diff --git a/CMakeLists.txt b/CMakeLists.txt
index 51769088a992..3152233587c6 100644
--- a/CMakeLists.txt
+++ b/CMakeLists.txt
@@ -7,14 +7,14 @@
 # An important constraint of the build is that it only produces libraries
 # based on the ability of the host toolchain to target various platforms.
 
+cmake_minimum_required(VERSION 3.4.3)
+
 # Check if compiler-rt is built as a standalone project.
 if (CMAKE_SOURCE_DIR STREQUAL CMAKE_CURRENT_SOURCE_DIR OR COMPILER_RT_STANDALONE_BUILD)
   project(CompilerRT C CXX ASM)
   set(COMPILER_RT_STANDALONE_BUILD TRUE)
 endif()
 
-cmake_minimum_required(VERSION 3.4.3)
-
 # Add path for custom compiler-rt modules.
 list(INSERT CMAKE_MODULE_PATH 0
   "${CMAKE_CURRENT_SOURCE_DIR}/cmake"
diff --git a/cmake/config-ix.cmake b/cmake/config-ix.cmake
index ae2a262a14a9..0da98b9a489d 100644
--- a/cmake/config-ix.cmake
+++ b/cmake/config-ix.cmake
@@ -179,7 +179,7 @@ set(ALL_UBSAN_SUPPORTED_ARCH ${X86} ${X86_64} ${ARM32} ${ARM64}
 set(ALL_SAFESTACK_SUPPORTED_ARCH ${X86} ${X86_64} ${ARM64} ${MIPS32} ${MIPS64})
 set(ALL_CFI_SUPPORTED_ARCH ${X86} ${X86_64} ${MIPS64})
 set(ALL_ESAN_SUPPORTED_ARCH ${X86_64} ${MIPS64})
-set(ALL_SCUDO_SUPPORTED_ARCH ${X86} ${X86_64} ${ARM32} ${ARM64})
+set(ALL_SCUDO_SUPPORTED_ARCH ${X86} ${X86_64} ${ARM32} ${ARM64} ${MIPS32} ${MIPS64})
 set(ALL_XRAY_SUPPORTED_ARCH ${X86_64} ${ARM32} ${ARM64} ${MIPS32} ${MIPS64} powerpc64le)
 
 if(APPLE)
diff --git a/include/xray/xray_interface.h b/include/xray/xray_interface.h
index dc0c277aa841..564613417069 100644
--- a/include/xray/xray_interface.h
+++ b/include/xray/xray_interface.h
@@ -60,7 +60,8 @@ extern int __xray_remove_handler();
 /// start logging their subsequent affected function calls (if patched).
 ///
 /// Returns 1 on success, 0 on error.
-extern int __xray_set_handler_arg1(void (*)(int32_t, XRayEntryType, uint64_t));
+extern int __xray_set_handler_arg1(void (*entry)(int32_t, XRayEntryType,
+                                                 uint64_t));
 
 /// Disables the XRay handler used to log first arguments of function calls.
 /// Returns 1 on success, 0 on error.
diff --git a/lib/asan/asan_allocator.cc b/lib/asan/asan_allocator.cc
index 1ded7794c9f4..57651f49cdb9 100644
--- a/lib/asan/asan_allocator.cc
+++ b/lib/asan/asan_allocator.cc
@@ -160,7 +160,7 @@ struct QuarantineCallback {
   }
 
   void *Allocate(uptr size) {
-    return get_allocator().Allocate(cache_, size, 1, false);
+    return get_allocator().Allocate(cache_, size, 1);
   }
 
   void Deallocate(void *p) {
@@ -266,7 +266,8 @@ struct Allocator {
   }
 
   void Initialize(const AllocatorOptions &options) {
-    allocator.Init(options.may_return_null, options.release_to_os_interval_ms);
+    SetAllocatorMayReturnNull(options.may_return_null);
+    allocator.Init(options.release_to_os_interval_ms);
     SharedInitCode(options);
   }
 
@@ -302,7 +303,7 @@ struct Allocator {
   }
 
   void ReInitialize(const AllocatorOptions &options) {
-    allocator.SetMayReturnNull(options.may_return_null);
+    SetAllocatorMayReturnNull(options.may_return_null);
     allocator.SetReleaseToOSIntervalMs(options.release_to_os_interval_ms);
     SharedInitCode(options);
 
@@ -323,7 +324,7 @@ struct Allocator {
     options->thread_local_quarantine_size_kb = quarantine.GetCacheSize() >> 10;
     options->min_redzone = atomic_load(&min_redzone, memory_order_acquire);
     options->max_redzone = atomic_load(&max_redzone, memory_order_acquire);
-    options->may_return_null = allocator.MayReturnNull();
+    options->may_return_null = AllocatorMayReturnNull();
     options->alloc_dealloc_mismatch =
         atomic_load(&alloc_dealloc_mismatch, memory_order_acquire);
     options->release_to_os_interval_ms = allocator.ReleaseToOSIntervalMs();
@@ -374,7 +375,7 @@ struct Allocator {
     if (UNLIKELY(!asan_inited))
       AsanInitFromRtl();
     if (RssLimitExceeded())
-      return allocator.ReturnNullOrDieOnOOM();
+      return AsanAllocator::FailureHandler::OnOOM();
     Flags &fl = *flags();
     CHECK(stack);
     const uptr min_alignment = SHADOW_GRANULARITY;
@@ -407,23 +408,21 @@ struct Allocator {
     if (size > kMaxAllowedMallocSize || needed_size > kMaxAllowedMallocSize) {
       Report("WARNING: AddressSanitizer failed to allocate 0x%zx bytes\n",
              (void*)size);
-      return allocator.ReturnNullOrDieOnBadRequest();
+      return AsanAllocator::FailureHandler::OnBadRequest();
     }
 
     AsanThread *t = GetCurrentThread();
     void *allocated;
     if (t) {
       AllocatorCache *cache = GetAllocatorCache(&t->malloc_storage());
-      allocated =
-          allocator.Allocate(cache, needed_size, 8, false);
+      allocated = allocator.Allocate(cache, needed_size, 8);
     } else {
       SpinMutexLock l(&fallback_mutex);
       AllocatorCache *cache = &fallback_allocator_cache;
-      allocated =
-          allocator.Allocate(cache, needed_size, 8, false);
+      allocated = allocator.Allocate(cache, needed_size, 8);
     }
-
-    if (!allocated) return allocator.ReturnNullOrDieOnOOM();
+    if (!allocated)
+      return nullptr;
 
     if (*(u8 *)MEM_TO_SHADOW((uptr)allocated) == 0 && CanPoisonMemory()) {
       // Heap poisoning is enabled, but the allocator provides an unpoisoned
@@ -634,7 +633,7 @@ struct Allocator {
 
   void *Calloc(uptr nmemb, uptr size, BufferedStackTrace *stack) {
     if (CallocShouldReturnNullDueToOverflow(size, nmemb))
-      return allocator.ReturnNullOrDieOnBadRequest();
+      return AsanAllocator::FailureHandler::OnBadRequest();
     void *ptr = Allocate(nmemb * size, 8, stack, FROM_MALLOC, false);
     // If the memory comes from the secondary allocator no need to clear it
     // as it comes directly from mmap.
diff --git a/lib/asan/asan_report.cc b/lib/asan/asan_report.cc
index f751b6184c6b..2e477f258b8d 100644
--- a/lib/asan/asan_report.cc
+++ b/lib/asan/asan_report.cc
@@ -204,6 +204,14 @@ class ScopedInErrorReport {
       error_report_callback(buffer_copy.data());
     }
 
