diff --git a/libsanitizer/MERGE b/libsanitizer/MERGE
index c4731d0866c..81d00f27de4 100644
--- a/libsanitizer/MERGE
+++ b/libsanitizer/MERGE
@@ -1,4 +1,4 @@
-f58e0513dd95944b81ce7a6e7b49ba656de7d75f
+7704fedfff6ef5676adb6415f3be0ac927d1a746
 
 The first line of this file holds the git revision number of the
 last merge done from the master library sources.
diff --git a/libsanitizer/asan/Makefile.am b/libsanitizer/asan/Makefile.am
index 2e6385509b4..74658ca7b9c 100644
--- a/libsanitizer/asan/Makefile.am
+++ b/libsanitizer/asan/Makefile.am
@@ -38,7 +38,6 @@ asan_files = \
 	asan_posix.cpp \
 	asan_premap_shadow.cpp \
 	asan_report.cpp \
-	asan_rtems.cpp \
 	asan_rtl.cpp \
 	asan_shadow_setup.cpp \
 	asan_stack.cpp \
diff --git a/libsanitizer/asan/Makefile.in b/libsanitizer/asan/Makefile.in
index 25c7fd7b759..53efe526f9c 100644
--- a/libsanitizer/asan/Makefile.in
+++ b/libsanitizer/asan/Makefile.in
@@ -156,9 +156,9 @@ am__objects_1 = asan_activation.lo asan_allocator.lo asan_debugging.lo \
 	asan_interceptors_memintrinsics.lo asan_linux.lo asan_mac.lo \
 	asan_malloc_linux.lo asan_malloc_mac.lo asan_malloc_win.lo \
 	asan_memory_profile.lo asan_new_delete.lo asan_poisoning.lo \
-	asan_posix.lo asan_premap_shadow.lo asan_report.lo \
-	asan_rtems.lo asan_rtl.lo asan_shadow_setup.lo asan_stack.lo \
-	asan_stats.lo asan_suppressions.lo asan_thread.lo asan_win.lo \
+	asan_posix.lo asan_premap_shadow.lo asan_report.lo asan_rtl.lo \
+	asan_shadow_setup.lo asan_stack.lo asan_stats.lo \
+	asan_suppressions.lo asan_thread.lo asan_win.lo \
 	asan_win_dll_thunk.lo asan_win_dynamic_runtime_thunk.lo \
 	asan_interceptors_vfork.lo
 am_libasan_la_OBJECTS = $(am__objects_1)
@@ -446,7 +446,6 @@ asan_files = \
 	asan_posix.cpp \
 	asan_premap_shadow.cpp \
 	asan_report.cpp \
-	asan_rtems.cpp \
 	asan_rtl.cpp \
 	asan_shadow_setup.cpp \
 	asan_stack.cpp \
@@ -604,7 +603,6 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_posix.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_premap_shadow.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_report.Plo@am__quote@
-@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_rtems.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_rtl.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_shadow_setup.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/asan_stack.Plo@am__quote@
diff --git a/libsanitizer/asan/asan_allocator.cpp b/libsanitizer/asan/asan_allocator.cpp
index 7c8bb504332..414fba3b427 100644
--- a/libsanitizer/asan/asan_allocator.cpp
+++ b/libsanitizer/asan/asan_allocator.cpp
@@ -852,12 +852,12 @@ struct Allocator {
     quarantine.PrintStats();
   }
 
-  void ForceLock() {
+  void ForceLock() ACQUIRE(fallback_mutex) {
     allocator.ForceLock();
     fallback_mutex.Lock();
   }
 
-  void ForceUnlock() {
+  void ForceUnlock() RELEASE(fallback_mutex) {
     fallback_mutex.Unlock();
     allocator.ForceUnlock();
   }
@@ -1081,11 +1081,9 @@ uptr asan_mz_size(const void *ptr) {
   return instance.AllocationSize(reinterpret_cast<uptr>(ptr));
 }
 
-void asan_mz_force_lock() {
-  instance.ForceLock();
-}
+void asan_mz_force_lock() NO_THREAD_SAFETY_ANALYSIS { instance.ForceLock(); }
 
-void asan_mz_force_unlock() {
+void asan_mz_force_unlock() NO_THREAD_SAFETY_ANALYSIS {
   instance.ForceUnlock();
 }
 
diff --git a/libsanitizer/asan/asan_errors.cpp b/libsanitizer/asan/asan_errors.cpp
index e68e6971f96..45166c06487 100644
--- a/libsanitizer/asan/asan_errors.cpp
+++ b/libsanitizer/asan/asan_errors.cpp
@@ -533,7 +533,6 @@ static void PrintLegend(InternalScopedString *str) {
   PrintShadowByte(str, "  ASan internal:           ", kAsanInternalHeapMagic);
   PrintShadowByte(str, "  Left alloca redzone:     ", kAsanAllocaLeftMagic);
   PrintShadowByte(str, "  Right alloca redzone:    ", kAsanAllocaRightMagic);
-  PrintShadowByte(str, "  Shadow gap:              ", kAsanShadowGap);
 }
 
 static void PrintShadowBytes(InternalScopedString *str, const char *before,
diff --git a/libsanitizer/asan/asan_fake_stack.cpp b/libsanitizer/asan/asan_fake_stack.cpp
index 1f873fec7d7..bf5c342ee59 100644
--- a/libsanitizer/asan/asan_fake_stack.cpp
+++ b/libsanitizer/asan/asan_fake_stack.cpp
@@ -187,7 +187,7 @@ void SetTLSFakeStack(FakeStack *fs) { }
 static FakeStack *GetFakeStack() {
   AsanThread *t = GetCurrentThread();
   if (!t) return nullptr;
-  return t->fake_stack();
+  return t->get_or_create_fake_stack();
 }
 
 static FakeStack *GetFakeStackFast() {
@@ -198,7 +198,13 @@ static FakeStack *GetFakeStackFast() {
   return GetFakeStack();
 }
 
-ALWAYS_INLINE uptr OnMalloc(uptr class_id, uptr size) {
+static FakeStack *GetFakeStackFastAlways() {
+  if (FakeStack *fs = GetTLSFakeStack())
+    return fs;
+  return GetFakeStack();
+}
+
+static ALWAYS_INLINE uptr OnMalloc(uptr class_id, uptr size) {
   FakeStack *fs = GetFakeStackFast();
   if (!fs) return 0;
   uptr local_stack;
@@ -210,7 +216,21 @@ ALWAYS_INLINE uptr OnMalloc(uptr class_id, uptr size) {
   return ptr;
 }
 
-ALWAYS_INLINE void OnFree(uptr ptr, uptr class_id, uptr size) {
+static ALWAYS_INLINE uptr OnMallocAlways(uptr class_id, uptr size) {
+  FakeStack *fs = GetFakeStackFastAlways();
+  if (!fs)
+    return 0;
+  uptr local_stack;
+  uptr real_stack = reinterpret_cast<uptr>(&local_stack);
+  FakeFrame *ff = fs->Allocate(fs->stack_size_log(), class_id, real_stack);
+  if (!ff)
+    return 0;  // Out of fake stack.
+  uptr ptr = reinterpret_cast<uptr>(ff);
+  SetShadow(ptr, size, class_id, 0);
+  return ptr;
+}
+
+static ALWAYS_INLINE void OnFree(uptr ptr, uptr class_id, uptr size) {
   FakeStack::Deallocate(ptr, class_id);
   SetShadow(ptr, size, class_id, kMagic8);
 }
@@ -219,14 +239,18 @@ ALWAYS_INLINE void OnFree(uptr ptr, uptr class_id, uptr size) {
 
 // ---------------------- Interface ---------------- {{{1
 using namespace __asan;
-#define DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(class_id)                       \
-  extern "C" SANITIZER_INTERFACE_ATTRIBUTE uptr                                \
-      __asan_stack_malloc_##class_id(uptr size) {                              \
-    return OnMalloc(class_id, size);                                           \
-  }                                                                            \
-  extern "C" SANITIZER_INTERFACE_ATTRIBUTE void __asan_stack_free_##class_id(  \
-      uptr ptr, uptr size) {                                                   \
-    OnFree(ptr, class_id, size);                                               \
+#define DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(class_id)                      \
+  extern "C" SANITIZER_INTERFACE_ATTRIBUTE uptr                               \
+      __asan_stack_malloc_##class_id(uptr size) {                             \
+    return OnMalloc(class_id, size);                                          \
+  }                                                                           \
+  extern "C" SANITIZER_INTERFACE_ATTRIBUTE uptr                               \
+      __asan_stack_malloc_always_##class_id(uptr size) {                      \
+    return OnMallocAlways(class_id, size);                                    \
+  }                                                                           \
+  extern "C" SANITIZER_INTERFACE_ATTRIBUTE void __asan_stack_free_##class_id( \
+      uptr ptr, uptr size) {                                                  \
+    OnFree(ptr, class_id, size);                                              \
   }
 
 DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(0)
@@ -240,7 +264,11 @@ DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(7)
 DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(8)
 DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(9)
 DEFINE_STACK_MALLOC_FREE_WITH_CLASS_ID(10)
+
 extern "C" {
+// TODO: remove this method and fix tests that use it by setting
+// -asan-use-after-return=never, after modal UAR flag lands
+// (https://github.com/google/sanitizers/issues/1394)
 SANITIZER_INTERFACE_ATTRIBUTE
 void *__asan_get_current_fake_stack() { return GetFakeStackFast(); }
 
diff --git a/libsanitizer/asan/asan_flags.cpp b/libsanitizer/asan/asan_flags.cpp
index cb6a89fe32c..c64e4647028 100644
--- a/libsanitizer/asan/asan_flags.cpp
+++ b/libsanitizer/asan/asan_flags.cpp
@@ -155,10 +155,6 @@ void InitializeFlags() {
   CHECK_LE(f->max_redzone, 2048);
   CHECK(IsPowerOfTwo(f->redzone));
   CHECK(IsPowerOfTwo(f->max_redzone));
-  if (SANITIZER_RTEMS) {
-    CHECK(!f->unmap_shadow_on_exit);
-    CHECK(!f->protect_shadow_gap);
-  }
 
   // quarantine_size is deprecated but we still honor it.
   // quarantine_size can not be used together with quarantine_size_mb.
diff --git a/libsanitizer/asan/asan_flags.inc b/libsanitizer/asan/asan_flags.inc
index 43c70dbca56..514b225c407 100644
--- a/libsanitizer/asan/asan_flags.inc
+++ b/libsanitizer/asan/asan_flags.inc
@@ -87,8 +87,7 @@ ASAN_FLAG(bool, check_malloc_usable_size, true,
           "295.*.")
 ASAN_FLAG(bool, unmap_shadow_on_exit, false,
           "If set, explicitly unmaps the (huge) shadow at exit.")
-ASAN_FLAG(bool, protect_shadow_gap, !SANITIZER_RTEMS,
-          "If set, mprotect the shadow gap")
+ASAN_FLAG(bool, protect_shadow_gap, true, "If set, mprotect the shadow gap")
 ASAN_FLAG(bool, print_stats, false,
           "Print various statistics after printing an error message or if "
           "atexit=1.")
diff --git a/libsanitizer/asan/asan_globals.cpp b/libsanitizer/asan/asan_globals.cpp
index e045c31cd1c..9d7dbc6f264 100644
--- a/libsanitizer/asan/asan_globals.cpp
+++ b/libsanitizer/asan/asan_globals.cpp
@@ -154,6 +154,23 @@ static void CheckODRViolationViaIndicator(const Global *g) {
   }
 }
 
+// Check ODR violation for given global G by checking if it's already poisoned.
+// We use this method in case compiler doesn't use private aliases for global
+// variables.
+static void CheckODRViolationViaPoisoning(const Global *g) {
+  if (__asan_region_is_poisoned(g->beg, g->size_with_redzone)) {
+    // This check may not be enough: if the first global is much larger
+    // the entire redzone of the second global may be within the first global.
+    for (ListOfGlobals *l = list_of_all_globals; l; l = l->next) {
+      if (g->beg == l->g->beg &&
+          (flags()->detect_odr_violation >= 2 || g->size != l->g->size) &&
+          !IsODRViolationSuppressed(g->name))
+        ReportODRViolation(g, FindRegistrationSite(g),
+                           l->g, FindRegistrationSite(l->g));
+    }
+  }
+}
+
 // Clang provides two different ways for global variables protection:
 // it can poison the global itself or its private alias. In former
 // case we may poison same symbol multiple times, that can help us to
@@ -199,6 +216,8 @@ static void RegisterGlobal(const Global *g) {
     // where two globals with the same name are defined in different modules.
     if (UseODRIndicator(g))
       CheckODRViolationViaIndicator(g);
+    else
+      CheckODRViolationViaPoisoning(g);
   }
   if (CanPoisonMemory())
     PoisonRedZones(*g);
diff --git a/libsanitizer/asan/asan_interceptors.cpp b/libsanitizer/asan/asan_interceptors.cpp
index 9db7db89fa1..d0a6dd48a74 100644
--- a/libsanitizer/asan/asan_interceptors.cpp
+++ b/libsanitizer/asan/asan_interceptors.cpp
@@ -23,25 +23,25 @@
 #include "lsan/lsan_common.h"
 #include "sanitizer_common/sanitizer_libc.h"
 
-// There is no general interception at all on Fuchsia and RTEMS.
+// There is no general interception at all on Fuchsia.
 // Only the functions in asan_interceptors_memintrinsics.cpp are
 // really defined to replace libc functions.
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#if !SANITIZER_FUCHSIA
 
-#if SANITIZER_POSIX
-#include "sanitizer_common/sanitizer_posix.h"
-#endif
+#  if SANITIZER_POSIX
+#    include "sanitizer_common/sanitizer_posix.h"
+#  endif
 
-#if ASAN_INTERCEPT__UNWIND_RAISEEXCEPTION || \
-    ASAN_INTERCEPT__SJLJ_UNWIND_RAISEEXCEPTION
-#include <unwind.h>
-#endif
+#  if ASAN_INTERCEPT__UNWIND_RAISEEXCEPTION || \
+      ASAN_INTERCEPT__SJLJ_UNWIND_RAISEEXCEPTION
+#    include <unwind.h>
+#  endif
 
-#if defined(__i386) && SANITIZER_LINUX
-#define ASAN_PTHREAD_CREATE_VERSION "GLIBC_2.1"
-#elif defined(__mips__) && SANITIZER_LINUX
-#define ASAN_PTHREAD_CREATE_VERSION "GLIBC_2.2"
-#endif
+#  if defined(__i386) && SANITIZER_LINUX
+#    define ASAN_PTHREAD_CREATE_VERSION "GLIBC_2.1"
+#  elif defined(__mips__) && SANITIZER_LINUX
+#    define ASAN_PTHREAD_CREATE_VERSION "GLIBC_2.2"
+#  endif
 
 namespace __asan {
 
diff --git a/libsanitizer/asan/asan_interceptors.h b/libsanitizer/asan/asan_interceptors.h
index 5bd5be6bac2..a9249dea45b 100644
--- a/libsanitizer/asan/asan_interceptors.h
+++ b/libsanitizer/asan/asan_interceptors.h
@@ -34,10 +34,10 @@ void InitializePlatformInterceptors();
 
 }  // namespace __asan
 
-// There is no general interception at all on Fuchsia and RTEMS.
+// There is no general interception at all on Fuchsia.
 // Only the functions in asan_interceptors_memintrinsics.h are
 // really defined to replace libc functions.
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#if !SANITIZER_FUCHSIA
 
 // Use macro to describe if specific function should be
 // intercepted on a given platform.
@@ -81,12 +81,7 @@ void InitializePlatformInterceptors();
 #if ASAN_HAS_EXCEPTIONS && !SANITIZER_WINDOWS && !SANITIZER_SOLARIS && \
     !SANITIZER_NETBSD
 # define ASAN_INTERCEPT___CXA_THROW 1
-# if ! defined(ASAN_HAS_CXA_RETHROW_PRIMARY_EXCEPTION) \
-     || ASAN_HAS_CXA_RETHROW_PRIMARY_EXCEPTION
-#   define ASAN_INTERCEPT___CXA_RETHROW_PRIMARY_EXCEPTION 1
-# else
-#   define ASAN_INTERCEPT___CXA_RETHROW_PRIMARY_EXCEPTION 0
-# endif
+# define ASAN_INTERCEPT___CXA_RETHROW_PRIMARY_EXCEPTION 1
 # if defined(_GLIBCXX_SJLJ_EXCEPTIONS) || (SANITIZER_IOS && defined(__arm__))
 #  define ASAN_INTERCEPT__UNWIND_SJLJ_RAISEEXCEPTION 1
 # else
diff --git a/libsanitizer/asan/asan_interceptors_memintrinsics.cpp b/libsanitizer/asan/asan_interceptors_memintrinsics.cpp
index ccdd5159042..9c316bb9574 100644
--- a/libsanitizer/asan/asan_interceptors_memintrinsics.cpp
+++ b/libsanitizer/asan/asan_interceptors_memintrinsics.cpp
@@ -30,9 +30,9 @@ void *__asan_memmove(void *to, const void *from, uptr size) {
   ASAN_MEMMOVE_IMPL(nullptr, to, from, size);
 }
 
-#if SANITIZER_FUCHSIA || SANITIZER_RTEMS
+#if SANITIZER_FUCHSIA
 
-// Fuchsia and RTEMS don't use sanitizer_common_interceptors.inc, but
+// Fuchsia doesn't use sanitizer_common_interceptors.inc, but
 // the only things there it wants are these three.  Just define them
 // as aliases here rather than repeating the contents.
 
@@ -40,4 +40,4 @@ extern "C" decltype(__asan_memcpy) memcpy[[gnu::alias("__asan_memcpy")]];
 extern "C" decltype(__asan_memmove) memmove[[gnu::alias("__asan_memmove")]];
 extern "C" decltype(__asan_memset) memset[[gnu::alias("__asan_memset")]];
 
-#endif  // SANITIZER_FUCHSIA || SANITIZER_RTEMS
+#endif  // SANITIZER_FUCHSIA
diff --git a/libsanitizer/asan/asan_interface.inc b/libsanitizer/asan/asan_interface.inc
index 94801043982..ea28fc8ae87 100644
--- a/libsanitizer/asan/asan_interface.inc
+++ b/libsanitizer/asan/asan_interface.inc
@@ -134,6 +134,17 @@ INTERFACE_FUNCTION(__asan_stack_malloc_7)
 INTERFACE_FUNCTION(__asan_stack_malloc_8)
 INTERFACE_FUNCTION(__asan_stack_malloc_9)
 INTERFACE_FUNCTION(__asan_stack_malloc_10)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_0)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_1)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_2)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_3)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_4)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_5)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_6)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_7)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_8)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_9)
+INTERFACE_FUNCTION(__asan_stack_malloc_always_10)
 INTERFACE_FUNCTION(__asan_store1)
 INTERFACE_FUNCTION(__asan_store2)
 INTERFACE_FUNCTION(__asan_store4)
diff --git a/libsanitizer/asan/asan_internal.h b/libsanitizer/asan/asan_internal.h
index cfb54927c6c..ad3320304d0 100644
--- a/libsanitizer/asan/asan_internal.h
+++ b/libsanitizer/asan/asan_internal.h
@@ -35,11 +35,11 @@
 // If set, values like allocator chunk size, as well as defaults for some flags
 // will be changed towards less memory overhead.
 #ifndef ASAN_LOW_MEMORY
-# if SANITIZER_IOS || SANITIZER_ANDROID || SANITIZER_RTEMS
-#  define ASAN_LOW_MEMORY 1
-# else
-#  define ASAN_LOW_MEMORY 0
-# endif
+#  if SANITIZER_IOS || SANITIZER_ANDROID
+#    define ASAN_LOW_MEMORY 1
+#  else
+#    define ASAN_LOW_MEMORY 0
+#  endif
 #endif
 
 #ifndef ASAN_DYNAMIC
@@ -77,7 +77,7 @@ void InitializeShadowMemory();
 // asan_malloc_linux.cpp / asan_malloc_mac.cpp
 void ReplaceSystemMalloc();
 
-// asan_linux.cpp / asan_mac.cpp / asan_rtems.cpp / asan_win.cpp
+// asan_linux.cpp / asan_mac.cpp / asan_win.cpp
 uptr FindDynamicShadowStart();
 void *AsanDoesNotSupportStaticLinkage();
 void AsanCheckDynamicRTPrereqs();
@@ -159,9 +159,6 @@ const int kAsanArrayCookieMagic = 0xac;
 const int kAsanIntraObjectRedzone = 0xbb;
 const int kAsanAllocaLeftMagic = 0xca;
 const int kAsanAllocaRightMagic = 0xcb;
-// Used to populate the shadow gap for systems without memory
-// protection there (i.e. Myriad).
-const int kAsanShadowGap = 0xcc;
 
 static const uptr kCurrentStackFrameMagic = 0x41B58AB3;
 static const uptr kRetiredStackFrameMagic = 0x45E0360E;
diff --git a/libsanitizer/asan/asan_malloc_linux.cpp b/libsanitizer/asan/asan_malloc_linux.cpp
index 9c3f0a5338e..c6bec8551bc 100644
--- a/libsanitizer/asan/asan_malloc_linux.cpp
+++ b/libsanitizer/asan/asan_malloc_linux.cpp
@@ -15,23 +15,22 @@
 
 #include "sanitizer_common/sanitizer_platform.h"
 #if SANITIZER_FREEBSD || SANITIZER_FUCHSIA || SANITIZER_LINUX || \
-    SANITIZER_NETBSD || SANITIZER_RTEMS || SANITIZER_SOLARIS
+    SANITIZER_NETBSD || SANITIZER_SOLARIS
 
-#include "sanitizer_common/sanitizer_allocator_checks.h"
-#include "sanitizer_common/sanitizer_errno.h"
-#include "sanitizer_common/sanitizer_tls_get_addr.h"
-#include "asan_allocator.h"
-#include "asan_interceptors.h"
-#include "asan_internal.h"
-#include "asan_malloc_local.h"
-#include "asan_stack.h"
+#  include "asan_allocator.h"
+#  include "asan_interceptors.h"
+#  include "asan_internal.h"
+#  include "asan_stack.h"
+#  include "sanitizer_common/sanitizer_allocator_checks.h"
+#  include "sanitizer_common/sanitizer_errno.h"
+#  include "sanitizer_common/sanitizer_tls_get_addr.h"
 
 // ---------------------- Replacement functions ---------------- {{{1
 using namespace __asan;
 
 static uptr allocated_for_dlsym;
 static uptr last_dlsym_alloc_size_in_words;
-static const uptr kDlsymAllocPoolSize = SANITIZER_RTEMS ? 4096 : 1024;
+static const uptr kDlsymAllocPoolSize = 1024;
 static uptr alloc_memory_for_dlsym[kDlsymAllocPoolSize];
 
 static inline bool IsInDlsymAllocPool(const void *ptr) {
@@ -82,27 +81,12 @@ static int PosixMemalignFromLocalPool(void **memptr, uptr alignment,
   return 0;
 }
 
-#if SANITIZER_RTEMS
-void* MemalignFromLocalPool(uptr alignment, uptr size) {
-  void *ptr = nullptr;
-  alignment = Max(alignment, kWordSize);
-  PosixMemalignFromLocalPool(&ptr, alignment, size);
-  return ptr;
-}
-
-bool IsFromLocalPool(const void *ptr) {
-  return IsInDlsymAllocPool(ptr);
-}
-#endif
-
 static inline bool MaybeInDlsym() {
   // Fuchsia doesn't use dlsym-based interceptors.
   return !SANITIZER_FUCHSIA && asan_init_is_running;
 }
 
-static inline bool UseLocalPool() {
-  return EarlyMalloc() || MaybeInDlsym();
-}
+static inline bool UseLocalPool() { return MaybeInDlsym(); }
 
 static void *ReallocFromLocalPool(void *ptr, uptr size) {
   const uptr offset = (uptr)ptr - (uptr)alloc_memory_for_dlsym;
diff --git a/libsanitizer/asan/asan_malloc_local.h b/libsanitizer/asan/asan_malloc_local.h
deleted file mode 100644
index e2c9be0379f..00000000000
--- a/libsanitizer/asan/asan_malloc_local.h
+++ /dev/null
@@ -1,52 +0,0 @@
-//===-- asan_malloc_local.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 AddressSanitizer, an address sanity checker.
-//
-// Provide interfaces to check for and handle local pool memory allocation.
-//===----------------------------------------------------------------------===//
-
-#ifndef ASAN_MALLOC_LOCAL_H
-#define ASAN_MALLOC_LOCAL_H
-
-#include "sanitizer_common/sanitizer_platform.h"
-#include "asan_internal.h"
-
-static inline bool EarlyMalloc() {
-  return SANITIZER_RTEMS &&
-         (!__asan::asan_inited || __asan::asan_init_is_running);
-}
-
-#if SANITIZER_RTEMS
-
-bool IsFromLocalPool(const void *ptr);
-void *MemalignFromLocalPool(uptr alignment, uptr size);
-
-// On RTEMS, we use the local pool to handle memory allocation when the ASan
-// run-time is not up. This macro is expanded in the context of the operator new
-// implementation.
-#define MAYBE_ALLOCATE_FROM_LOCAL_POOL(nothrow)                    \
-  do {                                                             \
-    if (UNLIKELY(EarlyMalloc())) {                                 \
-      void *res = MemalignFromLocalPool(SHADOW_GRANULARITY, size); \
-      if (!nothrow)                                                \
-        CHECK(res);                                                \
-      return res;                                                  \
-    }                                                              \
-  } while (0)
-
-#define IS_FROM_LOCAL_POOL(ptr) UNLIKELY(IsFromLocalPool(ptr))
-
-#else  // SANITIZER_RTEMS
-
-#define MAYBE_ALLOCATE_FROM_LOCAL_POOL(nothrow)
-#define IS_FROM_LOCAL_POOL(ptr) 0
-
-#endif  // SANITIZER_RTEMS
-
-#endif  // ASAN_MALLOC_LOCAL_H
diff --git a/libsanitizer/asan/asan_mapping.h b/libsanitizer/asan/asan_mapping.h
index 2511e7043df..e5a7f2007ae 100644
--- a/libsanitizer/asan/asan_mapping.h
+++ b/libsanitizer/asan/asan_mapping.h
@@ -150,17 +150,11 @@
 // || `[0x36000000, 0x39ffffff]` || ShadowGap  ||
 // || `[0x30000000, 0x35ffffff]` || LowShadow  ||
 // || `[0x00000000, 0x2fffffff]` || LowMem     ||
-//
-// Shadow mapping on Myriad2 (for shadow scale 5):
-// || `[0x9ff80000, 0x9fffffff]` || ShadowGap  ||
-// || `[0x9f000000, 0x9ff7ffff]` || LowShadow  ||
-// || `[0x80000000, 0x9effffff]` || LowMem     ||
-// || `[0x00000000, 0x7fffffff]` || Ignored    ||
 
 #if defined(ASAN_SHADOW_SCALE)
 static const u64 kDefaultShadowScale = ASAN_SHADOW_SCALE;
 #else
-static const u64 kDefaultShadowScale = SANITIZER_MYRIAD2 ? 5 : 3;
+static const u64 kDefaultShadowScale = 3;
 #endif
 static const u64 kDefaultShadowSentinel = ~(uptr)0;
 static const u64 kDefaultShadowOffset32 = 1ULL << 29;  // 0x20000000
@@ -171,7 +165,7 @@ static const u64 kAArch64_ShadowOffset64 = 1ULL << 36;
 static const u64 kRiscv64_ShadowOffset64 = 0xd55550000;
 static const u64 kMIPS32_ShadowOffset32 = 0x0aaa0000;
 static const u64 kMIPS64_ShadowOffset64 = 1ULL << 37;
-static const u64 kPPC64_ShadowOffset64 = 1ULL << 41;
+static const u64 kPPC64_ShadowOffset64 = 1ULL << 44;
 static const u64 kSystemZ_ShadowOffset64 = 1ULL << 52;
 static const u64 kSPARC64_ShadowOffset64 = 1ULL << 43;  // 0x80000000000
 static const u64 kFreeBSD_ShadowOffset32 = 1ULL << 30;  // 0x40000000
@@ -180,15 +174,6 @@ static const u64 kNetBSD_ShadowOffset32 = 1ULL << 30;  // 0x40000000
 static const u64 kNetBSD_ShadowOffset64 = 1ULL << 46;  // 0x400000000000
 static const u64 kWindowsShadowOffset32 = 3ULL << 28;  // 0x30000000
 
-static const u64 kMyriadMemoryOffset32 = 0x80000000ULL;
-static const u64 kMyriadMemorySize32 = 0x20000000ULL;
-static const u64 kMyriadMemoryEnd32 =
-    kMyriadMemoryOffset32 + kMyriadMemorySize32 - 1;
-static const u64 kMyriadShadowOffset32 =
-    (kMyriadMemoryOffset32 + kMyriadMemorySize32 -
-     (kMyriadMemorySize32 >> kDefaultShadowScale));
-static const u64 kMyriadCacheBitMask32 = 0x40000000ULL;
-
 #define SHADOW_SCALE kDefaultShadowScale
 
 #if SANITIZER_FUCHSIA
@@ -206,8 +191,6 @@ static const u64 kMyriadCacheBitMask32 = 0x40000000ULL;
 #    define SHADOW_OFFSET kWindowsShadowOffset32
 #  elif SANITIZER_IOS
 #    define SHADOW_OFFSET __asan_shadow_memory_dynamic_address
-#  elif SANITIZER_MYRIAD2
-#    define SHADOW_OFFSET kMyriadShadowOffset32
 #  else
 #    define SHADOW_OFFSET kDefaultShadowOffset32
 #  endif
@@ -278,10 +261,8 @@ extern uptr kHighMemEnd, kMidMemBeg, kMidMemEnd;  // Initialized in __asan_init.
 
 }  // namespace __asan
 
-#if SANITIZER_MYRIAD2
-#include "asan_mapping_myriad.h"
-#elif defined(__sparc__) && SANITIZER_WORDSIZE == 64
-#include "asan_mapping_sparc64.h"
+#if defined(__sparc__) && SANITIZER_WORDSIZE == 64
+#  include "asan_mapping_sparc64.h"
 #else
 #define MEM_TO_SHADOW(mem) (((mem) >> SHADOW_SCALE) + (SHADOW_OFFSET))
 
@@ -363,7 +344,7 @@ static inline bool AddrIsInShadowGap(uptr a) {
 
 }  // namespace __asan
 
-#endif  // SANITIZER_MYRIAD2
+#endif
 
 namespace __asan {
 
@@ -393,8 +374,6 @@ static inline bool AddrIsAlignedByGranularity(uptr a) {
 
 static inline bool AddressIsPoisoned(uptr a) {
   PROFILE_ASAN_MAPPING();
-  if (SANITIZER_MYRIAD2 && !AddrIsInMem(a) && !AddrIsInShadow(a))
-    return false;
   const uptr kAccessSize = 1;
   u8 *shadow_address = (u8*)MEM_TO_SHADOW(a);
   s8 shadow_value = *shadow_address;
diff --git a/libsanitizer/asan/asan_mapping_myriad.h b/libsanitizer/asan/asan_mapping_myriad.h
deleted file mode 100644
index 6969e3a4931..00000000000
--- a/libsanitizer/asan/asan_mapping_myriad.h
+++ /dev/null
@@ -1,85 +0,0 @@
-//===-- asan_mapping_myriad.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 AddressSanitizer, an address sanity checker.
-//
-// Myriad-specific definitions for ASan memory mapping.
-//===----------------------------------------------------------------------===//
-#ifndef ASAN_MAPPING_MYRIAD_H
-#define ASAN_MAPPING_MYRIAD_H
-
-#define RAW_ADDR(mem) ((mem) & ~kMyriadCacheBitMask32)
-#define MEM_TO_SHADOW(mem) \
-  (((RAW_ADDR(mem) - kLowMemBeg) >> SHADOW_SCALE) + (SHADOW_OFFSET))
-
-#define kLowMemBeg     kMyriadMemoryOffset32
-#define kLowMemEnd     (SHADOW_OFFSET - 1)
-
-#define kLowShadowBeg  SHADOW_OFFSET
-#define kLowShadowEnd  MEM_TO_SHADOW(kLowMemEnd)
-
-#define kHighMemBeg    0
-
-#define kHighShadowBeg 0
-#define kHighShadowEnd 0
-
-#define kMidShadowBeg  0
-#define kMidShadowEnd  0
-
-#define kShadowGapBeg  (kLowShadowEnd + 1)
-#define kShadowGapEnd  kMyriadMemoryEnd32
-
-#define kShadowGap2Beg 0
-#define kShadowGap2End 0
-
-#define kShadowGap3Beg 0
-#define kShadowGap3End 0
-
-namespace __asan {
-
-static inline bool AddrIsInLowMem(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  a = RAW_ADDR(a);
-  return a >= kLowMemBeg && a <= kLowMemEnd;
-}
-
-static inline bool AddrIsInLowShadow(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  a = RAW_ADDR(a);
-  return a >= kLowShadowBeg && a <= kLowShadowEnd;
-}
-
-static inline bool AddrIsInMidMem(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  return false;
-}
-
-static inline bool AddrIsInMidShadow(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  return false;
-}
-
-static inline bool AddrIsInHighMem(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  return false;
-}
-
-static inline bool AddrIsInHighShadow(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  return false;
-}
-
-static inline bool AddrIsInShadowGap(uptr a) {
-  PROFILE_ASAN_MAPPING();
-  a = RAW_ADDR(a);
-  return a >= kShadowGapBeg && a <= kShadowGapEnd;
-}
-
-}  // namespace __asan
-
-#endif  // ASAN_MAPPING_MYRIAD_H
diff --git a/libsanitizer/asan/asan_new_delete.cpp b/libsanitizer/asan/asan_new_delete.cpp
index 92a8648452b..da446072de1 100644
--- a/libsanitizer/asan/asan_new_delete.cpp
+++ b/libsanitizer/asan/asan_new_delete.cpp
@@ -11,16 +11,14 @@
 // Interceptors for operators new and delete.
 //===----------------------------------------------------------------------===//
 
+#include <stddef.h>
+
 #include "asan_allocator.h"
 #include "asan_internal.h"
-#include "asan_malloc_local.h"
 #include "asan_report.h"
 #include "asan_stack.h"
-
 #include "interception/interception.h"
 
-#include <stddef.h>
-
 // C++ operators can't have dllexport attributes on Windows. We export them
 // anyway by passing extra -export flags to the linker, which is exactly that
 // dllexport would normally do. We need to export them in order to make the
@@ -72,14 +70,12 @@ enum class align_val_t: size_t {};
 // For local pool allocation, align to SHADOW_GRANULARITY to match asan
 // allocator behavior.
 #define OPERATOR_NEW_BODY(type, nothrow)            \
-  MAYBE_ALLOCATE_FROM_LOCAL_POOL(nothrow);          \
   GET_STACK_TRACE_MALLOC;                           \
   void *res = asan_memalign(0, size, &stack, type); \
   if (!nothrow && UNLIKELY(!res))                   \
     ReportOutOfMemory(size, &stack);                \
   return res;
 #define OPERATOR_NEW_BODY_ALIGN(type, nothrow)                \
-  MAYBE_ALLOCATE_FROM_LOCAL_POOL(nothrow);                    \
   GET_STACK_TRACE_MALLOC;                                     \
   void *res = asan_memalign((uptr)align, size, &stack, type); \
   if (!nothrow && UNLIKELY(!res))                             \
@@ -135,23 +131,19 @@ INTERCEPTOR(void *, _ZnamRKSt9nothrow_t, size_t size, std::nothrow_t const&) {
 #endif  // !SANITIZER_MAC
 
 #define OPERATOR_DELETE_BODY(type) \
-  if (IS_FROM_LOCAL_POOL(ptr)) return;\
-  GET_STACK_TRACE_FREE;\
+  GET_STACK_TRACE_FREE;            \
   asan_delete(ptr, 0, 0, &stack, type);
 
 #define OPERATOR_DELETE_BODY_SIZE(type) \
-  if (IS_FROM_LOCAL_POOL(ptr)) return;\
-  GET_STACK_TRACE_FREE;\
+  GET_STACK_TRACE_FREE;                 \
   asan_delete(ptr, size, 0, &stack, type);
 
 #define OPERATOR_DELETE_BODY_ALIGN(type) \
-  if (IS_FROM_LOCAL_POOL(ptr)) return;\
-  GET_STACK_TRACE_FREE;\
+  GET_STACK_TRACE_FREE;                  \
   asan_delete(ptr, 0, static_cast<uptr>(align), &stack, type);
 
 #define OPERATOR_DELETE_BODY_SIZE_ALIGN(type) \
-  if (IS_FROM_LOCAL_POOL(ptr)) return;\
-  GET_STACK_TRACE_FREE;\
+  GET_STACK_TRACE_FREE;                       \
   asan_delete(ptr, size, static_cast<uptr>(align), &stack, type);
 
