libsanitizer: Merge with upstream

Merged revision: 7704fedfff6ef5676adb6415f3be0ac927d1a746

28 files changed, 219 insertions, 622 deletions

diff --git a/libsanitizer/tsan/Makefile.am b/libsanitizer/tsan/Makefile.am
index fb20ff7..dcb247e 100644
--- a/libsanitizer/tsan/Makefile.am
+++ b/libsanitizer/tsan/Makefile.am
@@ -43,13 +43,12 @@ tsan_files = \
 	tsan_rtl_report.cpp \
 	tsan_rtl_thread.cpp \
 	tsan_stack_trace.cpp \
-	tsan_stat.cpp \
 	tsan_suppressions.cpp \
 	tsan_symbolize.cpp \
 	tsan_sync.cpp 
 
 libtsan_la_SOURCES = $(tsan_files)
-EXTRA_libtsan_la_SOURCES = tsan_rtl_amd64.S tsan_rtl_aarch64.S tsan_rtl_mips64.S tsan_rtl_ppc64.S
+EXTRA_libtsan_la_SOURCES = tsan_rtl_amd64.S tsan_rtl_aarch64.S tsan_rtl_mips64.S tsan_rtl_ppc64.S tsan_rtl_s390x.S
 libtsan_la_LIBADD = $(top_builddir)/sanitizer_common/libsanitizer_common.la $(top_builddir)/interception/libinterception.la $(TSAN_TARGET_DEPENDENT_OBJECTS)
 libtsan_la_DEPENDENCIES = $(top_builddir)/sanitizer_common/libsanitizer_common.la $(top_builddir)/interception/libinterception.la $(TSAN_TARGET_DEPENDENT_OBJECTS)
 if LIBBACKTRACE_SUPPORTED
diff --git a/libsanitizer/tsan/Makefile.in b/libsanitizer/tsan/Makefile.in
index ce11d24..83617cf 100644
--- a/libsanitizer/tsan/Makefile.in
+++ b/libsanitizer/tsan/Makefile.in
@@ -155,8 +155,7 @@ am__objects_1 = tsan_clock.lo tsan_debugging.lo tsan_external.lo \
 	tsan_platform_posix.lo tsan_platform_windows.lo tsan_report.lo \
 	tsan_rtl.lo tsan_rtl_mutex.lo tsan_rtl_proc.lo \
 	tsan_rtl_report.lo tsan_rtl_thread.lo tsan_stack_trace.lo \
-	tsan_stat.lo tsan_suppressions.lo tsan_symbolize.lo \
-	tsan_sync.lo
+	tsan_suppressions.lo tsan_symbolize.lo tsan_sync.lo
 am_libtsan_la_OBJECTS = $(am__objects_1)
 libtsan_la_OBJECTS = $(am_libtsan_la_OBJECTS)
 AM_V_lt = $(am__v_lt_@AM_V@)
@@ -446,13 +445,12 @@ tsan_files = \
 	tsan_rtl_report.cpp \
 	tsan_rtl_thread.cpp \
 	tsan_stack_trace.cpp \
-	tsan_stat.cpp \
 	tsan_suppressions.cpp \
 	tsan_symbolize.cpp \
 	tsan_sync.cpp 
 
 libtsan_la_SOURCES = $(tsan_files)
-EXTRA_libtsan_la_SOURCES = tsan_rtl_amd64.S tsan_rtl_aarch64.S tsan_rtl_mips64.S tsan_rtl_ppc64.S
+EXTRA_libtsan_la_SOURCES = tsan_rtl_amd64.S tsan_rtl_aarch64.S tsan_rtl_mips64.S tsan_rtl_ppc64.S tsan_rtl_s390x.S
 libtsan_la_LIBADD =  \
 	$(top_builddir)/sanitizer_common/libsanitizer_common.la \
 	$(top_builddir)/interception/libinterception.la \
@@ -612,9 +610,9 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_rtl_ppc64.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_rtl_proc.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_rtl_report.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_rtl_s390x.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_rtl_thread.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_stack_trace.Plo@am__quote@
-@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_stat.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_suppressions.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_symbolize.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/tsan_sync.Plo@am__quote@
diff --git a/libsanitizer/tsan/tsan_clock.cpp b/libsanitizer/tsan/tsan_clock.cpp
index 8e51883..61848c2 100644
--- a/libsanitizer/tsan/tsan_clock.cpp
+++ b/libsanitizer/tsan/tsan_clock.cpp
@@ -80,14 +80,6 @@
 //   release-store operation by the thread with release_store_tid_ index.
 // release_store_reused_ - reuse count of release_store_tid_.
 
-// We don't have ThreadState in these methods, so this is an ugly hack that
-// works only in C++.
-#if !SANITIZER_GO
-# define CPP_STAT_INC(typ) StatInc(cur_thread(), typ)
-#else
-# define CPP_STAT_INC(typ) (void)0
-#endif
-
 namespace __tsan {
 
 static atomic_uint32_t *ref_ptr(ClockBlock *cb) {
@@ -138,14 +130,11 @@ void ThreadClock::ResetCached(ClockCache *c) {
 void ThreadClock::acquire(ClockCache *c, SyncClock *src) {
   DCHECK_LE(nclk_, kMaxTid);
   DCHECK_LE(src->size_, kMaxTid);
-  CPP_STAT_INC(StatClockAcquire);
 
   // Check if it's empty -> no need to do anything.
   const uptr nclk = src->size_;
-  if (nclk == 0) {
-    CPP_STAT_INC(StatClockAcquireEmpty);
+  if (nclk == 0)
     return;
-  }
 
   bool acquired = false;
   for (unsigned i = 0; i < kDirtyTids; i++) {
@@ -162,7 +151,6 @@ void ThreadClock::acquire(ClockCache *c, SyncClock *src) {
   // Check if we've already acquired src after the last release operation on src
   if (tid_ >= nclk || src->elem(tid_).reused != reused_) {
     // O(N) acquire.
-    CPP_STAT_INC(StatClockAcquireFull);
     nclk_ = max(nclk_, nclk);
     u64 *dst_pos = &clk_[0];
     for (ClockElem &src_elem : *src) {
@@ -180,7 +168,6 @@ void ThreadClock::acquire(ClockCache *c, SyncClock *src) {
   }
 
   if (acquired) {
-    CPP_STAT_INC(StatClockAcquiredSomething);
     last_acquire_ = clk_[tid_];
     ResetCached(c);
   }
@@ -223,7 +210,6 @@ void ThreadClock::releaseStoreAcquire(ClockCache *c, SyncClock *sc) {
   sc->release_store_reused_ = 0;
 
   if (acquired) {
-    CPP_STAT_INC(StatClockAcquiredSomething);
     last_acquire_ = clk_[tid_];
     ResetCached(c);
   }
@@ -240,7 +226,6 @@ void ThreadClock::release(ClockCache *c, SyncClock *dst) {
     return;
   }
 
-  CPP_STAT_INC(StatClockRelease);
   // Check if we need to resize dst.
   if (dst->size_ < nclk_)
     dst->Resize(c, nclk_);
@@ -257,12 +242,9 @@ void ThreadClock::release(ClockCache *c, SyncClock *dst) {
   }
 
   // O(N) release.
-  CPP_STAT_INC(StatClockReleaseFull);
   dst->Unshare(c);
   // First, remember whether we've acquired dst.
   bool acquired = IsAlreadyAcquired(dst);
-  if (acquired)
-    CPP_STAT_INC(StatClockReleaseAcquired);
   // Update dst->clk_.
   dst->FlushDirty();
   uptr i = 0;
@@ -272,8 +254,6 @@ void ThreadClock::release(ClockCache *c, SyncClock *dst) {
     i++;
   }
   // Clear 'acquired' flag in the remaining elements.
-  if (nclk_ < dst->size_)
-    CPP_STAT_INC(StatClockReleaseClearTail);
   dst->release_store_tid_ = kInvalidTid;
   dst->release_store_reused_ = 0;
   // If we've acquired dst, remember this fact,
@@ -285,7 +265,6 @@ void ThreadClock::release(ClockCache *c, SyncClock *dst) {
 void ThreadClock::ReleaseStore(ClockCache *c, SyncClock *dst) {
   DCHECK_LE(nclk_, kMaxTid);
   DCHECK_LE(dst->size_, kMaxTid);
-  CPP_STAT_INC(StatClockStore);
 
   if (dst->size_ == 0 && cached_idx_ != 0) {
     // Reuse the cached clock.
@@ -320,13 +299,11 @@ void ThreadClock::ReleaseStore(ClockCache *c, SyncClock *dst) {
   if (dst->release_store_tid_ == tid_ &&
       dst->release_store_reused_ == reused_ &&
       !HasAcquiredAfterRelease(dst)) {
-    CPP_STAT_INC(StatClockStoreFast);
     UpdateCurrentThread(c, dst);
     return;
   }
 
   // O(N) release-store.
-  CPP_STAT_INC(StatClockStoreFull);
   dst->Unshare(c);
   // Note: dst can be larger than this ThreadClock.
   // This is fine since clk_ beyond size is all zeros.
@@ -358,7 +335,6 @@ void ThreadClock::ReleaseStore(ClockCache *c, SyncClock *dst) {
 }
 
 void ThreadClock::acq_rel(ClockCache *c, SyncClock *dst) {
-  CPP_STAT_INC(StatClockAcquireRelease);
   acquire(c, dst);
   ReleaseStore(c, dst);
 }
@@ -370,7 +346,6 @@ void ThreadClock::UpdateCurrentThread(ClockCache *c, SyncClock *dst) const {
     SyncClock::Dirty *dirty = &dst->dirty_[i];
     const unsigned tid = dirty->tid();
     if (tid == tid_ || tid == kInvalidTid) {
-      CPP_STAT_INC(StatClockReleaseFast);
       dirty->set_tid(tid_);
       dirty->epoch = clk_[tid_];
       return;
@@ -379,7 +354,6 @@ void ThreadClock::UpdateCurrentThread(ClockCache *c, SyncClock *dst) const {
   // Reset all 'acquired' flags, O(N).
   // We are going to touch dst elements, so we need to unshare it.
   dst->Unshare(c);
-  CPP_STAT_INC(StatClockReleaseSlow);
   dst->elem(tid_).epoch = clk_[tid_];
   for (uptr i = 0; i < dst->size_; i++)
     dst->elem(i).reused = 0;
@@ -456,7 +430,6 @@ void SyncClock::ResetImpl() {
 }
 
 void SyncClock::Resize(ClockCache *c, uptr nclk) {
-  CPP_STAT_INC(StatClockReleaseResize);
   Unshare(c);
   if (nclk <= capacity()) {
     // Memory is already allocated, just increase the size.
diff --git a/libsanitizer/tsan/tsan_defs.h b/libsanitizer/tsan/tsan_defs.h
index f53787a..5c8f280 100644
--- a/libsanitizer/tsan/tsan_defs.h
+++ b/libsanitizer/tsan/tsan_defs.h
@@ -15,7 +15,6 @@
 
