tsan: align ThreadState to cache line

[dvyukov]

Summary

There are 2 reasons to do this:

1. We place hot data in the first cache line of ThreadState,

this assumed that it's cache-line-aligned but we never actually
enforced it (or it was lost at some point).

2. The new vector clock uses vector instructions and requires

data alignment. Later the new vector clock will be embedded in
ThreadState, then ensuring vector clock alignment will be
impossible w/o ThreadState alignment.

Depends on D110519.

[melver]

compiler-rt/lib/tsan/rtl/tsan_rtl.cpp

148

I assume this could only fail if the allocator messes things up?

[dvyukov]

compiler-rt/lib/tsan/rtl/tsan_rtl.cpp

148

Yes, if something messes things up. Maybe linker or maybe one of our custom allocations (we frequently allocate things in strange ways).
I had some failures related to ThreadState alignment in tests, but I can't reproduce them and not sure if there was something real, or just an inconsistent build (cmake seems to mishandle some of test dependencies sometimes).

[nemanjai]

Either this patch or https://reviews.llvm.org/D110519 has broken almost all the PPC bots. Please revert and let us know how we can help you figure out what needs to be done to investigate/fix this on the PPC bots.

Bot failures:
https://lab.llvm.org/buildbot/#/builders/52/builds/10854
https://lab.llvm.org/buildbot/#/builders/100/builds/9308
https://lab.llvm.org/buildbot/#/builders/93/builds/5134
https://lab.llvm.org/buildbot/#/builders/76/builds/4004
https://lab.llvm.org/buildbot/#/builders/105/builds/15435
https://lab.llvm.org/buildbot/#/builders/121/builds/11854

[dvyukov]

Hi Nemanja,

I hope this will fix the breakages:
https://reviews.llvm.org/D110629
If not, I think we need to remove the CHECK.
Sorry for the breakage.

compiler-rt/lib/tsan/rtl/tsan_rtl.h

#endif
   // PC of the last memory access, used to compute PC deltas in the trace.
   uptr trace_prev_pc;
   Sid sid;
   Epoch epoch;
 
   explicit ThreadState(Context *ctx, Tid tid, int unique_id, u64 epoch,
                        unsigned reuse_count, uptr stk_addr, uptr stk_size,
                        uptr tls_addr, uptr tls_size);
-};
+} ALIGNED(SANITIZER_CACHE_LINE_SIZE);
 
 #if !SANITIZER_GO
 #if SANITIZER_MAC || SANITIZER_ANDROID
 ThreadState *cur_thread();
 void set_cur_thread(ThreadState *thr);
 void cur_thread_finalize();
 inline void cur_thread_init() { }
 #else

compiler-rt/lib/tsan/rtl/tsan_rtl.cpp

#endif
       stk_size(stk_size),
       tls_addr(tls_addr),
       tls_size(tls_size)
 #if !SANITIZER_GO
       ,
       last_sleep_clock(tid)
 #endif
 {
+  CHECK_EQ(reinterpret_cast<uptr>(this) % SANITIZER_CACHE_LINE_SIZE, 0);
 #if !SANITIZER_GO
   shadow_stack_pos = shadow_stack;
   shadow_stack_end = shadow_stack + kShadowStackSize;
 #else
   // Setup dynamic shadow stack.
   const int kInitStackSize = 8;
   shadow_stack = (uptr *)Alloc(kInitStackSize * sizeof(uptr));
   shadow_stack_pos = shadow_stack;