Bug 1271169 - Move EME/GMP device binding code into GMPDeviceBinding.h/cpp. r=gerald

I want the EME device binding/nodeId code to be callable from gtests, as well
as from in plugin-container.

First step is to move the device binding code into a discrete file, so I can
also link that into gtests, and call it from there to compare the result with
what's in the GMP process.

MozReview-Commit-ID: 9xT2rp3hWW

--HG--
extra : rebase_source : 824c7a9841bce83c438decad48ce210f6c2a5571
This commit is contained in:
Chris Pearce 2016-05-11 19:56:42 +12:00
parent f89be2268a
commit 9c711e8fb4
4 changed files with 235 additions and 167 deletions

View File

@ -15,34 +15,9 @@
#ifdef XP_WIN
#include "windows.h"
#ifdef MOZ_SANDBOX
#include <intrin.h>
#include <assert.h>
#endif
#endif
#ifdef XP_MACOSX
#include <assert.h>
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
#include <unistd.h>
#include <mach/mach.h>
#include <mach/mach_vm.h>
#endif
#endif
#if defined(HASH_NODE_ID_WITH_DEVICE_ID)
// In order to provide EME plugins with a "device binding" capability,
// in the parent we generate and store some random bytes as salt for every
// (origin, urlBarOrigin) pair that uses EME. We store these bytes so
// that every time we revisit the same origin we get the same salt.
// We send this salt to the child on startup. The child collects some
// device specific data and munges that with the salt to create the
// "node id" that we expose to EME plugins. It then overwrites the device
// specific data, and activates the sandbox.
#include "rlz/lib/machine_id.h"
#include "rlz/lib/string_utils.h"
#include "sha256.h"
#endif
#include "GMPDeviceBinding.h"
namespace mozilla {
namespace gmp {
@ -144,94 +119,6 @@ private:
PRLibrary* mLib = nullptr;
};
#if defined(XP_WIN) && defined(HASH_NODE_ID_WITH_DEVICE_ID)
MOZ_NEVER_INLINE
static bool
GetStackAfterCurrentFrame(uint8_t** aOutTop, uint8_t** aOutBottom)
{
// "Top" of the free space on the stack is directly after the memory
// holding our return address.
uint8_t* top = (uint8_t*)_AddressOfReturnAddress();
// Look down the stack until we find the guard page...
MEMORY_BASIC_INFORMATION memInfo = {0};
uint8_t* bottom = top;
while (1) {
if (!VirtualQuery(bottom, &memInfo, sizeof(memInfo))) {
return false;
}
if ((memInfo.Protect & PAGE_GUARD) == PAGE_GUARD) {
bottom = (uint8_t*)memInfo.BaseAddress + memInfo.RegionSize;
#ifdef DEBUG
if (!VirtualQuery(bottom, &memInfo, sizeof(memInfo))) {
return false;
}
assert(!(memInfo.Protect & PAGE_GUARD)); // Should have found boundary.
#endif
break;
} else if (memInfo.State != MEM_COMMIT ||
(memInfo.AllocationProtect & PAGE_READWRITE) != PAGE_READWRITE) {
return false;
}
bottom = (uint8_t*)memInfo.BaseAddress - 1;
}
*aOutTop = top;
*aOutBottom = bottom;
return true;
}
#endif
#if defined(XP_MACOSX) && defined(HASH_NODE_ID_WITH_DEVICE_ID)
static mach_vm_address_t
RegionContainingAddress(mach_vm_address_t aAddress)
{
mach_port_t task;
kern_return_t kr = task_for_pid(mach_task_self(), getpid(), &task);
if (kr != KERN_SUCCESS) {
return 0;
}
mach_vm_address_t address = aAddress;
mach_vm_size_t size;
vm_region_basic_info_data_64_t info;
mach_msg_type_number_t count = VM_REGION_BASIC_INFO_COUNT_64;
mach_port_t object_name;
