This patch does a few cleanup things:
1. The non-standalone scudo has a problem where GWP-ASan allocations
may not meet alignment requirements where Scudo was requested to have
alignment >= 16. Use the new GWP-ASan API to fix this.
2. The standalone variant loses some debugging information inside of
GWP-ASan because we ask GWP-ASan to allocate an aligned size in the
frontend. This means reports end up with 'UaF on a 16-byte allocation'
for a 1-byte allocation with 16-byte alignment. Also use the new API to
fix this.
3. Add post-alloc hooks for GWP-ASan intercepted allocations, and add
stats tracking for GWP-ASan allocations.
4. Add a small test that checks the alignment of the frontend
allocator, so that it can be used under GWP-ASan torture mode.
5. Add GWP-ASan torture mode as a testing configuration to catch these
regressions.
Depends on D94830, D95889.
Reviewed By: cryptoad
Differential Revision: https://reviews.llvm.org/D95884
GWP-ASan is the "production" variant as compiled by compiler-rt, and it's useful to be able to benchmark changes in GWP-ASan or Scudo's GWP-ASan hooks across versions. GWP-ASan is sampled, and sampled allocations are much slower, but given the amount of allocations that happen under test here - we actually get a reasonable representation of GWP-ASan's negligent performance impact between runs.
Reviewed By: cryptoad
Differential Revision: https://reviews.llvm.org/D101865
According to:
https://docs.python.org/3/library/subprocess.html#subprocess.Popen.poll
poll can return None if the process hasn't terminated.
I'm not quite sure how addr2line could end up closing the pipe without
terminating but we did see this happen on one of our bots:
```
<...>scripts/asan_symbolize.py",
line 211, in symbolize
logging.debug("addr2line exited early (broken pipe), returncode=%d"
% self.pipe.poll())
TypeError: %d format: a number is required, not NoneType
```
Handle None by printing a message that we couldn't get the return
code.
Reviewed By: delcypher
Differential Revision: https://reviews.llvm.org/D101891
Address sanitizer can detect stack exhaustion via its SEGV handler, which is
executed on a separate stack using the sigaltstack mechanism. When libFuzzer is
used with address sanitizer, it installs its own signal handlers which defer to
those put in place by the sanitizer before performing additional actions. In the
particular case of a stack overflow, the current setup fails because libFuzzer
doesn't preserve the flag for executing the signal handler on a separate stack:
when we run out of stack space, the operating system can't run the SEGV handler,
so address sanitizer never reports the issue. See the included test for an
example.
This commit fixes the issue by making libFuzzer preserve the SA_ONSTACK flag
when installing its signal handlers; the dedicated signal-handler stack set up
by the sanitizer runtime appears to be large enough to support the additional
frames from the fuzzer.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D101824
Fixes compilation on Android which has a TSDSharedRegistry object in the config.
Reviewed By: cryptoad, vitalybuka
Differential Revision: https://reviews.llvm.org/D101951
Operator new must align allocations for types with large alignment.
Before c++17 behavior was implementation defined and both clang and gc++
before 11 ignored alignment. Miss-aligned objects mysteriously crashed
tests on Ubuntu 14.
Alternatives are compile with -std=c++17 or -faligned-new, but they were
discarded as less portable.
Reviewed By: hctim
Differential Revision: https://reviews.llvm.org/D101874
The problem was introduced in D100348.
It's really hard to trigger the bug in a stress test - the race is just too
narrow - but the new checks in Thread::Init should at least provide usable
diagnostic if the problem ever returns.
Differential Revision: https://reviews.llvm.org/D101881
This relates to https://reviews.llvm.org/D95835.
In DFSan origin tracking we use StackDepot to record
stack traces and origin traces (like MSan origin tracking).
For at least two reasons, we wanted to control StackDepot's memory cost
1) We may use DFSan origin tracking to monitor programs that run for
many days. This may eventually use too much memory for StackDepot.
2) DFSan supports flush shadow memory to reduce overhead. After flush,
all existing IDs in StackDepot are not valid because no one will
refer to them.
Currently, the position hint of an entry in the persistent auto
dictionary is fixed to 1. As a consequence, with a 50% chance, the entry
is applied right after the first byte of the input. As the position 1
does not appear to have any particular significance, this is likely a
bug that may have been caused by confusing the constructor parameter
with a success count.
This commit resolves the issue by preserving any existing position hint
or disabling the hint if the original entry didn't have one.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D101686
Code patterns like this are common, `#` at the line beginning
(https://google.github.io/styleguide/cppguide.html#Preprocessor_Directives),
one space indentation for if/elif/else directives.
```
#if SANITIZER_LINUX
# if defined(__aarch64__)
# endif
#endif
```
However, currently clang-format wants to reformat the code to
```
#if SANITIZER_LINUX
#if defined(__aarch64__)
#endif
#endif
```
This significantly harms readability in my review. Use `IndentPPDirectives:
AfterHash` to defeat the diagnostic. clang-format will now suggest:
```
#if SANITIZER_LINUX
# if defined(__aarch64__)
# endif
#endif
```
Unfortunately there is no clang-format option using indent with 1 for
just preprocessor directives. However, this is still one step forward
from the current behavior.
