libsanitizer: merge from master

Merged revision: f58e0513dd95944b81ce7a6e7b49ba656de7d75f

153 files changed, 2538 insertions, 1239 deletions

diff --git a/libsanitizer/MERGE b/libsanitizer/MERGE
index 0fb64a9..c4731d0 100644
--- a/libsanitizer/MERGE
+++ b/libsanitizer/MERGE
@@ -1,4 +1,4 @@
-6e7dd1e3e1170080b76b5dcc5716bdd974343233
+f58e0513dd95944b81ce7a6e7b49ba656de7d75f
 
 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/asan_allocator.cpp b/libsanitizer/asan/asan_allocator.cpp
index 58b496a..7c8bb50 100644
--- a/libsanitizer/asan/asan_allocator.cpp
+++ b/libsanitizer/asan/asan_allocator.cpp
@@ -476,7 +476,7 @@ struct Allocator {
       return false;
     if (m->Beg() != addr) return false;
     AsanThread *t = GetCurrentThread();
-    m->SetAllocContext(t ? t->tid() : 0, StackDepotPut(*stack));
+    m->SetAllocContext(t ? t->tid() : kMainTid, StackDepotPut(*stack));
     return true;
   }
 
@@ -570,7 +570,7 @@ struct Allocator {
     m->SetUsedSize(size);
     m->user_requested_alignment_log = user_requested_alignment_log;
 
-    m->SetAllocContext(t ? t->tid() : 0, StackDepotPut(*stack));
+    m->SetAllocContext(t ? t->tid() : kMainTid, StackDepotPut(*stack));
 
     uptr size_rounded_down_to_granularity =
         RoundDownTo(size, SHADOW_GRANULARITY);
@@ -1183,6 +1183,34 @@ IgnoreObjectResult IgnoreObjectLocked(const void *p) {
   m->lsan_tag = __lsan::kIgnored;
   return kIgnoreObjectSuccess;
 }
+
+void GetAdditionalThreadContextPtrs(ThreadContextBase *tctx, void *ptrs) {
+  // Look for the arg pointer of threads that have been created or are running.
+  // This is necessary to prevent false positive leaks due to the AsanThread
+  // holding the only live reference to a heap object.  This can happen because
+  // the `pthread_create()` interceptor doesn't wait for the child thread to
+  // start before returning and thus loosing the the only live reference to the
+  // heap object on the stack.
+
+  __asan::AsanThreadContext *atctx =
+      reinterpret_cast<__asan::AsanThreadContext *>(tctx);
+  __asan::AsanThread *asan_thread = atctx->thread;
+
+  // Note ThreadStatusRunning is required because there is a small window where
+  // the thread status switches to `ThreadStatusRunning` but the `arg` pointer
+  // still isn't on the stack yet.
+  if (atctx->status != ThreadStatusCreated &&
+      atctx->status != ThreadStatusRunning)
+    return;
+
+  uptr thread_arg = reinterpret_cast<uptr>(asan_thread->get_arg());
+  if (!thread_arg)
+    return;
+
+  auto ptrsVec = reinterpret_cast<InternalMmapVector<uptr> *>(ptrs);
+  ptrsVec->push_back(thread_arg);
+}
+
 }  // namespace __lsan
 
 // ---------------------- Interface ---------------- {{{1
diff --git a/libsanitizer/asan/asan_descriptions.cpp b/libsanitizer/asan/asan_descriptions.cpp
index 153c874..2ba8a02 100644
--- a/libsanitizer/asan/asan_descriptions.cpp
+++ b/libsanitizer/asan/asan_descriptions.cpp
@@ -44,11 +44,11 @@ void DescribeThread(AsanThreadContext *context) {
   CHECK(context);
   asanThreadRegistry().CheckLocked();
   // No need to announce the main thread.
-  if (context->tid == 0 || context->announced) {
+  if (context->tid == kMainTid || context->announced) {
     return;
   }
   context->announced = true;
-  InternalScopedString str(1024);
+  InternalScopedString str;
   str.append("Thread %s", AsanThreadIdAndName(context).c_str());
   if (context->parent_tid == kInvalidTid) {
     str.append(" created by unknown thread\n");
@@ -77,7 +77,6 @@ static bool GetShadowKind(uptr addr, ShadowKind *shadow_kind) {
   } else if (AddrIsInLowShadow(addr)) {
     *shadow_kind = kShadowKindLow;
   } else {
-    CHECK(0 && "Address is not in memory and not in shadow?");
     return false;
   }
   return true;
@@ -126,7 +125,7 @@ static void GetAccessToHeapChunkInformation(ChunkAccess *descr,
 
 static void PrintHeapChunkAccess(uptr addr, const ChunkAccess &descr) {
   Decorator d;
-  InternalScopedString str(4096);
+  InternalScopedString str;
   str.append("%s", d.Location());
   switch (descr.access_type) {
     case kAccessTypeLeft:
@@ -243,7 +242,7 @@ static void PrintAccessAndVarIntersection(const StackVarDescr &var, uptr addr,
     else if (addr >= prev_var_end && addr - prev_var_end >= var.beg - addr_end)
       pos_descr = "underflows";
   }
-  InternalScopedString str(1024);
+  InternalScopedString str;
   str.append("    [%zd, %zd)", var.beg, var_end);
   // Render variable name.
   str.append(" '");
@@ -276,7 +275,7 @@ bool DescribeAddressIfStack(uptr addr, uptr access_size) {
 // Global descriptions
 static void DescribeAddressRelativeToGlobal(uptr addr, uptr access_size,
                                             const __asan_global &g) {
-  InternalScopedString str(4096);
+  InternalScopedString str;
   Decorator d;
   str.append("%s", d.Location());
   if (addr < g.beg) {
@@ -464,7 +463,13 @@ AddressDescription::AddressDescription(uptr addr, uptr access_size,
     return;
   }
   data.kind = kAddressKindWild;
-  addr = 0;
+  data.wild.addr = addr;
+  data.wild.access_size = access_size;
+}
+
+void WildAddressDescription::Print() const {
+  Printf("Address %p is a wild pointer inside of access range of size %p.\n",
+         addr, access_size);
 }
 
 void PrintAddressDescription(uptr addr, uptr access_size,
diff --git a/libsanitizer/asan/asan_descriptions.h b/libsanitizer/asan/asan_descriptions.h
index ee0e206..650e2eb 100644
--- a/libsanitizer/asan/asan_descriptions.h
+++ b/libsanitizer/asan/asan_descriptions.h
@@ -146,6 +146,13 @@ struct StackAddressDescription {
 bool GetStackAddressInformation(uptr addr, uptr access_size,
                                 StackAddressDescription *descr);
 
+struct WildAddressDescription {
+  uptr addr;
+  uptr access_size;
+
+  void Print() const;
+};
+
 struct GlobalAddressDescription {
   uptr addr;
   // Assume address is close to at most four globals.
@@ -193,7 +200,7 @@ class AddressDescription {
       HeapAddressDescription heap;
       StackAddressDescription stack;
       GlobalAddressDescription global;
-      uptr addr;
+      WildAddressDescription wild;
     };
   };
 
@@ -211,7 +218,7 @@ class AddressDescription {
   uptr Address() const {
     switch (data.kind) {
       case kAddressKindWild:
-        return data.addr;
+        return data.wild.addr;
       case kAddressKindShadow:
         return data.shadow.addr;
       case kAddressKindHeap:
@@ -226,7 +233,7 @@ class AddressDescription {
   void Print(const char *bug_descr = nullptr) const {
     switch (data.kind) {
       case kAddressKindWild:
-        Printf("Address %p is a wild pointer.\n", data.addr);
+        data.wild.Print();
         return;
       case kAddressKindShadow:
         return data.shadow.Print();
diff --git a/libsanitizer/asan/asan_errors.cpp b/libsanitizer/asan/asan_errors.cpp
index 541c6e0..e68e697 100644
--- a/libsanitizer/asan/asan_errors.cpp
+++ b/libsanitizer/asan/asan_errors.cpp
@@ -343,7 +343,8 @@ void ErrorODRViolation::Print() {
   Report("ERROR: AddressSanitizer: %s (%p):\n", scariness.GetDescription(),
          global1.beg);
   Printf("%s", d.Default());
-  InternalScopedString g1_loc(256), g2_loc(256);
+  InternalScopedString g1_loc;
+  InternalScopedString g2_loc;
   PrintGlobalLocation(&g1_loc, global1);
   PrintGlobalLocation(&g2_loc, global2);
   Printf("  [1] size=%zd '%s' %s\n", global1.size,
@@ -360,7 +361,7 @@ void ErrorODRViolation::Print() {
   Report(
       "HINT: if you don't care about these errors you may set "
       "ASAN_OPTIONS=detect_odr_violation=0\n");
-  InternalScopedString error_msg(256);
+  InternalScopedString error_msg;
   error_msg.append("%s: global '%s' at %s", scariness.GetDescription(),
                    MaybeDemangleGlobalName(global1.name), g1_loc.data());
   ReportErrorSummary(error_msg.data());
@@ -554,7 +555,7 @@ static void PrintShadowMemoryForAddress(uptr addr) {
   uptr shadow_addr = MemToShadow(addr);
   const uptr n_bytes_per_row = 16;
   uptr aligned_shadow = shadow_addr & ~(n_bytes_per_row - 1);
-  InternalScopedString str(4096 * 8);
+  InternalScopedString str;
   str.append("Shadow bytes around the buggy address:\n");
   for (int i = -5; i <= 5; i++) {
     uptr row_shadow_addr = aligned_shadow + i * n_bytes_per_row;
diff --git a/libsanitizer/asan/asan_fake_stack.cpp b/libsanitizer/asan/asan_fake_stack.cpp
index 295e6de..1f873fe 100644
--- a/libsanitizer/asan/asan_fake_stack.cpp
+++ b/libsanitizer/asan/asan_fake_stack.cpp
@@ -65,7 +65,7 @@ FakeStack *FakeStack::Create(uptr stack_size_log) {
 void FakeStack::Destroy(int tid) {
   PoisonAll(0);
   if (Verbosity() >= 2) {
-    InternalScopedString str(kNumberOfSizeClasses * 50);
+    InternalScopedString str;
     for (uptr class_id = 0; class_id < kNumberOfSizeClasses; class_id++)
       str.append("%zd: %zd/%zd; ", class_id, hint_position_[class_id],
                  NumberOfFrames(stack_size_log(), class_id));
diff --git a/libsanitizer/asan/asan_fuchsia.cpp b/libsanitizer/asan/asan_fuchsia.cpp
index 6c61344..b0c7255 100644
--- a/libsanitizer/asan/asan_fuchsia.cpp
+++ b/libsanitizer/asan/asan_fuchsia.cpp
@@ -81,7 +81,7 @@ void AsanTSDInit(void (*destructor)(void *tsd)) {
 void PlatformTSDDtor(void *tsd) { UNREACHABLE(__func__); }
 
 static inline size_t AsanThreadMmapSize() {
-  return RoundUpTo(sizeof(AsanThread), PAGE_SIZE);
+  return RoundUpTo(sizeof(AsanThread), _zx_system_get_page_size());
 }
 
 struct AsanThread::InitOptions {
diff --git a/libsanitizer/asan/asan_globals.cpp b/libsanitizer/asan/asan_globals.cpp
index e045c31..9d7dbc6 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 4e68b3b..9db7db8 100644
--- a/libsanitizer/asan/asan_interceptors.cpp
+++ b/libsanitizer/asan/asan_interceptors.cpp
@@ -191,20 +191,11 @@ DECLARE_REAL_AND_INTERCEPTOR(void, free, void *)
 #include "sanitizer_common/sanitizer_common_syscalls.inc"
 #include "sanitizer_common/sanitizer_syscalls_netbsd.inc"
 
-struct ThreadStartParam {
-  atomic_uintptr_t t;
-  atomic_uintptr_t is_registered;
-};
-
 #if ASAN_INTERCEPT_PTHREAD_CREATE
 static thread_return_t THREAD_CALLING_CONV asan_thread_start(void *arg) {
-  ThreadStartParam *param = reinterpret_cast<ThreadStartParam *>(arg);
-  AsanThread *t = nullptr;
-  while ((t = reinterpret_cast<AsanThread *>(
-              atomic_load(&param->t, memory_order_acquire))) == nullptr)
-    internal_sched_yield();
+  AsanThread *t = (AsanThread *)arg;
   SetCurrentThread(t);
-  return t->ThreadStart(GetTid(), &param->is_registered);
+  return t->ThreadStart(GetTid());
 }
 
 INTERCEPTOR(int, pthread_create, void *thread,
@@ -217,9 +208,11 @@ INTERCEPTOR(int, pthread_create, void *thread,
   int detached = 0;
   if (attr)
     REAL(pthread_attr_getdetachstate)(attr, &detached);
-  ThreadStartParam param;
-  atomic_store(&param.t, 0, memory_order_relaxed);
-  atomic_store(&param.is_registered, 0, memory_order_relaxed);
+
+  u32 current_tid = GetCurrentTidOrInvalid();
+  AsanThread *t =
+      AsanThread::Create(start_routine, arg, current_tid, &stack, detached);
+
   int result;
   {
     // Ignore all allocations made by pthread_create: thread stack/TLS may be
@@ -229,21 +222,13 @@ INTERCEPTOR(int, pthread_create, void *thread,
 #if CAN_SANITIZE_LEAKS
     __lsan::ScopedInterceptorDisabler disabler;
 #endif
-    result = REAL(pthread_create)(thread, attr, asan_thread_start, &param);
+    result = REAL(pthread_create)(thread, attr, asan_thread_start, t);
   }
-  if (result == 0) {
-    u32 current_tid = GetCurrentTidOrInvalid();
-    AsanThread *t =
-        AsanThread::Create(start_routine, arg, current_tid, &stack, detached);
-    atomic_store(&param.t, reinterpret_cast<uptr>(t), memory_order_release);
-    // Wait until the AsanThread object is initialized and the ThreadRegistry
-    // entry is in "started" state. One reason for this is that after this
-    // interceptor exits, the child thread's stack may be the only thing holding
-    // the |arg| pointer. This may cause LSan to report a leak if leak checking
-    // happens at a point when the interceptor has already exited, but the stack
-    // range for the child thread is not yet known.
-    while (atomic_load(&param.is_registered, memory_order_acquire) == 0)
-      internal_sched_yield();
+  if (result != 0) {
+    // If the thread didn't start delete the AsanThread to avoid leaking it.
+    // Note AsanThreadContexts never get destroyed so the AsanThreadContext
+    // that was just created for the AsanThread is wasted.
+    t->Destroy();
   }
   return result;
 }
diff --git a/libsanitizer/asan/asan_interceptors.h b/libsanitizer/asan/asan_interceptors.h
index 56dc34b..e8c58c2 100644
--- a/libsanitizer/asan/asan_interceptors.h
+++ b/libsanitizer/asan/asan_interceptors.h
@@ -60,7 +60,7 @@ void InitializePlatformInterceptors();
 # define ASAN_USE_ALIAS_ATTRIBUTE_FOR_INDEX 0
 #endif
 
-#if (SANITIZER_LINUX && !SANITIZER_ANDROID) || SANITIZER_SOLARIS
+#if SANITIZER_GLIBC || SANITIZER_SOLARIS
 # define ASAN_INTERCEPT_SWAPCONTEXT 1
 #else
 # define ASAN_INTERCEPT_SWAPCONTEXT 0
@@ -72,7 +72,7 @@ void InitializePlatformInterceptors();
 # define ASAN_INTERCEPT_SIGLONGJMP 0
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
 # define ASAN_INTERCEPT___LONGJMP_CHK 1
 #else
 # define ASAN_INTERCEPT___LONGJMP_CHK 0
@@ -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
@@ -111,7 +106,7 @@ void InitializePlatformInterceptors();
 # define ASAN_INTERCEPT_ATEXIT 0
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
 # define ASAN_INTERCEPT___STRDUP 1
 #else
 # define ASAN_INTERCEPT___STRDUP 0
@@ -139,10 +134,10 @@ DECLARE_REAL(uptr, strnlen, const char *s, uptr maxlen)
 DECLARE_REAL(char*, strstr, const char *s1, const char *s2)
 
 #if !SANITIZER_MAC
-#define ASAN_INTERCEPT_FUNC(name)                                         \
-  do {                                                                    \
-    if (!INTERCEPT_FUNCTION(name))                                        \
-      VReport(1, "AddressSanitizer: failed to intercept '%s'\n'", #name); \
+#define ASAN_INTERCEPT_FUNC(name)                                        \
+  do {                                                                   \
+    if (!INTERCEPT_FUNCTION(name))                                       \
+      VReport(1, "AddressSanitizer: failed to intercept '%s'\n", #name); \
   } while (0)
 #define ASAN_INTERCEPT_FUNC_VER(name, ver)                                  \
   do {                                                                      \
diff --git a/libsanitizer/asan/asan_linux.cpp b/libsanitizer/asan/asan_linux.cpp
index fb1a442..4bcbe5d 100644
--- a/libsanitizer/asan/asan_linux.cpp
+++ b/libsanitizer/asan/asan_linux.cpp
@@ -55,6 +55,7 @@ extern Elf_Dyn _DYNAMIC;
 #else
 #include <sys/ucontext.h>
 #include <link.h>
+extern ElfW(Dyn) _DYNAMIC[];
 #endif
 
 // x86-64 FreeBSD 9.2 and older define 'ucontext_t' incorrectly in
@@ -84,7 +85,7 @@ bool IsSystemHeapAddress (uptr addr) { return false; }
 
 void *AsanDoesNotSupportStaticLinkage() {
   // This will fail to link with -static.
-  return &_DYNAMIC;  // defined in link.h
+  return &_DYNAMIC;
 }
 
 #if ASAN_PREMAP_SHADOW
diff --git a/libsanitizer/asan/asan_mapping.h b/libsanitizer/asan/asan_mapping.h
index a7136de..455e236 100644
--- a/libsanitizer/asan/asan_mapping.h
+++ b/libsanitizer/asan/asan_mapping.h
@@ -72,6 +72,13 @@
 // || `[0x2000000000, 0x23ffffffff]` || LowShadow  ||
 // || `[0x0000000000, 0x1fffffffff]` || LowMem     ||
 //
+// Default Linux/RISCV64 Sv39 mapping:
+// || `[0x1555550000, 0x3fffffffff]` || HighMem    ||
+// || `[0x0fffffa000, 0x1555555fff]` || HighShadow ||
+// || `[0x0effffa000, 0x0fffff9fff]` || ShadowGap  ||
+// || `[0x0d55550000, 0x0effff9fff]` || LowShadow  ||
+// || `[0x0000000000, 0x0d5554ffff]` || LowMem     ||
+//
 // Default Linux/AArch64 (39-bit VMA) mapping:
 // || `[0x2000000000, 0x7fffffffff]` || highmem    ||
 // || `[0x1400000000, 0x1fffffffff]` || highshadow ||
@@ -79,20 +86,6 @@
 // || `[0x1000000000, 0x11ffffffff]` || lowshadow  ||
 // || `[0x0000000000, 0x0fffffffff]` || lowmem     ||
 //
-// RISC-V has only 38 bits for task size
-// Low mem size is set with kRiscv64_ShadowOffset64 in
-// compiler-rt/lib/asan/asan_allocator.h and in
-// llvm/lib/Transforms/Instrumentation/AddressSanitizer.cpp with
-// kRiscv64_ShadowOffset64, High mem top border is set with
-// GetMaxVirtualAddress() in
-// compiler-rt/lib/sanitizer_common/sanitizer_linux.cpp
-// Default Linux/RISCV64 Sv39/Sv48 mapping:
-// || `[0x000820000000, 0x003fffffffff]` || HighMem    ||
-// || `[0x000124000000, 0x00081fffffff]` || HighShadow ||
-// || `[0x000024000000, 0x000123ffffff]` || ShadowGap  ||
-// || `[0x000020000000, 0x000023ffffff]` || LowShadow  ||
-// || `[0x000000000000, 0x00001fffffff]` || LowMem     ||
-//
 // Default Linux/AArch64 (42-bit VMA) mapping:
 // || `[0x10000000000, 0x3ffffffffff]` || highmem    ||
 // || `[0x0a000000000, 0x0ffffffffff]` || highshadow ||
@@ -175,10 +168,10 @@ static const u64 kDefaultShadowOffset64 = 1ULL << 44;
 static const u64 kDefaultShort64bitShadowOffset =
     0x7FFFFFFF & (~0xFFFULL << kDefaultShadowScale);  // < 2G.
 static const u64 kAArch64_ShadowOffset64 = 1ULL << 36;
-static const u64 kRiscv64_ShadowOffset64 = 0x20000000;
+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
diff --git a/libsanitizer/asan/asan_new_delete.cpp b/libsanitizer/asan/asan_new_delete.cpp
index 5dfcc00..92a8648 100644
--- a/libsanitizer/asan/asan_new_delete.cpp
+++ b/libsanitizer/asan/asan_new_delete.cpp
@@ -45,7 +45,7 @@ COMMENT_EXPORT("??_V@YAXPAX@Z")                   // operator delete[]
 #endif
 #undef COMMENT_EXPORT
 #else
-#define CXX_OPERATOR_ATTRIBUTE INTERCEPTOR_ATTRIBUTE
+#define CXX_OPERATOR_ATTRIBUTE INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
 #endif
 
 using namespace __asan;
diff --git a/libsanitizer/asan/asan_poisoning.cpp b/libsanitizer/asan/asan_poisoning.cpp
index 44f872e..fa149ec 100644
--- a/libsanitizer/asan/asan_poisoning.cpp
+++ b/libsanitizer/asan/asan_poisoning.cpp
@@ -364,7 +364,7 @@ void __sanitizer_annotate_contiguous_container(const void *beg_p,
                                                  &stack);
   }
   CHECK_LE(end - beg,
-           FIRST_32_SECOND_64(1UL << 30, 1ULL << 34)); // Sanity check.
+           FIRST_32_SECOND_64(1UL << 30, 1ULL << 40)); // Sanity check.
 
   uptr a = RoundDownTo(Min(old_mid, new_mid), granularity);
   uptr c = RoundUpTo(Max(old_mid, new_mid), granularity);
diff --git a/libsanitizer/asan/asan_posix.cpp b/libsanitizer/asan/asan_posix.cpp
index d7f19d8..63ad735 100644
--- a/libsanitizer/asan/asan_posix.cpp
+++ b/libsanitizer/asan/asan_posix.cpp
@@ -56,7 +56,7 @@ bool PlatformUnpoisonStacks() {
   if (signal_stack.ss_flags != SS_ONSTACK)
     return false;
 
-  // Since we're on the signal altnerate stack, we cannot find the DEFAULT
+  // Since we're on the signal alternate stack, we cannot find the DEFAULT
   // stack bottom using a local variable.
   uptr default_bottom, tls_addr, tls_size, stack_size;
   GetThreadStackAndTls(/*main=*/false, &default_bottom, &stack_size, &tls_addr,
diff --git a/libsanitizer/asan/asan_rtl.cpp b/libsanitizer/asan/asan_rtl.cpp
index 7b5a929..e715d77 100644
--- a/libsanitizer/asan/asan_rtl.cpp
+++ b/libsanitizer/asan/asan_rtl.cpp
@@ -62,19 +62,9 @@ static void AsanDie() {
   }
 }
 
-static void AsanCheckFailed(const char *file, int line, const char *cond,
-                            u64 v1, u64 v2) {
-  Report("AddressSanitizer CHECK failed: %s:%d \"%s\" (0x%zx, 0x%zx)\n", file,
-         line, cond, (uptr)v1, (uptr)v2);
-
-  // Print a stack trace the first time we come here. Otherwise, we probably
-  // failed a CHECK during symbolization.
-  static atomic_uint32_t num_calls;
-  if (atomic_fetch_add(&num_calls, 1, memory_order_relaxed) == 0) {
-    PRINT_CURRENT_STACK_CHECK();
-  }
-
-  Die();
+static void CheckUnwind() {
+  GET_STACK_TRACE(kStackTraceMax, common_flags()->fast_unwind_on_check);
+  stack.Print();
 }
 
 // -------------------------- Globals --------------------- {{{1
@@ -432,7 +422,7 @@ static void AsanInitInternal() {
 
   // Install tool-specific callbacks in sanitizer_common.
   AddDieCallback(AsanDie);
-  SetCheckFailedCallback(AsanCheckFailed);
+  SetCheckUnwindCallback(CheckUnwind);
   SetPrintfAndReportCallback(AppendToErrorMessageBuffer);
 
   __sanitizer_set_report_path(common_flags()->log_path);
@@ -568,7 +558,7 @@ void UnpoisonStack(uptr bottom, uptr top, const char *type) {
         type, top, bottom, top - bottom, top - bottom);
     return;
   }
-  PoisonShadow(bottom, top - bottom, 0);
+  PoisonShadow(bottom, RoundUpTo(top - bottom, SHADOW_GRANULARITY), 0);
 }
 
 static void UnpoisonDefaultStack() {
diff --git a/libsanitizer/asan/asan_stack.h b/libsanitizer/asan/asan_stack.h
index 47ca85a..b9575d2 100644
--- a/libsanitizer/asan/asan_stack.h
+++ b/libsanitizer/asan/asan_stack.h
@@ -54,9 +54,6 @@ u32 GetMallocContextSize();
 #define GET_STACK_TRACE_FATAL_HERE                                \
   GET_STACK_TRACE(kStackTraceMax, common_flags()->fast_unwind_on_fatal)
 
-#define GET_STACK_TRACE_CHECK_HERE                                \
-  GET_STACK_TRACE(kStackTraceMax, common_flags()->fast_unwind_on_check)
-
 #define GET_STACK_TRACE_THREAD                                    \
   GET_STACK_TRACE(kStackTraceMax, true)
 
@@ -71,10 +68,4 @@ u32 GetMallocContextSize();
     stack.Print();              \
   }
 
-#define PRINT_CURRENT_STACK_CHECK() \
-  {                                 \
-    GET_STACK_TRACE_CHECK_HERE;     \
-    stack.Print();                  \
-  }
-
 #endif // ASAN_STACK_H
diff --git a/libsanitizer/asan/asan_thread.cpp b/libsanitizer/asan/asan_thread.cpp
index fb09af0..9c3c86f 100644
--- a/libsanitizer/asan/asan_thread.cpp
+++ b/libsanitizer/asan/asan_thread.cpp
@@ -100,18 +100,27 @@ void AsanThread::Destroy() {
   int tid = this->tid();
   VReport(1, "T%d exited\n", tid);
 
-  malloc_storage().CommitBack();
-  if (common_flags()->use_sigaltstack) UnsetAlternateSignalStack();
-  asanThreadRegistry().FinishThread(tid);
-  FlushToDeadThreadStats(&stats_);
-  // We also clear the shadow on thread destruction because
-  // some code may still be executing in later TSD destructors
-  // and we don't want it to have any poisoned stack.
-  ClearShadowForThreadStackAndTLS();
-  DeleteFakeStack(tid);
+  bool was_running =
+      (asanThreadRegistry().FinishThread(tid) == ThreadStatusRunning);
+  if (was_running) {
+    if (AsanThread *thread = GetCurrentThread())
+      CHECK_EQ(this, thread);
+    malloc_storage().CommitBack();
+    if (common_flags()->use_sigaltstack)
+      UnsetAlternateSignalStack();
+    FlushToDeadThreadStats(&stats_);
+    // We also clear the shadow on thread destruction because
+    // some code may still be executing in later TSD destructors
+    // and we don't want it to have any poisoned stack.
+    ClearShadowForThreadStackAndTLS();
+    DeleteFakeStack(tid);
+  } else {
+    CHECK_NE(this, GetCurrentThread());
+  }
   uptr size = RoundUpTo(sizeof(AsanThread), GetPageSizeCached());
   UnmapOrDie(this, size);
-  DTLS_Destroy();
+  if (was_running)
+    DTLS_Destroy();
 }
 
 void AsanThread::StartSwitchFiber(FakeStack **fake_stack_save, uptr bottom,
@@ -219,7 +228,7 @@ FakeStack *AsanThread::AsyncSignalSafeLazyInitFakeStack() {
 }
 
 void AsanThread::Init(const InitOptions *options) {
-  DCHECK_NE(tid(), ThreadRegistry::kUnknownTid);
+  DCHECK_NE(tid(), kInvalidTid);
   next_stack_top_ = next_stack_bottom_ = 0;
   atomic_store(&stack_switching_, false, memory_order_release);
   CHECK_EQ(this->stack_size(), 0U);
@@ -253,12 +262,9 @@ void AsanThread::Init(const InitOptions *options) {
 // SetThreadStackAndTls.
 #if !SANITIZER_FUCHSIA && !SANITIZER_RTEMS
 
-thread_return_t AsanThread::ThreadStart(
-    tid_t os_id, atomic_uintptr_t *signal_thread_is_registered) {
+thread_return_t AsanThread::ThreadStart(tid_t os_id) {
   Init();
   asanThreadRegistry().StartThread(tid(), os_id, ThreadType::Regular, nullptr);
-  if (signal_thread_is_registered)
-    atomic_store(signal_thread_is_registered, 1, memory_order_release);
 
   if (common_flags()->use_sigaltstack) SetAlternateSignalStack();
 
@@ -285,11 +291,10 @@ thread_return_t AsanThread::ThreadStart(
 
 AsanThread *CreateMainThread() {
   AsanThread *main_thread = AsanThread::Create(
-      /* start_routine */ nullptr, /* arg */ nullptr, /* parent_tid */ 0,
+      /* start_routine */ nullptr, /* arg */ nullptr, /* parent_tid */ kMainTid,
       /* stack */ nullptr, /* detached */ true);
   SetCurrentThread(main_thread);
-  main_thread->ThreadStart(internal_getpid(),
-                           /* signal_thread_is_registered */ nullptr);
+  main_thread->ThreadStart(internal_getpid());
   return main_thread;
 }
 
@@ -300,9 +305,9 @@ void AsanThread::SetThreadStackAndTls(const InitOptions *options) {
   DCHECK_EQ(options, nullptr);
   uptr tls_size = 0;
   uptr stack_size = 0;
-  GetThreadStackAndTls(tid() == 0, &stack_bottom_, &stack_size, &tls_begin_,
-                       &tls_size);
-  stack_top_ = stack_bottom_ + stack_size;
+  GetThreadStackAndTls(tid() == kMainTid, &stack_bottom_, &stack_size,
+                       &tls_begin_, &tls_size);
+  stack_top_ = RoundDownTo(stack_bottom_ + stack_size, SHADOW_GRANULARITY);
   tls_end_ = tls_begin_ + tls_size;
   dtls_ = DTLS_Get();
 
@@ -426,7 +431,7 @@ AsanThread *GetCurrentThread() {
       // address. We are not entirely sure that we have correct main thread
       // limits, so only do this magic on Android, and only if the found thread
       // is the main thread.
-      AsanThreadContext *tctx = GetThreadContextByTidLocked(0);
+      AsanThreadContext *tctx = GetThreadContextByTidLocked(kMainTid);
       if (tctx && ThreadStackContainsAddress(tctx, &context)) {
         SetCurrentThread(tctx->thread);
         return tctx->thread;
@@ -463,7 +468,7 @@ AsanThread *FindThreadByStackAddress(uptr addr) {
 void EnsureMainThreadIDIsCorrect() {
   AsanThreadContext *context =
       reinterpret_cast<AsanThreadContext *>(AsanTSDGet());
-  if (context && (context->tid == 0))
+  if (context && (context->tid == kMainTid))
     context->os_id = GetTid();
 }
 
diff --git a/libsanitizer/asan/asan_thread.h b/libsanitizer/asan/asan_thread.h
index ea58de4..200069c 100644
--- a/libsanitizer/asan/asan_thread.h
+++ b/libsanitizer/asan/asan_thread.h
@@ -28,7 +28,6 @@ struct DTLS;
 
 namespace __asan {
 
-const u32 kInvalidTid = 0xffffff;  // Must fit into 24 bits.
 const u32 kMaxNumberOfThreads = (1 << 22);  // 4M
 
 class AsanThread;
@@ -69,8 +68,7 @@ class AsanThread {
   struct InitOptions;
   void Init(const InitOptions *options = nullptr);
 
-  thread_return_t ThreadStart(tid_t os_id,
-                              atomic_uintptr_t *signal_thread_is_registered);
+  thread_return_t ThreadStart(tid_t os_id);
 
   uptr stack_top();
   uptr stack_bottom();
@@ -132,6 +130,8 @@ class AsanThread {
 
   void *extra_spill_area() { return &extra_spill_area_; }
 
+  void *get_arg() { return arg_; }
+
  private:
   // NOTE: There is no AsanThread constructor. It is allocated
   // via mmap() and *must* be valid in zero-initialized state.
diff --git a/libsanitizer/asan/asan_win.cpp b/libsanitizer/asan/asan_win.cpp
index 8044ae1..1577c83 100644
--- a/libsanitizer/asan/asan_win.cpp
+++ b/libsanitizer/asan/asan_win.cpp
@@ -134,7 +134,7 @@ INTERCEPTOR(int, _except_handler4, void *a, void *b, void *c, void *d) {
 static thread_return_t THREAD_CALLING_CONV asan_thread_start(void *arg) {
   AsanThread *t = (AsanThread *)arg;
   SetCurrentThread(t);
-  return t->ThreadStart(GetTid(), /* signal_thread_is_registered */ nullptr);
+  return t->ThreadStart(GetTid());
 }
 
 INTERCEPTOR_WINAPI(HANDLE, CreateThread, LPSECURITY_ATTRIBUTES security,
diff --git a/libsanitizer/builtins/assembly.h b/libsanitizer/builtins/assembly.h
index f437cb8..9c01505 100644
--- a/libsanitizer/builtins/assembly.h
+++ b/libsanitizer/builtins/assembly.h
@@ -14,8 +14,8 @@
 #ifndef COMPILERRT_ASSEMBLY_H
 #define COMPILERRT_ASSEMBLY_H
 
-#if defined(__POWERPC__) || defined(__powerpc__) || defined(__ppc__)
-#define SEPARATOR @
+#if defined(__APPLE__) && defined(__aarch64__)
+#define SEPARATOR %%
 #else
 #define SEPARATOR ;
 #endif
@@ -35,14 +35,14 @@
 #define HIDDEN(name) .hidden name
 #define LOCAL_LABEL(name) .L_##name
 #define FILE_LEVEL_DIRECTIVE
-#if defined(__arm__)
+#if defined(__arm__) || defined(__aarch64__)
 #define SYMBOL_IS_FUNC(name) .type name,%function
 #else
 #define SYMBOL_IS_FUNC(name) .type name,@function
 #endif
 #define CONST_SECTION .section .rodata
 
-#if defined(__GNU__) || defined(__FreeBSD__) || defined(__Fuchsia__) || \
+#if defined(__GNU__) || defined(__FreeBSD__) || defined(__Fuchsia__) ||        \
     defined(__linux__)
 #define NO_EXEC_STACK_DIRECTIVE .section .note.GNU-stack,"",%progbits
 #else
@@ -65,6 +65,68 @@
 
 #endif
 
+#if defined(__arm__) || defined(__aarch64__)
+#define FUNC_ALIGN                                                             \
+  .text SEPARATOR                                                              \
+  .balign 16 SEPARATOR
+#else
+#define FUNC_ALIGN
+#endif
+
+// BTI and PAC gnu property note
+#define NT_GNU_PROPERTY_TYPE_0 5
+#define GNU_PROPERTY_AARCH64_FEATURE_1_AND 0xc0000000
+#define GNU_PROPERTY_AARCH64_FEATURE_1_BTI 1
+#define GNU_PROPERTY_AARCH64_FEATURE_1_PAC 2
+
+#if defined(__ARM_FEATURE_BTI_DEFAULT)
+#define BTI_FLAG GNU_PROPERTY_AARCH64_FEATURE_1_BTI
+#else
+#define BTI_FLAG 0
+#endif
+
+#if __ARM_FEATURE_PAC_DEFAULT & 3
+#define PAC_FLAG GNU_PROPERTY_AARCH64_FEATURE_1_PAC
+#else
+#define PAC_FLAG 0
+#endif
+
+#define GNU_PROPERTY(type, value)                                              \
+  .pushsection .note.gnu.property, "a" SEPARATOR                               \
+  .p2align 3 SEPARATOR                                                         \
+  .word 4 SEPARATOR                                                            \
+  .word 16 SEPARATOR                                                           \
+  .word NT_GNU_PROPERTY_TYPE_0 SEPARATOR                                       \
+  .asciz "GNU" SEPARATOR                                                       \
+  .word type SEPARATOR                                                         \
+  .word 4 SEPARATOR                                                            \
+  .word value SEPARATOR                                                        \
+  .word 0 SEPARATOR                                                            \
+  .popsection
+
+#if BTI_FLAG != 0
+#define BTI_C hint #34
+#define BTI_J hint #36
+#else
+#define BTI_C
+#define BTI_J
+#endif
+
+#if (BTI_FLAG | PAC_FLAG) != 0
+#define GNU_PROPERTY_BTI_PAC                                                   \
+  GNU_PROPERTY(GNU_PROPERTY_AARCH64_FEATURE_1_AND, BTI_FLAG | PAC_FLAG)
+#else
+#define GNU_PROPERTY_BTI_PAC
+#endif
+
+#if defined(__clang__) || defined(__GCC_HAVE_DWARF2_CFI_ASM)
+#define CFI_START .cfi_startproc
+#define CFI_END .cfi_endproc
+#else
+#define CFI_START
+#define CFI_END
+#endif
+
 #if defined(__arm__)
 
 // Determine actual [ARM][THUMB[1][2]] ISA using compiler predefined macros:
@@ -131,15 +193,24 @@
 #define DEFINE_CODE_STATE
 #endif
 
-#define GLUE2(a, b) a##b
-#define GLUE(a, b) GLUE2(a, b)
+#define GLUE2_(a, b) a##b
+#define GLUE(a, b) GLUE2_(a, b)
+#define GLUE2(a, b) GLUE2_(a, b)
+#define GLUE3_(a, b, c) a##b##c
+#define GLUE3(a, b, c) GLUE3_(a, b, c)
+#define GLUE4_(a, b, c, d) a##b##c##d
+#define GLUE4(a, b, c, d) GLUE4_(a, b, c, d)
+
 #define SYMBOL_NAME(name) GLUE(__USER_LABEL_PREFIX__, name)
 
 #ifdef VISIBILITY_HIDDEN
 #define DECLARE_SYMBOL_VISIBILITY(name)                                        \
   HIDDEN(SYMBOL_NAME(name)) SEPARATOR
+#define DECLARE_SYMBOL_VISIBILITY_UNMANGLED(name) \
+  HIDDEN(name) SEPARATOR
 #else
 #define DECLARE_SYMBOL_VISIBILITY(name)
+#define DECLARE_SYMBOL_VISIBILITY_UNMANGLED(name)
 #endif
 
 #define DEFINE_COMPILERRT_FUNCTION(name)                                       \
@@ -177,6 +248,16 @@
   DECLARE_FUNC_ENCODING                                                        \
   name:
 
+#define DEFINE_COMPILERRT_OUTLINE_FUNCTION_UNMANGLED(name)                     \
+  DEFINE_CODE_STATE                                                            \
+  FUNC_ALIGN                                                                   \
+  .globl name SEPARATOR                                                        \
+  SYMBOL_IS_FUNC(name) SEPARATOR                                               \
+  DECLARE_SYMBOL_VISIBILITY_UNMANGLED(name) SEPARATOR                          \
+  CFI_START SEPARATOR                                                          \
+  DECLARE_FUNC_ENCODING                                                        \
+  name: SEPARATOR BTI_C
+
 #define DEFINE_COMPILERRT_FUNCTION_ALIAS(name, target)                         \
   .globl SYMBOL_NAME(name) SEPARATOR                                           \
   SYMBOL_IS_FUNC(SYMBOL_NAME(name)) SEPARATOR                                  \
@@ -193,8 +274,13 @@
 #ifdef __ELF__
 #define END_COMPILERRT_FUNCTION(name)                                          \
   .size SYMBOL_NAME(name), . - SYMBOL_NAME(name)
+#define END_COMPILERRT_OUTLINE_FUNCTION(name)                                  \
+  CFI_END SEPARATOR                                                            \
+  .size SYMBOL_NAME(name), . - SYMBOL_NAME(name)
 #else
 #define END_COMPILERRT_FUNCTION(name)
+#define END_COMPILERRT_OUTLINE_FUNCTION(name)                                  \
+  CFI_END
 #endif
 
 #endif // COMPILERRT_ASSEMBLY_H
diff --git a/libsanitizer/hwasan/hwasan.cpp b/libsanitizer/hwasan/hwasan.cpp
index c532211..8d6c252 100644
--- a/libsanitizer/hwasan/hwasan.cpp
+++ b/libsanitizer/hwasan/hwasan.cpp
@@ -128,16 +128,11 @@ static void InitializeFlags() {
   if (common_flags()->help) parser.PrintFlagDescriptions();
 }
 
-static void HWAsanCheckFailed(const char *file, int line, const char *cond,
-                              u64 v1, u64 v2) {
-  Report("HWAddressSanitizer CHECK failed: %s:%d \"%s\" (0x%zx, 0x%zx)\n", file,
-         line, cond, (uptr)v1, (uptr)v2);
-  PRINT_CURRENT_STACK_CHECK();
-  Die();
+static void CheckUnwind() {
+  GET_FATAL_STACK_TRACE_PC_BP(StackTrace::GetCurrentPc(), GET_CURRENT_FRAME());
+  stack.Print();
 }
 
-static constexpr uptr kMemoryUsageBufferSize = 4096;
-
 static void HwasanFormatMemoryUsage(InternalScopedString &s) {
   HwasanThreadList &thread_list = hwasanThreadList();
   auto thread_stats = thread_list.GetThreadStats();
@@ -155,6 +150,8 @@ static void HwasanFormatMemoryUsage(InternalScopedString &s) {
 }
 
 #if SANITIZER_ANDROID
+static constexpr uptr kMemoryUsageBufferSize = 4096;
+
 static char *memory_usage_buffer = nullptr;
 
 static void InitMemoryUsage() {
@@ -171,7 +168,7 @@ void UpdateMemoryUsage() {
     return;
   if (!memory_usage_buffer)
     InitMemoryUsage();
-  InternalScopedString s(kMemoryUsageBufferSize);
+  InternalScopedString s;
   HwasanFormatMemoryUsage(s);
   internal_strncpy(memory_usage_buffer, s.data(), kMemoryUsageBufferSize - 1);
   memory_usage_buffer[kMemoryUsageBufferSize - 1] = '\0';
@@ -271,7 +268,7 @@ void __hwasan_init() {
   InitializeFlags();
 
   // Install tool-specific callbacks in sanitizer_common.
-  SetCheckFailedCallback(HWAsanCheckFailed);
+  SetCheckUnwindCallback(CheckUnwind);
 
   __sanitizer_set_report_path(common_flags()->log_path);
 
@@ -493,7 +490,7 @@ extern "C" void *__hwasan_extra_spill_area() {
 }
 
 void __hwasan_print_memory_usage() {
-  InternalScopedString s(kMemoryUsageBufferSize);
+  InternalScopedString s;
   HwasanFormatMemoryUsage(s);
   Printf("%s\n", s.data());
 }
diff --git a/libsanitizer/hwasan/hwasan.h b/libsanitizer/hwasan/hwasan.h
index d4521ef..8515df5 100644
--- a/libsanitizer/hwasan/hwasan.h
+++ b/libsanitizer/hwasan/hwasan.h
@@ -14,11 +14,12 @@
 #ifndef HWASAN_H
 #define HWASAN_H
 
+#include "hwasan_flags.h"
+#include "hwasan_interface_internal.h"
+#include "sanitizer_common/sanitizer_common.h"
 #include "sanitizer_common/sanitizer_flags.h"
 #include "sanitizer_common/sanitizer_internal_defs.h"
 #include "sanitizer_common/sanitizer_stacktrace.h"
-#include "hwasan_interface_internal.h"
-#include "hwasan_flags.h"
 #include "ubsan/ubsan_platform.h"
 
 #ifndef HWASAN_CONTAINS_UBSAN
@@ -35,10 +36,31 @@
 
 typedef u8 tag_t;
 
+#if defined(__x86_64__)
+// Tags are done in middle bits using userspace aliasing.
+constexpr unsigned kAddressTagShift = 39;
+constexpr unsigned kTagBits = 3;
+
+// The alias region is placed next to the shadow so the upper bits of all
+// taggable addresses matches the upper bits of the shadow base.  This shift
+// value determines which upper bits must match.  It has a floor of 44 since the
+// shadow is always 8TB.
+// TODO(morehouse): In alias mode we can shrink the shadow and use a
+// simpler/faster shadow calculation.
+constexpr unsigned kTaggableRegionCheckShift =
+    __sanitizer::Max(kAddressTagShift + kTagBits + 1U, 44U);
+#else
 // TBI (Top Byte Ignore) feature of AArch64: bits [63:56] are ignored in address
 // translation and can be used to store a tag.
-const unsigned kAddressTagShift = 56;
-const uptr kAddressTagMask = 0xFFUL << kAddressTagShift;
+constexpr unsigned kAddressTagShift = 56;
+constexpr unsigned kTagBits = 8;
+#endif  // defined(__x86_64__)
+
+// Mask for extracting tag bits from the lower 8 bits.
+constexpr uptr kTagMask = (1UL << kTagBits) - 1;
+
+// Mask for extracting tag bits from full pointers.
+constexpr uptr kAddressTagMask = kTagMask << kAddressTagShift;
 
 // Minimal alignment of the shadow base address. Determines the space available
 // for threads and stack histories. This is an ABI constant.
@@ -50,7 +72,7 @@ const unsigned kRecordFPLShift = 4;
 const unsigned kRecordFPModulus = 1 << (64 - kRecordFPShift + kRecordFPLShift);
 
 static inline tag_t GetTagFromPointer(uptr p) {
-  return p >> kAddressTagShift;
+  return (p >> kAddressTagShift) & kTagMask;
 }
 
 static inline uptr UntagAddr(uptr tagged_addr) {
@@ -105,15 +127,6 @@ void InstallAtExitHandler();
   if (hwasan_inited)                                     \
     stack.Unwind(pc, bp, nullptr, common_flags()->fast_unwind_on_fatal)
 
-#define GET_FATAL_STACK_TRACE_HERE \
-  GET_FATAL_STACK_TRACE_PC_BP(StackTrace::GetCurrentPc(), GET_CURRENT_FRAME())
-
-#define PRINT_CURRENT_STACK_CHECK() \
-  {                                 \
-    GET_FATAL_STACK_TRACE_HERE;     \
-    stack.Print();                  \
-  }
-
 void HwasanTSDInit();
 void HwasanTSDThreadInit();
 
diff --git a/libsanitizer/hwasan/hwasan_allocator.cpp b/libsanitizer/hwasan/hwasan_allocator.cpp
index 0b6b734..a6fc794 100644
--- a/libsanitizer/hwasan/hwasan_allocator.cpp
+++ b/libsanitizer/hwasan/hwasan_allocator.cpp
@@ -29,8 +29,8 @@ static AllocatorCache fallback_allocator_cache;
 static SpinMutex fallback_mutex;
 static atomic_uint8_t hwasan_allocator_tagging_enabled;
 