kr = mach_vm_region(task, &address, &size, VM_REGION_BASIC_INFO_64,
reinterpret_cast<vm_region_info_t>(&info), &count,
&object_name);
if (kr != KERN_SUCCESS || size == 0
|| address > aAddress || address + size <= aAddress) {
// mach_vm_region failed, or couldn't find region at given address.
return 0;
}
return address;
}
MOZ_NEVER_INLINE
static bool
GetStackAfterCurrentFrame(uint8_t** aOutTop, uint8_t** aOutBottom)
{
mach_vm_address_t stackFrame =
reinterpret_cast<mach_vm_address_t>(__builtin_frame_address(0));
*aOutTop = reinterpret_cast<uint8_t*>(stackFrame);
// Kernel code shows that stack is always a single region.
*aOutBottom = reinterpret_cast<uint8_t*>(RegionContainingAddress(stackFrame));
return *aOutBottom && (*aOutBottom < *aOutTop);
}
#endif
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
static void SecureMemset(void* start, uint8_t value, size_t size)
{
// Inline instructions equivalent to RtlSecureZeroMemory().
for (size_t i = 0; i < size; ++i) {
volatile uint8_t* p = static_cast<volatile uint8_t*>(start) + i;
*p = value;
}
}
#endif
bool
GMPLoaderImpl::Load(const char* aUTF8LibPath,
uint32_t aUTF8LibPathLen,
@ -241,59 +128,8 @@ GMPLoaderImpl::Load(const char* aUTF8LibPath,
GMPAdapter* aAdapter)
{
std::string nodeId;
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
if (aOriginSaltLen > 0) {
std::vector<uint8_t> deviceId;
int volumeId;
if (!rlz_lib::GetRawMachineId(&deviceId, &volumeId)) {
return false;
}
SHA256Context ctx;
SHA256_Begin(&ctx);
SHA256_Update(&ctx, (const uint8_t*)aOriginSalt, aOriginSaltLen);
SHA256_Update(&ctx, deviceId.data(), deviceId.size());
SHA256_Update(&ctx, (const uint8_t*)&volumeId, sizeof(int));
uint8_t digest[SHA256_LENGTH] = {0};
unsigned int digestLen = 0;
SHA256_End(&ctx, digest, &digestLen, SHA256_LENGTH);
// Overwrite all data involved in calculation as it could potentially
// identify the user, so there's no chance a GMP can read it and use
// it for identity tracking.
SecureMemset(&ctx, 0, sizeof(ctx));
SecureMemset(aOriginSalt, 0, aOriginSaltLen);
SecureMemset(&volumeId, 0, sizeof(volumeId));
SecureMemset(deviceId.data(), '*', deviceId.size());
deviceId.clear();
if (!rlz_lib::BytesToString(digest, SHA256_LENGTH, &nodeId)) {
return false;
}
if (!PR_GetEnv("MOZ_GMP_DISABLE_NODE_ID_CLEANUP")) {
// We've successfully bound the origin salt to node id.
// rlz_lib::GetRawMachineId and/or the system functions it
// called could have left user identifiable data on the stack,
// so carefully zero the stack down to the guard page.
uint8_t* top;
uint8_t* bottom;
if (!GetStackAfterCurrentFrame(&top, &bottom)) {
return false;
}
assert(top >= bottom);
// Inline instructions equivalent to RtlSecureZeroMemory().
// We can't just use RtlSecureZeroMemory here directly, as in debug
// builds, RtlSecureZeroMemory() can't be inlined, and the stack
// memory it uses would get wiped by itself running, causing crashes.
for (volatile uint8_t* p = (volatile uint8_t*)bottom; p < top; p++) {
*p = 0;
}
}
} else
#endif
{
nodeId = std::string(aOriginSalt, aOriginSalt + aOriginSaltLen);
if (!CalculateGMPDeviceId(aOriginSalt, aOriginSaltLen, nodeId)) {
return false;
}
// Start the sandbox now that we've generated the device bound node id.