+    if (halt_on_error_ && common_flags()->abort_on_error) {
+      // On Android the message is truncated to 512 characters.
+      // FIXME: implement "compact" error format, possibly without, or with
+      // highly compressed stack traces?
+      // FIXME: or just use the summary line as abort message?
+      SetAbortMessage(buffer_copy.data());
+    }
+
     // In halt_on_error = false mode, reset the current error object (before
     // unlocking).
     if (!halt_on_error_)
diff --git a/lib/asan/scripts/asan_device_setup b/lib/asan/scripts/asan_device_setup
index c807df3cd252..79ac2f916569 100755
--- a/lib/asan/scripts/asan_device_setup
+++ b/lib/asan/scripts/asan_device_setup
@@ -410,15 +410,15 @@ if ! ( cd "$TMPDIRBASE" && diff -qr old/ new/ ) ; then
       install "$TMPDIR/asanwrapper" /system/bin 755
       install "$TMPDIR/asanwrapper64" /system/bin 755
 
-      adb_shell ln -s $ASAN_RT /system/lib/$ASAN_RT_SYMLINK
-      adb_shell ln -s $ASAN_RT64 /system/lib64/$ASAN_RT_SYMLINK
+      adb_shell ln -sf $ASAN_RT /system/lib/$ASAN_RT_SYMLINK
+      adb_shell ln -sf $ASAN_RT64 /system/lib64/$ASAN_RT_SYMLINK
     else
       install "$TMPDIR/$ASAN_RT" /system/lib 644
       install "$TMPDIR/app_process32" /system/bin 755 $CTX
       install "$TMPDIR/app_process.wrap" /system/bin 755 $CTX
       install "$TMPDIR/asanwrapper" /system/bin 755 $CTX
 
-      adb_shell ln -s $ASAN_RT /system/lib/$ASAN_RT_SYMLINK
+      adb_shell ln -sf $ASAN_RT /system/lib/$ASAN_RT_SYMLINK
 
       adb_shell rm /system/bin/app_process
       adb_shell ln -s /system/bin/app_process.wrap /system/bin/app_process
diff --git a/lib/asan/weak_symbols.txt b/lib/asan/weak_symbols.txt
index ba7b0272c2b1..fe680f8a9a4f 100644
--- a/lib/asan/weak_symbols.txt
+++ b/lib/asan/weak_symbols.txt
@@ -1,3 +1,12 @@
 ___asan_default_options
 ___asan_default_suppressions
 ___asan_on_error
+___asan_set_shadow_00
+___asan_set_shadow_f1
+___asan_set_shadow_f2
+___asan_set_shadow_f3
+___asan_set_shadow_f4
+___asan_set_shadow_f5
+___asan_set_shadow_f6
+___asan_set_shadow_f7
+___asan_set_shadow_f8
diff --git a/lib/interception/interception_win.cc b/lib/interception/interception_win.cc
index e4f3d358f40c..b2902d57f542 100644
--- a/lib/interception/interception_win.cc
+++ b/lib/interception/interception_win.cc
@@ -477,7 +477,7 @@ static size_t GetInstructionSize(uptr address, size_t* rel_offset = nullptr) {
   switch (*(u8*)address) {
     case 0xA1:  // A1 XX XX XX XX XX XX XX XX :
                 //   movabs eax, dword ptr ds:[XXXXXXXX]
-      return 8;
+      return 9;
   }
 
   switch (*(u16*)address) {
@@ -495,6 +495,11 @@ static size_t GetInstructionSize(uptr address, size_t* rel_offset = nullptr) {
     case 0x5741:  // push r15
     case 0x9066:  // Two-byte NOP
       return 2;
+
+    case 0x058B:  // 8B 05 XX XX XX XX : mov eax, dword ptr [XX XX XX XX]
+      if (rel_offset)
+        *rel_offset = 2;
+      return 6;
   }
 
   switch (0x00FFFFFF & *(u32*)address) {
diff --git a/lib/interception/tests/interception_win_test.cc b/lib/interception/tests/interception_win_test.cc
index a705768d6d5d..37ef994f83da 100644
--- a/lib/interception/tests/interception_win_test.cc
+++ b/lib/interception/tests/interception_win_test.cc
@@ -170,6 +170,13 @@ const u8 kPatchableCode5[] = {
     0x54,                                      // push    esp
 };
 
+#if SANITIZER_WINDOWS64
+u8 kLoadGlobalCode[] = {
+  0x8B, 0x05, 0x00, 0x00, 0x00, 0x00, // mov    eax [rip + global]
+  0xC3,                               // ret
+};
+#endif
+
 const u8 kUnpatchableCode1[] = {
     0xC3,                           // ret
 };
@@ -502,6 +509,10 @@ TEST(Interception, PatchableFunction) {
   EXPECT_TRUE(TestFunctionPatching(kPatchableCode4, override));
   EXPECT_TRUE(TestFunctionPatching(kPatchableCode5, override));
 
+#if SANITIZER_WINDOWS64
+  EXPECT_TRUE(TestFunctionPatching(kLoadGlobalCode, override));
+#endif
+
   EXPECT_FALSE(TestFunctionPatching(kUnpatchableCode1, override));
   EXPECT_FALSE(TestFunctionPatching(kUnpatchableCode2, override));
   EXPECT_FALSE(TestFunctionPatching(kUnpatchableCode3, override));
diff --git a/lib/lsan/CMakeLists.txt b/lib/lsan/CMakeLists.txt
index 55c825f5c6ea..bd3a96f328b1 100644
--- a/lib/lsan/CMakeLists.txt
+++ b/lib/lsan/CMakeLists.txt
@@ -13,6 +13,7 @@ set(LSAN_SOURCES
   lsan_allocator.cc
   lsan_linux.cc
   lsan_interceptors.cc
+  lsan_mac.cc
   lsan_malloc_mac.cc
   lsan_preinit.cc
   lsan_thread.cc)
diff --git a/lib/lsan/lsan.h b/lib/lsan/lsan.h
index 1061d2fcfde7..7446e9507e0d 100644
--- a/lib/lsan/lsan.h
+++ b/lib/lsan/lsan.h
@@ -38,6 +38,8 @@
   GET_STACK_TRACE(__sanitizer::common_flags()->malloc_context_size, \
                   common_flags()->fast_unwind_on_malloc)
 
