||||
0x66, 0x66, 0x66, // (padding)
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
i++;
|
||||
} else {
|
||||
write32s(ctx.got->get_tlsld_addr(ctx) + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_DTPOFF32:
|
||||
if (ctx.arg.relax && !ctx.arg.shared)
|
||||
write32s(S + A - ctx.tls_end);
|
||||
else
|
||||
write32s(S + A - ctx.tls_begin);
|
||||
continue;
|
||||
case R_X86_64_DTPOFF64:
|
||||
if (ctx.arg.relax && !ctx.arg.shared)
|
||||
*(u64 *)loc = S + A - ctx.tls_end;
|
||||
else
|
||||
*(u64 *)loc = S + A - ctx.tls_begin;
|
||||
continue;
|
||||
case R_X86_64_TPOFF32:
|
||||
write32s(S + A - ctx.tls_end);
|
||||
continue;
|
||||
case R_X86_64_TPOFF64:
|
||||
*(u64 *)loc = S + A - ctx.tls_end;
|
||||
continue;
|
||||
case R_X86_64_GOTTPOFF:
|
||||
if (sym.get_gottp_idx(ctx) == -1) {
|
||||
u32 insn = relax_gottpoff(loc - 3);
|
||||
loc[-3] = insn >> 16;
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(S + A - ctx.tls_end + 4);
|
||||
} else {
|
||||
write32s(sym.get_gottp_addr(ctx) + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_GOTPC32_TLSDESC:
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
static const u8 insn[] = {
|
||||
0x48, 0xc7, 0xc0, 0, 0, 0, 0, // mov $0, %rax
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
write32s(S + A - ctx.tls_end + 4);
|
||||
} else {
|
||||
write32s(sym.get_tlsdesc_addr(ctx) + A - P);
|
||||
}
|
||||
continue;
|
||||
case R_X86_64_SIZE32:
|
||||
write32(sym.esym().st_size + A);
|
||||
continue;
|
||||
case R_X86_64_SIZE64:
|
||||
*(u64 *)loc = sym.esym().st_size + A;
|
||||
continue;
|
||||
case R_X86_64_TLSDESC_CALL:
|
||||
if (ctx.relax_tlsdesc && !sym.is_imported) {
|
||||
// call *(%rax) -> nop
|
||||
loc[0] = 0x66;
|
||||
loc[1] = 0x90;
|
||||
}
|
||||
continue;
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
#undef P
|
||||
#undef G
|
||||
#undef GOT
|
||||
}
|
||||
}
|
||||
|
||||
// This function is responsible for applying relocations against
|
||||
// non-SHF_ALLOC sections (i.e. sections that are not mapped to memory
|
||||
// at runtime).
|
||||
//
|
||||
// Relocations against non-SHF_ALLOC sections are much easier to
|
||||
// handle than that against SHF_ALLOC sections. It is because, since
|
||||
// they are not mapped to memory, they don't contain any variable or
|
||||
// function and never need PLT or GOT. Non-SHF_ALLOC sections are
|
||||
// mostly debug info sections.
|
||||
//
|
||||
// Relocations against non-SHF_ALLOC sections are not scanned by
|
||||
// scan_relocations.
|
||||
template <>
|
||||
void InputSection<X86_64>::apply_reloc_nonalloc(Context<X86_64> &ctx, u8 *base) {
|
||||
std::span<ElfRel<X86_64>> rels = get_rels(ctx);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<X86_64> &rel = rels[i];
|
||||
if (rel.r_type == R_X86_64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<X86_64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
const SectionFragmentRef<X86_64> *ref = nullptr;
|
||||
if (rel_fragments && rel_fragments[frag_idx].idx == i)
|
||||
ref = &rel_fragments[frag_idx++];
|
||||
|
||||
auto overflow_check = [&](i64 val, i64 lo, i64 hi) {
|
||||
if (val < lo || hi <= val)
|
||||
Error(ctx) << *this << ": relocation " << rel << " against "
|
||||
<< sym << " out of range: " << val << " is not in ["
|
||||
<< lo << ", " << hi << ")";
|
||||
};
|
||||
|
||||
auto write8 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 8);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write8s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 7), 1 << 7);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write16 = [&](u64 val) {
|
||||
overflow_check(val, 0, 1 << 16);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write16s = [&](u64 val) {
|
||||
overflow_check(val, -(1 << 15), 1 << 15);
|
||||
*(u16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32 = [&](u64 val) {
|
||||
overflow_check(val, 0, (i64)1 << 32);
|
||||
*(u32 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32s = [&](u64 val) {
|
||||
overflow_check(val, -((i64)1 << 31), (i64)1 << 31);
|
||||
*(u32 *)loc = val;
|
||||
};
|
||||
|
||||
#define S (ref ? ref->frag->get_addr(ctx) : sym.get_addr(ctx))
|
||||
#define A (ref ? ref->addend : rel.r_addend)
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
write8(S + A);
|
||||
break;
|
||||
case R_X86_64_16:
|
||||
write16(S + A);
|
||||
break;
|
||||
case R_X86_64_32:
|
||||
write32(S + A);
|
||||
break;
|
||||
case R_X86_64_32S:
|
||||
write32s(S + A);
|
||||
break;
|
||||
case R_X86_64_64:
|
||||
*(u64 *)loc = S + A;
|
||||
break;
|
||||
case R_X86_64_DTPOFF32:
|
||||
write32s(S + A - ctx.tls_begin);
|
||||
break;
|
||||
case R_X86_64_DTPOFF64:
|
||||
*(u64 *)loc = S + A - ctx.tls_begin;
|
||||
break;
|
||||
case R_X86_64_SIZE32:
|
||||
write32(sym.esym().st_size + A);
|
||||
break;
|
||||
case R_X86_64_SIZE64:
|
||||
*(u64 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
break;
|
||||
}
|
||||
|
||||
#undef S
|
||||
#undef A
|
||||
}
|
||||
}
|
||||
|
||||
// Linker has to create data structures in an output file to apply
|
||||
// some type of relocations. For example, if a relocation refers a GOT
|
||||
// or a PLT entry of a symbol, linker has to create an entry in .got
|
||||
// or in .plt for that symbol. In order to fix the file layout, we
|
||||
// need to scan relocations.
|
||||
template <>
|
||||
void InputSection<X86_64>::scan_relocations(Context<X86_64> &ctx) {
|
||||
ASSERT(shdr.sh_flags & SHF_ALLOC);
|
||||
|
||||
this->reldyn_offset = file.num_dynrel * sizeof(ElfRel<X86_64>);
|
||||
std::span<ElfRel<X86_64>> rels = get_rels(ctx);
|
||||
bool is_writable = (shdr.sh_flags & SHF_WRITE);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<X86_64> &rel = rels[i];
|
||||
if (rel.r_type == R_X86_64_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<X86_64> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (!sym.file) {
|
||||
report_undef(ctx, sym);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (sym.get_type() == STT_GNU_IFUNC) {
|
||||
sym.flags |= NEEDS_GOT;
|
||||
sym.flags |= NEEDS_PLT;
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
case R_X86_64_16:
|
||||
case R_X86_64_32:
|
||||
case R_X86_64_32S: {
|
||||
// Dynamic linker does not support 8, 16 or 32-bit dynamic
|
||||
// relocations for these types of relocations. We report an
|
||||
// error if we cannot relocate them even at load-time.
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, ERROR, ERROR, ERROR }, // DSO
|
||||
{ NONE, ERROR, ERROR, ERROR }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_X86_64_64: {
|
||||
// Unlike the above, we can use R_X86_64_RELATIVE and R_86_64_64
|
||||
// relocations.
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // DSO
|
||||
{ NONE, BASEREL, DYNREL, DYNREL }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PC8:
|
||||
case R_X86_64_PC16:
|
||||
case R_X86_64_PC32: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ ERROR, NONE, ERROR, ERROR }, // DSO
|
||||
{ ERROR, NONE, COPYREL, PLT }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PC64: {
|
||||
Action table[][4] = {
|
||||
// Absolute Local Imported data Imported code
|
||||
{ BASEREL, NONE, ERROR, ERROR }, // DSO
|
||||
{ BASEREL, NONE, COPYREL, PLT }, // PIE
|
||||
{ NONE, NONE, COPYREL, PLT }, // PDE
|
||||
};
|
||||
dispatch(ctx, table, i);
|
||||
break;
|
||||
}
|
||||
case R_X86_64_GOT32:
|
||||
case R_X86_64_GOT64:
|
||||
case R_X86_64_GOTPC32:
|
||||
case R_X86_64_GOTPC64:
|
||||
case R_X86_64_GOTPCREL:
|
||||
case R_X86_64_GOTPCREL64:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_X86_64_GOTPCRELX: {
|
||||
if (rel.r_addend != -4)
|
||||
Fatal(ctx) << *this << ": bad r_addend for R_X86_64_GOTPCRELX";
|
||||
|
||||
bool do_relax = ctx.arg.relax && !sym.is_imported &&
|
||||
sym.is_relative(ctx) && relax_gotpcrelx(loc - 2);
|
||||
if (!do_relax)
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_REX_GOTPCRELX: {
|
||||
if (rel.r_addend != -4)
|
||||
Fatal(ctx) << *this << ": bad r_addend for R_X86_64_REX_GOTPCRELX";
|
||||
|
||||
bool do_relax = ctx.arg.relax && !sym.is_imported &&
|
||||
sym.is_relative(ctx) && relax_rex_gotpcrelx(loc - 3);
|
||||
if (!do_relax)
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PLT32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_X86_64_TLSGD:
|
||||
if (i + 1 == rels.size())
|
||||
Fatal(ctx) << *this
|
||||
<< ": TLSGD reloc must be followed by PLT32 or GOTPCREL";
|
||||
|
||||
if (ctx.arg.relax && !ctx.arg.shared && !sym.is_imported)
|
||||
i++;
|
||||
else
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_X86_64_TLSLD:
|
||||
if (i + 1 == rels.size())
|
||||
Fatal(ctx) << *this
|
||||
<< ": TLSGD reloc must be followed by PLT32 or GOTPCREL";
|
||||
if (sym.is_imported)
|
||||
Fatal(ctx) << *this << ": TLSLD reloc refers external symbol " << sym;
|
||||
|
||||
if (ctx.arg.relax && !ctx.arg.shared)
|
||||
i++;
|
||||
else
|
||||
sym.flags |= NEEDS_TLSLD;
|
||||
break;
|
||||
case R_X86_64_DTPOFF32:
|
||||
case R_X86_64_DTPOFF64:
|
||||
if (sym.is_imported)
|
||||
Fatal(ctx) << *this << ": DTPOFF reloc refers external symbol " << sym;
|
||||
break;
|
||||
case R_X86_64_GOTTPOFF: {
|
||||
ctx.has_gottp_rel = true;
|
||||
|
||||
bool do_relax = ctx.arg.relax && !ctx.arg.shared &&
|
||||
!sym.is_imported && relax_gottpoff(loc - 3);
|
||||
if (!do_relax)
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_GOTPC32_TLSDESC: {
|
||||
if (memcmp(loc - 3, "\x48\x8d\x05", 3))
|
||||
Fatal(ctx) << *this << ": GOTPC32_TLSDESC relocation is used"
|
||||
<< " against an invalid code sequence";
|
||||
|
||||
bool do_relax = ctx.relax_tlsdesc && !sym.is_imported;
|
||||
if (!do_relax)
|
||||
sym.flags |= NEEDS_TLSDESC;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_TPOFF32:
|
||||
case R_X86_64_TPOFF64:
|
||||
case R_X86_64_SIZE32:
|
||||
case R_X86_64_SIZE64:
|
||||
case R_X86_64_TLSDESC_CALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,123 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
struct ArHdr {
|
||||
char ar_name[16];
|
||||
char ar_date[12];
|
||||
char ar_uid[6];
|
||||
char ar_gid[6];
|
||||
char ar_mode[8];
|
||||
char ar_size[10];
|
||||
char ar_fmag[2];
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
std::vector<MemoryMappedFile<E> *>
|
||||
read_thin_archive_members(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
u8 *begin = mb->data(ctx);
|
||||
u8 *data = begin + 8;
|
||||
std::vector<MemoryMappedFile<E> *> vec;
|
||||
std::string_view strtab;
|
||||
|
||||
while (data < begin + mb->size()) {
|
||||
// Each header is aligned to a 2 byte boundary.
|
||||
if ((begin - data) % 2)
|
||||
data++;
|
||||
|
||||
ArHdr &hdr = *(ArHdr *)data;
|
||||
u8 *body = data + sizeof(hdr);
|
||||
u64 size = atol(hdr.ar_size);
|
||||
|
||||
// Read a string table.
|
||||
if (!memcmp(hdr.ar_name, "// ", 3)) {
|
||||
strtab = {(char *)body, size};
|
||||
data = body + size;
|
||||
continue;
|
||||
}
|
||||
|
||||
// Skip a symbol table.
|
||||
if (!memcmp(hdr.ar_name, "/ ", 2)) {
|
||||
data = body + size;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (hdr.ar_name[0] != '/')
|
||||
Fatal(ctx) << mb->name << ": filename is not stored as a long filename";
|
||||
|
||||
const char *start = strtab.data() + atoi(hdr.ar_name + 1);
|
||||
std::string name(start, (const char *)strstr(start, "/\n"));
|
||||
std::string path = name.starts_with('/') ?
|
||||
name : std::string(path_dirname(mb->name)) + "/" + name;
|
||||
vec.push_back(MemoryMappedFile<E>::must_open(ctx, path));
|
||||
data = body;
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<MemoryMappedFile<E> *>
|
||||
read_fat_archive_members(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
u8 *begin = mb->data(ctx);
|
||||
u8 *data = begin + 8;
|
||||
std::vector<MemoryMappedFile<E> *> vec;
|
||||
std::string_view strtab;
|
||||
|
||||
while (begin + mb->size() - data >= 2) {
|
||||
if ((begin - data) % 2)
|
||||
data++;
|
||||
|
||||
ArHdr &hdr = *(ArHdr *)data;
|
||||
u8 *body = data + sizeof(hdr);
|
||||
u64 size = atol(hdr.ar_size);
|
||||
data = body + size;
|
||||
|
||||
if (!memcmp(hdr.ar_name, "// ", 3)) {
|
||||
strtab = {(char *)body, size};
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!memcmp(hdr.ar_name, "/ ", 2) ||
|
||||
!memcmp(hdr.ar_name, "__.SYMDEF/", 10))
|
||||
continue;
|
||||
|
||||
std::string name;
|
||||
|
||||
if (hdr.ar_name[0] == '/') {
|
||||
const char *start = strtab.data() + atoi(hdr.ar_name + 1);
|
||||
name = {start, (const char *)strstr(start, "/\n")};
|
||||
} else {
|
||||
name = {hdr.ar_name, strchr(hdr.ar_name, '/')};
|
||||
}
|
||||
|
||||
vec.push_back(mb->slice(ctx, name, body - begin, size));
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<MemoryMappedFile<E> *>
|
||||
read_archive_members(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
switch (get_file_type(ctx, mb)) {
|
||||
case FileType::AR:
|
||||
return read_fat_archive_members(ctx, mb);
|
||||
case FileType::THIN_AR:
|
||||
return read_thin_archive_members(ctx, mb);
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template std::vector<MemoryMappedFile<E> *> \
|
||||
read_fat_archive_members(Context<E> &, MemoryMappedFile<E> *); \
|
||||
template std::vector<MemoryMappedFile<E> *> \
|
||||
read_thin_archive_members(Context<E> &, MemoryMappedFile<E> *); \
|
||||
template std::vector<MemoryMappedFile<E> *> \
|
||||
read_archive_members(Context<E> &, MemoryMappedFile<E> *);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+881
@@ -0,0 +1,881 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <sstream>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/types.h>
|
||||
#include <tbb/global_control.h>
|
||||
#include <unistd.h>
|
||||
#include <unordered_set>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static const char helpmsg[] = R"(
|
||||
Options:
|
||||
--help Report usage information
|
||||
-v, --version Report version information
|
||||
-V Report version and target information
|
||||
-(, --start-group Ignored
|
||||
-), --end-group Ignored
|
||||
-C DIR, --directory DIR Change to DIR before doing anything
|
||||
-E, --export-dynamic Put symbols in the dynamic symbol table
|
||||
--no-export-dynamic
|
||||
-F LIBNAME, --filter LIBNAME
|
||||
Set DT_FILTER to the specified value
|
||||
-I FILE, --dynamic-linker FILE
|
||||
Set dynamic linker path
|
||||
--no-dynamic-linker
|
||||
-L DIR, --library-path DIR Add DIR to library search path
|
||||
-M, --print-map Write map file to stdout
|
||||
-N, --omagic Do not page align data, do not make text readonly
|
||||
--no-omagic
|
||||
-O NUMBER Ignored
|
||||
-S, --strip-debug Strip .debug_* sections
|
||||
-T FILE, --script FILE Read linker script
|
||||
-X, --discard-locals Discard temporary local symbols
|
||||
-e SYMBOL, --entry SYMBOL Set program entry point
|
||||
-f SHLIB, --auxiliary SHLIB Set DT_AUXILIARY to the specified value
|
||||
-h LIBNAME, --soname LIBNAME
|
||||
Set shared library name
|
||||
-l LIBNAME Search for a given library
|
||||
-m EMULATION Ignored
|
||||
-o FILE, --output FILE Set output filename
|
||||
-r, --relocatable Generate relocatable output
|
||||
-s, --strip-all Strip .symtab section
|
||||
-u SYMBOL, --undefined SYMBOL
|
||||
Force to resolve SYMBOL
|
||||
--Bdynamic Link against shared libraries (default)
|
||||
--Bstatic Do not link against shared libraries
|
||||
--Bsymbolic Bind global symbols locally
|
||||
--Bsymbolic-functions Bind global functions locally
|
||||
--Map FILE Write map file to a given file
|
||||
--allow-multiple-definition Allow multiple definitions
|
||||
--as-needed Only set DT_NEEDED if used
|
||||
--no-as-needed
|
||||
--build-id [none,md5,sha1,sha256,uuid,HEXSTRING]
|
||||
Generate build ID
|
||||
--no-build-id
|
||||
--chroot DIR Set a given path to root directory
|
||||
--color-diagnostics Ignored
|
||||
--compress-debug-sections [none,zlib,zlib-gabi,zlib-gnu]
|
||||
Compress .debug_* sections
|
||||
--demangle Demangle C++ symbols in log messages (default)
|
||||
--no-demangle
|
||||
--disable-new-dtags Ignored
|
||||
--dynamic-list Read a list of dynamic symbols
|
||||
--eh-frame-hdr Create .eh_frame_hdr section
|
||||
--no-eh-frame-hdr
|
||||
--enable-new-dtags Ignored
|
||||
--exclude-libs LIB,LIB,.. Mark all symbols in given libraries hidden
|
||||
--fatal-warnings Ignored
|
||||
--no-fatal-warnings Ignored
|
||||
--fini SYMBOL Call SYMBOL at unload-time
|
||||
--fork Spawn a child process (default)
|
||||
--no-fork
|
||||
--gc-sections Remove unreferenced sections
|
||||
--no-gc-sections
|
||||
--gdb-index Ignored
|
||||
--hash-style [sysv,gnu,both]
|
||||
Set hash style
|
||||
--icf Fold identical code
|
||||
--no-icf
|
||||
--image-base ADDR Set the base address to a given value
|
||||
--init SYMBOL Call SYMBOl at load-time
|
||||
--no-undefined Report undefined symbols (even with --shared)
|
||||
--perf Print performance statistics
|
||||
--pie, --pic-executable Create a position independent executable
|
||||
--no-pie, --no-pic-executable
|
||||
--plugin Ignored
|
||||
--plugin-opt Ignored
|
||||
--pop-state Pop state of flags governing input file handling
|
||||
--preload
|
||||
--print-gc-sections Print removed unreferenced sections
|
||||
--no-print-gc-sections
|
||||
--print-icf-sections Print folded identical sections
|
||||
--no-print-icf-sections
|
||||
--push-state Pop state of flags governing input file handling
|
||||
--quick-exit Use quick_exit to exit (default)
|
||||
--no-quick-exit
|
||||
--relax Optimize instructions (default)
|
||||
--no-relax
|
||||
--repro Embed input files to .repro section
|
||||
--retain-symbols-file FILE Keep only symbols listed in FILE
|
||||
--rpath DIR Add DIR to runtime search path
|
||||
--rpath-link DIR Ignored
|
||||
--run COMMAND ARG... Run COMMAND with mold as /usr/bin/ld
|
||||
--shared, --Bshareable Create a share library
|
||||
--sort-common Ignored
|
||||
--sort-section Ignored
|
||||
--spare-dynamic-tags NUMBER Reserve give number of tags in .dynamic section
|
||||
--static Do not link against shared libraries
|
||||
--stats Print input statistics
|
||||
--sysroot DIR Set target system root directory
|
||||
--thread-count COUNT Use COUNT number of threads
|
||||
--threads Use multiple threads (default)
|
||||
--no-threads
|
||||
--trace Print name of each input file
|
||||
--unique PATTERN Don't merge input sections that match a given pattern
|
||||
--unresolved-symbols [report-all,ignore-all,ignore-in-object-files,ignore-in-shared-libs]
|
||||
How to handle unresolved symbols
|
||||
--version-script FILE Read version script
|
||||
--warn-common Warn about common symbols
|
||||
--no-warn-common
|
||||
--warn-unresolved-symbols Report unresolved symbols as warnings
|
||||
--error-unresolved-symbols
|
||||
Report unresolved symbols as errors (default)
|
||||
--whole-archive Include all objects from static archives
|
||||
--no-whole-archive
|
||||
--wrap SYMBOL Use wrapper function for a given symbol
|
||||
-z defs Report undefined symbols (even with --shared)
|
||||
-z nodefs
|
||||
-z execstack Require executable stack
|
||||
-z noexecstack
|
||||
-z initfirst Mark DSO to be initialized first at runtime
|
||||
-z interpose Mark object to interpose all DSOs but executable
|
||||
-z keep-text-section-prefix Keep .text.{hot,unknown,unlikely,startup,exit} as separate sections in the final binary
|
||||
-z nokeep-text-section-prefix
|
||||
-z lazy Enable lazy function resolution (default)
|
||||
-z nocopyreloc Do not create copy relocations
|
||||
-z nodelete Mark DSO non-deletable at runtime
|
||||
-z nodlopen Mark DSO not available to dlopen
|
||||
-z nodump Mark DSO not available to dldump
|
||||
-z now Disable lazy function resolution
|
||||
-z origin Mark object requiring immediate $ORIGIN processing at runtime
|
||||
-z relro Make some sections read-only after relocation (default)
|
||||
-z norelro
|
||||
-z text Report error if DT_TEXTREL is set
|
||||
-z notext
|
||||
-z textoff
|
||||
|
||||
mold: supported targets: elf32-i386 elf64-x86-64
|
||||
mold: supported emulations: elf_i386 elf_x86_64)";
|
||||
|
||||
template <typename E>
|
||||
static std::vector<std::string_view>
|
||||
read_response_file(Context<E> &ctx, std::string_view path) {
|
||||
std::vector<std::string_view> vec;
|
||||
MemoryMappedFile<E> *mb =
|
||||
MemoryMappedFile<E>::must_open(ctx, std::string(path));
|
||||
u8 *data = mb->data(ctx);
|
||||
|
||||
auto read_quoted = [&](i64 i, char quote) {
|
||||
std::string buf;
|
||||
while (i < mb->size() && data[i] != quote) {
|
||||
if (data[i] == '\\') {
|
||||
buf.append(1, data[i + 1]);
|
||||
i += 2;
|
||||
} else {
|
||||
buf.append(1, data[i++]);
|
||||
}
|
||||
}
|
||||
if (i >= mb->size())
|
||||
Fatal(ctx) << path << ": premature end of input";
|
||||
vec.push_back(save_string(ctx, buf));
|
||||
return i + 1;
|
||||
};
|
||||
|
||||
auto read_unquoted = [&](i64 i) {
|
||||
std::string buf;
|
||||
while (i < mb->size() && !isspace(data[i]))
|
||||
buf.append(1, data[i++]);
|
||||
vec.push_back(save_string(ctx, buf));
|
||||
return i;
|
||||
};
|
||||
|
||||
for (i64 i = 0; i < mb->size();) {
|
||||
if (isspace(data[i]))
|
||||
i++;
|
||||
else if (data[i] == '\'')
|
||||
i = read_quoted(i + 1, '\'');
|
||||
else if (data[i] == '\"')
|
||||
i = read_quoted(i + 1, '\"');
|
||||
else
|
||||
i = read_unquoted(i);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<std::string_view>
|
||||
expand_response_files(Context<E> &ctx, char **argv) {
|
||||
std::vector<std::string_view> vec;
|
||||
|
||||
for (i64 i = 0; argv[i]; i++) {
|
||||
if (argv[i][0] == '@')
|
||||
append(vec, read_response_file(ctx, argv[i] + 1));
|
||||
else
|
||||
vec.push_back(argv[i]);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
static std::vector<std::string> add_dashes(std::string name) {
|
||||
// Multi-letter linker options can be preceded by either a single
|
||||
// dash or double dashes except ones starting with "o", which must
|
||||
// be preceded by double dashes. For example, "-omagic" is
|
||||
// interpreted as "-o magic". If you really want to specify the
|
||||
// "omagic" option, you have to pass "--omagic".
|
||||
if (name[0] == 'o')
|
||||
return {"--" + name};
|
||||
return {"-" + name, "--" + name};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
bool read_arg(Context<E> &ctx, std::span<std::string_view> &args,
|
||||
std::string_view &arg, std::string name) {
|
||||
if (name.size() == 1) {
|
||||
if (args[0] == "-" + name) {
|
||||
if (args.size() == 1)
|
||||
Fatal(ctx) << "option -" << name << ": argument missing";
|
||||
arg = args[1];
|
||||
args = args.subspan(2);
|
||||
return true;
|
||||
}
|
||||
|
||||
if (args[0].starts_with("-" + name)) {
|
||||
arg = args[0].substr(name.size() + 1);
|
||||
args = args.subspan(1);
|
||||
return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
for (std::string opt : add_dashes(name)) {
|
||||
if (args[0] == opt) {
|
||||
if (args.size() == 1)
|
||||
Fatal(ctx) << "option -" << name << ": argument missing";
|
||||
arg = args[1];
|
||||
args = args.subspan(2);
|
||||
return true;
|
||||
}
|
||||
|
||||
if (args[0].starts_with(opt + "=")) {
|
||||
arg = args[0].substr(opt.size() + 1);
|
||||
args = args.subspan(1);
|
||||
return true;
|
||||
}
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
bool read_flag(std::span<std::string_view> &args, std::string name) {
|
||||
for (std::string opt : add_dashes(name)) {
|
||||
if (args[0] == opt) {
|
||||
args = args.subspan(1);
|
||||
return true;
|
||||
}
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
static bool read_z_flag(std::span<std::string_view> &args, std::string name) {
|
||||
if (args.size() >= 2 && args[0] == "-z" && args[1] == name) {
|
||||
args = args.subspan(2);
|
||||
return true;
|
||||
}
|
||||
|
||||
if (!args.empty() && args[0] == "-z" + name) {
|
||||
args = args.subspan(1);
|
||||
return true;
|
||||
}
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::string create_response_file(Context<E> &ctx) {
|
||||
std::string buf;
|
||||
std::stringstream out;
|
||||
|
||||
std::string cwd = get_current_dir();
|
||||
out << "-C " << cwd.substr(1) << "\n";
|
||||
|
||||
if (cwd != "/") {
|
||||
out << "--chroot ..";
|
||||
i64 depth = std::count(cwd.begin(), cwd.end(), '/');
|
||||
for (i64 i = 1; i < depth; i++)
|
||||
out << "/..";
|
||||
out << "\n";
|
||||
}
|
||||
|
||||
for (i64 i = 1; i < ctx.cmdline_args.size(); i++) {
|
||||
std::string_view arg = ctx.cmdline_args[i];
|
||||
|
||||
if (arg == "-repro" || arg == "--repro") {
|
||||
i++;
|
||||
continue;
|
||||
}
|
||||
|
||||
out << arg << "\n";
|
||||
}
|
||||
|
||||
return out.str();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 parse_hex(Context<E> &ctx, std::string opt, std::string_view value) {
|
||||
if (!value.starts_with("0x") && !value.starts_with("0X"))
|
||||
Fatal(ctx) << "option -" << opt << ": not a hexadecimal number";
|
||||
value = value.substr(2);
|
||||
if (value.find_first_not_of("0123456789abcdefABCDEF") != value.npos)
|
||||
Fatal(ctx) << "option -" << opt << ": not a hexadecimal number";
|
||||
return std::stol(std::string(value), nullptr, 16);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 parse_number(Context<E> &ctx, std::string opt,
|
||||
std::string_view value) {
|
||||
size_t nread;
|
||||
i64 ret = std::stol(std::string(value), &nread, 0);
|
||||
if (value.size() != nread)
|
||||
Fatal(ctx) << "option -" << opt << ": not a number: " << value;
|
||||
return ret;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::vector<u8> parse_hex_build_id(Context<E> &ctx,
|
||||
std::string_view arg) {
|
||||
ASSERT(arg.starts_with("0x") || arg.starts_with("0X"));
|
||||
|
||||
if (arg.size() % 2)
|
||||
Fatal(ctx) << "invalid build-id: " << arg;
|
||||
if (arg.substr(2).find_first_not_of("0123456789abcdefABCDEF") != arg.npos)
|
||||
Fatal(ctx) << "invalid build-id: " << arg;
|
||||
|
||||
arg = arg.substr(2);
|
||||
|
||||
auto fn = [](char c) {
|
||||
if ('0' <= c && c <= '9')
|
||||
return c - '0';
|
||||
if ('a' <= c && c <= 'f')
|
||||
return c - 'a' + 10;
|
||||
ASSERT('A' <= c && c <= 'F');
|
||||
return c - 'A' + 10;
|
||||
};
|
||||
|
||||
std::vector<u8> vec(arg.size() / 2);
|
||||
for (i64 i = 0; i < vec.size(); i++)
|
||||
vec[i] = (fn(arg[i * 2]) << 4) | fn(arg[i * 2 + 1]);
|
||||
return vec;
|
||||
}
|
||||
|
||||
static std::vector<std::string_view>
|
||||
split_by_comma_or_colon(std::string_view str) {
|
||||
std::vector<std::string_view> vec;
|
||||
|
||||
for (;;) {
|
||||
i64 pos = str.find_first_of(",:");
|
||||
if (pos == str.npos) {
|
||||
vec.push_back(str);
|
||||
break;
|
||||
}
|
||||
vec.push_back(str.substr(0, pos));
|
||||
str = str.substr(pos);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
static i64 get_default_thread_count() {
|
||||
// mold doesn't scale above 32 threads.
|
||||
int n = tbb::global_control::active_value(
|
||||
tbb::global_control::max_allowed_parallelism);
|
||||
return std::min(n, 32);
|
||||
}
|
||||
|
||||
static std::string_view trim(std::string_view str) {
|
||||
size_t pos = str.find_first_not_of(" \t");
|
||||
if (pos == str.npos)
|
||||
return "";
|
||||
str = str.substr(pos);
|
||||
|
||||
pos = str.find_last_not_of(" \t");
|
||||
if (pos == str.npos)
|
||||
return str;
|
||||
return str.substr(0, pos + 1);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void read_retain_symbols_file(Context<E> &ctx, std::string_view path) {
|
||||
MemoryMappedFile<E> *mb =
|
||||
MemoryMappedFile<E>::must_open(ctx, std::string(path));
|
||||
std::string_view data((char *)mb->data(ctx), mb->size());
|
||||
|
||||
ctx.arg.retain_symbols_file.reset(new std::unordered_set<std::string_view>);
|
||||
|
||||
while (!data.empty()) {
|
||||
size_t pos = data.find('\n');
|
||||
std::string_view name;
|
||||
|
||||
if (pos == data.npos) {
|
||||
name = data;
|
||||
data = "";
|
||||
} else {
|
||||
name = data.substr(0, pos);
|
||||
data = data.substr(pos + 1);
|
||||
}
|
||||
|
||||
name = trim(name);
|
||||
if (!name.empty())
|
||||
ctx.arg.retain_symbols_file->insert(name);
|
||||
}
|
||||
}
|
||||
|
||||
static bool is_file(std::string_view path) {
|
||||
struct stat st;
|
||||
return stat(std::string(path).c_str(), &st) == 0 &&
|
||||
(st.st_mode & S_IFMT) != S_IFDIR;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void parse_nonpositional_args(Context<E> &ctx,
|
||||
std::vector<std::string_view> &remaining) {
|
||||
std::span<std::string_view> args = ctx.cmdline_args;
|
||||
args = args.subspan(1);
|
||||
|
||||
ctx.arg.thread_count = get_default_thread_count();
|
||||
bool version_shown = false;
|
||||
|
||||
while (!args.empty()) {
|
||||
std::string_view arg;
|
||||
|
||||
if (read_flag(args, "help")) {
|
||||
SyncOut(ctx) << "Usage: " << ctx.cmdline_args[0]
|
||||
<< " [options] file...\n" << helpmsg;
|
||||
exit(0);
|
||||
}
|
||||
|
||||
if (read_arg(ctx, args, arg, "o") || read_arg(ctx, args, arg, "output")) {
|
||||
ctx.arg.output = arg;
|
||||
} else if (read_arg(ctx, args, arg, "dynamic-linker") ||
|
||||
read_arg(ctx, args, arg, "I")) {
|
||||
ctx.arg.dynamic_linker = arg;
|
||||
} else if (read_arg(ctx, args, arg, "no-dynamic-linker")) {
|
||||
ctx.arg.dynamic_linker = "";
|
||||
} else if (read_flag(args, "v")) {
|
||||
SyncOut(ctx) << get_version_string();
|
||||
version_shown = true;
|
||||
} else if (read_flag(args, "version")) {
|
||||
SyncOut(ctx) << get_version_string();
|
||||
exit(0);
|
||||
} else if (read_flag(args, "V")) {
|
||||
SyncOut(ctx) << get_version_string()
|
||||
<< "\n Supported emulations:\n elf_x86_64\n elf_i386";
|
||||
version_shown = true;
|
||||
} else if (read_arg(ctx, args, arg, "m")) {
|
||||
if (arg == "elf_x86_64") {
|
||||
ctx.arg.emulation = EM_X86_64;
|
||||
} else if (arg == "elf_i386") {
|
||||
ctx.arg.emulation = EM_386;
|
||||
} else if (arg == "aarch64linux") {
|
||||
ctx.arg.emulation = EM_AARCH64;
|
||||
} else {
|
||||
Fatal(ctx) << "unknown -m argument: " << arg;
|
||||
}
|
||||
} else if (read_flag(args, "export-dynamic") || read_flag(args, "E")) {
|
||||
ctx.arg.export_dynamic = true;
|
||||
} else if (read_flag(args, "no-export-dynamic")) {
|
||||
ctx.arg.export_dynamic = false;
|
||||
} else if (read_flag(args, "Bsymbolic")) {
|
||||
ctx.arg.Bsymbolic = true;
|
||||
} else if (read_flag(args, "Bsymbolic-functions")) {
|
||||
ctx.arg.Bsymbolic_functions = true;
|
||||
} else if (read_arg(ctx, args, arg, "exclude-libs")) {
|
||||
append(ctx.arg.exclude_libs, split_by_comma_or_colon(arg));
|
||||
} else if (read_arg(ctx, args, arg, "e") ||
|
||||
read_arg(ctx, args, arg, "entry")) {
|
||||
ctx.arg.entry = arg;
|
||||
} else if (read_arg(ctx, args, arg, "Map")) {
|
||||
ctx.arg.Map = arg;
|
||||
ctx.arg.print_map = true;
|
||||
} else if (read_flag(args, "print-map") || read_flag(args, "M")) {
|
||||
ctx.arg.print_map = true;
|
||||
} else if (read_flag(args, "static") || read_flag(args, "Bstatic")) {
|
||||
ctx.arg.is_static = true;
|
||||
remaining.push_back("-Bstatic");
|
||||
} else if (read_flag(args, "Bdynamic")) {
|
||||
ctx.arg.is_static = false;
|
||||
remaining.push_back("-Bdynamic");
|
||||
} else if (read_flag(args, "shared") || read_flag(args, "Bshareable")) {
|
||||
ctx.arg.shared = true;
|
||||
} else if (read_arg(ctx, args, arg, "spare-dynamic-tags")) {
|
||||
ctx.arg.spare_dynamic_tags = parse_number(ctx, "spare-dynamic-tags", arg);
|
||||
} else if (read_flag(args, "demangle")) {
|
||||
ctx.arg.demangle = true;
|
||||
} else if (read_flag(args, "no-demangle")) {
|
||||
ctx.arg.demangle = false;
|
||||
} else if (read_arg(ctx, args, arg, "y") ||
|
||||
read_arg(ctx, args, arg, "trace-symbol")) {
|
||||
ctx.arg.trace_symbol.push_back(arg);
|
||||
} else if (read_arg(ctx, args, arg, "filler")) {
|
||||
ctx.arg.filler = parse_hex(ctx, "filler", arg);
|
||||
} else if (read_arg(ctx, args, arg, "L") ||
|
||||
read_arg(ctx, args, arg, "library-path")) {
|
||||
ctx.arg.library_paths.push_back(std::string(arg));
|
||||
} else if (read_arg(ctx, args, arg, "sysroot")) {
|
||||
ctx.arg.sysroot = arg;
|
||||
} else if (read_arg(ctx, args, arg, "unique")) {
|
||||
ctx.arg.unique.reset(new std::regex(glob_to_regex(arg)));
|
||||
} else if (read_arg(ctx, args, arg, "unresolved-symbols")) {
|
||||
if (arg == "report-all" || arg == "ignore-in-shared-libs")
|
||||
ctx.arg.unresolved_symbols = UnresolvedKind::ERROR;
|
||||
else if (arg == "ignore-all" || arg == "ignore-in-object-files")
|
||||
ctx.arg.unresolved_symbols = UnresolvedKind::IGNORE;
|
||||
else
|
||||
Fatal(ctx) << "unknown --unresolved-symbols argument: " << arg;
|
||||
} else if (read_arg(ctx, args, arg, "u") ||
|
||||
read_arg(ctx, args, arg, "undefined")) {
|
||||
ctx.arg.undefined.push_back(arg);
|
||||
} else if (read_arg(ctx, args, arg, "init")) {
|
||||
ctx.arg.init = arg;
|
||||
} else if (read_arg(ctx, args, arg, "fini")) {
|
||||
ctx.arg.fini = arg;
|
||||
} else if (read_arg(ctx, args, arg, "hash-style")) {
|
||||
if (arg == "sysv") {
|
||||
ctx.arg.hash_style_sysv = true;
|
||||
ctx.arg.hash_style_gnu = false;
|
||||
} else if (arg == "gnu") {
|
||||
ctx.arg.hash_style_sysv = false;
|
||||
ctx.arg.hash_style_gnu = true;
|
||||
} else if (arg == "both") {
|
||||
ctx.arg.hash_style_sysv = true;
|
||||
ctx.arg.hash_style_gnu = true;
|
||||
} else {
|
||||
Fatal(ctx) << "invalid --hash-style argument: " << arg;
|
||||
}
|
||||
} else if (read_arg(ctx, args, arg, "soname") ||
|
||||
read_arg(ctx, args, arg, "h")) {
|
||||
ctx.arg.soname = arg;
|
||||
} else if (read_flag(args, "allow-multiple-definition")) {
|
||||
ctx.arg.allow_multiple_definition = true;
|
||||
} else if (read_flag(args, "trace")) {
|
||||
ctx.arg.trace = true;
|
||||
} else if (read_flag(args, "eh-frame-hdr")) {
|
||||
ctx.arg.eh_frame_hdr = true;
|
||||
} else if (read_flag(args, "no-eh-frame-hdr")) {
|
||||
ctx.arg.eh_frame_hdr = false;
|
||||
} else if (read_flag(args, "pie") || read_flag(args, "pic-executable")) {
|
||||
ctx.arg.pic = true;
|
||||
ctx.arg.pie = true;
|
||||
} else if (read_flag(args, "no-pie") ||
|
||||
read_flag(args, "no-pic-executable")) {
|
||||
ctx.arg.pic = false;
|
||||
ctx.arg.pie = false;
|
||||
} else if (read_flag(args, "relax")) {
|
||||
ctx.arg.relax = true;
|
||||
} else if (read_flag(args, "no-relax")) {
|
||||
ctx.arg.relax = false;
|
||||
} else if (read_flag(args, "r") || read_flag(args, "relocatable")) {
|
||||
ctx.arg.relocatable = true;
|
||||
} else if (read_flag(args, "perf")) {
|
||||
ctx.arg.perf = true;
|
||||
} else if (read_flag(args, "stats")) {
|
||||
ctx.arg.stats = true;
|
||||
Counter::enabled = true;
|
||||
} else if (read_arg(ctx, args, arg, "C") ||
|
||||
read_arg(ctx, args, arg, "directory")) {
|
||||
ctx.arg.directory = arg;
|
||||
} else if (read_arg(ctx, args, arg, "chroot")) {
|
||||
ctx.arg.chroot = arg;
|
||||
} else if (read_flag(args, "warn-common")) {
|
||||
ctx.arg.warn_common = true;
|
||||
} else if (read_flag(args, "no-warn-common")) {
|
||||
ctx.arg.warn_common = false;
|
||||
} else if (read_arg(ctx, args, arg, "compress-debug-sections")) {
|
||||
if (arg == "zlib" || arg == "zlib-gabi")
|
||||
ctx.arg.compress_debug_sections = COMPRESS_GABI;
|
||||
else if (arg == "zlib-gnu")
|
||||
ctx.arg.compress_debug_sections = COMPRESS_GNU;
|
||||
else if (arg == "none")
|
||||
ctx.arg.compress_debug_sections = COMPRESS_NONE;
|
||||
else
|
||||
Fatal(ctx) << "invalid --compress-debug-sections argument: " << arg;
|
||||
} else if (read_arg(ctx, args, arg, "wrap")) {
|
||||
ctx.arg.wrap.insert(arg);
|
||||
} else if (read_flag(args, "omagic") || read_flag(args, "N")) {
|
||||
ctx.arg.omagic = true;
|
||||
ctx.arg.is_static = true;
|
||||
} else if (read_flag(args, "no-omagic")) {
|
||||
ctx.arg.omagic = false;
|
||||
} else if (read_arg(ctx, args, arg, "retain-symbols-file")) {
|
||||
read_retain_symbols_file(ctx, arg);
|
||||
} else if (read_flag(args, "repro")) {
|
||||
ctx.arg.repro = true;
|
||||
} else if (read_z_flag(args, "now")) {
|
||||
ctx.arg.z_now = true;
|
||||
} else if (read_z_flag(args, "lazy")) {
|
||||
ctx.arg.z_now = false;
|
||||
} else if (read_z_flag(args, "execstack")) {
|
||||
ctx.arg.z_execstack = true;
|
||||
} else if (read_z_flag(args, "noexecstack")) {
|
||||
ctx.arg.z_execstack = false;
|
||||
} else if (read_z_flag(args, "relro")) {
|
||||
ctx.arg.z_relro = true;
|
||||
} else if (read_z_flag(args, "norelro")) {
|
||||
ctx.arg.z_relro = false;
|
||||
} else if (read_z_flag(args, "defs")) {
|
||||
ctx.arg.z_defs = true;
|
||||
} else if (read_z_flag(args, "nodefs")) {
|
||||
ctx.arg.z_defs = false;
|
||||
} else if (read_z_flag(args, "nodlopen")) {
|
||||
ctx.arg.z_dlopen = false;
|
||||
} else if (read_z_flag(args, "nodelete")) {
|
||||
ctx.arg.z_delete = false;
|
||||
} else if (read_z_flag(args, "nocopyreloc")) {
|
||||
ctx.arg.z_copyreloc = false;
|
||||
} else if (read_z_flag(args, "nodump")) {
|
||||
ctx.arg.z_dump = false;
|
||||
} else if (read_z_flag(args, "initfirst")) {
|
||||
ctx.arg.z_initfirst = true;
|
||||
} else if (read_z_flag(args, "interpose")) {
|
||||
ctx.arg.z_interpose = true;
|
||||
} else if (read_z_flag(args, "muldefs")) {
|
||||
ctx.arg.allow_multiple_definition = true;
|
||||
} else if (read_z_flag(args, "keep-text-section-prefix")) {
|
||||
ctx.arg.z_keep_text_section_prefix = true;
|
||||
} else if (read_z_flag(args, "nokeep-text-section-prefix")) {
|
||||
ctx.arg.z_keep_text_section_prefix = false;
|
||||
} else if (read_z_flag(args, "text")) {
|
||||
ctx.arg.z_text = true;
|
||||
} else if (read_z_flag(args, "notext") || read_z_flag(args, "textoff")) {
|
||||
ctx.arg.z_text = false;
|
||||
} else if (read_z_flag(args, "origin")) {
|
||||
ctx.arg.z_origin = true;
|
||||
} else if (read_flag(args, "no-undefined")) {
|
||||
ctx.arg.z_defs = true;
|
||||
} else if (read_flag(args, "fatal-warnings")) {
|
||||
ctx.arg.fatal_warnings = true;
|
||||
} else if (read_flag(args, "no-fatal-warnings")) {
|
||||
ctx.arg.fatal_warnings = false;
|
||||
} else if (read_flag(args, "fork")) {
|
||||
ctx.arg.fork = true;
|
||||
} else if (read_flag(args, "no-fork")) {
|
||||
ctx.arg.fork = false;
|
||||
} else if (read_flag(args, "gc-sections")) {
|
||||
ctx.arg.gc_sections = true;
|
||||
} else if (read_flag(args, "no-gc-sections")) {
|
||||
ctx.arg.gc_sections = false;
|
||||
} else if (read_flag(args, "print-gc-sections")) {
|
||||
ctx.arg.print_gc_sections = true;
|
||||
} else if (read_flag(args, "no-print-gc-sections")) {
|
||||
ctx.arg.print_gc_sections = false;
|
||||
} else if (read_arg(ctx, args, arg, "icf")) {
|
||||
if (arg == "all")
|
||||
ctx.arg.icf = true;
|
||||
else if (arg == "none")
|
||||
ctx.arg.icf = false;
|
||||
else
|
||||
Fatal(ctx) << "unknown --icf argument: " << arg;
|
||||
} else if (read_flag(args, "no-icf")) {
|
||||
ctx.arg.icf = false;
|
||||
} else if (read_arg(ctx, args, arg, "image-base")) {
|
||||
ctx.arg.image_base = parse_number(ctx, "image-base", arg);
|
||||
} else if (read_flag(args, "quick-exit")) {
|
||||
ctx.arg.quick_exit = true;
|
||||
} else if (read_flag(args, "no-quick-exit")) {
|
||||
ctx.arg.quick_exit = false;
|
||||
} else if (read_flag(args, "print-icf-sections")) {
|
||||
ctx.arg.print_icf_sections = true;
|
||||
} else if (read_flag(args, "no-print-icf-sections")) {
|
||||
ctx.arg.print_icf_sections = false;
|
||||
} else if (read_flag(args, "quick-exit")) {
|
||||
ctx.arg.quick_exit = true;
|
||||
} else if (read_flag(args, "no-quick-exit")) {
|
||||
ctx.arg.quick_exit = false;
|
||||
} else if (read_arg(ctx, args, arg, "thread-count")) {
|
||||
ctx.arg.thread_count = parse_number(ctx, "thread-count", arg);
|
||||
} else if (read_flag(args, "threads")) {
|
||||
ctx.arg.thread_count = get_default_thread_count();
|
||||
} else if (read_flag(args, "no-threads")) {
|
||||
ctx.arg.thread_count = 1;
|
||||
} else if (read_flag(args, "discard-all") || read_flag(args, "x")) {
|
||||
ctx.arg.discard_all = true;
|
||||
} else if (read_flag(args, "discard-locals") || read_flag(args, "X")) {
|
||||
ctx.arg.discard_locals = true;
|
||||
} else if (read_flag(args, "strip-all") || read_flag(args, "s")) {
|
||||
ctx.arg.strip_all = true;
|
||||
} else if (read_flag(args, "strip-debug") || read_flag(args, "S")) {
|
||||
ctx.arg.strip_all = true;
|
||||
} else if (read_flag(args, "warn-unresolved-symbols")) {
|
||||
ctx.arg.unresolved_symbols = UnresolvedKind::WARN;
|
||||
} else if (read_flag(args, "error-unresolved-symbols")) {
|
||||
ctx.arg.unresolved_symbols = UnresolvedKind::ERROR;
|
||||
} else if (read_arg(ctx, args, arg, "rpath")) {
|
||||
if (!ctx.arg.rpaths.empty())
|
||||
ctx.arg.rpaths += ":";
|
||||
ctx.arg.rpaths += arg;
|
||||
} else if (read_arg(ctx, args, arg, "R")) {
|
||||
if (is_file(arg))
|
||||
Fatal(ctx) << "-R" << arg
|
||||
<< ": -R as an alias for --just-symbols is not supported";
|
||||
|
||||
if (!ctx.arg.rpaths.empty())
|
||||
ctx.arg.rpaths += ":";
|
||||
ctx.arg.rpaths += arg;
|
||||
} else if (read_flag(args, "build-id")) {
|
||||
ctx.arg.build_id.kind = BuildId::HASH;
|
||||
ctx.arg.build_id.hash_size = 20;
|
||||
} else if (read_arg(ctx, args, arg, "build-id")) {
|
||||
if (arg == "none") {
|
||||
ctx.arg.build_id.kind = BuildId::NONE;
|
||||
} else if (arg == "uuid") {
|
||||
ctx.arg.build_id.kind = BuildId::UUID;
|
||||
} else if (arg == "md5") {
|
||||
ctx.arg.build_id.kind = BuildId::HASH;
|
||||
ctx.arg.build_id.hash_size = 16;
|
||||
} else if (arg == "sha1") {
|
||||
ctx.arg.build_id.kind = BuildId::HASH;
|
||||
ctx.arg.build_id.hash_size = 20;
|
||||
} else if (arg == "sha256") {
|
||||
ctx.arg.build_id.kind = BuildId::HASH;
|
||||
ctx.arg.build_id.hash_size = 32;
|
||||
} else if (arg.starts_with("0x") || arg.starts_with("0X")) {
|
||||
ctx.arg.build_id.kind = BuildId::HEX;
|
||||
ctx.arg.build_id.value = parse_hex_build_id(ctx, arg);
|
||||
} else {
|
||||
Fatal(ctx) << "invalid --build-id argument: " << arg;
|
||||
}
|
||||
} else if (read_flag(args, "no-build-id")) {
|
||||
ctx.arg.build_id.kind = BuildId::NONE;
|
||||
} else if (read_arg(ctx, args, arg, "auxiliary") ||
|
||||
read_arg(ctx, args, arg, "f")) {
|
||||
ctx.arg.auxiliary.push_back(arg);
|
||||
} else if (read_arg(ctx, args, arg, "filter") ||
|
||||
read_arg(ctx, args, arg, "F")) {
|
||||
ctx.arg.filter.push_back(arg);
|
||||
} else if (read_flag(args, "preload")) {
|
||||
ctx.arg.preload = true;
|
||||
} else if (read_arg(ctx, args, arg, "O")) {
|
||||
} else if (read_flag(args, "O0")) {
|
||||
} else if (read_flag(args, "O1")) {
|
||||
} else if (read_flag(args, "O2")) {
|
||||
} else if (read_flag(args, "verbose")) {
|
||||
} else if (read_arg(ctx, args, arg, "plugin")) {
|
||||
} else if (read_arg(ctx, args, arg, "plugin-opt")) {
|
||||
} else if (read_flag(args, "color-diagnostics")) {
|
||||
} else if (read_flag(args, "gdb-index")) {
|
||||
} else if (read_flag(args, "eh-frame-hdr")) {
|
||||
} else if (read_flag(args, "start-group")) {
|
||||
} else if (read_flag(args, "end-group")) {
|
||||
} else if (read_flag(args, "(")) {
|
||||
} else if (read_flag(args, ")")) {
|
||||
} else if (read_flag(args, "fatal-warnings")) {
|
||||
} else if (read_flag(args, "enable-new-dtags")) {
|
||||
} else if (read_flag(args, "disable-new-dtags")) {
|
||||
} else if (read_flag(args, "nostdlib")) {
|
||||
} else if (read_flag(args, "allow-shlib-undefined")) {
|
||||
} else if (read_flag(args, "no-allow-shlib-undefined")) {
|
||||
} else if (read_flag(args, "no-add-needed")) {
|
||||
} else if (read_flag(args, "no-copy-dt-needed-entries")) {
|
||||
} else if (read_flag(args, "no-undefined-version")) {
|
||||
} else if (read_arg(ctx, args, arg, "sort-section")) {
|
||||
} else if (read_flag(args, "sort-common")) {
|
||||
} else if (read_flag(args, "fix-cortex-a53-843419")) {
|
||||
} else if (read_flag(args, "EL")) {
|
||||
} else if (read_flag(args, "warn-once")) {
|
||||
} else if (read_flag(args, "nodefaultlibs")) {
|
||||
} else if (read_flag(args, "warn-constructors")) {
|
||||
} else if (read_arg(ctx, args, arg, "rpath-link")) {
|
||||
} else if (read_z_flag(args, "combreloc")) {
|
||||
} else if (read_z_flag(args, "nocombreloc")) {
|
||||
} else if (read_arg(ctx, args, arg, "version-script")) {
|
||||
remaining.push_back("--version-script");
|
||||
remaining.push_back(arg);
|
||||
} else if (read_arg(ctx, args, arg, "dynamic-list")) {
|
||||
remaining.push_back("--dynamic-list");
|
||||
remaining.push_back(arg);
|
||||
} else if (read_flag(args, "as-needed")) {
|
||||
remaining.push_back("-as-needed");
|
||||
} else if (read_flag(args, "no-as-needed")) {
|
||||
remaining.push_back("-no-as-needed");
|
||||
} else if (read_flag(args, "whole-archive")) {
|
||||
remaining.push_back("-whole-archive");
|
||||
} else if (read_flag(args, "no-whole-archive")) {
|
||||
remaining.push_back("-no-whole-archive");
|
||||
} else if (read_arg(ctx, args, arg, "l")) {
|
||||
remaining.push_back("-l");
|
||||
remaining.push_back(arg);
|
||||
} else if (read_arg(ctx, args, arg, "script") ||
|
||||
read_arg(ctx, args, arg, "T")) {
|
||||
remaining.push_back(arg);
|
||||
} else if (read_flag(args, "push-state")) {
|
||||
remaining.push_back("-push-state");
|
||||
} else if (read_flag(args, "pop-state")) {
|
||||
remaining.push_back("-pop-state");
|
||||
} else {
|
||||
if (args[0] == "-z" && args.size() >= 2)
|
||||
Fatal(ctx) << "mold: unknown command line option: -z " << args[1];
|
||||
if (args[0][0] == '-')
|
||||
Fatal(ctx) << "mold: unknown command line option: " << args[0];
|
||||
remaining.push_back(args[0]);
|
||||
args = args.subspan(1);
|
||||
}
|
||||
}
|
||||
|
||||
if (!ctx.arg.sysroot.empty()) {
|
||||
for (std::string &path : ctx.arg.library_paths) {
|
||||
if (std::string_view(path).starts_with('='))
|
||||
path = std::string(ctx.arg.sysroot) + path.substr(1);
|
||||
else if (std::string_view(path).starts_with("$SYSROOT"))
|
||||
path = std::string(ctx.arg.sysroot) + path.substr(8);
|
||||
}
|
||||
}
|
||||
|
||||
if (ctx.arg.shared) {
|
||||
ctx.arg.pic = true;
|
||||
ctx.arg.dynamic_linker = "";
|
||||
}
|
||||
|
||||
if (ctx.arg.pic)
|
||||
ctx.arg.image_base = 0;
|
||||
|
||||
if (ctx.arg.retain_symbols_file) {
|
||||
ctx.arg.strip_all = false;
|
||||
ctx.arg.discard_all = false;
|
||||
}
|
||||
|
||||
if (ctx.arg.relocatable)
|
||||
ctx.arg.is_static = true;
|
||||
|
||||
if (!ctx.arg.shared) {
|
||||
if (!ctx.arg.filter.empty())
|
||||
Fatal(ctx) << "-filter may not be used without -shared";
|
||||
if (!ctx.arg.auxiliary.empty())
|
||||
Fatal(ctx) << "-auxiliary may not be used without -shared";
|
||||
}
|
||||
|
||||
if (char *env = getenv("MOLD_REPRO"); env && env[0])
|
||||
ctx.arg.repro = true;
|
||||
|
||||
if (ctx.arg.output.empty())
|
||||
ctx.arg.output = "a.out";
|
||||
|
||||
// TLSDESC relocs must be always relaxed for statically-linked
|
||||
// executables even if -no-relax is given. It is because a
|
||||
// statically-linked executable doesn't contain a tranpoline
|
||||
// function needed for TLSDESC.
|
||||
ctx.relax_tlsdesc = ctx.arg.is_static || (ctx.arg.relax && !ctx.arg.shared);
|
||||
|
||||
if (version_shown && remaining.empty())
|
||||
exit(0);
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template \
|
||||
std::vector<std::string_view> \
|
||||
expand_response_files(Context<E> &ctx, char **argv); \
|
||||
\
|
||||
template \
|
||||
bool read_arg(Context<E> &ctx, std::span<std::string_view> &args, \
|
||||
std::string_view &arg, \
|
||||
std::string name); \
|
||||
\
|
||||
template std::string create_response_file(Context<E> &ctx); \
|
||||
\
|
||||
template \
|
||||
void parse_nonpositional_args(Context<E> &ctx, \
|
||||
std::vector<std::string_view> &remaining)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+166
@@ -0,0 +1,166 @@
|
||||
// This file implements a multi-threaded zlib compression routine.
|
||||
//
|
||||
// Multiple pieces of raw compressed data in zlib-format can be merged
|
||||
// just by concatenation as long as each zlib stream is flushed with
|
||||
// Z_SYNC_FLUSH. In this file, we split input data into multiple
|
||||
// shards, compress them individually and concatenate them. We then
|
||||
// append a header, a trailer and a checksum so that the concatenated
|
||||
// data is valid zlib-format data.
|
||||
//
|
||||
// Using threads to compress data has a downside. Since the dictionary
|
||||
// is reset on boundaries of shards, compression ratio is sacrificed
|
||||
// a little bit. However, if a shard size is large enough, that loss
|
||||
// is negligible in practice.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <zlib.h>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static constexpr i64 SHARD_SIZE = 1024 * 1024;
|
||||
|
||||
static std::vector<std::string_view> split(std::string_view input) {
|
||||
std::vector<std::string_view> shards;
|
||||
|
||||
while (input.size() >= SHARD_SIZE) {
|
||||
shards.push_back(input.substr(0, SHARD_SIZE));
|
||||
input = input.substr(SHARD_SIZE);
|
||||
}
|
||||
if (!input.empty())
|
||||
shards.push_back(input);
|
||||
return shards;
|
||||
}
|
||||
|
||||
static std::vector<u8> do_compress(std::string_view input) {
|
||||
// Initialize zlib stream. Since debug info is generally compressed
|
||||
// pretty well, we chose compression level 3.
|
||||
z_stream strm;
|
||||
strm.zalloc = Z_NULL;
|
||||
strm.zfree = Z_NULL;
|
||||
strm.opaque = Z_NULL;
|
||||
int r = deflateInit2(&strm, 3, Z_DEFLATED, -15, 8, Z_DEFAULT_STRATEGY);
|
||||
ASSERT(r == Z_OK);
|
||||
|
||||
// Set an input buffer
|
||||
strm.avail_in = input.size();
|
||||
strm.next_in = (u8 *)input.data();
|
||||
|
||||
// Set an output buffer. deflateBound() returns an upper bound
|
||||
// on the compression size. +16 for Z_SYNC_FLUSH.
|
||||
std::vector<u8> buf(deflateBound(&strm, strm.avail_in) + 16);
|
||||
|
||||
strm.avail_out = buf.size();
|
||||
strm.next_out = buf.data();
|
||||
|
||||
r = deflate(&strm, Z_SYNC_FLUSH);
|
||||
ASSERT(r == Z_OK);
|
||||
ASSERT(strm.avail_out > 0);
|
||||
|
||||
buf.resize(buf.size() - strm.avail_out);
|
||||
deflateEnd(&strm);
|
||||
return buf;
|
||||
}
|
||||
|
||||
ZlibCompressor::ZlibCompressor(std::string_view input) {
|
||||
std::vector<std::string_view> inputs = split(input);
|
||||
std::vector<u64> adlers(inputs.size());
|
||||
shards.resize(inputs.size());
|
||||
|
||||
// Compress each shard
|
||||
tbb::parallel_for((i64)0, (i64)inputs.size(), [&](i64 i) {
|
||||
adlers[i] = adler32(1, (u8 *)inputs[i].data(), inputs[i].size());
|
||||
shards[i] = do_compress(inputs[i]);
|
||||
});
|
||||
|
||||
// Combine checksums
|
||||
checksum = adlers[0];
|
||||
for (i64 i = 1; i < inputs.size(); i++)
|
||||
checksum = adler32_combine(checksum, adlers[i], inputs[i].size());
|
||||
}
|
||||
|
||||
i64 ZlibCompressor::size() const {
|
||||
i64 size = 2; // +2 for header
|
||||
for (const std::vector<u8> &shard : shards)
|
||||
size += shard.size();
|
||||
return size + 6; // +6 for trailer and checksum
|
||||
}
|
||||
|
||||
void ZlibCompressor::write_to(u8 *buf) {
|
||||
// Write a zlib-format header
|
||||
buf[0] = 0x78;
|
||||
buf[1] = 0x9c;
|
||||
|
||||
// Copy compressed data
|
||||
std::vector<i64> offsets(shards.size());
|
||||
offsets[0] = 2; // +2 for header
|
||||
for (i64 i = 1; i < shards.size(); i++)
|
||||
offsets[i] = offsets[i - 1] + shards[i - 1].size();
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)shards.size(), [&](i64 i) {
|
||||
memcpy(&buf[offsets[i]], shards[i].data(), shards[i].size());
|
||||
});
|
||||
|
||||
// Write a trailer
|
||||
u8 *end = buf + size();
|
||||
end[-6] = 3;
|
||||
end[-5] = 0;
|
||||
|
||||
// Write a checksum
|
||||
write32be(end - 4, checksum);
|
||||
}
|
||||
|
||||
GzipCompressor::GzipCompressor(std::string_view input) {
|
||||
std::vector<std::string_view> inputs = split(input);
|
||||
std::vector<u32> crc(inputs.size());
|
||||
shards.resize(inputs.size());
|
||||
|
||||
// Compress each shard
|
||||
tbb::parallel_for((i64)0, (i64)inputs.size(), [&](i64 i) {
|
||||
crc[i] = crc32(0, (u8 *)inputs[i].data(), inputs[i].size());
|
||||
shards[i] = do_compress(inputs[i]);
|
||||
});
|
||||
|
||||
// Combine checksums
|
||||
checksum = crc[0];
|
||||
for (i64 i = 1; i < inputs.size(); i++)
|
||||
checksum = crc32_combine(checksum, crc[i], inputs[i].size());
|
||||
|
||||
uncompressed_size = input.size();
|
||||
}
|
||||
|
||||
i64 GzipCompressor::size() const {
|
||||
i64 size = 10; // +10 for header
|
||||
for (const std::vector<u8> &shard : shards)
|
||||
size += shard.size();
|
||||
return size + 10; // +10 for trailer and checksum
|
||||
}
|
||||
|
||||
void GzipCompressor::write_to(u8 *buf) {
|
||||
// Write a zlib-format header
|
||||
memset(buf, 0, 10);
|
||||
buf[0] = 0x1f; // magic
|
||||
buf[1] = 0x8b; // magic
|
||||
buf[2] = 0x08; // compression method is zlib
|
||||
buf[9] = 0xff; // made on unknown OS
|
||||
|
||||
// Copy compressed data
|
||||
std::vector<i64> offsets(shards.size());
|
||||
offsets[0] = 10; // +10 for header
|
||||
for (i64 i = 1; i < shards.size(); i++)
|
||||
offsets[i] = offsets[i - 1] + shards[i - 1].size();
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)shards.size(), [&](i64 i) {
|
||||
memcpy(&buf[offsets[i]], shards[i].data(), shards[i].size());
|
||||
});
|
||||
|
||||
// Write a trailer
|
||||
u8 *end = buf + size();
|
||||
end[-10] = 0x3; // two-byte zlib stream terminator
|
||||
end[-9] = 0;
|
||||
*(u32 *)(end - 8) = checksum;
|
||||
*(u32 *)(end - 4) = uncompressed_size;
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,83 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static const char *locked = (char *)-1;
|
||||
|
||||
template <typename T>
|
||||
ConcurrentMap<T>::ConcurrentMap() {}
|
||||
|
||||
template <typename T>
|
||||
ConcurrentMap<T>::ConcurrentMap(i64 nbuckets) {
|
||||
resize(nbuckets);
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
void ConcurrentMap<T>::resize(i64 nbuckets) {
|
||||
this->~ConcurrentMap();
|
||||
|
||||
nbuckets = std::max<i64>(MIN_NBUCKETS, next_power_of_two(nbuckets));
|
||||
|
||||
this->nbuckets = nbuckets;
|
||||
keys = (std::atomic<const char *> *)calloc(nbuckets, sizeof(keys[0]));
|
||||
sizes = (u32 *)calloc(nbuckets, sizeof(sizes[0]));
|
||||
values = (T *)calloc(nbuckets, sizeof(values[0]));
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
ConcurrentMap<T>::~ConcurrentMap() {
|
||||
if (keys) {
|
||||
free((void *)keys);
|
||||
free((void *)sizes);
|
||||
free((void *)values);
|
||||
}
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
std::pair<T *, bool>
|
||||
ConcurrentMap<T>::insert(std::string_view key, u64 hash, const T &val) {
|
||||
if (!keys)
|
||||
return {nullptr, false};
|
||||
|
||||
ASSERT(__builtin_popcount(nbuckets) == 1);
|
||||
i64 idx = hash & (nbuckets - 1);
|
||||
i64 retry = 0;
|
||||
|
||||
while (retry < MAX_RETRY) {
|
||||
const char *ptr = keys[idx];
|
||||
if (ptr == locked) {
|
||||
#ifdef __x86_64__
|
||||
asm volatile("pause" ::: "memory");
|
||||
#endif
|
||||
continue;
|
||||
}
|
||||
|
||||
if (ptr == nullptr) {
|
||||
if (!keys[idx].compare_exchange_weak(ptr, locked))
|
||||
continue;
|
||||
new (values + idx) T(val);
|
||||
sizes[idx] = key.size();
|
||||
keys[idx] = key.data();
|
||||
return {values + idx, true};
|
||||
}
|
||||
|
||||
if (key.size() == sizes[idx] && memcmp(ptr, key.data(), sizes[idx]) == 0)
|
||||
return {values + idx, false};
|
||||
|
||||
u64 mask = nbuckets / NUM_SHARDS - 1;
|
||||
idx = (idx & ~mask) | ((idx + 1) & mask);
|
||||
retry++;
|
||||
}
|
||||
|
||||
ASSERT(false && "ConcurrentMap is full");
|
||||
return {nullptr, false};
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template class ConcurrentMap<SectionFragment<E>>;
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,968 @@
|
||||
#pragma once
|
||||
|
||||
#include <cstdint>
|
||||
#include <ostream>
|
||||
#include <string>
|
||||
|
||||
#ifdef __BIG_ENDIAN__
|
||||
#error "mold does not support big-endian hosts"
|
||||
#endif
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
typedef uint8_t u8;
|
||||
typedef uint16_t u16;
|
||||
typedef uint32_t u32;
|
||||
typedef uint64_t u64;
|
||||
|
||||
typedef int8_t i8;
|
||||
typedef int16_t i16;
|
||||
typedef int32_t i32;
|
||||
typedef int64_t i64;
|
||||
|
||||
struct X86_64;
|
||||
struct I386;
|
||||
struct AARCH64;
|
||||
|
||||
template <typename E> struct ElfSym;
|
||||
template <typename E> struct ElfShdr;
|
||||
template <typename E> struct ElfEhdr;
|
||||
template <typename E> struct ElfPhdr;
|
||||
template <typename E> struct ElfRel;
|
||||
template <typename E> struct ElfDyn;
|
||||
template <typename E> struct ElfVerneed;
|
||||
template <typename E> struct ElfVernaux;
|
||||
template <typename E> struct ElfVerdef;
|
||||
template <typename E> struct ElfVerdaux;
|
||||
template <typename E> struct ElfChdr;
|
||||
template <typename E> struct ElfNhdr;
|
||||
|
||||
template <typename E>
|
||||
std::string rel_to_string(u32 r_type);
|
||||
|
||||
template <typename E>
|
||||
std::ostream &operator<<(std::ostream &out, const ElfRel<E> &rel) {
|
||||
out << rel_to_string<E>(rel.r_type);
|
||||
return out;
|
||||
}
|
||||
|
||||
static constexpr u32 SHN_UNDEF = 0;
|
||||
static constexpr u32 SHN_ABS = 0xfff1;
|
||||
static constexpr u32 SHN_COMMON = 0xfff2;
|
||||
static constexpr u32 SHN_XINDEX = 0xffff;
|
||||
|
||||
static constexpr u32 SHT_NULL = 0;
|
||||
static constexpr u32 SHT_PROGBITS = 1;
|
||||
static constexpr u32 SHT_SYMTAB = 2;
|
||||
static constexpr u32 SHT_STRTAB = 3;
|
||||
static constexpr u32 SHT_RELA = 4;
|
||||
static constexpr u32 SHT_HASH = 5;
|
||||
static constexpr u32 SHT_DYNAMIC = 6;
|
||||
static constexpr u32 SHT_NOTE = 7;
|
||||
static constexpr u32 SHT_NOBITS = 8;
|
||||
static constexpr u32 SHT_REL = 9;
|
||||
static constexpr u32 SHT_SHLIB = 10;
|
||||
static constexpr u32 SHT_DYNSYM = 11;
|
||||
static constexpr u32 SHT_INIT_ARRAY = 14;
|
||||
static constexpr u32 SHT_FINI_ARRAY = 15;
|
||||
static constexpr u32 SHT_PREINIT_ARRAY = 16;
|
||||
static constexpr u32 SHT_GROUP = 17;
|
||||
static constexpr u32 SHT_SYMTAB_SHNDX = 18;
|
||||
static constexpr u32 SHT_GNU_HASH = 0x6ffffff6;
|
||||
static constexpr u32 SHT_GNU_VERDEF = 0x6ffffffd;
|
||||
static constexpr u32 SHT_GNU_VERNEED = 0x6ffffffe;
|
||||
static constexpr u32 SHT_GNU_VERSYM = 0x6fffffff;
|
||||
static constexpr u32 SHT_X86_64_UNWIND = 0x70000001;
|
||||
|
||||
static constexpr u32 SHF_WRITE = 0x1;
|
||||
static constexpr u32 SHF_ALLOC = 0x2;
|
||||
static constexpr u32 SHF_EXECINSTR = 0x4;
|
||||
static constexpr u32 SHF_MERGE = 0x10;
|
||||
static constexpr u32 SHF_STRINGS = 0x20;
|
||||
static constexpr u32 SHF_INFO_LINK = 0x40;
|
||||
static constexpr u32 SHF_LINK_ORDER = 0x80;
|
||||
static constexpr u32 SHF_GROUP = 0x200;
|
||||
static constexpr u32 SHF_TLS = 0x400;
|
||||
static constexpr u32 SHF_COMPRESSED = 0x800;
|
||||
static constexpr u32 SHF_EXCLUDE = 0x80000000;
|
||||
|
||||
static constexpr u32 GRP_COMDAT = 1;
|
||||
|
||||
static constexpr u32 STT_NOTYPE = 0;
|
||||
static constexpr u32 STT_OBJECT = 1;
|
||||
static constexpr u32 STT_FUNC = 2;
|
||||
static constexpr u32 STT_SECTION = 3;
|
||||
static constexpr u32 STT_FILE = 4;
|
||||
static constexpr u32 STT_COMMON = 5;
|
||||
static constexpr u32 STT_TLS = 6;
|
||||
static constexpr u32 STT_GNU_IFUNC = 10;
|
||||
|
||||
static constexpr u32 STB_LOCAL = 0;
|
||||
static constexpr u32 STB_GLOBAL = 1;
|
||||
static constexpr u32 STB_WEAK = 2;
|
||||
static constexpr u32 STB_GNU_UNIQUE = 10;
|
||||
|
||||
static constexpr u32 STV_DEFAULT = 0;
|
||||
static constexpr u32 STV_INTERNAL = 1;
|
||||
static constexpr u32 STV_HIDDEN = 2;
|
||||
static constexpr u32 STV_PROTECTED = 3;
|
||||
|
||||
static constexpr u32 VER_NDX_LOCAL = 0;
|
||||
static constexpr u32 VER_NDX_GLOBAL = 1;
|
||||
static constexpr u32 VER_NDX_LAST_RESERVED = 1;
|
||||
|
||||
static constexpr u32 VER_FLG_BASE = 1;
|
||||
static constexpr u32 VER_FLG_WEAK = 2;
|
||||
static constexpr u32 VER_FLG_INFO = 4;
|
||||
|
||||
static constexpr u32 VERSYM_HIDDEN = 0x8000;
|
||||
|
||||
static constexpr u32 PT_NULL = 0;
|
||||
static constexpr u32 PT_LOAD = 1;
|
||||
static constexpr u32 PT_DYNAMIC = 2;
|
||||
static constexpr u32 PT_INTERP = 3;
|
||||
static constexpr u32 PT_NOTE = 4;
|
||||
static constexpr u32 PT_SHLIB = 5;
|
||||
static constexpr u32 PT_PHDR = 6;
|
||||
static constexpr u32 PT_TLS = 7;
|
||||
static constexpr u32 PT_GNU_EH_FRAME = 0x6474e550;
|
||||
static constexpr u32 PT_GNU_STACK = 0x6474e551;
|
||||
static constexpr u32 PT_GNU_RELRO = 0x6474e552;
|
||||
|
||||
static constexpr u32 PF_X = 1;
|
||||
static constexpr u32 PF_W = 2;
|
||||
static constexpr u32 PF_R = 4;
|
||||
|
||||
static constexpr u32 ET_NONE = 0;
|
||||
static constexpr u32 ET_REL = 1;
|
||||
static constexpr u32 ET_EXEC = 2;
|
||||
static constexpr u32 ET_DYN = 3;
|
||||
|
||||
static constexpr u32 ELFDATA2LSB = 1;
|
||||
static constexpr u32 ELFDATA2MSB = 2;
|
||||
|
||||
static constexpr u32 ELFCLASS32 = 1;
|
||||
static constexpr u32 ELFCLASS64 = 2;
|
||||
|
||||
static constexpr u32 EV_CURRENT = 1;
|
||||
|
||||
static constexpr u32 EM_386 = 3;
|
||||
static constexpr u32 EM_X86_64 = 62;
|
||||
static constexpr u32 EM_AARCH64 = 183;
|
||||
|
||||
static constexpr u32 EI_CLASS = 4;
|
||||
static constexpr u32 EI_DATA = 5;
|
||||
static constexpr u32 EI_VERSION = 6;
|
||||
static constexpr u32 EI_OSABI = 7;
|
||||
static constexpr u32 EI_ABIVERSION = 8;
|
||||
|
||||
static constexpr u32 ELFOSABI_NONE = 0;
|
||||
static constexpr u32 ELFOSABI_GNU = 3;
|
||||
|
||||
static constexpr u32 DT_NULL = 0;
|
||||
static constexpr u32 DT_NEEDED = 1;
|
||||
static constexpr u32 DT_PLTRELSZ = 2;
|
||||
static constexpr u32 DT_PLTGOT = 3;
|
||||
static constexpr u32 DT_HASH = 4;
|
||||
static constexpr u32 DT_STRTAB = 5;
|
||||
static constexpr u32 DT_SYMTAB = 6;
|
||||
static constexpr u32 DT_RELA = 7;
|
||||
static constexpr u32 DT_RELASZ = 8;
|
||||
static constexpr u32 DT_RELAENT = 9;
|
||||
static constexpr u32 DT_STRSZ = 10;
|
||||
static constexpr u32 DT_SYMENT = 11;
|
||||
static constexpr u32 DT_INIT = 12;
|
||||
static constexpr u32 DT_FINI = 13;
|
||||
static constexpr u32 DT_SONAME = 14;
|
||||
static constexpr u32 DT_RPATH = 15;
|
||||
static constexpr u32 DT_SYMBOLIC = 16;
|
||||
static constexpr u32 DT_REL = 17;
|
||||
static constexpr u32 DT_RELSZ = 18;
|
||||
static constexpr u32 DT_RELENT = 19;
|
||||
static constexpr u32 DT_PLTREL = 20;
|
||||
static constexpr u32 DT_DEBUG = 21;
|
||||
static constexpr u32 DT_TEXTREL = 22;
|
||||
static constexpr u32 DT_JMPREL = 23;
|
||||
static constexpr u32 DT_BIND_NOW = 24;
|
||||
static constexpr u32 DT_INIT_ARRAY = 25;
|
||||
static constexpr u32 DT_FINI_ARRAY = 26;
|
||||
static constexpr u32 DT_INIT_ARRAYSZ = 27;
|
||||
static constexpr u32 DT_FINI_ARRAYSZ = 28;
|
||||
static constexpr u32 DT_RUNPATH = 29;
|
||||
static constexpr u32 DT_FLAGS = 30;
|
||||
static constexpr u32 DT_GNU_HASH = 0x6ffffef5;
|
||||
static constexpr u32 DT_VERSYM = 0x6ffffff0;
|
||||
static constexpr u32 DT_RELACOUNT = 0x6ffffff9;
|
||||
static constexpr u32 DT_RELCOUNT = 0x6ffffffa;
|
||||
static constexpr u32 DT_FLAGS_1 = 0x6ffffffb;
|
||||
static constexpr u32 DT_VERDEF = 0x6ffffffc;
|
||||
static constexpr u32 DT_VERDEFNUM = 0x6ffffffd;
|
||||
static constexpr u32 DT_VERNEED = 0x6ffffffe;
|
||||
static constexpr u32 DT_VERNEEDNUM = 0x6fffffff;
|
||||
static constexpr u32 DT_AUXILIARY = 0x7ffffffd;
|
||||
static constexpr u32 DT_FILTER = 0x7fffffff;
|
||||
|
||||
static constexpr u32 DF_ORIGIN = 0x01;
|
||||
static constexpr u32 DF_SYMBOLIC = 0x02;
|
||||
static constexpr u32 DF_TEXTREL = 0x04;
|
||||
static constexpr u32 DF_BIND_NOW = 0x08;
|
||||
static constexpr u32 DF_STATIC_TLS = 0x10;
|
||||
|
||||
static constexpr u32 DF_1_NOW = 0x00000001;
|
||||
static constexpr u32 DF_1_NODELETE = 0x00000008;
|
||||
static constexpr u32 DF_1_INITFIRST = 0x00000020;
|
||||
static constexpr u32 DF_1_NOOPEN = 0x00000040;
|
||||
static constexpr u32 DF_1_ORIGIN = 0x00000080;
|
||||
static constexpr u32 DF_1_INTERPOSE = 0x00000400;
|
||||
static constexpr u32 DF_1_NODUMP = 0x00001000;
|
||||
static constexpr u32 DF_1_PIE = 0x08000000;
|
||||
|
||||
static constexpr u32 NT_GNU_ABI_TAG = 1;
|
||||
static constexpr u32 NT_GNU_HWCAP = 2;
|
||||
static constexpr u32 NT_GNU_BUILD_ID = 3;
|
||||
static constexpr u32 NT_GNU_GOLD_VERSION = 4;
|
||||
static constexpr u32 NT_GNU_PROPERTY_TYPE_0 = 5;
|
||||
|
||||
static constexpr u32 GNU_PROPERTY_AARCH64_FEATURE_1_AND = 0xc0000000;
|
||||
static constexpr u32 GNU_PROPERTY_X86_FEATURE_1_AND = 0xc0000002;
|
||||
|
||||
static constexpr u32 ELFCOMPRESS_ZLIB = 1;
|
||||
|
||||
static constexpr u32 R_X86_64_NONE = 0;
|
||||
static constexpr u32 R_X86_64_64 = 1;
|
||||
static constexpr u32 R_X86_64_PC32 = 2;
|
||||
static constexpr u32 R_X86_64_GOT32 = 3;
|
||||
static constexpr u32 R_X86_64_PLT32 = 4;
|
||||
static constexpr u32 R_X86_64_COPY = 5;
|
||||
static constexpr u32 R_X86_64_GLOB_DAT = 6;
|
||||
static constexpr u32 R_X86_64_JUMP_SLOT = 7;
|
||||
static constexpr u32 R_X86_64_RELATIVE = 8;
|
||||
static constexpr u32 R_X86_64_GOTPCREL = 9;
|
||||
static constexpr u32 R_X86_64_32 = 10;
|
||||
static constexpr u32 R_X86_64_32S = 11;
|
||||
static constexpr u32 R_X86_64_16 = 12;
|
||||
static constexpr u32 R_X86_64_PC16 = 13;
|
||||
static constexpr u32 R_X86_64_8 = 14;
|
||||
static constexpr u32 R_X86_64_PC8 = 15;
|
||||
static constexpr u32 R_X86_64_DTPMOD64 = 16;
|
||||
static constexpr u32 R_X86_64_DTPOFF64 = 17;
|
||||
static constexpr u32 R_X86_64_TPOFF64 = 18;
|
||||
static constexpr u32 R_X86_64_TLSGD = 19;
|
||||
static constexpr u32 R_X86_64_TLSLD = 20;
|
||||
static constexpr u32 R_X86_64_DTPOFF32 = 21;
|
||||
static constexpr u32 R_X86_64_GOTTPOFF = 22;
|
||||
static constexpr u32 R_X86_64_TPOFF32 = 23;
|
||||
static constexpr u32 R_X86_64_PC64 = 24;
|
||||
static constexpr u32 R_X86_64_GOTOFF64 = 25;
|
||||
static constexpr u32 R_X86_64_GOTPC32 = 26;
|
||||
static constexpr u32 R_X86_64_GOT64 = 27;
|
||||
static constexpr u32 R_X86_64_GOTPCREL64 = 28;
|
||||
static constexpr u32 R_X86_64_GOTPC64 = 29;
|
||||
static constexpr u32 R_X86_64_GOTPLT64 = 30;
|
||||
static constexpr u32 R_X86_64_PLTOFF64 = 31;
|
||||
static constexpr u32 R_X86_64_SIZE32 = 32;
|
||||
static constexpr u32 R_X86_64_SIZE64 = 33;
|
||||
static constexpr u32 R_X86_64_GOTPC32_TLSDESC = 34;
|
||||
static constexpr u32 R_X86_64_TLSDESC_CALL = 35;
|
||||
static constexpr u32 R_X86_64_TLSDESC = 36;
|
||||
static constexpr u32 R_X86_64_IRELATIVE = 37;
|
||||
static constexpr u32 R_X86_64_GOTPCRELX = 41;
|
||||
static constexpr u32 R_X86_64_REX_GOTPCRELX = 42;
|
||||
|
||||
template <>
|
||||
inline std::string rel_to_string<X86_64>(u32 r_type) {
|
||||
switch (r_type) {
|
||||
case R_X86_64_NONE: return "R_X86_64_NONE";
|
||||
case R_X86_64_64: return "R_X86_64_64";
|
||||
case R_X86_64_PC32: return "R_X86_64_PC32";
|
||||
case R_X86_64_GOT32: return "R_X86_64_GOT32";
|
||||
case R_X86_64_PLT32: return "R_X86_64_PLT32";
|
||||
case R_X86_64_COPY: return "R_X86_64_COPY";
|
||||
case R_X86_64_GLOB_DAT: return "R_X86_64_GLOB_DAT";
|
||||
case R_X86_64_JUMP_SLOT: return "R_X86_64_JUMP_SLOT";
|
||||
case R_X86_64_RELATIVE: return "R_X86_64_RELATIVE";
|
||||
case R_X86_64_GOTPCREL: return "R_X86_64_GOTPCREL";
|
||||
case R_X86_64_32: return "R_X86_64_32";
|
||||
case R_X86_64_32S: return "R_X86_64_32S";
|
||||
case R_X86_64_16: return "R_X86_64_16";
|
||||
case R_X86_64_PC16: return "R_X86_64_PC16";
|
||||
case R_X86_64_8: return "R_X86_64_8";
|
||||
case R_X86_64_PC8: return "R_X86_64_PC8";
|
||||
case R_X86_64_DTPMOD64: return "R_X86_64_DTPMOD64";
|
||||
case R_X86_64_DTPOFF64: return "R_X86_64_DTPOFF64";
|
||||
case R_X86_64_TPOFF64: return "R_X86_64_TPOFF64";
|
||||
case R_X86_64_TLSGD: return "R_X86_64_TLSGD";
|
||||
case R_X86_64_TLSLD: return "R_X86_64_TLSLD";
|
||||
case R_X86_64_DTPOFF32: return "R_X86_64_DTPOFF32";
|
||||
case R_X86_64_GOTTPOFF: return "R_X86_64_GOTTPOFF";
|
||||
case R_X86_64_TPOFF32: return "R_X86_64_TPOFF32";
|
||||
case R_X86_64_PC64: return "R_X86_64_PC64";
|
||||
case R_X86_64_GOTOFF64: return "R_X86_64_GOTOFF64";
|
||||
case R_X86_64_GOTPC32: return "R_X86_64_GOTPC32";
|
||||
case R_X86_64_GOT64: return "R_X86_64_GOT64";
|
||||
case R_X86_64_GOTPCREL64: return "R_X86_64_GOTPCREL64";
|
||||
case R_X86_64_GOTPC64: return "R_X86_64_GOTPC64";
|
||||
case R_X86_64_GOTPLT64: return "R_X86_64_GOTPLT64";
|
||||
case R_X86_64_PLTOFF64: return "R_X86_64_PLTOFF64";
|
||||
case R_X86_64_SIZE32: return "R_X86_64_SIZE32";
|
||||
case R_X86_64_SIZE64: return "R_X86_64_SIZE64";
|
||||
case R_X86_64_GOTPC32_TLSDESC: return "R_X86_64_GOTPC32_TLSDESC";
|
||||
case R_X86_64_TLSDESC_CALL: return "R_X86_64_TLSDESC_CALL";
|
||||
case R_X86_64_TLSDESC: return "R_X86_64_TLSDESC";
|
||||
case R_X86_64_IRELATIVE: return "R_X86_64_IRELATIVE";
|
||||
case R_X86_64_GOTPCRELX: return "R_X86_64_GOTPCRELX";
|
||||
case R_X86_64_REX_GOTPCRELX: return "R_X86_64_REX_GOTPCRELX";
|
||||
}
|
||||
return "unknown (" + std::to_string(r_type) + ")";
|
||||
}
|
||||
|
||||
static constexpr u32 R_386_NONE = 0;
|
||||
static constexpr u32 R_386_32 = 1;
|
||||
static constexpr u32 R_386_PC32 = 2;
|
||||
static constexpr u32 R_386_GOT32 = 3;
|
||||
static constexpr u32 R_386_PLT32 = 4;
|
||||
static constexpr u32 R_386_COPY = 5;
|
||||
static constexpr u32 R_386_GLOB_DAT = 6;
|
||||
static constexpr u32 R_386_JUMP_SLOT = 7;
|
||||
static constexpr u32 R_386_RELATIVE = 8;
|
||||
static constexpr u32 R_386_GOTOFF = 9;
|
||||
static constexpr u32 R_386_GOTPC = 10;
|
||||
static constexpr u32 R_386_32PLT = 11;
|
||||
static constexpr u32 R_386_TLS_TPOFF = 14;
|
||||
static constexpr u32 R_386_TLS_IE = 15;
|
||||
static constexpr u32 R_386_TLS_GOTIE = 16;
|
||||
static constexpr u32 R_386_TLS_LE = 17;
|
||||
static constexpr u32 R_386_TLS_GD = 18;
|
||||
static constexpr u32 R_386_TLS_LDM = 19;
|
||||
static constexpr u32 R_386_16 = 20;
|
||||
static constexpr u32 R_386_PC16 = 21;
|
||||
static constexpr u32 R_386_8 = 22;
|
||||
static constexpr u32 R_386_PC8 = 23;
|
||||
static constexpr u32 R_386_TLS_GD_32 = 24;
|
||||
static constexpr u32 R_386_TLS_GD_PUSH = 25;
|
||||
static constexpr u32 R_386_TLS_GD_CALL = 26;
|
||||
static constexpr u32 R_386_TLS_GD_POP = 27;
|
||||
static constexpr u32 R_386_TLS_LDM_32 = 28;
|
||||
static constexpr u32 R_386_TLS_LDM_PUSH = 29;
|
||||
static constexpr u32 R_386_TLS_LDM_CALL = 30;
|
||||
static constexpr u32 R_386_TLS_LDM_POP = 31;
|
||||
static constexpr u32 R_386_TLS_LDO_32 = 32;
|
||||
static constexpr u32 R_386_TLS_IE_32 = 33;
|
||||
static constexpr u32 R_386_TLS_LE_32 = 34;
|
||||
static constexpr u32 R_386_TLS_DTPMOD32 = 35;
|
||||
static constexpr u32 R_386_TLS_DTPOFF32 = 36;
|
||||
static constexpr u32 R_386_TLS_TPOFF32 = 37;
|
||||
static constexpr u32 R_386_SIZE32 = 38;
|
||||
static constexpr u32 R_386_TLS_GOTDESC = 39;
|
||||
static constexpr u32 R_386_TLS_DESC_CALL = 40;
|
||||
static constexpr u32 R_386_TLS_DESC = 41;
|
||||
static constexpr u32 R_386_IRELATIVE = 42;
|
||||
static constexpr u32 R_386_GOT32X = 43;
|
||||
|
||||
template <>
|
||||
inline std::string rel_to_string<I386>(u32 r_type) {
|
||||
switch (r_type) {
|
||||
case R_386_NONE: return "R_386_NONE";
|
||||
case R_386_32: return "R_386_32";
|
||||
case R_386_PC32: return "R_386_PC32";
|
||||
case R_386_GOT32: return "R_386_GOT32";
|
||||
case R_386_PLT32: return "R_386_PLT32";
|
||||
case R_386_COPY: return "R_386_COPY";
|
||||
case R_386_GLOB_DAT: return "R_386_GLOB_DAT";
|
||||
case R_386_JUMP_SLOT: return "R_386_JUMP_SLOT";
|
||||
case R_386_RELATIVE: return "R_386_RELATIVE";
|
||||
case R_386_GOTOFF: return "R_386_GOTOFF";
|
||||
case R_386_GOTPC: return "R_386_GOTPC";
|
||||
case R_386_32PLT: return "R_386_32PLT";
|
||||
case R_386_TLS_TPOFF: return "R_386_TLS_TPOFF";
|
||||
case R_386_TLS_IE: return "R_386_TLS_IE";
|
||||
case R_386_TLS_GOTIE: return "R_386_TLS_GOTIE";
|
||||
case R_386_TLS_LE: return "R_386_TLS_LE";
|
||||
case R_386_TLS_GD: return "R_386_TLS_GD";
|
||||
case R_386_TLS_LDM: return "R_386_TLS_LDM";
|
||||
case R_386_16: return "R_386_16";
|
||||
case R_386_PC16: return "R_386_PC16";
|
||||
case R_386_8: return "R_386_8";
|
||||
case R_386_PC8: return "R_386_PC8";
|
||||
case R_386_TLS_GD_32: return "R_386_TLS_GD_32";
|
||||
case R_386_TLS_GD_PUSH: return "R_386_TLS_GD_PUSH";
|
||||
case R_386_TLS_GD_CALL: return "R_386_TLS_GD_CALL";
|
||||
case R_386_TLS_GD_POP: return "R_386_TLS_GD_POP";
|
||||
case R_386_TLS_LDM_32: return "R_386_TLS_LDM_32";
|
||||
case R_386_TLS_LDM_PUSH: return "R_386_TLS_LDM_PUSH";
|
||||
case R_386_TLS_LDM_CALL: return "R_386_TLS_LDM_CALL";
|
||||
case R_386_TLS_LDM_POP: return "R_386_TLS_LDM_POP";
|
||||
case R_386_TLS_LDO_32: return "R_386_TLS_LDO_32";
|
||||
case R_386_TLS_IE_32: return "R_386_TLS_IE_32";
|
||||
case R_386_TLS_LE_32: return "R_386_TLS_LE_32";
|
||||