View File

@ -0,0 +1,209 @@
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this file,
* You can obtain one at http://mozilla.org/MPL/2.0/. */
#include "GMPDeviceBinding.h"
#include "mozilla/Attributes.h"
#include "prenv.h"
#include <string>
#ifdef XP_WIN
#include "windows.h"
#ifdef MOZ_SANDBOX
#include <intrin.h>
#include <assert.h>
#endif
#endif
#if defined(HASH_NODE_ID_WITH_DEVICE_ID)
// In order to provide EME plugins with a "device binding" capability,
// in the parent we generate and store some random bytes as salt for every
// (origin, urlBarOrigin) pair that uses EME. We store these bytes so
// that every time we revisit the same origin we get the same salt.
// We send this salt to the child on startup. The child collects some
// device specific data and munges that with the salt to create the
// "node id" that we expose to EME plugins. It then overwrites the device
// specific data, and activates the sandbox.
#include "rlz/lib/machine_id.h"
#include "rlz/lib/string_utils.h"
#include "sha256.h"
#ifdef XP_WIN
#include "windows.h"
#ifdef MOZ_SANDBOX
#include <intrin.h>
#include <assert.h>
#endif
#endif
#ifdef XP_MACOSX
#include <assert.h>
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
#include <unistd.h>
#include <mach/mach.h>
#include <mach/mach_vm.h>
#endif
#endif
#endif // HASH_NODE_ID_WITH_DEVICE_ID
namespace mozilla {
namespace gmp {
#if defined(XP_WIN) && defined(HASH_NODE_ID_WITH_DEVICE_ID)
MOZ_NEVER_INLINE
static bool
GetStackAfterCurrentFrame(uint8_t** aOutTop, uint8_t** aOutBottom)
{
// "Top" of the free space on the stack is directly after the memory
// holding our return address.
uint8_t* top = (uint8_t*)_AddressOfReturnAddress();
// Look down the stack until we find the guard page...
MEMORY_BASIC_INFORMATION memInfo = {0};
uint8_t* bottom = top;
while (1) {
if (!VirtualQuery(bottom, &memInfo, sizeof(memInfo))) {
return false;
}
if ((memInfo.Protect & PAGE_GUARD) == PAGE_GUARD) {
bottom = (uint8_t*)memInfo.BaseAddress + memInfo.RegionSize;
#ifdef DEBUG
if (!VirtualQuery(bottom, &memInfo, sizeof(memInfo))) {
return false;
}
assert(!(memInfo.Protect & PAGE_GUARD)); // Should have found boundary.
#endif
break;
} else if (memInfo.State != MEM_COMMIT ||
(memInfo.AllocationProtect & PAGE_READWRITE) != PAGE_READWRITE) {
return false;
}
bottom = (uint8_t*)memInfo.BaseAddress - 1;
}
*aOutTop = top;
*aOutBottom = bottom;
return true;
}
#endif
#if defined(XP_MACOSX) && defined(HASH_NODE_ID_WITH_DEVICE_ID)
static mach_vm_address_t
RegionContainingAddress(mach_vm_address_t aAddress)
{
mach_port_t task;
kern_return_t kr = task_for_pid(mach_task_self(), getpid(), &task);
if (kr != KERN_SUCCESS) {
return 0;
}
mach_vm_address_t address = aAddress;
mach_vm_size_t size;
vm_region_basic_info_data_64_t info;
mach_msg_type_number_t count = VM_REGION_BASIC_INFO_COUNT_64;
mach_port_t object_name;
kr = mach_vm_region(task, &address, &size, VM_REGION_BASIC_INFO_64,
reinterpret_cast<vm_region_info_t>(&info), &count,
&object_name);
if (kr != KERN_SUCCESS || size == 0
|| address > aAddress || address + size <= aAddress) {
// mach_vm_region failed, or couldn't find region at given address.
return 0;
}
return address;
}
MOZ_NEVER_INLINE
static bool
GetStackAfterCurrentFrame(uint8_t** aOutTop, uint8_t** aOutBottom)
{
mach_vm_address_t stackFrame =
reinterpret_cast<mach_vm_address_t>(__builtin_frame_address(0));
*aOutTop = reinterpret_cast<uint8_t*>(stackFrame);