+#define GET_STACK_TRACE_THREAD GET_STACK_TRACE(kStackTraceMax, true)
+
 namespace __lsan {
 
 void InitializeInterceptors();
diff --git a/lib/lsan/lsan_allocator.cc b/lib/lsan/lsan_allocator.cc
index 011979eee396..f54e953731b4 100644
--- a/lib/lsan/lsan_allocator.cc
+++ b/lib/lsan/lsan_allocator.cc
@@ -38,8 +38,8 @@ typedef CombinedAllocator<PrimaryAllocator, AllocatorCache,
 static Allocator allocator;
 
 void InitializeAllocator() {
+  SetAllocatorMayReturnNull(common_flags()->allocator_may_return_null);
   allocator.InitLinkerInitialized(
-      common_flags()->allocator_may_return_null,
       common_flags()->allocator_release_to_os_interval_ms);
 }
 
@@ -76,7 +76,7 @@ void *Allocate(const StackTrace &stack, uptr size, uptr alignment,
     Report("WARNING: LeakSanitizer failed to allocate %zu bytes\n", size);
     return nullptr;
   }
-  void *p = allocator.Allocate(GetAllocatorCache(), size, alignment, false);
+  void *p = allocator.Allocate(GetAllocatorCache(), size, alignment);
   // Do not rely on the allocator to clear the memory (it's slow).
   if (cleared && allocator.FromPrimary(p))
     memset(p, 0, size);
diff --git a/lib/lsan/lsan_common_mac.cc b/lib/lsan/lsan_common_mac.cc
index 114dd8c9019e..adde3a1b4035 100644
--- a/lib/lsan/lsan_common_mac.cc
+++ b/lib/lsan/lsan_common_mac.cc
@@ -79,8 +79,7 @@ void EnableInThisThread() {
 
 u32 GetCurrentThread() {
   thread_local_data_t *data = get_tls_val(false);
-  CHECK(data);
-  return data->current_thread_id;
+  return data ? data->current_thread_id : kInvalidTid;
 }
 