case R_386_TLS_DTPMOD32: return "R_386_TLS_DTPMOD32";
|
||||
case R_386_TLS_DTPOFF32: return "R_386_TLS_DTPOFF32";
|
||||
case R_386_TLS_TPOFF32: return "R_386_TLS_TPOFF32";
|
||||
case R_386_SIZE32: return "R_386_SIZE32";
|
||||
case R_386_TLS_GOTDESC: return "R_386_TLS_GOTDESC";
|
||||
case R_386_TLS_DESC_CALL: return "R_386_TLS_DESC_CALL";
|
||||
case R_386_TLS_DESC: return "R_386_TLS_DESC";
|
||||
case R_386_IRELATIVE: return "R_386_IRELATIVE";
|
||||
case R_386_GOT32X: return "R_386_GOT32X";
|
||||
}
|
||||
return "unknown (" + std::to_string(r_type) + ")";
|
||||
}
|
||||
|
||||
static constexpr u32 R_AARCH64_NONE = 0;
|
||||
static constexpr u32 R_AARCH64_ABS64 = 0x101;
|
||||
static constexpr u32 R_AARCH64_ABS32 = 0x102;
|
||||
static constexpr u32 R_AARCH64_ABS16 = 0x103;
|
||||
static constexpr u32 R_AARCH64_PREL64 = 0x104;
|
||||
static constexpr u32 R_AARCH64_PREL32 = 0x105;
|
||||
static constexpr u32 R_AARCH64_PREL16 = 0x106;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G0 = 0x107;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G0_NC = 0x108;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G1 = 0x109;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G1_NC = 0x10a;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G2 = 0x10b;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G2_NC = 0x10c;
|
||||
static constexpr u32 R_AARCH64_MOVW_UABS_G3 = 0x10d;
|
||||
static constexpr u32 R_AARCH64_MOVW_SABS_G0 = 0x10e;
|
||||
static constexpr u32 R_AARCH64_MOVW_SABS_G1 = 0x10f;
|
||||
static constexpr u32 R_AARCH64_MOVW_SABS_G2 = 0x110;
|
||||
static constexpr u32 R_AARCH64_LD_PREL_LO19 = 0x111;
|
||||
static constexpr u32 R_AARCH64_ADR_PREL_LO21 = 0x112;
|
||||
static constexpr u32 R_AARCH64_ADR_PREL_PG_HI21 = 0x113;
|
||||
static constexpr u32 R_AARCH64_ADR_PREL_PG_HI21_NC = 0x114;
|
||||
static constexpr u32 R_AARCH64_ADD_ABS_LO12_NC = 0x115;
|
||||
static constexpr u32 R_AARCH64_LDST8_ABS_LO12_NC = 0x116;
|
||||
static constexpr u32 R_AARCH64_TSTBR14 = 0x117;
|
||||
static constexpr u32 R_AARCH64_CONDBR19 = 0x118;
|
||||
static constexpr u32 R_AARCH64_JUMP26 = 0x11a;
|
||||
static constexpr u32 R_AARCH64_CALL26 = 0x11b;
|
||||
static constexpr u32 R_AARCH64_LDST16_ABS_LO12_NC = 0x11c;
|
||||
static constexpr u32 R_AARCH64_LDST32_ABS_LO12_NC = 0x11d;
|
||||
static constexpr u32 R_AARCH64_LDST64_ABS_LO12_NC = 0x11e;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G0 = 0x11f;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G0_NC = 0x120;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G1 = 0x121;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G1_NC = 0x122;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G2 = 0x123;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G2_NC = 0x124;
|
||||
static constexpr u32 R_AARCH64_MOVW_PREL_G3 = 0x125;
|
||||
static constexpr u32 R_AARCH64_LDST128_ABS_LO12_NC = 0x12b;
|
||||
static constexpr u32 R_AARCH64_ADR_GOT_PAGE = 0x137;
|
||||
static constexpr u32 R_AARCH64_LD64_GOT_LO12_NC = 0x138;
|
||||
static constexpr u32 R_AARCH64_LD64_GOTPAGE_LO15 = 0x139;
|
||||
static constexpr u32 R_AARCH64_PLT32 = 0x13a;
|
||||
static constexpr u32 R_AARCH64_TLSGD_ADR_PREL21 = 0x200;
|
||||
static constexpr u32 R_AARCH64_TLSGD_ADR_PAGE21 = 0x201;
|
||||
static constexpr u32 R_AARCH64_TLSGD_ADD_LO12_NC = 0x202;
|
||||
static constexpr u32 R_AARCH64_TLSGD_MOVW_G1 = 0x203;
|
||||
static constexpr u32 R_AARCH64_TLSGD_MOVW_G0_NC = 0x204;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADR_PREL21 = 0x205;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADR_PAGE21 = 0x206;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADD_LO12_NC = 0x207;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_G1 = 0x208;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_G0_NC = 0x209;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LD_PREL19 = 0x20a;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_DTPREL_G2 = 0x20b;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_DTPREL_G1 = 0x20c;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_DTPREL_G1_NC = 0x20d;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_DTPREL_G0 = 0x20e;
|
||||
static constexpr u32 R_AARCH64_TLSLD_MOVW_DTPREL_G0_NC = 0x20f;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADD_DTPREL_HI12 = 0x210;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADD_DTPREL_LO12 = 0x211;
|
||||
static constexpr u32 R_AARCH64_TLSLD_ADD_DTPREL_LO12_NC = 0x212;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST8_DTPREL_LO12 = 0x213;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST8_DTPREL_LO12_NC = 0x214;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST16_DTPREL_LO12 = 0x215;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST16_DTPREL_LO12_NC = 0x216;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST32_DTPREL_LO12 = 0x217;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST32_DTPREL_LO12_NC = 0x218;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST64_DTPREL_LO12 = 0x219;
|
||||
static constexpr u32 R_AARCH64_TLSLD_LDST64_DTPREL_LO12_NC = 0x21a;
|
||||
static constexpr u32 R_AARCH64_TLSIE_MOVW_GOTTPREL_G1 = 0x21b;
|
||||
static constexpr u32 R_AARCH64_TLSIE_MOVW_GOTTPREL_G0_NC = 0x21c;
|
||||
static constexpr u32 R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21 = 0x21d;
|
||||
static constexpr u32 R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC = 0x21e;
|
||||
static constexpr u32 R_AARCH64_TLSIE_LD_GOTTPREL_PREL19 = 0x21f;
|
||||
static constexpr u32 R_AARCH64_TLSLE_MOVW_TPREL_G2 = 0x220;
|
||||
static constexpr u32 R_AARCH64_TLSLE_MOVW_TPREL_G1 = 0x221;
|
||||
static constexpr u32 R_AARCH64_TLSLE_MOVW_TPREL_G1_NC = 0x222;
|
||||
static constexpr u32 R_AARCH64_TLSLE_MOVW_TPREL_G0 = 0x223;
|
||||
static constexpr u32 R_AARCH64_TLSLE_MOVW_TPREL_G0_NC = 0x224;
|
||||
static constexpr u32 R_AARCH64_TLSLE_ADD_TPREL_HI12 = 0x225;
|
||||
static constexpr u32 R_AARCH64_TLSLE_ADD_TPREL_LO12 = 0x226;
|
||||
static constexpr u32 R_AARCH64_TLSLE_ADD_TPREL_LO12_NC = 0x227;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST8_TPREL_LO12 = 0x228;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST8_TPREL_LO12_NC = 0x229;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST16_TPREL_LO12 = 0x22a;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST16_TPREL_LO12_NC = 0x22b;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST32_TPREL_LO12 = 0x22c;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST32_TPREL_LO12_NC = 0x22d;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST64_TPREL_LO12 = 0x22e;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST64_TPREL_LO12_NC = 0x22f;
|
||||
static constexpr u32 R_AARCH64_TLSDESC_ADR_PAGE21 = 0x232;
|
||||
static constexpr u32 R_AARCH64_TLSDESC_LD64_LO12 = 0x233;
|
||||
static constexpr u32 R_AARCH64_TLSDESC_ADD_LO12 = 0x234;
|
||||
static constexpr u32 R_AARCH64_TLSDESC_CALL = 0x239;
|
||||
static constexpr u32 R_AARCH64_TLSLE_LDST128_TPREL_LO12_NC = 0x23b;
|
||||
static constexpr u32 R_AARCH64_COPY = 0x400;
|
||||
static constexpr u32 R_AARCH64_GLOB_DAT = 0x401;
|
||||
static constexpr u32 R_AARCH64_JUMP_SLOT = 0x402;
|
||||
static constexpr u32 R_AARCH64_RELATIVE = 0x403;
|
||||
static constexpr u32 R_AARCH64_TLS_DTPMOD64 = 0x404;
|
||||
static constexpr u32 R_AARCH64_TLS_DTPREL64 = 0x405;
|
||||
static constexpr u32 R_AARCH64_TLS_TPREL64 = 0x406;
|
||||
static constexpr u32 R_AARCH64_TLSDESC = 0x407;
|
||||
static constexpr u32 R_AARCH64_IRELATIVE = 0x408;
|
||||
|
||||
template <>
|
||||
inline std::string rel_to_string<AARCH64>(u32 r_type) {
|
||||
switch (r_type) {
|
||||
case R_AARCH64_NONE: return "R_AARCH64_NONE";
|
||||
case R_AARCH64_ABS64: return "R_AARCH64_ABS64";
|
||||
case R_AARCH64_ABS32: return "R_AARCH64_ABS32";
|
||||
case R_AARCH64_ABS16: return "R_AARCH64_ABS16";
|
||||
case R_AARCH64_PREL64: return "R_AARCH64_PREL64";
|
||||
case R_AARCH64_PREL32: return "R_AARCH64_PREL32";
|
||||
case R_AARCH64_PREL16: return "R_AARCH64_PREL16";
|
||||
case R_AARCH64_MOVW_UABS_G0: return "R_AARCH64_MOVW_UABS_G0";
|
||||
case R_AARCH64_MOVW_UABS_G0_NC: return "R_AARCH64_MOVW_UABS_G0_NC";
|
||||
case R_AARCH64_MOVW_UABS_G1: return "R_AARCH64_MOVW_UABS_G1";
|
||||
case R_AARCH64_MOVW_UABS_G1_NC: return "R_AARCH64_MOVW_UABS_G1_NC";
|
||||
case R_AARCH64_MOVW_UABS_G2: return "R_AARCH64_MOVW_UABS_G2";
|
||||
case R_AARCH64_MOVW_UABS_G2_NC: return "R_AARCH64_MOVW_UABS_G2_NC";
|
||||
case R_AARCH64_MOVW_UABS_G3: return "R_AARCH64_MOVW_UABS_G3";
|
||||
case R_AARCH64_MOVW_SABS_G0: return "R_AARCH64_MOVW_SABS_G0";
|
||||
case R_AARCH64_MOVW_SABS_G1: return "R_AARCH64_MOVW_SABS_G1";
|
||||
case R_AARCH64_MOVW_SABS_G2: return "R_AARCH64_MOVW_SABS_G2";
|
||||
case R_AARCH64_LD_PREL_LO19: return "R_AARCH64_LD_PREL_LO19";
|
||||
case R_AARCH64_ADR_PREL_LO21: return "R_AARCH64_ADR_PREL_LO21";
|
||||
case R_AARCH64_ADR_PREL_PG_HI21: return "R_AARCH64_ADR_PREL_PG_HI21";
|
||||
case R_AARCH64_ADR_PREL_PG_HI21_NC: return "R_AARCH64_ADR_PREL_PG_HI21_NC";
|
||||
case R_AARCH64_ADD_ABS_LO12_NC: return "R_AARCH64_ADD_ABS_LO12_NC";
|
||||
case R_AARCH64_LDST8_ABS_LO12_NC: return "R_AARCH64_LDST8_ABS_LO12_NC";
|
||||
case R_AARCH64_TSTBR14: return "R_AARCH64_TSTBR14";
|
||||
case R_AARCH64_CONDBR19: return "R_AARCH64_CONDBR19";
|
||||
case R_AARCH64_JUMP26: return "R_AARCH64_JUMP26";
|
||||
case R_AARCH64_CALL26: return "R_AARCH64_CALL26";
|
||||
case R_AARCH64_LDST16_ABS_LO12_NC: return "R_AARCH64_LDST16_ABS_LO12_NC";
|
||||
case R_AARCH64_LDST32_ABS_LO12_NC: return "R_AARCH64_LDST32_ABS_LO12_NC";
|
||||
case R_AARCH64_LDST64_ABS_LO12_NC: return "R_AARCH64_LDST64_ABS_LO12_NC";
|
||||
case R_AARCH64_MOVW_PREL_G0: return "R_AARCH64_MOVW_PREL_G0";
|
||||
case R_AARCH64_MOVW_PREL_G0_NC: return "R_AARCH64_MOVW_PREL_G0_NC";
|
||||
case R_AARCH64_MOVW_PREL_G1: return "R_AARCH64_MOVW_PREL_G1";
|
||||
case R_AARCH64_MOVW_PREL_G1_NC: return "R_AARCH64_MOVW_PREL_G1_NC";
|
||||
case R_AARCH64_MOVW_PREL_G2: return "R_AARCH64_MOVW_PREL_G2";
|
||||
case R_AARCH64_MOVW_PREL_G2_NC: return "R_AARCH64_MOVW_PREL_G2_NC";
|
||||
case R_AARCH64_MOVW_PREL_G3: return "R_AARCH64_MOVW_PREL_G3";
|
||||
case R_AARCH64_LDST128_ABS_LO12_NC: return "R_AARCH64_LDST128_ABS_LO12_NC";
|
||||
case R_AARCH64_ADR_GOT_PAGE: return "R_AARCH64_ADR_GOT_PAGE";
|
||||
case R_AARCH64_LD64_GOT_LO12_NC: return "R_AARCH64_LD64_GOT_LO12_NC";
|
||||
case R_AARCH64_LD64_GOTPAGE_LO15: return "R_AARCH64_LD64_GOTPAGE_LO15";
|
||||
case R_AARCH64_PLT32: return "R_AARCH64_PLT32";
|
||||
case R_AARCH64_TLSGD_ADR_PREL21: return "R_AARCH64_TLSGD_ADR_PREL21";
|
||||
case R_AARCH64_TLSGD_ADR_PAGE21: return "R_AARCH64_TLSGD_ADR_PAGE21";
|
||||
case R_AARCH64_TLSGD_ADD_LO12_NC: return "R_AARCH64_TLSGD_ADD_LO12_NC";
|
||||
case R_AARCH64_TLSGD_MOVW_G1: return "R_AARCH64_TLSGD_MOVW_G1";
|
||||
case R_AARCH64_TLSGD_MOVW_G0_NC: return "R_AARCH64_TLSGD_MOVW_G0_NC";
|
||||
case R_AARCH64_TLSLD_ADR_PREL21: return "R_AARCH64_TLSLD_ADR_PREL21";
|
||||
case R_AARCH64_TLSLD_ADR_PAGE21: return "R_AARCH64_TLSLD_ADR_PAGE21";
|
||||
case R_AARCH64_TLSLD_ADD_LO12_NC: return "R_AARCH64_TLSLD_ADD_LO12_NC";
|
||||
case R_AARCH64_TLSLD_MOVW_G1: return "R_AARCH64_TLSLD_MOVW_G1";
|
||||
case R_AARCH64_TLSLD_MOVW_G0_NC: return "R_AARCH64_TLSLD_MOVW_G0_NC";
|
||||
case R_AARCH64_TLSLD_LD_PREL19: return "R_AARCH64_TLSLD_LD_PREL19";
|
||||
case R_AARCH64_TLSLD_MOVW_DTPREL_G2: return "R_AARCH64_TLSLD_MOVW_DTPREL_G2";
|
||||
case R_AARCH64_TLSLD_MOVW_DTPREL_G1: return "R_AARCH64_TLSLD_MOVW_DTPREL_G1";
|
||||
case R_AARCH64_TLSLD_MOVW_DTPREL_G1_NC: return "R_AARCH64_TLSLD_MOVW_DTPREL_G1_NC";
|
||||
case R_AARCH64_TLSLD_MOVW_DTPREL_G0: return "R_AARCH64_TLSLD_MOVW_DTPREL_G0";
|
||||
case R_AARCH64_TLSLD_MOVW_DTPREL_G0_NC: return "R_AARCH64_TLSLD_MOVW_DTPREL_G0_NC";
|
||||
case R_AARCH64_TLSLD_ADD_DTPREL_HI12: return "R_AARCH64_TLSLD_ADD_DTPREL_HI12";
|
||||
case R_AARCH64_TLSLD_ADD_DTPREL_LO12: return "R_AARCH64_TLSLD_ADD_DTPREL_LO12";
|
||||
case R_AARCH64_TLSLD_ADD_DTPREL_LO12_NC: return "R_AARCH64_TLSLD_ADD_DTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLD_LDST8_DTPREL_LO12: return "R_AARCH64_TLSLD_LDST8_DTPREL_LO12";
|
||||
case R_AARCH64_TLSLD_LDST8_DTPREL_LO12_NC: return "R_AARCH64_TLSLD_LDST8_DTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLD_LDST16_DTPREL_LO12: return "R_AARCH64_TLSLD_LDST16_DTPREL_LO12";
|
||||
case R_AARCH64_TLSLD_LDST16_DTPREL_LO12_NC: return "R_AARCH64_TLSLD_LDST16_DTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLD_LDST32_DTPREL_LO12: return "R_AARCH64_TLSLD_LDST32_DTPREL_LO12";
|
||||
case R_AARCH64_TLSLD_LDST32_DTPREL_LO12_NC: return "R_AARCH64_TLSLD_LDST32_DTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLD_LDST64_DTPREL_LO12: return "R_AARCH64_TLSLD_LDST64_DTPREL_LO12";
|
||||
case R_AARCH64_TLSLD_LDST64_DTPREL_LO12_NC: return "R_AARCH64_TLSLD_LDST64_DTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSIE_MOVW_GOTTPREL_G1: return "R_AARCH64_TLSIE_MOVW_GOTTPREL_G1";
|
||||
case R_AARCH64_TLSIE_MOVW_GOTTPREL_G0_NC: return "R_AARCH64_TLSIE_MOVW_GOTTPREL_G0_NC";
|
||||
case R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21: return "R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21";
|
||||
case R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC: return "R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC";
|
||||
case R_AARCH64_TLSIE_LD_GOTTPREL_PREL19: return "R_AARCH64_TLSIE_LD_GOTTPREL_PREL19";
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G2: return "R_AARCH64_TLSLE_MOVW_TPREL_G2";
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1: return "R_AARCH64_TLSLE_MOVW_TPREL_G1";
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1_NC: return "R_AARCH64_TLSLE_MOVW_TPREL_G1_NC";
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0: return "R_AARCH64_TLSLE_MOVW_TPREL_G0";
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0_NC: return "R_AARCH64_TLSLE_MOVW_TPREL_G0_NC";
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_HI12: return "R_AARCH64_TLSLE_ADD_TPREL_HI12";
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12: return "R_AARCH64_TLSLE_ADD_TPREL_LO12";
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12_NC: return "R_AARCH64_TLSLE_ADD_TPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLE_LDST8_TPREL_LO12: return "R_AARCH64_TLSLE_LDST8_TPREL_LO12";
|
||||
case R_AARCH64_TLSLE_LDST8_TPREL_LO12_NC: return "R_AARCH64_TLSLE_LDST8_TPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLE_LDST16_TPREL_LO12: return "R_AARCH64_TLSLE_LDST16_TPREL_LO12";
|
||||
case R_AARCH64_TLSLE_LDST16_TPREL_LO12_NC: return "R_AARCH64_TLSLE_LDST16_TPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLE_LDST32_TPREL_LO12: return "R_AARCH64_TLSLE_LDST32_TPREL_LO12";
|
||||
case R_AARCH64_TLSLE_LDST32_TPREL_LO12_NC: return "R_AARCH64_TLSLE_LDST32_TPREL_LO12_NC";
|
||||
case R_AARCH64_TLSLE_LDST64_TPREL_LO12: return "R_AARCH64_TLSLE_LDST64_TPREL_LO12";
|
||||
case R_AARCH64_TLSLE_LDST64_TPREL_LO12_NC: return "R_AARCH64_TLSLE_LDST64_TPREL_LO12_NC";
|
||||
case R_AARCH64_TLSDESC_ADR_PAGE21: return "R_AARCH64_TLSDESC_ADR_PAGE21";
|
||||
case R_AARCH64_TLSDESC_LD64_LO12: return "R_AARCH64_TLSDESC_LD64_LO12";
|
||||
case R_AARCH64_TLSDESC_ADD_LO12: return "R_AARCH64_TLSDESC_ADD_LO12";
|
||||
case R_AARCH64_TLSDESC_CALL: return "R_AARCH64_TLSDESC_CALL";
|
||||
case R_AARCH64_TLSLE_LDST128_TPREL_LO12_NC: return "R_AARCH64_TLSLE_LDST128_TPREL_LO12_NC";
|
||||
case R_AARCH64_COPY: return "R_AARCH64_COPY";
|
||||
case R_AARCH64_GLOB_DAT: return "R_AARCH64_GLOB_DAT";
|
||||
case R_AARCH64_JUMP_SLOT: return "R_AARCH64_JUMP_SLOT";
|
||||
case R_AARCH64_RELATIVE: return "R_AARCH64_RELATIVE";
|
||||
case R_AARCH64_TLS_DTPMOD64: return "R_AARCH64_TLS_DTPMOD64";
|
||||
case R_AARCH64_TLS_DTPREL64: return "R_AARCH64_TLS_DTPREL64";
|
||||
case R_AARCH64_TLS_TPREL64: return "R_AARCH64_TLS_TPREL64";
|
||||
case R_AARCH64_TLSDESC: return "R_AARCH64_TLSDESC";
|
||||
case R_AARCH64_IRELATIVE: return "R_AARCH64_IRELATIVE";
|
||||
}
|
||||
return "unknown (" + std::to_string(r_type) + ")";
|
||||
}
|
||||
|
||||
static constexpr u32 DW_EH_PE_absptr = 0;
|
||||
static constexpr u32 DW_EH_PE_omit = 0xff;
|
||||
static constexpr u32 DW_EH_PE_uleb128 = 0x01;
|
||||
static constexpr u32 DW_EH_PE_udata2 = 0x02;
|
||||
static constexpr u32 DW_EH_PE_udata4 = 0x03;
|
||||
static constexpr u32 DW_EH_PE_udata8 = 0x04;
|
||||
static constexpr u32 DW_EH_PE_signed = 0x08;
|
||||
static constexpr u32 DW_EH_PE_sleb128 = 0x09;
|
||||
static constexpr u32 DW_EH_PE_sdata2 = 0x0a;
|
||||
static constexpr u32 DW_EH_PE_sdata4 = 0x0b;
|
||||
static constexpr u32 DW_EH_PE_sdata8 = 0x0c;
|
||||
static constexpr u32 DW_EH_PE_pcrel = 0x10;
|
||||
static constexpr u32 DW_EH_PE_textrel = 0x20;
|
||||
static constexpr u32 DW_EH_PE_datarel = 0x30;
|
||||
static constexpr u32 DW_EH_PE_funcrel = 0x40;
|
||||
static constexpr u32 DW_EH_PE_aligned = 0x50;
|
||||
|
||||
struct Elf64Sym {
|
||||
bool is_defined() const { return !is_undef(); }
|
||||
bool is_undef() const { return st_shndx == SHN_UNDEF; }
|
||||
bool is_abs() const { return st_shndx == SHN_ABS; }
|
||||
bool is_common() const { return st_shndx == SHN_COMMON; }
|
||||
bool is_weak() const { return st_bind == STB_WEAK; }
|
||||
|
||||
bool is_undef_strong() const {
|
||||
return st_shndx == SHN_UNDEF && st_bind != STB_WEAK;
|
||||
}
|
||||
|
||||
bool is_undef_weak() const {
|
||||
return st_shndx == SHN_UNDEF && st_bind == STB_WEAK;
|
||||
}
|
||||
|
||||
u32 st_name;
|
||||
u8 st_type : 4;
|
||||
u8 st_bind : 4;
|
||||
u8 st_visibility : 2;
|
||||
u16 st_shndx;
|
||||
u64 st_value;
|
||||
u64 st_size;
|
||||
};
|
||||
|
||||
struct Elf32Sym {
|
||||
bool is_defined() const { return !is_undef(); }
|
||||
bool is_undef() const { return st_shndx == SHN_UNDEF; }
|
||||
bool is_abs() const { return st_shndx == SHN_ABS; }
|
||||
bool is_common() const { return st_shndx == SHN_COMMON; }
|
||||
bool is_weak() const { return st_bind == STB_WEAK; }
|
||||
|
||||
bool is_undef_strong() const {
|
||||
return st_shndx == SHN_UNDEF && st_bind != STB_WEAK;
|
||||
}
|
||||
|
||||
bool is_undef_weak() const {
|
||||
return st_shndx == SHN_UNDEF && st_bind == STB_WEAK;
|
||||
}
|
||||
|
||||
u32 st_name;
|
||||
u32 st_value;
|
||||
u32 st_size;
|
||||
u8 st_type : 4;
|
||||
u8 st_bind : 4;
|
||||
u8 st_visibility : 2;
|
||||
u16 st_shndx;
|
||||
};
|
||||
|
||||
struct Elf64Shdr {
|
||||
u32 sh_name;
|
||||
u32 sh_type;
|
||||
u64 sh_flags;
|
||||
u64 sh_addr;
|
||||
u64 sh_offset;
|
||||
u64 sh_size;
|
||||
u32 sh_link;
|
||||
u32 sh_info;
|
||||
u64 sh_addralign;
|
||||
u64 sh_entsize;
|
||||
};
|
||||
|
||||
struct Elf32Shdr {
|
||||
u32 sh_name;
|
||||
u32 sh_type;
|
||||
u32 sh_flags;
|
||||
u32 sh_addr;
|
||||
u32 sh_offset;
|
||||
u32 sh_size;
|
||||
u32 sh_link;
|
||||
u32 sh_info;
|
||||
u32 sh_addralign;
|
||||
u32 sh_entsize;
|
||||
};
|
||||
|
||||
struct Elf64Ehdr {
|
||||
u8 e_ident[16];
|
||||
u16 e_type;
|
||||
u16 e_machine;
|
||||
u32 e_version;
|
||||
u64 e_entry;
|
||||
u64 e_phoff;
|
||||
u64 e_shoff;
|
||||
u32 e_flags;
|
||||
u16 e_ehsize;
|
||||
u16 e_phentsize;
|
||||
u16 e_phnum;
|
||||
u16 e_shentsize;
|
||||
u16 e_shnum;
|
||||
u16 e_shstrndx;
|
||||
};
|
||||
|
||||
struct Elf32Ehdr {
|
||||
u8 e_ident[16];
|
||||
u16 e_type;
|
||||
u16 e_machine;
|
||||
u32 e_version;
|
||||
u32 e_entry;
|
||||
u32 e_phoff;
|
||||
u32 e_shoff;
|
||||
u32 e_flags;
|
||||
u16 e_ehsize;
|
||||
u16 e_phentsize;
|
||||
u16 e_phnum;
|
||||
u16 e_shentsize;
|
||||
u16 e_shnum;
|
||||
u16 e_shstrndx;
|
||||
};
|
||||
|
||||
struct Elf64Phdr {
|
||||
u32 p_type;
|
||||
u32 p_flags;
|
||||
u64 p_offset;
|
||||
u64 p_vaddr;
|
||||
u64 p_paddr;
|
||||
u64 p_filesz;
|
||||
u64 p_memsz;
|
||||
u64 p_align;
|
||||
};
|
||||
|
||||
struct Elf32Phdr {
|
||||
u32 p_type;
|
||||
u32 p_offset;
|
||||
u32 p_vaddr;
|
||||
u32 p_paddr;
|
||||
u32 p_filesz;
|
||||
u32 p_memsz;
|
||||
u32 p_flags;
|
||||
u32 p_align;
|
||||
};
|
||||
|
||||
struct Elf64Rel {
|
||||
u64 r_offset;
|
||||
u32 r_type;
|
||||
u32 r_sym;
|
||||
};
|
||||
|
||||
struct Elf32Rel {
|
||||
u32 r_offset;
|
||||
u32 r_type : 8;
|
||||
u32 r_sym : 24;
|
||||
};
|
||||
|
||||
struct Elf64Rela {
|
||||
u64 r_offset;
|
||||
u32 r_type;
|
||||
u32 r_sym;
|
||||
i64 r_addend;
|
||||
};
|
||||
|
||||
struct Elf32Rela {
|
||||
u32 r_offset;
|
||||
u32 r_type : 8;
|
||||
u32 r_sym : 24;
|
||||
i32 r_addend;
|
||||
};
|
||||
|
||||
struct Elf64Dyn {
|
||||
u64 d_tag;
|
||||
u64 d_val;
|
||||
};
|
||||
|
||||
struct Elf32Dyn {
|
||||
u32 d_tag;
|
||||
u32 d_val;
|
||||
};
|
||||
|
||||
struct Elf64Verneed {
|
||||
u16 vn_version;
|
||||
u16 vn_cnt;
|
||||
u32 vn_file;
|
||||
u32 vn_aux;
|
||||
u32 vn_next;
|
||||
};
|
||||
|
||||
struct Elf64Vernaux {
|
||||
u32 vna_hash;
|
||||
u16 vna_flags;
|
||||
u16 vna_other;
|
||||
u32 vna_name;
|
||||
u32 vna_next;
|
||||
};
|
||||
|
||||
struct Elf64Verdef {
|
||||
u16 vd_version;
|
||||
u16 vd_flags;
|
||||
u16 vd_ndx;
|
||||
u16 vd_cnt;
|
||||
u32 vd_hash;
|
||||
u32 vd_aux;
|
||||
u32 vd_next;
|
||||
};
|
||||
|
||||
struct Elf64Verdaux {
|
||||
u32 vda_name;
|
||||
u32 vda_next;
|
||||
};
|
||||
|
||||
struct Elf64Chdr {
|
||||
u32 ch_type;
|
||||
u32 ch_reserved;
|
||||
u64 ch_size;
|
||||
u64 ch_addralign;
|
||||
};
|
||||
|
||||
struct Elf32Chdr {
|
||||
u32 ch_type;
|
||||
u32 ch_size;
|
||||
u32 ch_addralign;
|
||||
};
|
||||
|
||||
struct Elf64Nhdr {
|
||||
u32 n_namesz;
|
||||
u32 n_descsz;
|
||||
u32 n_type;
|
||||
};
|
||||
|
||||
struct X86_64 {
|
||||
typedef u64 WordTy;
|
||||
|
||||
static constexpr u32 R_NONE = R_X86_64_NONE;
|
||||
static constexpr u32 R_COPY = R_X86_64_COPY;
|
||||
static constexpr u32 R_GLOB_DAT = R_X86_64_GLOB_DAT;
|
||||
static constexpr u32 R_JUMP_SLOT = R_X86_64_JUMP_SLOT;
|
||||
static constexpr u32 R_RELATIVE = R_X86_64_RELATIVE;
|
||||
static constexpr u32 R_IRELATIVE = R_X86_64_IRELATIVE;
|
||||
static constexpr u32 R_DTPOFF = R_X86_64_DTPOFF64;
|
||||
static constexpr u32 R_TPOFF = R_X86_64_TPOFF64;
|
||||
static constexpr u32 R_DTPMOD = R_X86_64_DTPMOD64;
|
||||
static constexpr u32 R_TLSDESC = R_X86_64_TLSDESC;
|
||||
|
||||
static constexpr u32 wordsize = 8;
|
||||
static constexpr u32 e_machine = EM_X86_64;
|
||||
static constexpr u32 plt_hdr_size = 16;
|
||||
static constexpr u32 plt_size = 16;
|
||||
static constexpr u32 pltgot_size = 8;
|
||||
static constexpr bool is_rel = false;
|
||||
static constexpr bool is_le = true;
|
||||
};
|
||||
|
||||
template <> struct ElfSym<X86_64> : public Elf64Sym {};
|
||||
template <> struct ElfShdr<X86_64> : public Elf64Shdr {};
|
||||
template <> struct ElfEhdr<X86_64> : public Elf64Ehdr {};
|
||||
template <> struct ElfPhdr<X86_64> : public Elf64Phdr {};
|
||||
template <> struct ElfRel<X86_64> : public Elf64Rela {};
|
||||
template <> struct ElfDyn<X86_64> : public Elf64Dyn {};
|
||||
template <> struct ElfVerneed<X86_64> : public Elf64Verneed {};
|
||||
template <> struct ElfVernaux<X86_64> : public Elf64Vernaux {};
|
||||
template <> struct ElfVerdef<X86_64> : public Elf64Verdef {};
|
||||
template <> struct ElfVerdaux<X86_64> : public Elf64Verdaux {};
|
||||
template <> struct ElfChdr<X86_64> : public Elf64Chdr {};
|
||||
template <> struct ElfNhdr<X86_64> : public Elf64Nhdr {};
|
||||
|
||||
struct I386 {
|
||||
typedef u32 WordTy;
|
||||
|
||||
static constexpr u32 R_NONE = R_386_NONE;
|
||||
static constexpr u32 R_COPY = R_386_COPY;
|
||||
static constexpr u32 R_GLOB_DAT = R_386_GLOB_DAT;
|
||||
static constexpr u32 R_JUMP_SLOT = R_386_JUMP_SLOT;
|
||||
static constexpr u32 R_RELATIVE = R_386_RELATIVE;
|
||||
static constexpr u32 R_IRELATIVE = R_386_IRELATIVE;
|
||||
static constexpr u32 R_DTPOFF = R_386_TLS_DTPOFF32;
|
||||
static constexpr u32 R_TPOFF = R_386_TLS_TPOFF;
|
||||
static constexpr u32 R_DTPMOD = R_386_TLS_DTPMOD32;
|
||||
static constexpr u32 R_TLSDESC = R_386_TLS_DESC;
|
||||
|
||||
static constexpr u32 wordsize = 4;
|
||||
static constexpr u32 e_machine = EM_386;
|
||||
static constexpr u32 plt_hdr_size = 16;
|
||||
static constexpr u32 plt_size = 16;
|
||||
static constexpr u32 pltgot_size = 8;
|
||||
static constexpr bool is_rel = true;
|
||||
static constexpr bool is_le = true;
|
||||
};
|
||||
|
||||
template <> struct ElfSym<I386> : public Elf32Sym {};
|
||||
template <> struct ElfShdr<I386> : public Elf32Shdr {};
|
||||
template <> struct ElfEhdr<I386> : public Elf32Ehdr {};
|
||||
template <> struct ElfPhdr<I386> : public Elf32Phdr {};
|
||||
template <> struct ElfRel<I386> : public Elf32Rel {};
|
||||
template <> struct ElfDyn<I386> : public Elf32Dyn {};
|
||||
template <> struct ElfVerneed<I386> : public Elf64Verneed {};
|
||||
template <> struct ElfVernaux<I386> : public Elf64Vernaux {};
|
||||
template <> struct ElfVerdef<I386> : public Elf64Verdef {};
|
||||
template <> struct ElfVerdaux<I386> : public Elf64Verdaux {};
|
||||
template <> struct ElfChdr<I386> : public Elf32Chdr {};
|
||||
template <> struct ElfNhdr<I386> : public Elf64Nhdr {};
|
||||
|
||||
struct AARCH64 {
|
||||
typedef u64 WordTy;
|
||||
|
||||
static constexpr u32 R_NONE = R_AARCH64_NONE;
|
||||
static constexpr u32 R_COPY = R_AARCH64_COPY;
|
||||
static constexpr u32 R_GLOB_DAT = R_AARCH64_GLOB_DAT;
|
||||
static constexpr u32 R_JUMP_SLOT = R_AARCH64_JUMP_SLOT;
|
||||
static constexpr u32 R_RELATIVE = R_AARCH64_RELATIVE;
|
||||
static constexpr u32 R_IRELATIVE = R_AARCH64_IRELATIVE;
|
||||
static constexpr u32 R_DTPOFF = R_AARCH64_TLS_DTPREL64;
|
||||
static constexpr u32 R_TPOFF = R_AARCH64_TLS_TPREL64;
|
||||
static constexpr u32 R_DTPMOD = R_AARCH64_TLS_DTPMOD64;
|
||||
static constexpr u32 R_TLSDESC = R_AARCH64_TLSDESC;
|
||||
|
||||
static constexpr u32 wordsize = 8;
|
||||
static constexpr u32 e_machine = EM_AARCH64;
|
||||
static constexpr u32 plt_hdr_size = 32;
|
||||
static constexpr u32 plt_size = 16;
|
||||
static constexpr u32 pltgot_size = 16;
|
||||
static constexpr bool is_rel = false;
|
||||
static constexpr bool is_le = true;
|
||||
};
|
||||
|
||||
template <> struct ElfSym<AARCH64> : public Elf64Sym {};
|
||||
template <> struct ElfShdr<AARCH64> : public Elf64Shdr {};
|
||||
template <> struct ElfEhdr<AARCH64> : public Elf64Ehdr {};
|
||||
template <> struct ElfPhdr<AARCH64> : public Elf64Phdr {};
|
||||
template <> struct ElfRel<AARCH64> : public Elf64Rela {};
|
||||
template <> struct ElfDyn<AARCH64> : public Elf64Dyn {};
|
||||
template <> struct ElfVerneed<AARCH64> : public Elf64Verneed {};
|
||||
template <> struct ElfVernaux<AARCH64> : public Elf64Vernaux {};
|
||||
template <> struct ElfVerdef<AARCH64> : public Elf64Verdef {};
|
||||
template <> struct ElfVerdaux<AARCH64> : public Elf64Verdaux {};
|
||||
template <> struct ElfChdr<AARCH64> : public Elf64Chdr {};
|
||||
template <> struct ElfNhdr<AARCH64> : public Elf64Nhdr {};
|
||||
|
||||
} // namespace mold::elf
|
||||
+109
@@ -0,0 +1,109 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
std::string get_current_dir() {
|
||||
std::string buf(8192, '\0');
|
||||
getcwd(buf.data(), buf.size());
|
||||
buf.resize(buf.find_first_of('\0'));
|
||||
return buf;
|
||||
}
|
||||
|
||||
// Returns the directory part of a given path.
|
||||
// Returns '.' if path doesn't contain '/'.
|
||||
std::string_view path_dirname(std::string_view path) {
|
||||
i64 pos = path.find_last_of('/');
|
||||
if (pos == path.npos)
|
||||
return ".";
|
||||
return path.substr(0, pos);
|
||||
}
|
||||
|
||||
// Returns the filename part of a given path.
|
||||
// Returns '/' if path reperesents the root directory.
|
||||
// Returns '.' if path is empty.
|
||||
std::string_view path_filename(std::string_view path) {
|
||||
if (path.empty())
|
||||
return ".";
|
||||
|
||||
while (path.ends_with('/'))
|
||||
path = path.substr(0, path.size() - 2);
|
||||
|
||||
if (path.empty())
|
||||
return "/";
|
||||
|
||||
i64 pos = path.find_last_of('/');
|
||||
if (pos == path.npos)
|
||||
return path;
|
||||
return path.substr(pos + 1);
|
||||
}
|
||||
|
||||
// Returns the filename part of a given path without the file
|
||||
// extension.
|
||||
std::string_view path_basename(std::string_view path) {
|
||||
std::string_view name = path_filename(path);
|
||||
|
||||
i64 pos = name.find_last_of('.');
|
||||
if (pos == name.npos)
|
||||
return name;
|
||||
return name.substr(0, pos);
|
||||
}
|
||||
|
||||
std::string path_to_absolute(std::string_view path) {
|
||||
if (path.starts_with('/'))
|
||||
return std::string(path);
|
||||
return get_current_dir() + "/" + std::string(path);
|
||||
}
|
||||
|
||||
// Removes redundant '/..' or '/.' from a given path.
|
||||
// The transformation is done purely by lexical processing.
|
||||
// This function does not access file system.
|
||||
std::string path_clean(std::string_view path) {
|
||||
std::vector<std::string_view> components;
|
||||
bool is_rooted = path.starts_with('/');
|
||||
|
||||
while (!path.empty()) {
|
||||
size_t pos = path.find('/');
|
||||
std::string_view elem;
|
||||
|
||||
if (pos == path.npos) {
|
||||
elem = path;
|
||||
path = "";
|
||||
} else {
|
||||
elem = path.substr(0, pos);
|
||||
path = path.substr(pos + 1);
|
||||
}
|
||||
|
||||
if (elem.empty() || elem == ".")
|
||||
continue;
|
||||
|
||||
if (elem == "..") {
|
||||
if (components.empty()) {
|
||||
if (is_rooted)
|
||||
continue;
|
||||
components.push_back("..");
|
||||
continue;
|
||||
}
|
||||
|
||||
if (components.back() == "..")
|
||||
components.push_back("..");
|
||||
else
|
||||
components.pop_back();
|
||||
continue;
|
||||
}
|
||||
|
||||
components.push_back(elem);
|
||||
}
|
||||
|
||||
std::string ret;
|
||||
if (is_rooted)
|
||||
ret = "/";
|
||||
|
||||
for (i64 i = 0, end = components.size(); i < end; i++) {
|
||||
ret += components[i];
|
||||
if (i != end - 1)
|
||||
ret += "/";
|
||||
}
|
||||
return ret;
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,198 @@
|
||||
// This file implements a mark-sweep garbage collector for -gc-sections.
|
||||
// In this algorithm, vertices are sections and edges are relocations.
|
||||
// Any section that is reachable from a root section is considered alive.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <tbb/concurrent_vector.h>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
static bool is_init_fini(const InputSection<E> &isec) {
|
||||
return isec.shdr.sh_type == SHT_INIT_ARRAY ||
|
||||
isec.shdr.sh_type == SHT_FINI_ARRAY ||
|
||||
isec.shdr.sh_type == SHT_PREINIT_ARRAY ||
|
||||
isec.name().starts_with(".ctors") ||
|
||||
isec.name().starts_with(".dtors") ||
|
||||
isec.name().starts_with(".init") ||
|
||||
isec.name().starts_with(".fini");
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static bool mark_section(InputSection<E> *isec) {
|
||||
return isec && isec->is_alive && !isec->is_visited.exchange(true);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void visit(Context<E> &ctx, InputSection<E> *isec,
|
||||
tbb::feeder<InputSection<E> *> &feeder, i64 depth) {
|
||||
ASSERT(isec->is_visited);
|
||||
|
||||
// A relocation can refer either a section fragment (i.e. a piece of
|
||||
// string in a mergeable string section) or a symbol. Mark all
|
||||
// section fragments as alive.
|
||||
if (SectionFragmentRef<E> *refs = isec->rel_fragments.get())
|
||||
for (i64 i = 0; refs[i].idx >= 0; i++)
|
||||
refs[i].frag->is_alive.store(true, std::memory_order_relaxed);
|
||||
|
||||
// If this is a text section, .eh_frame may contain records
|
||||
// describing how to handle exceptions for that function.
|
||||
// We want to keep associated .eh_frame records.
|
||||
for (FdeRecord<E> &fde : isec->get_fdes())
|
||||
for (ElfRel<E> &rel : fde.get_rels().subspan(1))
|
||||
if (Symbol<E> *sym = isec->file.symbols[rel.r_sym])
|
||||
if (mark_section(sym->input_section))
|
||||
feeder.add(sym->input_section);
|
||||
|
||||
for (ElfRel<E> &rel : isec->get_rels(ctx)) {
|
||||
Symbol<E> &sym = *isec->file.symbols[rel.r_sym];
|
||||
|
||||
// Symbol can refer either a section fragment or an input section.
|
||||
// Mark a fragment as alive.
|
||||
if (SectionFragment<E> *frag = sym.get_frag()) {
|
||||
frag->is_alive.store(true, std::memory_order_relaxed);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!mark_section(sym.input_section))
|
||||
continue;
|
||||
|
||||
// Mark a section alive. For better performacne, we don't call
|
||||
// `feeder.add` too often.
|
||||
if (depth < 3)
|
||||
visit(ctx, sym.input_section, feeder, depth + 1);
|
||||
else
|
||||
feeder.add(sym.input_section);
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static tbb::concurrent_vector<InputSection<E> *>
|
||||
collect_root_set(Context<E> &ctx) {
|
||||
Timer t(ctx, "collect_root_set");
|
||||
tbb::concurrent_vector<InputSection<E> *> roots;
|
||||
|
||||
auto enqueue_section = [&](InputSection<E> *isec) {
|
||||
if (mark_section(isec))
|
||||
roots.push_back(isec);
|
||||
};
|
||||
|
||||
auto enqueue_symbol = [&](Symbol<E> *sym) {
|
||||
if (sym) {
|
||||
if (SectionFragment<E> *frag = sym->get_frag())
|
||||
frag->is_alive.store(true, std::memory_order_relaxed);
|
||||
else
|
||||
enqueue_section(sym->input_section);
|
||||
}
|
||||
};
|
||||
|
||||
// Add sections that are not subject to garbage collection.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (!isec || !isec->is_alive)
|
||||
continue;
|
||||
|
||||
// -gc-sections discards only SHF_ALLOC sections. If you want to
|
||||
// reduce the amount of non-memory-mapped segments, you should
|
||||
// use `strip` command, compile without debug info or use
|
||||
// -strip-all linker option.
|
||||
if (!(isec->shdr.sh_flags & SHF_ALLOC))
|
||||
isec->is_visited = true;
|
||||
|
||||
if (is_init_fini(*isec) || is_c_identifier(isec->name()) ||
|
||||
isec->shdr.sh_type == SHT_NOTE)
|
||||
enqueue_section(isec.get());
|
||||
}
|
||||
});
|
||||
|
||||
// Add sections containing exported symbols
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->symbols)
|
||||
if (sym->file == file && sym->is_exported)
|
||||
enqueue_symbol(sym);
|
||||
});
|
||||
|
||||
// Add sections referenced by root symbols.
|
||||
enqueue_symbol(Symbol<E>::intern(ctx, ctx.arg.entry));
|
||||
|
||||
for (std::string_view name : ctx.arg.undefined)
|
||||
enqueue_symbol(Symbol<E>::intern(ctx, name));
|
||||
|
||||
// .eh_frame consists of variable-length records called CIE and FDE
|
||||
// records, and they are a unit of inclusion or exclusion.
|
||||
// We just keep all CIEs and everything that are referenced by them.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (CieRecord<E> &cie : file->cies)
|
||||
for (ElfRel<E> &rel : cie.get_rels())
|
||||
enqueue_symbol(file->symbols[rel.r_sym]);
|
||||
});
|
||||
|
||||
return roots;
|
||||
}
|
||||
|
||||
// Mark all reachable sections
|
||||
template <typename E>
|
||||
static void mark(Context<E> &ctx,
|
||||
tbb::concurrent_vector<InputSection<E> *> &roots) {
|
||||
Timer t(ctx, "mark");
|
||||
|
||||
tbb::parallel_for_each(roots, [&](InputSection<E> *isec,
|
||||
tbb::feeder<InputSection<E> *> &feeder) {
|
||||
visit(ctx, isec, feeder, 0);
|
||||
});
|
||||
}
|
||||
|
||||
// Remove unreachable sections
|
||||
template <typename E>
|
||||
static void sweep(Context<E> &ctx) {
|
||||
Timer t(ctx, "sweep");
|
||||
static Counter counter("garbage_sections");
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (i64 i = 0; i < file->sections.size(); i++) {
|
||||
std::unique_ptr<InputSection<E>> &isec = file->sections[i];
|
||||
|
||||
if (isec && isec->is_alive && !isec->is_visited) {
|
||||
if (ctx.arg.print_gc_sections)
|
||||
SyncOut(ctx) << "removing unused section " << *isec;
|
||||
isec->kill();
|
||||
counter++;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// Non-alloc section fragments are not subject of garbage collection.
|
||||
// This function marks such fragments.
|
||||
template <typename E>
|
||||
static void mark_nonalloc_fragments(Context<E> &ctx) {
|
||||
Timer t(ctx, "mark_nonalloc_fragments");
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
for (SectionFragment<E> *frag : file->fragments)
|
||||
if (!(frag->output_section.shdr.sh_flags & SHF_ALLOC))
|
||||
frag->is_alive.store(true, std::memory_order_relaxed);
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void gc_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "gc");
|
||||
|
||||
mark_nonalloc_fragments(ctx);
|
||||
|
||||
tbb::concurrent_vector<InputSection<E> *> roots = collect_root_set(ctx);
|
||||
mark(ctx, roots);
|
||||
sweep(ctx);
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void gc_sections(Context<E> &ctx);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -4,17 +4,22 @@
|
||||
// For more info, read
|
||||
// https://engineering.fb.com/2018/12/13/data-infrastructure/hyperloglog
|
||||
|
||||
#include "lib.h"
|
||||
#include "mold.h"
|
||||
|
||||
#include <cmath>
|
||||
|
||||
namespace mold {
|
||||
namespace mold::elf {
|
||||
|
||||
i64 HyperLogLog::get_cardinality() const {
|
||||
double z = 0;
|
||||
for (i64 val : buckets)
|
||||
z += std::ldexp(1.0, -val);
|
||||
z += pow(2, -val);
|
||||
return ALPHA * NBUCKETS * NBUCKETS / z;
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
void HyperLogLog::merge(const HyperLogLog &other) {
|
||||
for (i64 i = 0; i < NBUCKETS; i++)
|
||||
merge_one(i, other.buckets[i]);
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
+160
-216
@@ -40,36 +40,21 @@
|
||||
// regular trees are equal. Given this formulation, we want to find as
|
||||
// many identical vertices as possible.
|
||||
//
|
||||
// Just like a lot of problems with graph, this problem doesn't have a
|
||||
// straightforward "optimal" solution, and we need to resort to heuristics.
|
||||
// Solving such problem is computationally intensive, but mold is quite fast.
|
||||
// For Chromium, mold's ICF finishes in less than 1 second with 20 threads.
|
||||
// This is contrary to lld and gold, which take about 5 and 50 seconds to
|
||||
// run ICF under the same condition, respectively.
|
||||
//
|
||||
// mold approaches this problem by hashing program trees with increasing depth
|
||||
// on each iteration.
|
||||
// For example, when we start, we only hash individual functions with
|
||||
// their call into other functions omitted. From the second iteration, we
|
||||
// put the function they call into the hash by appending the hash of those
|
||||
// functions from the previous iteration. This means that the nth iteration
|
||||
// hashes call chain up to (n-1) levels deep.
|
||||
// We use a cryptographic hash function, so the unique number of hashes will
|
||||
// only monotonically increase as we take into account of deeper trees with
|
||||
// iterations (otherwise, that means we have found a hash collision). We stop
|
||||
// when the unique number of hashes stop increasing; this is based on the fact
|
||||
// that once we observe an iteration with the same amount of unique hashes as
|
||||
// the previous iteration, it will remain unchanged for further iterations.
|
||||
// This is provable, but here we omit the proof for brevity.
|
||||
//
|
||||
// When compared to other approaches, mold's approach has a relatively cheaper
|
||||
// cost per iteration, and as a bonus, is highly parallelizable.
|
||||
// For Chromium, mold's ICF finishes in less than 1 second with 20 threads,
|
||||
// whereas lld takes 5 seconds and gold takes 50 seconds under the same
|
||||
// conditions.
|
||||
// mold's ICF is faster because we are using a better algorithm.
|
||||
// It's actually me who developed and implemented the lld's ICF algorithm,
|
||||
// and I can say that mold's algorithm is better than that in all aspects.
|
||||
// It scales better for number of available cores, require less overall
|
||||
// computation, and has a smaller working set. So, it's better with a single
|
||||
// thread and even better with multiple threads.
|
||||
|
||||
#include "mold.h"
|
||||
#include "../lib/siphash.h"
|
||||
|
||||
#include <array>
|
||||
#include <cstdio>
|
||||
#include <fstream>
|
||||
#include <tbb/concurrent_unordered_map.h>
|
||||
#include <tbb/concurrent_vector.h>
|
||||
#include <tbb/enumerable_thread_specific.h>
|
||||
@@ -77,72 +62,75 @@
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <tbb/parallel_sort.h>
|
||||
|
||||
#ifdef __APPLE__
|
||||
# define COMMON_DIGEST_FOR_OPENSSL
|
||||
# include <CommonCrypto/CommonDigest.h>
|
||||
#else
|
||||
# include <openssl/sha.h>
|
||||
#endif
|
||||
|
||||
static constexpr int64_t HASH_SIZE = 16;
|
||||
|
||||
using Digest = std::array<uint8_t, HASH_SIZE>;
|
||||
typedef std::array<uint8_t, HASH_SIZE> Digest;
|
||||
|
||||
namespace std {
|
||||
template <> struct hash<Digest> {
|
||||
template<> struct hash<Digest> {
|
||||
size_t operator()(const Digest &k) const {
|
||||
static_assert(sizeof(size_t) <= HASH_SIZE);
|
||||
size_t val;
|
||||
memcpy(&val, k.data(), sizeof(size_t));
|
||||
return val;
|
||||
return *(int64_t *)&k[0];
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
namespace mold {
|
||||
|
||||
static u8 hmac_key[16];
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
static void uniquify_cies(Context<E> &ctx) {
|
||||
Timer t(ctx, "uniquify_cies");
|
||||
std::vector<CieRecord<E> *> cies;
|
||||
|
||||
auto find = [&](CieRecord<E> &cie) -> i64 {
|
||||
for (i64 i = 0; i < cies.size(); i++)
|
||||
if (cie_equals(cie, *cies[i]))
|
||||
return i;
|
||||
return -1;
|
||||
};
|
||||
|
||||
for (ObjectFile<E> *file : ctx.objs) {
|
||||
for (CieRecord<E> &cie : file->cies) {
|
||||
if (i64 idx = find(cie); idx != -1) {
|
||||
cie.icf_idx = idx;
|
||||
} else {
|
||||
cie.icf_idx = cies.size();
|
||||
cies.push_back(&cie);
|
||||
for (i64 i = 0; i < cies.size(); i++) {
|
||||
if (cie.equals(*cies[i])) {
|
||||
cie.icf_idx = i;
|
||||
goto found;
|
||||
}
|
||||
}
|
||||
cie.icf_idx = cies.size();
|
||||
cies.push_back(&cie);
|
||||
found:;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static bool is_eligible(Context<E> &ctx, InputSection<E> &isec) {
|
||||
const ElfShdr<E> &shdr = isec.shdr();
|
||||
static bool is_eligible(InputSection<E> &isec) {
|
||||
const ElfShdr<E> &shdr = isec.shdr;
|
||||
std::string_view name = isec.name();
|
||||
|
||||
if (shdr.sh_size == 0 || !(shdr.sh_flags & SHF_ALLOC) ||
|
||||
shdr.sh_type == SHT_NOBITS || is_c_identifier(isec.name))
|
||||
return false;
|
||||
bool is_alloc = (shdr.sh_flags & SHF_ALLOC);
|
||||
bool is_executable = (shdr.sh_flags & SHF_EXECINSTR);
|
||||
bool is_relro = (name == ".data.rel.ro" ||
|
||||
name.starts_with(".data.rel.ro."));
|
||||
bool is_readonly = !(shdr.sh_flags & SHF_WRITE) || is_relro;
|
||||
bool is_bss = (shdr.sh_type == SHT_NOBITS);
|
||||
bool is_empty = (shdr.sh_size == 0);
|
||||
bool is_init = (shdr.sh_type == SHT_INIT_ARRAY || name == ".init");
|
||||
bool is_fini = (shdr.sh_type == SHT_FINI_ARRAY || name == ".fini");
|
||||
bool is_enumerable = is_c_identifier(name);
|
||||
|
||||
if (shdr.sh_flags & SHF_EXECINSTR)
|
||||
return (ctx.arg.icf_all || !isec.address_taken) &&
|
||||
isec.name != ".init" && isec.name != ".fini";
|
||||
return is_alloc && is_executable && is_readonly && !is_bss &&
|
||||
!is_empty && !is_init && !is_fini && !is_enumerable;
|
||||
}
|
||||
|
||||
// .gcc_except_table contains a compiler-generated table. Pointer
|
||||
// equality for the section is not significant because only the C++
|
||||
// exception handling code will use the table at runtime.
|
||||
if (isec.name == ".gcc_except_table" ||
|
||||
isec.name.starts_with(".gcc_except_table."))
|
||||
return true;
|
||||
static Digest digest_final(SHA256_CTX &sha) {
|
||||
u8 buf[SHA256_SIZE];
|
||||
int res = SHA256_Final(buf, &sha);
|
||||
ASSERT(res == 1);
|
||||
|
||||
bool is_readonly = !(shdr.sh_flags & SHF_WRITE);
|
||||
bool is_relro = isec.name.starts_with(".data.rel.ro");
|
||||
return (ctx.arg.ignore_data_address_equality || !isec.address_taken) &&
|
||||
(is_readonly || is_relro);
|
||||
Digest digest;
|
||||
memcpy(digest.data(), buf, HASH_SIZE);
|
||||
return digest;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
@@ -151,18 +139,22 @@ static bool is_leaf(Context<E> &ctx, InputSection<E> &isec) {
|
||||
return false;
|
||||
|
||||
for (FdeRecord<E> &fde : isec.get_fdes())
|
||||
if (fde.get_rels(isec.file).size() > 1)
|
||||
if (fde.get_rels().size() > 1)
|
||||
return false;
|
||||
|
||||
return true;
|
||||
}
|
||||
|
||||
static u64 combine_hash(u64 a, u64 b) {
|
||||
return a ^ (b + 0x9e3779b9 + (a << 6) + (a >> 2));
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
struct LeafHasher {
|
||||
size_t operator()(InputSection<E> *isec) const {
|
||||
size_t operator()(const InputSection<E> *isec) const {
|
||||
u64 h = hash_string(isec->contents);
|
||||
for (FdeRecord<E> &fde : isec->get_fdes()) {
|
||||
u64 h2 = hash_string(fde.get_contents(isec->file).substr(8));
|
||||
u64 h2 = hash_string(fde.get_contents().substr(8));
|
||||
h = combine_hash(h, h2);
|
||||
}
|
||||
return h;
|
||||
@@ -171,7 +163,7 @@ struct LeafHasher {
|
||||
|
||||
template <typename E>
|
||||
struct LeafEq {
|
||||
bool operator()(InputSection<E> *a, InputSection<E> *b) const {
|
||||
bool operator()(const InputSection<E> *a, const InputSection<E> *b) const {
|
||||
if (a->contents != b->contents)
|
||||
return false;
|
||||
|
||||
@@ -182,15 +174,12 @@ struct LeafEq {
|
||||
return false;
|
||||
|
||||
for (i64 i = 0; i < x.size(); i++)
|
||||
if (x[i].get_contents(a->file).substr(8) !=
|
||||
y[i].get_contents(b->file).substr(8))
|
||||
if (x[i].get_contents().substr(8) != y[i].get_contents().substr(8))
|
||||
return false;
|
||||
return true;
|
||||
}
|
||||
};
|
||||
|
||||
// Early merge of leaf nodes, which can be processed without constructing the
|
||||
// entire graph. This reduces the vertex count and improves memory efficiency.
|
||||
template <typename E>
|
||||
static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
Timer t(ctx, "merge_leaf_nodes");
|
||||
@@ -199,7 +188,7 @@ static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
static Counter non_eligible("icf_non_eligibles");
|
||||
static Counter leaf("icf_leaf_nodes");
|
||||
|
||||
tbb::concurrent_unordered_map<InputSection<E> *, Atomic<InputSection<E> *>,
|
||||
tbb::concurrent_unordered_map<InputSection<E> *, InputSection<E> *,
|
||||
LeafHasher<E>, LeafEq<E>> map;
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)ctx.objs.size(), [&](i64 i) {
|
||||
@@ -207,7 +196,7 @@ static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
if (!isec || !isec->is_alive)
|
||||
continue;
|
||||
|
||||
if (!is_eligible(ctx, *isec)) {
|
||||
if (!is_eligible(*isec)) {
|
||||
non_eligible++;
|
||||
continue;
|
||||
}
|
||||
@@ -216,11 +205,8 @@ static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
leaf++;
|
||||
isec->icf_leaf = true;
|
||||
auto [it, inserted] = map.insert({isec.get(), isec.get()});
|
||||
if (!inserted) {
|
||||
InputSection<E> *isec2 = it->second.load();
|
||||
while (isec->get_priority() < isec2->get_priority() &&
|
||||
!it->second.compare_exchange_strong(isec2, isec.get()));
|
||||
}
|
||||
if (!inserted && isec->get_priority() < it->second->get_priority())
|
||||
it->second = isec.get();
|
||||
} else {
|
||||
eligible++;
|
||||
isec->icf_eligible = true;
|
||||
@@ -232,7 +218,7 @@ static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : ctx.objs[i]->sections) {
|
||||
if (isec && isec->is_alive && isec->icf_leaf) {
|
||||
auto it = map.find(isec.get());
|
||||
assert(it != map.end());
|
||||
ASSERT(it != map.end());
|
||||
isec->leader = it->second;
|
||||
}
|
||||
}
|
||||
@@ -241,26 +227,29 @@ static void merge_leaf_nodes(Context<E> &ctx) {
|
||||
|
||||
template <typename E>
|
||||
static Digest compute_digest(Context<E> &ctx, InputSection<E> &isec) {
|
||||
SipHash13_128 hasher(hmac_key);
|
||||
SHA256_CTX sha;
|
||||
SHA256_Init(&sha);
|
||||
|
||||
u8 *buf = (u8 *)isec.contents.data();
|
||||
|
||||
auto hash = [&](auto val) {
|
||||
hasher.update((u8 *)&val, sizeof(val));
|
||||
SHA256_Update(&sha, &val, sizeof(val));
|
||||
};
|
||||
|
||||
auto hash_string = [&](std::string_view str) {
|
||||
hash(str.size());
|
||||
hasher.update((u8 *)str.data(), str.size());
|
||||
SHA256_Update(&sha, str.data(), str.size());
|
||||
};
|
||||
|
||||
auto hash_symbol = [&](Symbol<E> &sym) {
|
||||
InputSection<E> *isec = sym.get_input_section();
|
||||
InputSection<E> *isec = sym.input_section;
|
||||
|
||||
if (!sym.file || sym.is_imported) {
|
||||
if (!sym.file) {
|
||||
hash('1');
|
||||
hash((u64)&sym);
|
||||
} else if (SectionFragment<E> *frag = sym.get_frag()) {
|
||||
hash('2');
|
||||
hash((u64)frag);
|
||||
hash_string(frag->data);
|
||||
} else if (!isec) {
|
||||
hash('3');
|
||||
} else if (isec->leader) {
|
||||
@@ -276,38 +265,46 @@ static Digest compute_digest(Context<E> &ctx, InputSection<E> &isec) {
|
||||
};
|
||||
|
||||
hash_string(isec.contents);
|
||||
hash(isec.shdr().sh_flags);
|
||||
hash(isec.shdr.sh_flags);
|
||||
hash(isec.get_fdes().size());
|
||||
hash(isec.get_rels(ctx).size());
|
||||
|
||||
for (FdeRecord<E> &fde : isec.get_fdes()) {
|
||||
hash(isec.file.cies[fde.cie_idx].icf_idx);
|
||||
hash(fde.cie->icf_idx);
|
||||
|
||||
// Bytes 0 to 4 contain the length of this record, and
|
||||
// bytes 4 to 8 contain an offset to CIE.
|
||||
hash_string(fde.get_contents(isec.file).substr(8));
|
||||
hash_string(fde.get_contents().substr(8));
|
||||
|
||||
hash(fde.get_rels(isec.file).size());
|
||||
hash(fde.get_rels().size());
|
||||
|
||||
for (const ElfRel<E> &rel : fde.get_rels(isec.file).subspan(1)) {
|
||||
for (ElfRel<E> &rel : fde.get_rels().subspan(1)) {
|
||||
hash_symbol(*isec.file.symbols[rel.r_sym]);
|
||||
hash(rel.r_type);
|
||||
hash(rel.r_offset - fde.input_offset);
|
||||
hash(get_addend(isec.file.cies[fde.cie_idx].input_section, rel));
|
||||
hash(fde.cie->input_section.get_addend(rel));
|
||||
}
|
||||
}
|
||||
|
||||
i64 frag_idx = 0;
|
||||
|
||||
for (i64 i = 0; i < isec.get_rels(ctx).size(); i++) {
|
||||
const ElfRel<E> &rel = isec.get_rels(ctx)[i];
|
||||
ElfRel<E> &rel = isec.get_rels(ctx)[i];
|
||||
hash(rel.r_offset);
|
||||
hash(rel.r_type);
|
||||
hash(get_addend(isec, rel));
|
||||
hash_symbol(*isec.file.symbols[rel.r_sym]);
|
||||
hash(isec.get_addend(rel));
|
||||
|
||||
if (isec.rel_fragments && isec.rel_fragments[frag_idx].idx == i) {
|
||||
SectionFragmentRef<E> &ref = isec.rel_fragments[frag_idx++];
|
||||
hash('a');
|
||||
isec.get_addend(rel);
|
||||
hash_string(ref.frag->data);
|
||||
} else {
|
||||
hash_symbol(*isec.file.symbols[rel.r_sym]);
|
||||
}
|
||||
}
|
||||
|
||||
Digest digest;
|
||||
hasher.finish(digest.data());
|
||||
return digest;
|
||||
return digest_final(sha);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
@@ -357,9 +354,6 @@ compute_digests(Context<E> &ctx, std::span<InputSection<E> *> sections) {
|
||||
return digests;
|
||||
}
|
||||
|
||||
// Build a graph, treating every function as a vertex and every function call
|
||||
// as an edge. See the description at the top for a more detailed formulation.
|
||||
// We use u32 indices here to improve cache locality.
|
||||
template <typename E>
|
||||
static void gather_edges(Context<E> &ctx,
|
||||
std::span<InputSection<E> *> sections,
|
||||
@@ -367,23 +361,24 @@ static void gather_edges(Context<E> &ctx,
|
||||
std::vector<u32> &edge_indices) {
|
||||
Timer t(ctx, "gather_edges");
|
||||
|
||||
if (sections.empty())
|
||||
return;
|
||||
|
||||
std::vector<i64> num_edges(sections.size());
|
||||
edge_indices.resize(sections.size());
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)sections.size(), [&](i64 i) {
|
||||
InputSection<E> &isec = *sections[i];
|
||||
assert(isec.icf_eligible);
|
||||
ASSERT(isec.icf_eligible);
|
||||
i64 frag_idx = 0;
|
||||
|
||||
for (i64 j = 0; j < isec.get_rels(ctx).size(); j++) {
|
||||
const ElfRel<E> &rel = isec.get_rels(ctx)[j];
|
||||
Symbol<E> &sym = *isec.file.symbols[rel.r_sym];
|
||||
if (!sym.get_frag())
|
||||
if (InputSection<E> *isec = sym.get_input_section())
|
||||
if (isec->icf_eligible)
|
||||
num_edges[i]++;
|
||||
if (isec.rel_fragments && isec.rel_fragments[frag_idx].idx == j) {
|
||||
frag_idx++;
|
||||
} else {
|
||||
ElfRel<E> &rel = isec.get_rels(ctx)[j];
|
||||
Symbol<E> &sym = *isec.file.symbols[rel.r_sym];
|
||||
if (!sym.get_frag() && sym.input_section &&
|
||||
sym.input_section->icf_eligible)
|
||||
num_edges[i]++;
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
@@ -394,22 +389,26 @@ static void gather_edges(Context<E> &ctx,
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)num_edges.size(), [&](i64 i) {
|
||||
InputSection<E> &isec = *sections[i];
|
||||
i64 frag_idx = 0;
|
||||
i64 idx = edge_indices[i];
|
||||
|
||||
for (ElfRel<E> &rel : isec.get_rels(ctx)) {
|
||||
Symbol<E> &sym = *isec.file.symbols[rel.r_sym];
|
||||
if (InputSection<E> *isec = sym.get_input_section())
|
||||
if (isec->icf_eligible)
|
||||
edges[idx++] = isec->icf_idx;
|
||||
}
|
||||
for (i64 j = 0; j < isec.get_rels(ctx).size(); j++) {
|
||||
if (isec.rel_fragments && isec.rel_fragments[frag_idx].idx == j) {
|
||||
frag_idx++;
|
||||
ElfRel<E> &rel = isec.get_rels(ctx)[j];
|
||||
Symbol<E> &sym = *isec.file.symbols[rel.r_sym];
|
||||
if (!sym.get_frag() && sym.input_section &&
|
||||
sym.input_section->icf_eligible)
|
||||
edges[idx++] = sym.input_section->icf_idx;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 propagate(std::span<std::vector<Digest>> digests,
|
||||
std::span<u32> edges, std::span<u32> edge_indices,
|
||||
bool &slot, std::span<u8> converged,
|
||||
tbb::affinity_partitioner &ap) {
|
||||
bool &slot, tbb::affinity_partitioner &ap) {
|
||||
static Counter round("icf_round");
|
||||
round++;
|
||||
|
||||
@@ -417,27 +416,23 @@ static i64 propagate(std::span<std::vector<Digest>> digests,
|
||||
tbb::enumerable_thread_specific<i64> changed;
|
||||
|
||||
tbb::parallel_for((i64)0, num_digests, [&](i64 i) {
|
||||
if (converged[i])
|
||||
if (digests[slot][i] == digests[!slot][i])
|
||||
return;
|
||||
|
||||
SipHash13_128 hasher(hmac_key);
|
||||
hasher.update(digests[2][i].data(), HASH_SIZE);
|
||||
SHA256_CTX sha;
|
||||
SHA256_Init(&sha);
|
||||
SHA256_Update(&sha, digests[2][i].data(), HASH_SIZE);
|
||||
|
||||
i64 begin = edge_indices[i];
|
||||
i64 end = (i + 1 == num_digests) ? edges.size() : edge_indices[i + 1];
|
||||
|
||||
for (i64 j : edges.subspan(begin, end - begin))
|
||||
hasher.update(digests[slot][j].data(), HASH_SIZE);
|
||||
SHA256_Update(&sha, digests[slot][j].data(), HASH_SIZE);
|
||||
|
||||
hasher.finish(digests[!slot][i].data());
|
||||
digests[!slot][i] = digest_final(sha);
|
||||
|
||||
if (digests[slot][i] == digests[!slot][i]) {
|
||||
// This node has converged. Skip further iterations as it will
|
||||
// yield the same hash.
|
||||
converged[i] = true;
|
||||
} else {
|
||||
if (digests[slot][i] != digests[!slot][i])
|
||||
changed.local()++;
|
||||
}
|
||||
}, ap);
|
||||
|
||||
slot = !slot;
|
||||
@@ -475,45 +470,33 @@ static void print_icf_sections(Context<E> &ctx) {
|
||||
});
|
||||
|
||||
tbb::parallel_sort(leaders.begin(), leaders.end(),
|
||||
[](InputSection<E> *a, InputSection<E> *b) {
|
||||
[&](InputSection<E> *a, InputSection<E> *b) {
|
||||
return a->get_priority() < b->get_priority();
|
||||
});
|
||||
|
||||
std::ostream *out = &std::cout;
|
||||
std::ofstream file;
|
||||
|
||||
std::string &path = ctx.arg.print_icf_sections;
|
||||
if (path != "-") {
|
||||
file.open(path);
|
||||
if (file.fail())
|
||||
Fatal(ctx) << "--print-icf-sections: cannot open " << path << ": "
|
||||
<< errno_string();
|
||||
out = &file;
|
||||
}
|
||||
|
||||
i64 saved_bytes = 0;
|
||||
|
||||
for (InputSection<E> *leader : leaders) {
|
||||
auto [begin, end] = map.equal_range(leader);
|
||||
if (begin != end) {
|
||||
*out << "selected section " << *leader << '\n';
|
||||
for (auto it = begin; it != end; it++) {
|
||||
*out << " removing identical section " << *it->second << '\n';
|
||||
saved_bytes += leader->contents.size();
|
||||
}
|
||||
if (begin == end)
|
||||
continue;
|
||||
|
||||
SyncOut(ctx) << "selected section " << *leader;
|
||||
|
||||
i64 n = 0;
|
||||
for (auto it = begin; it != end; it++) {
|
||||
SyncOut(ctx) << " removing identical section " << *it->second;
|
||||
n++;
|
||||
}
|
||||
saved_bytes += leader->contents.size() * n;
|
||||
}
|
||||
|
||||
*out << "ICF saved " << saved_bytes << " bytes\n";
|
||||
SyncOut(ctx) << "ICF saved " << saved_bytes << " bytes";
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void icf_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "icf");
|
||||
if (ctx.objs.empty())
|
||||
return;
|
||||
|
||||
get_random_bytes(hmac_key, sizeof(hmac_key));
|
||||
|
||||
uniquify_cies(ctx);
|
||||
merge_leaf_nodes(ctx);
|
||||
@@ -521,15 +504,6 @@ void icf_sections(Context<E> &ctx) {
|
||||
// Prepare for the propagation rounds.
|
||||
std::vector<InputSection<E> *> sections = gather_sections(ctx);
|
||||
|
||||
// We allocate 3 arrays to store hashes for each vertex.
|
||||
//
|
||||
// Index 0 and 1 are used for tree hashes from the previous
|
||||
// iteration and the current iteration. They switch roles every
|
||||
// iteration. See `slot` below.
|
||||
//
|
||||
// Index 2 stores the initial, single-vertex hash. This is combined
|
||||
// with hashes from the connected vertices to form the tree hash
|
||||
// described above.
|
||||
std::vector<std::vector<Digest>> digests(3);
|
||||
digests[0] = compute_digests<E>(ctx, sections);
|
||||
digests[1].resize(digests[0].size());
|
||||
@@ -539,7 +513,6 @@ void icf_sections(Context<E> &ctx) {
|
||||
std::vector<u32> edge_indices;
|
||||
gather_edges<E>(ctx, sections, edges, edge_indices);
|
||||
|
||||
std::vector<u8> converged(digests[0].size());
|
||||
bool slot = 0;
|
||||
|
||||
// Execute the propagation rounds until convergence is obtained.
|
||||
@@ -547,33 +520,18 @@ void icf_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "propagate");
|
||||
tbb::affinity_partitioner ap;
|
||||
|
||||
// A cheap test that the graph hasn't converged yet.
|
||||
// The loop after this one uses a strict condition, but it's expensive
|
||||
// as it requires sorting the entire hash collection.
|
||||
//
|
||||
// For nodes that have a cycle in downstream (i.e. recursive
|
||||
// functions and functions that calls recursive functions) will always
|
||||
// change with the iterations. Nodes that doesn't (i.e. non-recursive
|
||||
// functions) will stop changing as soon as the propagation depth reaches
|
||||
// the call tree depth.
|
||||
// Here, we test whether we have reached sufficient depth for the latter,
|
||||
// which is a necessary (but not sufficient) condition for convergence.
|
||||
i64 num_changed = -1;
|
||||
for (;;) {
|
||||
i64 n = propagate<E>(digests, edges, edge_indices, slot, converged, ap);
|
||||
i64 n = propagate<E>(digests, edges, edge_indices, slot, ap);
|
||||
if (n == num_changed)
|
||||
break;
|
||||
num_changed = n;
|
||||
}
|
||||
|
||||
// Run the pass until the unique number of hashes stop increasing, at which
|
||||
// point we have achieved convergence (proof omitted for brevity).
|
||||
i64 num_classes = -1;
|
||||
for (;;) {
|
||||
// count_num_classes requires sorting which is O(n log n), so do a little
|
||||
// more work beforehand to amortize that log factor.
|
||||
for (i64 i = 0; i < 10; i++)
|
||||
propagate<E>(digests, edges, edge_indices, slot, converged, ap);
|
||||
propagate<E>(digests, edges, edge_indices, slot, ap);
|
||||
|
||||
i64 n = count_num_classes<E>(digests[slot], ap);
|
||||
if (n == num_classes)
|
||||
@@ -582,69 +540,55 @@ void icf_sections(Context<E> &ctx) {
|
||||
}
|
||||
}
|
||||
|
||||
// Group sections by hash values.
|
||||
// Group sections by SHA digest.
|
||||
{
|
||||
Timer t(ctx, "group");
|
||||
|
||||
auto *map =
|
||||
new tbb::concurrent_unordered_map<Digest, Atomic<InputSection<E> *>>;
|
||||
|
||||
auto *map = new tbb::concurrent_unordered_map<Digest, InputSection<E> *>;
|
||||
std::span<Digest> digest = digests[slot];
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)sections.size(), [&](i64 i) {
|
||||
InputSection<E> *isec = sections[i];
|
||||
auto [it, inserted] = map->insert({digest[i], isec});
|
||||
if (!inserted) {
|
||||
InputSection<E> *isec2 = it->second.load();
|
||||
while (isec->get_priority() < isec2->get_priority() &&
|
||||
!it->second.compare_exchange_strong(isec2, isec));
|
||||
}
|
||||
if (!inserted && isec->get_priority() < it->second->get_priority())
|
||||
it->second = isec;
|
||||
});
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)sections.size(), [&](i64 i) {
|
||||
auto it = map->find(digest[i]);
|
||||
assert(it != map->end());
|
||||
ASSERT(it != map->end());
|
||||
sections[i]->leader = it->second;
|
||||
});
|
||||
|
||||
// Since free'ing the map is slow, postpone it.
|
||||
ctx.on_exit.push_back([=] { delete map; });
|
||||
ctx.on_exit.push_back([=]() { delete map; });
|
||||
}
|
||||
|
||||
if (!ctx.arg.print_icf_sections.empty())
|
||||
if (ctx.arg.print_icf_sections)
|
||||
print_icf_sections(ctx);
|
||||
|
||||
// Update alignment of leaders.
|
||||
// Re-assign input sections to symbols.
|
||||
{
|
||||
Timer t(ctx, "update_alignment");
|
||||
Timer t(ctx, "reassign");
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (isec && isec->is_alive && isec->leader && isec->leader != isec.get()) {
|
||||
update_maximum(isec->leader->p2align, isec->p2align);
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// Eliminate duplicate sections.
|
||||
// Symbols pointing to eliminated sections will be redirected on the fly when
|
||||
// exporting to the symtab.
|
||||
{
|
||||
Timer t(ctx, "sweep");
|
||||
static Counter eliminated("icf_eliminated");
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (isec && isec->is_alive && isec->icf_removed()) {
|
||||
for (Symbol<E> *sym : file->symbols) {
|
||||
if (sym->file != file)
|
||||
continue;
|
||||
InputSection<E> *isec = sym->input_section;
|
||||
if (isec && isec->leader && isec->leader != isec) {
|
||||
sym->input_section = isec->leader;
|
||||
isec->kill();
|
||||
eliminated++;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
#define INSTANTIATE(E) \
|
||||
template void icf_sections(Context<E> &ctx);
|
||||
|
||||
template void icf_sections(Context<E> &ctx);
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,170 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <limits>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
InputSection<E>::InputSection(Context<E> &ctx, ObjectFile<E> &file,
|
||||
const ElfShdr<E> &shdr, std::string_view name,
|
||||
std::string_view contents, i64 section_idx)
|
||||
: file(file), shdr(shdr), nameptr(name.data()), namelen(name.size()),
|
||||
contents(contents), section_idx(section_idx) {
|
||||
// As a special case, we want to map .ctors and .dtors to
|
||||
// .init_array and .fini_array, respectively. However, old CRT
|
||||
// object files are not compatible with this translation, so we need
|
||||
// to keep them as-is if a section came from crtbegin.o or crtend.o.
|
||||
//
|
||||
// Yeah, this is an ugly hack, but the fundamental problem is that
|
||||
// we have two different mechanism, ctors/dtors and init_array/fini_array
|
||||
// for the same purpose. The latter was introduced to replace the
|
||||
// former, but as it is often the case, the former still lingers
|
||||
// around, so we need to keep this code to conver the old mechanism
|
||||
// to the new one.
|
||||
std::string_view stem = path_filename(file.filename);
|
||||
if (stem != "crtbegin.o" && stem != "crtend.o" &&
|
||||
stem != "crtbeginS.o" && stem != "crtendS.o" &&
|
||||
stem != "crtbeginT.o" && stem != "crtendT.o") {;
|
||||
if (name == ".ctors" || name.starts_with(".ctors."))
|
||||
name = ".init_array";
|
||||
else if (name == ".dtors" || name.starts_with(".dtors."))
|
||||
name = ".fini_array";
|
||||
}
|
||||
|
||||
output_section =
|
||||
OutputSection<E>::get_instance(ctx, name, shdr.sh_type, shdr.sh_flags);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void InputSection<E>::write_to(Context<E> &ctx, u8 *buf) {
|
||||
if (shdr.sh_type == SHT_NOBITS || shdr.sh_size == 0)
|
||||
return;
|
||||
|
||||
// Copy data
|
||||
memcpy(buf, contents.data(), contents.size());
|
||||
|
||||
// Apply relocations
|
||||
if (shdr.sh_flags & SHF_ALLOC)
|
||||
apply_reloc_alloc(ctx, buf);
|
||||
else
|
||||
apply_reloc_nonalloc(ctx, buf);
|
||||
|
||||
// As a special case, .ctors and .dtors section contents are
|
||||
// reversed. These sections are now obsolete and mapped to
|
||||
// .init_array and .fini_array, but they have to be reversed to
|
||||
// maintain the original semantics.
|
||||
bool init_fini = output_section->name == ".init_array" ||
|
||||
output_section->name == ".fini_array";
|
||||
bool ctors_dtors = name().starts_with(".ctors") ||
|
||||
name().starts_with(".dtors");
|
||||
if (init_fini && ctors_dtors)
|
||||
std::reverse((typename E::WordTy *)buf,
|
||||
(typename E::WordTy *)(buf + shdr.sh_size));
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 get_output_type(Context<E> &ctx) {
|
||||
if (ctx.arg.shared)
|
||||
return 0;
|
||||
if (ctx.arg.pie)
|
||||
return 1;
|
||||
return 2;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 get_sym_type(Context<E> &ctx, Symbol<E> &sym) {
|
||||
if (sym.is_absolute(ctx))
|
||||
return 0;
|
||||
if (!sym.is_imported)
|
||||
return 1;
|
||||
if (sym.get_type() != STT_FUNC)
|
||||
return 2;
|
||||
return 3;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void InputSection<E>::dispatch(Context<E> &ctx, Action table[3][4], i64 i) {
|
||||
std::span<ElfRel<E>> rels = get_rels(ctx);
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
bool is_code = (shdr.sh_flags & SHF_EXECINSTR);
|
||||
bool is_writable = (shdr.sh_flags & SHF_WRITE);
|
||||
Action action = table[get_output_type(ctx)][get_sym_type(ctx, sym)];
|
||||
|
||||
auto error = [&]() {
|
||||
Error(ctx) << *this << ": " << rel << " relocation against symbol `"
|
||||
<< sym << "' can not be used; recompile with -fPIC";
|
||||
};
|
||||
|
||||
switch (action) {
|
||||
case NONE:
|
||||
return;
|
||||
case ERROR:
|
||||
error();
|
||||
return;
|
||||
case COPYREL:
|
||||
if (!ctx.arg.z_copyreloc) {
|
||||
error();
|
||||
return;
|
||||
}
|
||||
if (sym.esym().st_visibility == STV_PROTECTED) {
|
||||
Error(ctx) << *this << ": cannot make copy relocation for "
|
||||
<< " protected symbol '" << sym << "', defined in "
|
||||
<< *sym.file;
|
||||
return;
|
||||
}
|
||||
sym.flags |= NEEDS_COPYREL;
|
||||
return;
|
||||
case PLT:
|
||||
sym.flags |= NEEDS_PLT;
|
||||
return;
|
||||
case DYNREL:
|
||||
if (!is_writable) {
|
||||
if (!is_code || ctx.arg.z_text) {
|
||||
error();
|
||||
return;
|
||||
}
|
||||
ctx.has_textrel = true;
|
||||
}
|
||||
sym.flags |= NEEDS_DYNSYM;
|
||||
rel_exprs[i] = R_DYN;
|
||||
file.num_dynrel++;
|
||||
return;
|
||||
case BASEREL:
|
||||
if (!is_writable) {
|
||||
if (!is_code || ctx.arg.z_text) {
|
||||
error();
|
||||
return;
|
||||
}
|
||||
ctx.has_textrel = true;
|
||||
}
|
||||
rel_exprs[i] = R_BASEREL;
|
||||
file.num_dynrel++;
|
||||
return;
|
||||
default:
|
||||
unreachable(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void InputSection<E>::report_undef(Context<E> &ctx, Symbol<E> &sym) {
|
||||
switch (ctx.arg.unresolved_symbols) {
|
||||
case UnresolvedKind::ERROR:
|
||||
Error(ctx) << "undefined symbol: " << file << ": " << sym;
|
||||
break;
|
||||
case UnresolvedKind::WARN:
|
||||
Warn(ctx) << "undefined symbol: " << file << ": " << sym;
|
||||
break;
|
||||
case UnresolvedKind::IGNORE:
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template class InputSection<E>;
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,387 @@
|
||||
// On Linux, /usr/lib/x86_64-linux-gnu/libc.so is not actually
|
||||
// a shared object file but an ASCII text file containing a linker
|
||||
// script to include a "real" libc.so file. Therefore, we need to
|
||||
// support a (very limited) subset of the linker script language.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <cctype>
|
||||
#include <iomanip>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
static thread_local MemoryMappedFile<E> *current_file;
|
||||
|
||||
template <typename E>
|
||||
void read_version_script(Context<E> &ctx, std::span<std::string_view> &tok);
|
||||
|
||||
static std::string_view get_line(std::string_view input, const char *pos) {
|
||||
ASSERT(input.data() <= pos);
|
||||
ASSERT(pos < input.data() + input.size());
|
||||
|
||||
i64 start = input.rfind('\n', pos - input.data());
|
||||
if (start == input.npos)
|
||||
start = 0;
|
||||
else
|
||||
start++;
|
||||
|
||||
i64 end = input.find('\n', pos - input.data());
|
||||
if (end == input.npos)
|
||||
end = input.size();
|
||||
|
||||
return input.substr(start, end - start);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
class SyntaxError {
|
||||
public:
|
||||
SyntaxError(Context<E> &ctx, std::string_view errpos) : out(ctx) {
|
||||
std::string_view contents = current_file<E>->get_contents(ctx);
|
||||
std::string_view line = get_line(contents, errpos.data());
|
||||
|
||||
i64 lineno = 1;
|
||||
for (i64 i = 0; contents.data() + i < line.data(); i++)
|
||||
if (contents[i] == '\n')
|
||||
lineno++;
|
||||
|
||||
i64 column = errpos.data() - line.data();
|
||||
|
||||
std::stringstream ss;
|
||||
ss << current_file<E>->name << ":" << lineno << ": ";
|
||||
i64 indent = (i64)ss.tellp() + strlen("mold: ");
|
||||
ss << line << "\n" << std::setw(indent + column) << " " << "^ ";
|
||||
out << ss.str();
|
||||
}
|
||||
|
||||
template <class T> SyntaxError &operator<<(T &&val) {
|
||||
out << std::forward<T>(val);
|
||||
return *this;
|
||||
}
|
||||
|
||||
[[noreturn]] ~SyntaxError() = default;
|
||||
|
||||
Fatal<E> out;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
static std::vector<std::string_view>
|
||||
tokenize(Context<E> &ctx, std::string_view input) {
|
||||
std::vector<std::string_view> vec;
|
||||
while (!input.empty()) {
|
||||
if (isspace(input[0])) {
|
||||
input = input.substr(1);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (input.starts_with("/*")) {
|
||||
i64 pos = input.find("*/", 2);
|
||||
if (pos == std::string_view::npos)
|
||||
SyntaxError(ctx, input) << "unclosed comment";
|
||||
input = input.substr(pos + 2);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (input[0] == '#') {
|
||||
i64 pos = input.find("\n", 1);
|
||||
if (pos == std::string_view::npos)
|
||||
break;
|
||||
input = input.substr(pos + 1);
|
||||
continue;
|
||||
}
|
||||
|
||||
if (input[0] == '"') {
|
||||
i64 pos = input.find('"', 1);
|
||||
if (pos == std::string_view::npos)
|
||||
SyntaxError(ctx, input) << "unclosed string literal";
|
||||
vec.push_back(input.substr(0, pos + 1));
|
||||
input = input.substr(pos + 1);
|
||||
continue;
|
||||
}
|
||||
|
||||
i64 pos = input.find_first_not_of(
|
||||
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz"
|
||||
"0123456789_.$/\\~=+[]*?-!^:");
|
||||
|
||||
if (pos == 0)
|
||||
pos = 1;
|
||||
else if (pos == input.npos)
|
||||
pos = input.size();
|
||||
|
||||
vec.push_back(input.substr(0, pos));
|
||||
input = input.substr(pos);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::span<std::string_view>
|
||||
skip(Context<E> &ctx, std::span<std::string_view> tok, std::string_view str) {
|
||||
if (tok.empty())
|
||||
Fatal(ctx) << current_file<E>->name << ": expected '" << str
|
||||
<< "', but got EOF";
|
||||
if (tok[0] != str)
|
||||
SyntaxError(ctx, tok[0]) << "expected '" << str << "'";
|
||||
return tok.subspan(1);
|
||||
}
|
||||
|
||||
static std::string_view unquote(std::string_view s) {
|
||||
if (s.size() > 0 && s[0] == '"') {
|
||||
ASSERT(s[s.size() - 1] == '"');
|
||||
return s.substr(1, s.size() - 2);
|
||||
}
|
||||
return s;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::span<std::string_view>
|
||||
read_output_format(Context<E> &ctx, std::span<std::string_view> tok) {
|
||||
tok = skip(ctx, tok, "(");
|
||||
while (!tok.empty() && tok[0] != ")")
|
||||
tok = tok.subspan(1);
|
||||
if (tok.empty())
|
||||
Fatal(ctx) << current_file<E>->name << ": expected ')', but got EOF";
|
||||
return tok.subspan(1);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static bool is_in_sysroot(Context<E> &ctx, std::string path) {
|
||||
std::string sysroot = path_clean(path_to_absolute(ctx.arg.sysroot));
|
||||
path = path_clean(path_to_absolute(path));
|
||||
return path_dirname(path).starts_with(sysroot);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static MemoryMappedFile<E> *resolve_path(Context<E> &ctx, std::string_view tok) {
|
||||
std::string str(unquote(tok));
|
||||
|
||||
// GNU ld prepends the sysroot if a pathname starts with '/' and the
|
||||
// script being processed is in the sysroot. We do the same.