 #include "sanitizer_common/sanitizer_internal_defs.h"
 #include "sanitizer_common/sanitizer_libc.h"
-#include "tsan_stat.h"
 #include "ubsan/ubsan_platform.h"
 
 // Setup defaults for compile definitions.
@@ -23,10 +22,6 @@
 # define TSAN_NO_HISTORY 0
 #endif
 
-#ifndef TSAN_COLLECT_STATS
-# define TSAN_COLLECT_STATS 0
-#endif
-
 #ifndef TSAN_CONTAINS_UBSAN
 # if CAN_SANITIZE_UB && !SANITIZER_GO
 #  define TSAN_CONTAINS_UBSAN 1
@@ -107,23 +102,12 @@ void build_consistency_debug();
 void build_consistency_release();
 #endif
 
-#if TSAN_COLLECT_STATS
-void build_consistency_stats();
-#else
-void build_consistency_nostats();
-#endif
-
 static inline void USED build_consistency() {
 #if SANITIZER_DEBUG
   build_consistency_debug();
 #else
   build_consistency_release();
 #endif
-#if TSAN_COLLECT_STATS
-  build_consistency_stats();
-#else
-  build_consistency_nostats();
-#endif
 }
 
 template<typename T>
diff --git a/libsanitizer/tsan/tsan_interceptors.h b/libsanitizer/tsan/tsan_interceptors.h
index 29576ea..c5716f5 100644
--- a/libsanitizer/tsan/tsan_interceptors.h
+++ b/libsanitizer/tsan/tsan_interceptors.h
@@ -30,14 +30,14 @@ inline bool in_symbolizer() {
 
 }  // namespace __tsan
 
-#define SCOPED_INTERCEPTOR_RAW(func, ...) \
-    cur_thread_init(); \
-    ThreadState *thr = cur_thread(); \
-    const uptr caller_pc = GET_CALLER_PC(); \
-    ScopedInterceptor si(thr, #func, caller_pc); \
-    const uptr pc = StackTrace::GetCurrentPc(); \
-    (void)pc; \
-/**/
+#define SCOPED_INTERCEPTOR_RAW(func, ...)      \
+  cur_thread_init();                           \
+  ThreadState *thr = cur_thread();             \
+  const uptr caller_pc = GET_CALLER_PC();      \
+  ScopedInterceptor si(thr, #func, caller_pc); \
+  const uptr pc = GET_CURRENT_PC();            \
+  (void)pc;                                    \
+  /**/
 
 #define SCOPED_TSAN_INTERCEPTOR(func, ...) \
     SCOPED_INTERCEPTOR_RAW(func, __VA_ARGS__); \
diff --git a/libsanitizer/tsan/tsan_interceptors_mac.cpp b/libsanitizer/tsan/tsan_interceptors_mac.cpp
index ed10fcc..2d400c7 100644
--- a/libsanitizer/tsan/tsan_interceptors_mac.cpp
+++ b/libsanitizer/tsan/tsan_interceptors_mac.cpp
@@ -44,8 +44,9 @@ namespace __tsan {
 // actually aliases of each other, and we cannot have different interceptors for
 // them, because they're actually the same function.  Thus, we have to stay
 // conservative and treat the non-barrier versions as mo_acq_rel.
-static const morder kMacOrderBarrier = mo_acq_rel;
-static const morder kMacOrderNonBarrier = mo_acq_rel;
+static constexpr morder kMacOrderBarrier = mo_acq_rel;
+static constexpr morder kMacOrderNonBarrier = mo_acq_rel;
+static constexpr morder kMacFailureOrder = mo_relaxed;
 
 #define OSATOMIC_INTERCEPTOR(return_t, t, tsan_t, f, tsan_atomic_f, mo) \
   TSAN_INTERCEPTOR(return_t, f, t x, volatile t *ptr) {                 \
@@ -110,7 +111,7 @@ OSATOMIC_INTERCEPTORS_BITWISE(OSAtomicXor, fetch_xor,
     SCOPED_TSAN_INTERCEPTOR(f, old_value, new_value, ptr);                  \
     return tsan_atomic_f##_compare_exchange_strong(                         \
         (volatile tsan_t *)ptr, (tsan_t *)&old_value, (tsan_t)new_value,    \
-        kMacOrderNonBarrier, kMacOrderNonBarrier);                          \
+        kMacOrderNonBarrier, kMacFailureOrder);                             \
   }                                                                         \
                                                                             \
   TSAN_INTERCEPTOR(bool, f##Barrier, t old_value, t new_value,              \
@@ -118,7 +119,7 @@ OSATOMIC_INTERCEPTORS_BITWISE(OSAtomicXor, fetch_xor,
     SCOPED_TSAN_INTERCEPTOR(f##Barrier, old_value, new_value, ptr);         \
     return tsan_atomic_f##_compare_exchange_strong(                         \
         (volatile tsan_t *)ptr, (tsan_t *)&old_value, (tsan_t)new_value,    \
-        kMacOrderBarrier, kMacOrderNonBarrier);                             \
+        kMacOrderBarrier, kMacFailureOrder);                                \
   }
 
 OSATOMIC_INTERCEPTORS_CAS(OSAtomicCompareAndSwapInt, __tsan_atomic32, a32, int)
diff --git a/libsanitizer/tsan/tsan_interceptors_posix.cpp b/libsanitizer/tsan/tsan_interceptors_posix.cpp
index 2651e22..6808f2e 100644
--- a/libsanitizer/tsan/tsan_interceptors_posix.cpp
+++ b/libsanitizer/tsan/tsan_interceptors_posix.cpp
@@ -71,7 +71,8 @@ struct ucontext_t {
 };
 #endif
 
-#if defined(__x86_64__) || defined(__mips__) || SANITIZER_PPC64V1
+#if defined(__x86_64__) || defined(__mips__) || SANITIZER_PPC64V1 || \
+    defined(__s390x__)
 #define PTHREAD_ABI_BASE  "GLIBC_2.3.2"
 #elif defined(__aarch64__) || SANITIZER_PPC64V2
 #define PTHREAD_ABI_BASE  "GLIBC_2.17"
@@ -2270,6 +2271,7 @@ static void HandleRecvmsg(ThreadState *thr, uptr pc,
 #define NEED_TLS_GET_ADDR
 #endif
 #undef SANITIZER_INTERCEPT_TLS_GET_ADDR
+#define SANITIZER_INTERCEPT_TLS_GET_OFFSET 1
 #undef SANITIZER_INTERCEPT_PTHREAD_SIGMASK
 
 #define COMMON_INTERCEPT_FUNCTION(name) INTERCEPT_FUNCTION(name)
@@ -2585,6 +2587,20 @@ static void syscall_post_fork(uptr pc, int pid) {
 #include "sanitizer_common/sanitizer_syscalls_netbsd.inc"
 
 #ifdef NEED_TLS_GET_ADDR
+
+static void handle_tls_addr(void *arg, void *res) {
+  ThreadState *thr = cur_thread();
+  if (!thr)
+    return;
+  DTLS::DTV *dtv = DTLS_on_tls_get_addr(arg, res, thr->tls_addr,
+                                        thr->tls_addr + thr->tls_size);
+  if (!dtv)
+    return;
+  // New DTLS block has been allocated.
+  MemoryResetRange(thr, 0, dtv->beg, dtv->size);
+}
+
+#if !SANITIZER_S390
 // Define own interceptor instead of sanitizer_common's for three reasons:
 // 1. It must not process pending signals.
 //    Signal handlers may contain MOVDQA instruction (see below).
@@ -2597,18 +2613,18 @@ static void syscall_post_fork(uptr pc, int pid) {
 // execute MOVDQA with stack addresses.
 TSAN_INTERCEPTOR(void *, __tls_get_addr, void *arg) {
   void *res = REAL(__tls_get_addr)(arg);
-  ThreadState *thr = cur_thread();
-  if (!thr)
-    return res;
-  DTLS::DTV *dtv = DTLS_on_tls_get_addr(arg, res, thr->tls_addr,
-                                        thr->tls_addr + thr->tls_size);
-  if (!dtv)
-    return res;
-  // New DTLS block has been allocated.
-  MemoryResetRange(thr, 0, dtv->beg, dtv->size);
+  handle_tls_addr(arg, res);
+  return res;
+}
+#else // SANITIZER_S390
+TSAN_INTERCEPTOR(uptr, __tls_get_addr_internal, void *arg) {
+  uptr res = __tls_get_offset_wrapper(arg, REAL(__tls_get_offset));
+  char *tp = static_cast<char *>(__builtin_thread_pointer());
+  handle_tls_addr(arg, res + tp);
   return res;
 }
 #endif
+#endif
 
 #if SANITIZER_NETBSD
 TSAN_INTERCEPTOR(void, _lwp_exit) {
@@ -2830,7 +2846,12 @@ void InitializeInterceptors() {
   TSAN_INTERCEPT(_exit);
 
 #ifdef NEED_TLS_GET_ADDR
+#if !SANITIZER_S390
   TSAN_INTERCEPT(__tls_get_addr);
+#else
+  TSAN_INTERCEPT(__tls_get_addr_internal);
+  TSAN_INTERCEPT(__tls_get_offset);
+#endif
 #endif
 