 void SetCurrentThread(u32 tid) { get_tls_val(true)->current_thread_id = tid; }
diff --git a/lib/lsan/lsan_mac.cc b/lib/lsan/lsan_mac.cc
new file mode 100644
index 000000000000..1a6f5f4899b1
--- /dev/null
+++ b/lib/lsan/lsan_mac.cc
@@ -0,0 +1,192 @@
+//===-- lsan_mac.cc -------------------------------------------------------===//
+//
+//                     The LLVM Compiler Infrastructure
+//
+// This file is distributed under the University of Illinois Open Source
+// License. See LICENSE.TXT for details.
+//
+//===----------------------------------------------------------------------===//
+//
+// This file is a part of LeakSanitizer, a memory leak checker.
+//
+// Mac-specific details.
+//===----------------------------------------------------------------------===//
+
+#include "sanitizer_common/sanitizer_platform.h"
+#if SANITIZER_MAC
+
+#include "interception/interception.h"
+#include "lsan.h"
+#include "lsan_allocator.h"
+#include "lsan_thread.h"
+
+#include <pthread.h>
+
+namespace __lsan {
+// Support for the following functions from libdispatch on Mac OS:
+//   dispatch_async_f()
+//   dispatch_async()
+//   dispatch_sync_f()
+//   dispatch_sync()
+//   dispatch_after_f()
+//   dispatch_after()
+//   dispatch_group_async_f()
+//   dispatch_group_async()
+// TODO(glider): libdispatch API contains other functions that we don't support
+// yet.
+//
+// dispatch_sync() and dispatch_sync_f() are synchronous, although chances are
+// they can cause jobs to run on a thread different from the current one.
+// TODO(glider): if so, we need a test for this (otherwise we should remove
+// them).
+//
+// The following functions use dispatch_barrier_async_f() (which isn't a library
+// function but is exported) and are thus supported:
+//   dispatch_source_set_cancel_handler_f()
+//   dispatch_source_set_cancel_handler()
+//   dispatch_source_set_event_handler_f()
+//   dispatch_source_set_event_handler()
+//
+// The reference manual for Grand Central Dispatch is available at
+//   http://developer.apple.com/library/mac/#documentation/Performance/Reference/GCD_libdispatch_Ref/Reference/reference.html
+// The implementation details are at
+//   http://libdispatch.macosforge.org/trac/browser/trunk/src/queue.c
+
+typedef void *dispatch_group_t;
+typedef void *dispatch_queue_t;
+typedef void *dispatch_source_t;
+typedef u64 dispatch_time_t;
+typedef void (*dispatch_function_t)(void *block);
+typedef void *(*worker_t)(void *block);
+
+// A wrapper for the ObjC blocks used to support libdispatch.
+typedef struct {
+  void *block;
+  dispatch_function_t func;
+  u32 parent_tid;
+} lsan_block_context_t;
+
+ALWAYS_INLINE
+void lsan_register_worker_thread(int parent_tid) {
+  if (GetCurrentThread() == kInvalidTid) {
+    u32 tid = ThreadCreate(parent_tid, 0, true);
+    ThreadStart(tid, GetTid());
+    SetCurrentThread(tid);
+  }
+}
+
+// For use by only those functions that allocated the context via
+// alloc_lsan_context().
+extern "C" void lsan_dispatch_call_block_and_release(void *block) {
+  lsan_block_context_t *context = (lsan_block_context_t *)block;
+  VReport(2,
+          "lsan_dispatch_call_block_and_release(): "
+          "context: %p, pthread_self: %p\n",
+          block, pthread_self());
+  lsan_register_worker_thread(context->parent_tid);
+  // Call the original dispatcher for the block.
+  context->func(context->block);
+  lsan_free(context);
+}
+
+}  // namespace __lsan
+
+using namespace __lsan;  // NOLINT
+
+// Wrap |ctxt| and |func| into an lsan_block_context_t.
+// The caller retains control of the allocated context.
+extern "C" lsan_block_context_t *alloc_lsan_context(void *ctxt,
+                                                    dispatch_function_t func) {
+  GET_STACK_TRACE_THREAD;
+  lsan_block_context_t *lsan_ctxt =
+      (lsan_block_context_t *)lsan_malloc(sizeof(lsan_block_context_t), stack);
+  lsan_ctxt->block = ctxt;
+  lsan_ctxt->func = func;
+  lsan_ctxt->parent_tid = GetCurrentThread();
+  return lsan_ctxt;
+}
+
+// Define interceptor for dispatch_*_f function with the three most common
+// parameters: dispatch_queue_t, context, dispatch_function_t.
+#define INTERCEPT_DISPATCH_X_F_3(dispatch_x_f)                        \
+  INTERCEPTOR(void, dispatch_x_f, dispatch_queue_t dq, void *ctxt,    \
+              dispatch_function_t func) {                             \
+    lsan_block_context_t *lsan_ctxt = alloc_lsan_context(ctxt, func); \
+    return REAL(dispatch_x_f)(dq, (void *)lsan_ctxt,                  \
+                              lsan_dispatch_call_block_and_release);  \
+  }
+
+INTERCEPT_DISPATCH_X_F_3(dispatch_async_f)
+INTERCEPT_DISPATCH_X_F_3(dispatch_sync_f)
+INTERCEPT_DISPATCH_X_F_3(dispatch_barrier_async_f)
+
+INTERCEPTOR(void, dispatch_after_f, dispatch_time_t when, dispatch_queue_t dq,
+            void *ctxt, dispatch_function_t func) {
+  lsan_block_context_t *lsan_ctxt = alloc_lsan_context(ctxt, func);
+  return REAL(dispatch_after_f)(when, dq, (void *)lsan_ctxt,
+                                lsan_dispatch_call_block_and_release);
+}
+
+INTERCEPTOR(void, dispatch_group_async_f, dispatch_group_t group,
+            dispatch_queue_t dq, void *ctxt, dispatch_function_t func) {
+  lsan_block_context_t *lsan_ctxt = alloc_lsan_context(ctxt, func);
+  REAL(dispatch_group_async_f)
+  (group, dq, (void *)lsan_ctxt, lsan_dispatch_call_block_and_release);
+}
+
+#if !defined(MISSING_BLOCKS_SUPPORT)
+extern "C" {
+void dispatch_async(dispatch_queue_t dq, void (^work)(void));
+void dispatch_group_async(dispatch_group_t dg, dispatch_queue_t dq,
+                          void (^work)(void));
+void dispatch_after(dispatch_time_t when, dispatch_queue_t queue,
+                    void (^work)(void));
+void dispatch_source_set_cancel_handler(dispatch_source_t ds,
+                                        void (^work)(void));
+void dispatch_source_set_event_handler(dispatch_source_t ds,
+                                       void (^work)(void));
+}
+
+#define GET_LSAN_BLOCK(work)                 \
+  void (^lsan_block)(void);                  \
+  int parent_tid = GetCurrentThread();       \
+  lsan_block = ^(void) {                     \
+    lsan_register_worker_thread(parent_tid); \
+    work();                                  \
+  }
+
+INTERCEPTOR(void, dispatch_async, dispatch_queue_t dq, void (^work)(void)) {
+  GET_LSAN_BLOCK(work);
+  REAL(dispatch_async)(dq, lsan_block);
+}
+
+INTERCEPTOR(void, dispatch_group_async, dispatch_group_t dg,
+            dispatch_queue_t dq, void (^work)(void)) {
+  GET_LSAN_BLOCK(work);
+  REAL(dispatch_group_async)(dg, dq, lsan_block);
+}
+
+INTERCEPTOR(void, dispatch_after, dispatch_time_t when, dispatch_queue_t queue,
+            void (^work)(void)) {
+  GET_LSAN_BLOCK(work);
+  REAL(dispatch_after)(when, queue, lsan_block);
+}
+
+INTERCEPTOR(void, dispatch_source_set_cancel_handler, dispatch_source_t ds,
+            void (^work)(void)) {
+  if (!work) {
+    REAL(dispatch_source_set_cancel_handler)(ds, work);
+    return;
+  }
+  GET_LSAN_BLOCK(work);
+  REAL(dispatch_source_set_cancel_handler)(ds, lsan_block);
+}
+
+INTERCEPTOR(void, dispatch_source_set_event_handler, dispatch_source_t ds,
+            void (^work)(void)) {
+  GET_LSAN_BLOCK(work);
+  REAL(dispatch_source_set_event_handler)(ds, lsan_block);
+}
+#endif
+
+#endif  // SANITIZER_MAC
diff --git a/lib/lsan/lsan_thread.cc b/lib/lsan/lsan_thread.cc
index 0ea7a6e97497..4404c8cc51d2 100644
--- a/lib/lsan/lsan_thread.cc
+++ b/lib/lsan/lsan_thread.cc
@@ -77,7 +77,7 @@ u32 ThreadCreate(u32 parent_tid, uptr user_id, bool detached) {
                                        /* arg */ nullptr);
 }
 
-void ThreadStart(u32 tid, tid_t os_id) {
+void ThreadStart(u32 tid, tid_t os_id, bool workerthread) {
   OnStartedArgs args;
   uptr stack_size = 0;
   uptr tls_size = 0;
@@ -87,7 +87,7 @@ void ThreadStart(u32 tid, tid_t os_id) {
   args.tls_end = args.tls_begin + tls_size;
   GetAllocatorCacheRange(&args.cache_begin, &args.cache_end);
   args.dtls = DTLS_Get();
-  thread_registry->StartThread(tid, os_id, /*workerthread*/ false, &args);
+  thread_registry->StartThread(tid, os_id, workerthread, &args);
 }
 
 void ThreadFinish() {
diff --git a/lib/lsan/lsan_thread.h b/lib/lsan/lsan_thread.h
index 73e080e26f76..b16d3d91537f 100644
--- a/lib/lsan/lsan_thread.h
+++ b/lib/lsan/lsan_thread.h
@@ -45,7 +45,7 @@ class ThreadContext : public ThreadContextBase {
 
 void InitializeThreadRegistry();
 
-void ThreadStart(u32 tid, tid_t os_id);
+void ThreadStart(u32 tid, tid_t os_id, bool workerthread = false);
 void ThreadFinish();
 u32 ThreadCreate(u32 tid, uptr uid, bool detached);
 void ThreadJoin(u32 tid);
diff --git a/lib/msan/msan_allocator.cc b/lib/msan/msan_allocator.cc
index 1be573faa412..d0f478afc9cc 100644
--- a/lib/msan/msan_allocator.cc
+++ b/lib/msan/msan_allocator.cc
@@ -119,9 +119,8 @@ static AllocatorCache fallback_allocator_cache;
 static SpinMutex fallback_mutex;
 
 void MsanAllocatorInit() {
-  allocator.Init(
-      common_flags()->allocator_may_return_null,
-      common_flags()->allocator_release_to_os_interval_ms);
+  SetAllocatorMayReturnNull(common_flags()->allocator_may_return_null);
+  allocator.Init(common_flags()->allocator_release_to_os_interval_ms);
 }
 