|
||||
if (str.starts_with('/') && is_in_sysroot(ctx, current_file<E>->name))
|
||||
return MemoryMappedFile<E>::must_open(ctx, ctx.arg.sysroot + str);
|
||||
|
||||
if (str.starts_with("-l"))
|
||||
return find_library(ctx, str.substr(2));
|
||||
|
||||
if (MemoryMappedFile<E> *mb = MemoryMappedFile<E>::open(ctx, str))
|
||||
return mb;
|
||||
|
||||
for (std::string_view dir : ctx.arg.library_paths) {
|
||||
std::string path = std::string(dir) + "/" + str;
|
||||
if (MemoryMappedFile<E> *mb = MemoryMappedFile<E>::open(ctx, path))
|
||||
return mb;
|
||||
}
|
||||
|
||||
SyntaxError(ctx, tok) << "library not found: " << str;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::span<std::string_view>
|
||||
read_group(Context<E> &ctx, std::span<std::string_view> tok) {
|
||||
tok = skip(ctx, tok, "(");
|
||||
|
||||
while (!tok.empty() && tok[0] != ")") {
|
||||
if (tok[0] == "AS_NEEDED") {
|
||||
bool orig = ctx.as_needed;
|
||||
ctx.as_needed = true;
|
||||
tok = read_group(ctx, tok.subspan(1));
|
||||
ctx.as_needed = orig;
|
||||
continue;
|
||||
}
|
||||
|
||||
MemoryMappedFile<E> *mb = resolve_path(ctx, tok[0]);
|
||||
read_file(ctx, mb);
|
||||
tok = tok.subspan(1);
|
||||
}
|
||||
|
||||
if (tok.empty())
|
||||
Fatal(ctx) << current_file<E>->name << ": expected ')', but got EOF";
|
||||
return tok.subspan(1);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void parse_linker_script(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
current_file<E> = mb;
|
||||
|
||||
std::vector<std::string_view> vec = tokenize(ctx, mb->get_contents(ctx));
|
||||
std::span<std::string_view> tok = vec;
|
||||
|
||||
while (!tok.empty()) {
|
||||
if (tok[0] == "OUTPUT_FORMAT") {
|
||||
tok = read_output_format(ctx, tok.subspan(1));
|
||||
} else if (tok[0] == "INPUT" || tok[0] == "GROUP") {
|
||||
tok = read_group(ctx, tok.subspan(1));
|
||||
} else if (tok[0] == "VERSION") {
|
||||
tok = tok.subspan(1);
|
||||
tok = skip(ctx, tok, "{");
|
||||
read_version_script(ctx, tok);
|
||||
tok = skip(ctx, tok, "}");
|
||||
} else if (tok[0] == ";") {
|
||||
tok = tok.subspan(1);
|
||||
} else {
|
||||
SyntaxError(ctx, tok[0]) << "unknown linker script token";
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
i64 get_script_output_type(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
current_file<E> = mb;
|
||||
|
||||
std::vector<std::string_view> vec = tokenize(ctx, mb->get_contents(ctx));
|
||||
std::span<std::string_view> tok = vec;
|
||||
|
||||
if (tok.size() >= 3 && tok[0] == "OUTPUT_FORMAT" && tok[1] == "(") {
|
||||
if (tok[2] == "elf64-x86-64")
|
||||
return EM_X86_64;
|
||||
if (tok[2] == "elf32-i386")
|
||||
return EM_386;
|
||||
}
|
||||
return -1;
|
||||
}
|
||||
|
||||
static bool read_label(std::span<std::string_view> &tok,
|
||||
std::string label) {
|
||||
if (tok.size() >= 1 && tok[0] == label + ":") {
|
||||
tok = tok.subspan(1);
|
||||
return true;
|
||||
}
|
||||
|
||||
if (tok.size() >= 2 && tok[0] == label && tok[1] == ":") {
|
||||
tok = tok.subspan(2);
|
||||
return true;
|
||||
}
|
||||
return false;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void read_version_script_commands(Context<E> &ctx,
|
||||
std::span<std::string_view> &tok,
|
||||
i16 &ver, bool is_extern_cpp) {
|
||||
bool is_global = true;
|
||||
|
||||
while (!tok.empty() && tok[0] != "}") {
|
||||
if (read_label(tok, "global")) {
|
||||
is_global = true;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (read_label(tok, "local")) {
|
||||
is_global = false;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (tok[0] == "extern") {
|
||||
tok = tok.subspan(1);
|
||||
|
||||
bool is_cpp;
|
||||
if (!tok.empty() && tok[0] == "\"C\"") {
|
||||
tok = tok.subspan(1);
|
||||
is_cpp = false;
|
||||
} else {
|
||||
tok = skip(ctx, tok, "\"C++\"");
|
||||
is_cpp = true;
|
||||
}
|
||||
|
||||
tok = skip(ctx, tok, "{");
|
||||
read_version_script_commands(ctx, tok, ver, is_cpp);
|
||||
tok = skip(ctx, tok, "}");
|
||||
tok = skip(ctx, tok, ";");
|
||||
continue;
|
||||
}
|
||||
|
||||
if (tok[0] == "*")
|
||||
ctx.arg.default_version = (is_global ? ver : VER_NDX_LOCAL);
|
||||
else
|
||||
ctx.arg.version_patterns.push_back({tok[0], ver, is_extern_cpp});
|
||||
tok = tok.subspan(1);
|
||||
|
||||
if (!tok.empty() && tok[0] == "}")
|
||||
return;
|
||||
tok = skip(ctx, tok, ";");
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void read_version_script(Context<E> &ctx, std::span<std::string_view> &tok) {
|
||||
i16 next_ver = VER_NDX_LAST_RESERVED + 1;
|
||||
|
||||
while (!tok.empty() && tok[0] != "}") {
|
||||
i16 ver = VER_NDX_GLOBAL;
|
||||
if (tok[0] != "{") {
|
||||
ver = next_ver++;
|
||||
ctx.arg.version_definitions.push_back(tok[0]);
|
||||
tok = tok.subspan(1);
|
||||
}
|
||||
|
||||
tok = skip(ctx, tok, "{");
|
||||
read_version_script_commands(ctx, tok, ver, false);
|
||||
tok = skip(ctx, tok, "}");
|
||||
if (!tok.empty() && tok[0] != ";")
|
||||
tok = tok.subspan(1);
|
||||
tok = skip(ctx, tok, ";");
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void parse_version_script(Context<E> &ctx, std::string path) {
|
||||
current_file<E> = MemoryMappedFile<E>::must_open(ctx, path);
|
||||
std::vector<std::string_view> vec =
|
||||
tokenize(ctx, current_file<E>->get_contents(ctx));
|
||||
std::span<std::string_view> tok = vec;
|
||||
read_version_script(ctx, tok);
|
||||
if (!tok.empty())
|
||||
SyntaxError(ctx, tok[0]) << "trailing garbage token";
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void parse_dynamic_list(Context<E> &ctx, std::string path) {
|
||||
current_file<E> = MemoryMappedFile<E>::must_open(ctx, path);
|
||||
std::vector<std::string_view> vec =
|
||||
tokenize(ctx, current_file<E>->get_contents(ctx));
|
||||
|
||||
std::span<std::string_view> tok = vec;
|
||||
|
||||
tok = skip(ctx, tok, "{");
|
||||
i16 ver = VER_NDX_GLOBAL;
|
||||
|
||||
while (!tok.empty() && tok[0] != "}") {
|
||||
if (read_label(tok, "global")) {
|
||||
ver = VER_NDX_GLOBAL;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (read_label(tok, "local")) {
|
||||
ver = VER_NDX_LOCAL;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (tok[0] == "*")
|
||||
ctx.arg.default_version = ver;
|
||||
else
|
||||
ctx.arg.version_patterns.push_back({tok[0], ver, false});
|
||||
tok = skip(ctx, tok.subspan(1), ";");
|
||||
}
|
||||
|
||||
tok = skip(ctx, tok, "}");
|
||||
tok = skip(ctx, tok, ";");
|
||||
|
||||
if (!tok.empty())
|
||||
SyntaxError(ctx, tok[0]) << "trailing garbage token";
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template \
|
||||
void parse_linker_script(Context<E> &ctx, MemoryMappedFile<E> *mb); \
|
||||
template \
|
||||
i64 get_script_output_type(Context<E> &ctx, MemoryMappedFile<E> *mb); \
|
||||
template \
|
||||
void parse_version_script(Context<E> &ctx, std::string path); \
|
||||
template \
|
||||
void parse_dynamic_list(Context<E> &ctx, std::string path)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+686
@@ -0,0 +1,686 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <cstring>
|
||||
#include <functional>
|
||||
#include <iomanip>
|
||||
#include <map>
|
||||
#include <signal.h>
|
||||
#include <tbb/global_control.h>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <unistd.h>
|
||||
#include <unordered_set>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
std::string errno_string() {
|
||||
char buf[500];
|
||||
strerror_r(errno, buf, sizeof(buf));
|
||||
return buf;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::string_view save_string(Context<E> &ctx, const std::string &str) {
|
||||
u8 *buf = new u8[str.size() + 1];
|
||||
memcpy(buf, str.data(), str.size());
|
||||
buf[str.size()] = '\0';
|
||||
ctx.owning_bufs.push_back(std::unique_ptr<u8[]>(buf));
|
||||
return {(char *)buf, str.size()};
|
||||
}
|
||||
|
||||
std::string get_version_string() {
|
||||
if (strlen(GIT_HASH) == 0)
|
||||
return "mold " MOLD_VERSION " (compatible with GNU ld and GNU gold)";
|
||||
return "mold " MOLD_VERSION " (" GIT_HASH
|
||||
"; compatible with GNU ld and GNU gold)";
|
||||
}
|
||||
|
||||
std::regex glob_to_regex(std::string_view pattern) {
|
||||
std::stringstream ss;
|
||||
for (u8 c : pattern) {
|
||||
if (c == '*')
|
||||
ss << ".*";
|
||||
else
|
||||
ss << "\\x" << std::hex << std::setw(2) << std::setfill('0') << (int)c;
|
||||
}
|
||||
return std::regex(ss.str(), std::regex::optimize);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static ObjectFile<E> *new_object_file(Context<E> &ctx, MemoryMappedFile<E> *mb,
|
||||
std::string archive_name) {
|
||||
static Counter count("parsed_objs");
|
||||
count++;
|
||||
|
||||
bool in_lib = (!archive_name.empty() && !ctx.whole_archive);
|
||||
ObjectFile<E> *file = ObjectFile<E>::create(ctx, mb, archive_name, in_lib);
|
||||
file->priority = ctx.file_priority++;
|
||||
ctx.tg.run([file, &ctx]() { file->parse(ctx); });
|
||||
if (ctx.arg.trace)
|
||||
SyncOut(ctx) << "trace: " << *file;
|
||||
return file;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static SharedFile<E> *new_shared_file(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
SharedFile<E> *file = SharedFile<E>::create(ctx, mb);
|
||||
file->priority = ctx.file_priority++;
|
||||
ctx.tg.run([file, &ctx]() { file->parse(ctx); });
|
||||
if (ctx.arg.trace)
|
||||
SyncOut(ctx) << "trace: " << *file;
|
||||
return file;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void read_file(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
if (ctx.visited.contains(mb->name))
|
||||
return;
|
||||
|
||||
if (ctx.is_preloading) {
|
||||
switch (get_file_type(ctx, mb)) {
|
||||
case FileType::OBJ:
|
||||
ctx.obj_cache.store(mb, new_object_file(ctx, mb, ""));
|
||||
return;
|
||||
case FileType::DSO:
|
||||
ctx.dso_cache.store(mb, new_shared_file(ctx, mb));
|
||||
return;
|
||||
case FileType::AR:
|
||||
for (MemoryMappedFile<E> *child : read_fat_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
ctx.obj_cache.store(mb, new_object_file(ctx, child, mb->name));
|
||||
return;
|
||||
case FileType::THIN_AR:
|
||||
for (MemoryMappedFile<E> *child : read_thin_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
ctx.obj_cache.store(child, new_object_file(ctx, child, mb->name));
|
||||
return;
|
||||
case FileType::TEXT:
|
||||
parse_linker_script(ctx, mb);
|
||||
return;
|
||||
case FileType::LLVM_BITCODE:
|
||||
Fatal(ctx) << mb->name << ": looks like this is an LLVM bitcode, but"
|
||||
<< " mold does not support LTO";
|
||||
default:
|
||||
Fatal(ctx) << mb->name << ": unknown file type";
|
||||
}
|
||||
}
|
||||
|
||||
switch (get_file_type(ctx, mb)) {
|
||||
case FileType::OBJ:
|
||||
if (ObjectFile<E> *obj = ctx.obj_cache.get_one(mb))
|
||||
ctx.objs.push_back(obj);
|
||||
else
|
||||
ctx.objs.push_back(new_object_file(ctx, mb, ""));
|
||||
return;
|
||||
case FileType::DSO:
|
||||
if (SharedFile<E> *obj = ctx.dso_cache.get_one(mb))
|
||||
ctx.dsos.push_back(obj);
|
||||
else
|
||||
ctx.dsos.push_back(new_shared_file(ctx, mb));
|
||||
ctx.visited.insert(mb->name);
|
||||
return;
|
||||
case FileType::AR: {
|
||||
std::vector<ObjectFile<E> *> objs = ctx.obj_cache.get(mb);
|
||||
if (!objs.empty()) {
|
||||
append(ctx.objs, objs);
|
||||
} else {
|
||||
for (MemoryMappedFile<E> *child : read_fat_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
ctx.objs.push_back(new_object_file(ctx, child, mb->name));
|
||||
}
|
||||
ctx.visited.insert(mb->name);
|
||||
return;
|
||||
}
|
||||
case FileType::THIN_AR:
|
||||
for (MemoryMappedFile<E> *child : read_thin_archive_members(ctx, mb)) {
|
||||
if (ObjectFile<E> *obj = ctx.obj_cache.get_one(child))
|
||||
ctx.objs.push_back(obj);
|
||||
else if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
ctx.objs.push_back(new_object_file(ctx, child, mb->name));
|
||||
}
|
||||
ctx.visited.insert(mb->name);
|
||||
return;
|
||||
case FileType::TEXT:
|
||||
parse_linker_script(ctx, mb);
|
||||
return;
|
||||
case FileType::LLVM_BITCODE:
|
||||
Fatal(ctx) << mb->name << ": looks like this is an LLVM bitcode, "
|
||||
<< "but mold does not support LTO";
|
||||
default:
|
||||
Fatal(ctx) << mb->name << ": unknown file type";
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void cleanup() {
|
||||
if (OutputFile<E>::tmpfile)
|
||||
unlink(OutputFile<E>::tmpfile);
|
||||
if (socket_tmpfile)
|
||||
unlink(socket_tmpfile);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void signal_handler(int) {
|
||||
cleanup<E>();
|
||||
_exit(1);
|
||||
}
|
||||
|
||||
// Read the beginning of a given file and returns its machine type
|
||||
// (e.g. EM_X86_64 or EM_386). Return -1 if unknown.
|
||||
template <typename E>
|
||||
static i64 get_machine_type(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
switch (get_file_type(ctx, mb)) {
|
||||
case FileType::DSO:
|
||||
return ((ElfEhdr<E> *)mb->data(ctx))->e_machine;
|
||||
case FileType::AR:
|
||||
for (MemoryMappedFile<E> *child : read_fat_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
return ((ElfEhdr<E> *)child->data(ctx))->e_machine;
|
||||
return -1;
|
||||
case FileType::THIN_AR:
|
||||
for (MemoryMappedFile<E> *child : read_thin_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
return ((ElfEhdr<E> *)child->data(ctx))->e_machine;
|
||||
return -1;
|
||||
case FileType::TEXT:
|
||||
return get_script_output_type(ctx, mb);
|
||||
default:
|
||||
return -1;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static MemoryMappedFile<E> *open_library(Context<E> &ctx, std::string path) {
|
||||
MemoryMappedFile<E> *mb = MemoryMappedFile<E>::open(ctx, path);
|
||||
if (!mb)
|
||||
return nullptr;
|
||||
|
||||
i64 type = get_machine_type(ctx, mb);
|
||||
if (type == -1 || type == E::e_machine)
|
||||
return mb;
|
||||
Warn(ctx) << path << ": skipping incompatible file"
|
||||
<< " " << (int)type << " " << (int)E::e_machine;
|
||||
return nullptr;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
MemoryMappedFile<E> *find_library(Context<E> &ctx, std::string name) {
|
||||
if (name.starts_with(':')) {
|
||||
for (std::string_view dir : ctx.arg.library_paths) {
|
||||
std::string path = std::string(dir) + "/" + name.substr(1);
|
||||
if (MemoryMappedFile<E> *mb = open_library(ctx, path))
|
||||
return mb;
|
||||
}
|
||||
Fatal(ctx) << "library not found: " << name;
|
||||
}
|
||||
|
||||
for (std::string_view dir : ctx.arg.library_paths) {
|
||||
std::string stem = std::string(dir) + "/lib" + name;
|
||||
if (!ctx.is_static)
|
||||
if (MemoryMappedFile<E> *mb = open_library(ctx, stem + ".so"))
|
||||
return mb;
|
||||
if (MemoryMappedFile<E> *mb = open_library(ctx, stem + ".a"))
|
||||
return mb;
|
||||
}
|
||||
Fatal(ctx) << "library not found: " << name;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void read_input_files(Context<E> &ctx, std::span<std::string_view> args) {
|
||||
std::vector<std::tuple<bool, bool, bool>> state;
|
||||
|
||||
while (!args.empty()) {
|
||||
std::string_view arg;
|
||||
|
||||
if (read_flag(args, "as-needed")) {
|
||||
ctx.as_needed = true;
|
||||
} else if (read_flag(args, "no-as-needed")) {
|
||||
ctx.as_needed = false;
|
||||
} else if (read_flag(args, "whole-archive")) {
|
||||
ctx.whole_archive = true;
|
||||
} else if (read_flag(args, "no-whole-archive")) {
|
||||
ctx.whole_archive = false;
|
||||
} else if (read_flag(args, "Bstatic")) {
|
||||
ctx.is_static = true;
|
||||
} else if (read_flag(args, "Bdynamic")) {
|
||||
ctx.is_static = false;
|
||||
} else if (read_arg(ctx, args, arg, "version-script")) {
|
||||
parse_version_script(ctx, std::string(arg));
|
||||
} else if (read_arg(ctx, args, arg, "dynamic-list")) {
|
||||
parse_dynamic_list(ctx, std::string(arg));
|
||||
} else if (read_flag(args, "push-state")) {
|
||||
state.push_back({ctx.as_needed, ctx.whole_archive, ctx.is_static});
|
||||
} else if (read_flag(args, "pop-state")) {
|
||||
if (state.empty())
|
||||
Fatal(ctx) << "no state pushed before popping";
|
||||
std::tie(ctx.as_needed, ctx.whole_archive, ctx.is_static) = state.back();
|
||||
state.pop_back();
|
||||
} else if (read_arg(ctx, args, arg, "l")) {
|
||||
MemoryMappedFile<E> *mb = find_library(ctx, std::string(arg));
|
||||
mb->given_fullpath = false;
|
||||
read_file(ctx, mb);
|
||||
} else {
|
||||
read_file(ctx, MemoryMappedFile<E>::must_open(ctx, std::string(args[0])));
|
||||
args = args.subspan(1);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void show_stats(Context<E> &ctx) {
|
||||
for (ObjectFile<E> *obj : ctx.objs) {
|
||||
static Counter defined("defined_syms");
|
||||
defined += obj->first_global - 1;
|
||||
|
||||
static Counter undefined("undefined_syms");
|
||||
undefined += obj->symbols.size() - obj->first_global;
|
||||
|
||||
for (std::unique_ptr<InputSection<E>> &sec : obj->sections) {
|
||||
if (!sec || !sec->is_alive)
|
||||
continue;
|
||||
|
||||
static Counter alloc("reloc_alloc");
|
||||
static Counter nonalloc("reloc_nonalloc");
|
||||
|
||||
if (sec->shdr.sh_flags & SHF_ALLOC)
|
||||
alloc += sec->get_rels(ctx).size();
|
||||
else
|
||||
nonalloc += sec->get_rels(ctx).size();
|
||||
}
|
||||
|
||||
static Counter comdats("comdats");
|
||||
comdats += obj->comdat_groups.size();
|
||||
|
||||
static Counter removed_comdats("removed_comdat_mem");
|
||||
for (auto &pair : obj->comdat_groups)
|
||||
if (ComdatGroup *group = pair.first; group->owner != obj->priority)
|
||||
removed_comdats += pair.second.size();
|
||||
|
||||
static Counter num_cies("num_cies");
|
||||
num_cies += obj->cies.size();
|
||||
|
||||
static Counter num_unique_cies("num_unique_cies");
|
||||
for (CieRecord<E> &cie : obj->cies)
|
||||
if (cie.is_leader)
|
||||
num_unique_cies++;
|
||||
|
||||
static Counter num_fdes("num_fdes");
|
||||
num_fdes += obj->fdes.size();
|
||||
}
|
||||
|
||||
static Counter num_bytes("total_input_bytes");
|
||||
for (std::unique_ptr<MemoryMappedFile<E>> &mb : ctx.owning_mbs)
|
||||
num_bytes += mb->size();
|
||||
|
||||
static Counter num_input_sections("input_sections");
|
||||
for (ObjectFile<E> *file : ctx.objs)
|
||||
num_input_sections += file->sections.size();
|
||||
|
||||
static Counter num_output_chunks("output_chunks", ctx.chunks.size());
|
||||
static Counter num_objs("num_objs", ctx.objs.size());
|
||||
static Counter num_dsos("num_dsos", ctx.dsos.size());
|
||||
|
||||
Counter::print();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
int do_main(int argc, char **argv) {
|
||||
Context<E> ctx;
|
||||
|
||||
// Process -run option first. process_run_subcommand() does not return.
|
||||
if (argc >= 2)
|
||||
if (std::string_view arg = argv[1]; arg == "-run" || arg == "--run")
|
||||
process_run_subcommand(ctx, argc, argv);
|
||||
|
||||
// Parse non-positional command line options
|
||||
ctx.cmdline_args = expand_response_files(ctx, argv);
|
||||
std::vector<std::string_view> file_args;
|
||||
parse_nonpositional_args(ctx, file_args);
|
||||
|
||||
// Redo if -m is not x86-64.
|
||||
if (ctx.arg.emulation != E::e_machine) {
|
||||
switch (ctx.arg.emulation) {
|
||||
case EM_386:
|
||||
return do_main<I386>(argc, argv);
|
||||
case EM_AARCH64:
|
||||
return do_main<AARCH64>(argc, argv);
|
||||
}
|
||||
unreachable(ctx);
|
||||
}
|
||||
|
||||
Timer t_all(ctx, "all");
|
||||
|
||||
if (ctx.arg.relocatable) {
|
||||
combine_objects(ctx, file_args);
|
||||
return 0;
|
||||
}
|
||||
|
||||
if (!ctx.arg.preload)
|
||||
try_resume_daemon(ctx);
|
||||
|
||||
tbb::global_control tbb_cont(tbb::global_control::max_allowed_parallelism,
|
||||
ctx.arg.thread_count);
|
||||
|
||||
signal(SIGINT, signal_handler<E>);
|
||||
signal(SIGTERM, signal_handler<E>);
|
||||
|
||||
if (!ctx.arg.directory.empty() && chdir(ctx.arg.directory.c_str()) == -1)
|
||||
Fatal(ctx) << "chdir failed: " << ctx.arg.directory
|
||||
<< ": " << errno_string();
|
||||
|
||||
// Handle --wrap options if any.
|
||||
for (std::string_view name : ctx.arg.wrap)
|
||||
Symbol<E>::intern(ctx, name)->wrap = true;
|
||||
|
||||
// Handle --retain-symbols-file options if any.
|
||||
if (ctx.arg.retain_symbols_file)
|
||||
for (std::string_view name : *ctx.arg.retain_symbols_file)
|
||||
Symbol<E>::intern(ctx, name)->write_to_symtab = true;
|
||||
|
||||
// Preload input files
|
||||
std::function<void()> on_complete;
|
||||
|
||||
if (ctx.arg.preload) {
|
||||
Timer t(ctx, "preload");
|
||||
std::function<void()> wait_for_client;
|
||||
daemonize(ctx, &wait_for_client, &on_complete);
|
||||
|
||||
ctx.reset_reader_context(true);
|
||||
read_input_files(ctx, file_args);
|
||||
ctx.tg.wait();
|
||||
t.stop();
|
||||
|
||||
Timer t2(ctx, "wait_for_client");
|
||||
wait_for_client();
|
||||
} else if (ctx.arg.fork) {
|
||||
on_complete = fork_child();
|
||||
}
|
||||
|
||||
for (std::string_view arg : ctx.arg.trace_symbol)
|
||||
Symbol<E>::intern(ctx, arg)->traced = true;
|
||||
|
||||
// Parse input files
|
||||
{
|
||||
Timer t(ctx, "parse");
|
||||
ctx.reset_reader_context(false);
|
||||
read_input_files(ctx, file_args);
|
||||
ctx.tg.wait();
|
||||
}
|
||||
|
||||
if (ctx.objs.empty())
|
||||
Fatal(ctx) << "no input files";
|
||||
|
||||
{
|
||||
Timer t(ctx, "register_section_pieces");
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->register_section_pieces(ctx);
|
||||
});
|
||||
}
|
||||
|
||||
// Uniquify shared object files by soname
|
||||
{
|
||||
std::unordered_set<std::string_view> seen;
|
||||
erase(ctx.dsos, [&](SharedFile<E> *file) {
|
||||
return !seen.insert(file->soname).second;
|
||||
});
|
||||
}
|
||||
|
||||
Timer t_total(ctx, "total");
|
||||
Timer t_before_copy(ctx, "before_copy");
|
||||
|
||||
// Apply -exclude-libs
|
||||
apply_exclude_libs(ctx);
|
||||
|
||||
// Create instances of linker-synthesized sections such as
|
||||
// .got or .plt.
|
||||
create_synthetic_sections(ctx);
|
||||
|
||||
// Resolve symbols and fix the set of object files that are
|
||||
// included to the final output.
|
||||
resolve_symbols(ctx);
|
||||
|
||||
// Remove redundant comdat sections (e.g. duplicate inline functions).
|
||||
eliminate_comdats(ctx);
|
||||
|
||||
// Create .bss sections for common symbols.
|
||||
convert_common_symbols(ctx);
|
||||
|
||||
// Apply version scripts.
|
||||
apply_version_script(ctx);
|
||||
|
||||
// Parse symbol version suffixes (e.g. "foo@ver1").
|
||||
parse_symbol_version(ctx);
|
||||
|
||||
// Set is_import and is_export bits for each symbol.
|
||||
compute_import_export(ctx);
|
||||
|
||||
// Garbage-collect unreachable sections.
|
||||
if (ctx.arg.gc_sections)
|
||||
gc_sections(ctx);
|
||||
|
||||
// Merge identical read-only sections.
|
||||
if (ctx.arg.icf)
|
||||
icf_sections(ctx);
|
||||
|
||||
// Compute sizes of sections containing mergeable strings.
|
||||
compute_merged_section_sizes(ctx);
|
||||
|
||||
// ctx input sections into output sections
|
||||
bin_sections(ctx);
|
||||
|
||||
// Get a list of output sections.
|
||||
append(ctx.chunks, collect_output_sections(ctx));
|
||||
|
||||
// Create a dummy file containing linker-synthesized symbols
|
||||
// (e.g. `__bss_start`).
|
||||
ctx.internal_obj = create_internal_file(ctx);
|
||||
ctx.internal_obj->resolve_regular_symbols(ctx);
|
||||
ctx.objs.push_back(ctx.internal_obj);
|
||||
|
||||
// Beyond this point, no new files will be added to ctx.objs
|
||||
// or ctx.dsos.
|
||||
|
||||
// If we are linking a .so file, remaining undefined symbols does
|
||||
// not cause a linker error. Instead, they are treated as if they
|
||||
// were imported symbols.
|
||||
//
|
||||
// If we are linking an executable, weak undefs are converted to
|
||||
// weakly imported symbol so that they'll have another chance to be
|
||||
// resolved.
|
||||
claim_unresolved_symbols(ctx);
|
||||
|
||||
// Beyond this point, no new symbols will be added to the result.
|
||||
|
||||
// Make sure that all symbols have been resolved.
|
||||
if (!ctx.arg.allow_multiple_definition)
|
||||
check_duplicate_symbols(ctx);
|
||||
|
||||
// .init_array and .fini_array contents have to be sorted by
|
||||
// a special rule. Sort them.
|
||||
sort_init_fini(ctx);
|
||||
|
||||
// Compute sizes of output sections while assigning offsets
|
||||
// within an output section to input sections.
|
||||
compute_section_sizes(ctx);
|
||||
|
||||
// Sort sections by section attributes so that we'll have to
|
||||
// create as few segments as possible.
|
||||
sort(ctx.chunks, [&](OutputChunk<E> *a, OutputChunk<E> *b) {
|
||||
return get_section_rank(ctx, a) < get_section_rank(ctx, b);
|
||||
});
|
||||
|
||||
// Copy string referred by .dynamic to .dynstr.
|
||||
for (SharedFile<E> *file : ctx.dsos)
|
||||
ctx.dynstr->add_string(file->soname);
|
||||
for (std::string_view str : ctx.arg.auxiliary)
|
||||
ctx.dynstr->add_string(str);
|
||||
for (std::string_view str : ctx.arg.filter)
|
||||
ctx.dynstr->add_string(str);
|
||||
if (!ctx.arg.rpaths.empty())
|
||||
ctx.dynstr->add_string(ctx.arg.rpaths);
|
||||
if (!ctx.arg.soname.empty())
|
||||
ctx.dynstr->add_string(ctx.arg.soname);
|
||||
|
||||
// Scan relocations to find symbols that need entries in .got, .plt,
|
||||
// .got.plt, .dynsym, .dynstr, etc.
|
||||
scan_rels(ctx);
|
||||
|
||||
// Reserve a space for dynamic symbol strings in .dynstr and sort
|
||||
// .dynsym contents if necessary. Beyond this point, no symbol will
|
||||
// be added to .dynsym.
|
||||
ctx.dynsym->finalize(ctx);
|
||||
|
||||
// Fill .gnu.version_d section contents.
|
||||
ctx.verdef->construct(ctx);
|
||||
|
||||
// Fill .gnu.version_r section contents.
|
||||
ctx.verneed->construct(ctx);
|
||||
|
||||
// Compute .symtab and .strtab sizes for each file.
|
||||
{
|
||||
Timer t(ctx, "compute_symtab");
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->compute_symtab(ctx);
|
||||
});
|
||||
}
|
||||
|
||||
// .eh_frame is a special section from the linker's point of view,
|
||||
// as its contents are parsed and reconstructed by the linker,
|
||||
// unlike other sections that are regarded as opaque bytes.
|
||||
// Here, we transplant .eh_frame sections from a regular output
|
||||
// section to the special EHFrameSection.
|
||||
{
|
||||
Timer t(ctx, "eh_frame");
|
||||
erase(ctx.chunks, [](OutputChunk<E> *chunk) {
|
||||
return chunk->kind == OutputChunk<E>::REGULAR &&
|
||||
chunk->name == ".eh_frame";
|
||||
});
|
||||
ctx.eh_frame->construct(ctx);
|
||||
}
|
||||
|
||||
// Now that we have computed sizes for all sections and assigned
|
||||
// section indices to them, so we can fix section header contents
|
||||
// for all output sections.
|
||||
for (OutputChunk<E> *chunk : ctx.chunks)
|
||||
chunk->update_shdr(ctx);
|
||||
|
||||
erase(ctx.chunks, [](OutputChunk<E> *chunk) {
|
||||
return chunk->kind == OutputChunk<E>::SYNTHETIC &&
|
||||
chunk->shdr.sh_size == 0;
|
||||
});
|
||||
|
||||
// Set section indices.
|
||||
for (i64 i = 0, shndx = 1; i < ctx.chunks.size(); i++)
|
||||
if (ctx.chunks[i]->kind != OutputChunk<E>::HEADER)
|
||||
ctx.chunks[i]->shndx = shndx++;
|
||||
|
||||
for (OutputChunk<E> *chunk : ctx.chunks)
|
||||
chunk->update_shdr(ctx);
|
||||
|
||||
// Assign offsets to output sections
|
||||
i64 filesize = set_osec_offsets(ctx);
|
||||
|
||||
// Fix linker-synthesized symbol addresses.
|
||||
fix_synthetic_symbols(ctx);
|
||||
|
||||
// If --compress-debug-sections is given, compress .debug_* sections
|
||||
// using zlib.
|
||||
if (ctx.arg.compress_debug_sections != COMPRESS_NONE) {
|
||||
compress_debug_sections(ctx);
|
||||
filesize = set_osec_offsets(ctx);
|
||||
}
|
||||
|
||||
// At this point, file layout is fixed.
|
||||
|
||||
// Beyond this, you can assume that symbol addresses including their
|
||||
// GOT or PLT addresses have a correct final value.
|
||||
|
||||
// Some types of relocations for TLS symbols need the TLS segment
|
||||
// address. Find it out now.
|
||||
for (ElfPhdr<E> phdr : create_phdr(ctx)) {
|
||||
if (phdr.p_type == PT_TLS) {
|
||||
ctx.tls_begin = phdr.p_vaddr;
|
||||
ctx.tls_end = align_to(phdr.p_vaddr + phdr.p_memsz, phdr.p_align);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
t_before_copy.stop();
|
||||
|
||||
// Create an output file
|
||||
ctx.output_file = OutputFile<E>::open(ctx, ctx.arg.output, filesize, 0777);
|
||||
ctx.buf = ctx.output_file->buf;
|
||||
|
||||
Timer t_copy(ctx, "copy");
|
||||
|
||||
// Copy input sections to the output file
|
||||
{
|
||||
Timer t(ctx, "copy_buf");
|
||||
|
||||
tbb::parallel_for_each(ctx.chunks, [&](OutputChunk<E> *chunk) {
|
||||
std::string name(chunk->name);
|
||||
if (name.empty())
|
||||
name = "(header)";
|
||||
Timer t2(ctx, name, &t);
|
||||
|
||||
chunk->copy_buf(ctx);
|
||||
});
|
||||
|
||||
Error<E>::checkpoint(ctx);
|
||||
}
|
||||
|
||||
// Dynamic linker works better with sorted .rela.dyn section,
|
||||
// so we sort them.
|
||||
ctx.reldyn->sort(ctx);
|
||||
|
||||
// Zero-clear paddings between sections
|
||||
clear_padding(ctx);
|
||||
|
||||
if (ctx.buildid) {
|
||||
Timer t(ctx, "build_id");
|
||||
ctx.buildid->write_buildid(ctx);
|
||||
}
|
||||
|
||||
t_copy.stop();
|
||||
|
||||
// Commit
|
||||
ctx.output_file->close(ctx);
|
||||
|
||||
t_total.stop();
|
||||
t_all.stop();
|
||||
|
||||
if (ctx.arg.print_map)
|
||||
print_map(ctx);
|
||||
|
||||
// Show stats numbers
|
||||
if (ctx.arg.stats)
|
||||
show_stats(ctx);
|
||||
|
||||
if (ctx.arg.perf)
|
||||
Timer<E>::print(ctx);
|
||||
|
||||
std::cout << std::flush;
|
||||
std::cerr << std::flush;
|
||||
if (on_complete)
|
||||
on_complete();
|
||||
|
||||
if (ctx.arg.quick_exit)
|
||||
_exit(0);
|
||||
|
||||
for (std::function<void()> &fn : ctx.on_exit)
|
||||
fn();
|
||||
return 0;
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void read_file(Context<E> &, MemoryMappedFile<E> *); \
|
||||
template std::string_view save_string(Context<E> &, const std::string &)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
return mold::elf::do_main<mold::elf::X86_64>(argc, argv);
|
||||
}
|
||||
+112
@@ -0,0 +1,112 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <fstream>
|
||||
#include <iomanip>
|
||||
#include <ios>
|
||||
#include <sstream>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <unordered_map>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
using Map =
|
||||
tbb::concurrent_hash_map<InputSection<E> *, std::vector<Symbol<E> *>>;
|
||||
|
||||
template <typename E>
|
||||
static std::unique_ptr<std::ofstream> open_output_file(Context<E> &ctx) {
|
||||
std::unique_ptr<std::ofstream> file(new std::ofstream);
|
||||
file->open(ctx.arg.Map.c_str());
|
||||
if (!file->is_open())
|
||||
Fatal(ctx) << "cannot open " << ctx.arg.Map << ": " << errno_string();
|
||||
return file;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static Map<E> get_map(Context<E> &ctx) {
|
||||
Map<E> map;
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->symbols) {
|
||||
if (sym->file == file && sym->input_section &&
|
||||
sym->get_type() != STT_SECTION) {
|
||||
ASSERT(file == &sym->input_section->file);
|
||||
|
||||
typename Map<E>::accessor acc;
|
||||
map.insert(acc, {sym->input_section, {}});
|
||||
acc->second.push_back(sym);
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
tbb::parallel_for(map.range(), [](const typename Map<E>::range_type &range) {
|
||||
for (auto it = range.begin(); it != range.end(); it++) {
|
||||
std::vector<Symbol<E> *> &vec = it->second;
|
||||
sort(vec, [](Symbol<E> *a, Symbol<E> *b) { return a->value < b->value; });
|
||||
}
|
||||
});
|
||||
return map;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void print_map(Context<E> &ctx) {
|
||||
|
||||
std::ostream *out = &std::cout;
|
||||
std::unique_ptr<std::ofstream> file;
|
||||
|
||||
if (!ctx.arg.Map.empty()) {
|
||||
file = open_output_file(ctx);
|
||||
out = file.get();
|
||||
}
|
||||
|
||||
// Construct a section-to-symbol map.
|
||||
Map<E> map = get_map(ctx);
|
||||
|
||||
// Print a mapfile.
|
||||
*out << " VMA Size Align Out In Symbol\n";
|
||||
|
||||
for (OutputChunk<E> *osec : ctx.chunks) {
|
||||
*out << std::setw(16) << (u64)osec->shdr.sh_addr
|
||||
<< std::setw(11) << (u64)osec->shdr.sh_size
|
||||
<< std::setw(6) << (u64)osec->shdr.sh_addralign
|
||||
<< " " << osec->name << "\n";
|
||||
|
||||
if (osec->kind != OutputChunk<E>::REGULAR)
|
||||
continue;
|
||||
|
||||
std::span<InputSection<E> *> members = ((OutputSection<E> *)osec)->members;
|
||||
std::vector<std::string> bufs(members.size());
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)members.size(), [&](i64 i) {
|
||||
InputSection<E> *mem = members[i];
|
||||
std::ostringstream ss;
|
||||
opt_demangle = ctx.arg.demangle;
|
||||
|
||||
ss << std::setw(16) << (osec->shdr.sh_addr + mem->offset)
|
||||
<< std::setw(11) << (u64)mem->shdr.sh_size
|
||||
<< std::setw(6) << (u64)mem->shdr.sh_addralign
|
||||
<< " " << *mem << "\n";
|
||||
|
||||
typename Map<E>::const_accessor acc;
|
||||
if (map.find(acc, mem))
|
||||
for (Symbol<E> *sym : acc->second)
|
||||
ss << std::setw(16) << sym->get_addr(ctx)
|
||||
<< " 0 0 "
|
||||
<< *sym << "\n";
|
||||
|
||||
bufs[i] = std::move(ss.str());
|
||||
});
|
||||
|
||||
for (std::string &str : bufs)
|
||||
*out << str;
|
||||
}
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void print_map(Context<E> &ctx);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,129 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <fcntl.h>
|
||||
#include <sys/mman.h>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/types.h>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
MemoryMappedFile<E> *
|
||||
MemoryMappedFile<E>::open(Context<E> &ctx, std::string path) {
|
||||
if (path.starts_with('/') && !ctx.arg.chroot.empty())
|
||||
path = ctx.arg.chroot + "/" + path_clean(path);
|
||||
|
||||
struct stat st;
|
||||
if (stat(path.c_str(), &st) == -1)
|
||||
return nullptr;
|
||||
|
||||
u64 mtime;
|
||||
#ifdef __APPLE__
|
||||
mtime = (u64)st.st_mtimespec.tv_sec * 1000000000 + st.st_mtimespec.tv_nsec;
|
||||
#else
|
||||
mtime = (u64)st.st_mtim.tv_sec * 1000000000 + st.st_mtim.tv_nsec;
|
||||
#endif
|
||||
|
||||
u8 *data = nullptr;
|
||||
if (st.st_size == 0) {
|
||||
// mmap(2) doesn't allow 0 as an argument for length parameter,
|
||||
// so this case has to be handled as a special case.
|
||||
static u8 dummy[1];
|
||||
data = dummy;
|
||||
}
|
||||
|
||||
MemoryMappedFile *mb = new MemoryMappedFile(path, data, st.st_size, mtime);
|
||||
ctx.owning_mbs.push_back(std::unique_ptr<MemoryMappedFile>(mb));
|
||||
return mb;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
MemoryMappedFile<E> *
|
||||
MemoryMappedFile<E>::must_open(Context<E> &ctx, std::string path) {
|
||||
if (MemoryMappedFile *mb = MemoryMappedFile::open(ctx, path))
|
||||
return mb;
|
||||
Fatal(ctx) << "cannot open " << path;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
u8 *MemoryMappedFile<E>::data(Context<E> &ctx) {
|
||||
if (data_)
|
||||
return data_;
|
||||
|
||||
std::lock_guard lock(mu);
|
||||
if (data_)
|
||||
return data_;
|
||||
|
||||
i64 fd = ::open(name.c_str(), O_RDONLY);
|
||||
if (fd == -1)
|
||||
Fatal(ctx) << name << ": cannot open: " << errno_string();
|
||||
|
||||
data_ = (u8 *)mmap(nullptr, size_, PROT_READ, MAP_PRIVATE, fd, 0);
|
||||
if (data_ == MAP_FAILED)
|
||||
Fatal(ctx) << name << ": mmap failed: " << errno_string();
|
||||
close(fd);
|
||||
return data_;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
MemoryMappedFile<E> *
|
||||
MemoryMappedFile<E>::slice(Context<E> &ctx, std::string name, u64 start,
|
||||
u64 size) {
|
||||
MemoryMappedFile *mb = new MemoryMappedFile<E>(name, data_ + start, size);
|
||||
ctx.owning_mbs.push_back(std::unique_ptr<MemoryMappedFile>(mb));
|
||||
mb->parent = this;
|
||||
return mb;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
MemoryMappedFile<E>::~MemoryMappedFile() {
|
||||
if (data_ && !parent)
|
||||
munmap(data_, size_);
|
||||
}
|
||||
|
||||
|
||||
template <typename E>
|
||||
static bool is_text_file(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
u8 *data = mb->data(ctx);
|
||||
return mb->size() >= 4 &&
|
||||
isprint(data[0]) &&
|
||||
isprint(data[1]) &&
|
||||
isprint(data[2]) &&
|
||||
isprint(data[3]);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
FileType get_file_type(Context<E> &ctx, MemoryMappedFile<E> *mb) {
|
||||
u8 *data = mb->data(ctx);
|
||||
|
||||
if (mb->size() >= 20 && memcmp(data, "\177ELF", 4) == 0) {
|
||||
ElfEhdr<E> &ehdr = *(ElfEhdr<E> *)data;
|
||||
if (ehdr.e_type == ET_REL)
|
||||
return FileType::OBJ;
|
||||
if (ehdr.e_type == ET_DYN)
|
||||
return FileType::DSO;
|
||||
return FileType::UNKNOWN;
|
||||
}
|
||||
|
||||
if (mb->size() >= 8 && memcmp(data, "!<arch>\n", 8) == 0)
|
||||
return FileType::AR;
|
||||
if (mb->size() >= 8 && memcmp(data, "!<thin>\n", 8) == 0)
|
||||
return FileType::THIN_AR;
|
||||
if (is_text_file(ctx, mb))
|
||||
return FileType::TEXT;
|
||||
if (mb->size() >= 4 && memcmp(data, "\xDE\xC0\x17\x0B", 4) == 0)
|
||||
return FileType::LLVM_BITCODE;
|
||||
if (mb->size() >= 4 && memcmp(data, "BC\xC0\xDE", 4) == 0)
|
||||
return FileType::LLVM_BITCODE;
|
||||
return FileType::UNKNOWN;
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template class MemoryMappedFile<E>; \
|
||||
template FileType get_file_type(Context<E> &, MemoryMappedFile<E> *)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -9,10 +9,6 @@
|
||||
#include <string.h>
|
||||
#include <unistd.h>
|
||||
|
||||
#if __has_include(<alloca.h>)
|
||||
# include <alloca.h>
|
||||
#endif
|
||||
|
||||
extern char **environ;
|
||||
|
||||
static char *get_mold_path() {
|
||||
@@ -23,7 +19,7 @@ static char *get_mold_path() {
|
||||
exit(1);
|
||||
}
|
||||
|
||||
static void debug_print(const char *fmt, ...) {
|
||||
static void debug_print(char *fmt, ...) {
|
||||
if (!getenv("MOLD_WRAPPER_DEBUG"))
|
||||
return;
|
||||
|
||||
@@ -35,25 +31,13 @@ static void debug_print(const char *fmt, ...) {
|
||||
va_end(ap);
|
||||
}
|
||||
|
||||
static int count_args(va_list *ap) {
|
||||
va_list aq;
|
||||
va_copy(aq, *ap);
|
||||
|
||||
int i = 0;
|
||||
while (va_arg(aq, char *))
|
||||
i++;
|
||||
va_end(aq);
|
||||
return i;
|
||||
}
|
||||
|
||||
static void copy_args(char **argv, const char *arg0, va_list *ap) {
|
||||
int i = 1;
|
||||
char *arg;
|
||||
while ((arg = va_arg(*ap, char *)))
|
||||
argv[i++] = arg;
|
||||
|
||||
((const char **)argv)[0] = arg0;
|
||||
((const char **)argv)[i] = NULL;
|
||||
static void get_args(va_list *ap, int argc, char **argv) {
|
||||
for (int i = 1; i < argc - 1; i++) {
|
||||
char *arg = va_arg(*ap, char *);
|
||||
if (!arg)
|
||||
break;
|
||||
argv[i] = arg;
|
||||
}
|
||||
}
|
||||
|
||||
static bool is_ld(const char *path) {
|
||||
@@ -62,15 +46,16 @@ static bool is_ld(const char *path) {
|
||||
ptr--;
|
||||
|
||||
return !strcmp(ptr, "ld") || !strcmp(ptr, "ld.lld") ||
|
||||
!strcmp(ptr, "ld.gold") || !strcmp(ptr, "ld.bfd") ||
|
||||
!strcmp(ptr, "ld.mold");
|
||||
!strcmp(ptr, "ld.gold");
|
||||
}
|
||||
|
||||
int execvpe(const char *file, char *const *argv, char *const *envp) {
|
||||
debug_print("execvpe %s\n", file);
|
||||
|
||||
if (!strcmp(file, "ld") || is_ld(file))
|
||||
if (!strcmp(file, "ld") || is_ld(file)) {
|
||||
file = get_mold_path();
|
||||
((const char **)argv)[0] = file;
|
||||
}
|
||||
|
||||
for (int i = 0; envp[i]; i++)
|
||||
putenv(envp[i]);
|
||||
@@ -81,8 +66,12 @@ int execvpe(const char *file, char *const *argv, char *const *envp) {
|
||||
|
||||
int execve(const char *path, char *const *argv, char *const *envp) {
|
||||
debug_print("execve %s\n", path);
|
||||
if (is_ld(path))
|
||||
|
||||
if (is_ld(path)) {
|
||||
path = get_mold_path();
|
||||
((const char **)argv)[0] = path;
|
||||
}
|
||||
|
||||
typeof(execve) *real = dlsym(RTLD_NEXT, "execve");
|
||||
return real(path, argv, envp);
|
||||
}
|
||||
@@ -90,28 +79,25 @@ int execve(const char *path, char *const *argv, char *const *envp) {
|
||||
int execl(const char *path, const char *arg0, ...) {
|
||||
va_list ap;
|
||||
va_start(ap, arg0);
|
||||
char **argv = alloca((count_args(&ap) + 2) * sizeof(char *));
|
||||
copy_args(argv, arg0, &ap);
|
||||
va_end(ap);
|
||||
char *argv[4096] = {(char *)arg0};
|
||||
get_args(&ap, 4096, argv);
|
||||
return execve(path, argv, environ);
|
||||
}
|
||||
|
||||
int execlp(const char *file, const char *arg0, ...) {
|
||||
va_list ap;
|
||||
va_start(ap, arg0);
|
||||
char **argv = alloca((count_args(&ap) + 2) * sizeof(char *));
|
||||
copy_args(argv, arg0, &ap);
|
||||
va_end(ap);
|
||||
char *argv[4096] = {(char *)arg0};
|
||||
get_args(&ap, 4096, argv);
|
||||
return execvpe(file, argv, environ);
|
||||
}
|
||||
|
||||
int execle(const char *path, const char *arg0, ...) {
|
||||
va_list ap;
|
||||
va_start(ap, arg0);
|
||||
char **argv = alloca((count_args(&ap) + 2) * sizeof(char *));
|
||||
copy_args(argv, arg0, &ap);
|
||||
char *argv[4096] = {(char *)arg0};
|
||||
get_args(&ap, 4096, argv);
|
||||
char **env = va_arg(ap, char **);
|
||||
va_end(ap);
|
||||
return execve(path, argv, env);
|
||||
}
|
||||
|
||||
@@ -128,19 +114,12 @@ int posix_spawn(pid_t *pid, const char *path,
|
||||
const posix_spawnattr_t *attrp,
|
||||
char *const *argv, char *const *envp) {
|
||||
debug_print("posix_spawn %s\n", path);
|
||||
if (is_ld(path))
|
||||
|
||||
if (is_ld(path)) {
|
||||
path = get_mold_path();
|
||||
((const char **)argv)[0] = path;
|
||||
}
|
||||
|
||||
typeof(posix_spawn) *real = dlsym(RTLD_NEXT, "posix_spawn");
|
||||
return real(pid, path, file_actions, attrp, argv, envp);
|
||||
}
|
||||
|
||||
int posix_spawnp(pid_t *pid, const char *file,
|
||||
const posix_spawn_file_actions_t *file_actions,
|
||||
const posix_spawnattr_t *attrp,
|
||||
char *const *argv, char *const *envp) {
|
||||
debug_print("posix_spawnp %s\n", file);
|
||||
if (is_ld(file))
|
||||
file = get_mold_path();
|
||||
typeof(posix_spawnp) *real = dlsym(RTLD_NEXT, "posix_spawnp");
|
||||
return real(pid, file, file_actions, attrp, argv, envp);
|
||||
}
|
||||
+2354
File diff suppressed because it is too large.
Load diff
+1439
File diff suppressed because it is too large.
Load diff
File diff suppressed because it is too large.
Load diff
@@ -0,0 +1,136 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <fcntl.h>
|
||||
#include <sys/mman.h>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/types.h>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static u32 get_umask() {
|
||||
u32 orig_umask = umask(0);
|
||||
umask(orig_umask);
|
||||
return orig_umask;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
class MemoryMappedOutputFile : public OutputFile<E> {
|
||||
public:
|
||||
MemoryMappedOutputFile(Context<E> &ctx, std::string path, i64 filesize,
|
||||
i64 perm)
|
||||
: OutputFile<E>(path, filesize, true) {
|
||||
std::string dir(path_dirname(path));
|
||||
this->tmpfile = (char *)save_string(ctx, dir + "/.mold-XXXXXX").data();
|
||||
i64 fd = mkstemp(this->tmpfile);
|
||||
if (fd == -1)
|
||||
Fatal(ctx) << "cannot open " << this->tmpfile << ": " << errno_string();
|
||||
|
||||
if (rename(path.c_str(), this->tmpfile) == 0) {
|
||||
::close(fd);
|
||||
fd = ::open(this->tmpfile, O_RDWR | O_CREAT, perm);
|
||||
if (fd == -1) {
|
||||
if (errno != ETXTBSY)
|
||||
Fatal(ctx) << "cannot open " << path << ": " << errno_string();
|
||||
unlink(this->tmpfile);
|
||||
fd = ::open(this->tmpfile, O_RDWR | O_CREAT, perm);
|
||||
if (fd == -1)
|
||||
Fatal(ctx) << "cannot open " << path << ": " << errno_string();
|
||||
}
|
||||
}
|
||||
|
||||
if (ftruncate(fd, filesize))
|
||||
Fatal(ctx) << "ftruncate failed";
|
||||
|
||||
if (fchmod(fd, (perm & ~get_umask())) == -1)
|
||||
Fatal(ctx) << "fchmod failed";
|
||||
|
||||
this->buf = (u8 *)mmap(nullptr, filesize, PROT_READ | PROT_WRITE,
|
||||
MAP_SHARED, fd, 0);
|
||||
if (this->buf == MAP_FAILED)
|
||||
Fatal(ctx) << path << ": mmap failed: " << errno_string();
|
||||
::close(fd);
|
||||
}
|
||||
|
||||
void close(Context<E> &ctx) override {
|
||||
Timer t(ctx, "close_file");
|
||||
|
||||
if (!ctx.buildid)
|
||||
munmap(this->buf, this->filesize);
|
||||
|
||||
if (rename(this->tmpfile, this->path.c_str()) == -1)
|
||||
Fatal(ctx) << this->path << ": rename failed: " << errno_string();
|
||||
this->tmpfile = nullptr;
|
||||
}
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class MallocOutputFile : public OutputFile<E> {
|
||||
public:
|
||||
MallocOutputFile(Context<E> &ctx, std::string path, i64 filesize, i64 perm)
|
||||
: OutputFile<E>(path, filesize, false), perm(perm) {
|
||||
this->buf = (u8 *)mmap(NULL, filesize, PROT_READ | PROT_WRITE,
|
||||
MAP_SHARED | MAP_ANONYMOUS, -1, 0);
|
||||
if (this->buf == MAP_FAILED)
|
||||
Fatal(ctx) << "mmap failed: " << errno_string();
|
||||
}
|
||||
|
||||
void close(Context<E> &ctx) override {
|
||||
Timer t(ctx, "close_file");
|
||||
|
||||
if (this->path == "-") {
|
||||
fwrite(this->buf, this->filesize, 1, stdout);
|
||||
fclose(stdout);
|
||||
return;
|
||||
}
|
||||
|
||||
i64 fd = ::open(this->path.c_str(), O_RDWR | O_CREAT, perm);
|
||||
if (fd == -1)
|
||||
Fatal(ctx) << "cannot open " << this->path << ": " << errno_string();
|
||||
|
||||
FILE *fp = fdopen(fd, "w");
|
||||
fwrite(this->buf, this->filesize, 1, fp);
|
||||
fclose(fp);
|
||||
}
|
||||
|
||||
private:
|
||||
i64 perm;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
std::unique_ptr<OutputFile<E>>
|
||||
OutputFile<E>::open(Context<E> &ctx, std::string path, i64 filesize, i64 perm) {
|
||||
Timer t(ctx, "open_file");
|
||||
|
||||
if (path.starts_with('/') && !ctx.arg.chroot.empty())
|
||||
path = ctx.arg.chroot + "/" + path_clean(path);
|
||||
|
||||
bool is_special = false;
|
||||
if (path == "-") {
|
||||
is_special = true;
|
||||
} else {
|
||||
struct stat st;
|
||||
if (stat(path.c_str(), &st) == 0 && (st.st_mode & S_IFMT) != S_IFREG)
|
||||
is_special = true;
|
||||
}
|
||||
|
||||
std::unique_ptr<OutputFile<E>> file;
|
||||
if (is_special)
|
||||
file = std::make_unique<MallocOutputFile<E>>(ctx, path, filesize, perm);
|
||||
else
|
||||
file = std::make_unique<MemoryMappedOutputFile<E>>(ctx, path, filesize, perm);
|
||||
|
||||
if (ctx.arg.filler != -1)
|
||||
memset(file->buf, ctx.arg.filler, filesize);
|
||||
return file;
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template \
|
||||
std::unique_ptr<OutputFile<E>> \
|
||||
OutputFile<E>::open(Context<E> &, std::string, i64, i64);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+986
@@ -0,0 +1,986 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <functional>
|
||||
#include <map>
|
||||
#include <regex>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <tbb/parallel_scan.h>
|
||||
#include <tbb/partitioner.h>
|
||||
#include <unordered_set>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E>
|
||||
void apply_exclude_libs(Context<E> &ctx) {
|
||||
Timer t(ctx, "apply_exclude_libs");
|
||||
|
||||
if (ctx.arg.exclude_libs.empty())
|
||||
return;
|
||||
|
||||
std::unordered_set<std::string_view> set(ctx.arg.exclude_libs.begin(),
|
||||
ctx.arg.exclude_libs.end());
|
||||
|
||||
for (ObjectFile<E> *file : ctx.objs)
|
||||
if (!file->archive_name.empty())
|
||||
if (set.contains("ALL") ||
|
||||
set.contains(path_filename(file->archive_name)))
|
||||
file->exclude_libs = true;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void create_synthetic_sections(Context<E> &ctx) {
|
||||
auto add = [&](auto &chunk) {
|
||||
ctx.chunks.push_back(chunk.get());
|
||||
};
|
||||
|
||||
add(ctx.ehdr = std::make_unique<OutputEhdr<E>>());
|
||||
add(ctx.phdr = std::make_unique<OutputPhdr<E>>());
|
||||
add(ctx.shdr = std::make_unique<OutputShdr<E>>());
|
||||
add(ctx.got = std::make_unique<GotSection<E>>());
|
||||
add(ctx.gotplt = std::make_unique<GotPltSection<E>>());
|
||||
add(ctx.relplt = std::make_unique<RelPltSection<E>>());
|
||||
add(ctx.strtab = std::make_unique<StrtabSection<E>>());
|
||||
add(ctx.shstrtab = std::make_unique<ShstrtabSection<E>>());
|
||||
add(ctx.plt = std::make_unique<PltSection<E>>());
|
||||
add(ctx.pltgot = std::make_unique<PltGotSection<E>>());
|
||||
add(ctx.symtab = std::make_unique<SymtabSection<E>>());
|
||||
add(ctx.dynsym = std::make_unique<DynsymSection<E>>());
|
||||
add(ctx.dynstr = std::make_unique<DynstrSection<E>>());
|
||||
add(ctx.eh_frame = std::make_unique<EhFrameSection<E>>());
|
||||
add(ctx.dynbss = std::make_unique<DynbssSection<E>>(false));
|
||||
add(ctx.dynbss_relro = std::make_unique<DynbssSection<E>>(true));
|
||||
|
||||
if (!ctx.arg.dynamic_linker.empty())
|
||||
add(ctx.interp = std::make_unique<InterpSection<E>>());
|
||||
if (ctx.arg.build_id.kind != BuildId::NONE)
|
||||
add(ctx.buildid = std::make_unique<BuildIdSection<E>>());
|
||||
if (ctx.arg.eh_frame_hdr)
|
||||
add(ctx.eh_frame_hdr = std::make_unique<EhFrameHdrSection<E>>());
|
||||
if (ctx.arg.hash_style_sysv)
|
||||
add(ctx.hash = std::make_unique<HashSection<E>>());
|
||||
if (ctx.arg.hash_style_gnu)
|
||||
add(ctx.gnu_hash = std::make_unique<GnuHashSection<E>>());
|
||||
if (!ctx.arg.version_definitions.empty())
|
||||
add(ctx.verdef = std::make_unique<VerdefSection<E>>());
|
||||
|
||||
add(ctx.reldyn = std::make_unique<RelDynSection<E>>());
|
||||
add(ctx.dynamic = std::make_unique<DynamicSection<E>>());
|
||||
add(ctx.versym = std::make_unique<VersymSection<E>>());
|
||||
add(ctx.verneed = std::make_unique<VerneedSection<E>>());
|
||||
add(ctx.note_property = std::make_unique<NotePropertySection<E>>());
|
||||
|
||||
if (ctx.arg.repro)
|
||||
add(ctx.repro = std::make_unique<ReproSection<E>>());
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void resolve_symbols(Context<E> &ctx) {
|
||||
Timer t(ctx, "resolve_obj_symbols");
|
||||
|
||||
// Register object symbols
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
if (file->is_in_lib)
|
||||
file->resolve_lazy_symbols(ctx);
|
||||
else
|
||||
file->resolve_regular_symbols(ctx);
|
||||
});
|
||||
|
||||
// Register DSO symbols
|
||||
tbb::parallel_for_each(ctx.dsos, [&](SharedFile<E> *file) {
|
||||
file->resolve_dso_symbols(ctx);
|
||||
});
|
||||
|
||||
// Mark reachable objects to decide which files to include
|
||||
// into an output.
|
||||
std::vector<ObjectFile<E> *> live_objs = ctx.objs;
|
||||
erase(live_objs, [](InputFile<E> *file) { return !file->is_alive; });
|
||||
|
||||
for (std::string_view name : ctx.arg.undefined)
|
||||
if (InputFile<E> *file = Symbol<E>::intern(ctx, name)->file)
|
||||
if (!file->is_alive.exchange(true) && !file->is_dso)
|
||||
live_objs.push_back((ObjectFile<E> *)file);
|
||||
|
||||
tbb::parallel_for_each(live_objs,
|
||||
[&](ObjectFile<E> *file,
|
||||
tbb::feeder<ObjectFile<E> *> &feeder) {
|
||||
file->mark_live_objects(ctx, [&](ObjectFile<E> *obj) { feeder.add(obj); });
|
||||
});
|
||||
|
||||
// Remove symbols of eliminated objects.
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
if (!file->is_alive)
|
||||
for (Symbol<E> *sym : file->get_global_syms())
|
||||
if (sym->file == file)
|
||||
new (sym) Symbol<E>(sym->name());
|
||||
});
|
||||
|
||||
// Eliminate unused archive members.
|
||||
erase(ctx.objs, [](InputFile<E> *file) { return !file->is_alive; });
|
||||
|
||||
// Mark live DSOs
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
for (i64 i = file->first_global; i < file->elf_syms.size(); i++) {
|
||||
const ElfSym<E> &esym = file->elf_syms[i];
|
||||
Symbol<E> &sym = *file->symbols[i];
|
||||
if (esym.is_undef_strong() && sym.file && sym.file->is_dso) {
|
||||
std::lock_guard lock(sym.mu);
|
||||
sym.file->is_alive = true;
|
||||
sym.is_weak = false;
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
// DSOs referenced by live DSOs are also alive.
|
||||
std::vector<SharedFile<E> *> live_dsos = ctx.dsos;
|
||||
erase(live_dsos, [](SharedFile<E> *file) { return !file->is_alive; });
|
||||
|
||||
tbb::parallel_for_each(live_dsos,
|
||||
[&](SharedFile<E> *file,
|
||||
tbb::feeder<SharedFile<E> *> &feeder) {
|
||||
for (Symbol<E> *sym : file->globals)
|
||||
if (sym->file && sym->file != file && sym->file->is_dso &&
|
||||
!sym->file->is_alive.exchange(true))
|
||||
feeder.add(file);
|
||||
});
|
||||
|
||||
// Remove symbols of unreferenced DSOs.
|
||||
tbb::parallel_for_each(ctx.dsos, [](SharedFile<E> *file) {
|
||||
if (!file->is_alive)
|
||||
for (Symbol<E> *sym : file->symbols)
|
||||
if (sym->file == file)
|
||||
new (sym) Symbol<E>(sym->name());
|
||||
});
|
||||
|
||||
// Remove unreferenced DSOs
|
||||
erase(ctx.dsos, [](InputFile<E> *file) { return !file->is_alive; });
|
||||
|
||||
// Register common symbols
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->resolve_common_symbols(ctx);
|
||||
});
|
||||
|
||||
if (Symbol<E> *sym = Symbol<E>::intern(ctx, "__gnu_lto_slim"); sym->file)
|
||||
Fatal(ctx) << *sym->file << ": looks like this file contains a GCC "
|
||||
<< "intermediate code, but mold does not support LTO";
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void eliminate_comdats(Context<E> &ctx) {
|
||||
Timer t(ctx, "eliminate_comdats");
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
file->resolve_comdat_groups();
|
||||
});
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
file->eliminate_duplicate_comdat_groups();
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void convert_common_symbols(Context<E> &ctx) {
|
||||
Timer t(ctx, "convert_common_symbols");
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->convert_common_symbols(ctx);
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::string get_cmdline_args(Context<E> &ctx) {
|
||||
std::stringstream ss;
|
||||
ss << ctx.cmdline_args[1];
|
||||
for (i64 i = 2; i < ctx.cmdline_args.size(); i++)
|
||||
ss << " " << ctx.cmdline_args[i];
|
||||
return ss.str();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void add_comment_string(Context<E> &ctx, std::string str) {
|
||||
std::string_view buf = save_string(ctx, str);
|
||||
MergedSection<E> *sec =
|
||||
MergedSection<E>::get_instance(ctx, ".comment", SHT_PROGBITS, 0);
|
||||
std::string_view data(buf.data(), buf.size() + 1);
|
||||
SectionFragment<E> *frag = sec->insert(data, hash_string(data), 1);
|
||||
frag->is_alive = true;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void compute_merged_section_sizes(Context<E> &ctx) {
|
||||
Timer t(ctx, "compute_merged_section_sizes");
|
||||
|
||||
// Mark section fragments referenced by live objects.
|
||||
if (!ctx.arg.gc_sections) {
|
||||
tbb::parallel_for_each(ctx.objs, [](ObjectFile<E> *file) {
|
||||
for (SectionFragment<E> *frag : file->fragments)
|
||||
frag->is_alive.store(true, std::memory_order_relaxed);
|
||||
});
|
||||
}
|
||||
|
||||
// Add an identification string to .comment.
|
||||
add_comment_string(ctx, get_version_string());
|
||||
|
||||
// Embed command line arguments for debugging.
|
||||
if (char *env = getenv("MOLD_DEBUG"); env && env[0])
|
||||
add_comment_string(ctx, "mold command line: " + get_cmdline_args(ctx));
|
||||
|
||||
Timer t2(ctx, "MergedSection assign_offsets");
|
||||
tbb::parallel_for_each(ctx.merged_sections,
|
||||
[&](std::unique_ptr<MergedSection<E>> &sec) {
|
||||
sec->assign_offsets(ctx);
|
||||
});
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
static std::vector<std::span<T>> split(std::vector<T> &input, i64 unit) {
|
||||
ASSERT(input.size() > 0);
|
||||
std::span<T> span(input);
|
||||
std::vector<std::span<T>> vec;
|
||||
|
||||
while (span.size() >= unit) {
|
||||
vec.push_back(span.subspan(0, unit));
|
||||
span = span.subspan(unit);
|
||||
}
|
||||
if (!span.empty())
|
||||
vec.push_back(span);
|
||||
return vec;
|
||||
}
|
||||
|
||||
// So far, each input section has a pointer to its corresponding
|
||||
// output section, but there's no reverse edge to get a list of
|
||||
// input sections from an output section. This function creates it.
|
||||
//
|
||||
// An output section may contain millions of input sections.
|
||||
// So, we append input sections to output sections in parallel.
|
||||
template <typename E>
|
||||
void bin_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "bin_sections");
|
||||
|
||||
static constexpr i64 num_shards = 128;
|
||||
i64 unit = (ctx.objs.size() + num_shards - 1) / num_shards;
|
||||
std::vector<std::span<ObjectFile<E> *>> slices = split(ctx.objs, unit);
|
||||
|
||||
i64 num_osec = ctx.output_sections.size();
|
||||
|
||||
std::vector<std::vector<std::vector<InputSection<E> *>>> groups(slices.size());
|
||||
for (i64 i = 0; i < groups.size(); i++)
|
||||
groups[i].resize(num_osec);
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)slices.size(), [&](i64 i) {
|
||||
for (ObjectFile<E> *file : slices[i])
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections)
|
||||
if (isec && isec->is_alive)
|
||||
groups[i][isec->output_section->idx].push_back(isec.get());
|
||||
});
|
||||
|
||||
std::vector<i64> sizes(num_osec);
|
||||
|
||||
for (std::span<std::vector<InputSection<E> *>> group : groups)
|
||||
for (i64 i = 0; i < group.size(); i++)
|
||||
sizes[i] += group[i].size();
|
||||
|
||||
tbb::parallel_for((i64)0, num_osec, [&](i64 j) {
|
||||
ctx.output_sections[j]->members.reserve(sizes[j]);
|
||||
for (i64 i = 0; i < groups.size(); i++)
|
||||
append(ctx.output_sections[j]->members, groups[i][j]);
|
||||
});
|
||||
}
|
||||
|
||||
// Create a dummy object file containing linker-synthesized
|
||||
// symbols.
|
||||
template <typename E>
|
||||
ObjectFile<E> *create_internal_file(Context<E> &ctx) {
|
||||
ObjectFile<E> *obj = new ObjectFile<E>;
|
||||
ctx.owning_objs.push_back(std::unique_ptr<ObjectFile<E>>(obj));
|
||||
|
||||
// Create linker-synthesized symbols.
|
||||
auto *esyms = new std::vector<ElfSym<E>>(1);
|
||||
obj->symbols.push_back(new Symbol<E>);
|
||||
obj->first_global = 1;
|
||||
obj->is_alive = true;
|
||||
obj->priority = 1;
|
||||
|
||||
auto add = [&](std::string_view name) {
|
||||
ElfSym<E> esym = {};
|
||||
esym.st_type = STT_NOTYPE;
|
||||
esym.st_shndx = SHN_ABS;
|
||||
esym.st_bind = STB_GLOBAL;
|
||||
esym.st_visibility = STV_HIDDEN;
|
||||
esyms->push_back(esym);
|
||||
|
||||
Symbol<E> *sym = Symbol<E>::intern(ctx, name);
|
||||
obj->symbols.push_back(sym);
|
||||
return sym;
|
||||
};
|
||||
|
||||
ctx.__ehdr_start = add("__ehdr_start");
|
||||
ctx.__init_array_start = add("__init_array_start");
|
||||
ctx.__init_array_end = add("__init_array_end");
|
||||
ctx.__fini_array_start = add("__fini_array_start");
|
||||
ctx.__fini_array_end = add("__fini_array_end");
|
||||
ctx.__preinit_array_start = add("__preinit_array_start");
|
||||
ctx.__preinit_array_end = add("__preinit_array_end");
|
||||
ctx._DYNAMIC = add("_DYNAMIC");
|
||||
ctx._GLOBAL_OFFSET_TABLE_ = add("_GLOBAL_OFFSET_TABLE_");
|
||||
ctx.__bss_start = add("__bss_start");
|
||||
ctx._end = add("_end");
|
||||
ctx._etext = add("_etext");
|
||||
ctx._edata = add("_edata");
|
||||
ctx.__executable_start = add("__executable_start");
|
||||
|
||||
ctx.__rel_iplt_start =
|
||||
add(E::is_rel ? "__rel_iplt_start" : "__rela_iplt_start");
|
||||
ctx.__rel_iplt_end =
|
||||
add(E::is_rel ? "__rel_iplt_end" : "__rela_iplt_end");
|
||||
|
||||
if (ctx.arg.eh_frame_hdr)
|
||||
ctx.__GNU_EH_FRAME_HDR = add("__GNU_EH_FRAME_HDR");
|
||||
|
||||
if (!Symbol<E>::intern(ctx, "end")->file)
|
||||
ctx.end = add("end");
|
||||
if (!Symbol<E>::intern(ctx, "etext")->file)
|
||||
ctx.etext = add("etext");
|
||||
if (!Symbol<E>::intern(ctx, "edata")->file)
|
||||
ctx.edata = add("edata");
|
||||
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
if (!is_c_identifier(chunk->name))
|
||||
continue;
|
||||
|
||||
add(save_string(ctx, "__start_" + std::string(chunk->name)));
|
||||
add(save_string(ctx, "__stop_" + std::string(chunk->name)));
|
||||
}
|
||||
|
||||
obj->elf_syms = *esyms;
|
||||
obj->sym_fragments.resize(obj->elf_syms.size());
|
||||
|
||||
i64 num_globals = obj->elf_syms.size() - obj->first_global;
|
||||
obj->symvers.resize(num_globals);
|
||||
|
||||
ctx.on_exit.push_back([=]() {
|
||||
delete esyms;
|
||||
delete obj->symbols[0];
|
||||
});
|
||||
|
||||
return obj;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void check_duplicate_symbols(Context<E> &ctx) {
|
||||
Timer t(ctx, "check_dup_syms");
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (i64 i = file->first_global; i < file->elf_syms.size(); i++) {
|
||||
const ElfSym<E> &esym = file->elf_syms[i];
|
||||
Symbol<E> &sym = *file->symbols[i];
|
||||
|
||||
if (sym.file == file || sym.file == ctx.internal_obj ||
|
||||
esym.is_undef() || esym.is_common() || (esym.st_bind == STB_WEAK))
|
||||
continue;
|
||||
|
||||
if (!esym.is_abs() && !file->get_section(esym)->is_alive)
|
||||
continue;
|
||||
|
||||
Error(ctx) << "duplicate symbol: " << *file << ": " << *sym.file
|
||||
<< ": " << sym;
|
||||
}
|
||||
});
|
||||
|
||||
Error<E>::checkpoint(ctx);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void sort_init_fini(Context<E> &ctx) {
|
||||
Timer t(ctx, "sort_init_fini");
|
||||
|
||||
auto get_priority = [](InputSection<E> *isec) {
|
||||
static std::regex re(R"(_array\.(\d+)$)", std::regex_constants::optimize);
|
||||
std::string name = isec->name().begin();
|
||||
std::smatch m;
|
||||
if (std::regex_search(name, m, re))
|
||||
return std::stoi(m[1]);
|
||||
return 65536;
|
||||
};
|
||||
|
||||
for (std::unique_ptr<OutputSection<E>> &osec : ctx.output_sections) {
|
||||
if (osec->name == ".init_array" || osec->name == ".fini_array") {
|
||||
sort(osec->members, [&](InputSection<E> *a, InputSection<E> *b) {
|
||||
return get_priority(a) < get_priority(b);
|
||||
});
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<OutputChunk<E> *> collect_output_sections(Context<E> &ctx) {
|
||||
std::vector<OutputChunk<E> *> vec;
|
||||
|
||||
for (std::unique_ptr<OutputSection<E>> &osec : ctx.output_sections)
|
||||
if (!osec->members.empty())
|
||||
vec.push_back(osec.get());
|
||||
for (std::unique_ptr<MergedSection<E>> &osec : ctx.merged_sections)
|
||||
if (osec->shdr.sh_size)
|
||||
vec.push_back(osec.get());
|
||||
|
||||
// Sections are added to the section lists in an arbitrary order because
|
||||
// they are created in parallel.
|
||||
// Sort them to to make the output deterministic.
|
||||
sort(vec, [](OutputChunk<E> *x, OutputChunk<E> *y) {
|
||||
return std::tuple(x->name, x->shdr.sh_type, x->shdr.sh_flags) <
|
||||
std::tuple(y->name, y->shdr.sh_type, y->shdr.sh_flags);
|
||||
});
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void compute_section_sizes(Context<E> &ctx) {
|
||||
Timer t(ctx, "compute_section_sizes");
|
||||
|
||||
tbb::parallel_for_each(ctx.output_sections,
|
||||
[&](std::unique_ptr<OutputSection<E>> &osec) {
|
||||
if (osec->members.empty())
|
||||
return;
|
||||
|
||||
struct T {
|
||||
i64 offset;
|
||||
i64 align;
|
||||
};
|
||||
|
||||
T sum = tbb::parallel_scan(
|
||||
tbb::blocked_range<i64>(0, osec->members.size(), 10000),
|
||||
T{0, 1},
|
||||
[&](const tbb::blocked_range<i64> &r, T sum, bool is_final) {
|
||||
for (i64 i = r.begin(); i < r.end(); i++) {
|
||||
InputSection<E> &isec = *osec->members[i];
|
||||
sum.offset = align_to(sum.offset, isec.shdr.sh_addralign);
|
||||
if (is_final)
|
||||
isec.offset = sum.offset;
|
||||
sum.offset += isec.shdr.sh_size;
|
||||
sum.align = std::max<i64>(sum.align, isec.shdr.sh_addralign);
|
||||
}
|
||||
return sum;
|
||||
},
|
||||
[](T lhs, T rhs) {
|
||||
i64 offset = align_to(lhs.offset, rhs.align) + rhs.offset;
|
||||
i64 align = std::max(lhs.align, rhs.align);
|
||||
return T{offset, align};
|
||||
},
|
||||
tbb::simple_partitioner());
|
||||
|
||||
osec->shdr.sh_size = sum.offset;
|
||||
osec->shdr.sh_addralign = sum.align;
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void claim_unresolved_symbols(Context<E> &ctx) {
|
||||
Timer t(ctx, "claim_unresolved_symbols");
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->claim_unresolved_symbols(ctx);
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void scan_rels(Context<E> &ctx) {
|
||||
Timer t(ctx, "scan_rels");
|
||||
|
||||
// Scan relocations to find dynamic symbols.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
file->scan_relocations(ctx);
|
||||
});
|
||||
|
||||
// Exit if there was a relocation that refers an undefined symbol.
|
||||
Error<E>::checkpoint(ctx);
|
||||
|
||||
// Add symbol aliases for COPYREL.
|
||||
tbb::parallel_for_each(ctx.dsos, [&](SharedFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->symbols)
|
||||
if (sym->file == file && (sym->flags & NEEDS_COPYREL))
|
||||
for (Symbol<E> *alias : file->find_aliases(sym))
|
||||
alias->flags |= NEEDS_DYNSYM;
|
||||
});
|
||||
|
||||
// Aggregate dynamic symbols to a single vector.
|
||||
std::vector<InputFile<E> *> files;
|
||||
append(files, ctx.objs);
|
||||
append(files, ctx.dsos);
|
||||
|
||||
std::vector<std::vector<Symbol<E> *>> vec(files.size());
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)files.size(), [&](i64 i) {
|
||||
for (Symbol<E> *sym : files[i]->symbols) {
|
||||
if (!files[i]->is_dso && (sym->is_imported || sym->is_exported))
|
||||
sym->flags |= NEEDS_DYNSYM;
|
||||
if (sym->file == files[i] && sym->flags)
|
||||
vec[i].push_back(sym);
|
||||
}
|
||||
});
|
||||
|
||||
std::vector<Symbol<E> *> syms = flatten(vec);
|
||||
|
||||
ctx.symbol_aux.resize(syms.size());
|
||||
for (i64 i = 0; i < syms.size(); i++)
|
||||
syms[i]->aux_idx = i;
|
||||
|
||||
// Assign offsets in additional tables for each dynamic symbol.
|
||||
for (Symbol<E> *sym : syms) {
|
||||
if (sym->flags & NEEDS_DYNSYM)
|
||||
ctx.dynsym->add_symbol(ctx, sym);
|
||||
|
||||
if (sym->flags & NEEDS_GOT)
|
||||
ctx.got->add_got_symbol(ctx, sym);
|
||||
|
||||
if (sym->flags & NEEDS_PLT) {
|
||||
// If we need to create a canonical PLT (i.e. ctx.arg.pic is false),
|
||||
// we can't use .plt.got because otherwise .plt.got and .plt would
|
||||
// refer each other, resulting in an infinite loop at runtime.
|
||||
if ((sym->flags & NEEDS_GOT) && !ctx.arg.pic)
|
||||
ctx.pltgot->add_symbol(ctx, sym);
|
||||
else
|
||||
ctx.plt->add_symbol(ctx, sym);
|
||||
}
|
||||
|
||||
if (sym->flags & NEEDS_GOTTP)
|
||||
ctx.got->add_gottp_symbol(ctx, sym);
|
||||
|
||||
if (sym->flags & NEEDS_TLSGD)
|
||||
ctx.got->add_tlsgd_symbol(ctx, sym);
|
||||
|
||||
if (sym->flags & NEEDS_TLSDESC)
|
||||
ctx.got->add_tlsdesc_symbol(ctx, sym);
|
||||
|
||||
if (sym->flags & NEEDS_TLSLD)
|
||||
ctx.got->add_tlsld(ctx);
|
||||
|
||||
if (sym->flags & NEEDS_COPYREL) {
|
||||
ASSERT(sym->file->is_dso);
|
||||
SharedFile<E> *file = (SharedFile<E> *)sym->file;
|
||||
sym->copyrel_readonly = file->is_readonly(ctx, sym);
|
||||
|
||||
if (sym->copyrel_readonly)
|
||||
ctx.dynbss_relro->add_symbol(ctx, sym);
|
||||
else
|
||||
ctx.dynbss->add_symbol(ctx, sym);
|
||||
|
||||
for (Symbol<E> *alias : file->find_aliases(sym)) {
|
||||
alias->has_copyrel = true;
|
||||
alias->value = sym->value;
|
||||
alias->copyrel_readonly = sym->copyrel_readonly;
|
||||
ctx.dynsym->add_symbol(ctx, alias);
|
||||
}
|
||||
}
|
||||
|
||||
sym->flags = 0;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void apply_version_script(Context<E> &ctx) {
|
||||
Timer t(ctx, "apply_version_script");
|
||||
|
||||
for (VersionPattern &elem : ctx.arg.version_patterns) {
|
||||
ASSERT(elem.pattern != "*");
|
||||
|
||||
if (!elem.is_extern_cpp &&
|
||||
elem.pattern.find('*') == elem.pattern.npos) {
|
||||
Symbol<E> *sym = Symbol<E>::intern(ctx, elem.pattern);
|
||||
if (sym->file && !sym->file->is_dso)
|
||||
sym->ver_idx = elem.ver_idx;
|
||||
continue;
|
||||
}
|
||||
|
||||
std::regex re = glob_to_regex(elem.pattern);
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->get_global_syms()) {
|
||||
if (sym->file == file) {
|
||||
std::string_view name = elem.is_extern_cpp
|
||||
? sym->get_demangled_name() : sym->name();
|
||||
if (std::regex_match(name.begin(), name.end(), re))
|
||||
sym->ver_idx = elem.ver_idx;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void parse_symbol_version(Context<E> &ctx) {
|
||||
if (!ctx.arg.shared)
|
||||
return;
|
||||
|
||||
Timer t(ctx, "parse_symbol_version");
|
||||
|
||||
std::unordered_map<std::string_view, u16> verdefs;
|
||||
for (i64 i = 0; i < ctx.arg.version_definitions.size(); i++)
|
||||
verdefs[ctx.arg.version_definitions[i]] = i + VER_NDX_LAST_RESERVED + 1;
|
||||
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (i64 i = 0; i < file->symbols.size() - file->first_global; i++) {
|
||||
if (!file->symvers[i])
|
||||
continue;
|
||||
|
||||
Symbol<E> *sym = file->symbols[i + file->first_global];
|
||||
if (sym->file != file)
|
||||
continue;
|
||||
|
||||
std::string_view ver = file->symvers[i];
|
||||
|
||||
bool is_default = false;
|
||||
if (ver.starts_with('@')) {
|
||||
is_default = true;
|
||||
ver = ver.substr(1);
|
||||
}
|
||||
|
||||
auto it = verdefs.find(ver);
|
||||
if (it == verdefs.end()) {
|
||||
Error(ctx) << *file << ": symbol " << *sym << " has undefined version "
|
||||
<< ver;
|
||||
continue;
|
||||
}
|
||||
|
||||
sym->ver_idx = it->second;
|
||||
if (!is_default)
|
||||
sym->ver_idx |= VERSYM_HIDDEN;
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void compute_import_export(Context<E> &ctx) {
|
||||
Timer t(ctx, "compute_import_export");
|
||||
|
||||
// Export symbols referenced by DSOs.
|
||||
if (!ctx.arg.shared) {
|
||||
tbb::parallel_for_each(ctx.dsos, [&](SharedFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->globals) {
|
||||
if (sym->file && !sym->file->is_dso && sym->visibility != STV_HIDDEN) {
|
||||
std::lock_guard lock(sym->mu);
|
||||
sym->is_exported = true;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// Global symbols are exported from DSO by default.
|
||||
if (ctx.arg.shared || ctx.arg.export_dynamic) {
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->get_global_syms()) {
|
||||
if (sym->file != file)
|
||||
continue;
|
||||
|
||||
if (sym->visibility == STV_HIDDEN || sym->ver_idx == VER_NDX_LOCAL)
|
||||
continue;
|
||||
|
||||
sym->is_exported = true;
|
||||
|
||||
if (ctx.arg.shared && sym->visibility != STV_PROTECTED &&
|
||||
!ctx.arg.Bsymbolic &&
|
||||
!(ctx.arg.Bsymbolic_functions && sym->get_type() == STT_FUNC))
|
||||
sym->is_imported = true;
|
||||
}
|
||||
});
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void clear_padding(Context<E> &ctx) {
|
||||
Timer t(ctx, "clear_padding");
|
||||
|
||||
auto zero = [&](OutputChunk<E> *chunk, i64 next_start) {
|
||||
i64 pos = chunk->shdr.sh_offset;
|
||||
if (chunk->shdr.sh_type != SHT_NOBITS)
|
||||
pos += chunk->shdr.sh_size;
|
||||
memset(ctx.buf + pos, 0, next_start - pos);
|
||||
};
|
||||
|
||||
for (i64 i = 1; i < ctx.chunks.size(); i++)
|
||||
zero(ctx.chunks[i - 1], ctx.chunks[i]->shdr.sh_offset);
|
||||
zero(ctx.chunks.back(), ctx.output_file->filesize);
|
||||
}
|
||||
|
||||
// We want to sort output chunks in the following order.
|
||||
//
|
||||
// ELF header
|
||||
// program header
|
||||
// .interp
|
||||
// note
|
||||
// alloc readonly data
|
||||
// alloc readonly code
|
||||
// alloc writable tdata
|
||||
// alloc writable tbss
|
||||
// alloc writable RELRO data
|
||||
// alloc writable RELRO bss
|
||||
// alloc writable non-RELRO data
|
||||
// alloc writable non-RELRO bss
|
||||
// nonalloc
|
||||
// section header
|
||||
template <typename E>
|
||||
i64 get_section_rank(Context<E> &ctx, OutputChunk<E> *chunk) {
|
||||
u64 type = chunk->shdr.sh_type;
|
||||
u64 flags = chunk->shdr.sh_flags;
|
||||
|
||||
if (chunk == ctx.ehdr.get())
|
||||
return -4;
|
||||
if (chunk == ctx.phdr.get())
|
||||
return -3;
|
||||
if (chunk == ctx.interp.get())
|
||||
return -2;
|
||||
if (type == SHT_NOTE && (flags & SHF_ALLOC))
|
||||
return -1;
|
||||
if (chunk == ctx.shdr.get())
|
||||
return 1 << 6;
|
||||
if (!(flags & SHF_ALLOC))
|
||||
return 1 << 5;
|
||||
|
||||
bool writable = (flags & SHF_WRITE);
|
||||
bool exec = (flags & SHF_EXECINSTR);
|
||||
bool tls = (flags & SHF_TLS);
|
||||
bool relro = is_relro(ctx, chunk);
|
||||
bool is_bss = (type == SHT_NOBITS);
|
||||
|
||||
return (writable << 4) | (exec << 3) | (!tls << 2) |
|
||||
(!relro << 1) | is_bss;
|
||||
}
|
||||
|
||||
// Returns the smallest number n such that
|
||||
// n >= val and n % align == skew.
|
||||
inline u64 align_with_skew(u64 val, u64 align, u64 skew) {
|
||||
return align_to(val + align - skew, align) - align + skew;
|
||||
}
|
||||
|
||||
// Assign virtual addresses and file offsets to output sections.
|
||||
template <typename E>
|
||||
i64 set_osec_offsets(Context<E> &ctx) {
|
||||
Timer t(ctx, "osec_offset");
|
||||
|
||||
u64 fileoff = 0;
|
||||
u64 vaddr = ctx.arg.image_base;
|
||||
|
||||
i64 i = 0;
|
||||
i64 end = 0;
|
||||
while (ctx.chunks[end]->shdr.sh_flags & SHF_ALLOC)
|
||||
end++;
|
||||
|
||||
while (i < end) {
|
||||
fileoff = align_with_skew(fileoff, COMMON_PAGE_SIZE, vaddr % COMMON_PAGE_SIZE);
|
||||
|
||||
// Each group consists of zero or more non-BSS sections followed
|
||||
// by zero or more BSS sections. Virtual addresses of non-BSS
|
||||
// sections need to be congruent to file offsets modulo the page size.
|
||||
// BSS sections don't increment file offsets.
