Preloading libmozsandbox allows the symbol interpositions used by
sandboxing to be defined there instead of statically linked into the
executable; this patch also does that.
MozReview-Commit-ID: FL1QWLSKA0S
--HG--
rename : security/sandbox/linux/interpose/SandboxHooks.cpp => security/sandbox/linux/SandboxHooks.cpp
This way they'll continue to be at the beginning of the symbol search
path after mozsandbox returns to being a shared library instead of
statically linked into plugin-container.
--HG--
rename : security/sandbox/linux/SandboxHooks.cpp => security/sandbox/linux/interpose/SandboxHooks.cpp
The patch is generated from following command:
rgrep -l unused.h|xargs sed -i -e s,mozilla/unused.h,mozilla/Unused.h,
MozReview-Commit-ID: AtLcWApZfES
--HG--
rename : mfbt/unused.h => mfbt/Unused.h
Adds content sandbox metadata to parent and child crash reports:
Includes the value of pref security.sandbox.content.level,
whether or not the system is capable of sandboxing, if the
sandbox was successfully turned on, and (on Linux systems)
the sandbox capabilities flags.
New crash report keys:
"ContentSandboxLevel" in parent and content
"ContentSandboxCapable" in parent
"ContentSandboxEnabled" in content
"ContentSandboxCapabilities" in content on Linux
It's an annotation that is used a lot, and should be used even more, so a
shorter name is better.
MozReview-Commit-ID: 1VS4Dney4WX
--HG--
extra : rebase_source : b26919c1b0fcb32e5339adeef5be5becae6032cf
Bonus fix: don't start the chroot helper unless we're going to use
it. For this to matter, you'd need a system with unprivileged user
namespaces but no seccomp-bpf (or fake it with env vars) *and* to set
media.gmp.insecure.allow, so this is more to set a good example for
future changes to this code than for functional reasons.
This gives us a logging macro that's safe to use in async signal context
(cf. bug 1046210, where we needed this and didn't have it).
This patch also changes one of the format strings to work with
SafeSPrintf's format string dialect; upstream would probably take a
patch to handle those letters, but this is easier.
This gives us a logging macro that's safe to use in async signal context
(cf. bug 1046210, where we needed this and didn't have it).
This patch also changes one of the format strings to work with
SafeSPrintf's format string dialect; upstream would probably take a
patch to handle those letters, but this is easier.
We can now keep the part of the policy implemented by upcalls to
userspace in the same place as the part of the policy that's handled
entirely in the kernel. This will become more useful in the future
(e.g., bug 930258).
This is more complicated than I'd like it to be, because we don't have
a good way to combine a specific trap function's knowledge that we want
to get a crash dump with the SIGSYS handler's copy of the unprocessed
signal info (which breakpad wants). The bpf_dsl interface requires a
specific trap function type (via the TrapRegistry superclass), so even
if we implement our own registry we can't change what's passed to it.
Normally we could use thread-local storage to get around that, but it's
not async signal safe.
As a result there is an imperfect compromise: the trap function returns
a failure with ENOSYS, Chromium's SIGSYS handler writes it into the
context, our SIGSYS handler reads it back out and uses a copy of
the original signal context for the crash dump. Other error codes
(and returning ENOSYS via the seccomp-bpf policy itself) are handled
normally.
This completely rewrites SandboxFilter.cpp and removes SandboxAssembler.
System calls are now loosely grouped by what they do, now that order
doesn't matter, and most of the intersection the content and media
plugin whitelists is moved into a common superclass. Hopefully this
improves the readability and comprehensibility of the syscall policies.
Also, the macros that take the syscall name are gone, because a plain
case label usually suffices now (the CASES_FOR_thing macros are a little
unsightly, but they're relatively simple), and at one point we saw
strange macro expansion issues with system header files that #define'd
some syscall names.
The signal handling is not migrated yet, so Trap() actions can't be used
yet; the next patch will take care of that, and to keep the intermediate
state working there's a minimal shim.
Bonus fix: non-const global variables use the "g" prefix; "s" is for
static class members and static variables in a function (where the
default is to allocate a separate copy per instance/activation).
This needs more unit tests for the various pieces of what's going on
here (LinuxCapabilities, SandboxChroot, UnshareUserNamespace()) but
that's nontrivial due to needing a single-threaded process -- and
currently they can't be run on Mozilla's CI anyway due to needing user
namespaces, and local testing can just try using GMP and manually
inspecting the child process. So that will be a followup.
This means that B2G plugin-container must (dynamically) link against
libmozsandbox in order to call into it before initializing Binder.
(Desktop Linux plugin-container already contains the sandbox code.)
Specifically:
* SandboxCrash() uses internal Gecko interfaces, so stays in libxul.
* SandboxInfo moves to libxul from libmozsandbox, which no longer exists.
* Where libxul calls Set*Sandbox(), it uses weak symbols.
* Everything remains as it was on mobile.
This changes the interface so that the code which determines the flags
can live in one place, but checking the flags doesn't need to call into
another library.
Also removes the no-op wrappers for Set*Sandbox when disabled at build
time; nothing used them, one of them was unusable due to having the wrong
type, and all they really accomplish is allowing sloppiness with ifdefs
(which could hide actual mistakes).
This adds "hasSeccompBPF" for seccomp-bpf support; other "has" keys
will be added in the future (e.g., user namespaces).
This also adds "canSandboxContent" and "canSandboxMedia", which are
absent if the corresponding type of sandboxing isn't enabled at build
type (or is disabled with environment variables), and otherwise present
as a boolean indicating whether that type of sandboxing is supported.
Currently this is always the same as hasSeccompBPF, but that could change
in the future.
Some changes have been made to the "mozilla/Sandbox.h" interface to
support this; the idea is that the MOZ_DISABLE_*_SANDBOX environment
variables should be equivalent to disabling MOZ_*_SANDBOX at build time.
This creates libmozsandbox.so on builds that use sandboxing
(MOZ_CONTENT_SANDBOX or MOZ_GMP_SANDBOX).
The unavoidably libxul-dependent parts, for invoking the crash reporter
and printing the JS context, are separated into glue/SandboxCrash.cpp
and invoked via a callback.