 AllocatorCache *GetAllocatorCache(MsanThreadLocalMallocStorage *ms) {
@@ -139,17 +138,17 @@ static void *MsanAllocate(StackTrace *stack, uptr size, uptr alignment,
   if (size > kMaxAllowedMallocSize) {
     Report("WARNING: MemorySanitizer failed to allocate %p bytes\n",
            (void *)size);
-    return allocator.ReturnNullOrDieOnBadRequest();
+    return Allocator::FailureHandler::OnBadRequest();
   }
   MsanThread *t = GetCurrentThread();
   void *allocated;
   if (t) {
     AllocatorCache *cache = GetAllocatorCache(&t->malloc_storage());
-    allocated = allocator.Allocate(cache, size, alignment, false);
+    allocated = allocator.Allocate(cache, size, alignment);
   } else {
     SpinMutexLock l(&fallback_mutex);
     AllocatorCache *cache = &fallback_allocator_cache;
-    allocated = allocator.Allocate(cache, size, alignment, false);
+    allocated = allocator.Allocate(cache, size, alignment);
   }
   Metadata *meta =
       reinterpret_cast<Metadata *>(allocator.GetMetaData(allocated));
@@ -197,7 +196,7 @@ void MsanDeallocate(StackTrace *stack, void *p) {
 
 void *MsanCalloc(StackTrace *stack, uptr nmemb, uptr size) {
   if (CallocShouldReturnNullDueToOverflow(size, nmemb))
-    return allocator.ReturnNullOrDieOnBadRequest();
+    return Allocator::FailureHandler::OnBadRequest();
   return MsanReallocate(stack, nullptr, nmemb * size, sizeof(u64), true);
 }
 
diff --git a/lib/sanitizer_common/sanitizer_allocator.cc b/lib/sanitizer_common/sanitizer_allocator.cc
index d47b5b41413c..db3ebb0336a9 100644
--- a/lib/sanitizer_common/sanitizer_allocator.cc
+++ b/lib/sanitizer_common/sanitizer_allocator.cc
@@ -94,8 +94,7 @@ InternalAllocator *internal_allocator() {
     SpinMutexLock l(&internal_alloc_init_mu);
     if (atomic_load(&internal_allocator_initialized, memory_order_relaxed) ==
         0) {
-      internal_allocator_instance->Init(
-          /* may_return_null */ false, kReleaseToOSIntervalNever);
+      internal_allocator_instance->Init(kReleaseToOSIntervalNever);
       atomic_store(&internal_allocator_initialized, 1, memory_order_release);
     }
   }
@@ -108,9 +107,9 @@ static void *RawInternalAlloc(uptr size, InternalAllocatorCache *cache,
   if (cache == 0) {
     SpinMutexLock l(&internal_allocator_cache_mu);
     return internal_allocator()->Allocate(&internal_allocator_cache, size,
-                                          alignment, false);
+                                          alignment);
   }
-  return internal_allocator()->Allocate(cache, size, alignment, false);
+  return internal_allocator()->Allocate(cache, size, alignment);
 }
 
 static void *RawInternalRealloc(void *ptr, uptr size,
@@ -162,7 +161,7 @@ void *InternalRealloc(void *addr, uptr size, InternalAllocatorCache *cache) {
 
 void *InternalCalloc(uptr count, uptr size, InternalAllocatorCache *cache) {
   if (CallocShouldReturnNullDueToOverflow(count, size))
-    return internal_allocator()->ReturnNullOrDieOnBadRequest();
+    return InternalAllocator::FailureHandler::OnBadRequest();
   void *p = InternalAlloc(count * size, cache);
   if (p) internal_memset(p, 0, count * size);
   return p;
@@ -209,12 +208,15 @@ bool CallocShouldReturnNullDueToOverflow(uptr size, uptr n) {
   return (max / size) < n;
 }
 
-static atomic_uint8_t reporting_out_of_memory = {0};
+static atomic_uint8_t allocator_out_of_memory = {0};
+static atomic_uint8_t allocator_may_return_null = {0};
 
-bool IsReportingOOM() { return atomic_load_relaxed(&reporting_out_of_memory); }
+bool IsAllocatorOutOfMemory() {
+  return atomic_load_relaxed(&allocator_out_of_memory);
+}
 
-void NORETURN ReportAllocatorCannotReturnNull(bool out_of_memory) {
-  if (out_of_memory) atomic_store_relaxed(&reporting_out_of_memory, 1);
+// Prints error message and kills the program.
+void NORETURN ReportAllocatorCannotReturnNull() {
   Report("%s's allocator is terminating the process instead of returning 0\n",
          SanitizerToolName);
   Report("If you don't like this behavior set allocator_may_return_null=1\n");
@@ -222,4 +224,35 @@ void NORETURN ReportAllocatorCannotReturnNull(bool out_of_memory) {
   Die();
 }
 
+bool AllocatorMayReturnNull() {
+  return atomic_load(&allocator_may_return_null, memory_order_relaxed);
+}
+
+void SetAllocatorMayReturnNull(bool may_return_null) {
+  atomic_store(&allocator_may_return_null, may_return_null,
+               memory_order_relaxed);
+}
+
+void *ReturnNullOrDieOnFailure::OnBadRequest() {
+  if (AllocatorMayReturnNull())
+    return nullptr;
+  ReportAllocatorCannotReturnNull();
+}
+
+void *ReturnNullOrDieOnFailure::OnOOM() {
+  atomic_store_relaxed(&allocator_out_of_memory, 1);
+  if (AllocatorMayReturnNull())
+    return nullptr;
+  ReportAllocatorCannotReturnNull();
+}
+
+void *DieOnFailure::OnBadRequest() {
+  ReportAllocatorCannotReturnNull();
+}
+
+void *DieOnFailure::OnOOM() {
+  atomic_store_relaxed(&allocator_out_of_memory, 1);
+  ReportAllocatorCannotReturnNull();
+}
+
 } // namespace __sanitizer
diff --git a/lib/sanitizer_common/sanitizer_allocator.h b/lib/sanitizer_common/sanitizer_allocator.h
index 9a37a2f2145f..f59c13d1c5a5 100644
--- a/lib/sanitizer_common/sanitizer_allocator.h
+++ b/lib/sanitizer_common/sanitizer_allocator.h
@@ -24,12 +24,28 @@
 
 namespace __sanitizer {
 
-// Returns true if ReportAllocatorCannotReturnNull(true) was called.
-// Can be use to avoid memory hungry operations.
-bool IsReportingOOM();
+// Since flags are immutable and allocator behavior can be changed at runtime
+// (unit tests or ASan on Android are some examples), allocator_may_return_null
+// flag value is cached here and can be altered later.
+bool AllocatorMayReturnNull();
+void SetAllocatorMayReturnNull(bool may_return_null);
 