 #if !SANITIZER_MAC
diff --git a/libsanitizer/asan/asan_poisoning.cpp b/libsanitizer/asan/asan_poisoning.cpp
index fa149ecfde6..5f215fe0f9b 100644
--- a/libsanitizer/asan/asan_poisoning.cpp
+++ b/libsanitizer/asan/asan_poisoning.cpp
@@ -173,17 +173,13 @@ int __asan_address_is_poisoned(void const volatile *addr) {
 }
 
 uptr __asan_region_is_poisoned(uptr beg, uptr size) {
-  if (!size) return 0;
+  if (!size)
+    return 0;
   uptr end = beg + size;
-  if (SANITIZER_MYRIAD2) {
-    // On Myriad, address not in DRAM range need to be treated as
-    // unpoisoned.
-    if (!AddrIsInMem(beg) && !AddrIsInShadow(beg)) return 0;
-    if (!AddrIsInMem(end) && !AddrIsInShadow(end)) return 0;
-  } else {
-    if (!AddrIsInMem(beg)) return beg;
-    if (!AddrIsInMem(end)) return end;
-  }
+  if (!AddrIsInMem(beg))
+    return beg;
+  if (!AddrIsInMem(end))
+    return end;
   CHECK_LT(beg, end);
   uptr aligned_b = RoundUpTo(beg, SHADOW_GRANULARITY);
   uptr aligned_e = RoundDownTo(end, SHADOW_GRANULARITY);
@@ -192,8 +188,7 @@ uptr __asan_region_is_poisoned(uptr beg, uptr size) {
   // First check the first and the last application bytes,
   // then check the SHADOW_GRANULARITY-aligned region by calling
   // mem_is_zero on the corresponding shadow.
-  if (!__asan::AddressIsPoisoned(beg) &&
-      !__asan::AddressIsPoisoned(end - 1) &&
+  if (!__asan::AddressIsPoisoned(beg) && !__asan::AddressIsPoisoned(end - 1) &&
       (shadow_end <= shadow_beg ||
        __sanitizer::mem_is_zero((const char *)shadow_beg,
                                 shadow_end - shadow_beg)))
diff --git a/libsanitizer/asan/asan_poisoning.h b/libsanitizer/asan/asan_poisoning.h
index 62dd9bd0edd..3d536f2d309 100644
--- a/libsanitizer/asan/asan_poisoning.h
+++ b/libsanitizer/asan/asan_poisoning.h
@@ -51,9 +51,6 @@ ALWAYS_INLINE void FastPoisonShadow(uptr aligned_beg, uptr aligned_size,
   // probably provide higher-level interface for these operations.
   // For now, just memset on Windows.
   if (value || SANITIZER_WINDOWS == 1 ||
-      // RTEMS doesn't have have pages, let alone a fast way to zero
-      // them, so default to memset.
-      SANITIZER_RTEMS == 1 ||
       shadow_end - shadow_beg < common_flags()->clear_shadow_mmap_threshold) {
     REAL(memset)((void*)shadow_beg, value, shadow_end - shadow_beg);
   } else {
diff --git a/libsanitizer/asan/asan_rtems.cpp b/libsanitizer/asan/asan_rtems.cpp
deleted file mode 100644
index ea0b4ad9db6..00000000000
--- a/libsanitizer/asan/asan_rtems.cpp
+++ /dev/null
@@ -1,266 +0,0 @@
-//===-- asan_rtems.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 AddressSanitizer, an address sanity checker.
-//
-// RTEMS-specific details.
-//===----------------------------------------------------------------------===//
-
-#include "sanitizer_common/sanitizer_rtems.h"
-#if SANITIZER_RTEMS
-
-#include "asan_internal.h"
-#include "asan_interceptors.h"
-#include "asan_mapping.h"
-#include "asan_poisoning.h"
-#include "asan_report.h"
-#include "asan_stack.h"
-#include "sanitizer_common/sanitizer_common.h"
-#include "sanitizer_common/sanitizer_libc.h"
-
-#include <pthread.h>
-#include <stdlib.h>
-
-namespace __asan {
-
-static void ResetShadowMemory() {
-  uptr shadow_start = SHADOW_OFFSET;
-  uptr shadow_end = MEM_TO_SHADOW(kMyriadMemoryEnd32);
-  uptr gap_start = MEM_TO_SHADOW(shadow_start);
-  uptr gap_end = MEM_TO_SHADOW(shadow_end);
-
-  REAL(memset)((void *)shadow_start, 0, shadow_end - shadow_start);
-  REAL(memset)((void *)gap_start, kAsanShadowGap, gap_end - gap_start);
-}
-
-void InitializeShadowMemory() {
-  kHighMemEnd = 0;
-  kMidMemBeg =  0;
-  kMidMemEnd =  0;
-
-  ResetShadowMemory();
-}
-
-void AsanApplyToGlobals(globals_op_fptr op, const void *needle) {
-  UNIMPLEMENTED();
-}
-
-void FlushUnneededASanShadowMemory(uptr p, uptr size) {
-  // Since asan's mapping is compacting, the shadow chunk may be
-  // not page-aligned, so we only flush the page-aligned portion.
-  ReleaseMemoryPagesToOS(MemToShadow(p), MemToShadow(p + size));
-}
-
-void AsanCheckDynamicRTPrereqs() {}
-void AsanCheckIncompatibleRT() {}
-void InitializeAsanInterceptors() {}
-void InitializePlatformInterceptors() {}
-void InitializePlatformExceptionHandlers() {}
-
-// RTEMS only support static linking; it sufficies to return with no
-// error.
-void *AsanDoesNotSupportStaticLinkage() { return nullptr; }
-
-void AsanOnDeadlySignal(int signo, void *siginfo, void *context) {
-  UNIMPLEMENTED();
-}
-
-bool PlatformUnpoisonStacks() { return false; }
-
-void EarlyInit() {
-  // Provide early initialization of shadow memory so that
-  // instrumented code running before full initialzation will not
-  // report spurious errors.
-  ResetShadowMemory();
-}
-
-// We can use a plain thread_local variable for TSD.
-static thread_local void *per_thread;
-
-void *AsanTSDGet() { return per_thread; }
-
-void AsanTSDSet(void *tsd) { per_thread = tsd; }
-
-// There's no initialization needed, and the passed-in destructor
-// will never be called.  Instead, our own thread destruction hook
-// (below) will call AsanThread::TSDDtor directly.
-void AsanTSDInit(void (*destructor)(void *tsd)) {
-  DCHECK(destructor == &PlatformTSDDtor);
-}
-
-void PlatformTSDDtor(void *tsd) { UNREACHABLE(__func__); }
-
-//
-// Thread registration.  We provide an API similar to the Fushia port.
-//
-
-struct AsanThread::InitOptions {
-  uptr stack_bottom, stack_size, tls_bottom, tls_size;
-};
-
-// Shared setup between thread creation and startup for the initial thread.
-static AsanThread *CreateAsanThread(StackTrace *stack, u32 parent_tid,
-                                    uptr user_id, bool detached,
-                                    uptr stack_bottom, uptr stack_size,
-                                    uptr tls_bottom, uptr tls_size) {
-  // In lieu of AsanThread::Create.
-  AsanThread *thread = (AsanThread *)MmapOrDie(sizeof(AsanThread), __func__);
-  AsanThreadContext::CreateThreadContextArgs args = {thread, stack};
-  asanThreadRegistry().CreateThread(user_id, detached, parent_tid, &args);
-
-  // On other systems, AsanThread::Init() is called from the new
-  // thread itself.  But on RTEMS we already know the stack address
-  // range beforehand, so we can do most of the setup right now.
-  const AsanThread::InitOptions options = {stack_bottom, stack_size,
-                                           tls_bottom, tls_size};
-  thread->Init(&options);
-  return thread;
-}
-
-// This gets the same arguments passed to Init by CreateAsanThread, above.
-// We're in the creator thread before the new thread is actually started, but
-// its stack and tls address range are already known.
-void AsanThread::SetThreadStackAndTls(const AsanThread::InitOptions *options) {
-  DCHECK_NE(GetCurrentThread(), this);
-  DCHECK_NE(GetCurrentThread(), nullptr);
-  CHECK_NE(options->stack_bottom, 0);
-  CHECK_NE(options->stack_size, 0);
-  stack_bottom_ = options->stack_bottom;
-  stack_top_ = options->stack_bottom + options->stack_size;
-  tls_begin_ = options->tls_bottom;
-  tls_end_ = options->tls_bottom + options->tls_size;
-}
-
-// Called by __asan::AsanInitInternal (asan_rtl.c).  Unlike other ports, the
-// main thread on RTEMS does not require special treatment; its AsanThread is
-// already created by the provided hooks.  This function simply looks up and
-// returns the created thread.
-AsanThread *CreateMainThread() {
-  return GetThreadContextByTidLocked(0)->thread;
-}
-
-// This is called before each thread creation is attempted.  So, in
-// its first call, the calling thread is the initial and sole thread.
-static void *BeforeThreadCreateHook(uptr user_id, bool detached,
-                                    uptr stack_bottom, uptr stack_size,
-                                    uptr tls_bottom, uptr tls_size) {
-  EnsureMainThreadIDIsCorrect();
-  // Strict init-order checking is thread-hostile.
-  if (flags()->strict_init_order) StopInitOrderChecking();
-
-  GET_STACK_TRACE_THREAD;
-  u32 parent_tid = GetCurrentTidOrInvalid();
-
-  return CreateAsanThread(&stack, parent_tid, user_id, detached,
-                          stack_bottom, stack_size, tls_bottom, tls_size);
-}
-
-// This is called after creating a new thread (in the creating thread),
-// with the pointer returned by BeforeThreadCreateHook (above).
-static void ThreadCreateHook(void *hook, bool aborted) {
-  AsanThread *thread = static_cast<AsanThread *>(hook);
-  if (!aborted) {
-    // The thread was created successfully.
-    // ThreadStartHook is already running in the new thread.
-  } else {
-    // The thread wasn't created after all.
-    // Clean up everything we set up in BeforeThreadCreateHook.
-    asanThreadRegistry().FinishThread(thread->tid());
-    UnmapOrDie(thread, sizeof(AsanThread));
-  }
-}
-
-// This is called (1) in the newly-created thread before it runs anything else,
-// with the pointer returned by BeforeThreadCreateHook (above).  (2) before a
-// thread restart.
-static void ThreadStartHook(void *hook, uptr os_id) {
-  if (!hook)
-    return;
-
-  AsanThread *thread = static_cast<AsanThread *>(hook);
-  SetCurrentThread(thread);
-
-  ThreadStatus status =
-      asanThreadRegistry().GetThreadLocked(thread->tid())->status;
-  DCHECK(status == ThreadStatusCreated || status == ThreadStatusRunning);
-  // Determine whether we are starting or restarting the thread.
-  if (status == ThreadStatusCreated) {
-    // In lieu of AsanThread::ThreadStart.
-    asanThreadRegistry().StartThread(thread->tid(), os_id, ThreadType::Regular,
-                                     nullptr);
-  } else {
-    // In a thread restart, a thread may resume execution at an
-    // arbitrary function entry point, with its stack and TLS state
-    // reset.  We unpoison the stack in that case.
-    PoisonShadow(thread->stack_bottom(), thread->stack_size(), 0);
-  }
-}
-
-// Each thread runs this just before it exits,
-// with the pointer returned by BeforeThreadCreateHook (above).
-// All per-thread destructors have already been called.
-static void ThreadExitHook(void *hook, uptr os_id) {
-  AsanThread *thread = static_cast<AsanThread *>(hook);
-  if (thread)
-    AsanThread::TSDDtor(thread->context());
-}
-
-static void HandleExit() {
-  // Disable ASan by setting it to uninitialized.  Also reset the
-  // shadow memory to avoid reporting errors after the run-time has
-  // been desroyed.
-  if (asan_inited) {
-    asan_inited = false;
-    ResetShadowMemory();
-  }
-}
-
-bool HandleDlopenInit() {
-  // Not supported on this platform.
-  static_assert(!SANITIZER_SUPPORTS_INIT_FOR_DLOPEN,
-                "Expected SANITIZER_SUPPORTS_INIT_FOR_DLOPEN to be false");
-  return false;
-}
-}  // namespace __asan
-
-// These are declared (in extern "C") by <some_path/sanitizer.h>.
-// The system runtime will call our definitions directly.
-
-extern "C" {
-void __sanitizer_early_init() {
-  __asan::EarlyInit();
-}
-
-void *__sanitizer_before_thread_create_hook(uptr thread, bool detached,
-                                            const char *name,
-                                            void *stack_base, size_t stack_size,
-                                            void *tls_base, size_t tls_size) {
-  return __asan::BeforeThreadCreateHook(
-      thread, detached,
-      reinterpret_cast<uptr>(stack_base), stack_size,
-      reinterpret_cast<uptr>(tls_base), tls_size);
-}
-
-void __sanitizer_thread_create_hook(void *handle, uptr thread, int status) {
-  __asan::ThreadCreateHook(handle, status != 0);
-}
-
-void __sanitizer_thread_start_hook(void *handle, uptr self) {
-  __asan::ThreadStartHook(handle, self);
-}
-
-void __sanitizer_thread_exit_hook(void *handle, uptr self) {
-  __asan::ThreadExitHook(handle, self);
-}
-
-void __sanitizer_exit() {
-  __asan::HandleExit();
-}
-}  // "C"
-
-#endif  // SANITIZER_RTEMS
diff --git a/libsanitizer/asan/asan_rtl.cpp b/libsanitizer/asan/asan_rtl.cpp
index e715d774228..e06a1113f4e 100644
--- a/libsanitizer/asan/asan_rtl.cpp
+++ b/libsanitizer/asan/asan_rtl.cpp
@@ -13,6 +13,7 @@
 
 #include "asan_activation.h"
 #include "asan_allocator.h"
+#include "asan_fake_stack.h"
 #include "asan_interceptors.h"
 #include "asan_interface_internal.h"
 #include "asan_internal.h"
@@ -23,11 +24,11 @@
 #include "asan_stats.h"
 #include "asan_suppressions.h"
 #include "asan_thread.h"
+#include "lsan/lsan_common.h"
 #include "sanitizer_common/sanitizer_atomic.h"
 #include "sanitizer_common/sanitizer_flags.h"
 #include "sanitizer_common/sanitizer_libc.h"
 #include "sanitizer_common/sanitizer_symbolizer.h"
-#include "lsan/lsan_common.h"
 #include "ubsan/ubsan_init.h"
 #include "ubsan/ubsan_platform.h"
 
@@ -137,24 +138,21 @@ ASAN_REPORT_ERROR_N(load, false)
 ASAN_REPORT_ERROR_N(store, true)
 
 #define ASAN_MEMORY_ACCESS_CALLBACK_BODY(type, is_write, size, exp_arg, fatal) \
-    if (SANITIZER_MYRIAD2 && !AddrIsInMem(addr) && !AddrIsInShadow(addr))      \
-      return;                                                                  \
-    uptr sp = MEM_TO_SHADOW(addr);                                             \
-    uptr s = size <= SHADOW_GRANULARITY ? *reinterpret_cast<u8 *>(sp)          \
-                                        : *reinterpret_cast<u16 *>(sp);        \
-    if (UNLIKELY(s)) {                                                         \
-      if (UNLIKELY(size >= SHADOW_GRANULARITY ||                               \
-                   ((s8)((addr & (SHADOW_GRANULARITY - 1)) + size - 1)) >=     \
-                       (s8)s)) {                                               \
-        if (__asan_test_only_reported_buggy_pointer) {                         \
-          *__asan_test_only_reported_buggy_pointer = addr;                     \
-        } else {                                                               \
-          GET_CALLER_PC_BP_SP;                                                 \
-          ReportGenericError(pc, bp, sp, addr, is_write, size, exp_arg,        \
-                              fatal);                                          \
-        }                                                                      \
+  uptr sp = MEM_TO_SHADOW(addr);                                               \
+  uptr s = size <= SHADOW_GRANULARITY ? *reinterpret_cast<u8 *>(sp)            \
+                                      : *reinterpret_cast<u16 *>(sp);          \
+  if (UNLIKELY(s)) {                                                           \
+    if (UNLIKELY(size >= SHADOW_GRANULARITY ||                                 \
+                 ((s8)((addr & (SHADOW_GRANULARITY - 1)) + size - 1)) >=       \
+                     (s8)s)) {                                                 \
+      if (__asan_test_only_reported_buggy_pointer) {                           \
+        *__asan_test_only_reported_buggy_pointer = addr;                       \
+      } else {                                                                 \
+        GET_CALLER_PC_BP_SP;                                                   \
+        ReportGenericError(pc, bp, sp, addr, is_write, size, exp_arg, fatal);  \
       }                                                                        \
-    }
+    }                                                                          \
+  }
 
 #define ASAN_MEMORY_ACCESS_CALLBACK(type, is_write, size)                      \
   extern "C" NOINLINE INTERFACE_ATTRIBUTE                                      \
@@ -305,7 +303,6 @@ static void asan_atexit() {
 }
 
 static void InitializeHighMemEnd() {
-#if !SANITIZER_MYRIAD2
 #if !ASAN_FIXED_MAPPING
   kHighMemEnd = GetMaxUserVirtualAddress();
   // Increase kHighMemEnd to make sure it's properly
@@ -313,7 +310,6 @@ static void InitializeHighMemEnd() {
   kHighMemEnd |= (GetMmapGranularity() << SHADOW_SCALE) - 1;
 #endif  // !ASAN_FIXED_MAPPING
   CHECK_EQ((kHighMemBeg % GetMmapGranularity()), 0);
-#endif  // !SANITIZER_MYRIAD2
 }
 
 void PrintAddressSpaceLayout() {
@@ -569,9 +565,6 @@ static void UnpoisonDefaultStack() {
     const uptr page_size = GetPageSizeCached();
     top = curr_thread->stack_top();
     bottom = ((uptr)&local_stack - page_size) & ~(page_size - 1);
-  } else if (SANITIZER_RTEMS) {
-    // Give up On RTEMS.
-    return;
   } else {
     CHECK(!SANITIZER_FUCHSIA);
     // If we haven't seen this thread, try asking the OS for stack bounds.
@@ -586,8 +579,12 @@ static void UnpoisonDefaultStack() {
 
 static void UnpoisonFakeStack() {
   AsanThread *curr_thread = GetCurrentThread();
-  if (curr_thread && curr_thread->has_fake_stack())
-    curr_thread->fake_stack()->HandleNoReturn();
+  if (!curr_thread)
+    return;
+  FakeStack *stack = curr_thread->get_fake_stack();
+  if (!stack)
+    return;
+  stack->HandleNoReturn();
 }
 
 }  // namespace __asan
diff --git a/libsanitizer/asan/asan_shadow_setup.cpp b/libsanitizer/asan/asan_shadow_setup.cpp
index 2ead4425add..6e6260d3413 100644
--- a/libsanitizer/asan/asan_shadow_setup.cpp
+++ b/libsanitizer/asan/asan_shadow_setup.cpp
@@ -13,12 +13,11 @@
 
 #include "sanitizer_common/sanitizer_platform.h"
 
-// asan_fuchsia.cpp and asan_rtems.cpp have their own
-// InitializeShadowMemory implementation.
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+// asan_fuchsia.cpp has their own InitializeShadowMemory implementation.
+#if !SANITIZER_FUCHSIA
 
-#include "asan_internal.h"
-#include "asan_mapping.h"
+#  include "asan_internal.h"
+#  include "asan_mapping.h"
 
 namespace __asan {
 
@@ -123,4 +122,4 @@ void InitializeShadowMemory() {
 
 }  // namespace __asan
 
-#endif  // !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#endif  // !SANITIZER_FUCHSIA
diff --git a/libsanitizer/asan/asan_stack.cpp b/libsanitizer/asan/asan_stack.cpp
index b7f4e6aeeab..048295d6928 100644
--- a/libsanitizer/asan/asan_stack.cpp
+++ b/libsanitizer/asan/asan_stack.cpp
@@ -74,7 +74,8 @@ void __sanitizer::BufferedStackTrace::UnwindImpl(
   if (SANITIZER_MIPS && t &&
       !IsValidFrame(bp, t->stack_top(), t->stack_bottom()))
     return;
-  Unwind(max_depth, pc, bp, context, 0, 0, false);
+  Unwind(max_depth, pc, bp, context, t ? t->stack_top() : 0,
+         t ? t->stack_bottom() : 0, false);
 }
 
 // ------------------ Interface -------------- {{{1
diff --git a/libsanitizer/asan/asan_thread.cpp b/libsanitizer/asan/asan_thread.cpp
index 9c3c86f5735..35d4467e7b5 100644
--- a/libsanitizer/asan/asan_thread.cpp
+++ b/libsanitizer/asan/asan_thread.cpp
@@ -60,8 +60,8 @@ ThreadRegistry &asanThreadRegistry() {
     // in TSD and can't reliably tell when no more TSD destructors will
     // be called. It would be wrong to reuse AsanThreadContext for another
     // thread before all TSD destructors will be called for it.
-    asan_thread_registry = new(thread_registry_placeholder) ThreadRegistry(
-        GetAsanThreadContext, kMaxNumberOfThreads, kMaxNumberOfThreads);
+    asan_thread_registry =
+        new (thread_registry_placeholder) ThreadRegistry(GetAsanThreadContext);
     initialized = true;
   }
   return *asan_thread_registry;
@@ -257,10 +257,9 @@ void AsanThread::Init(const InitOptions *options) {
           &local);
 }
 
-// Fuchsia and RTEMS don't use ThreadStart.
-// asan_fuchsia.c/asan_rtems.c define CreateMainThread and
-// SetThreadStackAndTls.
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+// Fuchsia doesn't use ThreadStart.
+// asan_fuchsia.c definies CreateMainThread and SetThreadStackAndTls.
+#if !SANITIZER_FUCHSIA
 
 thread_return_t AsanThread::ThreadStart(tid_t os_id) {
   Init();
@@ -317,7 +316,7 @@ void AsanThread::SetThreadStackAndTls(const InitOptions *options) {
   }
 }
 
-#endif  // !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#endif  // !SANITIZER_FUCHSIA
 
 void AsanThread::ClearShadowForThreadStackAndTLS() {
   if (stack_top_ != stack_bottom_)
@@ -339,8 +338,8 @@ bool AsanThread::GetStackFrameAccessByAddr(uptr addr,
   uptr bottom = 0;
   if (AddrIsInStack(addr)) {
     bottom = stack_bottom();
-  } else if (has_fake_stack()) {
-    bottom = fake_stack()->AddrIsInFakeStack(addr);
+  } else if (FakeStack *fake_stack = get_fake_stack()) {
+    bottom = fake_stack->AddrIsInFakeStack(addr);
     CHECK(bottom);
     access->offset = addr - bottom;
     access->frame_pc = ((uptr*)bottom)[2];
@@ -380,8 +379,8 @@ uptr AsanThread::GetStackVariableShadowStart(uptr addr) {
   uptr bottom = 0;
   if (AddrIsInStack(addr)) {
     bottom = stack_bottom();
-  } else if (has_fake_stack()) {
-    bottom = fake_stack()->AddrIsInFakeStack(addr);
+  } else if (FakeStack *fake_stack = get_fake_stack()) {
+    bottom = fake_stack->AddrIsInFakeStack(addr);
     if (bottom == 0) {
       return 0;
     }
@@ -409,19 +408,19 @@ bool AsanThread::AddrIsInStack(uptr addr) {
 
 static bool ThreadStackContainsAddress(ThreadContextBase *tctx_base,
                                        void *addr) {
-  AsanThreadContext *tctx = static_cast<AsanThreadContext*>(tctx_base);
+  AsanThreadContext *tctx = static_cast<AsanThreadContext *>(tctx_base);
   AsanThread *t = tctx->thread;
-  if (!t) return false;
-  if (t->AddrIsInStack((uptr)addr)) return true;
-  if (t->has_fake_stack() && t->fake_stack()->AddrIsInFakeStack((uptr)addr))
+  if (!t)
+    return false;
+  if (t->AddrIsInStack((uptr)addr))
     return true;
-  return false;
+  FakeStack *fake_stack = t->get_fake_stack();
+  if (!fake_stack)
+    return false;
+  return fake_stack->AddrIsInFakeStack((uptr)addr);
 }
 
 AsanThread *GetCurrentThread() {
-  if (SANITIZER_RTEMS && !asan_inited)
-    return nullptr;
-
   AsanThreadContext *context =
       reinterpret_cast<AsanThreadContext *>(AsanTSDGet());
   if (!context) {
@@ -503,8 +502,12 @@ void GetAllThreadAllocatorCachesLocked(InternalMmapVector<uptr> *caches) {}
 void ForEachExtraStackRange(tid_t os_id, RangeIteratorCallback callback,
                             void *arg) {
   __asan::AsanThread *t = __asan::GetAsanThreadByOsIDLocked(os_id);
-  if (t && t->has_fake_stack())
-    t->fake_stack()->ForEachFakeFrame(callback, arg);
+  if (!t)
+    return;
+  __asan::FakeStack *fake_stack = t->get_fake_stack();
+  if (!fake_stack)
+    return;
+  fake_stack->ForEachFakeFrame(callback, arg);
 }
 
 void LockThreadRegistry() {
diff --git a/libsanitizer/asan/asan_thread.h b/libsanitizer/asan/asan_thread.h
index 200069ce0dd..801a3960ec6 100644
--- a/libsanitizer/asan/asan_thread.h
+++ b/libsanitizer/asan/asan_thread.h
@@ -28,8 +28,6 @@ struct DTLS;
 
 namespace __asan {
 
-const u32 kMaxNumberOfThreads = (1 << 22);  // 4M
-
 class AsanThread;
 
 // These objects are created for every thread and are never deleted,
@@ -104,17 +102,18 @@ class AsanThread {
   void FinishSwitchFiber(FakeStack *fake_stack_save, uptr *bottom_old,
                          uptr *size_old);
 
-  bool has_fake_stack() {
-    return !atomic_load(&stack_switching_, memory_order_relaxed) &&
-           (reinterpret_cast<uptr>(fake_stack_) > 1);
-  }
-
-  FakeStack *fake_stack() {
-    if (!__asan_option_detect_stack_use_after_return)
-      return nullptr;
+  FakeStack *get_fake_stack() {
     if (atomic_load(&stack_switching_, memory_order_relaxed))
       return nullptr;
-    if (!has_fake_stack())
+    if (reinterpret_cast<uptr>(fake_stack_) <= 1)
+      return nullptr;
+    return fake_stack_;
+  }
+
+  FakeStack *get_or_create_fake_stack() {
+    if (atomic_load(&stack_switching_, memory_order_relaxed))
+      return nullptr;
+    if (reinterpret_cast<uptr>(fake_stack_) <= 1)
       return AsyncSignalSafeLazyInitFakeStack();
     return fake_stack_;
   }
diff --git a/libsanitizer/hwasan/Makefile.am b/libsanitizer/hwasan/Makefile.am
index 2226f68ab29..5e3a0f1b0a1 100644
--- a/libsanitizer/hwasan/Makefile.am
+++ b/libsanitizer/hwasan/Makefile.am
@@ -13,11 +13,13 @@ ACLOCAL_AMFLAGS = -I $(top_srcdir) -I $(top_srcdir)/config
 toolexeclib_LTLIBRARIES = libhwasan.la
 
 hwasan_files = \
+	hwasan_allocation_functions.cpp \
 	hwasan_allocator.cpp \
 	hwasan.cpp \
 	hwasan_dynamic_shadow.cpp \
 	hwasan_exceptions.cpp \
 	hwasan_flags.inc \
+	hwasan_fuchsia.cpp \
 	hwasan_globals.cpp \
 	hwasan_interceptors.cpp \
 	hwasan_interceptors_vfork.S \
diff --git a/libsanitizer/hwasan/Makefile.in b/libsanitizer/hwasan/Makefile.in
index 542af8f193e..22c5266a120 100644
--- a/libsanitizer/hwasan/Makefile.in
+++ b/libsanitizer/hwasan/Makefile.in
@@ -146,8 +146,9 @@ am__DEPENDENCIES_1 =
 libhwasan_la_DEPENDENCIES =  \
 	$(top_builddir)/sanitizer_common/libsanitizer_common.la \
 	$(am__append_1) $(am__append_2) $(am__DEPENDENCIES_1)
-am__objects_1 = hwasan_allocator.lo hwasan.lo hwasan_dynamic_shadow.lo \
-	hwasan_exceptions.lo hwasan_globals.lo hwasan_interceptors.lo \
+am__objects_1 = hwasan_allocation_functions.lo hwasan_allocator.lo \
+	hwasan.lo hwasan_dynamic_shadow.lo hwasan_exceptions.lo \
+	hwasan_fuchsia.lo hwasan_globals.lo hwasan_interceptors.lo \
 	hwasan_interceptors_vfork.lo hwasan_linux.lo \
 	hwasan_memintrinsics.lo hwasan_new_delete.lo \
 	hwasan_poisoning.lo hwasan_report.lo hwasan_setjmp.lo \
@@ -411,11 +412,13 @@ AM_CXXFLAGS = -Wall -W -Wno-unused-parameter -Wwrite-strings -pedantic \
 ACLOCAL_AMFLAGS = -I $(top_srcdir) -I $(top_srcdir)/config
 toolexeclib_LTLIBRARIES = libhwasan.la
 hwasan_files = \
+	hwasan_allocation_functions.cpp \
 	hwasan_allocator.cpp \
 	hwasan.cpp \
 	hwasan_dynamic_shadow.cpp \
 	hwasan_exceptions.cpp \
 	hwasan_flags.inc \
+	hwasan_fuchsia.cpp \
 	hwasan_globals.cpp \
 	hwasan_interceptors.cpp \
 	hwasan_interceptors_vfork.S \
@@ -554,9 +557,11 @@ distclean-compile:
 	-rm -f *.tab.c
 
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_allocation_functions.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_allocator.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_dynamic_shadow.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_exceptions.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_fuchsia.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_globals.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_interceptors.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/hwasan_interceptors_vfork.Plo@am__quote@
diff --git a/libsanitizer/hwasan/hwasan.cpp b/libsanitizer/hwasan/hwasan.cpp
index 8d6c25261b8..cbe0dee66dc 100644
--- a/libsanitizer/hwasan/hwasan.cpp
+++ b/libsanitizer/hwasan/hwasan.cpp
@@ -50,6 +50,11 @@ bool hwasan_init_is_running;
 
 int hwasan_report_count = 0;
 
+uptr kLowShadowStart;
+uptr kLowShadowEnd;
+uptr kHighShadowStart;
+uptr kHighShadowEnd;
+
 void Flags::SetDefaults() {
 #define HWASAN_FLAG(Type, Name, DefaultValue, Description) Name = DefaultValue;
 #include "hwasan_flags.inc"
@@ -177,6 +182,65 @@ void UpdateMemoryUsage() {
 void UpdateMemoryUsage() {}
 #endif
 
+void HwasanAtExit() {
+  if (common_flags()->print_module_map)
+    DumpProcessMap();
+  if (flags()->print_stats && (flags()->atexit || hwasan_report_count > 0))
+    ReportStats();
+  if (hwasan_report_count > 0) {
+    // ReportAtExitStatistics();
+    if (common_flags()->exitcode)
+      internal__exit(common_flags()->exitcode);
+  }
+}
+
+void HandleTagMismatch(AccessInfo ai, uptr pc, uptr frame, void *uc,
+                       uptr *registers_frame) {
+  InternalMmapVector<BufferedStackTrace> stack_buffer(1);
+  BufferedStackTrace *stack = stack_buffer.data();
+  stack->Reset();
+  stack->Unwind(pc, frame, uc, common_flags()->fast_unwind_on_fatal);
+
+  // The second stack frame contains the failure __hwasan_check function, as
+  // we have a stack frame for the registers saved in __hwasan_tag_mismatch that
+  // we wish to ignore. This (currently) only occurs on AArch64, as x64
+  // implementations use SIGTRAP to implement the failure, and thus do not go
+  // through the stack saver.
+  if (registers_frame && stack->trace && stack->size > 0) {
+    stack->trace++;
+    stack->size--;
+  }
+
+  bool fatal = flags()->halt_on_error || !ai.recover;
+  ReportTagMismatch(stack, ai.addr, ai.size, ai.is_store, fatal,
+                    registers_frame);
+}
+
+void HwasanTagMismatch(uptr addr, uptr access_info, uptr *registers_frame,
+                       size_t outsize) {
+  __hwasan::AccessInfo ai;
+  ai.is_store = access_info & 0x10;
+  ai.is_load = !ai.is_store;
+  ai.recover = access_info & 0x20;
+  ai.addr = addr;
+  if ((access_info & 0xf) == 0xf)
+    ai.size = outsize;
+  else
+    ai.size = 1 << (access_info & 0xf);
+
+  HandleTagMismatch(ai, (uptr)__builtin_return_address(0),
+                    (uptr)__builtin_frame_address(0), nullptr, registers_frame);
+  __builtin_unreachable();
+}
+
+Thread *GetCurrentThread() {
+  uptr *ThreadLongPtr = GetCurrentThreadLongPtr();
+  if (UNLIKELY(*ThreadLongPtr == 0))
+    return nullptr;
+  auto *R = (StackAllocationsRingBuffer *)ThreadLongPtr;
+  return hwasanThreadList().GetThreadByBufferAddress((uptr)R->Next());
+}
+
 } // namespace __hwasan
 
 using namespace __hwasan;
@@ -216,7 +280,7 @@ static void InitLoadedGlobals() {
 static void InitInstrumentation() {
   if (hwasan_instrumentation_inited) return;
 
-  InitPrctl();
+  InitializeOsSupport();
 
   if (!InitShadow()) {
     Printf("FATAL: HWAddressSanitizer cannot mmap the shadow memory.\n");
@@ -225,7 +289,6 @@ static void InitInstrumentation() {
   }
 
   InitThreads();
-  hwasanThreadList().CreateCurrentThread();
 
   hwasan_instrumentation_inited = 1;
 }
@@ -495,7 +558,7 @@ void __hwasan_print_memory_usage() {
   Printf("%s\n", s.data());
 }
 
-static const u8 kFallbackTag = 0xBB;
+static const u8 kFallbackTag = 0xBB & kTagMask;
 
 u8 __hwasan_generate_tag() {
   Thread *t = GetCurrentThread();
@@ -516,4 +579,12 @@ void __sanitizer_print_stack_trace() {
   GET_FATAL_STACK_TRACE_PC_BP(StackTrace::GetCurrentPc(), GET_CURRENT_FRAME());
   stack.Print();
 }
+
+// Entry point for interoperability between __hwasan_tag_mismatch (ASM) and the
+// rest of the mismatch handling code (C++).
+void __hwasan_tag_mismatch4(uptr addr, uptr access_info, uptr *registers_frame,
+                            size_t outsize) {
+  __hwasan::HwasanTagMismatch(addr, access_info, registers_frame, outsize);
+}
+
 } // extern "C"
diff --git a/libsanitizer/hwasan/hwasan.h b/libsanitizer/hwasan/hwasan.h
index 8515df559f3..7338b696ad3 100644
--- a/libsanitizer/hwasan/hwasan.h
+++ b/libsanitizer/hwasan/hwasan.h
@@ -36,7 +36,10 @@
 
 typedef u8 tag_t;
 
-#if defined(__x86_64__)
+#if defined(HWASAN_ALIASING_MODE)
+#  if !defined(__x86_64__)
+#    error Aliasing mode is only supported on x86_64
+#  endif
 // Tags are done in middle bits using userspace aliasing.
 constexpr unsigned kAddressTagShift = 39;
 constexpr unsigned kTagBits = 3;
@@ -49,12 +52,16 @@ constexpr unsigned kTagBits = 3;
 // simpler/faster shadow calculation.
 constexpr unsigned kTaggableRegionCheckShift =
     __sanitizer::Max(kAddressTagShift + kTagBits + 1U, 44U);
+#elif defined(__x86_64__)
+// Tags are done in upper bits using Intel LAM.
+constexpr unsigned kAddressTagShift = 57;
+constexpr unsigned kTagBits = 6;
 #else
 // TBI (Top Byte Ignore) feature of AArch64: bits [63:56] are ignored in address
 // translation and can be used to store a tag.
 constexpr unsigned kAddressTagShift = 56;
 constexpr unsigned kTagBits = 8;
-#endif  // defined(__x86_64__)
+#endif  // defined(HWASAN_ALIASING_MODE)
 
 // Mask for extracting tag bits from the lower 8 bits.
 constexpr uptr kTagMask = (1UL << kTagBits) - 1;
@@ -95,7 +102,7 @@ extern bool hwasan_init_is_running;
 extern int hwasan_report_count;
 
 bool InitShadow();
-void InitPrctl();
+void InitializeOsSupport();
 void InitThreads();
 void InitializeInterceptors();
 
@@ -129,6 +136,7 @@ void InstallAtExitHandler();
 
 void HwasanTSDInit();
 void HwasanTSDThreadInit();
+void HwasanAtExit();
 
 void HwasanOnDeadlySignal(int signo, void *info, void *context);
 
@@ -138,6 +146,26 @@ void AppendToErrorMessageBuffer(const char *buffer);
 
 void AndroidTestTlsSlot();
 
+// This is a compiler-generated struct that can be shared between hwasan
+// implementations.
+struct AccessInfo {
+  uptr addr;
+  uptr size;
+  bool is_store;
+  bool is_load;
+  bool recover;
+};
+
+// Given access info and frame information, unwind the stack and report the tag
+// mismatch.
+void HandleTagMismatch(AccessInfo ai, uptr pc, uptr frame, void *uc,
+                       uptr *registers_frame = nullptr);
+
+// This dispatches to HandleTagMismatch but sets up the AccessInfo, program
+// counter, and frame pointer.
+void HwasanTagMismatch(uptr addr, uptr access_info, uptr *registers_frame,
+                       size_t outsize);
+
 }  // namespace __hwasan
 
 #define HWASAN_MALLOC_HOOK(ptr, size)       \
@@ -175,4 +203,12 @@ typedef struct __hw_jmp_buf_struct __hw_jmp_buf[1];
 typedef struct __hw_jmp_buf_struct __hw_sigjmp_buf[1];
 #endif // HWASAN_WITH_INTERCEPTORS && __aarch64__
 