|
||||
for (; i < end && ctx.chunks[i]->shdr.sh_type != SHT_NOBITS; i++) {
|
||||
OutputChunk<E> &chunk = *ctx.chunks[i];
|
||||
u64 prev_vaddr = vaddr;
|
||||
|
||||
if (chunk.new_page)
|
||||
vaddr = align_to(vaddr, COMMON_PAGE_SIZE);
|
||||
vaddr = align_to(vaddr, chunk.shdr.sh_addralign);
|
||||
fileoff += vaddr - prev_vaddr;
|
||||
|
||||
chunk.shdr.sh_addr = vaddr;
|
||||
vaddr += chunk.shdr.sh_size;
|
||||
|
||||
chunk.shdr.sh_offset = fileoff;
|
||||
fileoff += chunk.shdr.sh_size;
|
||||
}
|
||||
|
||||
for (; i < end && ctx.chunks[i]->shdr.sh_type == SHT_NOBITS; i++) {
|
||||
OutputChunk<E> &chunk = *ctx.chunks[i];
|
||||
|
||||
if (chunk.new_page)
|
||||
vaddr = align_to(vaddr, COMMON_PAGE_SIZE);
|
||||
vaddr = align_to(vaddr, chunk.shdr.sh_addralign);
|
||||
fileoff = align_with_skew(fileoff, COMMON_PAGE_SIZE, vaddr % COMMON_PAGE_SIZE);
|
||||
|
||||
chunk.shdr.sh_addr = vaddr;
|
||||
chunk.shdr.sh_offset = fileoff;
|
||||
if (!(chunk.shdr.sh_flags & SHF_TLS))
|
||||
vaddr += chunk.shdr.sh_size;
|
||||
}
|
||||
}
|
||||
|
||||
for (; i < ctx.chunks.size(); i++) {
|
||||
OutputChunk<E> &chunk = *ctx.chunks[i];
|
||||
ASSERT(!(chunk.shdr.sh_flags & SHF_ALLOC));
|
||||
fileoff = align_to(fileoff, chunk.shdr.sh_addralign);
|
||||
chunk.shdr.sh_offset = fileoff;
|
||||
fileoff += chunk.shdr.sh_size;
|
||||
}
|
||||
return fileoff;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 get_num_irelative_relocs(Context<E> &ctx) {
|
||||
i64 n = 0;
|
||||
for (Symbol<E> *sym : ctx.got->got_syms)
|
||||
if (sym->get_type() == STT_GNU_IFUNC)
|
||||
n++;
|
||||
return n;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void fix_synthetic_symbols(Context<E> &ctx) {
|
||||
auto start = [](Symbol<E> *sym, auto &chunk) {
|
||||
if (sym && chunk) {
|
||||
sym->shndx = chunk->shndx;
|
||||
sym->value = chunk->shdr.sh_addr;
|
||||
}
|
||||
};
|
||||
|
||||
auto stop = [](Symbol<E> *sym, auto &chunk) {
|
||||
if (sym && chunk) {
|
||||
sym->shndx = chunk->shndx;
|
||||
sym->value = chunk->shdr.sh_addr + chunk->shdr.sh_size;
|
||||
}
|
||||
};
|
||||
|
||||
// __bss_start
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
if (chunk->kind == OutputChunk<E>::REGULAR && chunk->name == ".bss") {
|
||||
start(ctx.__bss_start, chunk);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// __ehdr_start and __executable_start
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
if (chunk->shndx == 1) {
|
||||
ctx.__ehdr_start->shndx = 1;
|
||||
ctx.__ehdr_start->value = ctx.ehdr->shdr.sh_addr;
|
||||
|
||||
ctx.__executable_start->shndx = 1;
|
||||
ctx.__executable_start->value = ctx.ehdr->shdr.sh_addr;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// __rel_iplt_start
|
||||
start(ctx.__rel_iplt_start, ctx.reldyn);
|
||||
|
||||
// __rel_iplt_end
|
||||
ctx.__rel_iplt_end->shndx = ctx.reldyn->shndx;
|
||||
ctx.__rel_iplt_end->value = ctx.reldyn->shdr.sh_addr +
|
||||
get_num_irelative_relocs(ctx) * sizeof(ElfRel<E>);
|
||||
|
||||
// __{init,fini}_array_{start,end}
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
switch (chunk->shdr.sh_type) {
|
||||
case SHT_INIT_ARRAY:
|
||||
start(ctx.__init_array_start, chunk);
|
||||
stop(ctx.__init_array_end, chunk);
|
||||
break;
|
||||
case SHT_FINI_ARRAY:
|
||||
start(ctx.__fini_array_start, chunk);
|
||||
stop(ctx.__fini_array_end, chunk);
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// _end, _etext, _edata and the like
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
if (chunk->kind == OutputChunk<E>::HEADER)
|
||||
continue;
|
||||
|
||||
if (chunk->shdr.sh_flags & SHF_ALLOC) {
|
||||
stop(ctx._end, chunk);
|
||||
stop(ctx.end, chunk);
|
||||
}
|
||||
|
||||
if (chunk->shdr.sh_flags & SHF_EXECINSTR) {
|
||||
stop(ctx._etext, chunk);
|
||||
stop(ctx.etext, chunk);
|
||||
}
|
||||
|
||||
if (chunk->shdr.sh_type != SHT_NOBITS &&
|
||||
(chunk->shdr.sh_flags & SHF_ALLOC)) {
|
||||
stop(ctx._edata, chunk);
|
||||
stop(ctx.edata, chunk);
|
||||
}
|
||||
}
|
||||
|
||||
// _DYNAMIC
|
||||
start(ctx._DYNAMIC, ctx.dynamic);
|
||||
|
||||
// _GLOBAL_OFFSET_TABLE_
|
||||
if (E::e_machine == EM_X86_64 || E::e_machine == EM_386)
|
||||
start(ctx._GLOBAL_OFFSET_TABLE_, ctx.gotplt);
|
||||
else if (E::e_machine == EM_AARCH64)
|
||||
start(ctx._GLOBAL_OFFSET_TABLE_, ctx.got);
|
||||
else
|
||||
unreachable(ctx);
|
||||
|
||||
// __GNU_EH_FRAME_HDR
|
||||
start(ctx.__GNU_EH_FRAME_HDR, ctx.eh_frame_hdr);
|
||||
|
||||
// __start_ and __stop_ symbols
|
||||
for (OutputChunk<E> *chunk : ctx.chunks) {
|
||||
if (is_c_identifier(chunk->name)) {
|
||||
std::string_view sym1 =
|
||||
save_string(ctx, "__start_" + std::string(chunk->name));
|
||||
std::string_view sym2 =
|
||||
save_string(ctx, "__stop_" + std::string(chunk->name));
|
||||
|
||||
start(Symbol<E>::intern(ctx, sym1), chunk);
|
||||
stop(Symbol<E>::intern(ctx, sym2), chunk);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void compress_debug_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "compress_debug_sections");
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)ctx.chunks.size(), [&](i64 i) {
|
||||
OutputChunk<E> &chunk = *ctx.chunks[i];
|
||||
|
||||
if ((chunk.shdr.sh_flags & SHF_ALLOC) || chunk.shdr.sh_size == 0 ||
|
||||
!chunk.name.starts_with(".debug"))
|
||||
return;
|
||||
|
||||
OutputChunk<E> *comp = nullptr;
|
||||
if (ctx.arg.compress_debug_sections == COMPRESS_GABI)
|
||||
comp = new GabiCompressedSection<E>(ctx, chunk);
|
||||
else if (ctx.arg.compress_debug_sections == COMPRESS_GNU)
|
||||
comp = new GnuCompressedSection<E>(ctx, chunk);
|
||||
ASSERT(comp);
|
||||
|
||||
ctx.output_chunks.push_back(std::unique_ptr<OutputChunk<E>>(comp));
|
||||
ctx.chunks[i] = comp;
|
||||
});
|
||||
|
||||
ctx.shstrtab->update_shdr(ctx);
|
||||
ctx.ehdr->update_shdr(ctx);
|
||||
ctx.shdr->update_shdr(ctx);
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void apply_exclude_libs(Context<E> &ctx); \
|
||||
template void create_synthetic_sections(Context<E> &ctx); \
|
||||
template void resolve_symbols(Context<E> &ctx); \
|
||||
template void eliminate_comdats(Context<E> &ctx); \
|
||||
template void convert_common_symbols(Context<E> &ctx); \
|
||||
template void compute_merged_section_sizes(Context<E> &ctx); \
|
||||
template void bin_sections(Context<E> &ctx); \
|
||||
template ObjectFile<E> *create_internal_file(Context<E> &ctx); \
|
||||
template void check_duplicate_symbols(Context<E> &ctx); \
|
||||
template void sort_init_fini(Context<E> &ctx); \
|
||||
template std::vector<OutputChunk<E> *> \
|
||||
collect_output_sections(Context<E> &ctx); \
|
||||
template void compute_section_sizes(Context<E> &ctx); \
|
||||
template void claim_unresolved_symbols(Context<E> &ctx); \
|
||||
template void scan_rels(Context<E> &ctx); \
|
||||
template void apply_version_script(Context<E> &ctx); \
|
||||
template void parse_symbol_version(Context<E> &ctx); \
|
||||
template void compute_import_export(Context<E> &ctx); \
|
||||
template void clear_padding(Context<E> &ctx); \
|
||||
template i64 get_section_rank(Context<E> &ctx, OutputChunk<E> *chunk); \
|
||||
template i64 set_osec_offsets(Context<E> &ctx); \
|
||||
template void fix_synthetic_symbols(Context<E> &ctx); \
|
||||
template void compress_debug_sections(Context<E> &ctx);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+55
-41
@@ -1,25 +1,21 @@
|
||||
#include "lib.h"
|
||||
#include "mold.h"
|
||||
|
||||
#include <functional>
|
||||
#include <iomanip>
|
||||
#include <ios>
|
||||
#include <tbb/concurrent_vector.h>
|
||||
|
||||
#ifdef _WIN32
|
||||
#include <windows.h>
|
||||
#else
|
||||
#include <sys/resource.h>
|
||||
#include <sys/time.h>
|
||||
#endif
|
||||
|
||||
namespace mold {
|
||||
namespace mold::elf {
|
||||
|
||||
i64 Counter::get_value() {
|
||||
return values.combine(std::plus());
|
||||
}
|
||||
|
||||
void Counter::print() {
|
||||
ranges::stable_sort(instances, {}, [](Counter *x) { return x->get_value(); });
|
||||
sort(instances, [](Counter *a, Counter *b) {
|
||||
return a->get_value() > b->get_value();
|
||||
});
|
||||
|
||||
for (Counter *c : instances)
|
||||
std::cout << std::setw(20) << std::right << c->name
|
||||
@@ -27,33 +23,24 @@ void Counter::print() {
|
||||
}
|
||||
|
||||
static i64 now_nsec() {
|
||||
return (i64)std::chrono::steady_clock::now().time_since_epoch().count();
|
||||
struct timespec t;
|
||||
clock_gettime(CLOCK_MONOTONIC, &t);
|
||||
return (i64)t.tv_sec * 1000000000 + t.tv_nsec;
|
||||
}
|
||||
|
||||
static std::pair<i64, i64> get_usage() {
|
||||
#ifdef _WIN32
|
||||
auto to_nsec = [](FILETIME t) -> i64 {
|
||||
return (((u64)t.dwHighDateTime << 32) + (u64)t.dwLowDateTime) * 100;
|
||||
};
|
||||
|
||||
FILETIME creation, exit, kernel, user;
|
||||
GetProcessTimes(GetCurrentProcess(), &creation, &exit, &kernel, &user);
|
||||
return {to_nsec(user), to_nsec(kernel)};
|
||||
#else
|
||||
auto to_nsec = [](struct timeval t) -> i64 {
|
||||
return (i64)t.tv_sec * 1'000'000'000 + t.tv_usec * 1'000;
|
||||
};
|
||||
|
||||
struct rusage ru;
|
||||
getrusage(RUSAGE_SELF, &ru);
|
||||
return {to_nsec(ru.ru_utime), to_nsec(ru.ru_stime)};
|
||||
#endif
|
||||
static i64 to_nsec(struct timeval t) {
|
||||
return (i64)t.tv_sec * 1000000000 + t.tv_usec * 1000;
|
||||
}
|
||||
|
||||
TimerRecord::TimerRecord(std::string name, TimerRecord *parent)
|
||||
: name(name), parent(parent) {
|
||||
struct rusage usage;
|
||||
getrusage(RUSAGE_SELF, &usage);
|
||||
|
||||
start = now_nsec();
|
||||
std::tie(user, sys) = get_usage();
|
||||
user = to_nsec(usage.ru_utime);
|
||||
sys = to_nsec(usage.ru_stime);
|
||||
|
||||
if (parent)
|
||||
parent->children.push_back(this);
|
||||
}
|
||||
@@ -63,31 +50,51 @@ void TimerRecord::stop() {
|
||||
return;
|
||||
stopped = true;
|
||||
|
||||
i64 user2;
|
||||
i64 sys2;
|
||||
std::tie(user2, sys2) = get_usage();
|
||||
struct rusage usage;
|
||||
getrusage(RUSAGE_SELF, &usage);
|
||||
|
||||
end = now_nsec();
|
||||
user = user2 - user;
|
||||
sys = sys2 - sys;
|
||||
user = to_nsec(usage.ru_utime) - user;
|
||||
sys = to_nsec(usage.ru_stime) - sys;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
Timer<E>::Timer(Context<E> &ctx, std::string name, Timer *parent) {
|
||||
record = new TimerRecord(name, parent ? parent->record : nullptr);
|
||||
ctx.timer_records.push_back(std::unique_ptr<TimerRecord>(record));
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
Timer<E>::~Timer() {
|
||||
record->stop();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void Timer<E>::stop() {
|
||||
record->stop();
|
||||
}
|
||||
|
||||
static void print_rec(TimerRecord &rec, i64 indent) {
|
||||
printf(" % 8.3f % 8.3f % 8.3f %s%s\n",
|
||||
((double)rec.user / 1'000'000'000),
|
||||
((double)rec.sys / 1'000'000'000),
|
||||
(((double)rec.end - rec.start) / 1'000'000'000),
|
||||
((double)rec.user / 1000000000),
|
||||
((double)rec.sys / 1000000000),
|
||||
(((double)rec.end - rec.start) / 1000000000),
|
||||
std::string(indent * 2, ' ').c_str(),
|
||||
rec.name.c_str());
|
||||
|
||||
ranges::stable_sort(rec.children, {}, &TimerRecord::start);
|
||||
sort(rec.children, [](TimerRecord *a, TimerRecord *b) {
|
||||
return a->start < b->start;
|
||||
});
|
||||
|
||||
for (TimerRecord *child : rec.children)
|
||||
print_rec(*child, indent + 1);
|
||||
}
|
||||
|
||||
void print_timer_records(
|
||||
tbb::concurrent_vector<std::unique_ptr<TimerRecord>> &records) {
|
||||
template <typename E>
|
||||
void Timer<E>::print(Context<E> &ctx) {
|
||||
tbb::concurrent_vector<std::unique_ptr<TimerRecord>> &records =
|
||||
ctx.timer_records;
|
||||
|
||||
for (i64 i = records.size() - 1; i >= 0; i--)
|
||||
records[i]->stop();
|
||||
|
||||
@@ -115,4 +122,11 @@ void print_timer_records(
|
||||
std::cout << std::flush;
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
#define INSTANTIATE(E) \
|
||||
template class Timer<E>;
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,596 @@
|
||||
// This file implements the -r or ---relocatable option. If the option
|
||||
// is given to a linker, the linker creates not an executable nor a
|
||||
// shared library file but instead create another object file by
|
||||
// combining input object files.
|
||||
//
|
||||
// The behavior of a linker with -r is pretty different from the one
|
||||
// without -r, and adding code for -r to the regular code path could
|
||||
// severely complicate it. Therefore, we implement -r as an
|
||||
// independent feature from others. That's why this file share only a
|
||||
// small amount of code with other files.
|
||||
//
|
||||
// Since the -r option is a minor feature, we don't aim for speed in
|
||||
// this file. This is a "better than nothing" implementation of the -r
|
||||
// option.
|
||||
//
|
||||
// Here is the strategy as to how to combine multiple object files
|
||||
// into one:
|
||||
//
|
||||
// - Regular sections containing opaque data (e.g. ".text" or
|
||||
// ".data") are just copied as-is without merging.
|
||||
//
|
||||
// - .symtab, .strtab and .shstrtab are merged.
|
||||
//
|
||||
// - COMDAT groups are uniquified.
|
||||
//
|
||||
// - Relocation sections are copied one by one, but we need to fix
|
||||
// symbol indices.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <unordered_map>
|
||||
#include <unordered_set>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
template <typename E> class RObjectFile;
|
||||
|
||||
template <typename E>
|
||||
class ROutputChunk {
|
||||
public:
|
||||
ROutputChunk() {
|
||||
out_shdr.sh_addralign = 1;
|
||||
}
|
||||
|
||||
virtual ~ROutputChunk() = default;
|
||||
virtual void update_shdr(Context<E> &ctx) {}
|
||||
virtual void write_to(Context<E> &ctx) = 0;
|
||||
|
||||
std::string_view name;
|
||||
i64 shndx = 0;
|
||||
ElfShdr<E> in_shdr = {};
|
||||
ElfShdr<E> out_shdr = {};
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class RInputSection : public ROutputChunk<E> {
|
||||
public:
|
||||
RInputSection(Context<E> &ctx, RObjectFile<E> &file, const ElfShdr<E> &shdr);
|
||||
void update_shdr(Context<E> &ctx) override;
|
||||
void write_to(Context<E> &ctx) override;
|
||||
|
||||
RObjectFile<E> &file;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class RSymtabSection : public ROutputChunk<E> {
|
||||
public:
|
||||
RSymtabSection() {
|
||||
this->name = ".symtab";
|
||||
this->out_shdr.sh_type = SHT_SYMTAB;
|
||||
this->out_shdr.sh_entsize = sizeof(ElfSym<E>);
|
||||
this->out_shdr.sh_addralign = E::wordsize;
|
||||
}
|
||||
|
||||
void add_local_symbol(Context<E> &ctx, RObjectFile<E> &file, i64 idx);
|
||||
void add_global_symbol(Context<E> &ctx, RObjectFile<E> &file, i64 idx);
|
||||
void update_shdr(Context<E> &ctx) override;
|
||||
void write_to(Context<E> &ctx) override;
|
||||
|
||||
std::unordered_map<std::string_view, i64> sym_map;
|
||||
std::vector<ElfSym<E>> syms{1};
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class RStrtabSection : public ROutputChunk<E> {
|
||||
public:
|
||||
RStrtabSection(std::string_view name) {
|
||||
this->name = name;
|
||||
this->out_shdr.sh_type = SHT_STRTAB;
|
||||
this->out_shdr.sh_size = 1;
|
||||
}
|
||||
|
||||
i64 add_string(std::string_view str);
|
||||
void write_to(Context<E> &ctx) override;
|
||||
|
||||
std::unordered_map<std::string_view, i64> strings;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class ROutputEhdr : public ROutputChunk<E> {
|
||||
public:
|
||||
ROutputEhdr() {
|
||||
this->out_shdr.sh_size = sizeof(ElfEhdr<E>);
|
||||
}
|
||||
|
||||
void write_to(Context<E> &ctx) override;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class ROutputShdr : public ROutputChunk<E> {
|
||||
public:
|
||||
ROutputShdr() {
|
||||
this->out_shdr.sh_size = sizeof(ElfShdr<E>);
|
||||
}
|
||||
|
||||
void update_shdr(Context<E> &ctx) override;
|
||||
void write_to(Context<E> &ctx) override;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
class RObjectFile {
|
||||
public:
|
||||
RObjectFile(Context<E> &ctx, MemoryMappedFile<E> &mb, bool is_alive);
|
||||
|
||||
void remove_comdats(Context<E> &ctx,
|
||||
std::unordered_set<std::string_view> &groups);
|
||||
|
||||
template <typename T>
|
||||
std::span<T> get_data(Context<E> &ctx, const ElfShdr<E> &shdr);
|
||||
|
||||
MemoryMappedFile<E> &mb;
|
||||
std::span<ElfShdr<E>> elf_sections;
|
||||
std::vector<std::unique_ptr<RInputSection<E>>> sections;
|
||||
std::span<const ElfSym<E>> syms;
|
||||
std::vector<i64> symidx;
|
||||
std::unordered_set<std::string_view> defined_syms;
|
||||
std::unordered_set<std::string_view> undef_syms;
|
||||
i64 symtab_shndx = 0;
|
||||
i64 first_global = 0;
|
||||
bool is_alive;
|
||||
const char *strtab = nullptr;
|
||||
const char *shstrtab = nullptr;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
void RSymtabSection<E>::add_local_symbol(Context<E> &ctx, RObjectFile<E> &file,
|
||||
i64 idx) {
|
||||
ElfSym<E> sym = file.syms[idx];
|
||||
ASSERT(sym.st_bind == STB_LOCAL);
|
||||
|
||||
if (!sym.is_undef() && !sym.is_abs() && !sym.is_common()) {
|
||||
if (!file.sections[sym.st_shndx])
|
||||
return;
|
||||
sym.st_shndx = file.sections[sym.st_shndx]->shndx;
|
||||
}
|
||||
|
||||
std::string_view name = file.strtab + sym.st_name;
|
||||
sym.st_name = ctx.r_strtab->add_string(name);
|
||||
|
||||
file.symidx[idx] = syms.size();
|
||||
syms.push_back(sym);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RSymtabSection<E>::add_global_symbol(Context<E> &ctx, RObjectFile<E> &file,
|
||||
i64 idx) {
|
||||
ElfSym<E> sym = file.syms[idx];
|
||||
ASSERT(sym.st_bind != STB_LOCAL);
|
||||
|
||||
std::string_view name = file.strtab + sym.st_name;
|
||||
auto [it, inserted] = sym_map.insert({name, syms.size()});
|
||||
|
||||
if (inserted) {
|
||||
if (!sym.is_undef() && !sym.is_abs() && !sym.is_common())
|
||||
sym.st_shndx = file.sections[sym.st_shndx]->shndx;
|
||||
sym.st_name = ctx.r_strtab->add_string(name);
|
||||
file.symidx[idx] = syms.size();
|
||||
syms.push_back(sym);
|
||||
return;
|
||||
}
|
||||
|
||||
file.symidx[idx] = it->second;
|
||||
|
||||
ElfSym<E> &existing = syms[it->second];
|
||||
if (existing.is_undef() && !sym.is_undef()) {
|
||||
if (!sym.is_abs() && !sym.is_common())
|
||||
sym.st_shndx = file.sections[sym.st_shndx]->shndx;
|
||||
sym.st_name = existing.st_name;
|
||||
existing = sym;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RSymtabSection<E>::update_shdr(Context<E> &ctx) {
|
||||
this->out_shdr.sh_size = syms.size() * sizeof(ElfSym<E>);
|
||||
this->out_shdr.sh_link = ctx.r_strtab->shndx;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RSymtabSection<E>::write_to(Context<E> &ctx) {
|
||||
ElfSym<E> *buf = (ElfSym<E> *)(ctx.buf + this->out_shdr.sh_offset);
|
||||
for (i64 i = 1; i < syms.size(); i++)
|
||||
buf[i] = syms[i];
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
RInputSection<E>::RInputSection(Context<E> &ctx, RObjectFile<E> &file,
|
||||
const ElfShdr<E> &shdr)
|
||||
: file(file) {
|
||||
this->name = file.shstrtab + shdr.sh_name;
|
||||
this->in_shdr = shdr;
|
||||
this->out_shdr = shdr;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RInputSection<E>::update_shdr(Context<E> &ctx) {
|
||||
switch (this->in_shdr.sh_type) {
|
||||
case SHT_GROUP:
|
||||
this->out_shdr.sh_link = ctx.r_symtab->shndx;
|
||||
this->out_shdr.sh_info = file.symidx[this->in_shdr.sh_info];
|
||||
break;
|
||||
case SHT_REL:
|
||||
case SHT_RELA:
|
||||
this->out_shdr.sh_link = ctx.r_symtab->shndx;
|
||||
this->out_shdr.sh_info = file.sections[this->in_shdr.sh_info]->shndx;
|
||||
break;
|
||||
default:
|
||||
if (this->in_shdr.sh_link) {
|
||||
std::unique_ptr<RInputSection<E>> &sec =
|
||||
file.sections[this->in_shdr.sh_info];
|
||||
|
||||
if (sec)
|
||||
this->out_shdr.sh_link = sec->shndx;
|
||||
else if (this->in_shdr.sh_link == file.symtab_shndx)
|
||||
this->out_shdr.sh_link = ctx.r_symtab->shndx;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RInputSection<E>::write_to(Context<E> &ctx) {
|
||||
if (this->in_shdr.sh_type == SHT_NOBITS)
|
||||
return;
|
||||
|
||||
std::span<u8> contents = file.template get_data<u8>(ctx, this->in_shdr);
|
||||
memcpy(ctx.buf + this->out_shdr.sh_offset, contents.data(), contents.size());
|
||||
|
||||
switch (this->in_shdr.sh_type) {
|
||||
case SHT_GROUP: {
|
||||
u32 *mem = (u32 *)(ctx.buf + this->out_shdr.sh_offset);
|
||||
for (i64 i = 1; i < this->out_shdr.sh_size / sizeof(u32); i++)
|
||||
mem[i] = file.sections[mem[i]]->shndx;
|
||||
break;
|
||||
}
|
||||
case SHT_REL:
|
||||
case SHT_RELA: {
|
||||
ElfRel<E> *rel = (ElfRel<E> *)(ctx.buf + this->out_shdr.sh_offset);
|
||||
i64 size = this->out_shdr.sh_size / sizeof(ElfRel<E>);
|
||||
|
||||
for (i64 i = 0; i < size; i++) {
|
||||
const ElfSym<E> &sym = file.syms[rel[i].r_sym];
|
||||
if (sym.is_undef() || sym.is_abs() || sym.is_common() ||
|
||||
file.sections[sym.st_shndx])
|
||||
rel[i].r_sym = file.symidx[rel[i].r_sym];
|
||||
else
|
||||
memset(rel + i, 0, sizeof(ElfRel<E>));
|
||||
}
|
||||
|
||||
i64 i = 0;
|
||||
i64 j = 0;
|
||||
for (; j < size; j++)
|
||||
if (rel[j].r_type)
|
||||
rel[i++] = rel[j];
|
||||
for (; i < size; i++)
|
||||
memset(rel + i, 0, sizeof(ElfRel<E>));
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
i64 RStrtabSection<E>::add_string(std::string_view str) {
|
||||
auto [it, inserted] = strings.insert({str, this->out_shdr.sh_size});
|
||||
if (inserted)
|
||||
this->out_shdr.sh_size += str.size() + 1;
|
||||
return it->second;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void RStrtabSection<E>::write_to(Context<E> &ctx) {
|
||||
for (auto [str, offset] : strings)
|
||||
memcpy(ctx.buf + this->out_shdr.sh_offset + offset, str.data(), str.size());
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void ROutputEhdr<E>::write_to(Context<E> &ctx) {
|
||||
ElfEhdr<E> &hdr = *(ElfEhdr<E> *)(ctx.buf + this->out_shdr.sh_offset);
|
||||
memcpy(&hdr.e_ident, "\177ELF", 4);
|
||||
hdr.e_ident[EI_CLASS] = (E::wordsize == 8) ? ELFCLASS64 : ELFCLASS32;
|
||||
hdr.e_ident[EI_DATA] = E::is_le ? ELFDATA2LSB : ELFDATA2MSB;
|
||||
hdr.e_ident[EI_VERSION] = EV_CURRENT;
|
||||
hdr.e_type = ET_REL;
|
||||
hdr.e_machine = E::e_machine;
|
||||
hdr.e_version = EV_CURRENT;
|
||||
hdr.e_shoff = ctx.r_shdr->out_shdr.sh_offset;
|
||||
hdr.e_ehsize = sizeof(ElfEhdr<E>);
|
||||
hdr.e_shentsize = sizeof(ElfShdr<E>);
|
||||
hdr.e_shstrndx = ctx.r_shstrtab->shndx;
|
||||
hdr.e_shnum = ctx.r_shdr->out_shdr.sh_size / sizeof(ElfShdr<E>);
|
||||
hdr.e_shstrndx = ctx.r_shstrtab->shndx;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void ROutputShdr<E>::update_shdr(Context<E> &ctx) {
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
if (chunk->shndx)
|
||||
this->out_shdr.sh_size += sizeof(ElfShdr<E>);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void ROutputShdr<E>::write_to(Context<E> &ctx) {
|
||||
ElfShdr<E> *hdr = (ElfShdr<E> *)(ctx.buf + this->out_shdr.sh_offset);
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
if (chunk->shndx)
|
||||
hdr[chunk->shndx] = chunk->out_shdr;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
RObjectFile<E>::RObjectFile(Context<E> &ctx, MemoryMappedFile<E> &mb,
|
||||
bool is_alive)
|
||||
: mb(mb), is_alive(is_alive) {
|
||||
// Read ELF header and section header
|
||||
ElfEhdr<E> &ehdr = *(ElfEhdr<E> *)mb.data(ctx);
|
||||
ElfShdr<E> *sh_begin = (ElfShdr<E> *)(mb.data(ctx) + ehdr.e_shoff);
|
||||
i64 num_sections = (ehdr.e_shnum == 0) ? sh_begin->sh_size : ehdr.e_shnum;
|
||||
elf_sections = {sh_begin, sh_begin + num_sections};
|
||||
sections.resize(num_sections);
|
||||
|
||||
// Read .shstrtab
|
||||
i64 shstrtab_idx = (ehdr.e_shstrndx == SHN_XINDEX)
|
||||
? sh_begin->sh_link : ehdr.e_shstrndx;
|
||||
shstrtab = (char *)(mb.data(ctx) + elf_sections[shstrtab_idx].sh_offset);
|
||||
|
||||
// Read .symtab
|
||||
for (i64 i = 1; i < elf_sections.size(); i++) {
|
||||
ElfShdr<E> &shdr = elf_sections[i];
|
||||
if (shdr.sh_type == SHT_SYMTAB) {
|
||||
syms = get_data<const ElfSym<E>>(ctx, shdr);
|
||||
strtab = (char *)(mb.data(ctx) + elf_sections[shdr.sh_link].sh_offset);
|
||||
symtab_shndx = i;
|
||||
first_global = shdr.sh_info;
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
symidx.resize(syms.size());
|
||||
|
||||
// Read sections
|
||||
for (i64 i = 1; i < elf_sections.size(); i++) {
|
||||
ElfShdr<E> &shdr = elf_sections[i];
|
||||
switch (shdr.sh_type) {
|
||||
case SHT_NULL:
|
||||
case SHT_SYMTAB:
|
||||
case SHT_STRTAB:
|
||||
break;
|
||||
default:
|
||||
sections[i].reset(new RInputSection(ctx, *this, shdr));
|
||||
}
|
||||
}
|
||||
|
||||
// Read global symbols
|
||||
for (i64 i = first_global; i < syms.size(); i++) {
|
||||
std::string_view name = strtab + syms[i].st_name;
|
||||
if (syms[i].is_defined())
|
||||
defined_syms.insert(name);
|
||||
else
|
||||
undef_syms.insert(name);
|
||||
}
|
||||
}
|
||||
|
||||
// Remove duplicate comdat groups
|
||||
template <typename E>
|
||||
void RObjectFile<E>::remove_comdats(Context<E> &ctx,
|
||||
std::unordered_set<std::string_view> &groups) {
|
||||
for (i64 i = 1; i < sections.size(); i++) {
|
||||
if (!sections[i])
|
||||
continue;
|
||||
ElfShdr<E> &shdr = sections[i]->in_shdr;
|
||||
if (shdr.sh_type != SHT_GROUP)
|
||||
continue;
|
||||
|
||||
// Get a comdat group signature and insert it into a set.
|
||||
const ElfSym<E> &sym = syms[shdr.sh_info];
|
||||
std::string_view signature = strtab + sym.st_name;
|
||||
if (groups.insert(signature).second)
|
||||
continue;
|
||||
|
||||
// If it is a duplicate, remove it and its members.
|
||||
for (i64 j : this->template get_data<u32>(ctx, shdr).subspan(1))
|
||||
sections[j] = nullptr;
|
||||
sections[i] = nullptr;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
template <typename T>
|
||||
std::span<T> RObjectFile<E>::get_data(Context<E> &ctx, const ElfShdr<E> &shdr) {
|
||||
T *begin = (T *)(mb.data(ctx) + shdr.sh_offset);
|
||||
T *end = (T *)(mb.data(ctx) + shdr.sh_offset + shdr.sh_size);
|
||||
return {begin, end};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::vector<std::unique_ptr<RObjectFile<E>>>
|
||||
open_files(Context<E> &ctx, std::span<std::string_view> args) {
|
||||
std::vector<std::unique_ptr<RObjectFile<E>>> files;
|
||||
bool whole_archive = false;
|
||||
|
||||
while (!args.empty()) {
|
||||
if (read_flag(args, "whole-archive")) {
|
||||
whole_archive = true;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (read_flag(args, "no-whole-archive")) {
|
||||
whole_archive = false;
|
||||
continue;
|
||||
}
|
||||
|
||||
std::string_view arg;
|
||||
if (read_arg(ctx, args, arg, "version-script") ||
|
||||
read_arg(ctx, args, arg, "dynamic-list"))
|
||||
continue;
|
||||
|
||||
MemoryMappedFile<E> *mb = nullptr;
|
||||
|
||||
if (read_arg(ctx, args, arg, "l")) {
|
||||
mb = find_library(ctx, std::string(arg));
|
||||
} else {
|
||||
if (arg.starts_with('-'))
|
||||
continue;
|
||||
arg = args[0];
|
||||
args = args.subspan(1);
|
||||
mb = MemoryMappedFile<E>::must_open(ctx, std::string(arg));
|
||||
}
|
||||
|
||||
switch (get_file_type(ctx, mb)) {
|
||||
case FileType::OBJ:
|
||||
files.emplace_back(new RObjectFile<E>(ctx, *mb, true));
|
||||
break;
|
||||
case FileType::AR:
|
||||
case FileType::THIN_AR:
|
||||
for (MemoryMappedFile<E> *child : read_archive_members(ctx, mb))
|
||||
if (get_file_type(ctx, child) == FileType::OBJ)
|
||||
files.emplace_back(new RObjectFile<E>(ctx, *child, whole_archive));
|
||||
break;
|
||||
default:
|
||||
break;
|
||||
}
|
||||
}
|
||||
return files;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 assign_offsets(Context<E> &ctx) {
|
||||
i64 offset = 0;
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks) {
|
||||
offset = align_to(offset, chunk->out_shdr.sh_addralign);
|
||||
chunk->out_shdr.sh_offset = offset;
|
||||
offset += chunk->out_shdr.sh_size;
|
||||
}
|
||||
return offset;
|
||||
}
|
||||
|
||||
static bool contains(std::unordered_set<std::string_view> &a,
|
||||
std::unordered_set<std::string_view> &b) {
|
||||
for (std::string_view x : b)
|
||||
if (a.contains(x))
|
||||
return true;
|
||||
return false;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void combine_objects(Context<E> &ctx, std::span<std::string_view> file_args) {
|
||||
// Read object files
|
||||
std::vector<std::unique_ptr<RObjectFile<E>>> files = open_files(ctx, file_args);
|
||||
|
||||
// Identify needed objects
|
||||
std::unordered_set<std::string_view> undef_syms;
|
||||
auto add_syms = [&](RObjectFile<E> &file) {
|
||||
undef_syms.insert(file.undef_syms.begin(), file.undef_syms.end());
|
||||
for (std::string_view name : file.defined_syms)
|
||||
undef_syms.erase(name);
|
||||
file.is_alive = true;
|
||||
};
|
||||
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files)
|
||||
if (file->is_alive)
|
||||
add_syms(*file);
|
||||
|
||||
for (;;) {
|
||||
bool added = false;
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files) {
|
||||
if (!file->is_alive && contains(undef_syms, file->defined_syms)) {
|
||||
add_syms(*file);
|
||||
added = true;
|
||||
}
|
||||
}
|
||||
if (!added)
|
||||
break;
|
||||
}
|
||||
|
||||
files.erase(std::remove_if(files.begin(), files.end(),
|
||||
[](std::unique_ptr<RObjectFile<E>> &file) {
|
||||
return !file->is_alive;
|
||||
}),
|
||||
files.end());
|
||||
|
||||
// Remove duplicate comdat groups
|
||||
std::unordered_set<std::string_view> comdat_groups;
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files)
|
||||
file->remove_comdats(ctx, comdat_groups);
|
||||
|
||||
// Create headers and linker-synthesized sections
|
||||
ROutputEhdr<E> ehdr;
|
||||
ROutputShdr<E> shdr;
|
||||
RSymtabSection<E> symtab;
|
||||
RStrtabSection<E> shstrtab(".shstrtab");
|
||||
RStrtabSection<E> strtab(".strtab");
|
||||
|
||||
ctx.r_chunks.push_back(&ehdr);
|
||||
ctx.r_chunks.push_back(&shstrtab);
|
||||
ctx.r_chunks.push_back(&strtab);
|
||||
|
||||
ctx.r_ehdr = &ehdr;
|
||||
ctx.r_shdr = &shdr;
|
||||
ctx.r_shstrtab = &shstrtab;
|
||||
ctx.r_strtab = &strtab;
|
||||
ctx.r_symtab = &symtab;
|
||||
|
||||
// Add input sections to output sections
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files)
|
||||
for (std::unique_ptr<RInputSection<E>> &sec : file->sections)
|
||||
if (sec)
|
||||
ctx.r_chunks.push_back(sec.get());
|
||||
|
||||
ctx.r_chunks.push_back(&symtab);
|
||||
ctx.r_chunks.push_back(&shdr);
|
||||
|
||||
// Assign output section indices
|
||||
i64 shndx = 1;
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
if (chunk != &ehdr && chunk != &shdr)
|
||||
chunk->shndx = shndx++;
|
||||
|
||||
// Add section names to .shstrtab
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
if (chunk->shndx)
|
||||
chunk->out_shdr.sh_name = shstrtab.add_string(chunk->name);
|
||||
|
||||
// Copy symbols from input objects to an output object
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files)
|
||||
for (i64 i = 1; i < file->first_global; i++)
|
||||
symtab.add_local_symbol(ctx, *file, i);
|
||||
|
||||
symtab.out_shdr.sh_info = symtab.syms.size();
|
||||
|
||||
for (std::unique_ptr<RObjectFile<E>> &file : files)
|
||||
for (i64 i = file->first_global; i < file->syms.size(); i++)
|
||||
symtab.add_global_symbol(ctx, *file, i);
|
||||
|
||||
// Finalize section header
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
chunk->update_shdr(ctx);
|
||||
|
||||
// Open an output file
|
||||
i64 filesize = assign_offsets(ctx);
|
||||
std::unique_ptr<OutputFile<E>> out =
|
||||
OutputFile<E>::open(ctx, ctx.arg.output, filesize, 0666);
|
||||
memset(out->buf, 0, filesize);
|
||||
ctx.buf = out->buf;
|
||||
|
||||
// Write to the output file
|
||||
for (ROutputChunk<E> *chunk : ctx.r_chunks)
|
||||
chunk->write_to(ctx);
|
||||
out->close(ctx);
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void combine_objects(Context<E> &, std::span<std::string_view>);
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,322 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <signal.h>
|
||||
#include <sys/socket.h>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/types.h>
|
||||
#include <sys/un.h>
|
||||
#include <sys/wait.h>
|
||||
#include <unistd.h>
|
||||
|
||||
#ifdef __APPLE__
|
||||
# define COMMON_DIGEST_FOR_OPENSSL
|
||||
# include <CommonCrypto/CommonDigest.h>
|
||||
#else
|
||||
# include <openssl/sha.h>
|
||||
#endif
|
||||
|
||||
#define DAEMON_TIMEOUT 30
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
// Exiting from a program with large memory usage is slow --
|
||||
// it may take a few hundred milliseconds. To hide the latency,
|
||||
// we fork a child and let it do the actual linking work.
|
||||
std::function<void()> fork_child() {
|
||||
int pipefd[2];
|
||||
if (pipe(pipefd) == -1) {
|
||||
perror("pipe");
|
||||
exit(1);
|
||||
}
|
||||
|
||||
pid_t pid = fork();
|
||||
if (pid == -1) {
|
||||
perror("fork");
|
||||
exit(1);
|
||||
}
|
||||
|
||||
if (pid > 0) {
|
||||
// Parent
|
||||
close(pipefd[1]);
|
||||
|
||||
char buf[1];
|
||||
if (read(pipefd[0], buf, 1) == 1)
|
||||
_exit(0);
|
||||
|
||||
int status;
|
||||
waitpid(pid, &status, 0);
|
||||
|
||||
if (WIFEXITED(status))
|
||||
_exit(WEXITSTATUS(status));
|
||||
if (WIFSIGNALED(status))
|
||||
raise(WTERMSIG(status));
|
||||
_exit(1);
|
||||
}
|
||||
|
||||
// Child
|
||||
close(pipefd[0]);
|
||||
|
||||
return [=]() {
|
||||
char buf[] = {1};
|
||||
int n = write(pipefd[1], buf, 1);
|
||||
ASSERT(n == 1);
|
||||
};
|
||||
}
|
||||
|
||||
static std::string base64(u8 *data, u64 size) {
|
||||
static const char chars[] =
|
||||
"ABCDEFGHIJKLMNOPQRSTUVWXYZabcdefghijklmnopqrstuvwxyz0123456789+_";
|
||||
|
||||
std::ostringstream out;
|
||||
|
||||
auto encode = [&](u32 x) {
|
||||
out << chars[x & 0b111111]
|
||||
<< chars[(x >> 6) & 0b111111]
|
||||
<< chars[(x >> 12) & 0b111111]
|
||||
<< chars[(x >> 18) & 0b111111];
|
||||
};
|
||||
|
||||
i64 i = 0;
|
||||
for (; i < size - 3; i += 3)
|
||||
encode((data[i + 2] << 16) | (data[i + 1] << 8) | data[i]);
|
||||
|
||||
if (i == size - 1)
|
||||
encode(data[i]);
|
||||
else if (i == size - 2)
|
||||
encode((data[i + 1] << 8) | data[i]);
|
||||
return out.str();
|
||||
}
|
||||
|
||||
static std::string compute_sha256(std::span<std::string_view> argv) {
|
||||
SHA256_CTX sha;
|
||||
SHA256_Init(&sha);
|
||||
|
||||
for (std::string_view arg : argv) {
|
||||
if (arg != "-preload" && arg != "--preload") {
|
||||
SHA256_Update(&sha, arg.data(), arg.size());
|
||||
char buf[] = {0};
|
||||
SHA256_Update(&sha, buf, 1);
|
||||
}
|
||||
}
|
||||
|
||||
u8 digest[SHA256_SIZE];
|
||||
SHA256_Final(digest, &sha);
|
||||
return base64(digest, SHA256_SIZE);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void send_fd(Context<E> &ctx, i64 conn, i64 fd) {
|
||||
struct iovec iov;
|
||||
char dummy = '1';
|
||||
iov.iov_base = &dummy;
|
||||
iov.iov_len = 1;
|
||||
|
||||
struct msghdr msg = {};
|
||||
msg.msg_iov = &iov;
|
||||
msg.msg_iovlen = 1;
|
||||
|
||||
char buf[CMSG_SPACE(sizeof(int))];
|
||||
msg.msg_control = buf;
|
||||
msg.msg_controllen = CMSG_LEN(sizeof(int));
|
||||
|
||||
struct cmsghdr *cmsg = CMSG_FIRSTHDR(&msg);
|
||||
cmsg->cmsg_level = SOL_SOCKET;
|
||||
cmsg->cmsg_type = SCM_RIGHTS;
|
||||
cmsg->cmsg_len = CMSG_LEN(sizeof(int));
|
||||
*(int *)CMSG_DATA(cmsg) = fd;
|
||||
|
||||
if (sendmsg(conn, &msg, 0) == -1)
|
||||
Fatal(ctx) << "sendmsg failed: " << errno_string();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static i64 recv_fd(Context<E> &ctx, i64 conn) {
|
||||
struct iovec iov;
|
||||
char buf[1];
|
||||
iov.iov_base = buf;
|
||||
iov.iov_len = sizeof(buf);
|
||||
|
||||
struct msghdr msg = {};
|
||||
msg.msg_iov = &iov;
|
||||
msg.msg_iovlen = 1;
|
||||
|
||||
char cmsgbuf[CMSG_SPACE(sizeof(int))];
|
||||
msg.msg_control = (caddr_t)cmsgbuf;
|
||||
msg.msg_controllen = sizeof(cmsgbuf);
|
||||
|
||||
i64 len = recvmsg(conn, &msg, 0);
|
||||
if (len <= 0)
|
||||
Fatal(ctx) << "recvmsg failed: " << errno_string();
|
||||
|
||||
struct cmsghdr *cmsg;
|
||||
cmsg = CMSG_FIRSTHDR(&msg);
|
||||
return *(int *)CMSG_DATA(cmsg);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void try_resume_daemon(Context<E> &ctx) {
|
||||
i64 conn = socket(AF_UNIX, SOCK_STREAM, 0);
|
||||
if (conn == -1)
|
||||
Fatal(ctx) << "socket failed: " << errno_string();
|
||||
|
||||
std::string path = "/tmp/mold-" + compute_sha256(ctx.cmdline_args);
|
||||
|
||||
struct sockaddr_un name = {};
|
||||
name.sun_family = AF_UNIX;
|
||||
memcpy(name.sun_path, path.data(), path.size());
|
||||
|
||||
if (connect(conn, (struct sockaddr *)&name, sizeof(name)) != 0) {
|
||||
close(conn);
|
||||
return;
|
||||
}
|
||||
|
||||
send_fd(ctx, conn, STDOUT_FILENO);
|
||||
send_fd(ctx, conn, STDERR_FILENO);
|
||||
|
||||
char buf[1];
|
||||
i64 r = read(conn, buf, 1);
|
||||
close(conn);
|
||||
if (r == 1)
|
||||
exit(0);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void daemonize(Context<E> &ctx, std::function<void()> *wait_for_client,
|
||||
std::function<void()> *on_complete) {
|
||||
if (daemon(1, 0) == -1)
|
||||
Fatal(ctx) << "daemon failed: " << errno_string();
|
||||
|
||||
i64 sock = socket(AF_UNIX, SOCK_STREAM, 0);
|
||||
if (sock == -1)
|
||||
Fatal(ctx) << "socket failed: " << errno_string();
|
||||
|
||||
socket_tmpfile =
|
||||
strdup(("/tmp/mold-" + compute_sha256(ctx.cmdline_args)).c_str());
|
||||
|
||||
struct sockaddr_un name = {};
|
||||
name.sun_family = AF_UNIX;
|
||||
strcpy(name.sun_path, socket_tmpfile);
|
||||
|
||||
u32 orig_mask = umask(0177);
|
||||
|
||||
if (bind(sock, (struct sockaddr *)&name, sizeof(name)) == -1) {
|
||||
if (errno != EADDRINUSE)
|
||||
Fatal(ctx) << "bind failed: " << errno_string();
|
||||
|
||||
unlink(socket_tmpfile);
|
||||
if (bind(sock, (struct sockaddr *)&name, sizeof(name)) == -1)
|
||||
Fatal(ctx) << "bind failed: " << errno_string();
|
||||
}
|
||||
|
||||
umask(orig_mask);
|
||||
|
||||
if (listen(sock, 0) == -1)
|
||||
Fatal(ctx) << "listen failed: " << errno_string();
|
||||
|
||||
static i64 conn = -1;
|
||||
|
||||
*wait_for_client = [=, &ctx]() {
|
||||
fd_set rfds;
|
||||
FD_ZERO(&rfds);
|
||||
FD_SET(sock, &rfds);
|
||||
|
||||
struct timeval tv;
|
||||
tv.tv_sec = DAEMON_TIMEOUT;
|
||||
tv.tv_usec = 0;
|
||||
|
||||
i64 res = select(sock + 1, &rfds, NULL, NULL, &tv);
|
||||
if (res == -1)
|
||||
Fatal(ctx) << "select failed: " << errno_string();
|
||||
|
||||
if (res == 0) {
|
||||
std::cout << "timeout\n";
|
||||
exit(0);
|
||||
}
|
||||
|
||||
conn = accept(sock, NULL, NULL);
|
||||
if (conn == -1)
|
||||
Fatal(ctx) << "accept failed: " << errno_string();
|
||||
unlink(socket_tmpfile);
|
||||
|
||||
dup2(recv_fd(ctx, conn), STDOUT_FILENO);
|
||||
dup2(recv_fd(ctx, conn), STDERR_FILENO);
|
||||
};
|
||||
|
||||
*on_complete = [=]() {
|
||||
char buf[] = {1};
|
||||
int n = write(conn, buf, 1);
|
||||
ASSERT(n == 1);
|
||||
};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::string get_self_path(Context<E> &ctx) {
|
||||
char buf[4096];
|
||||
size_t n = readlink("/proc/self/exe", buf, sizeof(buf));
|
||||
if (n == -1)
|
||||
Fatal(ctx) << "readlink(\"/proc/self/exe\") failed: " << errno_string();
|
||||
if (n == sizeof(buf))
|
||||
Fatal(ctx) << "readlink: path too long";
|
||||
return {buf, n};
|
||||
}
|
||||
|
||||
static bool is_regular_file(const std::string &path) {
|
||||
struct stat st;
|
||||
return !stat(path.c_str(), &st) && (st.st_mode & S_IFMT) == S_IFREG;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::string find_dso(Context<E> &ctx, const std::string &self) {
|
||||
// Look for mold-wrapper.so from the same directory as the executable is.
|
||||
std::string path = std::string(path_dirname(self)) + "/mold-wrapper.so";
|
||||
if (is_regular_file(path))
|
||||
return path;
|
||||
|
||||
// If not exist, mold might be installed as $PREFIX/bin/mold and the
|
||||
// DSO as $PREFIX/lib/mold/mold-wrapper.so.
|
||||
path = path_clean(self + "/../../lib/mold/mold-wrapper.so");
|
||||
if (is_regular_file(path))
|
||||
return path;
|
||||
|
||||
Fatal(ctx) << "mold-wrapper.so is missing";
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
[[noreturn]]
|
||||
void process_run_subcommand(Context<E> &ctx, int argc, char **argv) {
|
||||
std::string_view arg1 = argv[1];
|
||||
ASSERT(arg1 == "-run" || arg1 == "--run");
|
||||
if (!argv[2])
|
||||
Fatal(ctx) << "-run: argument missing";
|
||||
|
||||
// Get the mold-wrapper.so path
|
||||
std::string self = get_self_path(ctx);
|
||||
std::string dso_path = find_dso(ctx, self);
|
||||
|
||||
// Set environment variables
|
||||
putenv(strdup(("LD_PRELOAD=" + dso_path).c_str()));
|
||||
putenv(strdup(("MOLD_PATH=" + self).c_str()));
|
||||
|
||||
// If ld, ld.lld or ld.gold is specified, run mold itself
|
||||
if (std::string_view cmd = path_basename(argv[2]);
|
||||
cmd == "ld" || cmd == "ld.lld" || cmd == "ld.gold") {
|
||||
execv(self.c_str(), argv + 2);
|
||||
Fatal(ctx) << "mold -run failed: " << self << ": " << errno_string();
|
||||
}
|
||||
|
||||
// Execute a given command
|
||||
execvp(argv[2], argv + 2);
|
||||
Fatal(ctx) << "mold -run failed: " << argv[2] << ": " << errno_string();
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template void try_resume_daemon(Context<E> &); \
|
||||
template void daemonize(Context<E> &, std::function<void()> *, \
|
||||
std::function<void()> *); \
|
||||
template void process_run_subcommand(Context<E> &, int, char **)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
@@ -0,0 +1,51 @@
|
||||
#include "mold.h"
|
||||
|
||||
#include <cxxabi.h>
|
||||
#include <stdlib.h>
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
static thread_local char *demangle_buf;
|
||||
static thread_local size_t demangle_buf_len;
|
||||
|
||||
static bool is_mangled_name(std::string_view name) {
|
||||
return name.starts_with("_Z");
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::string_view Symbol<E>::get_demangled_name() const {
|
||||
if (is_mangled_name(name())) {
|
||||
char *mangled = new char[name().size() + 1];
|
||||
memcpy(mangled, name().data(), name().size());
|
||||
mangled[name().size()] = '\0';
|
||||
|
||||
size_t len = sizeof(demangle_buf);
|
||||
int status;
|
||||
demangle_buf =
|
||||
abi::__cxa_demangle(mangled, demangle_buf, &demangle_buf_len, &status);
|
||||
delete[](mangled);
|
||||
if (status == 0)
|
||||
return demangle_buf;
|
||||
}
|
||||
|
||||
return name();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::ostream &operator<<(std::ostream &out, const Symbol<E> &sym) {
|
||||
if (opt_demangle)
|
||||
out << sym.get_demangled_name();
|
||||
else
|
||||
out << sym.name();
|
||||
return out;
|
||||
}
|
||||
|
||||
#define INSTANTIATE(E) \
|
||||
template class Symbol<E>; \
|
||||
template std::ostream &operator<<(std::ostream &, const Symbol<E> &)
|
||||
|
||||
INSTANTIATE(X86_64);
|
||||
INSTANTIATE(I386);
|
||||
INSTANTIATE(AARCH64);
|
||||
|
||||
} // namespace mold::elf
|
||||
+102
@@ -0,0 +1,102 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold::elf {
|
||||
|
||||
// A tar file consists of one or more Ustar header followed by data.
|
||||
// Each Ustar header represents a single file in an archive.
|
||||
//
|
||||
// tar is an old file format, and its `name` field is only 100 bytes long.
|
||||
// If `name` is longer than 100 bytes, we can emit a PAX header before a
|
||||
// Ustar header to store a long filename.
|
||||
//
|
||||
// For simplicity, we always emit a PAX header even for a short filename.
|
||||
struct UstarHeader {
|
||||
void flush() {
|
||||
memset(checksum, ' ', sizeof(checksum));
|
||||
memcpy(magic, "ustar", 5);
|
||||
memcpy(version, "00", 2);
|
||||
|
||||
// Compute checksum
|
||||
int sum = 0;
|
||||
for (i64 i = 0; i < sizeof(*this); i++)
|
||||
sum += ((u8 *)this)[i];
|
||||
sprintf(checksum, "%06o", sum);
|
||||
}
|
||||
|
||||
char name[100];
|
||||
char mode[8];
|
||||
char uid[8];
|
||||
char gid[8];
|
||||
char size[12];
|
||||
char mtime[12];
|
||||
char checksum[8];
|
||||
char typeflag[1];
|
||||
char linkname[100];
|
||||
char magic[6];
|
||||
char version[2];
|
||||
char uname[32];
|
||||
char gname[32];
|
||||
char devmajor[8];
|
||||
char devminor[8];
|
||||
char prefix[155];
|
||||
char pad[12];
|
||||
};
|
||||
|
||||
static_assert(sizeof(UstarHeader) == TarFile::BLOCK_SIZE);
|
||||
|
||||
static std::string encode_path(const std::string &path) {
|
||||
// Construct a string which contains something like
|
||||
// "16 path=foo/bar\n" where 16 is the size of the string
|
||||
// including the size string itself.
|
||||
i64 len = std::string(" path=\n").size() + path.size();
|
||||
i64 total = std::to_string(len).size() + len;
|
||||
total = std::to_string(total).size() + len;
|
||||
return std::to_string(total) + " path=" + path + "\n";
|
||||
}
|
||||
|
||||
void TarFile::append(std::string path, std::string_view data) {
|
||||
contents.push_back({path, data});
|
||||
|
||||
size_ += BLOCK_SIZE * 2;
|
||||
size_ += align_to(path_clean(basedir + "/" + path).size(), BLOCK_SIZE);
|
||||
size_ += align_to(data.size(), BLOCK_SIZE);
|
||||
}
|
||||
|
||||
void TarFile::write_to(u8 *buf) {
|
||||
u8 *start = buf;
|
||||
memset(buf, 0, size_);
|
||||
|
||||
for (i64 i = 0; i < contents.size(); i++) {
|
||||
ASSERT(buf - start <= size_);
|
||||
|
||||
const std::string &path = contents[i].first;
|
||||
std::string_view data = contents[i].second;
|
||||
|
||||
// Write PAX header
|
||||
UstarHeader &pax = *(UstarHeader *)buf;
|
||||
buf += BLOCK_SIZE;
|
||||
|
||||
std::string attr = encode_path(path_clean(basedir + "/" + path));
|
||||
sprintf(pax.size, "%011zo", attr.size());
|
||||
pax.typeflag[0] = 'x';
|
||||
pax.flush();
|
||||
|
||||
// Write pathname
|
||||
memcpy(buf, attr.data(), attr.size());
|
||||
buf += align_to(attr.size(), BLOCK_SIZE);
|
||||
|
||||
// Write Ustar header
|
||||
UstarHeader &ustar = *(UstarHeader *)buf;
|
||||
buf += BLOCK_SIZE;
|
||||
|
||||
memcpy(ustar.mode, "0000664", 8);
|
||||
sprintf(ustar.size, "%011zo", data.size());
|
||||
ustar.flush();
|
||||
|
||||
// Write file contents
|
||||
memcpy(buf, data.data(), data.size());
|
||||
buf += align_to(data.size(), BLOCK_SIZE);
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold::elf
|
||||
Executable
+77
@@ -0,0 +1,77 @@
|
||||
#!/bin/bash
|
||||
#
|
||||
# This test script takes a Gentoo package name and tries to build it
|
||||
# using mold in a Docker environment. We chose Gentoo Linux as a test
|
||||
# target, because its source-based package allows us to build programs
|
||||
# locally and run their test suites without any hassle.
|
||||
#
|
||||
# You can get a complete list of Gentoo packages availalbe for testing
|
||||
# with the following command:
|
||||
#
|
||||
# docker run --rm mold-gentoo emerge --color n -s '' | \
|
||||
# perl -ne 'next unless m!^\*\s+(\S+/\S+)!; print "$1\n"'
|
||||
|
||||
if [ "$1" = "" ]; then
|
||||
echo "Usage: $0 gentoo-package-name"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
set -x
|
||||
|
||||
# Create a Docker image
|
||||
if ! docker image ls mold-gentoo | grep -q mold-gentoo; then
|
||||
set -e
|
||||
cat <<EOF | docker build -t mold-gentoo -
|
||||
FROM gentoo/stage3
|
||||
RUN emerge-webrsync
|
||||
RUN echo 'USE="elogind -systemd corefonts truetype jpeg jpeg2k tiff zstd static-libs binary"' >> /etc/portage/make.conf && \
|
||||
echo 'ACCEPT_LICENSE="* -@EULA"' >> /etc/portage/make.conf && \
|
||||
echo 'FEATURES="\${FEATURE} noclean nostrip ccache -ipc-sandbox -network-sandbox -pid-sandbox -sandbox"' >> /etc/portage/make.conf && \
|
||||
echo 'CCACHE_DIR="/ccache"' >> /etc/portage/make.conf
|
||||
RUN emerge gdb lld vim strace ccache && rm -rf /var/tmp/portage
|
||||
EOF
|
||||
set +e
|
||||
fi
|
||||
|
||||
# Build mold as a statically-linked executable
|
||||
if ! [ -f mold ] || ! ldd mold 2>&1 | grep -Pq 'statically linked|not a dynamic executable'; then
|
||||
make clean
|
||||
./build-static.sh
|
||||
fi
|
||||
|
||||
git_hash=$(./mold --version | perl -ne '/\((\w+)/; print $1;')
|
||||
|
||||
# Build a given package in Docker
|
||||
build() {
|
||||
package="$1"
|
||||
cmd1='(cd /usr/bin; ln -sf /mold/mold $(realpath ld))'
|
||||
cmd2="MAKEOPTS=-'j$(nproc) --load-average=100' emerge --onlydeps $package"
|
||||
cmd3="MAKEOPTS='-j$(nproc) --load-average=100' FEATURES=test emerge $package"
|
||||
filename=`echo "$package" | sed 's!/!_!g'`
|
||||
docker="docker run --rm --cap-add=SYS_PTRACE -v `pwd`:/mold -v /var/cache/ccache-gentoo:/ccache mold-gentoo timeout -v -k 15s 1h"
|
||||
dir=gentoo/$git_hash
|
||||
|
||||
mkdir -p $dir/success $dir/failure
|
||||
|
||||
$docker nice -n 19 bash -c "$cmd1 && $cmd2 && $cmd3" >& $dir/$filename.mold
|
||||
if [ $? = 0 ]; then
|
||||
mv $dir/$filename.mold $dir/success
|
||||
else
|
||||
mv $dir/$filename.mold $dir/failure
|
||||
fi
|
||||
|
||||
$docker nice -n 19 bash -c "$cmd2 && $cmd3" >& $dir/$filename.ld
|
||||
if [ $? = 0 ]; then
|
||||
mv $dir/$filename.ld $dir/success
|
||||
else
|
||||
mv $dir/$filename.ld $dir/failure
|
||||
fi
|
||||
}
|
||||
|
||||
if [ "$1" = dev-libs/concurrencykit ]; then
|
||||
echo "Skipping known broken package: $1"
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Build a package
|
||||
build "$1"
|
||||
@@ -1,56 +0,0 @@
|
||||
#!/bin/sh
|
||||
# This script installs binary packages needed to build mold.
|
||||
# Feel free to send me a PR if your OS is not on this list.
|
||||
|
||||
set -e
|
||||
. /etc/os-release
|
||||
|
||||
set -x
|
||||
|
||||
case "$ID" in
|
||||
ubuntu | pop | linuxmint | debian | raspbian | neon | zorin)
|
||||
apt-get update
|
||||
apt-get install -y cmake gcc g++ clang gdb
|
||||
;;
|
||||
fedora | fedora-* | amzn | rhel | centos)
|
||||
dnf install -y gcc-c++ cmake glibc-static libstdc++-static diffutils util-linux tar
|
||||
;;
|
||||
rocky | ol)
|
||||
dnf install -y gcc-c++ cmake diffutils util-linux
|
||||
;;
|
||||
opensuse-*)
|
||||
zypper install -y make cmake gcc-c++ glibc-devel-static tar diffutils util-linux gawk
|
||||
;;
|
||||
gentoo)
|
||||
emerge-webrsync
|
||||
FEATURES='getbinpkg binpkg-request-signature' emerge dev-build/cmake
|
||||
;;
|
||||
arch | archarm | artix | endeavouros | manjaro | cachyos)
|
||||
pacman -Sy --needed --noconfirm base-devel cmake util-linux
|
||||
;;
|
||||
void)
|
||||
xbps-install -Sy xbps bash make cmake gcc tar diffutils util-linux
|
||||
;;
|
||||
alpine)
|
||||
apk update
|
||||
apk add bash make linux-headers cmake gcc g++
|
||||
;;
|
||||
clear-linux-os)
|
||||
swupd update
|
||||
swupd bundle-add c-basic diffutils
|
||||
;;
|
||||
almalinux)
|
||||
dnf install -y gcc-toolset-13-gcc-c++ gcc-toolset-13-libstdc++-devel cmake diffutils
|
||||
;;
|
||||
altlinux)
|
||||
apt-get update
|
||||
apt-get install -y gcc-c++ make cmake ctest diffutils util-linux
|
||||
;;
|
||||
freebsd)
|
||||
pkg update
|
||||
pkg install -y cmake bash binutils gcc
|
||||
;;
|
||||
*)
|
||||
echo "Error: don't know anything about build dependencies on $ID-$VERSION_ID"
|
||||
exit 1
|
||||
esac
|
||||
@@ -1,19 +0,0 @@
|
||||
#!/bin/bash
|
||||
set -e
|
||||
. /etc/os-release
|
||||
|
||||
set -x
|
||||
|
||||
# This script install packages for -DMOLD_ENABLE_QEMU_TESTS=1
|
||||
# to enable cross-target tests.
|
||||
#
|
||||
# Feel free to send me a PR if your OS is not on this list.
|
||||
|
||||
case "$ID" in
|
||||
ubuntu | pop | linuxmint | debian | raspbian)
|
||||
apt-get install -y qemu-user {gcc,g++}-{i686,aarch64,riscv64,powerpc,powerpc64,powerpc64le,s390x,sparc64,m68k,sh4}-linux-gnu {gcc,g++}-arm-linux-gnueabihf
|
||||
;;
|
||||
*)
|
||||
echo "Error: don't know anything about build dependencies on $ID-$VERSION_ID"
|
||||
exit 1
|
||||
esac
|
||||
@@ -1,132 +0,0 @@
|
||||
// This file implements the Aho-Corasick algorithm to search multiple
|
||||
// strings within an input string simultaneously. It is essentially a
|
||||
// trie with additional links. For details, see
|
||||
// https://en.wikipedia.org/wiki/Aho-Corasick_algorithm.
|
||||
//
|
||||
// We use it for simple glob patterns in version scripts or dynamic
|
||||
// list files. Here are some examples of glob patterns:
|
||||
//
|
||||
// qt_private_api_tag*
|
||||
// *16QAccessibleCache*
|
||||
// *32QAbstractFileIconProviderPrivate*
|
||||
// *17QPixmapIconEngine*
|
||||
//
|
||||
// Aho-Corasick can do only substring search, so it cannot handle
|
||||
// complex glob patterns such as `*foo*bar*`. We handle such patterns
|
||||
// with the Glob class.
|
||||
|
||||
#include "lib.h"
|
||||
|
||||
#include <queue>
|
||||
|
||||
namespace mold {
|
||||
|
||||
bool AhoCorasick::can_handle(std::string_view pat) {
|
||||
if (pat.starts_with('*'))
|
||||
pat.remove_prefix(1);
|
||||
if (pat.ends_with('*'))
|
||||
pat.remove_suffix(1);
|
||||
return pat.find_first_of("*?[") == pat.npos;
|
||||
}
|
||||
|
||||
i64 AhoCorasick::find(std::string_view str) {
|
||||
if (nodes.empty())
|
||||
return -1;
|
||||
|
||||
i64 idx = 0;
|
||||
i64 val = -1;
|
||||
|
||||
auto walk = [&](u8 c) {
|
||||
for (i64 j = idx; j != -1; j = nodes[j].suffix_link) {
|
||||
i64 child = nodes[j].children[c];
|
||||
if (child != -1) {
|
||||
idx = child;
|
||||
val = std::max(val, nodes[child].value);
|
||||
return;
|
||||
}
|
||||
}
|
||||
idx = 0;
|
||||
};
|
||||
|
||||
walk('\0');
|
||||
for (u8 c : str)
|
||||
walk(c);
|
||||
walk('\0');
|
||||
return val;
|
||||
}
|
||||
|
||||
bool AhoCorasick::add(std::string_view pat, i64 val) {
|
||||
assert(can_handle(pat));
|
||||
if (nodes.empty())
|
||||
nodes.resize(1);
|
||||
i64 idx = 0;
|
||||
|
||||
auto walk = [&](u8 c) {
|
||||
if (nodes[idx].children[c] == -1) {
|
||||
nodes[idx].children[c] = nodes.size();
|
||||
nodes.resize(nodes.size() + 1);
|
||||
}
|
||||
idx = nodes[idx].children[c];
|
||||
};
|
||||
|
||||
// We handle "foo" as if "\0foo\0", "*foo" as if "foo\0", "foo*" as
|
||||
// if "\0foo", and "*foo*" as if "foo". Aho-Corasick can do only
|
||||
// substring matching, so we use \0 as a beginning/end-of-string
|
||||
// markers.
|
||||
if (!pat.starts_with('*'))
|
||||
walk('\0');
|
||||
for (u8 c : pat)
|
||||
if (c != '*')
|
||||
walk(c);
|
||||
if (!pat.ends_with('*'))
|
||||
walk('\0');
|
||||
|
||||
nodes[idx].value = std::max(nodes[idx].value, val);
|
||||
return true;
|
||||
}
|
||||
|
||||
void AhoCorasick::compile() {
|
||||
if (nodes.empty())
|
||||
return;
|
||||
fix_suffix_links(0);
|
||||
fix_values();
|
||||
}
|
||||
|
||||
void AhoCorasick::fix_suffix_links(i64 idx) {
|
||||
for (i64 i = 0; i < 256; i++) {
|
||||
i64 child = nodes[idx].children[i];
|
||||
if (child == -1)
|
||||
continue;
|
||||
|
||||
i64 j = nodes[idx].suffix_link;
|
||||
for (; j != -1; j = nodes[j].suffix_link) {
|
||||
if (nodes[j].children[i] != -1) {
|
||||
nodes[child].suffix_link = j;
|
||||
break;
|
||||
}
|
||||
}
|
||||
if (j == -1)
|
||||
nodes[child].suffix_link = 0;
|
||||
fix_suffix_links(child);
|
||||
}
|
||||
}
|
||||
|
||||
void AhoCorasick::fix_values() {
|
||||
std::queue<i64> queue;
|
||||
queue.push(0);
|
||||
|
||||
do {
|
||||
i64 idx = queue.front();
|
||||
queue.pop();
|
||||
|
||||
for (i64 child : nodes[idx].children) {
|
||||
if (child != -1) {
|
||||
i64 suffix = nodes[child].suffix_link;
|
||||
nodes[child].value = std::max(nodes[child].value, nodes[suffix].value);
|
||||
queue.push(child);
|
||||
}
|
||||
}
|
||||
} while (!queue.empty());
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,50 +0,0 @@
|
||||
// This is the same as std::atomic except that the default memory
|
||||
// order is relaxed instead of sequential consistency.
|
||||
|
||||
#pragma once
|
||||
|
||||
#include <atomic>
|
||||
|
||||
namespace mold {
|
||||
|
||||
template <typename T>
|
||||
struct Atomic : std::atomic<T> {
|
||||
static constexpr std::memory_order relaxed = std::memory_order_relaxed;
|
||||
|
||||
using std::atomic<T>::atomic;
|
||||
|
||||
Atomic(const Atomic<T> &other) : std::atomic<T>(other.load()) {}
|
||||
|
||||
Atomic<T> &operator=(const Atomic<T> &other) {
|
||||
store(other.load());
|
||||
return *this;
|
||||
}
|
||||
|
||||
void operator=(T val) { store(val); }
|
||||
operator T() const { return load(); }
|
||||
|
||||
void store(T val, std::memory_order order = relaxed) {
|
||||
std::atomic<T>::store(val, order);
|
||||
}
|
||||
|
||||
T load(std::memory_order order = relaxed) const {
|
||||
return std::atomic<T>::load(order);
|
||||
}
|
||||
|
||||
T exchange(T val) { return std::atomic<T>::exchange(val, relaxed); }
|
||||
T operator|=(T val) { return std::atomic<T>::fetch_or(val, relaxed); }
|
||||
T operator++() { return std::atomic<T>::fetch_add(1, relaxed) + 1; }
|
||||
T operator--() { return std::atomic<T>::fetch_sub(1, relaxed) - 1; }
|
||||
T operator++(int) { return std::atomic<T>::fetch_add(1, relaxed); }
|
||||
T operator--(int) { return std::atomic<T>::fetch_sub(1, relaxed); }
|
||||
|
||||
bool test_and_set() {
|
||||
// A relaxed load + branch (assuming miss) takes only around 20 cycles,
|
||||
// while an atomic RMW can easily take hundreds on x86. We note that it's
|
||||
// common that another thread beat us in marking, so doing an optimistic
|
||||
// early test tends to improve performance in the ~20% ballpark.
|
||||
return load() || exchange(true);
|
||||
}
|
||||
};
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,78 +0,0 @@
|
||||
#pragma once
|
||||
|
||||
#include "integers.h"
|
||||
|
||||
#include <cassert>
|
||||
#include <vector>
|
||||
|
||||
namespace mold {
|
||||
|
||||
class BitvectorProxy {
|
||||
public:
|
||||
BitvectorProxy(u64 &word, size_t pos)
|
||||
: word(word), mask(1ULL << pos) {}
|
||||
|
||||
BitvectorProxy &operator=(bool val) {
|
||||
if (val)
|
||||
word |= mask;
|
||||
else
|
||||
word &= ~mask;
|
||||
return *this;
|
||||
}
|
||||
|
||||
BitvectorProxy &operator=(const BitvectorProxy &other) {
|
||||
return *this = (bool)other;
|
||||
}
|
||||
|
||||
operator bool() const {
|
||||
return word & mask;
|
||||
}
|
||||
|
||||
private:
|
||||
u64 &word;
|
||||
u64 mask;
|
||||
};
|
||||
|
||||
class Bitvector {
|
||||
public:
|
||||
Bitvector() = default;
|
||||
Bitvector(i64 n) : size(n), words((n + 63) / 64) {}
|
||||
|
||||
void resize(i64 n) {
|
||||
words.clear();
|
||||
words.resize((n + 63) / 64);
|
||||
size = n;
|
||||
}
|
||||
|
||||
Bitvector &operator|=(const Bitvector &x) {
|
||||
assert(size == x.size);
|
||||
for (i64 i = 0; i < words.size(); i++)
|
||||
words[i] |= x.words[i];
|
||||
return *this;
|
||||
}
|
||||
|
||||
Bitvector &operator&=(const Bitvector &x) {
|
||||
assert(size == x.size);
|
||||
for (i64 i = 0; i < words.size(); i++)
|
||||
words[i] &= x.words[i];
|
||||
return *this;
|
||||
}
|
||||
|
||||
Bitvector &operator<<=(size_t n) {
|
||||
assert(n == 1);
|
||||
for (i64 i = words.size() - 1; i > 0; i--)
|
||||
words[i] = (words[i] << 1) | (words[i - 1] >> 63);
|
||||
words[0] <<= 1;
|
||||
return *this;
|
||||
}
|
||||
|
||||
BitvectorProxy operator[](size_t pos) {
|
||||
assert(pos < size);
|
||||
return {words[pos / 64], pos % 64};
|
||||
}
|
||||
|
||||
i64 size = 0;
|
||||
std::vector<u64> words;
|
||||
};
|
||||
|
||||
} // namespace mold
|
||||
-174
@@ -1,174 +0,0 @@
|
||||
// This file implements a multi-threaded zlib and zstd compression
|
||||
// routine.
|
||||
//
|
||||
// zlib-compressed data can be merged just by concatenation as long as
|
||||
// each piece of data is flushed with Z_SYNC_FLUSH. In this file, we
|
||||
// split input data into multiple shards, compress them individually
|
||||
// and concatenate them. We then append a header, a trailer and a
|
||||
// checksum so that the concatenated data is valid zlib-format data.
|
||||
//
|
||||
// zstd-compressed data can be merged in the same way.
|
||||
//
|
||||
// Using threads to compress data has a downside. Since the dictionary
|
||||
// is reset on boundaries of shards, compression ratio is sacrificed
|
||||
// a little bit. However, if a shard size is large enough, that loss
|
||||
// is negligible in practice.
|
||||
|
||||
#include "lib.h"
|
||||
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <zlib.h>
|
||||
#include <zstd.h>
|
||||
|
||||
#define CHECK(fn) \
|
||||
do { \
|
||||
[[maybe_unused]] int r = (fn); \
|
||||
assert(r == Z_OK); \
|
||||
} while (0)
|
||||
|
||||
namespace mold {
|
||||
|
||||
static constexpr i64 SHARD_SIZE = 1024 * 1024;
|
||||
|
||||
Compressor::~Compressor() {
|
||||
for (std::span<u8> shard : shards)
|
||||
delete[] shard.data();
|
||||
}
|
||||
|
||||
static std::vector<std::span<u8>> split(std::span<u8> input) {
|
||||
std::vector<std::span<u8>> vec;
|
||||
while (!input.empty()) {
|
||||
i64 sz = std::min<i64>(SHARD_SIZE, input.size());
|
||||
vec.push_back(input.subspan(0, sz));
|
||||
input = input.subspan(sz);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
static std::span<u8> zlib_compress(std::span<u8> input, int level) {
|
||||
// Initialize zlib stream. Since debug info is generally compressed
|
||||
// pretty well with lower compression levels, the default level is 1.
|
||||
z_stream strm = {};
|
||||
CHECK(deflateInit2(&strm, level, Z_DEFLATED, -15, 8, Z_DEFAULT_STRATEGY));
|
||||
|
||||
// Set an input buffer
|
||||
strm.avail_in = input.size();
|
||||
strm.next_in = input.data();
|
||||
|
||||
// Set an output buffer. deflateBound() returns an upper bound
|
||||
// on the compression size. +16 for Z_SYNC_FLUSH.
|
||||
i64 bufsize = deflateBound(&strm, strm.avail_in) + 16;
|
||||
u8 *buf = new u8[bufsize];
|
||||
|
||||
// Compress data. It writes all compressed bytes except the last
|
||||
// partial byte, so up to 7 bits can be held to be written to the
|
||||
// buffer.
|
||||
strm.avail_out = bufsize;
|
||||
strm.next_out = buf;
|
||||
CHECK(deflate(&strm, Z_BLOCK));
|
||||
|
||||
// This is a workaround for libbacktrace before 2022-04-06.
|
||||
//
|
||||
// Zlib is a bit stream, and what Z_SYNC_FLUSH does is to write a
|
||||
// three bit value indicating the start of an uncompressed data
|
||||
// block followed by four byte data 00 00 ff ff which indicates that
|
||||
// the length of the block is zero. libbacktrace uses its own zlib
|
||||
// inflate routine, and it had a bug that if that particular three
|
||||
// bit value happens to end at a byte boundary, it accidentally
|
||||
// skipped the next byte.
|
||||
//
|
||||
// In order to avoid triggering that bug, we should avoid calling
|
||||
// deflate() with Z_SYNC_FLUSH if the current bit position is 5.
|
||||
// If it's 5, we insert an empty block consisting of 10 bits so
|
||||
// that the bit position is 7 in the next byte.
|
||||
//
|
||||
// https://github.com/ianlancetaylor/libbacktrace/pull/87
|
||||
int nbits = 0;
|
||||
deflatePending(&strm, Z_NULL, &nbits);
|
||||
if (nbits == 5)
|
||||
CHECK(deflatePrime(&strm, 10, 2));
|
||||
CHECK(deflate(&strm, Z_SYNC_FLUSH));
|
||||
|
||||
deflateEnd(&strm);
|
||||
return {buf, (size_t)(bufsize - strm.avail_out)};
|
||||
}
|
||||
|
||||
ZlibCompressor::ZlibCompressor(u8 *buf, i64 size, i64 level) {
|
||||
std::vector<std::span<u8>> inputs = split(std::span(buf, size));
|
||||
std::vector<u32> adlers(inputs.size());
|
||||
shards.resize(inputs.size());
|
||||
|
||||
// Compress each shard
|
||||
tbb::parallel_for((i64)0, (i64)inputs.size(), [&](i64 i) {
|
||||
std::span<u8> in = inputs[i];
|
||||
adlers[i] = adler32(1, in.data(), in.size());
|
||||
shards[i] = zlib_compress(in, level);
|
||||
});
|
||||
|
||||
// Combine checksums
|
||||
checksum = adlers[0];
|
||||
for (i64 i = 1; i < inputs.size(); i++)
|
||||
checksum = adler32_combine(checksum, adlers[i], inputs[i].size());
|
||||
|
||||
// Comput the total size
|
||||
compressed_size = 8; // the header and the trailer
|
||||
for (std::span<u8> &shard : shards)
|
||||
compressed_size += shard.size();
|
||||
}
|
||||
|
||||
void ZlibCompressor::write_to(u8 *buf) {
|
||||
// Write a zlib-format header
|
||||
buf[0] = 0x78;
|
||||
buf[1] = 0x9c;
|
||||
|
||||
// Copy compressed data
|
||||
std::vector<i64> offsets(shards.size());
|
||||
offsets[0] = 2; // +2 for the header
|
||||
for (i64 i = 1; i < shards.size(); i++)
|
||||
offsets[i] = offsets[i - 1] + shards[i - 1].size();
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)shards.size(), [&](i64 i) {
|
||||
memcpy(buf + offsets[i], shards[i].data(), shards[i].size());
|
||||
});
|
||||
|
||||
// Write a trailer
|
||||
u8 *end = buf + compressed_size;
|
||||
end[-6] = 3;
|
||||
end[-5] = 0;
|
||||
*(ub32 *)(end - 4) = checksum;
|
||||
}
|
||||
|
||||
static std::span<u8> zstd_compress(std::span<u8> input, int level) {
|
||||
i64 bufsize = ZSTD_COMPRESSBOUND(input.size());
|
||||
u8 *buf = new u8[bufsize];
|
||||
size_t sz = ZSTD_compress(buf, bufsize, input.data(), input.size(), level);
|
||||
assert(!ZSTD_isError(sz));
|
||||
return {buf, sz};
|
||||
}
|
||||
|
||||
ZstdCompressor::ZstdCompressor(u8 *buf, i64 size, i64 level) {
|
||||
std::vector<std::span<u8>> inputs = split(std::span(buf, size));
|
||||
shards.resize(inputs.size());
|
||||
|
||||
// Compress each shard
|
||||
tbb::parallel_for((i64)0, (i64)inputs.size(), [&](i64 i) {
|
||||
shards[i] = zstd_compress(inputs[i], level);
|
||||
});
|
||||
|
||||
compressed_size = 0;
|
||||
for (std::span<u8> &shard : shards)
|
||||
compressed_size += shard.size();
|
||||
}
|
||||
|
||||
void ZstdCompressor::write_to(u8 *buf) {
|
||||
// Copy compressed data
|
||||
std::vector<i64> offsets(shards.size());
|
||||
for (i64 i = 1; i < shards.size(); i++)
|
||||
offsets[i] = offsets[i - 1] + shards[i - 1].size();
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)shards.size(), [&](i64 i) {
|
||||
memcpy(buf + offsets[i], shards[i].data(), shards[i].size());
|
||||
});
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,29 +0,0 @@
|
||||
#define MOLD_VERSION "@mold_VERSION@"
|
||||
#define MOLD_LIBDIR "@CMAKE_INSTALL_FULL_LIBDIR@"
|
||||
#define MOLD_FIRST_TARGET @MOLD_FIRST_TARGET@
|
||||
|
||||
#cmakedefine01 HAVE_MADVISE
|
||||
#cmakedefine01 HAVE_UNAME
|
||||
#cmakedefine01 MOLD_USE_MIMALLOC
|
||||
#cmakedefine01 MOLD_USE_SYSTEM_MIMALLOC
|
||||
|
||||
#cmakedefine01 HAVE_TARGET_X86_64
|
||||
#cmakedefine01 HAVE_TARGET_I386
|
||||
#cmakedefine01 HAVE_TARGET_ARM64LE
|
||||
#cmakedefine01 HAVE_TARGET_ARM64BE
|
||||
#cmakedefine01 HAVE_TARGET_ARM32LE
|
||||
#cmakedefine01 HAVE_TARGET_ARM32BE
|
||||
#cmakedefine01 HAVE_TARGET_RV32LE
|
||||
#cmakedefine01 HAVE_TARGET_RV32BE
|
||||
#cmakedefine01 HAVE_TARGET_RV64LE
|
||||
#cmakedefine01 HAVE_TARGET_RV64BE
|
||||
#cmakedefine01 HAVE_TARGET_PPC32
|
||||
#cmakedefine01 HAVE_TARGET_PPC64V1
|
||||
#cmakedefine01 HAVE_TARGET_PPC64V2
|
||||
#cmakedefine01 HAVE_TARGET_S390X
|
||||
#cmakedefine01 HAVE_TARGET_SPARC64
|
||||
#cmakedefine01 HAVE_TARGET_M68K
|
||||
#cmakedefine01 HAVE_TARGET_SH4LE
|
||||
#cmakedefine01 HAVE_TARGET_SH4BE
|
||||
#cmakedefine01 HAVE_TARGET_LOONGARCH64
|
||||
#cmakedefine01 HAVE_TARGET_LOONGARCH32
|
||||
@@ -1,60 +0,0 @@
|
||||
#include "lib.h"
|
||||
|
||||
#include <tbb/parallel_for_each.h>
|
||||
#include <zlib.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
// This function "forges" a CRC. That is, given the current and a desired
|
||||
// CRC32 value, crc32_solve() returns a binary blob to add to the end of
|
||||
// the original data to yield the desired CRC. Trailing garbage is ignored
|
||||
// by many bianry file formats, so you can create a file with a desired
|
||||
// CRC using crc32_solve(). We need it for --separate-debug-file.
|
||||
std::vector<u8> crc32_solve(u32 current, u32 desired) {
|
||||
constexpr u32 poly = 0xedb88320;
|
||||
u32 x = ~desired;
|
||||
|
||||
// Each iteration computes x = (x * x^-1) mod poly.
|
||||
for (i64 i = 0; i < 32; i++) {
|
||||
x = std::rotl(x, 1);
|
||||
x ^= (x & 1) * (poly << 1);
|
||||
}
|
||||
|
||||
x ^= ~current;
|
||||
|
||||
std::vector<u8> out(4);
|
||||
out[0] = x;
|
||||
out[1] = x >> 8;
|
||||
out[2] = x >> 16;
|
||||
out[3] = x >> 24;
|
||||
return out;
|
||||
}
|
||||
|
||||
// Compute a CRC for given data in parallel
|
||||
u32 compute_crc32(u32 crc, u8 *buf, i64 len) {
|
||||
struct Shard {
|
||||
u8 *buf;
|
||||
i64 len;
|
||||
u32 crc;
|
||||
};
|
||||
|
||||
constexpr i64 shard_size = 1024 * 1024; // 1 MiB
|
||||
std::vector<Shard> shards;
|
||||
|
||||
while (len > 0) {
|
||||
i64 sz = std::min(len, shard_size);
|
||||
shards.push_back({buf, sz, 0});
|
||||
buf += sz;
|
||||
len -= sz;
|
||||
}
|
||||
|
||||
tbb::parallel_for_each(shards, [](Shard &shard) {
|
||||
shard.crc = crc32(0, shard.buf, shard.len);
|
||||
});
|
||||
|
||||
for (Shard &shard : shards)
|
||||
crc = crc32_combine(crc, shard.crc, shard.len);
|
||||
return crc;
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,42 +0,0 @@
|
||||
#include "lib.h"
|
||||
|
||||
#include <cstdlib>
|
||||
|
||||
#ifndef _WIN32
|
||||
#include <cxxabi.h>
|
||||
#endif
|
||||
|
||||
#include "../third-party/rust-demangle/rust-demangle.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
std::optional<std::string_view> demangle_cpp(std::string_view name) {
|
||||
static thread_local char *buf;
|
||||
static thread_local size_t buflen;
|
||||
|
||||
// TODO(cwasser): Actually demangle Symbols on Windows using e.g.
|
||||
// `UnDecorateSymbolName` from Dbghelp, maybe even Itanium symbols?
|
||||
#ifndef _WIN32
|
||||
if (name.starts_with("_Z")) {
|
||||
int status;
|
||||
char *p = abi::__cxa_demangle(std::string(name).c_str(), buf, &buflen, &status);
|
||||
if (status == 0) {
|
||||
buf = p;
|
||||
return p;
|
||||
}
|
||||
}
|
||||
#endif
|
||||
|
||||
return {};
|
||||
}
|
||||
|
||||
std::optional<std::string_view> demangle_rust(std::string_view name) {
|
||||
static thread_local char *buf;
|
||||
free(buf);
|
||||
buf = rust_demangle(std::string(name).c_str(), 0);
|
||||
if (buf)
|
||||
return buf;
|
||||
return {};
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,42 +0,0 @@
|
||||
#include "lib.h"
|
||||
|
||||
#include <filesystem>
|
||||
#include <sys/stat.h>
|
||||
|
||||
#ifdef __APPLE__
|
||||
# include <mach-o/dyld.h>
|
||||
#endif
|
||||
|
||||
#ifdef __FreeBSD__
|
||||
# include <sys/sysctl.h>
|
||||
#endif
|
||||
|
||||
namespace mold {
|
||||
|
||||
// Returns the path of the mold executable itself
|
||||
std::string get_self_path() {
|
||||
#if __APPLE__ || _WIN32
|
||||
fprintf(stderr, "mold: get_self_path is not supported");
|
||||
exit(1);
|
||||
#elif __FreeBSD__
|
||||
// /proc may not be mounted on FreeBSD. The proper way to get the
|
||||
// current executable's path is to use sysctl(2).
|
||||
int mib[4];
|
||||
mib[0] = CTL_KERN;
|
||||
mib[1] = KERN_PROC;
|
||||
mib[2] = KERN_PROC_PATHNAME;
|
||||
mib[3] = -1;
|
||||
|
||||
size_t size;
|
||||
sysctl(mib, 4, NULL, &size, NULL, 0);
|
||||
|
||||
std::string path;
|
||||
path.resize(size);
|
||||
sysctl(mib, 4, path.data(), &size, NULL, 0);
|
||||
return path;
|
||||
#else
|
||||
return std::filesystem::read_symlink("/proc/self/exe").string();
|
||||
#endif
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,69 +0,0 @@
|
||||
#!/bin/bash
|
||||
#
|
||||
# This test script takes a Gentoo package name and tries to build it
|
||||
# using mold in a Podman environment. We chose Gentoo Linux as a test
|
||||
# target, because its source-based package allows us to build programs
|
||||
# locally and run their test suites without any hassle.
|
||||
#
|
||||
# You can get a complete list of Gentoo packages availalbe for testing
|
||||
# with the following command:
|
||||
#
|
||||
# podman run --rm mold-gentoo emerge --color n -s '' | \
|
||||
# perl -ne 'next unless m!^\*\s+(\S+/\S+)!; print "$1\n"'
|
||||
|
||||
package="$1"
|
||||
|
||||
if [ "$package" = "" ]; then
|
||||
echo "Usage: $0 gentoo-package-name"
|
||||
exit 1
|
||||
fi
|
||||
|
||||
set -x
|
||||
|
||||
# Create a Podman image
|
||||
if ! podman image ls mold-gentoo | grep -q mold-gentoo; then
|
||||
set -e
|
||||
cat <<EOF | podman build -t mold-gentoo -
|
||||
FROM docker.io/gentoo/stage3
|
||||
RUN emerge-webrsync
|
||||
RUN echo 'USE="X ssl elogind -systemd corefonts truetype jpeg jpeg2k tiff zstd static-libs binary -perl"' >> /etc/portage/make.conf && \
|
||||
echo 'ACCEPT_KEYWORDS="~amd64"' >> /etc/portage/make.conf && \
|
||||
echo 'ACCEPT_LICENSE="* -@EULA"' >> /etc/portage/make.conf && \
|
||||
echo 'FEATURES="\${FEATURE} noclean nostrip ccache -ipc-sandbox -network-sandbox -pid-sandbox -sandbox"' >> /etc/portage/make.conf && \
|
||||
echo 'CCACHE_DIR="/ccache"' >> /etc/portage/make.conf && \
|
||||
FEATURES='getbinpkg binpkg-request-signature' emerge gdb lld llvm-core/clang vim emacs strace ccache xeyes dev-build/cmake dev-vcs/git && \
|
||||
rm -rf /var/tmp/portage
|
||||
EOF
|
||||
set +e
|
||||
fi
|
||||
|
||||
git_hash=$(./dist/mold --version | perl -ne '/\((\w+)/; print $1;')
|
||||
|
||||
if [ "$package" = dev-libs/concurrencykit ]; then
|
||||
echo "Skipping known broken package: $package"
|
||||
exit 0
|
||||
fi
|
||||
|
||||
# Build a given package in Podman
|
||||
cmd1='(cd /usr/bin; ln -sf /mold/dist/mold $(realpath ld))'
|
||||
cmd2="MAKEOPTS=-'j$(nproc) --load-average=100' emerge --onlydeps $package"
|
||||
cmd3="MAKEOPTS='-j$(nproc) --load-average=100' FEATURES=test emerge $package"
|
||||
filename=`echo "$package" | sed 's!/!_!g'`
|
||||
podman="podman run --rm --pids-limit=-1 --cap-add=SYS_PTRACE -v `pwd`:/mold:ro -v /var/cache/ccache-gentoo:/ccache mold-gentoo timeout -v -k 15s 3h"
|
||||
dir=gentoo/$git_hash
|
||||
|
||||
mkdir -p "$dir"/success "$dir"/failure
|
||||
|
||||
$podman chrt --idle 0 nice -n 19 bash -c "$cmd1 && $cmd2 && $cmd3" >& "$dir"/"$filename".mold
|
||||
if [ $? = 0 ]; then
|
||||
mv "$dir"/"$filename".mold "$dir"/success
|
||||
else
|
||||
mv "$dir"/"$filename".mold "$dir"/failure
|
||||
fi
|
||||
|
||||
$podman chrt --idle 0 nice -n 19 bash -c "$cmd2 && $cmd3" >& "$dir"/"$filename".ld
|
||||
if [ $? = 0 ]; then
|
||||
mv "$dir"/"$filename".ld "$dir"/success
|
||||
else
|
||||
mv "$dir"/"$filename".ld "$dir"/failure
|
||||
fi
|
||||
-256
@@ -1,256 +0,0 @@
|
||||
// This file implements a glob matcher that can run multiple glob patterns
|
||||
// against an input string. mold uses the glob matcher for symbol name
|
||||
// patterns in a version script or a dynamic list file. Since we may need to
|
||||
// match hundreds of glob patterns against millions of symbol names, the
|
||||
// speed of the matcher is very important.
|
||||
//
|
||||
// The pattern match implemented in this file is NFA-based, although the
|
||||
// cost of the function is O(n*m), where n is the number of NFA states and m
|
||||
// is the length of the input string. We do not use recursion or
|
||||
// backtracking, unlike a generic NFA-based regular expression matcher. This
|
||||
// is doable because glob patterns are very limited subsets of regexes.
|
||||
//
|
||||
// Here is the explanation of the algorithm. Observe that the only "tricky"
|
||||
// meta-character in a glob pattern is "*", which matches zero or more
|
||||
// characters. Other characters and meta-characters always match a single
|
||||
// input character. So the key of the algorithm is to handle "*" efficiently.
|
||||
//
|
||||
// We can represent a glob pattern "a*b*" with three NFA states: q_start, q1
|
||||
// and q_accept, with the following transition functions:
|
||||
//
|
||||
// δ(q_start, "a") = q1
|
||||
// δ(q1, <any character>) = q1
|
||||
// δ(q1, "b") = q_accept
|
||||
// δ(q_accept, <any character>) = q_accept
|
||||
//
|
||||
// We can construct such an NFA in a straightforward manner. We maintain NFA
|
||||
// states as a list, with the initial contents being the start state. Each
|
||||
// character except for "*" creates a new NFA state, adds a transition from
|
||||
// the last state in the list to the new one, and appends the new state at
|
||||
// the end of the list. "*" sets the "is_star" flag on the last NFA state.
|
||||
// The flag indicates that the state machine can remain in the state for any
|
||||
// input character.
|
||||
//
|
||||
// An NFA constructed this way doesn't have any complicated loops,
|
||||
// ε-transitions, or anything like that. Each state has only one incoming
|
||||
// edge. The only loops in the state transition are the self-loops on states
|
||||
// followed by a "*". Aside from that, the state machine progresses linearly
|
||||
// from the start state to the accept state.
|
||||
//
|
||||
// Each state of an NFA can be represented by a single bit. If a bit is 1,
|
||||
// the non-deterministic state machine is in that state. Otherwise, it's
|
||||
// not. Observe that a state with the "is_star" flag will continued to be 1
|
||||
// once it becomes 1, since the state machine can loop over the state on any
|
||||
// input character.
|
||||
//
|
||||
// With that observation, we can represent an NFA with a bit vector of N
|
||||
// bits, where N is the number of NFA states. For each input character, bit
|
||||
// M becomes 1 if
|
||||
//
|
||||
// - bit M-1 is 1 and there's a transition from state_{M-1} to state_M
|
||||
// with the given character, or
|
||||
// - bit M is 1 and state_M's "is_star" flag is 1.
|
||||
//
|
||||
// Initially, the 0th bit is 1 for the start state. At each step, the bits
|
||||
// propagate from least significant to most significant positions, at most
|
||||
// one bit at a time. If the most significant bit is 1 after the entire
|
||||
// input has been processed, the string matches.
|
||||
//
|
||||
// This propagation can be implemented with bitwise OR, bitwise AND, and a
|
||||
// one-bit bit shift on the bit vector. All these operations are very cheap.
|
||||
//
|
||||
// We can combine multiple glob matchers into a single matcher by simply
|
||||
// concatenating the bit vectors of their state machines.