-static const tag_t kFallbackAllocTag = 0xBB;
-static const tag_t kFallbackFreeTag = 0xBC;
+static constexpr tag_t kFallbackAllocTag = 0xBB & kTagMask;
+static constexpr tag_t kFallbackFreeTag = 0xBC;
 
 enum RightAlignMode {
   kRightAlignNever,
@@ -84,7 +84,8 @@ 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);
+  allocator.Init(common_flags()->allocator_release_to_os_interval_ms,
+                 kAliasRegionStart);
   for (uptr i = 0; i < sizeof(tail_magic); i++)
     tail_magic[i] = GetCurrentThread()->GenerateRandomTag();
 }
@@ -148,7 +149,8 @@ static void *HwasanAllocate(StackTrace *stack, uptr orig_size, uptr alignment,
   // Tagging can only be skipped when both tag_in_malloc and tag_in_free are
   // false. When tag_in_malloc = false and tag_in_free = true malloc needs to
   // retag to 0.
-  if ((flags()->tag_in_malloc || flags()->tag_in_free) &&
+  if (InTaggableRegion(reinterpret_cast<uptr>(user_ptr)) &&
+      (flags()->tag_in_malloc || flags()->tag_in_free) &&
       atomic_load_relaxed(&hwasan_allocator_tagging_enabled)) {
     if (flags()->tag_in_malloc && malloc_bisect(stack, orig_size)) {
       tag_t tag = t ? t->GenerateRandomTag() : kFallbackAllocTag;
@@ -175,6 +177,8 @@ static void *HwasanAllocate(StackTrace *stack, uptr orig_size, uptr alignment,
 static bool PointerAndMemoryTagsMatch(void *tagged_ptr) {
   CHECK(tagged_ptr);
   uptr tagged_uptr = reinterpret_cast<uptr>(tagged_ptr);
+  if (!InTaggableRegion(tagged_uptr))
+    return true;
   tag_t mem_tag = *reinterpret_cast<tag_t *>(
       MemToShadow(reinterpret_cast<uptr>(UntagPtr(tagged_ptr))));
   return PossiblyShortTagMatches(mem_tag, tagged_uptr, 1);
@@ -187,7 +191,9 @@ static void HwasanDeallocate(StackTrace *stack, void *tagged_ptr) {
   if (!PointerAndMemoryTagsMatch(tagged_ptr))
     ReportInvalidFree(stack, reinterpret_cast<uptr>(tagged_ptr));
 
-  void *untagged_ptr = UntagPtr(tagged_ptr);
+  void *untagged_ptr = InTaggableRegion(reinterpret_cast<uptr>(tagged_ptr))
+                           ? UntagPtr(tagged_ptr)
+                           : tagged_ptr;
   void *aligned_ptr = reinterpret_cast<void *>(
       RoundDownTo(reinterpret_cast<uptr>(untagged_ptr), kShadowAlignment));
   Metadata *meta =
@@ -219,10 +225,14 @@ static void HwasanDeallocate(StackTrace *stack, void *tagged_ptr) {
         Min(TaggedSize(orig_size), (uptr)flags()->max_free_fill_size);
     internal_memset(aligned_ptr, flags()->free_fill_byte, fill_size);
   }
-  if (flags()->tag_in_free && malloc_bisect(stack, 0) &&
-      atomic_load_relaxed(&hwasan_allocator_tagging_enabled))
+  if (InTaggableRegion(reinterpret_cast<uptr>(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;
     TagMemoryAligned(reinterpret_cast<uptr>(aligned_ptr), TaggedSize(orig_size),
-                     t ? t->GenerateRandomTag() : kFallbackFreeTag);
+                     tag);
+  }
   if (t) {
     allocator.Deallocate(t->allocator_cache(), aligned_ptr);
     if (auto *ha = t->heap_allocations())
@@ -365,7 +375,7 @@ int hwasan_posix_memalign(void **memptr, uptr alignment, uptr size,
     // OOM error is already taken care of by HwasanAllocate.
     return errno_ENOMEM;
   CHECK(IsAligned((uptr)ptr, alignment));
-  *(void **)UntagPtr(memptr) = ptr;
+  *memptr = ptr;
   return 0;
 }
 
diff --git a/libsanitizer/hwasan/hwasan_allocator.h b/libsanitizer/hwasan/hwasan_allocator.h
index 43670a6..03bbcff 100644
--- a/libsanitizer/hwasan/hwasan_allocator.h
+++ b/libsanitizer/hwasan/hwasan_allocator.h
@@ -13,13 +13,15 @@
 #ifndef HWASAN_ALLOCATOR_H
 #define HWASAN_ALLOCATOR_H
 
+#include "hwasan.h"
+#include "hwasan_interface_internal.h"
+#include "hwasan_poisoning.h"
 #include "sanitizer_common/sanitizer_allocator.h"
 #include "sanitizer_common/sanitizer_allocator_checks.h"
 #include "sanitizer_common/sanitizer_allocator_interface.h"
 #include "sanitizer_common/sanitizer_allocator_report.h"
 #include "sanitizer_common/sanitizer_common.h"
 #include "sanitizer_common/sanitizer_ring_buffer.h"
-#include "hwasan_poisoning.h"
 
 #if !defined(__aarch64__) && !defined(__x86_64__)
 #error Unsupported platform
@@ -55,7 +57,12 @@ static const uptr kMaxAllowedMallocSize = 1UL << 40;  // 1T
 
 struct AP64 {
   static const uptr kSpaceBeg = ~0ULL;
+
+#if defined(__x86_64__)
+  static const uptr kSpaceSize = 1ULL << kAddressTagShift;
+#else
   static const uptr kSpaceSize = 0x2000000000ULL;
+#endif
   static const uptr kMetadataSize = sizeof(Metadata);
   typedef __sanitizer::VeryDenseSizeClassMap SizeClassMap;
   using AddressSpaceView = LocalAddressSpaceView;
@@ -102,6 +109,16 @@ typedef RingBuffer<HeapAllocationRecord> HeapAllocationsRingBuffer;
 
 void GetAllocatorStats(AllocatorStatCounters s);
 
+inline bool InTaggableRegion(uptr addr) {
+#if defined(__x86_64__)
+  // Aliases are mapped next to shadow so that the upper bits match the shadow
+  // base.
+  return (addr >> kTaggableRegionCheckShift) ==
+         (__hwasan_shadow_memory_dynamic_address >> kTaggableRegionCheckShift);
+#endif
+  return true;
+}
+
 } // namespace __hwasan
 
 #endif // HWASAN_ALLOCATOR_H
diff --git a/libsanitizer/hwasan/hwasan_checks.h b/libsanitizer/hwasan/hwasan_checks.h
index a8de0fe..ab543ea 100644
--- a/libsanitizer/hwasan/hwasan_checks.h
+++ b/libsanitizer/hwasan/hwasan_checks.h
@@ -13,6 +13,7 @@
 #ifndef HWASAN_CHECKS_H
 #define HWASAN_CHECKS_H
 
+#include "hwasan_allocator.h"
 #include "hwasan_mapping.h"
 #include "sanitizer_common/sanitizer_common.h"
 
@@ -81,6 +82,8 @@ enum class AccessType { Load, Store };
 
 template <ErrorAction EA, AccessType AT, unsigned LogSize>
 __attribute__((always_inline, nodebug)) static void CheckAddress(uptr p) {
+  if (!InTaggableRegion(p))
+    return;
   uptr ptr_raw = p & ~kAddressTagMask;
   tag_t mem_tag = *(tag_t *)MemToShadow(ptr_raw);
   if (UNLIKELY(!PossiblyShortTagMatches(mem_tag, p, 1 << LogSize))) {
@@ -94,7 +97,7 @@ __attribute__((always_inline, nodebug)) static void CheckAddress(uptr p) {
 template <ErrorAction EA, AccessType AT>
 __attribute__((always_inline, nodebug)) static void CheckAddressSized(uptr p,
                                                                       uptr sz) {
-  if (sz == 0)
+  if (sz == 0 || !InTaggableRegion(p))
     return;
   tag_t ptr_tag = GetTagFromPointer(p);
   uptr ptr_raw = p & ~kAddressTagMask;
diff --git a/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp b/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
index 12730b2..f53276e 100644
--- a/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
+++ b/libsanitizer/hwasan/hwasan_dynamic_shadow.cpp
@@ -12,15 +12,17 @@
 ///
 //===----------------------------------------------------------------------===//
 
-#include "hwasan.h"
 #include "hwasan_dynamic_shadow.h"
-#include "hwasan_mapping.h"
-#include "sanitizer_common/sanitizer_common.h"
-#include "sanitizer_common/sanitizer_posix.h"
 
 #include <elf.h>
 #include <link.h>
 
+#include "hwasan.h"
+#include "hwasan_mapping.h"
+#include "hwasan_thread_list.h"
+#include "sanitizer_common/sanitizer_common.h"
+#include "sanitizer_common/sanitizer_posix.h"
+
 // The code in this file needs to run in an unrelocated binary. It should not
 // access any external symbol, including its own non-hidden globals.
 
@@ -117,6 +119,12 @@ namespace __hwasan {
 void InitShadowGOT() {}
 
 uptr FindDynamicShadowStart(uptr shadow_size_bytes) {
+#if defined(__x86_64__)
+  constexpr uptr kAliasSize = 1ULL << kAddressTagShift;
+  constexpr uptr kNumAliases = 1ULL << kTagBits;
+  return MapDynamicShadowAndAliases(shadow_size_bytes, kAliasSize, kNumAliases,
+                                    RingBufferSize());
+#endif
   return MapDynamicShadow(shadow_size_bytes, kShadowScale, kShadowBaseAlignment,
                           kHighMemEnd);
 }
diff --git a/libsanitizer/hwasan/hwasan_flags.h b/libsanitizer/hwasan/hwasan_flags.h
index 0a6998f..b177501 100644
--- a/libsanitizer/hwasan/hwasan_flags.h
+++ b/libsanitizer/hwasan/hwasan_flags.h
@@ -12,6 +12,8 @@
 #ifndef HWASAN_FLAGS_H
 #define HWASAN_FLAGS_H
 
+#include "sanitizer_common/sanitizer_internal_defs.h"
+
 namespace __hwasan {
 
 struct Flags {
diff --git a/libsanitizer/hwasan/hwasan_flags.inc b/libsanitizer/hwasan/hwasan_flags.inc
index 8e431d9..18ea47f 100644
--- a/libsanitizer/hwasan/hwasan_flags.inc
+++ b/libsanitizer/hwasan/hwasan_flags.inc
@@ -72,3 +72,12 @@ HWASAN_FLAG(uptr, malloc_bisect_right, 0,
 HWASAN_FLAG(bool, malloc_bisect_dump, false,
             "Print all allocations within [malloc_bisect_left, "
             "malloc_bisect_right] range ")
+
+
+// Exit if we fail to enable the AArch64 kernel ABI relaxation which allows
+// tagged pointers in syscalls.  This is the default, but being able to disable
+// that behaviour is useful for running the testsuite on more platforms (the
+// testsuite can run since we manually ensure any pointer arguments to syscalls
+// are untagged before the call.
+HWASAN_FLAG(bool, fail_without_syscall_abi, true,
+            "Exit if fail to request relaxed syscall ABI.")
diff --git a/libsanitizer/hwasan/hwasan_interceptors.cpp b/libsanitizer/hwasan/hwasan_interceptors.cpp
index 44e569e..ad67e27 100644
--- a/libsanitizer/hwasan/hwasan_interceptors.cpp
+++ b/libsanitizer/hwasan/hwasan_interceptors.cpp
@@ -221,8 +221,7 @@ INTERCEPTOR(int, pthread_create, void *th, void *attr, void *(*callback)(void*),
   ThreadStartArg *A = reinterpret_cast<ThreadStartArg *> (MmapOrDie(
       GetPageSizeCached(), "pthread_create"));
   *A = {callback, param};
-  int res = REAL(pthread_create)(UntagPtr(th), UntagPtr(attr),
-                                 &HwasanThreadStartFunc, A);
+  int res = REAL(pthread_create)(th, attr, &HwasanThreadStartFunc, A);
   return res;
 }
 
diff --git a/libsanitizer/hwasan/hwasan_interceptors_vfork.S b/libsanitizer/hwasan/hwasan_interceptors_vfork.S
index 23d5659..fd20825 100644
--- a/libsanitizer/hwasan/hwasan_interceptors_vfork.S
+++ b/libsanitizer/hwasan/hwasan_interceptors_vfork.S
@@ -1,4 +1,5 @@
 #include "sanitizer_common/sanitizer_asm.h"
+#include "builtins/assembly.h"
 
 #if defined(__linux__) && HWASAN_WITH_INTERCEPTORS
 #define COMMON_INTERCEPTOR_SPILL_AREA __hwasan_extra_spill_area
@@ -9,3 +10,5 @@
 #endif
 
 NO_EXEC_STACK_DIRECTIVE
+
+GNU_PROPERTY_BTI_PAC
diff --git a/libsanitizer/hwasan/hwasan_interface_internal.h b/libsanitizer/hwasan/hwasan_interface_internal.h
index aedda31..25c0f94 100644
--- a/libsanitizer/hwasan/hwasan_interface_internal.h
+++ b/libsanitizer/hwasan/hwasan_interface_internal.h
@@ -222,6 +222,9 @@ SANITIZER_INTERFACE_ATTRIBUTE
 void *__hwasan_memset(void *s, int c, uptr n);
 SANITIZER_INTERFACE_ATTRIBUTE
 void *__hwasan_memmove(void *dest, const void *src, uptr n);
+
+SANITIZER_INTERFACE_ATTRIBUTE
+void __hwasan_set_error_report_callback(void (*callback)(const char *));
 }  // extern "C"
 
 #endif  // HWASAN_INTERFACE_INTERNAL_H
diff --git a/libsanitizer/hwasan/hwasan_linux.cpp b/libsanitizer/hwasan/hwasan_linux.cpp
index e99926d..8ce0ff7 100644
--- a/libsanitizer/hwasan/hwasan_linux.cpp
+++ b/libsanitizer/hwasan/hwasan_linux.cpp
@@ -76,6 +76,8 @@ 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);
 }
@@ -119,9 +121,11 @@ void InitPrctl() {
 #define PR_GET_TAGGED_ADDR_CTRL 56
 #define PR_TAGGED_ADDR_ENABLE (1UL << 0)
   // Check we're running on a kernel that can use the tagged address ABI.
-  if (internal_prctl(PR_GET_TAGGED_ADDR_CTRL, 0, 0, 0, 0) == (uptr)-1 &&
-      errno == EINVAL) {
-#if SANITIZER_ANDROID
+  int local_errno = 0;
+  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__)
     // 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.
@@ -129,17 +133,20 @@ void InitPrctl() {
     // case.
     return;
 #else
-    Printf(
-        "FATAL: "
-        "HWAddressSanitizer requires a kernel with tagged address ABI.\n");
-    Die();
+    if (flags()->fail_without_syscall_abi) {
+      Printf(
+          "FATAL: "
+          "HWAddressSanitizer requires a kernel with tagged address ABI.\n");
+      Die();
+    }
 #endif
   }
 
   // Turn on the tagged address ABI.
-  if (internal_prctl(PR_SET_TAGGED_ADDR_CTRL, PR_TAGGED_ADDR_ENABLE, 0, 0, 0) ==
-          (uptr)-1 ||
-      !internal_prctl(PR_GET_TAGGED_ADDR_CTRL, 0, 0, 0, 0)) {
+  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` "
@@ -174,6 +181,18 @@ 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);
@@ -217,7 +236,9 @@ void InitThreads() {
 }
 
 bool MemIsApp(uptr p) {
+#if !defined(__x86_64__)  // Memory outside the alias range has non-zero tags.
   CHECK(GetTagFromPointer(p) == 0);
+#endif
   return p >= kHighMemStart || (p >= kLowMemStart && p <= kLowMemEnd);
 }
 
diff --git a/libsanitizer/hwasan/hwasan_mapping.h b/libsanitizer/hwasan/hwasan_mapping.h
index c149687..8243d1e 100644
--- a/libsanitizer/hwasan/hwasan_mapping.h
+++ b/libsanitizer/hwasan/hwasan_mapping.h
@@ -48,6 +48,8 @@ extern uptr kHighShadowEnd;
 extern uptr kHighMemStart;
 extern uptr kHighMemEnd;
 
+extern uptr kAliasRegionStart;
+
 inline uptr MemToShadow(uptr untagged_addr) {
   return (untagged_addr >> kShadowScale) +
          __hwasan_shadow_memory_dynamic_address;
diff --git a/libsanitizer/hwasan/hwasan_memintrinsics.cpp b/libsanitizer/hwasan/hwasan_memintrinsics.cpp
index e82d77a..fab017a 100644
--- a/libsanitizer/hwasan/hwasan_memintrinsics.cpp
+++ b/libsanitizer/hwasan/hwasan_memintrinsics.cpp
@@ -24,7 +24,7 @@ using namespace __hwasan;
 void *__hwasan_memset(void *block, int c, uptr size) {
   CheckAddressSized<ErrorAction::Recover, AccessType::Store>(
       reinterpret_cast<uptr>(block), size);
-  return memset(UntagPtr(block), c, size);
+  return memset(block, c, size);
 }
 
 void *__hwasan_memcpy(void *to, const void *from, uptr size) {
@@ -32,7 +32,7 @@ void *__hwasan_memcpy(void *to, const void *from, uptr size) {
       reinterpret_cast<uptr>(to), size);
   CheckAddressSized<ErrorAction::Recover, AccessType::Load>(
       reinterpret_cast<uptr>(from), size);
-  return memcpy(UntagPtr(to), UntagPtr(from), size);
+  return memcpy(to, from, size);
 }
 
 void *__hwasan_memmove(void *to, const void *from, uptr size) {
diff --git a/libsanitizer/hwasan/hwasan_new_delete.cpp b/libsanitizer/hwasan/hwasan_new_delete.cpp
index 8d01d39..69cddda 100644
--- a/libsanitizer/hwasan/hwasan_new_delete.cpp
+++ b/libsanitizer/hwasan/hwasan_new_delete.cpp
@@ -27,6 +27,12 @@
   void *res = hwasan_malloc(size, &stack);\
   if (!nothrow && UNLIKELY(!res)) ReportOutOfMemory(size, &stack);\
   return res
+#define OPERATOR_NEW_ALIGN_BODY(nothrow)                                    \
+  GET_MALLOC_STACK_TRACE;                                                   \
+  void *res = hwasan_aligned_alloc(static_cast<uptr>(align), size, &stack); \
+  if (!nothrow && UNLIKELY(!res))                                           \
+    ReportOutOfMemory(size, &stack);                                        \
+  return res
 
 #define OPERATOR_DELETE_BODY \
   GET_MALLOC_STACK_TRACE; \
@@ -50,6 +56,7 @@ 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
 
 
@@ -66,6 +73,22 @@ 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 new(
+    size_t size, std::align_val_t align) {
+  OPERATOR_NEW_ALIGN_BODY(false /*nothrow*/);
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void *operator new[](
+    size_t size, std::align_val_t align) {
+  OPERATOR_NEW_ALIGN_BODY(false /*nothrow*/);
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void *operator new(
+    size_t size, std::align_val_t align, std::nothrow_t const &) {
+  OPERATOR_NEW_ALIGN_BODY(true /*nothrow*/);
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void *operator new[](
+    size_t size, std::align_val_t align, std::nothrow_t const &) {
+  OPERATOR_NEW_ALIGN_BODY(true /*nothrow*/);
+}
 
 INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE
 void operator delete(void *ptr) NOEXCEPT { OPERATOR_DELETE_BODY; }
@@ -77,5 +100,21 @@ 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;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete[](
+    void *ptr, std::align_val_t) NOEXCEPT {
+  OPERATOR_DELETE_BODY;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete(
+    void *ptr, std::align_val_t, std::nothrow_t const &) NOEXCEPT {
+  OPERATOR_DELETE_BODY;
+}
+INTERCEPTOR_ATTRIBUTE SANITIZER_WEAK_ATTRIBUTE void operator delete[](
+    void *ptr, std::align_val_t, std::nothrow_t const &) NOEXCEPT {
+  OPERATOR_DELETE_BODY;
+}
 
 #endif // OPERATOR_NEW_BODY
diff --git a/libsanitizer/hwasan/hwasan_report.cpp b/libsanitizer/hwasan/hwasan_report.cpp
index 894a149..c021779 100644
--- a/libsanitizer/hwasan/hwasan_report.cpp
+++ b/libsanitizer/hwasan/hwasan_report.cpp
@@ -43,12 +43,16 @@ class ScopedReport {
   }
 
   ~ScopedReport() {
+    void (*report_cb)(const char *);
     {
       BlockingMutexLock lock(&error_message_lock_);
-      if (fatal)
-        SetAbortMessage(error_message_.data());
+      report_cb = error_report_callback_;
       error_message_ptr_ = nullptr;
     }
+    if (report_cb)
+      report_cb(error_message_.data());
+    if (fatal)
+      SetAbortMessage(error_message_.data());
     if (common_flags()->print_module_map >= 2 ||
         (fatal && common_flags()->print_module_map))
       DumpProcessMap();
@@ -66,6 +70,12 @@ class ScopedReport {
     // overwrite old trailing '\0', keep new trailing '\0' untouched.
     internal_memcpy(&(*error_message_ptr_)[old_size - 1], msg, len);
   }
+
+  static void SetErrorReportCallback(void (*callback)(const char *)) {
+    BlockingMutexLock lock(&error_message_lock_);
+    error_report_callback_ = callback;
+  }
+
  private:
   ScopedErrorReportLock error_report_lock_;
   InternalMmapVector<char> error_message_;
@@ -73,10 +83,12 @@ class ScopedReport {
 
   static InternalMmapVector<char> *error_message_ptr_;
   static BlockingMutex error_message_lock_;
+  static void (*error_report_callback_)(const char *);
 };
 
 InternalMmapVector<char> *ScopedReport::error_message_ptr_;
 BlockingMutex ScopedReport::error_message_lock_;
+void (*ScopedReport::error_report_callback_)(const char *);
 
 // If there is an active ScopedReport, append to its error message.
 void AppendToErrorMessageBuffer(const char *buffer) {
@@ -212,7 +224,7 @@ static void PrintStackAllocations(StackAllocationsRingBuffer *sa,
 
   // We didn't find any locals. Most likely we don't have symbols, so dump
   // the information that we have for offline analysis.
-  InternalScopedString frame_desc(GetPageSizeCached() * 2);
+  InternalScopedString frame_desc;
   Printf("Previously allocated frames:\n");
   for (uptr i = 0; i < frames; i++) {
     const uptr *record_addr = &(*sa)[i];
@@ -447,7 +459,7 @@ static void PrintTagInfoAroundAddr(tag_t *tag_ptr, uptr num_rows,
       RoundDownTo(reinterpret_cast<uptr>(tag_ptr), row_len));
   tag_t *beg_row = center_row_beg - row_len * (num_rows / 2);
   tag_t *end_row = center_row_beg + row_len * ((num_rows + 1) / 2);
-  InternalScopedString s(GetPageSizeCached() * 8);
+  InternalScopedString s;
   for (tag_t *row = beg_row; row < end_row; row += row_len) {
     s.append("%s", row == center_row_beg ? "=>" : "  ");
     s.append("%p:", row);
@@ -535,7 +547,7 @@ void ReportTailOverwritten(StackTrace *stack, uptr tagged_addr, uptr orig_size,
     GetStackTraceFromId(chunk.GetAllocStackId()).Print();
   }
 
-  InternalScopedString s(GetPageSizeCached() * 8);
+  InternalScopedString s;
   CHECK_GT(tail_size, 0U);
   CHECK_LT(tail_size, kShadowAlignment);
   u8 *tail = reinterpret_cast<u8*>(untagged_addr + orig_size);
@@ -650,3 +662,7 @@ void ReportRegisters(uptr *frame, uptr pc) {
 }
 
 }  // namespace __hwasan
+
+void __hwasan_set_error_report_callback(void (*callback)(const char *)) {
+  __hwasan::ScopedReport::SetErrorReportCallback(callback);
+}
diff --git a/libsanitizer/hwasan/hwasan_setjmp.S b/libsanitizer/hwasan/hwasan_setjmp.S
index 0c13543..381af63 100644
--- a/libsanitizer/hwasan/hwasan_setjmp.S
+++ b/libsanitizer/hwasan/hwasan_setjmp.S
@@ -12,6 +12,7 @@
 //===----------------------------------------------------------------------===//
 
 #include "sanitizer_common/sanitizer_asm.h"
+#include "builtins/assembly.h"
 
 #if HWASAN_WITH_INTERCEPTORS && defined(__aarch64__)
 #include "sanitizer_common/sanitizer_platform.h"
@@ -34,6 +35,7 @@
 ASM_TYPE_FUNCTION(__interceptor_setjmp)
 __interceptor_setjmp:
   CFI_STARTPROC
+  BTI_C
   mov	x1, #0
   b	__interceptor_sigsetjmp
   CFI_ENDPROC
@@ -46,6 +48,7 @@ ASM_SIZE(__interceptor_setjmp)
 ASM_TYPE_FUNCTION(__interceptor_setjmp_bionic)
 __interceptor_setjmp_bionic:
   CFI_STARTPROC
+  BTI_C
   mov	x1, #1
   b	__interceptor_sigsetjmp
   CFI_ENDPROC
@@ -56,6 +59,7 @@ ASM_SIZE(__interceptor_setjmp_bionic)
 ASM_TYPE_FUNCTION(__interceptor_sigsetjmp)
 __interceptor_sigsetjmp:
   CFI_STARTPROC
+  BTI_C
   stp	x19, x20, [x0, #0<<3]
   stp	x21, x22, [x0, #2<<3]
   stp	x23, x24, [x0, #4<<3]
@@ -98,3 +102,5 @@ ALIAS __interceptor_setjmp, _setjmp
 
 // We do not need executable stack.
 NO_EXEC_STACK_DIRECTIVE
+
+GNU_PROPERTY_BTI_PAC
diff --git a/libsanitizer/hwasan/hwasan_tag_mismatch_aarch64.S b/libsanitizer/hwasan/hwasan_tag_mismatch_aarch64.S
index 08df127..bcb0df4 100644
--- a/libsanitizer/hwasan/hwasan_tag_mismatch_aarch64.S
+++ b/libsanitizer/hwasan/hwasan_tag_mismatch_aarch64.S
@@ -1,4 +1,5 @@
 #include "sanitizer_common/sanitizer_asm.h"
+#include "builtins/assembly.h"
 
 // The content of this file is AArch64-only:
 #if defined(__aarch64__)
@@ -74,6 +75,8 @@
 .global __hwasan_tag_mismatch
 .type __hwasan_tag_mismatch, %function
 __hwasan_tag_mismatch:
+  BTI_J
+
   // Compute the granule position one past the end of the access.
   mov x16, #1
   and x17, x1, #0xf
@@ -106,6 +109,7 @@ __hwasan_tag_mismatch:
 .type __hwasan_tag_mismatch_v2, %function
 __hwasan_tag_mismatch_v2:
   CFI_STARTPROC
+  BTI_J
 
   // Set the CFA to be the return address for caller of __hwasan_check_*. Note
   // that we do not emit CFI predicates to describe the contents of this stack
@@ -150,3 +154,5 @@ __hwasan_tag_mismatch_v2:
 
 // We do not need executable stack.
 NO_EXEC_STACK_DIRECTIVE
+
+GNU_PROPERTY_BTI_PAC
diff --git a/libsanitizer/hwasan/hwasan_thread.cpp b/libsanitizer/hwasan/hwasan_thread.cpp
index b81a635..bb4d56a 100644
--- a/libsanitizer/hwasan/hwasan_thread.cpp
+++ b/libsanitizer/hwasan/hwasan_thread.cpp
@@ -35,6 +35,10 @@ void Thread::InitRandomState() {
 }
 
 void Thread::Init(uptr stack_buffer_start, uptr stack_buffer_size) {
+  CHECK_EQ(0, unique_id_);  // try to catch bad stack reuse
+  CHECK_EQ(0, stack_top_);
+  CHECK_EQ(0, stack_bottom_);
+
   static u64 unique_id;
   unique_id_ = unique_id++;
   if (auto sz = flags()->heap_history_size)
@@ -113,18 +117,21 @@ static u32 xorshift(u32 state) {
 }
 
 // Generate a (pseudo-)random non-zero tag.
-tag_t Thread::GenerateRandomTag() {
+tag_t Thread::GenerateRandomTag(uptr num_bits) {
+  DCHECK_GT(num_bits, 0);
   if (tagging_disabled_) return 0;
   tag_t tag;
+  const uptr tag_mask = (1ULL << num_bits) - 1;
   do {
     if (flags()->random_tags) {
       if (!random_buffer_)
         random_buffer_ = random_state_ = xorshift(random_state_);
       CHECK(random_buffer_);
-      tag = random_buffer_ & 0xFF;
-      random_buffer_ >>= 8;
+      tag = random_buffer_ & tag_mask;
+      random_buffer_ >>= num_bits;
     } else {
-      tag = random_state_ = (random_state_ + 1) & 0xFF;
+      random_state_ += 1;
+      tag = random_state_ & tag_mask;
     }
   } while (!tag);
   return tag;
diff --git a/libsanitizer/hwasan/hwasan_thread.h b/libsanitizer/hwasan/hwasan_thread.h
index ebcdb79..1c71cab 100644
--- a/libsanitizer/hwasan/hwasan_thread.h
+++ b/libsanitizer/hwasan/hwasan_thread.h
@@ -42,7 +42,7 @@ class Thread {
   HeapAllocationsRingBuffer *heap_allocations() { return heap_allocations_; }
   StackAllocationsRingBuffer *stack_allocations() { return stack_allocations_; }
 
-  tag_t GenerateRandomTag();
+  tag_t GenerateRandomTag(uptr num_bits = kTagBits);
 
   void DisableTagging() { tagging_disabled_++; }
   void EnableTagging() { tagging_disabled_--; }
@@ -74,8 +74,6 @@ class Thread {
   HeapAllocationsRingBuffer *heap_allocations_;
   StackAllocationsRingBuffer *stack_allocations_;
 
-  Thread *next_;  // All live threads form a linked list.
-
   u64 unique_id_;  // counting from zero.
 
   u32 tagging_disabled_;  // if non-zero, malloc uses zero tag in this thread.
diff --git a/libsanitizer/hwasan/hwasan_thread_list.h b/libsanitizer/hwasan/hwasan_thread_list.h
index 914b632..11c5863 100644
--- a/libsanitizer/hwasan/hwasan_thread_list.h
+++ b/libsanitizer/hwasan/hwasan_thread_list.h
@@ -66,40 +66,6 @@ static uptr RingBufferSize() {
   return 0;
 }
 
-struct ThreadListHead {
-  Thread *list_;
-
-  ThreadListHead() : list_(nullptr) {}
-
-  void Push(Thread *t) {
-    t->next_ = list_;
-    list_ = t;
-  }
-
-  Thread *Pop() {
-    Thread *t = list_;
-    if (t)
-      list_ = t->next_;
-    return t;
-  }
-
-  void Remove(Thread *t) {
-    Thread **cur = &list_;
-    while (*cur != t) cur = &(*cur)->next_;
-    CHECK(*cur && "thread not found");
-    *cur = (*cur)->next_;
-  }
-
-  template <class CB>
-  void ForEach(CB cb) {
-    Thread *t = list_;
-    while (t) {
-      cb(t);
-      t = t->next_;
-    }
-  }
-};
-
 struct ThreadStats {
   uptr n_live_threads;
   uptr total_stack_size;
@@ -120,17 +86,23 @@ class HwasanThreadList {
   }
 
   Thread *CreateCurrentThread() {
-    Thread *t;
+    Thread *t = nullptr;
     {
-      SpinMutexLock l(&list_mutex_);
-      t = free_list_.Pop();
-      if (t) {
-        uptr start = (uptr)t - ring_buffer_size_;
-        internal_memset((void *)start, 0, ring_buffer_size_ + sizeof(Thread));
-      } else {
-        t = AllocThread();
+      SpinMutexLock l(&free_list_mutex_);
+      if (!free_list_.empty()) {
+        t = free_list_.back();
+        free_list_.pop_back();
       }
-      live_list_.Push(t);
+    }
+    if (t) {
+      uptr start = (uptr)t - ring_buffer_size_;
+      internal_memset((void *)start, 0, ring_buffer_size_ + sizeof(Thread));
+    } else {
+      t = AllocThread();
+    }
+    {
+      SpinMutexLock l(&live_list_mutex_);
+      live_list_.push_back(t);
     }
     t->Init((uptr)t - ring_buffer_size_, ring_buffer_size_);
     AddThreadStats(t);
@@ -142,13 +114,26 @@ class HwasanThreadList {
     ReleaseMemoryPagesToOS(start, start + thread_alloc_size_);
   }
 
+  void RemoveThreadFromLiveList(Thread *t) {
+    SpinMutexLock l(&live_list_mutex_);
+    for (Thread *&t2 : live_list_)
+      if (t2 == t) {
+        // To remove t2, copy the last element of the list in t2's position, and
+        // pop_back(). This works even if t2 is itself the last element.
+        t2 = live_list_.back();
+        live_list_.pop_back();
+        return;
+      }
+    CHECK(0 && "thread not found in live list");
+  }
+
   void ReleaseThread(Thread *t) {
     RemoveThreadStats(t);
     t->Destroy();
-    SpinMutexLock l(&list_mutex_);
-    live_list_.Remove(t);
-    free_list_.Push(t);
     DontNeedThread(t);
+    RemoveThreadFromLiveList(t);
+    SpinMutexLock l(&free_list_mutex_);
+    free_list_.push_back(t);
   }
 
   Thread *GetThreadByBufferAddress(uptr p) {
@@ -165,8 +150,8 @@ class HwasanThreadList {
 
   template <class CB>
   void VisitAllLiveThreads(CB cb) {
-    SpinMutexLock l(&list_mutex_);
-    live_list_.ForEach(cb);
+    SpinMutexLock l(&live_list_mutex_);
+    for (Thread *t : live_list_) cb(t);
   }
 
   void AddThreadStats(Thread *t) {
@@ -188,6 +173,7 @@ class HwasanThreadList {
 
  private:
   Thread *AllocThread() {
+    SpinMutexLock l(&free_space_mutex_);
     uptr align = ring_buffer_size_ * 2;
     CHECK(IsAligned(free_space_, align));
     Thread *t = (Thread *)(free_space_ + ring_buffer_size_);
@@ -196,14 +182,16 @@ class HwasanThreadList {
     return t;
   }
 
+  SpinMutex free_space_mutex_;
   uptr free_space_;
   uptr free_space_end_;
   uptr ring_buffer_size_;
   uptr thread_alloc_size_;
 
-  ThreadListHead free_list_;
-  ThreadListHead live_list_;
-  SpinMutex list_mutex_;
+  SpinMutex free_list_mutex_;
+  InternalMmapVector<Thread *> free_list_;
+  SpinMutex live_list_mutex_;
+  InternalMmapVector<Thread *> live_list_;
 
   ThreadStats stats_;
   SpinMutex stats_mutex_;
diff --git a/libsanitizer/include/sanitizer/common_interface_defs.h b/libsanitizer/include/sanitizer/common_interface_defs.h
index b4f977b..cd69285 100644
--- a/libsanitizer/include/sanitizer/common_interface_defs.h
+++ b/libsanitizer/include/sanitizer/common_interface_defs.h
@@ -43,6 +43,9 @@ void __sanitizer_set_report_path(const char *path);
 // Tell the tools to write their reports to the provided file descriptor
 // (casted to void *).
 void __sanitizer_set_report_fd(void *fd);
+// Get the current full report file path, if a path was specified by
+// an earlier call to __sanitizer_set_report_path. Returns null otherwise.
+const char *__sanitizer_get_report_path();
 
 // Notify the tools that the sandbox is going to be turned on. The reserved
 // parameter will be used in the future to hold a structure with functions
diff --git a/libsanitizer/include/sanitizer/dfsan_interface.h b/libsanitizer/include/sanitizer/dfsan_interface.h
index 18b2c81..40f9379 100644
--- a/libsanitizer/include/sanitizer/dfsan_interface.h
+++ b/libsanitizer/include/sanitizer/dfsan_interface.h
@@ -22,6 +22,7 @@ extern "C" {
 #endif
 
 typedef uint16_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
@@ -63,6 +64,12 @@ void dfsan_add_label(dfsan_label label, void *addr, size_t size);
 /// value.
 dfsan_label dfsan_get_label(long data);
 
+/// Retrieves the immediate origin associated with the given data. The returned
+/// origin may point to another origin.
+///
+/// The type of 'data' is arbitrary.
+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);
 
@@ -110,6 +117,15 @@ void dfsan_weak_hook_memcmp(void *caller_pc, const void *s1, const void *s2,
 void dfsan_weak_hook_strncmp(void *caller_pc, const char *s1, const char *s2,
                              size_t n, dfsan_label s1_label,
                              dfsan_label s2_label, dfsan_label n_label);
+
+/// Prints the origin trace of the label at the address addr to stderr. It also
+/// prints description at the beginning of the trace. If origin tracking is not
+/// on, or the address is not labeled, it prints nothing.
+void dfsan_print_origin_trace(const void *addr, const char *description);
+
+/// Retrieves the very first origin associated with the data at the given
+/// address.
+dfsan_origin dfsan_get_init_origin(const void *addr);
 #ifdef __cplusplus
 }  // extern "C"
 
diff --git a/libsanitizer/include/sanitizer/hwasan_interface.h b/libsanitizer/include/sanitizer/hwasan_interface.h
index 4c9ad13..14035c0 100644
--- a/libsanitizer/include/sanitizer/hwasan_interface.h
+++ b/libsanitizer/include/sanitizer/hwasan_interface.h
@@ -73,6 +73,9 @@ extern "C" {
    * accessed through the pointer in x, or -1 if the whole range is good. */
   intptr_t __hwasan_test_shadow(const volatile void *x, size_t size);
 
+  /* Sets the callback function to be called during HWASan error reporting. */
+  void __hwasan_set_error_report_callback(void (*callback)(const char *));
+
   int __sanitizer_posix_memalign(void **memptr, size_t alignment, size_t size);
   void * __sanitizer_memalign(size_t alignment, size_t size);
   void * __sanitizer_aligned_alloc(size_t alignment, size_t size);
diff --git a/libsanitizer/include/sanitizer/memprof_interface.h b/libsanitizer/include/sanitizer/memprof_interface.h
index a721260..76031de 100644
--- a/libsanitizer/include/sanitizer/memprof_interface.h
+++ b/libsanitizer/include/sanitizer/memprof_interface.h
@@ -53,6 +53,11 @@ void __memprof_print_accumulated_stats(void);
 /// \returns Default options string.
 const char *__memprof_default_options(void);
 
+/// Prints the memory profile to the current profile file.
+///
+/// \returns 0 on success.
+int __memprof_profile_dump(void);
+
 #ifdef __cplusplus
 } // extern "C"
 #endif
diff --git a/libsanitizer/include/sanitizer/tsan_interface.h b/libsanitizer/include/sanitizer/tsan_interface.h
index 96b8ad5..565aa39 100644
--- a/libsanitizer/include/sanitizer/tsan_interface.h
+++ b/libsanitizer/include/sanitizer/tsan_interface.h
@@ -67,6 +67,12 @@ static const unsigned __tsan_mutex_recursive_lock   = 1 << 6;
 // the corresponding __tsan_mutex_post_lock annotation.
 static const unsigned __tsan_mutex_recursive_unlock = 1 << 7;
 
+// Convenient composed constants.
+static const unsigned __tsan_mutex_try_read_lock =
+    __tsan_mutex_read_lock | __tsan_mutex_try_lock;
+static const unsigned __tsan_mutex_try_read_lock_failed =
+    __tsan_mutex_try_read_lock | __tsan_mutex_try_lock_failed;
+
 // Annotate creation of a mutex.
 // Supported flags: mutex creation flags.
 void __tsan_mutex_create(void *addr, unsigned flags);
@@ -141,7 +147,7 @@ void __tsan_external_write(void *addr, void *caller_pc, void *tag);
 //     and freed by __tsan_destroy_fiber.
 //   - TSAN context of current fiber or thread can be obtained
 //     by calling __tsan_get_current_fiber.
-//   - __tsan_switch_to_fiber should be called immediatly before switch
+//   - __tsan_switch_to_fiber should be called immediately before switch
 //     to fiber, such as call of swapcontext.
 //   - Fiber name can be set by __tsan_set_fiber_name.
 void *__tsan_get_current_fiber(void);
@@ -154,6 +160,15 @@ void __tsan_set_fiber_name(void *fiber, const char *name);
 // Do not establish a happens-before relation between fibers
 static const unsigned __tsan_switch_to_fiber_no_sync = 1 << 0;
 
+// User-provided callback invoked on TSan initialization.
+void __tsan_on_initialize();
+
+// User-provided callback invoked on TSan shutdown.
+// `failed` - Nonzero if TSan did detect issues, zero otherwise.
+// Return `0` if TSan should exit as if no issues were detected.  Return nonzero
+// if TSan should exit as if issues were detected.
+int __tsan_on_finalize(int failed);
+
 #ifdef __cplusplus
 }  // extern "C"
 #endif
diff --git a/libsanitizer/include/sanitizer/tsan_interface_atomic.h b/libsanitizer/include/sanitizer/tsan_interface_atomic.h
index 8052bc1..5e41e22 100644
--- a/libsanitizer/include/sanitizer/tsan_interface_atomic.h
+++ b/libsanitizer/include/sanitizer/tsan_interface_atomic.h
@@ -30,7 +30,7 @@ __extension__ typedef __int128 __tsan_atomic128;
 #endif
 
 // Part of ABI, do not change.
-// https://github.com/llvm/llvm-project/blob/master/libcxx/include/atomic
+// https://github.com/llvm/llvm-project/blob/main/libcxx/include/atomic
 typedef enum {
   __tsan_memory_order_relaxed,
   __tsan_memory_order_consume,
diff --git a/libsanitizer/interception/interception_linux.cpp b/libsanitizer/interception/interception_linux.cpp
index 6883608..5111a87 100644
--- a/libsanitizer/interception/interception_linux.cpp
+++ b/libsanitizer/interception/interception_linux.cpp
@@ -63,8 +63,8 @@ bool InterceptFunction(const char *name, uptr *ptr_to_real, uptr func,
   return addr && (func == wrapper);
 }
 
-// Android and Solaris do not have dlvsym
-#if !SANITIZER_ANDROID && !SANITIZER_SOLARIS
+// dlvsym is a GNU extension supported by some other platforms.
+#if SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD
 static void *GetFuncAddr(const char *name, const char *ver) {
   return dlvsym(RTLD_NEXT, name, ver);
 }
@@ -75,7 +75,7 @@ bool InterceptFunction(const char *name, const char *ver, uptr *ptr_to_real,
   *ptr_to_real = (uptr)addr;
   return addr && (func == wrapper);
 }
-#endif  // !SANITIZER_ANDROID
+#endif  // SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD
 
 }  // namespace __interception
 
diff --git a/libsanitizer/interception/interception_linux.h b/libsanitizer/interception/interception_linux.h
index 097375f..a08f8cb 100644
--- a/libsanitizer/interception/interception_linux.h
+++ b/libsanitizer/interception/interception_linux.h
@@ -35,8 +35,8 @@ bool InterceptFunction(const char *name, const char *ver, uptr *ptr_to_real,
       (::__interception::uptr) & (func),          \
       (::__interception::uptr) & WRAP(func))
 
-// Android and Solaris do not have dlvsym
-#if !SANITIZER_ANDROID && !SANITIZER_SOLARIS
+// dlvsym is a GNU extension supported by some other platforms.
+#if SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD
 #define INTERCEPT_FUNCTION_VER_LINUX_OR_FREEBSD(func, symver) \
   ::__interception::InterceptFunction(                        \
       #func, symver,                                          \
@@ -46,7 +46,7 @@ bool InterceptFunction(const char *name, const char *ver, uptr *ptr_to_real,
 #else
 #define INTERCEPT_FUNCTION_VER_LINUX_OR_FREEBSD(func, symver) \
   INTERCEPT_FUNCTION_LINUX_OR_FREEBSD(func)
-#endif  // !SANITIZER_ANDROID && !SANITIZER_SOLARIS
+#endif  // SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD
 
 #endif  // INTERCEPTION_LINUX_H
 #endif  // SANITIZER_LINUX || SANITIZER_FREEBSD || SANITIZER_NETBSD ||
diff --git a/libsanitizer/interception/interception_win.cpp b/libsanitizer/interception/interception_win.cpp
index 1a1c327..98bc756 100644
--- a/libsanitizer/interception/interception_win.cpp
+++ b/libsanitizer/interception/interception_win.cpp
@@ -136,7 +136,7 @@ namespace __interception {
 static const int kAddressLength = FIRST_32_SECOND_64(4, 8);
 static const int kJumpInstructionLength = 5;
 static const int kShortJumpInstructionLength = 2;
-static const int kIndirectJumpInstructionLength = 6;
+UNUSED static const int kIndirectJumpInstructionLength = 6;
 static const int kBranchLength =
     FIRST_32_SECOND_64(kJumpInstructionLength, kIndirectJumpInstructionLength);
 static const int kDirectBranchLength = kBranchLength + kAddressLength;
@@ -165,7 +165,7 @@ static uptr GetMmapGranularity() {
   return si.dwAllocationGranularity;
 }
 
-static uptr RoundUpTo(uptr size, uptr boundary) {
+UNUSED static uptr RoundUpTo(uptr size, uptr boundary) {
   return (size + boundary - 1) & ~(boundary - 1);
 }
 
@@ -309,7 +309,7 @@ struct TrampolineMemoryRegion {
   uptr max_size;
 };
 
-static const uptr kTrampolineScanLimitRange = 1 << 31;  // 2 gig
+UNUSED static const uptr kTrampolineScanLimitRange = 1 << 31;  // 2 gig
 static const int kMaxTrampolineRegion = 1024;
 static TrampolineMemoryRegion TrampolineRegions[kMaxTrampolineRegion];
 
diff --git a/libsanitizer/lsan/lsan_allocator.cpp b/libsanitizer/lsan/lsan_allocator.cpp
index d86c392..91e34eb 100644
--- a/libsanitizer/lsan/lsan_allocator.cpp
+++ b/libsanitizer/lsan/lsan_allocator.cpp
@@ -123,14 +123,18 @@ void Deallocate(void *p) {
 
 void *Reallocate(const StackTrace &stack, void *p, uptr new_size,
                  uptr alignment) {
-  RegisterDeallocation(p);
   if (new_size > max_malloc_size) {
-    allocator.Deallocate(GetAllocatorCache(), p);
-    return ReportAllocationSizeTooBig(new_size, stack);
+    ReportAllocationSizeTooBig(new_size, stack);
+    return nullptr;
   }
-  p = allocator.Reallocate(GetAllocatorCache(), p, new_size, alignment);
-  RegisterAllocation(stack, p, new_size);
-  return p;
+  RegisterDeallocation(p);
+  void *new_p =
+      allocator.Reallocate(GetAllocatorCache(), p, new_size, alignment);
+  if (new_p)
+    RegisterAllocation(stack, new_p, new_size);
+  else if (new_size != 0)
+    RegisterAllocation(stack, p, new_size);
+  return new_p;
 }
 
 void GetAllocatorCacheRange(uptr *begin, uptr *end) {
@@ -309,6 +313,16 @@ IgnoreObjectResult IgnoreObjectLocked(const void *p) {
     return kIgnoreObjectInvalid;
   }
 }
+
+void GetAdditionalThreadContextPtrs(ThreadContextBase *tctx, void *ptrs) {
+  // This function can be used to treat memory reachable from `tctx` as live.
+  // This is useful for threads that have been created but not yet started.
+
+  // This is currently a no-op because the LSan `pthread_create()` interceptor
+  // blocks until the child thread starts which keeps the thread's `arg` pointer
+  // live.
+}
+
 } // namespace __lsan
 
 using namespace __lsan;
diff --git a/libsanitizer/lsan/lsan_allocator.h b/libsanitizer/lsan/lsan_allocator.h
index 17e13cd0..9d76378 100644
--- a/libsanitizer/lsan/lsan_allocator.h
+++ b/libsanitizer/lsan/lsan_allocator.h
@@ -50,7 +50,7 @@ struct ChunkMetadata {
 };
 
