llvm-svn: 168692

llvm-svn: 168683

do the atomic operation under the sync object mutex
make acquire/release sync atomic with the operation itself
combine acquire and release into a single acq_rel operation

llvm-svn: 168682

this allows to build tsan runtime as dynamic library

llvm-svn: 168589

llvm-svn: 168584

llvm-svn: 168567

[asan/tsan] get rid of kPageSize completely in favor of GetPageSizeCached(). This makes the code friendly to more platforms

llvm-svn: 168537

llvm-svn: 168518

[asan] get rid of some of the uses of kPageSize. The intent is to get rid of it completely to support platforms with multiple possible page sizes.

llvm-svn: 168517

llvm-svn: 168506

llvm-svn: 168497

llvm-svn: 168496

llvm-svn: 168423

(do not execute synchronous signals in recursive interceptors)

llvm-svn: 168421

llvm-svn: 168150

llvm-svn: 168146

llvm-svn: 168060

llvm-svn: 168059

add interceptor for poll()
in addition process signals in every atomic op
in addition process signals in blocking libc functions

llvm-svn: 168050

llvm-svn: 167835

llvm-svn: 167834

llvm-svn: 167833

llvm-svn: 167832

llvm-svn: 167630

llvm-svn: 167617

llvm-svn: 167614

It's easier to call from compiler module.

llvm-svn: 167611

llvm-svn: 167580

llvm-svn: 167578

llvm-svn: 167575

llvm-svn: 167533

llvm-svn: 167532

llvm-svn: 167530

tsan: slightly relax requirements for lazy shadow memory (can overlap and may not be properly aligned)
it's problematic on windows where allocation granularity is much larger than page size

llvm-svn: 167466

llvm-svn: 167464

llvm-svn: 167458

llvm-svn: 167457

llvm-svn: 167456

llvm-svn: 167455

llvm-svn: 167454