|
||||
|
||||
#include "lib.h"
|
||||
|
||||
#include <cstring>
|
||||
|
||||
namespace mold {
|
||||
|
||||
static std::vector<MultiGlob::State> parse_glob(std::string_view pat) {
|
||||
std::vector<MultiGlob::State> vec(1);
|
||||
|
||||
while (!pat.empty()) {
|
||||
u8 c = pat[0];
|
||||
pat = pat.substr(1);
|
||||
std::bitset<256> chars;
|
||||
|
||||
switch (c) {
|
||||
case '*':
|
||||
vec.back().is_star = true;
|
||||
continue;
|
||||
case '?':
|
||||
chars.set();
|
||||
break;
|
||||
case '\\':
|
||||
if (pat.empty())
|
||||
return {};
|
||||
chars[pat[0]] = true;
|
||||
pat = pat.substr(1);
|
||||
break;
|
||||
case '[': {
|
||||
// Here are a few bracket pattern examples:
|
||||
//
|
||||
// [abc]: a, b or c
|
||||
// [$\]!]: $, ] or !
|
||||
// [a-czg-i]: a, b, c, z, g, h, or i
|
||||
// [!a-z]: Any character except lowercase letters
|
||||
//
|
||||
// Both `!` and `^` are accepted as negation markers. `!` is the
|
||||
// POSIX/shell convention used by other linkers. `^` was mold's
|
||||
// original syntax and is kept for backward compatibility.
|
||||
bool negate = false;
|
||||
bool closed = false;
|
||||
|
||||
if (!pat.empty() && (pat[0] == '!' || pat[0] == '^')) {
|
||||
negate = true;
|
||||
pat = pat.substr(1);
|
||||
}
|
||||
|
||||
while (!pat.empty()) {
|
||||
if (pat[0] == ']') {
|
||||
pat = pat.substr(1);
|
||||
closed = true;
|
||||
break;
|
||||
}
|
||||
|
||||
if (pat[0] == '\\') {
|
||||
pat = pat.substr(1);
|
||||
if (pat.empty())
|
||||
return {};
|
||||
}
|
||||
|
||||
if (pat.size() >= 3 && pat[1] == '-') {
|
||||
u8 start = pat[0];
|
||||
u8 end = pat[2];
|
||||
pat = pat.substr(3);
|
||||
|
||||
if (end == '\\') {
|
||||
if (pat.empty())
|
||||
return {};
|
||||
end = pat[0];
|
||||
pat = pat.substr(1);
|
||||
}
|
||||
|
||||
if (end < start)
|
||||
return {};
|
||||
|
||||
for (i64 i = start; i <= end; i++)
|
||||
chars[i] = true;
|
||||
} else {
|
||||
chars[pat[0]] = true;
|
||||
pat = pat.substr(1);
|
||||
}
|
||||
}
|
||||
|
||||
if (!closed)
|
||||
return {};
|
||||
|
||||
if (negate)
|
||||
chars.flip();
|
||||
break;
|
||||
}
|
||||
default:
|
||||
chars[c] = true;
|
||||
break;
|
||||
}
|
||||
|
||||
vec.push_back({chars, false});
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
// Instead of returning just a match/no match boolean value, our glob
|
||||
// matcher returns an integer value associated with each given pattern.
|
||||
// If multiple patterns match at the same time, the largest associated
|
||||
// value will be returned by find().
|
||||
bool MultiGlob::add(std::string_view pat, i64 val) {
|
||||
std::vector<State> vec = parse_glob(pat);
|
||||
if (vec.empty())
|
||||
return false;
|
||||
patterns.push_back({std::move(vec), val});
|
||||
return true;
|
||||
}
|
||||
|
||||
void MultiGlob::compile() {
|
||||
if (patterns.empty())
|
||||
return;
|
||||
|
||||
ranges::stable_sort(patterns, ranges::greater(), &GlobPattern::value);
|
||||
|
||||
std::vector<State> states;
|
||||
for (GlobPattern &p : patterns)
|
||||
append(states, p.states);
|
||||
i64 sz = states.size();
|
||||
|
||||
start_states.resize(sz);
|
||||
for (i64 pos = 0; GlobPattern &p : patterns) {
|
||||
start_states[pos] = true;
|
||||
pos += p.states.size();
|
||||
}
|
||||
|
||||
star_mask.resize(sz);
|
||||
for (i64 i = 0; i < sz; i++)
|
||||
if (states[i].is_star)
|
||||
star_mask[i] = true;
|
||||
|
||||
for (i64 i = 0; i < 256; i++) {
|
||||
char_mask[i].resize(sz);
|
||||
for (i64 j = 0; j < sz; j++)
|
||||
if (states[j].incoming_edge[i])
|
||||
char_mask[i][j] = true;
|
||||
}
|
||||
}
|
||||
|
||||
i64 MultiGlob::find(std::string_view str) {
|
||||
if (patterns.empty())
|
||||
return -1;
|
||||
|
||||
Bitvector bits = start_states;
|
||||
Bitvector tmp;
|
||||
|
||||
for (u8 c : str) {
|
||||
// This is equivalent to
|
||||
//
|
||||
// bits = (bits & star_mask) | ((bits << 1) & char_mask[c])
|
||||
//
|
||||
// but we update the existing objects in place to avoid allocating
|
||||
// temporary objects.
|
||||
tmp = bits;
|
||||
tmp &= star_mask;
|
||||
bits <<= 1;
|
||||
bits &= char_mask[c];
|
||||
bits |= tmp;
|
||||
}
|
||||
|
||||
for (i64 pos = 0; GlobPattern &p : patterns) {
|
||||
pos += p.states.size();
|
||||
if (bits[pos - 1])
|
||||
return p.value;
|
||||
}
|
||||
return -1;
|
||||
}
|
||||
|
||||
bool Glob::add(std::string_view pat, i64 val) {
|
||||
assert(val >= 0);
|
||||
assert(!is_compiled);
|
||||
|
||||
// If the pattern requires only a single substring search, the
|
||||
// Aho-Corasick algorithm is even faster than our glob matcher.
|
||||
if (aho_corasick.can_handle(pat))
|
||||
return aho_corasick.add(pat, val);
|
||||
return multi_glob.add(pat, val);
|
||||
}
|
||||
|
||||
i64 Glob::find(std::string_view str) {
|
||||
std::call_once(once, [&] {
|
||||
multi_glob.compile();
|
||||
aho_corasick.compile();
|
||||
is_compiled = true;
|
||||
});
|
||||
|
||||
return std::max(multi_glob.find(str), aho_corasick.find(str));
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
-127
@@ -1,127 +0,0 @@
|
||||
// This file defines integral types for file input/output. We need to use
|
||||
// these types instead of the plain integers (such as uint32_t or int32_t)
|
||||
// when reading from/writing to an mmap'ed file area for the following
|
||||
// reasons:
|
||||
//
|
||||
// 1. mold is always a cross linker and should not depend on what host it
|
||||
// is running on. For example, users should be able to run mold on a
|
||||
// little-endian x86 machine to create a big-endian s390x binary.
|
||||
//
|
||||
// 2. Even though data members in all ELF data strucutres are naturally
|
||||
// aligned, they are not guaranteed to be aligned on memory because of
|
||||
// archive files. Archive files (.a files) align each file only to a
|
||||
// 2 byte boundary, so anything larger than 2 bytes may be misaligned
|
||||
// in an mmap'ed memory. Misaligned access is an undefined behavior in
|
||||
// C/C++, so we shouldn't cast an arbitrary pointer to a uint32_t, for
|
||||
// example, to read a 32 bit value.
|
||||
//
|
||||
// The data types defined in this file are independent of the host byte
|
||||
// order and are designed to avoid unaligned access.
|
||||
//
|
||||
// Note that in C/C++, memcpy is a portable and efficient way to access
|
||||
// unaligned data, as it is typically treated as an intrinsic. Compilers
|
||||
// can easily optimize memcpy calls in this file into a single load or
|
||||
// store instruction.
|
||||
|
||||
#pragma once
|
||||
|
||||
#include <bit>
|
||||
#include <cstdint>
|
||||
#include <cstring>
|
||||
#include <type_traits>
|
||||
|
||||
#if !defined(__LITTLE_ENDIAN__) && !defined(__BIG_ENDIAN__)
|
||||
# if __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
|
||||
# define __LITTLE_ENDIAN__ 1
|
||||
# elif __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__
|
||||
# define __BIG_ENDIAN__ 1
|
||||
# else
|
||||
# error "unknown host byte order"
|
||||
# endif
|
||||
#endif
|
||||
|
||||
namespace mold {
|
||||
|
||||
template <typename T, bool is_le, int size = sizeof(T)>
|
||||
class Integer {
|
||||
public:
|
||||
constexpr Integer() = default;
|
||||
|
||||
constexpr Integer(T v) {
|
||||
if (std::is_constant_evaluated() || size == 3) {
|
||||
for (int i = 0; i < size; i++)
|
||||
buf[is_le ? i : (size - i - 1)] = v >> (i * 8);
|
||||
} else {
|
||||
if (!is_native)
|
||||
v = bswap(v);
|
||||
memcpy(buf, &v, size);
|
||||
}
|
||||
}
|
||||
|
||||
operator T() const {
|
||||
if (size == 3) {
|
||||
if (is_le)
|
||||
return buf[2] << 16 | buf[1] << 8 | buf[0];
|
||||
return buf[0] << 16 | buf[1] << 8 | buf[2];
|
||||
}
|
||||
|
||||
T v;
|
||||
memcpy(&v, buf, size);
|
||||
return is_native ? v : bswap(v);
|
||||
}
|
||||
|
||||
Integer &operator=(T v) { new (this) Integer(v); return *this; }
|
||||
Integer &operator++() { return *this = *this + 1; }
|
||||
Integer operator++(int) { auto x = *this; ++*this; return x; }
|
||||
Integer &operator--() { return *this = *this - 1; }
|
||||
Integer operator--(int) { auto x = *this; --*this; return x; }
|
||||
Integer &operator+=(T v) { return *this = *this + v; }
|
||||
Integer &operator-=(T v) { return *this = *this - v; }
|
||||
Integer &operator&=(T v) { return *this = *this & v; }
|
||||
Integer &operator|=(T v) { return *this = *this | v; }
|
||||
|
||||
private:
|
||||
static constexpr bool is_native =
|
||||
(std::endian::native == (is_le ? std::endian::little : std::endian::big));
|
||||
|
||||
static T bswap(T v) {
|
||||
switch (size) {
|
||||
case 2: return __builtin_bswap16(v);
|
||||
case 4: return __builtin_bswap32(v);
|
||||
case 8: return __builtin_bswap64(v);
|
||||
default: __builtin_unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
uint8_t buf[size];
|
||||
};
|
||||
|
||||
using i8 = int8_t;
|
||||
using i16 = int16_t;
|
||||
using i32 = int32_t;
|
||||
using i64 = int64_t;
|
||||
|
||||
using u8 = uint8_t;
|
||||
using u16 = uint16_t;
|
||||
using u32 = uint32_t;
|
||||
using u64 = uint64_t;
|
||||
|
||||
using il16 = Integer<i16, true>;
|
||||
using il32 = Integer<i32, true>;
|
||||
using il64 = Integer<i64, true>;
|
||||
|
||||
using ul16 = Integer<u16, true>;
|
||||
using ul24 = Integer<u32, true, 3>;
|
||||
using ul32 = Integer<u32, true>;
|
||||
using ul64 = Integer<u64, true>;
|
||||
|
||||
using ib16 = Integer<i16, false>;
|
||||
using ib32 = Integer<i32, false>;
|
||||
using ib64 = Integer<i64, false>;
|
||||
|
||||
using ub16 = Integer<u16, false>;
|
||||
using ub24 = Integer<u32, false, 3>;
|
||||
using ub32 = Integer<u32, false>;
|
||||
using ub64 = Integer<u64, false>;
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,709 +0,0 @@
|
||||
#pragma once
|
||||
|
||||
#include "atomics.h"
|
||||
#include "integers.h"
|
||||
#include "bitvector.h"
|
||||
|
||||
#include <array>
|
||||
#include <atomic>
|
||||
#include <bit>
|
||||
#include <bitset>
|
||||
#include <cassert>
|
||||
#include <cstdio>
|
||||
#include <cstdlib>
|
||||
#include <cstring>
|
||||
#include <fcntl.h>
|
||||
#include <filesystem>
|
||||
#include <iostream>
|
||||
#include <mutex>
|
||||
#include <optional>
|
||||
#include <span>
|
||||
#include <sstream>
|
||||
#include <string>
|
||||
#include <string_view>
|
||||
#include <sys/stat.h>
|
||||
#include <sys/types.h>
|
||||
#include <tbb/concurrent_vector.h>
|
||||
#include <tbb/enumerable_thread_specific.h>
|
||||
#include <tbb/parallel_for.h>
|
||||
#include <vector>
|
||||
|
||||
#ifdef _WIN32
|
||||
# include <io.h>
|
||||
#else
|
||||
# include <sys/mman.h>
|
||||
# include <unistd.h>
|
||||
#endif
|
||||
|
||||
#define XXH_INLINE_ALL 1
|
||||
#include "../third-party/xxhash/xxhash.h"
|
||||
|
||||
#ifdef NDEBUG
|
||||
# define unreachable() __builtin_unreachable()
|
||||
#else
|
||||
# define unreachable() assert(0 && "unreachable")
|
||||
#endif
|
||||
|
||||
inline uint64_t hash_string(std::string_view str) {
|
||||
return XXH3_64bits(str.data(), str.size());
|
||||
}
|
||||
|
||||
class HashCmp {
|
||||
public:
|
||||
static size_t hash(const std::string_view &k) {
|
||||
return hash_string(k);
|
||||
}
|
||||
|
||||
static bool equal(const std::string_view &k1, const std::string_view &k2) {
|
||||
return k1 == k2;
|
||||
}
|
||||
};
|
||||
|
||||
namespace mold {
|
||||
|
||||
namespace ranges = std::ranges;
|
||||
using namespace std::literals::string_literals;
|
||||
using namespace std::literals::string_view_literals;
|
||||
|
||||
inline u64 combine_hash(u64 a, u64 b) {
|
||||
return a ^ (b + 0x9e3779b9 + (a << 6) + (a >> 2));
|
||||
}
|
||||
|
||||
//
|
||||
// perf.cc
|
||||
//
|
||||
|
||||
// Counter is used to collect statistics numbers.
|
||||
class Counter {
|
||||
public:
|
||||
Counter(std::string_view name, i64 value = 0) : name(name), values(value) {
|
||||
static std::mutex mu;
|
||||
std::scoped_lock lock(mu);
|
||||
instances.push_back(this);
|
||||
}
|
||||
|
||||
Counter &operator++(int) {
|
||||
if (enabled) [[unlikely]]
|
||||
values.local()++;
|
||||
return *this;
|
||||
}
|
||||
|
||||
Counter &operator+=(int delta) {
|
||||
if (enabled) [[unlikely]]
|
||||
values.local() += delta;
|
||||
return *this;
|
||||
}
|
||||
|
||||
static void print();
|
||||
|
||||
static inline bool enabled = false;
|
||||
|
||||
private:
|
||||
i64 get_value();
|
||||
|
||||
std::string_view name;
|
||||
tbb::enumerable_thread_specific<i64> values;
|
||||
|
||||
static inline std::vector<Counter *> instances;
|
||||
};
|
||||
|
||||
// Timer and TimeRecord records elapsed time (wall clock time)
|
||||
// used by each pass of the linker.
|
||||
struct TimerRecord {
|
||||
TimerRecord(std::string name, TimerRecord *parent = nullptr);
|
||||
void stop();
|
||||
|
||||
std::string name;
|
||||
TimerRecord *parent;
|
||||
tbb::concurrent_vector<TimerRecord *> children;
|
||||
i64 start;
|
||||
i64 end;
|
||||
i64 user;
|
||||
i64 sys;
|
||||
bool stopped = false;
|
||||
};
|
||||
|
||||
void
|
||||
print_timer_records(tbb::concurrent_vector<std::unique_ptr<TimerRecord>> &);
|
||||
|
||||
template <typename Context>
|
||||
class Timer {
|
||||
public:
|
||||
Timer(Context &ctx, std::string name, Timer *parent = nullptr) {
|
||||
record = new TimerRecord(name, parent ? parent->record : nullptr);
|
||||
ctx.timer_records.emplace_back(record);
|
||||
}
|
||||
|
||||
Timer(const Timer &) = delete;
|
||||
|
||||
~Timer() {
|
||||
record->stop();
|
||||
}
|
||||
|
||||
void stop() {
|
||||
record->stop();
|
||||
}
|
||||
|
||||
private:
|
||||
TimerRecord *record;
|
||||
};
|
||||
|
||||
//
|
||||
// Utility functions
|
||||
//
|
||||
|
||||
// Some C++ libraries haven't implemented std::has_single_bit yet.
|
||||
inline bool has_single_bit(u64 val) {
|
||||
return std::popcount(val) == 1;
|
||||
}
|
||||
|
||||
// Some C++ libraries haven't implemented std::bit_ceil yet.
|
||||
inline u64 bit_ceil(u64 val) {
|
||||
if (has_single_bit(val))
|
||||
return val;
|
||||
return 1LL << (64 - std::countl_zero(val));
|
||||
}
|
||||
|
||||
inline u64 align_to(u64 val, u64 align) {
|
||||
if (align == 0)
|
||||
return val;
|
||||
assert(has_single_bit(align));
|
||||
return (val + align - 1) & ~(align - 1);
|
||||
}
|
||||
|
||||
inline u64 align_down(u64 val, u64 align) {
|
||||
assert(has_single_bit(align));
|
||||
return val & ~(align - 1);
|
||||
}
|
||||
|
||||
inline u64 bit(u64 val, i64 pos) {
|
||||
return (val >> pos) & 1;
|
||||
};
|
||||
|
||||
// Returns [hi:lo] bits of val.
|
||||
inline u64 bits(u64 val, u64 hi, u64 lo) {
|
||||
return (val >> lo) & ((1LL << (hi - lo + 1)) - 1);
|
||||
}
|
||||
|
||||
// Cast val to a signed N bit integer.
|
||||
// For example, sign_extend(x, 32) == (i32)x for any integer x.
|
||||
inline i64 sign_extend(u64 val, i64 n) {
|
||||
return (i64)(val << (64 - n)) >> (64 - n);
|
||||
}
|
||||
|
||||
inline bool is_int(u64 val, i64 n) {
|
||||
return sign_extend(val, n) == val;
|
||||
}
|
||||
|
||||
template <typename T, typename Compare = std::less<T>>
|
||||
void update_minimum(std::atomic<T> &atomic, u64 new_val, Compare cmp = {}) {
|
||||
T old_val = atomic.load(std::memory_order_relaxed);
|
||||
while (cmp(new_val, old_val) &&
|
||||
!atomic.compare_exchange_weak(old_val, new_val,
|
||||
std::memory_order_relaxed));
|
||||
}
|
||||
|
||||
template <typename T, typename Compare = std::less<T>>
|
||||
void update_maximum(std::atomic<T> &atomic, u64 new_val, Compare cmp = {}) {
|
||||
T old_val = atomic.load(std::memory_order_relaxed);
|
||||
while (cmp(old_val, new_val) &&
|
||||
!atomic.compare_exchange_weak(old_val, new_val,
|
||||
std::memory_order_relaxed));
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline void append(std::vector<T> &x, const auto &y) {
|
||||
x.insert(x.end(), y.begin(), y.end());
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline std::vector<T> flatten(std::vector<std::vector<T>> &vec) {
|
||||
i64 size = 0;
|
||||
for (std::vector<T> &v : vec)
|
||||
size += v.size();
|
||||
|
||||
std::vector<T> ret;
|
||||
ret.reserve(size);
|
||||
for (std::vector<T> &v : vec)
|
||||
append(ret, v);
|
||||
return ret;
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline void remove_duplicates(std::vector<T> &vec) {
|
||||
vec.erase(std::unique(vec.begin(), vec.end()), vec.end());
|
||||
}
|
||||
|
||||
inline i64 write_string(void *buf, std::string_view str) {
|
||||
memcpy(buf, str.data(), str.size());
|
||||
*((u8 *)buf + str.size()) = '\0';
|
||||
return str.size() + 1;
|
||||
}
|
||||
|
||||
template <typename T>
|
||||
inline void write_vector(void *buf, const std::vector<T> &vec) {
|
||||
if (!vec.empty())
|
||||
memcpy(buf, vec.data(), vec.size() * sizeof(T));
|
||||
}
|
||||
|
||||
inline void encode_uleb(std::vector<u8> &vec, u64 val) {
|
||||
do {
|
||||
u8 byte = val & 0x7f;
|
||||
val >>= 7;
|
||||
vec.push_back(val ? (byte | 0x80) : byte);
|
||||
} while (val);
|
||||
}
|
||||
|
||||
inline void encode_sleb(std::vector<u8> &vec, i64 val) {
|
||||
for (;;) {
|
||||
u8 byte = val & 0x7f;
|
||||
val >>= 7;
|
||||
|
||||
bool neg = (byte & 0x40);
|
||||
if ((val == 0 && !neg) || (val == -1 && neg)) {
|
||||
vec.push_back(byte);
|
||||
break;
|
||||
}
|
||||
vec.push_back(byte | 0x80);
|
||||
}
|
||||
}
|
||||
|
||||
inline i64 write_uleb(u8 *buf, u64 val) {
|
||||
i64 i = 0;
|
||||
do {
|
||||
u8 byte = val & 0x7f;
|
||||
val >>= 7;
|
||||
buf[i++] = val ? (byte | 0x80) : byte;
|
||||
} while (val);
|
||||
return i;
|
||||
}
|
||||
|
||||
inline u64 read_uleb(u8 **buf) {
|
||||
u64 val = 0;
|
||||
u8 shift = 0;
|
||||
u8 byte;
|
||||
do {
|
||||
byte = *(*buf)++;
|
||||
val |= (byte & 0x7f) << shift;
|
||||
shift += 7;
|
||||
} while (byte & 0x80);
|
||||
return val;
|
||||
}
|
||||
|
||||
inline u64 read_uleb(u8 *buf) {
|
||||
u8 *tmp = buf;
|
||||
return read_uleb(&tmp);
|
||||
}
|
||||
|
||||
inline i64 read_sleb(u8 **buf) {
|
||||
u64 val = 0;
|
||||
u8 shift = 0;
|
||||
u8 byte;
|
||||
do {
|
||||
byte = *(*buf)++;
|
||||
val |= (byte & 0x7f) << shift;
|
||||
shift += 7;
|
||||
} while (byte & 0x80);
|
||||
return sign_extend(val, shift);
|
||||
}
|
||||
|
||||
inline i64 read_sleb(u8 *buf) {
|
||||
u8 *tmp = buf;
|
||||
return read_sleb(&tmp);
|
||||
}
|
||||
|
||||
inline u64 read_uleb(std::string_view *str) {
|
||||
u8 *start = (u8 *)str->data();
|
||||
u8 *ptr = start;
|
||||
u64 val = read_uleb(&ptr);
|
||||
*str = str->substr(ptr - start);
|
||||
return val;
|
||||
}
|
||||
|
||||
inline u64 read_uleb(std::string_view str) {
|
||||
std::string_view tmp = str;
|
||||
return read_uleb(&tmp);
|
||||
}
|
||||
|
||||
inline i64 uleb_size(u64 val) {
|
||||
for (int i = 1; i < 9; i++)
|
||||
if (val < (1LL << (7 * i)))
|
||||
return i;
|
||||
return 9;
|
||||
}
|
||||
|
||||
inline void overwrite_uleb(u8 *loc, u64 val) {
|
||||
while (*loc & 0b1000'0000) {
|
||||
*loc++ = 0b1000'0000 | (val & 0b0111'1111);
|
||||
val >>= 7;
|
||||
}
|
||||
*loc = val & 0b0111'1111;
|
||||
}
|
||||
|
||||
static inline void pause() {
|
||||
#if defined(__x86_64__)
|
||||
asm volatile("pause");
|
||||
#elif defined(__aarch64__)
|
||||
asm volatile("yield");
|
||||
#elif defined(__ARM_ARCH_7A__) || defined(__ARM_ARCH_8A__)
|
||||
asm volatile("yield");
|
||||
#endif
|
||||
}
|
||||
|
||||
//
|
||||
// Concurrent Map
|
||||
//
|
||||
|
||||
// This is an implementation of a fast concurrent hash map. Unlike
|
||||
// ordinary hash tables, this impl just aborts if it becomes full.
|
||||
// So you need to give a correct estimation of the final size before
|
||||
// using it. We use this hash map to uniquify pieces of data in
|
||||
// mergeable sections.
|
||||
//
|
||||
// We've implemented this ourselves because the performance of
|
||||
// conrurent hash map is critical for our linker.
|
||||
template <typename T>
|
||||
class ConcurrentMap {
|
||||
public:
|
||||
ConcurrentMap() = default;
|
||||
|
||||
ConcurrentMap(i64 nbuckets) {
|
||||
resize(nbuckets);
|
||||
}
|
||||
|
||||
~ConcurrentMap() {
|
||||
if (entries) {
|
||||
#ifdef _WIN32
|
||||
_aligned_free(entries);
|
||||
#else
|
||||
munmap(entries, sizeof(Entry) * nbuckets);
|
||||
#endif
|
||||
}
|
||||
}
|
||||
|
||||
// In order to avoid unnecessary cache-line false sharing, we want
|
||||
// to make this object to be aligned to a reasonably large
|
||||
// power-of-two address.
|
||||
struct alignas(32) Entry {
|
||||
Atomic<const char *> key;
|
||||
u32 keylen;
|
||||
T value;
|
||||
};
|
||||
|
||||
void resize(i64 nbuckets) {
|
||||
assert(!entries);
|
||||
this->nbuckets = std::max<i64>(MIN_NBUCKETS, bit_ceil(nbuckets));
|
||||
i64 bufsize = sizeof(Entry) * this->nbuckets;
|
||||
|
||||
// Allocate a zero-initialized buffer. mmap() is faster than
|
||||
// malloc()+memset(), and MAP_POPULATE pre-faults pages here to
|
||||
// avoid CoW faults and TLB shootdowns under concurrent inserts.
|
||||
#ifdef _WIN32
|
||||
entries = (Entry *)_aligned_malloc(bufsize, alignof(Entry));
|
||||
memset((void *)entries, 0, bufsize);
|
||||
#elif MAP_POPULATE
|
||||
entries = (Entry *)mmap(nullptr, bufsize, PROT_READ | PROT_WRITE,
|
||||
MAP_ANONYMOUS | MAP_PRIVATE | MAP_POPULATE, -1, 0);
|
||||
#else
|
||||
entries = (Entry *)mmap(nullptr, bufsize, PROT_READ | PROT_WRITE,
|
||||
MAP_ANONYMOUS | MAP_PRIVATE, -1, 0);
|
||||
#endif
|
||||
}
|
||||
|
||||
std::pair<T *, bool> insert(std::string_view key, u64 hash, const T &val) {
|
||||
assert(has_single_bit(nbuckets));
|
||||
|
||||
u64 begin = hash & (nbuckets - 1);
|
||||
u64 mask = nbuckets / NUM_SHARDS - 1;
|
||||
|
||||
for (i64 i = 0; i < MAX_RETRY; i++) {
|
||||
u64 idx = (begin & ~mask) | ((begin + i) & mask);
|
||||
Entry &ent = entries[idx];
|
||||
|
||||
// It seems avoiding compare-and-swap is faster overall at least
|
||||
// on my Zen4 machine, so do it.
|
||||
if (const char *ptr = ent.key.load(std::memory_order_acquire);
|
||||
ptr != nullptr && ptr != (char *)-1) {
|
||||
if (key == std::string_view(ptr, ent.keylen))
|
||||
return {&ent.value, false};
|
||||
continue;
|
||||
}
|
||||
|
||||
// Otherwise, use CAS to atomically claim the ownership of the slot.
|
||||
const char *ptr = nullptr;
|
||||
bool claimed = ent.key.compare_exchange_strong(ptr, (char *)-1,
|
||||
std::memory_order_acquire);
|
||||
|
||||
// If we successfully claimed the ownership of the slot,
|
||||
// copy values to it.
|
||||
if (claimed) {
|
||||
new (&ent.value) T(val);
|
||||
ent.keylen = key.size();
|
||||
ent.key.store(key.data(), std::memory_order_release);
|
||||
return {&ent.value, true};
|
||||
}
|
||||
|
||||
// If someone is copying values to the slot, do busy wait.
|
||||
while (ptr == (char *)-1) {
|
||||
pause();
|
||||
ptr = ent.key.load(std::memory_order_acquire);
|
||||
}
|
||||
|
||||
// If the same key is already present, this is the slot we are
|
||||
// looking for.
|
||||
if (key == std::string_view(ptr, ent.keylen))
|
||||
return {&ent.value, false};
|
||||
}
|
||||
|
||||
std::cerr << "ConcurrentMap is full\n";
|
||||
abort();
|
||||
}
|
||||
|
||||
i64 get_idx(T *value) const {
|
||||
uintptr_t addr = (uintptr_t)value - (uintptr_t)value % sizeof(Entry);
|
||||
return (Entry *)addr - entries;
|
||||
}
|
||||
|
||||
// Return a list of map entries sorted in a deterministic order.
|
||||
std::vector<Entry *> get_sorted_entries(i64 shard_idx) {
|
||||
if (nbuckets == 0)
|
||||
return {};
|
||||
|
||||
i64 shard_size = nbuckets / NUM_SHARDS;
|
||||
i64 begin = shard_idx * shard_size;
|
||||
i64 end = begin + shard_size;
|
||||
|
||||
i64 sz = 0;
|
||||
for (i64 i = begin; i < end; i++)
|
||||
if (entries[i].key)
|
||||
sz++;
|
||||
|
||||
std::vector<Entry *> vec;
|
||||
vec.reserve(sz);
|
||||
|
||||
// Since the shard is circular, we need to handle the last entries
|
||||
// as if they were next to the first entries.
|
||||
while (begin < end && entries[end - 1].key)
|
||||
vec.push_back(entries + --end);
|
||||
|
||||
// Find entries contiguous in the buckets and sort them.
|
||||
i64 last = 0;
|
||||
for (i64 i = begin; i < end;) {
|
||||
while (i < end && entries[i].key)
|
||||
vec.push_back(entries + i++);
|
||||
|
||||
std::sort(vec.begin() + last, vec.end(), [](Entry *a, Entry *b) {
|
||||
if (a->keylen != b->keylen)
|
||||
return a->keylen < b->keylen;
|
||||
return memcmp(a->key, b->key, a->keylen) < 0;
|
||||
});
|
||||
|
||||
last = vec.size();
|
||||
|
||||
while (i < end && !entries[i].key)
|
||||
i++;
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
std::vector<Entry *> get_sorted_entries_all() {
|
||||
std::vector<std::vector<Entry *>> vec(NUM_SHARDS);
|
||||
tbb::parallel_for((i64)0, NUM_SHARDS, [&](i64 i) {
|
||||
vec[i] = get_sorted_entries(i);
|
||||
});
|
||||
return flatten(vec);
|
||||
}
|
||||
|
||||
static constexpr i64 MIN_NBUCKETS = 4096;
|
||||
static constexpr i64 NUM_SHARDS = 16;
|
||||
static constexpr i64 MAX_RETRY = 256;
|
||||
|
||||
Entry *entries = nullptr;
|
||||
u64 nbuckets = 0;
|
||||
};
|
||||
|
||||
//
|
||||
// random.cc
|
||||
//
|
||||
|
||||
void get_random_bytes(u8 *buf, i64 size);
|
||||
|
||||
//
|
||||
// hyperloglog.cc
|
||||
//
|
||||
|
||||
class HyperLogLog {
|
||||
public:
|
||||
void insert(u64 hash) {
|
||||
update_maximum(buckets[hash & (NBUCKETS - 1)], std::countl_zero(hash) + 1);
|
||||
}
|
||||
|
||||
i64 get_cardinality() const;
|
||||
|
||||
void merge(const HyperLogLog &other) {
|
||||
for (i64 i = 0; i < NBUCKETS; i++)
|
||||
update_maximum(buckets[i], other.buckets[i]);
|
||||
}
|
||||
|
||||
private:
|
||||
static constexpr i64 NBUCKETS = 2048;
|
||||
static constexpr double ALPHA = 0.79402;
|
||||
|
||||
Atomic<u8> buckets[NBUCKETS];
|
||||
};
|
||||
|
||||
//
|
||||
// aho-corasick.cc
|
||||
//
|
||||
|
||||
class AhoCorasick {
|
||||
public:
|
||||
bool add(std::string_view pat, i64 val);
|
||||
bool empty() const { return nodes.empty(); }
|
||||
void compile();
|
||||
i64 find(std::string_view str);
|
||||
|
||||
static bool can_handle(std::string_view str);
|
||||
|
||||
private:
|
||||
struct TrieNode {
|
||||
TrieNode() { children.fill(-1); }
|
||||
i64 value = -1;
|
||||
i32 suffix_link = -1;
|
||||
std::array<i32, 256> children;
|
||||
};
|
||||
|
||||
void fix_suffix_links(i64 idx);
|
||||
void fix_values();
|
||||
|
||||
std::vector<TrieNode> nodes;
|
||||
};
|
||||
|
||||
//
|
||||
// glob.cc
|
||||
//
|
||||
|
||||
class MultiGlob {
|
||||
public:
|
||||
bool add(std::string_view pat, i64 val);
|
||||
bool empty() const { return patterns.empty(); }
|
||||
void compile();
|
||||
i64 find(std::string_view str);
|
||||
|
||||
struct State {
|
||||
std::bitset<256> incoming_edge;
|
||||
bool is_star = false;
|
||||
};
|
||||
|
||||
struct GlobPattern {
|
||||
std::vector<State> states;
|
||||
i64 value = -1;
|
||||
};
|
||||
|
||||
private:
|
||||
std::vector<GlobPattern> patterns;
|
||||
Bitvector start_states;
|
||||
Bitvector star_mask;
|
||||
Bitvector char_mask[256];
|
||||
};
|
||||
|
||||
class Glob {
|
||||
public:
|
||||
bool add(std::string_view pat, i64 val);
|
||||
bool empty() const { return multi_glob.empty() && aho_corasick.empty(); }
|
||||
i64 find(std::string_view str);
|
||||
|
||||
private:
|
||||
std::once_flag once;
|
||||
bool is_compiled = false;
|
||||
|
||||
MultiGlob multi_glob;
|
||||
AhoCorasick aho_corasick;
|
||||
};
|
||||
|
||||
//
|
||||
// filepath.cc
|
||||
//
|
||||
|
||||
inline std::filesystem::path path_dirname(std::string_view path) {
|
||||
return std::filesystem::path(path).parent_path();
|
||||
}
|
||||
|
||||
inline std::string path_filename(std::string_view path) {
|
||||
return std::filesystem::path(path).filename().string();
|
||||
}
|
||||
|
||||
inline std::string path_clean(std::string_view path) {
|
||||
return std::filesystem::path(path).lexically_normal().string();
|
||||
}
|
||||
|
||||
std::string get_self_path();
|
||||
|
||||
//
|
||||
// demangle.cc
|
||||
//
|
||||
|
||||
std::optional<std::string_view> demangle_cpp(std::string_view name);
|
||||
std::optional<std::string_view> demangle_rust(std::string_view name);
|
||||
|
||||
//
|
||||
// crc32.cc
|
||||
//
|
||||
|
||||
u32 compute_crc32(u32 crc, u8 *buf, i64 len);
|
||||
std::vector<u8> crc32_solve(u32 current, u32 desired);
|
||||
|
||||
//
|
||||
// compress.cc
|
||||
//
|
||||
|
||||
class Compressor {
|
||||
public:
|
||||
virtual void write_to(u8 *buf) = 0;
|
||||
virtual ~Compressor();
|
||||
i64 compressed_size = 0;
|
||||
|
||||
protected:
|
||||
std::vector<std::span<u8>> shards;
|
||||
};
|
||||
|
||||
class ZlibCompressor : public Compressor {
|
||||
public:
|
||||
ZlibCompressor(u8 *buf, i64 size, i64 level);
|
||||
void write_to(u8 *buf) override;
|
||||
|
||||
private:
|
||||
u32 checksum = 0;
|
||||
};
|
||||
|
||||
class ZstdCompressor : public Compressor {
|
||||
public:
|
||||
ZstdCompressor(u8 *buf, i64 size, i64 level);
|
||||
void write_to(u8 *buf) override;
|
||||
};
|
||||
|
||||
//
|
||||
// tar.cc
|
||||
//
|
||||
|
||||
// TarFile is a class to create a tar file.
|
||||
//
|
||||
// If you pass `--repro` to mold, mold collects all input files and
|
||||
// put them into `<output-file-path>.repro.tar`, so that it is easy to
|
||||
// run the same command with the same command line arguments.
|
||||
class TarWriter {
|
||||
public:
|
||||
static std::unique_ptr<TarWriter>
|
||||
open(std::string output_path, std::string basedir);
|
||||
|
||||
~TarWriter();
|
||||
void append(std::string path, std::string_view data);
|
||||
|
||||
private:
|
||||
TarWriter(FILE *out, std::string basedir) : out(out), basedir(basedir) {}
|
||||
|
||||
FILE *out = nullptr;
|
||||
std::string basedir;
|
||||
};
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,20 +0,0 @@
|
||||
#include "lib.h"
|
||||
|
||||
#include <random>
|
||||
|
||||
namespace mold {
|
||||
|
||||
void get_random_bytes(u8 *buf, i64 size) {
|
||||
std::random_device rand;
|
||||
i64 i = 0;
|
||||
|
||||
for (; i < size - 4; i += 4) {
|
||||
u32 val = rand();
|
||||
memcpy(buf + i, &val, 4);
|
||||
}
|
||||
|
||||
u32 val = rand();
|
||||
memcpy(buf + i, &val, size - i);
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
-144
@@ -1,144 +0,0 @@
|
||||
// This is a header-only C++20 implementation of SipHash based on the
|
||||
// reference implementation. To use, just copy this header file into
|
||||
// your project and #include it.
|
||||
//
|
||||
// https://github.com/rui314/siphash/blob/main/siphash.h
|
||||
|
||||
#include <bit>
|
||||
#include <cstdint>
|
||||
#include <cstring>
|
||||
|
||||
template <int C_ROUNDS, int D_ROUNDS, int OUTLEN>
|
||||
class SipHashTmpl {
|
||||
public:
|
||||
static_assert(OUTLEN == 64 || OUTLEN == 128);
|
||||
|
||||
SipHashTmpl(void *key) {
|
||||
uint64_t k0 = read64(key);
|
||||
uint64_t k1 = read64((char *)key + 8);
|
||||
|
||||
v0 = 0x736f6d6570736575 ^ k0;
|
||||
v1 = 0x646f72616e646f6d ^ k1;
|
||||
v2 = 0x6c7967656e657261 ^ k0;
|
||||
v3 = 0x7465646279746573 ^ k1;
|
||||
|
||||
if (OUTLEN == 128)
|
||||
v1 ^= 0xee;
|
||||
}
|
||||
|
||||
void update(void *msgp, int64_t msglen) {
|
||||
char *msg = (char *)msgp;
|
||||
sum += msglen;
|
||||
|
||||
if (buflen) {
|
||||
if (buflen + msglen < 8) {
|
||||
memcpy(buf + buflen, msg, msglen);
|
||||
buflen += msglen;
|
||||
return;
|
||||
}
|
||||
|
||||
int j = 8 - buflen;
|
||||
memcpy(buf + buflen, msg, j);
|
||||
compress(read64(buf));
|
||||
|
||||
msg += j;
|
||||
msglen -= j;
|
||||
buflen = 0;
|
||||
}
|
||||
|
||||
while (msglen >= 8) {
|
||||
compress(read64(msg));
|
||||
msg += 8;
|
||||
msglen -= 8;
|
||||
}
|
||||
|
||||
memcpy(buf, msg, msglen);
|
||||
buflen = msglen;
|
||||
}
|
||||
|
||||
void finish(void *out) {
|
||||
memset(buf + buflen, 0, 8 - buflen);
|
||||
compress(((uint64_t)sum << 56) | read64(buf));
|
||||
|
||||
v2 ^= (OUTLEN == 128) ? 0xee : 0xff;
|
||||
finalize();
|
||||
write64(out, v0 ^ v1 ^ v2 ^ v3);
|
||||
|
||||
if (OUTLEN == 128) {
|
||||
v1 ^= 0xdd;
|
||||
finalize();
|
||||
write64((char *)out + 8, v0 ^ v1 ^ v2 ^ v3);
|
||||
}
|
||||
}
|
||||
|
||||
static void hash(void *out, void *key, void *in, int inlen) {
|
||||
SipHashTmpl<C_ROUNDS, D_ROUNDS, OUTLEN> h(key);
|
||||
h.update(in, inlen);
|
||||
h.finish(out);
|
||||
}
|
||||
|
||||
private:
|
||||
uint64_t v0, v1, v2, v3;
|
||||
uint8_t buf[8];
|
||||
uint8_t buflen = 0;
|
||||
uint8_t sum = 0;
|
||||
|
||||
uint64_t read64(void *loc) {
|
||||
uint64_t val;
|
||||
memcpy(&val, loc, 8);
|
||||
if (std::endian::native == std::endian::big)
|
||||
val = bswap(val);
|
||||
return val;
|
||||
}
|
||||
|
||||
void write64(void *loc, uint64_t val) {
|
||||
if (std::endian::native == std::endian::big)
|
||||
val = bswap(val);
|
||||
memcpy(loc, &val, 8);
|
||||
}
|
||||
|
||||
uint64_t bswap(uint64_t val) {
|
||||
return ((val << 56) & 0xff00000000000000) |
|
||||
((val << 40) & 0x00ff000000000000) |
|
||||
((val << 24) & 0x0000ff0000000000) |
|
||||
((val << 8) & 0x000000ff00000000) |
|
||||
((val >> 8) & 0x00000000ff000000) |
|
||||
((val >> 24) & 0x0000000000ff0000) |
|
||||
((val >> 40) & 0x000000000000ff00) |
|
||||
((val >> 56) & 0x00000000000000ff);
|
||||
}
|
||||
|
||||
void round() {
|
||||
v0 += v1;
|
||||
v1 = std::rotl(v1, 13);
|
||||
v1 ^= v0;
|
||||
v0 = std::rotl(v0, 32);
|
||||
v2 += v3;
|
||||
v3 = std::rotl(v3, 16);
|
||||
v3 ^= v2;
|
||||
v0 += v3;
|
||||
v3 = std::rotl(v3, 21);
|
||||
v3 ^= v0;
|
||||
v2 += v1;
|
||||
v1 = std::rotl(v1, 17);
|
||||
v1 ^= v2;
|
||||
v2 = std::rotl(v2, 32);
|
||||
}
|
||||
|
||||
void compress(uint64_t m) {
|
||||
v3 ^= m;
|
||||
for (int i = 0; i < C_ROUNDS; i++)
|
||||
round();
|
||||
v0 ^= m;
|
||||
}
|
||||
|
||||
void finalize() {
|
||||
for (int i = 0; i < D_ROUNDS; i++)
|
||||
round();
|
||||
}
|
||||
};
|
||||
|
||||
using SipHash = SipHashTmpl<2, 4, 64>;
|
||||
using SipHash128 = SipHashTmpl<2, 4, 128>;
|
||||
using SipHash13 = SipHashTmpl<1, 3, 64>;
|
||||
using SipHash13_128 = SipHashTmpl<1, 3, 128>;
|
||||
-119
@@ -1,119 +0,0 @@
|
||||
// This file contains functions to create a tar file.
|
||||
|
||||
#include "lib.h"
|
||||
|
||||
#ifdef _WIN32
|
||||
# define ftruncate _chsize_s
|
||||
#endif
|
||||
|
||||
namespace mold {
|
||||
|
||||
static constexpr i64 BLOCK_SIZE = 512;
|
||||
|
||||
// A tar file consists of one or more Ustar header followed by data.
|
||||
// Each Ustar header represents a single file in an archive.
|
||||
//
|
||||
// tar is an old file format, and its `name` field is only 100 bytes long.
|
||||
// If `name` is longer than 100 bytes, we can emit a PAX header before a
|
||||
// Ustar header to store a long filename.
|
||||
//
|
||||
// For simplicity, we always emit a PAX header even for a short filename.
|
||||
struct UstarHeader {
|
||||
char name[100];
|
||||
char mode[8];
|
||||
char uid[8];
|
||||
char gid[8];
|
||||
char size[12];
|
||||
char mtime[12];
|
||||
char checksum[8];
|
||||
char typeflag[1];
|
||||
char linkname[100];
|
||||
char magic[6];
|
||||
char version[2];
|
||||
char uname[32];
|
||||
char gname[32];
|
||||
char devmajor[8];
|
||||
char devminor[8];
|
||||
char prefix[155];
|
||||
char pad[12];
|
||||
};
|
||||
|
||||
static_assert(sizeof(UstarHeader) == BLOCK_SIZE);
|
||||
|
||||
static void finalize(UstarHeader &hdr) {
|
||||
memset(hdr.checksum, ' ', sizeof(hdr.checksum));
|
||||
memcpy(hdr.magic, "ustar", 5);
|
||||
memcpy(hdr.version, "00", 2);
|
||||
|
||||
// Compute checksum
|
||||
int sum = 0;
|
||||
for (i64 i = 0; i < sizeof(hdr); i++)
|
||||
sum += ((u8 *)&hdr)[i];
|
||||
|
||||
// We need to convince the compiler that sum isn't too big to silence
|
||||
// -Werror=format-truncation.
|
||||
if (sum >= 01'000'000)
|
||||
unreachable();
|
||||
snprintf(hdr.checksum, sizeof(hdr.checksum), "%06o", sum);
|
||||
}
|
||||
|
||||
static std::string encode_path(std::string basedir, std::string path) {
|
||||
path = path_clean(basedir + "/" + path);
|
||||
|
||||
// Construct a string which contains something like
|
||||
// "16 path=foo/bar\n" where 16 is the size of the string
|
||||
// including the size string itself.
|
||||
i64 len = std::string(" path=\n").size() + path.size();
|
||||
i64 total = std::to_string(len).size() + len;
|
||||
total = std::to_string(total).size() + len;
|
||||
return std::to_string(total) + " path=" + path + "\n";
|
||||
}
|
||||
|
||||
std::unique_ptr<TarWriter>
|
||||
TarWriter::open(std::string output_path, std::string basedir) {
|
||||
FILE *out = fopen(output_path.c_str(), "w");
|
||||
if (!out)
|
||||
return nullptr;
|
||||
return std::unique_ptr<TarWriter>(new TarWriter(out, basedir));
|
||||
}
|
||||
|
||||
TarWriter::~TarWriter() {
|
||||
fclose(out);
|
||||
}
|
||||
|
||||
void TarWriter::append(std::string path, std::string_view data) {
|
||||
// Write PAX header
|
||||
UstarHeader pax = {};
|
||||
|
||||
std::string attr = encode_path(basedir, path);
|
||||
|
||||
if (attr.size() >= 0100'000'000'000)
|
||||
unreachable();
|
||||
snprintf(pax.size, sizeof(pax.size), "%011zo", attr.size());
|
||||
pax.name[0] = '/';
|
||||
pax.typeflag[0] = 'x';
|
||||
finalize(pax);
|
||||
fwrite(&pax, sizeof(pax), 1, out);
|
||||
|
||||
// Write pathname
|
||||
fwrite(attr.data(), attr.size(), 1, out);
|
||||
fseek(out, align_to(ftell(out), BLOCK_SIZE), SEEK_SET);
|
||||
|
||||
// Write Ustar header
|
||||
UstarHeader ustar = {};
|
||||
memcpy(ustar.mode, "0000664", 8);
|
||||
if (data.size() >= 0100'000'000'000)
|
||||
unreachable();
|
||||
snprintf(ustar.size, sizeof(ustar.size), "%011zo", data.size());
|
||||
finalize(ustar);
|
||||
fwrite(&ustar, sizeof(ustar), 1, out);
|
||||
|
||||
// Write file contents
|
||||
fwrite(data.data(), data.size(), 1, out);
|
||||
fseek(out, align_to(ftell(out), BLOCK_SIZE), SEEK_SET);
|
||||
|
||||
// A tar file must ends with two empty blocks
|
||||
(void)!ftruncate(fileno(out), ftell(out) + BLOCK_SIZE * 2);
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,33 +0,0 @@
|
||||
# Get a git hash value. We do not want to use git command here
|
||||
# because we don't want to make git a build-time dependency.
|
||||
if(EXISTS "${SOURCE_DIR}/.git/HEAD")
|
||||
file(READ "${SOURCE_DIR}/.git/HEAD" HASH)
|
||||
string(STRIP "${HASH}" HASH)
|
||||
|
||||
if(HASH MATCHES "^ref: (.*)")
|
||||
set(HEAD "${CMAKE_MATCH_1}")
|
||||
if(EXISTS "${SOURCE_DIR}/.git/${HEAD}")
|
||||
file(READ "${SOURCE_DIR}/.git/${HEAD}" HASH)
|
||||
string(STRIP "${HASH}" HASH)
|
||||
else()
|
||||
file(READ "${SOURCE_DIR}/.git/packed-refs" PACKED_REFS)
|
||||
string(REGEX REPLACE ".*\n([0-9a-f]+) ${HEAD}\n.*" "\\1" HASH "\n${PACKED_REFS}")
|
||||
endif()
|
||||
endif()
|
||||
endif()
|
||||
|
||||
# Create new file contents and update a given file if necessary.
|
||||
if("${HASH}" STREQUAL "")
|
||||
set(NEW_CONTENTS "")
|
||||
else()
|
||||
set(NEW_CONTENTS "#define MOLD_GIT_HASH \"${HASH}\"\n")
|
||||
endif()
|
||||
|
||||
if(EXISTS "${OUTPUT_FILE}")
|
||||
file(READ "${OUTPUT_FILE}" OLD_CONTENTS)
|
||||
if(NOT "${NEW_CONTENTS}" STREQUAL "${OLD_CONTENTS}")
|
||||
file(WRITE "${OUTPUT_FILE}" "${NEW_CONTENTS}")
|
||||
endif()
|
||||
else()
|
||||
file(WRITE "${OUTPUT_FILE}" "${NEW_CONTENTS}")
|
||||
endif()
|
||||
@@ -1,916 +0,0 @@
|
||||
// ARM32 is a bit special from the linker's viewpoint because ARM
|
||||
// processors support two different instruction encodings: Thumb and
|
||||
// ARM (in a narrower sense). Thumb instructions are either 16 bits or
|
||||
// 32 bits, while ARM instructions are all 32 bits. Feature-wise,
|
||||
// Thumb is a subset of ARM, so not all ARM instructions are
|
||||
// representable in Thumb.
|
||||
//
|
||||
// ARM processors originally supported only ARM instructions. Thumb
|
||||
// instructions were later added to increase code density.
|
||||
//
|
||||
// ARM processors runs in either ARM mode or Thumb mode. The mode can
|
||||
// be switched using BX (branch and mode exchange)-family instructions.
|
||||
// We need to use that instructions to, for example, call a function
|
||||
// encoded in Thumb from a function encoded in ARM. Sometimes, the
|
||||
// linker even has to emit interworking thunk code to switch mode.
|
||||
//
|
||||
// ARM instructions are aligned to 4 byte boundaries. Thumb are to 2
|
||||
// byte boundaries. So the least significant bit of a function address
|
||||
// is always 0.
|
||||
//
|
||||
// To distinguish Thumb functions from ARM fucntions, the LSB of a
|
||||
// function address is repurposed as a boolean flag. If the LSB is 0,
|
||||
// the function referred to by the address is encoded in ARM;
|
||||
// otherwise, Thumb.
|
||||
//
|
||||
// For example, if a symbol `foo` is of type STT_FUNC and has value
|
||||
// 0x2001, `foo` is a function using Thumb instructions whose address
|
||||
// is 0x2000 (not 0x2001, as Thumb instructions are always 2-byte
|
||||
// aligned). Likewise, if a function pointer has value 0x2001, it
|
||||
// refers a Thumb function at 0x2000.
|
||||
//
|
||||
// https://github.com/ARM-software/abi-aa/blob/main/aaelf32/aaelf32.rst
|
||||
|
||||
#if MOLD_ARM32LE || MOLD_ARM32BE
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <tbb/parallel_for.h>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template <>
|
||||
i64 get_addend(u8 *loc, const ElfRel<E> &rel) {
|
||||
U32<E> *arm = (U32<E> *)loc;
|
||||
U16<E> *thm = (U16<E> *)loc;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_ARM_ABS32:
|
||||
case R_ARM_REL32:
|
||||
case R_ARM_BASE_PREL:
|
||||
case R_ARM_GOTOFF32:
|
||||
case R_ARM_GOT_PREL:
|
||||
case R_ARM_GOT_BREL:
|
||||
case R_ARM_TLS_GD32:
|
||||
case R_ARM_TLS_LDM32:
|
||||
case R_ARM_TLS_LDO32:
|
||||
case R_ARM_TLS_IE32:
|
||||
case R_ARM_TLS_LE32:
|
||||
case R_ARM_TLS_GOTDESC:
|
||||
case R_ARM_TARGET1:
|
||||
case R_ARM_TARGET2:
|
||||
return (I32<E>)*arm;
|
||||
case R_ARM_THM_JUMP8:
|
||||
return sign_extend(thm[0], 8) << 1;
|
||||
case R_ARM_THM_JUMP11:
|
||||
return sign_extend(thm[0], 11) << 1;
|
||||
case R_ARM_THM_JUMP19: {
|
||||
// https://developer.arm.com/documentation/ddi0597/2024-12/Base-Instructions/B--Branch-
|
||||
u32 S = bit(thm[0], 10);
|
||||
u32 J2 = bit(thm[1], 11);
|
||||
u32 J1 = bit(thm[1], 13);
|
||||
u32 imm6 = bits(thm[0], 5, 0);
|
||||
u32 imm11 = bits(thm[1], 10, 0);
|
||||
u32 val = (S << 20) | (J2 << 19) | (J1 << 18) | (imm6 << 12) | (imm11 << 1);
|
||||
return sign_extend(val, 21);
|
||||
}
|
||||
case R_ARM_THM_CALL:
|
||||
case R_ARM_THM_JUMP24:
|
||||
case R_ARM_THM_TLS_CALL: {
|
||||
// https://developer.arm.com/documentation/ddi0597/2024-12/Base-Instructions/BL--BLX--immediate---Branch-with-Link-and-optional-Exchange--immediate--
|
||||
u32 S = bit(thm[0], 10);
|
||||
u32 J1 = bit(thm[1], 13);
|
||||
u32 J2 = bit(thm[1], 11);
|
||||
u32 I1 = !(J1 ^ S);
|
||||
u32 I2 = !(J2 ^ S);
|
||||
u32 imm10 = bits(thm[0], 9, 0);
|
||||
u32 imm11 = bits(thm[1], 10, 0);
|
||||
u32 val = (S << 24) | (I1 << 23) | (I2 << 22) | (imm10 << 12) | (imm11 << 1);
|
||||
return sign_extend(val, 25);
|
||||
}
|
||||
case R_ARM_CALL:
|
||||
case R_ARM_JUMP24:
|
||||
case R_ARM_PLT32:
|
||||
case R_ARM_TLS_CALL:
|
||||
return sign_extend(*arm, 24) << 2;
|
||||
case R_ARM_MOVW_PREL_NC:
|
||||
case R_ARM_MOVW_ABS_NC:
|
||||
case R_ARM_MOVT_PREL:
|
||||
case R_ARM_MOVT_ABS: {
|
||||
// https://developer.arm.com/documentation/ddi0597/2024-12/Base-Instructions/MOV--MOVS--immediate---Move--immediate--
|
||||
u32 imm4 = bits(*arm, 19, 16);
|
||||
u32 imm12 = bits(*arm, 11, 0);
|
||||
u32 val = (imm4 << 12) | imm12;
|
||||
return sign_extend(val, 16);
|
||||
}
|
||||
case R_ARM_PREL31:
|
||||
return sign_extend(*arm, 31);
|
||||
case R_ARM_THM_MOVW_PREL_NC:
|
||||
case R_ARM_THM_MOVW_ABS_NC:
|
||||
case R_ARM_THM_MOVT_PREL:
|
||||
case R_ARM_THM_MOVT_ABS: {
|
||||
// https://developer.arm.com/documentation/ddi0597/2024-12/Base-Instructions/MOVT--Move-Top-
|
||||
u32 imm4 = bits(thm[0], 3, 0);
|
||||
u32 i = bit(thm[0], 10);
|
||||
u32 imm3 = bits(thm[1], 14, 12);
|
||||
u32 imm8 = bits(thm[1], 7, 0);
|
||||
u32 val = (imm4 << 12) | (i << 11) | (imm3 << 8) | imm8;
|
||||
return sign_extend(val, 16);
|
||||
}
|
||||
default:
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
static void write_arm_mov(u8 *loc, u32 val) {
|
||||
u32 imm12 = bits(val, 11, 0);
|
||||
u32 imm4 = bits(val, 15, 12);
|
||||
*(U32<E> *)loc = (*(U32<E> *)loc & 0xfff0'f000) | (imm4 << 16) | imm12;
|
||||
}
|
||||
|
||||
static void write_thm_b21(u8 *loc, u32 val) {
|
||||
u32 S = bit(val, 20);
|
||||
u32 J2 = bit(val, 19);
|
||||
u32 J1 = bit(val, 18);
|
||||
u32 imm6 = bits(val, 17, 12);
|
||||
u32 imm11 = bits(val, 11, 1);
|
||||
|
||||
U16<E> *buf = (U16<E> *)loc;
|
||||
buf[0] = (buf[0] & 0b1111'1011'1100'0000) | (S << 10) | imm6;
|
||||
buf[1] = (buf[1] & 0b1101'0000'0000'0000) | (J1 << 13) | (J2 << 11) | imm11;
|
||||
}
|
||||
|
||||
static void write_thm_b25(u8 *loc, u32 val) {
|
||||
u32 S = bit(val, 24);
|
||||
u32 I1 = bit(val, 23);
|
||||
u32 I2 = bit(val, 22);
|
||||
u32 J1 = !I1 ^ S;
|
||||
u32 J2 = !I2 ^ S;
|
||||
u32 imm10 = bits(val, 21, 12);
|
||||
u32 imm11 = bits(val, 11, 1);
|
||||
|
||||
U16<E> *buf = (U16<E> *)loc;
|
||||
buf[0] = (buf[0] & 0b1111'1000'0000'0000) | (S << 10) | imm10;
|
||||
buf[1] = (buf[1] & 0b1101'0000'0000'0000) | (J1 << 13) | (J2 << 11) | imm11;
|
||||
}
|
||||
|
||||
static void write_thm_mov(u8 *loc, u32 val) {
|
||||
u32 imm4 = bits(val, 15, 12);
|
||||
u32 i = bit(val, 11);
|
||||
u32 imm3 = bits(val, 10, 8);
|
||||
u32 imm8 = bits(val, 7, 0);
|
||||
|
||||
U16<E> *buf = (U16<E> *)loc;
|
||||
buf[0] = (buf[0] & 0b1111'1011'1111'0000) | (i << 10) | imm4;
|
||||
buf[1] = (buf[1] & 0b1000'1111'0000'0000) | (imm3 << 12) | imm8;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_addend(u8 *loc, i64 val, const ElfRel<E> &rel) {
|
||||
switch (rel.r_type) {
|
||||
case R_ARM_NONE:
|
||||
break;
|
||||
case R_ARM_ABS32:
|
||||
case R_ARM_REL32:
|
||||
case R_ARM_BASE_PREL:
|
||||
case R_ARM_GOTOFF32:
|
||||
case R_ARM_GOT_PREL:
|
||||
case R_ARM_GOT_BREL:
|
||||
case R_ARM_TLS_GD32:
|
||||
case R_ARM_TLS_LDM32:
|
||||
case R_ARM_TLS_LDO32:
|
||||
case R_ARM_TLS_IE32:
|
||||
case R_ARM_TLS_LE32:
|
||||
case R_ARM_TLS_GOTDESC:
|
||||
case R_ARM_TARGET1:
|
||||
case R_ARM_TARGET2:
|
||||
*(U32<E> *)loc = val;
|
||||
break;
|
||||
case R_ARM_THM_JUMP8:
|
||||
*(U16<E> *)loc = (*(U16<E> *)loc & 0xff00) | bits(val, 8, 1);
|
||||
break;
|
||||
case R_ARM_THM_JUMP11:
|
||||
*(U16<E> *)loc = (*(U16<E> *)loc & 0xf800) | bits(val, 11, 1);
|
||||
break;
|
||||
case R_ARM_THM_CALL:
|
||||
case R_ARM_THM_JUMP24:
|
||||
case R_ARM_THM_TLS_CALL:
|
||||
write_thm_b25(loc, val);
|
||||
break;
|
||||
case R_ARM_CALL:
|
||||
case R_ARM_JUMP24:
|
||||
case R_ARM_PLT32:
|
||||
*(U32<E> *)loc = (*(U32<E> *)loc & 0xff00'0000) | bits(val, 25, 2);
|
||||
break;
|
||||
case R_ARM_MOVW_PREL_NC:
|
||||
case R_ARM_MOVW_ABS_NC:
|
||||
case R_ARM_MOVT_PREL:
|
||||
case R_ARM_MOVT_ABS:
|
||||
write_arm_mov(loc, val);
|
||||
break;
|
||||
case R_ARM_PREL31:
|
||||
*(U32<E> *)loc = (*(U32<E> *)loc & 0x8000'0000) | (val & 0x7fff'ffff);
|
||||
break;
|
||||
case R_ARM_THM_MOVW_PREL_NC:
|
||||
case R_ARM_THM_MOVW_ABS_NC:
|
||||
case R_ARM_THM_MOVT_PREL:
|
||||
case R_ARM_THM_MOVT_ABS:
|
||||
write_thm_mov(loc, val);
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
constexpr ul32 insn[] = {
|
||||
0xe52d'e004, // push {lr}
|
||||
0xe59f'e004, // ldr lr, 2f
|
||||
0xe08f'e00e, // 1: add lr, pc, lr
|
||||
0xe5be'f008, // ldr pc, [lr, #8]!
|
||||
0x0000'0000, // 2: .word .got.plt - 1b - 8
|
||||
0x0000'0000, // (padding)
|
||||
0x0000'0000, // (padding)
|
||||
0x0000'0000, // (padding)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 16) = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 16;
|
||||
}
|
||||
|
||||
static constexpr ul32 plt_entry[] = {
|
||||
0xe59f'c004, // 1: ldr ip, 2f
|
||||
0xe08c'c00f, // add ip, ip, pc
|
||||
0xe59c'f000, // ldr pc, [ip]
|
||||
0x0000'0000, // 2: .word sym@GOT - 1b
|
||||
};
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
memcpy(buf, plt_entry, sizeof(plt_entry));
|
||||
*(U32<E> *)(buf + 12) = sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx) - 12;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
memcpy(buf, plt_entry, sizeof(plt_entry));
|
||||
*(U32<E> *)(buf + 12) = sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx) - 12;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_ARM_ABS32:
|
||||
*(U32<E> *)loc = val;
|
||||
break;
|
||||
case R_ARM_REL32:
|
||||
*(U32<E> *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || rel.r_type == R_ARM_V4BX)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = get_addend(*this, rel);
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 T = S & 1;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto get_thumb_thunk_addr = [&] { return sym.get_thunk_addr(ctx, P); };
|
||||
auto get_arm_thunk_addr = [&] { return sym.get_thunk_addr(ctx, P) + 4; };
|
||||
|
||||
auto get_tlsdesc_trampoline_addr = [&] {
|
||||
auto it = ranges::upper_bound(output_section->thunks, P, {},
|
||||
[](std::unique_ptr<Thunk<E>> &thunk) {
|
||||
return thunk->get_addr();
|
||||
});
|
||||
return (*it)->get_addr();
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_ARM_ABS32:
|
||||
case R_ARM_TARGET1:
|
||||
break;
|
||||
case R_ARM_REL32:
|
||||
*(U32<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_ARM_THM_CALL: {
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
// On ARM, calling an weak undefined symbol jumps to the
|
||||
// next instruction.
|
||||
*(U32<E> *)loc = 0x8000'f3af; // NOP.W
|
||||
break;
|
||||
}
|
||||
|
||||
// THM_CALL relocation refers to either BL or BLX instruction.
|
||||
// They are different in only one bit. We need to use BL if
|
||||
// the jump target is Thumb. Otherwise, use BLX.
|
||||
i64 val1 = S + A - P;
|
||||
i64 val2 = align_to(S + A - P, 4);
|
||||
|
||||
if (T && is_int(val1, 25)) {
|
||||
*(U16<E> *)(loc + 2) |= 0x1000; // BL
|
||||
write_thm_b25(loc, val1);
|
||||
} else if (!T && is_int(val2, 25)) {
|
||||
*(U16<E> *)(loc + 2) &= ~0x1000; // BLX
|
||||
write_thm_b25(loc, val2);
|
||||
} else {
|
||||
*(U16<E> *)(loc + 2) |= 0x1000; // BL
|
||||
write_thm_b25(loc, get_thumb_thunk_addr() + A - P);
|
||||
}
|
||||
break;
|
||||
}
|
||||
case R_ARM_BASE_PREL:
|
||||
*(U32<E> *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_ARM_GOTOFF32:
|
||||
*(U32<E> *)loc = ((S + A) | T) - GOT;
|
||||
break;
|
||||
case R_ARM_GOT_PREL:
|
||||
case R_ARM_TARGET2:
|
||||
*(U32<E> *)loc = GOT + G + A - P;
|
||||
break;
|
||||
case R_ARM_GOT_BREL:
|
||||
*(U32<E> *)loc = G + A;
|
||||
break;
|
||||
case R_ARM_CALL: {
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
*(U32<E> *)loc = 0xe320'f000; // NOP
|
||||
break;
|
||||
}
|
||||
|
||||
// Just like THM_CALL, ARM_CALL relocation refers to either BL or
|
||||
// BLX instruction. We may need to rewrite BL → BLX or BLX → BL.
|
||||
bool is_bl = ((*(U32<E> *)loc & 0xff00'0000) == 0xeb00'0000);
|
||||
bool is_blx = ((*(U32<E> *)loc & 0xfe00'0000) == 0xfa00'0000);
|
||||
if (!is_bl && !is_blx)
|
||||
Fatal(ctx) << *this << ": R_ARM_CALL refers to neither BL nor BLX";
|
||||
|
||||
i64 val = S + A - P;
|
||||
if (is_int(val, 26)) {
|
||||
if (T) {
|
||||
*(U32<E> *)loc = 0xfa00'0000; // BLX
|
||||
*(U32<E> *)loc |= (bit(val, 1) << 24) | bits(val, 25, 2);
|
||||
} else {
|
||||
*(U32<E> *)loc = 0xeb00'0000; // BL
|
||||
*(U32<E> *)loc |= bits(val, 25, 2);
|
||||
}
|
||||
} else {
|
||||
*(U32<E> *)loc = 0xeb00'0000; // BL
|
||||
*(U32<E> *)loc |= bits(get_arm_thunk_addr() + A - P, 25, 2);
|
||||
}
|
||||
break;
|
||||
}
|
||||
case R_ARM_JUMP24: {
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
*(U32<E> *)loc = 0xe320'f000; // NOP
|
||||
break;
|
||||
}
|
||||
|
||||
// These relocs refers to a B (unconditional branch) instruction.
|
||||
// Unlike BL or BLX, we can't rewrite B to BX in place when the
|
||||
// processor mode switch is required because BX doesn't takes an
|
||||
// immediate; it takes only a register. So if mode switch is
|
||||
// required, we jump to a linker-synthesized thunk which does the
|
||||
// job with a longer code sequence.
|
||||
i64 val = S + A - P;
|
||||
if (T || !is_int(val, 26))
|
||||
val = get_arm_thunk_addr() + A - P;
|
||||
*(U32<E> *)loc = (*(U32<E> *)loc & 0xff00'0000) | bits(val, 25, 2);
|
||||
break;
|
||||
}
|
||||
case R_ARM_PLT32:
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
*(U32<E> *)loc = 0xe320'f000; // NOP
|
||||
} else {
|
||||
u64 val = (T ? get_arm_thunk_addr() : S) + A - P;
|
||||
*(U32<E> *)loc = (*(U32<E> *)loc & 0xff00'0000) | bits(val, 25, 2);
|
||||
}
|
||||
break;
|
||||
case R_ARM_THM_JUMP8:
|
||||
check(S + A - P, -(1 << 8), 1 << 8);
|
||||
*(U16<E> *)loc &= 0xff00;
|
||||
*(U16<E> *)loc |= bits(S + A - P, 8, 1);
|
||||
break;
|
||||
case R_ARM_THM_JUMP11:
|
||||
check(S + A - P, -(1 << 11), 1 << 11);
|
||||
*(U16<E> *)loc &= 0xf800;
|
||||
*(U16<E> *)loc |= bits(S + A - P, 11, 1);
|
||||
break;
|
||||
case R_ARM_THM_JUMP19:
|
||||
check(S + A - P, -(1 << 20), 1 << 20);
|
||||
write_thm_b21(loc, S + A - P);
|
||||
break;
|
||||
case R_ARM_THM_JUMP24: {
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
*(U32<E> *)loc = 0x8000'f3af; // NOP
|
||||
break;
|
||||
}
|
||||
|
||||
// Just like R_ARM_JUMP24, we need to jump to a thunk if we need to
|
||||
// switch processor mode.
|
||||
i64 val = S + A - P;
|
||||
if (!T || !is_int(val, 25))
|
||||
val = get_thumb_thunk_addr() + A - P;
|
||||
write_thm_b25(loc, val);
|
||||
break;
|
||||
}
|
||||
case R_ARM_MOVW_PREL_NC:
|
||||
write_arm_mov(loc, ((S + A) | T) - P);
|
||||
break;
|
||||
case R_ARM_MOVW_ABS_NC:
|
||||
write_arm_mov(loc, (S + A) | T);
|
||||
break;
|
||||
case R_ARM_THM_MOVW_PREL_NC:
|
||||
write_thm_mov(loc, ((S + A) | T) - P);
|
||||
break;
|
||||
case R_ARM_PREL31:
|
||||
check(S + A - P, -(1LL << 30), 1LL << 30);
|
||||
*(U32<E> *)loc &= 0x8000'0000;
|
||||
*(U32<E> *)loc |= (S + A - P) & 0x7fff'ffff;
|
||||
break;
|
||||
case R_ARM_THM_MOVW_ABS_NC:
|
||||
write_thm_mov(loc, (S + A) | T);
|
||||
break;
|
||||
case R_ARM_MOVT_PREL:
|
||||
write_arm_mov(loc, (S + A - P) >> 16);
|
||||
break;
|
||||
case R_ARM_THM_MOVT_PREL:
|
||||
write_thm_mov(loc, (S + A - P) >> 16);
|
||||
break;
|
||||
case R_ARM_MOVT_ABS:
|
||||
write_arm_mov(loc, (S + A) >> 16);
|
||||
break;
|
||||
case R_ARM_THM_MOVT_ABS:
|
||||
write_thm_mov(loc, (S + A) >> 16);
|
||||
break;
|
||||
case R_ARM_TLS_GD32:
|
||||
*(U32<E> *)loc = sym.get_tlsgd_addr(ctx) + A - P;
|
||||
break;
|
||||
case R_ARM_TLS_LDM32:
|
||||
*(U32<E> *)loc = ctx.got->get_tlsld_addr(ctx) + A - P;
|
||||
break;
|
||||
case R_ARM_TLS_LDO32:
|
||||
*(U32<E> *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_ARM_TLS_IE32:
|
||||
*(U32<E> *)loc = sym.get_gottp_addr(ctx) + A - P;
|
||||
break;
|
||||
case R_ARM_TLS_LE32:
|
||||
*(U32<E> *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_ARM_TLS_GOTDESC:
|
||||
// ARM32 TLSDESC uses the following code sequence to materialize
|
||||
// a TP-relative address in r0.
|
||||
//
|
||||
// ldr r0, .L2
|
||||
// .L1: bl foo
|
||||
// R_ARM_TLS_CALL
|
||||
// .L2: .word foo + . - .L1
|
||||
// R_ARM_TLS_GOTDESC
|
||||
//
|
||||
// We may relax the instructions to the following if its TP-relative
|
||||
// address is known at link-time
|
||||
//
|
||||
// ldr r0, .L2
|
||||
// .L1: nop
|
||||
// ...
|
||||
// .L2: .word foo(tpoff)
|
||||
//
|
||||
// or to the following if the TP-relative address is known at
|
||||
// process startup time.
|
||||
//
|
||||
// ldr r0, .L2
|
||||
// .L1: ldr r0, [pc, r0]
|
||||
// ...
|
||||
// .L2: .word foo(gottpoff) + . - .L1
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
// A is odd if the corresponding TLS_CALL is Thumb.
|
||||
*(U32<E> *)loc = sym.get_tlsdesc_addr(ctx) - P + A - ((A & 1) ? 6 : 4);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
*(U32<E> *)loc = sym.get_gottp_addr(ctx) - P + A - ((A & 1) ? 5 : 8);
|
||||
} else {
|
||||
*(U32<E> *)loc = S - ctx.tp_addr;
|
||||
}
|
||||
break;
|
||||
case R_ARM_TLS_CALL:
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
*(U32<E> *)loc = 0xeb00'0000; // bl 0
|
||||
*(U32<E> *)loc |= bits(get_tlsdesc_trampoline_addr() - P - 8, 25, 2);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
*(U32<E> *)loc = 0xe79f'0000; // ldr r0, [pc, r0]
|
||||
} else {
|
||||
*(U32<E> *)loc = 0xe320'f000; // nop
|
||||
}
|
||||
break;
|
||||
case R_ARM_THM_TLS_CALL:
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
u64 val = align_to(get_tlsdesc_trampoline_addr() - P - 4, 4);
|
||||
write_thm_b25(loc, val);
|
||||
*(U16<E> *)(loc + 2) &= ~0x1000; // rewrite BL with BLX
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
// Since `ldr r0, [pc, r0]` is not representable in Thumb,
|
||||
// we use two instructions instead.
|
||||
*(U16<E> *)loc = 0x4478; // add r0, pc
|
||||
*(U16<E> *)(loc + 2) = 0x6800; // ldr r0, [r0]
|
||||
} else {
|
||||
*(U32<E> *)loc = 0x8000'f3af; // nop.w
|
||||
}
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : get_addend(*this, rel);
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_ARM_ABS32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(U32<E> *)loc = *val;
|
||||
else
|
||||
*(U32<E> *)loc = S + A;
|
||||
break;
|
||||
case R_ARM_TLS_LDO32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(U32<E> *)loc = *val;
|
||||
else
|
||||
*(U32<E> *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_ARM_MOVW_ABS_NC:
|
||||
case R_ARM_THM_MOVW_ABS_NC:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_ARM_THM_CALL:
|
||||
case R_ARM_CALL:
|
||||
case R_ARM_JUMP24:
|
||||
case R_ARM_PLT32:
|
||||
case R_ARM_THM_JUMP24:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_ARM_GOT_PREL:
|
||||
case R_ARM_GOT_BREL:
|
||||
case R_ARM_TARGET2:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_ARM_MOVT_PREL:
|
||||
case R_ARM_THM_MOVT_PREL:
|
||||
case R_ARM_PREL31:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_ARM_TLS_GD32:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_ARM_TLS_LDM32:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_ARM_TLS_IE32:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_ARM_TLS_CALL:
|
||||
case R_ARM_THM_TLS_CALL:
|
||||
scan_tlsdesc(ctx, sym);
|
||||
break;
|
||||
case R_ARM_TLS_LE32:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_ARM_ABS32:
|
||||
case R_ARM_TARGET1:
|
||||
case R_ARM_MOVT_ABS:
|
||||
case R_ARM_THM_MOVT_ABS:
|
||||
case R_ARM_REL32:
|
||||
case R_ARM_BASE_PREL:
|
||||
case R_ARM_GOTOFF32:
|
||||
case R_ARM_THM_JUMP8:
|
||||
case R_ARM_THM_JUMP11:
|
||||
case R_ARM_THM_JUMP19:
|
||||
case R_ARM_MOVW_PREL_NC:
|
||||
case R_ARM_THM_MOVW_PREL_NC:
|
||||
case R_ARM_TLS_LDO32:
|
||||
case R_ARM_V4BX:
|
||||
case R_ARM_TLS_GOTDESC:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Thunk<E>::copy_buf(Context<E> &ctx) {
|
||||
// TLS trampoline code. ARM32's TLSDESC is designed so that this
|
||||
// common piece of code is factored out from object files to reduce
|
||||
// output size. Since no one provide, the linker has to synthesize it.
|
||||
constexpr ul32 hdr[] = {
|
||||
0xe08e'0000, // add r0, lr, r0
|
||||
0xe590'1004, // ldr r1, [r0, #4]
|
||||
0xe12f'ff11, // bx r1
|
||||
0xe320'f000, // nop
|
||||
};
|
||||
|
||||
// This is a range extension and mode switch thunk.
|
||||
// It has two entry points: +0 for Thumb and +4 for ARM.
|
||||
constexpr u8 entry[] = {
|
||||
// .thumb
|
||||
0x78, 0x47, // bx pc # jumps to 1f
|
||||
0xc0, 0x46, // nop
|
||||
// .arm
|
||||
0x00, 0xc0, 0x9f, 0xe5, // 1: ldr ip, 3f
|
||||
0x0f, 0xf0, 0x8c, 0xe0, // 2: add pc, ip, pc
|
||||
0x00, 0x00, 0x00, 0x00, // 3: .word sym - 2b
|
||||
};
|
||||
|
||||
static_assert(E::thunk_hdr_size == sizeof(hdr));
|
||||
static_assert(E::thunk_size == sizeof(entry));
|
||||
|
||||
u8 *base = ctx.buf + output_section.shdr.sh_offset + offset;
|
||||
memcpy(base, hdr, sizeof(hdr));
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
u64 S = symbols[i]->get_addr(ctx);
|
||||
u64 P = get_addr() + offsets[i];
|
||||
u8 *buf = base + offsets[i];
|
||||
memcpy(buf, entry, sizeof(entry));
|
||||
*(U32<E> *)(buf + 12) = S - P - 16;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
u64 get_eflags(Context<E> &ctx) {
|
||||
if constexpr (E::is_le)
|
||||
return EF_ARM_EABI_VER5;
|
||||
else
|
||||
return EF_ARM_EABI_VER5 | EF_ARM_BE8;
|
||||
}
|
||||
|
||||
template <>
|
||||
void create_arm_exidx_section<E>(Context<E> &ctx) {
|
||||
for (i64 i = 0; i < ctx.chunks.size(); i++) {
|
||||
OutputSection<E> *osec = ctx.chunks[i]->to_osec();
|
||||
|
||||
if (osec && osec->shdr.sh_type == SHT_ARM_EXIDX) {
|
||||
auto *sec = new Arm32ExidxSection(*osec);
|
||||
ctx.extra.exidx = sec;
|
||||
ctx.chunks[i] = sec;
|
||||
ctx.chunk_pool.emplace_back(sec);
|
||||
|
||||
for (InputSection<E> *isec : osec->members)
|
||||
isec->is_alive = false;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Arm32ExidxSection<E>::compute_section_size(Context<E> &ctx) {
|
||||
output_section.compute_section_size(ctx);
|
||||
this->shdr.sh_size = output_section.shdr.sh_size + 8; // +8 for sentinel
|
||||
}
|
||||
|
||||
template <>
|
||||
void Arm32ExidxSection<E>::update_shdr(Context<E> &ctx) {
|
||||
// .ARM.exidx's sh_link should be set to the .text section index.
|
||||
// Runtime doesn't care about it, but the binutils's strip command does.
|
||||
if (Chunk<E> *chunk = find_chunk(ctx, ".text"))
|
||||
this->shdr.sh_link = chunk->shndx;
|
||||
}
|
||||
|
||||
// Returns the end of the text segment
|
||||
static u64 get_text_end(Context<E> &ctx) {
|
||||
u64 ret = 0;
|
||||
for (Chunk<E> *chunk : ctx.chunks)
|
||||
if (chunk->shdr.sh_flags & SHF_EXECINSTR)
|
||||
ret = std::max<u64>(ret, chunk->shdr.sh_addr + chunk->shdr.sh_size);
|
||||
return ret;
|
||||
}
|
||||
|
||||
// ARM executables use an .ARM.exidx section to look up an exception
|
||||
// handling record for the current instruction pointer. The table needs
|
||||
// to be sorted by their addresses.
|
||||
//
|
||||
// Other target uses .eh_frame_hdr instead for the same purpose.
|
||||
// I don't know why only ARM uses the different mechanism, but it's
|
||||
// likely that it's due to some historical reason.
|
||||
//
|
||||
// This function returns contents of .ARM.exidx.
|
||||
template <>
|
||||
std::vector<u8> Arm32ExidxSection<E>::get_contents(Context<E> &ctx) {
|
||||
// .ARM.exidx records consists of a signed 31-bit relative address
|
||||
// and a 32-bit value. The relative address indicates the start
|
||||
// address of a function that the record covers. The value is one of
|
||||
// the followings:
|
||||
//
|
||||
// 1. CANTUNWIND indicating that there's no unwinding info for the function,
|
||||
// 2. a compact unwinding record encoded into a 32-bit value, or
|
||||
// 3. a 31-bit relative address which points to a larger record in
|
||||
// the .ARM.extab section.
|
||||
//
|
||||
// CANTUNWIND is value 1. The most significant bit is set in (2) but
|
||||
// not in (3). So we can distinguished them just by looking at a value.
|
||||
const u32 CANTUNWIND = 1;
|
||||
|
||||
struct Entry {
|
||||
U32<E> addr;
|
||||
U32<E> val;
|
||||
};
|
||||
|
||||
// We reserve one extra slot for the sentinel
|
||||
i64 num_entries = output_section.shdr.sh_size / sizeof(Entry) + 1;
|
||||
std::vector<u8> buf(num_entries * sizeof(Entry));
|
||||
Entry *ent = (Entry *)buf.data();
|
||||
|
||||
// Write section contents to the buffer
|
||||
output_section.shdr.sh_addr = this->shdr.sh_addr;
|
||||
output_section.write_to(ctx, buf.data());
|
||||
|
||||
// Fill in sentinel fields
|
||||
u64 sentinel_addr = this->shdr.sh_addr + sizeof(Entry) * (num_entries - 1);
|
||||
ent[num_entries - 1].addr = get_text_end(ctx) - sentinel_addr;
|
||||
ent[num_entries - 1].val = CANTUNWIND;
|
||||
|
||||
// Entry's addresses are relative to themselves. In order to sort
|
||||
// records by address, we first translate them so that the addresses
|
||||
// are relative to the beginning of the section.
|
||||
auto is_relative = [](u32 val) {
|
||||
return val != CANTUNWIND && !(val & 0x8000'0000);
|
||||
};
|
||||
|
||||
tbb::parallel_for((i64)0, num_entries, [&](i64 i) {
|
||||
i64 offset = sizeof(Entry) * i;
|
||||
ent[i].addr = sign_extend(ent[i].addr, 31) + offset;
|
||||
if (is_relative(ent[i].val))
|
||||
ent[i].val = 0x7fff'ffff & (ent[i].val + offset);
|
||||
});
|
||||
|
||||
ranges::sort(ent, ent + num_entries, {}, &Entry::addr);
|
||||
|
||||
// Remove duplicate adjacent entries. That is, if two adjacent functions
|
||||
// have the same compact unwind info or are both CANTUNWIND, we can
|
||||
// merge them into a single address range.
|
||||
auto tail = ranges::unique(ent, ent + num_entries, {}, &Entry::val);
|
||||
num_entries -= tail.size();
|
||||
buf.resize(num_entries * sizeof(Entry));
|
||||
|
||||
// Make addresses relative to themselves.
|
||||
tbb::parallel_for((i64)0, num_entries, [&](i64 i) {
|
||||
i64 offset = sizeof(Entry) * i;
|
||||
ent[i].addr = 0x7fff'ffff & (ent[i].addr - offset);
|
||||
if (is_relative(ent[i].val))
|
||||
ent[i].val = 0x7fff'ffff & (ent[i].val - offset);
|
||||
});
|
||||
|
||||
return buf;
|
||||
}
|
||||
|
||||
template <>
|
||||
void Arm32ExidxSection<E>::remove_duplicate_entries(Context<E> &ctx) {
|
||||
this->shdr.sh_size = get_contents(ctx).size();
|
||||
}
|
||||
|
||||
template <>
|
||||
void Arm32ExidxSection<E>::copy_buf(Context<E> &ctx) {
|
||||
std::vector<u8> contents = get_contents(ctx);
|
||||
assert(this->shdr.sh_size == contents.size());
|
||||
write_vector(ctx.buf + this->shdr.sh_offset, contents);
|
||||
}
|
||||
|
||||
// Even though using ARM32 in big-endian mode is very rare, the processor
|
||||
// technically supports both little- and big-endian modes. There are two
|
||||
// variants of big-endian mode: BE32 and BE8. In BE32, instructions and
|
||||
// data are encoded in big-endian. In BE8, instructions are encoded in
|
||||
// little-endian, and only data is in big-endian. BE8 is the de facto
|
||||
// standard for ARMv6 or later. We support only BE8.
|
||||
//
|
||||
// A tricky thing is that instructions in an object file are always
|
||||
// big-endian if the file is compiled for big-endian mode. In other words,
|
||||
// the compiler always emit code in BE32 if -mbig-endian is specified. It
|
||||
// is the linker's responsibility to rewrite instructions from big-endian
|
||||
// to little-endian for an BE8 output. This function does that.
|
||||
//
|
||||
// The text section may contain a mix of 32-bit ARM instructions, 16-bit
|
||||
// Thumb instructions, and data. We need to distinguish them to swap 4
|
||||
// bytes, 2 bytes, or not swap bytes, respectively. The beginning of ARM
|
||||
// code, Thumb code, and data is labeled with a mapping symbol of $a, $t,
|
||||
// and $d, respectively. We use mapping symbols to determine what to do
|
||||
// with the text section.
|
||||
//
|
||||
// This function is called after we copy the input section contents to the
|
||||
// output file. We rewrite instructions in the output buffer in place.
|
||||
#if MOLD_ARM32BE
|
||||
void arm32be_swap_bytes(Context<E> &ctx) {
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
// Collect mapping symbols
|
||||
std::vector<Symbol<E> *> syms;
|
||||
|
||||
for (Symbol<E> *sym : file->get_local_syms())
|
||||
if (InputSection<E> *isec = sym->get_input_section())
|
||||
if (isec->is_alive && (isec->shdr().sh_flags & SHF_EXECINSTR))
|
||||
if (std::string_view x = sym->name();
|
||||
x == "$a" || x.starts_with("$a.") ||
|
||||
x == "$t" || x.starts_with("$t.") ||
|
||||
x == "$d" || x.starts_with("$d."))
|
||||
syms.push_back(sym);
|
||||
|
||||
// Group mapping symbols by input section and sort by address
|
||||
ranges::stable_sort(syms, {}, [](const Symbol<E> *sym) {
|
||||
return std::tuple{(uintptr_t)sym->get_input_section(), sym->value};
|
||||
});
|
||||
|
||||
// Swap bytes
|
||||
for (i64 i = 0; i < syms.size(); i++) {
|
||||
Symbol<E> &sym = *syms[i];
|
||||
if (sym.name().starts_with("$d"))
|
||||
continue;
|
||||
|
||||
InputSection<E> &isec = *sym.get_input_section();
|
||||
u8 *base = ctx.buf + isec.output_section->shdr.sh_offset + isec.offset;
|
||||
u8 *start = base + sym.value;
|
||||
u8 *end;
|
||||
|
||||
if (i + 1 < syms.size() && syms[i + 1]->get_input_section() == &isec)
|
||||
end = base + syms[i + 1]->value;
|
||||
else
|
||||
end = base + isec.sh_size;
|
||||
|
||||
i64 sz = sym.name().starts_with("$a") ? 4 : 2;
|
||||
for (u8 *p = start; p < end; p += sz)
|
||||
std::reverse(p, p + sz);
|
||||
}
|
||||
});
|
||||
}
|
||||
#endif
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,703 +0,0 @@
|
||||
// This file contains ARM64-specific code. Being new, the ARM64's ELF
|
||||
// psABI doesn't have anything peculiar. ARM64 is a clean RISC
|
||||
// instruction set that supports PC-relative load/store instructions.
|
||||
//
|
||||
// Unlike ARM32, instructions length doesn't vary. All ARM64
|
||||
// instructions are 4 bytes long.
|
||||
//
|
||||
// Branch instructions used for function call can jump within ±128 MiB.
|
||||
// We need to create range extension thunks to support binaries whose
|
||||
// .text is larger than that.
|
||||
//
|
||||
// Unlike most other targets, the TLSDESC access model is used by default
|
||||
// for -fPIC to access thread-local variables instead of the less
|
||||
// efficient GD model. You can still enable GD but it needs the
|
||||
// -mtls-dialect=trad flag. Since GD is used rarely, we don't need to
|
||||
// implement GD → LE relaxation.
|
||||
//
|
||||
// https://github.com/ARM-software/abi-aa/blob/main/aaelf64/aaelf64.rst
|
||||
|
||||
#if MOLD_ARM64LE || MOLD_ARM64BE
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
static void write_adrp(u8 *buf, u64 val) {
|
||||
*(ul32 *)buf |= (bits(val, 13, 12) << 29) | (bits(val, 32, 14) << 5);
|
||||
}
|
||||
|
||||
static void write_adr(u8 *buf, u64 val) {
|
||||
*(ul32 *)buf |= (bits(val, 1, 0) << 29) | (bits(val, 20, 2) << 5);
|
||||
}
|
||||
|
||||
static void write_movn_movz(u8 *buf, i64 val) {
|
||||
*(ul32 *)buf &= 0b0000'0000'0110'0000'0000'0000'0001'1111;
|
||||
|
||||
if (val >= 0)
|
||||
*(ul32 *)buf |= 0xd280'0000 | (bits(val, 15, 0) << 5); // rewrite to movz
|
||||
else
|
||||
*(ul32 *)buf |= 0x9280'0000 | (bits(~val, 15, 0) << 5); // rewrite to movn
|
||||
}
|
||||
|
||||
static u64 page(u64 val) {
|
||||
return val & 0xffff'ffff'ffff'f000;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
constexpr ul32 insn[] = {
|
||||
0xa9bf'7bf0, // stp x16, x30, [sp,#-16]!