 #if defined(__mips64) || defined(__aarch64__) || defined(__i386__) || \
-    defined(__arm__)
+    defined(__arm__) || SANITIZER_RISCV64
 template <typename AddressSpaceViewTy>
 struct AP32 {
   static const uptr kSpaceBeg = 0;
diff --git a/libsanitizer/lsan/lsan_common.cpp b/libsanitizer/lsan/lsan_common.cpp
index 9e23aa9..74400d2 100644
--- a/libsanitizer/lsan/lsan_common.cpp
+++ b/libsanitizer/lsan/lsan_common.cpp
@@ -65,8 +65,34 @@ void RegisterLsanFlags(FlagParser *parser, Flags *f) {
     if (flags()->log_threads) Report(__VA_ARGS__); \
   } while (0)
 
-ALIGNED(64) static char suppression_placeholder[sizeof(SuppressionContext)];
-static SuppressionContext *suppression_ctx = nullptr;
+class LeakSuppressionContext {
+  bool parsed = false;
+  SuppressionContext context;
+  bool suppressed_stacks_sorted = true;
+  InternalMmapVector<u32> suppressed_stacks;
+
+  Suppression *GetSuppressionForAddr(uptr addr);
+  void LazyInit();
+
+ public:
+  LeakSuppressionContext(const char *supprression_types[],
+                         int suppression_types_num)
+      : context(supprression_types, suppression_types_num) {}
+
+  Suppression *GetSuppressionForStack(u32 stack_trace_id);
+
+  const InternalMmapVector<u32> &GetSortedSuppressedStacks() {
+    if (!suppressed_stacks_sorted) {
+      suppressed_stacks_sorted = true;
+      SortAndDedup(suppressed_stacks);
+    }
+    return suppressed_stacks;
+  }
+  void PrintMatchedSuppressions();
+};
+
+ALIGNED(64) static char suppression_placeholder[sizeof(LeakSuppressionContext)];
+static LeakSuppressionContext *suppression_ctx = nullptr;
 static const char kSuppressionLeak[] = "leak";
 static const char *kSuppressionTypes[] = { kSuppressionLeak };
 static const char kStdSuppressions[] =
@@ -86,14 +112,20 @@ static const char kStdSuppressions[] =
 void InitializeSuppressions() {
   CHECK_EQ(nullptr, suppression_ctx);
   suppression_ctx = new (suppression_placeholder)
-      SuppressionContext(kSuppressionTypes, ARRAY_SIZE(kSuppressionTypes));
-  suppression_ctx->ParseFromFile(flags()->suppressions);
-  if (&__lsan_default_suppressions)
-    suppression_ctx->Parse(__lsan_default_suppressions());
-  suppression_ctx->Parse(kStdSuppressions);
+      LeakSuppressionContext(kSuppressionTypes, ARRAY_SIZE(kSuppressionTypes));
+}
+
+void LeakSuppressionContext::LazyInit() {
+  if (!parsed) {
+    parsed = true;
+    context.ParseFromFile(flags()->suppressions);
+    if (&__lsan_default_suppressions)
+      context.Parse(__lsan_default_suppressions());
+    context.Parse(kStdSuppressions);
+  }
 }
 
-static SuppressionContext *GetSuppressionContext() {
+static LeakSuppressionContext *GetSuppressionContext() {
   CHECK(suppression_ctx);
   return suppression_ctx;
 }
@@ -221,6 +253,27 @@ extern "C" SANITIZER_WEAK_ATTRIBUTE void __libc_iterate_dynamic_tls(
     pid_t, void (*cb)(void *, void *, uptr, void *), void *);
 #endif
 
+static void ProcessThreadRegistry(Frontier *frontier) {
+  InternalMmapVector<uptr> ptrs;
+  GetThreadRegistryLocked()->RunCallbackForEachThreadLocked(
+      GetAdditionalThreadContextPtrs, &ptrs);
+
+  for (uptr i = 0; i < ptrs.size(); ++i) {
+    void *ptr = reinterpret_cast<void *>(ptrs[i]);
+    uptr chunk = PointsIntoChunk(ptr);
+    if (!chunk)
+      continue;
+    LsanMetadata m(chunk);
+    if (!m.allocated())
+      continue;
+
+    // Mark as reachable and add to frontier.
+    LOG_POINTERS("Treating pointer %p from ThreadContext as reachable\n", ptr);
+    m.set_tag(kReachable);
+    frontier->push_back(chunk);
+  }
+}
+
 // Scans thread data (stacks and TLS) for heap pointers.
 static void ProcessThreads(SuspendedThreadsList const &suspended_threads,
                            Frontier *frontier) {
@@ -315,15 +368,15 @@ static void ProcessThreads(SuspendedThreadsList const &suspended_threads,
       __libc_iterate_dynamic_tls(os_id, cb, frontier);
 #else
       if (dtls && !DTLSInDestruction(dtls)) {
-        for (uptr j = 0; j < dtls->dtv_size; ++j) {
-          uptr dtls_beg = dtls->dtv[j].beg;
-          uptr dtls_end = dtls_beg + dtls->dtv[j].size;
+        ForEachDVT(dtls, [&](const DTLS::DTV &dtv, int id) {
+          uptr dtls_beg = dtv.beg;
+          uptr dtls_end = dtls_beg + dtv.size;
           if (dtls_beg < dtls_end) {
-            LOG_THREADS("DTLS %zu at %p-%p.\n", j, dtls_beg, dtls_end);
+            LOG_THREADS("DTLS %zu at %p-%p.\n", id, dtls_beg, dtls_end);
             ScanRangeForPointers(dtls_beg, dtls_end, frontier, "DTLS",
                                  kReachable);
           }
-        }
+        });
       } else {
         // We are handling a thread with DTLS under destruction. Log about
         // this and continue.
@@ -332,6 +385,9 @@ static void ProcessThreads(SuspendedThreadsList const &suspended_threads,
 #endif
     }
   }
+
+  // Add pointers reachable from ThreadContexts
+  ProcessThreadRegistry(frontier);
 }
 
 #endif  // SANITIZER_FUCHSIA
@@ -390,6 +446,24 @@ static void MarkIndirectlyLeakedCb(uptr chunk, void *arg) {
   }
 }
 
+static void IgnoredSuppressedCb(uptr chunk, void *arg) {
+  CHECK(arg);
+  chunk = GetUserBegin(chunk);
+  LsanMetadata m(chunk);
+  if (!m.allocated() || m.tag() == kIgnored)
+    return;
+
+  const InternalMmapVector<u32> &suppressed =
+      *static_cast<const InternalMmapVector<u32> *>(arg);
+  uptr idx = InternalLowerBound(suppressed, m.stack_trace_id());
+  if (idx >= suppressed.size() || m.stack_trace_id() != suppressed[idx])
+    return;
+
+  LOG_POINTERS("Suppressed: chunk %p-%p of size %zu.\n", chunk,
+               chunk + m.requested_size(), m.requested_size());
+  m.set_tag(kIgnored);
+}
+
 // ForEachChunk callback. If chunk is marked as ignored, adds its address to
 // frontier.
 static void CollectIgnoredCb(uptr chunk, void *arg) {
@@ -473,6 +547,12 @@ void ProcessPC(Frontier *frontier) {
 // Sets the appropriate tag on each chunk.
 static void ClassifyAllChunks(SuspendedThreadsList const &suspended_threads,
                               Frontier *frontier) {
+  const InternalMmapVector<u32> &suppressed_stacks =
+      GetSuppressionContext()->GetSortedSuppressedStacks();
+  if (!suppressed_stacks.empty()) {
+    ForEachChunk(IgnoredSuppressedCb,
+                 const_cast<InternalMmapVector<u32> *>(&suppressed_stacks));
+  }
   ForEachChunk(CollectIgnoredCb, frontier);
   ProcessGlobalRegions(frontier);
   ProcessThreads(suspended_threads, frontier);
@@ -532,18 +612,20 @@ static void CollectLeaksCb(uptr chunk, void *arg) {
   }
 }
 
-static void PrintMatchedSuppressions() {
+void LeakSuppressionContext::PrintMatchedSuppressions() {
   InternalMmapVector<Suppression *> matched;
-  GetSuppressionContext()->GetMatched(&matched);
+  context.GetMatched(&matched);
   if (!matched.size())
     return;
   const char *line = "-----------------------------------------------------";
   Printf("%s\n", line);
   Printf("Suppressions used:\n");
   Printf("  count      bytes template\n");
-  for (uptr i = 0; i < matched.size(); i++)
-    Printf("%7zu %10zu %s\n", static_cast<uptr>(atomic_load_relaxed(
-        &matched[i]->hit_count)), matched[i]->weight, matched[i]->templ);
+  for (uptr i = 0; i < matched.size(); i++) {
+    Printf("%7zu %10zu %s\n",
+           static_cast<uptr>(atomic_load_relaxed(&matched[i]->hit_count)),
+           matched[i]->weight, matched[i]->templ);
+  }
   Printf("%s\n\n", line);
 }
 
@@ -551,8 +633,7 @@ static void ReportIfNotSuspended(ThreadContextBase *tctx, void *arg) {
   const InternalMmapVector<tid_t> &suspended_threads =
       *(const InternalMmapVector<tid_t> *)arg;
   if (tctx->status == ThreadStatusRunning) {
-    uptr i = InternalLowerBound(suspended_threads, 0, suspended_threads.size(),
-                                tctx->os_id, CompareLess<int>());
+    uptr i = InternalLowerBound(suspended_threads, tctx->os_id);
     if (i >= suspended_threads.size() || suspended_threads[i] != tctx->os_id)
       Report("Running thread %d was not suspended. False leaks are possible.\n",
              tctx->os_id);
@@ -595,43 +676,68 @@ static void CheckForLeaksCallback(const SuspendedThreadsList &suspended_threads,
   param->success = true;
 }
 
-static bool CheckForLeaks() {
-  if (&__lsan_is_turned_off && __lsan_is_turned_off())
-    return false;
-  EnsureMainThreadIDIsCorrect();
-  CheckForLeaksParam param;
-  LockStuffAndStopTheWorld(CheckForLeaksCallback, &param);
-
-  if (!param.success) {
-    Report("LeakSanitizer has encountered a fatal error.\n");
-    Report(
-        "HINT: For debugging, try setting environment variable "
-        "LSAN_OPTIONS=verbosity=1:log_threads=1\n");
-    Report(
-        "HINT: LeakSanitizer does not work under ptrace (strace, gdb, etc)\n");
-    Die();
-  }
-  param.leak_report.ApplySuppressions();
-  uptr unsuppressed_count = param.leak_report.UnsuppressedLeakCount();
-  if (unsuppressed_count > 0) {
+static bool PrintResults(LeakReport &report) {
+  uptr unsuppressed_count = report.UnsuppressedLeakCount();
+  if (unsuppressed_count) {
     Decorator d;
-    Printf("\n"
-           "================================================================="
-           "\n");
+    Printf(
+        "\n"
+        "================================================================="
+        "\n");
     Printf("%s", d.Error());
     Report("ERROR: LeakSanitizer: detected memory leaks\n");
     Printf("%s", d.Default());
-    param.leak_report.ReportTopLeaks(flags()->max_leaks);
+    report.ReportTopLeaks(flags()->max_leaks);
   }
   if (common_flags()->print_suppressions)
-    PrintMatchedSuppressions();
+    GetSuppressionContext()->PrintMatchedSuppressions();
   if (unsuppressed_count > 0) {
-    param.leak_report.PrintSummary();
+    report.PrintSummary();
     return true;
   }
   return false;
 }
 
+static bool CheckForLeaks() {
+  if (&__lsan_is_turned_off && __lsan_is_turned_off())
+    return false;
+  // Inside LockStuffAndStopTheWorld we can't run symbolizer, so we can't match
+  // suppressions. However if a stack id was previously suppressed, it should be
+  // suppressed in future checks as well.
+  for (int i = 0;; ++i) {
+    EnsureMainThreadIDIsCorrect();
+    CheckForLeaksParam param;
+    LockStuffAndStopTheWorld(CheckForLeaksCallback, &param);
+    if (!param.success) {
+      Report("LeakSanitizer has encountered a fatal error.\n");
+      Report(
+          "HINT: For debugging, try setting environment variable "
+          "LSAN_OPTIONS=verbosity=1:log_threads=1\n");
+      Report(
+          "HINT: LeakSanitizer does not work under ptrace (strace, gdb, "
+          "etc)\n");
+      Die();
+    }
+    // No new suppressions stacks, so rerun will not help and we can report.
+    if (!param.leak_report.ApplySuppressions())
+      return PrintResults(param.leak_report);
+
+    // No indirect leaks to report, so we are done here.
+    if (!param.leak_report.IndirectUnsuppressedLeakCount())
+      return PrintResults(param.leak_report);
+
+    if (i >= 8) {
+      Report("WARNING: LeakSanitizer gave up on indirect leaks suppression.\n");
+      return PrintResults(param.leak_report);
+    }
+
+    // We found a new previously unseen suppressed call stack. Rerun to make
+    // sure it does not hold indirect leaks.
+    VReport(1, "Rerun with %zu suppressed stacks.",
+            GetSuppressionContext()->GetSortedSuppressedStacks().size());
+  }
+}
+
 static bool has_reported_leaks = false;
 bool HasReportedLeaks() { return has_reported_leaks; }
 
@@ -652,21 +758,20 @@ static int DoRecoverableLeakCheck() {
 
 void DoRecoverableLeakCheckVoid() { DoRecoverableLeakCheck(); }
 
-static Suppression *GetSuppressionForAddr(uptr addr) {
+Suppression *LeakSuppressionContext::GetSuppressionForAddr(uptr addr) {
   Suppression *s = nullptr;
 
   // Suppress by module name.
-  SuppressionContext *suppressions = GetSuppressionContext();
   if (const char *module_name =
           Symbolizer::GetOrInit()->GetModuleNameForPc(addr))
-    if (suppressions->Match(module_name, kSuppressionLeak, &s))
+    if (context.Match(module_name, kSuppressionLeak, &s))
       return s;
 
   // Suppress by file or function name.
   SymbolizedStack *frames = Symbolizer::GetOrInit()->SymbolizePC(addr);
   for (SymbolizedStack *cur = frames; cur; cur = cur->next) {
-    if (suppressions->Match(cur->info.function, kSuppressionLeak, &s) ||
-        suppressions->Match(cur->info.file, kSuppressionLeak, &s)) {
+    if (context.Match(cur->info.function, kSuppressionLeak, &s) ||
+        context.Match(cur->info.file, kSuppressionLeak, &s)) {
       break;
     }
   }
@@ -674,12 +779,18 @@ static Suppression *GetSuppressionForAddr(uptr addr) {
   return s;
 }
 
-static Suppression *GetSuppressionForStack(u32 stack_trace_id) {
+Suppression *LeakSuppressionContext::GetSuppressionForStack(
+    u32 stack_trace_id) {
+  LazyInit();
   StackTrace stack = StackDepotGet(stack_trace_id);
   for (uptr i = 0; i < stack.size; i++) {
     Suppression *s = GetSuppressionForAddr(
         StackTrace::GetPreviousInstructionPc(stack.trace[i]));
-    if (s) return s;
+    if (s) {
+      suppressed_stacks_sorted = false;
+      suppressed_stacks.push_back(stack_trace_id);
+      return s;
+    }
   }
   return nullptr;
 }
@@ -784,22 +895,27 @@ void LeakReport::PrintSummary() {
       bytes += leaks_[i].total_size;
       allocations += leaks_[i].hit_count;
   }
-  InternalScopedString summary(kMaxSummaryLength);
+  InternalScopedString summary;
   summary.append("%zu byte(s) leaked in %zu allocation(s).", bytes,
                  allocations);
   ReportErrorSummary(summary.data());
 }
 
-void LeakReport::ApplySuppressions() {
+uptr LeakReport::ApplySuppressions() {
+  LeakSuppressionContext *suppressions = GetSuppressionContext();
+  uptr new_suppressions = false;
   for (uptr i = 0; i < leaks_.size(); i++) {
-    Suppression *s = GetSuppressionForStack(leaks_[i].stack_trace_id);
+    Suppression *s =
+        suppressions->GetSuppressionForStack(leaks_[i].stack_trace_id);
     if (s) {
       s->weight += leaks_[i].total_size;
       atomic_store_relaxed(&s->hit_count, atomic_load_relaxed(&s->hit_count) +
           leaks_[i].hit_count);
       leaks_[i].is_suppressed = true;
+      ++new_suppressions;
     }
   }
+  return new_suppressions;
 }
 
 uptr LeakReport::UnsuppressedLeakCount() {
@@ -809,6 +925,14 @@ uptr LeakReport::UnsuppressedLeakCount() {
   return result;
 }
 
+uptr LeakReport::IndirectUnsuppressedLeakCount() {
+  uptr result = 0;
+  for (uptr i = 0; i < leaks_.size(); i++)
+    if (!leaks_[i].is_suppressed && !leaks_[i].is_directly_leaked)
+      result++;
+  return result;
+}
+
 } // namespace __lsan
 #else // CAN_SANITIZE_LEAKS
 namespace __lsan {
diff --git a/libsanitizer/lsan/lsan_common.h b/libsanitizer/lsan/lsan_common.h
index 1fdce08..fe855cf 100644
--- a/libsanitizer/lsan/lsan_common.h
+++ b/libsanitizer/lsan/lsan_common.h
@@ -41,6 +41,8 @@
 #define CAN_SANITIZE_LEAKS 1
 #elif defined(__arm__) && SANITIZER_LINUX
 #define CAN_SANITIZE_LEAKS 1
+#elif SANITIZER_RISCV64 && SANITIZER_LINUX
+#define CAN_SANITIZE_LEAKS 1
 #elif SANITIZER_NETBSD || SANITIZER_FUCHSIA
 #define CAN_SANITIZE_LEAKS 1
 #else
@@ -50,6 +52,7 @@
 namespace __sanitizer {
 class FlagParser;
 class ThreadRegistry;
+class ThreadContextBase;
 struct DTLS;
 }
 
@@ -63,8 +66,6 @@ enum ChunkTag {
   kIgnored = 3
 };
 
-const u32 kInvalidTid = (u32) -1;
-
 struct Flags {
 #define LSAN_FLAG(Type, Name, DefaultValue, Description) Type Name;
 #include "lsan_flags.inc"
@@ -103,8 +104,9 @@ class LeakReport {
                       ChunkTag tag);
   void ReportTopLeaks(uptr max_leaks);
   void PrintSummary();
-  void ApplySuppressions();
+  uptr ApplySuppressions();
   uptr UnsuppressedLeakCount();
+  uptr IndirectUnsuppressedLeakCount();
 
  private:
   void PrintReportForLeak(uptr index);
@@ -141,6 +143,7 @@ InternalMmapVector<RootRegion> const *GetRootRegions();
 void ScanRootRegion(Frontier *frontier, RootRegion const &region,
                     uptr region_begin, uptr region_end, bool is_readable);
 void ForEachExtraStackRangeCb(uptr begin, uptr end, void* arg);
+void GetAdditionalThreadContextPtrs(ThreadContextBase *tctx, void *ptrs);
 // Run stoptheworld while holding any platform-specific locks, as well as the
 // allocator and thread registry locks.
 void LockStuffAndStopTheWorld(StopTheWorldCallback callback,
diff --git a/libsanitizer/lsan/lsan_common_fuchsia.cpp b/libsanitizer/lsan/lsan_common_fuchsia.cpp
index 3c62c94..2d35fa5 100644
--- a/libsanitizer/lsan/lsan_common_fuchsia.cpp
+++ b/libsanitizer/lsan/lsan_common_fuchsia.cpp
@@ -107,9 +107,7 @@ void LockStuffAndStopTheWorld(StopTheWorldCallback callback,
     auto params = static_cast<const Params *>(data);
     uptr begin = reinterpret_cast<uptr>(chunk);
     uptr end = begin + size;
-    auto i = __sanitizer::InternalLowerBound(params->allocator_caches, 0,
-                                             params->allocator_caches.size(),
-                                             begin, CompareLess<uptr>());
+    auto i = __sanitizer::InternalLowerBound(params->allocator_caches, begin);
     if (i < params->allocator_caches.size() &&
         params->allocator_caches[i] >= begin &&
         end - params->allocator_caches[i] <= sizeof(AllocatorCache)) {
diff --git a/libsanitizer/lsan/lsan_fuchsia.h b/libsanitizer/lsan/lsan_fuchsia.h
index 65d20ea..e730d8f 100644
--- a/libsanitizer/lsan/lsan_fuchsia.h
+++ b/libsanitizer/lsan/lsan_fuchsia.h
@@ -23,7 +23,7 @@
 
 namespace __lsan {
 
-class ThreadContext : public ThreadContextLsanBase {
+class ThreadContext final : public ThreadContextLsanBase {
  public:
   explicit ThreadContext(int tid);
   void OnCreated(void *arg) override;
diff --git a/libsanitizer/lsan/lsan_interceptors.cpp b/libsanitizer/lsan/lsan_interceptors.cpp
index bf8d316..90a90a5 100644
--- a/libsanitizer/lsan/lsan_interceptors.cpp
+++ b/libsanitizer/lsan/lsan_interceptors.cpp
@@ -460,7 +460,7 @@ INTERCEPTOR(int, pthread_create, void *th, void *attr,
   if (res == 0) {
     int tid = ThreadCreate(GetCurrentThread(), *(uptr *)th,
                            IsStateDetached(detached));
-    CHECK_NE(tid, 0);
+    CHECK_NE(tid, kMainTid);
     atomic_store(&p.tid, tid, memory_order_release);
     while (atomic_load(&p.tid, memory_order_acquire) != 0)
       internal_sched_yield();
diff --git a/libsanitizer/lsan/lsan_posix.cpp b/libsanitizer/lsan/lsan_posix.cpp
index 8e05915..5d1c3f6 100644
--- a/libsanitizer/lsan/lsan_posix.cpp
+++ b/libsanitizer/lsan/lsan_posix.cpp
@@ -48,7 +48,7 @@ void ThreadStart(u32 tid, tid_t os_id, ThreadType thread_type) {
   OnStartedArgs args;
   uptr stack_size = 0;
   uptr tls_size = 0;
-  GetThreadStackAndTls(tid == 0, &args.stack_begin, &stack_size,
+  GetThreadStackAndTls(tid == kMainTid, &args.stack_begin, &stack_size,
                        &args.tls_begin, &tls_size);
   args.stack_end = args.stack_begin + stack_size;
   args.tls_end = args.tls_begin + tls_size;
@@ -75,8 +75,8 @@ bool GetThreadRangesLocked(tid_t os_id, uptr *stack_begin, uptr *stack_end,
 }
 
 void InitializeMainThread() {
-  u32 tid = ThreadCreate(0, 0, true);
-  CHECK_EQ(tid, 0);
+  u32 tid = ThreadCreate(kMainTid, 0, true);
+  CHECK_EQ(tid, kMainTid);
   ThreadStart(tid, GetTid());
 }
 
diff --git a/libsanitizer/lsan/lsan_thread.cpp b/libsanitizer/lsan/lsan_thread.cpp
index 371a1f2..8efb54a 100644
--- a/libsanitizer/lsan/lsan_thread.cpp
+++ b/libsanitizer/lsan/lsan_thread.cpp
@@ -94,7 +94,7 @@ void ThreadJoin(u32 tid) {
 }
 
 void EnsureMainThreadIDIsCorrect() {
-  if (GetCurrentThread() == 0)
+  if (GetCurrentThread() == kMainTid)
     CurrentThreadContext()->os_id = GetTid();
 }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h b/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
index 33f89d6..eb836bc 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_combined.h
@@ -35,9 +35,9 @@ class CombinedAllocator {
     secondary_.InitLinkerInitialized();
   }
 
-  void Init(s32 release_to_os_interval_ms) {
+  void Init(s32 release_to_os_interval_ms, uptr heap_start = 0) {
     stats_.Init();
-    primary_.Init(release_to_os_interval_ms);
+    primary_.Init(release_to_os_interval_ms, heap_start);
     secondary_.Init();
   }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h b/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
index b90dabb..fb5394c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_primary32.h
@@ -119,7 +119,8 @@ class SizeClassAllocator32 {
   typedef SizeClassAllocator32<Params> ThisT;
   typedef SizeClassAllocator32LocalCache<ThisT> AllocatorCache;
 
-  void Init(s32 release_to_os_interval_ms) {
+  void Init(s32 release_to_os_interval_ms, uptr heap_start = 0) {
+    CHECK(!heap_start);
     possible_regions.Init();
     internal_memset(size_class_info_array, 0, sizeof(size_class_info_array));
   }
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h b/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
index 0a18b0c..db30e13 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_primary64.h
@@ -19,7 +19,7 @@ template<class SizeClassAllocator> struct SizeClassAllocator64LocalCache;
 // The template parameter Params is a class containing the actual parameters.
 //
 // Space: a portion of address space of kSpaceSize bytes starting at SpaceBeg.
-// If kSpaceBeg is ~0 then SpaceBeg is chosen dynamically my mmap.
+// If kSpaceBeg is ~0 then SpaceBeg is chosen dynamically by mmap.
 // Otherwise SpaceBeg=kSpaceBeg (fixed address).
 // kSpaceSize is a power of two.
 // At the beginning the entire space is mprotect-ed, then small parts of it
@@ -69,25 +69,45 @@ class SizeClassAllocator64 {
     return base + (static_cast<uptr>(ptr32) << kCompactPtrScale);
   }
 
-  void Init(s32 release_to_os_interval_ms) {
+  // If heap_start is nonzero, assumes kSpaceSize bytes are already mapped R/W
+  // at heap_start and places the heap there.  This mode requires kSpaceBeg ==
+  // ~(uptr)0.
+  void Init(s32 release_to_os_interval_ms, uptr heap_start = 0) {
     uptr TotalSpaceSize = kSpaceSize + AdditionalSize();
-    if (kUsingConstantSpaceBeg) {
-      CHECK(IsAligned(kSpaceBeg, SizeClassMap::kMaxSize));
-      CHECK_EQ(kSpaceBeg, address_range.Init(TotalSpaceSize,
-                                             PrimaryAllocatorName, kSpaceBeg));
+    PremappedHeap = heap_start != 0;
+    if (PremappedHeap) {
+      CHECK(!kUsingConstantSpaceBeg);
+      NonConstSpaceBeg = heap_start;
+      uptr RegionInfoSize = AdditionalSize();
+      RegionInfoSpace =
+          address_range.Init(RegionInfoSize, PrimaryAllocatorName);
+      CHECK_NE(RegionInfoSpace, ~(uptr)0);
+      CHECK_EQ(RegionInfoSpace,
+               address_range.MapOrDie(RegionInfoSpace, RegionInfoSize,
+                                      "SizeClassAllocator: region info"));
+      MapUnmapCallback().OnMap(RegionInfoSpace, RegionInfoSize);
     } else {
-      // Combined allocator expects that an 2^N allocation is always aligned to
-      // 2^N. For this to work, the start of the space needs to be aligned as
-      // high as the largest size class (which also needs to be a power of 2).
-      NonConstSpaceBeg = address_range.InitAligned(
-          TotalSpaceSize, SizeClassMap::kMaxSize, PrimaryAllocatorName);
-      CHECK_NE(NonConstSpaceBeg, ~(uptr)0);
+      if (kUsingConstantSpaceBeg) {
+        CHECK(IsAligned(kSpaceBeg, SizeClassMap::kMaxSize));
+        CHECK_EQ(kSpaceBeg,
+                 address_range.Init(TotalSpaceSize, PrimaryAllocatorName,
+                                    kSpaceBeg));
+      } else {
+        // Combined allocator expects that an 2^N allocation is always aligned
+        // to 2^N. For this to work, the start of the space needs to be aligned
+        // as high as the largest size class (which also needs to be a power of
+        // 2).
+        NonConstSpaceBeg = address_range.InitAligned(
+            TotalSpaceSize, SizeClassMap::kMaxSize, PrimaryAllocatorName);
+        CHECK_NE(NonConstSpaceBeg, ~(uptr)0);
+      }
+      RegionInfoSpace = SpaceEnd();
+      MapWithCallbackOrDie(RegionInfoSpace, AdditionalSize(),
+                           "SizeClassAllocator: region info");
     }
     SetReleaseToOSIntervalMs(release_to_os_interval_ms);
-    MapWithCallbackOrDie(SpaceEnd(), AdditionalSize(),
-                         "SizeClassAllocator: region info");
     // Check that the RegionInfo array is aligned on the CacheLine size.
-    DCHECK_EQ(SpaceEnd() % kCacheLineSize, 0);
+    DCHECK_EQ(RegionInfoSpace % kCacheLineSize, 0);
   }
 
   s32 ReleaseToOSIntervalMs() const {
@@ -144,6 +164,17 @@ class SizeClassAllocator64 {
     CompactPtrT *free_array = GetFreeArray(region_beg);
 
     BlockingMutexLock l(&region->mutex);
+#if SANITIZER_WINDOWS
+    /* On Windows unmapping of memory during __sanitizer_purge_allocator is
+    explicit and immediate, so unmapped regions must be explicitly mapped back
+    in when they are accessed again. */
+    if (region->rtoi.last_released_bytes > 0) {
+      MmapFixedOrDie(region_beg, region->mapped_user,
+                                      "SizeClassAllocator: region data");
+      region->rtoi.n_freed_at_last_release = 0;
+      region->rtoi.last_released_bytes = 0;
+    }
+#endif
     if (UNLIKELY(region->num_freed_chunks < n_chunks)) {
       if (UNLIKELY(!PopulateFreeArray(stat, class_id, region,
                                       n_chunks - region->num_freed_chunks)))
@@ -360,8 +391,7 @@ class SizeClassAllocator64 {
     }
     ~PackedCounterArray() {
       if (buffer) {
-        memory_mapper->UnmapPackedCounterArrayBuffer(
-            reinterpret_cast<uptr>(buffer), buffer_size);
+        memory_mapper->UnmapPackedCounterArrayBuffer(buffer, buffer_size);
       }
     }
 
@@ -586,6 +616,11 @@ class SizeClassAllocator64 {
 
   atomic_sint32_t release_to_os_interval_ms_;
 
+  uptr RegionInfoSpace;
+
+  // True if the user has already mapped the entire heap R/W.
+  bool PremappedHeap;
+
   struct Stats {
     uptr n_allocated;
     uptr n_freed;
@@ -615,7 +650,7 @@ class SizeClassAllocator64 {
 
   RegionInfo *GetRegionInfo(uptr class_id) const {
     DCHECK_LT(class_id, kNumClasses);
-    RegionInfo *regions = reinterpret_cast<RegionInfo *>(SpaceEnd());
+    RegionInfo *regions = reinterpret_cast<RegionInfo *>(RegionInfoSpace);
     return &regions[class_id];
   }
 
@@ -640,6 +675,9 @@ class SizeClassAllocator64 {
   }
 
   bool MapWithCallback(uptr beg, uptr size, const char *name) {
+    if (PremappedHeap)
+      return beg >= NonConstSpaceBeg &&
+             beg + size <= NonConstSpaceBeg + kSpaceSize;
     uptr mapped = address_range.Map(beg, size, name);
     if (UNLIKELY(!mapped))
       return false;
@@ -649,11 +687,18 @@ class SizeClassAllocator64 {
   }
 
   void MapWithCallbackOrDie(uptr beg, uptr size, const char *name) {
+    if (PremappedHeap) {
+      CHECK_GE(beg, NonConstSpaceBeg);
+      CHECK_LE(beg + size, NonConstSpaceBeg + kSpaceSize);
+      return;
+    }
     CHECK_EQ(beg, address_range.MapOrDie(beg, size, name));
     MapUnmapCallback().OnMap(beg, size);
   }
 
   void UnmapWithCallbackOrDie(uptr beg, uptr size) {
+    if (PremappedHeap)
+      return;
     MapUnmapCallback().OnUnmap(beg, size);
     address_range.Unmap(beg, size);
   }
@@ -792,17 +837,16 @@ class SizeClassAllocator64 {
       return released_bytes;
     }
 
-    uptr MapPackedCounterArrayBuffer(uptr buffer_size) {
+    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 reinterpret_cast<uptr>(
-          MmapOrDieOnFatalError(buffer_size, "ReleaseToOSPageCounters"));
+      return MmapOrDieOnFatalError(buffer_size, "ReleaseToOSPageCounters");
     }
 
-    void UnmapPackedCounterArrayBuffer(uptr buffer, uptr buffer_size) {
-      UnmapOrDie(reinterpret_cast<void *>(buffer), buffer_size);
+    void UnmapPackedCounterArrayBuffer(void *buffer, uptr buffer_size) {
+      UnmapOrDie(buffer, buffer_size);
     }
 
     // Releases [from, to) range of pages back to OS.
@@ -823,6 +867,9 @@ class SizeClassAllocator64 {
 
   // 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) {
     RegionInfo *region = GetRegionInfo(class_id);
     const uptr chunk_size = ClassIdToSize(class_id);
diff --git a/libsanitizer/sanitizer_common/sanitizer_allocator_size_class_map.h b/libsanitizer/sanitizer_common/sanitizer_allocator_size_class_map.h
index 12d8c89..c50d133 100644
--- a/libsanitizer/sanitizer_common/sanitizer_allocator_size_class_map.h
+++ b/libsanitizer/sanitizer_common/sanitizer_allocator_size_class_map.h
@@ -24,7 +24,7 @@
 //             E.g. with kNumBits==3 all size classes after 2^kMidSizeLog
 //             look like 0b1xx0..0, where x is either 0 or 1.
 //
-// Example: kNumBits=3, kMidSizeLog=4, kMidSizeLog=8, kMaxSizeLog=17:
+// Example: kNumBits=3, kMinSizeLog=4, kMidSizeLog=8, kMaxSizeLog=17:
 //
 // Classes 1 - 16 correspond to sizes 16 to 256 (size = class_id * 16).
 // Next 4 classes: 256 + i * 64  (i = 1 to 4).
diff --git a/libsanitizer/sanitizer_common/sanitizer_atomic_clang_mips.h b/libsanitizer/sanitizer_common/sanitizer_atomic_clang_mips.h
index 59155e9..2b39097 100644
--- a/libsanitizer/sanitizer_common/sanitizer_atomic_clang_mips.h
+++ b/libsanitizer/sanitizer_common/sanitizer_atomic_clang_mips.h
@@ -41,7 +41,7 @@ inline atomic_uint64_t::Type atomic_fetch_add(volatile atomic_uint64_t *ptr,
                                               atomic_uint64_t::Type val,
                                               memory_order mo) {
   DCHECK(mo &
-         (memory_order_relaxed | memory_order_releasae | memory_order_seq_cst));
+         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
   DCHECK(!((uptr)ptr % sizeof(*ptr)));
 
   atomic_uint64_t::Type ret;
@@ -67,7 +67,7 @@ inline bool atomic_compare_exchange_strong(volatile atomic_uint64_t *ptr,
                                            atomic_uint64_t::Type xchg,
                                            memory_order mo) {
   DCHECK(mo &
-         (memory_order_relaxed | memory_order_releasae | memory_order_seq_cst));
+         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
   DCHECK(!((uptr)ptr % sizeof(*ptr)));
 
   typedef atomic_uint64_t::Type Type;
@@ -90,7 +90,7 @@ template <>
 inline atomic_uint64_t::Type atomic_load(const volatile atomic_uint64_t *ptr,
                                          memory_order mo) {
   DCHECK(mo &
-         (memory_order_relaxed | memory_order_releasae | memory_order_seq_cst));
+         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
   DCHECK(!((uptr)ptr % sizeof(*ptr)));
 
   atomic_uint64_t::Type zero = 0;
@@ -103,7 +103,7 @@ template <>
 inline void atomic_store(volatile atomic_uint64_t *ptr, atomic_uint64_t::Type v,
                          memory_order mo) {
   DCHECK(mo &
-         (memory_order_relaxed | memory_order_releasae | memory_order_seq_cst));
+         (memory_order_relaxed | memory_order_release | memory_order_seq_cst));
   DCHECK(!((uptr)ptr % sizeof(*ptr)));
 
   __spin_lock(&lock.lock);
diff --git a/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.cpp b/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.cpp
new file mode 100644
index 0000000..250ac39
--- /dev/null
+++ b/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.cpp
@@ -0,0 +1,108 @@
+//===-- sanitizer_chained_origin_depot.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
+//
+//===----------------------------------------------------------------------===//
+//
+// A storage for chained origins.
+//===----------------------------------------------------------------------===//
+
+#include "sanitizer_chained_origin_depot.h"
+
+namespace __sanitizer {
+
+bool ChainedOriginDepot::ChainedOriginDepotNode::eq(
+    u32 hash, const args_type &args) const {
+  return here_id == args.here_id && prev_id == args.prev_id;
+}
+
+uptr ChainedOriginDepot::ChainedOriginDepotNode::storage_size(
+    const args_type &args) {
+  return sizeof(ChainedOriginDepotNode);
+}
+
+/* This is murmur2 hash for the 64->32 bit case.
+   It does not behave all that well because the keys have a very biased
+   distribution (I've seen 7-element buckets with the table only 14% full).
+
+   here_id is built of
+   * (1 bits) Reserved, zero.
+   * (8 bits) Part id = bits 13..20 of the hash value of here_id's key.
+   * (23 bits) Sequential number (each part has each own sequence).
+
+   prev_id has either the same distribution as here_id (but with 3:8:21)
+   split, or one of two reserved values (-1) or (-2). Either case can
+   dominate depending on the workload.
+*/
+u32 ChainedOriginDepot::ChainedOriginDepotNode::hash(const args_type &args) {
+  const u32 m = 0x5bd1e995;
+  const u32 seed = 0x9747b28c;
+  const u32 r = 24;
+  u32 h = seed;
+  u32 k = args.here_id;
+  k *= m;
+  k ^= k >> r;
+  k *= m;
+  h *= m;
+  h ^= k;
+
+  k = args.prev_id;
+  k *= m;
+  k ^= k >> r;
+  k *= m;
+  h *= m;
+  h ^= k;
+
+  h ^= h >> 13;
+  h *= m;
+  h ^= h >> 15;
+  return h;
+}
+
+bool ChainedOriginDepot::ChainedOriginDepotNode::is_valid(
+    const args_type &args) {
+  return true;
+}
+
+void ChainedOriginDepot::ChainedOriginDepotNode::store(const args_type &args,
+                                                       u32 other_hash) {
+  here_id = args.here_id;
+  prev_id = args.prev_id;
+}
+
+ChainedOriginDepot::ChainedOriginDepotNode::args_type
+ChainedOriginDepot::ChainedOriginDepotNode::load() const {
+  args_type ret = {here_id, prev_id};
+  return ret;
+}
+
+ChainedOriginDepot::ChainedOriginDepotNode::Handle
+ChainedOriginDepot::ChainedOriginDepotNode::get_handle() {
+  return Handle(this);
+}
+
+ChainedOriginDepot::ChainedOriginDepot() {}
+
+StackDepotStats *ChainedOriginDepot::GetStats() { return depot.GetStats(); }
+
+bool ChainedOriginDepot::Put(u32 here_id, u32 prev_id, u32 *new_id) {
+  ChainedOriginDepotDesc desc = {here_id, prev_id};
+  bool inserted;
+  ChainedOriginDepotNode::Handle h = depot.Put(desc, &inserted);
+  *new_id = h.valid() ? h.id() : 0;
+  return inserted;
+}
+
+u32 ChainedOriginDepot::Get(u32 id, u32 *other) {
+  ChainedOriginDepotDesc desc = depot.Get(id);
+  *other = desc.prev_id;
+  return desc.here_id;
+}
+
+void ChainedOriginDepot::LockAll() { depot.LockAll(); }
+
+void ChainedOriginDepot::UnlockAll() { depot.UnlockAll(); }
+
+}  // namespace __sanitizer
diff --git a/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.h b/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.h
new file mode 100644
index 0000000..453cdf6
--- /dev/null
+++ b/libsanitizer/sanitizer_common/sanitizer_chained_origin_depot.h
@@ -0,0 +1,88 @@
+//===-- sanitizer_chained_origin_depot.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
+//
+//===----------------------------------------------------------------------===//
+//
+// A storage for chained origins.
+//===----------------------------------------------------------------------===//
+
+#ifndef SANITIZER_CHAINED_ORIGIN_DEPOT_H
+#define SANITIZER_CHAINED_ORIGIN_DEPOT_H
+
+#include "sanitizer_common.h"
+#include "sanitizer_stackdepotbase.h"
+
+namespace __sanitizer {
+
+class ChainedOriginDepot {
+ public:
+  ChainedOriginDepot();
+
+  // Gets the statistic of the origin chain storage.
+  StackDepotStats *GetStats();
+
+  // Stores a chain with StackDepot ID here_id and previous chain ID prev_id.
+  // If successful, returns true and the new chain id new_id.
+  // If the same element already exists, returns false and sets new_id to the
+  // existing ID.
+  bool Put(u32 here_id, u32 prev_id, u32 *new_id);
+
+  // Retrieves the stored StackDepot ID for the given origin ID.
+  u32 Get(u32 id, u32 *other);
+
+  void LockAll();
+  void UnlockAll();
+
+ private:
+  struct ChainedOriginDepotDesc {
+    u32 here_id;
+    u32 prev_id;
+  };
+
+  struct ChainedOriginDepotNode {
+    ChainedOriginDepotNode *link;
+    u32 id;
+    u32 here_id;
+    u32 prev_id;
+
+    typedef ChainedOriginDepotDesc args_type;
+
+    bool eq(u32 hash, const args_type &args) const;
+
+    static uptr storage_size(const args_type &args);
+
+    static u32 hash(const args_type &args);
+
+    static bool is_valid(const args_type &args);
+
+    void store(const args_type &args, u32 other_hash);
+
+    args_type load() const;
+
+    struct Handle {
+      ChainedOriginDepotNode *node_;
+      Handle() : node_(nullptr) {}
+      explicit Handle(ChainedOriginDepotNode *node) : node_(node) {}
+      bool valid() { return node_; }
+      u32 id() { return node_->id; }
+      int here_id() { return node_->here_id; }
+      int prev_id() { return node_->prev_id; }
+    };
+
+    Handle get_handle();
+
+    typedef Handle handle_type;
+  };
+
+  StackDepotBase<ChainedOriginDepotNode, 4, 20> depot;
+
+  ChainedOriginDepot(const ChainedOriginDepot &) = delete;
+  void operator=(const ChainedOriginDepot &) = delete;
+};
+
+}  // namespace __sanitizer
+
+#endif  // SANITIZER_CHAINED_ORIGIN_DEPOT_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_common.cpp b/libsanitizer/sanitizer_common/sanitizer_common.cpp
index 87efda5..33960d9 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_common.cpp
@@ -87,7 +87,7 @@ const char *StripModuleName(const char *module) {
 void ReportErrorSummary(const char *error_message, const char *alt_tool_name) {
   if (!common_flags()->print_summary)
     return;
-  InternalScopedString buff(kMaxSummaryLength);
+  InternalScopedString buff;
   buff.append("SUMMARY: %s: %s",
               alt_tool_name ? alt_tool_name : SanitizerToolName, error_message);
   __sanitizer_report_error_summary(buff.data());
@@ -274,6 +274,14 @@ uptr ReadBinaryNameCached(/*out*/char *buf, uptr buf_len) {
   return name_len;
 }
 
+uptr ReadBinaryDir(/*out*/ char *buf, uptr buf_len) {
+  ReadBinaryNameCached(buf, buf_len);
+  const char *exec_name_pos = StripModuleName(buf);
+  uptr name_len = exec_name_pos - buf;
+  buf[name_len] = '\0';
+  return name_len;
+}
+
 #if !SANITIZER_GO
 void PrintCmdline() {
   char **argv = GetArgv();
diff --git a/libsanitizer/sanitizer_common/sanitizer_common.h b/libsanitizer/sanitizer_common/sanitizer_common.h
index bce24d6..7b65dd7 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common.h
+++ b/libsanitizer/sanitizer_common/sanitizer_common.h
@@ -44,7 +44,7 @@ const uptr kMaxPathLength = 4096;
 
 const uptr kMaxThreadStackSize = 1 << 30;  // 1Gb
 
-static const uptr kErrorMessageBufferSize = 1 << 16;
+const uptr kErrorMessageBufferSize = 1 << 16;
 
 // Denotes fake PC values that come from JIT/JAVA/etc.
 // For such PC values __tsan_symbolize_external_ex() will be called.
@@ -135,6 +135,15 @@ void UnmapFromTo(uptr from, uptr to);
 uptr MapDynamicShadow(uptr shadow_size_bytes, uptr shadow_scale,
                       uptr min_shadow_base_alignment, uptr &high_mem_end);
 
+// Let S = max(shadow_size, num_aliases * alias_size, ring_buffer_size).
+// Reserves 2*S bytes of address space to the right of the returned address and
+// ring_buffer_size bytes to the left.  The returned address is aligned to 2*S.
+// Also creates num_aliases regions of accessible memory starting at offset S
+// from the returned address.  Each region has size alias_size and is backed by
+// the same physical memory.
+uptr MapDynamicShadowAndAliases(uptr shadow_size, uptr alias_size,
+                                uptr num_aliases, uptr ring_buffer_size);
+
 // Reserve memory range [beg, end]. If madvise_shadow is true then apply
 // madvise (e.g. hugepages, core dumping) requested by options.
 void ReserveShadowMemoryRange(uptr beg, uptr end, const char *name,
@@ -248,6 +257,7 @@ const char *StripModuleName(const char *module);
 // OS
 uptr ReadBinaryName(/*out*/char *buf, uptr buf_len);
 uptr ReadBinaryNameCached(/*out*/char *buf, uptr buf_len);
+uptr ReadBinaryDir(/*out*/ char *buf, uptr buf_len);
 uptr ReadLongProcessName(/*out*/ char *buf, uptr buf_len);
 const char *GetProcessName();
 void UpdateProcessName();
@@ -294,8 +304,8 @@ void NORETURN ReportMmapFailureAndDie(uptr size, const char *mem_type,
                                       const char *mmap_type, error_t err,
                                       bool raw_report = false);
 
-// Specific tools may override behavior of "Die" and "CheckFailed" functions
-// to do tool-specific job.
+// Specific tools may override behavior of "Die" function to do tool-specific
+// job.
 typedef void (*DieCallbackType)(void);
 
 // It's possible to add several callbacks that would be run when "Die" is
@@ -307,9 +317,7 @@ bool RemoveDieCallback(DieCallbackType callback);
 
 void SetUserDieCallback(DieCallbackType callback);
 