   TSAN_MAYBE_INTERCEPT__LWP_EXIT;
diff --git a/libsanitizer/tsan/tsan_interface.h b/libsanitizer/tsan/tsan_interface.h
index 6e022b5..124aa2f 100644
--- a/libsanitizer/tsan/tsan_interface.h
+++ b/libsanitizer/tsan/tsan_interface.h
@@ -196,7 +196,8 @@ typedef unsigned short a16;
 typedef unsigned int       a32;
 typedef unsigned long long a64;
 #if !SANITIZER_GO && (defined(__SIZEOF_INT128__) \
-    || (__clang_major__ * 100 + __clang_minor__ >= 302)) && !defined(__mips64)
+    || (__clang_major__ * 100 + __clang_minor__ >= 302)) && \
+    !defined(__mips64) && !defined(__s390x__)
 __extension__ typedef __int128 a128;
 # define __TSAN_HAS_INT128 1
 #else
diff --git a/libsanitizer/tsan/tsan_interface_ann.cpp b/libsanitizer/tsan/tsan_interface_ann.cpp
index 99516d9..175855f 100644
--- a/libsanitizer/tsan/tsan_interface_ann.cpp
+++ b/libsanitizer/tsan/tsan_interface_ann.cpp
@@ -49,8 +49,6 @@ class ScopedAnnotation {
       return ret; \
     ThreadState *thr = cur_thread(); \
     const uptr caller_pc = (uptr)__builtin_return_address(0); \
-    StatInc(thr, StatAnnotation); \
-    StatInc(thr, Stat##typ); \
     ScopedAnnotation sa(thr, __func__, caller_pc); \
     const uptr pc = StackTrace::GetCurrentPc(); \
     (void)pc; \
@@ -77,9 +75,7 @@ struct DynamicAnnContext {
   ExpectRace expect;
   ExpectRace benign;
 
-  DynamicAnnContext()
-    : mtx(MutexTypeAnnotations, StatMtxAnnotations) {
-  }
+  DynamicAnnContext() : mtx(MutexTypeAnnotations) {}
 };
 
 static DynamicAnnContext *dyn_ann_ctx;
diff --git a/libsanitizer/tsan/tsan_interface_atomic.cpp b/libsanitizer/tsan/tsan_interface_atomic.cpp
index 3f459af..21fe4a1 100644
--- a/libsanitizer/tsan/tsan_interface_atomic.cpp
+++ b/libsanitizer/tsan/tsan_interface_atomic.cpp
@@ -402,34 +402,45 @@ static T NoTsanAtomicCAS(volatile T *a, T c, T v, morder mo, morder fmo) {
 template<typename T>
 static bool AtomicCAS(ThreadState *thr, uptr pc,
     volatile T *a, T *c, T v, morder mo, morder fmo) {
-  (void)fmo;  // Unused because llvm does not pass it yet.
+  // 31.7.2.18: "The failure argument shall not be memory_order_release
+  // nor memory_order_acq_rel". LLVM (2021-05) fallbacks to Monotonic
+  // (mo_relaxed) when those are used.
+  CHECK(IsLoadOrder(fmo));
+
   MemoryWriteAtomic(thr, pc, (uptr)a, SizeLog<T>());
   SyncVar *s = 0;
-  bool write_lock = mo != mo_acquire && mo != mo_consume;
-  if (mo != mo_relaxed) {
+  bool write_lock = IsReleaseOrder(mo);
+
+  if (mo != mo_relaxed || fmo != mo_relaxed)
     s = ctx->metamap.GetOrCreateAndLock(thr, pc, (uptr)a, write_lock);
+
+  T cc = *c;
+  T pr = func_cas(a, cc, v);
+  bool success = pr == cc;
+  if (!success) {
+    *c = pr;
+    mo = fmo;
+  }
+
+  if (s) {
     thr->fast_state.IncrementEpoch();
     // Can't increment epoch w/o writing to the trace as well.
     TraceAddEvent(thr, thr->fast_state, EventTypeMop, 0);
-    if (IsAcqRelOrder(mo))
+
+    if (success && IsAcqRelOrder(mo))
       AcquireReleaseImpl(thr, pc, &s->clock);
-    else if (IsReleaseOrder(mo))
+    else if (success && IsReleaseOrder(mo))
       ReleaseImpl(thr, pc, &s->clock);
     else if (IsAcquireOrder(mo))
       AcquireImpl(thr, pc, &s->clock);
-  }
-  T cc = *c;
-  T pr = func_cas(a, cc, v);
-  if (s) {
+
     if (write_lock)
       s->mtx.Unlock();
     else
       s->mtx.ReadUnlock();
   }
-  if (pr == cc)
-    return true;
-  *c = pr;
-  return false;
+
+  return success;
 }
 
 template<typename T>
@@ -481,7 +492,6 @@ static morder convert_morder(morder mo) {
     const uptr callpc = (uptr)__builtin_return_address(0); \
     uptr pc = StackTrace::GetCurrentPc(); \
     mo = convert_morder(mo); \
-    AtomicStatInc(thr, sizeof(*a), mo, StatAtomic##func); \
     ScopedAtomic sa(thr, callpc, a, mo, __func__); \
     return Atomic##func(thr, pc, __VA_ARGS__); \
 /**/
@@ -502,22 +512,6 @@ class ScopedAtomic {
   ThreadState *thr_;
 };
 
-static void AtomicStatInc(ThreadState *thr, uptr size, morder mo, StatType t) {
-  StatInc(thr, StatAtomic);
-  StatInc(thr, t);
-  StatInc(thr, size == 1 ? StatAtomic1
-             : size == 2 ? StatAtomic2
-             : size == 4 ? StatAtomic4
-             : size == 8 ? StatAtomic8
-             :             StatAtomic16);
-  StatInc(thr, mo == mo_relaxed ? StatAtomicRelaxed
-             : mo == mo_consume ? StatAtomicConsume
-             : mo == mo_acquire ? StatAtomicAcquire
-             : mo == mo_release ? StatAtomicRelease
-             : mo == mo_acq_rel ? StatAtomicAcq_Rel
-             :                    StatAtomicSeq_Cst);
-}
-
 extern "C" {
 SANITIZER_INTERFACE_ATTRIBUTE
 a8 __tsan_atomic8_load(const volatile a8 *a, morder mo) {
diff --git a/libsanitizer/tsan/tsan_mman.cpp b/libsanitizer/tsan/tsan_mman.cpp
index 45a39f0..7765bc0 100644
--- a/libsanitizer/tsan/tsan_mman.cpp
+++ b/libsanitizer/tsan/tsan_mman.cpp
@@ -70,10 +70,7 @@ struct GlobalProc {
   Mutex mtx;
   Processor *proc;
 
-  GlobalProc()
-      : mtx(MutexTypeGlobalProc, StatMtxGlobalProc)
-      , proc(ProcCreate()) {
-  }
+  GlobalProc() : mtx(MutexTypeGlobalProc), proc(ProcCreate()) {}
 };
 
 static char global_proc_placeholder[sizeof(GlobalProc)] ALIGNED(64);
diff --git a/libsanitizer/tsan/tsan_mutex.cpp b/libsanitizer/tsan/tsan_mutex.cpp
index 7a0918f..d8b1826 100644
--- a/libsanitizer/tsan/tsan_mutex.cpp
+++ b/libsanitizer/tsan/tsan_mutex.cpp
@@ -207,15 +207,12 @@ class Backoff {
   static const int kActiveSpinCnt = 20;
 };
 
-Mutex::Mutex(MutexType type, StatType stat_type) {
+Mutex::Mutex(MutexType type) {
   CHECK_GT(type, MutexTypeInvalid);
   CHECK_LT(type, MutexTypeCount);
 #if SANITIZER_DEBUG
   type_ = type;
 #endif
-#if TSAN_COLLECT_STATS
-  stat_type_ = stat_type;
-#endif
   atomic_store(&state_, kUnlocked, memory_order_relaxed);
 }
 
@@ -236,9 +233,6 @@ void Mutex::Lock() {
       cmp = kUnlocked;
       if (atomic_compare_exchange_weak(&state_, &cmp, kWriteLock,
                                        memory_order_acquire)) {
-#if TSAN_COLLECT_STATS && !SANITIZER_GO
-        StatInc(cur_thread(), stat_type_, backoff.Contention());
-#endif
         return;
       }
     }
@@ -264,9 +258,6 @@ void Mutex::ReadLock() {
   for (Backoff backoff; backoff.Do();) {
     prev = atomic_load(&state_, memory_order_acquire);
     if ((prev & kWriteLock) == 0) {
-#if TSAN_COLLECT_STATS && !SANITIZER_GO
-      StatInc(cur_thread(), stat_type_, backoff.Contention());
-#endif
       return;
     }
   }
diff --git a/libsanitizer/tsan/tsan_mutex.h b/libsanitizer/tsan/tsan_mutex.h
index 80fdc6e..9a579ea 100644
--- a/libsanitizer/tsan/tsan_mutex.h
+++ b/libsanitizer/tsan/tsan_mutex.h
@@ -41,7 +41,7 @@ enum MutexType {
 
 class Mutex {
  public:
-  explicit Mutex(MutexType type, StatType stat_type);
+  explicit Mutex(MutexType type);
   ~Mutex();
 
   void Lock();
@@ -57,9 +57,6 @@ class Mutex {
 #if SANITIZER_DEBUG
   MutexType type_;
 #endif
-#if TSAN_COLLECT_STATS
-  StatType stat_type_;
-#endif
 
   Mutex(const Mutex&);
   void operator = (const Mutex&);
diff --git a/libsanitizer/tsan/tsan_platform.h b/libsanitizer/tsan/tsan_platform.h
index 101522d..8bd218e 100644
--- a/libsanitizer/tsan/tsan_platform.h
+++ b/libsanitizer/tsan/tsan_platform.h
@@ -365,6 +365,38 @@ struct Mapping47 {
 
 // Indicates the runtime will define the memory regions at runtime.
 #define TSAN_RUNTIME_VMA 1
+#elif defined(__s390x__)
+/*
+C/C++ on linux/s390x
+While the kernel provides a 64-bit address space, we have to restrict ourselves
+to 48 bits due to how e.g. SyncVar::GetId() works.
+0000 0000 1000 - 0e00 0000 0000: binary, modules, stacks - 14 TiB
+0e00 0000 0000 - 4000 0000 0000: -
+4000 0000 0000 - 8000 0000 0000: shadow - 64TiB (4 * app)
+8000 0000 0000 - 9000 0000 0000: -
+9000 0000 0000 - 9800 0000 0000: metainfo - 8TiB (0.5 * app)
+9800 0000 0000 - a000 0000 0000: -
+a000 0000 0000 - b000 0000 0000: traces - 16TiB (max history * 128k threads)
+b000 0000 0000 - be00 0000 0000: -
+be00 0000 0000 - c000 0000 0000: heap - 2TiB (max supported by the allocator)
+*/
+struct Mapping {
+  static const uptr kMetaShadowBeg = 0x900000000000ull;
+  static const uptr kMetaShadowEnd = 0x980000000000ull;
+  static const uptr kTraceMemBeg   = 0xa00000000000ull;
+  static const uptr kTraceMemEnd   = 0xb00000000000ull;
+  static const uptr kShadowBeg     = 0x400000000000ull;
+  static const uptr kShadowEnd     = 0x800000000000ull;
+  static const uptr kHeapMemBeg    = 0xbe0000000000ull;
+  static const uptr kHeapMemEnd    = 0xc00000000000ull;
+  static const uptr kLoAppMemBeg   = 0x000000001000ull;
+  static const uptr kLoAppMemEnd   = 0x0e0000000000ull;
+  static const uptr kHiAppMemBeg   = 0xc00000004000ull;
+  static const uptr kHiAppMemEnd   = 0xc00000004000ull;
+  static const uptr kAppMemMsk     = 0xb00000000000ull;
+  static const uptr kAppMemXor     = 0x100000000000ull;
+  static const uptr kVdsoBeg       = 0xfffffffff000ull;
+};
 #endif
 