// Kernel code shows that stack is always a single region.
*aOutBottom = reinterpret_cast<uint8_t*>(RegionContainingAddress(stackFrame));
return *aOutBottom && (*aOutBottom < *aOutTop);
}
#endif
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
static void SecureMemset(void* start, uint8_t value, size_t size)
{
// Inline instructions equivalent to RtlSecureZeroMemory().
for (size_t i = 0; i < size; ++i) {
volatile uint8_t* p = static_cast<volatile uint8_t*>(start) + i;
*p = value;
}
}
#endif
bool
CalculateGMPDeviceId(char* aOriginSalt,
uint32_t aOriginSaltLen,
std::string& aOutNodeId)
{
#ifdef HASH_NODE_ID_WITH_DEVICE_ID
if (aOriginSaltLen > 0) {
std::vector<uint8_t> deviceId;
int volumeId;
if (!rlz_lib::GetRawMachineId(&deviceId, &volumeId)) {
return false;
}
SHA256Context ctx;
SHA256_Begin(&ctx);
SHA256_Update(&ctx, (const uint8_t*)aOriginSalt, aOriginSaltLen);
SHA256_Update(&ctx, deviceId.data(), deviceId.size());
SHA256_Update(&ctx, (const uint8_t*)&volumeId, sizeof(int));
uint8_t digest[SHA256_LENGTH] = {0};
unsigned int digestLen = 0;
SHA256_End(&ctx, digest, &digestLen, SHA256_LENGTH);
// Overwrite all data involved in calculation as it could potentially
// identify the user, so there's no chance a GMP can read it and use
// it for identity tracking.
SecureMemset(&ctx, 0, sizeof(ctx));
SecureMemset(aOriginSalt, 0, aOriginSaltLen);
SecureMemset(&volumeId, 0, sizeof(volumeId));
SecureMemset(deviceId.data(), '*', deviceId.size());
deviceId.clear();
if (!rlz_lib::BytesToString(digest, SHA256_LENGTH, &aOutNodeId)) {
return false;
}
if (!PR_GetEnv("MOZ_GMP_DISABLE_NODE_ID_CLEANUP")) {
// We've successfully bound the origin salt to node id.
// rlz_lib::GetRawMachineId and/or the system functions it
// called could have left user identifiable data on the stack,
// so carefully zero the stack down to the guard page.
uint8_t* top;
uint8_t* bottom;
if (!GetStackAfterCurrentFrame(&top, &bottom)) {
return false;
}
assert(top >= bottom);
// Inline instructions equivalent to RtlSecureZeroMemory().
// We can't just use RtlSecureZeroMemory here directly, as in debug
// builds, RtlSecureZeroMemory() can't be inlined, and the stack
// memory it uses would get wiped by itself running, causing crashes.
for (volatile uint8_t* p = (volatile uint8_t*)bottom; p < top; p++) {
*p = 0;
}
}
} else
#endif
{
aOutNodeId = std::string(aOriginSalt, aOriginSalt + aOriginSaltLen);
}
return true;
}
} // namespace gmp
} // namespace mozilla

View File

@ -0,0 +1,22 @@
/* -*- Mode: C++; tab-width: 8; indent-tabs-mode: nil; c-basic-offset: 2 -*- */
/* vim: set ts=8 sts=2 et sw=2 tw=80: */
/* This Source Code Form is subject to the terms of the Mozilla Public
* License, v. 2.0. If a copy of the MPL was not distributed with this file,
* You can obtain one at http://mozilla.org/MPL/2.0/. */
#ifndef GMP_DEVICE_BINDING_h_
#define GMP_DEVICE_BINDING_h_
#include <string>
namespace mozilla {
namespace gmp {
bool CalculateGMPDeviceId(char* aOriginSalt,
uint32_t aOriginSaltLen,
std::string& aOutNodeId);
} // namespace gmp
} // namespace mozilla
#endif // GMP_DEVICE_BINDING_h_

View File

@ -8,6 +8,7 @@
#include <IOKit/network/IOEthernetInterface.h>
#include <IOKit/network/IONetworkInterface.h>
#include <vector>
#include <string>
// Note: The original machine_id_mac.cc code is in namespace rlz_lib below.
// It depends on some external files, which would bring in a log of Chromium