-typedef void (*CheckFailedCallbackType)(const char *, int, const char *,
-                                       u64, u64);
-void SetCheckFailedCallback(CheckFailedCallbackType callback);
+void SetCheckUnwindCallback(void (*callback)());
 
 // Callback will be called if soft_rss_limit_mb is given and the limit is
 // exceeded (exceeded==true) or if rss went down below the limit
@@ -343,8 +351,6 @@ void ReportDeadlySignal(const SignalContext &sig, u32 tid,
 void SetAlternateSignalStack();
 void UnsetAlternateSignalStack();
 
-// We don't want a summary too long.
-const int kMaxSummaryLength = 1024;
 // Construct a one-line string:
 //   SUMMARY: SanitizerToolName: error_message
 // and pass it to __sanitizer_report_error_summary.
@@ -441,8 +447,14 @@ inline uptr Log2(uptr x) {
 
 // Don't use std::min, std::max or std::swap, to minimize dependency
 // on libstdc++.
-template<class T> T Min(T a, T b) { return a < b ? a : b; }
-template<class T> T Max(T a, T b) { return a > b ? a : b; }
+template <class T>
+constexpr T Min(T a, T b) {
+  return a < b ? a : b;
+}
+template <class T>
+constexpr T Max(T a, T b) {
+  return a > b ? a : b;
+}
 template<class T> void Swap(T& a, T& b) {
   T tmp = a;
   a = b;
@@ -467,6 +479,7 @@ inline int ToLower(int c) {
 template<typename T>
 class InternalMmapVectorNoCtor {
  public:
+  using value_type = T;
   void Initialize(uptr initial_capacity) {
     capacity_bytes_ = 0;
     size_ = 0;
@@ -590,21 +603,21 @@ class InternalMmapVector : public InternalMmapVectorNoCtor<T> {
   InternalMmapVector &operator=(InternalMmapVector &&) = delete;
 };
 
-class InternalScopedString : public InternalMmapVector<char> {
+class InternalScopedString {
  public:
-  explicit InternalScopedString(uptr max_length)
-      : InternalMmapVector<char>(max_length), length_(0) {
-    (*this)[0] = '\0';
-  }
-  uptr length() { return length_; }
+  InternalScopedString() : buffer_(1) { buffer_[0] = '\0'; }
+
+  uptr length() const { return buffer_.size() - 1; }
   void clear() {
-    (*this)[0] = '\0';
-    length_ = 0;
+    buffer_.resize(1);
+    buffer_[0] = '\0';
   }
   void append(const char *format, ...);
+  const char *data() const { return buffer_.data(); }
+  char *data() { return buffer_.data(); }
 
  private:
-  uptr length_;
+  InternalMmapVector<char> buffer_;
 };
 
 template <class T>
@@ -651,9 +664,13 @@ void Sort(T *v, uptr size, Compare comp = {}) {
 
 // Works like std::lower_bound: finds the first element that is not less
 // than the val.
-template <class Container, class Value, class Compare>
-uptr InternalLowerBound(const Container &v, uptr first, uptr last,
-                        const Value &val, Compare comp) {
+template <class Container,
+          class Compare = CompareLess<typename Container::value_type>>
+uptr InternalLowerBound(const Container &v,
+                        const typename Container::value_type &val,
+                        Compare comp = {}) {
+  uptr first = 0;
+  uptr last = v.size();
   while (last > first) {
     uptr mid = (first + last) / 2;
     if (comp(v[mid], val))
@@ -677,6 +694,27 @@ enum ModuleArch {
   kModuleArchRISCV64
 };
 
+// Sorts and removes duplicates from the container.
+template <class Container,
+          class Compare = CompareLess<typename Container::value_type>>
+void SortAndDedup(Container &v, Compare comp = {}) {
+  Sort(v.data(), v.size(), comp);
+  uptr size = v.size();
+  if (size < 2)
+    return;
+  uptr last = 0;
+  for (uptr i = 1; i < size; ++i) {
+    if (comp(v[last], v[i])) {
+      ++last;
+      if (last != i)
+        v[last] = v[i];
+    } else {
+      CHECK(!comp(v[i], v[last]));
+    }
+  }
+  v.resize(last + 1);
+}
+
 // Opens the file 'file_name" and reads up to 'max_len' bytes.
 // The resulting buffer is mmaped and stored in '*buff'.
 // Returns true if file was successfully opened and read.
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc b/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
index 2f2787e..7867fcc 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_common_interceptors.inc
@@ -239,7 +239,7 @@ extern const short *_tolower_tab_;
     COMMON_INTERCEPT_FUNCTION(fn)
 #endif
 
-#ifdef __GLIBC__
+#if SANITIZER_GLIBC
 // If we could not find the versioned symbol, fall back to an unversioned
 // lookup. This is needed to work around a GLibc bug that causes dlsym
 // with RTLD_NEXT to return the oldest versioned symbol.
@@ -2195,6 +2195,7 @@ INTERCEPTOR(int, clock_gettime, u32 clk_id, void *tp) {
   }
   return res;
 }
+#if SANITIZER_GLIBC
 namespace __sanitizer {
 extern "C" {
 int real_clock_gettime(u32 clk_id, void *tp) {
@@ -2204,6 +2205,7 @@ int real_clock_gettime(u32 clk_id, void *tp) {
 }
 }  // extern "C"
 }  // namespace __sanitizer
+#endif
 INTERCEPTOR(int, clock_settime, u32 clk_id, const void *tp) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, clock_settime, clk_id, tp);
@@ -3355,7 +3357,7 @@ INTERCEPTOR(char *, setlocale, int category, char *locale) {
     COMMON_INTERCEPTOR_READ_RANGE(ctx, locale, REAL(strlen)(locale) + 1);
   char *res = REAL(setlocale)(category, locale);
   if (res) {
-    COMMON_INTERCEPTOR_WRITE_RANGE(ctx, res, REAL(strlen)(res) + 1);
+    COMMON_INTERCEPTOR_INITIALIZE_RANGE(res, REAL(strlen)(res) + 1);
     unpoison_ctype_arrays(ctx);
   }
   return res;
@@ -4030,7 +4032,7 @@ INTERCEPTOR(int, sigwait, __sanitizer_sigset_t *set, int *sig) {
   // FIXME: under ASan the call below may write to freed memory and corrupt
   // its metadata. See
   // https://github.com/google/sanitizers/issues/321.
-  int res = REAL(sigwait)(set, sig);
+  int res = COMMON_INTERCEPTOR_BLOCK_REAL(sigwait)(set, sig);
   if (!res && sig) COMMON_INTERCEPTOR_WRITE_RANGE(ctx, sig, sizeof(*sig));
   return res;
 }
@@ -4047,7 +4049,7 @@ INTERCEPTOR(int, sigwaitinfo, __sanitizer_sigset_t *set, void *info) {
   // FIXME: under ASan the call below may write to freed memory and corrupt
   // its metadata. See
   // https://github.com/google/sanitizers/issues/321.
-  int res = REAL(sigwaitinfo)(set, info);
+  int res = COMMON_INTERCEPTOR_BLOCK_REAL(sigwaitinfo)(set, info);
   if (res > 0 && info) COMMON_INTERCEPTOR_WRITE_RANGE(ctx, info, siginfo_t_sz);
   return res;
 }
@@ -4066,7 +4068,7 @@ INTERCEPTOR(int, sigtimedwait, __sanitizer_sigset_t *set, void *info,
   // FIXME: under ASan the call below may write to freed memory and corrupt
   // its metadata. See
   // https://github.com/google/sanitizers/issues/321.
-  int res = REAL(sigtimedwait)(set, info, timeout);
+  int res = COMMON_INTERCEPTOR_BLOCK_REAL(sigtimedwait)(set, info, timeout);
   if (res > 0 && info) COMMON_INTERCEPTOR_WRITE_RANGE(ctx, info, siginfo_t_sz);
   return res;
 }
@@ -5995,6 +5997,9 @@ void unpoison_file(__sanitizer_FILE *fp) {
   if (fp->_IO_read_base && fp->_IO_read_base < fp->_IO_read_end)
     COMMON_INTERCEPTOR_INITIALIZE_RANGE(fp->_IO_read_base,
                                         fp->_IO_read_end - fp->_IO_read_base);
+  if (fp->_IO_write_base && fp->_IO_write_base < fp->_IO_write_end)
+    COMMON_INTERCEPTOR_INITIALIZE_RANGE(fp->_IO_write_base,
+                                        fp->_IO_write_end - fp->_IO_write_base);
 #endif
 #endif  // SANITIZER_HAS_STRUCT_FILE
 }
@@ -6221,6 +6226,8 @@ INTERCEPTOR(void, _obstack_newchunk, __sanitizer_obstack *obstack, int length) {
 INTERCEPTOR(int, fflush, __sanitizer_FILE *fp) {
   void *ctx;
   COMMON_INTERCEPTOR_ENTER(ctx, fflush, fp);
+  if (fp)
+    unpoison_file(fp);
   int res = REAL(fflush)(fp);
   // FIXME: handle fp == NULL
   if (fp) {
@@ -6240,6 +6247,8 @@ INTERCEPTOR(int, fclose, __sanitizer_FILE *fp) {
   COMMON_INTERCEPTOR_ENTER(ctx, fclose, fp);
   COMMON_INTERCEPTOR_FILE_CLOSE(ctx, fp);
   const FileMetadata *m = GetInterceptorMetadata(fp);
+  if (fp)
+    unpoison_file(fp);
   int res = REAL(fclose)(fp);
   if (m) {
     COMMON_INTERCEPTOR_INITIALIZE_RANGE(*m->addr, *m->size);
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_interceptors_ioctl.inc b/libsanitizer/sanitizer_common/sanitizer_common_interceptors_ioctl.inc
index 490a04b..7f18125 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_interceptors_ioctl.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_common_interceptors_ioctl.inc
@@ -330,13 +330,17 @@ static void ioctl_table_fill() {
   _(SOUND_PCM_WRITE_CHANNELS, WRITE, sizeof(int));
   _(SOUND_PCM_WRITE_FILTER, WRITE, sizeof(int));
   _(TCFLSH, NONE, 0);
+#if SANITIZER_GLIBC
   _(TCGETA, WRITE, struct_termio_sz);
+#endif
   _(TCGETS, WRITE, struct_termios_sz);
   _(TCSBRK, NONE, 0);
   _(TCSBRKP, NONE, 0);
+#if SANITIZER_GLIBC
   _(TCSETA, READ, struct_termio_sz);
   _(TCSETAF, READ, struct_termio_sz);
   _(TCSETAW, READ, struct_termio_sz);
+#endif
   _(TCSETS, READ, struct_termios_sz);
   _(TCSETSF, READ, struct_termios_sz);
   _(TCSETSW, READ, struct_termios_sz);
@@ -364,7 +368,7 @@ static void ioctl_table_fill() {
   _(VT_WAITACTIVE, NONE, 0);
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
   // _(SIOCDEVPLIP, WRITE, struct_ifreq_sz); // the same as EQL_ENSLAVE
   _(CYGETDEFTHRESH, WRITE, sizeof(int));
   _(CYGETDEFTIMEOUT, WRITE, sizeof(int));
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_interceptors_vfork_aarch64.inc.S b/libsanitizer/sanitizer_common/sanitizer_common_interceptors_vfork_aarch64.inc.S
index 20f42f1..72e4827 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_interceptors_vfork_aarch64.inc.S
+++ b/libsanitizer/sanitizer_common/sanitizer_common_interceptors_vfork_aarch64.inc.S
@@ -1,6 +1,7 @@
 #if defined(__aarch64__) && defined(__linux__)
 
 #include "sanitizer_common/sanitizer_asm.h"
+#include "builtins/assembly.h"
 
 ASM_HIDDEN(COMMON_INTERCEPTOR_SPILL_AREA)
 
@@ -9,6 +10,7 @@ ASM_HIDDEN(COMMON_INTERCEPTOR_SPILL_AREA)
 ASM_TYPE_FUNCTION(ASM_WRAPPER_NAME(vfork))
 ASM_WRAPPER_NAME(vfork):
         // Save x30 in the off-stack spill area.
+        hint    #25 // paciasp
         stp     xzr, x30, [sp, #-16]!
         bl      COMMON_INTERCEPTOR_SPILL_AREA
         ldp     xzr, x30, [sp], 16
@@ -33,6 +35,7 @@ ASM_WRAPPER_NAME(vfork):
         bl     COMMON_INTERCEPTOR_SPILL_AREA
         ldr    x30, [x0]
         ldp    x0, xzr, [sp], 16
+        hint   #29 // autiasp
 
         ret
 ASM_SIZE(vfork)
@@ -40,4 +43,6 @@ ASM_SIZE(vfork)
 .weak vfork
 .set vfork, ASM_WRAPPER_NAME(vfork)
 
+GNU_PROPERTY_BTI_PAC
+
 #endif
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_interface.inc b/libsanitizer/sanitizer_common/sanitizer_common_interface.inc
index c78b6e1..932e547 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_interface.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_common_interface.inc
@@ -13,6 +13,7 @@ INTERFACE_FUNCTION(__sanitizer_contiguous_container_find_bad_address)
 INTERFACE_FUNCTION(__sanitizer_set_death_callback)
 INTERFACE_FUNCTION(__sanitizer_set_report_path)
 INTERFACE_FUNCTION(__sanitizer_set_report_fd)
+INTERFACE_FUNCTION(__sanitizer_get_report_path)
 INTERFACE_FUNCTION(__sanitizer_verify_contiguous_container)
 INTERFACE_WEAK_FUNCTION(__sanitizer_on_print)
 INTERFACE_WEAK_FUNCTION(__sanitizer_report_error_summary)
diff --git a/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
index 047c5a1..1037938 100644
--- a/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_common_libcdep.cpp
@@ -92,14 +92,13 @@ void *BackgroundThread(void *arg) {
 #endif
 
 void WriteToSyslog(const char *msg) {
-  InternalScopedString msg_copy(kErrorMessageBufferSize);
+  InternalScopedString msg_copy;
   msg_copy.append("%s", msg);
-  char *p = msg_copy.data();
-  char *q;
+  const char *p = msg_copy.data();
 
   // Print one line at a time.
   // syslog, at least on Android, has an implicit message length limit.
-  while ((q = internal_strchr(p, '\n'))) {
+  while (char* q = internal_strchr(p, '\n')) {
     *q = '\0';
     WriteOneLineToSyslog(p);
     p = q + 1;
diff --git a/libsanitizer/sanitizer_common/sanitizer_file.cpp b/libsanitizer/sanitizer_common/sanitizer_file.cpp
index 7cce609..0b92dcc 100644
--- a/libsanitizer/sanitizer_common/sanitizer_file.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_file.cpp
@@ -58,6 +58,9 @@ void ReportFile::ReopenIfNecessary() {
   } else {
     internal_snprintf(full_path, kMaxPathLength, "%s.%zu", path_prefix, pid);
   }
+  if (common_flags()->log_suffix) {
+    internal_strlcat(full_path, common_flags()->log_suffix, kMaxPathLength);
+  }
   error_t err;
   fd = OpenFile(full_path, WrOnly, &err);
   if (fd == kInvalidFd) {
@@ -95,6 +98,12 @@ void ReportFile::SetReportPath(const char *path) {
   }
 }
 
+const char *ReportFile::GetReportPath() {
+  SpinMutexLock l(mu);
+  ReopenIfNecessary();
+  return full_path;
+}
+
 bool ReadFileToBuffer(const char *file_name, char **buff, uptr *buff_size,
                       uptr *read_len, uptr max_len, error_t *errno_p) {
   *buff = nullptr;
@@ -213,6 +222,10 @@ void __sanitizer_set_report_fd(void *fd) {
   report_file.fd = (fd_t)reinterpret_cast<uptr>(fd);
   report_file.fd_pid = internal_getpid();
 }
+
+const char *__sanitizer_get_report_path() {
+  return report_file.GetReportPath();
+}
 } // extern "C"
 
 #endif  // !SANITIZER_FUCHSIA
diff --git a/libsanitizer/sanitizer_common/sanitizer_file.h b/libsanitizer/sanitizer_common/sanitizer_file.h
index 26681f0..08671ab 100644
--- a/libsanitizer/sanitizer_common/sanitizer_file.h
+++ b/libsanitizer/sanitizer_common/sanitizer_file.h
@@ -26,6 +26,7 @@ struct ReportFile {
   void Write(const char *buffer, uptr length);
   bool SupportsColors();
   void SetReportPath(const char *path);
+  const char *GetReportPath();
 
   // Don't use fields directly. They are only declared public to allow
   // aggregate initialization.
diff --git a/libsanitizer/sanitizer_common/sanitizer_flags.cpp b/libsanitizer/sanitizer_common/sanitizer_flags.cpp
index 21048be..d52e96a 100644
--- a/libsanitizer/sanitizer_common/sanitizer_flags.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_flags.cpp
@@ -35,6 +35,7 @@ void CommonFlags::CopyFrom(const CommonFlags &other) {
 // Copy the string from "s" to "out", making the following substitutions:
 // %b = binary basename
 // %p = pid
+// %d = binary directory
 void SubstituteForFlagValue(const char *s, char *out, uptr out_size) {
   char *out_end = out + out_size;
   while (*s && out < out_end - 1) {
@@ -64,6 +65,12 @@ void SubstituteForFlagValue(const char *s, char *out, uptr out_size) {
         s += 2; // skip "%p"
         break;
       }
+      case 'd': {
+        uptr len = ReadBinaryDir(out, out_end - out);
+        out += len;
+        s += 2;  // skip "%d"
+        break;
+      }
       default:
         *out++ = *s++;
         break;
diff --git a/libsanitizer/sanitizer_common/sanitizer_flags.inc b/libsanitizer/sanitizer_common/sanitizer_flags.inc
index cfb5822..3bc44c6 100644
--- a/libsanitizer/sanitizer_common/sanitizer_flags.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_flags.inc
@@ -59,6 +59,8 @@ COMMON_FLAG(
     bool, log_exe_name, false,
     "Mention name of executable when reporting error and "
     "append executable name to logs (as in \"log_path.exe_name.pid\").")
+COMMON_FLAG(const char *, log_suffix, nullptr,
+            "String to append to log file name, e.g. \".txt\".")
 COMMON_FLAG(
     bool, log_to_syslog, (bool)SANITIZER_ANDROID || (bool)SANITIZER_MAC,
     "Write all sanitizer output to syslog in addition to other means of "
diff --git a/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp b/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
index 7200ffd..4f692f9 100644
--- a/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_fuchsia.cpp
@@ -14,7 +14,6 @@
 #include "sanitizer_fuchsia.h"
 #if SANITIZER_FUCHSIA
 
-#include <limits.h>
 #include <pthread.h>
 #include <stdlib.h>
 #include <unistd.h>
@@ -69,9 +68,7 @@ uptr internal_getpid() {
   return pid;
 }
 
-int internal_dlinfo(void *handle, int request, void *p) {
-  UNIMPLEMENTED();
-}
+int internal_dlinfo(void *handle, int request, void *p) { UNIMPLEMENTED(); }
 
 uptr GetThreadSelf() { return reinterpret_cast<uptr>(thrd_current()); }
 
@@ -153,9 +150,9 @@ void BlockingMutex::CheckLocked() {
   CHECK_NE(MtxUnlocked, atomic_load(m, memory_order_relaxed));
 }
 
-uptr GetPageSize() { return PAGE_SIZE; }
+uptr GetPageSize() { return _zx_system_get_page_size(); }
 
-uptr GetMmapGranularity() { return PAGE_SIZE; }
+uptr GetMmapGranularity() { return _zx_system_get_page_size(); }
 
 sanitizer_shadow_bounds_t ShadowBounds;
 
@@ -168,7 +165,7 @@ uptr GetMaxVirtualAddress() { return GetMaxUserVirtualAddress(); }
 
 static void *DoAnonymousMmapOrDie(uptr size, const char *mem_type,
                                   bool raw_report, bool die_for_nomem) {
-  size = RoundUpTo(size, PAGE_SIZE);
+  size = RoundUpTo(size, GetPageSize());
 
   zx_handle_t vmo;
   zx_status_t status = _zx_vmo_create(size, 0, &vmo);
@@ -214,15 +211,14 @@ void *MmapOrDieOnFatalError(uptr size, const char *mem_type) {
 
 uptr ReservedAddressRange::Init(uptr init_size, const char *name,
                                 uptr fixed_addr) {
-  init_size = RoundUpTo(init_size, PAGE_SIZE);
+  init_size = RoundUpTo(init_size, GetPageSize());
   DCHECK_EQ(os_handle_, ZX_HANDLE_INVALID);
   uintptr_t base;
   zx_handle_t vmar;
-  zx_status_t status =
-      _zx_vmar_allocate(
-          _zx_vmar_root_self(),
-          ZX_VM_CAN_MAP_READ | ZX_VM_CAN_MAP_WRITE | ZX_VM_CAN_MAP_SPECIFIC,
-          0, init_size, &vmar, &base);
+  zx_status_t status = _zx_vmar_allocate(
+      _zx_vmar_root_self(),
+      ZX_VM_CAN_MAP_READ | ZX_VM_CAN_MAP_WRITE | ZX_VM_CAN_MAP_SPECIFIC, 0,
+      init_size, &vmar, &base);
   if (status != ZX_OK)
     ReportMmapFailureAndDie(init_size, name, "zx_vmar_allocate", status);
   base_ = reinterpret_cast<void *>(base);
@@ -236,7 +232,7 @@ uptr ReservedAddressRange::Init(uptr init_size, const char *name,
 static uptr DoMmapFixedOrDie(zx_handle_t vmar, uptr fixed_addr, uptr map_size,
                              void *base, const char *name, bool die_for_nomem) {
   uptr offset = fixed_addr - reinterpret_cast<uptr>(base);
-  map_size = RoundUpTo(map_size, PAGE_SIZE);
+  map_size = RoundUpTo(map_size, GetPageSize());
   zx_handle_t vmo;
   zx_status_t status = _zx_vmo_create(map_size, 0, &vmo);
   if (status != ZX_OK) {
@@ -264,19 +260,19 @@ static uptr DoMmapFixedOrDie(zx_handle_t vmar, uptr fixed_addr, uptr map_size,
 
 uptr ReservedAddressRange::Map(uptr fixed_addr, uptr map_size,
                                const char *name) {
-  return DoMmapFixedOrDie(os_handle_, fixed_addr, map_size, base_,
-                          name_, false);
+  return DoMmapFixedOrDie(os_handle_, fixed_addr, map_size, base_, name_,
+                          false);
 }
 
 uptr ReservedAddressRange::MapOrDie(uptr fixed_addr, uptr map_size,
                                     const char *name) {
-  return DoMmapFixedOrDie(os_handle_, fixed_addr, map_size, base_,
-                          name_, true);
+  return DoMmapFixedOrDie(os_handle_, fixed_addr, map_size, base_, name_, true);
 }
 
 void UnmapOrDieVmar(void *addr, uptr size, zx_handle_t target_vmar) {
-  if (!addr || !size) return;
-  size = RoundUpTo(size, PAGE_SIZE);
+  if (!addr || !size)
+    return;
+  size = RoundUpTo(size, GetPageSize());
 
   zx_status_t status =
       _zx_vmar_unmap(target_vmar, reinterpret_cast<uintptr_t>(addr), size);
@@ -316,7 +312,7 @@ void *MmapFixedNoAccess(uptr fixed_addr, uptr size, const char *name) {
 
 void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
                                    const char *mem_type) {
-  CHECK_GE(size, PAGE_SIZE);
+  CHECK_GE(size, GetPageSize());
   CHECK(IsPowerOfTwo(size));
   CHECK(IsPowerOfTwo(alignment));
 
@@ -356,7 +352,8 @@ void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
             _zx_vmar_root_self(),
             ZX_VM_PERM_READ | ZX_VM_PERM_WRITE | ZX_VM_SPECIFIC_OVERWRITE,
             addr - info.base, vmo, 0, size, &new_addr);
-        if (status == ZX_OK) CHECK_EQ(new_addr, addr);
+        if (status == ZX_OK)
+          CHECK_EQ(new_addr, addr);
       }
     }
     if (status == ZX_OK && addr != map_addr)
@@ -381,9 +378,18 @@ void UnmapOrDie(void *addr, uptr size) {
   UnmapOrDieVmar(addr, size, _zx_vmar_root_self());
 }
 
-// This is used on the shadow mapping, which cannot be changed.
-// Zircon doesn't have anything like MADV_DONTNEED.
-void ReleaseMemoryPagesToOS(uptr beg, uptr end) {}
+void ReleaseMemoryPagesToOS(uptr beg, uptr end) {
+  uptr beg_aligned = RoundUpTo(beg, GetPageSize());
+  uptr end_aligned = RoundDownTo(end, GetPageSize());
+  if (beg_aligned < end_aligned) {
+    zx_handle_t root_vmar = _zx_vmar_root_self();
+    CHECK_NE(root_vmar, ZX_HANDLE_INVALID);
+    zx_status_t status =
+        _zx_vmar_op_range(root_vmar, ZX_VMAR_OP_DECOMMIT, beg_aligned,
+                          end_aligned - beg_aligned, nullptr, 0);
+    CHECK_EQ(status, ZX_OK);
+  }
+}
 
 void DumpProcessMap() {
   // TODO(mcgrathr): write it
@@ -412,8 +418,9 @@ bool ReadFileToBuffer(const char *file_name, char **buff, uptr *buff_size,
     uint64_t vmo_size;
     status = _zx_vmo_get_size(vmo, &vmo_size);
     if (status == ZX_OK) {
-      if (vmo_size < max_len) max_len = vmo_size;
-      size_t map_size = RoundUpTo(max_len, PAGE_SIZE);
+      if (vmo_size < max_len)
+        max_len = vmo_size;
+      size_t map_size = RoundUpTo(max_len, GetPageSize());
       uintptr_t addr;
       status = _zx_vmar_map(_zx_vmar_root_self(), ZX_VM_PERM_READ, 0, vmo, 0,
                             map_size, &addr);
@@ -425,7 +432,8 @@ bool ReadFileToBuffer(const char *file_name, char **buff, uptr *buff_size,
     }
     _zx_handle_close(vmo);
   }
-  if (status != ZX_OK && errno_p) *errno_p = status;
+  if (status != ZX_OK && errno_p)
+    *errno_p = status;
   return status == ZX_OK;
 }
 
@@ -499,9 +507,7 @@ bool GetRandom(void *buffer, uptr length, bool blocking) {
   return true;
 }
 
-u32 GetNumberOfCPUs() {
-  return zx_system_get_num_cpus();
-}
+u32 GetNumberOfCPUs() { return zx_system_get_num_cpus(); }
 
 uptr GetRSS() { UNIMPLEMENTED(); }
 
@@ -529,6 +535,10 @@ void __sanitizer_set_report_path(const char *path) {
 void __sanitizer_set_report_fd(void *fd) {
   UNREACHABLE("not available on Fuchsia");
 }
+
+const char *__sanitizer_get_report_path() {
+  UNREACHABLE("not available on Fuchsia");
+}
 }  // extern "C"
 
 #endif  // SANITIZER_FUCHSIA
diff --git a/libsanitizer/sanitizer_common/sanitizer_interface_internal.h b/libsanitizer/sanitizer_common/sanitizer_interface_internal.h
index be8023e..0b001c1 100644
--- a/libsanitizer/sanitizer_common/sanitizer_interface_internal.h
+++ b/libsanitizer/sanitizer_common/sanitizer_interface_internal.h
@@ -28,6 +28,10 @@ extern "C" {
   // (casted to void *).
   SANITIZER_INTERFACE_ATTRIBUTE
   void __sanitizer_set_report_fd(void *fd);
+  // Get the current full report file path, if a path was specified by
+  // an earlier call to __sanitizer_set_report_path. Returns null otherwise.
+  SANITIZER_INTERFACE_ATTRIBUTE
+  const char *__sanitizer_get_report_path();
 
   typedef struct {
       int coverage_sandboxed;
diff --git a/libsanitizer/sanitizer_common/sanitizer_internal_defs.h b/libsanitizer/sanitizer_common/sanitizer_internal_defs.h
index d8f0540..84053fe 100644
--- a/libsanitizer/sanitizer_common/sanitizer_internal_defs.h
+++ b/libsanitizer/sanitizer_common/sanitizer_internal_defs.h
@@ -409,6 +409,9 @@ inline void Trap() {
     (void)enable_fp;                      \
   } while (0)
 
+constexpr u32 kInvalidTid = -1;
+constexpr u32 kMainTid = 0;
+
 }  // namespace __sanitizer
 
 namespace __asan {
diff --git a/libsanitizer/sanitizer_common/sanitizer_libignore.cpp b/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
index 9ea19bc..a65d3d8 100644
--- a/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_libignore.cpp
@@ -38,7 +38,7 @@ void LibIgnore::AddIgnoredLibrary(const char *name_templ) {
 void LibIgnore::OnLibraryLoaded(const char *name) {
   BlockingMutexLock lock(&mutex_);
   // Try to match suppressions with symlink target.
-  InternalScopedString buf(kMaxPathLength);
+  InternalMmapVector<char> buf(kMaxPathLength);
   if (name && internal_readlink(name, buf.data(), buf.size() - 1) > 0 &&
       buf[0]) {
     for (uptr i = 0; i < count_; i++) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_linux.cpp b/libsanitizer/sanitizer_common/sanitizer_linux.cpp
index 379f6d9..b371477 100644
--- a/libsanitizer/sanitizer_common/sanitizer_linux.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_linux.cpp
@@ -183,6 +183,14 @@ uptr internal_munmap(void *addr, uptr length) {
   return internal_syscall(SYSCALL(munmap), (uptr)addr, length);
 }
 
+#if SANITIZER_LINUX
+uptr internal_mremap(void *old_address, uptr old_size, uptr new_size, int flags,
+                     void *new_address) {
+  return internal_syscall(SYSCALL(mremap), (uptr)old_address, old_size,
+                          new_size, flags, (uptr)new_address);
+}
+#endif
+
 int internal_mprotect(void *addr, uptr length, int prot) {
   return internal_syscall(SYSCALL(mprotect), (uptr)addr, length, prot);
 }
@@ -489,22 +497,24 @@ int TgKill(pid_t pid, tid_t tid, int sig) {
 }
 #endif
 
-#if !SANITIZER_SOLARIS && !SANITIZER_NETBSD
+#if SANITIZER_GLIBC
 u64 NanoTime() {
-#if SANITIZER_FREEBSD
-  timeval tv;
-#else
   kernel_timeval tv;
-#endif
   internal_memset(&tv, 0, sizeof(tv));
   internal_syscall(SYSCALL(gettimeofday), &tv, 0);
-  return (u64)tv.tv_sec * 1000*1000*1000 + tv.tv_usec * 1000;
+  return (u64)tv.tv_sec * 1000 * 1000 * 1000 + tv.tv_usec * 1000;
 }
-
+// Used by real_clock_gettime.
 uptr internal_clock_gettime(__sanitizer_clockid_t clk_id, void *tp) {
   return internal_syscall(SYSCALL(clock_gettime), clk_id, tp);
 }
-#endif  // !SANITIZER_SOLARIS && !SANITIZER_NETBSD
+#elif !SANITIZER_SOLARIS && !SANITIZER_NETBSD
+u64 NanoTime() {
+  struct timespec ts;
+  clock_gettime(CLOCK_REALTIME, &ts);
+  return (u64)ts.tv_sec * 1000 * 1000 * 1000 + ts.tv_nsec;
+}
+#endif
 
 // Like getenv, but reads env directly from /proc (on Linux) or parses the
 // 'environ' array (on some others) and does not use libc. This function
@@ -1334,50 +1344,42 @@ uptr internal_clone(int (*fn)(void *), void *child_stack, int flags, void *arg,
 #elif SANITIZER_RISCV64
 uptr internal_clone(int (*fn)(void *), void *child_stack, int flags, void *arg,
                     int *parent_tidptr, void *newtls, int *child_tidptr) {
-  long long res;
   if (!fn || !child_stack)
     return -EINVAL;
+
   CHECK_EQ(0, (uptr)child_stack % 16);
-  child_stack = (char *)child_stack - 2 * sizeof(unsigned long long);
-  ((unsigned long long *)child_stack)[0] = (uptr)fn;
-  ((unsigned long long *)child_stack)[1] = (uptr)arg;
 
-  register int (*__fn)(void *) __asm__("a0") = fn;
+  register int res __asm__("a0");
+  register int __flags __asm__("a0") = flags;
   register void *__stack __asm__("a1") = child_stack;
-  register int __flags __asm__("a2") = flags;
-  register void *__arg __asm__("a3") = arg;
-  register int *__ptid __asm__("a4") = parent_tidptr;
-  register void *__tls __asm__("a5") = newtls;
-  register int *__ctid __asm__("a6") = child_tidptr;
+  register int *__ptid __asm__("a2") = parent_tidptr;
+  register void *__tls __asm__("a3") = newtls;
+  register int *__ctid __asm__("a4") = child_tidptr;
+  register int (*__fn)(void *) __asm__("a5") = fn;
+  register void *__arg __asm__("a6") = arg;
+  register int nr_clone __asm__("a7") = __NR_clone;
 
   __asm__ __volatile__(
-      "mv a0,a2\n"          /* flags  */
-      "mv a2,a4\n"          /* ptid  */
-      "mv a3,a5\n"          /* tls  */
-      "mv a4,a6\n"          /* ctid  */
-      "addi a7, zero, %9\n" /* clone  */
-
       "ecall\n"
 
-      /* if (%r0 != 0)
-       *   return %r0;
+      /* if (a0 != 0)
+       *   return a0;
        */
       "bnez a0, 1f\n"
 
-      /* In the child, now. Call "fn(arg)". */
-      "ld  a0,  8(sp)\n"
-      "ld  a1, 16(sp)\n"
-      "jalr a1\n"
+      // In the child, now. Call "fn(arg)".
+      "mv a0, a6\n"
+      "jalr a5\n"
 
-      /* Call _exit(%r0).  */
-      "addi  a7, zero, %10\n"
+      // Call _exit(a0).
+      "addi a7, zero, %9\n"
       "ecall\n"
       "1:\n"
 
       : "=r"(res)
-      : "i"(-EINVAL), "r"(__fn), "r"(__stack), "r"(__flags), "r"(__arg),
-        "r"(__ptid), "r"(__tls), "r"(__ctid), "i"(__NR_clone), "i"(__NR_exit)
-      : "ra", "memory");
+      : "0"(__flags), "r"(__stack), "r"(__ptid), "r"(__tls), "r"(__ctid),
+        "r"(__fn), "r"(__arg), "r"(nr_clone), "i"(__NR_exit)
+      : "memory");
   return res;
 }
 #elif defined(__aarch64__)
diff --git a/libsanitizer/sanitizer_common/sanitizer_linux.h b/libsanitizer/sanitizer_common/sanitizer_linux.h
index 24902d1..9a23fcf 100644
--- a/libsanitizer/sanitizer_common/sanitizer_linux.h
+++ b/libsanitizer/sanitizer_common/sanitizer_linux.h
@@ -49,7 +49,9 @@ uptr internal_getdents(fd_t fd, struct linux_dirent *dirp, unsigned int count);
 uptr internal_sigaltstack(const void* ss, void* oss);
 uptr internal_sigprocmask(int how, __sanitizer_sigset_t *set,
     __sanitizer_sigset_t *oldset);
+#if SANITIZER_GLIBC
 uptr internal_clock_gettime(__sanitizer_clockid_t clk_id, void *tp);
+#endif
 
 // Linux-only syscalls.
 #if SANITIZER_LINUX
@@ -96,7 +98,6 @@ class ThreadLister {
 // Exposed for testing.
 uptr ThreadDescriptorSize();
 uptr ThreadSelf();
-uptr ThreadSelfOffset();
 
 // Matches a library's file name against a base name (stripping path and version
 // information).
diff --git a/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
index bc10e89..572aa86 100644
--- a/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_linux_libcdep.cpp
@@ -36,6 +36,7 @@
 #include <link.h>
 #include <pthread.h>
 #include <signal.h>
+#include <sys/mman.h>
 #include <sys/resource.h>
 #include <syslog.h>
 
@@ -48,6 +49,10 @@
 #include <osreldate.h>
 #include <sys/sysctl.h>
 #define pthread_getattr_np pthread_attr_get_np
+// The MAP_NORESERVE define has been removed in FreeBSD 11.x, and even before
+// that, it was never implemented. So just define it to zero.
+#undef MAP_NORESERVE
+#define MAP_NORESERVE 0
 #endif
 
 #if SANITIZER_NETBSD
@@ -183,85 +188,35 @@ __attribute__((unused)) static bool GetLibcVersion(int *major, int *minor,
 #endif
 }
 
-#if !SANITIZER_FREEBSD && !SANITIZER_ANDROID && !SANITIZER_GO && \
-    !SANITIZER_NETBSD && !SANITIZER_SOLARIS
-static uptr g_tls_size;
+// True if we can use dlpi_tls_data. glibc before 2.25 may leave NULL (BZ
+// #19826) so dlpi_tls_data cannot be used.
+//
+// musl before 1.2.3 and FreeBSD as of 12.2 incorrectly set dlpi_tls_data to
+// the TLS initialization image
+// https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=254774
+__attribute__((unused)) static int g_use_dlpi_tls_data;
 
-#ifdef __i386__
-#define CHECK_GET_TLS_STATIC_INFO_VERSION (!__GLIBC_PREREQ(2, 27))
-#else
-#define CHECK_GET_TLS_STATIC_INFO_VERSION 0
-#endif
+#if SANITIZER_GLIBC && !SANITIZER_GO
+__attribute__((unused)) static uptr g_tls_size;
+void InitTlsSize() {
+  int major, minor, patch;
+  g_use_dlpi_tls_data =
+      GetLibcVersion(&major, &minor, &patch) && major == 2 && minor >= 25;
 
-#if CHECK_GET_TLS_STATIC_INFO_VERSION
-#define DL_INTERNAL_FUNCTION __attribute__((regparm(3), stdcall))
-#else
-#define DL_INTERNAL_FUNCTION
+#if 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);
 #endif
-
-namespace {
-struct GetTlsStaticInfoCall {
-  typedef void (*get_tls_func)(size_t*, size_t*);
-};
-struct GetTlsStaticInfoRegparmCall {
-  typedef void (*get_tls_func)(size_t*, size_t*) DL_INTERNAL_FUNCTION;
-};
-
-template <typename T>
-void CallGetTls(void* ptr, size_t* size, size_t* align) {
-  typename T::get_tls_func get_tls;
-  CHECK_EQ(sizeof(get_tls), sizeof(ptr));
-  internal_memcpy(&get_tls, &ptr, sizeof(ptr));
-  CHECK_NE(get_tls, 0);
-  get_tls(size, align);
-}
-
-bool CmpLibcVersion(int major, int minor, int patch) {
-  int ma;
-  int mi;
-  int pa;
-  if (!GetLibcVersion(&ma, &mi, &pa))
-    return false;
-  if (ma > major)
-    return true;
-  if (ma < major)
-    return false;
-  if (mi > minor)
-    return true;
-  if (mi < minor)
-    return false;
-  return pa >= patch;
-}
-
-}  // namespace
-
-void InitTlsSize() {
-  // all current supported platforms have 16 bytes stack alignment
-  const size_t kStackAlign = 16;
-  void *get_tls_static_info_ptr = dlsym(RTLD_NEXT, "_dl_get_tls_static_info");
-  size_t tls_size = 0;
-  size_t tls_align = 0;
-  // On i?86, _dl_get_tls_static_info used to be internal_function, i.e.
-  // __attribute__((regparm(3), stdcall)) before glibc 2.27 and is normal
-  // function in 2.27 and later.
-  if (CHECK_GET_TLS_STATIC_INFO_VERSION && !CmpLibcVersion(2, 27, 0))
-    CallGetTls<GetTlsStaticInfoRegparmCall>(get_tls_static_info_ptr,
-                                            &tls_size, &tls_align);
-  else
-    CallGetTls<GetTlsStaticInfoCall>(get_tls_static_info_ptr,
-                                     &tls_size, &tls_align);
-  if (tls_align < kStackAlign)
-    tls_align = kStackAlign;
-  g_tls_size = RoundUpTo(tls_size, tls_align);
 }
 #else
 void InitTlsSize() { }
-#endif
+#endif  // SANITIZER_GLIBC && !SANITIZER_GO
 
-#if (defined(__x86_64__) || defined(__i386__) || defined(__mips__) ||       \
-     defined(__aarch64__) || defined(__powerpc64__) || defined(__s390__) || \
-     defined(__arm__) || SANITIZER_RISCV64) &&                              \
-    SANITIZER_LINUX && !SANITIZER_ANDROID
+// On glibc x86_64, ThreadDescriptorSize() needs to be precise due to the usage
+// of g_tls_size. On other targets, ThreadDescriptorSize() is only used by lsan
+// to get the pointer to thread-specific data keys in the thread control block.
+#if (SANITIZER_FREEBSD || SANITIZER_LINUX) && !SANITIZER_ANDROID
 // sizeof(struct pthread) from glibc.
 static atomic_uintptr_t thread_descriptor_size;
 
@@ -294,9 +249,18 @@ uptr ThreadDescriptorSize() {
       val = FIRST_32_SECOND_64(1168, 2288);
     else if (minor <= 14)
       val = FIRST_32_SECOND_64(1168, 2304);
-    else
+    else if (minor < 32)  // Unknown version
       val = FIRST_32_SECOND_64(1216, 2304);
+    else  // minor == 32
+      val = FIRST_32_SECOND_64(1344, 2496);
   }
+#elif defined(__s390__) || defined(__sparc__)
+  // The size of a prefix of TCB including pthread::{specific_1stblock,specific}
+  // suffices. Just return offsetof(struct pthread, specific_used), which hasn't
+  // changed since 2007-05. Technically this applies to i386/x86_64 as well but
+  // we call _dl_get_tls_static_info and need the precise size of struct
+  // pthread.
+  return FIRST_32_SECOND_64(524, 1552);
 #elif defined(__mips__)
   // TODO(sagarthakur): add more values as per different glibc versions.
   val = FIRST_32_SECOND_64(1152, 1776);
@@ -320,21 +284,12 @@ uptr ThreadDescriptorSize() {
   val = 1776;
 #elif defined(__powerpc64__)
   val = 1776; // from glibc.ppc64le 2.20-8.fc21
-#elif defined(__s390__)
-  val = FIRST_32_SECOND_64(1152, 1776); // valid for glibc 2.22
 #endif
   if (val)
     atomic_store_relaxed(&thread_descriptor_size, val);
   return val;
 }
 
-// The offset at which pointer to self is located in the thread descriptor.
-const uptr kThreadSelfOffset = FIRST_32_SECOND_64(8, 16);
-
-uptr ThreadSelfOffset() {
-  return kThreadSelfOffset;
-}
-
 #if defined(__mips__) || defined(__powerpc64__) || SANITIZER_RISCV64
 // TlsPreTcbSize includes size of struct pthread_descr and size of tcb
 // head structure. It lies before the static tls blocks.
@@ -353,68 +308,74 @@ static uptr TlsPreTcbSize() {
 }
 #endif
 
-uptr ThreadSelf() {
-  uptr descr_addr;
-#if defined(__i386__)
-  asm("mov %%gs:%c1,%0" : "=r"(descr_addr) : "i"(kThreadSelfOffset));
-#elif defined(__x86_64__)
-  asm("mov %%fs:%c1,%0" : "=r"(descr_addr) : "i"(kThreadSelfOffset));
-#elif defined(__mips__)
-  // MIPS uses TLS variant I. The thread pointer (in hardware register $29)
-  // points to the end of the TCB + 0x7000. The pthread_descr structure is
-  // immediately in front of the TCB. TlsPreTcbSize() includes the size of the
-  // TCB and the size of pthread_descr.
-  const uptr kTlsTcbOffset = 0x7000;
-  uptr thread_pointer;
-  asm volatile(".set push;\
-                .set mips64r2;\
-                rdhwr %0,$29;\
-                .set pop" : "=r" (thread_pointer));
-  descr_addr = thread_pointer - kTlsTcbOffset - TlsPreTcbSize();
-#elif defined(__aarch64__) || defined(__arm__)
-  descr_addr = reinterpret_cast<uptr>(__builtin_thread_pointer()) -
-                                      ThreadDescriptorSize();
-#elif SANITIZER_RISCV64
-  // https://github.com/riscv/riscv-elf-psabi-doc/issues/53
-  uptr thread_pointer = reinterpret_cast<uptr>(__builtin_thread_pointer());
-  descr_addr = thread_pointer - TlsPreTcbSize();
-#elif defined(__s390__)
-  descr_addr = reinterpret_cast<uptr>(__builtin_thread_pointer());
-#elif defined(__powerpc64__)
-  // PPC64LE uses TLS variant I. The thread pointer (in GPR 13)
-  // points to the end of the TCB + 0x7000. The pthread_descr structure is
-  // immediately in front of the TCB. TlsPreTcbSize() includes the size of the
-  // TCB and the size of pthread_descr.
-  const uptr kTlsTcbOffset = 0x7000;
-  uptr thread_pointer;
-  asm("addi %0,13,%1" : "=r"(thread_pointer) : "I"(-kTlsTcbOffset));
-  descr_addr = thread_pointer - TlsPreTcbSize();
-#else
-#error "unsupported CPU arch"
-#endif
-  return descr_addr;
-}
-#endif  // (x86_64 || i386 || MIPS) && SANITIZER_LINUX
+#if !SANITIZER_GO
+namespace {
+struct TlsBlock {
+  uptr begin, end, align;
+  size_t tls_modid;
+  bool operator<(const TlsBlock &rhs) const { return begin < rhs.begin; }
+};
+}  // namespace
 
-#if SANITIZER_FREEBSD
-static void **ThreadSelfSegbase() {
-  void **segbase = 0;
-#if defined(__i386__)
-  // sysarch(I386_GET_GSBASE, segbase);
-  __asm __volatile("mov %%gs:0, %0" : "=r" (segbase));
-#elif defined(__x86_64__)
-  // sysarch(AMD64_GET_FSBASE, segbase);
-  __asm __volatile("movq %%fs:0, %0" : "=r" (segbase));
-#else
-#error "unsupported CPU arch"
+extern "C" void *__tls_get_addr(size_t *);
+
+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.
+    size_t mod_and_off[2] = {info->dlpi_tls_modid, 0};
+    begin = (uptr)__tls_get_addr(mod_and_off);
+  }
 #endif
-  return segbase;
+  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(
+          TlsBlock{begin, begin + info->dlpi_phdr[i].p_memsz,
+                   info->dlpi_phdr[i].p_align, info->dlpi_tls_modid});
+      break;
+    }
+  return 0;
 }
 
