Details
Diff Detail
- Repository
- rG LLVM Github Monorepo
Event Timeline
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
302 | do we know the corresponding x86-64 patch? |
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
302 | https://android.googlesource.com/platform/system/core/+/master/init/security.cpp#100 points to 9e08f57d684a x86: mm: support ARCH_MMAP_RND_BITS as the equivalent for x86. | |
319 | This function seems to be needed only for aarch64 [1]? Add a separate #if guard around this call with the old check #if SANITIZER_LINUX && defined(__aarch64__). [1] https://github.com/llvm/llvm-project/blob/main/compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp#L432 |
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
319 | For clarity, I'd duplicate the SANITIZER_LINUX here instead of nesting the pre-processor checks - because the ASLR personality check is unrelated to initializing the longjmp key. |
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
306 | nit |
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
302 | Is this 39-bit VA specific? I guess it's android only? I was recently notoced internal that ASLR is not working on TSAN arm64 linux, and want to undo this reexec |
compiler-rt/lib/tsan/rtl/tsan_platform_linux.cpp | ||
---|---|---|
302 |
If I'm reading this correctly, TSAN for arm64-linux doesn't need the re-exec? It's possible the address mappings here are for a different VA address space. Can you confirm if it's ok to relax the re-exec for all linux systems? |
do we know the corresponding x86-64 patch?