Asan crashes on malloc

[avaer]

I am getting this stack on calling a WASM function; the error comes from ASAN:

Uncaught (in promise) RuntimeError: memory access out of bounds
  | __sanitizer::ReportFile::Write(char const*, unsigned long) | @ | PsAssert.cpp:69
-- | -- | -- | --
  | __sanitizer::RawWrite(char const*) | @ | PsAssert.cpp:69
  | __sanitizer::MmapNoReserveOrDie(unsigned long, char const*) | @ | PsAssert.cpp:69
  | __asan::FakeStack::Create(unsigned long) | @ | PsAssert.cpp:69
  | __asan::AsanThread::AsyncSignalSafeLazyInitFakeStack() | @ | PsAssert.cpp:69
  | __asan_stack_malloc_0 | @ | PsAssert.cpp:69
  | sleep | @ | PsAssert.cpp:69
  | __sanitizer::SleepForSeconds(int) | @ | PsAssert.cpp:69
  | __sanitizer::CheckFailed(char const*, int, char const*, unsigned long long, unsigned long long) | @ | PsAssert.cpp:69
  | __sanitizer::LargeMmapAllocator<__asan::AsanMapUnmapCallback, __sanitizer::LargeMmapAllocatorPtrArrayStatic, __sanitizer::LocalAddressSpaceView>::Allocate(__sanitizer::AllocatorStats*, unsigned long, unsigned long) | @ | PsAssert.cpp:69
  | __sanitizer::CombinedAllocator<__sanitizer::SizeClassAllocator32<__asan::AP32<__sanitizer::LocalAddressSpaceView> >, __sanitizer::LargeMmapAllocatorPtrArrayStatic>::Allocate(__sanitizer::SizeClassAllocator32LocalCache<__sanitizer::SizeClassAllocator32<__asan::AP32<__sanitizer::LocalAddressSpaceView> > >*, unsigned long, unsigned long) | @ | PsAssert.cpp:69
  | __asan::Allocator::Allocate(unsigned long, unsigned long, __sanitizer::BufferedStackTrace*, __asan::AllocType, bool) | @ | PsAssert.cpp:69
  | __asan::asan_malloc(unsigned long, __sanitizer::BufferedStackTrace*) | @ | PsAssert.cpp:69
  | malloc | @ | PsAssert.cpp:69
  | ArenaAllocator::ArenaAllocator(unsigned int) | @ | subparcel.cc:44
  | makeArenaAllocator

Build command:

emcc -s WASM=1 -s USE_PTHREADS=1 -s PTHREAD_POOL_SIZE=1 -s TOTAL_MEMORY=599982080 -s MODULARIZE=0 -O3 -g3 -s ASSERTIONS=1 -fsanitize=address -s DETERMINISTIC=1 -s FORCE_FILESYSTEM=1 -s INVOKE_RUN=0 \
  -IPhysX/physx/include -IPhysX/pxshared/include \
  -IPhysX/physx/source/foundation/include \
  -IPhysX/physx/source/pvd/include \
  -IPhysX/physx/source/simulationcontroller/include -IPhysX/physx/source/lowlevel/api/include \
  -IPhysX/physx/source/geomutils/include \
  -IPhysX/physx/source/scenequery/include \
  -IPhysX/physx/source/lowleveldynamics/include \
  -IPhysX/physx/source/lowlevel/software/include \
  -IPhysX/physx/source/lowlevelaabb/include \
  -IPhysX/physx/source/lowlevel/common/include/pipeline \
  -IPhysX/physx/source/lowlevel/common/include/utils \
  -IPhysX/physx/source/lowlevel/common/include/collision \
  -IPhysX/physx/source/geomutils/src -IPhysX/physx/source/geomutils/src/common -IPhysX/physx/source/geomutils/src/mesh -IPhysX/physx/source/geomutils/src/hf -IPhysX/physx/source/geomutils/src/convex -IPhysX/physx/source/geomutils/src/gjk \
  -IPhysX/physx/source/common/src \
  -IPhysX/physx/source/physx/src/buffering \
  -IPhysX/physx/source/physx/src \
  -IPhysX/physx/source/physxcooking/src/convex \
  -IPhysX/physx/source/physxcooking/src/mesh \
  -IPhysX/physx/source/physxextensions/src/serialization/File \
  -IPhysX/physx/source/physxcooking/src \
  -IPhysX/physx/source/simulationcontroller/src \
  -IPhysX/physx/source/geomutils/src/intersection \
  -IPhysX//physx/source/geomutils/src/ccd \
  -IPhysX/physx/source/geomutils/src/contact \
  -IPhysX/physx/source/geomutils/src/pcm \
  -IPhysX/physx/source/geomutils/src/distance \
  -IPhysX/physx/source/geomutils/src/sweep \
  -Iearcut \
  -IRectBinPack/include \
  -Iconcaveman \
  objectize.cc vector.cc subparcel.cc geometry.cc collide.cc \
  FastNoise.cpp noise.cc march.cc biomes.cc \
  draco.o physx.o concaveman.o \
  -DNDEBUG -DPX_SIMD_DISABLED -DPX_EMSCRIPTEN=1 -DPX_COOKING \
  -I. \
  -o bin/geometry.js

[avaer]

I might have blown the memory limits; this seems to make a difference:

-s ASAN_SHADOW_SIZE=134217728

It should probably still not crash like this though.

[tlively]

I'll take a look at how the default ASan shadow size is computed to see if we can make it better or have it output some sort of diagnostic.

[stale[bot]]

This issue has been automatically marked as stale because there has been no activity in the past year. It will be closed automatically if no further activity occurs in the next 30 days. Feel free to re-open at any time if this issue is still relevant.

[tlively]

It looks like we fixed this in #12249.