+#define ENSURE_HWASAN_INITED()      \
+  do {                              \
+    CHECK(!hwasan_init_is_running); \
+    if (!hwasan_inited) {           \
+      __hwasan_init();              \
+    }                               \
+  } while (0)
+
 #endif  // HWASAN_H
diff --git a/libsanitizer/hwasan/hwasan_allocation_functions.cpp b/libsanitizer/hwasan/hwasan_allocation_functions.cpp
new file mode 100644
index 00000000000..6c2a6077866
--- /dev/null
+++ b/libsanitizer/hwasan/hwasan_allocation_functions.cpp
@@ -0,0 +1,172 @@
+//===-- hwasan_allocation_functions.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 HWAddressSanitizer.
+//
+// Definitions for __sanitizer allocation functions.
+//
+//===----------------------------------------------------------------------===//
+
+#include "hwasan.h"
+#include "interception/interception.h"
+#include "sanitizer_common/sanitizer_allocator_interface.h"
+#include "sanitizer_common/sanitizer_tls_get_addr.h"
+
+using namespace __hwasan;
+
+static uptr allocated_for_dlsym;
+static const uptr kDlsymAllocPoolSize = 1024;
+static uptr alloc_memory_for_dlsym[kDlsymAllocPoolSize];
+
+static bool IsInDlsymAllocPool(const void *ptr) {
+  uptr off = (uptr)ptr - (uptr)alloc_memory_for_dlsym;
+  return off < sizeof(alloc_memory_for_dlsym);
+}
+
+static void *AllocateFromLocalPool(uptr size_in_bytes) {
+  uptr size_in_words = RoundUpTo(size_in_bytes, kWordSize) / kWordSize;
+  void *mem = (void *)&alloc_memory_for_dlsym[allocated_for_dlsym];
+  allocated_for_dlsym += size_in_words;
+  CHECK_LT(allocated_for_dlsym, kDlsymAllocPoolSize);
+  return mem;
+}
+
+int __sanitizer_posix_memalign(void **memptr, uptr alignment, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  CHECK_NE(memptr, 0);
+  int res = hwasan_posix_memalign(memptr, alignment, size, &stack);
+  return res;
+}
+
+void *__sanitizer_memalign(uptr alignment, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  return hwasan_memalign(alignment, size, &stack);
+}
+
+void *__sanitizer_aligned_alloc(uptr alignment, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  return hwasan_aligned_alloc(alignment, size, &stack);
+}
+
+void *__sanitizer___libc_memalign(uptr alignment, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  void *ptr = hwasan_memalign(alignment, size, &stack);
+  if (ptr)
+    DTLS_on_libc_memalign(ptr, size);
+  return ptr;
+}
+
+void *__sanitizer_valloc(uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  return hwasan_valloc(size, &stack);
+}
+
+void *__sanitizer_pvalloc(uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  return hwasan_pvalloc(size, &stack);
+}
+
+void __sanitizer_free(void *ptr) {
+  GET_MALLOC_STACK_TRACE;
+  if (!ptr || UNLIKELY(IsInDlsymAllocPool(ptr)))
+    return;
+  hwasan_free(ptr, &stack);
+}
+
+void __sanitizer_cfree(void *ptr) {
+  GET_MALLOC_STACK_TRACE;
+  if (!ptr || UNLIKELY(IsInDlsymAllocPool(ptr)))
+    return;
+  hwasan_free(ptr, &stack);
+}
+
+uptr __sanitizer_malloc_usable_size(const void *ptr) {
+  return __sanitizer_get_allocated_size(ptr);
+}
+
+struct __sanitizer_struct_mallinfo __sanitizer_mallinfo() {
+  __sanitizer_struct_mallinfo sret;
+  internal_memset(&sret, 0, sizeof(sret));
+  return sret;
+}
+
+int __sanitizer_mallopt(int cmd, int value) { return 0; }
+
+void __sanitizer_malloc_stats(void) {
+  // FIXME: implement, but don't call REAL(malloc_stats)!
+}
+
+void *__sanitizer_calloc(uptr nmemb, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  if (UNLIKELY(!hwasan_inited))
+    // Hack: dlsym calls calloc before REAL(calloc) is retrieved from dlsym.
+    return AllocateFromLocalPool(nmemb * size);
+  return hwasan_calloc(nmemb, size, &stack);
+}
+
+void *__sanitizer_realloc(void *ptr, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  if (UNLIKELY(IsInDlsymAllocPool(ptr))) {
+    uptr offset = (uptr)ptr - (uptr)alloc_memory_for_dlsym;
+    uptr copy_size = Min(size, kDlsymAllocPoolSize - offset);
+    void *new_ptr;
+    if (UNLIKELY(!hwasan_inited)) {
+      new_ptr = AllocateFromLocalPool(copy_size);
+    } else {
+      copy_size = size;
+      new_ptr = hwasan_malloc(copy_size, &stack);
+    }
+    internal_memcpy(new_ptr, ptr, copy_size);
+    return new_ptr;
+  }
+  return hwasan_realloc(ptr, size, &stack);
+}
+
+void *__sanitizer_reallocarray(void *ptr, uptr nmemb, uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  return hwasan_reallocarray(ptr, nmemb, size, &stack);
+}
+
+void *__sanitizer_malloc(uptr size) {
+  GET_MALLOC_STACK_TRACE;
+  if (UNLIKELY(!hwasan_init_is_running))
+    ENSURE_HWASAN_INITED();
+  if (UNLIKELY(!hwasan_inited))
+    // Hack: dlsym calls malloc before REAL(malloc) is retrieved from dlsym.
+    return AllocateFromLocalPool(size);
+  return hwasan_malloc(size, &stack);
+}
+
+#if HWASAN_WITH_INTERCEPTORS
+#  define INTERCEPTOR_ALIAS(RET, FN, ARGS...)                                 \
+    extern "C" SANITIZER_INTERFACE_ATTRIBUTE RET WRAP(FN)(ARGS)               \
+        ALIAS("__sanitizer_" #FN);                                            \
+    extern "C" SANITIZER_INTERFACE_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE RET FN( \
+        ARGS) ALIAS("__sanitizer_" #FN)
+
+INTERCEPTOR_ALIAS(int, posix_memalign, void **memptr, SIZE_T alignment,
+                  SIZE_T size);
+INTERCEPTOR_ALIAS(void *, aligned_alloc, SIZE_T alignment, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, __libc_memalign, SIZE_T alignment, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, valloc, SIZE_T size);
+INTERCEPTOR_ALIAS(void, free, void *ptr);
+INTERCEPTOR_ALIAS(uptr, malloc_usable_size, const void *ptr);
+INTERCEPTOR_ALIAS(void *, calloc, SIZE_T nmemb, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, realloc, void *ptr, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, reallocarray, void *ptr, SIZE_T nmemb, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, malloc, SIZE_T size);
+
+#  if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
+INTERCEPTOR_ALIAS(void *, memalign, SIZE_T alignment, SIZE_T size);
+INTERCEPTOR_ALIAS(void *, pvalloc, SIZE_T size);
+INTERCEPTOR_ALIAS(void, cfree, void *ptr);
+INTERCEPTOR_ALIAS(__sanitizer_struct_mallinfo, mallinfo);
+INTERCEPTOR_ALIAS(int, mallopt, int cmd, int value);
+INTERCEPTOR_ALIAS(void, malloc_stats, void);
+#  endif
+#endif  // #if HWASAN_WITH_INTERCEPTORS
diff --git a/libsanitizer/hwasan/hwasan_allocator.cpp b/libsanitizer/hwasan/hwasan_allocator.cpp
index a6fc794082a..ef6d4d6c767 100644
--- a/libsanitizer/hwasan/hwasan_allocator.cpp
+++ b/libsanitizer/hwasan/hwasan_allocator.cpp
@@ -80,12 +80,29 @@ void GetAllocatorStats(AllocatorStatCounters s) {
   allocator.GetStats(s);
 }
 
+uptr GetAliasRegionStart() {
+#if defined(HWASAN_ALIASING_MODE)
+  constexpr uptr kAliasRegionOffset = 1ULL << (kTaggableRegionCheckShift - 1);
+  uptr AliasRegionStart =
+      __hwasan_shadow_memory_dynamic_address + kAliasRegionOffset;
+
+  CHECK_EQ(AliasRegionStart >> kTaggableRegionCheckShift,
+           __hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
+  CHECK_EQ(
+      (AliasRegionStart + kAliasRegionOffset - 1) >> kTaggableRegionCheckShift,
+      __hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
+  return AliasRegionStart;
+#else
+  return 0;
+#endif
+}
+
 void HwasanAllocatorInit() {
   atomic_store_relaxed(&hwasan_allocator_tagging_enabled,
                        !flags()->disable_allocator_tagging);
   SetAllocatorMayReturnNull(common_flags()->allocator_may_return_null);
   allocator.Init(common_flags()->allocator_release_to_os_interval_ms,
-                 kAliasRegionStart);
+                 GetAliasRegionStart());
   for (uptr i = 0; i < sizeof(tail_magic); i++)
     tail_magic[i] = GetCurrentThread()->GenerateRandomTag();
 }
@@ -196,6 +213,7 @@ static void HwasanDeallocate(StackTrace *stack, void *tagged_ptr) {
                            : tagged_ptr;
   void *aligned_ptr = reinterpret_cast<void *>(
       RoundDownTo(reinterpret_cast<uptr>(untagged_ptr), kShadowAlignment));
+  tag_t pointer_tag = GetTagFromPointer(reinterpret_cast<uptr>(tagged_ptr));
   Metadata *meta =
       reinterpret_cast<Metadata *>(allocator.GetMetaData(aligned_ptr));
   uptr orig_size = meta->get_requested_size();
@@ -229,7 +247,20 @@ static void HwasanDeallocate(StackTrace *stack, void *tagged_ptr) {
       flags()->tag_in_free && malloc_bisect(stack, 0) &&
       atomic_load_relaxed(&hwasan_allocator_tagging_enabled)) {
     // Always store full 8-bit tags on free to maximize UAF detection.
-    tag_t tag = t ? t->GenerateRandomTag(/*num_bits=*/8) : kFallbackFreeTag;
+    tag_t tag;
+    if (t) {
+      // Make sure we are not using a short granule tag as a poison tag. This
+      // would make us attempt to read the memory on a UaF.
+      // The tag can be zero if tagging is disabled on this thread.
+      do {
+        tag = t->GenerateRandomTag(/*num_bits=*/8);
+      } while (
+          UNLIKELY((tag < kShadowAlignment || tag == pointer_tag) && tag != 0));
+    } else {
+      static_assert(kFallbackFreeTag >= kShadowAlignment,
+                    "fallback tag must not be a short granule tag.");
+      tag = kFallbackFreeTag;
+    }
     TagMemoryAligned(reinterpret_cast<uptr>(aligned_ptr), TaggedSize(orig_size),
                      tag);
   }
diff --git a/libsanitizer/hwasan/hwasan_allocator.h b/libsanitizer/hwasan/hwasan_allocator.h
index 03bbcff3f0f..35c3d6b4bf4 100644
--- a/libsanitizer/hwasan/hwasan_allocator.h
+++ b/libsanitizer/hwasan/hwasan_allocator.h
@@ -15,6 +15,7 @@
 
 #include "hwasan.h"
 #include "hwasan_interface_internal.h"
+#include "hwasan_mapping.h"
 #include "hwasan_poisoning.h"
 #include "sanitizer_common/sanitizer_allocator.h"
 #include "sanitizer_common/sanitizer_allocator_checks.h"
@@ -58,7 +59,7 @@ static const uptr kMaxAllowedMallocSize = 1UL << 40;  // 1T
 struct AP64 {
   static const uptr kSpaceBeg = ~0ULL;
 
-#if defined(__x86_64__)
+#if defined(HWASAN_ALIASING_MODE)
   static const uptr kSpaceSize = 1ULL << kAddressTagShift;
 #else
   static const uptr kSpaceSize = 0x2000000000ULL;
@@ -110,11 +111,11 @@ typedef RingBuffer<HeapAllocationRecord> HeapAllocationsRingBuffer;
 void GetAllocatorStats(AllocatorStatCounters s);
 
 inline bool InTaggableRegion(uptr addr) {
-#if defined(__x86_64__)
+#if defined(HWASAN_ALIASING_MODE)
   // Aliases are mapped next to shadow so that the upper bits match the shadow
   // base.
   return (addr >> kTaggableRegionCheckShift) ==
-         (__hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
+         (GetShadowOffset() >> kTaggableRegionCheckShift);
 #endif
   return true;
 }
diff --git a/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp b/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
index f53276e330d..bde22dfa4bc 100644
--- a/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
+++ b/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
@@ -119,12 +119,12 @@ namespace __hwasan {
 void InitShadowGOT() {}
 
 uptr FindDynamicShadowStart(uptr shadow_size_bytes) {
-#if defined(__x86_64__)
+#  if defined(HWASAN_ALIASING_MODE)
   constexpr uptr kAliasSize = 1ULL << kAddressTagShift;
   constexpr uptr kNumAliases = 1ULL << kTagBits;
   return MapDynamicShadowAndAliases(shadow_size_bytes, kAliasSize, kNumAliases,
                                     RingBufferSize());
-#endif
+#  endif
   return MapDynamicShadow(shadow_size_bytes, kShadowScale, kShadowBaseAlignment,
                           kHighMemEnd);
 }
diff --git a/libsanitizer/hwasan/hwasan_fuchsia.cpp b/libsanitizer/hwasan/hwasan_fuchsia.cpp
new file mode 100644
index 00000000000..e61f6ada72f
--- /dev/null
+++ b/libsanitizer/hwasan/hwasan_fuchsia.cpp
@@ -0,0 +1,192 @@
+//===-- hwasan_fuchsia.cpp --------------------------------------*- 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
+//
+//===----------------------------------------------------------------------===//
+///
+/// \file
+/// This file is a part of HWAddressSanitizer and contains Fuchsia-specific
+/// code.
+///
+//===----------------------------------------------------------------------===//
+
+#include "sanitizer_common/sanitizer_fuchsia.h"
+#if SANITIZER_FUCHSIA
+
+#include "hwasan.h"
+#include "hwasan_interface_internal.h"
+#include "hwasan_report.h"
+#include "hwasan_thread.h"
+#include "hwasan_thread_list.h"
+
+// This TLS variable contains the location of the stack ring buffer and can be
+// used to always find the hwasan thread object associated with the current
+// running thread.
+[[gnu::tls_model("initial-exec")]]
+SANITIZER_INTERFACE_ATTRIBUTE
+THREADLOCAL uptr __hwasan_tls;
+
+namespace __hwasan {
+
+bool InitShadow() {
+  __sanitizer::InitShadowBounds();
+  CHECK_NE(__sanitizer::ShadowBounds.shadow_limit, 0);
+
+  return true;
+}
+
+bool MemIsApp(uptr p) {
+  CHECK(GetTagFromPointer(p) == 0);
+  return __sanitizer::ShadowBounds.shadow_limit <= p &&
+         p <= (__sanitizer::ShadowBounds.memory_limit - 1);
+}
+
+// These are known parameters passed to the hwasan runtime on thread creation.
+struct Thread::InitState {
+  uptr stack_bottom, stack_top;
+};
+
+static void FinishThreadInitialization(Thread *thread);
+
+void InitThreads() {
+  // This is the minimal alignment needed for the storage where hwasan threads
+  // and their stack ring buffers are placed. This alignment is necessary so the
+  // stack ring buffer can perform a simple calculation to get the next element
+  // in the RB. The instructions for this calculation are emitted by the
+  // compiler. (Full explanation in hwasan_thread_list.h.)
+  uptr alloc_size = UINT64_C(1) << kShadowBaseAlignment;
+  uptr thread_start = reinterpret_cast<uptr>(
+      MmapAlignedOrDieOnFatalError(alloc_size, alloc_size, __func__));
+
+  InitThreadList(thread_start, alloc_size);
+
+  // Create the hwasan thread object for the current (main) thread. Stack info
+  // for this thread is known from information passed via
+  // __sanitizer_startup_hook.
+  const Thread::InitState state = {
+      .stack_bottom = __sanitizer::MainThreadStackBase,
+      .stack_top =
+          __sanitizer::MainThreadStackBase + __sanitizer::MainThreadStackSize,
+  };
+  FinishThreadInitialization(hwasanThreadList().CreateCurrentThread(&state));
+}
+
+uptr *GetCurrentThreadLongPtr() { return &__hwasan_tls; }
+
+// This is called from the parent thread before the new thread is created. Here
+// we can propagate known info like the stack bounds to Thread::Init before
+// jumping into the thread. We cannot initialize the stack ring buffer yet since
+// we have not entered the new thread.
+static void *BeforeThreadCreateHook(uptr user_id, bool detached,
+                                    const char *name, uptr stack_bottom,
+                                    uptr stack_size) {
+  const Thread::InitState state = {
+      .stack_bottom = stack_bottom,
+      .stack_top = stack_bottom + stack_size,
+  };
+  return hwasanThreadList().CreateCurrentThread(&state);
+}
+
+// This sets the stack top and bottom according to the InitState passed to
+// CreateCurrentThread above.
+void Thread::InitStackAndTls(const InitState *state) {
+  CHECK_NE(state->stack_bottom, 0);
+  CHECK_NE(state->stack_top, 0);
+  stack_bottom_ = state->stack_bottom;
+  stack_top_ = state->stack_top;
+  tls_end_ = tls_begin_ = 0;
+}
+
+// This is called after creating a new thread with the pointer returned by
+// BeforeThreadCreateHook. We are still in the creating thread and should check
+// if it was actually created correctly.
+static void ThreadCreateHook(void *hook, bool aborted) {
+  Thread *thread = static_cast<Thread *>(hook);
+  if (!aborted) {
+    // The thread was created successfully.
+    // ThreadStartHook can already be running in the new thread.
+  } else {
+    // The thread wasn't created after all.
+    // Clean up everything we set up in BeforeThreadCreateHook.
+    atomic_signal_fence(memory_order_seq_cst);
+    hwasanThreadList().ReleaseThread(thread);
+  }
+}
+
+// This is called in the newly-created thread before it runs anything else,
+// with the pointer returned by BeforeThreadCreateHook (above). Here we can
+// setup the stack ring buffer.
+static void ThreadStartHook(void *hook, thrd_t self) {
+  Thread *thread = static_cast<Thread *>(hook);
+  FinishThreadInitialization(thread);
+  thread->InitRandomState();
+}
+
+// This is the function that sets up the stack ring buffer and enables us to use
+// GetCurrentThread. This function should only be called while IN the thread
+// that we want to create the hwasan thread object for so __hwasan_tls can be
+// properly referenced.
+static void FinishThreadInitialization(Thread *thread) {
+  CHECK_NE(thread, nullptr);
+
+  // The ring buffer is located immediately before the thread object.
+  uptr stack_buffer_size = hwasanThreadList().GetRingBufferSize();
+  uptr stack_buffer_start = reinterpret_cast<uptr>(thread) - stack_buffer_size;
+  thread->InitStackRingBuffer(stack_buffer_start, stack_buffer_size);
+}
+
+static void ThreadExitHook(void *hook, thrd_t self) {
+  Thread *thread = static_cast<Thread *>(hook);
+  atomic_signal_fence(memory_order_seq_cst);
+  hwasanThreadList().ReleaseThread(thread);
+}
+
+// Not implemented because Fuchsia does not use signal handlers.
+void HwasanOnDeadlySignal(int signo, void *info, void *context) {}
+
+// Not implemented because Fuchsia does not use interceptors.
+void InitializeInterceptors() {}
+
+// Not implemented because this is only relevant for Android.
+void AndroidTestTlsSlot() {}
+
+// TSD was normally used on linux as a means of calling the hwasan thread exit
+// handler passed to pthread_key_create. This is not needed on Fuchsia because
+// we will be using __sanitizer_thread_exit_hook.
+void HwasanTSDInit() {}
+void HwasanTSDThreadInit() {}
+
+// On linux, this just would call `atexit(HwasanAtExit)`. The functions in
+// HwasanAtExit are unimplemented for Fuchsia and effectively no-ops, so this
+// function is unneeded.
+void InstallAtExitHandler() {}
+
+}  // namespace __hwasan
+
+extern "C" {
+
+void *__sanitizer_before_thread_create_hook(thrd_t thread, bool detached,
+                                            const char *name, void *stack_base,
+                                            size_t stack_size) {
+  return __hwasan::BeforeThreadCreateHook(
+      reinterpret_cast<uptr>(thread), detached, name,
+      reinterpret_cast<uptr>(stack_base), stack_size);
+}
+
+void __sanitizer_thread_create_hook(void *hook, thrd_t thread, int error) {
+  __hwasan::ThreadCreateHook(hook, error != thrd_success);
+}
+
+void __sanitizer_thread_start_hook(void *hook, thrd_t self) {
+  __hwasan::ThreadStartHook(hook, reinterpret_cast<uptr>(self));
+}
+
+void __sanitizer_thread_exit_hook(void *hook, thrd_t self) {
+  __hwasan::ThreadExitHook(hook, self);
+}
+
+}  // extern "C"
+
+#endif  // SANITIZER_FUCHSIA
diff --git a/libsanitizer/hwasan/hwasan_interceptors.cpp b/libsanitizer/hwasan/hwasan_interceptors.cpp
index ad67e2787d3..68f8adec077 100644
--- a/libsanitizer/hwasan/hwasan_interceptors.cpp
+++ b/libsanitizer/hwasan/hwasan_interceptors.cpp
@@ -16,192 +16,14 @@
 
 #include "interception/interception.h"
 #include "hwasan.h"
-#include "hwasan_allocator.h"
-#include "hwasan_mapping.h"
 #include "hwasan_thread.h"
-#include "hwasan_poisoning.h"
-#include "hwasan_report.h"
-#include "sanitizer_common/sanitizer_platform_limits_posix.h"
-#include "sanitizer_common/sanitizer_allocator.h"
-#include "sanitizer_common/sanitizer_allocator_interface.h"
-#include "sanitizer_common/sanitizer_allocator_internal.h"
-#include "sanitizer_common/sanitizer_atomic.h"
-#include "sanitizer_common/sanitizer_common.h"
-#include "sanitizer_common/sanitizer_errno.h"
 #include "sanitizer_common/sanitizer_stackdepot.h"
-#include "sanitizer_common/sanitizer_libc.h"
-#include "sanitizer_common/sanitizer_linux.h"
-#include "sanitizer_common/sanitizer_tls_get_addr.h"
 
-#include <stdarg.h>
-// ACHTUNG! No other system header includes in this file.
-// Ideally, we should get rid of stdarg.h as well.
+#if !SANITIZER_FUCHSIA
 
 using namespace __hwasan;
 
-using __sanitizer::memory_order;
-using __sanitizer::atomic_load;
-using __sanitizer::atomic_store;
-using __sanitizer::atomic_uintptr_t;
-
-static uptr allocated_for_dlsym;
-static const uptr kDlsymAllocPoolSize = 1024;
-static uptr alloc_memory_for_dlsym[kDlsymAllocPoolSize];
-
-static bool IsInDlsymAllocPool(const void *ptr) {
-  uptr off = (uptr)ptr - (uptr)alloc_memory_for_dlsym;
-  return off < sizeof(alloc_memory_for_dlsym);
-}
-
-static void *AllocateFromLocalPool(uptr size_in_bytes) {
-  uptr size_in_words = RoundUpTo(size_in_bytes, kWordSize) / kWordSize;
-  void *mem = (void *)&alloc_memory_for_dlsym[allocated_for_dlsym];
-  allocated_for_dlsym += size_in_words;
-  CHECK_LT(allocated_for_dlsym, kDlsymAllocPoolSize);
-  return mem;
-}
-
-#define ENSURE_HWASAN_INITED() do { \
-  CHECK(!hwasan_init_is_running); \
-  if (!hwasan_inited) { \
-    __hwasan_init(); \
-  } \
-} while (0)
-
-
-int __sanitizer_posix_memalign(void **memptr, uptr alignment, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  CHECK_NE(memptr, 0);
-  int res = hwasan_posix_memalign(memptr, alignment, size, &stack);
-  return res;
-}
-
-void * __sanitizer_memalign(uptr alignment, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  return hwasan_memalign(alignment, size, &stack);
-}
-
-void * __sanitizer_aligned_alloc(uptr alignment, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  return hwasan_aligned_alloc(alignment, size, &stack);
-}
-
-void * __sanitizer___libc_memalign(uptr alignment, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  void *ptr = hwasan_memalign(alignment, size, &stack);
-  if (ptr)
-    DTLS_on_libc_memalign(ptr, size);
-  return ptr;
-}
-
-void * __sanitizer_valloc(uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  return hwasan_valloc(size, &stack);
-}
-
-void * __sanitizer_pvalloc(uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  return hwasan_pvalloc(size, &stack);
-}
-
-void __sanitizer_free(void *ptr) {
-  GET_MALLOC_STACK_TRACE;
-  if (!ptr || UNLIKELY(IsInDlsymAllocPool(ptr))) return;
-  hwasan_free(ptr, &stack);
-}
-
-void __sanitizer_cfree(void *ptr) {
-  GET_MALLOC_STACK_TRACE;
-  if (!ptr || UNLIKELY(IsInDlsymAllocPool(ptr))) return;
-  hwasan_free(ptr, &stack);
-}
-
-uptr __sanitizer_malloc_usable_size(const void *ptr) {
-  return __sanitizer_get_allocated_size(ptr);
-}
-
-struct __sanitizer_struct_mallinfo __sanitizer_mallinfo() {
-  __sanitizer_struct_mallinfo sret;
-  internal_memset(&sret, 0, sizeof(sret));
-  return sret;
-}
-
-int __sanitizer_mallopt(int cmd, int value) {
-  return 0;
-}
-
-void __sanitizer_malloc_stats(void) {
-  // FIXME: implement, but don't call REAL(malloc_stats)!
-}
-
-void * __sanitizer_calloc(uptr nmemb, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  if (UNLIKELY(!hwasan_inited))
-    // Hack: dlsym calls calloc before REAL(calloc) is retrieved from dlsym.
-    return AllocateFromLocalPool(nmemb * size);
-  return hwasan_calloc(nmemb, size, &stack);
-}
-
-void * __sanitizer_realloc(void *ptr, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  if (UNLIKELY(IsInDlsymAllocPool(ptr))) {
-    uptr offset = (uptr)ptr - (uptr)alloc_memory_for_dlsym;
-    uptr copy_size = Min(size, kDlsymAllocPoolSize - offset);
-    void *new_ptr;
-    if (UNLIKELY(!hwasan_inited)) {
-      new_ptr = AllocateFromLocalPool(copy_size);
-    } else {
-      copy_size = size;
-      new_ptr = hwasan_malloc(copy_size, &stack);
-    }
-    internal_memcpy(new_ptr, ptr, copy_size);
-    return new_ptr;
-  }
-  return hwasan_realloc(ptr, size, &stack);
-}
-
-void * __sanitizer_reallocarray(void *ptr, uptr nmemb, uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  return hwasan_reallocarray(ptr, nmemb, size, &stack);
-}
-
-void * __sanitizer_malloc(uptr size) {
-  GET_MALLOC_STACK_TRACE;
-  if (UNLIKELY(!hwasan_init_is_running))
-    ENSURE_HWASAN_INITED();
-  if (UNLIKELY(!hwasan_inited))
-    // Hack: dlsym calls malloc before REAL(malloc) is retrieved from dlsym.
-    return AllocateFromLocalPool(size);
-  return hwasan_malloc(size, &stack);
-}
-
 #if HWASAN_WITH_INTERCEPTORS
-#define INTERCEPTOR_ALIAS(RET, FN, ARGS...)                                  \
-  extern "C" SANITIZER_INTERFACE_ATTRIBUTE RET WRAP(FN)(ARGS)                \
-      ALIAS("__sanitizer_" #FN);                                             \
-  extern "C" SANITIZER_INTERFACE_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE RET FN(  \
-      ARGS) ALIAS("__sanitizer_" #FN)
-
-INTERCEPTOR_ALIAS(int, posix_memalign, void **memptr, SIZE_T alignment,
-                  SIZE_T size);
-INTERCEPTOR_ALIAS(void *, aligned_alloc, SIZE_T alignment, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, __libc_memalign, SIZE_T alignment, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, valloc, SIZE_T size);
-INTERCEPTOR_ALIAS(void, free, void *ptr);
-INTERCEPTOR_ALIAS(uptr, malloc_usable_size, const void *ptr);
-INTERCEPTOR_ALIAS(void *, calloc, SIZE_T nmemb, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, realloc, void *ptr, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, reallocarray, void *ptr, SIZE_T nmemb, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, malloc, SIZE_T size);
-
-#if !SANITIZER_FREEBSD && !SANITIZER_NETBSD
-INTERCEPTOR_ALIAS(void *, memalign, SIZE_T alignment, SIZE_T size);
-INTERCEPTOR_ALIAS(void *, pvalloc, SIZE_T size);
-INTERCEPTOR_ALIAS(void, cfree, void *ptr);
-INTERCEPTOR_ALIAS(__sanitizer_struct_mallinfo, mallinfo);
-INTERCEPTOR_ALIAS(int, mallopt, int cmd, int value);
-INTERCEPTOR_ALIAS(void, malloc_stats, void);
-#endif
 
 struct ThreadStartArg {
   thread_callback_t callback;
@@ -346,3 +168,5 @@ void InitializeInterceptors() {
   inited = 1;
 }
 } // namespace __hwasan
+
+#endif  // #if !SANITIZER_FUCHSIA
diff --git a/libsanitizer/hwasan/hwasan_linux.cpp b/libsanitizer/hwasan/hwasan_linux.cpp
index 8ce0ff7da95..e22723529f4 100644
--- a/libsanitizer/hwasan/hwasan_linux.cpp
+++ b/libsanitizer/hwasan/hwasan_linux.cpp
@@ -69,15 +69,9 @@ static void ProtectGap(uptr addr, uptr size) {
 
 uptr kLowMemStart;
 uptr kLowMemEnd;
-uptr kLowShadowEnd;
-uptr kLowShadowStart;
-uptr kHighShadowStart;
-uptr kHighShadowEnd;
 uptr kHighMemStart;
 uptr kHighMemEnd;
 
-uptr kAliasRegionStart;  // Always 0 on non-x86.
-
 static void PrintRange(uptr start, uptr end, const char *name) {
   Printf("|| [%p, %p] || %.*s ||\n", (void *)start, (void *)end, 10, name);
 }
@@ -116,7 +110,7 @@ static void InitializeShadowBaseAddress(uptr shadow_size_bytes) {
       FindDynamicShadowStart(shadow_size_bytes);
 }
 
-void InitPrctl() {
+void InitializeOsSupport() {
 #define PR_SET_TAGGED_ADDR_CTRL 55
 #define PR_GET_TAGGED_ADDR_CTRL 56
 #define PR_TAGGED_ADDR_ENABLE (1UL << 0)
@@ -125,33 +119,50 @@ void InitPrctl() {
   if (internal_iserror(internal_prctl(PR_GET_TAGGED_ADDR_CTRL, 0, 0, 0, 0),
                        &local_errno) &&
       local_errno == EINVAL) {
-#if SANITIZER_ANDROID || defined(__x86_64__)
+#  if SANITIZER_ANDROID || defined(HWASAN_ALIASING_MODE)
     // Some older Android kernels have the tagged pointer ABI on
     // unconditionally, and hence don't have the tagged-addr prctl while still
     // allow the ABI.
     // If targeting Android and the prctl is not around we assume this is the
     // case.
     return;
-#else
+#  else
     if (flags()->fail_without_syscall_abi) {
       Printf(
           "FATAL: "
           "HWAddressSanitizer requires a kernel with tagged address ABI.\n");
       Die();
     }
-#endif
+#  endif
   }
 
   // Turn on the tagged address ABI.
   if ((internal_iserror(internal_prctl(PR_SET_TAGGED_ADDR_CTRL,
                                        PR_TAGGED_ADDR_ENABLE, 0, 0, 0)) ||
-       !internal_prctl(PR_GET_TAGGED_ADDR_CTRL, 0, 0, 0, 0)) &&
-      flags()->fail_without_syscall_abi) {
-    Printf(
-        "FATAL: HWAddressSanitizer failed to enable tagged address syscall "
-        "ABI.\nSuggest check `sysctl abi.tagged_addr_disabled` "
-        "configuration.\n");
-    Die();
+       !internal_prctl(PR_GET_TAGGED_ADDR_CTRL, 0, 0, 0, 0))) {
+#  if defined(__x86_64__) && !defined(HWASAN_ALIASING_MODE)
+    // Try the new prctl API for Intel LAM.  The API is based on a currently
+    // unsubmitted patch to the Linux kernel (as of May 2021) and is thus
+    // subject to change.  Patch is here:
+    // https://lore.kernel.org/linux-mm/20210205151631.43511-12-kirill.shutemov@linux.intel.com/
+    int tag_bits = kTagBits;
+    int tag_shift = kAddressTagShift;
+    if (!internal_iserror(
+            internal_prctl(PR_SET_TAGGED_ADDR_CTRL, PR_TAGGED_ADDR_ENABLE,
+                           reinterpret_cast<unsigned long>(&tag_bits),
+                           reinterpret_cast<unsigned long>(&tag_shift), 0))) {
+      CHECK_EQ(tag_bits, kTagBits);
+      CHECK_EQ(tag_shift, kAddressTagShift);
+      return;
+    }
+#  endif  // defined(__x86_64__) && !defined(HWASAN_ALIASING_MODE)
+    if (flags()->fail_without_syscall_abi) {
+      Printf(
+          "FATAL: HWAddressSanitizer failed to enable tagged address syscall "
+          "ABI.\nSuggest check `sysctl abi.tagged_addr_disabled` "
+          "configuration.\n");
+      Die();
+    }
   }
 #undef PR_SET_TAGGED_ADDR_CTRL
 #undef PR_GET_TAGGED_ADDR_CTRL
@@ -181,18 +192,6 @@ bool InitShadow() {
   // High memory starts where allocated shadow allows.
   kHighMemStart = ShadowToMem(kHighShadowStart);
 
-#if defined(__x86_64__)
-  constexpr uptr kAliasRegionOffset = 1ULL << (kTaggableRegionCheckShift - 1);
-  kAliasRegionStart =
-      __hwasan_shadow_memory_dynamic_address + kAliasRegionOffset;
-
-  CHECK_EQ(kAliasRegionStart >> kTaggableRegionCheckShift,
-           __hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
-  CHECK_EQ(
-      (kAliasRegionStart + kAliasRegionOffset - 1) >> kTaggableRegionCheckShift,
-      __hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
-#endif
-
   // Check the sanity of the defined memory ranges (there might be gaps).
   CHECK_EQ(kHighMemStart % GetMmapGranularity(), 0);
   CHECK_GT(kHighMemStart, kHighShadowEnd);
@@ -233,25 +232,16 @@ void InitThreads() {
   ProtectGap(thread_space_end,
              __hwasan_shadow_memory_dynamic_address - thread_space_end);
   InitThreadList(thread_space_start, thread_space_end - thread_space_start);
+  hwasanThreadList().CreateCurrentThread();
 }
 
 bool MemIsApp(uptr p) {
-#if !defined(__x86_64__)  // Memory outside the alias range has non-zero tags.
+// Memory outside the alias range has non-zero tags.
+#  if !defined(HWASAN_ALIASING_MODE)
   CHECK(GetTagFromPointer(p) == 0);
-#endif
-  return p >= kHighMemStart || (p >= kLowMemStart && p <= kLowMemEnd);
-}
+#  endif
 
-static void HwasanAtExit(void) {
-  if (common_flags()->print_module_map)
-    DumpProcessMap();
-  if (flags()->print_stats && (flags()->atexit || hwasan_report_count > 0))
-    ReportStats();
-  if (hwasan_report_count > 0) {
-    // ReportAtExitStatistics();
-    if (common_flags()->exitcode)
-      internal__exit(common_flags()->exitcode);
-  }
+  return p >= kHighMemStart || (p >= kLowMemStart && p <= kLowMemEnd);
 }
 
 void InstallAtExitHandler() {
@@ -330,22 +320,6 @@ void AndroidTestTlsSlot() {
 void AndroidTestTlsSlot() {}
 #endif
 
-Thread *GetCurrentThread() {
-  uptr *ThreadLongPtr = GetCurrentThreadLongPtr();
-  if (UNLIKELY(*ThreadLongPtr == 0))
-    return nullptr;
-  auto *R = (StackAllocationsRingBuffer *)ThreadLongPtr;
-  return hwasanThreadList().GetThreadByBufferAddress((uptr)R->Next());
-}
-
-struct AccessInfo {
-  uptr addr;
-  uptr size;
-  bool is_store;
-  bool is_load;
-  bool recover;
-};
-
 static AccessInfo GetAccessInfo(siginfo_t *info, ucontext_t *uc) {
   // Access type is passed in a platform dependent way (see below) and encoded
   // as 0xXY, where X&1 is 1 for store, 0 for load, and X&2 is 1 if the error is
@@ -396,28 +370,6 @@ static AccessInfo GetAccessInfo(siginfo_t *info, ucontext_t *uc) {
   return AccessInfo{addr, size, is_store, !is_store, recover};
 }
 
-static void HandleTagMismatch(AccessInfo ai, uptr pc, uptr frame,
-                              ucontext_t *uc, uptr *registers_frame = nullptr) {
-  InternalMmapVector<BufferedStackTrace> stack_buffer(1);
-  BufferedStackTrace *stack = stack_buffer.data();
-  stack->Reset();
-  stack->Unwind(pc, frame, uc, common_flags()->fast_unwind_on_fatal);
-
-  // The second stack frame contains the failure __hwasan_check function, as
-  // we have a stack frame for the registers saved in __hwasan_tag_mismatch that
-  // we wish to ignore. This (currently) only occurs on AArch64, as x64
-  // implementations use SIGTRAP to implement the failure, and thus do not go
-  // through the stack saver.
-  if (registers_frame && stack->trace && stack->size > 0) {
-    stack->trace++;
-    stack->size--;
-  }
-
-  bool fatal = flags()->halt_on_error || !ai.recover;
-  ReportTagMismatch(stack, ai.addr, ai.size, ai.is_store, fatal,
-                    registers_frame);
-}
-
 static bool HwasanOnSIGTRAP(int signo, siginfo_t *info, ucontext_t *uc) {
   AccessInfo ai = GetAccessInfo(info, uc);
   if (!ai.is_store && !ai.is_load)
@@ -450,27 +402,39 @@ void HwasanOnDeadlySignal(int signo, void *info, void *context) {
   HandleDeadlySignal(info, context, GetTid(), &OnStackUnwind, nullptr);
 }
 
+void Thread::InitStackAndTls(const InitState *) {
+  uptr tls_size;
+  uptr stack_size;
+  GetThreadStackAndTls(IsMainThread(), &stack_bottom_, &stack_size, &tls_begin_,
+                       &tls_size);
+  stack_top_ = stack_bottom_ + stack_size;
+  tls_end_ = tls_begin_ + tls_size;
+}
+
+uptr TagMemoryAligned(uptr p, uptr size, tag_t tag) {
+  CHECK(IsAligned(p, kShadowAlignment));
+  CHECK(IsAligned(size, kShadowAlignment));
+  uptr shadow_start = MemToShadow(p);
+  uptr shadow_size = MemToShadowSize(size);
+
+  uptr page_size = GetPageSizeCached();
+  uptr page_start = RoundUpTo(shadow_start, page_size);
+  uptr page_end = RoundDownTo(shadow_start + shadow_size, page_size);
+  uptr threshold = common_flags()->clear_shadow_mmap_threshold;
+  if (SANITIZER_LINUX &&
+      UNLIKELY(page_end >= page_start + threshold && tag == 0)) {
+    internal_memset((void *)shadow_start, tag, page_start - shadow_start);
+    internal_memset((void *)page_end, tag,
+                    shadow_start + shadow_size - page_end);
+    // For an anonymous private mapping MADV_DONTNEED will return a zero page on
+    // Linux.
+    ReleaseMemoryPagesToOSAndZeroFill(page_start, page_end);
+  } else {
+    internal_memset((void *)shadow_start, tag, shadow_size);
+  }
+  return AddTagToPointer(p, tag);
+}
 