 #elif SANITIZER_GO && !SANITIZER_WINDOWS && HAS_48_BIT_ADDRESS_SPACE
@@ -528,6 +560,28 @@ struct Mapping47 {
 
 #define TSAN_RUNTIME_VMA 1
 
+#elif SANITIZER_GO && defined(__s390x__)
+/*
+Go on linux/s390x
+0000 0000 1000 - 1000 0000 0000: executable and heap - 16 TiB
+1000 0000 0000 - 4000 0000 0000: -
+4000 0000 0000 - 8000 0000 0000: shadow - 64TiB (4 * app)
+8000 0000 0000 - 9000 0000 0000: -
+9000 0000 0000 - 9800 0000 0000: metainfo - 8TiB (0.5 * app)
+9800 0000 0000 - a000 0000 0000: -
+a000 0000 0000 - b000 0000 0000: traces - 16TiB (max history * 128k threads)
+*/
+struct Mapping {
+  static const uptr kMetaShadowBeg = 0x900000000000ull;
+  static const uptr kMetaShadowEnd = 0x980000000000ull;
+  static const uptr kTraceMemBeg   = 0xa00000000000ull;
+  static const uptr kTraceMemEnd   = 0xb00000000000ull;
+  static const uptr kShadowBeg     = 0x400000000000ull;
+  static const uptr kShadowEnd     = 0x800000000000ull;
+  static const uptr kAppMemBeg     = 0x000000001000ull;
+  static const uptr kAppMemEnd     = 0x100000000000ull;
+};
+
 #else
 # error "Unknown platform"
 #endif
diff --git a/libsanitizer/tsan/tsan_platform_linux.cpp b/libsanitizer/tsan/tsan_platform_linux.cpp
index e5b6690..cfe597e 100644
--- a/libsanitizer/tsan/tsan_platform_linux.cpp
+++ b/libsanitizer/tsan/tsan_platform_linux.cpp
@@ -391,6 +391,10 @@ static uptr UnmangleLongJmpSp(uptr mangled_sp) {
   return mangled_sp ^ xor_key;
 #elif defined(__mips__)
   return mangled_sp;
+#elif defined(__s390x__)
+  // tcbhead_t.stack_guard
+  uptr xor_key = ((uptr *)__builtin_thread_pointer())[5];
+  return mangled_sp ^ xor_key;
 #else
   #error "Unknown platform"
 #endif
@@ -411,6 +415,8 @@ static uptr UnmangleLongJmpSp(uptr mangled_sp) {
 #  define LONG_JMP_SP_ENV_SLOT 13
 # elif defined(__mips64)
 #  define LONG_JMP_SP_ENV_SLOT 1
+# elif defined(__s390x__)
+#  define LONG_JMP_SP_ENV_SLOT 9
 # else
 #  define LONG_JMP_SP_ENV_SLOT 6
 # endif
diff --git a/libsanitizer/tsan/tsan_platform_posix.cpp b/libsanitizer/tsan/tsan_platform_posix.cpp
index 73e1d45..1c6198c 100644
--- a/libsanitizer/tsan/tsan_platform_posix.cpp
+++ b/libsanitizer/tsan/tsan_platform_posix.cpp
@@ -72,11 +72,15 @@ void InitializeShadowMemory() {
   InitializeShadowMemoryPlatform();
 }
 
-static void ProtectRange(uptr beg, uptr end) {
+static bool TryProtectRange(uptr beg, uptr end) {
   CHECK_LE(beg, end);
   if (beg == end)
-    return;
-  if (beg != (uptr)MmapFixedNoAccess(beg, end - beg)) {
+    return true;
+  return beg == (uptr)MmapFixedNoAccess(beg, end - beg);
+}
+
+static void ProtectRange(uptr beg, uptr end) {
+  if (!TryProtectRange(beg, end)) {
     Printf("FATAL: ThreadSanitizer can not protect [%zx,%zx]\n", beg, end);
     Printf("FATAL: Make sure you are not using unlimited stack\n");
     Die();
@@ -118,6 +122,16 @@ void CheckAndProtect() {
   ProtectRange(TraceMemEnd(), HeapMemBeg());
   ProtectRange(HeapEnd(), HiAppMemBeg());
 #endif
+
+#if defined(__s390x__)
+  // Protect the rest of the address space.
+  const uptr user_addr_max_l4 = 0x0020000000000000ull;
+  const uptr user_addr_max_l5 = 0xfffffffffffff000ull;
+  // All the maintained s390x kernels support at least 4-level page tables.
+  ProtectRange(HiAppMemEnd(), user_addr_max_l4);
+  // Older s390x kernels may not support 5-level page tables.
+  TryProtectRange(user_addr_max_l4, user_addr_max_l5);
+#endif
 }
 #endif
 
diff --git a/libsanitizer/tsan/tsan_rtl.cpp b/libsanitizer/tsan/tsan_rtl.cpp
index 0efa997..bcf489a 100644
--- a/libsanitizer/tsan/tsan_rtl.cpp
+++ b/libsanitizer/tsan/tsan_rtl.cpp
@@ -77,7 +77,7 @@ void OnInitialize() {
 }
 #endif
 
-static char thread_registry_placeholder[sizeof(ThreadRegistry)];
+static ALIGNED(64) char thread_registry_placeholder[sizeof(ThreadRegistry)];
 
 static ThreadContextBase *CreateThreadContext(u32 tid) {
   // Map thread trace when context is created.
@@ -115,15 +115,15 @@ static const u32 kThreadQuarantineSize = 64;
 
 Context::Context()
     : initialized(),
-      report_mtx(MutexTypeReport, StatMtxReport),
+      report_mtx(MutexTypeReport),
       nreported(),
       nmissed_expected(),
       thread_registry(new (thread_registry_placeholder) ThreadRegistry(
           CreateThreadContext, kMaxTid, kThreadQuarantineSize, kMaxTidReuse)),
-      racy_mtx(MutexTypeRacy, StatMtxRacy),
+      racy_mtx(MutexTypeRacy),
       racy_stacks(),
       racy_addresses(),
-      fired_suppressions_mtx(MutexTypeFired, StatMtxFired),
+      fired_suppressions_mtx(MutexTypeFired),
       clock_alloc(LINKER_INITIALIZED, "clock allocator") {
   fired_suppressions.reserve(8);
 }
@@ -522,18 +522,14 @@ int Finalize(ThreadState *thr) {
 
   failed = OnFinalize(failed);
 
-#if TSAN_COLLECT_STATS
-  StatAggregate(ctx->stat, thr->stat);
-  StatOutput(ctx->stat);
-#endif
-
   return failed ? common_flags()->exitcode : 0;
 }
 
 #if !SANITIZER_GO
-void ForkBefore(ThreadState *thr, uptr pc) {
+void ForkBefore(ThreadState *thr, uptr pc) NO_THREAD_SAFETY_ANALYSIS {
   ctx->thread_registry->Lock();
   ctx->report_mtx.Lock();
+  ScopedErrorReportLock::Lock();
   // Suppress all reports in the pthread_atfork callbacks.
   // Reports will deadlock on the report_mtx.
   // We could ignore sync operations as well,
@@ -545,16 +541,18 @@ void ForkBefore(ThreadState *thr, uptr pc) {
   thr->ignore_interceptors++;
 }
 
-void ForkParentAfter(ThreadState *thr, uptr pc) {
+void ForkParentAfter(ThreadState *thr, uptr pc) NO_THREAD_SAFETY_ANALYSIS {
   thr->suppress_reports--;  // Enabled in ForkBefore.
   thr->ignore_interceptors--;
+  ScopedErrorReportLock::Unlock();
   ctx->report_mtx.Unlock();
   ctx->thread_registry->Unlock();
 }
 
-void ForkChildAfter(ThreadState *thr, uptr pc) {
+void ForkChildAfter(ThreadState *thr, uptr pc) NO_THREAD_SAFETY_ANALYSIS {
   thr->suppress_reports--;  // Enabled in ForkBefore.
   thr->ignore_interceptors--;
+  ScopedErrorReportLock::Unlock();
   ctx->report_mtx.Unlock();
   ctx->thread_registry->Unlock();
 
@@ -693,9 +691,6 @@ ALWAYS_INLINE
 void MemoryAccessImpl1(ThreadState *thr, uptr addr,
     int kAccessSizeLog, bool kAccessIsWrite, bool kIsAtomic,
     u64 *shadow_mem, Shadow cur) {
-  StatInc(thr, StatMop);
-  StatInc(thr, kAccessIsWrite ? StatMopWrite : StatMopRead);
-  StatInc(thr, (StatType)(StatMop1 + kAccessSizeLog));
 
   // This potentially can live in an MMX/SSE scratch register.
   // The required intrinsics are:
@@ -752,7 +747,6 @@ void MemoryAccessImpl1(ThreadState *thr, uptr addr,
     return;
   // choose a random candidate slot and replace it
   StoreShadow(shadow_mem + (cur.epoch() % kShadowCnt), store_word);
-  StatInc(thr, StatShadowReplace);
   return;
  RACE:
   HandleRace(thr, shadow_mem, cur, old);
@@ -891,19 +885,11 @@ void MemoryAccess(ThreadState *thr, uptr pc, uptr addr,
   if (!SANITIZER_GO && !kAccessIsWrite && *shadow_mem == kShadowRodata) {
     // Access to .rodata section, no races here.
     // Measurements show that it can be 10-20% of all memory accesses.
-    StatInc(thr, StatMop);
-    StatInc(thr, kAccessIsWrite ? StatMopWrite : StatMopRead);
-    StatInc(thr, (StatType)(StatMop1 + kAccessSizeLog));
-    StatInc(thr, StatMopRodata);
     return;
   }
 
   FastState fast_state = thr->fast_state;
   if (UNLIKELY(fast_state.GetIgnoreBit())) {
-    StatInc(thr, StatMop);
-    StatInc(thr, kAccessIsWrite ? StatMopWrite : StatMopRead);
-    StatInc(thr, (StatType)(StatMop1 + kAccessSizeLog));
-    StatInc(thr, StatMopIgnored);
     return;
   }
 