Reviewed By: #sanitizers, vitalybuka
Differential Revision: https://reviews.llvm.org/D100238
The Scudo C unit tests are currently non-hermetic. In particular, adding
or removing a transfer batch is a global state of the allocator that
persists between tests. This can cause flakiness in
ScudoWrappersCTest.MallInfo, because the creation or teardown of a batch
causes mallinfo's uordblks or fordblks to move up or down by the size of
a transfer batch on malloc/free.
It's my opinion that uordblks and fordblks should track the statistics
related to the user's malloc() and free() usage, and not the state of
the internal allocator structures. Thus, excluding the transfer batches
from stat collection does the trick and makes these tests pass.
Repro instructions of the bug:
1. ninja ./projects/compiler-rt/lib/scudo/standalone/tests/ScudoCUnitTest-x86_64-Test
2. ./projects/compiler-rt/lib/scudo/standalone/tests/ScudoCUnitTest-x86_64-Test --gtest_filter=ScudoWrappersCTest.MallInfo
Reviewed By: cryptoad
Differential Revision: https://reviews.llvm.org/D101653
In the overwrite branch of MutationDispatcher::ApplyDictionaryEntry in
FuzzerMutate.cpp, the index Idx at which W.size() bytes are overwritten
with the word W is chosen uniformly at random in the interval
[0, Size - W.size()). This means that Idx + W.size() will always be
strictly less than Size, i.e., the last byte of the current unit will
never be overwritten.
This is fixed by adding 1 to the exclusive upper bound.
Addresses https://bugs.llvm.org/show_bug.cgi?id=49989.
Reviewed By: morehouse
Differential Revision: https://reviews.llvm.org/D101625
Currently we have a bit of a mess related to tids:
- sanitizers re-declare kInvalidTid multiple times
- some call it kUnknownTid
- implicit assumptions that main tid is 0
- asan/memprof claim their tids need to fit into 24 bits,
but this does not seem to be true anymore
- inconsistent use of u32/int to store tids
Introduce kInvalidTid/kMainTid in sanitizer_common
and use them consistently.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101428
Commit efd254b636 ("tsan: fix deadlock in pthread_atfork callbacks")
fixed another deadlock related to atfork handling.
But builders with DCHECKs enabled reported failures of
pthread_atfork_deadlock2.c and pthread_atfork_deadlock3.c tests
related to the fact that we hold runtime locks on interceptor exit:
https://lab.llvm.org/buildbot/#/builders/70/builds/6727
This issue is somewhat inherent to the current approach,
we indeed execute user code (atfork callbacks) with runtime lock held.
Refactor fork handling to not run user code (atfork callbacks)
with runtime locks held. This change does this by installing
own atfork callbacks during runtime initialization.
Atfork callbacks run in LIFO order, so the expectation is that
our callbacks run last, right before the actual fork.
This way we lock runtime mutexes around fork, but not around
user callbacks.
Extend tests to also install after fork callbacks just to cover
more scenarios. Some tests also started reporting real races
that we previously suppressed.
Also extend tests to cover fork syscall support.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101517
This is to help review refactor the allocator code.
So it is easy to see which are the real public interfaces.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101586
To see how to extract a shared allocator interface for D101204,
found some unused code. Tests passed. Are they safe to remove?
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101559
We've got a user report about heap block allocator overflow.
Bump the L1 capacity of all dense slab allocators to maximum
and be careful to not page the whole L1 array in from .bss.
If OS uses huge pages, this still may cause a limited RSS increase
due to boundary huge pages, but avoiding that looks hard.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101161
While implementing support for the float128 routines on x86_64, I noticed
that __builtin_isinf() was returning true for 128-bit floating point
values that are not infinite when compiling with GCC and using the
compiler-rt implementation of the soft-float comparison functions.
After stepping through the assembly, I discovered that this was caused by
GCC assuming a sign-extended 64-bit -1 result, but our implementation
returns an enum (which then has zeroes in the upper bits) and therefore
causes the comparison with -1 to fail.
Fix this by using a CMP_RESULT typedef and add a static_assert that it
matches the GCC soft-float comparison return type when compiling with GCC
(GCC has a __libgcc_cmp_return__ mode that can be used for this purpose).
Also move the 3 copies of the same code to a shared .inc file.
Reviewed By: compnerd
Differential Revision: https://reviews.llvm.org/D98205
COMPILER_RT_TSAN_DEBUG_OUTPUT enables TSAN_COLLECT_STATS,
which changes layout of runtime structs (some structs contain
stats when the option is enabled).
It's not OK to build runtime with the define, but tests without it.
The error is detected by build_consistency_stats/nostats.