 } // namespace __hwasan
 
-// Entry point for interoperability between __hwasan_tag_mismatch (ASM) and the
-// rest of the mismatch handling code (C++).
-void __hwasan_tag_mismatch4(uptr addr, uptr access_info, uptr *registers_frame,
-                            size_t outsize) {
-  __hwasan::AccessInfo ai;
-  ai.is_store = access_info & 0x10;
-  ai.is_load = !ai.is_store;
-  ai.recover = access_info & 0x20;
-  ai.addr = addr;
-  if ((access_info & 0xf) == 0xf)
-    ai.size = outsize;
-  else
-    ai.size = 1 << (access_info & 0xf);
-
-  __hwasan::HandleTagMismatch(ai, (uptr)__builtin_return_address(0),
-                              (uptr)__builtin_frame_address(0), nullptr,
-                              registers_frame);
-  __builtin_unreachable();
-}
-
 #endif // SANITIZER_FREEBSD || SANITIZER_LINUX || SANITIZER_NETBSD
diff --git a/libsanitizer/hwasan/hwasan_mapping.h b/libsanitizer/hwasan/hwasan_mapping.h
index 8243d1ec7ed..79a143632f6 100644
--- a/libsanitizer/hwasan/hwasan_mapping.h
+++ b/libsanitizer/hwasan/hwasan_mapping.h
@@ -48,14 +48,14 @@ extern uptr kHighShadowEnd;
 extern uptr kHighMemStart;
 extern uptr kHighMemEnd;
 
-extern uptr kAliasRegionStart;
-
+inline uptr GetShadowOffset() {
+  return SANITIZER_FUCHSIA ? 0 : __hwasan_shadow_memory_dynamic_address;
+}
 inline uptr MemToShadow(uptr untagged_addr) {
-  return (untagged_addr >> kShadowScale) +
-         __hwasan_shadow_memory_dynamic_address;
+  return (untagged_addr >> kShadowScale) + GetShadowOffset();
 }
 inline uptr ShadowToMem(uptr shadow_addr) {
-  return (shadow_addr - __hwasan_shadow_memory_dynamic_address) << kShadowScale;
+  return (shadow_addr - GetShadowOffset()) << kShadowScale;
 }
 inline uptr MemToShadowSize(uptr size) {
   return size >> kShadowScale;
@@ -63,6 +63,13 @@ inline uptr MemToShadowSize(uptr size) {
 
 bool MemIsApp(uptr p);
 
+inline bool MemIsShadow(uptr p) {
+  return (kLowShadowStart <= p && p <= kLowShadowEnd) ||
+         (kHighShadowStart <= p && p <= kHighShadowEnd);
+}
+
+uptr GetAliasRegionStart();
+
 }  // namespace __hwasan
 
 #endif  // HWASAN_MAPPING_H
diff --git a/libsanitizer/hwasan/hwasan_new_delete.cpp b/libsanitizer/hwasan/hwasan_new_delete.cpp
index 69cddda736e..4e057a651e1 100644
--- a/libsanitizer/hwasan/hwasan_new_delete.cpp
+++ b/libsanitizer/hwasan/hwasan_new_delete.cpp
@@ -56,7 +56,6 @@ using namespace __hwasan;
 // Fake std::nothrow_t to avoid including <new>.
 namespace std {
   struct nothrow_t {};
-  enum class align_val_t : size_t {};
 }  // namespace std
 
 
@@ -73,6 +72,32 @@ INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
 void *operator new[](size_t size, std::nothrow_t const&) {
   OPERATOR_NEW_BODY(true /*nothrow*/);
 }
+
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete(void *ptr)
+    NOEXCEPT {
+  OPERATOR_DELETE_BODY;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete[](
+    void *ptr) NOEXCEPT {
+  OPERATOR_DELETE_BODY;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete(
+    void *ptr, std::nothrow_t const &) {
+  OPERATOR_DELETE_BODY;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete[](
+    void *ptr, std::nothrow_t const &) {
+  OPERATOR_DELETE_BODY;
+}
+
+#endif  // OPERATOR_NEW_BODY
+
+#ifdef OPERATOR_NEW_ALIGN_BODY
+
+namespace std {
+enum class align_val_t : size_t {};
+}  // namespace std
+
 INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void *operator new(
     size_t size, std::align_val_t align) {
   OPERATOR_NEW_ALIGN_BODY(false /*nothrow*/);
@@ -90,16 +115,6 @@ INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void *operator new[](
   OPERATOR_NEW_ALIGN_BODY(true /*nothrow*/);
 }
 
-INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
-void operator delete(void *ptr) NOEXCEPT { OPERATOR_DELETE_BODY; }
-INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
-void operator delete[](void *ptr) NOEXCEPT { OPERATOR_DELETE_BODY; }
-INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
-void operator delete(void *ptr, std::nothrow_t const&) { OPERATOR_DELETE_BODY; }
-INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
-void operator delete[](void *ptr, std::nothrow_t const&) {
-  OPERATOR_DELETE_BODY;
-}
 INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete(
     void *ptr, std::align_val_t align) NOEXCEPT {
   OPERATOR_DELETE_BODY;
@@ -117,4 +132,4 @@ INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete[](
   OPERATOR_DELETE_BODY;
 }
 
-#endif // OPERATOR_NEW_BODY
+#endif  // OPERATOR_NEW_ALIGN_BODY
diff --git a/libsanitizer/hwasan/hwasan_poisoning.cpp b/libsanitizer/hwasan/hwasan_poisoning.cpp
index 2a0816428e7..5aafdb1884b 100644
--- a/libsanitizer/hwasan/hwasan_poisoning.cpp
+++ b/libsanitizer/hwasan/hwasan_poisoning.cpp
@@ -19,30 +19,6 @@
 
 namespace __hwasan {
 
-uptr TagMemoryAligned(uptr p, uptr size, tag_t tag) {
-  CHECK(IsAligned(p, kShadowAlignment));
-  CHECK(IsAligned(size, kShadowAlignment));
-  uptr shadow_start = MemToShadow(p);
-  uptr shadow_size = MemToShadowSize(size);
-
-  uptr page_size = GetPageSizeCached();
-  uptr page_start = RoundUpTo(shadow_start, page_size);
-  uptr page_end = RoundDownTo(shadow_start + shadow_size, page_size);
-  uptr threshold = common_flags()->clear_shadow_mmap_threshold;
-  if (SANITIZER_LINUX &&
-      UNLIKELY(page_end >= page_start + threshold && tag == 0)) {
-    internal_memset((void *)shadow_start, tag, page_start - shadow_start);
-    internal_memset((void *)page_end, tag,
-                    shadow_start + shadow_size - page_end);
-    // For an anonymous private mapping MADV_DONTNEED will return a zero page on
-    // Linux.
-    ReleaseMemoryPagesToOSAndZeroFill(page_start, page_end);
-  } else {
-    internal_memset((void *)shadow_start, tag, shadow_size);
-  }
-  return AddTagToPointer(p, tag);
-}
-
 uptr TagMemory(uptr p, uptr size, tag_t tag) {
   uptr start = RoundDownTo(p, kShadowAlignment);
   uptr end = RoundUpTo(p + size, kShadowAlignment);
diff --git a/libsanitizer/hwasan/hwasan_report.cpp b/libsanitizer/hwasan/hwasan_report.cpp
index c0217799391..44047c9fdaf 100644
--- a/libsanitizer/hwasan/hwasan_report.cpp
+++ b/libsanitizer/hwasan/hwasan_report.cpp
@@ -236,12 +236,12 @@ static void PrintStackAllocations(StackAllocationsRingBuffer *sa,
     frame_desc.append("  record_addr:0x%zx record:0x%zx",
                       reinterpret_cast<uptr>(record_addr), record);
     if (SymbolizedStack *frame = Symbolizer::GetOrInit()->SymbolizePC(pc)) {
-      RenderFrame(&frame_desc, " %F %L\n", 0, frame->info.address, &frame->info,
+      RenderFrame(&frame_desc, " %F %L", 0, frame->info.address, &frame->info,
                   common_flags()->symbolize_vs_style,
                   common_flags()->strip_path_prefix);
       frame->ClearAll();
     }
-    Printf("%s", frame_desc.data());
+    Printf("%s\n", frame_desc.data());
     frame_desc.clear();
   }
 }
@@ -296,6 +296,75 @@ static uptr GetGlobalSizeFromDescriptor(uptr ptr) {
   return 0;
 }
 
+static void ShowHeapOrGlobalCandidate(uptr untagged_addr, tag_t *candidate,
+                                      tag_t *left, tag_t *right) {
+  Decorator d;
+  uptr mem = ShadowToMem(reinterpret_cast<uptr>(candidate));
+  HwasanChunkView chunk = FindHeapChunkByAddress(mem);
+  if (chunk.IsAllocated()) {
+    uptr offset;
+    const char *whence;
+    if (untagged_addr < chunk.End() && untagged_addr >= chunk.Beg()) {
+      offset = untagged_addr - chunk.Beg();
+      whence = "inside";
+    } else if (candidate == left) {
+      offset = untagged_addr - chunk.End();
+      whence = "to the right of";
+    } else {
+      offset = chunk.Beg() - untagged_addr;
+      whence = "to the left of";
+    }
+    Printf("%s", d.Error());
+    Printf("\nCause: heap-buffer-overflow\n");
+    Printf("%s", d.Default());
+    Printf("%s", d.Location());
+    Printf("%p is located %zd bytes %s %zd-byte region [%p,%p)\n",
+           untagged_addr, offset, whence, chunk.UsedSize(), chunk.Beg(),
+           chunk.End());
+    Printf("%s", d.Allocation());
+    Printf("allocated here:\n");
+    Printf("%s", d.Default());
+    GetStackTraceFromId(chunk.GetAllocStackId()).Print();
+    return;
+  }
+  // Check whether the address points into a loaded library. If so, this is
+  // most likely a global variable.
+  const char *module_name;
+  uptr module_address;
+  Symbolizer *sym = Symbolizer::GetOrInit();
+  if (sym->GetModuleNameAndOffsetForPC(mem, &module_name, &module_address)) {
+    Printf("%s", d.Error());
+    Printf("\nCause: global-overflow\n");
+    Printf("%s", d.Default());
+    DataInfo info;
+    Printf("%s", d.Location());
+    if (sym->SymbolizeData(mem, &info) && info.start) {
+      Printf(
+          "%p is located %zd bytes to the %s of %zd-byte global variable "
+          "%s [%p,%p) in %s\n",
+          untagged_addr,
+          candidate == left ? untagged_addr - (info.start + info.size)
+                            : info.start - untagged_addr,
+          candidate == left ? "right" : "left", info.size, info.name,
+          info.start, info.start + info.size, module_name);
+    } else {
+      uptr size = GetGlobalSizeFromDescriptor(mem);
+      if (size == 0)
+        // We couldn't find the size of the global from the descriptors.
+        Printf("%p is located to the %s of a global variable in (%s+0x%x)\n",
+               untagged_addr, candidate == left ? "right" : "left", module_name,
+               module_address);
+      else
+        Printf(
+            "%p is located to the %s of a %zd-byte global variable in "
+            "(%s+0x%x)\n",
+            untagged_addr, candidate == left ? "right" : "left", size,
+            module_name, module_address);
+    }
+    Printf("%s", d.Default());
+  }
+}
+
 void PrintAddressDescription(
     uptr tagged_addr, uptr access_size,
     StackAllocationsRingBuffer *current_stack_allocations) {
@@ -317,78 +386,59 @@ void PrintAddressDescription(
            d.Default());
   }
 
+  tag_t addr_tag = GetTagFromPointer(tagged_addr);
+
+  bool on_stack = false;
+  // Check stack first. If the address is on the stack of a live thread, we
+  // know it cannot be a heap / global overflow.
+  hwasanThreadList().VisitAllLiveThreads([&](Thread *t) {
+    if (t->AddrIsInStack(untagged_addr)) {
+      on_stack = true;
+      // TODO(fmayer): figure out how to distinguish use-after-return and
+      // stack-buffer-overflow.
+      Printf("%s", d.Error());
+      Printf("\nCause: stack tag-mismatch\n");
+      Printf("%s", d.Location());
+      Printf("Address %p is located in stack of thread T%zd\n", untagged_addr,
+             t->unique_id());
+      Printf("%s", d.Default());
+      t->Announce();
+
+      auto *sa = (t == GetCurrentThread() && current_stack_allocations)
+                     ? current_stack_allocations
+                     : t->stack_allocations();
+      PrintStackAllocations(sa, addr_tag, untagged_addr);
+      num_descriptions_printed++;
+    }
+  });
+
   // Check if this looks like a heap buffer overflow by scanning
   // the shadow left and right and looking for the first adjacent
   // object with a different memory tag. If that tag matches addr_tag,
   // check the allocator if it has a live chunk there.
-  tag_t addr_tag = GetTagFromPointer(tagged_addr);
   tag_t *tag_ptr = reinterpret_cast<tag_t*>(MemToShadow(untagged_addr));
   tag_t *candidate = nullptr, *left = tag_ptr, *right = tag_ptr;
-  for (int i = 0; i < 1000; i++) {
-    if (TagsEqual(addr_tag, left)) {
+  uptr candidate_distance = 0;
+  for (; candidate_distance < 1000; candidate_distance++) {
+    if (MemIsShadow(reinterpret_cast<uptr>(left)) &&
+        TagsEqual(addr_tag, left)) {
       candidate = left;
       break;
     }
     --left;
-    if (TagsEqual(addr_tag, right)) {
+    if (MemIsShadow(reinterpret_cast<uptr>(right)) &&
+        TagsEqual(addr_tag, right)) {
       candidate = right;
       break;
     }
     ++right;
   }
 
-  if (candidate) {
-    uptr mem = ShadowToMem(reinterpret_cast<uptr>(candidate));
-    HwasanChunkView chunk = FindHeapChunkByAddress(mem);
-    if (chunk.IsAllocated()) {
-      Printf("%s", d.Location());
-      Printf("%p is located %zd bytes to the %s of %zd-byte region [%p,%p)\n",
-             untagged_addr,
-             candidate == left ? untagged_addr - chunk.End()
-                               : chunk.Beg() - untagged_addr,
-             candidate == left ? "right" : "left", chunk.UsedSize(),
-             chunk.Beg(), chunk.End());
-      Printf("%s", d.Allocation());
-      Printf("allocated here:\n");
-      Printf("%s", d.Default());
-      GetStackTraceFromId(chunk.GetAllocStackId()).Print();
-      num_descriptions_printed++;
-    } else {
-      // Check whether the address points into a loaded library. If so, this is
-      // most likely a global variable.
-      const char *module_name;
-      uptr module_address;
-      Symbolizer *sym = Symbolizer::GetOrInit();
-      if (sym->GetModuleNameAndOffsetForPC(mem, &module_name,
-                                           &module_address)) {
-        DataInfo info;
-        if (sym->SymbolizeData(mem, &info) && info.start) {
-          Printf(
-              "%p is located %zd bytes to the %s of %zd-byte global variable "
-              "%s [%p,%p) in %s\n",
-              untagged_addr,
-              candidate == left ? untagged_addr - (info.start + info.size)
-                                : info.start - untagged_addr,
-              candidate == left ? "right" : "left", info.size, info.name,
-              info.start, info.start + info.size, module_name);
-        } else {
-          uptr size = GetGlobalSizeFromDescriptor(mem);
-          if (size == 0)
-            // We couldn't find the size of the global from the descriptors.
-            Printf(
-                "%p is located to the %s of a global variable in (%s+0x%x)\n",
-                untagged_addr, candidate == left ? "right" : "left",
-                module_name, module_address);
-          else
-            Printf(
-                "%p is located to the %s of a %zd-byte global variable in "
-                "(%s+0x%x)\n",
-                untagged_addr, candidate == left ? "right" : "left", size,
-                module_name, module_address);
-        }
-        num_descriptions_printed++;
-      }
-    }
+  constexpr auto kCloseCandidateDistance = 1;
+
+  if (!on_stack && candidate && candidate_distance <= kCloseCandidateDistance) {
+    ShowHeapOrGlobalCandidate(untagged_addr, candidate, left, right);
+    num_descriptions_printed++;
   }
 
   hwasanThreadList().VisitAllLiveThreads([&](Thread *t) {
@@ -398,6 +448,8 @@ void PrintAddressDescription(
     if (FindHeapAllocation(t->heap_allocations(), tagged_addr, &har,
                            &ring_index, &num_matching_addrs,
                            &num_matching_addrs_4b)) {
+      Printf("%s", d.Error());
+      Printf("\nCause: use-after-free\n");
       Printf("%s", d.Location());
       Printf("%p is located %zd bytes inside of %zd-byte region [%p,%p)\n",
              untagged_addr, untagged_addr - UntagAddr(har.tagged_addr),
@@ -424,29 +476,25 @@ void PrintAddressDescription(
       t->Announce();
       num_descriptions_printed++;
     }
-
-    // Very basic check for stack memory.
-    if (t->AddrIsInStack(untagged_addr)) {
-      Printf("%s", d.Location());
-      Printf("Address %p is located in stack of thread T%zd\n", untagged_addr,
-             t->unique_id());
-      Printf("%s", d.Default());
-      t->Announce();
-
-      auto *sa = (t == GetCurrentThread() && current_stack_allocations)
-                     ? current_stack_allocations
-                     : t->stack_allocations();
-      PrintStackAllocations(sa, addr_tag, untagged_addr);
-      num_descriptions_printed++;
-    }
   });
 
+  if (candidate && num_descriptions_printed == 0) {
+    ShowHeapOrGlobalCandidate(untagged_addr, candidate, left, right);
+    num_descriptions_printed++;
+  }
+
   // Print the remaining threads, as an extra information, 1 line per thread.
   hwasanThreadList().VisitAllLiveThreads([&](Thread *t) { t->Announce(); });
 
   if (!num_descriptions_printed)
     // We exhausted our possibilities. Bail out.
     Printf("HWAddressSanitizer can not describe address in more detail.\n");
+  if (num_descriptions_printed > 1) {
+    Printf(
+        "There are %d potential causes, printed above in order "
+        "of likeliness.\n",
+        num_descriptions_printed);
+  }
 }
 
 void ReportStats() {}
@@ -538,6 +586,12 @@ void ReportTailOverwritten(StackTrace *stack, uptr tagged_addr, uptr orig_size,
   Report("ERROR: %s: %s; heap object [%p,%p) of size %zd\n", SanitizerToolName,
          bug_type, untagged_addr, untagged_addr + orig_size, orig_size);
   Printf("\n%s", d.Default());
+  Printf(
+      "Stack of invalid access unknown. Issue detected at deallocation "
+      "time.\n");
+  Printf("%s", d.Allocation());
+  Printf("deallocated here:\n");
+  Printf("%s", d.Default());
   stack->Print();
   HwasanChunkView chunk = FindHeapChunkByAddress(untagged_addr);
   if (chunk.Beg()) {
@@ -657,8 +711,10 @@ void ReportRegisters(uptr *frame, uptr pc) {
        frame[20], frame[21], frame[22], frame[23]);
   Printf("    x24 %016llx  x25 %016llx  x26 %016llx  x27 %016llx\n",
        frame[24], frame[25], frame[26], frame[27]);
-  Printf("    x28 %016llx  x29 %016llx  x30 %016llx\n",
-       frame[28], frame[29], frame[30]);
+  // hwasan_check* reduces the stack pointer by 256, then __hwasan_tag_mismatch
+  // passes it to this function.
+  Printf("    x28 %016llx  x29 %016llx  x30 %016llx   sp %016llx\n", frame[28],
+         frame[29], frame[30], reinterpret_cast<u8 *>(frame) + 256);
 }
 
 }  // namespace __hwasan
diff --git a/libsanitizer/hwasan/hwasan_thread.cpp b/libsanitizer/hwasan/hwasan_thread.cpp
index bb4d56abed0..ee747a3beea 100644
--- a/libsanitizer/hwasan/hwasan_thread.cpp
+++ b/libsanitizer/hwasan/hwasan_thread.cpp
@@ -34,7 +34,8 @@ void Thread::InitRandomState() {
     stack_allocations_->push(0);
 }
 
-void Thread::Init(uptr stack_buffer_start, uptr stack_buffer_size) {
+void Thread::Init(uptr stack_buffer_start, uptr stack_buffer_size,
+                  const InitState *state) {
   CHECK_EQ(0, unique_id_);  // try to catch bad stack reuse
   CHECK_EQ(0, stack_top_);
   CHECK_EQ(0, stack_bottom_);
@@ -44,6 +45,17 @@ void Thread::Init(uptr stack_buffer_start, uptr stack_buffer_size) {
   if (auto sz = flags()->heap_history_size)
     heap_allocations_ = HeapAllocationsRingBuffer::New(sz);
 
+  InitStackAndTls(state);
+#if !SANITIZER_FUCHSIA
+  // Do not initialize the stack ring buffer just yet on Fuchsia. Threads will
+  // be initialized before we enter the thread itself, so we will instead call
+  // this later.
+  InitStackRingBuffer(stack_buffer_start, stack_buffer_size);
+#endif
+}
+
+void Thread::InitStackRingBuffer(uptr stack_buffer_start,
+                                 uptr stack_buffer_size) {
   HwasanTSDThreadInit();  // Only needed with interceptors.
   uptr *ThreadLong = GetCurrentThreadLongPtr();
   // The following implicitly sets (this) as the current thread.
@@ -55,13 +67,6 @@ void Thread::Init(uptr stack_buffer_start, uptr stack_buffer_size) {
   // ScopedTaggingDisable needs GetCurrentThread to be set up.
   ScopedTaggingDisabler disabler;
 
-  uptr tls_size;
-  uptr stack_size;
-  GetThreadStackAndTls(IsMainThread(), &stack_bottom_, &stack_size, &tls_begin_,
-                       &tls_size);
-  stack_top_ = stack_bottom_ + stack_size;
-  tls_end_ = tls_begin_ + tls_size;
-
   if (stack_bottom_) {
     int local;
     CHECK(AddrIsInStack((uptr)&local));
diff --git a/libsanitizer/hwasan/hwasan_thread.h b/libsanitizer/hwasan/hwasan_thread.h
index 1c71cab41c4..9f20afe1dc7 100644
--- a/libsanitizer/hwasan/hwasan_thread.h
+++ b/libsanitizer/hwasan/hwasan_thread.h
@@ -23,8 +23,17 @@ typedef __sanitizer::CompactRingBuffer<uptr> StackAllocationsRingBuffer;
 
 class Thread {
  public:
-  void Init(uptr stack_buffer_start, uptr stack_buffer_size);  // Must be called from the thread itself.
+  // These are optional parameters that can be passed to Init.
+  struct InitState;
+
+  void Init(uptr stack_buffer_start, uptr stack_buffer_size,
+            const InitState *state = nullptr);
   void InitRandomState();
+  void InitStackAndTls(const InitState *state = nullptr);
+
+  // Must be called from the thread itself.
+  void InitStackRingBuffer(uptr stack_buffer_start, uptr stack_buffer_size);
+
   void Destroy();
 
   uptr stack_top() { return stack_top_; }
diff --git a/libsanitizer/hwasan/hwasan_thread_list.cpp b/libsanitizer/hwasan/hwasan_thread_list.cpp
index a31eee84ed9..fa46e658b69 100644
--- a/libsanitizer/hwasan/hwasan_thread_list.cpp
+++ b/libsanitizer/hwasan/hwasan_thread_list.cpp
@@ -12,4 +12,4 @@ void InitThreadList(uptr storage, uptr size) {
       new (thread_list_placeholder) HwasanThreadList(storage, size);
 }
 
-} // namespace
+} // namespace __hwasan
diff --git a/libsanitizer/hwasan/hwasan_thread_list.h b/libsanitizer/hwasan/hwasan_thread_list.h
index 11c586314ce..15916a802d6 100644
--- a/libsanitizer/hwasan/hwasan_thread_list.h
+++ b/libsanitizer/hwasan/hwasan_thread_list.h
@@ -85,7 +85,7 @@ class HwasanThreadList {
         RoundUpTo(ring_buffer_size_ + sizeof(Thread), ring_buffer_size_ * 2);
   }
 
-  Thread *CreateCurrentThread() {
+  Thread *CreateCurrentThread(const Thread::InitState *state = nullptr) {
     Thread *t = nullptr;
     {
       SpinMutexLock l(&free_list_mutex_);
@@ -104,7 +104,7 @@ class HwasanThreadList {
       SpinMutexLock l(&live_list_mutex_);
       live_list_.push_back(t);
     }
-    t->Init((uptr)t - ring_buffer_size_, ring_buffer_size_);
+    t->Init((uptr)t - ring_buffer_size_, ring_buffer_size_, state);
     AddThreadStats(t);
     return t;
   }
@@ -171,6 +171,8 @@ class HwasanThreadList {
     return stats_;
   }
 
+  uptr GetRingBufferSize() const { return ring_buffer_size_; }
+
  private:
   Thread *AllocThread() {
     SpinMutexLock l(&free_space_mutex_);
@@ -200,4 +202,4 @@ class HwasanThreadList {
 void InitThreadList(uptr storage, uptr size);
 HwasanThreadList &hwasanThreadList();
 
-} // namespace
+} // namespace __hwasan
diff --git a/libsanitizer/include/sanitizer/dfsan_interface.h b/libsanitizer/include/sanitizer/dfsan_interface.h
index 40f9379b557..cd3b6d6e2b1 100644
--- a/libsanitizer/include/sanitizer/dfsan_interface.h
+++ b/libsanitizer/include/sanitizer/dfsan_interface.h
@@ -21,34 +21,15 @@
 extern "C" {
 #endif
 
-typedef uint16_t dfsan_label;
+typedef uint8_t dfsan_label;
 typedef uint32_t dfsan_origin;
 
-/// Stores information associated with a specific label identifier.  A label
-/// may be a base label created using dfsan_create_label, with associated
-/// text description and user data, or an automatically created union label,
-/// which represents the union of two label identifiers (which may themselves
-/// be base or union labels).
-struct dfsan_label_info {
-  // Fields for union labels, set to 0 for base labels.
-  dfsan_label l1;
-  dfsan_label l2;
-
-  // Fields for base labels.
-  const char *desc;
-  void *userdata;
-};
-
 /// Signature of the callback argument to dfsan_set_write_callback().
 typedef void (*dfsan_write_callback_t)(int fd, const void *buf, size_t count);
 
-/// Computes the union of \c l1 and \c l2, possibly creating a union label in
-/// the process.
+/// Computes the union of \c l1 and \c l2, resulting in a union label.
 dfsan_label dfsan_union(dfsan_label l1, dfsan_label l2);
 
-/// Creates and returns a base label with the given description and user data.
-dfsan_label dfsan_create_label(const char *desc, void *userdata);
-
 /// Sets the label for each address in [addr,addr+size) to \c label.
 void dfsan_set_label(dfsan_label label, void *addr, size_t size);
 
@@ -73,19 +54,9 @@ dfsan_origin dfsan_get_origin(long data);
 /// Retrieves the label associated with the data at the given address.
 dfsan_label dfsan_read_label(const void *addr, size_t size);
 
-/// Retrieves a pointer to the dfsan_label_info struct for the given label.
-const struct dfsan_label_info *dfsan_get_label_info(dfsan_label label);
-
 /// Returns whether the given label label contains the label elem.
 int dfsan_has_label(dfsan_label label, dfsan_label elem);
 
-/// If the given label label contains a label with the description desc, returns
-/// that label, else returns 0.
-dfsan_label dfsan_has_label_with_desc(dfsan_label label, const char *desc);
-
-/// Returns the number of labels allocated.
-size_t dfsan_get_label_count(void);
-
 /// Flushes the DFSan shadow, i.e. forgets about all labels currently associated
 /// with the application memory.  Use this call to start over the taint tracking
 /// within the same process.
@@ -99,12 +70,6 @@ void dfsan_flush(void);
 /// callback executes.  Pass in NULL to remove any callback.
 void dfsan_set_write_callback(dfsan_write_callback_t labeled_write_callback);
 
-/// Writes the labels currently used by the program to the given file
-/// descriptor. The lines of the output have the following format:
-///
-/// <label> <parent label 1> <parent label 2> <label description if any>
-void dfsan_dump_labels(int fd);
-
 /// Interceptor hooks.
 /// Whenever a dfsan's custom function is called the corresponding
 /// hook is called it non-zero. The hooks should be defined by the user.
@@ -123,9 +88,65 @@ void dfsan_weak_hook_strncmp(void *caller_pc, const char *s1, const char *s2,
 /// on, or the address is not labeled, it prints nothing.
 void dfsan_print_origin_trace(const void *addr, const char *description);
 
+/// Prints the origin trace of the label at the address \p addr to a
+/// pre-allocated output buffer. If origin tracking is not on, or the address is
+/// not labeled, it prints nothing.
+///
+/// Typical usage:
+/// \code
+///   char kDescription[] = "...";
+///   char buf[1024];
+///   dfsan_sprint_origin_trace(&tainted_var, kDescription, buf, sizeof(buf));
+/// \endcode
+///
+/// Typical usage that handles truncation:
+/// \code
+///   char buf[1024];
+///   int len = dfsan_sprint_origin_trace(&var, nullptr, buf, sizeof(buf));
+///
+///   if (len < sizeof(buf)) {
+///     ProcessOriginTrace(buf);
+///   } else {
+///     char *tmpbuf = new char[len + 1];
+///     dfsan_sprint_origin_trace(&var, nullptr, tmpbuf, len + 1);
+///     ProcessOriginTrace(tmpbuf);
+///     delete[] tmpbuf;
+///   }
+/// \endcode
+///
+/// \param addr The tainted memory address whose origin we are printing.
+/// \param description A description printed at the beginning of the trace.
+/// \param [out] out_buf The output buffer to write the results to.
+/// \param out_buf_size The size of \p out_buf.
+///
+/// \returns The number of symbols that should have been written to \p out_buf
+/// (not including trailing null byte '\0'). Thus, the string is truncated iff
+/// return value is not less than \p out_buf_size.
+size_t dfsan_sprint_origin_trace(const void *addr, const char *description,
+                                 char *out_buf, size_t out_buf_size);
+
+/// Prints the stack trace leading to this call to a pre-allocated output
+/// buffer.
+///
+/// For usage examples, see dfsan_sprint_origin_trace.
+///
+/// \param [out] out_buf The output buffer to write the results to.
+/// \param out_buf_size The size of \p out_buf.
+///
+/// \returns The number of symbols that should have been written to \p out_buf
+/// (not including trailing null byte '\0'). Thus, the string is truncated iff
+/// return value is not less than \p out_buf_size.
+size_t dfsan_sprint_stack_trace(char *out_buf, size_t out_buf_size);
+
 /// Retrieves the very first origin associated with the data at the given
 /// address.
 dfsan_origin dfsan_get_init_origin(const void *addr);
+
+/// Returns the value of -dfsan-track-origins.
+/// * 0: do not track origins.
+/// * 1: track origins at memory store operations.
+/// * 2: track origins at memory load and store operations.
+int dfsan_get_track_origins(void);
 #ifdef __cplusplus
 }  // extern "C"
 
diff --git a/libsanitizer/interception/interception.h b/libsanitizer/interception/interception.h
index cb0b5284ed2..d8dc092c45f 100644
--- a/libsanitizer/interception/interception.h
+++ b/libsanitizer/interception/interception.h
@@ -16,10 +16,10 @@
 
 #include "sanitizer_common/sanitizer_internal_defs.h"
 
-#if !SANITIZER_LINUX && !SANITIZER_FREEBSD && !SANITIZER_MAC && \
-    !SANITIZER_NETBSD && !SANITIZER_WINDOWS && \
-    !SANITIZER_FUCHSIA && !SANITIZER_RTEMS && !SANITIZER_SOLARIS
-# error "Interception doesn't work on this operating system."
+#if !SANITIZER_LINUX && !SANITIZER_FREEBSD && !SANITIZER_MAC &&      \
+    !SANITIZER_NETBSD && !SANITIZER_WINDOWS && !SANITIZER_FUCHSIA && \
+    !SANITIZER_SOLARIS
+#  error "Interception doesn't work on this operating system."
 #endif
 
 // These typedefs should be used only in the interceptor definitions to replace
@@ -130,11 +130,6 @@ const interpose_substitution substitution_##func_name[] \
     extern "C" ret_type func(__VA_ARGS__);
 # define DECLARE_WRAPPER_WINAPI(ret_type, func, ...) \
     extern "C" __declspec(dllimport) ret_type __stdcall func(__VA_ARGS__);
-#elif SANITIZER_RTEMS
-# define WRAP(x) x
-# define WRAPPER_NAME(x) #x
-# define INTERCEPTOR_ATTRIBUTE
-# define DECLARE_WRAPPER(ret_type, func, ...)
 #elif SANITIZER_FREEBSD || SANITIZER_NETBSD
 # define WRAP(x) __interceptor_ ## x
 # define WRAPPER_NAME(x) "__interceptor_" #x
@@ -162,10 +157,6 @@ const interpose_substitution substitution_##func_name[] \
 # define INTERCEPTOR_ATTRIBUTE __attribute__((visibility("default")))
 # define REAL(x) __unsanitized_##x
 # define DECLARE_REAL(ret_type, func, ...)
-#elif SANITIZER_RTEMS
-# define REAL(x) __real_ ## x
-# define DECLARE_REAL(ret_type, func, ...) \
-    extern "C" ret_type REAL(func)(__VA_ARGS__);
 #elif !SANITIZER_MAC
 # define PTR_TO_REAL(x) real_##x
 # define REAL(x) __interception::PTR_TO_REAL(x)
@@ -184,10 +175,10 @@ const interpose_substitution substitution_##func_name[] \
 # define ASSIGN_REAL(x, y)
 #endif  // SANITIZER_MAC
 
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
-# define DECLARE_REAL_AND_INTERCEPTOR(ret_type, func, ...) \
-  DECLARE_REAL(ret_type, func, __VA_ARGS__) \
-  extern "C" ret_type WRAP(func)(__VA_ARGS__);
+#if !SANITIZER_FUCHSIA
+#  define DECLARE_REAL_AND_INTERCEPTOR(ret_type, func, ...) \
+    DECLARE_REAL(ret_type, func, __VA_ARGS__)               \
+    extern "C" ret_type WRAP(func)(__VA_ARGS__);
 // Declare an interceptor and its wrapper defined in a different translation
 // unit (ex. asm).
 # define DECLARE_EXTERN_INTERCEPTOR_AND_WRAPPER(ret_type, func, ...)    \
@@ -202,11 +193,11 @@ const interpose_substitution substitution_##func_name[] \
 // macros does its job. In exceptional cases you may need to call REAL(foo)
 // without defining INTERCEPTOR(..., foo, ...). For example, if you override
 // foo with an interceptor for other function.
-#if !SANITIZER_MAC && !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
-# define DEFINE_REAL(ret_type, func, ...) \
+#if !SANITIZER_MAC && !SANITIZER_FUCHSIA
+#  define DEFINE_REAL(ret_type, func, ...)            \
     typedef ret_type (*FUNC_TYPE(func))(__VA_ARGS__); \
-    namespace __interception { \
-      FUNC_TYPE(func) PTR_TO_REAL(func); \
+    namespace __interception {                        \
+    FUNC_TYPE(func) PTR_TO_REAL(func);                \
     }
 #else
 # define DEFINE_REAL(ret_type, func, ...)
diff --git a/libsanitizer/lsan/lsan.cpp b/libsanitizer/lsan/lsan.cpp
index 2c0a3bf0787..b6adc248157 100644
--- a/libsanitizer/lsan/lsan.cpp
+++ b/libsanitizer/lsan/lsan.cpp
@@ -35,18 +35,14 @@ void __sanitizer::BufferedStackTrace::UnwindImpl(
     uptr pc, uptr bp, void *context, bool request_fast, u32 max_depth) {
   using namespace __lsan;
   uptr stack_top = 0, stack_bottom = 0;
-  ThreadContext *t;
-  if (StackTrace::WillUseFastUnwind(request_fast) &&
-      (t = CurrentThreadContext())) {
+  if (ThreadContext *t = CurrentThreadContext()) {
     stack_top = t->stack_end();
     stack_bottom = t->stack_begin();
   }
-  if (!SANITIZER_MIPS || IsValidFrame(bp, stack_top, stack_bottom)) {
-    if (StackTrace::WillUseFastUnwind(request_fast))
-      Unwind(max_depth, pc, bp, nullptr, stack_top, stack_bottom, true);
-    else
-      Unwind(max_depth, pc, 0, context, 0, 0, false);
-  }
+  if (SANITIZER_MIPS && !IsValidFrame(bp, stack_top, stack_bottom))
+    return;
+  bool fast = StackTrace::WillUseFastUnwind(request_fast);
+  Unwind(max_depth, pc, bp, context, stack_top, stack_bottom, fast);
 }
 
 using namespace __lsan;
diff --git a/libsanitizer/lsan/lsan_common.h b/libsanitizer/lsan/lsan_common.h
index fe855cf3754..776ca60b1e9 100644
--- a/libsanitizer/lsan/lsan_common.h
+++ b/libsanitizer/lsan/lsan_common.h
@@ -221,8 +221,8 @@ void UnlockAllocator();
 // Returns true if [addr, addr + sizeof(void *)) is poisoned.
 bool WordIsPoisoned(uptr addr);
 // Wrappers for ThreadRegistry access.
-void LockThreadRegistry();
-void UnlockThreadRegistry();
+void LockThreadRegistry() NO_THREAD_SAFETY_ANALYSIS;
+void UnlockThreadRegistry() NO_THREAD_SAFETY_ANALYSIS;
 ThreadRegistry *GetThreadRegistryLocked();
 bool GetThreadRangesLocked(tid_t os_id, uptr *stack_begin, uptr *stack_end,
                            uptr *tls_begin, uptr *tls_end, uptr *cache_begin,
diff --git a/libsanitizer/lsan/lsan_thread.cpp b/libsanitizer/lsan/lsan_thread.cpp
index 8efb54a6fb7..1d224ebca69 100644
--- a/libsanitizer/lsan/lsan_thread.cpp
+++ b/libsanitizer/lsan/lsan_thread.cpp
@@ -30,13 +30,10 @@ static ThreadContextBase *CreateThreadContext(u32 tid) {
   return new (mem) ThreadContext(tid);
 }
 