-// Prints error message and kills the program.
-void NORETURN ReportAllocatorCannotReturnNull(bool out_of_memory);
+// Allocator failure handling policies:
+// Implements AllocatorMayReturnNull policy, returns null when the flag is set,
+// dies otherwise.
+struct ReturnNullOrDieOnFailure {
+  static void *OnBadRequest();
+  static void *OnOOM();
+};
+// Always dies on the failure.
+struct DieOnFailure {
+  static void *OnBadRequest();
+  static void *OnOOM();
+};
+
+// Returns true if allocator detected OOM condition. Can be used to avoid memory
+// hungry operations. Set when AllocatorReturnNullOrDieOnOOM() is called.
+bool IsAllocatorOutOfMemory();
 
 // Allocators call these callbacks on mmap/munmap.
 struct NoOpMapUnmapCallback {
diff --git a/lib/sanitizer_common/sanitizer_allocator_combined.h b/lib/sanitizer_common/sanitizer_allocator_combined.h
index 2c2390b3d2c6..efd25cadfe74 100644
--- a/lib/sanitizer_common/sanitizer_allocator_combined.h
+++ b/lib/sanitizer_common/sanitizer_allocator_combined.h
@@ -24,31 +24,26 @@ template <class PrimaryAllocator, class AllocatorCache,
           class SecondaryAllocator>  // NOLINT
 class CombinedAllocator {
  public:
-  void InitCommon(bool may_return_null, s32 release_to_os_interval_ms) {
+  typedef typename SecondaryAllocator::FailureHandler FailureHandler;
+
+  void InitLinkerInitialized(s32 release_to_os_interval_ms) {
     primary_.Init(release_to_os_interval_ms);
-    atomic_store(&may_return_null_, may_return_null, memory_order_relaxed);
-  }
-
-  void InitLinkerInitialized(
-      bool may_return_null, s32 release_to_os_interval_ms) {
-    secondary_.InitLinkerInitialized(may_return_null);
+    secondary_.InitLinkerInitialized();
     stats_.InitLinkerInitialized();
-    InitCommon(may_return_null, release_to_os_interval_ms);
   }
 
-  void Init(bool may_return_null, s32 release_to_os_interval_ms) {
-    secondary_.Init(may_return_null);
+  void Init(s32 release_to_os_interval_ms) {
+    primary_.Init(release_to_os_interval_ms);
+    secondary_.Init();
     stats_.Init();
-    InitCommon(may_return_null, release_to_os_interval_ms);
   }
 
-  void *Allocate(AllocatorCache *cache, uptr size, uptr alignment,
-                 bool cleared = false) {
+  void *Allocate(AllocatorCache *cache, uptr size, uptr alignment) {
     // Returning 0 on malloc(0) may break a lot of code.
     if (size == 0)
       size = 1;
     if (size + alignment < size)
-      return ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
     uptr original_size = size;
     // If alignment requirements are to be fulfilled by the frontend allocator
     // rather than by the primary or secondary, passing an alignment lower than
@@ -56,49 +51,24 @@ class CombinedAllocator {
     // alignment check.
     if (alignment > 8)
       size = RoundUpTo(size, alignment);
-    void *res;
-    bool from_primary = primary_.CanAllocate(size, alignment);
     // The primary allocator should return a 2^x aligned allocation when
     // requested 2^x bytes, hence using the rounded up 'size' when being
     // serviced by the primary (this is no longer true when the primary is
     // using a non-fixed base address). The secondary takes care of the
     // alignment without such requirement, and allocating 'size' would use
     // extraneous memory, so we employ 'original_size'.
-    if (from_primary)
+    void *res;
+    if (primary_.CanAllocate(size, alignment))
       res = cache->Allocate(&primary_, primary_.ClassID(size));
     else
       res = secondary_.Allocate(&stats_, original_size, alignment);
+    if (!res)
+      return FailureHandler::OnOOM();
     if (alignment > 8)
       CHECK_EQ(reinterpret_cast<uptr>(res) & (alignment - 1), 0);
-    // When serviced by the secondary, the chunk comes from a mmap allocation
-    // and will be zero'd out anyway. We only need to clear our the chunk if
-    // it was serviced by the primary, hence using the rounded up 'size'.
-    if (cleared && res && from_primary)
-      internal_bzero_aligned16(res, RoundUpTo(size, 16));
     return res;
   }
 
-  bool MayReturnNull() const {
-    return atomic_load(&may_return_null_, memory_order_acquire);
-  }
-
-  void *ReturnNullOrDieOnBadRequest() {
-    if (MayReturnNull())
-      return nullptr;
-    ReportAllocatorCannotReturnNull(false);
-  }
-
-  void *ReturnNullOrDieOnOOM() {
-    if (MayReturnNull())
-      return nullptr;
-    ReportAllocatorCannotReturnNull(true);
-  }
-
-  void SetMayReturnNull(bool may_return_null) {
-    secondary_.SetMayReturnNull(may_return_null);
-    atomic_store(&may_return_null_, may_return_null, memory_order_release);
-  }
-
   s32 ReleaseToOSIntervalMs() const {
     return primary_.ReleaseToOSIntervalMs();
   }
@@ -219,6 +189,5 @@ class CombinedAllocator {
   PrimaryAllocator primary_;
   SecondaryAllocator secondary_;
   AllocatorGlobalStats stats_;
-  atomic_uint8_t may_return_null_;
 };
 
diff --git a/lib/sanitizer_common/sanitizer_allocator_internal.h b/lib/sanitizer_common/sanitizer_allocator_internal.h
index d1890f20f810..a791d0d94894 100644
--- a/lib/sanitizer_common/sanitizer_allocator_internal.h
+++ b/lib/sanitizer_common/sanitizer_allocator_internal.h
@@ -47,7 +47,8 @@ typedef SizeClassAllocatorLocalCache<PrimaryInternalAllocator>
     InternalAllocatorCache;
 
 typedef CombinedAllocator<PrimaryInternalAllocator, InternalAllocatorCache,
-                          LargeMmapAllocator<> > InternalAllocator;
+                          LargeMmapAllocator<NoOpMapUnmapCallback, DieOnFailure>
+                         > InternalAllocator;
 