@@ -914,10 +900,6 @@ void MemoryAccess(ThreadState *thr, uptr pc, uptr addr,
 
   if (LIKELY(ContainsSameAccess(shadow_mem, cur.raw(),
       thr->fast_synch_epoch, kAccessIsWrite))) {
-    StatInc(thr, StatMop);
-    StatInc(thr, kAccessIsWrite ? StatMopWrite : StatMopRead);
-    StatInc(thr, (StatType)(StatMop1 + kAccessSizeLog));
-    StatInc(thr, StatMopSame);
     return;
   }
 
@@ -939,10 +921,6 @@ void MemoryAccessImpl(ThreadState *thr, uptr addr,
     u64 *shadow_mem, Shadow cur) {
   if (LIKELY(ContainsSameAccess(shadow_mem, cur.raw(),
       thr->fast_synch_epoch, kAccessIsWrite))) {
-    StatInc(thr, StatMop);
-    StatInc(thr, kAccessIsWrite ? StatMopWrite : StatMopRead);
-    StatInc(thr, (StatType)(StatMop1 + kAccessSizeLog));
-    StatInc(thr, StatMopSame);
     return;
   }
 
@@ -999,7 +977,6 @@ static void MemoryRangeSet(ThreadState *thr, uptr pc, uptr addr, uptr size,
     // Reset middle part.
     u64 *p1 = p;
     p = RoundDown(end, kPageSize);
-    UnmapOrDie((void*)p1, (uptr)p - (uptr)p1);
     if (!MmapFixedSuperNoReserve((uptr)p1, (uptr)p - (uptr)p1))
       Die();
     // Set the ending.
@@ -1059,7 +1036,6 @@ void MemoryRangeImitateWriteOrResetRange(ThreadState *thr, uptr pc, uptr addr,
 
 ALWAYS_INLINE USED
 void FuncEntry(ThreadState *thr, uptr pc) {
-  StatInc(thr, StatFuncEnter);
   DPrintf2("#%d: FuncEntry %p\n", (int)thr->fast_state.tid(), (void*)pc);
   if (kCollectHistory) {
     thr->fast_state.IncrementEpoch();
@@ -1081,7 +1057,6 @@ void FuncEntry(ThreadState *thr, uptr pc) {
 
 ALWAYS_INLINE USED
 void FuncExit(ThreadState *thr) {
-  StatInc(thr, StatFuncExit);
   DPrintf2("#%d: FuncExit\n", (int)thr->fast_state.tid());
   if (kCollectHistory) {
     thr->fast_state.IncrementEpoch();
@@ -1156,12 +1131,6 @@ void build_consistency_debug() {}
 void build_consistency_release() {}
 #endif
 
-#if TSAN_COLLECT_STATS
-void build_consistency_stats() {}
-#else
-void build_consistency_nostats() {}
-#endif
-
 }  // namespace __tsan
 
 #if !SANITIZER_GO
diff --git a/libsanitizer/tsan/tsan_rtl.h b/libsanitizer/tsan/tsan_rtl.h
index 3ae519d..6576d40 100644
--- a/libsanitizer/tsan/tsan_rtl.h
+++ b/libsanitizer/tsan/tsan_rtl.h
@@ -400,9 +400,6 @@ struct ThreadState {
   Vector<JmpBuf> jmp_bufs;
   int ignore_interceptors;
 #endif
-#if TSAN_COLLECT_STATS
-  u64 stat[StatCnt];
-#endif
   const u32 tid;
   const int unique_id;
   bool in_symbolizer;
@@ -550,7 +547,6 @@ struct Context {
 
   Flags flags;
 
-  u64 stat[StatCnt];
   u64 int_alloc_cnt[MBlockTypeCount];
   u64 int_alloc_siz[MBlockTypeCount];
 };
@@ -658,22 +654,6 @@ void ObtainCurrentStack(ThreadState *thr, uptr toppc, StackTraceTy *stack,
   ObtainCurrentStack(thr, pc, &stack); \
   stack.ReverseOrder();
 
-#if TSAN_COLLECT_STATS
-void StatAggregate(u64 *dst, u64 *src);
-void StatOutput(u64 *stat);
-#endif
-
-void ALWAYS_INLINE StatInc(ThreadState *thr, StatType typ, u64 n = 1) {
-#if TSAN_COLLECT_STATS
-  thr->stat[typ] += n;
-#endif
-}
-void ALWAYS_INLINE StatSet(ThreadState *thr, StatType typ, u64 n) {
-#if TSAN_COLLECT_STATS
-  thr->stat[typ] = n;
-#endif
-}
-
 void MapShadow(uptr addr, uptr size);
 void MapThreadTrace(uptr addr, uptr size, const char *name);
 void DontNeedShadowFor(uptr addr, uptr size);
@@ -854,7 +834,6 @@ void ALWAYS_INLINE TraceAddEvent(ThreadState *thr, FastState fs,
   DCHECK_GE((int)typ, 0);
   DCHECK_LE((int)typ, 7);
   DCHECK_EQ(GetLsb(addr, kEventPCBits), addr);
-  StatInc(thr, StatEvents);
   u64 pos = fs.GetTracePos();
   if (UNLIKELY((pos % kTracePartSize) == 0)) {
 #if !SANITIZER_GO
diff --git a/libsanitizer/tsan/tsan_rtl_mutex.cpp b/libsanitizer/tsan/tsan_rtl_mutex.cpp
index 0a8f3aa..a214a33 100644
--- a/libsanitizer/tsan/tsan_rtl_mutex.cpp
+++ b/libsanitizer/tsan/tsan_rtl_mutex.cpp
@@ -65,7 +65,6 @@ static void ReportMutexMisuse(ThreadState *thr, uptr pc, ReportType typ,
 
 void MutexCreate(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
   DPrintf("#%d: MutexCreate %zx flagz=0x%x\n", thr->tid, addr, flagz);
-  StatInc(thr, StatMutexCreate);
   if (!(flagz & MutexFlagLinkerInit) && IsAppMem(addr)) {
     CHECK(!thr->is_freeing);
     thr->is_freeing = true;
@@ -81,7 +80,6 @@ void MutexCreate(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
 
 void MutexDestroy(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
   DPrintf("#%d: MutexDestroy %zx\n", thr->tid, addr);
-  StatInc(thr, StatMutexDestroy);
   SyncVar *s = ctx->metamap.GetIfExistsAndLock(addr, true);
   if (s == 0)
     return;
@@ -183,11 +181,9 @@ void MutexPostLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz, int rec) {
   const bool first = s->recursion == 0;
   s->recursion += rec;
   if (first) {
-    StatInc(thr, StatMutexLock);
     AcquireImpl(thr, pc, &s->clock);
     AcquireImpl(thr, pc, &s->read_clock);
   } else if (!s->IsFlagSet(MutexFlagWriteReentrant)) {
-    StatInc(thr, StatMutexRecLock);
   }
   thr->mset.Add(s->GetId(), true, thr->fast_state.epoch());
   bool pre_lock = false;
@@ -229,11 +225,9 @@ int MutexUnlock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
     rec = (flagz & MutexFlagRecursiveUnlock) ? s->recursion : 1;
     s->recursion -= rec;
     if (s->recursion == 0) {
-      StatInc(thr, StatMutexUnlock);
       s->owner_tid = kInvalidTid;
       ReleaseStoreImpl(thr, pc, &s->clock);
     } else {
-      StatInc(thr, StatMutexRecUnlock);
     }
   }
   thr->mset.Del(s->GetId(), true);
@@ -268,7 +262,6 @@ void MutexPreReadLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
 
 void MutexPostReadLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
   DPrintf("#%d: MutexPostReadLock %zx flagz=0x%x\n", thr->tid, addr, flagz);
-  StatInc(thr, StatMutexReadLock);
   if (IsAppMem(addr))
     MemoryReadAtomic(thr, pc, addr, kSizeLog1);
   SyncVar *s = ctx->metamap.GetOrCreateAndLock(thr, pc, addr, false);
@@ -308,7 +301,6 @@ void MutexPostReadLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
 
 void MutexReadUnlock(ThreadState *thr, uptr pc, uptr addr) {
   DPrintf("#%d: MutexReadUnlock %zx\n", thr->tid, addr);
-  StatInc(thr, StatMutexReadUnlock);
   if (IsAppMem(addr))
     MemoryReadAtomic(thr, pc, addr, kSizeLog1);
   SyncVar *s = ctx->metamap.GetOrCreateAndLock(thr, pc, addr, true);
@@ -348,7 +340,6 @@ void MutexReadOrWriteUnlock(ThreadState *thr, uptr pc, uptr addr) {
   if (s->owner_tid == kInvalidTid) {
     // Seems to be read unlock.
     write = false;
-    StatInc(thr, StatMutexReadUnlock);
     thr->fast_state.IncrementEpoch();
     TraceAddEvent(thr, thr->fast_state, EventTypeRUnlock, s->GetId());
     ReleaseImpl(thr, pc, &s->read_clock);
@@ -359,11 +350,9 @@ void MutexReadOrWriteUnlock(ThreadState *thr, uptr pc, uptr addr) {
     CHECK_GT(s->recursion, 0);
     s->recursion--;
     if (s->recursion == 0) {
-      StatInc(thr, StatMutexUnlock);
       s->owner_tid = kInvalidTid;
       ReleaseStoreImpl(thr, pc, &s->clock);
     } else {
-      StatInc(thr, StatMutexRecUnlock);
     }
   } else if (!s->IsFlagSet(MutexFlagBroken)) {
     s->SetFlags(MutexFlagBroken);
@@ -494,7 +483,6 @@ void AcquireImpl(ThreadState *thr, uptr pc, SyncClock *c) {
     return;
   thr->clock.set(thr->fast_state.epoch());
   thr->clock.acquire(&thr->proc()->clock_cache, c);
-  StatInc(thr, StatSyncAcquire);
 }
 
 void ReleaseStoreAcquireImpl(ThreadState *thr, uptr pc, SyncClock *c) {
@@ -503,7 +491,6 @@ void ReleaseStoreAcquireImpl(ThreadState *thr, uptr pc, SyncClock *c) {
   thr->clock.set(thr->fast_state.epoch());
   thr->fast_synch_epoch = thr->fast_state.epoch();
   thr->clock.releaseStoreAcquire(&thr->proc()->clock_cache, c);
-  StatInc(thr, StatSyncReleaseStoreAcquire);
 }
 