-static const uptr kMaxThreads = 1 << 13;
-static const uptr kThreadQuarantineSize = 64;
-
 void InitializeThreadRegistry() {
   static ALIGNED(64) char thread_registry_placeholder[sizeof(ThreadRegistry)];
-  thread_registry = new (thread_registry_placeholder)
-      ThreadRegistry(CreateThreadContext, kMaxThreads, kThreadQuarantineSize);
+  thread_registry =
+      new (thread_registry_placeholder) ThreadRegistry(CreateThreadContext);
 }
 
 ThreadContextLsanBase::ThreadContextLsanBase(int tid)
diff --git a/libsanitizer/sanitizer_common/Makefile.am b/libsanitizer/sanitizer_common/Makefile.am
index 7cba5342647..d04f2d8bd16 100644
--- a/libsanitizer/sanitizer_common/Makefile.am
+++ b/libsanitizer/sanitizer_common/Makefile.am
@@ -41,6 +41,7 @@ sanitizer_common_files = \
 	sanitizer_linux_s390.cpp \
 	sanitizer_mac.cpp \
 	sanitizer_mac_libcdep.cpp \
+	sanitizer_mutex.cpp \
 	sanitizer_netbsd.cpp \
 	sanitizer_openbsd.cpp \
 	sanitizer_persistent_allocator.cpp \
@@ -57,7 +58,6 @@ sanitizer_common_files = \
 	sanitizer_procmaps_linux.cpp \
 	sanitizer_procmaps_mac.cpp \
 	sanitizer_procmaps_solaris.cpp \
-	sanitizer_rtems.cpp \
 	sanitizer_solaris.cpp \
 	sanitizer_stackdepot.cpp \
 	sanitizer_stacktrace.cpp \
diff --git a/libsanitizer/sanitizer_common/Makefile.in b/libsanitizer/sanitizer_common/Makefile.in
index 7e5555cc9df..2856894d62b 100644
--- a/libsanitizer/sanitizer_common/Makefile.in
+++ b/libsanitizer/sanitizer_common/Makefile.in
@@ -128,8 +128,9 @@ am__objects_1 = sancov_flags.lo sanitizer_allocator.lo \
 	sanitizer_file.lo sanitizer_flags.lo sanitizer_flag_parser.lo \
 	sanitizer_libc.lo sanitizer_libignore.lo sanitizer_linux.lo \
 	sanitizer_linux_libcdep.lo sanitizer_linux_s390.lo \
-	sanitizer_mac.lo sanitizer_mac_libcdep.lo sanitizer_netbsd.lo \
-	sanitizer_openbsd.lo sanitizer_persistent_allocator.lo \
+	sanitizer_mac.lo sanitizer_mac_libcdep.lo sanitizer_mutex.lo \
+	sanitizer_netbsd.lo sanitizer_openbsd.lo \
+	sanitizer_persistent_allocator.lo \
 	sanitizer_platform_limits_freebsd.lo \
 	sanitizer_platform_limits_linux.lo \
 	sanitizer_platform_limits_openbsd.lo \
@@ -138,11 +139,11 @@ am__objects_1 = sancov_flags.lo sanitizer_allocator.lo \
 	sanitizer_posix_libcdep.lo sanitizer_printf.lo \
 	sanitizer_procmaps_bsd.lo sanitizer_procmaps_common.lo \
 	sanitizer_procmaps_linux.lo sanitizer_procmaps_mac.lo \
-	sanitizer_procmaps_solaris.lo sanitizer_rtems.lo \
-	sanitizer_solaris.lo sanitizer_stackdepot.lo \
-	sanitizer_stacktrace.lo sanitizer_stacktrace_libcdep.lo \
-	sanitizer_stacktrace_sparc.lo sanitizer_symbolizer_mac.lo \
-	sanitizer_symbolizer_report.lo sanitizer_stacktrace_printer.lo \
+	sanitizer_procmaps_solaris.lo sanitizer_solaris.lo \
+	sanitizer_stackdepot.lo sanitizer_stacktrace.lo \
+	sanitizer_stacktrace_libcdep.lo sanitizer_stacktrace_sparc.lo \
+	sanitizer_symbolizer_mac.lo sanitizer_symbolizer_report.lo \
+	sanitizer_stacktrace_printer.lo \
 	sanitizer_stoptheworld_linux_libcdep.lo \
 	sanitizer_stoptheworld_mac.lo sanitizer_suppressions.lo \
 	sanitizer_symbolizer.lo sanitizer_symbolizer_libbacktrace.lo \
@@ -400,6 +401,7 @@ sanitizer_common_files = \
 	sanitizer_linux_s390.cpp \
 	sanitizer_mac.cpp \
 	sanitizer_mac_libcdep.cpp \
+	sanitizer_mutex.cpp \
 	sanitizer_netbsd.cpp \
 	sanitizer_openbsd.cpp \
 	sanitizer_persistent_allocator.cpp \
@@ -416,7 +418,6 @@ sanitizer_common_files = \
 	sanitizer_procmaps_linux.cpp \
 	sanitizer_procmaps_mac.cpp \
 	sanitizer_procmaps_solaris.cpp \
-	sanitizer_rtems.cpp \
 	sanitizer_solaris.cpp \
 	sanitizer_stackdepot.cpp \
 	sanitizer_stacktrace.cpp \
@@ -557,6 +558,7 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_linux_s390.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_mac.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_mac_libcdep.Plo@am__quote@
+@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_mutex.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_netbsd.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_openbsd.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_persistent_allocator.Plo@am__quote@
@@ -573,7 +575,6 @@ distclean-compile:
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_procmaps_linux.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_procmaps_mac.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_procmaps_solaris.Plo@am__quote@
-@AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_rtems.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_solaris.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_stackdepot.Plo@am__quote@
 @AMDEP_TRUE@@am__include@ @am__quote@./$(DEPDIR)/sanitizer_stacktrace.Plo@am__quote@
diff --git a/libsanitizer/sanitizer_common/sanitizer_addrhashmap.h b/libsanitizer/sanitizer_common/sanitizer_addrhashmap.h
index a033e788cbf..15f81a04350 100644
--- a/libsanitizer/sanitizer_common/sanitizer_addrhashmap.h
+++ b/libsanitizer/sanitizer_common/sanitizer_addrhashmap.h
@@ -162,8 +162,8 @@ AddrHashMap<T, kSize>::AddrHashMap() {
   table_ = (Bucket*)MmapOrDie(kSize * sizeof(table_[0]), "AddrHashMap");
 }
 
-template<typename T, uptr kSize>
-void AddrHashMap<T, kSize>::acquire(Handle *h) {
+template <typename T, uptr kSize>
+void AddrHashMap<T, kSize>::acquire(Handle *h) NO_THREAD_SAFETY_ANALYSIS {
   uptr addr = h->addr_;
   uptr hash = calcHash(addr);
   Bucket *b = &table_[hash];
@@ -289,57 +289,57 @@ void AddrHashMap<T, kSize>::acquire(Handle *h) {
   CHECK_EQ(atomic_load(&c->addr, memory_order_relaxed), 0);
   h->addidx_ = i;
   h->cell_ = c;
-}
+ }
 
-template<typename T, uptr kSize>
-void AddrHashMap<T, kSize>::release(Handle *h) {
-  if (!h->cell_)
-    return;
-  Bucket *b = h->bucket_;
-  Cell *c = h->cell_;
-  uptr addr1 = atomic_load(&c->addr, memory_order_relaxed);
-  if (h->created_) {
-    // Denote completion of insertion.
-    CHECK_EQ(addr1, 0);
-    // After the following store, the element becomes available
-    // for lock-free reads.
-    atomic_store(&c->addr, h->addr_, memory_order_release);
-    b->mtx.Unlock();
-  } else if (h->remove_) {
-    // Denote that the cell is empty now.
-    CHECK_EQ(addr1, h->addr_);
-    atomic_store(&c->addr, 0, memory_order_release);
-    // See if we need to compact the bucket.
-    AddBucket *add = (AddBucket*)atomic_load(&b->add, memory_order_relaxed);
-    if (h->addidx_ == -1U) {
-      // Removed from embed array, move an add element into the freed cell.
-      if (add && add->size != 0) {
-        uptr last = --add->size;
-        Cell *c1 = &add->cells[last];
-        c->val = c1->val;
-        uptr addr1 = atomic_load(&c1->addr, memory_order_relaxed);
-        atomic_store(&c->addr, addr1, memory_order_release);
-        atomic_store(&c1->addr, 0, memory_order_release);
-      }
-    } else {
-      // Removed from add array, compact it.
-      uptr last = --add->size;
-      Cell *c1 = &add->cells[last];
-      if (c != c1) {
-        *c = *c1;
-        atomic_store(&c1->addr, 0, memory_order_relaxed);
-      }
-    }
-    if (add && add->size == 0) {
-      // FIXME(dvyukov): free add?
-    }
-    b->mtx.Unlock();
-  } else {
-    CHECK_EQ(addr1, h->addr_);
-    if (h->addidx_ != -1U)
-      b->mtx.ReadUnlock();
-  }
-}
+ template <typename T, uptr kSize>
+ void AddrHashMap<T, kSize>::release(Handle *h) NO_THREAD_SAFETY_ANALYSIS {
+   if (!h->cell_)
+     return;
+   Bucket *b = h->bucket_;
+   Cell *c = h->cell_;
+   uptr addr1 = atomic_load(&c->addr, memory_order_relaxed);
+   if (h->created_) {
+     // Denote completion of insertion.
+     CHECK_EQ(addr1, 0);
+     // After the following store, the element becomes available
+     // for lock-free reads.
+     atomic_store(&c->addr, h->addr_, memory_order_release);
+     b->mtx.Unlock();
+   } else if (h->remove_) {
+     // Denote that the cell is empty now.
+     CHECK_EQ(addr1, h->addr_);
+     atomic_store(&c->addr, 0, memory_order_release);
+     // See if we need to compact the bucket.
+     AddBucket *add = (AddBucket *)atomic_load(&b->add, memory_order_relaxed);
+     if (h->addidx_ == -1U) {
+       // Removed from embed array, move an add element into the freed cell.
+       if (add && add->size != 0) {
+         uptr last = --add->size;
+         Cell *c1 = &add->cells[last];
+         c->val = c1->val;
+         uptr addr1 = atomic_load(&c1->addr, memory_order_relaxed);
+         atomic_store(&c->addr, addr1, memory_order_release);
+         atomic_store(&c1->addr, 0, memory_order_release);
+       }
+     } else {
+       // Removed from add array, compact it.
+       uptr last = --add->size;
+       Cell *c1 = &add->cells[last];
+       if (c != c1) {
+         *c = *c1;
+         atomic_store(&c1->addr, 0, memory_order_relaxed);
+       }
+     }
+     if (add && add->size == 0) {
+       // FIXME(dvyukov): free add?
+     }
+     b->mtx.Unlock();
+   } else {
+     CHECK_EQ(addr1, h->addr_);
+     if (h->addidx_ != -1U)
+       b->mtx.ReadUnlock();
+   }
+ }
 
 template<typename T, uptr kSize>
 uptr AddrHashMap<T, kSize>::calcHash(uptr addr) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator.cpp b/libsanitizer/sanitizer_common/sanitizer_allocator.cpp
index 3157b35ffaf..bcb7370a790 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator.cpp
@@ -137,14 +137,6 @@ static void RawInternalFree(void *ptr, InternalAllocatorCache *cache) {
 
 #endif  // SANITIZER_GO || defined(SANITIZER_USE_MALLOC)
 
-namespace {
-const u64 kBlockMagic = 0x6A6CB03ABCEBC041ull;
-
-struct BlockHeader {
-  u64 magic;
-};
-}  // namespace
-
 static void NORETURN ReportInternalAllocatorOutOfMemory(uptr requested_size) {
   SetAllocatorOutOfMemory();
   Report("FATAL: %s: internal allocator is out of memory trying to allocate "
@@ -153,28 +145,17 @@ static void NORETURN ReportInternalAllocatorOutOfMemory(uptr requested_size) {
 }
 
 void *InternalAlloc(uptr size, InternalAllocatorCache *cache, uptr alignment) {
-  uptr s = size + sizeof(BlockHeader);
-  if (s < size)
-    return nullptr;
-  BlockHeader *p = (BlockHeader *)RawInternalAlloc(s, cache, alignment);
+  void *p = RawInternalAlloc(size, cache, alignment);
   if (UNLIKELY(!p))
-    ReportInternalAllocatorOutOfMemory(s);
-  p->magic = kBlockMagic;
-  return p + 1;
+    ReportInternalAllocatorOutOfMemory(size);
+  return p;
 }
 
 void *InternalRealloc(void *addr, uptr size, InternalAllocatorCache *cache) {
-  if (!addr)
-    return InternalAlloc(size, cache);
-  uptr s = size + sizeof(BlockHeader);
-  if (s < size)
-    return nullptr;
-  BlockHeader *p = (BlockHeader *)addr - 1;
-  CHECK_EQ(kBlockMagic, p->magic);
-  p = (BlockHeader *)RawInternalRealloc(p, s, cache);
+  void *p = RawInternalRealloc(addr, size, cache);
   if (UNLIKELY(!p))
-    ReportInternalAllocatorOutOfMemory(s);
-  return p + 1;
+    ReportInternalAllocatorOutOfMemory(size);
+  return p;
 }
 
 void *InternalReallocArray(void *addr, uptr count, uptr size,
@@ -203,12 +184,7 @@ void *InternalCalloc(uptr count, uptr size, InternalAllocatorCache *cache) {
 }
 
 void InternalFree(void *addr, InternalAllocatorCache *cache) {
-  if (!addr)
-    return;
-  BlockHeader *p = (BlockHeader *)addr - 1;
-  CHECK_EQ(kBlockMagic, p->magic);
-  p->magic = 0;
-  RawInternalFree(p, cache);
+  RawInternalFree(addr, cache);
 }
 
 // LowLevelAllocator
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h b/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
index eb836bc4787..0e81e6764f9 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
@@ -177,12 +177,12 @@ class CombinedAllocator {
 
   // ForceLock() and ForceUnlock() are needed to implement Darwin malloc zone
   // introspection API.
-  void ForceLock() {
+  void ForceLock() NO_THREAD_SAFETY_ANALYSIS {
     primary_.ForceLock();
     secondary_.ForceLock();
   }
 
-  void ForceUnlock() {
+  void ForceUnlock() NO_THREAD_SAFETY_ANALYSIS {
     secondary_.ForceUnlock();
     primary_.ForceUnlock();
   }
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_local_cache.h b/libsanitizer/sanitizer_common/sanitizer_allocator_local_cache.h
index 108dfc231a2..e495c56f037 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_local_cache.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_local_cache.h
@@ -17,6 +17,7 @@
 template <class SizeClassAllocator>
 struct SizeClassAllocator64LocalCache {
   typedef SizeClassAllocator Allocator;
+  typedef MemoryMapper<Allocator> MemoryMapperT;
 
   void Init(AllocatorGlobalStats *s) {
     stats_.Init();
@@ -53,7 +54,7 @@ struct SizeClassAllocator64LocalCache {
     PerClass *c = &per_class_[class_id];
     InitCache(c);
     if (UNLIKELY(c->count == c->max_count))
-      Drain(c, allocator, class_id, c->max_count / 2);
+      DrainHalfMax(c, allocator, class_id);
     CompactPtrT chunk = allocator->PointerToCompactPtr(
         allocator->GetRegionBeginBySizeClass(class_id),
         reinterpret_cast<uptr>(p));
@@ -62,10 +63,10 @@ struct SizeClassAllocator64LocalCache {
   }
 
   void Drain(SizeClassAllocator *allocator) {
+    MemoryMapperT memory_mapper(*allocator);
     for (uptr i = 1; i < kNumClasses; i++) {
       PerClass *c = &per_class_[i];
-      while (c->count > 0)
-        Drain(c, allocator, i, c->count);
+      while (c->count > 0) Drain(&memory_mapper, c, allocator, i, c->count);
     }
   }
 
@@ -106,12 +107,18 @@ struct SizeClassAllocator64LocalCache {
     return true;
   }
 
-  NOINLINE void Drain(PerClass *c, SizeClassAllocator *allocator, uptr class_id,
-                      uptr count) {
+  NOINLINE void DrainHalfMax(PerClass *c, SizeClassAllocator *allocator,
+                             uptr class_id) {
+    MemoryMapperT memory_mapper(*allocator);
+    Drain(&memory_mapper, c, allocator, class_id, c->max_count / 2);
+  }
+
+  void Drain(MemoryMapperT *memory_mapper, PerClass *c,
+             SizeClassAllocator *allocator, uptr class_id, uptr count) {
     CHECK_GE(c->count, count);
     const uptr first_idx_to_drain = c->count - count;
     c->count -= count;
-    allocator->ReturnToAllocator(&stats_, class_id,
+    allocator->ReturnToAllocator(memory_mapper, &stats_, class_id,
                                  &c->chunks[first_idx_to_drain], count);
   }
 };
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h b/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
index fb5394cd39c..38d2a7d117f 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
@@ -237,13 +237,13 @@ class SizeClassAllocator32 {
 
   // ForceLock() and ForceUnlock() are needed to implement Darwin malloc zone
   // introspection API.
-  void ForceLock() {
+  void ForceLock() NO_THREAD_SAFETY_ANALYSIS {
     for (uptr i = 0; i < kNumClasses; i++) {
       GetSizeClassInfo(i)->mutex.Lock();
     }
   }
 
-  void ForceUnlock() {
+  void ForceUnlock() NO_THREAD_SAFETY_ANALYSIS {
     for (int i = kNumClasses - 1; i >= 0; i--) {
       GetSizeClassInfo(i)->mutex.Unlock();
     }
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h b/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
index db30e138154..b142ee0131b 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
@@ -42,6 +42,44 @@ struct SizeClassAllocator64FlagMasks {  //  Bit masks.
   };
 };
 
+template <typename Allocator>
+class MemoryMapper {
+ public:
+  typedef typename Allocator::CompactPtrT CompactPtrT;
+
+  explicit MemoryMapper(const Allocator &allocator) : allocator_(allocator) {}
+
+  bool GetAndResetStats(uptr &ranges, uptr &bytes) {
+    ranges = released_ranges_count_;
+    released_ranges_count_ = 0;
+    bytes = released_bytes_;
+    released_bytes_ = 0;
+    return ranges != 0;
+  }
+
+  u64 *MapPackedCounterArrayBuffer(uptr count) {
+    buffer_.clear();
+    buffer_.resize(count);
+    return buffer_.data();
+  }
+
+  // Releases [from, to) range of pages back to OS.
+  void ReleasePageRangeToOS(uptr class_id, CompactPtrT from, CompactPtrT to) {
+    const uptr region_base = allocator_.GetRegionBeginBySizeClass(class_id);
+    const uptr from_page = allocator_.CompactPtrToPointer(region_base, from);
+    const uptr to_page = allocator_.CompactPtrToPointer(region_base, to);
+    ReleaseMemoryPagesToOS(from_page, to_page);
+    released_ranges_count_++;
+    released_bytes_ += to_page - from_page;
+  }
+
+ private:
+  const Allocator &allocator_;
+  uptr released_ranges_count_ = 0;
+  uptr released_bytes_ = 0;
+  InternalMmapVector<u64> buffer_;
+};
+
 template <class Params>
 class SizeClassAllocator64 {
  public:
@@ -57,6 +95,7 @@ class SizeClassAllocator64 {
 
   typedef SizeClassAllocator64<Params> ThisT;
   typedef SizeClassAllocator64LocalCache<ThisT> AllocatorCache;
+  typedef MemoryMapper<ThisT> MemoryMapperT;
 
   // When we know the size class (the region base) we can represent a pointer
   // as a 4-byte integer (offset from the region start shifted right by 4).
@@ -120,9 +159,10 @@ class SizeClassAllocator64 {
   }
 
   void ForceReleaseToOS() {
+    MemoryMapperT memory_mapper(*this);
     for (uptr class_id = 1; class_id < kNumClasses; class_id++) {
       BlockingMutexLock l(&GetRegionInfo(class_id)->mutex);
-      MaybeReleaseToOS(class_id, true /*force*/);
+      MaybeReleaseToOS(&memory_mapper, class_id, true /*force*/);
     }
   }
 
@@ -131,7 +171,8 @@ class SizeClassAllocator64 {
       alignment <= SizeClassMap::kMaxSize;
   }
 
-  NOINLINE void ReturnToAllocator(AllocatorStats *stat, uptr class_id,
+  NOINLINE void ReturnToAllocator(MemoryMapperT *memory_mapper,
+                                  AllocatorStats *stat, uptr class_id,
                                   const CompactPtrT *chunks, uptr n_chunks) {
     RegionInfo *region = GetRegionInfo(class_id);
     uptr region_beg = GetRegionBeginBySizeClass(class_id);
@@ -154,7 +195,7 @@ class SizeClassAllocator64 {
     region->num_freed_chunks = new_num_freed_chunks;
     region->stats.n_freed += n_chunks;
 
-    MaybeReleaseToOS(class_id, false /*force*/);
+    MaybeReleaseToOS(memory_mapper, class_id, false /*force*/);
   }
 
   NOINLINE bool GetFromAllocator(AllocatorStats *stat, uptr class_id,
@@ -312,13 +353,13 @@ class SizeClassAllocator64 {
 
   // ForceLock() and ForceUnlock() are needed to implement Darwin malloc zone
   // introspection API.
-  void ForceLock() {
+  void ForceLock() NO_THREAD_SAFETY_ANALYSIS {
     for (uptr i = 0; i < kNumClasses; i++) {
       GetRegionInfo(i)->mutex.Lock();
     }
   }
 
-  void ForceUnlock() {
+  void ForceUnlock() NO_THREAD_SAFETY_ANALYSIS {
     for (int i = (int)kNumClasses - 1; i >= 0; i--) {
       GetRegionInfo(i)->mutex.Unlock();
     }
@@ -362,11 +403,11 @@ class SizeClassAllocator64 {
   // For the performance sake, none of the accessors check the validity of the
   // arguments, it is assumed that index is always in [0, n) range and the value
   // is not incremented past max_value.
-  template<class MemoryMapperT>
   class PackedCounterArray {
    public:
-    PackedCounterArray(u64 num_counters, u64 max_value, MemoryMapperT *mapper)
-        : n(num_counters), memory_mapper(mapper) {
+    template <typename MemoryMapper>
+    PackedCounterArray(u64 num_counters, u64 max_value, MemoryMapper *mapper)
+        : n(num_counters) {
       CHECK_GT(num_counters, 0);
       CHECK_GT(max_value, 0);
       constexpr u64 kMaxCounterBits = sizeof(*buffer) * 8ULL;
@@ -383,16 +424,8 @@ class SizeClassAllocator64 {
       packing_ratio_log = Log2(packing_ratio);
       bit_offset_mask = packing_ratio - 1;
 
-      buffer_size =
-          (RoundUpTo(n, 1ULL << packing_ratio_log) >> packing_ratio_log) *
-          sizeof(*buffer);
-      buffer = reinterpret_cast<u64*>(
-          memory_mapper->MapPackedCounterArrayBuffer(buffer_size));
-    }
-    ~PackedCounterArray() {
-      if (buffer) {
-        memory_mapper->UnmapPackedCounterArrayBuffer(buffer, buffer_size);
-      }
+      buffer = mapper->MapPackedCounterArrayBuffer(
+          RoundUpTo(n, 1ULL << packing_ratio_log) >> packing_ratio_log);
     }
 
     bool IsAllocated() const {
@@ -429,19 +462,16 @@ class SizeClassAllocator64 {
     u64 counter_mask;
     u64 packing_ratio_log;
     u64 bit_offset_mask;
-
-    MemoryMapperT* const memory_mapper;
-    u64 buffer_size;
     u64* buffer;
   };
 
-  template<class MemoryMapperT>
+  template <class MemoryMapperT>
   class FreePagesRangeTracker {
    public:
-    explicit FreePagesRangeTracker(MemoryMapperT* mapper)
+    FreePagesRangeTracker(MemoryMapperT *mapper, uptr class_id)
         : memory_mapper(mapper),
-          page_size_scaled_log(Log2(GetPageSizeCached() >> kCompactPtrScale)),
-          in_the_range(false), current_page(0), current_range_start_page(0) {}
+          class_id(class_id),
+          page_size_scaled_log(Log2(GetPageSizeCached() >> kCompactPtrScale)) {}
 
     void NextPage(bool freed) {
       if (freed) {
@@ -463,28 +493,30 @@ class SizeClassAllocator64 {
     void CloseOpenedRange() {
       if (in_the_range) {
         memory_mapper->ReleasePageRangeToOS(
-            current_range_start_page << page_size_scaled_log,
+            class_id, current_range_start_page << page_size_scaled_log,
             current_page << page_size_scaled_log);
         in_the_range = false;
       }
     }
 
-    MemoryMapperT* const memory_mapper;
-    const uptr page_size_scaled_log;
-    bool in_the_range;
-    uptr current_page;
-    uptr current_range_start_page;
+    MemoryMapperT *const memory_mapper = nullptr;
+    const uptr class_id = 0;
+    const uptr page_size_scaled_log = 0;
+    bool in_the_range = false;
+    uptr current_page = 0;
+    uptr current_range_start_page = 0;
   };
 
   // Iterates over the free_array to identify memory pages containing freed
   // chunks only and returns these pages back to OS.
   // allocated_pages_count is the total number of pages allocated for the
   // current bucket.
-  template<class MemoryMapperT>
+  template <typename MemoryMapper>
   static void ReleaseFreeMemoryToOS(CompactPtrT *free_array,
                                     uptr free_array_count, uptr chunk_size,
                                     uptr allocated_pages_count,
-                                    MemoryMapperT *memory_mapper) {
+                                    MemoryMapper *memory_mapper,
+                                    uptr class_id) {
     const uptr page_size = GetPageSizeCached();
 
     // Figure out the number of chunks per page and whether we can take a fast
@@ -520,9 +552,8 @@ class SizeClassAllocator64 {
       UNREACHABLE("All chunk_size/page_size ratios must be handled.");
     }
 
-    PackedCounterArray<MemoryMapperT> counters(allocated_pages_count,
-                                               full_pages_chunk_count_max,
-                                               memory_mapper);
+    PackedCounterArray counters(allocated_pages_count,
+                                full_pages_chunk_count_max, memory_mapper);
     if (!counters.IsAllocated())
       return;
 
@@ -547,7 +578,7 @@ class SizeClassAllocator64 {
 
     // Iterate over pages detecting ranges of pages with chunk counters equal
     // to the expected number of chunks for the particular page.
-    FreePagesRangeTracker<MemoryMapperT> range_tracker(memory_mapper);
+    FreePagesRangeTracker<MemoryMapper> range_tracker(memory_mapper, class_id);
     if (same_chunk_count_per_page) {
       // Fast path, every page has the same number of chunks affecting it.
       for (uptr i = 0; i < counters.GetCount(); i++)
@@ -586,7 +617,7 @@ class SizeClassAllocator64 {
   }
 
  private:
-  friend class MemoryMapper;
+  friend class MemoryMapper<ThisT>;
 
   ReservedAddressRange address_range;
 
@@ -820,57 +851,13 @@ class SizeClassAllocator64 {
     return true;
   }
 
-  class MemoryMapper {
-   public:
-    MemoryMapper(const ThisT& base_allocator, uptr class_id)
-        : allocator(base_allocator),
-          region_base(base_allocator.GetRegionBeginBySizeClass(class_id)),
-          released_ranges_count(0),
-          released_bytes(0) {
-    }
-
-    uptr GetReleasedRangesCount() const {
-      return released_ranges_count;
-    }
-
-    uptr GetReleasedBytes() const {
-      return released_bytes;
-    }
-
-    void *MapPackedCounterArrayBuffer(uptr buffer_size) {
-      // TODO(alekseyshl): The idea to explore is to check if we have enough
-      // space between num_freed_chunks*sizeof(CompactPtrT) and
-      // mapped_free_array to fit buffer_size bytes and use that space instead
-      // of mapping a temporary one.
-      return MmapOrDieOnFatalError(buffer_size, "ReleaseToOSPageCounters");
-    }
-
-    void UnmapPackedCounterArrayBuffer(void *buffer, uptr buffer_size) {
-      UnmapOrDie(buffer, buffer_size);
-    }
-
-    // Releases [from, to) range of pages back to OS.
-    void ReleasePageRangeToOS(CompactPtrT from, CompactPtrT to) {
-      const uptr from_page = allocator.CompactPtrToPointer(region_base, from);
-      const uptr to_page = allocator.CompactPtrToPointer(region_base, to);
-      ReleaseMemoryPagesToOS(from_page, to_page);
-      released_ranges_count++;
-      released_bytes += to_page - from_page;
-    }
-
-   private:
-    const ThisT& allocator;
-    const uptr region_base;
-    uptr released_ranges_count;
-    uptr released_bytes;
-  };
-
   // Attempts to release RAM occupied by freed chunks back to OS. The region is
   // expected to be locked.
   //
   // TODO(morehouse): Support a callback on memory release so HWASan can release
   // aliases as well.
-  void MaybeReleaseToOS(uptr class_id, bool force) {
+  void MaybeReleaseToOS(MemoryMapperT *memory_mapper, uptr class_id,
+                        bool force) {
     RegionInfo *region = GetRegionInfo(class_id);
     const uptr chunk_size = ClassIdToSize(class_id);
     const uptr page_size = GetPageSizeCached();
@@ -894,17 +881,16 @@ class SizeClassAllocator64 {
       }
     }
 
-    MemoryMapper memory_mapper(*this, class_id);
-
-    ReleaseFreeMemoryToOS<MemoryMapper>(
+    ReleaseFreeMemoryToOS(
         GetFreeArray(GetRegionBeginBySizeClass(class_id)), n, chunk_size,
-        RoundUpTo(region->allocated_user, page_size) / page_size,
-        &memory_mapper);
+        RoundUpTo(region->allocated_user, page_size) / page_size, memory_mapper,
+        class_id);
 
-    if (memory_mapper.GetReleasedRangesCount() > 0) {
+    uptr ranges, bytes;
+    if (memory_mapper->GetAndResetStats(ranges, bytes)) {
       region->rtoi.n_freed_at_last_release = region->stats.n_freed;
-      region->rtoi.num_releases += memory_mapper.GetReleasedRangesCount();
-      region->rtoi.last_released_bytes = memory_mapper.GetReleasedBytes();
+      region->rtoi.num_releases += ranges;
+      region->rtoi.last_released_bytes = bytes;
     }
     region->rtoi.last_release_at_ns = MonotonicNanoTime();
   }
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_secondary.h b/libsanitizer/sanitizer_common/sanitizer_allocator_secondary.h
index 61fb9874237..dd34fe85cc3 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_secondary.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_secondary.h
@@ -267,13 +267,9 @@ class LargeMmapAllocator {
 
   // ForceLock() and ForceUnlock() are needed to implement Darwin malloc zone
   // introspection API.
-  void ForceLock() {
-    mutex_.Lock();
-  }
+  void ForceLock() ACQUIRE(mutex_) { mutex_.Lock(); }
 
-  void ForceUnlock() {
-    mutex_.Unlock();
-  }
+  void ForceUnlock() RELEASE(mutex_) { mutex_.Unlock(); }
 
   // Iterate over all existing chunks.
   // The allocator must be locked when calling this function.
diff --git a/libsanitizer/sanitizer_common/sanitizer_common.cpp b/libsanitizer/sanitizer_common/sanitizer_common.cpp
index 33960d94a2f..5fae8e33b90 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_common.cpp
@@ -37,10 +37,9 @@ void NORETURN ReportMmapFailureAndDie(uptr size, const char *mem_type,
                                       const char *mmap_type, error_t err,
                                       bool raw_report) {
   static int recursion_count;
-  if (SANITIZER_RTEMS || raw_report || recursion_count) {
-    // If we are on RTEMS or raw report is requested or we went into recursion,
-    // just die.  The Report() and CHECK calls below may call mmap recursively
-    // and fail.
+  if (raw_report || recursion_count) {
+    // If raw report is requested or we went into recursion just die.  The
+    // Report() and CHECK calls below may call mmap recursively and fail.
     RawWrite("ERROR: Failed to mmap\n");
     Die();
   }
@@ -331,6 +330,14 @@ static int InstallMallocFreeHooks(void (*malloc_hook)(const void *, uptr),
   return 0;
 }
 
+void internal_sleep(unsigned seconds) {
+  internal_usleep((u64)seconds * 1000 * 1000);
+}
+void SleepForSeconds(unsigned seconds) {
+  internal_usleep((u64)seconds * 1000 * 1000);
+}
+void SleepForMillis(unsigned millis) { internal_usleep((u64)millis * 1000); }
+
 } // namespace __sanitizer
 
 using namespace __sanitizer;
diff --git a/libsanitizer/sanitizer_common/sanitizer_common.h b/libsanitizer/sanitizer_common/sanitizer_common.h
index 7b65dd7dfb8..cbdbb0c4c4b 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common.h
+++ b/libsanitizer/sanitizer_common/sanitizer_common.h
@@ -237,10 +237,16 @@ void SetPrintfAndReportCallback(void (*callback)(const char *));
 // Lock sanitizer error reporting and protects against nested errors.
 class ScopedErrorReportLock {
  public:
-  ScopedErrorReportLock();
-  ~ScopedErrorReportLock();
+  ScopedErrorReportLock() ACQUIRE(mutex_) { Lock(); }
+  ~ScopedErrorReportLock() RELEASE(mutex_) { Unlock(); }
 
-  static void CheckLocked();
+  static void Lock() ACQUIRE(mutex_);
+  static void Unlock() RELEASE(mutex_);
+  static void CheckLocked() CHECK_LOCKED(mutex_);
+
+ private:
+  static atomic_uintptr_t reporting_thread_;
+  static StaticSpinMutex mutex_;
 };
 
 extern uptr stoptheworld_tracer_pid;
@@ -288,8 +294,8 @@ void InitTlsSize();
 uptr GetTlsSize();
 
 // Other
-void SleepForSeconds(int seconds);
-void SleepForMillis(int millis);
+void SleepForSeconds(unsigned seconds);
+void SleepForMillis(unsigned millis);
 u64 NanoTime();
 u64 MonotonicNanoTime();
 int Atexit(void (*function)(void));
@@ -1057,6 +1063,13 @@ class ArrayRef {
   T *end_ = nullptr;
 };
 
+#define PRINTF_128(v)                                                         \
+  (*((u8 *)&v + 0)), (*((u8 *)&v + 1)), (*((u8 *)&v + 2)), (*((u8 *)&v + 3)), \
+      (*((u8 *)&v + 4)), (*((u8 *)&v + 5)), (*((u8 *)&v + 6)),                \
+      (*((u8 *)&v + 7)), (*((u8 *)&v + 8)), (*((u8 *)&v + 9)),                \
+      (*((u8 *)&v + 10)), (*((u8 *)&v + 11)), (*((u8 *)&v + 12)),             \
+      (*((u8 *)&v + 13)), (*((u8 *)&v + 14)), (*((u8 *)&v + 15))
+
 }  // namespace __sanitizer
 
 inline void *operator new(__sanitizer::operator_new_size_type size,
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc b/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
index 7867fccde39..6205d853a4c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
@@ -134,11 +134,11 @@ extern const short *_tolower_tab_;
 
 // Platform-specific options.
 #if SANITIZER_MAC
-#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE false
+#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE 0
 #elif SANITIZER_WINDOWS64
-#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE false
+#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE 0
 #else
-#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE true
+#define PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE 1
 #endif  // SANITIZER_MAC
 
 #ifndef COMMON_INTERCEPTOR_INITIALIZE_RANGE
@@ -823,11 +823,11 @@ INTERCEPTOR(void *, memcpy, void *dst, const void *src, uptr size) {
   // N.B.: If we switch this to internal_ we'll have to use internal_memmove
   // due to memcpy being an alias of memmove on OS X.
   void *ctx;
-  if (PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE) {
+#if PLATFORM_HAS_DIFFERENT_MEMCPY_AND_MEMMOVE
     COMMON_INTERCEPTOR_MEMCPY_IMPL(ctx, dst, src, size);
-  } else {
+#else
     COMMON_INTERCEPTOR_MEMMOVE_IMPL(ctx, dst, src, size);
-  }
+#endif
 }
 
 #define INIT_MEMCPY                                  \
@@ -957,6 +957,7 @@ INTERCEPTOR(double, frexp, double x, int *exp) {
   // Assuming frexp() always writes to |exp|.
   COMMON_INTERCEPTOR_WRITE_RANGE(ctx, exp, sizeof(*exp));
   double res = REAL(frexp)(x, exp);
+  COMMON_INTERCEPTOR_INITIALIZE_RANGE(exp, sizeof(*exp));
   return res;
 }
 
@@ -969,22 +970,18 @@ INTERCEPTOR(double, frexp, double x, int *exp) {
 INTERCEPTOR(float, frexpf, float x, int *exp) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, frexpf, x, exp);
-  // FIXME: under ASan the call below may write to freed memory and corrupt
-  // its metadata. See
-  // https://github.com/google/sanitizers/issues/321.
-  float res = REAL(frexpf)(x, exp);
   COMMON_INTERCEPTOR_WRITE_RANGE(ctx, exp, sizeof(*exp));
+  float res = REAL(frexpf)(x, exp);
+  COMMON_INTERCEPTOR_INITIALIZE_RANGE(exp, sizeof(*exp));
   return res;
 }
 
 INTERCEPTOR(long double, frexpl, long double x, int *exp) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, frexpl, x, exp);
-  // FIXME: under ASan the call below may write to freed memory and corrupt
-  // its metadata. See
-  // https://github.com/google/sanitizers/issues/321.
-  long double res = REAL(frexpl)(x, exp);
   COMMON_INTERCEPTOR_WRITE_RANGE(ctx, exp, sizeof(*exp));
+  long double res = REAL(frexpl)(x, exp);
+  COMMON_INTERCEPTOR_INITIALIZE_RANGE(exp, sizeof(*exp));
   return res;
 }
 