|
||||
0x9000'0010, // adrp x16, .got.plt[2]
|
||||
0xf940'0211, // ldr x17, [x16, .got.plt[2]]
|
||||
0x9100'0210, // add x16, x16, .got.plt[2]
|
||||
0xd61f'0220, // br x17
|
||||
0xd420'7d00, // brk
|
||||
0xd420'7d00, // brk
|
||||
0xd420'7d00, // brk
|
||||
};
|
||||
|
||||
u64 gotplt = ctx.gotplt->shdr.sh_addr + 16;
|
||||
u64 plt = ctx.plt->shdr.sh_addr;
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
write_adrp(buf + 4, page(gotplt) - page(plt + 4));
|
||||
*(ul32 *)(buf + 8) |= bits(gotplt, 11, 3) << 10;
|
||||
*(ul32 *)(buf + 12) |= (gotplt & 0xfff) << 10;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
constexpr ul32 insn[] = {
|
||||
0x9000'0010, // adrp x16, .got.plt[n]
|
||||
0xf940'0211, // ldr x17, [x16, .got.plt[n]]
|
||||
0x9100'0210, // add x16, x16, .got.plt[n]
|
||||
0xd61f'0220, // br x17
|
||||
};
|
||||
|
||||
u64 gotplt = sym.get_gotplt_addr(ctx);
|
||||
u64 plt = sym.get_plt_addr(ctx);
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
write_adrp(buf, page(gotplt) - page(plt));
|
||||
*(ul32 *)(buf + 4) |= bits(gotplt, 11, 3) << 10;
|
||||
*(ul32 *)(buf + 8) |= (gotplt & 0xfff) << 10;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
constexpr ul32 insn[] = {
|
||||
0x9000'0010, // adrp x16, GOT[n]
|
||||
0xf940'0211, // ldr x17, [x16, GOT[n]]
|
||||
0xd61f'0220, // br x17
|
||||
0xd420'7d00, // brk
|
||||
};
|
||||
|
||||
u64 got = sym.get_got_pltgot_addr(ctx);
|
||||
u64 plt = sym.get_plt_addr(ctx);
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
write_adrp(buf, page(got) - page(plt));
|
||||
*(ul32 *)(buf + 4) |= bits(got, 11, 3) << 10;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_AARCH64_ABS64:
|
||||
*(U64<E> *)loc = val;
|
||||
break;
|
||||
case R_AARCH64_PREL32:
|
||||
*(U32<E> *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
case R_AARCH64_PREL64:
|
||||
*(U64<E> *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
static bool is_adrp(u8 *loc) {
|
||||
// https://developer.arm.com/documentation/ddi0596/2021-12/Base-Instructions/ADRP--Form-PC-relative-address-to-4KB-page-
|
||||
u32 insn = *(ul32 *)loc;
|
||||
return (bits(insn, 31, 24) & 0b1001'1111) == 0b1001'0000;
|
||||
}
|
||||
|
||||
static bool is_ldr(u8 *loc) {
|
||||
// https://developer.arm.com/documentation/ddi0596/2021-12/Base-Instructions/LDR--immediate---Load-Register--immediate--
|
||||
u32 insn = *(ul32 *)loc;
|
||||
return (bits(insn, 31, 20) & 0b1111'1111'1100) == 0b1111'1001'0100;
|
||||
}
|
||||
|
||||
static bool is_add(u8 *loc) {
|
||||
// https://developer.arm.com/documentation/ddi0596/2021-12/Base-Instructions/ADD--immediate---Add--immediate--
|
||||
u32 insn = *(ul32 *)loc;
|
||||
return (bits(insn, 31, 20) & 0b1111'1111'1100) == 0b1001'0001'0000;
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64:
|
||||
break;
|
||||
case R_AARCH64_LDST8_ABS_LO12_NC:
|
||||
case R_AARCH64_ADD_ABS_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 11, 0) << 10;
|
||||
break;
|
||||
case R_AARCH64_LDST16_ABS_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 11, 1) << 10;
|
||||
break;
|
||||
case R_AARCH64_LDST32_ABS_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 11, 2) << 10;
|
||||
break;
|
||||
case R_AARCH64_LDST64_ABS_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 11, 3) << 10;
|
||||
break;
|
||||
case R_AARCH64_LDST128_ABS_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 11, 4) << 10;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G0:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ul32 *)loc |= bits(S + A, 15, 0) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G0_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 15, 0) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G1:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ul32 *)loc |= bits(S + A, 31, 16) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G1_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 31, 16) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G2:
|
||||
check(S + A, 0, 1LL << 48);
|
||||
*(ul32 *)loc |= bits(S + A, 47, 32) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G2_NC:
|
||||
*(ul32 *)loc |= bits(S + A, 47, 32) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_UABS_G3:
|
||||
*(ul32 *)loc |= bits(S + A, 63, 48) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G0:
|
||||
check(S + A - P, -(1 << 15), 1 << 15);
|
||||
write_movn_movz(loc, S + A - P);
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G0_NC:
|
||||
*(ul32 *)loc |= bits(S + A - P, 15, 0) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G1:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 31);
|
||||
write_movn_movz(loc, (i64)(S + A - P) >> 16);
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G1_NC:
|
||||
*(ul32 *)loc |= bits(S + A - P, 31, 16) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G2:
|
||||
check(S + A - P, -(1LL << 47), 1LL << 47);
|
||||
write_movn_movz(loc, (i64)(S + A - P) >> 32);
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G2_NC:
|
||||
*(ul32 *)loc |= bits(S + A - P, 47, 32) << 5;
|
||||
break;
|
||||
case R_AARCH64_MOVW_PREL_G3:
|
||||
*(ul32 *)loc |= bits(S + A - P, 63, 48) << 5;
|
||||
break;
|
||||
case R_AARCH64_ADR_GOT_PAGE:
|
||||
if (sym.has_got(ctx)) {
|
||||
i64 val = page(G + GOT + A) - page(P);
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
} else {
|
||||
// Relax GOT-loading ADRP+LDR to an immediate ADRP+ADD
|
||||
i64 val = page(S + A) - page(P);
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
|
||||
u32 reg = bits(*(ul32 *)loc, 4, 0);
|
||||
*(ul32 *)(loc + 4) = 0x9100'0000 | (reg << 5) | reg; // ADD
|
||||
*(ul32 *)(loc + 4) |= bits(S + A, 11, 0) << 10;
|
||||
i++;
|
||||
}
|
||||
break;
|
||||
case R_AARCH64_ADR_PREL_PG_HI21:
|
||||
case R_AARCH64_ADR_PREL_PG_HI21_NC: {
|
||||
// The ARM64 psABI defines that an `ADRP x0, foo` and `ADD x0, x0,
|
||||
// :lo12: foo` instruction pair to materialize a PC-relative address
|
||||
// in a register can be relaxed to `NOP` followed by `ADR x0, foo`
|
||||
// if foo is in PC ± 1 MiB.
|
||||
if (ctx.arg.relax && sym.is_pcrel_linktime_const(ctx) &&
|
||||
i + 1 < rels.size()) {
|
||||
i64 val = S + A - P - 4;
|
||||
const ElfRel<E> &rel2 = rels[i + 1];
|
||||
if (is_int(val, 21) &&
|
||||
rel2.r_type == R_AARCH64_ADD_ABS_LO12_NC &&
|
||||
rel2.r_sym == rel.r_sym &&
|
||||
rel2.r_offset == rel.r_offset + 4 &&
|
||||
rel2.r_addend == rel.r_addend &&
|
||||
is_adrp(loc) &&
|
||||
is_add(loc + 4)) {
|
||||
u32 reg1 = bits(*(ul32 *)loc, 4, 0);
|
||||
u32 reg2 = bits(*(ul32 *)(loc + 4), 4, 0);
|
||||
if (reg1 == reg2) {
|
||||
*(ul32 *)loc = 0xd503'201f; // nop
|
||||
*(ul32 *)(loc + 4) = 0x1000'0000 | reg1; // adr
|
||||
write_adr(loc + 4, val);
|
||||
i++;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
i64 val = page(S + A) - page(P);
|
||||
if (rel.r_type == R_AARCH64_ADR_PREL_PG_HI21)
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_ADR_PREL_LO21:
|
||||
check(S + A - P, -(1LL << 20), 1LL << 20);
|
||||
write_adr(loc, S + A - P);
|
||||
break;
|
||||
case R_AARCH64_CALL26:
|
||||
case R_AARCH64_JUMP26: {
|
||||
if (sym.is_remaining_undef_weak()) {
|
||||
// On ARM, calling an weak undefined symbol jumps to the
|
||||
// next instruction.
|
||||
*(ul32 *)loc = 0xd503'201f; // nop
|
||||
break;
|
||||
}
|
||||
|
||||
i64 val = S + A - P;
|
||||
if (!is_int(val, 28))
|
||||
val = sym.get_thunk_addr(ctx, P) + A - P;
|
||||
*(ul32 *)loc |= bits(val, 27, 2);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_PLT32:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 31);
|
||||
*(U32<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_AARCH64_GOTPCREL32:
|
||||
check(G + GOT + A - P, -(1LL << 31), 1LL << 31);
|
||||
*(U32<E> *)loc = G + GOT + A - P;
|
||||
break;
|
||||
case R_AARCH64_CONDBR19:
|
||||
case R_AARCH64_LD_PREL_LO19:
|
||||
check(S + A - P, -(1LL << 20), 1LL << 20);
|
||||
*(ul32 *)loc |= bits(S + A - P, 20, 2) << 5;
|
||||
break;
|
||||
case R_AARCH64_PREL16:
|
||||
check(S + A - P, -(1LL << 15), 1LL << 16);
|
||||
*(U16<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_AARCH64_PREL32:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 32);
|
||||
*(U32<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_AARCH64_PREL64:
|
||||
*(U64<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_AARCH64_LD64_GOT_LO12_NC:
|
||||
*(ul32 *)loc |= bits(G + GOT + A, 11, 3) << 10;
|
||||
break;
|
||||
case R_AARCH64_LD64_GOTPAGE_LO15: {
|
||||
i64 val = G + GOT + A - page(GOT);
|
||||
check(val, 0, 1 << 15);
|
||||
*(ul32 *)loc |= bits(val, 14, 3) << 10;
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21: {
|
||||
i64 val = page(sym.get_gottp_addr(ctx) + A) - page(P);
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC:
|
||||
*(ul32 *)loc |= bits(sym.get_gottp_addr(ctx) + A, 11, 3) << 10;
|
||||
break;
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0: {
|
||||
i64 val = S + A - ctx.tp_addr;
|
||||
check(val, -(1 << 15), 1 << 15);
|
||||
write_movn_movz(loc, val);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0_NC:
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 15, 0) << 5;
|
||||
break;
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1: {
|
||||
i64 val = S + A - ctx.tp_addr;
|
||||
check(val, -(1LL << 31), 1LL << 31);
|
||||
write_movn_movz(loc, val >> 16);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1_NC:
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 31, 16) << 5;
|
||||
break;
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G2: {
|
||||
i64 val = S + A - ctx.tp_addr;
|
||||
check(val, -(1LL << 47), 1LL << 47);
|
||||
write_movn_movz(loc, val >> 32);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_HI12: {
|
||||
i64 val = S + A - ctx.tp_addr;
|
||||
check(val, 0, 1LL << 24);
|
||||
*(ul32 *)loc |= bits(val, 23, 12) << 10;
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12:
|
||||
check(S + A - ctx.tp_addr, 0, 1 << 12);
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 11, 0) << 10;
|
||||
break;
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12_NC:
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 11, 0) << 10;
|
||||
break;
|
||||
case R_AARCH64_TLSGD_ADR_PAGE21: {
|
||||
i64 val = page(sym.get_tlsgd_addr(ctx) + A) - page(P);
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
break;
|
||||
}
|
||||
case R_AARCH64_TLSGD_ADD_LO12_NC:
|
||||
*(ul32 *)loc |= bits(sym.get_tlsgd_addr(ctx) + A, 11, 0) << 10;
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_ADR_PAGE21:
|
||||
// ARM64 TLSDESC uses the following code sequence to materialize
|
||||
// a TP-relative address in x0.
|
||||
//
|
||||
// adrp x0, 0
|
||||
// R_AARCH64_TLSDESC_ADR_PAGE21 foo
|
||||
// ldr x1, [x0]
|
||||
// R_AARCH64_TLSDESC_LD64_LO12 foo
|
||||
// add x0, x0, #0
|
||||
// R_AARCH64_TLSDESC_ADD_LO12 foo
|
||||
// blr x1
|
||||
// R_AARCH64_TLSDESC_CALL foo
|
||||
//
|
||||
// We may relax the instructions to the following if its TP-relative
|
||||
// address is known at link-time
|
||||
//
|
||||
// nop
|
||||
// nop
|
||||
// movz x0, :tls_offset_hi:foo, lsl #16
|
||||
// movk x0, :tls_offset_lo:foo
|
||||
//
|
||||
// or to the following if the TP-relative address is known at
|
||||
// process startup time.
|
||||
//
|
||||
// nop
|
||||
// nop
|
||||
// adrp x0, :gottprel:foo
|
||||
// ldr x0, [x0, :gottprel_lo12:foo]
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
i64 val = page(sym.get_tlsdesc_addr(ctx) + A) - page(P);
|
||||
check(val, -(1LL << 32), 1LL << 32);
|
||||
write_adrp(loc, val);
|
||||
} else {
|
||||
*(ul32 *)loc = 0xd503'201f; // nop
|
||||
}
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_LD64_LO12:
|
||||
if (sym.has_tlsdesc(ctx))
|
||||
*(ul32 *)loc |= bits(sym.get_tlsdesc_addr(ctx) + A, 11, 3) << 10;
|
||||
else
|
||||
*(ul32 *)loc = 0xd503'201f; // nop
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_ADD_LO12:
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
*(ul32 *)loc |= bits(sym.get_tlsdesc_addr(ctx) + A, 11, 0) << 10;
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
*(ul32 *)loc = 0x9000'0000; // adrp x0, 0
|
||||
write_adrp(loc, page(sym.get_gottp_addr(ctx) + A) - page(P));
|
||||
} else {
|
||||
*(ul32 *)loc = 0xd2a0'0000; // movz x0, 0, lsl #16
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 32, 16) << 5;
|
||||
}
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_CALL:
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
// Do nothing
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
*(ul32 *)loc = 0xf940'0000; // ldr x0, [x0, 0]
|
||||
*(ul32 *)loc |= bits(sym.get_gottp_addr(ctx) + A, 11, 3) << 10;
|
||||
} else {
|
||||
*(ul32 *)loc = 0xf280'0000; // movk x0, 0
|
||||
*(ul32 *)loc |= bits(S + A - ctx.tp_addr, 15, 0) << 5;
|
||||
}
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, val, i, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_ABS64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(U64<E> *)loc = *val;
|
||||
else
|
||||
*(U64<E> *)loc = S + A;
|
||||
break;
|
||||
case R_AARCH64_ABS32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(U32<E> *)loc = S + A;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_AARCH64_MOVW_UABS_G3:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_AARCH64_ADR_GOT_PAGE:
|
||||
// An ADR_GOT_PAGE and GOT_LO12_NC relocation pair is used to load a
|
||||
// symbol's address from GOT. If the GOT value is a link-time
|
||||
// constant, we may be able to rewrite the ADRP+LDR instruction pair
|
||||
// with an ADRP+ADD, eliminating a GOT memory load.
|
||||
if (ctx.arg.relax && sym.is_pcrel_linktime_const(ctx) &&
|
||||
i + 1 < rels.size()) {
|
||||
// ADRP+LDR must be consecutive and use the same register to relax.
|
||||
const ElfRel<E> &rel2 = rels[i + 1];
|
||||
if (rel2.r_type == R_AARCH64_LD64_GOT_LO12_NC &&
|
||||
rel2.r_offset == rel.r_offset + 4 &&
|
||||
rel2.r_sym == rel.r_sym &&
|
||||
rel.r_addend == 0 &&
|
||||
rel2.r_addend == 0 &&
|
||||
is_adrp(loc) &&
|
||||
is_ldr(loc + 4)) {
|
||||
u32 rd = bits(*(ul32 *)loc, 4, 0);
|
||||
u32 rn = bits(*(ul32 *)(loc + 4), 9, 5);
|
||||
u32 rt = bits(*(ul32 *)(loc + 4), 4, 0);
|
||||
if (rd == rn && rn == rt) {
|
||||
i++;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_AARCH64_LD64_GOT_LO12_NC:
|
||||
case R_AARCH64_LD64_GOTPAGE_LO15:
|
||||
case R_AARCH64_GOTPCREL32:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_AARCH64_CALL26:
|
||||
case R_AARCH64_JUMP26:
|
||||
case R_AARCH64_PLT32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_AARCH64_TLSIE_ADR_GOTTPREL_PAGE21:
|
||||
case R_AARCH64_TLSIE_LD64_GOTTPREL_LO12_NC:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_AARCH64_ADR_PREL_PG_HI21:
|
||||
case R_AARCH64_ADR_PREL_PG_HI21_NC:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_AARCH64_TLSGD_ADR_PAGE21:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_AARCH64_TLSDESC_CALL:
|
||||
scan_tlsdesc(ctx, sym);
|
||||
break;
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G2:
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12:
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_LO12_NC:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_AARCH64_ABS64:
|
||||
case R_AARCH64_ADD_ABS_LO12_NC:
|
||||
case R_AARCH64_ADR_PREL_LO21:
|
||||
case R_AARCH64_CONDBR19:
|
||||
case R_AARCH64_LD_PREL_LO19:
|
||||
case R_AARCH64_LDST16_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST32_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST64_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST128_ABS_LO12_NC:
|
||||
case R_AARCH64_LDST8_ABS_LO12_NC:
|
||||
case R_AARCH64_MOVW_UABS_G0:
|
||||
case R_AARCH64_MOVW_UABS_G0_NC:
|
||||
case R_AARCH64_MOVW_UABS_G1:
|
||||
case R_AARCH64_MOVW_UABS_G1_NC:
|
||||
case R_AARCH64_MOVW_UABS_G2:
|
||||
case R_AARCH64_MOVW_UABS_G2_NC:
|
||||
case R_AARCH64_MOVW_PREL_G0:
|
||||
case R_AARCH64_MOVW_PREL_G0_NC:
|
||||
case R_AARCH64_MOVW_PREL_G1:
|
||||
case R_AARCH64_MOVW_PREL_G1_NC:
|
||||
case R_AARCH64_MOVW_PREL_G2:
|
||||
case R_AARCH64_MOVW_PREL_G2_NC:
|
||||
case R_AARCH64_MOVW_PREL_G3:
|
||||
case R_AARCH64_PREL16:
|
||||
case R_AARCH64_PREL32:
|
||||
case R_AARCH64_PREL64:
|
||||
case R_AARCH64_TLSGD_ADD_LO12_NC:
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0:
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G0_NC:
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1:
|
||||
case R_AARCH64_TLSLE_MOVW_TPREL_G1_NC:
|
||||
case R_AARCH64_TLSLE_ADD_TPREL_HI12:
|
||||
case R_AARCH64_TLSDESC_ADR_PAGE21:
|
||||
case R_AARCH64_TLSDESC_LD64_LO12:
|
||||
case R_AARCH64_TLSDESC_ADD_LO12:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// The size of a thunk entry varies on ARM64 depending on the distance to
|
||||
// the branch target. This function computes the size of each thunk entry.
|
||||
template <>
|
||||
void Thunk<E>::shrink_size(Context<E> &ctx) {
|
||||
offsets.clear();
|
||||
offsets.push_back(0);
|
||||
i64 off = 0;
|
||||
|
||||
// The distance between S and P is only reduced by shrink_size(), but
|
||||
// page(S) – page(P) may still increase by one page due to address
|
||||
// changes, so we add a safety margin.
|
||||
//
|
||||
// For example, page(0x1200) – page(0x1000) is 0, whereas
|
||||
// page(0x1100) – page(0xfff) is 0x1000, even though the latter
|
||||
// distance is shorter than the former.
|
||||
auto is_small = [](i64 prel) {
|
||||
return is_int(prel + 0x1000, 33) && is_int(prel - 0x1000, 33);
|
||||
};
|
||||
|
||||
for (Symbol<E> *sym : symbols) {
|
||||
u64 S = sym->get_addr(ctx);
|
||||
u64 P = get_addr() + off;
|
||||
i64 prel = page(S) - page(P);
|
||||
off += is_small(prel) ? 16 : 32;
|
||||
offsets.push_back(off);
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Thunk<E>::copy_buf(Context<E> &ctx) {
|
||||
// Short thunk with a 33 bit displacement
|
||||
constexpr ul32 insn1[] = {
|
||||
0x9000'0010, // adrp x16, 0
|
||||
0x9100'0210, // add x16, x16
|
||||
0xd61f'0200, // br x16
|
||||
0xd420'7d00, // brk
|
||||
};
|
||||
|
||||
// Long thunk with a 64 bit displacement
|
||||
constexpr ul32 insn2[] = {
|
||||
0x1000'0010, // adr x16, 0
|
||||
0xd2a0'0011, // movz x17, 0, lsl #16
|
||||
0xf2c0'0011, // movk x17, 0, lsl #32
|
||||
0xf2e0'0011, // movk x17, 0, lsl #48
|
||||
0x8b11'0210, // add x16, x16, x17
|
||||
0xd61f'0200, // br x16
|
||||
0xd420'7d00, // brk
|
||||
0xd420'7d00, // brk
|
||||
};
|
||||
|
||||
u8 *base = ctx.buf + output_section.shdr.sh_offset + offset;
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
u64 S = symbols[i]->get_addr(ctx);
|
||||
u64 P = get_addr() + offsets[i];
|
||||
u8 *buf = base + offsets[i];
|
||||
|
||||
if (offsets[i + 1] - offsets[i] == 16) {
|
||||
i64 prel = page(S) - page(P);
|
||||
assert(is_int(prel, 33));
|
||||
memcpy(buf, insn1, sizeof(insn1));
|
||||
write_adrp(buf, prel);
|
||||
*(ul32 *)(buf + 4) |= bits(S, 11, 0) << 10;
|
||||
} else {
|
||||
memcpy(buf, insn2, sizeof(insn2));
|
||||
write_adr(buf, bits(S - P, 15, 0));
|
||||
*(ul32 *)(buf + 4) |= bits(S - P, 31, 16) << 5;
|
||||
*(ul32 *)(buf + 8) |= bits(S - P, 47, 32) << 5;
|
||||
*(ul32 *)(buf + 12) |= bits(S - P, 63, 48) << 5;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template class Thunk<E>;
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,602 +0,0 @@
|
||||
// i386 is similar to x86-64 but lacks PC-relative memory access
|
||||
// instructions. So it's not straightforward to support position-
|
||||
// independent code (PIC) on that target.
|
||||
//
|
||||
// If an object file is compiled with -fPIC, a function that needs to load
|
||||
// a value from memory first obtains its own address with the following
|
||||
// code
|
||||
//
|
||||
// call __x86.get_pc_thunk.bx
|
||||
//
|
||||
// where __x86.get_pc_thunk.bx is defined as
|
||||
//
|
||||
// __x86.get_pc_thunk.bx:
|
||||
// mov (%esp), %ebx # move the return address to %ebx
|
||||
// ret
|
||||
//
|
||||
// . With the function's own address (or, more precisely, the address
|
||||
// immediately after the call instruction), the function can compute an
|
||||
// absolute address of a variable with its address + link-time constant.
|
||||
//
|
||||
// Executing call-mov-ret isn't very cheap, and allocating one register to
|
||||
// store PC isn't cheap too, especially given that i386 has only 8
|
||||
// general-purpose registers. But that's the cost of PIC on i386. You need
|
||||
// to pay it when creating a .so and a position-independent executable.
|
||||
//
|
||||
// When a position-independent function calls another function, it sets
|
||||
// %ebx to the address of .got. Position-independent PLT entries use that
|
||||
// register to load values from .got.plt/.got.
|
||||
//
|
||||
// If we are creating a position-dependent executable (PDE), we can't
|
||||
// assume that %ebx is set to .got. For PDE, we need to create position-
|
||||
// dependent PLT entries which don't use %ebx.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/i386.pdf
|
||||
|
||||
#if MOLD_I386
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = I386;
|
||||
|
||||
template <>
|
||||
i64 get_addend(u8 *loc, const ElfRel<E> &rel) {
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
case R_386_PC8:
|
||||
return *(i8 *)loc;
|
||||
case R_386_16:
|
||||
case R_386_PC16:
|
||||
return *(il16 *)loc;
|
||||
case R_386_32:
|
||||
case R_386_PC32:
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X:
|
||||
case R_386_PLT32:
|
||||
case R_386_GOTOFF:
|
||||
case R_386_GOTPC:
|
||||
case R_386_TLS_LDM:
|
||||
case R_386_TLS_GOTIE:
|
||||
case R_386_TLS_LE:
|
||||
case R_386_TLS_IE:
|
||||
case R_386_TLS_GD:
|
||||
case R_386_TLS_LDO_32:
|
||||
case R_386_SIZE32:
|
||||
case R_386_TLS_GOTDESC:
|
||||
return *(il32 *)loc;
|
||||
default:
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_addend(u8 *loc, i64 val, const ElfRel<E> &rel) {
|
||||
switch (rel.r_type) {
|
||||
case R_386_NONE:
|
||||
break;
|
||||
case R_386_8:
|
||||
case R_386_PC8:
|
||||
*loc = val;
|
||||
break;
|
||||
case R_386_16:
|
||||
case R_386_PC16:
|
||||
*(ul16 *)loc = val;
|
||||
break;
|
||||
case R_386_32:
|
||||
case R_386_PC32:
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X:
|
||||
case R_386_PLT32:
|
||||
case R_386_GOTOFF:
|
||||
case R_386_GOTPC:
|
||||
case R_386_TLS_LDM:
|
||||
case R_386_TLS_GOTIE:
|
||||
case R_386_TLS_LE:
|
||||
case R_386_TLS_IE:
|
||||
case R_386_TLS_GD:
|
||||
case R_386_TLS_LDO_32:
|
||||
case R_386_SIZE32:
|
||||
case R_386_TLS_GOTDESC:
|
||||
*(ul32 *)loc = val;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 insn[] = {
|
||||
0x51, // push %ecx
|
||||
0x8d, 0x8b, 0, 0, 0, 0, // lea GOTPLT+4(%ebx), %ecx
|
||||
0xff, 0x31, // push (%ecx)
|
||||
0xff, 0x61, 0x04, // jmp *0x4(%ecx)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 3) = ctx.gotplt->shdr.sh_addr - ctx.got->shdr.sh_addr + 4;
|
||||
} else {
|
||||
static const u8 insn[] = {
|
||||
0x51, // push %ecx
|
||||
0xb9, 0, 0, 0, 0, // mov GOTPLT+4, %ecx
|
||||
0xff, 0x31, // push (%ecx)
|
||||
0xff, 0x61, 0x04, // jmp *0x4(%ecx)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 2) = ctx.gotplt->shdr.sh_addr + 4;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 insn[] = {
|
||||
0xb9, 0, 0, 0, 0, // mov $reloc_offset, %ecx
|
||||
0xff, 0xa3, 0, 0, 0, 0, // jmp *foo@GOT(%ebx)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 1) = sym.get_plt_idx(ctx) * sizeof(ElfRel<E>);
|
||||
*(ul32 *)(buf + 7) = sym.get_gotplt_addr(ctx) - ctx.got->shdr.sh_addr;
|
||||
} else {
|
||||
static const u8 insn[] = {
|
||||
0xb9, 0, 0, 0, 0, // mov $reloc_offset, %ecx
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0xcc, 0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 1) = sym.get_plt_idx(ctx) * sizeof(ElfRel<E>);
|
||||
*(ul32 *)(buf + 7) = sym.get_gotplt_addr(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
if (ctx.arg.pic) {
|
||||
static const u8 insn[] = {
|
||||
0xff, 0xa3, 0, 0, 0, 0, // jmp *foo@GOT(%ebx)
|
||||
0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 2) = sym.get_got_pltgot_addr(ctx) - ctx.got->shdr.sh_addr;
|
||||
} else {
|
||||
static const u8 insn[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0xcc, 0xcc, // (padding)
|
||||
};
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 2) = sym.get_got_pltgot_addr(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_386_32:
|
||||
*(ul32 *)loc = val;
|
||||
break;
|
||||
case R_386_PC32:
|
||||
*(ul32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
static u32 relax_got32x(u8 *loc) {
|
||||
// mov imm(%reg1), %reg2 -> lea imm(%reg1), %reg2
|
||||
if (loc[0] == 0x8b)
|
||||
return 0x8d00 | loc[1];
|
||||
return 0;
|
||||
}
|
||||
|
||||
// Relax GD to LE
|
||||
static void relax_gd_to_le(u8 *loc, ElfRel<E> rel, u64 val) {
|
||||
static const u8 insn[] = {
|
||||
0x65, 0xa1, 0, 0, 0, 0, // mov %gs:0, %eax
|
||||
0x81, 0xc0, 0, 0, 0, 0, // add $tp_offset, %eax
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_PLT32:
|
||||
case R_386_PC32:
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 5) = val;
|
||||
break;
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X:
|
||||
memcpy(loc - 2, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 6) = val;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
// Relax LD to LE
|
||||
static void relax_ld_to_le(u8 *loc, ElfRel<E> rel, u64 tls_size) {
|
||||
switch (rel.r_type) {
|
||||
case R_386_PLT32:
|
||||
case R_386_PC32: {
|
||||
static const u8 insn[] = {
|
||||
0x65, 0xa1, 0, 0, 0, 0, // mov %gs:0, %eax
|
||||
0x2d, 0, 0, 0, 0, // sub $tls_size, %eax
|
||||
};
|
||||
memcpy(loc - 2, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 5) = tls_size;
|
||||
break;
|
||||
}
|
||||
case R_386_GOT32:
|
||||
case R_386_GOT32X: {
|
||||
static const u8 insn[] = {
|
||||
0x65, 0xa1, 0, 0, 0, 0, // mov %gs:0, %eax
|
||||
0x81, 0xe8, 0, 0, 0, 0, // sub $tls_size, %eax
|
||||
};
|
||||
memcpy(loc - 2, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 6) = tls_size;
|
||||
break;
|
||||
}
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
static u32 relax_tlsdesc_to_ie(u8 *loc) {
|
||||
switch ((loc[0] << 8) | loc[1]) {
|
||||
case 0x8d83: return 0x8b83; // lea 0(%ebx), %eax -> mov 0(%ebx), %eax
|
||||
case 0x8d9b: return 0x8b9b; // lea 0(%ebx), %ebx -> mov 0(%ebx), %ebx
|
||||
case 0x8d8b: return 0x8b8b; // lea 0(%ebx), %ecx -> mov 0(%ebx), %ecx
|
||||
case 0x8d93: return 0x8b93; // lea 0(%ebx), %edx -> mov 0(%ebx), %edx
|
||||
case 0x8db3: return 0x8bb3; // lea 0(%ebx), %esi -> mov 0(%ebx), %esi
|
||||
case 0x8dbb: return 0x8bbb; // lea 0(%ebx), %edi -> mov 0(%ebx), %edi
|
||||
case 0x8da3: return 0x8ba3; // lea 0(%ebx), %esp -> mov 0(%ebx), %esp
|
||||
case 0x8dab: return 0x8bab; // lea 0(%ebx), %ebp -> mov 0(%ebx), %ebp
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_tlsdesc_to_le(u8 *loc) {
|
||||
switch ((loc[0] << 8) | loc[1]) {
|
||||
case 0x8d83: return 0x90b8; // lea 0(%ebx), %eax -> mov $0, %eax
|
||||
case 0x8d9b: return 0x90bb; // lea 0(%ebx), %ebx -> mov $0, %ebx
|
||||
case 0x8d8b: return 0x90b9; // lea 0(%ebx), %ecx -> mov $0, %ecx
|
||||
case 0x8d93: return 0x90ba; // lea 0(%ebx), %edx -> mov $0, %edx
|
||||
case 0x8db3: return 0x90be; // lea 0(%ebx), %esi -> mov $0, %esi
|
||||
case 0x8dbb: return 0x90bf; // lea 0(%ebx), %edi -> mov $0, %edi
|
||||
case 0x8da3: return 0x90bc; // lea 0(%ebx), %esp -> mov $0, %esp
|
||||
case 0x8dab: return 0x90bd; // lea 0(%ebx), %ebp -> mov $0, %ebp
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = get_addend(*this, rel);
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_386_16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ul16 *)loc = S + A;
|
||||
break;
|
||||
case R_386_32:
|
||||
break;
|
||||
case R_386_PC8:
|
||||
check(S + A - P, -(1 << 7), 1 << 7);
|
||||
*loc = S + A - P;
|
||||
break;
|
||||
case R_386_PC16:
|
||||
check(S + A - P, -(1 << 15), 1 << 15);
|
||||
*(ul16 *)loc = S + A - P;
|
||||
break;
|
||||
case R_386_PC32:
|
||||
case R_386_PLT32:
|
||||
*(ul32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_386_GOT32:
|
||||
*(ul32 *)loc = G + A;
|
||||
break;
|
||||
case R_386_GOT32X:
|
||||
if (sym.has_got(ctx)) {
|
||||
*(ul32 *)loc = G + A;
|
||||
} else {
|
||||
u32 insn = relax_got32x(loc - 2);
|
||||
assert(insn);
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
*(ul32 *)loc = S + A - GOT;
|
||||
}
|
||||
break;
|
||||
case R_386_GOTOFF:
|
||||
*(ul32 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_386_GOTPC:
|
||||
*(ul32 *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_386_TLS_GOTIE:
|
||||
*(ul32 *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_386_TLS_LE:
|
||||
*(ul32 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_386_TLS_IE:
|
||||
*(ul32 *)loc = sym.get_gottp_addr(ctx) + A;
|
||||
break;
|
||||
case R_386_TLS_GD:
|
||||
if (sym.has_tlsgd(ctx))
|
||||
*(ul32 *)loc = sym.get_tlsgd_addr(ctx) + A - GOT;
|
||||
else
|
||||
relax_gd_to_le(loc, rels[++i], S - ctx.tp_addr);
|
||||
break;
|
||||
case R_386_TLS_LDM:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
*(ul32 *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOT;
|
||||
else
|
||||
relax_ld_to_le(loc, rels[++i], ctx.tp_addr - ctx.tls_begin);
|
||||
break;
|
||||
case R_386_TLS_LDO_32:
|
||||
*(ul32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_386_SIZE32:
|
||||
*(ul32 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
case R_386_TLS_GOTDESC:
|
||||
// i386 TLSDESC uses the following code sequence to materialize
|
||||
// a TP-relative address in %eax.
|
||||
//
|
||||
// lea 0(%ebx), %eax
|
||||
// R_386_TLS_GOTDESC foo
|
||||
// call *(%eax)
|
||||
// R_386_TLS_DESC_CALL foo
|
||||
//
|
||||
// We may relax the instructions to the following if its TP-relative
|
||||
// address is known at link-time
|
||||
//
|
||||
// mov $foo@TPOFF, %eax
|
||||
// nop
|
||||
//
|
||||
// or to the following if the TP-relative address is known at
|
||||
// process startup time.
|
||||
//
|
||||
// mov foo@GOTTPOFF(%ebx), %eax
|
||||
// nop
|
||||
//
|
||||
// We allow the following alternative code sequence too because
|
||||
// LLVM emits such code.
|
||||
//
|
||||
// lea 0(%ebx), %reg
|
||||
// R_386_TLS_GOTDESC foo
|
||||
// mov %reg, %eax
|
||||
// call *(%eax)
|
||||
// R_386_TLS_DESC_CALL foo
|
||||
//
|
||||
// Note that the compiler always uses the local-exec TLS model
|
||||
// for -fno-pic, so TLSDESC code is always PIC (i.e. uses %ebx to
|
||||
// store the address of GOT.)
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
*(ul32 *)loc = sym.get_tlsdesc_addr(ctx) + A - GOT;
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
u32 insn = relax_tlsdesc_to_ie(loc - 2);
|
||||
if (!insn)
|
||||
Fatal(ctx) << *this << ": illegal instruction sequence for TLSDESC";
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
*(ul32 *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
} else {
|
||||
u32 insn = relax_tlsdesc_to_le(loc - 2);
|
||||
if (!insn)
|
||||
Fatal(ctx) << *this << ": illegal instruction sequence for TLSDESC";
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
*(ul32 *)loc = S + A - ctx.tp_addr;
|
||||
}
|
||||
break;
|
||||
case R_386_TLS_DESC_CALL:
|
||||
if (!sym.has_tlsdesc(ctx)) {
|
||||
// call *(%eax) -> nop
|
||||
loc[0] = 0x66;
|
||||
loc[1] = 0x90;
|
||||
}
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : get_addend(*this, rel);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, val, i, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_386_16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ul16 *)loc = S + A;
|
||||
break;
|
||||
case R_386_32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul32 *)loc = *val;
|
||||
else
|
||||
*(ul32 *)loc = S + A;
|
||||
break;
|
||||
case R_386_PC8:
|
||||
check(S + A, -(1 << 7), 1 << 7);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_386_PC16:
|
||||
check(S + A, -(1 << 15), 1 << 15);
|
||||
*(ul16 *)loc = S + A;
|
||||
break;
|
||||
case R_386_PC32:
|
||||
*(ul32 *)loc = S + A;
|
||||
break;
|
||||
case R_386_GOTPC:
|
||||
*(ul32 *)loc = GOT + A;
|
||||
break;
|
||||
case R_386_GOTOFF:
|
||||
*(ul32 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_386_TLS_LDO_32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul32 *)loc = *val;
|
||||
else
|
||||
*(ul32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_386_SIZE32:
|
||||
*(ul32 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
if (rel.r_type == R_386_TLS_GD || rel.r_type == R_386_TLS_LDM) {
|
||||
if (i + 1 == rels.size())
|
||||
Fatal(ctx) << *this << ": " << rel << " must be followed by PLT or GOT32";
|
||||
|
||||
if (u32 ty = rels[i + 1].r_type;
|
||||
ty != R_386_PLT32 && ty != R_386_PC32 &&
|
||||
ty != R_386_GOT32 && ty != R_386_GOT32X)
|
||||
Fatal(ctx) << *this << ": " << rel << " must be followed by PLT or GOT32";
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_386_8:
|
||||
case R_386_16:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_386_PC8:
|
||||
case R_386_PC16:
|
||||
case R_386_PC32:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_386_GOT32:
|
||||
case R_386_GOTPC:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_386_GOT32X:
|
||||
// We always want to relax GOT32X even if --no-relax is given
|
||||
// because static PIE doesn't work without it.
|
||||
if (sym.is_pcrel_linktime_const(ctx) && relax_got32x(loc - 2)) {
|
||||
// Do nothing
|
||||
} else {
|
||||
sym.flags |= NEEDS_GOT;
|
||||
}
|
||||
break;
|
||||
case R_386_PLT32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_386_TLS_GOTIE:
|
||||
case R_386_TLS_IE:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_386_TLS_GD:
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && sym.is_tprel_linktime_const(ctx)))
|
||||
i++;
|
||||
else
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_386_TLS_LDM:
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && !ctx.arg.shared))
|
||||
i++;
|
||||
else
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_386_TLS_GOTDESC:
|
||||
scan_tlsdesc(ctx, sym);
|
||||
break;
|
||||
case R_386_TLS_LE:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_386_32:
|
||||
case R_386_GOTOFF:
|
||||
case R_386_TLS_LDO_32:
|
||||
case R_386_SIZE32:
|
||||
case R_386_TLS_DESC_CALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
File diff suppressed because it is too large.
Load diff
@@ -1,316 +0,0 @@
|
||||
// This file contains code for the Motorola 68000 series microprocessors,
|
||||
// which is often abbreviated as m68k. Running a Unix-like system on a
|
||||
// m68k-based machine today is a retro-computing hobby activity, but the
|
||||
// processor was a popular choice to build Unix computers during '80s.
|
||||
// Early Sun workstations for example used m68k. Macintosh until 1994 were
|
||||
// based on m68k as well until they switched to PowerPC (and then to x86
|
||||
// and to ARM.)
|
||||
//
|
||||
// From the linker's point of view, it is not hard to support m68k. It's
|
||||
// just a 32-bit big-endian CISC ISA. Compared to comtemporary i386,
|
||||
// m68k's psABI is actually simpler because m68k has PC-relative memory
|
||||
// access instructions and therefore can support position-independent
|
||||
// code without too much hassle.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/m68k.pdf
|
||||
|
||||
#if MOLD_M68K
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = M68K;
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
static const u8 insn[] = {
|
||||
0x2f, 0x00, // move.l %d0, -(%sp)
|
||||
0x2f, 0x3b, 0x01, 0x70, 0, 0, 0, 0, // move.l (GOTPLT+4, %pc), -(%sp)
|
||||
0x4e, 0xfb, 0x01, 0x71, 0, 0, 0, 0, // jmp ([GOTPLT+8, %pc])
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 6) = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr;
|
||||
*(ub32 *)(buf + 14) = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 4;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static const u8 insn[] = {
|
||||
0x20, 0x3c, 0, 0, 0, 0, // move.l PLT_OFFSET, %d0
|
||||
0x4e, 0xfb, 0x01, 0x71, 0, 0, 0, 0, // jmp ([GOTPLT_ENTRY, %pc])
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 2) = sym.get_plt_idx(ctx) * sizeof(ElfRel<E>);
|
||||
*(ub32 *)(buf + 10) = sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx) - 8;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static const u8 insn[] = {
|
||||
0x4e, 0xfb, 0x01, 0x71, 0, 0, 0, 0, // jmp ([GOT_ENTRY, %pc])
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 4) = sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx) - 2;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_68K_32:
|
||||
*(ub32 *)loc = val;
|
||||
break;
|
||||
case R_68K_PC32:
|
||||
*(ub32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto write16 = [&](u64 val) {
|
||||
check(val, 0, 1 << 16);
|
||||
*(ub16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write16s = [&](u64 val) {
|
||||
check(val, -(1 << 15), 1 << 15);
|
||||
*(ub16 *)loc = val;
|
||||
};
|
||||
|
||||
auto write8 = [&](u64 val) {
|
||||
check(val, 0, 1 << 8);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
auto write8s = [&](u64 val) {
|
||||
check(val, -(1 << 7), 1 << 7);
|
||||
*loc = val;
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_68K_32:
|
||||
break;
|
||||
case R_68K_16:
|
||||
write16(S + A);
|
||||
break;
|
||||
case R_68K_8:
|
||||
write8(S + A);
|
||||
break;
|
||||
case R_68K_PC32:
|
||||
case R_68K_PLT32:
|
||||
*(ub32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_68K_PC16:
|
||||
case R_68K_PLT16:
|
||||
write16s(S + A - P);
|
||||
break;
|
||||
case R_68K_PC8:
|
||||
case R_68K_PLT8:
|
||||
write8s(S + A - P);
|
||||
break;
|
||||
case R_68K_GOTPCREL32:
|
||||
*(ub32 *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_68K_GOTPCREL16:
|
||||
write16s(GOT + A - P);
|
||||
break;
|
||||
case R_68K_GOTPCREL8:
|
||||
write8s(GOT + A - P);
|
||||
break;
|
||||
case R_68K_GOTOFF32:
|
||||
*(ub32 *)loc = G + A;
|
||||
break;
|
||||
case R_68K_GOTOFF16:
|
||||
write16(G + A);
|
||||
break;
|
||||
case R_68K_GOTOFF8:
|
||||
write8(G + A);
|
||||
break;
|
||||
case R_68K_TLS_GD32:
|
||||
*(ub32 *)loc = sym.get_tlsgd_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_68K_TLS_GD16:
|
||||
write16(sym.get_tlsgd_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_GD8:
|
||||
write8(sym.get_tlsgd_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_LDM32:
|
||||
*(ub32 *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_68K_TLS_LDM16:
|
||||
write16(ctx.got->get_tlsld_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_LDM8:
|
||||
write8(ctx.got->get_tlsld_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_LDO32:
|
||||
*(ub32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_68K_TLS_LDO16:
|
||||
write16s(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_68K_TLS_LDO8:
|
||||
write8s(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_68K_TLS_IE32:
|
||||
*(ub32 *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_68K_TLS_IE16:
|
||||
write16(sym.get_gottp_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_IE8:
|
||||
write8(sym.get_gottp_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_68K_TLS_LE32:
|
||||
*(ub32 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_68K_TLS_LE16:
|
||||
write16(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_68K_TLS_LE8:
|
||||
write8(S + A - ctx.tp_addr);
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_68K_32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub32 *)loc = *val;
|
||||
else
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
Error(ctx) << sym << ": GNU ifunc symbol is not supported on m68k";
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_68K_16:
|
||||
case R_68K_8:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_68K_PC32:
|
||||
case R_68K_PC16:
|
||||
case R_68K_PC8:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_68K_GOTPCREL32:
|
||||
case R_68K_GOTPCREL16:
|
||||
case R_68K_GOTPCREL8:
|
||||
case R_68K_GOTOFF32:
|
||||
case R_68K_GOTOFF16:
|
||||
case R_68K_GOTOFF8:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_68K_PLT32:
|
||||
case R_68K_PLT16:
|
||||
case R_68K_PLT8:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_68K_TLS_GD32:
|
||||
case R_68K_TLS_GD16:
|
||||
case R_68K_TLS_GD8:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_68K_TLS_LDM32:
|
||||
case R_68K_TLS_LDM16:
|
||||
case R_68K_TLS_LDM8:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_68K_TLS_IE32:
|
||||
case R_68K_TLS_IE16:
|
||||
case R_68K_TLS_IE8:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_68K_TLS_LE32:
|
||||
case R_68K_TLS_LE16:
|
||||
case R_68K_TLS_LE8:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_68K_32:
|
||||
case R_68K_TLS_LDO32:
|
||||
case R_68K_TLS_LDO16:
|
||||
case R_68K_TLS_LDO8:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,451 +0,0 @@
|
||||
// This file implements the PowerPC 32-bit ISA. For 64-bit PowerPC, see
|
||||
// arch-ppc64v1.cpp and arch-ppc64v2.cpp.
|
||||
//
|
||||
// PPC32 is a RISC ISA. It has 32 general-purpose registers (GPRs).
|
||||
// r0, r11 and r12 are reserved for static linkers, so we can use these
|
||||
// registers in PLTs and range extension thunks. In addition to that, it
|
||||
// has a few special registers. Notable ones are LR which holds a return
|
||||
// address and CTR which we can use to store a branch target address.
|
||||
//
|
||||
// It feels that the PPC32 psABI is unnecessarily complicated at first
|
||||
// glance, but that is mainly stemmed from the fact that the ISA lacks
|
||||
// PC-relative load/store instructions. Since machine instructions cannot
|
||||
// load data relative to its own address, it is not straightforward to
|
||||
// support position-independent code (PIC) on PPC32.
|
||||
//
|
||||
// A position-independent function typically contains the following code
|
||||
// in the prologue to obtain its own address:
|
||||
//
|
||||
// mflr r0 // save the current return address to %r0
|
||||
// bcl 20, 31, 4 // call the next instruction as if it were a function
|
||||
// mtlr r12 // save the return address to %r12
|
||||
// mtlr r0 // restore the original return address
|
||||
//
|
||||
// An object file compiled with -fPIC contains a data section named
|
||||
// `.got2` to store addresses of locally-defined global variables and
|
||||
// constants. A PIC function usually computes its .got2+0x8000 and set it
|
||||
// to %r30. This scheme allows the function to access global objects
|
||||
// defined in the same input file with a single %r30-relative load/store
|
||||
// instruction with a 16-bit offset, given that .got2 is smaller than
|
||||
// 0x10000 (or 65536) bytes.
|
||||
//
|
||||
// Since each object file has its own .got2, %r30 refers to different
|
||||
// places in a merged .got2 for two functions that came from different
|
||||
// input files. Therefore, %r30 makes sense only within a single function.
|
||||
//
|
||||
// Technically, we can reuse a %r30 value in our PLT if we create a PLT
|
||||
// _for each input file_ (that's what GNU ld seems to be doing), but that
|
||||
// doesn't seems to be worth its complexity. Our PLT simply doesn't rely
|
||||
// on a %r30 value.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/ppc32.pdf
|
||||
|
||||
#if MOLD_PPC32
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = PPC32;
|
||||
|
||||
static u64 lo(u64 x) { return x & 0xffff; }
|
||||
static u64 hi(u64 x) { return x >> 16; }
|
||||
static u64 ha(u64 x) { return (x + 0x8000) >> 16; }
|
||||
static u64 high(u64 x) { return (x >> 16) & 0xffff; }
|
||||
static u64 higha(u64 x) { return ((x + 0x8000) >> 16) & 0xffff; }
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
constexpr ub32 insn[] = {
|
||||
// Get the address of this PLT section
|
||||
0x7c08'02a6, // mflr r0
|
||||
0x429f'0005, // bcl 20, 31, 4
|
||||
0x7d88'02a6, // 1: mflr r12
|
||||
0x7c08'03a6, // mtlr r0
|
||||
|
||||
// Compute the runtime address of GOTPLT+12
|
||||
0x3d8c'0000, // addis r12, r12, (GOTPLT - 1b)@higha
|
||||
0x398c'0000, // addi r12, r12, (GOTPLT - 1b)@lo
|
||||
|
||||
// Compute the PLT entry offset
|
||||
0x7d6c'5850, // sub r11, r11, r12
|
||||
0x1d6b'0003, // mulli r11, r11, 3
|
||||
|
||||
// Load GOTPLT[2] and branch to GOTPLT[1]
|
||||
0x800c'fff8, // lwz r0, -8(r12)
|
||||
0x7c09'03a6, // mtctr r0
|
||||
0x818c'fffc, // lwz r12, -4(r12)
|
||||
0x4e80'0420, // bctr
|
||||
0x6000'0000, // nop
|
||||
0x6000'0000, // nop
|
||||
0x6000'0000, // nop
|
||||
0x6000'0000, // nop
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_hdr_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
|
||||
ub32 *loc = (ub32 *)buf;
|
||||
loc[4] |= higha(ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr + 4);
|
||||
loc[5] |= lo(ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr + 4);
|
||||
}
|
||||
|
||||
static constexpr ub32 plt_entry[] = {
|
||||
// Get the address of this PLT entry
|
||||
0x7c08'02a6, // mflr r0
|
||||
0x429f'0005, // bcl 20, 31, 4
|
||||
0x7d88'02a6, // mflr r12
|
||||
0x7c08'03a6, // mtlr r0
|
||||
|
||||
// Load an address from the GOT/GOTPLT entry and jump to that address
|
||||
0x3d6c'0000, // addis r11, r12, OFFSET@higha
|
||||
0x396b'0000, // addi r11, r11, OFFSET@lo
|
||||
0x818b'0000, // lwz r12, 0(r11)
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static_assert(E::plt_size == sizeof(plt_entry));
|
||||
memcpy(buf, plt_entry, sizeof(plt_entry));
|
||||
|
||||
ub32 *loc = (ub32 *)buf;
|
||||
i64 offset = sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx) - 8;
|
||||
loc[4] |= higha(offset);
|
||||
loc[5] |= lo(offset);
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static_assert(E::pltgot_size == sizeof(plt_entry));
|
||||
memcpy(buf, plt_entry, sizeof(plt_entry));
|
||||
|
||||
ub32 *loc = (ub32 *)buf;
|
||||
i64 offset = sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx) - 8;
|
||||
loc[4] |= higha(offset);
|
||||
loc[5] |= lo(offset);
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_PPC_ADDR32:
|
||||
*(ub32 *)loc = val;
|
||||
break;
|
||||
case R_PPC_REL32:
|
||||
*(ub32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
u64 GOT2 = file.extra.got2 ? file.extra.got2->get_addr() : 0;
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC_ADDR14:
|
||||
*(ub32 *)loc |= bits(S + A, 15, 2) << 2;
|
||||
break;
|
||||
case R_PPC_ADDR16:
|
||||
case R_PPC_UADDR16:
|
||||
case R_PPC_ADDR16_LO:
|
||||
*(ub16 *)loc = lo(S + A);
|
||||
break;
|
||||
case R_PPC_ADDR16_HI:
|
||||
*(ub16 *)loc = hi(S + A);
|
||||
break;
|
||||
case R_PPC_ADDR16_HA:
|
||||
*(ub16 *)loc = ha(S + A);
|
||||
break;
|
||||
case R_PPC_ADDR24:
|
||||
*(ub32 *)loc |= bits(S + A, 25, 2) << 2;
|
||||
break;
|
||||
case R_PPC_ADDR30:
|
||||
*(ub32 *)loc |= bits(S + A, 31, 2) << 2;
|
||||
break;
|
||||
case R_PPC_PLT16_LO:
|
||||
*(ub16 *)loc = lo(G + GOT - A - GOT2);
|
||||
break;
|
||||
case R_PPC_PLT16_HI:
|
||||
*(ub16 *)loc = hi(G + GOT - A - GOT2);
|
||||
break;
|
||||
case R_PPC_PLT16_HA:
|
||||
*(ub16 *)loc = ha(G + GOT - A - GOT2);
|
||||
break;
|
||||
case R_PPC_PLT32:
|
||||
*(ub32 *)loc = G + GOT - A - GOT2;
|
||||
break;
|
||||
case R_PPC_REL14:
|
||||
*(ub32 *)loc |= bits(S + A - P, 15, 2) << 2;
|
||||
break;
|
||||
case R_PPC_REL16:
|
||||
case R_PPC_REL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - P);
|
||||
break;
|
||||
case R_PPC_REL16_HI:
|
||||
*(ub16 *)loc = hi(S + A - P);
|
||||
break;
|
||||
case R_PPC_REL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - P);
|
||||
break;
|
||||
case R_PPC_REL24:
|
||||
case R_PPC_LOCAL24PC: {
|
||||
i64 val = S + A - P;
|
||||
if (!is_int(val, 26))
|
||||
val = sym.get_thunk_addr(ctx, P) - P;
|
||||
*(ub32 *)loc |= bits(val, 25, 2) << 2;
|
||||
break;
|
||||
}
|
||||
case R_PPC_PLTREL24: {
|
||||
i64 val = S - P;
|
||||
if (sym.has_plt(ctx) || !is_int(val, 26))
|
||||
val = sym.get_thunk_addr(ctx, P) - P;
|
||||
*(ub32 *)loc |= bits(val, 25, 2) << 2;
|
||||
break;
|
||||
}
|
||||
case R_PPC_REL32:
|
||||
case R_PPC_PLTREL32:
|
||||
*(ub32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_PPC_GOT16:
|
||||
case R_PPC_GOT16_LO:
|
||||
*(ub16 *)loc = lo(G + A);
|
||||
break;
|
||||
case R_PPC_GOT16_HI:
|
||||
*(ub16 *)loc = hi(G + A);
|
||||
break;
|
||||
case R_PPC_GOT16_HA:
|
||||
*(ub16 *)loc = ha(G + A);
|
||||
break;
|
||||
case R_PPC_TPREL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC_TPREL16_HI:
|
||||
*(ub16 *)loc = hi(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC_TPREL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC_DTPREL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC_DTPREL16_HI:
|
||||
*(ub16 *)loc = hi(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC_DTPREL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC_GOT_TLSGD16:
|
||||
*(ub16 *)loc = sym.get_tlsgd_addr(ctx) - GOT;
|
||||
break;
|
||||
case R_PPC_GOT_TLSLD16:
|
||||
*(ub16 *)loc = ctx.got->get_tlsld_addr(ctx) - GOT;
|
||||
break;
|
||||
case R_PPC_GOT_TPREL16:
|
||||
*(ub16 *)loc = sym.get_gottp_addr(ctx) - GOT;
|
||||
break;
|
||||
case R_PPC_ADDR32:
|
||||
case R_PPC_UADDR32:
|
||||
case R_PPC_TLS:
|
||||
case R_PPC_TLSGD:
|
||||
case R_PPC_TLSLD:
|
||||
case R_PPC_PLTSEQ:
|
||||
case R_PPC_PLTCALL:
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC_ADDR32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub32 *)loc = *val;
|
||||
else
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_PPC_DTPREL32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub32 *)loc = *val;
|
||||
else
|
||||
*(ub32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC_ADDR14:
|
||||
case R_PPC_ADDR16:
|
||||
case R_PPC_UADDR16:
|
||||
case R_PPC_ADDR16_LO:
|
||||
case R_PPC_ADDR16_HI:
|
||||
case R_PPC_ADDR16_HA:
|
||||
case R_PPC_ADDR24:
|
||||
case R_PPC_ADDR30:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_PPC_REL14:
|
||||
case R_PPC_REL16:
|
||||
case R_PPC_REL16_LO:
|
||||
case R_PPC_REL16_HI:
|
||||
case R_PPC_REL16_HA:
|
||||
case R_PPC_REL32:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_PPC_GOT16:
|
||||
case R_PPC_GOT16_LO:
|
||||
case R_PPC_GOT16_HI:
|
||||
case R_PPC_GOT16_HA:
|
||||
case R_PPC_PLT16_LO:
|
||||
case R_PPC_PLT16_HI:
|
||||
case R_PPC_PLT16_HA:
|
||||
case R_PPC_PLT32:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_PPC_REL24:
|
||||
case R_PPC_PLTREL24:
|
||||
case R_PPC_PLTREL32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_PPC_GOT_TLSGD16:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_PPC_GOT_TLSLD16:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_PPC_GOT_TPREL16:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_PPC_TPREL16_LO:
|
||||
case R_PPC_TPREL16_HI:
|
||||
case R_PPC_TPREL16_HA:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_PPC_ADDR32:
|
||||
case R_PPC_UADDR32:
|
||||
case R_PPC_LOCAL24PC:
|
||||
case R_PPC_TLS:
|
||||
case R_PPC_TLSGD:
|
||||
case R_PPC_TLSLD:
|
||||
case R_PPC_DTPREL16_LO:
|
||||
case R_PPC_DTPREL16_HI:
|
||||
case R_PPC_DTPREL16_HA:
|
||||
case R_PPC_PLTSEQ:
|
||||
case R_PPC_PLTCALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Thunk<E>::copy_buf(Context<E> &ctx) {
|
||||
constexpr ub32 local_thunk[] = {
|
||||
// Get this thunk's address
|
||||
0x7c08'02a6, // mflr r0
|
||||
0x429f'0005, // bcl 20, 31, 4
|
||||
0x7d88'02a6, // mflr r12
|
||||
0x7c08'03a6, // mtlr r0
|
||||
|
||||
// Materialize the destination's address in %r11 and jump to that address
|
||||
0x3d6c'0000, // addis r11, r12, OFFSET@higha
|
||||
0x396b'0000, // addi r11, r11, OFFSET@lo
|
||||
0x7d69'03a6, // mtctr r11
|
||||
0x4e80'0420, // bctr
|
||||
0x6000'0000, // nop
|
||||
};
|
||||
|
||||
static_assert(E::thunk_size == sizeof(plt_entry));
|
||||
static_assert(E::thunk_size == sizeof(local_thunk));
|
||||
|
||||
u8 *base = ctx.buf + output_section.shdr.sh_offset + offset;
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
Symbol<E> &sym = *symbols[i];
|
||||
u64 P = get_addr() + offsets[i];
|
||||
u8 *buf = base + offsets[i];
|
||||
|
||||
if (sym.has_plt(ctx)) {
|
||||
u64 got = sym.has_got(ctx) ? sym.get_got_addr(ctx) : sym.get_gotplt_addr(ctx);
|
||||
i64 val = got - P - 8;
|
||||
memcpy(buf, plt_entry, sizeof(plt_entry));
|
||||
*(ub32 *)(buf + 16) |= higha(val);
|
||||
*(ub32 *)(buf + 20) |= lo(val);
|
||||
} else {
|
||||
i64 val = sym.get_addr(ctx) - P - 8;
|
||||
memcpy(buf, local_thunk, sizeof(local_thunk));
|
||||
*(ub32 *)(buf + 16) |= higha(val);
|
||||
*(ub32 *)(buf + 20) |= lo(val);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,673 +0,0 @@
|
||||
// This file contains code for the 64-bit PowerPC ELFv1 ABI that is
|
||||
// commonly used for big-endian PPC systems. Modern PPC systems that use
|
||||
// the processor in the little-endian mode use the ELFv2 ABI instead. For
|
||||
// ELFv2, see arch-ppc64v2.cc.
|
||||
//
|
||||
// Even though they are similiar, ELFv1 isn't only different from ELFv2 in
|
||||
// endianness. The most notable difference is, in ELFv1, a function
|
||||
// pointer doesn't directly refer to the entry point of a function but
|
||||
// instead refers to a data structure so-called "function descriptor".
|
||||
//
|
||||
// The function descriptor is essentially a pair of a function entry point
|
||||
// address and a value that should be set to %r2 before calling that
|
||||
// function. There is also a third member for "the environment pointer for
|
||||
// languages such as Pascal and PL/1" according to the psABI, but it looks
|
||||
// like no one acutally uses it. In total, the function descriptor is 24
|
||||
// bytes long. Here is why we need it.
|
||||
//
|
||||
// PPC generally lacks PC-relative data access instructions. Position-
|
||||
// independent code sets GOT + 0x8000 to %r2 and access global variables
|
||||
// relative to %r2.
|
||||
//
|
||||
// Each ELF file has its own GOT. If a function calls another function in
|
||||
// the same ELF file, it doesn't have to reset %r2. However, if it is in
|
||||
// other file (e.g. other .so), it has to set a new value to %r2 so that
|
||||
// the register contains the callee's GOT + 0x8000.
|
||||
//
|
||||
// In this way, you can't call a function just by knowing the function's
|
||||
// entry point address. You also need to know a proper %r2 value for the
|
||||
// function. This is why a function pointer refers to a tuple of an
|
||||
// address and a %r2 value.
|
||||
//
|
||||
// If a function call is made through PLT, PLT takes care of restoring %r2.
|
||||
// Therefore, the caller has to restore %r2 only for function calls
|
||||
// through function pointers.
|
||||
//
|
||||
// .opd (short for "official procedure descriptors") contains function
|
||||
// descriptors.
|
||||
//
|
||||
// You can think OPD as this: even in other targets, a function can have a
|
||||
// few different addresses for different purposes. It may not only have an
|
||||
// entry point address but may also have PLT and/or GOT addresses.
|
||||
// In PPCV1, it may have an OPD address in addition to these. OPD address
|
||||
// is used for relocations that refers to the address of a function as a
|
||||
// function pointer.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/ppc64v1.pdf
|
||||
|
||||
#if MOLD_PPC64V1
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <algorithm>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = PPC64V1;
|
||||
|
||||
static u64 lo(u64 x) { return x & 0xffff; }
|
||||
static u64 hi(u64 x) { return x >> 16; }
|
||||
static u64 ha(u64 x) { return (x + 0x8000) >> 16; }
|
||||
static u64 high(u64 x) { return (x >> 16) & 0xffff; }
|
||||
static u64 higha(u64 x) { return ((x + 0x8000) >> 16) & 0xffff; }
|
||||
|
||||
// .plt is used only for lazy symbol resolution on PPC64. All PLT
|
||||
// calls are made via range extension thunks even if they are within
|
||||
// reach. Thunks read addresses from .got.plt and jump there.
|
||||
// Therefore, once PLT symbols are resolved and final addresses are
|
||||
// written to .got.plt, thunks just skip .plt and directly jump to the
|
||||
// resolved addresses.
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
constexpr ub32 insn[] = {
|
||||
0x7d88'02a6, // mflr r12
|
||||
0x429f'0005, // bcl 20, 31, 4 // obtain PC
|
||||
0x7d68'02a6, // mflr r11
|
||||
0x7d88'03a6, // mtlr r12
|
||||
0x3d6b'0000, // addis r11, r11, GOTPLT_OFFSET@ha
|
||||
0x396b'0000, // addi r11, r11, GOTPLT_OFFSET@lo
|
||||
0xe98b'0000, // ld r12,0(r11)
|
||||
0xe84b'0008, // ld r2,8(r11)
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0xe96b'0010, // ld r11,16(r11)
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_hdr_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
|
||||
i64 val = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 8;
|
||||
*(ub32 *)(buf + 16) |= higha(val);
|
||||
*(ub32 *)(buf + 20) |= lo(val);
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
ub32 *loc = (ub32 *)buf;
|
||||
i64 idx = sym.get_plt_idx(ctx);
|
||||
|
||||
// The PPC64 ELFv1 ABI requires PLT entries to be vary in size depending
|
||||
// on their indices. Unlike other targets, .got.plt is filled not by us
|
||||
// but by the loader, so we don't have a control over where the initial
|
||||
// call to the PLT entry jumps to. So we need to strictly follow the PLT
|
||||
// section layout as the loader expect it to be.
|
||||
if (idx < 0x8000) {
|
||||
constexpr ub32 insn[] = {
|
||||
0x3800'0000, // li r0, PLT_INDEX
|
||||
0x4b00'0000, // b plt0
|
||||
};
|
||||
|
||||
memcpy(loc, insn, sizeof(insn));
|
||||
loc[0] |= idx;
|
||||
loc[1] |= (ctx.plt->shdr.sh_addr - sym.get_plt_addr(ctx) - 4) & 0x00ff'ffff;
|
||||
} else {
|
||||
constexpr ub32 insn[] = {
|
||||
0x3c00'0000, // lis r0, PLT_INDEX@high
|
||||
0x6000'0000, // ori r0, r0, PLT_INDEX@lo
|
||||
0x4b00'0000, // b plt0
|
||||
};
|
||||
|
||||
memcpy(loc, insn, sizeof(insn));
|
||||
loc[0] |= high(idx);
|
||||
loc[1] |= lo(idx);
|
||||
loc[2] |= (ctx.plt->shdr.sh_addr - sym.get_plt_addr(ctx) - 8) & 0x00ff'ffff;
|
||||
}
|
||||
}
|
||||
|
||||
// .plt.got is not necessary on PPC64 because range extension thunks
|
||||
// directly read GOT entries and jump there.
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
*(ub64 *)loc = val;
|
||||
break;
|
||||
case R_PPC64_REL32:
|
||||
*(ub32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
case R_PPC64_REL64:
|
||||
*(ub64 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
u64 TOC = ctx.extra.TOC->value;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_TOC:
|
||||
break;
|
||||
case R_PPC64_TOC16_HA:
|
||||
*(ub16 *)loc = ha(S + A - TOC);
|
||||
break;
|
||||
case R_PPC64_TOC16_LO:
|
||||
*(ub16 *)loc = lo(S + A - TOC);
|
||||
break;
|
||||
case R_PPC64_TOC16_DS:
|
||||
check(S + A - TOC, -(1 << 15), 1 << 15);
|
||||
*(ub16 *)loc |= (S + A - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_TOC16_LO_DS:
|
||||
*(ub16 *)loc |= (S + A - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_REL24: {
|
||||
i64 val = sym.get_addr(ctx, NO_OPD) + A - P;
|
||||
if (sym.has_plt(ctx) || !is_int(val, 26))
|
||||
val = sym.get_thunk_addr(ctx, P) + A - P;
|
||||
|
||||
check(val, -(1 << 25), 1 << 25);
|
||||
*(ub32 *)loc |= bits(val, 25, 2) << 2;
|
||||
|
||||
// If a callee is an external function, PLT saves %r2 to the
|
||||
// caller's r2 save slot. We need to restore it after function
|
||||
// return. To do so, there's usually a NOP as a placeholder
|
||||
// after a BL. 0x6000'0000 is a NOP.
|
||||
if (sym.has_plt(ctx) && *(ub32 *)(loc + 4) == 0x6000'0000)
|
||||
*(ub32 *)(loc + 4) = 0xe841'0028; // ld r2, 40(r1)
|
||||
break;
|
||||
}
|
||||
case R_PPC64_REL32:
|
||||
*(ub32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_PPC64_REL64:
|
||||
*(ub64 *)loc = S + A - P;
|
||||
break;
|
||||
case R_PPC64_REL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - P);
|
||||
break;
|
||||
case R_PPC64_REL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - P);
|
||||
break;
|
||||
case R_PPC64_GOT16:
|
||||
*(ub16 *)loc = G - TOC;
|
||||
break;
|
||||
case R_PPC64_PLT16_HA:
|
||||
*(ub16 *)loc = ha(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_HI:
|
||||
*(ub16 *)loc = hi(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_LO:
|
||||
*(ub16 *)loc = lo(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_LO_DS:
|
||||
*(ub16 *)loc |= (G + GOT - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_GOT_TPREL16_HA:
|
||||
*(ub16 *)loc = ha(sym.get_gottp_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_HA:
|
||||
*(ub16 *)loc = ha(sym.get_tlsgd_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_LO:
|
||||
*(ub16 *)loc = lo(sym.get_tlsgd_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_HA:
|
||||
*(ub16 *)loc = ha(ctx.got->get_tlsld_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_LO:
|
||||
*(ub16 *)loc = lo(ctx.got->get_tlsld_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_LO_DS:
|
||||
*(ub16 *)loc |= (S + A - ctx.dtp_addr) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_TPREL16_HA:
|
||||
*(ub16 *)loc = ha(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC64_TPREL16_LO:
|
||||
*(ub16 *)loc = lo(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC64_TPREL16_LO_DS:
|
||||
*(ub16 *)loc |= (S + A - ctx.tp_addr) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_GOT_TPREL16_LO_DS:
|
||||
*(ub16 *)loc |= (sym.get_gottp_addr(ctx) - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
case R_PPC64_PLTSEQ:
|
||||
case R_PPC64_PLTCALL:
|
||||
case R_PPC64_TLS:
|
||||
case R_PPC64_TLSGD:
|
||||
case R_PPC64_TLSLD:
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, val, i, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_ADDR64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub64 *)loc = *val;
|
||||
else
|
||||
*(ub64 *)loc = S + A;
|
||||
break;
|
||||
case R_PPC64_ADDR32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_PPC64_DTPREL64:
|
||||
*(ub64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT | NEEDS_PPC_OPD;
|
||||
|
||||
// Any relocation except R_PPC64_REL24 is considered as an
|
||||
// address-taking relocation.
|
||||
if (rel.r_type != R_PPC64_REL24 && sym.get_type() == STT_FUNC)
|
||||
sym.flags |= NEEDS_PPC_OPD;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_GOT_TPREL16_HA:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_PPC64_REL24:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_PPC64_GOT16:
|
||||
case R_PPC64_PLT16_HA:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_HA:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_HA:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_PPC64_TPREL16_HA:
|
||||
case R_PPC64_TPREL16_LO:
|
||||
case R_PPC64_TPREL16_LO_DS:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
case R_PPC64_TOC:
|
||||
case R_PPC64_REL32:
|
||||
case R_PPC64_REL64:
|
||||
case R_PPC64_TOC16_HA:
|
||||
case R_PPC64_TOC16_LO:
|
||||
case R_PPC64_TOC16_LO_DS:
|
||||
case R_PPC64_TOC16_DS:
|
||||
case R_PPC64_REL16_HA:
|
||||
case R_PPC64_REL16_LO:
|
||||
case R_PPC64_PLT16_HI:
|
||||
case R_PPC64_PLT16_LO:
|
||||
case R_PPC64_PLT16_LO_DS:
|
||||
case R_PPC64_PLTSEQ:
|
||||
case R_PPC64_PLTCALL:
|
||||
case R_PPC64_GOT_TPREL16_LO_DS:
|
||||
case R_PPC64_GOT_TLSGD16_LO:
|
||||
case R_PPC64_GOT_TLSLD16_LO:
|
||||
case R_PPC64_TLS:
|
||||
case R_PPC64_TLSGD:
|
||||
case R_PPC64_TLSLD:
|
||||
case R_PPC64_DTPREL16_HA:
|
||||
case R_PPC64_DTPREL16_LO:
|
||||
case R_PPC64_DTPREL16_LO_DS:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Thunk<E>::copy_buf(Context<E> &ctx) {
|
||||
// If the destination is .plt.got, we save the current r2, read an
|
||||
// address of a function descriptor from .got, restore %r2 and jump
|
||||
// to the function.
|
||||
constexpr ub32 pltgot_thunk[] = {
|
||||
// Store the caller's %r2
|
||||
0xf841'0028, // std %r2, 40(%r1)
|
||||
|
||||
// Load an address of a function descriptor
|
||||
0x3d82'0000, // addis %r12, %r2, foo@got@toc@ha
|
||||
0xe98c'0000, // ld %r12, foo@got@toc@lo(%r12)
|
||||
|
||||
// Restore the callee's %r2
|
||||
0xe84c'0008, // ld %r2, 8(%r12)
|
||||
|
||||
// Jump to the function
|
||||
0xe98c'0000, // ld %r12, 0(%r12)
|
||||
0x7d89'03a6, // mtctr %r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
// If the destination is .plt, read a function descriptor from .got.plt.
|
||||
constexpr ub32 plt_thunk[] = {
|
||||
// Store the caller's %r2
|
||||
0xf841'0028, // std %r2, 40(%r1)
|
||||
|
||||
// Materialize an address of a function descriptor
|
||||
0x3d82'0000, // addis %r12, %r2, foo@gotplt@toc@ha
|
||||
0x398c'0000, // addi %r12, %r12, foo@gotplt@toc@lo
|
||||
|
||||
// Restore the callee's %r2
|
||||
0xe84c'0008, // ld %r2, 8(%r12)
|
||||
|
||||
// Jump to the function
|
||||
0xe98c'0000, // ld %r12, 0(%r12)
|
||||
0x7d89'03a6, // mtctr %r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
// If the destination is a non-imported function, we directly jump
|
||||
// to the function entry address.
|
||||
constexpr ub32 local_thunk[] = {
|
||||
0x3d82'0000, // addis r12, r2, foo@toc@ha
|
||||
0x398c'0000, // addi r12, r12, foo@toc@lo
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
0x6000'0000, // nop
|
||||
0x6000'0000, // nop
|
||||
0x6000'0000, // nop
|
||||
};
|
||||
|
||||
static_assert(E::thunk_size == sizeof(pltgot_thunk));
|
||||
static_assert(E::thunk_size == sizeof(plt_thunk));
|
||||
static_assert(E::thunk_size == sizeof(local_thunk));
|
||||
|
||||
u8 *base = ctx.buf + output_section.shdr.sh_offset + offset;
|
||||
u64 TOC = ctx.extra.TOC->value;
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
Symbol<E> &sym = *symbols[i];
|
||||
u8 *buf = base + offsets[i];
|
||||
|
||||
if (sym.has_got(ctx)) {
|
||||
memcpy(buf, pltgot_thunk, sizeof(pltgot_thunk));
|
||||
i64 val = sym.get_got_addr(ctx) - TOC;
|
||||
*(ub32 *)(buf + 4) |= higha(val);
|
||||
*(ub32 *)(buf + 8) |= lo(val);
|
||||
} else if(sym.has_plt(ctx)) {
|
||||
memcpy(buf, plt_thunk, sizeof(plt_thunk));
|
||||
i64 val = sym.get_gotplt_addr(ctx) - TOC;
|
||||
*(ub32 *)(buf + 4) |= higha(val);
|
||||
*(ub32 *)(buf + 8) |= lo(val);
|
||||
} else {
|
||||
memcpy(buf, local_thunk, sizeof(local_thunk));
|
||||
i64 val = sym.get_addr(ctx, NO_OPD) - TOC;
|
||||
*(ub32 *)buf |= higha(val);
|
||||
*(ub32 *)(buf + 4) |= lo(val);
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
static InputSection<E> *get_opd_section(ObjectFile<E> &file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file.sections)
|
||||
if (isec && isec->name == ".opd")
|
||||
return isec.get();
|
||||
return nullptr;
|
||||
}
|
||||
|
||||
static ElfRel<E> *
|
||||
get_relocation_at(Context<E> &ctx, InputSection<E> &isec, u64 offset) {
|
||||
std::span<ElfRel<E>> rels = isec.get_rels(ctx);
|
||||
auto it = ranges::lower_bound(rels, offset, {}, [](const ElfRel<E> &r) {
|
||||
return (u64)r.r_offset;
|
||||
});
|
||||
if (it == rels.end() || it->r_offset != offset)
|
||||
return nullptr;
|
||||
return &*it;
|
||||
}
|
||||
|
||||
namespace {
|
||||
struct OpdSymbol {
|
||||
u64 r_offset = 0;
|
||||
Symbol<E> *sym = nullptr;
|
||||
};
|
||||
}
|
||||
|
||||
static Symbol<E> *
|
||||
get_opd_sym_at(std::span<OpdSymbol> syms, u64 offset) {
|
||||
auto it = ranges::lower_bound(syms, offset, {}, &OpdSymbol::r_offset);
|
||||
if (it == syms.end() || it->r_offset != offset)
|
||||
return nullptr;
|
||||
return it->sym;
|
||||
}
|
||||
|
||||
// Compiler creates an .opd entry for each function symbol. The intention
|
||||
// is to make it possible to create an output .opd section just by linking
|
||||
// input .opd sections in the same manner as we do to other normal input
|
||||
// sections.
|
||||
//
|
||||
// However, in reality, .opd isn't a normal input section. It needs many
|
||||
// special treatments as follows:
|
||||
//
|
||||
// 1. A function symbol refers to not a .text but an .opd. Its address
|
||||
// works fine for address-taking relocations such as R_PPC64_ADDR64.
|
||||
// However, R_PPC64_REL24 (which is used for branch instruction) needs
|
||||
// a function's real address instead of the function's .opd address.
|
||||
// We need to read .opd contents to find out a function entry point
|
||||
// address to apply R_PPC64_REL24.
|
||||
//
|
||||
// 2. Output .opd entries are needed only for functions whose addresses
|
||||
// are taken. Just copying input .opd sections to an output would
|
||||
// produces lots of dead .opd entries.
|
||||
//
|
||||
// 3. In this design, all function symbols refer to an .opd section, and
|
||||
// that doesn't work well with graph traversal optimizations such as
|
||||
// garbage collection or identical comdat folding. For example, garbage
|
||||
// collector would mark an .opd alive which in turn mark all functions
|
||||
// thatare referenced by .opd as alive, effectively keeping all
|
||||
// functions as alive.
|
||||
//
|
||||
// The problem is that the compiler creates a half-baked .opd section, and
|
||||
// the linker has to figure out what all these .opd entries and
|
||||
// relocations are trying to achieve. It's like the compiler would emit a
|
||||
// half-baked .plt section in an object file and the linker has to deal
|
||||
// with that. That's not a good design.
|
||||
//
|
||||
// So, in this function, we undo what the compiler did to .opd. We remove
|
||||
// function symbols from .opd and reattach them to their function entry
|
||||
// points. We also rewrite relocations that directly refer to an input
|
||||
// .opd section so that they refer to function symbols instead. We then
|
||||
// mark input .opd sections as dead.
|
||||
//
|
||||
// After this function, we mark symbols with the NEEDS_PPC_OPD flag if the
|
||||
// symbol needs an .opd entry. We then create an output .opd just like we
|
||||
// do for .plt or .got.
|
||||
void ppc64v1_rewrite_opd(Context<E> &ctx) {
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
InputSection<E> *opd = get_opd_section(*file);
|
||||
if (!opd)
|
||||
return;
|
||||
opd->is_alive = false;
|
||||
|
||||
// Move symbols from .opd to .text.
|
||||
std::vector<OpdSymbol> opd_syms;
|
||||
|
||||
for (Symbol<E> *sym : file->symbols) {
|
||||
if (sym->file != file || sym->get_input_section() != opd)
|
||||
continue;
|
||||
|
||||
if (u32 ty = sym->get_type(); ty != STT_FUNC && ty != STT_GNU_IFUNC)
|
||||
continue;
|
||||
|
||||
ElfRel<E> *rel = get_relocation_at(ctx, *opd, sym->value);
|
||||
if (!rel)
|
||||
Fatal(ctx) << *file << ": cannot find a relocation in .opd for "
|
||||
<< *sym << " at offset 0x" << std::hex << (u64)sym->value;
|
||||
|
||||
Symbol<E> *sym2 = file->symbols[rel->r_sym];
|
||||
if (sym2->get_type() != STT_SECTION)
|
||||
Fatal(ctx) << *file << ": bad relocation in .opd referring to " << *sym2;
|
||||
|
||||
opd_syms.push_back({sym->value, sym});
|
||||
|
||||
sym->set_input_section(sym2->get_input_section());
|
||||
sym->value = rel->r_addend;
|
||||
}
|
||||
|
||||
// Sort symbols so that get_opd_sym_at() can do binary search.
|
||||
ranges::stable_sort(opd_syms, {}, &OpdSymbol::r_offset);
|
||||
|
||||
// Rewrite relocations so that they directly refer to .opd.
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (!isec || !isec->is_alive || isec.get() == opd)
|
||||
continue;
|
||||
|
||||
for (ElfRel<E> &r : isec->get_rels(ctx)) {
|
||||
Symbol<E> &sym = *file->symbols[r.r_sym];
|
||||
if (sym.get_input_section() != opd)
|
||||
continue;
|
||||
|
||||
Symbol<E> *real_sym = get_opd_sym_at(opd_syms, r.r_addend);
|
||||
if (!real_sym)
|
||||
Fatal(ctx) << *isec << ": cannot find a symbol in .opd for " << r
|
||||
<< " at offset 0x" << std::hex << (u64)r.r_addend;
|
||||
|
||||
r.r_sym = real_sym->sym_idx;
|
||||
r.r_addend = 0;
|
||||
}
|
||||
}
|
||||
});
|
||||
}
|
||||
|
||||
// When a function is exported, the dynamic symbol for the function should
|
||||
// refers to the function's .opd entry. This function marks such symbols
|
||||
// with NEEDS_PPC_OPD.
|
||||
void ppc64v1_scan_symbols(Context<E> &ctx) {
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->symbols)
|
||||
if (sym->file == file && sym->is_exported)
|
||||
if (u32 ty = sym->get_type(); ty == STT_FUNC || ty == STT_GNU_IFUNC)
|
||||
sym->flags |= NEEDS_PPC_OPD;
|
||||
});
|
||||
|
||||
// Functions referenced by the ELF header also have to have .opd entries.
|
||||
if (!ctx.arg.entry->is_imported)
|
||||
ctx.arg.entry->flags |= NEEDS_PPC_OPD;
|
||||
|
||||
if (!ctx.arg.init->is_imported)
|
||||
ctx.arg.init->flags |= NEEDS_PPC_OPD;
|
||||
|
||||
if (!ctx.arg.fini->is_imported)
|
||||
ctx.arg.fini->flags |= NEEDS_PPC_OPD;
|
||||
}
|
||||
|
||||
void PPC64OpdSection::add_symbol(Context<E> &ctx, Symbol<E> *sym) {
|
||||
sym->set_opd_idx(ctx, symbols.size());
|
||||
symbols.push_back(sym);
|
||||
this->shdr.sh_size += ENTRY_SIZE;
|
||||
}
|
||||
|
||||
std::vector<ElfRel<E>> PPC64OpdSection::collect_dynrels(Context<E> &ctx) const {
|
||||
if (!ctx.arg.pic)
|
||||
return {};
|
||||
|
||||
std::vector<ElfRel<E>> rels;
|
||||
|
||||
for (Symbol<E> *sym : symbols) {
|
||||
u64 loc = sym->get_opd_addr(ctx);
|
||||
rels.emplace_back(loc, E::R_RELATIVE, 0, sym->get_addr(ctx, NO_PLT | NO_OPD));
|
||||
rels.emplace_back(loc + 8, E::R_RELATIVE, 0, ctx.extra.TOC->value);
|
||||
}
|
||||
return rels;
|
||||
}
|
||||
|
||||
void PPC64OpdSection::copy_buf(Context<E> &ctx) {
|
||||
ub64 *buf = (ub64 *)(ctx.buf + this->shdr.sh_offset);
|
||||
|
||||
for (Symbol<E> *sym : symbols) {
|
||||
u64 addr = sym->get_addr(ctx, NO_PLT | NO_OPD);
|
||||
*buf++ = addr;
|
||||
*buf++ = ctx.extra.TOC->value;
|
||||
*buf++ = 0;
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,682 +0,0 @@
|
||||
// This file implements the PowerPC ELFv2 ABI which was standardized in
|
||||
// 2014. Modern little-endian PowerPC systems are based on this ABI.
|
||||
// The ABI is often referred to as "ppc64le". This shouldn't be confused
|
||||
// with "ppc64" which refers to the original, big-endian PowerPC systems.
|
||||
//
|
||||
// PPC64 is a bit tricky to support because PC-relative load/store
|
||||
// instructions hadn't been available until Power10 which debuted in 2021.
|
||||
// Prior to Power10, it wasn't trivial for position-independent code (PIC)
|
||||
// to load a value from, for example, .got, as we can't do that with [PC +
|
||||
// the offset to the .got entry].
|
||||
//
|
||||
// In the following, I'll explain how PIC is supported on pre-Power10
|
||||
// systems first and then explain what has changed with Power10.
|
||||
//
|
||||
//
|
||||
// Position-independent code on Power9 or earlier:
|
||||
//
|
||||
// We can get the program counter on older PPC64 systems with the
|
||||
// following four instructions
|
||||
//
|
||||
// mflr r1 // save the current link register to r1
|
||||
// bl .+4 // branch to the next instruction as if it were a function
|
||||
// mflr r0 // copy the return address to r0
|
||||
// mtlr r1 // restore the original link register value
|
||||
//
|
||||
// , but it's too expensive to do if we do this for each load/store.
|
||||
//
|
||||
// As a workaround, most functions are compiled in such a way that r2 is
|
||||
// assumed to always contain the address of .got + 0x8000. With this, we
|
||||
// can for example load the first entry of .got with a single instruction
|
||||
// `lw r0, -0x8000(r2)`. r2 is called the TOC pointer.
|
||||
//
|
||||
// There's only one .got for each ELF module. Therefore, if a callee is in
|
||||
// the same ELF module, r2 doesn't have to be recomputed. Most function
|
||||
// calls are usually within the same ELF module, so this mechanism is
|
||||
// efficient.
|
||||
//
|
||||
// A function compiled for pre-Power10 usually has two entry points,
|
||||
// global and local. The global entry point usually 8 bytes precedes
|
||||
// the local entry point. In between is the following instructions:
|
||||
//
|
||||
// addis r2, r12, .TOC.@ha
|
||||
// addi r2, r2, .TOC.@lo + 4;
|
||||
//
|
||||
// The global entry point assumes that the address of itself is in r12,
|
||||
// and it computes its own TOC pointer from r12. It's easy to do so for
|
||||
// the callee because the offset between its .got + 0x8000 and the
|
||||
// function is known at link-time. The above code sequence then falls
|
||||
// through to the local entry point that assumes r2 is .got + 0x8000.
|
||||
//
|
||||
// So, if a callee's TOC pointer is different from the current one
|
||||
// (e.g. calling a function in another .so), we first load the callee's
|
||||
// address to r12 (e.g. from .got.plt with a r2-relative load) and branch
|
||||
// to that address. Then the callee computes its own TOC pointer using
|
||||
// r12.
|
||||
//
|
||||
//
|
||||
// Position-independent code on Power10:
|
||||
//
|
||||
// Power10 added 8-bytes-long instructions to the ISA. Some of them are
|
||||
// PC-relative load/store instructions that take 34 bits offsets.
|
||||
// Functions compiled with `-mcpu=power10` use these instructions for PIC.
|
||||
// r2 does not have a special meaning in such fucntions.
|
||||
//
|
||||
// When a fucntion compiled for Power10 calls a function that uses the TOC
|
||||
// pointer, we need to compute a correct value for TOC and set it to r2
|
||||
// before transferring the control to the callee. Thunks are responsible
|
||||
// for doing it.
|
||||
//
|
||||
// `_NOTOC` relocations such as `R_PPC64_REL24_NOTOC` indicate that the
|
||||
// callee does not use TOC (i.e. compiled with `-mcpu=power10`). If a
|
||||
// function using TOC is referenced via a `_NOTOC` relocation, that call
|
||||
// is made through a range extension thunk.
|
||||
//
|
||||
//
|
||||
// Note on section names: the PPC64 psABI uses a weird naming convention
|
||||
// which calls .got.plt .plt. We ignored that part because it's just
|
||||
// confusing. Since the runtime only cares about segments, we should be
|
||||
// able to name sections whatever we want.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/ppc64v2.pdf
|
||||
|
||||
#if MOLD_PPC64V2
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = PPC64V2;
|
||||
|
||||
static u64 lo(u64 x) { return x & 0xffff; }
|
||||
static u64 hi(u64 x) { return x >> 16; }
|
||||
static u64 ha(u64 x) { return (x + 0x8000) >> 16; }
|
||||
static u64 high(u64 x) { return (x >> 16) & 0xffff; }
|
||||
static u64 higha(u64 x) { return ((x + 0x8000) >> 16) & 0xffff; }
|
||||
|
||||
static void write34(u8 *loc, u64 x) {
|
||||
ul32 *buf = (ul32 *)loc;
|
||||
buf[0] = (buf[0] & 0xfffc'0000) | bits(x, 33, 16);
|
||||
buf[1] = (buf[1] & 0xffff'0000) | bits(x, 15, 0);
|
||||
}
|
||||
|
||||
// .plt is used only for lazy symbol resolution on PPC64. All PLT
|
||||
// calls are made via range extension thunks even if they are within
|
||||
// reach. Thunks read addresses from .got.plt and jump there.
|
||||
// Therefore, once PLT symbols are resolved and final addresses are
|
||||
// written to .got.plt, thunks just skip .plt and directly jump to the
|
||||
// resolved addresses.
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
constexpr ul32 insn[] = {
|
||||
// Get PC
|
||||
0x7c08'02a6, // mflr r0
|
||||
0x429f'0005, // bcl 20, 31, 4 // obtain PC
|
||||
0x7d68'02a6, // mflr r11
|
||||
0x7c08'03a6, // mtlr r0
|
||||
|
||||
// Compute the PLT entry index
|
||||
0x398c'ffd4, // addi r12, r12, -44
|
||||
0x7c0b'6050, // subf r0, r11, r12
|
||||
0x7800'f082, // rldicl r0, r0, 62, 2
|
||||
|
||||
// Compute the address of .got.plt
|
||||
0x3d6b'0000, // addis r11, r11, GOTPLT_OFFSET@ha
|
||||
0x396b'0000, // addi r11, r11, GOTPLT_OFFSET@lo
|
||||
|
||||
// Load .got.plt[0] and .got.plt[1] and branch to .got.plt[0]
|
||||
0xe98b'0000, // ld r12, 0(r11)
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0xe96b'0008, // ld r11, 8(r11)
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
|
||||
i64 val = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 8;
|
||||
*(ul32 *)(buf + 28) |= higha(val);
|
||||
*(ul32 *)(buf + 32) |= lo(val);
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
// When the control is transferred to a PLT entry, the PLT entry's
|
||||
// address is already set to %r12 by the caller.
|
||||
i64 offset = ctx.plt->shdr.sh_addr - sym.get_plt_addr(ctx);
|
||||
*(ul32 *)buf = 0x4b00'0000 | (offset & 0x00ff'ffff); // b plt0
|
||||
}
|
||||
|
||||
// .plt.got is not necessary on PPC64 because range extension thunks
|
||||
// directly read GOT entries and jump there.