-uptr ThreadSelf() {
-  return (uptr)ThreadSelfSegbase()[2];
-}
-#endif  // SANITIZER_FREEBSD
+__attribute__((unused)) static void GetStaticTlsBoundary(uptr *addr, uptr *size,
+                                                         uptr *align) {
+  InternalMmapVector<TlsBlock> ranges;
+  dl_iterate_phdr(CollectStaticTlsBlocks, &ranges);
+  uptr len = ranges.size();
+  Sort(ranges.begin(), len);
+  // Find the range with tls_modid=1. For glibc, because libc.so uses PT_TLS,
+  // this module is guaranteed to exist and is one of the initially loaded
+  // modules.
+  uptr one = 0;
+  while (one != len && ranges[one].tls_modid != 1) ++one;
+  if (one == len) {
+    // This may happen with musl if no module uses PT_TLS.
+    *addr = 0;
+    *size = 0;
+    *align = 1;
+    return;
+  }
+  // Find the maximum consecutive ranges. We consider two modules consecutive if
+  // the gap is smaller than the alignment. The dynamic loader places static TLS
+  // blocks this way not to waste space.
+  uptr l = one;
+  *align = ranges[l].align;
+  while (l != 0 && ranges[l].begin < ranges[l - 1].end + ranges[l - 1].align)
+    *align = Max(*align, ranges[--l].align);
+  uptr r = one + 1;
+  while (r != len && ranges[r].begin < ranges[r - 1].end + ranges[r - 1].align)
+    *align = Max(*align, ranges[r++].align);
+  *addr = ranges[l].begin;
+  *size = ranges[r - 1].end - ranges[l].begin;
+}
+#endif  // !SANITIZER_GO
+#endif  // (x86_64 || i386 || mips || ...) && (SANITIZER_FREEBSD ||
+        // SANITIZER_LINUX) && !SANITIZER_ANDROID
 
 #if SANITIZER_NETBSD
 static struct tls_tcb * ThreadSelfTlsTcb() {
@@ -465,33 +426,67 @@ static void GetTls(uptr *addr, uptr *size) {
     *addr = 0;
     *size = 0;
   }
-#elif SANITIZER_LINUX
-#if defined(__x86_64__) || defined(__i386__) || defined(__s390__)
-  *addr = ThreadSelf();
-  *size = GetTlsSize();
+#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.
+  asm("mov %%fs:16,%0" : "=r"(*addr));
+  *size = g_tls_size;
   *addr -= *size;
   *addr += ThreadDescriptorSize();
-#elif defined(__mips__) || defined(__aarch64__) || defined(__powerpc64__) || \
-    defined(__arm__) || SANITIZER_RISCV64
-  *addr = ThreadSelf();
-  *size = GetTlsSize();
+#elif SANITIZER_GLIBC && defined(__powerpc64__)
+  // Workaround for glibc<2.25(?). 2.27 is known to not need this.
+  uptr tp;
+  asm("addi %0,13,-0x7000" : "=r"(tp));
+  const uptr pre_tcb_size = TlsPreTcbSize();
+  *addr = tp - pre_tcb_size;
+  *size = g_tls_size + pre_tcb_size;
+#elif SANITIZER_FREEBSD || SANITIZER_LINUX
+  uptr align;
+  GetStaticTlsBoundary(addr, size, &align);
+#if defined(__x86_64__) || defined(__i386__) || defined(__s390__) || \
+    defined(__sparc__)
+  if (SANITIZER_GLIBC) {
+#if defined(__x86_64__) || defined(__i386__)
+    align = Max<uptr>(align, 64);
 #else
-  *addr = 0;
-  *size = 0;
+    align = Max<uptr>(align, 16);
 #endif
-#elif SANITIZER_FREEBSD
-  void** segbase = ThreadSelfSegbase();
-  *addr = 0;
-  *size = 0;
-  if (segbase != 0) {
-    // tcbalign = 16
-    // tls_size = round(tls_static_space, tcbalign);
-    // dtv = segbase[1];
-    // dtv[2] = segbase - tls_static_space;
-    void **dtv = (void**) segbase[1];
-    *addr = (uptr) dtv[2];
-    *size = (*addr == 0) ? 0 : ((uptr) segbase[0] - (uptr) dtv[2]);
   }
+  const uptr tp = RoundUpTo(*addr + *size, align);
+
+  // lsan requires the range to additionally cover the static TLS surplus
+  // (elf/dl-tls.c defines 1664). Otherwise there may be false positives for
+  // allocations only referenced by tls in dynamically loaded modules.
+  if (SANITIZER_GLIBC)
+    *size += 1644;
+  else if (SANITIZER_FREEBSD)
+    *size += 128;  // RTLD_STATIC_TLS_EXTRA
+
+  // Extend the range to include the thread control block. On glibc, lsan needs
+  // the range to include pthread::{specific_1stblock,specific} so that
+  // allocations only referenced by pthread_setspecific can be scanned. This may
+  // underestimate by at most TLS_TCB_ALIGN-1 bytes but it should be fine
+  // because the number of bytes after pthread::specific is larger.
+  *addr = tp - RoundUpTo(*size, align);
+  *size = tp - *addr + ThreadDescriptorSize();
+#else
+  if (SANITIZER_GLIBC)
+    *size += 1664;
+  else if (SANITIZER_FREEBSD)
+    *size += 128;  // RTLD_STATIC_TLS_EXTRA
+#if defined(__mips__) || defined(__powerpc64__) || SANITIZER_RISCV64
+  const uptr pre_tcb_size = TlsPreTcbSize();
+  *addr -= pre_tcb_size;
+  *size += pre_tcb_size;
+#else
+  // arm and aarch64 reserve two words at TP, so this underestimates the range.
+  // However, this is sufficient for the purpose of finding the pointers to
+  // thread-specific data keys.
+  const uptr tcb_size = ThreadDescriptorSize();
+  *addr -= tcb_size;
+  *size += tcb_size;
+#endif
+#endif
 #elif SANITIZER_NETBSD
   struct tls_tcb * const tcb = ThreadSelfTlsTcb();
   *addr = 0;
@@ -518,15 +513,13 @@ static void GetTls(uptr *addr, uptr *size) {
 
 #if !SANITIZER_GO
 uptr GetTlsSize() {
-#if SANITIZER_FREEBSD || SANITIZER_ANDROID || SANITIZER_NETBSD || \
+#if SANITIZER_FREEBSD || SANITIZER_LINUX || SANITIZER_NETBSD || \
     SANITIZER_SOLARIS
   uptr addr, size;
   GetTls(&addr, &size);
   return size;
-#elif defined(__mips__) || defined(__powerpc64__) || SANITIZER_RISCV64
-  return RoundUpTo(g_tls_size + TlsPreTcbSize(), 16);
 #else
-  return g_tls_size;
+  return 0;
 #endif
 }
 #endif
@@ -547,10 +540,9 @@ void GetThreadStackAndTls(bool main, uptr *stk_addr, uptr *stk_size,
   if (!main) {
     // If stack and tls intersect, make them non-intersecting.
     if (*tls_addr > *stk_addr && *tls_addr < *stk_addr + *stk_size) {
-      CHECK_GT(*tls_addr + *tls_size, *stk_addr);
-      CHECK_LE(*tls_addr + *tls_size, *stk_addr + *stk_size);
-      *stk_size -= *tls_size;
-      *tls_addr = *stk_addr + *stk_size;
+      if (*stk_addr + *stk_size < *tls_addr + *tls_size)
+        *tls_size = *stk_addr + *stk_size - *tls_addr;
+      *stk_size = *tls_addr - *stk_addr;
     }
   }
 #endif
@@ -569,20 +561,12 @@ struct DlIteratePhdrData {
   bool first;
 };
 
-static int dl_iterate_phdr_cb(dl_phdr_info *info, size_t size, void *arg) {
-  DlIteratePhdrData *data = (DlIteratePhdrData*)arg;
-  InternalScopedString module_name(kMaxPathLength);
-  if (data->first) {
-    data->first = false;
-    // First module is the binary itself.
-    ReadBinaryNameCached(module_name.data(), module_name.size());
-  } else if (info->dlpi_name) {
-    module_name.append("%s", info->dlpi_name);
-  }
+static int AddModuleSegments(const char *module_name, dl_phdr_info *info,
+                             InternalMmapVectorNoCtor<LoadedModule> *modules) {
   if (module_name[0] == '\0')
     return 0;
   LoadedModule cur_module;
-  cur_module.set(module_name.data(), info->dlpi_addr);
+  cur_module.set(module_name, info->dlpi_addr);
   for (int i = 0; i < (int)info->dlpi_phnum; i++) {
     const Elf_Phdr *phdr = &info->dlpi_phdr[i];
     if (phdr->p_type == PT_LOAD) {
@@ -594,7 +578,26 @@ static int dl_iterate_phdr_cb(dl_phdr_info *info, size_t size, void *arg) {
                                  writable);
     }
   }
-  data->modules->push_back(cur_module);
+  modules->push_back(cur_module);
+  return 0;
+}
+
+static int dl_iterate_phdr_cb(dl_phdr_info *info, size_t size, void *arg) {
+  DlIteratePhdrData *data = (DlIteratePhdrData *)arg;
+  if (data->first) {
+    InternalMmapVector<char> module_name(kMaxPathLength);
+    data->first = false;
+    // First module is the binary itself.
+    ReadBinaryNameCached(module_name.data(), module_name.size());
+    return AddModuleSegments(module_name.data(), info, data->modules);
+  }
+
+  if (info->dlpi_name) {
+    InternalScopedString module_name;
+    module_name.append("%s", info->dlpi_name);
+    return AddModuleSegments(module_name.data(), info, data->modules);
+  }
+
   return 0;
 }
 
@@ -729,13 +732,9 @@ 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
 }
 
@@ -802,20 +801,13 @@ void LogMessageOnPrintf(const char *str) {
 
 #endif  // SANITIZER_LINUX
 
-#if SANITIZER_LINUX && !SANITIZER_GO
+#if SANITIZER_GLIBC && !SANITIZER_GO
 // glibc crashes when using clock_gettime from a preinit_array function as the
 // vDSO function pointers haven't been initialized yet. __progname is
 // initialized after the vDSO function pointers, so if it exists, is not null
 // and is not empty, we can use clock_gettime.
 extern "C" SANITIZER_WEAK_ATTRIBUTE char *__progname;
-inline bool CanUseVDSO() {
-  // Bionic is safe, it checks for the vDSO function pointers to be initialized.
-  if (SANITIZER_ANDROID)
-    return true;
-  if (&__progname && __progname && *__progname)
-    return true;
-  return false;
-}
+inline bool CanUseVDSO() { return &__progname && __progname && *__progname; }
 
 // MonotonicNanoTime is a timing function that can leverage the vDSO by calling
 // clock_gettime. real_clock_gettime only exists if clock_gettime is
@@ -835,13 +827,13 @@ u64 MonotonicNanoTime() {
   return (u64)ts.tv_sec * (1000ULL * 1000 * 1000) + ts.tv_nsec;
 }
 #else
-// Non-Linux & Go always use the syscall.
+// Non-glibc & Go always use the regular function.
 u64 MonotonicNanoTime() {
   timespec ts;
-  internal_clock_gettime(CLOCK_MONOTONIC, &ts);
+  clock_gettime(CLOCK_MONOTONIC, &ts);
   return (u64)ts.tv_sec * (1000ULL * 1000 * 1000) + ts.tv_nsec;
 }
-#endif  // SANITIZER_LINUX && !SANITIZER_GO
+#endif  // SANITIZER_GLIBC && !SANITIZER_GO
 
 void ReExec() {
   const char *pathname = "/proc/self/exe";
@@ -910,6 +902,65 @@ uptr MapDynamicShadow(uptr shadow_size_bytes, uptr shadow_scale,
   return shadow_start;
 }
 
+static uptr MmapSharedNoReserve(uptr addr, uptr size) {
+  return internal_mmap(
+      reinterpret_cast<void *>(addr), size, PROT_READ | PROT_WRITE,
+      MAP_FIXED | MAP_SHARED | MAP_ANONYMOUS | MAP_NORESERVE, -1, 0);
+}
+
+static uptr MremapCreateAlias(uptr base_addr, uptr alias_addr,
+                              uptr alias_size) {
+#if SANITIZER_LINUX
+  return internal_mremap(reinterpret_cast<void *>(base_addr), 0, alias_size,
+                         MREMAP_MAYMOVE | MREMAP_FIXED,
+                         reinterpret_cast<void *>(alias_addr));
+#else
+  CHECK(false && "mremap is not supported outside of Linux");
+  return 0;
+#endif
+}
+
+static void CreateAliases(uptr start_addr, uptr alias_size, uptr num_aliases) {
+  uptr total_size = alias_size * num_aliases;
+  uptr mapped = MmapSharedNoReserve(start_addr, total_size);
+  CHECK_EQ(mapped, start_addr);
+
+  for (uptr i = 1; i < num_aliases; ++i) {
+    uptr alias_addr = start_addr + i * alias_size;
+    CHECK_EQ(MremapCreateAlias(start_addr, alias_addr, alias_size), alias_addr);
+  }
+}
+
+uptr MapDynamicShadowAndAliases(uptr shadow_size, uptr alias_size,
+                                uptr num_aliases, uptr ring_buffer_size) {
+  CHECK_EQ(alias_size & (alias_size - 1), 0);
+  CHECK_EQ(num_aliases & (num_aliases - 1), 0);
+  CHECK_EQ(ring_buffer_size & (ring_buffer_size - 1), 0);
+
+  const uptr granularity = GetMmapGranularity();
+  shadow_size = RoundUpTo(shadow_size, granularity);
+  CHECK_EQ(shadow_size & (shadow_size - 1), 0);
+
+  const uptr alias_region_size = alias_size * num_aliases;
+  const uptr alignment =
+      2 * Max(Max(shadow_size, alias_region_size), ring_buffer_size);
+  const uptr left_padding = ring_buffer_size;
+
+  const uptr right_size = alignment;
+  const uptr map_size = left_padding + 2 * alignment;
+
+  const uptr map_start = reinterpret_cast<uptr>(MmapNoAccess(map_size));
+  CHECK_NE(map_start, static_cast<uptr>(-1));
+  const uptr right_start = RoundUpTo(map_start + left_padding, alignment);
+
+  UnmapFromTo(map_start, right_start - left_padding);
+  UnmapFromTo(right_start + right_size, map_start + map_size);
+
+  CreateAliases(right_start + right_size / 2, alias_size, num_aliases);
+
+  return right_start;
+}
+
 void InitializePlatformCommonFlags(CommonFlags *cf) {
 #if SANITIZER_ANDROID
   if (&__libc_get_static_tls_bounds == nullptr)
diff --git a/libsanitizer/sanitizer_common/sanitizer_local_address_space_view.h b/libsanitizer/sanitizer_common/sanitizer_local_address_space_view.h
index 5d1b526..0e19c4d 100644
--- a/libsanitizer/sanitizer_common/sanitizer_local_address_space_view.h
+++ b/libsanitizer/sanitizer_common/sanitizer_local_address_space_view.h
@@ -7,7 +7,7 @@
 //===----------------------------------------------------------------------===//
 //
 // `LocalAddressSpaceView` provides the local (i.e. target and current address
-// space are the same) implementation of the `AddressSpaveView` interface which
+// space are the same) implementation of the `AddressSpaceView` interface which
 // provides a simple interface to load memory from another process (i.e.
 // out-of-process)
 //
diff --git a/libsanitizer/sanitizer_common/sanitizer_mac.cpp b/libsanitizer/sanitizer_common/sanitizer_mac.cpp
index 011ec6f..f4c6b44 100644
--- a/libsanitizer/sanitizer_common/sanitizer_mac.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_mac.cpp
@@ -37,13 +37,21 @@
 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
 #define SANITIZER_OS_TRACE 0
 #endif
 
+// import new crash reporting api
+#if defined(__has_include) && __has_include(<CrashReporterClient.h>)
+#define HAVE_CRASHREPORTERCLIENT_H 1
+#include <CrashReporterClient.h>
+#else
+#define HAVE_CRASHREPORTERCLIENT_H 0
+#endif
+
 #if !SANITIZER_IOS
 #include <crt_externs.h>  // for _NSGetArgv and _NSGetEnviron
 #else
@@ -62,6 +70,7 @@ extern "C" {
 #include <mach/mach_time.h>
 #include <mach/vm_statistics.h>
 #include <malloc/malloc.h>
+#include <os/log.h>
 #include <pthread.h>
 #include <sched.h>
 #include <signal.h>
@@ -133,6 +142,12 @@ uptr internal_munmap(void *addr, uptr length) {
   return munmap(addr, length);
 }
 
+uptr internal_mremap(void *old_address, uptr old_size, uptr new_size, int flags,
+                     void *new_address) {
+  CHECK(false && "internal_mremap is unimplemented on Mac");
+  return 0;
+}
+
 int internal_mprotect(void *addr, uptr length, int prot) {
   return mprotect(addr, length, prot);
 }
@@ -444,7 +459,7 @@ uptr ReadBinaryName(/*out*/char *buf, uptr buf_len) {
   // On OS X the executable path is saved to the stack by dyld. Reading it
   // from there is much faster than calling dladdr, especially for large
   // binaries with symbols.
-  InternalScopedString exe_path(kMaxPathLength);
+  InternalMmapVector<char> exe_path(kMaxPathLength);
   uint32_t size = exe_path.size();
   if (_NSGetExecutablePath(exe_path.data(), &size) == 0 &&
       realpath(exe_path.data(), buf) != 0) {
@@ -620,6 +635,23 @@ constexpr u16 GetOSMajorKernelOffset() {
 
 using VersStr = char[64];
 
+static uptr ApproximateOSVersionViaKernelVersion(VersStr vers) {
+  u16 kernel_major = GetDarwinKernelVersion().major;
+  u16 offset = GetOSMajorKernelOffset();
+  CHECK_GE(kernel_major, offset);
+  u16 os_major = kernel_major - offset;
+
+  const char *format = "%d.0";
+  if (TARGET_OS_OSX) {
+    if (os_major >= 16) {  // macOS 11+
+      os_major -= 5;
+    } else {  // macOS 10.15 and below
+      format = "10.%d";
+    }
+  }
+  return internal_snprintf(vers, sizeof(VersStr), format, os_major);
+}
+
 static void GetOSVersion(VersStr vers) {
   uptr len = sizeof(VersStr);
   if (SANITIZER_IOSSIM) {
@@ -633,17 +665,19 @@ static void GetOSVersion(VersStr vers) {
   } else {
     int res =
         internal_sysctlbyname("kern.osproductversion", vers, &len, nullptr, 0);
-    if (res) {
-      // Fallback for XNU 17 (macOS 10.13) and below that do not provide the
-      // `kern.osproductversion` property.
-      u16 kernel_major = GetDarwinKernelVersion().major;
-      u16 offset = GetOSMajorKernelOffset();
-      CHECK_LE(kernel_major, 17);
-      CHECK_GE(kernel_major, offset);
-      u16 os_major = kernel_major - offset;
-
-      auto format = TARGET_OS_OSX ? "10.%d" : "%d.0";
-      len = internal_snprintf(vers, len, format, os_major);
+
+    // XNU 17 (macOS 10.13) and below do not provide the sysctl
+    // `kern.osproductversion` entry (res != 0).
+    bool no_os_version = res != 0;
+
+    // For launchd, sanitizer initialization runs before sysctl is setup
+    // (res == 0 && len != strlen(vers), vers is not a valid version).  However,
+    // the kernel version `kern.osrelease` is available.
+    bool launchd = (res == 0 && internal_strlen(vers) < 3);
+    if (launchd) CHECK_EQ(internal_getpid(), 1);
+
+    if (no_os_version || launchd) {
+      len = ApproximateOSVersionViaKernelVersion(vers);
     }
   }
   CHECK_LT(len, sizeof(VersStr));
@@ -681,7 +715,7 @@ static void MapToMacos(u16 *major, u16 *minor) {
 }
 
 static MacosVersion GetMacosAlignedVersionInternal() {
-  VersStr vers;
+  VersStr vers = {};
   GetOSVersion(vers);
 
   u16 major, minor;
@@ -707,7 +741,7 @@ MacosVersion GetMacosAlignedVersion() {
 }
 
 DarwinKernelVersion GetDarwinKernelVersion() {
-  VersStr vers;
+  VersStr vers = {};
   uptr len = sizeof(VersStr);
   int res = internal_sysctlbyname("kern.osrelease", vers, &len, nullptr, 0);
   CHECK_EQ(res, 0);
@@ -751,7 +785,51 @@ static BlockingMutex syslog_lock(LINKER_INITIALIZED);
 void WriteOneLineToSyslog(const char *s) {
 #if !SANITIZER_GO
   syslog_lock.CheckLocked();
-  asl_log(nullptr, nullptr, ASL_LEVEL_ERR, "%s", s);
+  if (GetMacosAlignedVersion() >= MacosVersion(10, 12)) {
+    os_log_error(OS_LOG_DEFAULT, "%{public}s", s);
+  } else {
+    asl_log(nullptr, nullptr, ASL_LEVEL_ERR, "%s", s);
+  }
+#endif
+}
+
+// buffer to store crash report application information
+static char crashreporter_info_buff[__sanitizer::kErrorMessageBufferSize] = {};
+static BlockingMutex crashreporter_info_mutex(LINKER_INITIALIZED);
+
+extern "C" {
+// Integrate with crash reporter libraries.
+#if HAVE_CRASHREPORTERCLIENT_H
+CRASH_REPORTER_CLIENT_HIDDEN
+struct crashreporter_annotations_t gCRAnnotations
+    __attribute__((section("__DATA," CRASHREPORTER_ANNOTATIONS_SECTION))) = {
+        CRASHREPORTER_ANNOTATIONS_VERSION,
+        0,
+        0,
+        0,
+        0,
+        0,
+        0,
+#if CRASHREPORTER_ANNOTATIONS_VERSION > 4
+        0,
+#endif
+};
+
+#else
+// fall back to old crashreporter api
+static const char *__crashreporter_info__ __attribute__((__used__)) =
+    &crashreporter_info_buff[0];
+asm(".desc ___crashreporter_info__, 0x10");
+#endif
+
+}  // extern "C"
+
+static void CRAppendCrashLogMessage(const char *msg) {
+  BlockingMutexLock l(&crashreporter_info_mutex);
+  internal_strlcat(crashreporter_info_buff, msg,
+                   sizeof(crashreporter_info_buff));
+#if HAVE_CRASHREPORTERCLIENT_H
+  (void)CRSetCrashLogMessage(crashreporter_info_buff);
 #endif
 }
 
@@ -947,7 +1025,7 @@ void MaybeReexec() {
   if (DyldNeedsEnvVariable() && !lib_is_in_env) {
     // DYLD_INSERT_LIBRARIES is not set or does not contain the runtime
     // library.
-    InternalScopedString program_name(1024);
+    InternalMmapVector<char> program_name(1024);
     uint32_t buf_size = program_name.size();
     _NSGetExecutablePath(program_name.data(), &buf_size);
     char *new_env = const_cast<char*>(info.dli_fname);
@@ -1066,7 +1144,7 @@ char **GetArgv() {
   return *_NSGetArgv();
 }
 
-#if SANITIZER_IOS
+#if SANITIZER_IOS && !SANITIZER_IOSSIM
 // The task_vm_info struct is normally provided by the macOS SDK, but we need
 // fields only available in 10.12+. Declare the struct manually to be able to
 // build against older SDKs.
@@ -1106,26 +1184,35 @@ static uptr GetTaskInfoMaxAddress() {
 
 uptr GetMaxUserVirtualAddress() {
   static uptr max_vm = GetTaskInfoMaxAddress();
-  if (max_vm != 0)
-    return max_vm - 1;
+  if (max_vm != 0) {
+    const uptr ret_value = max_vm - 1;
+    CHECK_LE(ret_value, SANITIZER_MMAP_RANGE_SIZE);
+    return ret_value;
+  }
 
   // xnu cannot provide vm address limit
 # if SANITIZER_WORDSIZE == 32
-  return 0xffe00000 - 1;
+  constexpr uptr fallback_max_vm = 0xffe00000 - 1;
 # else
-  return 0x200000000 - 1;
+  constexpr uptr fallback_max_vm = 0x200000000 - 1;
 # endif
+  static_assert(fallback_max_vm <= SANITIZER_MMAP_RANGE_SIZE,
+                "Max virtual address must be less than mmap range size.");
+  return fallback_max_vm;
 }
 
 #else // !SANITIZER_IOS
 
 uptr GetMaxUserVirtualAddress() {
 # if SANITIZER_WORDSIZE == 64
-  return (1ULL << 47) - 1;  // 0x00007fffffffffffUL;
+  constexpr uptr max_vm = (1ULL << 47) - 1;  // 0x00007fffffffffffUL;
 # else // SANITIZER_WORDSIZE == 32
   static_assert(SANITIZER_WORDSIZE == 32, "Wrong wordsize");
-  return (1ULL << 32) - 1;  // 0xffffffff;
+  constexpr uptr max_vm = (1ULL << 32) - 1;  // 0xffffffff;
 # endif
+  static_assert(max_vm <= SANITIZER_MMAP_RANGE_SIZE,
+                "Max virtual address must be less than mmap range size.");
+  return max_vm;
 }
 #endif
 
@@ -1180,6 +1267,12 @@ uptr MapDynamicShadow(uptr shadow_size_bytes, uptr shadow_scale,
   return shadow_start;
 }
 
+uptr MapDynamicShadowAndAliases(uptr shadow_size, uptr alias_size,
+                                uptr num_aliases, uptr ring_buffer_size) {
+  CHECK(false && "HWASan aliasing is unimplemented on Mac");
+  return 0;
+}
+
 uptr FindAvailableMemoryRange(uptr size, uptr alignment, uptr left_padding,
                               uptr *largest_gap_found,
                               uptr *max_occupied_addr) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_mac.h b/libsanitizer/sanitizer_common/sanitizer_mac.h
index a2c42b3..0b6af5a 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"
 
@@ -84,22 +64,5 @@ void RestrictMemoryToMaxAddress(uptr max_address);
 
 }  // namespace __sanitizer
 
-extern "C" {
-static char __crashreporter_info_buff__[__sanitizer::kErrorMessageBufferSize] =
-  {};
-static const char *__crashreporter_info__ __attribute__((__used__)) =
-  &__crashreporter_info_buff__[0];
-asm(".desc ___crashreporter_info__, 0x10");
-} // extern "C"
-
-namespace __sanitizer {
-static BlockingMutex crashreporter_info_mutex(LINKER_INITIALIZED);
-
-inline void CRAppendCrashLogMessage(const char *msg) {
-  BlockingMutexLock l(&crashreporter_info_mutex);
-  internal_strlcat(__crashreporter_info_buff__, msg,
-                   sizeof(__crashreporter_info_buff__)); }
-}  // namespace __sanitizer
-
 #endif  // SANITIZER_MAC
 #endif  // SANITIZER_MAC_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_malloc_mac.inc b/libsanitizer/sanitizer_common/sanitizer_malloc_mac.inc
index 647bcdf..e3b664f 100644
--- a/libsanitizer/sanitizer_common/sanitizer_malloc_mac.inc
+++ b/libsanitizer/sanitizer_common/sanitizer_malloc_mac.inc
@@ -120,11 +120,7 @@ INTERCEPTOR(int, malloc_make_nonpurgeable, void *ptr) {
 
 INTERCEPTOR(void, malloc_set_zone_name, malloc_zone_t *zone, const char *name) {
   COMMON_MALLOC_ENTER();
-  // Allocate |sizeof(COMMON_MALLOC_ZONE_NAME "-") + internal_strlen(name)|
-  // bytes.
-  size_t buflen =
-      sizeof(COMMON_MALLOC_ZONE_NAME "-") + (name ? internal_strlen(name) : 0);
-  InternalScopedString new_name(buflen);
+  InternalScopedString new_name;
   if (name && zone->introspect == sanitizer_zone.introspect) {
     new_name.append(COMMON_MALLOC_ZONE_NAME "-%s", name);
     name = new_name.data();
diff --git a/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp b/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
index 98ac736..ac20f91 100644
--- a/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_netbsd.cpp
@@ -105,6 +105,12 @@ uptr internal_munmap(void *addr, uptr length) {
   return _REAL(munmap, addr, length);
 }
 
+uptr internal_mremap(void *old_address, uptr old_size, uptr new_size, int flags,
+                     void *new_address) {
+  CHECK(false && "internal_mremap is unimplemented on NetBSD");
+  return 0;
+}
+
 int internal_mprotect(void *addr, uptr length, int prot) {
   DEFINE__REAL(int, mprotect, void *a, uptr b, int c);
   return _REAL(mprotect, addr, length, prot);
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform.h b/libsanitizer/sanitizer_common/sanitizer_platform.h
index b2372a0..2f64584 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform.h
+++ b/libsanitizer/sanitizer_common/sanitizer_platform.h
@@ -19,12 +19,25 @@
 # error "This operating system is not supported"
 #endif
 
+// Get __GLIBC__ on a glibc platform. Exclude Android: features.h includes C
+// function declarations into a .S file which doesn't compile.
+// https://crbug.com/1162741
+#if __has_include(<features.h>) && !defined(__ANDROID__)
+#include <features.h>
+#endif
+
 #if defined(__linux__)
 # define SANITIZER_LINUX   1
 #else
 # define SANITIZER_LINUX   0
 #endif
 
+#if defined(__GLIBC__)
+# define SANITIZER_GLIBC   1
+#else
+# define SANITIZER_GLIBC   0
+#endif
+
 #if defined(__FreeBSD__)
 # define SANITIZER_FREEBSD 1
 #else
@@ -46,6 +59,11 @@
 #if defined(__APPLE__)
 # define SANITIZER_MAC     1
 # include <TargetConditionals.h>
+# if TARGET_OS_OSX
+#  define SANITIZER_OSX    1
+# else
+#  define SANITIZER_OSX    0
+# endif
 # if TARGET_OS_IPHONE
 #  define SANITIZER_IOS    1
 # else
@@ -60,6 +78,7 @@
 # define SANITIZER_MAC     0
 # define SANITIZER_IOS     0
 # define SANITIZER_IOSSIM  0
+# define SANITIZER_OSX     0
 #endif
 
 #if defined(__APPLE__) && TARGET_OS_IPHONE && TARGET_OS_WATCH
@@ -247,8 +266,12 @@
 #define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 38)
 #elif defined(__aarch64__)
 # if SANITIZER_MAC
-// Darwin iOS/ARM64 has a 36-bit VMA, 64GiB VM
-#  define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 36)
+#  if SANITIZER_OSX || SANITIZER_IOSSIM
+#   define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 47)
+#  else
+    // Darwin iOS/ARM64 has a 36-bit VMA, 64GiB VM
+#   define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 36)
+#  endif
 # else
 #  define SANITIZER_MMAP_RANGE_SIZE FIRST_32_SECOND_64(1ULL << 32, 1ULL << 48)
 # endif
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h b/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
index 18bab34..731df71 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_interceptors.h
@@ -46,6 +46,12 @@
 #define SI_LINUX_NOT_ANDROID 0
 #endif
 
+#if SANITIZER_GLIBC
+#define SI_GLIBC 1
+#else
+#define SI_GLIBC 0
+#endif
+
 #if SANITIZER_ANDROID
 #define SI_ANDROID 1
 #else
@@ -159,7 +165,7 @@
   SANITIZER_INTERCEPT_MEMCMP &&  \
       ((SI_POSIX && _GNU_SOURCE) || SI_NETBSD || SI_FREEBSD)
 #define SANITIZER_INTERCEPT_STRNDUP SI_POSIX
-#define SANITIZER_INTERCEPT___STRNDUP SI_LINUX_NOT_FREEBSD
+#define SANITIZER_INTERCEPT___STRNDUP SI_GLIBC
 #if defined(__ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__) && \
     __ENVIRONMENT_MAC_OS_X_VERSION_MIN_REQUIRED__ < 1070
 #define SI_MAC_DEPLOYMENT_BELOW_10_7 1
@@ -183,8 +189,8 @@
 #define SANITIZER_INTERCEPT_FPUTS SI_POSIX
 #define SANITIZER_INTERCEPT_PUTS SI_POSIX
 
-#define SANITIZER_INTERCEPT_PREAD64 SI_LINUX_NOT_ANDROID || SI_SOLARIS32
-#define SANITIZER_INTERCEPT_PWRITE64 SI_LINUX_NOT_ANDROID || SI_SOLARIS32
+#define SANITIZER_INTERCEPT_PREAD64 (SI_GLIBC || SI_SOLARIS32)
+#define SANITIZER_INTERCEPT_PWRITE64 (SI_GLIBC || SI_SOLARIS32)
 
 #define SANITIZER_INTERCEPT_READV SI_POSIX
 #define SANITIZER_INTERCEPT_WRITEV SI_POSIX
@@ -192,8 +198,8 @@
 #define SANITIZER_INTERCEPT_PREADV \
   (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID)
 #define SANITIZER_INTERCEPT_PWRITEV SI_LINUX_NOT_ANDROID
-#define SANITIZER_INTERCEPT_PREADV64 SI_LINUX_NOT_ANDROID
-#define SANITIZER_INTERCEPT_PWRITEV64 SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_PREADV64 SI_GLIBC
+#define SANITIZER_INTERCEPT_PWRITEV64 SI_GLIBC
 
 #define SANITIZER_INTERCEPT_PRCTL SI_LINUX
 
@@ -201,16 +207,16 @@
 #define SANITIZER_INTERCEPT_STRPTIME SI_POSIX
 
 #define SANITIZER_INTERCEPT_SCANF SI_POSIX
-#define SANITIZER_INTERCEPT_ISOC99_SCANF SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_ISOC99_SCANF SI_GLIBC
 
 #ifndef SANITIZER_INTERCEPT_PRINTF
 #define SANITIZER_INTERCEPT_PRINTF SI_POSIX
 #define SANITIZER_INTERCEPT_PRINTF_L (SI_FREEBSD || SI_NETBSD)
-#define SANITIZER_INTERCEPT_ISOC99_PRINTF SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_ISOC99_PRINTF SI_GLIBC
 #endif
 
 #define SANITIZER_INTERCEPT___PRINTF_CHK \
-  (SANITIZER_INTERCEPT_PRINTF && SI_LINUX_NOT_ANDROID)
+  (SANITIZER_INTERCEPT_PRINTF && SI_GLIBC)
 
 #define SANITIZER_INTERCEPT_FREXP SI_NOT_FUCHSIA
 #define SANITIZER_INTERCEPT_FREXPF_FREXPL SI_POSIX
@@ -220,13 +226,11 @@
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_GETPWENT \
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_FGETGRENT_R \
-  (SI_FREEBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_FGETGRENT_R (SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_FGETPWENT SI_LINUX_NOT_ANDROID || SI_SOLARIS
 #define SANITIZER_INTERCEPT_GETPWENT_R \
-  (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_FGETPWENT_R \
-  (SI_FREEBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+  (SI_FREEBSD || SI_NETBSD || SI_GLIBC || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_FGETPWENT_R (SI_FREEBSD || SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_SETPWENT \
   (SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_CLOCK_GETTIME \
@@ -234,8 +238,8 @@
 #define SANITIZER_INTERCEPT_CLOCK_GETCPUCLOCKID SI_LINUX
 #define SANITIZER_INTERCEPT_GETITIMER SI_POSIX
 #define SANITIZER_INTERCEPT_TIME SI_POSIX
-#define SANITIZER_INTERCEPT_GLOB SI_LINUX_NOT_ANDROID || SI_SOLARIS
-#define SANITIZER_INTERCEPT_GLOB64 SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_GLOB (SI_GLIBC || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_GLOB64 SI_GLIBC
 #define SANITIZER_INTERCEPT_WAIT SI_POSIX
 #define SANITIZER_INTERCEPT_INET SI_POSIX
 #define SANITIZER_INTERCEPT_PTHREAD_GETSCHEDPARAM SI_POSIX
@@ -250,8 +254,7 @@
   (SI_FREEBSD || SI_LINUX_NOT_ANDROID)
 #define SANITIZER_INTERCEPT_GETHOSTBYADDR_R \
   (SI_FREEBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_GETHOSTENT_R \
-  (SI_FREEBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_GETHOSTENT_R (SI_FREEBSD || SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_GETSOCKOPT SI_POSIX
 #define SANITIZER_INTERCEPT_ACCEPT SI_POSIX
 #define SANITIZER_INTERCEPT_ACCEPT4 (SI_LINUX_NOT_ANDROID || SI_NETBSD)
@@ -296,8 +299,7 @@
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_TCGETATTR SI_LINUX_NOT_ANDROID || SI_SOLARIS
 #define SANITIZER_INTERCEPT_REALPATH SI_POSIX
-#define SANITIZER_INTERCEPT_CANONICALIZE_FILE_NAME \
-  (SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_CANONICALIZE_FILE_NAME (SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_CONFSTR \
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_SCHED_GETAFFINITY SI_LINUX_NOT_ANDROID
@@ -324,7 +326,7 @@
 #define SANITIZER_INTERCEPT_SIGPROCMASK SI_POSIX
 #define SANITIZER_INTERCEPT_PTHREAD_SIGMASK SI_POSIX
 #define SANITIZER_INTERCEPT_BACKTRACE \
-  (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+  (SI_FREEBSD || SI_NETBSD || SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_GETMNTENT SI_LINUX
 #define SANITIZER_INTERCEPT_GETMNTENT_R SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT_STATFS \
@@ -342,11 +344,11 @@
 #define SANITIZER_INTERCEPT_SHMCTL                                       \
   (((SI_FREEBSD || SI_LINUX_NOT_ANDROID) && SANITIZER_WORDSIZE == 64) || \
    SI_NETBSD || SI_SOLARIS)  // NOLINT
-#define SANITIZER_INTERCEPT_RANDOM_R SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_RANDOM_R SI_GLIBC
 #define SANITIZER_INTERCEPT_PTHREAD_ATTR_GET SI_POSIX
 #define SANITIZER_INTERCEPT_PTHREAD_ATTR_GETINHERITSCHED \
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_PTHREAD_ATTR_GETAFFINITY_NP SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_PTHREAD_ATTR_GETAFFINITY_NP SI_GLIBC
 #define SANITIZER_INTERCEPT_PTHREAD_ATTR_GET_SCHED SI_POSIX
 #define SANITIZER_INTERCEPT_PTHREAD_MUTEXATTR_GETPSHARED \
   (SI_POSIX && !SI_NETBSD)
@@ -360,7 +362,7 @@
 #define SANITIZER_INTERCEPT_PTHREAD_MUTEXATTR_GETROBUST_NP SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT_PTHREAD_RWLOCKATTR_GETPSHARED \
   (SI_POSIX && !SI_NETBSD)
-#define SANITIZER_INTERCEPT_PTHREAD_RWLOCKATTR_GETKIND_NP SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_PTHREAD_RWLOCKATTR_GETKIND_NP SI_GLIBC
 #define SANITIZER_INTERCEPT_PTHREAD_CONDATTR_GETPSHARED (SI_POSIX && !SI_NETBSD)
 #define SANITIZER_INTERCEPT_PTHREAD_CONDATTR_GETCLOCK \
   (SI_LINUX_NOT_ANDROID || SI_SOLARIS)
@@ -368,7 +370,7 @@
   (SI_LINUX_NOT_ANDROID && !SI_NETBSD)
 #define SANITIZER_INTERCEPT_THR_EXIT SI_FREEBSD
 #define SANITIZER_INTERCEPT_TMPNAM SI_POSIX
-#define SANITIZER_INTERCEPT_TMPNAM_R SI_LINUX_NOT_ANDROID || SI_SOLARIS
+#define SANITIZER_INTERCEPT_TMPNAM_R (SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_PTSNAME SI_LINUX
 #define SANITIZER_INTERCEPT_PTSNAME_R SI_LINUX
 #define SANITIZER_INTERCEPT_TTYNAME SI_POSIX
@@ -381,7 +383,7 @@
 #define SANITIZER_INTERCEPT_LGAMMAL (SI_POSIX && !SI_NETBSD)
 #define SANITIZER_INTERCEPT_LGAMMA_R (SI_FREEBSD || SI_LINUX || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_LGAMMAL_R SI_LINUX_NOT_ANDROID || SI_SOLARIS
-#define SANITIZER_INTERCEPT_DRAND48_R SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_DRAND48_R SI_GLIBC
 #define SANITIZER_INTERCEPT_RAND_R \
   (SI_FREEBSD || SI_NETBSD || SI_MAC || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_ICONV \
@@ -396,12 +398,12 @@
   (SI_LINUX || SI_FREEBSD || SI_NETBSD || SI_MAC || SI_SOLARIS)
 
 #define SANITIZER_INTERCEPT_PTHREAD_MUTEX SI_POSIX
-#define SANITIZER_INTERCEPT___PTHREAD_MUTEX SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT___PTHREAD_MUTEX SI_GLIBC
 #define SANITIZER_INTERCEPT___LIBC_MUTEX SI_NETBSD
 #define SANITIZER_INTERCEPT_PTHREAD_SETNAME_NP \
-  (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+  (SI_FREEBSD || SI_NETBSD || SI_GLIBC || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_PTHREAD_GETNAME_NP \
-  (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
+  (SI_FREEBSD || SI_NETBSD || SI_GLIBC || SI_SOLARIS)
 
 #define SANITIZER_INTERCEPT_TLS_GET_ADDR \
   (SI_FREEBSD || SI_NETBSD || SI_LINUX_NOT_ANDROID || SI_SOLARIS)
@@ -419,19 +421,19 @@
 #else
 #define SANITIZER_INTERCEPT_AEABI_MEM 0
 #endif
-#define SANITIZER_INTERCEPT___BZERO SI_MAC || SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT___BZERO SI_MAC || SI_GLIBC
 #define SANITIZER_INTERCEPT_BZERO SI_LINUX_NOT_ANDROID
 #define SANITIZER_INTERCEPT_FTIME (!SI_FREEBSD && !SI_NETBSD && SI_POSIX)
-#define SANITIZER_INTERCEPT_XDR SI_LINUX_NOT_ANDROID || SI_SOLARIS
-#define SANITIZER_INTERCEPT_XDRREC SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_XDR (SI_GLIBC || SI_SOLARIS)
+#define SANITIZER_INTERCEPT_XDRREC SI_GLIBC
 #define SANITIZER_INTERCEPT_TSEARCH \
   (SI_LINUX_NOT_ANDROID || SI_MAC || SI_NETBSD || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_LIBIO_INTERNALS SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_LIBIO_INTERNALS SI_GLIBC
 #define SANITIZER_INTERCEPT_FOPEN SI_POSIX
-#define SANITIZER_INTERCEPT_FOPEN64 SI_LINUX_NOT_ANDROID || SI_SOLARIS32
+#define SANITIZER_INTERCEPT_FOPEN64 (SI_GLIBC || SI_SOLARIS32)
 #define SANITIZER_INTERCEPT_OPEN_MEMSTREAM \
   (SI_LINUX_NOT_ANDROID || SI_NETBSD || SI_SOLARIS)
-#define SANITIZER_INTERCEPT_OBSTACK SI_LINUX_NOT_ANDROID
+#define SANITIZER_INTERCEPT_OBSTACK SI_GLIBC
 #define SANITIZER_INTERCEPT_FFLUSH SI_POSIX
 #define SANITIZER_INTERCEPT_FCLOSE SI_POSIX
 
@@ -456,7 +458,7 @@
 #define SANITIZER_INTERCEPT_CTERMID_R (SI_MAC || SI_FREEBSD || SI_SOLARIS)
 
 #define SANITIZER_INTERCEPTOR_HOOKS \
-  (SI_LINUX || SI_MAC || SI_WINDOWS || SI_NETBSD)
+  (SI_LINUX || SI_MAC || SI_WINDOWS || SI_FREEBSD || SI_NETBSD || SI_SOLARIS)
 #define SANITIZER_INTERCEPT_RECV_RECVFROM SI_POSIX
 #define SANITIZER_INTERCEPT_SEND_SENDTO SI_POSIX
 #define SANITIZER_INTERCEPT_EVENTFD_READ_WRITE SI_LINUX
@@ -479,20 +481,12 @@
 
 #define SANITIZER_INTERCEPT_MMAP SI_POSIX
 #define SANITIZER_INTERCEPT_MMAP64 SI_LINUX_NOT_ANDROID
-#define SANITIZER_INTERCEPT_MALLOPT_AND_MALLINFO                             \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && SI_NOT_RTEMS && \
-   !SI_SOLARIS)  // NOLINT
+#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___LIBC_MEMALIGN                               \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && !SI_OPENBSD && SI_NOT_RTEMS && \
-   !SI_ANDROID)  // NOLINT
-#define SANITIZER_INTERCEPT_PVALLOC                                          \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && SI_NOT_RTEMS && \
-   !SI_SOLARIS)  // NOLINT
-#define SANITIZER_INTERCEPT_CFREE                                            \
-  (!SI_FREEBSD && !SI_MAC && !SI_NETBSD && SI_NOT_FUCHSIA && SI_NOT_RTEMS && \
-   !SI_SOLARIS && !SANITIZER_ANDROID)  // NOLINT
+#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_MALLOC_USABLE_SIZE (!SI_MAC && !SI_NETBSD)
@@ -532,7 +526,7 @@
 #define SANITIZER_INTERCEPT_STRMODE (SI_NETBSD || SI_FREEBSD)
 #define SANITIZER_INTERCEPT_TTYENT SI_NETBSD
 #define SANITIZER_INTERCEPT_PROTOENT (SI_NETBSD || SI_LINUX)
-#define SANITIZER_INTERCEPT_PROTOENT_R (SI_LINUX_NOT_ANDROID)
+#define SANITIZER_INTERCEPT_PROTOENT_R SI_GLIBC
 #define SANITIZER_INTERCEPT_NETENT SI_NETBSD
 #define SANITIZER_INTERCEPT_SETVBUF \
   (SI_NETBSD || SI_FREEBSD || SI_LINUX || SI_MAC)
@@ -583,7 +577,7 @@
 #define SANITIZER_INTERCEPT_GETENTROPY SI_FREEBSD
 #define SANITIZER_INTERCEPT_QSORT \
   (SI_POSIX && !SI_IOSSIM && !SI_WATCHOS && !SI_TVOS && !SI_ANDROID)
-#define SANITIZER_INTERCEPT_QSORT_R (SI_LINUX && !SI_ANDROID)
+#define SANITIZER_INTERCEPT_QSORT_R SI_GLIBC
 // sigaltstack on i386 macOS cannot be intercepted due to setjmp()
 // calling it and assuming that it does not clobber registers.
 #define SANITIZER_INTERCEPT_SIGALTSTACK \
@@ -591,4 +585,25 @@
 #define SANITIZER_INTERCEPT_UNAME (SI_POSIX && !SI_FREEBSD)
 #define SANITIZER_INTERCEPT___XUNAME 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
+// to do two things:
+// 1. Build compiler-rt with -DSANITIZER_OVERRIDE_INTERCEPTORS.
+// 2. Provide a header file named sanitizer_intercept_overriders.h in the
+//    include path for their compiler-rt build.
+// An example of an overrider for strlen interceptor that one can list in
+// sanitizer_intercept_overriders.h is as follows:
+//
+// #ifdef SANITIZER_INTERCEPT_STRLEN
+// #undef SANITIZER_INTERCEPT_STRLEN
+// #define SANITIZER_INTERCEPT_STRLEN <value of choice>
+// #endif
+//
+// This "feature" is useful for downstream users who do not want some of
+// their libc funtions to be intercepted. They can selectively disable
+// interception of those functions.
+#ifdef SANITIZER_OVERRIDE_INTERCEPTORS
+#include <sanitizer_intercept_overriders.h>
+#endif
+
 #endif  // #ifndef SANITIZER_PLATFORM_INTERCEPTORS_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_limits_freebsd.cpp b/libsanitizer/sanitizer_common/sanitizer_platform_limits_freebsd.cpp
index b1c15be..b5a45ae 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_limits_freebsd.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_limits_freebsd.cpp
@@ -35,7 +35,10 @@
 #include <sys/stat.h>
 #include <sys/statvfs.h>
 #include <sys/time.h>
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-W#warnings"
 #include <sys/timeb.h>
+#pragma clang diagnostic pop
 #include <sys/times.h>
 #include <sys/timespec.h>
 #include <sys/types.h>
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp b/libsanitizer/sanitizer_common/sanitizer_platform_limits_linux.cpp
index f22f503..c51327e 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.cpp b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.cpp
index 1427cec..35a690c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.cpp
@@ -11,19 +11,20 @@
 // Sizes and layouts of platform-specific POSIX data structures.
 //===----------------------------------------------------------------------===//
 