@@ -5303,6 +5300,12 @@ INTERCEPTOR(__sanitizer_clock_t, times, void *tms) {
 #define INIT_TIMES
 #endif
 
+#if SANITIZER_S390 && \
+    (SANITIZER_INTERCEPT_TLS_GET_ADDR || SANITIZER_INTERCEPT_TLS_GET_OFFSET)
+extern "C" uptr __tls_get_offset_wrapper(void *arg, uptr (*fn)(void *arg));
+DEFINE_REAL(uptr, __tls_get_offset, void *arg)
+#endif
+
 #if SANITIZER_INTERCEPT_TLS_GET_ADDR
 #if !SANITIZER_S390
 #define INIT_TLS_GET_ADDR COMMON_INTERCEPT_FUNCTION(__tls_get_addr)
@@ -5342,11 +5345,7 @@ void *__tls_get_addr_opt(void *arg);
 //   descriptor offset as an argument instead of a pointer.  GOT address
 //   is passed in r12, so it's necessary to write it in assembly.  This is
 //   the function used by the compiler.
-extern "C" uptr __tls_get_offset_wrapper(void *arg, uptr (*fn)(void *arg));
 #define INIT_TLS_GET_ADDR COMMON_INTERCEPT_FUNCTION(__tls_get_offset)
-DEFINE_REAL(uptr, __tls_get_offset, void *arg)
-extern "C" uptr __tls_get_offset(void *arg);
-extern "C" uptr __interceptor___tls_get_offset(void *arg);
 INTERCEPTOR(uptr, __tls_get_addr_internal, void *arg) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, __tls_get_addr_internal, arg);
@@ -5362,6 +5361,15 @@ INTERCEPTOR(uptr, __tls_get_addr_internal, void *arg) {
   }
   return res;
 }
+#endif // SANITIZER_S390
+#else
+#define INIT_TLS_GET_ADDR
+#endif
+
+#if SANITIZER_S390 && \
+    (SANITIZER_INTERCEPT_TLS_GET_ADDR || SANITIZER_INTERCEPT_TLS_GET_OFFSET)
+extern "C" uptr __tls_get_offset(void *arg);
+extern "C" uptr __interceptor___tls_get_offset(void *arg);
 // We need a hidden symbol aliasing the above, so that we can jump
 // directly to it from the assembly below.
 extern "C" __attribute__((alias("__interceptor___tls_get_addr_internal"),
@@ -5400,9 +5408,6 @@ asm(
   "br %r3\n"
   ".size __tls_get_offset_wrapper, .-__tls_get_offset_wrapper\n"
 );
-#endif // SANITIZER_S390
-#else
-#define INIT_TLS_GET_ADDR
 #endif
 
 #if SANITIZER_INTERCEPT_LISTXATTR
@@ -6099,6 +6104,40 @@ INTERCEPTOR(__sanitizer_FILE *, freopen, const char *path, const char *mode,
 #define INIT_FOPEN
 #endif
 
+#if SANITIZER_INTERCEPT_FLOPEN
+INTERCEPTOR(int, flopen, const char *path, int flags, ...) {
+  void *ctx;
+  va_list ap;
+  va_start(ap, flags);
+  u16 mode = static_cast<u16>(va_arg(ap, u32));
+  va_end(ap);
+  COMMON_INTERCEPTOR_ENTER(ctx, flopen, path, flags, mode);
+  if (path) {
+    COMMON_INTERCEPTOR_READ_RANGE(ctx, path, REAL(strlen)(path) + 1);
+  }
+  return REAL(flopen)(path, flags, mode);
+}
+
+INTERCEPTOR(int, flopenat, int dirfd, const char *path, int flags, ...) {
+  void *ctx;
+  va_list ap;
+  va_start(ap, flags);
+  u16 mode = static_cast<u16>(va_arg(ap, u32));
+  va_end(ap);
+  COMMON_INTERCEPTOR_ENTER(ctx, flopen, path, flags, mode);
+  if (path) {
+    COMMON_INTERCEPTOR_READ_RANGE(ctx, path, REAL(strlen)(path) + 1);
+  }
+  return REAL(flopenat)(dirfd, path, flags, mode);
+}
+
+#define INIT_FLOPEN    \
+  COMMON_INTERCEPT_FUNCTION(flopen); \
+  COMMON_INTERCEPT_FUNCTION(flopenat);
+#else
+#define INIT_FLOPEN
+#endif
+
 #if SANITIZER_INTERCEPT_FOPEN64
 INTERCEPTOR(__sanitizer_FILE *, fopen64, const char *path, const char *mode) {
   void *ctx;
@@ -6463,7 +6502,7 @@ INTERCEPTOR(int, sem_wait, __sanitizer_sem_t *s) {
 INTERCEPTOR(int, sem_trywait, __sanitizer_sem_t *s) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, sem_trywait, s);
-  int res = COMMON_INTERCEPTOR_BLOCK_REAL(sem_trywait)(s);
+  int res = REAL(sem_trywait)(s);
   if (res == 0) {
     COMMON_INTERCEPTOR_ACQUIRE(ctx, (uptr)s);
   }
@@ -10264,6 +10303,7 @@ static void InitializeCommonInterceptors() {
   INIT_LIBIO_INTERNALS;
   INIT_FOPEN;
   INIT_FOPEN64;
+  INIT_FLOPEN;
   INIT_OPEN_MEMSTREAM;
   INIT_OBSTACK;
   INIT_FFLUSH;
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
index 1037938f3d3..01ccacc6f32 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
@@ -138,7 +138,7 @@ uptr ReservedAddressRange::InitAligned(uptr size, uptr align,
   return start;
 }
 
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#if !SANITIZER_FUCHSIA
 
 // Reserve memory range [beg, end].
 // We need to use inclusive range because end+1 may not be representable.
@@ -189,7 +189,7 @@ void ProtectGap(uptr addr, uptr size, uptr zero_base_shadow_start,
   Die();
 }
 
-#endif  // !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
+#endif  // !SANITIZER_FUCHSIA
 
 }  // namespace __sanitizer
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_nolibc.cpp b/libsanitizer/sanitizer_common/sanitizer_common_nolibc.cpp
index 487a634a165..9a4e5388f24 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_nolibc.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_common_nolibc.cpp
@@ -25,7 +25,6 @@ void LogMessageOnPrintf(const char *str) {}
 #endif
 void WriteToSyslog(const char *buffer) {}
 void Abort() { internal__exit(1); }
-void SleepForSeconds(int seconds) { internal_sleep(seconds); }
 #endif // !SANITIZER_WINDOWS
 
 #if !SANITIZER_WINDOWS && !SANITIZER_MAC
diff --git a/libsanitizer/sanitizer_common/sanitizer_deadlock_detector1.cpp b/libsanitizer/sanitizer_common/sanitizer_deadlock_detector1.cpp
index 2c924f5d396..ccb7065b07a 100644
--- a/libsanitizer/sanitizer_common/sanitizer_deadlock_detector1.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_deadlock_detector1.cpp
@@ -136,7 +136,7 @@ void DD::ReportDeadlock(DDCallback *cb, DDMutex *m) {
     DDMutex *m0 = (DDMutex*)dd.getData(from);
     DDMutex *m1 = (DDMutex*)dd.getData(to);
 
-    u32 stk_from = -1U, stk_to = -1U;
+    u32 stk_from = 0, stk_to = 0;
     int unique_tid = 0;
     dd.findEdge(from, to, &stk_from, &stk_to, &unique_tid);
     // Printf("Edge: %zd=>%zd: %u/%u T%d\n", from, to, stk_from, stk_to,
diff --git a/libsanitizer/sanitizer_common/sanitizer_deadlock_detector2.cpp b/libsanitizer/sanitizer_common/sanitizer_deadlock_detector2.cpp
index e3f8e1b1276..1fbbbcccfa9 100644
--- a/libsanitizer/sanitizer_common/sanitizer_deadlock_detector2.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_deadlock_detector2.cpp
@@ -73,7 +73,7 @@ struct DDLogicalThread {
   int         nlocked;
 };
 
-struct Mutex {
+struct MutexState {
   StaticSpinMutex mtx;
   u32 seq;
   int nlink;
@@ -101,12 +101,12 @@ struct DD final : public DDetector {
   void CycleCheck(DDPhysicalThread *pt, DDLogicalThread *lt, DDMutex *mtx);
   void Report(DDPhysicalThread *pt, DDLogicalThread *lt, int npath);
   u32 allocateId(DDCallback *cb);
-  Mutex *getMutex(u32 id);
-  u32 getMutexId(Mutex *m);
+  MutexState *getMutex(u32 id);
+  u32 getMutexId(MutexState *m);
 
   DDFlags flags;
 
-  Mutex* mutex[kL1Size];
+  MutexState *mutex[kL1Size];
 
   SpinMutex mtx;
   InternalMmapVector<u32> free_id;
@@ -152,13 +152,11 @@ void DD::MutexInit(DDCallback *cb, DDMutex *m) {
   atomic_store(&m->owner, 0, memory_order_relaxed);
 }
 
-Mutex *DD::getMutex(u32 id) {
-  return &mutex[id / kL2Size][id % kL2Size];
-}
+MutexState *DD::getMutex(u32 id) { return &mutex[id / kL2Size][id % kL2Size]; }
 
-u32 DD::getMutexId(Mutex *m) {
+u32 DD::getMutexId(MutexState *m) {
   for (int i = 0; i < kL1Size; i++) {
-    Mutex *tab = mutex[i];
+    MutexState *tab = mutex[i];
     if (tab == 0)
       break;
     if (m >= tab && m < tab + kL2Size)
@@ -176,8 +174,8 @@ u32 DD::allocateId(DDCallback *cb) {
   } else {
     CHECK_LT(id_gen, kMaxMutex);
     if ((id_gen % kL2Size) == 0) {
-      mutex[id_gen / kL2Size] = (Mutex*)MmapOrDie(kL2Size * sizeof(Mutex),
-          "deadlock detector (mutex table)");
+      mutex[id_gen / kL2Size] = (MutexState *)MmapOrDie(
+          kL2Size * sizeof(MutexState), "deadlock detector (mutex table)");
     }
     id = id_gen++;
   }
@@ -216,11 +214,11 @@ void DD::MutexBeforeLock(DDCallback *cb, DDMutex *m, bool wlock) {
   }
 
   bool added = false;
-  Mutex *mtx = getMutex(m->id);
+  MutexState *mtx = getMutex(m->id);
   for (int i = 0; i < lt->nlocked - 1; i++) {
     u32 id1 = lt->locked[i].id;
     u32 stk1 = lt->locked[i].stk;
-    Mutex *mtx1 = getMutex(id1);
+    MutexState *mtx1 = getMutex(id1);
     SpinMutexLock l(&mtx1->mtx);
     if (mtx1->nlink == kMaxLink) {
       // FIXME(dvyukov): check stale links
@@ -342,7 +340,7 @@ void DD::MutexDestroy(DDCallback *cb, DDMutex *m) {
 
   // Clear and invalidate the mutex descriptor.
   {
-    Mutex *mtx = getMutex(m->id);
+    MutexState *mtx = getMutex(m->id);
     SpinMutexLock l(&mtx->mtx);
     mtx->seq++;
     mtx->nlink = 0;
@@ -361,7 +359,7 @@ void DD::CycleCheck(DDPhysicalThread *pt, DDLogicalThread *lt,
   int npath = 0;
   int npending = 0;
   {
-    Mutex *mtx = getMutex(m->id);
+    MutexState *mtx = getMutex(m->id);
     SpinMutexLock l(&mtx->mtx);
     for (int li = 0; li < mtx->nlink; li++)
       pt->pending[npending++] = mtx->link[li];
@@ -374,7 +372,7 @@ void DD::CycleCheck(DDPhysicalThread *pt, DDLogicalThread *lt,
     }
     if (pt->visited[link.id])
       continue;
-    Mutex *mtx1 = getMutex(link.id);
+    MutexState *mtx1 = getMutex(link.id);
     SpinMutexLock l(&mtx1->mtx);
     if (mtx1->seq != link.seq)
       continue;
@@ -387,7 +385,7 @@ void DD::CycleCheck(DDPhysicalThread *pt, DDLogicalThread *lt,
       return Report(pt, lt, npath);  // Bingo!
     for (int li = 0; li < mtx1->nlink; li++) {
       Link *link1 = &mtx1->link[li];
-      // Mutex *mtx2 = getMutex(link->id);
+      // MutexState *mtx2 = getMutex(link->id);
       // FIXME(dvyukov): fast seq check
       // FIXME(dvyukov): fast nlink != 0 check
       // FIXME(dvyukov): fast pending check?
diff --git a/libsanitizer/sanitizer_common/sanitizer_errno.h b/libsanitizer/sanitizer_common/sanitizer_errno.h
index 94f16b6e873..70a6e88dbaa 100644
--- a/libsanitizer/sanitizer_common/sanitizer_errno.h
+++ b/libsanitizer/sanitizer_common/sanitizer_errno.h
@@ -23,8 +23,7 @@
 
 #if SANITIZER_FREEBSD || SANITIZER_MAC
 #  define __errno_location __error
-#elif SANITIZER_ANDROID || SANITIZER_NETBSD || \
-  SANITIZER_RTEMS
+#elif SANITIZER_ANDROID || SANITIZER_NETBSD
 #  define __errno_location __errno
 #elif SANITIZER_SOLARIS
 #  define __errno_location ___errno
diff --git a/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp b/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
index 4f692f99c20..65bc398656c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
@@ -36,16 +36,11 @@ uptr internal_sched_yield() {
   return 0;  // Why doesn't this return void?
 }
 
-static void internal_nanosleep(zx_time_t ns) {
-  zx_status_t status = _zx_nanosleep(_zx_deadline_after(ns));
+void internal_usleep(u64 useconds) {
+  zx_status_t status = _zx_nanosleep(_zx_deadline_after(ZX_USEC(useconds)));
   CHECK_EQ(status, ZX_OK);
 }
 
-unsigned int internal_sleep(unsigned int seconds) {
-  internal_nanosleep(ZX_SEC(seconds));
-  return 0;
-}
-
 u64 NanoTime() {
   zx_handle_t utc_clock = _zx_utc_reference_get();
   CHECK_NE(utc_clock, ZX_HANDLE_INVALID);
@@ -78,10 +73,6 @@ void Abort() { abort(); }
 
 int Atexit(void (*function)(void)) { return atexit(function); }
 
-void SleepForSeconds(int seconds) { internal_sleep(seconds); }
-
-void SleepForMillis(int millis) { internal_nanosleep(ZX_MSEC(millis)); }
-
 void GetThreadStackTopAndBottom(bool, uptr *stack_top, uptr *stack_bottom) {
   pthread_attr_t attr;
   CHECK_EQ(pthread_getattr_np(pthread_self(), &attr), 0);
@@ -109,6 +100,18 @@ bool SignalContext::IsStackOverflow() const { return false; }
 void SignalContext::DumpAllRegisters(void *context) { UNIMPLEMENTED(); }
 const char *SignalContext::Describe() const { UNIMPLEMENTED(); }
 
+void FutexWait(atomic_uint32_t *p, u32 cmp) {
+  zx_status_t status = _zx_futex_wait(reinterpret_cast<zx_futex_t *>(p), cmp,
+                                      ZX_HANDLE_INVALID, ZX_TIME_INFINITE);
+  if (status != ZX_ERR_BAD_STATE)  // Normal race.
+    CHECK_EQ(status, ZX_OK);
+}
+
+void FutexWake(atomic_uint32_t *p, u32 count) {
+  zx_status_t status = _zx_futex_wake(reinterpret_cast<zx_futex_t *>(p), count);
+  CHECK_EQ(status, ZX_OK);
+}
+
 enum MutexState : int { MtxUnlocked = 0, MtxLocked = 1, MtxSleeping = 2 };
 
 BlockingMutex::BlockingMutex() {
@@ -145,8 +148,8 @@ void BlockingMutex::Unlock() {
   }
 }
 
-void BlockingMutex::CheckLocked() {
-  atomic_uint32_t *m = reinterpret_cast<atomic_uint32_t *>(&opaque_storage_);
+void BlockingMutex::CheckLocked() const {
+  auto m = reinterpret_cast<atomic_uint32_t const *>(&opaque_storage_);
   CHECK_NE(MtxUnlocked, atomic_load(m, memory_order_relaxed));
 }
 
@@ -156,8 +159,10 @@ uptr GetMmapGranularity() { return _zx_system_get_page_size(); }
 
 sanitizer_shadow_bounds_t ShadowBounds;
 
+void InitShadowBounds() { ShadowBounds = __sanitizer_shadow_bounds(); }
+
 uptr GetMaxUserVirtualAddress() {
-  ShadowBounds = __sanitizer_shadow_bounds();
+  InitShadowBounds();
   return ShadowBounds.memory_limit - 1;
 }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_fuchsia.h b/libsanitizer/sanitizer_common/sanitizer_fuchsia.h
index 96f9cde7ef1..26c1deab9e5 100644
--- a/libsanitizer/sanitizer_common/sanitizer_fuchsia.h
+++ b/libsanitizer/sanitizer_common/sanitizer_fuchsia.h
@@ -30,6 +30,8 @@ struct MemoryMappingLayoutData {
   size_t current;  // Current index into the vector.
 };
 
+void InitShadowBounds();
+
 }  // namespace __sanitizer
 
 #endif  // SANITIZER_FUCHSIA
diff --git a/libsanitizer/sanitizer_common/sanitizer_libc.h b/libsanitizer/sanitizer_common/sanitizer_libc.h
index ec0a6ded009..bcb81ebbc80 100644
--- a/libsanitizer/sanitizer_common/sanitizer_libc.h
+++ b/libsanitizer/sanitizer_common/sanitizer_libc.h
@@ -67,7 +67,8 @@ uptr internal_ftruncate(fd_t fd, uptr size);
 
 // OS
 void NORETURN internal__exit(int exitcode);
-unsigned int internal_sleep(unsigned int seconds);
+void internal_sleep(unsigned seconds);
+void internal_usleep(u64 useconds);
 
 uptr internal_getpid();
 uptr internal_getppid();
diff --git a/libsanitizer/sanitizer_common/sanitizer_libignore.cpp b/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
index a65d3d896e3..431efc574fa 100644
--- a/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
@@ -84,6 +84,7 @@ void LibIgnore::OnLibraryLoaded(const char *name) {
         ignored_code_ranges_[idx].begin = range.beg;
         ignored_code_ranges_[idx].end = range.end;
         atomic_store(&ignored_ranges_count_, idx + 1, memory_order_release);
+        atomic_store(&enabled_, 1, memory_order_release);
         break;
       }
     }
@@ -114,6 +115,7 @@ void LibIgnore::OnLibraryLoaded(const char *name) {
         instrumented_code_ranges_[idx].end = range.end;
         atomic_store(&instrumented_ranges_count_, idx + 1,
                      memory_order_release);
+        atomic_store(&enabled_, 1, memory_order_release);
       }
     }
   }
@@ -123,6 +125,29 @@ void LibIgnore::OnLibraryUnloaded() {
   OnLibraryLoaded(nullptr);
 }
 
+bool LibIgnore::IsIgnoredSlow(uptr pc, bool *pc_in_ignored_lib) const {
+  const uptr n = atomic_load(&ignored_ranges_count_, memory_order_acquire);
+  for (uptr i = 0; i < n; i++) {
+    if (IsInRange(pc, ignored_code_ranges_[i])) {
+      *pc_in_ignored_lib = true;
+      return true;
+    }
+  }
+  *pc_in_ignored_lib = false;
+  if (track_instrumented_libs_ && !IsPcInstrumented(pc))
+    return true;
+  return false;
+}
+
+bool LibIgnore::IsPcInstrumented(uptr pc) const {
+  const uptr n = atomic_load(&instrumented_ranges_count_, memory_order_acquire);
+  for (uptr i = 0; i < n; i++) {
+    if (IsInRange(pc, instrumented_code_ranges_[i]))
+      return true;
+  }
+  return false;
+}
+
 } // namespace __sanitizer
 
 #endif  // SANITIZER_FREEBSD || SANITIZER_LINUX || SANITIZER_MAC ||
diff --git a/libsanitizer/sanitizer_common/sanitizer_libignore.h b/libsanitizer/sanitizer_common/sanitizer_libignore.h
index 256f685979f..85452e57ba3 100644
--- a/libsanitizer/sanitizer_common/sanitizer_libignore.h
+++ b/libsanitizer/sanitizer_common/sanitizer_libignore.h
@@ -45,9 +45,6 @@ class LibIgnore {
   // "pc_in_ignored_lib" if the PC is in an ignored library, false otherwise.
   bool IsIgnored(uptr pc, bool *pc_in_ignored_lib) const;
 
-  // Checks whether the provided PC belongs to an instrumented module.
-  bool IsPcInstrumented(uptr pc) const;
-
  private:
   struct Lib {
     char *templ;
@@ -61,6 +58,10 @@ class LibIgnore {
     uptr end;
   };
 
+  // Checks whether the provided PC belongs to an instrumented module.
+  bool IsPcInstrumented(uptr pc) const;
+  bool IsIgnoredSlow(uptr pc, bool *pc_in_ignored_lib) const;
+
   inline bool IsInRange(uptr pc, const LibCodeRange &range) const {
     return (pc >= range.begin && pc < range.end);
   }
@@ -70,6 +71,8 @@ class LibIgnore {
   static const uptr kMaxLibs = 1024;
 
   // Hot part:
+  atomic_uintptr_t enabled_;
+
   atomic_uintptr_t ignored_ranges_count_;
   LibCodeRange ignored_code_ranges_[kMaxIgnoredRanges];
 
@@ -87,27 +90,11 @@ class LibIgnore {
   void operator = (const LibIgnore&);  // not implemented
 };
 
-inline bool LibIgnore::IsIgnored(uptr pc, bool *pc_in_ignored_lib) const {
-  const uptr n = atomic_load(&ignored_ranges_count_, memory_order_acquire);
-  for (uptr i = 0; i < n; i++) {
-    if (IsInRange(pc, ignored_code_ranges_[i])) {
-      *pc_in_ignored_lib = true;
-      return true;
-    }
-  }
-  *pc_in_ignored_lib = false;
-  if (track_instrumented_libs_ && !IsPcInstrumented(pc))
-    return true;
-  return false;
-}
-
-inline bool LibIgnore::IsPcInstrumented(uptr pc) const {
-  const uptr n = atomic_load(&instrumented_ranges_count_, memory_order_acquire);
-  for (uptr i = 0; i < n; i++) {
-    if (IsInRange(pc, instrumented_code_ranges_[i]))
-      return true;
-  }
-  return false;
+ALWAYS_INLINE
+bool LibIgnore::IsIgnored(uptr pc, bool *pc_in_ignored_lib) const {
+  if (LIKELY(atomic_load(&enabled_, memory_order_acquire) == 0))
+    return false;
+  return IsIgnoredSlow(pc, pc_in_ignored_lib);
 }
 
 }  // namespace __sanitizer
diff --git a/libsanitizer/sanitizer_common/sanitizer_linux.cpp b/libsanitizer/sanitizer_common/sanitizer_linux.cpp
index b371477755f..9b7d87eb85e 100644
--- a/libsanitizer/sanitizer_common/sanitizer_linux.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_linux.cpp
@@ -430,13 +430,11 @@ uptr internal_sched_yield() {
   return internal_syscall(SYSCALL(sched_yield));
 }
 
-unsigned int internal_sleep(unsigned int seconds) {
+void internal_usleep(u64 useconds) {
   struct timespec ts;
-  ts.tv_sec = seconds;
-  ts.tv_nsec = 0;
-  int res = internal_syscall(SYSCALL(nanosleep), &ts, &ts);
-  if (res) return ts.tv_sec;
-  return 0;
+  ts.tv_sec = useconds / 1000000;
+  ts.tv_nsec = (useconds % 1000000) * 1000;
+  internal_syscall(SYSCALL(nanosleep), &ts, &ts);
 }
 
 uptr internal_execve(const char *filename, char *const argv[],
@@ -641,11 +639,27 @@ char **GetEnviron() {
 }
 
 #if !SANITIZER_SOLARIS
-enum MutexState {
-  MtxUnlocked = 0,
-  MtxLocked = 1,
-  MtxSleeping = 2
-};
+void FutexWait(atomic_uint32_t *p, u32 cmp) {
+#    if SANITIZER_FREEBSD
+  _umtx_op(p, UMTX_OP_WAIT_UINT, cmp, 0, 0);
+#    elif SANITIZER_NETBSD
+  sched_yield();   /* No userspace futex-like synchronization */
+#    else
+  internal_syscall(SYSCALL(futex), (uptr)p, FUTEX_WAIT_PRIVATE, cmp, 0, 0, 0);
+#    endif
+}
+
+void FutexWake(atomic_uint32_t *p, u32 count) {
+#    if SANITIZER_FREEBSD
+  _umtx_op(p, UMTX_OP_WAKE, count, 0, 0);
+#    elif SANITIZER_NETBSD
+                   /* No userspace futex-like synchronization */
+#    else
+  internal_syscall(SYSCALL(futex), (uptr)p, FUTEX_WAKE_PRIVATE, count, 0, 0, 0);
+#    endif
+}
+
+enum { MtxUnlocked = 0, MtxLocked = 1, MtxSleeping = 2 };
 
 BlockingMutex::BlockingMutex() {
   internal_memset(this, 0, sizeof(*this));
@@ -683,11 +697,11 @@ void BlockingMutex::Unlock() {
   }
 }
 
-void BlockingMutex::CheckLocked() {
-  atomic_uint32_t *m = reinterpret_cast<atomic_uint32_t *>(&opaque_storage_);
+void BlockingMutex::CheckLocked() const {
+  auto m = reinterpret_cast<atomic_uint32_t const *>(&opaque_storage_);
   CHECK_NE(MtxUnlocked, atomic_load(m, memory_order_relaxed));
 }
-#endif // !SANITIZER_SOLARIS
+#  endif  // !SANITIZER_SOLARIS
 
 // ----------------- sanitizer_linux.h
 // The actual size of this structure is specified by d_reclen.
@@ -884,7 +898,7 @@ void internal_sigdelset(__sanitizer_sigset_t *set, int signum) {
   __sanitizer_kernel_sigset_t *k_set = (__sanitizer_kernel_sigset_t *)set;
   const uptr idx = signum / (sizeof(k_set->sig[0]) * 8);
   const uptr bit = signum % (sizeof(k_set->sig[0]) * 8);
-  k_set->sig[idx] &= ~(1 << bit);
+  k_set->sig[idx] &= ~((uptr)1 << bit);
 }
 
 bool internal_sigismember(__sanitizer_sigset_t *set, int signum) {
@@ -894,7 +908,7 @@ bool internal_sigismember(__sanitizer_sigset_t *set, int signum) {
   __sanitizer_kernel_sigset_t *k_set = (__sanitizer_kernel_sigset_t *)set;
   const uptr idx = signum / (sizeof(k_set->sig[0]) * 8);
   const uptr bit = signum % (sizeof(k_set->sig[0]) * 8);
-  return k_set->sig[idx] & (1 << bit);
+  return k_set->sig[idx] & ((uptr)1 << bit);
 }
 #elif SANITIZER_FREEBSD
 void internal_sigdelset(__sanitizer_sigset_t *set, int signum) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
index 4f9577a97e2..7ce9e25da34 100644
--- a/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
@@ -203,7 +203,7 @@ void InitTlsSize() {
   g_use_dlpi_tls_data =
       GetLibcVersion(&major, &minor, &patch) && major == 2 && minor >= 25;
 
-#if defined(__x86_64__) || defined(__powerpc64__)
+#if defined(__aarch64__) || defined(__x86_64__) || defined(__powerpc64__)
   void *get_tls_static_info = dlsym(RTLD_NEXT, "_dl_get_tls_static_info");
   size_t tls_align;
   ((void (*)(size_t *, size_t *))get_tls_static_info)(&g_tls_size, &tls_align);
@@ -317,21 +317,44 @@ struct TlsBlock {
 };
 }  // namespace
 
+#ifdef __s390__
+extern "C" uptr __tls_get_offset(void *arg);
+
+static uptr TlsGetOffset(uptr ti_module, uptr ti_offset) {
+  // The __tls_get_offset ABI requires %r12 to point to GOT and %r2 to be an
+  // offset of a struct tls_index inside GOT. We don't possess either of the
+  // two, so violate the letter of the "ELF Handling For Thread-Local
+  // Storage" document and assume that the implementation just dereferences
+  // %r2 + %r12.
+  uptr tls_index[2] = {ti_module, ti_offset};
+  register uptr r2 asm("2") = 0;
+  register void *r12 asm("12") = tls_index;
+  asm("basr %%r14, %[__tls_get_offset]"
+      : "+r"(r2)
+      : [__tls_get_offset] "r"(__tls_get_offset), "r"(r12)
+      : "memory", "cc", "0", "1", "3", "4", "5", "14");
+  return r2;
+}
+#else
 extern "C" void *__tls_get_addr(size_t *);
+#endif
 
 static int CollectStaticTlsBlocks(struct dl_phdr_info *info, size_t size,
                                   void *data) {
   if (!info->dlpi_tls_modid)
     return 0;
   uptr begin = (uptr)info->dlpi_tls_data;
-#ifndef __s390__
   if (!g_use_dlpi_tls_data) {
     // Call __tls_get_addr as a fallback. This forces TLS allocation on glibc
     // and FreeBSD.
+#ifdef __s390__
+    begin = (uptr)__builtin_thread_pointer() +
+            TlsGetOffset(info->dlpi_tls_modid, 0);
+#else
     size_t mod_and_off[2] = {info->dlpi_tls_modid, 0};
     begin = (uptr)__tls_get_addr(mod_and_off);
-  }
 #endif
+  }
   for (unsigned i = 0; i != info->dlpi_phnum; ++i)
     if (info->dlpi_phdr[i].p_type == PT_TLS) {
       static_cast<InternalMmapVector<TlsBlock> *>(data)->push_back(
@@ -427,12 +450,16 @@ static void GetTls(uptr *addr, uptr *size) {
     *size = 0;
   }
 #elif SANITIZER_GLIBC && defined(__x86_64__)
-  // For x86-64, use an O(1) approach which requires precise
-  // ThreadDescriptorSize. g_tls_size was initialized in InitTlsSize.
+  // For aarch64 and x86-64, use an O(1) approach which requires relatively
+  // precise ThreadDescriptorSize. g_tls_size was initialized in InitTlsSize.
   asm("mov %%fs:16,%0" : "=r"(*addr));
   *size = g_tls_size;
   *addr -= *size;
   *addr += ThreadDescriptorSize();
+#elif SANITIZER_GLIBC && defined(__aarch64__)
+  *addr = reinterpret_cast<uptr>(__builtin_thread_pointer()) -
+          ThreadDescriptorSize();
+  *size = g_tls_size + ThreadDescriptorSize();
 #elif SANITIZER_GLIBC && defined(__powerpc64__)
   // Workaround for glibc<2.25(?). 2.27 is known to not need this.
   uptr tp;
@@ -732,13 +759,9 @@ u32 GetNumberOfCPUs() {
 #elif SANITIZER_SOLARIS
   return sysconf(_SC_NPROCESSORS_ONLN);
 #else
-#if defined(CPU_COUNT)
   cpu_set_t CPUs;
   CHECK_EQ(sched_getaffinity(0, sizeof(cpu_set_t), &CPUs), 0);
   return CPU_COUNT(&CPUs);
-#else
-  return 1;
-#endif
 #endif
 }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_mac.cpp b/libsanitizer/sanitizer_common/sanitizer_mac.cpp
index 30a94fcba14..125ecac8b12 100644
--- a/libsanitizer/sanitizer_common/sanitizer_mac.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_mac.cpp
@@ -37,7 +37,7 @@
 extern char **environ;
 #endif
 
-#if defined(__has_include) && __has_include(<os/trace.h>) && defined(__BLOCKS__)
+#if defined(__has_include) && __has_include(<os/trace.h>)
 #define SANITIZER_OS_TRACE 1
 #include <os/trace.h>
 #else
@@ -70,15 +70,7 @@ extern "C" {
 #include <mach/mach_time.h>
 #include <mach/vm_statistics.h>
 #include <malloc/malloc.h>
-#if defined(__has_builtin) && __has_builtin(__builtin_os_log_format)
-# include <os/log.h>
-#else
-   /* Without support for __builtin_os_log_format, fall back to the older
-      method.  */
-# define OS_LOG_DEFAULT 0
-# define os_log_error(A,B,C) \
-  asl_log(nullptr, nullptr, ASL_LEVEL_ERR, "%s", (C));
-#endif
+#include <os/log.h>
 #include <pthread.h>
 #include <sched.h>
 #include <signal.h>
@@ -227,9 +219,7 @@ void internal__exit(int exitcode) {
   _exit(exitcode);
 }
 
-unsigned int internal_sleep(unsigned int seconds) {
-  return sleep(seconds);
-}
+void internal_usleep(u64 useconds) { usleep(useconds); }
 
 uptr internal_getpid() {
   return getpid();
@@ -519,6 +509,13 @@ void MprotectMallocZones(void *addr, int prot) {
   }
 }
 
+void FutexWait(atomic_uint32_t *p, u32 cmp) {
+  // FIXME: implement actual blocking.
+  sched_yield();
+}
+
+void FutexWake(atomic_uint32_t *p, u32 count) {}
+
 BlockingMutex::BlockingMutex() {
   internal_memset(this, 0, sizeof(*this));
 }
@@ -534,7 +531,7 @@ void BlockingMutex::Unlock() {
   OSSpinLockUnlock((OSSpinLock*)&opaque_storage_);
 }
 
-void BlockingMutex::CheckLocked() {
+void BlockingMutex::CheckLocked() const {
   CHECK_NE(*(OSSpinLock*)&opaque_storage_, 0);
 }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_mac.h b/libsanitizer/sanitizer_common/sanitizer_mac.h
index 96a5986a47a..0b6af5a3c0e 100644
--- a/libsanitizer/sanitizer_common/sanitizer_mac.h
+++ b/libsanitizer/sanitizer_common/sanitizer_mac.h
@@ -14,26 +14,6 @@
 
 #include "sanitizer_common.h"
 #include "sanitizer_platform.h"
-
-/* TARGET_OS_OSX is not present in SDKs before Darwin16 (macOS 10.12) use
-   TARGET_OS_MAC (we have no support for iOS in any form for these versions,
-   so there's no ambiguity).  */
-#if !defined(TARGET_OS_OSX) && TARGET_OS_MAC
-# define TARGET_OS_OSX 1
-#endif
-
-/* Other TARGET_OS_xxx are not present on earlier versions, define them to
-   0 (we have no support for them; they are not valid targets anyway).  */
-#ifndef TARGET_OS_IOS
-#define TARGET_OS_IOS 0
-#endif
-#ifndef TARGET_OS_TV
-#define TARGET_OS_TV 0
-#endif
-#ifndef TARGET_OS_WATCH
-#define TARGET_OS_WATCH 0
-#endif
-
 #if SANITIZER_MAC
 #include "sanitizer_posix.h"
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_mutex.cpp b/libsanitizer/sanitizer_common/sanitizer_mutex.cpp
new file mode 100644
index 00000000000..bc2d83c42c1
--- /dev/null
+++ b/libsanitizer/sanitizer_common/sanitizer_mutex.cpp
@@ -0,0 +1,39 @@
+//===-- sanitizer_mutex.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 shared between AddressSanitizer and ThreadSanitizer
+// run-time libraries.
+//===----------------------------------------------------------------------===//
+
+#include "sanitizer_mutex.h"
+
+#include "sanitizer_common.h"
+
+namespace __sanitizer {
+
+void Semaphore::Wait() {
+  u32 count = atomic_load(&state_, memory_order_relaxed);
+  for (;;) {
+    if (count == 0) {
+      FutexWait(&state_, 0);
+      count = atomic_load(&state_, memory_order_relaxed);
+      continue;
+    }
+    if (atomic_compare_exchange_weak(&state_, &count, count - 1,
+                                     memory_order_acquire))
+      break;
+  }
+}
+
+void Semaphore::Post(u32 count) {
+  CHECK_NE(count, 0);
+  atomic_fetch_add(&state_, count, memory_order_release);
+  FutexWake(&state_, count);
+}
+
+}  // namespace __sanitizer
diff --git a/libsanitizer/sanitizer_common/sanitizer_mutex.h b/libsanitizer/sanitizer_common/sanitizer_mutex.h
index 40a65914299..e3ff650b2c5 100644
--- a/libsanitizer/sanitizer_common/sanitizer_mutex.h
+++ b/libsanitizer/sanitizer_common/sanitizer_mutex.h
@@ -16,30 +16,29 @@
 #include "sanitizer_atomic.h"
 #include "sanitizer_internal_defs.h"
 #include "sanitizer_libc.h"
+#include "sanitizer_thread_safety.h"
 
 namespace __sanitizer {
 
-class StaticSpinMutex {
+class MUTEX StaticSpinMutex {
  public:
   void Init() {
     atomic_store(&state_, 0, memory_order_relaxed);
   }
 
-  void Lock() {
+  void Lock() ACQUIRE() {
     if (TryLock())
       return;
     LockSlow();
   }
 
-  bool TryLock() {
+  bool TryLock() TRY_ACQUIRE(true) {
     return atomic_exchange(&state_, 1, memory_order_acquire) == 0;
   }
 
-  void Unlock() {
-    atomic_store(&state_, 0, memory_order_release);
-  }
+  void Unlock() RELEASE() { atomic_store(&state_, 0, memory_order_release); }
 
-  void CheckLocked() {
+  void CheckLocked() const CHECK_LOCKED() {
     CHECK_EQ(atomic_load(&state_, memory_order_relaxed), 1);
   }
 
@@ -59,24 +58,223 @@ class StaticSpinMutex {
   }
 };
 
-class SpinMutex : public StaticSpinMutex {
+class MUTEX SpinMutex : public StaticSpinMutex {
  public:
   SpinMutex() {
     Init();
   }
 
  private:
-  SpinMutex(const SpinMutex&);
-  void operator=(const SpinMutex&);
+  SpinMutex(const SpinMutex &) = delete;
+  void operator=(const SpinMutex &) = delete;
 };
 