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
*(ul64 *)loc = val;
|
||||
break;
|
||||
case R_PPC64_REL32:
|
||||
*(ul32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
case R_PPC64_REL64:
|
||||
*(ul64 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
static u64 get_local_entry_offset(Context<E> &ctx, Symbol<E> &sym) {
|
||||
i64 val = sym.esym().ppc64_local_entry;
|
||||
assert(val <= 7);
|
||||
if (val == 7)
|
||||
Fatal(ctx) << sym << ": local entry offset 7 is reserved";
|
||||
|
||||
if (val == 0 || val == 1)
|
||||
return 0;
|
||||
return 1 << val;
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
u64 TOC = ctx.extra.TOC->value;
|
||||
|
||||
auto r2save_thunk_addr = [&] { return sym.get_thunk_addr(ctx, P); };
|
||||
auto no_r2save_thunk_addr = [&] { return sym.get_thunk_addr(ctx, P) + 8; };
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_TOC16_HA:
|
||||
*(ul16 *)loc = ha(S + A - TOC);
|
||||
break;
|
||||
case R_PPC64_TOC16_LO:
|
||||
*(ul16 *)loc = lo(S + A - TOC);
|
||||
break;
|
||||
case R_PPC64_TOC16_DS:
|
||||
case R_PPC64_TOC16_LO_DS:
|
||||
*(ul16 *)loc |= (S + A - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_REL24:
|
||||
if (sym.has_plt(ctx) || !sym.esym().ppc64_preserves_r2()) {
|
||||
i64 val = r2save_thunk_addr() + A - P;
|
||||
*(ul32 *)loc |= bits(val, 25, 2) << 2;
|
||||
|
||||
// The thunk saves %r2 to the caller's r2 save slot. We need to
|
||||
// restore it after function return. To do so, there's usually a
|
||||
// NOP as a placeholder after a BL. 0x6000'0000 is a NOP.
|
||||
if (*(ul32 *)(loc + 4) == 0x6000'0000)
|
||||
*(ul32 *)(loc + 4) = 0xe841'0018; // ld r2, 24(r1)
|
||||
} else {
|
||||
i64 val = S + get_local_entry_offset(ctx, sym) + A - P;
|
||||
if (!is_int(val, 26))
|
||||
val = no_r2save_thunk_addr() + A - P;
|
||||
*(ul32 *)loc |= bits(val, 25, 2) << 2;
|
||||
}
|
||||
break;
|
||||
case R_PPC64_REL24_NOTOC:
|
||||
if (sym.has_plt(ctx) || sym.esym().ppc64_uses_toc()) {
|
||||
i64 val = no_r2save_thunk_addr() + A - P;
|
||||
*(ul32 *)loc |= bits(val, 25, 2) << 2;
|
||||
} else {
|
||||
i64 val = S + A - P;
|
||||
if (!is_int(val, 26))
|
||||
val = no_r2save_thunk_addr() + A - P;
|
||||
*(ul32 *)loc |= bits(val, 25, 2) << 2;
|
||||
}
|
||||
break;
|
||||
case R_PPC64_REL14:
|
||||
*(ul32 *)loc |= bits(S + A - P, 15, 2) << 2;
|
||||
break;
|
||||
case R_PPC64_REL32:
|
||||
*(ul32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_PPC64_REL64:
|
||||
*(ul64 *)loc = S + A - P;
|
||||
break;
|
||||
case R_PPC64_REL16_HA:
|
||||
*(ul16 *)loc = ha(S + A - P);
|
||||
break;
|
||||
case R_PPC64_REL16_LO:
|
||||
*(ul16 *)loc = lo(S + A - P);
|
||||
break;
|
||||
case R_PPC64_GOT16:
|
||||
*(ul16 *)loc = G - TOC;
|
||||
break;
|
||||
case R_PPC64_GOT16_LO:
|
||||
*(ul16 *)loc = lo(G - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT16_HI:
|
||||
*(ul16 *)loc = hi(G - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT16_HA:
|
||||
*(ul16 *)loc = ha(G - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_HA:
|
||||
*(ul16 *)loc = ha(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_HI:
|
||||
*(ul16 *)loc = hi(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_LO:
|
||||
*(ul16 *)loc = lo(G + GOT - TOC);
|
||||
break;
|
||||
case R_PPC64_PLT16_LO_DS:
|
||||
*(ul16 *)loc |= (G + GOT - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_PLT_PCREL34:
|
||||
case R_PPC64_PLT_PCREL34_NOTOC:
|
||||
case R_PPC64_GOT_PCREL34:
|
||||
write34(loc, G + GOT - P);
|
||||
break;
|
||||
case R_PPC64_PCREL34:
|
||||
write34(loc, S + A - P);
|
||||
break;
|
||||
case R_PPC64_GOT_TPREL16_HA:
|
||||
*(ul16 *)loc = ha(sym.get_gottp_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TPREL16_LO_DS:
|
||||
*(ul16 *)loc |= (sym.get_gottp_addr(ctx) - TOC) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_GOT_TPREL_PCREL34:
|
||||
write34(loc, sym.get_gottp_addr(ctx) - P);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_HA:
|
||||
*(ul16 *)loc = ha(sym.get_tlsgd_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_LO:
|
||||
*(ul16 *)loc = lo(sym.get_tlsgd_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD_PCREL34:
|
||||
write34(loc, sym.get_tlsgd_addr(ctx) - P);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_HA:
|
||||
*(ul16 *)loc = ha(ctx.got->get_tlsld_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_LO:
|
||||
*(ul16 *)loc = lo(ctx.got->get_tlsld_addr(ctx) - TOC);
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD_PCREL34:
|
||||
write34(loc, ctx.got->get_tlsld_addr(ctx) - P);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_HA:
|
||||
*(ul16 *)loc = ha(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_LO:
|
||||
*(ul16 *)loc = lo(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC64_DTPREL16_LO_DS:
|
||||
*(ul16 *)loc |= (S + A - ctx.dtp_addr) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_DTPREL34:
|
||||
write34(loc, S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_PPC64_TPREL16_HA:
|
||||
*(ul16 *)loc = ha(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC64_TPREL16_LO:
|
||||
*(ul16 *)loc = lo(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC64_TPREL16_LO_DS:
|
||||
*(ul16 *)loc |= (S + A - ctx.tp_addr) & 0xfffc;
|
||||
break;
|
||||
case R_PPC64_TPREL34:
|
||||
write34(loc, S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
case R_PPC64_PLTSEQ:
|
||||
case R_PPC64_PLTSEQ_NOTOC:
|
||||
case R_PPC64_PLTCALL:
|
||||
case R_PPC64_PLTCALL_NOTOC:
|
||||
case R_PPC64_TLS:
|
||||
case R_PPC64_TLSGD:
|
||||
case R_PPC64_TLSLD:
|
||||
case R_PPC64_ENTRY:
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_ADDR64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul64 *)loc = *val;
|
||||
else
|
||||
*(ul64 *)loc = S + A;
|
||||
break;
|
||||
case R_PPC64_ADDR32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ul32 *)loc = S + A;
|
||||
break;
|
||||
case R_PPC64_DTPREL64:
|
||||
*(ul64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_PPC64_GOT_TPREL16_HA:
|
||||
case R_PPC64_GOT_TPREL_PCREL34:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_PPC64_REL24:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_PPC64_REL24_NOTOC:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
ctx.extra.is_power10 = true;
|
||||
break;
|
||||
case R_PPC64_GOT16:
|
||||
case R_PPC64_GOT16_LO:
|
||||
case R_PPC64_GOT16_HI:
|
||||
case R_PPC64_GOT16_HA:
|
||||
case R_PPC64_PLT16_HA:
|
||||
case R_PPC64_PLT_PCREL34:
|
||||
case R_PPC64_PLT_PCREL34_NOTOC:
|
||||
case R_PPC64_GOT_PCREL34:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_PPC64_GOT_TLSGD16_HA:
|
||||
case R_PPC64_GOT_TLSGD_PCREL34:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_PPC64_GOT_TLSLD16_HA:
|
||||
case R_PPC64_GOT_TLSLD_PCREL34:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_PPC64_TPREL16_HA:
|
||||
case R_PPC64_TPREL34:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_PPC64_ADDR64:
|
||||
case R_PPC64_REL14:
|
||||
case R_PPC64_REL32:
|
||||
case R_PPC64_REL64:
|
||||
case R_PPC64_TOC16_HA:
|
||||
case R_PPC64_TOC16_LO:
|
||||
case R_PPC64_TOC16_LO_DS:
|
||||
case R_PPC64_TOC16_DS:
|
||||
case R_PPC64_REL16_HA:
|
||||
case R_PPC64_REL16_LO:
|
||||
case R_PPC64_PLT16_HI:
|
||||
case R_PPC64_PLT16_LO:
|
||||
case R_PPC64_PLT16_LO_DS:
|
||||
case R_PPC64_PCREL34:
|
||||
case R_PPC64_PLTSEQ:
|
||||
case R_PPC64_PLTSEQ_NOTOC:
|
||||
case R_PPC64_PLTCALL:
|
||||
case R_PPC64_PLTCALL_NOTOC:
|
||||
case R_PPC64_GOT_TPREL16_LO_DS:
|
||||
case R_PPC64_GOT_TLSGD16_LO:
|
||||
case R_PPC64_GOT_TLSLD16_LO:
|
||||
case R_PPC64_TPREL16_LO:
|
||||
case R_PPC64_TPREL16_LO_DS:
|
||||
case R_PPC64_TLS:
|
||||
case R_PPC64_TLSGD:
|
||||
case R_PPC64_TLSLD:
|
||||
case R_PPC64_DTPREL16_HA:
|
||||
case R_PPC64_DTPREL16_LO:
|
||||
case R_PPC64_DTPREL16_LO_DS:
|
||||
case R_PPC64_DTPREL34:
|
||||
case R_PPC64_ENTRY:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void Thunk<E>::copy_buf(Context<E> &ctx) {
|
||||
// If the destination is PLT, we read an address from .got.plt or .got
|
||||
// and jump there.
|
||||
constexpr ul32 plt_thunk[] = {
|
||||
0xf841'0018, // std r2, 24(r1)
|
||||
0x6000'0000, // nop
|
||||
0x3d82'0000, // addis r12, r2, foo@gotplt@toc@ha
|
||||
0xe98c'0000, // ld r12, foo@gotplt@toc@lo(r12)
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
constexpr ul32 plt_thunk_power10[] = {
|
||||
0xf841'0018, // std r2, 24(r1)
|
||||
0x6000'0000, // nop
|
||||
0x0410'0000, // pld r12, foo@gotplt@pcrel
|
||||
0xe580'0000,
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
// If the destination is a non-imported function, we directly jump
|
||||
// to its local entry point.
|
||||
constexpr ul32 local_thunk[] = {
|
||||
0xf841'0018, // std r2, 24(r1)
|
||||
0x6000'0000, // nop
|
||||
0x3d82'0000, // addis r12, r2, foo@toc@ha
|
||||
0x398c'0000, // addi r12, r12, foo@toc@lo
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
constexpr ul32 local_thunk_power10[] = {
|
||||
0xf841'0018, // std r2, 24(r1)
|
||||
0x6000'0000, // nop
|
||||
0x0610'0000, // pla r12, foo@pcrel
|
||||
0x3980'0000,
|
||||
0x7d89'03a6, // mtctr r12
|
||||
0x4e80'0420, // bctr
|
||||
};
|
||||
|
||||
static_assert(E::thunk_size == sizeof(plt_thunk));
|
||||
static_assert(E::thunk_size == sizeof(plt_thunk_power10));
|
||||
static_assert(E::thunk_size == sizeof(local_thunk));
|
||||
static_assert(E::thunk_size == sizeof(local_thunk_power10));
|
||||
|
||||
u8 *base = ctx.buf + output_section.shdr.sh_offset + offset;
|
||||
u64 TOC = ctx.extra.TOC->value;
|
||||
|
||||
for (i64 i = 0; i < symbols.size(); i++) {
|
||||
Symbol<E> &sym = *symbols[i];
|
||||
u64 P = get_addr() + offsets[i];
|
||||
u8 *buf = base + offsets[i];
|
||||
|
||||
if (sym.has_plt(ctx)) {
|
||||
u64 got = sym.has_got(ctx) ? sym.get_got_addr(ctx) : sym.get_gotplt_addr(ctx);
|
||||
if (ctx.extra.is_power10) {
|
||||
memcpy(buf, plt_thunk_power10, E::thunk_size);
|
||||
write34(buf + 8, got - P - 8);
|
||||
} else {
|
||||
memcpy(buf, plt_thunk, E::thunk_size);
|
||||
*(ul32 *)(buf + 8) |= higha(got - TOC);
|
||||
*(ul32 *)(buf + 12) |= lo(got - TOC);
|
||||
}
|
||||
} else {
|
||||
u64 S = sym.get_addr(ctx);
|
||||
if (ctx.extra.is_power10) {
|
||||
memcpy(buf, local_thunk_power10, E::thunk_size);
|
||||
write34(buf + 8, S - P - 8);
|
||||
} else {
|
||||
memcpy(buf, local_thunk, E::thunk_size);
|
||||
*(ul32 *)(buf + 8) |= higha(S - TOC);
|
||||
*(ul32 *)(buf + 12) |= lo(S - TOC);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// GCC may emit references to the following functions in function prologue
|
||||
// and epiilogue if -Os is specified. For some reason, these functions are
|
||||
// not in libgcc.a and expected to be synthesized by the linker.
|
||||
const std::vector<std::pair<std::string_view, u32>>
|
||||
ppc64_save_restore_insns = {
|
||||
{ "_savegpr0_14", 0xf9c1ff70 }, // std r14,-144(r1)
|
||||
{ "_savegpr0_15", 0xf9e1ff78 }, // std r15,-136(r1)
|
||||
{ "_savegpr0_16", 0xfa01ff80 }, // std r16,-128(r1)
|
||||
{ "_savegpr0_17", 0xfa21ff88 }, // std r17,-120(r1)
|
||||
{ "_savegpr0_18", 0xfa41ff90 }, // std r18,-112(r1)
|
||||
{ "_savegpr0_19", 0xfa61ff98 }, // std r19,-104(r1)
|
||||
{ "_savegpr0_20", 0xfa81ffa0 }, // std r20,-96(r1)
|
||||
{ "_savegpr0_21", 0xfaa1ffa8 }, // std r21,-88(r1)
|
||||
{ "_savegpr0_22", 0xfac1ffb0 }, // std r22,-80(r1)
|
||||
{ "_savegpr0_23", 0xfae1ffb8 }, // std r23,-72(r1)
|
||||
{ "_savegpr0_24", 0xfb01ffc0 }, // std r24,-64(r1)
|
||||
{ "_savegpr0_25", 0xfb21ffc8 }, // std r25,-56(r1)
|
||||
{ "_savegpr0_26", 0xfb41ffd0 }, // std r26,-48(r1)
|
||||
{ "_savegpr0_27", 0xfb61ffd8 }, // std r27,-40(r1)
|
||||
{ "_savegpr0_28", 0xfb81ffe0 }, // std r28,-32(r1)
|
||||
{ "_savegpr0_29", 0xfba1ffe8 }, // std r29,-24(r1)
|
||||
{ "_savegpr0_30", 0xfbc1fff0 }, // std r30,-16(r1)
|
||||
{ "_savegpr0_31", 0xfbe1fff8 }, // std r31,-8(r1)
|
||||
{ "", 0xf8010010 }, // std r0,16(r1)
|
||||
{ "", 0x4e800020 }, // blr
|
||||
|
||||
{ "_restgpr0_14", 0xe9c1ff70 }, // ld r14,-144(r1)
|
||||
{ "_restgpr0_15", 0xe9e1ff78 }, // ld r15,-136(r1)
|
||||
{ "_restgpr0_16", 0xea01ff80 }, // ld r16,-128(r1)
|
||||
{ "_restgpr0_17", 0xea21ff88 }, // ld r17,-120(r1)
|
||||
{ "_restgpr0_18", 0xea41ff90 }, // ld r18,-112(r1)
|
||||
{ "_restgpr0_19", 0xea61ff98 }, // ld r19,-104(r1)
|
||||
{ "_restgpr0_20", 0xea81ffa0 }, // ld r20,-96(r1)
|
||||
{ "_restgpr0_21", 0xeaa1ffa8 }, // ld r21,-88(r1)
|
||||
{ "_restgpr0_22", 0xeac1ffb0 }, // ld r22,-80(r1)
|
||||
{ "_restgpr0_23", 0xeae1ffb8 }, // ld r23,-72(r1)
|
||||
{ "_restgpr0_24", 0xeb01ffc0 }, // ld r24,-64(r1)
|
||||
{ "_restgpr0_25", 0xeb21ffc8 }, // ld r25,-56(r1)
|
||||
{ "_restgpr0_26", 0xeb41ffd0 }, // ld r26,-48(r1)
|
||||
{ "_restgpr0_27", 0xeb61ffd8 }, // ld r27,-40(r1)
|
||||
{ "_restgpr0_28", 0xeb81ffe0 }, // ld r28,-32(r1)
|
||||
{ "_restgpr0_29", 0xe8010010 }, // ld r0,16(r1)
|
||||
{ "", 0xeba1ffe8 }, // ld r29,-24(r1)
|
||||
{ "", 0x7c0803a6 }, // mtlr r0
|
||||
{ "", 0xebc1fff0 }, // ld r30,-16(r1)
|
||||
{ "", 0xebe1fff8 }, // ld r31,-8(r1)
|
||||
{ "", 0x4e800020 }, // blr
|
||||
{ "_restgpr0_30", 0xebc1fff0 }, // ld r30,-16(r1)
|
||||
{ "_restgpr0_31", 0xe8010010 }, // ld r0,16(r1)
|
||||
{ "", 0xebe1fff8 }, // ld r31,-8(r1)
|
||||
{ "", 0x7c0803a6 }, // mtlr r0
|
||||
{ "", 0x4e800020 }, // blr
|
||||
|
||||
{ "_savegpr1_14", 0xf9ccff70 }, // std r14,-144(r12)
|
||||
{ "_savegpr1_15", 0xf9ecff78 }, // std r15,-136(r12)
|
||||
{ "_savegpr1_16", 0xfa0cff80 }, // std r16,-128(r12)
|
||||
{ "_savegpr1_17", 0xfa2cff88 }, // std r17,-120(r12)
|
||||
{ "_savegpr1_18", 0xfa4cff90 }, // std r18,-112(r12)
|
||||
{ "_savegpr1_19", 0xfa6cff98 }, // std r19,-104(r12)
|
||||
{ "_savegpr1_20", 0xfa8cffa0 }, // std r20,-96(r12)
|
||||
{ "_savegpr1_21", 0xfaacffa8 }, // std r21,-88(r12)
|
||||
{ "_savegpr1_22", 0xfaccffb0 }, // std r22,-80(r12)
|
||||
{ "_savegpr1_23", 0xfaecffb8 }, // std r23,-72(r12)
|
||||
{ "_savegpr1_24", 0xfb0cffc0 }, // std r24,-64(r12)
|
||||
{ "_savegpr1_25", 0xfb2cffc8 }, // std r25,-56(r12)
|
||||
{ "_savegpr1_26", 0xfb4cffd0 }, // std r26,-48(r12)
|
||||
{ "_savegpr1_27", 0xfb6cffd8 }, // std r27,-40(r12)
|
||||
{ "_savegpr1_28", 0xfb8cffe0 }, // std r28,-32(r12)
|
||||
{ "_savegpr1_29", 0xfbacffe8 }, // std r29,-24(r12)
|
||||
{ "_savegpr1_30", 0xfbccfff0 }, // std r30,-16(r12)
|
||||
{ "_savegpr1_31", 0xfbecfff8 }, // std r31,-8(r12)
|
||||
{ "", 0x4e800020 }, // blr
|
||||
|
||||
{ "_restgpr1_14", 0xe9ccff70 }, // ld r14,-144(r12)
|
||||
{ "_restgpr1_15", 0xe9ecff78 }, // ld r15,-136(r12)
|
||||
{ "_restgpr1_16", 0xea0cff80 }, // ld r16,-128(r12)
|
||||
{ "_restgpr1_17", 0xea2cff88 }, // ld r17,-120(r12)
|
||||
{ "_restgpr1_18", 0xea4cff90 }, // ld r18,-112(r12)
|
||||
{ "_restgpr1_19", 0xea6cff98 }, // ld r19,-104(r12)
|
||||
{ "_restgpr1_20", 0xea8cffa0 }, // ld r20,-96(r12)
|
||||
{ "_restgpr1_21", 0xeaacffa8 }, // ld r21,-88(r12)
|
||||
{ "_restgpr1_22", 0xeaccffb0 }, // ld r22,-80(r12)
|
||||
{ "_restgpr1_23", 0xeaecffb8 }, // ld r23,-72(r12)
|
||||
{ "_restgpr1_24", 0xeb0cffc0 }, // ld r24,-64(r12)
|
||||
{ "_restgpr1_25", 0xeb2cffc8 }, // ld r25,-56(r12)
|
||||
{ "_restgpr1_26", 0xeb4cffd0 }, // ld r26,-48(r12)
|
||||
{ "_restgpr1_27", 0xeb6cffd8 }, // ld r27,-40(r12)
|
||||
{ "_restgpr1_28", 0xeb8cffe0 }, // ld r28,-32(r12)
|
||||
{ "_restgpr1_29", 0xebacffe8 }, // ld r29,-24(r12)
|
||||
{ "_restgpr1_30", 0xebccfff0 }, // ld r30,-16(r12)
|
||||
{ "_restgpr1_31", 0xebecfff8 }, // ld r31,-8(r12)
|
||||
{ "", 0x4e800020 }, // blr
|
||||
};
|
||||
|
||||
void PPC64SaveRestoreSection::copy_buf(Context<E> &ctx) {
|
||||
ul32 *buf = (ul32 *)(ctx.buf + this->shdr.sh_offset);
|
||||
for (auto [label, insn] : ppc64_save_restore_insns)
|
||||
*buf++ = insn;
|
||||
}
|
||||
|
||||
template <>
|
||||
u64 get_eflags(Context<E> &ctx) {
|
||||
return 2;
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
-1277
File diff suppressed because it is too large.
Load diff
@@ -1,495 +0,0 @@
|
||||
// This file contains code for the IBM z/Architecture 64-bit ISA, which is
|
||||
// commonly referred to as "s390x" on Linux.
|
||||
//
|
||||
// z/Architecture is a 64-bit CISC ISA developed by IBM around 2000 for
|
||||
// IBM's "big iron" mainframe computers. The computers are direct
|
||||
// descendents of IBM System/360 all the way back in 1966. I've never
|
||||
// actually seen a mainframe, and you probaly haven't either, but it looks
|
||||
// like the mainframe market is still large enough to sustain its ecosystem.
|
||||
// Ubuntu for example provides the official support for s390x as of 2022.
|
||||
// Since they are being actively maintained, we need to support them.
|
||||
//
|
||||
// As an instruction set, s390x isn't particularly odd. It has 16 general-
|
||||
// purpose registers. Instructions are 2, 4 or 6 bytes long and always
|
||||
// aligned to 2 bytes boundaries. Despite unfamiliarty, I found that it
|
||||
// just feels like an x86-64 in a parallel universe.
|
||||
//
|
||||
// Here is the register usage in this ABI:
|
||||
//
|
||||
// r0-r1: reserved as scratch registers so we can use them in our PLT
|
||||
// r2: parameter passing and return values
|
||||
// r3-r6: parameter passing
|
||||
// r12: address of GOT if position-independent code
|
||||
// r14: return address
|
||||
// r15: stack pointer
|
||||
// a1: upper 32 bits of TP (thread pointer)
|
||||
// a2: lower 32 bits of TP (thread pointer)
|
||||
//
|
||||
// Thread-local storage (TLS) is supported on s390x in the same way as it
|
||||
// is on other targets with one exeption. On other targets, __tls_get_addr
|
||||
// is used to get an address of a thread-local variable. On s390x,
|
||||
// __tls_get_offset is used instead. The difference is __tls_get_offset
|
||||
// returns an address of a thread-local variable as an offset from TP. So
|
||||
// we need to add TP to a return value before use. I don't know why it is
|
||||
// different, but that is the way it is.
|
||||
//
|
||||
// https://github.com/IBM/s390x-abi/releases/download/v1.6.1/lzsabi_s390x.pdf
|
||||
|
||||
#if MOLD_S390X
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = S390X;
|
||||
|
||||
static void write_mid20(u8 *loc, u64 val) {
|
||||
*(ub32 *)loc |= (bits(val, 11, 0) << 16) | (bits(val, 19, 12) << 8);
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
static u8 insn[] = {
|
||||
// Compute the offset into .rela.plt. This is equivalent to
|
||||
// (%r0 - %r1 - 48 - 14) * 3/2 where %r0 is the PLT entry address
|
||||
// plus 14, %r1 is the start address of .plt, and 48 is the size
|
||||
// of this PLT header. We multiply by 3/2 because each PLT entry
|
||||
// is 16 bytes, whereas each .rela.plt entry is 24 bytes.
|
||||
0xb9, 0x09, 0x00, 0x01, // sgr %r0, %r1
|
||||
0xa7, 0x0b, 0xff, 0xc2, // aghi %r0, -62
|
||||
0xeb, 0x10, 0x00, 0x01, 0x00, 0x0c, // srlg %r1, %r0, 1
|
||||
0xb9, 0x08, 0x00, 0x01, // agr %r0, %r1
|
||||
// Store the computed value to 56(%r15) and .got.plt[1] to 48(%15)
|
||||
// where %r15 is the stack pointer.
|
||||
0xe3, 0x00, 0xf0, 0x38, 0x00, 0x24, // stg %r0, 56(%r15)
|
||||
0xc0, 0x10, 0, 0, 0, 0, // larl %r1, GOTPLT_OFFSET
|
||||
0xd2, 0x07, 0xf0, 0x30, 0x10, 0x08, // mvc 48(8, %r15), 8(%r1)
|
||||
// Branch to _dl_runtime_resolve
|
||||
0xe3, 0x10, 0x10, 0x10, 0x00, 0x04, // lg %r1, 16(%r1)
|
||||
0x07, 0xf1, // br %r1
|
||||
0x00, 0x00, 0x00, 0x00, // (filler)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 26) =
|
||||
(ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 24) >> 1;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static u8 insn[] = {
|
||||
0xc0, 0x10, 0, 0, 0, 0, // larl %r1, GOTPLT_ENTRY_OFFSET
|
||||
0xe3, 0x10, 0x10, 0x00, 0x00, 0x04, // lg %r1, (%r1)
|
||||
0x0d, 0x01, // basr %r0, %r1
|
||||
0x00, 0x00, // (filler)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 2) = (sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx)) >> 1;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static u8 insn[] = {
|
||||
0xc0, 0x10, 0, 0, 0, 0, // larl %r1, GOT_ENTRY_OFFSET
|
||||
0xe3, 0x10, 0x10, 0x00, 0x00, 0x04, // lg %r1, (%r1)
|
||||
0x07, 0xf1, // br %r1
|
||||
0x00, 0x00, // (filler)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 2) = (sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx)) >> 1;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_390_PC32:
|
||||
*(ub32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
case R_390_64:
|
||||
*(ub64 *)loc = val;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto check_dbl = [&](i64 val, i64 lo, i64 hi) {
|
||||
// R_390_*DBL relocs should never refer to a symbol at an odd address
|
||||
check(val, lo, hi);
|
||||
if (val & 1)
|
||||
Error(ctx) << *this << ": misaligned symbol " << sym
|
||||
<< " for relocation " << rel;
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_390_64:
|
||||
break;
|
||||
case R_390_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_390_12:
|
||||
check(S + A, 0, 1 << 12);
|
||||
*(ul16 *)loc |= bits(S + A, 11, 0);
|
||||
break;
|
||||
case R_390_16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ub16 *)loc = S + A;
|
||||
break;
|
||||
case R_390_20:
|
||||
check(S + A, 0, 1 << 20);
|
||||
write_mid20(loc, S + A);
|
||||
break;
|
||||
case R_390_32:
|
||||
case R_390_PLT32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_390_PC12DBL:
|
||||
case R_390_PLT12DBL:
|
||||
check_dbl(S + A - P, -(1 << 12), 1 << 12);
|
||||
*(ul16 *)loc |= bits(S + A - P, 12, 1);
|
||||
break;
|
||||
case R_390_PC16:
|
||||
check(S + A - P, -(1 << 15), 1 << 15);
|
||||
*(ub16 *)loc = S + A - P;
|
||||
break;
|
||||
case R_390_PC32:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 31);
|
||||
*(ub32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_390_PC64:
|
||||
case R_390_PLT64:
|
||||
*(ub64 *)loc = S + A - P;
|
||||
break;
|
||||
case R_390_PC16DBL:
|
||||
case R_390_PLT16DBL:
|
||||
check_dbl(S + A - P, -(1 << 16), 1 << 16);
|
||||
*(ub16 *)loc = (S + A - P) >> 1;
|
||||
break;
|
||||
case R_390_PC24DBL:
|
||||
case R_390_PLT24DBL:
|
||||
check_dbl(S + A - P, -(1 << 24), 1 << 24);
|
||||
*(ub32 *)loc |= bits(S + A - P, 24, 1);
|
||||
break;
|
||||
case R_390_PC32DBL:
|
||||
case R_390_PLT32DBL:
|
||||
check_dbl(S + A - P, -(1LL << 32), 1LL << 32);
|
||||
*(ub32 *)loc = (S + A - P) >> 1;
|
||||
break;
|
||||
case R_390_GOT12:
|
||||
case R_390_GOTPLT12:
|
||||
check(G + A, 0, 1 << 12);
|
||||
*(ul16 *)loc |= bits(G + A, 11, 0);
|
||||
break;
|
||||
case R_390_GOT16:
|
||||
case R_390_GOTPLT16:
|
||||
check(G + A, 0, 1 << 16);
|
||||
*(ub16 *)loc = G + A;
|
||||
break;
|
||||
case R_390_GOT20:
|
||||
case R_390_GOTPLT20:
|
||||
check(G + A, 0, 1 << 20);
|
||||
write_mid20(loc, G + A);
|
||||
break;
|
||||
case R_390_GOT32:
|
||||
case R_390_GOTPLT32:
|
||||
check(G + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = G + A;
|
||||
break;
|
||||
case R_390_GOT64:
|
||||
case R_390_GOTPLT64:
|
||||
*(ub64 *)loc = G + A;
|
||||
break;
|
||||
case R_390_GOTOFF16:
|
||||
case R_390_PLTOFF16:
|
||||
check(S + A - GOT, -(1 << 15), 1 << 15);
|
||||
*(ub16 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_390_GOTOFF32:
|
||||
case R_390_PLTOFF32:
|
||||
check(S + A - GOT, -(1LL << 31), 1LL << 31);
|
||||
*(ub32 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_390_GOTOFF64:
|
||||
case R_390_PLTOFF64:
|
||||
*(ub64 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_390_GOTPC:
|
||||
*(ub64 *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_390_GOTPCDBL:
|
||||
check_dbl(GOT + A - P, -(1LL << 32), 1LL << 32);
|
||||
*(ub32 *)loc = (GOT + A - P) >> 1;
|
||||
break;
|
||||
case R_390_GOTENT:
|
||||
// If we can relax a GOT-loading LGRL to an address-materializing
|
||||
// LARL, do that. The format of LGRL is 0xc 0x4 <reg> 0x8 followed
|
||||
// by a 32-bit offset. LARL is 0xc 0x0 <reg> 0x0.
|
||||
if (ctx.arg.relax && sym.is_pcrel_linktime_const(ctx)) {
|
||||
u64 op = *(ub16 *)(loc - 2);
|
||||
u64 val = S + A - P;
|
||||
if ((op & 0xff0f) == 0xc408 && A == 2 && (val & 1) == 0 &&
|
||||
is_int(val, 33)) {
|
||||
*(ub16 *)(loc - 2) = 0xc000 | (op & 0x00f0);
|
||||
*(ub32 *)loc = val >> 1;
|
||||
break;
|
||||
}
|
||||
}
|
||||
check_dbl(GOT + G + A - P, -(1LL << 32), 1LL << 32);
|
||||
*(ub32 *)loc = (GOT + G + A - P) >> 1;
|
||||
break;
|
||||
case R_390_TLS_LE32:
|
||||
*(ub32 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_LE64:
|
||||
*(ub64 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_GOTIE20:
|
||||
write_mid20(loc, sym.get_gottp_addr(ctx) + A - GOT);
|
||||
break;
|
||||
case R_390_TLS_IEENT:
|
||||
*(ub32 *)loc = (sym.get_gottp_addr(ctx) + A - P) >> 1;
|
||||
break;
|
||||
case R_390_TLS_GD32:
|
||||
if (sym.has_tlsgd(ctx))
|
||||
*(ub32 *)loc = sym.get_tlsgd_addr(ctx) + A - GOT;
|
||||
else if (sym.has_gottp(ctx))
|
||||
*(ub32 *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
else
|
||||
*(ub32 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_GD64:
|
||||
if (sym.has_tlsgd(ctx))
|
||||
*(ub64 *)loc = sym.get_tlsgd_addr(ctx) + A - GOT;
|
||||
else if (sym.has_gottp(ctx))
|
||||
*(ub64 *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
else
|
||||
*(ub64 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_GDCALL:
|
||||
if (sym.has_tlsgd(ctx)) {
|
||||
// do nothing
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
// lg %r2, 0(%r2, %r12)
|
||||
static u8 insn[] = { 0xe3, 0x22, 0xc0, 0x00, 0x00, 0x04 };
|
||||
memcpy(loc, insn, sizeof(insn));
|
||||
} else {
|
||||
// nop
|
||||
static u8 insn[] = { 0xc0, 0x04, 0x00, 0x00, 0x00, 0x00 };
|
||||
memcpy(loc, insn, sizeof(insn));
|
||||
}
|
||||
break;
|
||||
case R_390_TLS_LDM32:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
*(ub32 *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOT;
|
||||
else
|
||||
*(ub32 *)loc = ctx.dtp_addr - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_LDM64:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
*(ub64 *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOT;
|
||||
else
|
||||
*(ub64 *)loc = ctx.dtp_addr - ctx.tp_addr;
|
||||
break;
|
||||
case R_390_TLS_LDCALL:
|
||||
if (!ctx.got->has_tlsld(ctx)) {
|
||||
// nop
|
||||
static u8 insn[] = { 0xc0, 0x04, 0x00, 0x00, 0x00, 0x00 };
|
||||
memcpy(loc, insn, sizeof(insn));
|
||||
}
|
||||
break;
|
||||
case R_390_TLS_LDO32:
|
||||
*(ub32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_390_TLS_LDO64:
|
||||
*(ub64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, val, i, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_390_32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_390_64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub64 *)loc = *val;
|
||||
else
|
||||
*(ub64 *)loc = S + A;
|
||||
break;
|
||||
case R_390_TLS_LDO64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub64 *)loc = *val;
|
||||
else
|
||||
*(ub64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": apply_reloc_nonalloc: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_390_8:
|
||||
case R_390_12:
|
||||
case R_390_16:
|
||||
case R_390_20:
|
||||
case R_390_32:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_390_PC12DBL:
|
||||
case R_390_PC16:
|
||||
case R_390_PC16DBL:
|
||||
case R_390_PC24DBL:
|
||||
case R_390_PC32:
|
||||
case R_390_PC32DBL:
|
||||
case R_390_PC64:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_390_GOT12:
|
||||
case R_390_GOT16:
|
||||
case R_390_GOT20:
|
||||
case R_390_GOT32:
|
||||
case R_390_GOT64:
|
||||
case R_390_GOTOFF16:
|
||||
case R_390_GOTOFF32:
|
||||
case R_390_GOTOFF64:
|
||||
case R_390_GOTPLT12:
|
||||
case R_390_GOTPLT16:
|
||||
case R_390_GOTPLT20:
|
||||
case R_390_GOTPLT32:
|
||||
case R_390_GOTPLT64:
|
||||
case R_390_GOTPC:
|
||||
case R_390_GOTPCDBL:
|
||||
case R_390_GOTENT:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_390_PLT12DBL:
|
||||
case R_390_PLT16DBL:
|
||||
case R_390_PLT24DBL:
|
||||
case R_390_PLT32:
|
||||
case R_390_PLT32DBL:
|
||||
case R_390_PLT64:
|
||||
case R_390_PLTOFF16:
|
||||
case R_390_PLTOFF32:
|
||||
case R_390_PLTOFF64:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_390_TLS_GOTIE20:
|
||||
case R_390_TLS_IEENT:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_390_TLS_GD32:
|
||||
case R_390_TLS_GD64:
|
||||
// We always want to relax calls to __tls_get_offset() in statically-
|
||||
// linked executables because __tls_get_offset() in libc.a just calls
|
||||
// abort().
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && sym.is_tprel_linktime_const(ctx))) {
|
||||
// Do nothing
|
||||
} else if (ctx.arg.relax && sym.is_tprel_runtime_const(ctx)) {
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
} else {
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
}
|
||||
break;
|
||||
case R_390_TLS_LDM32:
|
||||
case R_390_TLS_LDM64:
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && !ctx.arg.shared)) {
|
||||
// Do nothing
|
||||
} else {
|
||||
ctx.needs_tlsld = true;
|
||||
}
|
||||
break;
|
||||
case R_390_TLS_LE32:
|
||||
case R_390_TLS_LE64:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_390_64:
|
||||
case R_390_TLS_LDO32:
|
||||
case R_390_TLS_LDO64:
|
||||
case R_390_TLS_GDCALL:
|
||||
case R_390_TLS_LDCALL:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
-370
@@ -1,370 +0,0 @@
|
||||
// SH-4 (SuperH 4) is a 32-bit RISC ISA developed by Hitachi in the early
|
||||
// '90s. Some relatively powerful systems were developed with SH-4.
|
||||
// A notable example is Sega's Dreamcast game console which debuted in 1998.
|
||||
// Hitachi later spun off its semiconductor division as an independent
|
||||
// company, Renesas, and Renesas is still selling SH-4 processors for the
|
||||
// embedded market. It has never been as popular as ARM is, and its
|
||||
// popularity continues to decline though.
|
||||
//
|
||||
// SH-4's most distinctive feature compared to other RISC ISAs is that its
|
||||
// instructions are 16 bits in length instead of more common 32 bits for
|
||||
// better code density. This difference affects various aspects of its
|
||||
// instruction set as shown below:
|
||||
//
|
||||
// - SH-4 has 16 general-purpose registers (GPRs) instead of the most
|
||||
// commmon 32 GPR configuration to save one bit to specify a register.
|
||||
//
|
||||
// - Binary instructions such as ADD normally take three register in
|
||||
// RISC ISAs (e.g. x ← y ⊕ z where x, y and z are registers), but
|
||||
// SH-4's instructions take only two registers. The result of an
|
||||
// operation is written to one of the source registers (e.g. x ← x ⊕ y).
|
||||
//
|
||||
// - Usual RISC ISAs have "load high" and "load low" instructions to set
|
||||
// an immediate to most significant and least significant bits in a
|
||||
// register to construct a full 32-bit value in a register. This
|
||||
// technique is hard to use in SH-4, as 16 bit instructions are too
|
||||
// small to contain large immediates. On SH-4, large immediates are
|
||||
// loaded from memory using `mov.l` PC-relative load instruction.
|
||||
//
|
||||
// - Many RISC ISAs are, despite their name, actually fairly complex.
|
||||
// They tend to have hundreds if not thousands of different instructions.
|
||||
// SH-4 doesn't really have that many instructions because its 16-bit
|
||||
// machine code simply can't encode many different opcodes. As a
|
||||
// result, the number of relocations the linker has to support is also
|
||||
// small.
|
||||
//
|
||||
// Beside these, SH-4 has a delay branch slot just like contemporary MIPS
|
||||
// and SPARC. That is, one instruction after a branch instruction will
|
||||
// always be executed even if the branch is taken. Delay branch slot allows
|
||||
// a pipelined CPU to start and finish executing an instruction after a
|
||||
// branch regardless of the branch's condition, simplifying the processor's
|
||||
// implementation. It's considered a bad premature optimization nowadays,
|
||||
// though. Modern RISC processors don't have it.
|
||||
//
|
||||
// Here are notes about the SH-4 psABI:
|
||||
//
|
||||
// - If a source file is compiled with -fPIC, each function starts
|
||||
// with a piece of code to store the address of .got to %r12.
|
||||
// We can use the register in our PLT for position-independent output.
|
||||
//
|
||||
// - Even though it uses the RELA-type relocations, relocation addends
|
||||
// are stored not to the r_addend field but to the relocated section
|
||||
// contents for some reason. Therefore, it's effectively REL.
|
||||
//
|
||||
// - It looks like the ecosystem has bit-rotted. Some tests, especially
|
||||
// one using C++ exceptions, don't pass even with GNU ld.
|
||||
//
|
||||
// - GCC/SH4 tends to write dynamically-relocated data into .text, so the
|
||||
// output from the linker contains lots of text relocations. That's not
|
||||
// a problem with embedded programming, I guess.
|
||||
|
||||
#if MOLD_SH4LE || MOLD_SH4BE
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
// Even though SH-4 uses RELA-type relocations, addends are stored to
|
||||
// relocated places for some reason.
|
||||
template <>
|
||||
i64 get_addend(u8 *loc, const ElfRel<E> &rel) {
|
||||
switch (rel.r_type) {
|
||||
case R_SH_DIR32:
|
||||
case R_SH_REL32:
|
||||
case R_SH_TLS_GD_32:
|
||||
case R_SH_TLS_LD_32:
|
||||
case R_SH_TLS_LDO_32:
|
||||
case R_SH_TLS_IE_32:
|
||||
case R_SH_TLS_LE_32:
|
||||
case R_SH_TLS_DTPMOD32:
|
||||
case R_SH_TLS_DTPOFF32:
|
||||
case R_SH_TLS_TPOFF32:
|
||||
case R_SH_GOT32:
|
||||
case R_SH_PLT32:
|
||||
case R_SH_GOTOFF:
|
||||
case R_SH_GOTPC:
|
||||
case R_SH_GOTPLT32:
|
||||
return *(U32<E> *)loc;
|
||||
default:
|
||||
return 0;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_addend(u8 *loc, i64 val, const ElfRel<E> &rel) {
|
||||
switch (rel.r_type) {
|
||||
case R_SH_DIR32:
|
||||
case R_SH_REL32:
|
||||
case R_SH_TLS_GD_32:
|
||||
case R_SH_TLS_LD_32:
|
||||
case R_SH_TLS_LDO_32:
|
||||
case R_SH_TLS_IE_32:
|
||||
case R_SH_TLS_LE_32:
|
||||
case R_SH_TLS_DTPMOD32:
|
||||
case R_SH_TLS_DTPOFF32:
|
||||
case R_SH_TLS_TPOFF32:
|
||||
case R_SH_GOT32:
|
||||
case R_SH_PLT32:
|
||||
case R_SH_GOTOFF:
|
||||
case R_SH_GOTPC:
|
||||
case R_SH_GOTPLT32:
|
||||
*(U32<E> *)loc = val;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
if (ctx.arg.pic) {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd202, // mov.l 1f, r2
|
||||
0x32cc, // add r12, r2
|
||||
0x5022, // mov.l @(8, r2), r0
|
||||
0x5221, // mov.l @(4, r2), r2
|
||||
0x402b, // jmp @r0
|
||||
0xe000, // mov #0, r0
|
||||
0, 0, // 1: .long GOTPLT
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_hdr_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 12) = ctx.gotplt->shdr.sh_addr - ctx.got->shdr.sh_addr;
|
||||
} else {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd202, // mov.l 1f, r2
|
||||
0x5022, // mov.l @(8, r2), r0
|
||||
0x5221, // mov.l @(4, r2), r2
|
||||
0x402b, // jmp @r0
|
||||
0xe000, // mov #0, r0
|
||||
0xfffd, // (illegal)
|
||||
0, 0, // 1: .long GOTPLT
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_hdr_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 12) = ctx.gotplt->shdr.sh_addr;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
if (ctx.arg.pic) {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd001, // mov.l 1f, r0
|
||||
0x00ce, // mov.l @(r0, r12), r0
|
||||
0x402b, // jmp @r0
|
||||
0xd101, // mov.l 2f, r1
|
||||
0, 0, // 1: .long GOTPLT_ENTRY
|
||||
0, 0, // 2: .long INDEX_IN_RELPLT
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 8) = sym.get_gotplt_addr(ctx) - ctx.got->shdr.sh_addr;
|
||||
*(U32<E> *)(buf + 12) = sym.get_plt_idx(ctx) * sizeof(ElfRel<E>);
|
||||
} else {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd001, // mov.l 1f, r0
|
||||
0x6002, // mov.l @r0, r0
|
||||
0x402b, // jmp @r0
|
||||
0xd101, // mov.l 2f, r1
|
||||
0, 0, // 1: .long GOTPLT_ENTRY
|
||||
0, 0, // 2: .long INDEX_IN_RELPLT
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::plt_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 8) = sym.get_gotplt_addr(ctx);
|
||||
*(U32<E> *)(buf + 12) = sym.get_plt_idx(ctx) * sizeof(ElfRel<E>);
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
if (ctx.arg.pic) {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd001, // mov.l 1f, r0
|
||||
0x00ce, // mov.l @(r0, r12), r0
|
||||
0x402b, // jmp @r0
|
||||
0x0009, // nop
|
||||
0, 0, // 1: .long GOT_ENTRY
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::pltgot_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 8) = sym.get_got_pltgot_addr(ctx) - ctx.got->shdr.sh_addr;
|
||||
} else {
|
||||
constexpr U16<E> insn[] = {
|
||||
0xd001, // mov.l 1f, r0
|
||||
0x6002, // mov.l @r0, r0
|
||||
0x402b, // jmp @r0
|
||||
0x0009, // nop
|
||||
0, 0, // 1: .long GOT_ENTRY
|
||||
};
|
||||
|
||||
static_assert(sizeof(insn) == E::pltgot_size);
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(U32<E> *)(buf + 8) = sym.get_got_pltgot_addr(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_SH_DIR32:
|
||||
*(U32<E> *)loc = val;
|
||||
break;
|
||||
case R_SH_REL32:
|
||||
*(U32<E> *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = get_addend(loc, rel);
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SH_DIR32:
|
||||
break;
|
||||
case R_SH_REL32:
|
||||
case R_SH_PLT32:
|
||||
*(U32<E> *)loc = S + A - P;
|
||||
break;
|
||||
case R_SH_GOT32:
|
||||
*(U32<E> *)loc = G;
|
||||
break;
|
||||
case R_SH_GOTPC:
|
||||
*(U32<E> *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_SH_GOTOFF:
|
||||
*(U32<E> *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_SH_TLS_GD_32:
|
||||
*(U32<E> *)loc = sym.get_tlsgd_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_SH_TLS_LD_32:
|
||||
*(U32<E> *)loc = ctx.got->get_tlsld_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_SH_TLS_LDO_32:
|
||||
*(U32<E> *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_SH_TLS_IE_32:
|
||||
*(U32<E> *)loc = sym.get_gottp_addr(ctx) + A - GOT;
|
||||
break;
|
||||
case R_SH_TLS_LE_32:
|
||||
*(U32<E> *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : get_addend(loc, rel);
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SH_DIR32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(U32<E> *)loc = *val;
|
||||
else
|
||||
*(U32<E> *)loc = S + A;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
Error(ctx) << sym << ": GNU ifunc symbol is not supported on sh4";
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SH_REL32:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_SH_GOT32:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_SH_PLT32:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_SH_TLS_GD_32:
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
break;
|
||||
case R_SH_TLS_LD_32:
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_SH_TLS_IE_32:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_SH_TLS_LE_32:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_SH_DIR32:
|
||||
case R_SH_GOTPC:
|
||||
case R_SH_GOTOFF:
|
||||
case R_SH_TLS_LDO_32:
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,701 +0,0 @@
|
||||
// SPARC is a RISC ISA developed by Sun Microsystems.
|
||||
//
|
||||
// The byte order of the processor is big-endian. Anything larger than a
|
||||
// byte is stored in the "reverse" order compared to little-endian
|
||||
// processors such as x86-64.
|
||||
//
|
||||
// All instructions are 4 bytes long and aligned to 4 bytes boundaries.
|
||||
//
|
||||
// A notable feature of SPARC is that, unlike other RISC ISAs, it doesn't
|
||||
// need range extension thunks. It is because the SPARC's CALL instruction
|
||||
// contains a whopping 30 bits immediate. The processor scales it by 4 to
|
||||
// extend it to 32 bits (this is doable because all instructions are
|
||||
// aligned to 4 bytes boundaries, so the least significant two bits are
|
||||
// always zero). That means CALL's reach is PC ± 2 GiB, elinating the
|
||||
// need of range extension thunks. It comes with the cost that the CALL
|
||||
// instruction alone takes 1/4th of the instruction encoding space,
|
||||
// though.
|
||||
//
|
||||
// SPARC has 32 general purpose registers. CALL instruction saves a return
|
||||
// address to %o7, which is an alias for %r15. Thread pointer is stored to
|
||||
// %g7 which is %r7.
|
||||
//
|
||||
// SPARC does not have PC-relative load/store instructions. To access data
|
||||
// in the position-independent manner, we usually first set the address of
|
||||
// .got to, for example, %l7, with the following piece of code
|
||||
//
|
||||
// sethi %hi(. - _GLOBAL_OFFSET_TABLE_), %l7
|
||||
// add %l7, %lo(. - _GLOBAL_OFFSET_TABLE_), %l7
|
||||
// call __sparc_get_pc_thunk.l7
|
||||
// nop
|
||||
//
|
||||
// where __sparc_get_pc_thunk.l7 is defined as
|
||||
//
|
||||
// retl
|
||||
// add %o7, %l7, %l7
|
||||
//
|
||||
// . SETHI and the following ADD materialize a 32 bits offset to .got.
|
||||
// CALL instruction sets a return address to $o7, and the subsequent ADD
|
||||
// adds it to the GOT offset to materialize the absolute address of .got.
|
||||
//
|
||||
// Note that we have a NOP after CALL and an ADD after RETL because of
|
||||
// SPARC's delay branch slots. That is, the SPARC processor always
|
||||
// executes one instruction after a branch even if the branch is taken.
|
||||
// This may seem like an odd behavior, and indeed it is considered as such
|
||||
// (that's a premature optimization for the early pipelined SPARC
|
||||
// processors), but that's been a part of the ISA's spec so that's what it
|
||||
// is.
|
||||
//
|
||||
// Note also that the .got address obtained this way is not shared between
|
||||
// functions, so functions can use an arbitrary register to hold the .got
|
||||
// address. That also means each function needs to execute the above piece
|
||||
// of code to become position-independent.
|
||||
//
|
||||
// https://github.com/rui314/psabi/blob/main/sparc.pdf
|
||||
|
||||
#if MOLD_SPARC64
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = SPARC64;
|
||||
|
||||
// SPARC's PLT section is writable despite containing executable code.
|
||||
// We don't need to write the PLT header entry because the dynamic loader
|
||||
// will do that for us.
|
||||
//
|
||||
// We also don't need a .got.plt section to store the result of lazy PLT
|
||||
// symbol resolution because the dynamic symbol resolver directly mutates
|
||||
// instructions in PLT so that they jump to the right places next time.
|
||||
// That's why each PLT entry contains lots of NOPs; they are a placeholder
|
||||
// for the runtime to add more instructions.
|
||||
//
|
||||
// Self-modifying code is nowadays considered really bad from the security
|
||||
// point of view, though.
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
memset(buf, 0, E::plt_hdr_size);
|
||||
}
|
||||
|
||||
// Returns the byte offset within .plt of the data pointer for a large SPARC
|
||||
// PLT entry. See write_plt_entry below for the block layout this assumes.
|
||||
i64 sparc_plt_ptr_offset(Context<E> &ctx, i64 pltidx) {
|
||||
i64 i = pltidx - sparc_num_small_plt;
|
||||
i64 block = i / 160;
|
||||
i64 nlarge = (i64)ctx.plt->symbols.size() - sparc_num_small_plt;
|
||||
i64 nstub = std::min<i64>(160, nlarge - block * 160);
|
||||
return 0x100000 + block * 5120 + nstub * 24 + (i % 160) * 8;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
// SPARC uses two PLT entry formats. A "small" entry branches directly to
|
||||
// the resolver stub (.PLT1) with a BPcc instruction whose reach is only
|
||||
// ±1 MiB. Once the PLT grows past that (0x100000 bytes), we switch to a
|
||||
// "large" format: rather than branching to the resolver, a large entry
|
||||
// loads a 64-bit value from a nearby data pointer and jumps to (that value
|
||||
// + its own address).
|
||||
//
|
||||
// Large entries are grouped into blocks of 160: each block is 160 code
|
||||
// stubs followed by 160 8-byte data pointers, one per stub. The stubs must
|
||||
// sit on a fixed grid, 24 bytes apart, because that is how the loader finds
|
||||
// them: on a lazy bind the stub jumps to the resolver (.PLT0) with only its
|
||||
// own address, and the loader derives which symbol to resolve from that
|
||||
// address. Nothing may sit between two stubs, so each stub's pointer lives
|
||||
// in the block's pointer region, which it reaches with a signed 13-bit ldx
|
||||
// offset (see to_plt_offset). This layout is dictated by the loader; we
|
||||
// cannot rearrange or simplify it.
|
||||
i64 idx = sym.get_plt_idx(ctx);
|
||||
|
||||
if (idx < sparc_num_small_plt) {
|
||||
static ub32 insn[] = {
|
||||
0x0300'0000, // sethi (. - .PLT0), %g1
|
||||
0x3068'0000, // ba,a %xcc, .PLT1
|
||||
0x0100'0000, // nop
|
||||
0x0100'0000, // nop
|
||||
0x0100'0000, // nop
|
||||
0x0100'0000, // nop
|
||||
0x0100'0000, // nop
|
||||
0x0100'0000, // nop
|
||||
};
|
||||
|
||||
u64 plt0 = ctx.plt->shdr.sh_addr;
|
||||
u64 plt1 = ctx.plt->shdr.sh_addr + E::plt_size;
|
||||
u64 entry = sym.get_plt_addr(ctx);
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)buf |= bits(entry - plt0, 21, 0);
|
||||
*(ub32 *)(buf + 4) |= bits(plt1 - entry - 4, 20, 2);
|
||||
} else {
|
||||
static ub32 insn[] = {
|
||||
0x8a10'000f, // mov %o7, %g5
|
||||
0x4000'0002, // call . + 8
|
||||
0x0100'0000, // nop
|
||||
0xc25b'e000, // ldx [ %o7 + .ptr ], %g1
|
||||
0x83c3'c001, // jmpl %o7 + %g1, %g1
|
||||
0x9e10'0005, // mov %g5, %o7
|
||||
};
|
||||
|
||||
u64 plt = ctx.plt->shdr.sh_addr;
|
||||
u64 call = sym.get_plt_addr(ctx) + 4;
|
||||
i64 ptroff = sparc_plt_ptr_offset(ctx, idx);
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ub32 *)(buf + 12) |= bits(plt + ptroff - call, 12, 0);
|
||||
|
||||
// The data pointer initially holds (.PLT0 - call) so that the first call
|
||||
// jumps to .PLT0, where the loader's lazy resolver lives. The resolver
|
||||
// later overwrites it with (target - call).
|
||||
*(ub64 *)(ctx.buf + ctx.plt->shdr.sh_offset + ptroff) = plt - call;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static ub32 entry[] = {
|
||||
0x8a10'000f, // mov %o7, %g5
|
||||
0x4000'0002, // call . + 8
|
||||
0xc25b'e014, // ldx [ %o7 + 20 ], %g1
|
||||
0xc25b'c001, // ldx [ %o7 + %g1 ], %g1
|
||||
0x81c0'4000, // jmp %g1
|
||||
0x9e10'0005, // mov %g5, %o7
|
||||
0x0000'0000, // .quad $plt_entry - $got_entry
|
||||
0x0000'0000,
|
||||
};
|
||||
|
||||
memcpy(buf, entry, sizeof(entry));
|
||||
*(ub64 *)(buf + 24) = sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx) - 4;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_SPARC_64:
|
||||
case R_SPARC_UA64:
|
||||
*(ub64 *)loc = val;
|
||||
break;
|
||||
case R_SPARC_DISP32:
|
||||
*(ub32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// We iterate over relocations in reverse order so that it is easy
|
||||
// to swap instructions for R_SPARC_TLS_GD_CALL.
|
||||
for (i64 i = (i64)rels.size() - 1; i >= 0; i--) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_idx(ctx) * sizeof(Word<E>);
|
||||
u64 GOT = ctx.got->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto rs1 = [&] { return *(ub32 *)loc & (0b11111 << 14); };
|
||||
auto rs2 = [&] { return *(ub32 *)loc & 0b11111; };
|
||||
auto rd = [&] { return *(ub32 *)loc & (0b11111 << 25); };
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SPARC_5:
|
||||
check(S + A, 0, 1 << 5);
|
||||
*(ub32 *)loc |= bits(S + A, 4, 0);
|
||||
break;
|
||||
case R_SPARC_6:
|
||||
check(S + A, 0, 1 << 6);
|
||||
*(ub32 *)loc |= bits(S + A, 5, 0);
|
||||
break;
|
||||
case R_SPARC_7:
|
||||
check(S + A, 0, 1 << 7);
|
||||
*(ub32 *)loc |= bits(S + A, 6, 0);
|
||||
break;
|
||||
case R_SPARC_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_SPARC_10:
|
||||
check(S + A, 0, 1 << 10);
|
||||
*(ub32 *)loc |= bits(S + A, 9, 0);
|
||||
break;
|
||||
case R_SPARC_LO10:
|
||||
case R_SPARC_LOPLT10:
|
||||
*(ub32 *)loc |= bits(S + A, 9, 0);
|
||||
break;
|
||||
case R_SPARC_11:
|
||||
check(S + A, 0, 1 << 11);
|
||||
*(ub32 *)loc |= bits(S + A, 10, 0);
|
||||
break;
|
||||
case R_SPARC_13:
|
||||
check(S + A, 0, 1 << 13);
|
||||
*(ub32 *)loc |= bits(S + A, 12, 0);
|
||||
break;
|
||||
case R_SPARC_16:
|
||||
case R_SPARC_UA16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ub16 *)loc = S + A;
|
||||
break;
|
||||
case R_SPARC_22:
|
||||
check(S + A, 0, 1 << 22);
|
||||
*(ub32 *)loc |= bits(S + A, 21, 0);
|
||||
break;
|
||||
case R_SPARC_32:
|
||||
case R_SPARC_UA32:
|
||||
case R_SPARC_PLT32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_SPARC_PLT64:
|
||||
case R_SPARC_REGISTER:
|
||||
*(ub64 *)loc = S + A;
|
||||
break;
|
||||
case R_SPARC_DISP8:
|
||||
check(S + A - P, -(1 << 7), 1 << 7);
|
||||
*loc = S + A - P;
|
||||
break;
|
||||
case R_SPARC_DISP16:
|
||||
check(S + A - P, -(1 << 15), 1 << 15);
|
||||
*(ub16 *)loc = S + A - P;
|
||||
break;
|
||||
case R_SPARC_DISP32:
|
||||
case R_SPARC_PCPLT32:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 31);
|
||||
*(ub32 *)loc = S + A - P;
|
||||
break;
|
||||
case R_SPARC_DISP64:
|
||||
*(ub64 *)loc = S + A - P;
|
||||
break;
|
||||
case R_SPARC_WDISP16: {
|
||||
i64 val = S + A - P;
|
||||
check(val, -(1 << 16), 1 << 16);
|
||||
*(ub16 *)loc |= (bit(val, 16) << 21) | bits(val, 15, 2);
|
||||
break;
|
||||
}
|
||||
case R_SPARC_WDISP19:
|
||||
check(S + A - P, -(1 << 20), 1 << 20);
|
||||
*(ub32 *)loc |= bits(S + A - P, 20, 2);
|
||||
break;
|
||||
case R_SPARC_WDISP22:
|
||||
check(S + A - P, -(1 << 23), 1 << 23);
|
||||
*(ub32 *)loc |= bits(S + A - P, 23, 2);
|
||||
break;
|
||||
case R_SPARC_WDISP30:
|
||||
case R_SPARC_WPLT30:
|
||||
check(S + A - P, -(1LL << 31), 1LL << 31);
|
||||
*(ub32 *)loc |= bits(S + A - P, 31, 2);
|
||||
break;
|
||||
case R_SPARC_HI22:
|
||||
case R_SPARC_HIPLT22:
|
||||
case R_SPARC_LM22:
|
||||
*(ub32 *)loc |= bits(S + A, 31, 10);
|
||||
break;
|
||||
case R_SPARC_GOT10:
|
||||
*(ub32 *)loc |= bits(G, 9, 0);
|
||||
break;
|
||||
case R_SPARC_GOT13:
|
||||
check(G, 0, 1 << 12);
|
||||
*(ub32 *)loc |= bits(G, 12, 0);
|
||||
break;
|
||||
case R_SPARC_GOT22:
|
||||
*(ub32 *)loc |= bits(G, 31, 10);
|
||||
break;
|
||||
case R_SPARC_GOTDATA_HIX22: {
|
||||
i64 val = S + A - GOT;
|
||||
*(ub32 *)loc |= bits(val < 0 ? ~val : val, 31, 10);
|
||||
break;
|
||||
}
|
||||
case R_SPARC_GOTDATA_LOX10: {
|
||||
i64 val = S + A - GOT;
|
||||
*(ub32 *)loc |= bits(val, 9, 0) | (val < 0 ? 0b1'1100'0000'0000 : 0);
|
||||
break;
|
||||
}
|
||||
case R_SPARC_GOTDATA_OP_HIX22:
|
||||
// We always have to relax a GOT load to a load immediate if a
|
||||
// symbol is local, because R_SPARC_GOTDATA_OP cannot represent
|
||||
// an addend for a local symbol.
|
||||
if (sym.is_absolute()) {
|
||||
i64 val = S + A;
|
||||
*(ub32 *)loc |= bits(val < 0 ? ~val : val, 31, 10);
|
||||
} else if (sym.is_pcrel_linktime_const(ctx)) {
|
||||
i64 val = S + A - GOT;
|
||||
*(ub32 *)loc |= bits(val < 0 ? ~val : val, 31, 10);
|
||||
} else {
|
||||
*(ub32 *)loc |= bits(G, 31, 10);
|
||||
}
|
||||
break;
|
||||
case R_SPARC_GOTDATA_OP_LOX10:
|
||||
if (sym.is_absolute()) {
|
||||
i64 val = S + A;
|
||||
*(ub32 *)loc |= bits(val, 9, 0) | (val < 0 ? 0b1'1100'0000'0000 : 0);
|
||||
} else if (sym.is_pcrel_linktime_const(ctx)) {
|
||||
i64 val = S + A - GOT;
|
||||
*(ub32 *)loc |= bits(val, 9, 0) | (val < 0 ? 0b1'1100'0000'0000 : 0);
|
||||
} else {
|
||||
*(ub32 *)loc |= bits(G, 9, 0);
|
||||
}
|
||||
break;
|
||||
case R_SPARC_GOTDATA_OP:
|
||||
if (sym.is_absolute()) {
|
||||
// ldx [ %rs1 + %rs2 ], %rd → mov %rs2, %rd
|
||||
*(ub32 *)loc = 0x8010'0000 | rs2() | rd();
|
||||
} else if (sym.is_pcrel_linktime_const(ctx)) {
|
||||
// ldx [ %rs1 + %rs2 ], %rd → add %rs1, %rs2, %rd
|
||||
*(ub32 *)loc = 0x8000'0000 | rs1() | rs2() | rd();
|
||||
}
|
||||
break;
|
||||
case R_SPARC_PC10:
|
||||
case R_SPARC_PCPLT10:
|
||||
*(ub32 *)loc |= bits(S + A - P, 9, 0);
|
||||
break;
|
||||
case R_SPARC_PC22:
|
||||
case R_SPARC_PCPLT22:
|
||||
case R_SPARC_PC_LM22:
|
||||
*(ub32 *)loc |= bits(S + A - P, 31, 10);
|
||||
break;
|
||||
case R_SPARC_OLO10:
|
||||
*(ub32 *)loc |= bits(bits(S + A, 9, 0) + rel.r_type_data, 12, 0);
|
||||
break;
|
||||
case R_SPARC_HH22:
|
||||
*(ub32 *)loc |= bits(S + A, 63, 42);
|
||||
break;
|
||||
case R_SPARC_HM10:
|
||||
*(ub32 *)loc |= bits(S + A, 41, 32);
|
||||
break;
|
||||
case R_SPARC_PC_HH22:
|
||||
*(ub32 *)loc |= bits(S + A - P, 63, 42);
|
||||
break;
|
||||
case R_SPARC_PC_HM10:
|
||||
*(ub32 *)loc |= bits(S + A - P, 41, 32);
|
||||
break;
|
||||
case R_SPARC_HIX22:
|
||||
*(ub32 *)loc |= bits(~(S + A), 31, 10);
|
||||
break;
|
||||
case R_SPARC_LOX10:
|
||||
*(ub32 *)loc |= bits(S + A, 9, 0) | 0b1'1100'0000'0000;
|
||||
break;
|
||||
case R_SPARC_H44:
|
||||
*(ub32 *)loc |= bits(S + A, 43, 22);
|
||||
break;
|
||||
case R_SPARC_M44:
|
||||
*(ub32 *)loc |= bits(S + A, 21, 12);
|
||||
break;
|
||||
case R_SPARC_L44:
|
||||
*(ub32 *)loc |= bits(S + A, 11, 0);
|
||||
break;
|
||||
case R_SPARC_TLS_GD_HI22:
|
||||
if (sym.has_tlsgd(ctx)) {
|
||||
*(ub32 *)loc |= bits(sym.get_tlsgd_addr(ctx) + A - GOT, 31, 10);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
*(ub32 *)loc |= bits(sym.get_gottp_addr(ctx) + A - GOT, 31, 10);
|
||||
} else {
|
||||
*(ub32 *)loc |= bits(~(S + A - ctx.tp_addr), 31, 10);
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_GD_LO10:
|
||||
if (sym.has_tlsgd(ctx)) {
|
||||
*(ub32 *)loc |= bits(sym.get_tlsgd_addr(ctx) + A - GOT, 9, 0);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
// add %rs1, %rs2, %rd → or %rs1, $imm, %rd
|
||||
*(ub32 *)loc = 0x8010'2000 | rs1() | rd();
|
||||
*(ub32 *)loc |= bits(sym.get_gottp_addr(ctx) + A - GOT, 9, 0);
|
||||
} else {
|
||||
// add %rs1, %rs2, %rd → xor %rs1, $imm, %rd
|
||||
*(ub32 *)loc = 0x8018'2000 | rs1() | rd();
|
||||
*(ub32 *)loc |= bits(S + A - ctx.tp_addr, 9, 0) | 0b1'1100'0000'0000;
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_GD_ADD:
|
||||
if (sym.has_tlsgd(ctx)) {
|
||||
// do nothing
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
// add %rs1, %rs2, %rd → ldx [ %rs1 + %rs2 ], %rd
|
||||
*(ub32 *)loc = 0xc058'0000 | rs1() | rs2() | rd();
|
||||
} else {
|
||||
// add %rs1, %rs2, %rd → add %g7, %rs2, %rd
|
||||
*(ub32 *)loc = 0x8001'c000 | rs2() | rd();
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_GD_CALL:
|
||||
if (sym.has_tlsgd(ctx)) {
|
||||
u64 addr = ctx.extra.tls_get_addr->get_addr(ctx);
|
||||
*(ub32 *)loc |= bits(addr + A - P, 31, 2);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
// When we rewrite a branch instruction with a non-branch one,
|
||||
// we need to swap the instruction and the following one so that
|
||||
// the original execution order, which is inverted due to the
|
||||
// branch delay slot, is preserved.
|
||||
//
|
||||
// Since we apply relocations from the end to the beginning,
|
||||
// the instruction at loc + 4 is already complete.
|
||||
memcpy(loc, loc + 4, 4);
|
||||
*(ub32 *)(loc + 4) = 0x9001'c008; // add %g7, %o0, %o0
|
||||
} else {
|
||||
*(ub32 *)loc = 0x0100'0000; // call → nop
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_LDM_HI22:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
*(ub32 *)loc |= bits(ctx.got->get_tlsld_addr(ctx) + A - GOT, 31, 10);
|
||||
else
|
||||
*(ub32 *)loc |= bits(ctx.tp_addr - ctx.tls_begin, 31, 10);
|
||||
break;
|
||||
case R_SPARC_TLS_LDM_LO10:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
*(ub32 *)loc |= bits(ctx.got->get_tlsld_addr(ctx) + A - GOT, 9, 0);
|
||||
else
|
||||
*(ub32 *)loc |= bits(ctx.tp_addr - ctx.tls_begin, 9, 0);
|
||||
break;
|
||||
case R_SPARC_TLS_LDM_ADD:
|
||||
if (!ctx.got->has_tlsld(ctx))
|
||||
*(ub32 *)loc = 0x8021'c000 | rs2() | rd(); // sub %g7, %rs2, %rd
|
||||
break;
|
||||
case R_SPARC_TLS_LDM_CALL:
|
||||
if (ctx.got->has_tlsld(ctx)) {
|
||||
u64 addr = ctx.extra.tls_get_addr->get_addr(ctx);
|
||||
*(ub32 *)loc |= bits(addr + A - P, 31, 2);
|
||||
} else {
|
||||
*(ub32 *)loc = 0x0100'0000; // nop
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_LDO_HIX22:
|
||||
*(ub32 *)loc |= bits(S + A - ctx.dtp_addr, 31, 10);
|
||||
break;
|
||||
case R_SPARC_TLS_LDO_LOX10:
|
||||
*(ub32 *)loc |= bits(S + A - ctx.dtp_addr, 9, 0);
|
||||
break;
|
||||
case R_SPARC_TLS_IE_HI22:
|
||||
*(ub32 *)loc |= bits(sym.get_gottp_addr(ctx) + A - GOT, 31, 10);
|
||||
break;
|
||||
case R_SPARC_TLS_IE_LO10:
|
||||
*(ub32 *)loc |= bits(sym.get_gottp_addr(ctx) + A - GOT, 9, 0);
|
||||
break;
|
||||
case R_SPARC_TLS_LE_HIX22:
|
||||
*(ub32 *)loc |= bits(~(S + A - ctx.tp_addr), 31, 10);
|
||||
break;
|
||||
case R_SPARC_TLS_LE_LOX10:
|
||||
*(ub32 *)loc |= bits(S + A - ctx.tp_addr, 9, 0) | 0b1'1100'0000'0000;
|
||||
break;
|
||||
case R_SPARC_SIZE32:
|
||||
*(ub32 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
case R_SPARC_64:
|
||||
case R_SPARC_UA64:
|
||||
case R_SPARC_TLS_LDO_ADD:
|
||||
case R_SPARC_TLS_IE_LD:
|
||||
case R_SPARC_TLS_IE_LDX:
|
||||
case R_SPARC_TLS_IE_ADD:
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, val, i, lo, hi);
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SPARC_64:
|
||||
case R_SPARC_UA64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ub64 *)loc = *val;
|
||||
else
|
||||
*(ub64 *)loc = S + A;
|
||||
break;
|
||||
case R_SPARC_32:
|
||||
case R_SPARC_UA32:
|
||||
check(S + A, 0, 1LL << 32);
|
||||
*(ub32 *)loc = S + A;
|
||||
break;
|
||||
case R_SPARC_TLS_DTPOFF32:
|
||||
*(ub32 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_SPARC_TLS_DTPOFF64:
|
||||
*(ub64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": apply_reloc_nonalloc: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
bool needs_tlsgd = false;
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_SPARC_8:
|
||||
case R_SPARC_5:
|
||||
case R_SPARC_6:
|
||||
case R_SPARC_7:
|
||||
case R_SPARC_10:
|
||||
case R_SPARC_11:
|
||||
case R_SPARC_13:
|
||||
case R_SPARC_16:
|
||||
case R_SPARC_22:
|
||||
case R_SPARC_32:
|
||||
case R_SPARC_REGISTER:
|
||||
case R_SPARC_UA16:
|
||||
case R_SPARC_UA32:
|
||||
case R_SPARC_PC_HM10:
|
||||
case R_SPARC_OLO10:
|
||||
case R_SPARC_LOX10:
|
||||
case R_SPARC_HM10:
|
||||
case R_SPARC_M44:
|
||||
case R_SPARC_HIX22:
|
||||
case R_SPARC_LO10:
|
||||
case R_SPARC_L44:
|
||||
case R_SPARC_LM22:
|
||||
case R_SPARC_HI22:
|
||||
case R_SPARC_H44:
|
||||
case R_SPARC_HH22:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_SPARC_PLT32:
|
||||
case R_SPARC_WPLT30:
|
||||
case R_SPARC_WDISP30:
|
||||
case R_SPARC_HIPLT22:
|
||||
case R_SPARC_LOPLT10:
|
||||
case R_SPARC_PCPLT32:
|
||||
case R_SPARC_PCPLT22:
|
||||
case R_SPARC_PCPLT10:
|
||||
case R_SPARC_PLT64:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_SPARC_GOT13:
|
||||
case R_SPARC_GOT10:
|
||||
case R_SPARC_GOT22:
|
||||
case R_SPARC_GOTDATA_HIX22:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_SPARC_GOTDATA_OP_HIX22:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_SPARC_DISP16:
|
||||
case R_SPARC_DISP32:
|
||||
case R_SPARC_DISP64:
|
||||
case R_SPARC_DISP8:
|
||||
case R_SPARC_PC10:
|
||||
case R_SPARC_PC22:
|
||||
case R_SPARC_PC_LM22:
|
||||
case R_SPARC_WDISP16:
|
||||
case R_SPARC_WDISP19:
|
||||
case R_SPARC_WDISP22:
|
||||
case R_SPARC_PC_HH22:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_SPARC_TLS_GD_HI22:
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && sym.is_tprel_linktime_const(ctx))) {
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
} else if (ctx.arg.relax && sym.is_tprel_runtime_const(ctx)) {
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
} else {
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
needs_tlsgd = true;
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_LDM_HI22:
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && !ctx.arg.shared)) {
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
} else {
|
||||
ctx.needs_tlsld = true;
|
||||
}
|
||||
break;
|
||||
case R_SPARC_TLS_IE_HI22:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_SPARC_TLS_LE_HIX22:
|
||||
case R_SPARC_TLS_LE_LOX10:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_SPARC_64:
|
||||
case R_SPARC_UA64:
|
||||
case R_SPARC_GOTDATA_OP_LOX10:
|
||||
case R_SPARC_GOTDATA_OP:
|
||||
case R_SPARC_GOTDATA_LOX10:
|
||||
case R_SPARC_TLS_GD_LO10:
|
||||
case R_SPARC_TLS_GD_ADD:
|
||||
case R_SPARC_TLS_GD_CALL:
|
||||
case R_SPARC_TLS_LDM_LO10:
|
||||
case R_SPARC_TLS_LDM_ADD:
|
||||
case R_SPARC_TLS_LDM_CALL:
|
||||
case R_SPARC_TLS_LDO_HIX22:
|
||||
case R_SPARC_TLS_LDO_LOX10:
|
||||
case R_SPARC_TLS_LDO_ADD:
|
||||
case R_SPARC_TLS_IE_ADD:
|
||||
case R_SPARC_TLS_IE_LD:
|
||||
case R_SPARC_TLS_IE_LDX:
|
||||
case R_SPARC_TLS_IE_LO10:
|
||||
case R_SPARC_SIZE32:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
// TLS_GD_CALL and TLS_LDM_CALL relocations implicitly refer to
|
||||
// __tls_get_addr, which may be dynamically linked from libc.so.
|
||||
Symbol<E> &sym = *ctx.extra.tls_get_addr;
|
||||
if (sym.is_imported && (needs_tlsgd || ctx.needs_tlsld))
|
||||
sym.flags |= NEEDS_PLT;
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,955 +0,0 @@
|
||||
// Supporting x86-64 is straightforward. Unlike its predecessor, i386,
|
||||
// x86-64 supports PC-relative addressing for position-independent code.
|
||||
// Being CISC, its instructions are variable in size. Branch instructions
|
||||
// take 4 bytes offsets, so we don't need range extension thunks.
|
||||
//
|
||||
// The psABI specifies %r11 as neither caller- nor callee-saved. It's
|
||||
// intentionally left out so that we can use it as a scratch register in
|
||||
// PLT.
|
||||
//
|
||||
// Thread Pointer (TP) is stored not to a general-purpose register but to
|
||||
// FS segment register. Segment register is a 64-bits register which can
|
||||
// be used as a base address for memory access. Each thread has a unique
|
||||
// FS value, and they access their thread-local variables relative to FS
|
||||
// as %fs:offset_from_tp.
|
||||
//
|
||||
// The value of a segment register itself is not generally readable from
|
||||
// the user space. As a workaround, libc initializes %fs:0 (the first word
|
||||
// referenced by FS) to the value of %fs itself. So we can obtain TP just
|
||||
// by `mov %fs:0, %rax` if we need it.
|
||||
//
|
||||
// For historical reasons, TP points past the end of the TLS block on x86.
|
||||
// This is contrary to other psABIs which usually use the beginning of the
|
||||
// TLS block as TP (with some addend). As a result, offsets from TP to
|
||||
// thread-local variables (TLVs) in the main executable are all negative.
|
||||
//
|
||||
// https://gitlab.com/x86-psABIs/x86-64-ABI
|
||||
|
||||
#if MOLD_X86_64
|
||||
|
||||
#include "mold.h"
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
using E = X86_64;
|
||||
|
||||
// This is a security-enhanced version of the regular PLT. The PLT
|
||||
// header and each PLT entry starts with endbr64 for the Intel's
|
||||
// control-flow enforcement security mechanism.
|
||||
//
|
||||
// Note that our IBT-enabled PLT instruction sequence is different
|
||||
// from the one used in GNU ld. GNU's IBTPLT implementation uses two
|
||||
// separate sections (.plt and .plt.sec) in which one PLT entry takes
|
||||
// 32 bytes in total. Our IBTPLT consists of just .plt and each entry
|
||||
// is 16 bytes long.
|
||||
//
|
||||
// Our PLT entry clobbers %r11, but that's fine because the resolver
|
||||
// function (_dl_runtime_resolve) clobbers %r11 anyway.
|
||||
template <>
|
||||
void write_plt_header(Context<E> &ctx, u8 *buf) {
|
||||
static const u8 insn[] = {
|
||||
0xf3, 0x0f, 0x1e, 0xfa, // endbr64
|
||||
0x41, 0x53, // push %r11
|
||||
0xff, 0x35, 0, 0, 0, 0, // push GOTPLT+8(%rip)
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *GOTPLT+16(%rip)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
0xcc, 0xcc, // (padding)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 8) = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 4;
|
||||
*(ul32 *)(buf + 14) = ctx.gotplt->shdr.sh_addr - ctx.plt->shdr.sh_addr - 2;
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_plt_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
// Only a canonical PLT can be address-taken; there's no way to take
|
||||
// an address of a non-canonical PLT. Therefore, a non-canonical PLT
|
||||
// doesn't have to start with an endbr64.
|
||||
if (sym.is_canonical) {
|
||||
static const u8 insn[] = {
|
||||
0xf3, 0x0f, 0x1e, 0xfa, // endbr64
|
||||
0x41, 0xbb, 0, 0, 0, 0, // mov $index_in_relplt, %r11d
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOTPLT
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 6) = sym.get_plt_idx(ctx);
|
||||
*(ul32 *)(buf + 12) = sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx) - 16;
|
||||
} else {
|
||||
static const u8 insn[] = {
|
||||
0x41, 0xbb, 0, 0, 0, 0, // mov $index_in_relplt, %r11d
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOTPLT
|
||||
0xcc, 0xcc, 0xcc, 0xcc, // (padding)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 2) = sym.get_plt_idx(ctx);
|
||||
*(ul32 *)(buf + 8) = sym.get_gotplt_addr(ctx) - sym.get_plt_addr(ctx) - 12;
|
||||
}
|
||||
}
|
||||
|
||||
template <>
|
||||
void write_pltgot_entry(Context<E> &ctx, u8 *buf, Symbol<E> &sym) {
|
||||
static const u8 insn[] = {
|
||||
0xff, 0x25, 0, 0, 0, 0, // jmp *foo@GOT
|
||||
0xcc, 0xcc, // (padding)
|
||||
};
|
||||
|
||||
memcpy(buf, insn, sizeof(insn));
|
||||
*(ul32 *)(buf + 2) = sym.get_got_pltgot_addr(ctx) - sym.get_plt_addr(ctx) - 6;
|
||||
}
|
||||
|
||||
template <>
|
||||
void EhFrameSection<E>::apply_eh_reloc(Context<E> &ctx, const ElfRel<E> &rel,
|
||||
u64 offset, u64 val) {
|
||||
u8 *loc = ctx.buf + this->shdr.sh_offset + offset;
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_NONE:
|
||||
break;
|
||||
case R_X86_64_32:
|
||||
*(ul32 *)loc = val;
|
||||
break;
|
||||
case R_X86_64_64:
|
||||
*(ul64 *)loc = val;
|
||||
break;
|
||||
case R_X86_64_PC32:
|
||||
*(ul32 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
case R_X86_64_PC64:
|
||||
*(ul64 *)loc = val - this->shdr.sh_addr - offset;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "unsupported relocation in .eh_frame: " << rel;
|
||||
}
|
||||
}
|
||||
|
||||
static u32 relax_gotpcrelx(u8 *loc, const ElfRel<E> &rel) {
|
||||
if (rel.r_type == R_X86_64_GOTPCRELX) {
|
||||
switch ((loc[-2] << 8) | loc[-1]) {
|
||||
case 0xff15: return 0x40e8; // call *0(%rip) -> call 0
|
||||
case 0xff25: return 0x40e9; // jmp *0(%rip) -> jmp 0
|
||||
}
|
||||
} else {
|
||||
assert(rel.r_type == R_X86_64_REX_GOTPCRELX ||
|
||||
rel.r_type == R_X86_64_CODE_4_GOTPCRELX);
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488b05: return 0x8d05; // mov 0(%rip), %rax -> lea 0(%rip), %rax
|
||||
case 0x488b0d: return 0x8d0d; // mov 0(%rip), %rcx -> lea 0(%rip), %rcx
|
||||
case 0x488b15: return 0x8d15; // mov 0(%rip), %rdx -> lea 0(%rip), %rdx
|
||||
case 0x488b1d: return 0x8d1d; // mov 0(%rip), %rbx -> lea 0(%rip), %rbx
|
||||
case 0x488b25: return 0x8d25; // mov 0(%rip), %rsp -> lea 0(%rip), %rsp
|
||||
case 0x488b2d: return 0x8d2d; // mov 0(%rip), %rbp -> lea 0(%rip), %rbp
|
||||
case 0x488b35: return 0x8d35; // mov 0(%rip), %rsi -> lea 0(%rip), %rsi
|
||||
case 0x488b3d: return 0x8d3d; // mov 0(%rip), %rdi -> lea 0(%rip), %rdi
|
||||
case 0x4c8b05: return 0x8d05; // mov 0(%rip), %r8 -> lea 0(%rip), %r8
|
||||
case 0x4c8b0d: return 0x8d0d; // mov 0(%rip), %r9 -> lea 0(%rip), %r9
|
||||
case 0x4c8b15: return 0x8d15; // mov 0(%rip), %r10 -> lea 0(%rip), %r10
|
||||
case 0x4c8b1d: return 0x8d1d; // mov 0(%rip), %r11 -> lea 0(%rip), %r11
|
||||
case 0x4c8b25: return 0x8d25; // mov 0(%rip), %r12 -> lea 0(%rip), %r12
|
||||
case 0x4c8b2d: return 0x8d2d; // mov 0(%rip), %r13 -> lea 0(%rip), %r13
|
||||
case 0x4c8b35: return 0x8d35; // mov 0(%rip), %r14 -> lea 0(%rip), %r14
|
||||
case 0x4c8b3d: return 0x8d3d; // mov 0(%rip), %r15 -> lea 0(%rip), %r15
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_gottpoff(u8 *loc, const ElfRel<E> &rel) {
|
||||
if (rel.r_type == R_X86_64_GOTTPOFF) {
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488b05: return 0x48c7c0; // mov 0(%rip), %rax -> mov $0, %rax
|
||||
case 0x488b0d: return 0x48c7c1; // mov 0(%rip), %rcx -> mov $0, %rcx
|
||||
case 0x488b15: return 0x48c7c2; // mov 0(%rip), %rdx -> mov $0, %rdx
|
||||
case 0x488b1d: return 0x48c7c3; // mov 0(%rip), %rbx -> mov $0, %rbx
|
||||
case 0x488b25: return 0x48c7c4; // mov 0(%rip), %rsp -> mov $0, %rsp
|
||||
case 0x488b2d: return 0x48c7c5; // mov 0(%rip), %rbp -> mov $0, %rbp
|
||||
case 0x488b35: return 0x48c7c6; // mov 0(%rip), %rsi -> mov $0, %rsi
|
||||
case 0x488b3d: return 0x48c7c7; // mov 0(%rip), %rdi -> mov $0, %rdi
|
||||
case 0x4c8b05: return 0x49c7c0; // mov 0(%rip), %r8 -> mov $0, %r8
|
||||
case 0x4c8b0d: return 0x49c7c1; // mov 0(%rip), %r9 -> mov $0, %r9
|
||||
case 0x4c8b15: return 0x49c7c2; // mov 0(%rip), %r10 -> mov $0, %r10
|
||||
case 0x4c8b1d: return 0x49c7c3; // mov 0(%rip), %r11 -> mov $0, %r11
|
||||
case 0x4c8b25: return 0x49c7c4; // mov 0(%rip), %r12 -> mov $0, %r12
|
||||
case 0x4c8b2d: return 0x49c7c5; // mov 0(%rip), %r13 -> mov $0, %r13
|
||||
case 0x4c8b35: return 0x49c7c6; // mov 0(%rip), %r14 -> mov $0, %r14
|
||||
case 0x4c8b3d: return 0x49c7c7; // mov 0(%rip), %r15 -> mov $0, %r15
|
||||
}
|
||||
} else {
|
||||
assert(rel.r_type == R_X86_64_CODE_4_GOTTPOFF);
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488b05: return 0x18c7c0; // mov 0(%rip), %r16 -> mov $0, %r16
|
||||
case 0x488b0d: return 0x18c7c1; // mov 0(%rip), %r17 -> mov $0, %r17
|
||||
case 0x488b15: return 0x18c7c2; // mov 0(%rip), %r18 -> mov $0, %r18
|
||||
case 0x488b1d: return 0x18c7c3; // mov 0(%rip), %r19 -> mov $0, %r19
|
||||
case 0x488b25: return 0x18c7c4; // mov 0(%rip), %r20 -> mov $0, %r20
|
||||
case 0x488b2d: return 0x18c7c5; // mov 0(%rip), %r21 -> mov $0, %r21
|
||||
case 0x488b35: return 0x18c7c6; // mov 0(%rip), %r22 -> mov $0, %r22
|
||||
case 0x488b3d: return 0x18c7c7; // mov 0(%rip), %r23 -> mov $0, %r23
|
||||
case 0x4c8b05: return 0x19c7c0; // mov 0(%rip), %r24 -> mov $0, %r24
|
||||
case 0x4c8b0d: return 0x19c7c1; // mov 0(%rip), %r25 -> mov $0, %r25
|
||||
case 0x4c8b15: return 0x19c7c2; // mov 0(%rip), %r26 -> mov $0, %r26
|
||||
case 0x4c8b1d: return 0x19c7c3; // mov 0(%rip), %r27 -> mov $0, %r27
|
||||
case 0x4c8b25: return 0x19c7c4; // mov 0(%rip), %r28 -> mov $0, %r28
|
||||
case 0x4c8b2d: return 0x19c7c5; // mov 0(%rip), %r29 -> mov $0, %r29
|
||||
case 0x4c8b35: return 0x19c7c6; // mov 0(%rip), %r30 -> mov $0, %r30
|
||||
case 0x4c8b3d: return 0x19c7c7; // mov 0(%rip), %r31 -> mov $0, %r31
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_tlsdesc_to_ie(u8 *loc, const ElfRel<E> &rel) {
|
||||
if (rel.r_type == R_X86_64_GOTPC32_TLSDESC) {
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488d05: return 0x488b05; // lea 0(%rip), %rax -> mov 0(%rip), %rax
|
||||
case 0x488d0d: return 0x488b0d; // lea 0(%rip), %rcx -> mov 0(%rip), %rcx
|
||||
case 0x488d15: return 0x488b15; // lea 0(%rip), %rdx -> mov 0(%rip), %rdx
|
||||
case 0x488d1d: return 0x488b1d; // lea 0(%rip), %rbx -> mov 0(%rip), %rbx
|
||||
case 0x488d25: return 0x488b25; // lea 0(%rip), %rsp -> mov 0(%rip), %rsp
|
||||
case 0x488d2d: return 0x488b2d; // lea 0(%rip), %rbp -> mov 0(%rip), %rbp
|
||||
case 0x488d35: return 0x488b35; // lea 0(%rip), %rsi -> mov 0(%rip), %rsi
|
||||
case 0x488d3d: return 0x488b3d; // lea 0(%rip), %rdi -> mov 0(%rip), %rdi
|
||||
case 0x4c8d05: return 0x4c8b05; // lea 0(%rip), %r8 -> mov 0(%rip), %r8
|
||||
case 0x4c8d0d: return 0x4c8b0d; // lea 0(%rip), %r9 -> mov 0(%rip), %r9
|
||||
case 0x4c8d15: return 0x4c8b15; // lea 0(%rip), %r10 -> mov 0(%rip), %r10
|
||||
case 0x4c8d1d: return 0x4c8b1d; // lea 0(%rip), %r11 -> mov 0(%rip), %r11
|
||||
case 0x4c8d25: return 0x4c8b25; // lea 0(%rip), %r12 -> mov 0(%rip), %r12
|
||||
case 0x4c8d2d: return 0x4c8b2d; // lea 0(%rip), %r13 -> mov 0(%rip), %r13
|
||||
case 0x4c8d35: return 0x4c8b35; // lea 0(%rip), %r14 -> mov 0(%rip), %r14
|
||||
case 0x4c8d3d: return 0x4c8b3d; // lea 0(%rip), %r15 -> mov 0(%rip), %r15
|
||||
}
|
||||
} else {
|
||||
assert(rel.r_type == R_X86_64_CODE_4_GOTPC32_TLSDESC);
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488d05: return 0x488b05; // lea 0(%rip), %r16 -> mov 0(%rip), %r16
|
||||
case 0x488d0d: return 0x488b0d; // lea 0(%rip), %r17 -> mov 0(%rip), %r17
|
||||
case 0x488d15: return 0x488b15; // lea 0(%rip), %r18 -> mov 0(%rip), %r18
|
||||
case 0x488d1d: return 0x488b1d; // lea 0(%rip), %r19 -> mov 0(%rip), %r19
|
||||
case 0x488d25: return 0x488b25; // lea 0(%rip), %r20 -> mov 0(%rip), %r20
|
||||
case 0x488d2d: return 0x488b2d; // lea 0(%rip), %r21 -> mov 0(%rip), %r21
|
||||
case 0x488d35: return 0x488b35; // lea 0(%rip), %r22 -> mov 0(%rip), %r22
|
||||
case 0x488d3d: return 0x488b3d; // lea 0(%rip), %r23 -> mov 0(%rip), %r23
|
||||
case 0x4c8d05: return 0x4c8b05; // lea 0(%rip), %r24 -> mov 0(%rip), %r24
|
||||
case 0x4c8d0d: return 0x4c8b0d; // lea 0(%rip), %r25 -> mov 0(%rip), %r25
|
||||
case 0x4c8d15: return 0x4c8b15; // lea 0(%rip), %r26 -> mov 0(%rip), %r26
|
||||
case 0x4c8d1d: return 0x4c8b1d; // lea 0(%rip), %r27 -> mov 0(%rip), %r27
|
||||
case 0x4c8d25: return 0x4c8b25; // lea 0(%rip), %r28 -> mov 0(%rip), %r28
|
||||
case 0x4c8d2d: return 0x4c8b2d; // lea 0(%rip), %r29 -> mov 0(%rip), %r29
|
||||
case 0x4c8d35: return 0x4c8b35; // lea 0(%rip), %r30 -> mov 0(%rip), %r30
|
||||
case 0x4c8d3d: return 0x4c8b3d; // lea 0(%rip), %r31 -> mov 0(%rip), %r31
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
static u32 relax_tlsdesc_to_le(u8 *loc, const ElfRel<E> &rel) {
|
||||
if (rel.r_type == R_X86_64_GOTPC32_TLSDESC) {
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488d05: return 0x48c7c0; // lea 0(%rip), %rax -> mov $0, %rax
|
||||
case 0x488d0d: return 0x48c7c1; // lea 0(%rip), %rcx -> mov $0, %rcx
|
||||
case 0x488d15: return 0x48c7c2; // lea 0(%rip), %rdx -> mov $0, %rdx
|
||||
case 0x488d1d: return 0x48c7c3; // lea 0(%rip), %rbx -> mov $0, %rbx
|
||||
case 0x488d25: return 0x48c7c4; // lea 0(%rip), %rsp -> mov $0, %rsp
|
||||
case 0x488d2d: return 0x48c7c5; // lea 0(%rip), %rbp -> mov $0, %rbp
|
||||
case 0x488d35: return 0x48c7c6; // lea 0(%rip), %rsi -> mov $0, %rsi
|
||||
case 0x488d3d: return 0x48c7c7; // lea 0(%rip), %rdi -> mov $0, %rdi
|
||||
case 0x4c8d05: return 0x49c7c0; // lea 0(%rip), %r8 -> mov $0, %r8
|
||||
case 0x4c8d0d: return 0x49c7c1; // lea 0(%rip), %r9 -> mov $0, %r9
|
||||
case 0x4c8d15: return 0x49c7c2; // lea 0(%rip), %r10 -> mov $0, %r10
|
||||
case 0x4c8d1d: return 0x49c7c3; // lea 0(%rip), %r11 -> mov $0, %r11
|
||||
case 0x4c8d25: return 0x49c7c4; // lea 0(%rip), %r12 -> mov $0, %r12
|
||||
case 0x4c8d2d: return 0x49c7c5; // lea 0(%rip), %r13 -> mov $0, %r13
|
||||
case 0x4c8d35: return 0x49c7c6; // lea 0(%rip), %r14 -> mov $0, %r14
|
||||
case 0x4c8d3d: return 0x49c7c7; // lea 0(%rip), %r15 -> mov $0, %r15
|
||||
}
|
||||
} else {
|
||||
assert(rel.r_type == R_X86_64_CODE_4_GOTPC32_TLSDESC);
|
||||
switch ((loc[-3] << 16) | (loc[-2] << 8) | loc[-1]) {
|
||||
case 0x488d05: return 0x18c7c0; // lea 0(%rip), %r16 -> mov $0, %r16
|
||||
case 0x488d0d: return 0x18c7c1; // lea 0(%rip), %r17 -> mov $0, %r17
|
||||
case 0x488d15: return 0x18c7c2; // lea 0(%rip), %r18 -> mov $0, %r18
|
||||
case 0x488d1d: return 0x18c7c3; // lea 0(%rip), %r19 -> mov $0, %r19
|
||||
case 0x488d25: return 0x18c7c4; // lea 0(%rip), %r20 -> mov $0, %r20
|
||||
case 0x488d2d: return 0x18c7c5; // lea 0(%rip), %r21 -> mov $0, %r21
|
||||
case 0x488d35: return 0x18c7c6; // lea 0(%rip), %r22 -> mov $0, %r22
|
||||
case 0x488d3d: return 0x18c7c7; // lea 0(%rip), %r23 -> mov $0, %r23
|
||||
case 0x4c8d05: return 0x19c7c0; // lea 0(%rip), %r24 -> mov $0, %r24
|
||||
case 0x4c8d0d: return 0x19c7c1; // lea 0(%rip), %r25 -> mov $0, %r25
|
||||
case 0x4c8d15: return 0x19c7c2; // lea 0(%rip), %r26 -> mov $0, %r26
|
||||
case 0x4c8d1d: return 0x19c7c3; // lea 0(%rip), %r27 -> mov $0, %r27
|
||||
case 0x4c8d25: return 0x19c7c4; // lea 0(%rip), %r28 -> mov $0, %r28
|
||||
case 0x4c8d2d: return 0x19c7c5; // lea 0(%rip), %r29 -> mov $0, %r29
|
||||
case 0x4c8d35: return 0x19c7c6; // lea 0(%rip), %r30 -> mov $0, %r30
|
||||
case 0x4c8d3d: return 0x19c7c7; // lea 0(%rip), %r31 -> mov $0, %r31
|
||||
}
|
||||
}
|
||||
return 0;
|
||||
}
|
||||
|
||||
// Rewrite a function call to __tls_get_addr to a cheaper instruction
|
||||
// sequence. We can do this when we know the thread-local variable's TP-
|
||||
// relative address at link-time.
|
||||
static void relax_gd_to_le(u8 *loc, const ElfRel<E> &rel, u64 val) {
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_PLT32:
|
||||
case R_X86_64_PC32:
|
||||
case R_X86_64_GOTPCREL:
|
||||
case R_X86_64_GOTPCRELX: {
|
||||
// The original instructions are the following:
|
||||
//
|
||||
// 66 48 8d 3d 00 00 00 00 lea foo@tlsgd(%rip), %rdi
|
||||
// 66 66 48 e8 00 00 00 00 call __tls_get_addr
|
||||
//
|
||||
// or
|
||||
//
|
||||
// 66 48 8d 3d 00 00 00 00 lea foo@tlsgd(%rip), %rdi
|
||||
// 66 48 ff 15 00 00 00 00 call *__tls_get_addr@GOT(%rip)
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x81, 0xc0, 0, 0, 0, 0, // add $tp_offset, %rax
|
||||
};
|
||||
memcpy(loc - 4, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 8) = val;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PLTOFF64: {
|
||||
// The original instructions are the following:
|
||||
//
|
||||
// 48 8d 3d 00 00 00 00 lea foo@tlsgd(%rip), %rdi
|
||||
// 48 b8 00 00 00 00 00 00 00 00 movabs __tls_get_addr, %rax
|
||||
// 48 01 d8 add %rbx, %rax
|
||||
// ff d0 call *%rax
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x81, 0xc0, 0, 0, 0, 0, // add $tp_offset, %rax
|
||||
0x66, 0x0f, 0x1f, 0x44, 0x00, 0x00, // nop
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 9) = val;
|
||||
break;
|
||||
}
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
static void relax_gd_to_ie(u8 *loc, const ElfRel<E> &rel, u64 val) {
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_PLT32:
|
||||
case R_X86_64_PC32:
|
||||
case R_X86_64_GOTPCREL:
|
||||
case R_X86_64_GOTPCRELX: {
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x03, 0x05, 0, 0, 0, 0, // add foo@gottpoff(%rip), %rax
|
||||
};
|
||||
memcpy(loc - 4, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 8) = val - 12;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PLTOFF64: {
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x03, 0x05, 0, 0, 0, 0, // add foo@gottpoff(%rip), %rax
|
||||
0x66, 0x0f, 0x1f, 0x44, 0x00, 0x00, // nop
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 9) = val - 13;
|
||||
break;
|
||||
}
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
// Rewrite a function call to __tls_get_addr to a cheaper instruction
|
||||
// sequence. The difference from relax_gd_to_le is that we are materializing
|
||||
// the address of the beginning of TLS block instead of an address of a
|
||||
// particular thread-local variable.
|
||||
static void relax_ld_to_le(u8 *loc, const ElfRel<E> &rel, i64 tls_size) {
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_PLT32:
|
||||
case R_X86_64_PC32: {
|
||||
// The original instructions are the following:
|
||||
//
|
||||
// 48 8d 3d 00 00 00 00 lea foo@tlsld(%rip), %rdi
|
||||
// e8 00 00 00 00 call __tls_get_addr
|
||||
//
|
||||
// Because the original instruction sequence is so short that we need a
|
||||
// little bit of code golfing here. "mov %fs:0, %rax" is 9 byte long, so
|
||||
// xor + mov is shorter. Note that `xor %eax, %eax` zero-clears %eax.
|
||||
static const u8 insn[] = {
|
||||
0x31, 0xc0, // xor %eax, %eax
|
||||
0x64, 0x48, 0x8b, 0x00, // mov %fs:(%rax), %rax
|
||||
0x48, 0x2d, 0, 0, 0, 0, // sub $tls_size, %rax
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 5) = tls_size;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_GOTPCREL:
|
||||
case R_X86_64_GOTPCRELX: {
|
||||
// The original instructions are the following:
|
||||
//
|
||||
// 48 8d 3d 00 00 00 00 lea foo@tlsld(%rip), %rdi
|
||||
// ff 15 00 00 00 00 call *__tls_get_addr@GOT(%rip)
|
||||
static const u8 insn[] = {
|
||||
0x48, 0x31, 0xc0, // xor %rax, %rax
|
||||
0x64, 0x48, 0x8b, 0x00, // mov %fs:(%rax), %rax
|
||||
0x48, 0x2d, 0, 0, 0, 0, // sub $tls_size, %rax
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 6) = tls_size;
|
||||
break;
|
||||
}
|
||||
case R_X86_64_PLTOFF64: {
|
||||
// The original instructions are the following:
|
||||
//
|
||||
// 48 8d 3d 00 00 00 00 lea foo@tlsld(%rip), %rdi
|
||||
// 48 b8 00 00 00 00 00 00 00 00 movabs __tls_get_addr@GOTOFF, %rax
|
||||
// 48 01 d8 add %rbx, %rax
|
||||
// ff d0 call *%rax
|
||||
static const u8 insn[] = {
|
||||
0x64, 0x48, 0x8b, 0x04, 0x25, 0, 0, 0, 0, // mov %fs:0, %rax
|
||||
0x48, 0x2d, 0, 0, 0, 0, // sub $tls_size, %rax
|
||||
0x0f, 0x1f, 0x80, 0x00, 0x00, 0x00, 0x00, // nop
|
||||
};
|
||||
memcpy(loc - 3, insn, sizeof(insn));
|
||||
*(ul32 *)(loc + 8) = tls_size;
|
||||
break;
|
||||
}
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
|
||||
// Apply relocations to SHF_ALLOC sections (i.e. sections that are
|
||||
// mapped to memory at runtime) based on the result of
|
||||
// scan_relocations().