 void ReleaseImpl(ThreadState *thr, uptr pc, SyncClock *c) {
@@ -512,7 +499,6 @@ void ReleaseImpl(ThreadState *thr, uptr pc, SyncClock *c) {
   thr->clock.set(thr->fast_state.epoch());
   thr->fast_synch_epoch = thr->fast_state.epoch();
   thr->clock.release(&thr->proc()->clock_cache, c);
-  StatInc(thr, StatSyncRelease);
 }
 
 void ReleaseStoreImpl(ThreadState *thr, uptr pc, SyncClock *c) {
@@ -521,7 +507,6 @@ void ReleaseStoreImpl(ThreadState *thr, uptr pc, SyncClock *c) {
   thr->clock.set(thr->fast_state.epoch());
   thr->fast_synch_epoch = thr->fast_state.epoch();
   thr->clock.ReleaseStore(&thr->proc()->clock_cache, c);
-  StatInc(thr, StatSyncRelease);
 }
 
 void AcquireReleaseImpl(ThreadState *thr, uptr pc, SyncClock *c) {
@@ -530,8 +515,6 @@ void AcquireReleaseImpl(ThreadState *thr, uptr pc, SyncClock *c) {
   thr->clock.set(thr->fast_state.epoch());
   thr->fast_synch_epoch = thr->fast_state.epoch();
   thr->clock.acq_rel(&thr->proc()->clock_cache, c);
-  StatInc(thr, StatSyncAcquire);
-  StatInc(thr, StatSyncRelease);
 }
 
 void ReportDeadlock(ThreadState *thr, uptr pc, DDReport *r) {
diff --git a/libsanitizer/tsan/tsan_rtl_ppc64.S b/libsanitizer/tsan/tsan_rtl_ppc64.S
index 9e533a7..8285e21 100644
--- a/libsanitizer/tsan/tsan_rtl_ppc64.S
+++ b/libsanitizer/tsan/tsan_rtl_ppc64.S
@@ -1,6 +1,5 @@
 #include "tsan_ppc_regs.h"
 
-        .machine altivec
         .section .text
         .hidden __tsan_setjmp
         .globl _setjmp
diff --git a/libsanitizer/tsan/tsan_rtl_s390x.S b/libsanitizer/tsan/tsan_rtl_s390x.S
new file mode 100644
index 0000000..fcff35f
--- /dev/null
+++ b/libsanitizer/tsan/tsan_rtl_s390x.S
@@ -0,0 +1,47 @@
+#include "sanitizer_common/sanitizer_asm.h"
+
+#define CFA_OFFSET 160
+#define R2_REL_OFFSET 16
+#define R3_REL_OFFSET 24
+#define R14_REL_OFFSET 112
+#define R15_REL_OFFSET 120
+#define FRAME_SIZE 160
+
+.text
+
+ASM_HIDDEN(__tsan_setjmp)
+
+.macro intercept symbol, real
+.comm \real, 8, 8
+.globl ASM_SYMBOL_INTERCEPTOR(\symbol)
+ASM_TYPE_FUNCTION(ASM_SYMBOL_INTERCEPTOR(\symbol))
+ASM_SYMBOL_INTERCEPTOR(\symbol):
+  CFI_STARTPROC
+  stmg %r2, %r3, R2_REL_OFFSET(%r15)
+  CFI_REL_OFFSET(%r2, R2_REL_OFFSET)
+  CFI_REL_OFFSET(%r3, R3_REL_OFFSET)
+  stmg %r14, %r15, R14_REL_OFFSET(%r15)
+  CFI_REL_OFFSET(%r14, R14_REL_OFFSET)
+  CFI_REL_OFFSET(%r15, R15_REL_OFFSET)
+  aghi %r15, -FRAME_SIZE
+  CFI_ADJUST_CFA_OFFSET(FRAME_SIZE)
+  la %r2, FRAME_SIZE(%r15)
+  brasl %r14, ASM_SYMBOL(__tsan_setjmp)
+  lmg %r14, %r15, FRAME_SIZE + R14_REL_OFFSET(%r15)
+  CFI_RESTORE(%r14)
+  CFI_RESTORE(%r15)
+  CFI_DEF_CFA_OFFSET(CFA_OFFSET)
+  lmg %r2, %r3, R2_REL_OFFSET(%r15)
+  CFI_RESTORE(%r2)
+  CFI_RESTORE(%r3)
+  larl %r1, \real
+  lg %r1, 0(%r1)
+  br %r1
+  CFI_ENDPROC
+  ASM_SIZE(ASM_SYMBOL_INTERCEPTOR(\symbol))
+.endm
+
+intercept setjmp, _ZN14__interception11real_setjmpE
+intercept _setjmp, _ZN14__interception12real__setjmpE
+intercept sigsetjmp, _ZN14__interception14real_sigsetjmpE
+intercept __sigsetjmp, _ZN14__interception16real___sigsetjmpE
diff --git a/libsanitizer/tsan/tsan_rtl_thread.cpp b/libsanitizer/tsan/tsan_rtl_thread.cpp
index 6d1ccd8..cdb6e60 100644
--- a/libsanitizer/tsan/tsan_rtl_thread.cpp
+++ b/libsanitizer/tsan/tsan_rtl_thread.cpp
@@ -61,8 +61,6 @@ void ThreadContext::OnCreated(void *arg) {
   TraceAddEvent(args->thr, args->thr->fast_state, EventTypeMop, 0);
   ReleaseImpl(args->thr, 0, &sync);
   creation_stack_id = CurrentStackId(args->thr, args->pc);
-  if (reuse_count == 0)
-    StatInc(args->thr, StatThreadMaxTid);
 }
 
 void ThreadContext::OnReset() {
@@ -115,7 +113,6 @@ void ThreadContext::OnStarted(void *arg) {
 
   thr->fast_synch_epoch = epoch0;
   AcquireImpl(thr, 0, &sync);
-  StatInc(thr, StatSyncAcquire);
   sync.Reset(&thr->proc()->clock_cache);
   thr->is_inited = true;
   DPrintf("#%d: ThreadStart epoch=%zu stk_addr=%zx stk_size=%zx "
@@ -149,9 +146,6 @@ void ThreadContext::OnFinished() {
   PlatformCleanUpThreadState(thr);
 #endif
   thr->~ThreadState();
-#if TSAN_COLLECT_STATS
-  StatAggregate(ctx->stat, thr->stat);
-#endif
   thr = 0;
 }
 
@@ -232,13 +226,11 @@ int ThreadCount(ThreadState *thr) {
 }
 
 int ThreadCreate(ThreadState *thr, uptr pc, uptr uid, bool detached) {
-  StatInc(thr, StatThreadCreate);
   OnCreatedArgs args = { thr, pc };
   u32 parent_tid = thr ? thr->tid : kInvalidTid;  // No parent for GCD workers.
   int tid =
       ctx->thread_registry->CreateThread(uid, detached, parent_tid, &args);
   DPrintf("#%d: ThreadCreate tid=%d uid=%zu\n", parent_tid, tid, uid);
-  StatSet(thr, StatThreadMaxAlive, ctx->thread_registry->GetMaxAliveThreads());
   return tid;
 }
 
@@ -280,7 +272,6 @@ void ThreadStart(ThreadState *thr, int tid, tid_t os_id,
 
 void ThreadFinish(ThreadState *thr) {
   ThreadCheckIgnore(thr);
-  StatInc(thr, StatThreadFinish);
   if (thr->stk_addr && thr->stk_size)
     DontNeedShadowFor(thr->stk_addr, thr->stk_size);
   if (thr->tls_addr && thr->tls_size)
@@ -372,13 +363,10 @@ void MemoryAccessRange(ThreadState *thr, uptr pc, uptr addr,
   }
 #endif
 
-  StatInc(thr, StatMopRange);
-
   if (*shadow_mem == kShadowRodata) {
     DCHECK(!is_write);
     // Access to .rodata section, no races here.
     // Measurements show that it can be 10-20% of all memory accesses.
-    StatInc(thr, StatMopRangeRodata);
     return;
   }
 