-class BlockingMutex {
+// Semaphore provides an OS-dependent way to park/unpark threads.
+// The last thread returned from Wait can destroy the object
+// (destruction-safety).
+class Semaphore {
+ public:
+  constexpr Semaphore() {}
+  Semaphore(const Semaphore &) = delete;
+  void operator=(const Semaphore &) = delete;
+
+  void Wait();
+  void Post(u32 count = 1);
+
+ private:
+  atomic_uint32_t state_ = {0};
+};
+
+// Reader-writer mutex.
+class MUTEX Mutex2 {
+ public:
+  constexpr Mutex2() {}
+
+  void Lock() ACQUIRE() {
+    u64 reset_mask = ~0ull;
+    u64 state = atomic_load_relaxed(&state_);
+    const uptr kMaxSpinIters = 1500;
+    for (uptr spin_iters = 0;; spin_iters++) {
+      u64 new_state;
+      bool locked = (state & (kWriterLock | kReaderLockMask)) != 0;
+      if (LIKELY(!locked)) {
+        // The mutex is not read-/write-locked, try to lock.
+        new_state = (state | kWriterLock) & reset_mask;
+      } else if (spin_iters > kMaxSpinIters) {
+        // We've spun enough, increment waiting writers count and block.
+        // The counter will be decremented by whoever wakes us.
+        new_state = (state + kWaitingWriterInc) & reset_mask;
+      } else if ((state & kWriterSpinWait) == 0) {
+        // Active spinning, but denote our presence so that unlocking
+        // thread does not wake up other threads.
+        new_state = state | kWriterSpinWait;
+      } else {
+        // Active spinning.
+        state = atomic_load(&state_, memory_order_relaxed);
+        continue;
+      }
+      if (UNLIKELY(!atomic_compare_exchange_weak(&state_, &state, new_state,
+                                                 memory_order_acquire)))
+        continue;
+      if (LIKELY(!locked))
+        return;  // We've locked the mutex.
+      if (spin_iters > kMaxSpinIters) {
+        // We've incremented waiting writers, so now block.
+        writers_.Wait();
+        spin_iters = 0;
+        state = atomic_load(&state_, memory_order_relaxed);
+        DCHECK_NE(state & kWriterSpinWait, 0);
+      } else {
+        // We've set kWriterSpinWait, but we are still in active spinning.
+      }
+      // We either blocked and were unblocked,
+      // or we just spun but set kWriterSpinWait.
+      // Either way we need to reset kWriterSpinWait
+      // next time we take the lock or block again.
+      reset_mask = ~kWriterSpinWait;
+    }
+  }
+
+  void Unlock() RELEASE() {
+    bool wake_writer;
+    u64 wake_readers;
+    u64 new_state;
+    u64 state = atomic_load_relaxed(&state_);
+    do {
+      DCHECK_NE(state & kWriterLock, 0);
+      DCHECK_EQ(state & kReaderLockMask, 0);
+      new_state = state & ~kWriterLock;
+      wake_writer =
+          (state & kWriterSpinWait) == 0 && (state & kWaitingWriterMask) != 0;
+      if (wake_writer)
+        new_state = (new_state - kWaitingWriterInc) | kWriterSpinWait;
+      wake_readers =
+          (state & (kWriterSpinWait | kWaitingWriterMask)) != 0
+              ? 0
+              : ((state & kWaitingReaderMask) >> kWaitingReaderShift);
+      if (wake_readers)
+        new_state = (new_state & ~kWaitingReaderMask) +
+                    (wake_readers << kReaderLockShift);
+    } while (UNLIKELY(!atomic_compare_exchange_weak(&state_, &state, new_state,
+                                                    memory_order_release)));
+    if (UNLIKELY(wake_writer))
+      writers_.Post();
+    else if (UNLIKELY(wake_readers))
+      readers_.Post(wake_readers);
+  }
+
+  void ReadLock() ACQUIRE_SHARED() {
+    bool locked;
+    u64 new_state;
+    u64 state = atomic_load_relaxed(&state_);
+    do {
+      locked =
+          (state & kReaderLockMask) == 0 &&
+          (state & (kWriterLock | kWriterSpinWait | kWaitingWriterMask)) != 0;
+      if (LIKELY(!locked))
+        new_state = state + kReaderLockInc;
+      else
+        new_state = state + kWaitingReaderInc;
+    } while (UNLIKELY(!atomic_compare_exchange_weak(&state_, &state, new_state,
+                                                    memory_order_acquire)));
+    if (UNLIKELY(locked))
+      readers_.Wait();
+    DCHECK_EQ(atomic_load_relaxed(&state_) & kWriterLock, 0);
+    DCHECK_NE(atomic_load_relaxed(&state_) & kReaderLockMask, 0);
+  }
+
+  void ReadUnlock() RELEASE_SHARED() {
+    bool wake;
+    u64 new_state;
+    u64 state = atomic_load_relaxed(&state_);
+    do {
+      DCHECK_NE(state & kReaderLockMask, 0);
+      DCHECK_EQ(state & (kWaitingReaderMask | kWriterLock), 0);
+      new_state = state - kReaderLockInc;
+      wake = (new_state & (kReaderLockMask | kWriterSpinWait)) == 0 &&
+             (new_state & kWaitingWriterMask) != 0;
+      if (wake)
+        new_state = (new_state - kWaitingWriterInc) | kWriterSpinWait;
+    } while (UNLIKELY(!atomic_compare_exchange_weak(&state_, &state, new_state,
+                                                    memory_order_release)));
+    if (UNLIKELY(wake))
+      writers_.Post();
+  }
+
+  // This function does not guarantee an explicit check that the calling thread
+  // is the thread which owns the mutex. This behavior, while more strictly
+  // correct, causes problems in cases like StopTheWorld, where a parent thread
+  // owns the mutex but a child checks that it is locked. Rather than
+  // maintaining complex state to work around those situations, the check only
+  // checks that the mutex is owned.
+  void CheckWriteLocked() const CHECK_LOCKED() {
+    CHECK(atomic_load(&state_, memory_order_relaxed) & kWriterLock);
+  }
+
+  void CheckLocked() const CHECK_LOCKED() { CheckWriteLocked(); }
+
+  void CheckReadLocked() const CHECK_LOCKED() {
+    CHECK(atomic_load(&state_, memory_order_relaxed) & kReaderLockMask);
+  }
+
+ private:
+  atomic_uint64_t state_ = {0};
+  Semaphore writers_;
+  Semaphore readers_;
+
+  // The state has 3 counters:
+  //  - number of readers holding the lock,
+  //    if non zero, the mutex is read-locked
+  //  - number of waiting readers,
+  //    if not zero, the mutex is write-locked
+  //  - number of waiting writers,
+  //    if non zero, the mutex is read- or write-locked
+  // And 2 flags:
+  //  - writer lock
+  //    if set, the mutex is write-locked
+  //  - a writer is awake and spin-waiting
+  //    the flag is used to prevent thundering herd problem
+  //    (new writers are not woken if this flag is set)
+  //
+  // Writer support active spinning, readers does not.
+  // But readers are more aggressive and always take the mutex
+  // if there are any other readers.
+  // Writers hand off the mutex to readers: after wake up readers
+  // already assume ownership of the mutex (don't need to do any
+  // state updates). But the mutex is not handed off to writers,
+  // after wake up writers compete to lock the mutex again.
+  // This is needed to allow repeated write locks even in presence
+  // of other blocked writers.
+  static constexpr u64 kCounterWidth = 20;
+  static constexpr u64 kReaderLockShift = 0;
+  static constexpr u64 kReaderLockInc = 1ull << kReaderLockShift;
+  static constexpr u64 kReaderLockMask = ((1ull << kCounterWidth) - 1)
+                                         << kReaderLockShift;
+  static constexpr u64 kWaitingReaderShift = kCounterWidth;
+  static constexpr u64 kWaitingReaderInc = 1ull << kWaitingReaderShift;
+  static constexpr u64 kWaitingReaderMask = ((1ull << kCounterWidth) - 1)
+                                            << kWaitingReaderShift;
+  static constexpr u64 kWaitingWriterShift = 2 * kCounterWidth;
+  static constexpr u64 kWaitingWriterInc = 1ull << kWaitingWriterShift;
+  static constexpr u64 kWaitingWriterMask = ((1ull << kCounterWidth) - 1)
+                                            << kWaitingWriterShift;
+  static constexpr u64 kWriterLock = 1ull << (3 * kCounterWidth);
+  static constexpr u64 kWriterSpinWait = 1ull << (3 * kCounterWidth + 1);
+
+  Mutex2(const Mutex2 &) = delete;
+  void operator=(const Mutex2 &) = delete;
+};
+
+void FutexWait(atomic_uint32_t *p, u32 cmp);
+void FutexWake(atomic_uint32_t *p, u32 count);
+
+class MUTEX BlockingMutex {
  public:
   explicit constexpr BlockingMutex(LinkerInitialized)
       : opaque_storage_ {0, }, owner_ {0} {}
   BlockingMutex();
-  void Lock();
-  void Unlock();
+  void Lock() ACQUIRE();
+  void Unlock() RELEASE();
 
   // This function does not guarantee an explicit check that the calling thread
   // is the thread which owns the mutex. This behavior, while more strictly
@@ -85,7 +283,7 @@ class BlockingMutex {
   // maintaining complex state to work around those situations, the check only
   // checks that the mutex is owned, and assumes callers to be generally
   // well-behaved.
-  void CheckLocked();
+  void CheckLocked() const CHECK_LOCKED();
 
  private:
   // Solaris mutex_t has a member that requires 64-bit alignment.
@@ -94,7 +292,7 @@ class BlockingMutex {
 };
 
 // Reader-writer spin mutex.
-class RWMutex {
+class MUTEX RWMutex {
  public:
   RWMutex() {
     atomic_store(&state_, kUnlocked, memory_order_relaxed);
@@ -104,7 +302,7 @@ class RWMutex {
     CHECK_EQ(atomic_load(&state_, memory_order_relaxed), kUnlocked);
   }
 
-  void Lock() {
+  void Lock() ACQUIRE() {
     u32 cmp = kUnlocked;
     if (atomic_compare_exchange_strong(&state_, &cmp, kWriteLock,
                                        memory_order_acquire))
@@ -112,27 +310,27 @@ class RWMutex {
     LockSlow();
   }
 
-  void Unlock() {
+  void Unlock() RELEASE() {
     u32 prev = atomic_fetch_sub(&state_, kWriteLock, memory_order_release);
     DCHECK_NE(prev & kWriteLock, 0);
     (void)prev;
   }
 
-  void ReadLock() {
+  void ReadLock() ACQUIRE_SHARED() {
     u32 prev = atomic_fetch_add(&state_, kReadLock, memory_order_acquire);
     if ((prev & kWriteLock) == 0)
       return;
     ReadLockSlow();
   }
 
-  void ReadUnlock() {
+  void ReadUnlock() RELEASE_SHARED() {
     u32 prev = atomic_fetch_sub(&state_, kReadLock, memory_order_release);
     DCHECK_EQ(prev & kWriteLock, 0);
     DCHECK_GT(prev & ~kWriteLock, 0);
     (void)prev;
   }
 
-  void CheckLocked() {
+  void CheckLocked() const CHECK_LOCKED() {
     CHECK_NE(atomic_load(&state_, memory_order_relaxed), kUnlocked);
   }
 
@@ -171,46 +369,40 @@ class RWMutex {
     }
   }
 
-  RWMutex(const RWMutex&);
-  void operator = (const RWMutex&);
+  RWMutex(const RWMutex &) = delete;
+  void operator=(const RWMutex &) = delete;
 };
 
-template<typename MutexType>
-class GenericScopedLock {
+template <typename MutexType>
+class SCOPED_LOCK GenericScopedLock {
  public:
-  explicit GenericScopedLock(MutexType *mu)
-      : mu_(mu) {
+  explicit GenericScopedLock(MutexType *mu) ACQUIRE(mu) : mu_(mu) {
     mu_->Lock();
   }
 
-  ~GenericScopedLock() {
-    mu_->Unlock();
-  }
+  ~GenericScopedLock() RELEASE() { mu_->Unlock(); }
 
  private:
   MutexType *mu_;
 
-  GenericScopedLock(const GenericScopedLock&);
-  void operator=(const GenericScopedLock&);
+  GenericScopedLock(const GenericScopedLock &) = delete;
+  void operator=(const GenericScopedLock &) = delete;
 };
 
-template<typename MutexType>
-class GenericScopedReadLock {
+template <typename MutexType>
+class SCOPED_LOCK GenericScopedReadLock {
  public:
-  explicit GenericScopedReadLock(MutexType *mu)
-      : mu_(mu) {
+  explicit GenericScopedReadLock(MutexType *mu) ACQUIRE(mu) : mu_(mu) {
     mu_->ReadLock();
   }
 
-  ~GenericScopedReadLock() {
-    mu_->ReadUnlock();
-  }
+  ~GenericScopedReadLock() RELEASE() { mu_->ReadUnlock(); }
 
  private:
   MutexType *mu_;
 
-  GenericScopedReadLock(const GenericScopedReadLock&);
-  void operator=(const GenericScopedReadLock&);
+  GenericScopedReadLock(const GenericScopedReadLock &) = delete;
+  void operator=(const GenericScopedReadLock &) = delete;
 };
 
 typedef GenericScopedLock<StaticSpinMutex> SpinMutexLock;
diff --git a/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp b/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
index ac20f915fef..5e601bdcde1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
@@ -215,15 +215,12 @@ void internal__exit(int exitcode) {
   Die();  // Unreachable.
 }
 
-unsigned int internal_sleep(unsigned int seconds) {
+void internal_usleep(u64 useconds) {
   struct timespec ts;
-  ts.tv_sec = seconds;
-  ts.tv_nsec = 0;
+  ts.tv_sec = useconds / 1000000;
+  ts.tv_nsec = (useconds % 1000000) * 1000;
   CHECK(&_sys___nanosleep50);
-  int res = _sys___nanosleep50(&ts, &ts);
-  if (res)
-    return ts.tv_sec;
-  return 0;
+  _sys___nanosleep50(&ts, &ts);
 }
 
 uptr internal_execve(const char *filename, char *const argv[],
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform.h b/libsanitizer/sanitizer_common/sanitizer_platform.h
index 2f6458431c8..4d3c08893c1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform.h
+++ b/libsanitizer/sanitizer_common/sanitizer_platform.h
@@ -13,10 +13,9 @@
 #define SANITIZER_PLATFORM_H
 
 #if !defined(__linux__) && !defined(__FreeBSD__) && !defined(__NetBSD__) && \
-  !defined(__APPLE__) && !defined(_WIN32) && \
-  !defined(__Fuchsia__) && !defined(__rtems__) && \
-  !(defined(__sun__) && defined(__svr4__))
-# error "This operating system is not supported"
+    !defined(__APPLE__) && !defined(_WIN32) && !defined(__Fuchsia__) &&     \
+    !(defined(__sun__) && defined(__svr4__))
+#  error "This operating system is not supported"
 #endif
 
 // Get __GLIBC__ on a glibc platform. Exclude Android: features.h includes C
@@ -117,12 +116,6 @@
 # define SANITIZER_FUCHSIA 0
 #endif
 
-#if defined(__rtems__)
-# define SANITIZER_RTEMS 1
-#else
-# define SANITIZER_RTEMS 0
-#endif
-
 #define SANITIZER_POSIX \
   (SANITIZER_FREEBSD || SANITIZER_LINUX || SANITIZER_MAC || \
     SANITIZER_NETBSD || SANITIZER_SOLARIS)
@@ -226,12 +219,6 @@
 # define SANITIZER_SOLARIS32 0
 #endif
 
-#if defined(__myriad2__)
-# define SANITIZER_MYRIAD2 1
-#else
-# define SANITIZER_MYRIAD2 0
-#endif
-
 #if defined(__riscv) && (__riscv_xlen == 64)
 #define SANITIZER_RISCV64 1
 #else
@@ -374,9 +361,9 @@
 # define SANITIZER_CACHE_LINE_SIZE 64
 #endif
 
-// Enable offline markup symbolizer for Fuchsia and RTEMS.
-#if SANITIZER_FUCHSIA || SANITIZER_RTEMS
-#define SANITIZER_SYMBOLIZER_MARKUP 1
+// Enable offline markup symbolizer for Fuchsia.
+#if SANITIZER_FUCHSIA
+#  define SANITIZER_SYMBOLIZER_MARKUP 1
 #else
 #define SANITIZER_SYMBOLIZER_MARKUP 0
 #endif
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h b/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
index 731df710df5..5b710c23fd0 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
@@ -114,12 +114,6 @@
 #define SI_NOT_FUCHSIA 1
 #endif
 
-#if SANITIZER_RTEMS
-#define SI_NOT_RTEMS 0
-#else
-#define SI_NOT_RTEMS 1
-#endif
-
 #if SANITIZER_SOLARIS
 #define SI_SOLARIS 1
 #else
@@ -482,13 +476,12 @@
 #define SANITIZER_INTERCEPT_MMAP SI_POSIX
 #define SANITIZER_INTERCEPT_MMAP64 SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT_MALLOPT_AND_MALLINFO (SI_GLIBC || SI_ANDROID)
-#define SANITIZER_INTERCEPT_MEMALIGN \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_RTEMS)
+#define SANITIZER_INTERCEPT_MEMALIGN (!SI_FREEBSD && !SI_MAC && !SI_NETBSD)
 #define SANITIZER_INTERCEPT___LIBC_MEMALIGN SI_GLIBC
 #define SANITIZER_INTERCEPT_PVALLOC (SI_GLIBC || SI_ANDROID)
 #define SANITIZER_INTERCEPT_CFREE (SI_GLIBC && !SANITIZER_RISCV64)
 #define SANITIZER_INTERCEPT_REALLOCARRAY SI_POSIX
-#define SANITIZER_INTERCEPT_ALIGNED_ALLOC (!SI_MAC && SI_NOT_RTEMS)
+#define SANITIZER_INTERCEPT_ALIGNED_ALLOC (!SI_MAC)
 #define SANITIZER_INTERCEPT_MALLOC_USABLE_SIZE (!SI_MAC && !SI_NETBSD)
 #define SANITIZER_INTERCEPT_MCHECK_MPROBE SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT_WCSCAT SI_POSIX
@@ -584,6 +577,7 @@
   (SI_POSIX && !(SANITIZER_MAC && SANITIZER_I386))
 #define SANITIZER_INTERCEPT_UNAME (SI_POSIX && !SI_FREEBSD)
 #define SANITIZER_INTERCEPT___XUNAME SI_FREEBSD
+#define SANITIZER_INTERCEPT_FLOPEN SI_FREEBSD
 
 // This macro gives a way for downstream users to override the above
 // interceptor macros irrespective of the platform they are on. They have
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp b/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp
index f22f5039128..c51327e1269 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp
@@ -26,12 +26,9 @@
 
 // With old kernels (and even new kernels on powerpc) asm/stat.h uses types that
 // are not defined anywhere in userspace headers. Fake them. This seems to work
-// fine with newer headers, too.  Beware that with <sys/stat.h>, struct stat
-// takes the form of struct stat64 on 32-bit platforms if _FILE_OFFSET_BITS=64.
-// Also, for some platforms (e.g. mips) there are additional members in the
-// <sys/stat.h> struct stat:s.
+// fine with newer headers, too.
 #include <linux/posix_types.h>
-#if defined(__x86_64__)
+#if defined(__x86_64__) ||  defined(__mips__)
 #include <sys/stat.h>
 #else
 #define ino_t __kernel_ino_t
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h
index cba41ba5494..4dd27644ed1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h
@@ -83,7 +83,7 @@ const unsigned struct_kernel_stat64_sz = 104;
 #elif defined(__mips__)
 const unsigned struct_kernel_stat_sz = SANITIZER_ANDROID
                                            ? FIRST_32_SECOND_64(104, 128)
-                                           : FIRST_32_SECOND_64(144, 216);
+                                           : FIRST_32_SECOND_64(160, 216);
 const unsigned struct_kernel_stat64_sz = 104;
 #elif defined(__s390__) && !defined(__s390x__)
 const unsigned struct_kernel_stat_sz = 64;
@@ -650,15 +650,15 @@ struct __sanitizer_sigaction {
 #endif // !SANITIZER_ANDROID
 
 #if defined(__mips__)
-struct __sanitizer_kernel_sigset_t {
-  uptr sig[2];
-};
+#define __SANITIZER_KERNEL_NSIG 128
 #else
-struct __sanitizer_kernel_sigset_t {
-  u8 sig[8];
-};
+#define __SANITIZER_KERNEL_NSIG 64
 #endif
 
+struct __sanitizer_kernel_sigset_t {
+  uptr sig[__SANITIZER_KERNEL_NSIG / (sizeof(uptr) * 8)];
+};
+
 // Linux system headers define the 'sa_handler' and 'sa_sigaction' macros.
 #if SANITIZER_MIPS
 struct __sanitizer_kernel_sigaction_t {
diff --git a/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
index d1d8e509c4d..ddf6844bed1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
@@ -128,14 +128,6 @@ void SetAddressSpaceUnlimited() {
   CHECK(AddressSpaceIsUnlimited());
 }
 
-void SleepForSeconds(int seconds) {
-  sleep(seconds);
-}
-
-void SleepForMillis(int millis) {
-  usleep(millis * 1000);
-}
-
 void Abort() {
 #if !SANITIZER_GO
   // If we are handling SIGABRT, unhandle it first.
@@ -166,9 +158,10 @@ bool SupportsColoredOutput(fd_t fd) {
 #if !SANITIZER_GO
 // TODO(glider): different tools may require different altstack size.
 static uptr GetAltStackSize() {
-  // SIGSTKSZ is not enough.
-  static const uptr kAltStackSize = SIGSTKSZ * 4;
-  return kAltStackSize;
+  // Note: since GLIBC_2.31, SIGSTKSZ may be a function call, so this may be
+  // more costly that you think. However GetAltStackSize is only call 2-3 times
+  // per thread so don't cache the evaluation.
+  return SIGSTKSZ * 4;
 }
 
 void SetAlternateSignalStack() {
diff --git a/libsanitizer/sanitizer_common/sanitizer_printf.cpp b/libsanitizer/sanitizer_common/sanitizer_printf.cpp
index 5d16dfde678..b913c92e16f 100644
--- a/libsanitizer/sanitizer_common/sanitizer_printf.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_printf.cpp
@@ -20,6 +20,10 @@
 #include <stdio.h>
 #include <stdarg.h>
 
+#if defined(__x86_64__)
+#  include <emmintrin.h>
+#endif
+
 #if SANITIZER_WINDOWS && defined(_MSC_VER) && _MSC_VER < 1800 &&               \
       !defined(va_copy)
 # define va_copy(dst, src) ((dst) = (src))
@@ -128,7 +132,7 @@ static int AppendPointer(char **buff, const char *buff_end, u64 ptr_value) {
 int VSNPrintf(char *buff, int buff_length,
               const char *format, va_list args) {
   static const char *kPrintfFormatsHelp =
-      "Supported Printf formats: %([0-9]*)?(z|ll)?{d,u,x,X}; %p; "
+      "Supported Printf formats: %([0-9]*)?(z|ll)?{d,u,x,X,V}; %p; "
       "%[-]([0-9]*)?(\\.\\*)?s; %c\n";
   RAW_CHECK(format);
   RAW_CHECK(buff_length > 0);
@@ -162,17 +166,15 @@ int VSNPrintf(char *buff, int buff_length,
     cur += have_z;
     bool have_ll = !have_z && (cur[0] == 'l' && cur[1] == 'l');
     cur += have_ll * 2;
-    s64 dval;
-    u64 uval;
     const bool have_length = have_z || have_ll;
     const bool have_flags = have_width || have_length;
     // At the moment only %s supports precision and left-justification.
     CHECK(!((precision >= 0 || left_justified) && *cur != 's'));
     switch (*cur) {
       case 'd': {
-        dval = have_ll ? va_arg(args, s64)
-             : have_z ? va_arg(args, sptr)
-             : va_arg(args, int);
+        s64 dval = have_ll  ? va_arg(args, s64)
+                   : have_z ? va_arg(args, sptr)
+                            : va_arg(args, int);
         result += AppendSignedDecimal(&buff, buff_end, dval, width,
                                       pad_with_zero);
         break;
@@ -180,14 +182,21 @@ int VSNPrintf(char *buff, int buff_length,
       case 'u':
       case 'x':
       case 'X': {
-        uval = have_ll ? va_arg(args, u64)
-             : have_z ? va_arg(args, uptr)
-             : va_arg(args, unsigned);
+        u64 uval = have_ll  ? va_arg(args, u64)
+                   : have_z ? va_arg(args, uptr)
+                            : va_arg(args, unsigned);
         bool uppercase = (*cur == 'X');
         result += AppendUnsigned(&buff, buff_end, uval, (*cur == 'u') ? 10 : 16,
                                  width, pad_with_zero, uppercase);
         break;
       }
+      case 'V': {
+        for (uptr i = 0; i < 16; i++) {
+          unsigned x = va_arg(args, unsigned);
+          result += AppendUnsigned(&buff, buff_end, x, 16, 2, true, false);
+        }
+        break;
+      }
       case 'p': {
         RAW_CHECK_MSG(!have_flags, kPrintfFormatsHelp);
         result += AppendPointer(&buff, buff_end, va_arg(args, uptr));
diff --git a/libsanitizer/sanitizer_common/sanitizer_quarantine.h b/libsanitizer/sanitizer_common/sanitizer_quarantine.h
index 992f23152c6..1a074d2bb70 100644
--- a/libsanitizer/sanitizer_common/sanitizer_quarantine.h
+++ b/libsanitizer/sanitizer_common/sanitizer_quarantine.h
@@ -149,7 +149,8 @@ class Quarantine {
   Cache cache_;
   char pad2_[kCacheLineSize];
 
-  void NOINLINE Recycle(uptr min_size, Callback cb) {
+  void NOINLINE Recycle(uptr min_size, Callback cb) REQUIRES(recycle_mutex_)
+      RELEASE(recycle_mutex_) {
     Cache tmp;
     {
       SpinMutexLock l(&cache_mutex_);
diff --git a/libsanitizer/sanitizer_common/sanitizer_rtems.cpp b/libsanitizer/sanitizer_common/sanitizer_rtems.cpp
deleted file mode 100644
index d58bd08fb1a..00000000000
--- a/libsanitizer/sanitizer_common/sanitizer_rtems.cpp
+++ /dev/null
@@ -1,281 +0,0 @@
-//===-- sanitizer_rtems.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 shared between various sanitizers' runtime libraries and
-// implements RTEMS-specific functions.
-//===----------------------------------------------------------------------===//
-
-#include "sanitizer_rtems.h"
-#if SANITIZER_RTEMS
-
-#define posix_memalign __real_posix_memalign
-#define free __real_free
-#define memset __real_memset
-
-#include "sanitizer_file.h"
-#include "sanitizer_symbolizer.h"
-#include <errno.h>
-#include <fcntl.h>
-#include <pthread.h>
-#include <sched.h>
-#include <stdio.h>
-#include <stdlib.h>
-#include <string.h>
-#include <unistd.h>
-
-// There is no mmap on RTEMS.  Use memalign, etc.
-#define __mmap_alloc_aligned posix_memalign
-#define __mmap_free free
-#define __mmap_memset memset
-
-namespace __sanitizer {
-
-#include "sanitizer_syscall_generic.inc"
-
-void NORETURN internal__exit(int exitcode) {
-  _exit(exitcode);
-}
-
-uptr internal_sched_yield() {
-  return sched_yield();
-}
-
-uptr internal_getpid() {
-  return getpid();
-}
-
-int internal_dlinfo(void *handle, int request, void *p) {
-  UNIMPLEMENTED();
-}
-
-bool FileExists(const char *filename) {
-  struct stat st;
-  if (stat(filename, &st))
-    return false;
-  // Sanity check: filename is a regular file.
-  return S_ISREG(st.st_mode);
-}
-
-uptr GetThreadSelf() { return static_cast<uptr>(pthread_self()); }
-
-tid_t GetTid() { return GetThreadSelf(); }
-
-void Abort() { abort(); }
-
-int Atexit(void (*function)(void)) { return atexit(function); }
-
-void SleepForSeconds(int seconds) { sleep(seconds); }
-
-void SleepForMillis(int millis) { usleep(millis * 1000); }
-
-bool SupportsColoredOutput(fd_t fd) { return false; }
-
-void GetThreadStackTopAndBottom(bool at_initialization,
-                                uptr *stack_top, uptr *stack_bottom) {
-  pthread_attr_t attr;
-  pthread_attr_init(&attr);
-  CHECK_EQ(pthread_getattr_np(pthread_self(), &attr), 0);
-  void *base = nullptr;
-  size_t size = 0;
-  CHECK_EQ(pthread_attr_getstack(&attr, &base, &size), 0);
-  CHECK_EQ(pthread_attr_destroy(&attr), 0);
-
-  *stack_bottom = reinterpret_cast<uptr>(base);
-  *stack_top = *stack_bottom + size;
-}
-
-void GetThreadStackAndTls(bool main, uptr *stk_addr, uptr *stk_size,
-                          uptr *tls_addr, uptr *tls_size) {
-  uptr stack_top, stack_bottom;
-  GetThreadStackTopAndBottom(main, &stack_top, &stack_bottom);
-  *stk_addr = stack_bottom;
-  *stk_size = stack_top - stack_bottom;
-  *tls_addr = *tls_size = 0;
-}
-
-void InitializePlatformEarly() {}
-void MaybeReexec() {}
-void CheckASLR() {}
-void CheckMPROTECT() {}
-void DisableCoreDumperIfNecessary() {}
-void InstallDeadlySignalHandlers(SignalHandlerType handler) {}
-void SetAlternateSignalStack() {}
-void UnsetAlternateSignalStack() {}
-void InitTlsSize() {}
-
-void SignalContext::DumpAllRegisters(void *context) {}
-const char *DescribeSignalOrException(int signo) { UNIMPLEMENTED(); }
-
-enum MutexState { MtxUnlocked = 0, MtxLocked = 1, MtxSleeping = 2 };
-
-BlockingMutex::BlockingMutex() {
-  internal_memset(this, 0, sizeof(*this));
-}
-
-void BlockingMutex::Lock() {
-  CHECK_EQ(owner_, 0);
-  atomic_uint32_t *m = reinterpret_cast<atomic_uint32_t *>(&opaque_storage_);
-  if (atomic_exchange(m, MtxLocked, memory_order_acquire) == MtxUnlocked)
-    return;
-  while (atomic_exchange(m, MtxSleeping, memory_order_acquire) != MtxUnlocked) {
-    internal_sched_yield();
-  }
-}
-
-void BlockingMutex::Unlock() {
-  atomic_uint32_t *m = reinterpret_cast<atomic_uint32_t *>(&opaque_storage_);
-  u32 v = atomic_exchange(m, MtxUnlocked, memory_order_release);
-  CHECK_NE(v, MtxUnlocked);
-}
-
-void BlockingMutex::CheckLocked() {
-  atomic_uint32_t *m = reinterpret_cast<atomic_uint32_t *>(&opaque_storage_);
-  CHECK_NE(MtxUnlocked, atomic_load(m, memory_order_relaxed));
-}
-
-uptr GetPageSize() { return getpagesize(); }
-
-uptr GetMmapGranularity() { return GetPageSize(); }
-
-uptr GetMaxVirtualAddress() {
-  return (1ULL << 32) - 1;  // 0xffffffff
-}
-
-void *MmapOrDie(uptr size, const char *mem_type, bool raw_report) {
-  void* ptr = 0;
-  int res = __mmap_alloc_aligned(&ptr, GetPageSize(), size);
-  if (UNLIKELY(res))
-    ReportMmapFailureAndDie(size, mem_type, "allocate", res, raw_report);
-  __mmap_memset(ptr, 0, size);
-  IncreaseTotalMmap(size);
-  return ptr;
-}
-
-void *MmapOrDieOnFatalError(uptr size, const char *mem_type) {
-  void* ptr = 0;
-  int res = __mmap_alloc_aligned(&ptr, GetPageSize(), size);
-  if (UNLIKELY(res)) {
-    if (res == ENOMEM)
-      return nullptr;
-    ReportMmapFailureAndDie(size, mem_type, "allocate", false);
-  }
-  __mmap_memset(ptr, 0, size);
-  IncreaseTotalMmap(size);
-  return ptr;
-}
-
-void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
-                                   const char *mem_type) {
-  CHECK(IsPowerOfTwo(size));
-  CHECK(IsPowerOfTwo(alignment));
-  void* ptr = 0;
-  int res = __mmap_alloc_aligned(&ptr, alignment, size);
-  if (res)
-    ReportMmapFailureAndDie(size, mem_type, "align allocate", res, false);
-  __mmap_memset(ptr, 0, size);
-  IncreaseTotalMmap(size);
-  return ptr;
-}
-
-void *MmapNoReserveOrDie(uptr size, const char *mem_type) {
-  return MmapOrDie(size, mem_type, false);
-}
-
-void UnmapOrDie(void *addr, uptr size) {
-  if (!addr || !size) return;
-  __mmap_free(addr);
-  DecreaseTotalMmap(size);
-}
-
-fd_t OpenFile(const char *filename, FileAccessMode mode, error_t *errno_p) {
-  int flags;
-  switch (mode) {
-    case RdOnly: flags = O_RDONLY; break;
-    case WrOnly: flags = O_WRONLY | O_CREAT | O_TRUNC; break;
-    case RdWr: flags = O_RDWR | O_CREAT; break;
-  }
-  fd_t res = open(filename, flags, 0660);
-  if (internal_iserror(res, errno_p))
-    return kInvalidFd;
-  return res;
-}
-
-void CloseFile(fd_t fd) {
-  close(fd);
-}
-
-bool ReadFromFile(fd_t fd, void *buff, uptr buff_size, uptr *bytes_read,
-                  error_t *error_p) {
-  uptr res = read(fd, buff, buff_size);
-  if (internal_iserror(res, error_p))
-    return false;
-  if (bytes_read)
-    *bytes_read = res;
-  return true;
-}
-
-bool WriteToFile(fd_t fd, const void *buff, uptr buff_size, uptr *bytes_written,
-                 error_t *error_p) {
-  uptr res = write(fd, buff, buff_size);
-  if (internal_iserror(res, error_p))
-    return false;
-  if (bytes_written)
-    *bytes_written = res;
-  return true;
-}
-
-void ReleaseMemoryPagesToOS(uptr beg, uptr end) {}
-void DumpProcessMap() {}
-
-// There is no page protection so everything is "accessible."
-bool IsAccessibleMemoryRange(uptr beg, uptr size) {
-  return true;
-}
-
-char **GetArgv() { return nullptr; }
-char **GetEnviron() { return nullptr; }
-
-const char *GetEnv(const char *name) {
-  return getenv(name);
-}
-
-uptr ReadBinaryName(/*out*/char *buf, uptr buf_len) {
-  internal_strncpy(buf, "StubBinaryName", buf_len);
-  return internal_strlen(buf);
-}
-
-uptr ReadLongProcessName(/*out*/ char *buf, uptr buf_len) {
-  internal_strncpy(buf, "StubProcessName", buf_len);
-  return internal_strlen(buf);
-}
-
-bool IsPathSeparator(const char c) {
-  return c == '/';
-}
-
-bool IsAbsolutePath(const char *path) {
-  return path != nullptr && IsPathSeparator(path[0]);
-}
-
-void ReportFile::Write(const char *buffer, uptr length) {
-  SpinMutexLock l(mu);
-  static const char *kWriteError =
-      "ReportFile::Write() can't output requested buffer!\n";
-  ReopenIfNecessary();
-  if (length != write(fd, buffer, length)) {
-    write(fd, kWriteError, internal_strlen(kWriteError));
-    Die();
-  }
-}
-
-uptr MainThreadStackBase, MainThreadStackSize;
-uptr MainThreadTlsBase, MainThreadTlsSize;
-
-} // namespace __sanitizer
-
-#endif  // SANITIZER_RTEMS
diff --git a/libsanitizer/sanitizer_common/sanitizer_rtems.h b/libsanitizer/sanitizer_common/sanitizer_rtems.h
deleted file mode 100644
index e8adfd500df..00000000000
--- a/libsanitizer/sanitizer_common/sanitizer_rtems.h
+++ /dev/null
@@ -1,20 +0,0 @@
-//===-- sanitizer_rtems.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 shared between various sanitizers' runtime libraries and
-// provides definitions for RTEMS-specific functions.
-//===----------------------------------------------------------------------===//
-#ifndef SANITIZER_RTEMS_H
-#define SANITIZER_RTEMS_H
-
-#include "sanitizer_platform.h"
-#if SANITIZER_RTEMS
-#include "sanitizer_common.h"
-
-#endif  // SANITIZER_RTEMS
-#endif  // SANITIZER_RTEMS_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_solaris.cpp b/libsanitizer/sanitizer_common/sanitizer_solaris.cpp
index 8789dcd10a9..cb53eab8da1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_solaris.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_solaris.cpp
@@ -160,6 +160,13 @@ DECLARE__REAL_AND_INTERNAL(uptr, sched_yield, void) {
   return sched_yield();
 }
 
+DECLARE__REAL_AND_INTERNAL(void, usleep, u64 useconds) {
+  struct timespec ts;
+  ts.tv_sec = useconds / 1000000;
+  ts.tv_nsec = (useconds % 1000000) * 1000;
+  nanosleep(&ts, nullptr);
+}
+
 DECLARE__REAL_AND_INTERNAL(uptr, execve, const char *filename,
                            char *const argv[], char *const envp[]) {
   return _REAL(execve)(filename, argv, envp);
@@ -211,6 +218,13 @@ uptr internal_clock_gettime(__sanitizer_clockid_t clk_id, void *tp) {
 }
 
 // ----------------- sanitizer_common.h
+void FutexWait(atomic_uint32_t *p, u32 cmp) {
+  // FIXME: implement actual blocking.
+  sched_yield();
+}
+
+void FutexWake(atomic_uint32_t *p, u32 count) {}
+
 BlockingMutex::BlockingMutex() {
   CHECK(sizeof(mutex_t) <= sizeof(opaque_storage_));
   internal_memset(this, 0, sizeof(*this));
@@ -231,9 +245,7 @@ void BlockingMutex::Unlock() {
   CHECK_EQ(mutex_unlock((mutex_t *)&opaque_storage_), 0);
 }
 