-#include "sanitizer_platform.h"
-
-#if SANITIZER_LINUX || SANITIZER_MAC
+#if defined(__linux__) || defined(__APPLE__)
 // Tests in this file assume that off_t-dependent data structures match the
 // libc ABI. For example, struct dirent here is what readdir() function (as
 // exported from libc) returns, and not the user-facing "dirent", which
 // depends on _FILE_OFFSET_BITS setting.
 // To get this "true" dirent definition, we undefine _FILE_OFFSET_BITS below.
-#ifdef _FILE_OFFSET_BITS
 #undef _FILE_OFFSET_BITS
 #endif
 
 // Must go after undef _FILE_OFFSET_BITS.
+#include "sanitizer_platform.h"
+
+#if SANITIZER_LINUX || SANITIZER_MAC
+// Must go after undef _FILE_OFFSET_BITS.
 #include "sanitizer_glibc_version.h"
 
 #include <arpa/inet.h>
@@ -37,6 +38,7 @@
 #include <pwd.h>
 #include <signal.h>
 #include <stddef.h>
+#include <stdio.h>
 #include <sys/mman.h>
 #include <sys/resource.h>
 #include <sys/socket.h>
@@ -58,7 +60,6 @@
 #endif
 
 #if !SANITIZER_ANDROID
-#include <fstab.h>
 #include <sys/mount.h>
 #include <sys/timeb.h>
 #include <utmpx.h>
@@ -110,20 +111,31 @@ typedef struct user_fpregs elf_fpregset_t;
 #include <wordexp.h>
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
-#include <glob.h>
-#include <obstack.h>
-#include <mqueue.h>
+#if SANITIZER_LINUX
+#if SANITIZER_GLIBC
+#include <fstab.h>
 #include <net/if_ppp.h>
 #include <netax25/ax25.h>
 #include <netipx/ipx.h>
 #include <netrom/netrom.h>
+#include <obstack.h>
 #if HAVE_RPC_XDR_H
 # include <rpc/xdr.h>
 #endif
 #include <scsi/scsi.h>
-#include <sys/mtio.h>
+#else
+#include <linux/if_ppp.h>
+#include <linux/kd.h>
+#include <linux/ppp_defs.h>
+#endif  // SANITIZER_GLIBC
+
+#if SANITIZER_ANDROID
+#include <linux/mtio.h>
+#else
+#include <glob.h>
+#include <mqueue.h>
 #include <sys/kd.h>
+#include <sys/mtio.h>
 #include <sys/shm.h>
 #include <sys/statvfs.h>
 #include <sys/timex.h>
@@ -142,20 +154,14 @@ typedef struct user_fpregs elf_fpregset_t;
 #include <sys/msg.h>
 #include <sys/ipc.h>
 #include <crypt.h>
-#endif // SANITIZER_LINUX && !SANITIZER_ANDROID
+#endif  // SANITIZER_ANDROID
 
-#if SANITIZER_ANDROID
-#include <linux/kd.h>
-#include <linux/mtio.h>
-#include <linux/ppp_defs.h>
-#include <linux/if_ppp.h>
-#endif
-
-#if SANITIZER_LINUX
 #include <link.h>
 #include <sys/vfs.h>
 #include <sys/epoll.h>
 #include <linux/capability.h>
+#else
+#include <fstab.h>
 #endif // SANITIZER_LINUX
 
 #if SANITIZER_MAC
@@ -202,8 +208,11 @@ namespace __sanitizer {
   unsigned struct_statfs64_sz = sizeof(struct statfs64);
 #endif // (SANITIZER_MAC && !TARGET_CPU_ARM64) && !SANITIZER_IOS
 
-#if !SANITIZER_ANDROID
+#if SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD || SANITIZER_MAC
   unsigned struct_fstab_sz = sizeof(struct fstab);
+#endif  // SANITIZER_GLIBC || SANITIZER_FREEBSD || SANITIZER_NETBSD ||
+        // SANITIZER_MAC
+#if !SANITIZER_ANDROID
   unsigned struct_statfs_sz = sizeof(struct statfs);
   unsigned struct_sockaddr_sz = sizeof(struct sockaddr);
   unsigned ucontext_t_sz = sizeof(ucontext_t);
@@ -299,7 +308,7 @@ unsigned struct_ElfW_Phdr_sz = sizeof(ElfW(Phdr));
 unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
   int glob_nomatch = GLOB_NOMATCH;
   int glob_altdirfunc = GLOB_ALTDIRFUNC;
 #endif
@@ -422,7 +431,9 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned struct_input_id_sz = sizeof(struct input_id);
   unsigned struct_mtpos_sz = sizeof(struct mtpos);
   unsigned struct_rtentry_sz = sizeof(struct rtentry);
+#if SANITIZER_GLIBC || SANITIZER_ANDROID
   unsigned struct_termio_sz = sizeof(struct termio);
+#endif
   unsigned struct_vt_consize_sz = sizeof(struct vt_consize);
   unsigned struct_vt_sizes_sz = sizeof(struct vt_sizes);
   unsigned struct_vt_stat_sz = sizeof(struct vt_stat);
@@ -447,7 +458,7 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned struct_vt_mode_sz = sizeof(struct vt_mode);
 #endif // SANITIZER_LINUX
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
   unsigned struct_ax25_parms_struct_sz = sizeof(struct ax25_parms_struct);
   unsigned struct_cyclades_monitor_sz = sizeof(struct cyclades_monitor);
 #if EV_VERSION > (0x010000)
@@ -470,12 +481,10 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned struct_sockaddr_ax25_sz = sizeof(struct sockaddr_ax25);
   unsigned struct_unimapdesc_sz = sizeof(struct unimapdesc);
   unsigned struct_unimapinit_sz = sizeof(struct unimapinit);
-#endif // SANITIZER_LINUX && !SANITIZER_ANDROID
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
   unsigned struct_audio_buf_info_sz = sizeof(struct audio_buf_info);
   unsigned struct_ppp_stats_sz = sizeof(struct ppp_stats);
-#endif // (SANITIZER_LINUX || SANITIZER_FREEBSD) && !SANITIZER_ANDROID
+#endif  // SANITIZER_GLIBC
 
 #if !SANITIZER_ANDROID && !SANITIZER_MAC
   unsigned struct_sioc_sg_req_sz = sizeof(struct sioc_sg_req);
@@ -881,6 +890,7 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned IOCTL_PIO_UNIMAP = PIO_UNIMAP;
   unsigned IOCTL_PIO_UNIMAPCLR = PIO_UNIMAPCLR;
   unsigned IOCTL_PIO_UNISCRNMAP = PIO_UNISCRNMAP;
+#if SANITIZER_GLIBC
   unsigned IOCTL_SCSI_IOCTL_GET_IDLUN = SCSI_IOCTL_GET_IDLUN;
   unsigned IOCTL_SCSI_IOCTL_PROBE_HOST = SCSI_IOCTL_PROBE_HOST;
   unsigned IOCTL_SCSI_IOCTL_TAGGED_DISABLE = SCSI_IOCTL_TAGGED_DISABLE;
@@ -899,6 +909,7 @@ unsigned struct_ElfW_Phdr_sz = sizeof(Elf_Phdr);
   unsigned IOCTL_SIOCNRGETPARMS = SIOCNRGETPARMS;
   unsigned IOCTL_SIOCNRRTCTL = SIOCNRRTCTL;
   unsigned IOCTL_SIOCNRSETPARMS = SIOCNRSETPARMS;
+#endif
   unsigned IOCTL_TIOCGSERIAL = TIOCGSERIAL;
   unsigned IOCTL_TIOCSERGETMULTI = TIOCSERGETMULTI;
   unsigned IOCTL_TIOCSERSETMULTI = TIOCSERSETMULTI;
@@ -969,7 +980,7 @@ CHECK_SIZE_AND_OFFSET(dl_phdr_info, dlpi_phdr);
 CHECK_SIZE_AND_OFFSET(dl_phdr_info, dlpi_phnum);
 #endif // SANITIZER_LINUX || SANITIZER_FREEBSD
 
-#if (SANITIZER_LINUX || SANITIZER_FREEBSD) && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC || SANITIZER_FREEBSD
 CHECK_TYPE_SIZE(glob_t);
 CHECK_SIZE_AND_OFFSET(glob_t, gl_pathc);
 CHECK_SIZE_AND_OFFSET(glob_t, gl_pathv);
@@ -980,7 +991,7 @@ CHECK_SIZE_AND_OFFSET(glob_t, gl_readdir);
 CHECK_SIZE_AND_OFFSET(glob_t, gl_opendir);
 CHECK_SIZE_AND_OFFSET(glob_t, gl_lstat);
 CHECK_SIZE_AND_OFFSET(glob_t, gl_stat);
-#endif
+#endif  // SANITIZER_GLIBC || SANITIZER_FREEBSD
 
 CHECK_TYPE_SIZE(addrinfo);
 CHECK_SIZE_AND_OFFSET(addrinfo, ai_flags);
@@ -1003,17 +1014,27 @@ CHECK_TYPE_SIZE(iovec);
 CHECK_SIZE_AND_OFFSET(iovec, iov_base);
 CHECK_SIZE_AND_OFFSET(iovec, iov_len);
 
+// In POSIX, int msg_iovlen; socklen_t msg_controllen; socklen_t cmsg_len; but
+// many implementations don't conform to the standard. Since we pick the
+// non-conforming glibc definition, exclude the checks for musl (incompatible
+// sizes but compatible offsets).
 CHECK_TYPE_SIZE(msghdr);
 CHECK_SIZE_AND_OFFSET(msghdr, msg_name);
 CHECK_SIZE_AND_OFFSET(msghdr, msg_namelen);
 CHECK_SIZE_AND_OFFSET(msghdr, msg_iov);
+#if SANITIZER_GLIBC || SANITIZER_ANDROID
 CHECK_SIZE_AND_OFFSET(msghdr, msg_iovlen);
+#endif
 CHECK_SIZE_AND_OFFSET(msghdr, msg_control);
+#if SANITIZER_GLIBC || SANITIZER_ANDROID
 CHECK_SIZE_AND_OFFSET(msghdr, msg_controllen);
+#endif
 CHECK_SIZE_AND_OFFSET(msghdr, msg_flags);
 
 CHECK_TYPE_SIZE(cmsghdr);
+#if SANITIZER_GLIBC || SANITIZER_ANDROID
 CHECK_SIZE_AND_OFFSET(cmsghdr, cmsg_len);
+#endif
 CHECK_SIZE_AND_OFFSET(cmsghdr, cmsg_level);
 CHECK_SIZE_AND_OFFSET(cmsghdr, cmsg_type);
 
@@ -1121,7 +1142,7 @@ CHECK_SIZE_AND_OFFSET(mntent, mnt_passno);
 
 CHECK_TYPE_SIZE(ether_addr);
 
-#if (SANITIZER_LINUX || SANITIZER_FREEBSD) && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC || SANITIZER_FREEBSD
 CHECK_TYPE_SIZE(ipc_perm);
 # if SANITIZER_FREEBSD
 CHECK_SIZE_AND_OFFSET(ipc_perm, key);
@@ -1183,7 +1204,7 @@ CHECK_SIZE_AND_OFFSET(ifaddrs, ifa_dstaddr);
 CHECK_SIZE_AND_OFFSET(ifaddrs, ifa_data);
 #endif
 
-#if SANITIZER_LINUX
+#if SANITIZER_GLIBC || SANITIZER_ANDROID
 COMPILER_CHECK(sizeof(__sanitizer_struct_mallinfo) == sizeof(struct mallinfo));
 #endif
 
@@ -1233,7 +1254,7 @@ COMPILER_CHECK(__sanitizer_XDR_DECODE == XDR_DECODE);
 COMPILER_CHECK(__sanitizer_XDR_FREE == XDR_FREE);
 #endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
+#if SANITIZER_GLIBC
 COMPILER_CHECK(sizeof(__sanitizer_FILE) <= sizeof(FILE));
 CHECK_SIZE_AND_OFFSET(FILE, _flags);
 CHECK_SIZE_AND_OFFSET(FILE, _IO_read_ptr);
@@ -1250,9 +1271,7 @@ CHECK_SIZE_AND_OFFSET(FILE, _IO_save_end);
 CHECK_SIZE_AND_OFFSET(FILE, _markers);
 CHECK_SIZE_AND_OFFSET(FILE, _chain);
 CHECK_SIZE_AND_OFFSET(FILE, _fileno);
-#endif
 
-#if SANITIZER_LINUX && !SANITIZER_ANDROID
 COMPILER_CHECK(sizeof(__sanitizer__obstack_chunk) <= sizeof(_obstack_chunk));
 CHECK_SIZE_AND_OFFSET(_obstack_chunk, limit);
 CHECK_SIZE_AND_OFFSET(_obstack_chunk, prev);
@@ -1267,7 +1286,7 @@ CHECK_SIZE_AND_OFFSET(cookie_io_functions_t, read);
 CHECK_SIZE_AND_OFFSET(cookie_io_functions_t, write);
 CHECK_SIZE_AND_OFFSET(cookie_io_functions_t, seek);
 CHECK_SIZE_AND_OFFSET(cookie_io_functions_t, close);
-#endif
+#endif  // SANITIZER_GLIBC
 
 #if SANITIZER_LINUX || SANITIZER_FREEBSD
 CHECK_TYPE_SIZE(sem_t);
diff --git a/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h b/libsanitizer/sanitizer_common/sanitizer_platform_limits_posix.h
index 0812039..836b178 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;
@@ -443,6 +443,8 @@ struct __sanitizer_cmsghdr {
   int cmsg_type;
 };
 #else
+// In POSIX, int msg_iovlen; socklen_t msg_controllen; socklen_t cmsg_len; but
+// many implementations don't conform to the standard.
 struct __sanitizer_msghdr {
   void *msg_name;
   unsigned msg_namelen;
diff --git a/libsanitizer/sanitizer_common/sanitizer_posix.cpp b/libsanitizer/sanitizer_common/sanitizer_posix.cpp
index 2e08009..f8457a6 100644
--- a/libsanitizer/sanitizer_common/sanitizer_posix.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_posix.cpp
@@ -275,8 +275,8 @@ void ReportFile::Write(const char *buffer, uptr length) {
 
 bool GetCodeRangeForFile(const char *module, uptr *start, uptr *end) {
   MemoryMappingLayout proc_maps(/*cache_enabled*/false);
-  InternalScopedString buff(kMaxPathLength);
-  MemoryMappedSegment segment(buff.data(), kMaxPathLength);
+  InternalMmapVector<char> buff(kMaxPathLength);
+  MemoryMappedSegment segment(buff.data(), buff.size());
   while (proc_maps.Next(&segment)) {
     if (segment.IsExecutable() &&
         internal_strcmp(module, segment.filename) == 0) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_posix.h b/libsanitizer/sanitizer_common/sanitizer_posix.h
index e1a2b48..b65dae6 100644
--- a/libsanitizer/sanitizer_common/sanitizer_posix.h
+++ b/libsanitizer/sanitizer_common/sanitizer_posix.h
@@ -40,6 +40,10 @@ uptr internal_write(fd_t fd, const void *buf, uptr count);
 uptr internal_mmap(void *addr, uptr length, int prot, int flags,
                    int fd, u64 offset);
 uptr internal_munmap(void *addr, uptr length);
+#if SANITIZER_LINUX
+uptr internal_mremap(void *old_address, uptr old_size, uptr new_size, int flags,
+                     void *new_address);
+#endif
 int internal_mprotect(void *addr, uptr length, int prot);
 int internal_madvise(uptr addr, uptr length, int advice);
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
index 7ff48c3..d1d8e50 100644
--- a/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_posix_libcdep.cpp
@@ -143,7 +143,7 @@ void Abort() {
   if (GetHandleSignalMode(SIGABRT) != kHandleSignalNo) {
     struct sigaction sigact;
     internal_memset(&sigact, 0, sizeof(sigact));
-    sigact.sa_sigaction = (sa_sigaction_t)SIG_DFL;
+    sigact.sa_handler = SIG_DFL;
     internal_sigaction(SIGABRT, &sigact, nullptr);
   }
 #endif
diff --git a/libsanitizer/sanitizer_common/sanitizer_printf.cpp b/libsanitizer/sanitizer_common/sanitizer_printf.cpp
index a032787..5d16dfd 100644
--- a/libsanitizer/sanitizer_common/sanitizer_printf.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_printf.cpp
@@ -249,26 +249,21 @@ static void NOINLINE SharedPrintfCodeNoBuffer(bool append_pid,
                                               va_list args) {
   va_list args2;
   va_copy(args2, args);
-  const int kLen = 16 * 1024;
-  int needed_length;
+  InternalMmapVector<char> v;
+  int needed_length = 0;
   char *buffer = local_buffer;
   // First try to print a message using a local buffer, and then fall back to
   // mmaped buffer.
-  for (int use_mmap = 0; use_mmap < 2; use_mmap++) {
+  for (int use_mmap = 0;; use_mmap++) {
     if (use_mmap) {
       va_end(args);
       va_copy(args, args2);
-      buffer = (char*)MmapOrDie(kLen, "Report");
-      buffer_size = kLen;
+      v.resize(needed_length + 1);
+      buffer_size = v.capacity();
+      v.resize(buffer_size);
+      buffer = &v[0];
     }
     needed_length = 0;
-    // Check that data fits into the current buffer.
-#   define CHECK_NEEDED_LENGTH \
-      if (needed_length >= buffer_size) { \
-        if (!use_mmap) continue; \
-        RAW_CHECK_MSG(needed_length < kLen, \
-                      "Buffer in Report is too short!\n"); \
-      }
     // Fuchsia's logging infrastructure always keeps track of the logging
     // process, thread, and timestamp, so never prepend such information.
     if (!SANITIZER_FUCHSIA && append_pid) {
@@ -277,18 +272,20 @@ static void NOINLINE SharedPrintfCodeNoBuffer(bool append_pid,
       if (common_flags()->log_exe_name && exe_name) {
         needed_length += internal_snprintf(buffer, buffer_size,
                                            "==%s", exe_name);
-        CHECK_NEEDED_LENGTH
+        if (needed_length >= buffer_size)
+          continue;
       }
       needed_length += internal_snprintf(
           buffer + needed_length, buffer_size - needed_length, "==%d==", pid);
-      CHECK_NEEDED_LENGTH
+      if (needed_length >= buffer_size)
+        continue;
     }
     needed_length += VSNPrintf(buffer + needed_length,
                                buffer_size - needed_length, format, args);
-    CHECK_NEEDED_LENGTH
+    if (needed_length >= buffer_size)
+      continue;
     // If the message fit into the buffer, print it and exit.
     break;
-#   undef CHECK_NEEDED_LENGTH
   }
   RawWrite(buffer);
 
@@ -297,9 +294,6 @@ static void NOINLINE SharedPrintfCodeNoBuffer(bool append_pid,
   CallPrintfAndReportCallback(buffer);
   LogMessageOnPrintf(buffer);
 
-  // If we had mapped any memory, clean up.
-  if (buffer != local_buffer)
-    UnmapOrDie((void *)buffer, buffer_size);
   va_end(args2);
 }
 
@@ -346,13 +340,24 @@ int internal_snprintf(char *buffer, uptr length, const char *format, ...) {
 
 FORMAT(2, 3)
 void InternalScopedString::append(const char *format, ...) {
-  CHECK_LT(length_, size());
-  va_list args;
-  va_start(args, format);
-  VSNPrintf(data() + length_, size() - length_, format, args);
-  va_end(args);
-  length_ += internal_strlen(data() + length_);
-  CHECK_LT(length_, size());
+  uptr prev_len = length();
+
+  while (true) {
+    buffer_.resize(buffer_.capacity());
+
+    va_list args;
+    va_start(args, format);
+    uptr sz = VSNPrintf(buffer_.data() + prev_len, buffer_.size() - prev_len,
+                        format, args);
+    va_end(args);
+    if (sz < buffer_.size() - prev_len) {
+      buffer_.resize(prev_len + sz + 1);
+      break;
+    }
+
+    buffer_.reserve(buffer_.capacity() * 2);
+  }
+  CHECK_EQ(buffer_[length()], '\0');
 }
 
 } // namespace __sanitizer
diff --git a/libsanitizer/sanitizer_common/sanitizer_procmaps_common.cpp b/libsanitizer/sanitizer_common/sanitizer_procmaps_common.cpp
index f2cfcff..1b7dd46 100644
--- a/libsanitizer/sanitizer_common/sanitizer_procmaps_common.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_procmaps_common.cpp
@@ -120,7 +120,7 @@ void MemoryMappingLayout::LoadFromCache() {
 void MemoryMappingLayout::DumpListOfModules(
     InternalMmapVectorNoCtor<LoadedModule> *modules) {
   Reset();
-  InternalScopedString module_name(kMaxPathLength);
+  InternalMmapVector<char> module_name(kMaxPathLength);
   MemoryMappedSegment segment(module_name.data(), module_name.size());
   for (uptr i = 0; Next(&segment); i++) {
     const char *cur_name = segment.filename;
diff --git a/libsanitizer/sanitizer_common/sanitizer_procmaps_mac.cpp b/libsanitizer/sanitizer_common/sanitizer_procmaps_mac.cpp
index d02afcf..1f53e3e 100644
--- a/libsanitizer/sanitizer_common/sanitizer_procmaps_mac.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_procmaps_mac.cpp
@@ -354,8 +354,8 @@ bool MemoryMappingLayout::Next(MemoryMappedSegment *segment) {
 void MemoryMappingLayout::DumpListOfModules(
     InternalMmapVectorNoCtor<LoadedModule> *modules) {
   Reset();
-  InternalScopedString module_name(kMaxPathLength);
-  MemoryMappedSegment segment(module_name.data(), kMaxPathLength);
+  InternalMmapVector<char> module_name(kMaxPathLength);
+  MemoryMappedSegment segment(module_name.data(), module_name.size());
   MemoryMappedSegmentData data;
   segment.data_ = &data;
   while (Next(&segment)) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_procmaps_solaris.cpp b/libsanitizer/sanitizer_common/sanitizer_procmaps_solaris.cpp
index 4063ec8..bf813f23 100644
--- a/libsanitizer/sanitizer_common/sanitizer_procmaps_solaris.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_procmaps_solaris.cpp
@@ -9,13 +9,13 @@
 // Information about the process mappings (Solaris-specific parts).
 //===----------------------------------------------------------------------===//
 
+// Before Solaris 11.4, <procfs.h> doesn't work in a largefile environment.
+#undef _FILE_OFFSET_BITS
 #include "sanitizer_platform.h"
 #if SANITIZER_SOLARIS
 #include "sanitizer_common.h"
 #include "sanitizer_procmaps.h"
 
-// Before Solaris 11.4, <procfs.h> doesn't work in a largefile environment.
-#undef _FILE_OFFSET_BITS
 #include <procfs.h>
 #include <limits.h>
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_ptrauth.h b/libsanitizer/sanitizer_common/sanitizer_ptrauth.h
index a288068..5200354 100644
--- a/libsanitizer/sanitizer_common/sanitizer_ptrauth.h
+++ b/libsanitizer/sanitizer_common/sanitizer_ptrauth.h
@@ -11,6 +11,24 @@
 
 #if __has_feature(ptrauth_calls)
 #include <ptrauth.h>
+#elif defined(__ARM_FEATURE_PAC_DEFAULT) && !defined(__APPLE__)
+inline unsigned long ptrauth_strip(void* __value, unsigned int __key) {
+  // On the stack the link register is protected with Pointer
+  // Authentication Code when compiled with -mbranch-protection.
+  // Let's stripping the PAC unconditionally because xpaclri is in
+  // the NOP space so will do nothing when it is not enabled or not available.
+  unsigned long ret;
+  asm volatile(
+      "mov x30, %1\n\t"
+      "hint #7\n\t"  // xpaclri
+      "mov %0, x30\n\t"
+      : "=r"(ret)
+      : "r"(__value)
+      : "x30");
+  return ret;
+}
+#define ptrauth_auth_data(__value, __old_key, __old_data) __value
+#define ptrauth_string_discriminator(__string) ((int)0)
 #else
 // Copied from <ptrauth.h>
 #define ptrauth_strip(__value, __key) __value
@@ -18,6 +36,6 @@
 #define ptrauth_string_discriminator(__string) ((int)0)
 #endif
 
-#define STRIP_PC(pc) ((uptr)ptrauth_strip(pc, 0))
+#define STRIP_PAC_PC(pc) ((uptr)ptrauth_strip(pc, 0))
 
 #endif // SANITIZER_PTRAUTH_H
diff --git a/libsanitizer/sanitizer_common/sanitizer_stackdepot.cpp b/libsanitizer/sanitizer_common/sanitizer_stackdepot.cpp
index 4692f50..44a9521 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stackdepot.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stackdepot.cpp
@@ -145,8 +145,7 @@ StackTrace StackDepotReverseMap::Get(u32 id) {
   if (!map_.size())
     return StackTrace();
   IdDescPair pair = {id, nullptr};
-  uptr idx =
-      InternalLowerBound(map_, 0, map_.size(), pair, IdDescPair::IdComparator);
+  uptr idx = InternalLowerBound(map_, pair, IdDescPair::IdComparator);
   if (idx > map_.size() || map_[idx].id != id)
     return StackTrace();
   return map_[idx].desc->load();
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp b/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
index b28fc1c..07e4409 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace.cpp
@@ -15,6 +15,7 @@
 #include "sanitizer_common.h"
 #include "sanitizer_flags.h"
 #include "sanitizer_platform.h"
+#include "sanitizer_ptrauth.h"
 
 namespace __sanitizer {
 
@@ -84,8 +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
@@ -108,28 +109,21 @@ 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)
     // frame[-1] contains the return address
     uhwptr pc1 = frame[-1];
 #else
-    uhwptr pc1 = frame[1];
+    uhwptr pc1 = STRIP_PAC_PC((void *)frame[1]);
 #endif
     // Let's assume that any pointer in the 0th page (i.e. <0x1000 on i386 and
     // x86_64) is invalid and stop unwinding here.  If we're adding support for
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace.h b/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
index 0350fe8..15616f8 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace.h
@@ -67,8 +67,6 @@ struct StackTrace {
   static uptr GetCurrentPc();
   static inline uptr GetPreviousInstructionPc(uptr pc);
   static uptr GetNextInstructionPc(uptr pc);
-  typedef bool (*SymbolizeCallback)(const void *pc, char *out_buffer,
-                                    int out_size);
 };
 
 // Performance-critical, must be in the header.
diff --git a/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
index 7808ba9..7386332 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stacktrace_libcdep.cpp
@@ -23,8 +23,8 @@ void StackTrace::Print() const {
     Printf("    <empty stack>\n\n");
     return;
   }
-  InternalScopedString frame_desc(GetPageSizeCached() * 2);
-  InternalScopedString dedup_token(GetPageSizeCached());
+  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;
@@ -125,7 +125,7 @@ void __sanitizer_symbolize_pc(uptr pc, const char *fmt, char *out_buf,
     out_buf[out_buf_size - 1] = 0;
     return;
   }
-  InternalScopedString frame_desc(GetPageSizeCached());
+  InternalScopedString frame_desc;
   uptr frame_num = 0;
   // Reserve one byte for the final 0.
   char *out_end = out_buf + out_buf_size - 1;
@@ -156,7 +156,7 @@ void __sanitizer_symbolize_global(uptr data_addr, const char *fmt,
   out_buf[0] = 0;
   DataInfo DI;
   if (!Symbolizer::GetOrInit()->SymbolizeData(data_addr, &DI)) return;
-  InternalScopedString data_desc(GetPageSizeCached());
+  InternalScopedString data_desc;
   RenderData(&data_desc, fmt, &DI, common_flags()->strip_path_prefix);
   internal_strncpy(out_buf, data_desc.data(), out_buf_size);
   out_buf[out_buf_size - 1] = 0;
diff --git a/libsanitizer/sanitizer_common/sanitizer_stoptheworld_linux_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_stoptheworld_linux_libcdep.cpp
index 0f1cadf..53cfddc 100644
--- a/libsanitizer/sanitizer_common/sanitizer_stoptheworld_linux_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_stoptheworld_linux_libcdep.cpp
@@ -490,6 +490,9 @@ typedef user_regs_struct regs_struct;
 #ifndef NT_X86_XSTATE
 #define NT_X86_XSTATE 0x202
 #endif
+#ifndef PTRACE_GETREGSET
+#define PTRACE_GETREGSET 0x4204
+#endif
 // Compiler may use FP registers to store pointers.
 static constexpr uptr kExtraRegs[] = {NT_X86_XSTATE, NT_FPREGSET};
 
@@ -513,6 +516,8 @@ static constexpr uptr kExtraRegs[] = {0};
 
 #elif SANITIZER_RISCV64
 typedef struct user_regs_struct regs_struct;
+// sys/ucontext.h already defines REG_SP as 2. Undefine it first.
+#undef REG_SP
 #define REG_SP sp
 static constexpr uptr kExtraRegs[] = {0};
 #define ARCH_IOVEC_FOR_GETREGSET
diff --git a/libsanitizer/sanitizer_common/sanitizer_suppressions.cpp b/libsanitizer/sanitizer_common/sanitizer_suppressions.cpp
index 44c83a6..a674034 100644
--- a/libsanitizer/sanitizer_common/sanitizer_suppressions.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_suppressions.cpp
@@ -34,7 +34,7 @@ SuppressionContext::SuppressionContext(const char *suppression_types[],
 static bool GetPathAssumingFileIsRelativeToExec(const char *file_path,
                                                 /*out*/char *new_file_path,
                                                 uptr new_file_path_size) {
-  InternalScopedString exec(kMaxPathLength);
+  InternalMmapVector<char> exec(kMaxPathLength);
   if (ReadBinaryNameCached(exec.data(), exec.size())) {
     const char *file_name_pos = StripModuleName(exec.data());
     uptr path_to_exec_len = file_name_pos - exec.data();
@@ -69,7 +69,7 @@ void SuppressionContext::ParseFromFile(const char *filename) {
   if (filename[0] == '\0')
     return;
 
-  InternalScopedString new_file_path(kMaxPathLength);
+  InternalMmapVector<char> new_file_path(kMaxPathLength);
   filename = FindFile(filename, new_file_path.data(), new_file_path.size());
 
   // Read the file.
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_libcdep.cpp
index 710da4c..98418b4 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_libcdep.cpp
@@ -356,7 +356,7 @@ void ParseSymbolizePCOutput(const char *str, SymbolizedStack *res) {
       InternalFree(info->function);
       info->function = 0;
     }
-    if (0 == internal_strcmp(info->file, "??")) {
+    if (info->file && 0 == internal_strcmp(info->file, "??")) {
       InternalFree(info->file);
       info->file = 0;
     }
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
index 30cba08..01edef9 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_markup.cpp
@@ -54,6 +54,10 @@ bool Symbolizer::GetModuleNameAndOffsetForPC(uptr pc, const char **module_name,
   return false;
 }
 
+// This is mainly used by hwasan for online symbolization. This isn't needed
+// since hwasan can always just dump stack frames for offline symbolization.
+bool Symbolizer::SymbolizeFrame(uptr addr, FrameInfo *info) { return false; }
+
 // This is used in some places for suppression checking, which we
 // don't really support for Fuchsia.  It's also used in UBSan to
 // identify a PC location to a function name, so we always fill in
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cpp
index 4dd5cc3..4cd4b46 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cpp
@@ -400,11 +400,20 @@ const char *Symbolizer::PlatformDemangle(const char *name) {
 
 static SymbolizerTool *ChooseExternalSymbolizer(LowLevelAllocator *allocator) {
   const char *path = common_flags()->external_symbolizer_path;
+
+  if (path && internal_strchr(path, '%')) {
+    char *new_path = (char *)InternalAlloc(kMaxPathLength);
+    SubstituteForFlagValue(path, new_path, kMaxPathLength);
+    path = new_path;
+  }
+
   const char *binary_name = path ? StripModuleName(path) : "";
+  static const char kLLVMSymbolizerPrefix[] = "llvm-symbolizer";
   if (path && path[0] == '\0') {
     VReport(2, "External symbolizer is explicitly disabled.\n");
     return nullptr;
-  } else if (!internal_strcmp(binary_name, "llvm-symbolizer")) {
+  } else if (!internal_strncmp(binary_name, kLLVMSymbolizerPrefix,
+                               internal_strlen(kLLVMSymbolizerPrefix))) {
     VReport(2, "Using llvm-symbolizer at user-specified path: %s\n", path);
     return new(*allocator) LLVMSymbolizer(path, allocator);
   } else if (!internal_strcmp(binary_name, "atos")) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
index 06301b8..9287993 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_report.cpp
@@ -31,7 +31,7 @@ namespace __sanitizer {
 void ReportErrorSummary(const char *error_type, const AddressInfo &info,
                         const char *alt_tool_name) {
   if (!common_flags()->print_summary) return;
-  InternalScopedString buff(kMaxSummaryLength);
+  InternalScopedString buff;
   buff.append("%s ", error_type);
   RenderFrame(&buff, "%L %F", 0, info.address, &info,
               common_flags()->symbolize_vs_style,
@@ -150,7 +150,7 @@ static void PrintMemoryByte(InternalScopedString *str, const char *before,
 static void MaybeDumpInstructionBytes(uptr pc) {
   if (!common_flags()->dump_instruction_bytes || (pc < GetPageSizeCached()))
     return;
-  InternalScopedString str(1024);
+  InternalScopedString str;
   str.append("First 16 instruction bytes at pc: ");
   if (IsAccessibleMemoryRange(pc, 16)) {
     for (int i = 0; i < 16; ++i) {
@@ -211,7 +211,7 @@ static void ReportDeadlySignalImpl(const SignalContext &sig, u32 tid,
     Report("The signal is caused by a %s memory access.\n", access_type);
     if (!sig.is_true_faulting_addr)
       Report("Hint: this fault was caused by a dereference of a high value "
-             "address (see register values below).  Dissassemble the provided "
+             "address (see register values below).  Disassemble the provided "
              "pc to learn which register was used.\n");
     else if (sig.addr < GetPageSizeCached())
       Report("Hint: address points to the zero page.\n");
diff --git a/libsanitizer/sanitizer_common/sanitizer_symbolizer_win.cpp b/libsanitizer/sanitizer_common/sanitizer_symbolizer_win.cpp
index 48fa2d1..702d901 100644
--- a/libsanitizer/sanitizer_common/sanitizer_symbolizer_win.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_symbolizer_win.cpp
@@ -136,9 +136,10 @@ void InitializeDbgHelpIfNeeded() {
 bool WinSymbolizerTool::SymbolizePC(uptr addr, SymbolizedStack *frame) {
   InitializeDbgHelpIfNeeded();
 
-  // See http://msdn.microsoft.com/en-us/library/ms680578(VS.85).aspx
-  char buffer[sizeof(SYMBOL_INFO) + MAX_SYM_NAME * sizeof(CHAR)];
-  PSYMBOL_INFO symbol = (PSYMBOL_INFO)buffer;
+  // See https://docs.microsoft.com/en-us/windows/win32/debug/retrieving-symbol-information-by-address
+  InternalMmapVector<char> buffer(sizeof(SYMBOL_INFO) +
+                                  MAX_SYM_NAME * sizeof(CHAR));
+  PSYMBOL_INFO symbol = (PSYMBOL_INFO)&buffer[0];
   symbol->SizeOfStruct = sizeof(SYMBOL_INFO);
   symbol->MaxNameLen = MAX_SYM_NAME;
   DWORD64 offset = 0;
@@ -223,7 +224,7 @@ bool SymbolizerProcess::StartSymbolizerSubprocess() {
   // Compute the command line. Wrap double quotes around everything.
   const char *argv[kArgVMax];
   GetArgV(path_, argv);
-  InternalScopedString command_line(kMaxPathLength * 3);
+  InternalScopedString command_line;
   for (int i = 0; argv[i]; i++) {
     const char *arg = argv[i];
     int arglen = internal_strlen(arg);
@@ -281,8 +282,15 @@ static void ChooseSymbolizerTools(IntrusiveList<SymbolizerTool> *list,
     return;
   }
 
-  // Add llvm-symbolizer in case the binary has dwarf.
+  // Add llvm-symbolizer.
   const char *user_path = common_flags()->external_symbolizer_path;
+
+  if (user_path && internal_strchr(user_path, '%')) {
+    char *new_path = (char *)InternalAlloc(kMaxPathLength);
+    SubstituteForFlagValue(user_path, new_path, kMaxPathLength);
+    user_path = new_path;
+  }
+
   const char *path =
       user_path ? user_path : FindPathToBinary("llvm-symbolizer.exe");
   if (path) {
diff --git a/libsanitizer/sanitizer_common/sanitizer_termination.cpp b/libsanitizer/sanitizer_common/sanitizer_termination.cpp
index 84be6fc..6a54734 100644
--- a/libsanitizer/sanitizer_common/sanitizer_termination.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_termination.cpp
@@ -59,26 +59,31 @@ void NORETURN Die() {
   internal__exit(common_flags()->exitcode);
 }
 
-static CheckFailedCallbackType CheckFailedCallback;
-void SetCheckFailedCallback(CheckFailedCallbackType callback) {
-  CheckFailedCallback = callback;
+static void (*CheckUnwindCallback)();
+void SetCheckUnwindCallback(void (*callback)()) {
+  CheckUnwindCallback = callback;
 }
 
-const int kSecondsToSleepWhenRecursiveCheckFailed = 2;
-
 void NORETURN CheckFailed(const char *file, int line, const char *cond,
                           u64 v1, u64 v2) {
-  static atomic_uint32_t num_calls;
-  if (atomic_fetch_add(&num_calls, 1, memory_order_relaxed) > 10) {
-    SleepForSeconds(kSecondsToSleepWhenRecursiveCheckFailed);
+  u32 tid = GetTid();
+  Printf("%s: CHECK failed: %s:%d \"%s\" (0x%zx, 0x%zx) (tid=%u)\n",
+         SanitizerToolName, StripModuleName(file), line, cond, (uptr)v1,
+         (uptr)v2, tid);
+  static atomic_uint32_t first_tid;
+  u32 cmp = 0;
+  if (!atomic_compare_exchange_strong(&first_tid, &cmp, tid,
+                                      memory_order_relaxed)) {
+    if (cmp == tid) {
+      // Recursing into CheckFailed.
+    } else {
+      // Another thread fails already, let it print the stack and terminate.
+      SleepForSeconds(2);
+    }
     Trap();
   }
-
-  if (CheckFailedCallback) {
-    CheckFailedCallback(file, line, cond, v1, v2);
-  }
-  Report("Sanitizer CHECK failed: %s:%d %s (%lld, %lld)\n", file, line, cond,
-                                                            v1, v2);
+  if (CheckUnwindCallback)
+    CheckUnwindCallback();
   Die();
 }
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp b/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
index f2c6f27..3273da3 100644
--- a/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_thread_registry.cpp
@@ -85,7 +85,7 @@ void ThreadContextBase::SetCreated(uptr _user_id, u64 _unique_id,
   unique_id = _unique_id;
   detached = _detached;
   // Parent tid makes no sense for the main thread.
-  if (tid != 0)
+  if (tid != kMainTid)
     parent_tid = _parent_tid;
   OnCreated(arg);
 }
@@ -99,8 +99,6 @@ void ThreadContextBase::Reset() {
 
 // ThreadRegistry implementation.
 
-const u32 ThreadRegistry::kUnknownTid = ~0U;
-
 ThreadRegistry::ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
                                u32 thread_quarantine_size, u32 max_reuse)
     : context_factory_(factory),
@@ -135,7 +133,7 @@ uptr ThreadRegistry::GetMaxAliveThreads() {
 u32 ThreadRegistry::CreateThread(uptr user_id, bool detached, u32 parent_tid,
                                  void *arg) {
   BlockingMutexLock l(&mtx_);
-  u32 tid = kUnknownTid;
+  u32 tid = kInvalidTid;
   ThreadContextBase *tctx = QuarantinePop();
   if (tctx) {
     tid = tctx->tid;
@@ -155,7 +153,7 @@ u32 ThreadRegistry::CreateThread(uptr user_id, bool detached, u32 parent_tid,
     Die();
   }
   CHECK_NE(tctx, 0);
-  CHECK_NE(tid, kUnknownTid);
+  CHECK_NE(tid, kInvalidTid);
   CHECK_LT(tid, max_threads_);
   CHECK_EQ(tctx->status, ThreadStatusInvalid);
   alive_threads_++;
@@ -186,7 +184,7 @@ u32 ThreadRegistry::FindThread(FindThreadCallback cb, void *arg) {
     if (tctx != 0 && cb(tctx, arg))
       return tctx->tid;
   }
-  return kUnknownTid;
+  return kInvalidTid;
 }
 
 ThreadContextBase *
@@ -278,7 +276,7 @@ void ThreadRegistry::JoinThread(u32 tid, void *arg) {
 // really started.  We just did CreateThread for a prospective new
 // thread before trying to create it, and then failed to actually
 // create it, and so never called StartThread.
-void ThreadRegistry::FinishThread(u32 tid) {
+ThreadStatus ThreadRegistry::FinishThread(u32 tid) {
   BlockingMutexLock l(&mtx_);
   CHECK_GT(alive_threads_, 0);
   alive_threads_--;
@@ -286,6 +284,7 @@ void ThreadRegistry::FinishThread(u32 tid) {
   ThreadContextBase *tctx = threads_[tid];
   CHECK_NE(tctx, 0);
   bool dead = tctx->detached;
+  ThreadStatus prev_status = tctx->status;
   if (tctx->status == ThreadStatusRunning) {
     CHECK_GT(running_threads_, 0);
     running_threads_--;
@@ -300,6 +299,7 @@ void ThreadRegistry::FinishThread(u32 tid) {
     QuarantinePush(tctx);
   }
   tctx->SetDestroyed();
+  return prev_status;
 }
 
 void ThreadRegistry::StartThread(u32 tid, tid_t os_id, ThreadType thread_type,
diff --git a/libsanitizer/sanitizer_common/sanitizer_thread_registry.h b/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
index 85c522a..dcd445c 100644
--- a/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
+++ b/libsanitizer/sanitizer_common/sanitizer_thread_registry.h
@@ -87,8 +87,6 @@ typedef ThreadContextBase* (*ThreadContextFactory)(u32 tid);
 
 class ThreadRegistry {
  public:
-  static const u32 kUnknownTid;
-
   ThreadRegistry(ThreadContextFactory factory, u32 max_threads,
                  u32 thread_quarantine_size, u32 max_reuse = 0);
   void GetNumberOfThreads(uptr *total = nullptr, uptr *running = nullptr,
@@ -113,7 +111,7 @@ class ThreadRegistry {
   void RunCallbackForEachThreadLocked(ThreadCallback cb, void *arg);
 
   typedef bool (*FindThreadCallback)(ThreadContextBase *tctx, void *arg);
-  // Finds a thread using the provided callback. Returns kUnknownTid if no
+  // Finds a thread using the provided callback. Returns kInvalidTid if no
   // thread is found.
   u32 FindThread(FindThreadCallback cb, void *arg);
   // Should be guarded by ThreadRegistryLock. Return 0 if no thread
@@ -126,7 +124,8 @@ class ThreadRegistry {
   void SetThreadNameByUserId(uptr user_id, const char *name);
   void DetachThread(u32 tid, void *arg);
   void JoinThread(u32 tid, void *arg);
-  void FinishThread(u32 tid);
+  // Finishes thread and returns previous status.
+  ThreadStatus FinishThread(u32 tid);
   void StartThread(u32 tid, tid_t os_id, ThreadType thread_type, void *arg);
   void SetThreadUserId(u32 tid, uptr user_id);
 
diff --git a/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.cpp b/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.cpp
index 10748f9..1f664b6 100644
--- a/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.cpp
@@ -12,6 +12,7 @@
 
 #include "sanitizer_tls_get_addr.h"
 
+#include "sanitizer_atomic.h"
 #include "sanitizer_flags.h"
 #include "sanitizer_platform_interceptors.h"
 
@@ -42,39 +43,54 @@ static atomic_uintptr_t number_of_live_dtls;
 
 static const uptr kDestroyedThread = -1;
 
-static inline void DTLS_Deallocate(DTLS::DTV *dtv, uptr size) {
-  if (!size) return;
-  VReport(2, "__tls_get_addr: DTLS_Deallocate %p %zd\n", dtv, size);
-  UnmapOrDie(dtv, size * sizeof(DTLS::DTV));
+static void DTLS_Deallocate(DTLS::DTVBlock *block) {
+  VReport(2, "__tls_get_addr: DTLS_Deallocate %p %zd\n", block);
+  UnmapOrDie(block, sizeof(DTLS::DTVBlock));
   atomic_fetch_sub(&number_of_live_dtls, 1, memory_order_relaxed);
 }
 
-static inline void DTLS_Resize(uptr new_size) {
-  if (dtls.dtv_size >= new_size) return;
-  new_size = RoundUpToPowerOfTwo(new_size);
-  new_size = Max(new_size, 4096UL / sizeof(DTLS::DTV));
-  DTLS::DTV *new_dtv =
-      (DTLS::DTV *)MmapOrDie(new_size * sizeof(DTLS::DTV), "DTLS_Resize");
+static DTLS::DTVBlock *DTLS_NextBlock(atomic_uintptr_t *cur) {
+  uptr v = atomic_load(cur, memory_order_acquire);
+  if (v == kDestroyedThread)
+    return nullptr;
+  DTLS::DTVBlock *next = (DTLS::DTVBlock *)v;
+  if (next)
+    return next;
+  DTLS::DTVBlock *new_dtv =
+      (DTLS::DTVBlock *)MmapOrDie(sizeof(DTLS::DTVBlock), "DTLS_NextBlock");
+  uptr prev = 0;
+  if (!atomic_compare_exchange_strong(cur, &prev, (uptr)new_dtv,
+                                      memory_order_seq_cst)) {
+    UnmapOrDie(new_dtv, sizeof(DTLS::DTVBlock));
+    return (DTLS::DTVBlock *)prev;
+  }
   uptr num_live_dtls =
       atomic_fetch_add(&number_of_live_dtls, 1, memory_order_relaxed);
-  VReport(2, "__tls_get_addr: DTLS_Resize %p %zd\n", &dtls, num_live_dtls);
-  CHECK_LT(num_live_dtls, 1 << 20);
-  uptr old_dtv_size = dtls.dtv_size;
-  DTLS::DTV *old_dtv = dtls.dtv;
-  if (old_dtv_size)
-    internal_memcpy(new_dtv, dtls.dtv, dtls.dtv_size * sizeof(DTLS::DTV));
-  dtls.dtv = new_dtv;
-  dtls.dtv_size = new_size;
-  if (old_dtv_size)
-    DTLS_Deallocate(old_dtv, old_dtv_size);
+  VReport(2, "__tls_get_addr: DTLS_NextBlock %p %zd\n", &dtls, num_live_dtls);
+  return new_dtv;
+}
+
+static DTLS::DTV *DTLS_Find(uptr id) {
+  VReport(2, "__tls_get_addr: DTLS_Find %p %zd\n", &dtls, id);
+  static constexpr uptr kPerBlock = ARRAY_SIZE(DTLS::DTVBlock::dtvs);
+  DTLS::DTVBlock *cur = DTLS_NextBlock(&dtls.dtv_block);
+  if (!cur)
+    return nullptr;
+  for (; id >= kPerBlock; id -= kPerBlock) cur = DTLS_NextBlock(&cur->next);
+  return cur->dtvs + id;
 }
 
 void DTLS_Destroy() {
   if (!common_flags()->intercept_tls_get_addr) return;
-  VReport(2, "__tls_get_addr: DTLS_Destroy %p %zd\n", &dtls, dtls.dtv_size);
-  uptr s = dtls.dtv_size;
-  dtls.dtv_size = kDestroyedThread;  // Do this before unmap for AS-safety.
-  DTLS_Deallocate(dtls.dtv, s);
+  VReport(2, "__tls_get_addr: DTLS_Destroy %p\n", &dtls);
+  DTLS::DTVBlock *block = (DTLS::DTVBlock *)atomic_exchange(
+      &dtls.dtv_block, kDestroyedThread, memory_order_release);
+  while (block) {
+    DTLS::DTVBlock *next =
+        (DTLS::DTVBlock *)atomic_load(&block->next, memory_order_acquire);
+    DTLS_Deallocate(block);
+    block = next;
+  }
 }
 