diff --git a/libsanitizer/tsan/tsan_stack_trace.cpp b/libsanitizer/tsan/tsan_stack_trace.cpp
index 403a21a..6c703d7 100644
--- a/libsanitizer/tsan/tsan_stack_trace.cpp
+++ b/libsanitizer/tsan/tsan_stack_trace.cpp
@@ -54,10 +54,8 @@ void __sanitizer::BufferedStackTrace::UnwindImpl(
     uptr pc, uptr bp, void *context, bool request_fast, u32 max_depth) {
   uptr top = 0;
   uptr bottom = 0;
-  if (StackTrace::WillUseFastUnwind(request_fast)) {
-    GetThreadStackTopAndBottom(false, &top, &bottom);
-    Unwind(max_depth, pc, bp, nullptr, top, bottom, true);
-  } else
-    Unwind(max_depth, pc, 0, context, 0, 0, false);
+  GetThreadStackTopAndBottom(false, &top, &bottom);
+  bool fast = StackTrace::WillUseFastUnwind(request_fast);
+  Unwind(max_depth, pc, bp, context, top, bottom, fast);
 }
 #endif  // SANITIZER_GO
diff --git a/libsanitizer/tsan/tsan_stat.cpp b/libsanitizer/tsan/tsan_stat.cpp
deleted file mode 100644
index 78f3cce..0000000
--- a/libsanitizer/tsan/tsan_stat.cpp
+++ /dev/null
@@ -1,186 +0,0 @@
-//===-- tsan_stat.cpp -----------------------------------------------------===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file is a part of ThreadSanitizer (TSan), a race detector.
-//
-//===----------------------------------------------------------------------===//
-#include "tsan_stat.h"
-#include "tsan_rtl.h"
-
-namespace __tsan {
-
-#if TSAN_COLLECT_STATS
-
-void StatAggregate(u64 *dst, u64 *src) {
-  for (int i = 0; i < StatCnt; i++)
-    dst[i] += src[i];
-}
-
-void StatOutput(u64 *stat) {
-  stat[StatShadowNonZero] = stat[StatShadowProcessed] - stat[StatShadowZero];
-
-  static const char *name[StatCnt] = {};
-  name[StatMop]                          = "Memory accesses                   ";
-  name[StatMopRead]                      = "  Including reads                 ";
-  name[StatMopWrite]                     = "            writes                ";
-  name[StatMop1]                         = "  Including size 1                ";
-  name[StatMop2]                         = "            size 2                ";
-  name[StatMop4]                         = "            size 4                ";
-  name[StatMop8]                         = "            size 8                ";
-  name[StatMopSame]                      = "  Including same                  ";
-  name[StatMopIgnored]                   = "  Including ignored               ";
-  name[StatMopRange]                     = "  Including range                 ";
-  name[StatMopRodata]                    = "  Including .rodata               ";
-  name[StatMopRangeRodata]               = "  Including .rodata range         ";
-  name[StatShadowProcessed]              = "Shadow processed                  ";
-  name[StatShadowZero]                   = "  Including empty                 ";
-  name[StatShadowNonZero]                = "  Including non empty             ";
-  name[StatShadowSameSize]               = "  Including same size             ";
-  name[StatShadowIntersect]              = "            intersect             ";
-  name[StatShadowNotIntersect]           = "            not intersect         ";
-  name[StatShadowSameThread]             = "  Including same thread           ";
-  name[StatShadowAnotherThread]          = "            another thread        ";
-  name[StatShadowReplace]                = "  Including evicted               ";
-
-  name[StatFuncEnter]                    = "Function entries                  ";
-  name[StatFuncExit]                     = "Function exits                    ";
-  name[StatEvents]                       = "Events collected                  ";
-
-  name[StatThreadCreate]                 = "Total threads created             ";
-  name[StatThreadFinish]                 = "  threads finished                ";
-  name[StatThreadReuse]                  = "  threads reused                  ";
-  name[StatThreadMaxTid]                 = "  max tid                         ";
-  name[StatThreadMaxAlive]               = "  max alive threads               ";
-
-  name[StatMutexCreate]                  = "Mutexes created                   ";
-  name[StatMutexDestroy]                 = "  destroyed                       ";
-  name[StatMutexLock]                    = "  lock                            ";
-  name[StatMutexUnlock]                  = "  unlock                          ";
-  name[StatMutexRecLock]                 = "  recursive lock                  ";
-  name[StatMutexRecUnlock]               = "  recursive unlock                ";
-  name[StatMutexReadLock]                = "  read lock                       ";
-  name[StatMutexReadUnlock]              = "  read unlock                     ";
-
-  name[StatSyncCreated]                  = "Sync objects created              ";
-  name[StatSyncDestroyed]                = "             destroyed            ";
-  name[StatSyncAcquire]                  = "             acquired             ";
-  name[StatSyncRelease]                  = "             released             ";
-
-  name[StatClockAcquire]                 = "Clock acquire                     ";
-  name[StatClockAcquireEmpty]            = "  empty clock                     ";
-  name[StatClockAcquireFastRelease]      = "  fast from release-store         ";
-  name[StatClockAcquireFull]             = "  full (slow)                     ";
-  name[StatClockAcquiredSomething]       = "  acquired something              ";
-  name[StatClockRelease]                 = "Clock release                     ";
-  name[StatClockReleaseResize]           = "  resize                          ";
-  name[StatClockReleaseFast]             = "  fast                            ";
-  name[StatClockReleaseSlow]             = "  dirty overflow (slow)           ";
-  name[StatClockReleaseFull]             = "  full (slow)                     ";
-  name[StatClockReleaseAcquired]         = "  was acquired                    ";
-  name[StatClockReleaseClearTail]        = "  clear tail                      ";
-  name[StatClockStore]                   = "Clock release store               ";
-  name[StatClockStoreResize]             = "  resize                          ";
-  name[StatClockStoreFast]               = "  fast                            ";
-  name[StatClockStoreFull]               = "  slow                            ";
-  name[StatClockStoreTail]               = "  clear tail                      ";
-  name[StatClockAcquireRelease]          = "Clock acquire-release             ";
-
-  name[StatAtomic]                       = "Atomic operations                 ";
-  name[StatAtomicLoad]                   = "  Including load                  ";
-  name[StatAtomicStore]                  = "            store                 ";
-  name[StatAtomicExchange]               = "            exchange              ";
-  name[StatAtomicFetchAdd]               = "            fetch_add             ";
-  name[StatAtomicFetchSub]               = "            fetch_sub             ";
-  name[StatAtomicFetchAnd]               = "            fetch_and             ";
-  name[StatAtomicFetchOr]                = "            fetch_or              ";
-  name[StatAtomicFetchXor]               = "            fetch_xor             ";
-  name[StatAtomicFetchNand]              = "            fetch_nand            ";
-  name[StatAtomicCAS]                    = "            compare_exchange      ";
-  name[StatAtomicFence]                  = "            fence                 ";
-  name[StatAtomicRelaxed]                = "  Including relaxed               ";
-  name[StatAtomicConsume]                = "            consume               ";
-  name[StatAtomicAcquire]                = "            acquire               ";
-  name[StatAtomicRelease]                = "            release               ";
-  name[StatAtomicAcq_Rel]                = "            acq_rel               ";
-  name[StatAtomicSeq_Cst]                = "            seq_cst               ";
-  name[StatAtomic1]                      = "  Including size 1                ";
-  name[StatAtomic2]                      = "            size 2                ";
-  name[StatAtomic4]                      = "            size 4                ";
-  name[StatAtomic8]                      = "            size 8                ";
-  name[StatAtomic16]                     = "            size 16               ";
-
-  name[StatAnnotation]                   = "Dynamic annotations               ";
-  name[StatAnnotateHappensBefore]        = "  HappensBefore                   ";
-  name[StatAnnotateHappensAfter]         = "  HappensAfter                    ";
-  name[StatAnnotateCondVarSignal]        = "  CondVarSignal                   ";
-  name[StatAnnotateCondVarSignalAll]     = "  CondVarSignalAll                ";
-  name[StatAnnotateMutexIsNotPHB]        = "  MutexIsNotPHB                   ";
-  name[StatAnnotateCondVarWait]          = "  CondVarWait                     ";
-  name[StatAnnotateRWLockCreate]         = "  RWLockCreate                    ";
-  name[StatAnnotateRWLockCreateStatic]   = "  StatAnnotateRWLockCreateStatic  ";
-  name[StatAnnotateRWLockDestroy]        = "  RWLockDestroy                   ";
-  name[StatAnnotateRWLockAcquired]       = "  RWLockAcquired                  ";
-  name[StatAnnotateRWLockReleased]       = "  RWLockReleased                  ";
-  name[StatAnnotateTraceMemory]          = "  TraceMemory                     ";
-  name[StatAnnotateFlushState]           = "  FlushState                      ";
-  name[StatAnnotateNewMemory]            = "  NewMemory                       ";
-  name[StatAnnotateNoOp]                 = "  NoOp                            ";
-  name[StatAnnotateFlushExpectedRaces]   = "  FlushExpectedRaces              ";
-  name[StatAnnotateEnableRaceDetection]  = "  EnableRaceDetection             ";
-  name[StatAnnotateMutexIsUsedAsCondVar] = "  MutexIsUsedAsCondVar            ";
-  name[StatAnnotatePCQGet]               = "  PCQGet                          ";
-  name[StatAnnotatePCQPut]               = "  PCQPut                          ";
-  name[StatAnnotatePCQDestroy]           = "  PCQDestroy                      ";
-  name[StatAnnotatePCQCreate]            = "  PCQCreate                       ";
-  name[StatAnnotateExpectRace]           = "  ExpectRace                      ";
-  name[StatAnnotateBenignRaceSized]      = "  BenignRaceSized                 ";
-  name[StatAnnotateBenignRace]           = "  BenignRace                      ";
-  name[StatAnnotateIgnoreReadsBegin]     = "  IgnoreReadsBegin                ";
-  name[StatAnnotateIgnoreReadsEnd]       = "  IgnoreReadsEnd                  ";
-  name[StatAnnotateIgnoreWritesBegin]    = "  IgnoreWritesBegin               ";
-  name[StatAnnotateIgnoreWritesEnd]      = "  IgnoreWritesEnd                 ";
-  name[StatAnnotateIgnoreSyncBegin]      = "  IgnoreSyncBegin                 ";
-  name[StatAnnotateIgnoreSyncEnd]        = "  IgnoreSyncEnd                   ";
-  name[StatAnnotatePublishMemoryRange]   = "  PublishMemoryRange              ";
-  name[StatAnnotateUnpublishMemoryRange] = "  UnpublishMemoryRange            ";
-  name[StatAnnotateThreadName]           = "  ThreadName                      ";
-  name[Stat__tsan_mutex_create]          = "  __tsan_mutex_create             ";
-  name[Stat__tsan_mutex_destroy]         = "  __tsan_mutex_destroy            ";
-  name[Stat__tsan_mutex_pre_lock]        = "  __tsan_mutex_pre_lock           ";
-  name[Stat__tsan_mutex_post_lock]       = "  __tsan_mutex_post_lock          ";
-  name[Stat__tsan_mutex_pre_unlock]      = "  __tsan_mutex_pre_unlock         ";
-  name[Stat__tsan_mutex_post_unlock]     = "  __tsan_mutex_post_unlock        ";
-  name[Stat__tsan_mutex_pre_signal]      = "  __tsan_mutex_pre_signal         ";
-  name[Stat__tsan_mutex_post_signal]     = "  __tsan_mutex_post_signal        ";
-  name[Stat__tsan_mutex_pre_divert]      = "  __tsan_mutex_pre_divert         ";
-  name[Stat__tsan_mutex_post_divert]     = "  __tsan_mutex_post_divert        ";
-
-  name[StatMtxTotal]                     = "Contentionz                       ";
-  name[StatMtxTrace]                     = "  Trace                           ";
-  name[StatMtxThreads]                   = "  Threads                         ";
-  name[StatMtxReport]                    = "  Report                          ";
-  name[StatMtxSyncVar]                   = "  SyncVar                         ";
-  name[StatMtxSyncTab]                   = "  SyncTab                         ";
-  name[StatMtxSlab]                      = "  Slab                            ";
-  name[StatMtxAtExit]                    = "  Atexit                          ";
-  name[StatMtxAnnotations]               = "  Annotations                     ";
-  name[StatMtxMBlock]                    = "  MBlock                          ";
-  name[StatMtxDeadlockDetector]          = "  DeadlockDetector                ";
-  name[StatMtxFired]                     = "  FiredSuppressions               ";
-  name[StatMtxRacy]                      = "  RacyStacks                      ";
-  name[StatMtxFD]                        = "  FD                              ";
-  name[StatMtxGlobalProc]                = "  GlobalProc                      ";
-
-  Printf("Statistics:\n");
-  for (int i = 0; i < StatCnt; i++)
-    Printf("%s: %16zu\n", name[i], (uptr)stat[i]);
-}
-
-#endif
-
-}  // namespace __tsan
diff --git a/libsanitizer/tsan/tsan_stat.h b/libsanitizer/tsan/tsan_stat.h
deleted file mode 100644
index 8b26a59..0000000
--- a/libsanitizer/tsan/tsan_stat.h
+++ /dev/null
@@ -1,191 +0,0 @@
-//===-- tsan_stat.h ---------------------------------------------*- C++ -*-===//
-//
-// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
-// See https://llvm.org/LICENSE.txt for license information.
-// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
-//
-//===----------------------------------------------------------------------===//
-//
-// This file is a part of ThreadSanitizer (TSan), a race detector.
-//
-//===----------------------------------------------------------------------===//
-
-#ifndef TSAN_STAT_H
-#define TSAN_STAT_H
-
-namespace __tsan {
-
-enum StatType {
-  // Memory access processing related stuff.
-  StatMop,
-  StatMopRead,
-  StatMopWrite,
-  StatMop1,  // These must be consequtive.
-  StatMop2,
-  StatMop4,
-  StatMop8,
-  StatMopSame,
-  StatMopIgnored,
-  StatMopRange,
-  StatMopRodata,
-  StatMopRangeRodata,
-  StatShadowProcessed,
-  StatShadowZero,
-  StatShadowNonZero,  // Derived.
-  StatShadowSameSize,
-  StatShadowIntersect,
-  StatShadowNotIntersect,
-  StatShadowSameThread,
-  StatShadowAnotherThread,
-  StatShadowReplace,
-
-  // Func processing.
-  StatFuncEnter,
-  StatFuncExit,
-
-  // Trace processing.
-  StatEvents,
-
-  // Threads.
-  StatThreadCreate,
-  StatThreadFinish,
-  StatThreadReuse,
-  StatThreadMaxTid,
-  StatThreadMaxAlive,
-
-  // Mutexes.
-  StatMutexCreate,
-  StatMutexDestroy,
-  StatMutexLock,
-  StatMutexUnlock,
-  StatMutexRecLock,
-  StatMutexRecUnlock,
-  StatMutexReadLock,
-  StatMutexReadUnlock,
-
-  // Synchronization.
-  StatSyncCreated,
-  StatSyncDestroyed,
-  StatSyncAcquire,
-  StatSyncRelease,
-  StatSyncReleaseStoreAcquire,
-
-  // Clocks - acquire.
-  StatClockAcquire,
-  StatClockAcquireEmpty,
-  StatClockAcquireFastRelease,
-  StatClockAcquireFull,
-  StatClockAcquiredSomething,
-  // Clocks - release.
-  StatClockRelease,
-  StatClockReleaseResize,
-  StatClockReleaseFast,
-  StatClockReleaseSlow,
-  StatClockReleaseFull,
-  StatClockReleaseAcquired,
-  StatClockReleaseClearTail,
-  // Clocks - release store.
-  StatClockStore,
-  StatClockStoreResize,
-  StatClockStoreFast,
-  StatClockStoreFull,
-  StatClockStoreTail,
-  // Clocks - acquire-release.
-  StatClockAcquireRelease,
-
-  // Atomics.
-  StatAtomic,
-  StatAtomicLoad,
-  StatAtomicStore,
-  StatAtomicExchange,
-  StatAtomicFetchAdd,
-  StatAtomicFetchSub,
-  StatAtomicFetchAnd,
-  StatAtomicFetchOr,
-  StatAtomicFetchXor,
-  StatAtomicFetchNand,
-  StatAtomicCAS,
-  StatAtomicFence,
-  StatAtomicRelaxed,
-  StatAtomicConsume,
-  StatAtomicAcquire,
-  StatAtomicRelease,
-  StatAtomicAcq_Rel,
-  StatAtomicSeq_Cst,
-  StatAtomic1,
-  StatAtomic2,
-  StatAtomic4,
-  StatAtomic8,
-  StatAtomic16,
-
-  // Dynamic annotations.
-  StatAnnotation,
-  StatAnnotateHappensBefore,
-  StatAnnotateHappensAfter,
-  StatAnnotateCondVarSignal,
-  StatAnnotateCondVarSignalAll,
-  StatAnnotateMutexIsNotPHB,
-  StatAnnotateCondVarWait,
-  StatAnnotateRWLockCreate,
-  StatAnnotateRWLockCreateStatic,
-  StatAnnotateRWLockDestroy,
-  StatAnnotateRWLockAcquired,
-  StatAnnotateRWLockReleased,
-  StatAnnotateTraceMemory,
-  StatAnnotateFlushState,
-  StatAnnotateNewMemory,
-  StatAnnotateNoOp,
-  StatAnnotateFlushExpectedRaces,
-  StatAnnotateEnableRaceDetection,
-  StatAnnotateMutexIsUsedAsCondVar,
-  StatAnnotatePCQGet,
-  StatAnnotatePCQPut,
-  StatAnnotatePCQDestroy,
-  StatAnnotatePCQCreate,
-  StatAnnotateExpectRace,
-  StatAnnotateBenignRaceSized,
-  StatAnnotateBenignRace,
-  StatAnnotateIgnoreReadsBegin,
-  StatAnnotateIgnoreReadsEnd,
-  StatAnnotateIgnoreWritesBegin,
-  StatAnnotateIgnoreWritesEnd,
-  StatAnnotateIgnoreSyncBegin,
-  StatAnnotateIgnoreSyncEnd,
-  StatAnnotatePublishMemoryRange,
-  StatAnnotateUnpublishMemoryRange,
-  StatAnnotateThreadName,
-  Stat__tsan_mutex_create,
-  Stat__tsan_mutex_destroy,
-  Stat__tsan_mutex_pre_lock,
-  Stat__tsan_mutex_post_lock,
-  Stat__tsan_mutex_pre_unlock,
-  Stat__tsan_mutex_post_unlock,
-  Stat__tsan_mutex_pre_signal,
-  Stat__tsan_mutex_post_signal,
-  Stat__tsan_mutex_pre_divert,
-  Stat__tsan_mutex_post_divert,
-
-  // Internal mutex contentionz.
-  StatMtxTotal,
-  StatMtxTrace,
-  StatMtxThreads,
-  StatMtxReport,
-  StatMtxSyncVar,
-  StatMtxSyncTab,
-  StatMtxSlab,
-  StatMtxAnnotations,
-  StatMtxAtExit,
-  StatMtxMBlock,
-  StatMtxDeadlockDetector,
-  StatMtxFired,
-  StatMtxRacy,
-  StatMtxFD,
-  StatMtxGlobalProc,
-
-  // This must be the last.
-  StatCnt
-};
-
-}  // namespace __tsan
-
-#endif  // TSAN_STAT_H
diff --git a/libsanitizer/tsan/tsan_sync.cpp b/libsanitizer/tsan/tsan_sync.cpp
index ba24f98..d25434a 100644
--- a/libsanitizer/tsan/tsan_sync.cpp
+++ b/libsanitizer/tsan/tsan_sync.cpp
@@ -18,10 +18,7 @@ namespace __tsan {
 