-void BlockingMutex::CheckLocked() {
-  CHECK_EQ((uptr)thr_self(), owner_);
-}
+void BlockingMutex::CheckLocked() const { CHECK_EQ((uptr)thr_self(), owner_); }
 
 }  // namespace __sanitizer
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp b/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
index 515deddf6b2..07e4409f4a5 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
@@ -85,8 +85,8 @@ static inline uhwptr *GetCanonicFrame(uptr bp,
   // Nope, this does not look right either. This means the frame after next does
   // not have a valid frame pointer, but we can still extract the caller PC.
   // Unfortunately, there is no way to decide between GCC and LLVM frame
-  // layouts. Assume GCC.
-  return bp_prev - 1;
+  // layouts. Assume LLVM.
+  return bp_prev;
 #else
   return (uhwptr*)bp;
 #endif
@@ -109,21 +109,14 @@ void BufferedStackTrace::UnwindFast(uptr pc, uptr bp, uptr stack_top,
          IsAligned((uptr)frame, sizeof(*frame)) &&
          size < max_depth) {
 #ifdef __powerpc__
-    // PowerPC ABIs specify that the return address is saved on the
-    // *caller's* stack frame.  Thus we must dereference the back chain
-    // to find the caller frame before extracting it.
+    // PowerPC ABIs specify that the return address is saved at offset
+    // 16 of the *caller's* stack frame.  Thus we must dereference the
+    // back chain to find the caller frame before extracting it.
     uhwptr *caller_frame = (uhwptr*)frame[0];
     if (!IsValidFrame((uptr)caller_frame, stack_top, bottom) ||
         !IsAligned((uptr)caller_frame, sizeof(uhwptr)))
       break;
-    // For most ABIs the offset where the return address is saved is two
-    // register sizes.  The exception is the SVR4 ABI, which uses an
-    // offset of only one register size.
-#ifdef _CALL_SYSV
-    uhwptr pc1 = caller_frame[1];
-#else
     uhwptr pc1 = caller_frame[2];
-#endif
 #elif defined(__s390__)
     uhwptr pc1 = frame[14];
 #elif defined(__riscv)
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace.h b/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
index 15616f899d0..ea330f36f7d 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
@@ -12,6 +12,7 @@
 #ifndef SANITIZER_STACKTRACE_H
 #define SANITIZER_STACKTRACE_H
 
+#include "sanitizer_common.h"
 #include "sanitizer_internal_defs.h"
 #include "sanitizer_platform.h"
 
@@ -32,8 +33,8 @@ static const u32 kStackTraceMax = 256;
 // Fast unwind is the only option on Mac for now; we will need to
 // revisit this macro when slow unwind works on Mac, see
 // https://github.com/google/sanitizers/issues/137
-#if SANITIZER_MAC || SANITIZER_RTEMS
-# define SANITIZER_CAN_SLOW_UNWIND 0
+#if SANITIZER_MAC
+#  define SANITIZER_CAN_SLOW_UNWIND 0
 #else
 # define SANITIZER_CAN_SLOW_UNWIND 1
 #endif
@@ -56,6 +57,16 @@ struct StackTrace {
   // Prints a symbolized stacktrace, followed by an empty line.
   void Print() const;
 
+  // Prints a symbolized stacktrace to the output string, followed by an empty
+  // line.
+  void PrintTo(InternalScopedString *output) const;
+
+  // Prints a symbolized stacktrace to the output buffer, followed by an empty
+  // line. Returns the number of symbols that should have been written to buffer
+  // (not including trailing '\0'). Thus, the string is truncated iff return
+  // value is not less than "out_buf_size".
+  uptr PrintTo(char *out_buf, uptr out_buf_size) const;
+
   static bool WillUseFastUnwind(bool request_fast_unwind) {
     if (!SANITIZER_CAN_FAST_UNWIND)
       return false;
@@ -185,5 +196,26 @@ static inline bool IsValidFrame(uptr frame, uptr stack_top, uptr stack_bottom) {
   uptr local_stack;                           \
   uptr sp = (uptr)&local_stack
 
+// GET_CURRENT_PC() is equivalent to StackTrace::GetCurrentPc().
+// Optimized x86 version is faster than GetCurrentPc because
+// it does not involve a function call, instead it reads RIP register.
+// Reads of RIP by an instruction return RIP pointing to the next
+// instruction, which is exactly what we want here, thus 0 offset.
+// It needs to be a macro because otherwise we will get the name
+// of this function on the top of most stacks. Attribute artificial
+// does not do what it claims to do, unfortunatley. And attribute
+// __nodebug__ is clang-only. If we would have an attribute that
+// would remove this function from debug info, we could simply make
+// StackTrace::GetCurrentPc() faster.
+#if defined(__x86_64__)
+#  define GET_CURRENT_PC()                \
+    ({                                    \
+      uptr pc;                            \
+      asm("lea 0(%%rip), %0" : "=r"(pc)); \
+      pc;                                 \
+    })
+#else
+#  define GET_CURRENT_PC() StackTrace::GetCurrentPc()
+#endif
 
 #endif  // SANITIZER_STACKTRACE_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
index 738633209f0..f60ea773174 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
@@ -18,46 +18,119 @@
 
 namespace __sanitizer {
 
-void StackTrace::Print() const {
+namespace {
+
+class StackTraceTextPrinter {
+ public:
+  StackTraceTextPrinter(const char *stack_trace_fmt, char frame_delimiter,
+                        InternalScopedString *output,
+                        InternalScopedString *dedup_token)
+      : stack_trace_fmt_(stack_trace_fmt),
+        frame_delimiter_(frame_delimiter),
+        output_(output),
+        dedup_token_(dedup_token),
+        symbolize_(RenderNeedsSymbolization(stack_trace_fmt)) {}
+
+  bool ProcessAddressFrames(uptr pc) {
+    SymbolizedStack *frames = symbolize_
+                                  ? Symbolizer::GetOrInit()->SymbolizePC(pc)
+                                  : SymbolizedStack::New(pc);
+    if (!frames)
+      return false;
+
+    for (SymbolizedStack *cur = frames; cur; cur = cur->next) {
+      uptr prev_len = output_->length();
+      RenderFrame(output_, stack_trace_fmt_, frame_num_++, cur->info.address,
+                  symbolize_ ? &cur->info : nullptr,
+                  common_flags()->symbolize_vs_style,
+                  common_flags()->strip_path_prefix);
+
+      if (prev_len != output_->length())
+        output_->append("%c", frame_delimiter_);
+
+      ExtendDedupToken(cur);
+    }
+    frames->ClearAll();
+    return true;
+  }
+
+ private:
+  // Extend the dedup token by appending a new frame.
+  void ExtendDedupToken(SymbolizedStack *stack) {
+    if (!dedup_token_)
+      return;
+
+    if (dedup_frames_-- > 0) {
+      if (dedup_token_->length())
+        dedup_token_->append("--");
+      if (stack->info.function != nullptr)
+        dedup_token_->append(stack->info.function);
+    }
+  }
+
+  const char *stack_trace_fmt_;
+  const char frame_delimiter_;
+  int dedup_frames_ = common_flags()->dedup_token_length;
+  uptr frame_num_ = 0;
+  InternalScopedString *output_;
+  InternalScopedString *dedup_token_;
+  const bool symbolize_ = false;
+};
+
+static void CopyStringToBuffer(const InternalScopedString &str, char *out_buf,
+                               uptr out_buf_size) {
+  if (!out_buf_size)
+    return;
+
+  CHECK_GT(out_buf_size, 0);
+  uptr copy_size = Min(str.length(), out_buf_size - 1);
+  internal_memcpy(out_buf, str.data(), copy_size);
+  out_buf[copy_size] = '\0';
+}
+
+}  // namespace
+
+void StackTrace::PrintTo(InternalScopedString *output) const {
+  CHECK(output);
+
+  InternalScopedString dedup_token;
+  StackTraceTextPrinter printer(common_flags()->stack_trace_format, '\n',
+                                output, &dedup_token);
+
   if (trace == nullptr || size == 0) {
-    Printf("    <empty stack>\n\n");
+    output->append("    <empty stack>\n\n");
     return;
   }
-  InternalScopedString frame_desc;
-  InternalScopedString dedup_token;
-  int dedup_frames = common_flags()->dedup_token_length;
-  bool symbolize = RenderNeedsSymbolization(common_flags()->stack_trace_format);
-  uptr frame_num = 0;
+
   for (uptr i = 0; i < size && trace[i]; i++) {
     // PCs in stack traces are actually the return addresses, that is,
     // addresses of the next instructions after the call.
     uptr pc = GetPreviousInstructionPc(trace[i]);
-    SymbolizedStack *frames;
-    if (symbolize)
-      frames = Symbolizer::GetOrInit()->SymbolizePC(pc);
-    else
-      frames = SymbolizedStack::New(pc);
-    CHECK(frames);
-    for (SymbolizedStack *cur = frames; cur; cur = cur->next) {
-      frame_desc.clear();
-      RenderFrame(&frame_desc, common_flags()->stack_trace_format, frame_num++,
-                  cur->info.address, symbolize ? &cur->info : nullptr,
-                  common_flags()->symbolize_vs_style,
-                  common_flags()->strip_path_prefix);
-      Printf("%s\n", frame_desc.data());
-      if (dedup_frames-- > 0) {
-        if (dedup_token.length())
-          dedup_token.append("--");
-        if (cur->info.function != nullptr)
-          dedup_token.append(cur->info.function);
-      }
-    }
-    frames->ClearAll();
+    CHECK(printer.ProcessAddressFrames(pc));
   }
-  // Always print a trailing empty line after stack trace.
-  Printf("\n");
+
+  // Always add a trailing empty line after stack trace.
+  output->append("\n");
+
+  // Append deduplication token, if non-empty.
   if (dedup_token.length())
-    Printf("DEDUP_TOKEN: %s\n", dedup_token.data());
+    output->append("DEDUP_TOKEN: %s\n", dedup_token.data());
+}
+
+uptr StackTrace::PrintTo(char *out_buf, uptr out_buf_size) const {
+  CHECK(out_buf);
+
+  InternalScopedString output;
+  PrintTo(&output);
+  CopyStringToBuffer(output, out_buf, out_buf_size);
+
+  return output.length();
+}
+
+void StackTrace::Print() const {
+  InternalScopedString output;
+  PrintTo(&output);
+  Printf("%s", output.data());
 }
 
 void BufferedStackTrace::Unwind(u32 max_depth, uptr pc, uptr bp, void *context,
@@ -82,12 +155,15 @@ void BufferedStackTrace::Unwind(u32 max_depth, uptr pc, uptr bp, void *context,
       UnwindSlow(pc, context, max_depth);
     else
       UnwindSlow(pc, max_depth);
+    // If there are too few frames, the program may be built with
+    // -fno-asynchronous-unwind-tables. Fall back to fast unwinder below.
+    if (size > 2 || size >= max_depth)
+      return;
 #else
     UNREACHABLE("slow unwind requested but not available");
 #endif
-  } else {
-    UnwindFast(pc, bp, stack_top, stack_bottom, max_depth);
   }
+  UnwindFast(pc, bp, stack_top, stack_bottom, max_depth);
 }
 
 static int GetModuleAndOffsetForPc(uptr pc, char *module_name,
@@ -112,41 +188,18 @@ extern "C" {
 SANITIZER_INTERFACE_ATTRIBUTE
 void __sanitizer_symbolize_pc(uptr pc, const char *fmt, char *out_buf,
                               uptr out_buf_size) {
-  if (!out_buf_size) return;
-  pc = StackTrace::GetPreviousInstructionPc(pc);
-  SymbolizedStack *frame;
-  bool symbolize = RenderNeedsSymbolization(fmt);
-  if (symbolize)
-    frame = Symbolizer::GetOrInit()->SymbolizePC(pc);
-  else
-    frame = SymbolizedStack::New(pc);
-  if (!frame) {
-    internal_strncpy(out_buf, "<can't symbolize>", out_buf_size);
-    out_buf[out_buf_size - 1] = 0;
+  if (!out_buf_size)
     return;
+
+  pc = StackTrace::GetPreviousInstructionPc(pc);
+
+  InternalScopedString output;
+  StackTraceTextPrinter printer(fmt, '\0', &output, nullptr);
+  if (!printer.ProcessAddressFrames(pc)) {
+    output.clear();
+    output.append("<can't symbolize>");
   }
-  InternalScopedString frame_desc;
-  uptr frame_num = 0;
-  // Reserve one byte for the final 0.
-  char *out_end = out_buf + out_buf_size - 1;
-  for (SymbolizedStack *cur = frame; cur && out_buf < out_end;
-       cur = cur->next) {
-    frame_desc.clear();
-    RenderFrame(&frame_desc, fmt, frame_num++, cur->info.address,
-                symbolize ? &cur->info : nullptr,
-                common_flags()->symbolize_vs_style,
-                common_flags()->strip_path_prefix);
-    if (!frame_desc.length())
-      continue;
-    // Reserve one byte for the terminating 0.
-    uptr n = out_end - out_buf - 1;
-    internal_strncpy(out_buf, frame_desc.data(), n);
-    out_buf += __sanitizer::Min<uptr>(n, frame_desc.length());
-    *out_buf++ = 0;
-  }
-  CHECK(out_buf <= out_end);
-  *out_buf = 0;
-  frame->ClearAll();
+  CopyStringToBuffer(output, out_buf, out_buf_size);
 }
 
 SANITIZER_INTERFACE_ATTRIBUTE
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
index 01edef9c1aa..9a5b4a8c54c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
@@ -16,14 +16,13 @@
 
 #if SANITIZER_FUCHSIA
 #include "sanitizer_symbolizer_fuchsia.h"
-#elif SANITIZER_RTEMS
-#include "sanitizer_symbolizer_rtems.h"
-#endif
-#include "sanitizer_stacktrace.h"
-#include "sanitizer_symbolizer.h"
+#  endif
 
-#include <limits.h>
-#include <unwind.h>
+#  include <limits.h>
+#  include <unwind.h>
+
+#  include "sanitizer_stacktrace.h"
+#  include "sanitizer_symbolizer.h"
 
 namespace __sanitizer {
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
index 9287993e665..f330ed36640 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
@@ -120,7 +120,7 @@ void ReportMmapWriteExec(int prot) {
 #endif
 }
 
-#if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS && !SANITIZER_GO
+#if !SANITIZER_FUCHSIA && !SANITIZER_GO
 void StartReportDeadlySignal() {
   // Write the first message using fd=2, just in case.
   // It may actually fail to write in case stderr is closed.
@@ -250,17 +250,17 @@ void HandleDeadlySignal(void *siginfo, void *context, u32 tid,
 
 #endif  // !SANITIZER_FUCHSIA && !SANITIZER_GO
 
-static atomic_uintptr_t reporting_thread = {0};
-static StaticSpinMutex CommonSanitizerReportMutex;
+atomic_uintptr_t ScopedErrorReportLock::reporting_thread_ = {0};
+StaticSpinMutex ScopedErrorReportLock::mutex_;
 
-ScopedErrorReportLock::ScopedErrorReportLock() {
+void ScopedErrorReportLock::Lock() {
   uptr current = GetThreadSelf();
   for (;;) {
     uptr expected = 0;
-    if (atomic_compare_exchange_strong(&reporting_thread, &expected, current,
+    if (atomic_compare_exchange_strong(&reporting_thread_, &expected, current,
                                        memory_order_relaxed)) {
       // We've claimed reporting_thread so proceed.
-      CommonSanitizerReportMutex.Lock();
+      mutex_.Lock();
       return;
     }
 
@@ -282,13 +282,11 @@ ScopedErrorReportLock::ScopedErrorReportLock() {
   }
 }
 
-ScopedErrorReportLock::~ScopedErrorReportLock() {
-  CommonSanitizerReportMutex.Unlock();
-  atomic_store_relaxed(&reporting_thread, 0);
+void ScopedErrorReportLock::Unlock() {
+  mutex_.Unlock();
+  atomic_store_relaxed(&reporting_thread_, 0);
 }
 
-void ScopedErrorReportLock::CheckLocked() {
-  CommonSanitizerReportMutex.CheckLocked();
-}
+void ScopedErrorReportLock::CheckLocked() { mutex_.CheckLocked(); }
 
 }  // namespace __sanitizer
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_rtems.h b/libsanitizer/sanitizer_common/sanitizer_symbolizer_rtems.h
deleted file mode 100644
index 3371092e068..00000000000
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_rtems.h
+++ /dev/null
@@ -1,40 +0,0 @@
-//===-- sanitizer_symbolizer_rtems.h -----------------------------------===//
-//
-// 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 shared between various sanitizers' runtime libraries.
-//
-// Define RTEMS's string formats and limits for the markup symbolizer.
-//===----------------------------------------------------------------------===//
-#ifndef SANITIZER_SYMBOLIZER_RTEMS_H
-#define SANITIZER_SYMBOLIZER_RTEMS_H
-
-#include "sanitizer_internal_defs.h"
-
-namespace __sanitizer {
-
-// The Myriad RTEMS symbolizer currently only parses backtrace lines,
-// so use a format that the symbolizer understands.  For other
-// markups, keep them the same as the Fuchsia's.
-
-// This is used by UBSan for type names, and by ASan for global variable names.
-constexpr const char *kFormatDemangle = "{{{symbol:%s}}}";
-constexpr uptr kFormatDemangleMax = 1024;  // Arbitrary.
-
-// Function name or equivalent from PC location.
-constexpr const char *kFormatFunction = "{{{pc:%p}}}";
-constexpr uptr kFormatFunctionMax = 64;  // More than big enough for 64-bit hex.
-
-// Global variable name or equivalent from data memory address.
-constexpr const char *kFormatData = "{{{data:%p}}}";
-
-// One frame in a backtrace (printed on a line by itself).
-constexpr const char *kFormatFrame = "    [%u] IP: %p";
-
-}  // namespace __sanitizer
-
-#endif  // SANITIZER_SYMBOLIZER_RTEMS_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp b/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
index 3273da38bfd..745fbf76b01 100644
--- a/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
@@ -99,6 +99,9 @@ void ThreadContextBase::Reset() {
 
 // ThreadRegistry implementation.
 
+ThreadRegistry::ThreadRegistry(ThreadContextFactory factory)
+    : ThreadRegistry(factory, UINT32_MAX, UINT32_MAX, 0) {}
+
 ThreadRegistry::ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
                                u32 thread_quarantine_size, u32 max_reuse)
     : context_factory_(factory),
@@ -106,13 +109,10 @@ ThreadRegistry::ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
       thread_quarantine_size_(thread_quarantine_size),
       max_reuse_(max_reuse),
       mtx_(),
-      n_contexts_(0),
       total_threads_(0),
       alive_threads_(0),
       max_alive_threads_(0),
       running_threads_(0) {
-  threads_ = (ThreadContextBase **)MmapOrDie(max_threads_ * sizeof(threads_[0]),
-                                             "ThreadRegistry");
   dead_threads_.clear();
   invalid_threads_.clear();
 }
@@ -120,7 +120,8 @@ ThreadRegistry::ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
 void ThreadRegistry::GetNumberOfThreads(uptr *total, uptr *running,
                                         uptr *alive) {
   BlockingMutexLock l(&mtx_);
-  if (total) *total = n_contexts_;
+  if (total)
+    *total = threads_.size();
   if (running) *running = running_threads_;
   if (alive) *alive = alive_threads_;
 }
@@ -137,11 +138,11 @@ u32 ThreadRegistry::CreateThread(uptr user_id, bool detached, u32 parent_tid,
   ThreadContextBase *tctx = QuarantinePop();
   if (tctx) {
     tid = tctx->tid;
-  } else if (n_contexts_ < max_threads_) {
+  } else if (threads_.size() < max_threads_) {
     // Allocate new thread context and tid.
-    tid = n_contexts_++;
+    tid = threads_.size();
     tctx = context_factory_(tid);
-    threads_[tid] = tctx;
+    threads_.push_back(tctx);
   } else {
 #if !SANITIZER_GO
     Report("%s: Thread limit (%u threads) exceeded. Dying.\n",
@@ -169,7 +170,7 @@ u32 ThreadRegistry::CreateThread(uptr user_id, bool detached, u32 parent_tid,
 void ThreadRegistry::RunCallbackForEachThreadLocked(ThreadCallback cb,
                                                     void *arg) {
   CheckLocked();
-  for (u32 tid = 0; tid < n_contexts_; tid++) {
+  for (u32 tid = 0; tid < threads_.size(); tid++) {
     ThreadContextBase *tctx = threads_[tid];
     if (tctx == 0)
       continue;
@@ -179,7 +180,7 @@ void ThreadRegistry::RunCallbackForEachThreadLocked(ThreadCallback cb,
 
 u32 ThreadRegistry::FindThread(FindThreadCallback cb, void *arg) {
   BlockingMutexLock l(&mtx_);
-  for (u32 tid = 0; tid < n_contexts_; tid++) {
+  for (u32 tid = 0; tid < threads_.size(); tid++) {
     ThreadContextBase *tctx = threads_[tid];
     if (tctx != 0 && cb(tctx, arg))
       return tctx->tid;
@@ -190,7 +191,7 @@ u32 ThreadRegistry::FindThread(FindThreadCallback cb, void *arg) {
 ThreadContextBase *
 ThreadRegistry::FindThreadContextLocked(FindThreadCallback cb, void *arg) {
   CheckLocked();
-  for (u32 tid = 0; tid < n_contexts_; tid++) {
+  for (u32 tid = 0; tid < threads_.size(); tid++) {
     ThreadContextBase *tctx = threads_[tid];
     if (tctx != 0 && cb(tctx, arg))
       return tctx;
@@ -211,7 +212,6 @@ ThreadContextBase *ThreadRegistry::FindThreadContextByOsIDLocked(tid_t os_id) {
 
 void ThreadRegistry::SetThreadName(u32 tid, const char *name) {
   BlockingMutexLock l(&mtx_);
-  CHECK_LT(tid, n_contexts_);
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   CHECK_EQ(SANITIZER_FUCHSIA ? ThreadStatusCreated : ThreadStatusRunning,
@@ -221,7 +221,7 @@ void ThreadRegistry::SetThreadName(u32 tid, const char *name) {
 
 void ThreadRegistry::SetThreadNameByUserId(uptr user_id, const char *name) {
   BlockingMutexLock l(&mtx_);
-  for (u32 tid = 0; tid < n_contexts_; tid++) {
+  for (u32 tid = 0; tid < threads_.size(); tid++) {
     ThreadContextBase *tctx = threads_[tid];
     if (tctx != 0 && tctx->user_id == user_id &&
         tctx->status != ThreadStatusInvalid) {
@@ -233,7 +233,6 @@ void ThreadRegistry::SetThreadNameByUserId(uptr user_id, const char *name) {
 
 void ThreadRegistry::DetachThread(u32 tid, void *arg) {
   BlockingMutexLock l(&mtx_);
-  CHECK_LT(tid, n_contexts_);
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   if (tctx->status == ThreadStatusInvalid) {
@@ -254,7 +253,6 @@ void ThreadRegistry::JoinThread(u32 tid, void *arg) {
   do {
     {
       BlockingMutexLock l(&mtx_);
-      CHECK_LT(tid, n_contexts_);
       ThreadContextBase *tctx = threads_[tid];
       CHECK_NE(tctx, 0);
       if (tctx->status == ThreadStatusInvalid) {
@@ -280,7 +278,6 @@ ThreadStatus ThreadRegistry::FinishThread(u32 tid) {
   BlockingMutexLock l(&mtx_);
   CHECK_GT(alive_threads_, 0);
   alive_threads_--;
-  CHECK_LT(tid, n_contexts_);
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   bool dead = tctx->detached;
@@ -306,7 +303,6 @@ void ThreadRegistry::StartThread(u32 tid, tid_t os_id, ThreadType thread_type,
                                  void *arg) {
   BlockingMutexLock l(&mtx_);
   running_threads_++;
-  CHECK_LT(tid, n_contexts_);
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   CHECK_EQ(ThreadStatusCreated, tctx->status);
@@ -339,7 +335,6 @@ ThreadContextBase *ThreadRegistry::QuarantinePop() {
 
 void ThreadRegistry::SetThreadUserId(u32 tid, uptr user_id) {
   BlockingMutexLock l(&mtx_);
-  CHECK_LT(tid, n_contexts_);
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   CHECK_NE(tctx->status, ThreadStatusInvalid);
diff --git a/libsanitizer/sanitizer_common/sanitizer_thread_registry.h b/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
index dcd445c28ae..0b28bbe6ddf 100644
--- a/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
+++ b/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
@@ -85,22 +85,22 @@ class ThreadContextBase {
 
 typedef ThreadContextBase* (*ThreadContextFactory)(u32 tid);
 
-class ThreadRegistry {
+class MUTEX ThreadRegistry {
  public:
+  ThreadRegistry(ThreadContextFactory factory);
   ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
-                 u32 thread_quarantine_size, u32 max_reuse = 0);
+                 u32 thread_quarantine_size, u32 max_reuse);
   void GetNumberOfThreads(uptr *total = nullptr, uptr *running = nullptr,
                           uptr *alive = nullptr);
   uptr GetMaxAliveThreads();
 
-  void Lock() { mtx_.Lock(); }
-  void CheckLocked() { mtx_.CheckLocked(); }
-  void Unlock() { mtx_.Unlock(); }
+  void Lock() ACQUIRE() { mtx_.Lock(); }
+  void CheckLocked() const CHECK_LOCKED() { mtx_.CheckLocked(); }
+  void Unlock() RELEASE() { mtx_.Unlock(); }
 
   // Should be guarded by ThreadRegistryLock.
   ThreadContextBase *GetThreadLocked(u32 tid) {
-    DCHECK_LT(tid, n_contexts_);
-    return threads_[tid];
+    return threads_.empty() ? nullptr : threads_[tid];
   }
 
   u32 CreateThread(uptr user_id, bool detached, u32 parent_tid, void *arg);
@@ -137,15 +137,13 @@ class ThreadRegistry {
 
   BlockingMutex mtx_;
 
-  u32 n_contexts_;      // Number of created thread contexts,
-                        // at most max_threads_.
   u64 total_threads_;   // Total number of created threads. May be greater than
                         // max_threads_ if contexts were reused.
   uptr alive_threads_;  // Created or running.
   uptr max_alive_threads_;
   uptr running_threads_;
 
-  ThreadContextBase **threads_;  // Array of thread contexts is leaked.
+  InternalMmapVector<ThreadContextBase *> threads_;
   IntrusiveList<ThreadContextBase> dead_threads_;
   IntrusiveList<ThreadContextBase> invalid_threads_;
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_thread_safety.h b/libsanitizer/sanitizer_common/sanitizer_thread_safety.h
new file mode 100644
index 00000000000..52b25edaa7a
--- /dev/null
+++ b/libsanitizer/sanitizer_common/sanitizer_thread_safety.h
@@ -0,0 +1,42 @@
+//===-- sanitizer_thread_safety.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 shared between sanitizer tools.
+//
+// Wrappers around thread safety annotations.
+// https://clang.llvm.org/docs/ThreadSafetyAnalysis.html
+//===----------------------------------------------------------------------===//
+
+#ifndef SANITIZER_THREAD_SAFETY_H
+#define SANITIZER_THREAD_SAFETY_H
+
+#if defined(__clang__)
+#  define THREAD_ANNOTATION(x) __attribute__((x))
+#else
+#  define THREAD_ANNOTATION(x)
+#endif
+
+#define MUTEX THREAD_ANNOTATION(capability("mutex"))
+#define SCOPED_LOCK THREAD_ANNOTATION(scoped_lockable)
+#define GUARDED_BY(x) THREAD_ANNOTATION(guarded_by(x))
+#define PT_GUARDED_BY(x) THREAD_ANNOTATION(pt_guarded_by(x))
+#define REQUIRES(...) THREAD_ANNOTATION(requires_capability(__VA_ARGS__))
+#define REQUIRES_SHARED(...) \
+  THREAD_ANNOTATION(requires_shared_capability(__VA_ARGS__))
+#define ACQUIRE(...) THREAD_ANNOTATION(acquire_capability(__VA_ARGS__))
+#define ACQUIRE_SHARED(...) \
+  THREAD_ANNOTATION(acquire_shared_capability(__VA_ARGS__))
+#define TRY_ACQUIRE(...) THREAD_ANNOTATION(try_acquire_capability(__VA_ARGS__))
+#define RELEASE(...) THREAD_ANNOTATION(release_capability(__VA_ARGS__))
+#define RELEASE_SHARED(...) \
+  THREAD_ANNOTATION(release_shared_capability(__VA_ARGS__))
+#define EXCLUDES(...) THREAD_ANNOTATION(locks_excluded(__VA_ARGS__))
+#define CHECK_LOCKED(...) THREAD_ANNOTATION(assert_capability(__VA_ARGS__))
+#define NO_THREAD_SAFETY_ANALYSIS THREAD_ANNOTATION(no_thread_safety_analysis)
+
+#endif
diff --git a/libsanitizer/sanitizer_common/sanitizer_win.cpp b/libsanitizer/sanitizer_common/sanitizer_win.cpp
index f383e130fa5..dddd885a45d 100644
--- a/libsanitizer/sanitizer_common/sanitizer_win.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_win.cpp
@@ -44,6 +44,9 @@ TRACELOGGING_DEFINE_PROVIDER(g_asan_provider, "AddressSanitizerLoggingProvider",
 #define TraceLoggingUnregister(x)
 #endif
 
+// For WaitOnAddress
+#  pragma comment(lib, "synchronization.lib")
+
 // A macro to tell the compiler that this part of the code cannot be reached,
 // if the compiler supports this feature. Since we're using this in
 // code that is called when terminating the process, the expansion of the
@@ -541,13 +544,7 @@ bool IsAbsolutePath(const char *path) {
          IsPathSeparator(path[2]);
 }
 
-void SleepForSeconds(int seconds) {
-  Sleep(seconds * 1000);
-}
-
-void SleepForMillis(int millis) {
-  Sleep(millis);
-}
+void internal_usleep(u64 useconds) { Sleep(useconds / 1000); }
 
 u64 NanoTime() {
   static LARGE_INTEGER frequency = {};
@@ -819,6 +816,17 @@ uptr GetRSS() {
 void *internal_start_thread(void *(*func)(void *arg), void *arg) { return 0; }
 void internal_join_thread(void *th) { }
 
+void FutexWait(atomic_uint32_t *p, u32 cmp) {
+  WaitOnAddress(p, &cmp, sizeof(cmp), INFINITE);
+}
+
+void FutexWake(atomic_uint32_t *p, u32 count) {
+  if (count == 1)
+    WakeByAddressSingle(p);
+  else
+    WakeByAddressAll(p);
+}
+
 // ---------------------- BlockingMutex ---------------- {{{1
 
 BlockingMutex::BlockingMutex() {
@@ -838,9 +846,7 @@ void BlockingMutex::Unlock() {
   ReleaseSRWLockExclusive((PSRWLOCK)opaque_storage_);
 }
 
-void BlockingMutex::CheckLocked() {
-  CHECK_EQ(owner_, GetThreadSelf());
-}
+void BlockingMutex::CheckLocked() const { CHECK_EQ(owner_, GetThreadSelf()); }
 
 uptr GetTlsSize() {
   return 0;
diff --git a/libsanitizer/tsan/Makefile.am b/libsanitizer/tsan/Makefile.am
index fb20ff7e311..dcb247ec032 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 ce11d2497b7..83617cf6ab7 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 8e5188392ca..61848c21d16 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 f53787aeba9..5c8f2801b0c 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 29576ea2d49..c5716f53a32 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 ed10fccc980..2d400c7e709 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 2651e22c39f..6808f2e0e2d 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).
@@ -2597,17 +2613,17 @@ 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
@@ -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 6e022b56850..124aa2fd214 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 99516d94bba..175855f66c8 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 3f459aff532..21fe4a19619 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 45a39f0f8ec..7765bc07052 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 7a0918f2a2c..d8b1826ee47 100644
--- a/libsanitizer/tsan/tsan_mutex.cpp
+++ b/libsanitizer/tsan/tsan_mutex.cpp
@@ -207,14 +207,11 @@ 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 80fdc6ed57b..9a579eaf914 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 101522d8fa4..8bd218e25fd 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 e5b6690edfd..cfe597e5380 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 73e1d4577c2..1c6198cefcd 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 0efa99788ab..bcf489a71d5 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 3ae519d34da..6576d40aa48 100644
--- a/libsanitizer/tsan/tsan_rtl.h
+++ b/libsanitizer/tsan/tsan_rtl.h
@@ -399,9 +399,6 @@ struct ThreadState {
 #if !SANITIZER_GO
   Vector<JmpBuf> jmp_bufs;
   int ignore_interceptors;
-#endif
-#if TSAN_COLLECT_STATS
-  u64 stat[StatCnt];
 #endif
   const u32 tid;
   const int unique_id;
@@ -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 0a8f3aa3ddb..a214a336d69 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 9e533a71a9c..8285e21aa1e 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 00000000000..fcff35fbc7e
--- /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 6d1ccd8c9c7..cdb6e60ebbd 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 403a21ae4ae..6c703d7f2b1 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 78f3cce9138..00000000000
--- 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 8b26a59bb2e..00000000000
--- 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 ba24f98ae9f..d25434af52a 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 fbd0f72db6e..9f2677b778d 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 056c3aa2032..d23dfb0ba06 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);
diff --git a/libsanitizer/ubsan/ubsan_diag_standalone.cpp b/libsanitizer/ubsan/ubsan_diag_standalone.cpp
index 300179adae2..5526ae05165 100644
--- a/libsanitizer/ubsan/ubsan_diag_standalone.cpp
+++ b/libsanitizer/ubsan/ubsan_diag_standalone.cpp
@@ -20,11 +20,9 @@ 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, bp, context, 0, 0, false);
+  GetThreadStackTopAndBottom(false, &top, &bottom);
+  bool fast = StackTrace::WillUseFastUnwind(request_fast);
+  Unwind(max_depth, pc, bp, context, top, bottom, fast);
 }
 
 extern "C" {
diff --git a/libsanitizer/ubsan/ubsan_flags.cpp b/libsanitizer/ubsan/ubsan_flags.cpp
index 9a66bd37518..25cefd46ce2 100644
--- a/libsanitizer/ubsan/ubsan_flags.cpp
+++ b/libsanitizer/ubsan/ubsan_flags.cpp
@@ -50,7 +50,6 @@ void InitializeFlags() {
   {
     CommonFlags cf;
     cf.CopyFrom(*common_flags());
-    cf.print_summary = false;
     cf.external_symbolizer_path = GetFlag("UBSAN_SYMBOLIZER_PATH");
     OverrideCommonFlags(cf);
   }
diff --git a/libsanitizer/ubsan/ubsan_handlers.cpp b/libsanitizer/ubsan/ubsan_handlers.cpp
index 2184625aa6e..e201e6bba22 100644
--- a/libsanitizer/ubsan/ubsan_handlers.cpp
+++ b/libsanitizer/ubsan/ubsan_handlers.cpp
@@ -894,21 +894,6 @@ void __ubsan_handle_cfi_bad_type(CFICheckFailData *Data, ValueHandle Vtable,
 
 }  // namespace __ubsan
 
-void __ubsan::__ubsan_handle_cfi_bad_icall(CFIBadIcallData *CallData,
-                                           ValueHandle Function) {
-  GET_REPORT_OPTIONS(false);
-  CFICheckFailData Data = {CFITCK_ICall, CallData->Loc, CallData->Type};
-  handleCFIBadIcall(&Data, Function, Opts);
-}
-
-void __ubsan::__ubsan_handle_cfi_bad_icall_abort(CFIBadIcallData *CallData,
-                                                 ValueHandle Function) {
-  GET_REPORT_OPTIONS(true);
-  CFICheckFailData Data = {CFITCK_ICall, CallData->Loc, CallData->Type};
-  handleCFIBadIcall(&Data, Function, Opts);
-  Die();
-}
-
 void __ubsan::__ubsan_handle_cfi_check_fail(CFICheckFailData *Data,
                                             ValueHandle Value,
                                             uptr ValidVtable) {
diff --git a/libsanitizer/ubsan/ubsan_handlers.h b/libsanitizer/ubsan/ubsan_handlers.h
index 9f412353fc0..219fb15de55 100644
--- a/libsanitizer/ubsan/ubsan_handlers.h
+++ b/libsanitizer/ubsan/ubsan_handlers.h
@@ -215,20 +215,12 @@ enum CFITypeCheckKind : unsigned char {
   CFITCK_VMFCall,
 };
 
-struct CFIBadIcallData {
-  SourceLocation Loc;
-  const TypeDescriptor &Type;
-};
-
 struct CFICheckFailData {
   CFITypeCheckKind CheckKind;
   SourceLocation Loc;
   const TypeDescriptor &Type;
 };
 
-/// \brief Handle control flow integrity failure for indirect function calls.
-RECOVERABLE(cfi_bad_icall, CFIBadIcallData *Data, ValueHandle Function)
-
 /// \brief Handle control flow integrity failures.
 RECOVERABLE(cfi_check_fail, CFICheckFailData *Data, ValueHandle Function,
             uptr VtableIsValid)
diff --git a/libsanitizer/ubsan/ubsan_platform.h b/libsanitizer/ubsan/ubsan_platform.h
index 42944a0b3d0..d2cc2e10bd2 100644
--- a/libsanitizer/ubsan/ubsan_platform.h
+++ b/libsanitizer/ubsan/ubsan_platform.h
@@ -12,16 +12,14 @@
 #ifndef UBSAN_PLATFORM_H
 #define UBSAN_PLATFORM_H
 
-#ifndef CAN_SANITIZE_UB
 // Other platforms should be easy to add, and probably work as-is.
 #if defined(__linux__) || defined(__FreeBSD__) || defined(__APPLE__) ||        \
-    defined(__NetBSD__) || defined(__DragonFly__) || \
-    (defined(__sun__) && defined(__svr4__)) || \
-    defined(_WIN32) || defined(__Fuchsia__) || defined(__rtems__)
-# define CAN_SANITIZE_UB 1
+    defined(__NetBSD__) || defined(__DragonFly__) ||                           \
+    (defined(__sun__) && defined(__svr4__)) || defined(_WIN32) ||              \
+    defined(__Fuchsia__)
+#define CAN_SANITIZE_UB 1
 #else
 # define CAN_SANITIZE_UB 0
 #endif
-#endif //CAN_SANITIZE_UB
 
 #endif