 void *InternalAlloc(uptr size, InternalAllocatorCache *cache = nullptr,
                     uptr alignment = 0);
diff --git a/lib/sanitizer_common/sanitizer_allocator_local_cache.h b/lib/sanitizer_common/sanitizer_allocator_local_cache.h
index b3729bf55dbb..8fa62a3bf829 100644
--- a/lib/sanitizer_common/sanitizer_allocator_local_cache.h
+++ b/lib/sanitizer_common/sanitizer_allocator_local_cache.h
@@ -144,8 +144,10 @@ struct SizeClassAllocator32LocalCache {
     CHECK_NE(class_id, 0UL);
     CHECK_LT(class_id, kNumClasses);
     PerClass *c = &per_class_[class_id];
-    if (UNLIKELY(c->count == 0))
-      Refill(allocator, class_id);
+    if (UNLIKELY(c->count == 0)) {
+      if (UNLIKELY(!Refill(allocator, class_id)))
+        return nullptr;
+    }
     stats_.Add(AllocatorStatAllocated, c->class_size);
     void *res = c->batch[--c->count];
     PREFETCH(c->batch[c->count - 1]);
@@ -227,14 +229,17 @@ struct SizeClassAllocator32LocalCache {
       Deallocate(allocator, batch_class_id, b);
   }
 
-  NOINLINE void Refill(SizeClassAllocator *allocator, uptr class_id) {
+  NOINLINE bool Refill(SizeClassAllocator *allocator, uptr class_id) {
     InitCache();
     PerClass *c = &per_class_[class_id];
     TransferBatch *b = allocator->AllocateBatch(&stats_, this, class_id);
+    if (UNLIKELY(!b))
+      return false;
     CHECK_GT(b->Count(), 0);
     b->CopyToArray(c->batch);
     c->count = b->Count();
     DestroyBatch(class_id, allocator, b);
+    return true;
   }
 
   NOINLINE void Drain(SizeClassAllocator *allocator, uptr class_id) {
@@ -244,6 +249,10 @@ struct SizeClassAllocator32LocalCache {
     uptr first_idx_to_drain = c->count - cnt;
     TransferBatch *b = CreateBatch(
         class_id, allocator, (TransferBatch *)c->batch[first_idx_to_drain]);
+    // Failure to allocate a batch while releasing memory is non recoverable.
+    // TODO(alekseys): Figure out how to do it without allocating a new batch.
+    if (UNLIKELY(!b))
+      DieOnFailure::OnOOM();
     b->SetFromArray(allocator->GetRegionBeginBySizeClass(class_id),
                     &c->batch[first_idx_to_drain], cnt);
     c->count -= cnt;
diff --git a/lib/sanitizer_common/sanitizer_allocator_primary32.h b/lib/sanitizer_common/sanitizer_allocator_primary32.h
index e13510ba33b9..d3949cc05734 100644
--- a/lib/sanitizer_common/sanitizer_allocator_primary32.h
+++ b/lib/sanitizer_common/sanitizer_allocator_primary32.h
@@ -24,7 +24,8 @@ template<class SizeClassAllocator> struct SizeClassAllocator32LocalCache;
 // be returned by MmapOrDie().
 //
 // Region:
-//   a result of a single call to MmapAlignedOrDie(kRegionSize, kRegionSize).
+//   a result of a single call to MmapAlignedOrDieOnFatalError(kRegionSize,
+//                                                             kRegionSize).
 // Since the regions are aligned by kRegionSize, there are exactly
 // kNumPossibleRegions possible regions in the address space and so we keep
 // a ByteMap possible_regions to store the size classes of each Region.
@@ -149,8 +150,9 @@ class SizeClassAllocator32 {
     CHECK_LT(class_id, kNumClasses);
     SizeClassInfo *sci = GetSizeClassInfo(class_id);
     SpinMutexLock l(&sci->mutex);
-    if (sci->free_list.empty())
-      PopulateFreeList(stat, c, sci, class_id);
+    if (sci->free_list.empty() &&
+        UNLIKELY(!PopulateFreeList(stat, c, sci, class_id)))
+      return nullptr;
     CHECK(!sci->free_list.empty());
     TransferBatch *b = sci->free_list.front();
     sci->free_list.pop_front();
@@ -277,8 +279,10 @@ class SizeClassAllocator32 {
 
   uptr AllocateRegion(AllocatorStats *stat, uptr class_id) {
     CHECK_LT(class_id, kNumClasses);
-    uptr res = reinterpret_cast<uptr>(MmapAlignedOrDie(kRegionSize, kRegionSize,
-                                      "SizeClassAllocator32"));
+    uptr res = reinterpret_cast<uptr>(MmapAlignedOrDieOnFatalError(
+        kRegionSize, kRegionSize, "SizeClassAllocator32"));
+    if (UNLIKELY(!res))
+      return 0;
     MapUnmapCallback().OnMap(res, kRegionSize);
     stat->Add(AllocatorStatMapped, kRegionSize);
     CHECK_EQ(0U, (res & (kRegionSize - 1)));
@@ -291,16 +295,20 @@ class SizeClassAllocator32 {
     return &size_class_info_array[class_id];
   }
 
-  void PopulateFreeList(AllocatorStats *stat, AllocatorCache *c,
+  bool PopulateFreeList(AllocatorStats *stat, AllocatorCache *c,
                         SizeClassInfo *sci, uptr class_id) {
     uptr size = ClassIdToSize(class_id);
     uptr reg = AllocateRegion(stat, class_id);
+    if (UNLIKELY(!reg))
+      return false;
     uptr n_chunks = kRegionSize / (size + kMetadataSize);
     uptr max_count = TransferBatch::MaxCached(class_id);
     TransferBatch *b = nullptr;
     for (uptr i = reg; i < reg + n_chunks * size; i += size) {
       if (!b) {
         b = c->CreateBatch(class_id, this, (TransferBatch*)i);
+        if (!b)
+          return false;
         b->Clear();
       }
       b->Add((void*)i);
@@ -314,6 +322,7 @@ class SizeClassAllocator32 {
       CHECK_GT(b->Count(), 0);
       sci->free_list.push_back(b);
     }
+    return true;
   }
 
   ByteMap possible_regions;
diff --git a/lib/sanitizer_common/sanitizer_allocator_secondary.h b/lib/sanitizer_common/sanitizer_allocator_secondary.h
index 2c69f47ec4e6..261dfb5e1a28 100644
--- a/lib/sanitizer_common/sanitizer_allocator_secondary.h
+++ b/lib/sanitizer_common/sanitizer_allocator_secondary.h
@@ -17,17 +17,19 @@
 // This class can (de)allocate only large chunks of memory using mmap/unmap.
 // The main purpose of this allocator is to cover large and rare allocation
 // sizes not covered by more efficient allocators (e.g. SizeClassAllocator64).
-template <class MapUnmapCallback = NoOpMapUnmapCallback>
+template <class MapUnmapCallback = NoOpMapUnmapCallback,
+          class FailureHandlerT = ReturnNullOrDieOnFailure>
 class LargeMmapAllocator {
  public:
-  void InitLinkerInitialized(bool may_return_null) {
+  typedef FailureHandlerT FailureHandler;
+
+  void InitLinkerInitialized() {
     page_size_ = GetPageSizeCached();
-    atomic_store(&may_return_null_, may_return_null, memory_order_relaxed);
   }
 