 void DDMutexInit(ThreadState *thr, uptr pc, SyncVar *s);
 
-SyncVar::SyncVar()
-    : mtx(MutexTypeSyncVar, StatMtxSyncVar) {
-  Reset(0);
-}
+SyncVar::SyncVar() : mtx(MutexTypeSyncVar) { Reset(0); }
 
 void SyncVar::Init(ThreadState *thr, uptr pc, uptr addr, u64 uid) {
   this->addr = addr;
diff --git a/libsanitizer/tsan/tsan_trace.h b/libsanitizer/tsan/tsan_trace.h
index fbd0f72..9f2677b 100644
--- a/libsanitizer/tsan/tsan_trace.h
+++ b/libsanitizer/tsan/tsan_trace.h
@@ -65,9 +65,7 @@ struct Trace {
   // CreateThreadContext.
   TraceHeader headers[kTraceParts];
 
-  Trace()
-    : mtx(MutexTypeTrace, StatMtxTrace) {
-  }
+  Trace() : mtx(MutexTypeTrace) {}
 };
 
 }  // namespace __tsan
diff --git a/libsanitizer/tsan/tsan_update_shadow_word_inl.h b/libsanitizer/tsan/tsan_update_shadow_word_inl.h
index 056c3aa..d23dfb0 100644
--- a/libsanitizer/tsan/tsan_update_shadow_word_inl.h
+++ b/libsanitizer/tsan/tsan_update_shadow_word_inl.h
@@ -13,12 +13,10 @@
 // produce sligtly less efficient code.
 //===----------------------------------------------------------------------===//
 do {
-  StatInc(thr, StatShadowProcessed);
   const unsigned kAccessSize = 1 << kAccessSizeLog;
   u64 *sp = &shadow_mem[idx];
   old = LoadShadow(sp);
   if (LIKELY(old.IsZero())) {
-    StatInc(thr, StatShadowZero);
     if (!stored) {
       StoreIfNotYetStored(sp, &store_word);
       stored = true;
@@ -27,17 +25,14 @@ do {
   }
   // is the memory access equal to the previous?
   if (LIKELY(Shadow::Addr0AndSizeAreEqual(cur, old))) {
-    StatInc(thr, StatShadowSameSize);
     // same thread?
     if (LIKELY(Shadow::TidsAreEqual(old, cur))) {
-      StatInc(thr, StatShadowSameThread);
       if (LIKELY(old.IsRWWeakerOrEqual(kAccessIsWrite, kIsAtomic))) {
         StoreIfNotYetStored(sp, &store_word);
         stored = true;
       }
       break;
     }
-    StatInc(thr, StatShadowAnotherThread);
     if (HappensBefore(old, thr)) {
       if (old.IsRWWeakerOrEqual(kAccessIsWrite, kIsAtomic)) {
         StoreIfNotYetStored(sp, &store_word);
@@ -51,12 +46,8 @@ do {
   }
   // Do the memory access intersect?
   if (Shadow::TwoRangesIntersect(old, cur, kAccessSize)) {
-    StatInc(thr, StatShadowIntersect);
-    if (Shadow::TidsAreEqual(old, cur)) {
-      StatInc(thr, StatShadowSameThread);
+    if (Shadow::TidsAreEqual(old, cur))
       break;
-    }
-    StatInc(thr, StatShadowAnotherThread);
     if (old.IsBothReadsOrAtomic(kAccessIsWrite, kIsAtomic))
       break;
     if (LIKELY(HappensBefore(old, thr)))
@@ -64,6 +55,5 @@ do {
     goto RACE;
   }
   // The accesses do not intersect.
-  StatInc(thr, StatShadowNotIntersect);
   break;
 } while (0);