Fix this by defining TSAN_COLLECT_STATS for tests to match the runtime.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101386
Commit efd254b636 ("tsan: fix deadlock in pthread_atfork callbacks")
fixed another deadlock related to atfork handling.
But builders with DCHECKs enabled reported failures of
pthread_atfork_deadlock2.c and pthread_atfork_deadlock3.c tests
related to the fact that we hold runtime locks on interceptor exit:
https://lab.llvm.org/buildbot/#/builders/70/builds/6727
This issue is somewhat inherent to the current approach,
we indeed execute user code (atfork callbacks) with runtime lock held.
Refactor fork handling to not run user code (atfork callbacks)
with runtime locks held. This change does this by installing
own atfork callbacks during runtime initialization.
Atfork callbacks run in LIFO order, so the expectation is that
our callbacks run last, right before the actual fork.
This way we lock runtime mutexes around fork, but not around
user callbacks.
Extend tests to also install after fork callbacks just to cover
more scenarios. Some tests also started reporting real races
that we previously suppressed.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101385
We take report/thread_registry locks around fork.
This means we cannot report any bugs in atfork handlers.
We resolved this by enabling per-thread ignores around fork.
This resolved some of the cases, but not all.
The added test triggers a race report from a signal handler
called from atfork callback, we reset per-thread ignores
around signal handlers, so we tried to report it and deadlocked.
But there are more cases: a signal handler can be called
synchronously if it's sent to itself. Or any other report
types would cause deadlocks as well: mutex misuse,
signal handler spoiling errno, etc.
Disable all reports for the duration of fork with
thr->suppress_reports and don't re-enable them around
signal handlers.
Reviewed By: vitalybuka
Differential Revision: https://reviews.llvm.org/D101154
This is undefined if SANITIZER_SYMBOLIZER_MARKUP is 1, which is the case for
Fuchsia, and will result in a undefined symbol error. This function is needed
by hwasan for online symbolization, but is not needed for us since we do
offline symbolization.
Differential Revision: https://reviews.llvm.org/D99386
This reapplies 1e1d75b190, which was reverted in ce1a4d5323 due to build
failures.
The unconditional dependencies on clang and llvm-jitlink in
compiler-rt/test/orc/CMakeLists.txt have been removed -- they don't appear to
be necessary, and I suspect they're the cause of the build failures seen
earlier.
Some builders failed with a missing clang dependency. E.g.
CMake Error at /Users/buildslave/jenkins/workspace/clang-stage1-RA/clang-build \
/lib/cmake/llvm/AddLLVM.cmake:1786 (add_dependencies):
The dependency target "clang" of target "check-compiler-rt" does not exist.
Reverting while I investigate.
This reverts commit 1e1d75b190.
This patch does a few cleanup things:
1. The non-standalone scudo has a problem where GWP-ASan allocations
may not meet alignment requirements where Scudo was requested to have
alignment >= 16. Use the new GWP-ASan API to fix this.
2. The standalone variant loses some debugging information inside of
GWP-ASan because we ask GWP-ASan to allocate an aligned size in the
frontend. This means reports end up with 'UaF on a 16-byte allocation'
for a 1-byte allocation with 16-byte alignment. Also use the new API to
fix this.
3. Add post-alloc hooks for GWP-ASan intercepted allocations, and add
stats tracking for GWP-ASan allocations.
4. Add a small test that checks the alignment of the frontend
allocator, so that it can be used under GWP-ASan torture mode.
5. Add GWP-ASan torture mode as a testing configuration to catch these
regressions.
Depends on D94830, D95889.
Reviewed By: cryptoad
Differential Revision: https://reviews.llvm.org/D95884
From a cache perspective it's better to store the header immediately
after loading it. If we delay this operation until after we've
retagged it's more likely that our header will have been evicted from
the cache and we'll need to fetch it again in order to perform the
compare-exchange operation.
For similar reasons, store the deallocation stack before retagging
instead of afterwards.
Differential Revision: https://reviews.llvm.org/D101137
It looks like there's some old version of gcc that doesn't like this
static_assert (I couldn't reproduce the issue with gcc 8, 9 or 10).
Work around the issue by only checking the static_assert under clang,
which should provide sufficient coverage.
Should hopefully fix this buildbot:
https://lab.llvm.org/buildbot/#/builders/112/builds/5356
With AndroidSizeClassMap all of the LSBs are in the range 4-6 so we
only need 2 bits of information per size class. Furthermore we have
32 size classes, which conveniently lets us fit all of the information
into a 64-bit integer. Do so if possible so that we can avoid a table
lookup entirely.
Differential Revision: https://reviews.llvm.org/D101105
In the most common case we call computeOddEvenMaskForPointerMaybe()
from quarantineOrDeallocateChunk(), in which case we need to look up
the class size from the SizeClassMap in order to compute the LSB. Since
we need to do a lookup anyway, we may as well look up the LSB itself
and avoid computing it every time.
While here, switch to a slightly more efficient way of computing the
odd/even mask.
Differential Revision: https://reviews.llvm.org/D101018
Mitch Phillips