 #if defined(__powerpc64__) || defined(__mips__)
@@ -96,9 +112,9 @@ DTLS::DTV *DTLS_on_tls_get_addr(void *arg_void, void *res,
   if (!common_flags()->intercept_tls_get_addr) return 0;
   TlsGetAddrParam *arg = reinterpret_cast<TlsGetAddrParam *>(arg_void);
   uptr dso_id = arg->dso_id;
-  if (dtls.dtv_size == kDestroyedThread) return 0;
-  DTLS_Resize(dso_id + 1);
-  if (dtls.dtv[dso_id].beg) return 0;
+  DTLS::DTV *dtv = DTLS_Find(dso_id);
+  if (!dtv || dtv->beg)
+    return 0;
   uptr tls_size = 0;
   uptr tls_beg = reinterpret_cast<uptr>(res) - arg->offset - kDtvOffset;
   VReport(2, "__tls_get_addr: %p {%p,%p} => %p; tls_beg: %p; sp: %p "
@@ -126,9 +142,9 @@ DTLS::DTV *DTLS_on_tls_get_addr(void *arg_void, void *res,
     // This may happen inside the DTOR of main thread, so just ignore it.
     tls_size = 0;
   }
-  dtls.dtv[dso_id].beg = tls_beg;
-  dtls.dtv[dso_id].size = tls_size;
-  return dtls.dtv + dso_id;
+  dtv->beg = tls_beg;
+  dtv->size = tls_size;
+  return dtv;
 }
 
 void DTLS_on_libc_memalign(void *ptr, uptr size) {
@@ -141,7 +157,8 @@ void DTLS_on_libc_memalign(void *ptr, uptr size) {
 DTLS *DTLS_Get() { return &dtls; }
 
 bool DTLSInDestruction(DTLS *dtls) {
-  return dtls->dtv_size == kDestroyedThread;
+  return atomic_load(&dtls->dtv_block, memory_order_relaxed) ==
+         kDestroyedThread;
 }
 
 #else
diff --git a/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.h b/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.h
index c7cd5a8..a599c0b 100644
--- a/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.h
+++ b/libsanitizer/sanitizer_common/sanitizer_tls_get_addr.h
@@ -28,6 +28,7 @@
 #ifndef SANITIZER_TLS_GET_ADDR_H
 #define SANITIZER_TLS_GET_ADDR_H
 
+#include "sanitizer_atomic.h"
 #include "sanitizer_common.h"
 
 namespace __sanitizer {
@@ -38,15 +39,31 @@ struct DTLS {
   struct DTV {
     uptr beg, size;
   };
+  struct DTVBlock {
+    atomic_uintptr_t next;
+    DTV dtvs[(4096UL - sizeof(next)) / sizeof(DTLS::DTV)];
+  };
+
+  static_assert(sizeof(DTVBlock) <= 4096UL, "Unexpected block size");
 
-  uptr dtv_size;
-  DTV *dtv;  // dtv_size elements, allocated by MmapOrDie.
+  atomic_uintptr_t dtv_block;
 
   // Auxiliary fields, don't access them outside sanitizer_tls_get_addr.cpp
   uptr last_memalign_size;
   uptr last_memalign_ptr;
 };
 
+template <typename Fn>
+void ForEachDVT(DTLS *dtls, const Fn &fn) {
+  DTLS::DTVBlock *block =
+      (DTLS::DTVBlock *)atomic_load(&dtls->dtv_block, memory_order_acquire);
+  while (block) {
+    int id = 0;
+    for (auto &d : block->dtvs) fn(d, id++);
+    block = (DTLS::DTVBlock *)atomic_load(&block->next, memory_order_acquire);
+  }
+}
+
 // Returns pointer and size of a linker-allocated TLS block.
 // Each block is returned exactly once.
 DTLS::DTV *DTLS_on_tls_get_addr(void *arg, void *res, uptr static_tls_begin,
diff --git a/libsanitizer/sanitizer_common/sanitizer_unwind_win.cpp b/libsanitizer/sanitizer_common/sanitizer_unwind_win.cpp
index e2edf42..7e01c81 100644
--- a/libsanitizer/sanitizer_common/sanitizer_unwind_win.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_unwind_win.cpp
@@ -43,6 +43,10 @@ void BufferedStackTrace::UnwindSlow(uptr pc, u32 max_depth) {
   trace_buffer[0] = pc;
 }
 
+#ifdef __clang__
+#pragma clang diagnostic push
+#pragma clang diagnostic ignored "-Wframe-larger-than="
+#endif
 void BufferedStackTrace::UnwindSlow(uptr pc, void *context, u32 max_depth) {
   CHECK(context);
   CHECK_GE(max_depth, 2);
@@ -74,6 +78,9 @@ void BufferedStackTrace::UnwindSlow(uptr pc, void *context, u32 max_depth) {
     trace_buffer[size++] = (uptr)stack_frame.AddrPC.Offset;
   }
 }
+#ifdef __clang__
+#pragma clang diagnostic pop
+#endif
 #endif  // #if !SANITIZER_GO
 
 #endif  // SANITIZER_WINDOWS
diff --git a/libsanitizer/sanitizer_common/sanitizer_win.cpp b/libsanitizer/sanitizer_common/sanitizer_win.cpp
index 85ac263..f383e13 100644
--- a/libsanitizer/sanitizer_common/sanitizer_win.cpp
+++ b/libsanitizer/sanitizer_common/sanitizer_win.cpp
@@ -334,8 +334,12 @@ bool MprotectNoAccess(uptr addr, uptr size) {
 }
 
 void ReleaseMemoryPagesToOS(uptr beg, uptr end) {
-  // This is almost useless on 32-bits.
-  // FIXME: add madvise-analog when we move to 64-bits.
+  uptr beg_aligned = RoundDownTo(beg, GetPageSizeCached()),
+       end_aligned = RoundDownTo(end, GetPageSizeCached());
+  CHECK(beg < end);                // make sure the region is sane
+  if (beg_aligned == end_aligned)  // make sure we're freeing at least 1 page;
+    return;
+  UnmapOrDie((void *)beg, end_aligned - beg_aligned);
 }
 
 void SetShadowRegionHugePageMode(uptr addr, uptr size) {
@@ -386,6 +390,12 @@ uptr FindAvailableMemoryRange(uptr size, uptr alignment, uptr left_padding,
   return 0;
 }
 
+uptr MapDynamicShadowAndAliases(uptr shadow_size, uptr alias_size,
+                                uptr num_aliases, uptr ring_buffer_size) {
+  CHECK(false && "HWASan aliasing is unimplemented on Windows");
+  return 0;
+}
+
 bool MemoryRangeIsAvailable(uptr range_start, uptr range_end) {
   MEMORY_BASIC_INFORMATION mbi;
   CHECK(VirtualQuery((void *)range_start, &mbi, sizeof(mbi)));
@@ -564,7 +574,7 @@ void Abort() {
 // load the image at this address. Therefore, we call it the preferred base. Any
 // addresses in the DWARF typically assume that the object has been loaded at
 // this address.
-static uptr GetPreferredBase(const char *modname) {
+static uptr GetPreferredBase(const char *modname, char *buf, size_t buf_size) {
   fd_t fd = OpenFile(modname, RdOnly, nullptr);
   if (fd == kInvalidFd)
     return 0;
@@ -586,12 +596,10 @@ static uptr GetPreferredBase(const char *modname) {
   // IMAGE_FILE_HEADER
   // IMAGE_OPTIONAL_HEADER
   // Seek to e_lfanew and read all that data.
-  char buf[4 + sizeof(IMAGE_FILE_HEADER) + sizeof(IMAGE_OPTIONAL_HEADER)];
   if (::SetFilePointer(fd, dos_header.e_lfanew, nullptr, FILE_BEGIN) ==
       INVALID_SET_FILE_POINTER)
     return 0;
-  if (!ReadFromFile(fd, &buf[0], sizeof(buf), &bytes_read) ||
-      bytes_read != sizeof(buf))
+  if (!ReadFromFile(fd, buf, buf_size, &bytes_read) || bytes_read != buf_size)
     return 0;
 
   // Check for "PE\0\0" before the PE header.
@@ -633,6 +641,10 @@ void ListOfModules::init() {
     }
   }
 
+  InternalMmapVector<char> buf(4 + sizeof(IMAGE_FILE_HEADER) +
+                               sizeof(IMAGE_OPTIONAL_HEADER));
+  InternalMmapVector<wchar_t> modname_utf16(kMaxPathLength);
+  InternalMmapVector<char> module_name(kMaxPathLength);
   // |num_modules| is the number of modules actually present,
   size_t num_modules = bytes_required / sizeof(HMODULE);
   for (size_t i = 0; i < num_modules; ++i) {
@@ -642,15 +654,13 @@ void ListOfModules::init() {
       continue;
 
     // Get the UTF-16 path and convert to UTF-8.
-    wchar_t modname_utf16[kMaxPathLength];
     int modname_utf16_len =
-        GetModuleFileNameW(handle, modname_utf16, kMaxPathLength);
+        GetModuleFileNameW(handle, &modname_utf16[0], kMaxPathLength);
     if (modname_utf16_len == 0)
       modname_utf16[0] = '\0';
-    char module_name[kMaxPathLength];
-    int module_name_len =
-        ::WideCharToMultiByte(CP_UTF8, 0, modname_utf16, modname_utf16_len + 1,
-                              &module_name[0], kMaxPathLength, NULL, NULL);
+    int module_name_len = ::WideCharToMultiByte(
+        CP_UTF8, 0, &modname_utf16[0], modname_utf16_len + 1, &module_name[0],
+        kMaxPathLength, NULL, NULL);
     module_name[module_name_len] = '\0';
 
     uptr base_address = (uptr)mi.lpBaseOfDll;
@@ -660,15 +670,16 @@ void ListOfModules::init() {
     // RVA when computing the module offset. This helps llvm-symbolizer find the
     // right DWARF CU. In the common case that the image is loaded at it's
     // preferred address, we will now print normal virtual addresses.
-    uptr preferred_base = GetPreferredBase(&module_name[0]);
+    uptr preferred_base =
+        GetPreferredBase(&module_name[0], &buf[0], buf.size());
     uptr adjusted_base = base_address - preferred_base;
 
-    LoadedModule cur_module;
-    cur_module.set(module_name, adjusted_base);
+    modules_.push_back(LoadedModule());
+    LoadedModule &cur_module = modules_.back();
+    cur_module.set(&module_name[0], adjusted_base);
     // We add the whole module as one single address range.
     cur_module.addAddressRange(base_address, end_address, /*executable*/ true,
                                /*writable*/ true);
-    modules_.push_back(cur_module);
   }
   UnmapOrDie(hmodules, modules_buffer_size);
 }
@@ -956,22 +967,27 @@ void SignalContext::InitPcSpBp() {
 
 uptr SignalContext::GetAddress() const {
   EXCEPTION_RECORD *exception_record = (EXCEPTION_RECORD *)siginfo;
-  return exception_record->ExceptionInformation[1];
+  if (exception_record->ExceptionCode == EXCEPTION_ACCESS_VIOLATION)
+    return exception_record->ExceptionInformation[1];
+  return (uptr)exception_record->ExceptionAddress;
 }
 
 bool SignalContext::IsMemoryAccess() const {
-  return GetWriteFlag() != SignalContext::UNKNOWN;
+  return ((EXCEPTION_RECORD *)siginfo)->ExceptionCode ==
+         EXCEPTION_ACCESS_VIOLATION;
 }
 
-bool SignalContext::IsTrueFaultingAddress() const {
-  // FIXME: Provide real implementation for this. See Linux and Mac variants.
-  return IsMemoryAccess();
-}
+bool SignalContext::IsTrueFaultingAddress() const { return true; }
 
 SignalContext::WriteFlag SignalContext::GetWriteFlag() const {
   EXCEPTION_RECORD *exception_record = (EXCEPTION_RECORD *)siginfo;
+
+  // The write flag is only available for access violation exceptions.
+  if (exception_record->ExceptionCode != EXCEPTION_ACCESS_VIOLATION)
+    return SignalContext::UNKNOWN;
+
   // The contents of this array are documented at
-  // https://msdn.microsoft.com/en-us/library/windows/desktop/aa363082(v=vs.85).aspx
+  // https://docs.microsoft.com/en-us/windows/win32/api/winnt/ns-winnt-exception_record
   // The first element indicates read as 0, write as 1, or execute as 8.  The
   // second element is the faulting address.
   switch (exception_record->ExceptionInformation[0]) {
@@ -1037,10 +1053,24 @@ const char *SignalContext::Describe() const {
 }
 
 uptr ReadBinaryName(/*out*/char *buf, uptr buf_len) {
-  // FIXME: Actually implement this function.
-  CHECK_GT(buf_len, 0);
-  buf[0] = 0;
-  return 0;
+  if (buf_len == 0)
+    return 0;
+
+  // Get the UTF-16 path and convert to UTF-8.
+  InternalMmapVector<wchar_t> binname_utf16(kMaxPathLength);
+  int binname_utf16_len =
+      GetModuleFileNameW(NULL, &binname_utf16[0], kMaxPathLength);
+  if (binname_utf16_len == 0) {
+    buf[0] = '\0';
+    return 0;
+  }
+  int binary_name_len =
+      ::WideCharToMultiByte(CP_UTF8, 0, &binname_utf16[0], binname_utf16_len,
+                            buf, buf_len, NULL, NULL);
+  if ((unsigned)binary_name_len == buf_len)
+    --binary_name_len;
+  buf[binary_name_len] = '\0';
+  return binary_name_len;
 }
 
 uptr ReadLongProcessName(/*out*/char *buf, uptr buf_len) {
diff --git a/libsanitizer/tsan/tsan_clock.cpp b/libsanitizer/tsan/tsan_clock.cpp
index c91b29c..8e51883 100644
--- a/libsanitizer/tsan/tsan_clock.cpp
+++ b/libsanitizer/tsan/tsan_clock.cpp
@@ -150,7 +150,7 @@ void ThreadClock::acquire(ClockCache *c, SyncClock *src) {
   bool acquired = false;
   for (unsigned i = 0; i < kDirtyTids; i++) {
     SyncClock::Dirty dirty = src->dirty_[i];
-    unsigned tid = dirty.tid;
+    unsigned tid = dirty.tid();
     if (tid != kInvalidTid) {
       if (clk_[tid] < dirty.epoch) {
         clk_[tid] = dirty.epoch;
@@ -299,10 +299,10 @@ void ThreadClock::ReleaseStore(ClockCache *c, SyncClock *dst) {
     dst->tab_idx_ = cached_idx_;
     dst->size_ = cached_size_;
     dst->blocks_ = cached_blocks_;
-    CHECK_EQ(dst->dirty_[0].tid, kInvalidTid);
+    CHECK_EQ(dst->dirty_[0].tid(), kInvalidTid);
     // The cached clock is shared (immutable),
     // so this is where we store the current clock.
-    dst->dirty_[0].tid = tid_;
+    dst->dirty_[0].set_tid(tid_);
     dst->dirty_[0].epoch = clk_[tid_];
     dst->release_store_tid_ = tid_;
     dst->release_store_reused_ = reused_;
@@ -336,8 +336,7 @@ void ThreadClock::ReleaseStore(ClockCache *c, SyncClock *dst) {
     ce.reused = 0;
     i++;
   }
-  for (uptr i = 0; i < kDirtyTids; i++)
-    dst->dirty_[i].tid = kInvalidTid;
+  for (uptr i = 0; i < kDirtyTids; i++) dst->dirty_[i].set_tid(kInvalidTid);
   dst->release_store_tid_ = tid_;
   dst->release_store_reused_ = reused_;
   // Rememeber that we don't need to acquire it in future.
@@ -369,10 +368,10 @@ void ThreadClock::UpdateCurrentThread(ClockCache *c, SyncClock *dst) const {
   // Update the threads time, but preserve 'acquired' flag.
   for (unsigned i = 0; i < kDirtyTids; i++) {
     SyncClock::Dirty *dirty = &dst->dirty_[i];
-    const unsigned tid = dirty->tid;
+    const unsigned tid = dirty->tid();
     if (tid == tid_ || tid == kInvalidTid) {
       CPP_STAT_INC(StatClockReleaseFast);
-      dirty->tid = tid_;
+      dirty->set_tid(tid_);
       dirty->epoch = clk_[tid_];
       return;
     }
@@ -393,8 +392,8 @@ bool ThreadClock::IsAlreadyAcquired(const SyncClock *src) const {
     return false;
   for (unsigned i = 0; i < kDirtyTids; i++) {
     SyncClock::Dirty dirty = src->dirty_[i];
-    if (dirty.tid != kInvalidTid) {
-      if (clk_[dirty.tid] < dirty.epoch)
+    if (dirty.tid() != kInvalidTid) {
+      if (clk_[dirty.tid()] < dirty.epoch)
         return false;
     }
   }
@@ -453,8 +452,7 @@ void SyncClock::ResetImpl() {
   blocks_ = 0;
   release_store_tid_ = kInvalidTid;
   release_store_reused_ = 0;
-  for (uptr i = 0; i < kDirtyTids; i++)
-    dirty_[i].tid = kInvalidTid;
+  for (uptr i = 0; i < kDirtyTids; i++) dirty_[i].set_tid(kInvalidTid);
 }
 
 void SyncClock::Resize(ClockCache *c, uptr nclk) {
@@ -503,10 +501,10 @@ void SyncClock::Resize(ClockCache *c, uptr nclk) {
 void SyncClock::FlushDirty() {
   for (unsigned i = 0; i < kDirtyTids; i++) {
     Dirty *dirty = &dirty_[i];
-    if (dirty->tid != kInvalidTid) {
-      CHECK_LT(dirty->tid, size_);
-      elem(dirty->tid).epoch = dirty->epoch;
-      dirty->tid = kInvalidTid;
+    if (dirty->tid() != kInvalidTid) {
+      CHECK_LT(dirty->tid(), size_);
+      elem(dirty->tid()).epoch = dirty->epoch;
+      dirty->set_tid(kInvalidTid);
     }
   }
 }
@@ -559,7 +557,7 @@ ALWAYS_INLINE bool SyncClock::Cachable() const {
   if (size_ == 0)
     return false;
   for (unsigned i = 0; i < kDirtyTids; i++) {
-    if (dirty_[i].tid != kInvalidTid)
+    if (dirty_[i].tid() != kInvalidTid)
       return false;
   }
   return atomic_load_relaxed(ref_ptr(tab_)) == 1;
@@ -606,7 +604,7 @@ ALWAYS_INLINE void SyncClock::append_block(u32 idx) {
 u64 SyncClock::get(unsigned tid) const {
   for (unsigned i = 0; i < kDirtyTids; i++) {
     Dirty dirty = dirty_[i];
-    if (dirty.tid == tid)
+    if (dirty.tid() == tid)
       return dirty.epoch;
   }
   return elem(tid).epoch;
@@ -625,9 +623,8 @@ void SyncClock::DebugDump(int(*printf)(const char *s, ...)) {
   for (uptr i = 0; i < size_; i++)
     printf("%s%llu", i == 0 ? "" : ",", elem(i).reused);
   printf("] release_store_tid=%d/%d dirty_tids=%d[%llu]/%d[%llu]",
-      release_store_tid_, release_store_reused_,
-      dirty_[0].tid, dirty_[0].epoch,
-      dirty_[1].tid, dirty_[1].epoch);
+         release_store_tid_, release_store_reused_, dirty_[0].tid(),
+         dirty_[0].epoch, dirty_[1].tid(), dirty_[1].epoch);
 }
 
 void SyncClock::Iter::Next() {
diff --git a/libsanitizer/tsan/tsan_clock.h b/libsanitizer/tsan/tsan_clock.h
index 736cdae..31376a1 100644
--- a/libsanitizer/tsan/tsan_clock.h
+++ b/libsanitizer/tsan/tsan_clock.h
@@ -17,7 +17,7 @@
 
 namespace __tsan {
 
-typedef DenseSlabAlloc<ClockBlock, 1<<16, 1<<10> ClockAlloc;
+typedef DenseSlabAlloc<ClockBlock, 1 << 22, 1 << 10> ClockAlloc;
 typedef DenseSlabAllocCache ClockCache;
 
 // The clock that lives in sync variables (mutexes, atomics, etc).
@@ -65,10 +65,20 @@ class SyncClock {
   static const uptr kDirtyTids = 2;
 
   struct Dirty {
-    u64 epoch  : kClkBits;
-    u64 tid : 64 - kClkBits;  // kInvalidId if not active
+    u32 tid() const { return tid_ == kShortInvalidTid ? kInvalidTid : tid_; }
+    void set_tid(u32 tid) {
+      tid_ = tid == kInvalidTid ? kShortInvalidTid : tid;
+    }
+    u64 epoch : kClkBits;
+
+   private:
+    // Full kInvalidTid won't fit into Dirty::tid.
+    static const u64 kShortInvalidTid = (1ull << (64 - kClkBits)) - 1;
+    u64 tid_ : 64 - kClkBits;  // kInvalidId if not active
   };
 
+  static_assert(sizeof(Dirty) == 8, "Dirty is not 64bit");
+
   unsigned release_store_tid_;
   unsigned release_store_reused_;
   Dirty dirty_[kDirtyTids];
diff --git a/libsanitizer/tsan/tsan_defs.h b/libsanitizer/tsan/tsan_defs.h
index 293d7de..f53787a 100644
--- a/libsanitizer/tsan/tsan_defs.h
+++ b/libsanitizer/tsan/tsan_defs.h
@@ -98,8 +98,6 @@ const bool kCollectHistory = false;
 const bool kCollectHistory = true;
 #endif
 
-const u16 kInvalidTid = kMaxTid + 1;
-
 // The following "build consistency" machinery ensures that all source files
 // are built in the same configuration. Inconsistent builds lead to
 // hard to debug crashes.
diff --git a/libsanitizer/tsan/tsan_dense_alloc.h b/libsanitizer/tsan/tsan_dense_alloc.h
index 64fc50e..6c89e40 100644
--- a/libsanitizer/tsan/tsan_dense_alloc.h
+++ b/libsanitizer/tsan/tsan_dense_alloc.h
@@ -29,28 +29,40 @@ class DenseSlabAllocCache {
   typedef u32 IndexT;
   uptr pos;
   IndexT cache[kSize];
-  template<typename T, uptr kL1Size, uptr kL2Size> friend class DenseSlabAlloc;
+  template <typename, uptr, uptr, u64>
+  friend class DenseSlabAlloc;
 };
 
-template<typename T, uptr kL1Size, uptr kL2Size>
+template <typename T, uptr kL1Size, uptr kL2Size, u64 kReserved = 0>
 class DenseSlabAlloc {
  public:
   typedef DenseSlabAllocCache Cache;
   typedef typename Cache::IndexT IndexT;
 
-  explicit DenseSlabAlloc(const char *name) {
-    // Check that kL1Size and kL2Size are sane.
-    CHECK_EQ(kL1Size & (kL1Size - 1), 0);
-    CHECK_EQ(kL2Size & (kL2Size - 1), 0);
-    CHECK_GE(1ull << (sizeof(IndexT) * 8), kL1Size * kL2Size);
-    // Check that it makes sense to use the dense alloc.
-    CHECK_GE(sizeof(T), sizeof(IndexT));
-    internal_memset(map_, 0, sizeof(map_));
+  static_assert((kL1Size & (kL1Size - 1)) == 0,
+                "kL1Size must be a power-of-two");
+  static_assert((kL2Size & (kL2Size - 1)) == 0,
+                "kL2Size must be a power-of-two");
+  static_assert((kL1Size * kL2Size) <= (1ull << (sizeof(IndexT) * 8)),
+                "kL1Size/kL2Size are too large");
+  static_assert(((kL1Size * kL2Size - 1) & kReserved) == 0,
+                "reserved bits don't fit");
+  static_assert(sizeof(T) > sizeof(IndexT),
+                "it doesn't make sense to use dense alloc");
+
+  explicit DenseSlabAlloc(LinkerInitialized, const char *name) {
     freelist_ = 0;
     fillpos_ = 0;
     name_ = name;
   }
 
+  explicit DenseSlabAlloc(const char *name)
+      : DenseSlabAlloc(LINKER_INITIALIZED, name) {
+    // It can be very large.
+    // Don't page it in for linker initialized objects.
+    internal_memset(map_, 0, sizeof(map_));
+  }
+
   ~DenseSlabAlloc() {
     for (uptr i = 0; i < kL1Size; i++) {
       if (map_[i] != 0)
diff --git a/libsanitizer/tsan/tsan_external.cpp b/libsanitizer/tsan/tsan_external.cpp
index 466b2bf..a87e12f 100644
--- a/libsanitizer/tsan/tsan_external.cpp
+++ b/libsanitizer/tsan/tsan_external.cpp
@@ -111,12 +111,12 @@ void __tsan_external_assign_tag(void *addr, void *tag) {
 
 SANITIZER_INTERFACE_ATTRIBUTE
 void __tsan_external_read(void *addr, void *caller_pc, void *tag) {
-  ExternalAccess(addr, STRIP_PC(caller_pc), tag, MemoryRead);
+  ExternalAccess(addr, STRIP_PAC_PC(caller_pc), tag, MemoryRead);
 }
 
 SANITIZER_INTERFACE_ATTRIBUTE
 void __tsan_external_write(void *addr, void *caller_pc, void *tag) {
-  ExternalAccess(addr, STRIP_PC(caller_pc), tag, MemoryWrite);
+  ExternalAccess(addr, STRIP_PAC_PC(caller_pc), tag, MemoryWrite);
 }
 }  // extern "C"
 
diff --git a/libsanitizer/tsan/tsan_interceptors_mac.cpp b/libsanitizer/tsan/tsan_interceptors_mac.cpp
index aa29536..ed10fcc 100644
--- a/libsanitizer/tsan/tsan_interceptors_mac.cpp
+++ b/libsanitizer/tsan/tsan_interceptors_mac.cpp
@@ -438,6 +438,7 @@ struct fake_shared_weak_count {
   virtual void on_zero_shared() = 0;
   virtual void _unused_0x18() = 0;
   virtual void on_zero_shared_weak() = 0;
+  virtual ~fake_shared_weak_count() = 0;  // suppress -Wnon-virtual-dtor
 };
 }  // namespace
 
diff --git a/libsanitizer/tsan/tsan_interceptors_posix.cpp b/libsanitizer/tsan/tsan_interceptors_posix.cpp
index aa04d8d..2651e22 100644
--- a/libsanitizer/tsan/tsan_interceptors_posix.cpp
+++ b/libsanitizer/tsan/tsan_interceptors_posix.cpp
@@ -54,10 +54,6 @@ using namespace __tsan;
 #define vfork __vfork14
 #endif
 
-#if SANITIZER_ANDROID
-#define mallopt(a, b)
-#endif
-
 #ifdef __mips__
 const int kSigCount = 129;
 #else
@@ -85,6 +81,8 @@ extern "C" int pthread_attr_init(void *attr);
 extern "C" int pthread_attr_destroy(void *attr);
 DECLARE_REAL(int, pthread_attr_getdetachstate, void *, void *)
 extern "C" int pthread_attr_setstacksize(void *attr, uptr stacksize);
+extern "C" int pthread_atfork(void (*prepare)(void), void (*parent)(void),
+                              void (*child)(void));
 extern "C" int pthread_key_create(unsigned *key, void (*destructor)(void* v));
 extern "C" int pthread_setspecific(unsigned key, const void *v);
 DECLARE_REAL(int, pthread_mutexattr_gettype, void *, void *)
@@ -97,7 +95,7 @@ extern "C" void _exit(int status);
 extern "C" int fileno_unlocked(void *stream);
 extern "C" int dirfd(void *dirp);
 #endif
-#if !SANITIZER_FREEBSD && !SANITIZER_ANDROID && !SANITIZER_NETBSD
+#if SANITIZER_GLIBC
 extern "C" int mallopt(int param, int value);
 #endif
 #if SANITIZER_NETBSD
@@ -659,8 +657,11 @@ TSAN_INTERCEPTOR(void*, malloc, uptr size) {
   return p;
 }
 
+// In glibc<2.25, dynamic TLS blocks are allocated by __libc_memalign. Intercept
+// __libc_memalign so that (1) we can detect races (2) free will not be called
+// on libc internally allocated blocks.
 TSAN_INTERCEPTOR(void*, __libc_memalign, uptr align, uptr sz) {
-  SCOPED_TSAN_INTERCEPTOR(__libc_memalign, align, sz);
+  SCOPED_INTERCEPTOR_RAW(__libc_memalign, align, sz);
   return user_memalign(thr, pc, align, sz);
 }
 
@@ -773,6 +774,11 @@ static void *mmap_interceptor(ThreadState *thr, uptr pc, Mmap real_mmap,
   if (!fix_mmap_addr(&addr, sz, flags)) return MAP_FAILED;
   void *res = real_mmap(addr, sz, prot, flags, fd, off);
   if (res != MAP_FAILED) {
+    if (!IsAppMem((uptr)res) || !IsAppMem((uptr)res + sz - 1)) {
+      Report("ThreadSanitizer: mmap at bad address: addr=%p size=%p res=%p\n",
+             addr, (void*)sz, res);
+      Die();
+    }
     if (fd > 0) FdAccess(thr, pc, fd);
     MemoryRangeImitateWriteOrResetRange(thr, pc, (uptr)res, sz);
   }
@@ -1122,27 +1128,37 @@ static void *init_cond(void *c, bool force = false) {
   return (void*)cond;
 }
 
+namespace {
+
+template <class Fn>
 struct CondMutexUnlockCtx {
   ScopedInterceptor *si;
   ThreadState *thr;
   uptr pc;
   void *m;
+  void *c;
+  const Fn &fn;
+
+  int Cancel() const { return fn(); }
+  void Unlock() const;
 };
 
-static void cond_mutex_unlock(CondMutexUnlockCtx *arg) {
+template <class Fn>
+void CondMutexUnlockCtx<Fn>::Unlock() const {
   // pthread_cond_wait interceptor has enabled async signal delivery
   // (see BlockingCall below). Disable async signals since we are running
   // tsan code. Also ScopedInterceptor and BlockingCall destructors won't run
   // since the thread is cancelled, so we have to manually execute them
   // (the thread still can run some user code due to pthread_cleanup_push).
-  ThreadSignalContext *ctx = SigCtx(arg->thr);
+  ThreadSignalContext *ctx = SigCtx(thr);
   CHECK_EQ(atomic_load(&ctx->in_blocking_func, memory_order_relaxed), 1);
   atomic_store(&ctx->in_blocking_func, 0, memory_order_relaxed);
-  MutexPostLock(arg->thr, arg->pc, (uptr)arg->m, MutexFlagDoPreLockOnPostLock);
+  MutexPostLock(thr, pc, (uptr)m, MutexFlagDoPreLockOnPostLock);
   // Undo BlockingCall ctor effects.
-  arg->thr->ignore_interceptors--;
-  arg->si->~ScopedInterceptor();
+  thr->ignore_interceptors--;
+  si->~ScopedInterceptor();
 }
+}  // namespace
 
 INTERCEPTOR(int, pthread_cond_init, void *c, void *a) {
   void *cond = init_cond(c, true);
@@ -1151,20 +1167,24 @@ INTERCEPTOR(int, pthread_cond_init, void *c, void *a) {
   return REAL(pthread_cond_init)(cond, a);
 }
 
-static int cond_wait(ThreadState *thr, uptr pc, ScopedInterceptor *si,
-                     int (*fn)(void *c, void *m, void *abstime), void *c,
-                     void *m, void *t) {
+template <class Fn>
+int cond_wait(ThreadState *thr, uptr pc, ScopedInterceptor *si, const Fn &fn,
+              void *c, void *m) {
   MemoryAccessRange(thr, pc, (uptr)c, sizeof(uptr), false);
   MutexUnlock(thr, pc, (uptr)m);
-  CondMutexUnlockCtx arg = {si, thr, pc, m};
   int res = 0;
   // This ensures that we handle mutex lock even in case of pthread_cancel.
   // See test/tsan/cond_cancel.cpp.
   {
     // Enable signal delivery while the thread is blocked.
     BlockingCall bc(thr);
+    CondMutexUnlockCtx<Fn> arg = {si, thr, pc, m, c, fn};
     res = call_pthread_cancel_with_cleanup(
-        fn, c, m, t, (void (*)(void *arg))cond_mutex_unlock, &arg);
+        [](void *arg) -> int {
+          return ((const CondMutexUnlockCtx<Fn> *)arg)->Cancel();
+        },
+        [](void *arg) { ((const CondMutexUnlockCtx<Fn> *)arg)->Unlock(); },
+        &arg);
   }
   if (res == errno_EOWNERDEAD) MutexRepair(thr, pc, (uptr)m);
   MutexPostLock(thr, pc, (uptr)m, MutexFlagDoPreLockOnPostLock);
@@ -1174,25 +1194,46 @@ static int cond_wait(ThreadState *thr, uptr pc, ScopedInterceptor *si,
 INTERCEPTOR(int, pthread_cond_wait, void *c, void *m) {
   void *cond = init_cond(c);
   SCOPED_TSAN_INTERCEPTOR(pthread_cond_wait, cond, m);
-  return cond_wait(thr, pc, &si, (int (*)(void *c, void *m, void *abstime))REAL(
-                                     pthread_cond_wait),
-                   cond, m, 0);
+  return cond_wait(
+      thr, pc, &si, [=]() { return REAL(pthread_cond_wait)(cond, m); }, cond,
+      m);
 }
 
 INTERCEPTOR(int, pthread_cond_timedwait, void *c, void *m, void *abstime) {
   void *cond = init_cond(c);
   SCOPED_TSAN_INTERCEPTOR(pthread_cond_timedwait, cond, m, abstime);
-  return cond_wait(thr, pc, &si, REAL(pthread_cond_timedwait), cond, m,
-                   abstime);
+  return cond_wait(
+      thr, pc, &si,
+      [=]() { return REAL(pthread_cond_timedwait)(cond, m, abstime); }, cond,
+      m);
 }
 
+#if SANITIZER_LINUX
+INTERCEPTOR(int, pthread_cond_clockwait, void *c, void *m,
+            __sanitizer_clockid_t clock, void *abstime) {
+  void *cond = init_cond(c);
+  SCOPED_TSAN_INTERCEPTOR(pthread_cond_clockwait, cond, m, clock, abstime);
+  return cond_wait(
+      thr, pc, &si,
+      [=]() { return REAL(pthread_cond_clockwait)(cond, m, clock, abstime); },
+      cond, m);
+}
+#define TSAN_MAYBE_PTHREAD_COND_CLOCKWAIT TSAN_INTERCEPT(pthread_cond_clockwait)
+#else
+#define TSAN_MAYBE_PTHREAD_COND_CLOCKWAIT
+#endif
+
 #if SANITIZER_MAC
 INTERCEPTOR(int, pthread_cond_timedwait_relative_np, void *c, void *m,
             void *reltime) {
   void *cond = init_cond(c);
   SCOPED_TSAN_INTERCEPTOR(pthread_cond_timedwait_relative_np, cond, m, reltime);
-  return cond_wait(thr, pc, &si, REAL(pthread_cond_timedwait_relative_np), cond,
-                   m, reltime);
+  return cond_wait(
+      thr, pc, &si,
+      [=]() {
+        return REAL(pthread_cond_timedwait_relative_np)(cond, m, reltime);
+      },
+      cond, m);
 }
 #endif
 
@@ -1937,7 +1978,8 @@ static void CallUserSignalHandler(ThreadState *thr, bool sync, bool acquire,
   // because in async signal processing case (when handler is called directly
   // from rtl_generic_sighandler) we have not yet received the reraised
   // signal; and it looks too fragile to intercept all ways to reraise a signal.
-  if (flags()->report_bugs && !sync && sig != SIGTERM && errno != 99) {
+  if (ShouldReport(thr, ReportTypeErrnoInSignal) && !sync && sig != SIGTERM &&
+      errno != 99) {
     VarSizeStackTrace stack;
     // StackTrace::GetNestInstructionPc(pc) is used because return address is
     // expected, OutputReport() will undo this.
@@ -2107,26 +2149,32 @@ TSAN_INTERCEPTOR(int, fork, int fake) {
   if (in_symbolizer())
     return REAL(fork)(fake);
   SCOPED_INTERCEPTOR_RAW(fork, fake);
+  return REAL(fork)(fake);
+}
+
+void atfork_prepare() {
+  if (in_symbolizer())
+    return;
+  ThreadState *thr = cur_thread();
+  const uptr pc = StackTrace::GetCurrentPc();
   ForkBefore(thr, pc);
-  int pid;
-  {
-    // On OS X, REAL(fork) can call intercepted functions (OSSpinLockLock), and
-    // we'll assert in CheckNoLocks() unless we ignore interceptors.
-    ScopedIgnoreInterceptors ignore;
-    pid = REAL(fork)(fake);
-  }
-  if (pid == 0) {
-    // child
-    ForkChildAfter(thr, pc);
-    FdOnFork(thr, pc);
-  } else if (pid > 0) {
-    // parent
-    ForkParentAfter(thr, pc);
-  } else {
-    // error
-    ForkParentAfter(thr, pc);
-  }
-  return pid;
+}
+
+void atfork_parent() {
+  if (in_symbolizer())
+    return;
+  ThreadState *thr = cur_thread();
+  const uptr pc = StackTrace::GetCurrentPc();
+  ForkParentAfter(thr, pc);
+}
+
+void atfork_child() {
+  if (in_symbolizer())
+    return;
+  ThreadState *thr = cur_thread();
+  const uptr pc = StackTrace::GetCurrentPc();
+  ForkChildAfter(thr, pc);
+  FdOnFork(thr, pc);
 }
 
 TSAN_INTERCEPTOR(int, vfork, int fake) {
@@ -2479,13 +2527,10 @@ static USED void syscall_fd_release(uptr pc, int fd) {
   FdRelease(thr, pc, fd);
 }
 
-static void syscall_pre_fork(uptr pc) {
-  TSAN_SYSCALL();
-  ForkBefore(thr, pc);
-}
+static void syscall_pre_fork(uptr pc) { ForkBefore(cur_thread(), pc); }
 
 static void syscall_post_fork(uptr pc, int pid) {
-  TSAN_SYSCALL();
+  ThreadState *thr = cur_thread();
   if (pid == 0) {
     // child
     ForkChildAfter(thr, pc);
@@ -2635,7 +2680,7 @@ void InitializeInterceptors() {
 #endif
 
   // Instruct libc malloc to consume less memory.
-#if SANITIZER_LINUX
+#if SANITIZER_GLIBC
   mallopt(1, 0);  // M_MXFAST
   mallopt(-3, 32*1024);  // M_MMAP_THRESHOLD
 #endif
@@ -2698,6 +2743,8 @@ void InitializeInterceptors() {
   TSAN_INTERCEPT_VER(pthread_cond_timedwait, PTHREAD_ABI_BASE);
   TSAN_INTERCEPT_VER(pthread_cond_destroy, PTHREAD_ABI_BASE);
 
+  TSAN_MAYBE_PTHREAD_COND_CLOCKWAIT;
+
   TSAN_INTERCEPT(pthread_mutex_init);
   TSAN_INTERCEPT(pthread_mutex_destroy);
   TSAN_INTERCEPT(pthread_mutex_trylock);
@@ -2799,6 +2846,10 @@ void InitializeInterceptors() {
     Printf("ThreadSanitizer: failed to setup atexit callback\n");
     Die();
   }
+  if (pthread_atfork(atfork_prepare, atfork_parent, atfork_child)) {
+    Printf("ThreadSanitizer: failed to setup atfork callbacks\n");
+    Die();
+  }
 
 #if !SANITIZER_MAC && !SANITIZER_NETBSD && !SANITIZER_FREEBSD
   if (pthread_key_create(&interceptor_ctx()->finalize_key, &thread_finalize)) {
diff --git a/libsanitizer/tsan/tsan_interface.cpp b/libsanitizer/tsan/tsan_interface.cpp
index 55f1c98..9bd0e85 100644
--- a/libsanitizer/tsan/tsan_interface.cpp
+++ b/libsanitizer/tsan/tsan_interface.cpp
@@ -40,13 +40,13 @@ void __tsan_write16(void *addr) {
 }
 
 void __tsan_read16_pc(void *addr, void *pc) {
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog8);
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr + 8, kSizeLog8);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog8);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr + 8, kSizeLog8);
 }
 
 void __tsan_write16_pc(void *addr, void *pc) {
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog8);
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr + 8, kSizeLog8);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog8);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr + 8, kSizeLog8);
 }
 
 // __tsan_unaligned_read/write calls are emitted by compiler.
diff --git a/libsanitizer/tsan/tsan_interface.h b/libsanitizer/tsan/tsan_interface.h
index 6d7286c..6e022b5 100644
--- a/libsanitizer/tsan/tsan_interface.h
+++ b/libsanitizer/tsan/tsan_interface.h
@@ -204,7 +204,7 @@ __extension__ typedef __int128 a128;
 #endif
 
 // Part of ABI, do not change.
-// https://github.com/llvm/llvm-project/blob/master/libcxx/include/atomic
+// https://github.com/llvm/llvm-project/blob/main/libcxx/include/atomic
 typedef enum {
   mo_relaxed,
   mo_consume,
@@ -415,6 +415,13 @@ void __tsan_go_atomic32_compare_exchange(ThreadState *thr, uptr cpc, uptr pc,
 SANITIZER_INTERFACE_ATTRIBUTE
 void __tsan_go_atomic64_compare_exchange(ThreadState *thr, uptr cpc, uptr pc,
                                          u8 *a);
+
+SANITIZER_INTERFACE_ATTRIBUTE
+void __tsan_on_initialize();
+
+SANITIZER_INTERFACE_ATTRIBUTE
+int __tsan_on_finalize(int failed);
+
 }  // extern "C"
 
 }  // namespace __tsan
diff --git a/libsanitizer/tsan/tsan_interface_inl.h b/libsanitizer/tsan/tsan_interface_inl.h
index f5d743c..5e77d4d 100644
--- a/libsanitizer/tsan/tsan_interface_inl.h
+++ b/libsanitizer/tsan/tsan_interface_inl.h
@@ -51,35 +51,35 @@ void __tsan_write8(void *addr) {
 }
 
 void __tsan_read1_pc(void *addr, void *pc) {
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog1);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog1);
 }
 
 void __tsan_read2_pc(void *addr, void *pc) {
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog2);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog2);
 }
 
 void __tsan_read4_pc(void *addr, void *pc) {
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog4);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog4);
 }
 
 void __tsan_read8_pc(void *addr, void *pc) {
-  MemoryRead(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog8);
+  MemoryRead(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog8);
 }
 
 void __tsan_write1_pc(void *addr, void *pc) {
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog1);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog1);
 }
 
 void __tsan_write2_pc(void *addr, void *pc) {
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog2);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog2);
 }
 
 void __tsan_write4_pc(void *addr, void *pc) {
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog4);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog4);
 }
 
 void __tsan_write8_pc(void *addr, void *pc) {
-  MemoryWrite(cur_thread(), STRIP_PC(pc), (uptr)addr, kSizeLog8);
+  MemoryWrite(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, kSizeLog8);
 }
 
 void __tsan_vptr_update(void **vptr_p, void *new_val) {
@@ -101,7 +101,7 @@ void __tsan_vptr_read(void **vptr_p) {
 }
 
 void __tsan_func_entry(void *pc) {
-  FuncEntry(cur_thread(), STRIP_PC(pc));
+  FuncEntry(cur_thread(), STRIP_PAC_PC(pc));
 }
 
 void __tsan_func_exit() {
@@ -125,9 +125,9 @@ void __tsan_write_range(void *addr, uptr size) {
 }
 
 void __tsan_read_range_pc(void *addr, uptr size, void *pc) {
-  MemoryAccessRange(cur_thread(), STRIP_PC(pc), (uptr)addr, size, false);
+  MemoryAccessRange(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, size, false);
 }
 
 void __tsan_write_range_pc(void *addr, uptr size, void *pc) {
-  MemoryAccessRange(cur_thread(), STRIP_PC(pc), (uptr)addr, size, true);
+  MemoryAccessRange(cur_thread(), STRIP_PAC_PC(pc), (uptr)addr, size, true);
 }
diff --git a/libsanitizer/tsan/tsan_mman.cpp b/libsanitizer/tsan/tsan_mman.cpp
index 743e67bf..45a39f0 100644
--- a/libsanitizer/tsan/tsan_mman.cpp
+++ b/libsanitizer/tsan/tsan_mman.cpp
@@ -145,7 +145,7 @@ void AllocatorPrintStats() {
 
 static void SignalUnsafeCall(ThreadState *thr, uptr pc) {
   if (atomic_load_relaxed(&thr->in_signal_handler) == 0 ||
-      !flags()->report_signal_unsafe)
+      !ShouldReport(thr, ReportTypeSignalUnsafe))
     return;
   VarSizeStackTrace stack;
   ObtainCurrentStack(thr, pc, &stack);
diff --git a/libsanitizer/tsan/tsan_platform.h b/libsanitizer/tsan/tsan_platform.h
index 16169ca..101522d 100644
--- a/libsanitizer/tsan/tsan_platform.h
+++ b/libsanitizer/tsan/tsan_platform.h
@@ -23,9 +23,21 @@
 
 namespace __tsan {
 
+#if defined(__x86_64__)
+#define HAS_48_BIT_ADDRESS_SPACE 1
+#elif SANITIZER_IOSSIM // arm64 iOS simulators (order of #if matters)
+#define HAS_48_BIT_ADDRESS_SPACE 1
+#elif SANITIZER_IOS // arm64 iOS devices (order of #if matters)
+#define HAS_48_BIT_ADDRESS_SPACE 0
+#elif SANITIZER_MAC // arm64 macOS (order of #if matters)
+#define HAS_48_BIT_ADDRESS_SPACE 1
+#else
+#define HAS_48_BIT_ADDRESS_SPACE 0
+#endif
+
 #if !SANITIZER_GO
 
