Compiler-rt part of MSan implementation of advanced origin tracking,
when we record not only creation point, but all locations where
an uninitialized value was stored to memory, too.
llvm-svn: 204152
Before we did it lazily on the first stack unwind in the thread.
It resulted in deadlock when the unwind was caused by memory allocation
inside pthread_getattr_np:
pthread_getattr_np <<< not reentable
GetThreadStackTopAndBottom
__interceptor_realloc
pthread_getattr_np
llvm-svn: 197026
Origin copying may destroy valid origin info. This is caused by
__msan_copy_origin widening the address range to the nearest 4-byte aligned
addresses both on the left and on the right. If the target buffer is
uninitialized and the source is fully initialized, this will result in
overriding valid origin of target buffer with stale (possibly 0) origin of the
source buffer.
With this change the widened origin is copied only if corresponding shadow
values are non zero.
llvm-svn: 193338
MSan poisons deallocated memory but it used to give it an invalid origin value,
resulting in confusing reports. This change associates deallocation stack trace
with such memory.
Note that MSan does not have quarantine, and use-after-free detection is very
limited.
llvm-svn: 190781
Summary:
The loader does not call mmap() through the PLT because it has to
bootstrap the process before libc is present. Hooking dlopen() isn't
enough either because the loader runs module initializers before
returning, and they could run arbitrary msan instrumented code.
If msandr is present, then we can intercept the mmaps from dlopen at the
syscall layer and clear the shadow there. If msandr is missing, we
clear the shadow after dlopen() and hope any initializers are trivial.
Reviewers: eugenis
CC: kcc, llvm-commits
Differential Revision: http://llvm-reviews.chandlerc.com/D509
llvm-svn: 176818
Moved everything users are not supposed to use to a private interface header.
Documented all public interfaces. Made them safe to use even if built without
MemorySanitizer.
llvm-svn: 173800