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_alloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE)
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
u64 S = sym.get_addr(ctx);
|
||||
u64 A = rel.r_addend;
|
||||
u64 P = get_addr() + rel.r_offset;
|
||||
u64 G = sym.get_got_addr(ctx) - ctx.gotplt->shdr.sh_addr;
|
||||
u64 GOT = ctx.gotplt->shdr.sh_addr;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto write32 = [&](u64 val) {
|
||||
check(val, 0, 1LL << 32);
|
||||
*(ul32 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32s = [&](u64 val) {
|
||||
check(val, -(1LL << 31), 1LL << 31);
|
||||
*(ul32 *)loc = val;
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_X86_64_16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ul16 *)loc = S + A;
|
||||
break;
|
||||
case R_X86_64_32:
|
||||
write32(S + A);
|
||||
break;
|
||||
case R_X86_64_32S:
|
||||
write32s(S + A);
|
||||
break;
|
||||
case R_X86_64_64:
|
||||
break;
|
||||
case R_X86_64_PC8:
|
||||
check(S + A - P, -(1 << 7), 1 << 7);
|
||||
*loc = S + A - P;
|
||||
break;
|
||||
case R_X86_64_PC16:
|
||||
check(S + A - P, -(1 << 15), 1 << 15);
|
||||
*(ul16 *)loc = S + A - P;
|
||||
break;
|
||||
case R_X86_64_PC32:
|
||||
case R_X86_64_PLT32:
|
||||
write32s(S + A - P);
|
||||
break;
|
||||
case R_X86_64_PC64:
|
||||
*(ul64 *)loc = S + A - P;
|
||||
break;
|
||||
case R_X86_64_GOT32:
|
||||
write32(G + A);
|
||||
break;
|
||||
case R_X86_64_GOT64:
|
||||
*(ul64 *)loc = G + A;
|
||||
break;
|
||||
case R_X86_64_GOTOFF64:
|
||||
case R_X86_64_PLTOFF64:
|
||||
*(ul64 *)loc = S + A - GOT;
|
||||
break;
|
||||
case R_X86_64_GOTPC32:
|
||||
write32s(GOT + A - P);
|
||||
break;
|
||||
case R_X86_64_GOTPC64:
|
||||
*(ul64 *)loc = GOT + A - P;
|
||||
break;
|
||||
case R_X86_64_GOTPCREL:
|
||||
write32s(G + GOT + A - P);
|
||||
break;
|
||||
case R_X86_64_GOTPCREL64:
|
||||
*(ul64 *)loc = G + GOT + A - P;
|
||||
break;
|
||||
case R_X86_64_GOTPCRELX:
|
||||
case R_X86_64_REX_GOTPCRELX:
|
||||
case R_X86_64_CODE_4_GOTPCRELX:
|
||||
// We always want to relax GOTPCRELX relocs even if --no-relax
|
||||
// was given because some static PIE runtime code depends on these
|
||||
// relaxations.
|
||||
if (sym.is_pcrel_linktime_const(ctx) && is_int(S + A - P, 32)) {
|
||||
if (u32 insn = relax_gotpcrelx(loc, rel)) {
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
*(ul32 *)loc = S + A - P;
|
||||
break;
|
||||
}
|
||||
}
|
||||
write32s(G + GOT + A - P);
|
||||
break;
|
||||
case R_X86_64_TLSGD:
|
||||
if (sym.has_tlsgd(ctx))
|
||||
write32s(sym.get_tlsgd_addr(ctx) + A - P);
|
||||
else if (sym.has_gottp(ctx))
|
||||
relax_gd_to_ie(loc, rels[++i], sym.get_gottp_addr(ctx) - P);
|
||||
else
|
||||
relax_gd_to_le(loc, rels[++i], S - ctx.tp_addr);
|
||||
break;
|
||||
case R_X86_64_TLSLD:
|
||||
if (ctx.got->has_tlsld(ctx))
|
||||
write32s(ctx.got->get_tlsld_addr(ctx) + A - P);
|
||||
else
|
||||
relax_ld_to_le(loc, rels[++i], ctx.tp_addr - ctx.tls_begin);
|
||||
break;
|
||||
case R_X86_64_DTPOFF32:
|
||||
write32s(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_X86_64_DTPOFF64:
|
||||
*(ul64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_X86_64_TPOFF32:
|
||||
write32s(S + A - ctx.tp_addr);
|
||||
break;
|
||||
case R_X86_64_TPOFF64:
|
||||
*(ul64 *)loc = S + A - ctx.tp_addr;
|
||||
break;
|
||||
case R_X86_64_GOTTPOFF:
|
||||
case R_X86_64_CODE_4_GOTTPOFF:
|
||||
if (sym.has_gottp(ctx)) {
|
||||
write32s(sym.get_gottp_addr(ctx) + A - P);
|
||||
} else {
|
||||
u32 insn = relax_gottpoff(loc, rel);
|
||||
loc[-3] = insn >> 16;
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(S - ctx.tp_addr);
|
||||
}
|
||||
break;
|
||||
case R_X86_64_CODE_6_GOTTPOFF:
|
||||
write32s(sym.get_gottp_addr(ctx) + A - P);
|
||||
break;
|
||||
case R_X86_64_GOTPC32_TLSDESC:
|
||||
case R_X86_64_CODE_4_GOTPC32_TLSDESC:
|
||||
// x86-64 TLSDESC uses the following code sequence to materialize
|
||||
// a TP-relative address in %rax.
|
||||
//
|
||||
// lea 0(%rip), %rax
|
||||
// R_X86_64_GOTPC32_TLSDESC foo
|
||||
// call *(%rax)
|
||||
// R_X86_64_TLSDESC_CALL foo
|
||||
//
|
||||
// We may relax the instructions to the following if its TP-relative
|
||||
// address is known at link-time
|
||||
//
|
||||
// mov $foo@TPOFF, %rax
|
||||
// nop
|
||||
//
|
||||
// or to the following if the TP-relative address is known at
|
||||
// process startup time.
|
||||
//
|
||||
// mov foo@GOTTPOFF(%rip), %rax
|
||||
// nop
|
||||
//
|
||||
// We allow the following alternative code sequence too because
|
||||
// LLVM emits such code.
|
||||
//
|
||||
// lea 0(%rip), %reg
|
||||
// R_X86_64_GOTPC32_TLSDESC foo
|
||||
// mov %reg, %rax
|
||||
// call *(%rax)
|
||||
// R_X86_64_TLSDESC_CALL foo
|
||||
if (sym.has_tlsdesc(ctx)) {
|
||||
write32s(sym.get_tlsdesc_addr(ctx) + A - P);
|
||||
} else if (sym.has_gottp(ctx)) {
|
||||
u32 insn = relax_tlsdesc_to_ie(loc, rel);
|
||||
if (!insn)
|
||||
Fatal(ctx) << *this << ": illegal instruction sequence for " << rel;
|
||||
loc[-3] = insn >> 16;
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(sym.get_gottp_addr(ctx) + A - P);
|
||||
} else {
|
||||
u32 insn = relax_tlsdesc_to_le(loc, rel);
|
||||
if (!insn)
|
||||
Fatal(ctx) << *this << ": illegal instruction sequence for " << rel;
|
||||
loc[-3] = insn >> 16;
|
||||
loc[-2] = insn >> 8;
|
||||
loc[-1] = insn;
|
||||
write32s(S - ctx.tp_addr);
|
||||
}
|
||||
break;
|
||||
case R_X86_64_TLSDESC_CALL:
|
||||
if (!sym.has_tlsdesc(ctx)) {
|
||||
// call *(%rax) -> nop
|
||||
loc[0] = 0x66;
|
||||
loc[1] = 0x90;
|
||||
}
|
||||
break;
|
||||
case R_X86_64_SIZE32:
|
||||
write32(sym.esym().st_size + A);
|
||||
break;
|
||||
case R_X86_64_SIZE64:
|
||||
*(ul64 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
default:
|
||||
unreachable();
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// This function is responsible for applying relocations against
|
||||
// non-SHF_ALLOC sections (i.e. sections that are not mapped to memory
|
||||
// at runtime).
|
||||
//
|
||||
// Relocations against non-SHF_ALLOC sections are much easier to
|
||||
// handle than that against SHF_ALLOC sections. It is because, since
|
||||
// they are not mapped to memory, they don't contain any variable or
|
||||
// function and never need PLT or GOT. Non-SHF_ALLOC sections are
|
||||
// mostly debug info sections.
|
||||
//
|
||||
// Relocations against non-SHF_ALLOC sections are not scanned by
|
||||
// scan_relocations.
|
||||
template <>
|
||||
void InputSection<E>::apply_reloc_nonalloc(Context<E> &ctx, u8 *base) {
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = base + rel.r_offset;
|
||||
|
||||
SectionFragment<E> *frag;
|
||||
i64 frag_addend;
|
||||
std::tie(frag, frag_addend) = get_fragment(ctx, rel);
|
||||
|
||||
u64 S = frag ? frag->get_addr(ctx) : sym.get_addr(ctx);
|
||||
u64 A = frag ? frag_addend : (i64)rel.r_addend;
|
||||
|
||||
auto check = [&](i64 val, i64 lo, i64 hi) {
|
||||
check_range(ctx, i, val, lo, hi);
|
||||
};
|
||||
|
||||
auto write32 = [&](u64 val) {
|
||||
check(val, 0, 1LL << 32);
|
||||
*(ul32 *)loc = val;
|
||||
};
|
||||
|
||||
auto write32s = [&](u64 val) {
|
||||
check(val, -(1LL << 31), 1LL << 31);
|
||||
*(ul32 *)loc = val;
|
||||
};
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
check(S + A, 0, 1 << 8);
|
||||
*loc = S + A;
|
||||
break;
|
||||
case R_X86_64_16:
|
||||
check(S + A, 0, 1 << 16);
|
||||
*(ul16 *)loc = S + A;
|
||||
break;
|
||||
case R_X86_64_32:
|
||||
write32(S + A);
|
||||
break;
|
||||
case R_X86_64_32S:
|
||||
write32s(S + A);
|
||||
break;
|
||||
case R_X86_64_64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul64 *)loc = *val;
|
||||
else
|
||||
*(ul64 *)loc = S + A;
|
||||
break;
|
||||
case R_X86_64_DTPOFF32:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul32 *)loc = *val;
|
||||
else
|
||||
write32s(S + A - ctx.dtp_addr);
|
||||
break;
|
||||
case R_X86_64_DTPOFF64:
|
||||
if (std::optional<u64> val = get_tombstone(sym, frag))
|
||||
*(ul64 *)loc = *val;
|
||||
else
|
||||
*(ul64 *)loc = S + A - ctx.dtp_addr;
|
||||
break;
|
||||
case R_X86_64_GOTOFF64:
|
||||
*(ul64 *)loc = S + A - ctx.gotplt->shdr.sh_addr;
|
||||
break;
|
||||
case R_X86_64_GOTPC64:
|
||||
// PC-relative relocation doesn't make sense for non-memory-allocated
|
||||
// section, but GCC 6.3.0 seems to create this reloc for
|
||||
// _GLOBAL_OFFSET_TABLE_.
|
||||
*(ul64 *)loc = ctx.gotplt->shdr.sh_addr + A;
|
||||
break;
|
||||
case R_X86_64_SIZE32:
|
||||
write32(sym.esym().st_size + A);
|
||||
break;
|
||||
case R_X86_64_SIZE64:
|
||||
*(ul64 *)loc = sym.esym().st_size + A;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << *this << ": invalid relocation for non-allocated sections: "
|
||||
<< rel;
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Linker has to create data structures in an output file to apply
|
||||
// some type of relocations. For example, if a relocation refers a GOT
|
||||
// or a PLT entry of a symbol, linker has to create an entry in .got
|
||||
// or in .plt for that symbol. In order to fix the file layout, we
|
||||
// need to scan relocations.
|
||||
template <>
|
||||
void InputSection<E>::scan_relocations(Context<E> &ctx) {
|
||||
assert(shdr().sh_flags & SHF_ALLOC);
|
||||
std::span<const ElfRel<E>> rels = get_rels(ctx);
|
||||
|
||||
// Scan relocations
|
||||
for (i64 i = 0; i < rels.size(); i++) {
|
||||
const ElfRel<E> &rel = rels[i];
|
||||
if (rel.r_type == R_NONE || record_undef_error(ctx, rel))
|
||||
continue;
|
||||
|
||||
Symbol<E> &sym = *file.symbols[rel.r_sym];
|
||||
u8 *loc = (u8 *)(contents.data() + rel.r_offset);
|
||||
|
||||
if (sym.is_ifunc())
|
||||
sym.flags |= NEEDS_GOT | NEEDS_PLT;
|
||||
|
||||
if (rel.r_type == R_X86_64_TLSGD || rel.r_type == R_X86_64_TLSLD) {
|
||||
if (i + 1 == rels.size())
|
||||
Fatal(ctx) << *this << ": " << rel
|
||||
<< " must be followed by PLT or GOTPCREL";
|
||||
|
||||
if (u32 ty = rels[i + 1].r_type;
|
||||
ty != R_X86_64_PLT32 && ty != R_X86_64_PC32 &&
|
||||
ty != R_X86_64_PLTOFF64 && ty != R_X86_64_GOTPCREL &&
|
||||
ty != R_X86_64_GOTPCRELX)
|
||||
Fatal(ctx) << *this << ": " << rel
|
||||
<< " must be followed by PLT or GOTPCREL";
|
||||
}
|
||||
|
||||
switch (rel.r_type) {
|
||||
case R_X86_64_8:
|
||||
case R_X86_64_16:
|
||||
case R_X86_64_32:
|
||||
case R_X86_64_32S:
|
||||
scan_absrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_X86_64_PC8:
|
||||
case R_X86_64_PC16:
|
||||
case R_X86_64_PC32:
|
||||
case R_X86_64_PC64:
|
||||
scan_pcrel(ctx, sym, rel);
|
||||
break;
|
||||
case R_X86_64_GOT32:
|
||||
case R_X86_64_GOT64:
|
||||
case R_X86_64_GOTPC32:
|
||||
case R_X86_64_GOTPC64:
|
||||
case R_X86_64_GOTPCREL:
|
||||
case R_X86_64_GOTPCREL64:
|
||||
case R_X86_64_GOTPCRELX:
|
||||
case R_X86_64_REX_GOTPCRELX:
|
||||
case R_X86_64_CODE_4_GOTPCRELX:
|
||||
sym.flags |= NEEDS_GOT;
|
||||
break;
|
||||
case R_X86_64_PLT32:
|
||||
case R_X86_64_PLTOFF64:
|
||||
if (sym.is_imported)
|
||||
sym.flags |= NEEDS_PLT;
|
||||
break;
|
||||
case R_X86_64_TLSGD:
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && sym.is_tprel_linktime_const(ctx))) {
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
i++;
|
||||
} else if (ctx.arg.relax && sym.is_tprel_runtime_const(ctx)) {
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
i++;
|
||||
} else {
|
||||
sym.flags |= NEEDS_TLSGD;
|
||||
}
|
||||
break;
|
||||
case R_X86_64_TLSLD:
|
||||
// We always relax if -static because libc.a doesn't contain
|
||||
// __tls_get_addr().
|
||||
if (ctx.arg.static_ || (ctx.arg.relax && !ctx.arg.shared))
|
||||
i++;
|
||||
else
|
||||
ctx.needs_tlsld = true;
|
||||
break;
|
||||
case R_X86_64_GOTTPOFF:
|
||||
case R_X86_64_CODE_4_GOTTPOFF:
|
||||
if (!ctx.arg.relax || !sym.is_tprel_linktime_const(ctx) ||
|
||||
!relax_gottpoff(loc, rel))
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_X86_64_CODE_6_GOTTPOFF:
|
||||
sym.flags |= NEEDS_GOTTP;
|
||||
break;
|
||||
case R_X86_64_TLSDESC_CALL:
|
||||
scan_tlsdesc(ctx, sym);
|
||||
break;
|
||||
case R_X86_64_TPOFF32:
|
||||
case R_X86_64_TPOFF64:
|
||||
check_tlsle(ctx, sym, rel);
|
||||
break;
|
||||
case R_X86_64_64:
|
||||
case R_X86_64_GOTOFF64:
|
||||
case R_X86_64_DTPOFF32:
|
||||
case R_X86_64_DTPOFF64:
|
||||
case R_X86_64_SIZE32:
|
||||
case R_X86_64_SIZE64:
|
||||
case R_X86_64_GOTPC32_TLSDESC:
|
||||
case R_X86_64_CODE_4_GOTPC32_TLSDESC:
|
||||
break;
|
||||
default:
|
||||
Error(ctx) << *this << ": unknown relocation: " << rel;
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Intel CET is a relatively new CPU feature to enhance security by
|
||||
// protecting control flow integrity. If the feature is enabled, indirect
|
||||
// branches (i.e. branch instructions that take a register instead of an
|
||||
// immediate) must land on a "landing pad" instruction, or a CPU-level fault
|
||||
// will raise. That prevents an attacker to branch to a middle of a random
|
||||
// function, making ROP or JOP much harder to conduct.
|
||||
//
|
||||
// On x86-64, the landing pad instruction is ENDBR64. That is actually a
|
||||
// repurposed NOP instruction to provide binary compatibility with older
|
||||
// hardware that doesn't support CET.
|
||||
//
|
||||
// The problem here is that the compiler always emits a landing pad at the
|
||||
// beginning fo a global function because it doesn't know whether or not the
|
||||
// function's address is taken in other translation units. As a result, the
|
||||
// resulting binary contains more landing pads than necessary.
|
||||
//
|
||||
// This function rewrites a landing pad with a nop if the function's address
|
||||
// was not actually taken. We can do what the compiler cannot because we
|
||||
// know about all translation units.
|
||||
void rewrite_endbr(Context<E> &ctx) {
|
||||
Timer t(ctx, "rewrite_endbr");
|
||||
|
||||
constexpr u8 endbr64[] = {0xf3, 0x0f, 0x1e, 0xfa};
|
||||
constexpr u8 nop[] = {0x0f, 0x1f, 0x40, 0x00};
|
||||
|
||||
// Rewrite all endbr64 instructions referred to by function symbols with
|
||||
// NOPs. We handle only global symbols because the compiler doesn't emit
|
||||
// an endbr64 for a file-scoped function in the first place if its address
|
||||
// is not taken within the file.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->get_global_syms()) {
|
||||
if (sym->file == file && sym->esym().st_type == STT_FUNC) {
|
||||
if (InputSection<E> *isec = sym->get_input_section();
|
||||
isec && (isec->shdr().sh_flags & SHF_EXECINSTR)) {
|
||||
if (OutputSection<E> *osec = isec->output_section) {
|
||||
u8 *buf = ctx.buf + osec->shdr.sh_offset + isec->offset + sym->value;
|
||||
if (memcmp(buf, endbr64, 4) == 0)
|
||||
memcpy(buf, nop, 4);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
auto write_back = [&](InputSection<E> *isec, i64 offset) {
|
||||
// If isec has an endbr64 at a given offset, copy that instruction to
|
||||
// the output buffer, possibly overwriting a nop written in the above
|
||||
// loop.
|
||||
if (isec && isec->output_section &&
|
||||
(isec->shdr().sh_flags & SHF_EXECINSTR) &&
|
||||
0 <= offset && offset <= isec->contents.size() - 4 &&
|
||||
memcmp(isec->contents.data() + offset, endbr64, 4) == 0)
|
||||
memcpy(ctx.buf + isec->output_section->shdr.sh_offset + isec->offset + offset,
|
||||
endbr64, 4);
|
||||
};
|
||||
|
||||
// Write back endbr64 instructions if they are referred to by address-taking
|
||||
// relocations.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (isec && isec->is_alive && (isec->shdr().sh_flags & SHF_ALLOC)) {
|
||||
for (const ElfRel<E> &rel : isec->get_rels(ctx)) {
|
||||
if (!is_func_call_rel(rel)) {
|
||||
Symbol<E> *sym = file->symbols[rel.r_sym];
|
||||
if (sym->esym().st_type == STT_SECTION)
|
||||
write_back(sym->get_input_section(), rel.r_addend);
|
||||
else
|
||||
write_back(sym->get_input_section(), sym->value);
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
// We record addresses of some symbols in the ELF header, .dynamic or in
|
||||
// .dynsym. We need to retain endbr64s for such symbols.
|
||||
auto keep = [&](Symbol<E> *sym) {
|
||||
if (sym)
|
||||
write_back(sym->get_input_section(), sym->value);
|
||||
};
|
||||
|
||||
keep(ctx.arg.entry);
|
||||
keep(ctx.arg.init);
|
||||
keep(ctx.arg.fini);
|
||||
|
||||
if (ctx.dynsym)
|
||||
for (Symbol<E> *sym : ctx.dynsym->symbols)
|
||||
if (sym && sym->is_exported)
|
||||
keep(sym);
|
||||
}
|
||||
|
||||
} // namespace mold
|
||||
|
||||
#endif
|
||||
@@ -1,188 +0,0 @@
|
||||
// This file contains functions to read an archive file (.a file).
|
||||
// An archive file is just a bundle of object files. It's similar to
|
||||
// tar or zip, but the contents are not compressed.
|
||||
//
|
||||
// An archive file is either "regular" or "thin". A regular archive
|
||||
// contains object files directly, while a thin archive contains only
|
||||
// pathnames. In the latter case, actual file contents have to be read
|
||||
// from given pathnames. A regular archive is sometimes called "fat"
|
||||
// archive as opposed to "thin".
|
||||
//
|
||||
// If an archive file is given to the linker, the linker pulls out
|
||||
// object files that are needed to resolve undefined symbols. So,
|
||||
// bunding object files as an archive and giving that archive to the
|
||||
// linker has a different meaning than directly giving the same set of
|
||||
// object files to the linker. The former links only needed object
|
||||
// files, while the latter links all the given object files.
|
||||
//
|
||||
// Therefore, if you link libc.a for example, not all the libc
|
||||
// functions are linked to your binary. Instead, only object files
|
||||
// that provides functions and variables used in your program get
|
||||
// linked. To make this efficient, static library functions are
|
||||
// usually separated to each object file in an archive file. You can
|
||||
// see the contents of libc.a by running `ar t
|
||||
// /usr/lib/x86_64-linux-gnu/libc.a`.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
namespace {
|
||||
struct ArHdr {
|
||||
char ar_name[16];
|
||||
char ar_date[12];
|
||||
char ar_uid[6];
|
||||
char ar_gid[6];
|
||||
char ar_mode[8];
|
||||
char ar_size[10];
|
||||
char ar_fmag[2];
|
||||
|
||||
bool starts_with(std::string_view s) const {
|
||||
return std::string_view(ar_name, s.size()) == s;
|
||||
}
|
||||
|
||||
bool is_strtab() const {
|
||||
return starts_with("// ");
|
||||
}
|
||||
|
||||
bool is_symtab() const {
|
||||
return starts_with("/ ") || starts_with("/SYM64/ ");
|
||||
}
|
||||
|
||||
std::string read_name(std::string_view strtab, u8 *&ptr) const {
|
||||
// BSD-style long filename
|
||||
if (starts_with("#1/")) {
|
||||
int namelen = atoi(ar_name + 3);
|
||||
std::string name{(char *)ptr, (size_t)namelen};
|
||||
ptr += namelen;
|
||||
|
||||
if (size_t pos = name.find('\0'))
|
||||
name = name.substr(0, pos);
|
||||
return name;
|
||||
}
|
||||
|
||||
// SysV-style long filename
|
||||
if (starts_with("/")) {
|
||||
const char *start = strtab.data() + atoi(ar_name + 1);
|
||||
return {start, (const char *)strstr(start, "/\n")};
|
||||
}
|
||||
|
||||
// Short fileanme
|
||||
if (const char *end = (char *)memchr(ar_name, '/', sizeof(ar_name)))
|
||||
return {ar_name, end};
|
||||
return {ar_name, sizeof(ar_name)};
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<MappedFile *>
|
||||
read_thin_archive_members(Context<E> &ctx, MappedFile *mf) {
|
||||
u8 *begin = mf->data;
|
||||
u8 *data = begin + 8;
|
||||
std::vector<MappedFile *> vec;
|
||||
std::string_view strtab;
|
||||
|
||||
while (data < begin + mf->size) {
|
||||
// Each header is aligned to a 2 byte boundary.
|
||||
if ((begin - data) % 2)
|
||||
data++;
|
||||
|
||||
ArHdr &hdr = *(ArHdr *)data;
|
||||
u8 *body = data + sizeof(hdr);
|
||||
u64 size = atol(hdr.ar_size);
|
||||
|
||||
// Read a string table.
|
||||
if (hdr.is_strtab()) {
|
||||
strtab = {(char *)body, (size_t)size};
|
||||
data = body + size;
|
||||
continue;
|
||||
}
|
||||
|
||||
// Skip a symbol table.
|
||||
if (hdr.is_symtab()) {
|
||||
data = body + size;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (!hdr.starts_with("#1/") && !hdr.starts_with("/"))
|
||||
Fatal(ctx) << mf->name << ": filename is not stored as a long filename";
|
||||
|
||||
std::string name = hdr.read_name(strtab, body);
|
||||
|
||||
// Skip if symbol table
|
||||
if (name == "__.SYMDEF" || name == "__.SYMDEF SORTED")
|
||||
continue;
|
||||
|
||||
std::string path = name.starts_with('/') ?
|
||||
name : (path_dirname(mf->name) / name).string();
|
||||
vec.push_back(must_open_file(ctx, path));
|
||||
vec.back()->thin_parent = mf;
|
||||
data = body;
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<MappedFile *>
|
||||
read_fat_archive_members(Context<E> &ctx, MappedFile *mf) {
|
||||
u8 *begin = mf->data;
|
||||
u8 *data = begin + 8;
|
||||
std::vector<MappedFile *> vec;
|
||||
std::string_view strtab;
|
||||
|
||||
while (begin + mf->size - data >= 2) {
|
||||
if ((begin - data) % 2)
|
||||
data++;
|
||||
|
||||
ArHdr &hdr = *(ArHdr *)data;
|
||||
u8 *body = data + sizeof(hdr);
|
||||
u64 size = atol(hdr.ar_size);
|
||||
data = body + size;
|
||||
|
||||
// Read if string table
|
||||
if (hdr.is_strtab()) {
|
||||
strtab = {(char *)body, (size_t)size};
|
||||
continue;
|
||||
}
|
||||
|
||||
// Skip if symbol table
|
||||
if (hdr.is_symtab())
|
||||
continue;
|
||||
|
||||
// Read the name field
|
||||
std::string name = hdr.read_name(strtab, body);
|
||||
|
||||
// Skip if symbol table
|
||||
if (name == "__.SYMDEF" || name == "__.SYMDEF SORTED")
|
||||
continue;
|
||||
|
||||
vec.push_back(mf->slice(ctx, name, body - begin, data - body));
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::vector<MappedFile *>
|
||||
read_archive_members(Context<E> &ctx, MappedFile *mf) {
|
||||
std::string_view str = mf->get_contents();
|
||||
|
||||
if (str.starts_with("!<arch>\n"))
|
||||
return read_fat_archive_members(ctx, mf);
|
||||
|
||||
assert(str.starts_with("!<thin>\n"));
|
||||
return read_thin_archive_members(ctx, mf);
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template std::vector<MappedFile *>
|
||||
read_thin_archive_members(Context<E> &, MappedFile *);
|
||||
|
||||
template std::vector<MappedFile *>
|
||||
read_fat_archive_members(Context<E> &, MappedFile *);
|
||||
|
||||
template std::vector<MappedFile *>
|
||||
read_archive_members(Context<E> &, MappedFile *);
|
||||
|
||||
} // namespace mold
|
||||
-1631
File diff suppressed because it is too large.
Load diff
-1048
File diff suppressed because it is too large.
Load diff
@@ -1,32 +0,0 @@
|
||||
#include "config.h"
|
||||
#include "mold.h"
|
||||
#include "mold-git-hash.h"
|
||||
|
||||
#if MOLD_USE_MIMALLOC
|
||||
# include <mimalloc.h>
|
||||
#endif
|
||||
|
||||
#if MOLD_USE_SYSTEM_MIMALLOC
|
||||
// Including mimalloc-new-delete.h overrides the new/delete operators.
|
||||
// We need it only when using mimalloc as a dynamic library.
|
||||
// This header should be included in only one source file, so we do
|
||||
// it in this file.
|
||||
# include <mimalloc-new-delete.h>
|
||||
#endif
|
||||
|
||||
std::string mold::mold_version =
|
||||
#ifdef MOLD_GIT_HASH
|
||||
"mold " MOLD_VERSION " (" MOLD_GIT_HASH "; compatible with GNU ld)";
|
||||
#else
|
||||
"mold " MOLD_VERSION " (compatible with GNU ld)";
|
||||
#endif
|
||||
|
||||
int main(int argc, char **argv) {
|
||||
#if MOLD_USE_MIMALLOC
|
||||
// Silence mimalloc warnings that users can ignore
|
||||
mi_option_disable(mi_option_verbose);
|
||||
mi_option_disable(mi_option_show_errors);
|
||||
#endif
|
||||
|
||||
return mold::mold_main<mold::MOLD_FIRST_TARGET>(argc, argv);
|
||||
}
|
||||
@@ -1,55 +0,0 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
static std::string_view fatal_mono = "mold: fatal: ";
|
||||
static std::string_view fatal_color = "mold: \033[0;1;31mfatal:\033[0m ";
|
||||
static std::string_view error_mono = "mold: error: ";
|
||||
static std::string_view error_color = "mold: \033[0;1;31merror:\033[0m ";
|
||||
static std::string_view warning_mono = "mold: warning: ";
|
||||
static std::string_view warning_color = "mold: \033[0;1;35mwarning:\033[0m ";
|
||||
|
||||
template <typename E>
|
||||
Fatal<E>::Fatal(Context<E> &ctx) {
|
||||
out << (ctx.arg.color_diagnostics ? fatal_color : fatal_mono);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
[[noreturn]] Fatal<E>::~Fatal() {
|
||||
out.emit();
|
||||
cleanup();
|
||||
_exit(1);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
Error<E>::Error(Context<E> &ctx) {
|
||||
if (ctx.arg.noinhibit_exec) {
|
||||
out << (ctx.arg.color_diagnostics ? warning_color : warning_mono);
|
||||
} else {
|
||||
out << (ctx.arg.color_diagnostics ? error_color : error_mono);
|
||||
ctx.has_error = true;
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
Warn<E>::Warn(Context<E> &ctx) {
|
||||
if (ctx.arg.suppress_warnings)
|
||||
return;
|
||||
|
||||
out.emplace(std::cerr);
|
||||
|
||||
if (ctx.arg.fatal_warnings) {
|
||||
*out << (ctx.arg.color_diagnostics ? error_color : error_mono);
|
||||
ctx.has_error = true;
|
||||
} else {
|
||||
*out << (ctx.arg.color_diagnostics ? warning_color : warning_mono);
|
||||
}
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template class Fatal<E>;
|
||||
template class Error<E>;
|
||||
template class Warn<E>;
|
||||
|
||||
} // namespace mold
|
||||
-207
@@ -1,207 +0,0 @@
|
||||
#include "mold.h"
|
||||
|
||||
namespace mold {
|
||||
|
||||
static bool is_text_file(MappedFile *mf) {
|
||||
auto istext = [](char c) {
|
||||
return isprint(c) || c == '\n' || c == '\t';
|
||||
};
|
||||
|
||||
u8 *data = mf->data;
|
||||
return mf->size >= 4 && istext(data[0]) && istext(data[1]) &&
|
||||
istext(data[2]) && istext(data[3]);
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static bool is_gcc_lto_obj(MappedFile *mf, bool has_gcc_plugin) {
|
||||
const char *data = mf->get_contents().data();
|
||||
ElfEhdr<E> &ehdr = *(ElfEhdr<E> *)data;
|
||||
ElfShdr<E> *sh_begin = (ElfShdr<E> *)(data + ehdr.e_shoff);
|
||||
std::span<ElfShdr<E>> shdrs{(ElfShdr<E> *)(data + ehdr.e_shoff), ehdr.e_shnum};
|
||||
|
||||
// e_shstrndx is a 16-bit field. If .shstrtab's section index is
|
||||
// too large, the actual number is stored to sh_link field.
|
||||
i64 shstrtab_idx = (ehdr.e_shstrndx == SHN_XINDEX)
|
||||
? sh_begin->sh_link : ehdr.e_shstrndx;
|
||||
|
||||
for (ElfShdr<E> &sec : shdrs) {
|
||||
// GCC FAT LTO objects contain both regular ELF sections and GCC-
|
||||
// specific LTO sections, so that they can be linked as LTO objects if
|
||||
// the LTO linker plugin is available and falls back as regular
|
||||
// objects otherwise. GCC FAT LTO object can be identified by the
|
||||
// presence of `.gcc.lto_.symtab` section.
|
||||
if (has_gcc_plugin) {
|
||||
std::string_view name = data + shdrs[shstrtab_idx].sh_offset + sec.sh_name;
|
||||
if (name.starts_with(".gnu.lto_.symtab."))
|
||||
return true;
|
||||
}
|
||||
|
||||
if (sec.sh_type != SHT_SYMTAB)
|
||||
continue;
|
||||
|
||||
// GCC non-FAT LTO object contains only sections symbols followed by
|
||||
// a common symbol whose name is `__gnu_lto_slim` (or `__gnu_lto_v1`
|
||||
// for older GCC releases).
|
||||
std::span<ElfSym<E>> elf_syms{(ElfSym<E> *)(data + sec.sh_offset),
|
||||
(size_t)sec.sh_size / sizeof(ElfSym<E>)};
|
||||
|
||||
auto skip = [](u8 type) {
|
||||
return type == STT_NOTYPE || type == STT_FILE || type == STT_SECTION;
|
||||
};
|
||||
|
||||
i64 i = 1;
|
||||
while (i < elf_syms.size() && skip(elf_syms[i].st_type))
|
||||
i++;
|
||||
|
||||
if (i < elf_syms.size() && elf_syms[i].st_shndx == SHN_COMMON) {
|
||||
std::string_view name =
|
||||
data + shdrs[sec.sh_link].sh_offset + elf_syms[i].st_name;
|
||||
if (name.starts_with("__gnu_lto_"))
|
||||
return true;
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
return false;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
FileType get_file_type(Context<E> &ctx, MappedFile *mf) {
|
||||
std::string_view data = mf->get_contents();
|
||||
if (data.empty())
|
||||
return FileType::EMPTY;
|
||||
|
||||
// GCC FAT LTO objects can be linked as regular ELF objects. If the active
|
||||
// plugin is LLVM's, treat them as regular objects so that we can fall back
|
||||
// to native code instead of routing them through GCC LTO handling.
|
||||
bool has_gcc_plugin = !ctx.arg.plugin.empty() &&
|
||||
ctx.arg.plugin.find("LLVMgold.") == std::string_view::npos;
|
||||
|
||||
if (data.starts_with("\177ELF")) {
|
||||
u8 byte_order = ((ElfEhdr<I386> *)data.data())->e_ident[EI_DATA];
|
||||
|
||||
if (byte_order == ELFDATA2LSB) {
|
||||
auto &ehdr = *(ElfEhdr<I386> *)data.data();
|
||||
|
||||
if (ehdr.e_type == ET_REL) {
|
||||
if (ehdr.e_ident[EI_CLASS] == ELFCLASS32) {
|
||||
if (is_gcc_lto_obj<I386>(mf, has_gcc_plugin))
|
||||
return FileType::GCC_LTO_OBJ;
|
||||
} else {
|
||||
if (is_gcc_lto_obj<X86_64>(mf, has_gcc_plugin))
|
||||
return FileType::GCC_LTO_OBJ;
|
||||
}
|
||||
return FileType::ELF_OBJ;
|
||||
}
|
||||
|
||||
if (ehdr.e_type == ET_DYN)
|
||||
return FileType::ELF_DSO;
|
||||
} else {
|
||||
auto &ehdr = *(ElfEhdr<M68K> *)data.data();
|
||||
|
||||
if (ehdr.e_type == ET_REL) {
|
||||
if (ehdr.e_ident[EI_CLASS] == ELFCLASS32) {
|
||||
if (is_gcc_lto_obj<M68K>(mf, has_gcc_plugin))
|
||||
return FileType::GCC_LTO_OBJ;
|
||||
} else {
|
||||
if (is_gcc_lto_obj<SPARC64>(mf, has_gcc_plugin))
|
||||
return FileType::GCC_LTO_OBJ;
|
||||
}
|
||||
return FileType::ELF_OBJ;
|
||||
}
|
||||
|
||||
if (ehdr.e_type == ET_DYN)
|
||||
return FileType::ELF_DSO;
|
||||
}
|
||||
return FileType::UNKNOWN;
|
||||
}
|
||||
|
||||
if (data.starts_with("!<arch>\n"))
|
||||
return FileType::AR;
|
||||
if (data.starts_with("!<thin>\n"))
|
||||
return FileType::THIN_AR;
|
||||
if (is_text_file(mf))
|
||||
return FileType::TEXT;
|
||||
if (data.starts_with("\xde\xc0\x17\x0b"))
|
||||
return FileType::LLVM_BITCODE;
|
||||
if (data.starts_with("BC\xc0\xde"))
|
||||
return FileType::LLVM_BITCODE;
|
||||
return FileType::UNKNOWN;
|
||||
}
|
||||
|
||||
static std::string_view get_elf_type(u8 *buf) {
|
||||
bool is_le = (buf[EI_DATA] == ELFDATA2LSB);
|
||||
bool is_64 = (buf[EI_CLASS] == ELFCLASS64);
|
||||
|
||||
auto *ehdr_le = (ElfEhdr<I386> *)buf;
|
||||
auto *ehdr_be = (ElfEhdr<M68K> *)buf;
|
||||
|
||||
switch (is_le ? ehdr_le->e_machine : ehdr_be->e_machine) {
|
||||
case EM_386:
|
||||
return I386::name;
|
||||
case EM_X86_64:
|
||||
return X86_64::name;
|
||||
case EM_ARM:
|
||||
return is_le ? ARM32LE::name : ARM32BE::name;
|
||||
case EM_AARCH64:
|
||||
return is_le ? ARM64LE::name : ARM64BE::name;
|
||||
case EM_RISCV:
|
||||
if (is_le)
|
||||
return is_64 ? RV64LE::name : RV32LE::name;
|
||||
return is_64 ? RV64BE::name : RV32BE::name;
|
||||
case EM_PPC:
|
||||
return PPC32::name;
|
||||
case EM_PPC64:
|
||||
return is_le ? PPC64V2::name : PPC64V1::name;
|
||||
case EM_S390X:
|
||||
return S390X::name;
|
||||
case EM_SPARC64:
|
||||
return SPARC64::name;
|
||||
case EM_68K:
|
||||
return M68K::name;
|
||||
case EM_SH:
|
||||
return is_le ? SH4LE::name : SH4BE::name;
|
||||
case EM_LOONGARCH:
|
||||
return is_64 ? LOONGARCH64::name : LOONGARCH32::name;
|
||||
default:
|
||||
return "";
|
||||
}
|
||||
}
|
||||
|
||||
// Read the beginning of a given file and returns its machine type
|
||||
// (e.g. EM_X86_64 or EM_386).
|
||||
template <typename E>
|
||||
std::string_view
|
||||
get_machine_type(Context<E> &ctx, ReaderContext &rctx, MappedFile *mf) {
|
||||
switch (get_file_type(ctx, mf)) {
|
||||
case FileType::ELF_OBJ:
|
||||
case FileType::ELF_DSO:
|
||||
case FileType::GCC_LTO_OBJ:
|
||||
return get_elf_type(mf->data);
|
||||
case FileType::AR:
|
||||
for (MappedFile *child : read_fat_archive_members(ctx, mf))
|
||||
if (FileType ty = get_file_type(ctx, child);
|
||||
ty == FileType::ELF_OBJ || ty == FileType::GCC_LTO_OBJ)
|
||||
return get_elf_type(child->data);
|
||||
return "";
|
||||
case FileType::THIN_AR:
|
||||
for (MappedFile *child : read_thin_archive_members(ctx, mf))
|
||||
if (FileType ty = get_file_type(ctx, child);
|
||||
ty == FileType::ELF_OBJ || ty == FileType::GCC_LTO_OBJ)
|
||||
return get_elf_type(child->data);
|
||||
return "";
|
||||
case FileType::TEXT:
|
||||
return Script(ctx, mf, &rctx).get_script_output_type();
|
||||
default:
|
||||
return "";
|
||||
}
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template FileType get_file_type(Context<E> &, MappedFile *);
|
||||
|
||||
template std::string_view
|
||||
get_machine_type(Context<E> &, ReaderContext &, MappedFile *);
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,249 +0,0 @@
|
||||
// This file implements a mark-sweep garbage collector for -gc-sections.
|
||||
// In this algorithm, vertices are sections and edges are relocations.
|
||||
// Any section that is reachable from a root section is considered alive.
|
||||
|
||||
#include "mold.h"
|
||||
|
||||
#include <fstream>
|
||||
#include <tbb/concurrent_unordered_map.h>
|
||||
#include <tbb/concurrent_vector.h>
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
template <typename E>
|
||||
static bool should_keep(Context<E> &ctx, const InputSection<E> &isec) {
|
||||
u32 type = isec.shdr().sh_type;
|
||||
u32 flags = isec.shdr().sh_flags;
|
||||
if constexpr (is_ppc32<E>)
|
||||
if (isec.name == ".got2")
|
||||
return true;
|
||||
|
||||
// KEEP'ed by a linker script
|
||||
if (!isec.file.script_match.empty()) {
|
||||
i32 rank = isec.file.script_match[isec.shndx];
|
||||
if (rank != -1 && ctx.script_ranks[rank].keep)
|
||||
return true;
|
||||
}
|
||||
|
||||
return (flags & SHF_GNU_RETAIN) ||
|
||||
type == SHT_NOTE ||
|
||||
type == SHT_INIT_ARRAY ||
|
||||
type == SHT_FINI_ARRAY ||
|
||||
type == SHT_PREINIT_ARRAY ||
|
||||
isec.name.starts_with(".ctors") ||
|
||||
isec.name.starts_with(".dtors") ||
|
||||
isec.name.starts_with(".init") ||
|
||||
isec.name.starts_with(".fini");
|
||||
}
|
||||
|
||||
// Sections whose names are valid C identifiers can be referenced via
|
||||
// __start_<name>/__stop_<name> symbols, which the linker synthesizes.
|
||||
// Such sections must be kept alive only if such a marker symbol is
|
||||
// referenced from a live section. This map lets us find all sections
|
||||
// of a given name when we encounter such a reference during marking.
|
||||
template <typename E>
|
||||
using StartStopMap =
|
||||
tbb::concurrent_unordered_multimap<std::string_view, InputSection<E> *>;
|
||||
|
||||
template <typename E>
|
||||
static StartStopMap<E> build_start_stop_map(Context<E> &ctx) {
|
||||
StartStopMap<E> map;
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections)
|
||||
if (isec && isec->is_alive && (isec->shdr().sh_flags & SHF_ALLOC) &&
|
||||
is_c_identifier(isec->name))
|
||||
map.insert({isec->name, isec.get()});
|
||||
});
|
||||
return map;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static bool mark_section(InputSection<E> *isec) {
|
||||
return isec && isec->is_alive && !isec->is_visited.test_and_set();
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static tbb::concurrent_vector<InputSection<E> *>
|
||||
collect_root_set(Context<E> &ctx) {
|
||||
Timer t(ctx, "collect_root_set");
|
||||
tbb::concurrent_vector<InputSection<E> *> rootset;
|
||||
|
||||
auto enqueue_section = [&](InputSection<E> *isec) {
|
||||
if (mark_section(isec))
|
||||
rootset.push_back(isec);
|
||||
};
|
||||
|
||||
auto enqueue_symbol = [&](Symbol<E> *sym) {
|
||||
if (sym) {
|
||||
if (SectionFragment<E> *frag = sym->get_frag())
|
||||
frag->is_alive = true;
|
||||
else
|
||||
enqueue_section(sym->get_input_section());
|
||||
}
|
||||
};
|
||||
|
||||
// Add sections that are not subject to garbage collection.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file->sections) {
|
||||
if (!isec || !isec->is_alive)
|
||||
continue;
|
||||
|
||||
// --gc-sections discards only SHF_ALLOC sections. If you want to
|
||||
// reduce the amount of non-memory-mapped segments, you should
|
||||
// use `strip` command, compile without debug info or use
|
||||
// --strip-all linker option.
|
||||
if (!(isec->shdr().sh_flags & SHF_ALLOC)) {
|
||||
isec->is_visited = true;
|
||||
continue;
|
||||
}
|
||||
|
||||
if (should_keep(ctx, *isec))
|
||||
enqueue_section(isec.get());
|
||||
}
|
||||
});
|
||||
|
||||
// Add sections containing gc root or exported symbols
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (Symbol<E> *sym : file->symbols)
|
||||
if (sym->file == file && (sym->gc_root || sym->is_exported))
|
||||
enqueue_symbol(sym);
|
||||
});
|
||||
|
||||
// .eh_frame consists of variable-length records called CIE and FDE
|
||||
// records, and they are a unit of inclusion or exclusion.
|
||||
// We just keep all CIEs and everything that are referenced by them.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
for (CieRecord<E> &cie : file->cies)
|
||||
for (const ElfRel<E> &rel : cie.get_rels())
|
||||
enqueue_symbol(file->symbols[rel.r_sym]);
|
||||
});
|
||||
|
||||
return rootset;
|
||||
}
|
||||
|
||||
static std::string_view start_stop_name(std::string_view sym) {
|
||||
if (sym.starts_with("__start_"))
|
||||
return sym.substr(8);
|
||||
if (sym.starts_with("__stop_"))
|
||||
return sym.substr(7);
|
||||
return "";
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static void visit_section(Context<E> &ctx, InputSection<E> *isec,
|
||||
tbb::feeder<InputSection<E> *> &feeder, i64 depth,
|
||||
const StartStopMap<E> &start_stop_map) {
|
||||
assert(isec->is_visited);
|
||||
|
||||
// Mark a section alive. For better performacne, we don't call
|
||||
// `feeder.add` too often.
|
||||
auto mark = [&](InputSection<E> *sec) {
|
||||
if (mark_section(sec)) {
|
||||
if (depth < 3)
|
||||
visit_section(ctx, sec, feeder, depth + 1, start_stop_map);
|
||||
else
|
||||
feeder.add(sec);
|
||||
}
|
||||
};
|
||||
|
||||
// If this is a text section, .eh_frame may contain records
|
||||
// describing how to handle exceptions for that function.
|
||||
// We want to keep associated .eh_frame records.
|
||||
for (FdeRecord<E> &fde : isec->get_fdes())
|
||||
for (const ElfRel<E> &rel : fde.get_rels(isec->file).subspan(1))
|
||||
if (Symbol<E> *sym = isec->file.symbols[rel.r_sym])
|
||||
mark(sym->get_input_section());
|
||||
|
||||
for (const ElfRel<E> &rel : isec->get_rels(ctx)) {
|
||||
// Symbol can refer to either a section fragment or an input section.
|
||||
Symbol<E> &sym = *isec->file.symbols[rel.r_sym];
|
||||
if (SectionFragment<E> *frag = sym.get_frag()) {
|
||||
frag->is_alive = true;
|
||||
continue;
|
||||
}
|
||||
|
||||
mark(sym.get_input_section());
|
||||
|
||||
// A reference to __start_<name> or __stop_<name> keeps every section
|
||||
// named <name> alive, mirroring how those symbols are defined.
|
||||
if (std::string_view sec = start_stop_name(sym.name());
|
||||
!sec.empty()) {
|
||||
auto [i, end] = start_stop_map.equal_range(sec);
|
||||
for (; i != end; ++i)
|
||||
mark(i->second);
|
||||
}
|
||||
}
|
||||
|
||||
if constexpr (is_arm32<E>)
|
||||
mark(isec->extra.exidx);
|
||||
}
|
||||
|
||||
// Mark all reachable sections
|
||||
template <typename E>
|
||||
static void mark(Context<E> &ctx,
|
||||
tbb::concurrent_vector<InputSection<E> *> &rootset,
|
||||
const StartStopMap<E> &start_stop_map) {
|
||||
Timer t(ctx, "mark");
|
||||
|
||||
tbb::parallel_for_each(rootset, [&](InputSection<E> *isec,
|
||||
tbb::feeder<InputSection<E> *> &feeder) {
|
||||
visit_section(ctx, isec, feeder, 0, start_stop_map);
|
||||
});
|
||||
}
|
||||
|
||||
// Remove unreachable sections
|
||||
template <typename E>
|
||||
static void sweep(Context<E> &ctx) {
|
||||
Timer t(ctx, "sweep");
|
||||
|
||||
std::vector<std::vector<InputSection<E> *>> sections(ctx.objs.size());
|
||||
|
||||
tbb::parallel_for((i64)0, (i64)ctx.objs.size(), [&](i64 i) {
|
||||
ObjectFile<E> &file = *ctx.objs[i];
|
||||
|
||||
for (std::unique_ptr<InputSection<E>> &isec : file.sections) {
|
||||
if (isec && isec->is_alive && !isec->is_visited) {
|
||||
isec->kill();
|
||||
sections[i].push_back(isec.get());
|
||||
}
|
||||
}
|
||||
});
|
||||
|
||||
std::string &path = ctx.arg.print_gc_sections;
|
||||
|
||||
if (!path.empty()) {
|
||||
std::ostream *out = &std::cout;
|
||||
std::ofstream file;
|
||||
|
||||
if (path != "-") {
|
||||
file.open(path);
|
||||
if (file.fail())
|
||||
Fatal(ctx) << "--print-gc-sections: cannot open " << path << ": "
|
||||
<< errno_string();
|
||||
out = &file;
|
||||
}
|
||||
|
||||
i64 saved_bytes = 0;
|
||||
for (std::span<InputSection<E> *> vec : sections)
|
||||
for (InputSection<E> *isec : vec)
|
||||
*out << "removing unused section " << *isec << '\n';
|
||||
|
||||
*out << "GC saved " << saved_bytes << " bytes\n";
|
||||
}
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void gc_sections(Context<E> &ctx) {
|
||||
Timer t(ctx, "gc");
|
||||
tbb::concurrent_vector<InputSection<E> *> rootset = collect_root_set(ctx);
|
||||
StartStopMap<E> start_stop_map = build_start_stop_map(ctx);
|
||||
mark(ctx, rootset, start_stop_map);
|
||||
sweep(ctx);
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template void gc_sections(Context<E> &ctx);
|
||||
|
||||
} // namespace mold
|
||||
@@ -1,808 +0,0 @@
|
||||
// This file contains code to read DWARF debug info to create .gdb_index.
|
||||
//
|
||||
// .gdb_index is an optional section to speed up GNU debugger. It contains
|
||||
// two maps: 1) a map from function/variable/type names to compunits, and
|
||||
// 2) a map from function address ranges to compunits. gdb uses these
|
||||
// maps to quickly find a compunit given a name or an instruction pointer.
|
||||
//
|
||||
// (Terminology: a compilation unit, often abbreviated as compunit or
|
||||
// CU, is a unit of debug info. An input .debug_info section usually
|
||||
// contains one compunit, and thus an output .debug_info contains as
|
||||
// many compunits as the number of input files.)
|
||||
//
|
||||
// .gdb_index is not mandatory. All the information in .gdb_index is
|
||||
// also in other debug info sections. You can actually create an
|
||||
// executable without .gdb_index and later add it using the
|
||||
// `gdb-add-index` post-processing tool that comes with gdb.
|
||||
//
|
||||
// Post-relocated debug section contents are needed to create a
|
||||
// .gdb_index. Therefore, we create it after relocating all the other
|
||||
// sections. The size of the section is also hard to estimate before
|
||||
// applying relocations to debug info sections, so a .gdb_index is
|
||||
// placed at the very end of the output file, even after the section
|
||||
// header.
|
||||
//
|
||||
// The mapping from names to compunits is 1:n while the mapping from
|
||||
// address ranges to compunits is 1:1. That is, two object files may
|
||||
// define the same type name, while there should be no two functions
|
||||
// that overlap with each other in memory.
|
||||
//
|
||||
// .gdb_index contains an on-disk hash table for names, so gdb can
|
||||
// lookup names without loading all strings into memory and construct an
|
||||
// in-memory hash table.
|
||||
//
|
||||
// Names are in .debug_gnu_pubnames and .debug_gnu_pubtypes input
|
||||
// sections. These sections are created if `-ggnu-pubnames` is given.
|
||||
// Besides names, these sections contain attributes for each name so
|
||||
// that gdb can distinguish type names from function names, for example.
|
||||
//
|
||||
// A compunit contains one or more function address ranges. If an
|
||||
// object file is compiled without -ffunction-sections, it contains
|
||||
// only one .text section and therefore contains a single address range.
|
||||
// Such range is typically stored directly to the compunit.
|
||||
//
|
||||
// If an object file is compiled with -ffunction-sections, it contains
|
||||
// more than one .text section, and it has as many address ranges as
|
||||
// the number of .text sections. Such discontiguous address ranges are
|
||||
// stored to .debug_ranges in DWARF 2/3/4 and .debug_rnglists/.debug_addr
|
||||
// in DWARF 5.
|
||||
//
|
||||
// .debug_info section contains DWARF debug info. Although we don't need
|
||||
// to parse the whole .debug_info section to read address ranges, we
|
||||
// have to do a little bit. DWARF is complicated and often handled using
|
||||
// a library such as libdwarf. But we don't use any library because we
|
||||
// don't want to add an extra run-time dependency just for --gdb-index.
|
||||
//
|
||||
// This page explains the format of .gdb_index:
|
||||
// https://sourceware.org/gdb/onlinedocs/gdb/Index-Section-Format.html
|
||||
|
||||
#include "mold.h"
|
||||
#include <tbb/parallel_for_each.h>
|
||||
|
||||
namespace mold {
|
||||
|
||||
enum DwarfKind { DWARF2_32, DWARF5_32, DWARF2_64, DWARF5_64 };
|
||||
|
||||
template <typename E>
|
||||
struct CuHdrDwarf2_32 {
|
||||
U32<E> size;
|
||||
U16<E> version;
|
||||
U32<E> abbrev_offset;
|
||||
u8 address_size;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
struct CuHdrDwarf5_32 {
|
||||
U32<E> size;
|
||||
U16<E> version;
|
||||
u8 unit_type;
|
||||
u8 address_size;
|
||||
U32<E> abbrev_offset;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
struct CuHdrDwarf2_64 {
|
||||
U32<E> magic;
|
||||
U64<E> size;
|
||||
U16<E> version;
|
||||
U64<E> abbrev_offset;
|
||||
u8 address_size;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
struct CuHdrDwarf5_64 {
|
||||
U32<E> magic;
|
||||
U64<E> size;
|
||||
U16<E> version;
|
||||
u8 unit_type;
|
||||
u8 address_size;
|
||||
U64<E> abbrev_offset;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
struct PubnamesHdr32 {
|
||||
U32<E> size;
|
||||
U16<E> version;
|
||||
U32<E> debug_info_offset;
|
||||
U32<E> debug_info_size;
|
||||
};
|
||||
|
||||
template <typename E>
|
||||
struct PubnamesHdr64 {
|
||||
U32<E> magic;
|
||||
U64<E> size;
|
||||
U16<E> version;
|
||||
U64<E> debug_info_offset;
|
||||
U64<E> debug_info_size;
|
||||
};
|
||||
|
||||
struct SectionHeader {
|
||||
ul32 version = 7;
|
||||
ul32 cu_list_offset = 0;
|
||||
ul32 cu_types_offset = 0;
|
||||
ul32 ranges_offset = 0;
|
||||
ul32 symtab_offset = 0;
|
||||
ul32 const_pool_offset = 0;
|
||||
};
|
||||
|
||||
struct NameType {
|
||||
auto operator<=>(const NameType &) const = default;
|
||||
u64 hash;
|
||||
u8 type;
|
||||
std::string_view name;
|
||||
};
|
||||
|
||||
struct MapValue {
|
||||
u32 gdb_hash = 0;
|
||||
Atomic<u32> count;
|
||||
u32 name_offset = 0;
|
||||
u32 type_offset = 0;
|
||||
};
|
||||
|
||||
struct Compunit {
|
||||
DwarfKind kind;
|
||||
i64 offset;
|
||||
i64 size;
|
||||
std::vector<std::pair<u64, u64>> ranges;
|
||||
std::vector<NameType> nametypes;
|
||||
std::vector<MapValue *> entries;
|
||||
};
|
||||
|
||||
// The hash function for .gdb_index.
|
||||
static u32 gdb_hash(std::string_view name) {
|
||||
u32 h = 0;
|
||||
for (u8 c : name) {
|
||||
if ('A' <= c && c <= 'Z')
|
||||
c = 'a' + c - 'A';
|
||||
h = h * 67 + c - 113;
|
||||
}
|
||||
return h;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static DwarfKind get_dwarf_kind(Context<E> &ctx, u8 *p) {
|
||||
if (*(U32<E> *)p == 0xffff'ffff) {
|
||||
CuHdrDwarf2_64<E> &hdr = *(CuHdrDwarf2_64<E> *)p;
|
||||
if (hdr.version > 5)
|
||||
Fatal(ctx) << "--gdb-index: DWARF version " << hdr.version
|
||||
<< " is not supported";
|
||||
return (hdr.version == 5) ? DWARF5_64 : DWARF2_64;
|
||||
}
|
||||
|
||||
CuHdrDwarf2_32<E> &hdr = *(CuHdrDwarf2_32<E> *)p;
|
||||
if (hdr.version > 5)
|
||||
Fatal(ctx) << "--gdb-index: DWARF version " << hdr.version
|
||||
<< " is not supported";
|
||||
return (hdr.version == 5) ? DWARF5_32 : DWARF2_32;
|
||||
}
|
||||
|
||||
template <typename E, typename CuHdr>
|
||||
u8 *find_cu_abbrev(Context<E> &ctx, u8 **p, const CuHdr &hdr) {
|
||||
if (hdr.address_size != sizeof(Word<E>))
|
||||
Fatal(ctx) << "--gdb-index: unsupported address size " << hdr.address_size;
|
||||
|
||||
if constexpr (requires { hdr.unit_type; }) {
|
||||
switch (hdr.unit_type) {
|
||||
case DW_UT_compile:
|
||||
case DW_UT_partial:
|
||||
break;
|
||||
case DW_UT_skeleton:
|
||||
case DW_UT_split_compile:
|
||||
*p += 8;
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "--gdb-index: unknown unit type: 0x"
|
||||
<< std::hex << hdr.unit_type;
|
||||
}
|
||||
}
|
||||
|
||||
i64 abbrev_code = read_uleb(p);
|
||||
|
||||
// Find a .debug_abbrev record corresponding to the .debug_info record.
|
||||
// We assume the .debug_info record at a given offset is of
|
||||
// DW_TAG_compile_unit which describes a compunit.
|
||||
u8 *abbrev = &ctx.debug_abbrev[0] + hdr.abbrev_offset;
|
||||
|
||||
for (;;) {
|
||||
u32 code = read_uleb(&abbrev);
|
||||
if (code == 0)
|
||||
Fatal(ctx) << "--gdb-index: .debug_abbrev does not contain"
|
||||
<< " a record for the first .debug_info record";
|
||||
|
||||
if (code == abbrev_code) {
|
||||
// Found a record
|
||||
u64 abbrev_tag = read_uleb(&abbrev);
|
||||
if (abbrev_tag != DW_TAG_compile_unit && abbrev_tag != DW_TAG_skeleton_unit)
|
||||
Fatal(ctx) << "--gdb-index: the first entry's tag is not"
|
||||
<< " DW_TAG_compile_unit/DW_TAG_skeleton_unit but 0x"
|
||||
<< std::hex << abbrev_tag;
|
||||
break;
|
||||
}
|
||||
|
||||
// Skip an uninteresting record
|
||||
read_uleb(&abbrev); // tag
|
||||
abbrev++; // has_children byte
|
||||
for (;;) {
|
||||
u64 name = read_uleb(&abbrev);
|
||||
u64 form = read_uleb(&abbrev);
|
||||
if (name == 0 && form == 0)
|
||||
break;
|
||||
if (form == DW_FORM_implicit_const)
|
||||
read_uleb(&abbrev);
|
||||
}
|
||||
}
|
||||
|
||||
abbrev++; // skip has_children byte
|
||||
return abbrev;
|
||||
}
|
||||
|
||||
// .debug_info contains variable-length fields.
|
||||
// This function reads one scalar value from a given location.
|
||||
template <typename E, typename Offset>
|
||||
u64 read_scalar(Context<E> &ctx, u8 **p, u64 form) {
|
||||
switch (form) {
|
||||
case DW_FORM_flag_present:
|
||||
return 0;
|
||||
case DW_FORM_data1:
|
||||
case DW_FORM_flag:
|
||||
case DW_FORM_strx1:
|
||||
case DW_FORM_addrx1:
|
||||
case DW_FORM_ref1:
|
||||
return *(*p)++;
|
||||
case DW_FORM_data2:
|
||||
case DW_FORM_strx2:
|
||||
case DW_FORM_addrx2:
|
||||
case DW_FORM_ref2: {
|
||||
u64 val = *(U16<E> *)*p;
|
||||
*p += 2;
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_strx3:
|
||||
case DW_FORM_addrx3: {
|
||||
u64 val = *(U24<E> *)*p;
|
||||
*p += 3;
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_data4:
|
||||
case DW_FORM_strx4:
|
||||
case DW_FORM_addrx4:
|
||||
case DW_FORM_ref4: {
|
||||
u64 val = *(U32<E> *)*p;
|
||||
*p += 4;
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_data8:
|
||||
case DW_FORM_ref8: {
|
||||
u64 val = *(U64<E> *)*p;
|
||||
*p += 8;
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_strp:
|
||||
case DW_FORM_sec_offset:
|
||||
case DW_FORM_line_strp: {
|
||||
u64 val = *(Offset *)*p;
|
||||
*p += sizeof(Offset);
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_addr:
|
||||
case DW_FORM_ref_addr: {
|
||||
u64 val = *(Word<E> *)*p;
|
||||
*p += sizeof(Word<E>);
|
||||
return val;
|
||||
}
|
||||
case DW_FORM_strx:
|
||||
case DW_FORM_addrx:
|
||||
case DW_FORM_udata:
|
||||
case DW_FORM_ref_udata:
|
||||
case DW_FORM_loclistx:
|
||||
case DW_FORM_rnglistx:
|
||||
return read_uleb(p);
|
||||
case DW_FORM_string:
|
||||
*p += strlen((char *)*p) + 1;
|
||||
return 0;
|
||||
default:
|
||||
Fatal(ctx) << "--gdb-index: unhandled debug info form: 0x"
|
||||
<< std::hex << form;
|
||||
}
|
||||
}
|
||||
|
||||
// Read a range list from .debug_ranges starting at the given offset.
|
||||
template <typename E>
|
||||
static std::vector<std::pair<u64, u64>>
|
||||
read_debug_range(Word<E> *range, u64 base) {
|
||||
std::vector<std::pair<u64, u64>> vec;
|
||||
|
||||
for (i64 i = 0; range[i] || range[i + 1]; i += 2) {
|
||||
if (range[i] + 1 == 0)
|
||||
base = range[i + 1];
|
||||
else
|
||||
vec.emplace_back(range[i] + base, range[i + 1] + base);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
// Read a range list from .debug_rnglists starting at the given offset.
|
||||
template <typename E>
|
||||
static void
|
||||
read_rnglist_range(std::vector<std::pair<u64, u64>> &vec, u8 *p,
|
||||
Word<E> *addrx, u64 base) {
|
||||
for (;;) {
|
||||
switch (*p++) {
|
||||
case DW_RLE_end_of_list:
|
||||
return;
|
||||
case DW_RLE_base_addressx:
|
||||
base = addrx[read_uleb(&p)];
|
||||
break;
|
||||
case DW_RLE_startx_endx: {
|
||||
u64 val1 = read_uleb(&p);
|
||||
u64 val2 = read_uleb(&p);
|
||||
vec.emplace_back(addrx[val1], addrx[val2]);
|
||||
break;
|
||||
}
|
||||
case DW_RLE_startx_length: {
|
||||
u64 val1 = read_uleb(&p);
|
||||
u64 val2 = read_uleb(&p);
|
||||
vec.emplace_back(addrx[val1], addrx[val1] + val2);
|
||||
break;
|
||||
}
|
||||
case DW_RLE_offset_pair: {
|
||||
u64 val1 = read_uleb(&p);
|
||||
u64 val2 = read_uleb(&p);
|
||||
|
||||
// If the base is 0, this address range is for an eliminated
|
||||
// section. We only emit it if it's alive.
|
||||
if (base)
|
||||
vec.emplace_back(base + val1, base + val2);
|
||||
break;
|
||||
}
|
||||
case DW_RLE_base_address:
|
||||
base = *(Word<E> *)p;
|
||||
p += sizeof(Word<E>);
|
||||
break;
|
||||
case DW_RLE_start_end: {
|
||||
u64 val1 = ((Word<E> *)p)[0];
|
||||
u64 val2 = ((Word<E> *)p)[1];
|
||||
p += sizeof(Word<E>) * 2;
|
||||
vec.emplace_back(val1, val2);
|
||||
break;
|
||||
}
|
||||
case DW_RLE_start_length: {
|
||||
u64 val1 = *(Word<E> *)p;
|
||||
p += sizeof(Word<E>);
|
||||
u64 val2 = read_uleb(&p);
|
||||
vec.emplace_back(val1, val1 + val2);
|
||||
break;
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
// Returns a list of address ranges explained by a compunit at the
|
||||
// `offset` in an output .debug_info section.
|
||||
//
|
||||
// .debug_info contains DWARF debug info records, so this function
|
||||
// parses DWARF. If a designated compunit contains multiple ranges, the
|
||||
// ranges are read from .debug_ranges (or .debug_rnglists for DWARF5).
|
||||
// Otherwise, a range is read directly from .debug_info (or possibly
|
||||
// from .debug_addr for DWARF5).
|
||||
template <typename E, typename CuHdr>
|
||||
static std::vector<std::pair<u64, u64>>
|
||||
read_address_ranges(Context<E> &ctx, const Compunit &cu) {
|
||||
// Read .debug_info to find the record at a given offset.
|
||||
u8 *p = &ctx.debug_info[0] + cu.offset;
|
||||
CuHdr &hdr = *(CuHdr *)p;
|
||||
p += sizeof(hdr);
|
||||
|
||||
u8 *abbrev = find_cu_abbrev(ctx, &p, hdr);
|
||||
|
||||
// Now, read debug info records.
|
||||
struct Record {
|
||||
u64 form = 0;
|
||||
u64 value = 0;
|
||||
};
|
||||
|
||||
using Offset = decltype(hdr.size);
|
||||
|
||||
Record low_pc;
|
||||
Record high_pc;
|
||||
Record ranges;
|
||||
u64 rnglists_base = -1;
|
||||
Word<E> *addrx = nullptr;
|
||||
|
||||
// Read all interesting debug records.
|
||||
for (;;) {
|
||||
u64 name = read_uleb(&abbrev);
|
||||
u64 form = read_uleb(&abbrev);
|
||||
if (name == 0 && form == 0)
|
||||
break;
|
||||
|
||||
u64 val = read_scalar<E, Offset>(ctx, &p, form);
|
||||
|
||||
switch (name) {
|
||||
case DW_AT_low_pc:
|
||||
low_pc = {form, val};
|
||||
break;
|
||||
case DW_AT_high_pc:
|
||||
high_pc = {form, val};
|
||||
break;
|
||||
case DW_AT_rnglists_base:
|
||||
rnglists_base = val;
|
||||
break;
|
||||
case DW_AT_addr_base:
|
||||
addrx = (Word<E> *)(&ctx.debug_addr[0] + val);
|
||||
break;
|
||||
case DW_AT_ranges:
|
||||
ranges = {form, val};
|
||||
break;
|
||||
}
|
||||
}
|
||||
|
||||
// Handle non-contiguous address ranges.
|
||||
if (ranges.form) {
|
||||
if (hdr.version <= 4) {
|
||||
Word<E> *p = (Word<E> *)(&ctx.debug_ranges[0] + ranges.value);
|
||||
return read_debug_range<E>(p, low_pc.value);
|
||||
}
|
||||
|
||||
assert(hdr.version == 5);
|
||||
|
||||
std::vector<std::pair<u64, u64>> vec;
|
||||
u8 *buf = &ctx.debug_rnglists[0];
|
||||
|
||||
if (ranges.form == DW_FORM_sec_offset) {
|
||||
read_rnglist_range<E>(vec, buf + ranges.value, addrx, low_pc.value);
|
||||
} else {
|
||||
if (rnglists_base == -1)
|
||||
Fatal(ctx) << "--gdb-index: missing DW_AT_rnglists_base";
|
||||
|
||||
u8 *base = buf + rnglists_base;
|
||||
Offset *offsets = (Offset *)base;
|
||||
read_rnglist_range<E>(vec, base + offsets[ranges.value], addrx, low_pc.value);
|
||||
}
|
||||
return vec;
|
||||
}
|
||||
|
||||
// Handle a contiguous address range.
|
||||
if (low_pc.form && high_pc.form) {
|
||||
u64 lo;
|
||||
|
||||
switch (low_pc.form) {
|
||||
case DW_FORM_addr:
|
||||
lo = low_pc.value;
|
||||
break;
|
||||
case DW_FORM_addrx:
|
||||
case DW_FORM_addrx1:
|
||||
case DW_FORM_addrx2:
|
||||
case DW_FORM_addrx4:
|
||||
lo = addrx[low_pc.value];
|
||||
break;
|
||||
default:
|
||||
Fatal(ctx) << "--gdb-index: unhandled form for DW_AT_low_pc: 0x"
|
||||
<< std::hex << high_pc.form;
|
||||
}
|
||||
|
||||
switch (high_pc.form) {
|
||||
case DW_FORM_addr:
|
||||
return {{lo, high_pc.value}};
|
||||
case DW_FORM_addrx:
|
||||
case DW_FORM_addrx1:
|
||||
case DW_FORM_addrx2:
|
||||
case DW_FORM_addrx4:
|
||||
return {{lo, addrx[high_pc.value]}};
|
||||
case DW_FORM_udata:
|
||||
case DW_FORM_data1:
|
||||
case DW_FORM_data2:
|
||||
case DW_FORM_data4:
|
||||
case DW_FORM_data8:
|
||||
return {{lo, lo + high_pc.value}};
|
||||
default:
|
||||
Fatal(ctx) << "--gdb-index: unhandled form for DW_AT_high_pc: 0x"
|
||||
<< std::hex << high_pc.form;
|
||||
}
|
||||
}
|
||||
|
||||
return {};
|
||||
}
|
||||
|
||||
template <typename E, typename PubnamesHdr>
|
||||
static i64 read_pubnames_cu(Context<E> &ctx, const PubnamesHdr &hdr,
|
||||
std::vector<Compunit> &cus, ObjectFile<E> &file) {
|
||||
using Offset = decltype(hdr.size);
|
||||
|
||||
auto get_cu = [&](i64 offset) {
|
||||
for (i64 i = 0; i < cus.size(); i++)
|
||||
if (cus[i].offset == offset)
|
||||
return &cus[i];
|
||||
Fatal(ctx) << file << ": corrupted debug_info_offset";
|
||||
};
|
||||
|
||||
Compunit *cu = get_cu(file.debug_info->offset + hdr.debug_info_offset);
|
||||
i64 size = hdr.size + offsetof(PubnamesHdr, size) + sizeof(hdr.size);
|
||||
u8 *p = (u8 *)&hdr + sizeof(hdr);
|
||||
u8 *end = (u8 *)&hdr + size;
|
||||
|
||||
while (p < end) {
|
||||
if (*(Offset *)p == 0)
|
||||
break;
|
||||
p += sizeof(Offset);
|
||||
|
||||
u8 type = *p++;
|
||||
std::string_view name = (char *)p;
|
||||
p += name.size() + 1;
|
||||
cu->nametypes.push_back(NameType{hash_string(name), type, name});
|
||||
}
|
||||
|
||||
return size;
|
||||
}
|
||||
|
||||
// Parses .debug_gnu_pubnames and .debug_gnu_pubtypes. These sections
|
||||
// start with a 14 bytes header followed by (4-byte offset, 1-byte type,
|
||||
// null-terminated string) tuples.
|
||||
//
|
||||
// The 4-byte offset is an offset into .debug_info that contains details
|
||||
// about the name. The 1-byte type is a type of the corresponding name
|
||||
// (e.g. function, variable or datatype). The string is a name of a
|
||||
// function, a variable or a type.
|
||||
template <typename E>
|
||||
static void read_pubnames(Context<E> &ctx, std::vector<Compunit> &cus,
|
||||
ObjectFile<E> &file) {
|
||||
for (InputSection<E> *isec : { file.debug_pubnames, file.debug_pubtypes }) {
|
||||
if (!isec)
|
||||
continue;
|
||||
|
||||
isec->uncompress(ctx);
|
||||
if (isec->contents.empty())
|
||||
continue;
|
||||
|
||||
u8 *p = (u8*)&isec->contents[0];
|
||||
u8 *end = p + isec->contents.size();
|
||||
|
||||
while (p < end) {
|
||||
if (*(U32<E> *)p == 0xffff'ffff)
|
||||
p += read_pubnames_cu(ctx, *(PubnamesHdr64<E> *)p, cus, file);
|
||||
else
|
||||
p += read_pubnames_cu(ctx, *(PubnamesHdr32<E> *)p, cus, file);
|
||||
}
|
||||
};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
static std::vector<Compunit> read_compunits(Context<E> &ctx) {
|
||||
std::vector<Compunit> cus;
|
||||
|
||||
// Read compunits from the output .debug_info section.
|
||||
u8 *begin = &ctx.debug_info[0];
|
||||
u8 *end = begin + ctx.debug_info.size();
|
||||
|
||||
for (u8 *p = begin; p < end;) {
|
||||
DwarfKind kind = get_dwarf_kind(ctx, p);
|
||||
i64 size;
|
||||
if (kind == DWARF2_32 || kind == DWARF5_32)
|
||||
size = ((CuHdrDwarf2_32<E> *)p)->size + 4;
|
||||
else
|
||||
size = ((CuHdrDwarf2_64<E> *)p)->size + 12;
|
||||
|
||||
cus.push_back(Compunit{kind, p - begin, size});
|
||||
p += size;
|
||||
}
|
||||
|
||||
// Read address ranges for each compunit.
|
||||
tbb::parallel_for_each(cus, [&](Compunit &cu) {
|
||||
switch (cu.kind) {
|
||||
case DWARF2_32:
|
||||
cu.ranges = read_address_ranges<E, CuHdrDwarf2_32<E>>(ctx, cu);
|
||||
break;
|
||||
case DWARF5_32:
|
||||
cu.ranges = read_address_ranges<E, CuHdrDwarf5_32<E>>(ctx, cu);
|
||||
break;
|
||||
case DWARF2_64:
|
||||
cu.ranges = read_address_ranges<E, CuHdrDwarf2_64<E>>(ctx, cu);
|
||||
break;
|
||||
case DWARF5_64:
|
||||
cu.ranges = read_address_ranges<E, CuHdrDwarf5_64<E>>(ctx, cu);
|
||||
break;
|
||||
}
|
||||
|
||||
// Remove empty ranges
|
||||
std::erase_if(cu.ranges, [](std::pair<u64, u64> p) {
|
||||
return p.first == 0 || p.first == p.second;
|
||||
});
|
||||
});
|
||||
|
||||
// Read symbols from .debug_gnu_pubnames and .debug_gnu_pubtypes.
|
||||
tbb::parallel_for_each(ctx.objs, [&](ObjectFile<E> *file) {
|
||||
read_pubnames(ctx, cus, *file);
|
||||
});
|
||||
|
||||
// Uniquify elements because GCC 11 seems to emit one record for each
|
||||
// comdat group which results in having a lot of duplicate records.
|
||||
tbb::parallel_for_each(cus, [](Compunit &cu) {
|
||||
ranges::stable_sort(cu.nametypes);
|
||||
remove_duplicates(cu.nametypes);
|
||||
});
|
||||
|
||||
return cus;
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
std::span<u8> get_buffer(Context<E> &ctx, Chunk<E> *chunk) {
|
||||
if (chunk->is_compressed) {
|
||||
CompressedSection<E> &sec = *(CompressedSection<E> *)chunk;
|
||||
return {sec.uncompressed_data.get(), (size_t)sec.chdr.ch_size};
|
||||
}
|
||||
return {ctx.buf + chunk->shdr.sh_offset, (size_t)chunk->shdr.sh_size};
|
||||
}
|
||||
|
||||
template <typename E>
|
||||
void write_gdb_index(Context<E> &ctx) {
|
||||
Timer t(ctx, "write_gdb_index");
|
||||
|
||||
// Find debug info sections
|
||||
for (Chunk<E> *chunk : ctx.chunks) {
|
||||
std::string_view name = chunk->name;
|
||||
if (name == ".debug_info")
|
||||
ctx.debug_info = get_buffer(ctx, chunk);
|
||||
if (name == ".debug_abbrev")
|
||||
ctx.debug_abbrev = get_buffer(ctx, chunk);
|
||||
if (name == ".debug_ranges")
|
||||
ctx.debug_ranges = get_buffer(ctx, chunk);
|
||||
if (name == ".debug_addr")
|
||||
ctx.debug_addr = get_buffer(ctx, chunk);
|
||||
if (name == ".debug_rnglists")
|
||||
ctx.debug_rnglists = get_buffer(ctx, chunk);
|
||||
}
|
||||
|
||||
if (ctx.debug_info.empty())
|
||||
return;
|
||||
|
||||
// Read debug info
|
||||
std::vector<Compunit> cus = read_compunits(ctx);
|
||||
|
||||
// Uniquify symbols
|
||||
HyperLogLog estimator;
|
||||
|
||||
tbb::parallel_for_each(cus, [&](Compunit &cu) {
|
||||
HyperLogLog e;
|
||||
for (NameType &nt : cu.nametypes)
|
||||
e.insert(nt.hash);
|
||||
estimator.merge(e);
|
||||
});
|
||||
|
||||
ConcurrentMap<MapValue> map(estimator.get_cardinality() * 3 / 2);
|
||||
|
||||
tbb::parallel_for_each(cus, [&](Compunit &cu) {
|
||||
cu.entries.reserve(cu.nametypes.size());
|
||||
for (NameType &nt : cu.nametypes) {
|
||||
MapValue *ent;
|
||||
bool inserted;
|
||||
std::tie(ent, inserted) = map.insert(nt.name, nt.hash,
|
||||
MapValue{gdb_hash(nt.name)});
|
||||
ent->count++;
|
||||
cu.entries.push_back(ent);
|
||||
}
|
||||
});
|
||||
|
||||
// Sort symbols for build reproducibility
|
||||
using Entry = typename decltype(map)::Entry;
|
||||
std::vector<Entry *> entries = map.get_sorted_entries_all();
|
||||
|
||||
// Compute sizes of each components
|
||||
SectionHeader hdr;
|
||||
hdr.cu_list_offset = sizeof(hdr);
|
||||
hdr.cu_types_offset = hdr.cu_list_offset + cus.size() * 16;
|
||||
hdr.ranges_offset = hdr.cu_types_offset;
|
||||
|
||||
hdr.symtab_offset = hdr.ranges_offset;
|
||||
for (Compunit &cu : cus)
|
||||
hdr.symtab_offset += cu.ranges.size() * 20;
|
||||
|
||||
i64 ht_size = bit_ceil(entries.size() * 5 / 4 + 1);
|
||||
hdr.const_pool_offset = hdr.symtab_offset + ht_size * 8;
|
||||
|
||||
i64 offset = 0;
|
||||
for (Entry *ent : entries) {
|
||||
ent->value.type_offset = offset;
|
||||
offset += ent->value.count * 4 + 4;
|
||||
}
|
||||
|
||||
for (Entry *ent : entries) {
|
||||
ent->value.name_offset = offset;
|
||||
offset += ent->keylen + 1;
|
||||
}
|
||||
|
||||
i64 bufsize = hdr.const_pool_offset + offset;
|
||||
|
||||
// Allocate an output buffer. We use malloc instead of vector to
|
||||
// avoid zero-initializing the entire buffer.
|
||||
ctx.output_file->buf2 = (u8 *)malloc(bufsize);
|
||||
ctx.output_file->buf2_size = bufsize;
|
||||
u8 *buf = ctx.output_file->buf2;
|
||||
|
||||
// Write a section header
|
||||
memcpy(buf, &hdr, sizeof(hdr));
|
||||
|
||||
// Write a CU list
|
||||
u8 *p = buf + sizeof(hdr);
|
||||
|
||||
for (Compunit &cu : cus) {
|
||||
*(ul64 *)p = cu.offset;
|
||||
*(ul64 *)(p + 8) = cu.size;
|
||||
p += 16;
|
||||
}
|
||||
|
||||
// Write address areas
|
||||
std::vector<i64> range_offsets(cus.size());
|
||||
for (i64 i = 1; i < cus.size(); i++)
|
||||
range_offsets[i] = range_offsets[i - 1] + cus[i - 1].ranges.size() * 20;
|
||||
|
||||
tbb::parallel_for_each(cus, [&](Compunit &cu) {
|
||||
i64 i = &cu - cus.data();
|
||||
u8 *p = buf + hdr.ranges_offset + range_offsets[i];
|
||||
for (std::pair<u64, u64> range : cu.ranges) {
|
||||
*(ul64 *)p = range.first;
|
||||
*(ul64 *)(p + 8) = range.second;
|
||||
*(ul32 *)(p + 16) = i;
|
||||
p += 20;
|
||||
}
|
||||
});
|
||||
|
||||
// Write a symbol table
|
||||
u32 mask = ht_size - 1;
|
||||
ul32 *ht = (ul32 *)(buf + hdr.symtab_offset);
|
||||
memset(ht, 0, ht_size * 8);
|
||||
|
||||
for (Entry *ent : entries) {
|
||||
u32 hash = ent->value.gdb_hash;
|
||||
u32 step = ((hash * 17) & mask) | 1;
|
||||
u32 j = hash & mask;
|
||||
|
||||
while (ht[j * 2] || ht[j * 2 + 1])
|
||||
j = (j + step) & mask;
|
||||
|
||||
ht[j * 2] = ent->value.name_offset;
|
||||
ht[j * 2 + 1] = ent->value.type_offset;
|
||||
}
|
||||
|
||||
// Write types. Use MapValue::count as an atomic slot counter.
|
||||
u8 *base = buf + hdr.const_pool_offset;
|
||||
|
||||
for (Entry *ent : entries)
|
||||
ent->value.count = 0;
|
||||
|
||||
tbb::parallel_for_each(cus, [&](Compunit &cu) {
|
||||
i64 i = &cu - cus.data();
|
||||
for (i64 j = 0; j < cu.nametypes.size(); j++) {
|
||||
MapValue *ent = cu.entries[j];
|
||||
ul32 *p = (ul32 *)(base + ent->type_offset);
|
||||
i64 idx = ++ent->count;
|
||||
p[idx] = (cu.nametypes[j].type << 24) | i;
|
||||
}
|
||||
});
|
||||
|
||||
// Write the final counts into the buffer.
|
||||
tbb::parallel_for_each(entries, [&](Entry *ent) {
|
||||
ul32 *p = (ul32 *)(base + ent->value.type_offset);
|
||||
p[0] = ent->value.count;
|
||||
// Sort entries for deterministic output
|
||||
std::sort(p + 1, p + 1 + ent->value.count);
|
||||
});
|
||||
|
||||
// Write names
|
||||
tbb::parallel_for_each(entries, [&](Entry *ent) {
|
||||
u8 *dst = buf + hdr.const_pool_offset + ent->value.name_offset;
|
||||
memcpy(dst, ent->key, ent->keylen);
|
||||
dst[ent->keylen] = '\0';
|
||||
});
|
||||
|
||||
// Update the section size and rewrite the section header
|
||||
if (ctx.shdr) {
|
||||
ctx.gdb_index->shdr.sh_size = bufsize;
|
||||
ctx.shdr->copy_buf(ctx);
|
||||
}
|
||||
}
|
||||
|
||||
using E = MOLD_TARGET;
|
||||
|
||||
template void write_gdb_index(Context<E> &);
|
||||
|
||||
} // namespace mold
|
||||
-1729
File diff suppressed because it is too large.
Load diff
Loaded 100 of 2812 files, more files were not shown because too many files have changed in this diff.
Show more
Reference in new issue
Block a user