-#if defined(__x86_64__)
+#if HAS_48_BIT_ADDRESS_SPACE
 /*
 C/C++ on linux/x86_64 and freebsd/x86_64
 0000 0000 1000 - 0080 0000 0000: main binary and/or MAP_32BIT mappings (512GB)
@@ -93,7 +105,7 @@ fe00 0000 00 - ff00 0000 00: heap                                        (4 GB)
 ff00 0000 00 - ff80 0000 00: -                                           (2 GB)
 ff80 0000 00 - ffff ffff ff: modules and main thread stack              (<2 GB)
 */
-struct Mapping {
+struct Mapping40 {
   static const uptr kMetaShadowBeg = 0x4000000000ull;
   static const uptr kMetaShadowEnd = 0x5000000000ull;
   static const uptr kTraceMemBeg   = 0xb000000000ull;
@@ -114,6 +126,7 @@ struct Mapping {
 };
 
 #define TSAN_MID_APP_RANGE 1
+#define TSAN_RUNTIME_VMA 1
 #elif defined(__aarch64__) && defined(__APPLE__)
 /*
 C/C++ on Darwin/iOS/ARM64 (36-bit VMA, 64 GB VM)
@@ -146,7 +159,7 @@ struct Mapping {
   static const uptr kVdsoBeg       = 0x7000000000000000ull;
 };
 
-#elif defined(__aarch64__)
+#elif defined(__aarch64__) && !defined(__APPLE__)
 // AArch64 supports multiple VMA which leads to multiple address transformation
 // functions.  To support these multiple VMAS transformations and mappings TSAN
 // runtime for AArch64 uses an external memory read (vmaSize) to select which
@@ -354,7 +367,7 @@ struct Mapping47 {
 #define TSAN_RUNTIME_VMA 1
 #endif
 
-#elif SANITIZER_GO && !SANITIZER_WINDOWS && defined(__x86_64__)
+#elif SANITIZER_GO && !SANITIZER_WINDOWS && HAS_48_BIT_ADDRESS_SPACE
 
 /* Go on linux, darwin and freebsd on x86_64
 0000 0000 1000 - 0000 1000 0000: executable
@@ -502,7 +515,7 @@ Go on linux/mips64 (47-bit VMA)
 6000 0000 0000 - 6200 0000 0000: traces
 6200 0000 0000 - 8000 0000 0000: -
 */
-struct Mapping {
+struct Mapping47 {
   static const uptr kMetaShadowBeg = 0x300000000000ull;
   static const uptr kMetaShadowEnd = 0x400000000000ull;
   static const uptr kTraceMemBeg = 0x600000000000ull;
@@ -512,6 +525,9 @@ struct Mapping {
   static const uptr kAppMemBeg = 0x000000001000ull;
   static const uptr kAppMemEnd = 0x00e000000000ull;
 };
+
+#define TSAN_RUNTIME_VMA 1
+
 #else
 # error "Unknown platform"
 #endif
@@ -592,6 +608,16 @@ uptr MappingArchImpl(void) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return MappingImpl<Mapping40, Type>();
+#else
+    case 47: return MappingImpl<Mapping47, Type>();
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return MappingImpl<Mapping, Type>();
 #endif
@@ -749,6 +775,16 @@ bool IsAppMem(uptr mem) {
   }
   DCHECK(0);
   return false;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return IsAppMemImpl<Mapping40>(mem);
+#else
+    case 47: return IsAppMemImpl<Mapping47>(mem);
+#endif
+  }
+  DCHECK(0);
+  return false;
 #else
   return IsAppMemImpl<Mapping>(mem);
 #endif
@@ -780,6 +816,16 @@ bool IsShadowMem(uptr mem) {
   }
   DCHECK(0);
   return false;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return IsShadowMemImpl<Mapping40>(mem);
+#else
+    case 47: return IsShadowMemImpl<Mapping47>(mem);
+#endif
+  }
+  DCHECK(0);
+  return false;
 #else
   return IsShadowMemImpl<Mapping>(mem);
 #endif
@@ -811,6 +857,16 @@ bool IsMetaMem(uptr mem) {
   }
   DCHECK(0);
   return false;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return IsMetaMemImpl<Mapping40>(mem);
+#else
+    case 47: return IsMetaMemImpl<Mapping47>(mem);
+#endif
+  }
+  DCHECK(0);
+  return false;
 #else
   return IsMetaMemImpl<Mapping>(mem);
 #endif
@@ -852,6 +908,16 @@ uptr MemToShadow(uptr x) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return MemToShadowImpl<Mapping40>(x);
+#else
+    case 47: return MemToShadowImpl<Mapping47>(x);
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return MemToShadowImpl<Mapping>(x);
 #endif
@@ -895,6 +961,16 @@ u32 *MemToMeta(uptr x) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return MemToMetaImpl<Mapping40>(x);
+#else
+    case 47: return MemToMetaImpl<Mapping47>(x);
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return MemToMetaImpl<Mapping>(x);
 #endif
@@ -951,6 +1027,16 @@ uptr ShadowToMem(uptr s) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return ShadowToMemImpl<Mapping40>(s);
+#else
+    case 47: return ShadowToMemImpl<Mapping47>(s);
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return ShadowToMemImpl<Mapping>(s);
 #endif
@@ -990,6 +1076,16 @@ uptr GetThreadTrace(int tid) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return GetThreadTraceImpl<Mapping40>(tid);
+#else
+    case 47: return GetThreadTraceImpl<Mapping47>(tid);
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return GetThreadTraceImpl<Mapping>(tid);
 #endif
@@ -1024,6 +1120,16 @@ uptr GetThreadTraceHeader(int tid) {
   }
   DCHECK(0);
   return 0;
+#elif defined(__mips64)
+  switch (vmaSize) {
+#if !SANITIZER_GO
+    case 40: return GetThreadTraceHeaderImpl<Mapping40>(tid);
+#else
+    case 47: return GetThreadTraceHeaderImpl<Mapping47>(tid);
+#endif
+  }
+  DCHECK(0);
+  return 0;
 #else
   return GetThreadTraceHeaderImpl<Mapping>(tid);
 #endif
@@ -1040,9 +1146,8 @@ int ExtractRecvmsgFDs(void *msg, int *fds, int nfd);
 uptr ExtractLongJmpSp(uptr *env);
 void ImitateTlsWrite(ThreadState *thr, uptr tls_addr, uptr tls_size);
 
-int call_pthread_cancel_with_cleanup(int(*fn)(void *c, void *m,
-    void *abstime), void *c, void *m, void *abstime,
-    void(*cleanup)(void *arg), void *arg);
+int call_pthread_cancel_with_cleanup(int (*fn)(void *arg),
+                                     void (*cleanup)(void *arg), void *arg);
 
 void DestroyThreadState();
 void PlatformCleanUpThreadState(ThreadState *thr);
diff --git a/libsanitizer/tsan/tsan_platform_linux.cpp b/libsanitizer/tsan/tsan_platform_linux.cpp
index d136dcb..e5b6690 100644
--- a/libsanitizer/tsan/tsan_platform_linux.cpp
+++ b/libsanitizer/tsan/tsan_platform_linux.cpp
@@ -250,6 +250,20 @@ void InitializePlatformEarly() {
     Die();
   }
 # endif
+#elif defined(__mips64)
+# if !SANITIZER_GO
+  if (vmaSize != 40) {
+    Printf("FATAL: ThreadSanitizer: unsupported VMA range\n");
+    Printf("FATAL: Found %zd - Supported 40\n", vmaSize);
+    Die();
+  }
+# else
+  if (vmaSize != 47) {
+    Printf("FATAL: ThreadSanitizer: unsupported VMA range\n");
+    Printf("FATAL: Found %zd - Supported 47\n", vmaSize);
+    Die();
+  }
+# endif
 #endif
 #endif
 }
@@ -443,14 +457,13 @@ void ImitateTlsWrite(ThreadState *thr, uptr tls_addr, uptr tls_size) {
 
 // Note: this function runs with async signals enabled,
 // so it must not touch any tsan state.
-int call_pthread_cancel_with_cleanup(int(*fn)(void *c, void *m,
-    void *abstime), void *c, void *m, void *abstime,
-    void(*cleanup)(void *arg), void *arg) {
+int call_pthread_cancel_with_cleanup(int (*fn)(void *arg),
+                                     void (*cleanup)(void *arg), void *arg) {
   // pthread_cleanup_push/pop are hardcore macros mess.
   // We can't intercept nor call them w/o including pthread.h.
   int res;
   pthread_cleanup_push(cleanup, arg);
-  res = fn(c, m, abstime);
+  res = fn(arg);
   pthread_cleanup_pop(0);
   return res;
 }
@@ -484,7 +497,7 @@ ThreadState *cur_thread() {
         dead_thread_state->fast_state.SetIgnoreBit();
         dead_thread_state->ignore_interceptors = 1;
         dead_thread_state->is_dead = true;
-        *const_cast<int*>(&dead_thread_state->tid) = -1;
+        *const_cast<u32*>(&dead_thread_state->tid) = -1;
         CHECK_EQ(0, internal_mprotect(dead_thread_state, sizeof(ThreadState),
                                       PROT_READ));
       }
diff --git a/libsanitizer/tsan/tsan_platform_mac.cpp b/libsanitizer/tsan/tsan_platform_mac.cpp
index ec2c5fb..d9719a1 100644
--- a/libsanitizer/tsan/tsan_platform_mac.cpp
+++ b/libsanitizer/tsan/tsan_platform_mac.cpp
@@ -234,7 +234,7 @@ static void my_pthread_introspection_hook(unsigned int event, pthread_t thread,
 #endif
 
 void InitializePlatformEarly() {
-#if !SANITIZER_GO && defined(__aarch64__)
+#if !SANITIZER_GO && !HAS_48_BIT_ADDRESS_SPACE
   uptr max_vm = GetMaxUserVirtualAddress() + 1;
   if (max_vm != Mapping::kHiAppMemEnd) {
     Printf("ThreadSanitizer: unsupported vm address limit %p, expected %p.\n",
@@ -306,14 +306,13 @@ void ImitateTlsWrite(ThreadState *thr, uptr tls_addr, uptr tls_size) {
 #if !SANITIZER_GO
 // Note: this function runs with async signals enabled,
 // so it must not touch any tsan state.
-int call_pthread_cancel_with_cleanup(int(*fn)(void *c, void *m,
-    void *abstime), void *c, void *m, void *abstime,
-    void(*cleanup)(void *arg), void *arg) {
+int call_pthread_cancel_with_cleanup(int (*fn)(void *arg),
+                                     void (*cleanup)(void *arg), void *arg) {
   // pthread_cleanup_push/pop are hardcore macros mess.
   // We can't intercept nor call them w/o including pthread.h.
   int res;
   pthread_cleanup_push(cleanup, arg);
-  res = fn(c, m, abstime);
+  res = fn(arg);
   pthread_cleanup_pop(0);
   return res;
 }
diff --git a/libsanitizer/tsan/tsan_platform_posix.cpp b/libsanitizer/tsan/tsan_platform_posix.cpp
index d56b6c3..73e1d45 100644
--- a/libsanitizer/tsan/tsan_platform_posix.cpp
+++ b/libsanitizer/tsan/tsan_platform_posix.cpp
@@ -99,7 +99,7 @@ void CheckAndProtect() {
     Die();
   }
 
-#if defined(__aarch64__) && defined(__APPLE__)
+#if defined(__aarch64__) && defined(__APPLE__) && !HAS_48_BIT_ADDRESS_SPACE
   ProtectRange(HeapMemEnd(), ShadowBeg());
   ProtectRange(ShadowEnd(), MetaShadowBeg());
   ProtectRange(MetaShadowEnd(), TraceMemBeg());
diff --git a/libsanitizer/tsan/tsan_report.cpp b/libsanitizer/tsan/tsan_report.cpp
index 968c7b9..8ef9f0c 100644
--- a/libsanitizer/tsan/tsan_report.cpp
+++ b/libsanitizer/tsan/tsan_report.cpp
@@ -69,7 +69,7 @@ ReportDesc::~ReportDesc() {
 
 const int kThreadBufSize = 32;
 const char *thread_name(char *buf, int tid) {
-  if (tid == 0)
+  if (tid == kMainTid)
     return "main thread";
   internal_snprintf(buf, kThreadBufSize, "thread T%d", tid);
   return buf;
@@ -127,7 +127,7 @@ void PrintStack(const ReportStack *ent) {
   }
   SymbolizedStack *frame = ent->frames;
   for (int i = 0; frame && frame->info.address; frame = frame->next, i++) {
-    InternalScopedString res(2 * GetPageSizeCached());
+    InternalScopedString res;
     RenderFrame(&res, common_flags()->stack_trace_format, i,
                 frame->info.address, &frame->info,
                 common_flags()->symbolize_vs_style,
@@ -250,7 +250,7 @@ static void PrintMutex(const ReportMutex *rm) {
 
 static void PrintThread(const ReportThread *rt) {
   Decorator d;
-  if (rt->id == 0)  // Little sense in describing the main thread.
+  if (rt->id == kMainTid)  // Little sense in describing the main thread.
     return;
   Printf("%s", d.ThreadDescription());
   Printf("  Thread T%d", rt->id);
@@ -394,7 +394,7 @@ void PrintReport(const ReportDesc *rep) {
 
 #else  // #if !SANITIZER_GO
 
-const int kMainThreadId = 1;
+const u32 kMainGoroutineId = 1;
 
 void PrintStack(const ReportStack *ent) {
   if (ent == 0 || ent->frames == 0) {
@@ -415,7 +415,7 @@ static void PrintMop(const ReportMop *mop, bool first) {
   Printf("%s at %p by ",
       (first ? (mop->write ? "Write" : "Read")
              : (mop->write ? "Previous write" : "Previous read")), mop->addr);
-  if (mop->tid == kMainThreadId)
+  if (mop->tid == kMainGoroutineId)
     Printf("main goroutine:\n");
   else
     Printf("goroutine %d:\n", mop->tid);
@@ -428,7 +428,7 @@ static void PrintLocation(const ReportLocation *loc) {
     Printf("\n");
     Printf("Heap block of size %zu at %p allocated by ",
         loc->heap_chunk_size, loc->heap_chunk_start);
-    if (loc->tid == kMainThreadId)
+    if (loc->tid == kMainGoroutineId)
       Printf("main goroutine:\n");
     else
       Printf("goroutine %d:\n", loc->tid);
@@ -448,7 +448,7 @@ static void PrintLocation(const ReportLocation *loc) {
 }
 
 static void PrintThread(const ReportThread *rt) {
-  if (rt->id == kMainThreadId)
+  if (rt->id == kMainGoroutineId)
     return;
   Printf("\n");
   Printf("Goroutine %d (%s) created at:\n",
diff --git a/libsanitizer/tsan/tsan_rtl.cpp b/libsanitizer/tsan/tsan_rtl.cpp
index 3d721eb..0efa997 100644
--- a/libsanitizer/tsan/tsan_rtl.cpp
+++ b/libsanitizer/tsan/tsan_rtl.cpp
@@ -11,17 +11,19 @@
 // Main file (entry points) for the TSan run-time.
 //===----------------------------------------------------------------------===//
 
+#include "tsan_rtl.h"
+
 #include "sanitizer_common/sanitizer_atomic.h"
 #include "sanitizer_common/sanitizer_common.h"
 #include "sanitizer_common/sanitizer_file.h"
 #include "sanitizer_common/sanitizer_libc.h"
-#include "sanitizer_common/sanitizer_stackdepot.h"
 #include "sanitizer_common/sanitizer_placement_new.h"
+#include "sanitizer_common/sanitizer_stackdepot.h"
 #include "sanitizer_common/sanitizer_symbolizer.h"
 #include "tsan_defs.h"
-#include "tsan_platform.h"
-#include "tsan_rtl.h"
+#include "tsan_interface.h"
 #include "tsan_mman.h"
+#include "tsan_platform.h"
 #include "tsan_suppressions.h"
 #include "tsan_symbolize.h"
 #include "ubsan/ubsan_init.h"
@@ -56,12 +58,23 @@ Context *ctx;
 bool OnFinalize(bool failed);
 void OnInitialize();
 #else
+#include <dlfcn.h>
 SANITIZER_WEAK_CXX_DEFAULT_IMPL
 bool OnFinalize(bool failed) {
+#if !SANITIZER_GO
+  if (auto *ptr = dlsym(RTLD_DEFAULT, "__tsan_on_finalize"))
+    return reinterpret_cast<decltype(&__tsan_on_finalize)>(ptr)(failed);
+#endif
   return failed;
 }
 SANITIZER_WEAK_CXX_DEFAULT_IMPL
-void OnInitialize() {}
+void OnInitialize() {
+#if !SANITIZER_GO
+  if (auto *ptr = dlsym(RTLD_DEFAULT, "__tsan_on_initialize")) {
+    return reinterpret_cast<decltype(&__tsan_on_initialize)>(ptr)();
+  }
+#endif
+}
 #endif
 
 static char thread_registry_placeholder[sizeof(ThreadRegistry)];
@@ -77,12 +90,19 @@ static ThreadContextBase *CreateThreadContext(u32 tid) {
   new((void*)hdr) Trace();
   // We are going to use only a small part of the trace with the default
   // value of history_size. However, the constructor writes to the whole trace.
-  // Unmap the unused part.
+  // Release the unused part.
   uptr hdr_end = hdr + sizeof(Trace);
   hdr_end -= sizeof(TraceHeader) * (kTraceParts - TraceParts());
   hdr_end = RoundUp(hdr_end, GetPageSizeCached());
-  if (hdr_end < hdr + sizeof(Trace))
-    UnmapOrDie((void*)hdr_end, hdr + sizeof(Trace) - hdr_end);
+  if (hdr_end < hdr + sizeof(Trace)) {
+    ReleaseMemoryPagesToOS(hdr_end, hdr + sizeof(Trace));
+    uptr unused = hdr + sizeof(Trace) - hdr_end;
+    if (hdr_end != (uptr)MmapFixedNoAccess(hdr_end, unused)) {
+      Report("ThreadSanitizer: failed to mprotect(%p, %p)\n",
+          hdr_end, unused);
+      CHECK("unable to mprotect" && 0);
+    }
+  }
   void *mem = internal_alloc(MBlockThreadContex, sizeof(ThreadContext));
   return new(mem) ThreadContext(tid);
 }
@@ -94,42 +114,45 @@ static const u32 kThreadQuarantineSize = 64;
 #endif
 
 Context::Context()
-  : initialized()
-  , report_mtx(MutexTypeReport, StatMtxReport)
-  , nreported()
-  , nmissed_expected()
-  , thread_registry(new(thread_registry_placeholder) ThreadRegistry(
-      CreateThreadContext, kMaxTid, kThreadQuarantineSize, kMaxTidReuse))
-  , racy_mtx(MutexTypeRacy, StatMtxRacy)
-  , racy_stacks()
-  , racy_addresses()
-  , fired_suppressions_mtx(MutexTypeFired, StatMtxFired)
-  , clock_alloc("clock allocator") {
+    : initialized(),
+      report_mtx(MutexTypeReport, StatMtxReport),
+      nreported(),
+      nmissed_expected(),
+      thread_registry(new (thread_registry_placeholder) ThreadRegistry(
+          CreateThreadContext, kMaxTid, kThreadQuarantineSize, kMaxTidReuse)),
+      racy_mtx(MutexTypeRacy, StatMtxRacy),
+      racy_stacks(),
+      racy_addresses(),
+      fired_suppressions_mtx(MutexTypeFired, StatMtxFired),
+      clock_alloc(LINKER_INITIALIZED, "clock allocator") {
   fired_suppressions.reserve(8);
 }
 
 // The objects are allocated in TLS, so one may rely on zero-initialization.
-ThreadState::ThreadState(Context *ctx, int tid, int unique_id, u64 epoch,
-                         unsigned reuse_count,
-                         uptr stk_addr, uptr stk_size,
+ThreadState::ThreadState(Context *ctx, u32 tid, int unique_id, u64 epoch,
+                         unsigned reuse_count, uptr stk_addr, uptr stk_size,
                          uptr tls_addr, uptr tls_size)
-  : fast_state(tid, epoch)
-  // Do not touch these, rely on zero initialization,
-  // they may be accessed before the ctor.
-  // , ignore_reads_and_writes()
-  // , ignore_interceptors()
-  , clock(tid, reuse_count)
+    : fast_state(tid, epoch)
+      // Do not touch these, rely on zero initialization,
+      // they may be accessed before the ctor.
+      // , ignore_reads_and_writes()
+      // , ignore_interceptors()
+      ,
+      clock(tid, reuse_count)
 #if !SANITIZER_GO
-  , jmp_bufs()
+      ,
+      jmp_bufs()
 #endif
-  , tid(tid)
-  , unique_id(unique_id)
-  , stk_addr(stk_addr)
-  , stk_size(stk_size)
-  , tls_addr(tls_addr)
-  , tls_size(tls_size)
+      ,
+      tid(tid),
+      unique_id(unique_id),
+      stk_addr(stk_addr),
+      stk_size(stk_size),
+      tls_addr(tls_addr),
+      tls_size(tls_size)
 #if !SANITIZER_GO
-  , last_sleep_clock(tid)
+      ,
+      last_sleep_clock(tid)
 #endif
 {
 }
@@ -160,12 +183,12 @@ static void *BackgroundThread(void *arg) {
     } else if (internal_strcmp(flags()->profile_memory, "stderr") == 0) {
       mprof_fd = 2;
     } else {
-      InternalScopedString filename(kMaxPathLength);
+      InternalScopedString filename;
       filename.append("%s.%d", flags()->profile_memory, (int)internal_getpid());
       fd_t fd = OpenFile(filename.data(), WrOnly);
       if (fd == kInvalidFd) {
         Printf("ThreadSanitizer: failed to open memory profile file '%s'\n",
-            &filename[0]);
+               filename.data());
       } else {
         mprof_fd = fd;
       }
@@ -351,6 +374,18 @@ static void TsanOnDeadlySignal(int signo, void *siginfo, void *context) {
 }
 #endif
 
+void CheckUnwind() {
+  // There is high probability that interceptors will check-fail as well,
+  // on the other hand there is no sense in processing interceptors
+  // since we are going to die soon.
+  ScopedIgnoreInterceptors ignore;
+#if !SANITIZER_GO
+  cur_thread()->ignore_sync++;
+  cur_thread()->ignore_reads_and_writes++;
+#endif
+  PrintCurrentStackSlow(StackTrace::GetCurrentPc());
+}
+
 void Initialize(ThreadState *thr) {
   // Thread safe because done before all threads exist.
   static bool is_initialized = false;
@@ -361,7 +396,7 @@ void Initialize(ThreadState *thr) {
   ScopedIgnoreInterceptors ignore;
   SanitizerToolName = "ThreadSanitizer";
   // Install tool-specific callbacks in sanitizer_common.
-  SetCheckFailedCallback(TsanCheckFailed);
+  SetCheckUnwindCallback(CheckUnwind);
 
   ctx = new(ctx_placeholder) Context;
   const char *env_name = SANITIZER_GO ? "GORACE" : "TSAN_OPTIONS";
@@ -499,23 +534,27 @@ int Finalize(ThreadState *thr) {
 void ForkBefore(ThreadState *thr, uptr pc) {
   ctx->thread_registry->Lock();
   ctx->report_mtx.Lock();
-  // Ignore memory accesses in the pthread_atfork callbacks.
-  // If any of them triggers a data race we will deadlock
-  // on the report_mtx.
-  // We could ignore interceptors and sync operations as well,
+  // Suppress all reports in the pthread_atfork callbacks.
+  // Reports will deadlock on the report_mtx.
+  // We could ignore sync operations as well,
   // but so far it's unclear if it will do more good or harm.
   // Unnecessarily ignoring things can lead to false positives later.
-  ThreadIgnoreBegin(thr, pc);
+  thr->suppress_reports++;
+  // On OS X, REAL(fork) can call intercepted functions (OSSpinLockLock), and
+  // we'll assert in CheckNoLocks() unless we ignore interceptors.
+  thr->ignore_interceptors++;
 }
 
 void ForkParentAfter(ThreadState *thr, uptr pc) {
-  ThreadIgnoreEnd(thr, pc);  // Begin is in ForkBefore.
+  thr->suppress_reports--;  // Enabled in ForkBefore.
+  thr->ignore_interceptors--;
   ctx->report_mtx.Unlock();
   ctx->thread_registry->Unlock();
 }
 
 void ForkChildAfter(ThreadState *thr, uptr pc) {
-  ThreadIgnoreEnd(thr, pc);  // Begin is in ForkBefore.
+  thr->suppress_reports--;  // Enabled in ForkBefore.
+  thr->ignore_interceptors--;
   ctx->report_mtx.Unlock();
   ctx->thread_registry->Unlock();
 
diff --git a/libsanitizer/tsan/tsan_rtl.h b/libsanitizer/tsan/tsan_rtl.h
index 04d474e..3ae519d 100644
--- a/libsanitizer/tsan/tsan_rtl.h
+++ b/libsanitizer/tsan/tsan_rtl.h
@@ -84,9 +84,6 @@ typedef Allocator::AllocatorCache AllocatorCache;
 Allocator *allocator();
 #endif
 
-void TsanCheckFailed(const char *file, int line, const char *cond,
-                     u64 v1, u64 v2);
-
 const u64 kShadowRodata = (u64)-1;  // .rodata shadow marker
 
 // FastState (from most significant bit):
@@ -406,7 +403,7 @@ struct ThreadState {
 #if TSAN_COLLECT_STATS
   u64 stat[StatCnt];
 #endif
-  const int tid;
+  const u32 tid;
   const int unique_id;
   bool in_symbolizer;
   bool in_ignored_lib;
@@ -447,9 +444,8 @@ struct ThreadState {
 
   const ReportDesc *current_report;
 
-  explicit ThreadState(Context *ctx, int tid, int unique_id, u64 epoch,
-                       unsigned reuse_count,
-                       uptr stk_addr, uptr stk_size,
+  explicit ThreadState(Context *ctx, u32 tid, int unique_id, u64 epoch,
+                       unsigned reuse_count, uptr stk_addr, uptr stk_size,
                        uptr tls_addr, uptr tls_size);
 };
 
@@ -624,6 +620,7 @@ class ScopedReport : public ScopedReportBase {
   ScopedErrorReportLock lock_;
 };
 
+bool ShouldReport(ThreadState *thr, ReportType typ);
 ThreadContext *IsThreadStackOrTls(uptr addr, bool *is_stack);
 void RestoreStack(int tid, const u64 epoch, VarSizeStackTrace *stk,
                   MutexSet *mset, uptr *tag = nullptr);
diff --git a/libsanitizer/tsan/tsan_rtl_mutex.cpp b/libsanitizer/tsan/tsan_rtl_mutex.cpp
index 27897f0..0a8f3aa 100644
--- a/libsanitizer/tsan/tsan_rtl_mutex.cpp
+++ b/libsanitizer/tsan/tsan_rtl_mutex.cpp
@@ -51,6 +51,8 @@ static void ReportMutexMisuse(ThreadState *thr, uptr pc, ReportType typ,
   // or false positives (e.g. unlock in a different thread).
   if (SANITIZER_GO)
     return;
+  if (!ShouldReport(thr, typ))
+    return;
   ThreadRegistryLock l(ctx->thread_registry);
   ScopedReport rep(typ);
   rep.AddMutex(mid);
@@ -96,9 +98,8 @@ void MutexDestroy(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
     ctx->dd->MutexInit(&cb, &s->dd);
   }
   bool unlock_locked = false;
-  if (flags()->report_destroy_locked
-      && s->owner_tid != SyncVar::kInvalidTid
-      && !s->IsFlagSet(MutexFlagBroken)) {
+  if (flags()->report_destroy_locked && s->owner_tid != kInvalidTid &&
+      !s->IsFlagSet(MutexFlagBroken)) {
     s->SetFlags(MutexFlagBroken);
     unlock_locked = true;
   }
@@ -107,7 +108,7 @@ void MutexDestroy(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
   if (!unlock_locked)
     s->Reset(thr->proc());  // must not reset it before the report is printed
   s->mtx.Unlock();
-  if (unlock_locked) {
+  if (unlock_locked && ShouldReport(thr, ReportTypeMutexDestroyLocked)) {
     ThreadRegistryLock l(ctx->thread_registry);
     ScopedReport rep(ReportTypeMutexDestroyLocked);
     rep.AddMutex(mid);
@@ -169,7 +170,7 @@ void MutexPostLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz, int rec) {
   thr->fast_state.IncrementEpoch();
   TraceAddEvent(thr, thr->fast_state, EventTypeLock, s->GetId());
   bool report_double_lock = false;
-  if (s->owner_tid == SyncVar::kInvalidTid) {
+  if (s->owner_tid == kInvalidTid) {
     CHECK_EQ(s->recursion, 0);
     s->owner_tid = thr->tid;
     s->last_lock = thr->fast_state.raw();
@@ -229,7 +230,7 @@ int MutexUnlock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
     s->recursion -= rec;
     if (s->recursion == 0) {
       StatInc(thr, StatMutexUnlock);
-      s->owner_tid = SyncVar::kInvalidTid;
+      s->owner_tid = kInvalidTid;
       ReleaseStoreImpl(thr, pc, &s->clock);
     } else {
       StatInc(thr, StatMutexRecUnlock);
@@ -275,7 +276,7 @@ void MutexPostReadLock(ThreadState *thr, uptr pc, uptr addr, u32 flagz) {
   thr->fast_state.IncrementEpoch();
   TraceAddEvent(thr, thr->fast_state, EventTypeRLock, s->GetId());
   bool report_bad_lock = false;
-  if (s->owner_tid != SyncVar::kInvalidTid) {
+  if (s->owner_tid != kInvalidTid) {
     if (flags()->report_mutex_bugs && !s->IsFlagSet(MutexFlagBroken)) {
       s->SetFlags(MutexFlagBroken);
       report_bad_lock = true;
@@ -314,7 +315,7 @@ void MutexReadUnlock(ThreadState *thr, uptr pc, uptr addr) {
   thr->fast_state.IncrementEpoch();
   TraceAddEvent(thr, thr->fast_state, EventTypeRUnlock, s->GetId());
   bool report_bad_unlock = false;
-  if (s->owner_tid != SyncVar::kInvalidTid) {
+  if (s->owner_tid != kInvalidTid) {
     if (flags()->report_mutex_bugs && !s->IsFlagSet(MutexFlagBroken)) {
       s->SetFlags(MutexFlagBroken);
       report_bad_unlock = true;
@@ -344,7 +345,7 @@ void MutexReadOrWriteUnlock(ThreadState *thr, uptr pc, uptr addr) {
   SyncVar *s = ctx->metamap.GetOrCreateAndLock(thr, pc, addr, true);
   bool write = true;
   bool report_bad_unlock = false;
-  if (s->owner_tid == SyncVar::kInvalidTid) {
+  if (s->owner_tid == kInvalidTid) {
     // Seems to be read unlock.
     write = false;
     StatInc(thr, StatMutexReadUnlock);
@@ -359,7 +360,7 @@ void MutexReadOrWriteUnlock(ThreadState *thr, uptr pc, uptr addr) {
     s->recursion--;
     if (s->recursion == 0) {
       StatInc(thr, StatMutexUnlock);
-      s->owner_tid = SyncVar::kInvalidTid;
+      s->owner_tid = kInvalidTid;
       ReleaseStoreImpl(thr, pc, &s->clock);
     } else {
       StatInc(thr, StatMutexRecUnlock);
@@ -387,7 +388,7 @@ void MutexReadOrWriteUnlock(ThreadState *thr, uptr pc, uptr addr) {
 void MutexRepair(ThreadState *thr, uptr pc, uptr addr) {
   DPrintf("#%d: MutexRepair %zx\n", thr->tid, addr);
   SyncVar *s = ctx->metamap.GetOrCreateAndLock(thr, pc, addr, true);
-  s->owner_tid = SyncVar::kInvalidTid;
+  s->owner_tid = kInvalidTid;
   s->recursion = 0;
   s->mtx.Unlock();
 }
@@ -534,7 +535,7 @@ void AcquireReleaseImpl(ThreadState *thr, uptr pc, SyncClock *c) {
 }
 
 void ReportDeadlock(ThreadState *thr, uptr pc, DDReport *r) {
-  if (r == 0)
+  if (r == 0 || !ShouldReport(thr, ReportTypeDeadlock))
     return;
   ThreadRegistryLock l(ctx->thread_registry);
   ScopedReport rep(ReportTypeDeadlock);
diff --git a/libsanitizer/tsan/tsan_rtl_ppc64.S b/libsanitizer/tsan/tsan_rtl_ppc64.S
index 9e533a7..8285e21 100644
--- a/libsanitizer/tsan/tsan_rtl_ppc64.S
+++ b/libsanitizer/tsan/tsan_rtl_ppc64.S
@@ -1,6 +1,5 @@
 #include "tsan_ppc_regs.h"
 
-        .machine altivec
         .section .text
         .hidden __tsan_setjmp
         .globl _setjmp
diff --git a/libsanitizer/tsan/tsan_rtl_report.cpp b/libsanitizer/tsan/tsan_rtl_report.cpp
index 208d0df..706794f 100644
--- a/libsanitizer/tsan/tsan_rtl_report.cpp
+++ b/libsanitizer/tsan/tsan_rtl_report.cpp
@@ -31,23 +31,6 @@ using namespace __sanitizer;
 
 static ReportStack *SymbolizeStack(StackTrace trace);
 
-void TsanCheckFailed(const char *file, int line, const char *cond,
-                     u64 v1, u64 v2) {
-  // There is high probability that interceptors will check-fail as well,
-  // on the other hand there is no sense in processing interceptors
-  // since we are going to die soon.
-  ScopedIgnoreInterceptors ignore;
-#if !SANITIZER_GO
-  cur_thread()->ignore_sync++;
-  cur_thread()->ignore_reads_and_writes++;
-#endif
-  Printf("FATAL: ThreadSanitizer CHECK failed: "
-         "%s:%d \"%s\" (0x%zx, 0x%zx)\n",
-         file, line, cond, (uptr)v1, (uptr)v2);
-  PrintCurrentStackSlow(StackTrace::GetCurrentPc());
-  Die();
-}
-
 // Can be overriden by an application/test to intercept reports.
 #ifdef TSAN_EXTERNAL_HOOKS
 bool OnReport(const ReportDesc *rep, bool suppressed);
@@ -142,6 +125,34 @@ static ReportStack *SymbolizeStack(StackTrace trace) {
   return stack;
 }
 
+bool ShouldReport(ThreadState *thr, ReportType typ) {
+  // We set thr->suppress_reports in the fork context.
+  // Taking any locking in the fork context can lead to deadlocks.
+  // If any locks are already taken, it's too late to do this check.
+  CheckNoLocks(thr);
+  // For the same reason check we didn't lock thread_registry yet.
+  if (SANITIZER_DEBUG)
+    ThreadRegistryLock l(ctx->thread_registry);
+  if (!flags()->report_bugs || thr->suppress_reports)
+    return false;
+  switch (typ) {
+    case ReportTypeSignalUnsafe:
+      return flags()->report_signal_unsafe;
+    case ReportTypeThreadLeak:
+#if !SANITIZER_GO
+      // It's impossible to join phantom threads
+      // in the child after fork.
+      if (ctx->after_multithreaded_fork)
+        return false;
+#endif
+      return flags()->report_thread_leaks;
+    case ReportTypeMutexDestroyLocked:
+      return flags()->report_destroy_locked;
+    default:
+      return true;
+  }
+}
+
 ScopedReportBase::ScopedReportBase(ReportType typ, uptr tag) {
   ctx->thread_registry->CheckLocked();
   void *mem = internal_alloc(MBlockReport, sizeof(ReportDesc));
@@ -497,8 +508,10 @@ static bool HandleRacyAddress(ThreadState *thr, uptr addr_min, uptr addr_max) {
 }
 
 bool OutputReport(ThreadState *thr, const ScopedReport &srep) {
-  if (!flags()->report_bugs || thr->suppress_reports)
-    return false;
+  // These should have been checked in ShouldReport.
+  // It's too late to check them here, we have already taken locks.
+  CHECK(flags()->report_bugs);
+  CHECK(!thr->suppress_reports);
   atomic_store_relaxed(&ctx->last_symbolize_time_ns, NanoTime());
   const ReportDesc *rep = srep.GetReport();
   CHECK_EQ(thr->current_report, nullptr);
@@ -589,7 +602,7 @@ void ReportRace(ThreadState *thr) {
   // at best it will cause deadlocks on internal mutexes.
   ScopedIgnoreInterceptors ignore;
 
-  if (!flags()->report_bugs)
+  if (!ShouldReport(thr, ReportTypeRace))
     return;
   if (!flags()->report_atomic_races && !RaceBetweenAtomicAndFree(thr))
     return;
@@ -722,8 +735,7 @@ void PrintCurrentStack(ThreadState *thr, uptr pc) {
 // However, this solution is not reliable enough, please see dvyukov's comment
 // http://reviews.llvm.org/D19148#406208
 // Also see PR27280 comment 2 and 3 for breaking examples and analysis.
-ALWAYS_INLINE
-void PrintCurrentStackSlow(uptr pc) {
+ALWAYS_INLINE USED void PrintCurrentStackSlow(uptr pc) {
 #if !SANITIZER_GO
   uptr bp = GET_CURRENT_FRAME();
   BufferedStackTrace *ptrace =
diff --git a/libsanitizer/tsan/tsan_rtl_thread.cpp b/libsanitizer/tsan/tsan_rtl_thread.cpp
index d801467..6d1ccd8 100644
--- a/libsanitizer/tsan/tsan_rtl_thread.cpp
+++ b/libsanitizer/tsan/tsan_rtl_thread.cpp
@@ -51,7 +51,7 @@ struct OnCreatedArgs {
 
 void ThreadContext::OnCreated(void *arg) {
   thr = 0;
-  if (tid == 0)
+  if (tid == kMainTid)
     return;
   OnCreatedArgs *args = static_cast<OnCreatedArgs *>(arg);
   if (!args->thr)  // GCD workers don't have a parent thread.
@@ -179,7 +179,7 @@ static void MaybeReportThreadLeak(ThreadContextBase *tctx_base, void *arg) {
 
 #if !SANITIZER_GO
 static void ReportIgnoresEnabled(ThreadContext *tctx, IgnoreSet *set) {
-  if (tctx->tid == 0) {
+  if (tctx->tid == kMainTid) {
     Printf("ThreadSanitizer: main thread finished with ignores enabled\n");
   } else {
     Printf("ThreadSanitizer: thread T%d %s finished with ignores enabled,"
@@ -210,7 +210,7 @@ static void ThreadCheckIgnore(ThreadState *thr) {}
 void ThreadFinalize(ThreadState *thr) {
   ThreadCheckIgnore(thr);
 #if !SANITIZER_GO
-  if (!flags()->report_thread_leaks)
+  if (!ShouldReport(thr, ReportTypeThreadLeak))
     return;
   ThreadRegistryLock l(ctx->thread_registry);
   Vector<ThreadLeak> leaks;
@@ -250,9 +250,10 @@ void ThreadStart(ThreadState *thr, int tid, tid_t os_id,
   uptr tls_size = 0;
 #if !SANITIZER_GO
   if (thread_type != ThreadType::Fiber)
-    GetThreadStackAndTls(tid == 0, &stk_addr, &stk_size, &tls_addr, &tls_size);
+    GetThreadStackAndTls(tid == kMainTid, &stk_addr, &stk_size, &tls_addr,
+                         &tls_size);
 
-  if (tid) {
+  if (tid != kMainTid) {
     if (stk_addr && stk_size)
       MemoryRangeImitateWrite(thr, /*pc=*/ 1, stk_addr, stk_size);
 
@@ -313,7 +314,7 @@ static bool ConsumeThreadByUid(ThreadContextBase *tctx, void *arg) {
 int ThreadConsumeTid(ThreadState *thr, uptr pc, uptr uid) {
   ConsumeThreadContext findCtx = {uid, nullptr};
   ctx->thread_registry->FindThread(ConsumeThreadByUid, &findCtx);
-  int tid = findCtx.tctx ? findCtx.tctx->tid : ThreadRegistry::kUnknownTid;
+  int tid = findCtx.tctx ? findCtx.tctx->tid : kInvalidTid;
   DPrintf("#%d: ThreadTid uid=%zu tid=%d\n", thr->tid, uid, tid);
   return tid;
 }
diff --git a/libsanitizer/tsan/tsan_sync.cpp b/libsanitizer/tsan/tsan_sync.cpp
index 17ddd50f..ba24f98 100644
--- a/libsanitizer/tsan/tsan_sync.cpp
+++ b/libsanitizer/tsan/tsan_sync.cpp
@@ -53,8 +53,8 @@ void SyncVar::Reset(Processor *proc) {
 }
 
 MetaMap::MetaMap()
-    : block_alloc_("heap block allocator")
-    , sync_alloc_("sync allocator") {
+    : block_alloc_(LINKER_INITIALIZED, "heap block allocator"),
+      sync_alloc_(LINKER_INITIALIZED, "sync allocator") {
   atomic_store(&uid_gen_, 0, memory_order_relaxed);
 }
 
diff --git a/libsanitizer/tsan/tsan_sync.h b/libsanitizer/tsan/tsan_sync.h
index 47f2739..c4056f6 100644
--- a/libsanitizer/tsan/tsan_sync.h
+++ b/libsanitizer/tsan/tsan_sync.h
@@ -50,13 +50,11 @@ enum MutexFlags {
 struct SyncVar {
   SyncVar();
 
-  static const int kInvalidTid = -1;
-
   uptr addr;  // overwritten by DenseSlabAlloc freelist
   Mutex mtx;
   u64 uid;  // Globally unique id.
   u32 creation_stack_id;
-  int owner_tid;  // Set only by exclusive owners.
+  u32 owner_tid;  // Set only by exclusive owners.
   u64 last_lock;
   int recursion;
   atomic_uint32_t flags;
@@ -130,8 +128,8 @@ class MetaMap {
   static const u32 kFlagMask  = 3u << 30;
   static const u32 kFlagBlock = 1u << 30;
   static const u32 kFlagSync  = 2u << 30;
-  typedef DenseSlabAlloc<MBlock, 1<<16, 1<<12> BlockAlloc;
-  typedef DenseSlabAlloc<SyncVar, 1<<16, 1<<10> SyncAlloc;
+  typedef DenseSlabAlloc<MBlock, 1 << 18, 1 << 12, kFlagMask> BlockAlloc;
+  typedef DenseSlabAlloc<SyncVar, 1 << 20, 1 << 10, kFlagMask> SyncAlloc;
   BlockAlloc block_alloc_;
   SyncAlloc sync_alloc_;
   atomic_uint64_t uid_gen_;
diff --git a/libsanitizer/ubsan/ubsan_diag.cpp b/libsanitizer/ubsan/ubsan_diag.cpp
index 1b2828d..ef2e495 100644
--- a/libsanitizer/ubsan/ubsan_diag.cpp
+++ b/libsanitizer/ubsan/ubsan_diag.cpp
@@ -278,7 +278,7 @@ static void PrintMemorySnippet(const Decorator &Decor, MemoryLocation Loc,
   }
 
   // Emit data.
-  InternalScopedString Buffer(1024);
+  InternalScopedString Buffer;
   for (uptr P = Min; P != Max; ++P) {
     unsigned char C = *reinterpret_cast<const unsigned char*>(P);
     Buffer.append("%s%02x", (P % 8 == 0) ? "  " : " ", C);
@@ -346,7 +346,7 @@ Diag::~Diag() {
   // All diagnostics should be printed under report mutex.
   ScopedReport::CheckLocked();
   Decorator Decor;
-  InternalScopedString Buffer(1024);
+  InternalScopedString Buffer;
 
   // Prepare a report that a monitor process can inspect.
   if (Level == DL_Error) {
@@ -388,6 +388,10 @@ ScopedReport::ScopedReport(ReportOptions Opts, Location SummaryLoc,
 ScopedReport::~ScopedReport() {
   MaybePrintStackTrace(Opts.pc, Opts.bp);
   MaybeReportErrorSummary(SummaryLoc, Type);
+
+  if (common_flags()->print_module_map >= 2)
+    DumpProcessMap();
+
   if (flags()->halt_on_error)
     Die();
 }
diff --git a/libsanitizer/ubsan/ubsan_flags.cpp b/libsanitizer/ubsan/ubsan_flags.cpp
index 9a66bd3..25cefd4 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 2184625..e201e6b 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 9f41235..219fb15 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_init.cpp b/libsanitizer/ubsan/ubsan_init.cpp
index e0be5a7..9931d85 100644
--- a/libsanitizer/ubsan/ubsan_init.cpp
+++ b/libsanitizer/ubsan/ubsan_init.cpp
@@ -33,6 +33,11 @@ static void CommonInit() {
   InitializeSuppressions();
 }
 
+static void UbsanDie() {
+  if (common_flags()->print_module_map >= 1)
+    DumpProcessMap();
+}
+
 static void CommonStandaloneInit() {
   SanitizerToolName = GetSanititizerToolName();
   CacheBinaryName();
@@ -42,6 +47,10 @@ static void CommonStandaloneInit() {
   AndroidLogInit();
   InitializeCoverage(common_flags()->coverage, common_flags()->coverage_dir);
   CommonInit();
+
+  // Only add die callback when running in standalone mode to avoid printing
+  // the same information from multiple sanitizers' output
+  AddDieCallback(UbsanDie);
   Symbolizer::LateInitialize();
 }
 
diff --git a/libsanitizer/ubsan/ubsan_monitor.cpp b/libsanitizer/ubsan/ubsan_monitor.cpp
index d064e95..69dd986 100644
--- a/libsanitizer/ubsan/ubsan_monitor.cpp
+++ b/libsanitizer/ubsan/ubsan_monitor.cpp
@@ -17,7 +17,7 @@ using namespace __ubsan;
 UndefinedBehaviorReport::UndefinedBehaviorReport(const char *IssueKind,
                                                  Location &Loc,
                                                  InternalScopedString &Msg)
-    : IssueKind(IssueKind), Loc(Loc), Buffer(Msg.length() + 1) {
+    : IssueKind(IssueKind), Loc(Loc) {
   // We have the common sanitizer reporting lock, so it's safe to register a
   // new UB report.
   RegisterUndefinedBehaviorReport(this);
@@ -52,9 +52,9 @@ void __ubsan::__ubsan_get_current_report_data(const char **OutIssueKind,
 
   // Ensure that the first character of the diagnostic text can't start with a
   // lowercase letter.
-  char FirstChar = Buf.data()[0];
+  char FirstChar = *Buf.data();
   if (FirstChar >= 'a' && FirstChar <= 'z')
-    Buf.data()[0] = FirstChar - 'a' + 'A';
+    *Buf.data() += 'A' - 'a';
 
   *OutIssueKind = CurrentUBR->IssueKind;
   *OutMessage = Buf.data();
diff --git a/libsanitizer/ubsan/ubsan_platform.h b/libsanitizer/ubsan/ubsan_platform.h
index 98542fce..51e535d 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(__NetBSD__) || defined(__DragonFly__) || \
     (defined(__sun__) && defined(__svr4__)) || \
     defined(_WIN32) || defined(__Fuchsia__) || defined(__rtems__)
 # define CAN_SANITIZE_UB 1
 #else
 # define CAN_SANITIZE_UB 0
 #endif
-#endif //CAN_SANITIZE_UB
 
 #endif