-  void Init(bool may_return_null) {
+  void Init() {
     internal_memset(this, 0, sizeof(*this));
-    InitLinkerInitialized(may_return_null);
+    InitLinkerInitialized();
   }
 
   void *Allocate(AllocatorStats *stat, uptr size, uptr alignment) {
@@ -37,11 +39,11 @@ class LargeMmapAllocator {
       map_size += alignment;
     // Overflow.
     if (map_size < size)
-      return ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
     uptr map_beg = reinterpret_cast<uptr>(
         MmapOrDieOnFatalError(map_size, "LargeMmapAllocator"));
     if (!map_beg)
-      return ReturnNullOrDieOnOOM();
+      return FailureHandler::OnOOM();
     CHECK(IsAligned(map_beg, page_size_));
     MapUnmapCallback().OnMap(map_beg, map_size);
     uptr map_end = map_beg + map_size;
@@ -75,24 +77,6 @@ class LargeMmapAllocator {
     return reinterpret_cast<void*>(res);
   }
 
-  bool MayReturnNull() const {
-    return atomic_load(&may_return_null_, memory_order_acquire);
-  }
-
-  void *ReturnNullOrDieOnBadRequest() {
-    if (MayReturnNull()) return nullptr;
-    ReportAllocatorCannotReturnNull(false);
-  }
-
-  void *ReturnNullOrDieOnOOM() {
-    if (MayReturnNull()) return nullptr;
-    ReportAllocatorCannotReturnNull(true);
-  }
-
-  void SetMayReturnNull(bool may_return_null) {
-    atomic_store(&may_return_null_, may_return_null, memory_order_release);
-  }
-
   void Deallocate(AllocatorStats *stat, void *p) {
     Header *h = GetHeader(p);
     {
@@ -278,7 +262,6 @@ class LargeMmapAllocator {
   struct Stats {
     uptr n_allocs, n_frees, currently_allocated, max_allocated, by_size_log[64];
   } stats;
-  atomic_uint8_t may_return_null_;
   SpinMutex mutex_;
 };
 
diff --git a/lib/sanitizer_common/sanitizer_atomic_clang.h b/lib/sanitizer_common/sanitizer_atomic_clang.h
index 38363e875560..65b3a38f0d51 100644
--- a/lib/sanitizer_common/sanitizer_atomic_clang.h
+++ b/lib/sanitizer_common/sanitizer_atomic_clang.h
@@ -71,16 +71,25 @@ INLINE typename T::Type atomic_exchange(volatile T *a,
   return v;
 }
 
-template<typename T>
-INLINE bool atomic_compare_exchange_strong(volatile T *a,
-                                           typename T::Type *cmp,
+template <typename T>
+INLINE bool atomic_compare_exchange_strong(volatile T *a, typename T::Type *cmp,
                                            typename T::Type xchg,
                                            memory_order mo) {
   typedef typename T::Type Type;
   Type cmpv = *cmp;
-  Type prev = __sync_val_compare_and_swap(&a->val_dont_use, cmpv, xchg);
-  if (prev == cmpv)
-    return true;
+  Type prev;
+#if defined(_MIPS_SIM) && _MIPS_SIM == _ABIO32
+  if (sizeof(*a) == 8) {
+    Type volatile *val_ptr = const_cast<Type volatile *>(&a->val_dont_use);
+    prev = __mips_sync_val_compare_and_swap<u64>(
+        reinterpret_cast<u64 volatile *>(val_ptr), (u64)cmpv, (u64)xchg);
+  } else {
+    prev = __sync_val_compare_and_swap(&a->val_dont_use, cmpv, xchg);
+  }
+#else
+  prev = __sync_val_compare_and_swap(&a->val_dont_use, cmpv, xchg);
+#endif
+  if (prev == cmpv) return true;
   *cmp = prev;
   return false;
 }
diff --git a/lib/sanitizer_common/sanitizer_atomic_clang_other.h b/lib/sanitizer_common/sanitizer_atomic_clang_other.h
index 099b9f7ec8c6..d2acc311bf7d 100644
--- a/lib/sanitizer_common/sanitizer_atomic_clang_other.h
+++ b/lib/sanitizer_common/sanitizer_atomic_clang_other.h
@@ -17,6 +17,56 @@
 
 namespace __sanitizer {
 
+// MIPS32 does not support atomic > 4 bytes. To address this lack of
+// functionality, the sanitizer library provides helper methods which use an
+// internal spin lock mechanism to emulate atomic oprations when the size is
+// 8 bytes.
+#if defined(_MIPS_SIM) && _MIPS_SIM == _ABIO32
+static void __spin_lock(volatile int *lock) {
+  while (__sync_lock_test_and_set(lock, 1))
+    while (*lock) {
+    }
+}
+
+static void __spin_unlock(volatile int *lock) { __sync_lock_release(lock); }
+
+
+// Make sure the lock is on its own cache line to prevent false sharing.
+// Put it inside a struct that is aligned and padded to the typical MIPS
+// cacheline which is 32 bytes.
+static struct {
+  int lock;
+  char pad[32 - sizeof(int)];
+} __attribute__((aligned(32))) lock = {0};
+
+template <class T>
+T __mips_sync_fetch_and_add(volatile T *ptr, T val) {
+  T ret;
+
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  *ptr = ret + val;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+
+template <class T>
+T __mips_sync_val_compare_and_swap(volatile T *ptr, T oldval, T newval) {
+  T ret;
+  __spin_lock(&lock.lock);
+
+  ret = *ptr;
+  if (ret == oldval) *ptr = newval;
+
+  __spin_unlock(&lock.lock);
+
+  return ret;
+}
+#endif
+
 INLINE void proc_yield(int cnt) {
   __asm__ __volatile__("" ::: "memory");
 }
@@ -53,8 +103,15 @@ INLINE typename T::Type atomic_load(
     // 64-bit load on 32-bit platform.
     // Gross, but simple and reliable.
     // Assume that it is not in read-only memory.
+#if defined(_MIPS_SIM) && _MIPS_SIM == _ABIO32
+    typename T::Type volatile *val_ptr =
+        const_cast<typename T::Type volatile *>(&a->val_dont_use);
+    v = __mips_sync_fetch_and_add<u64>(
+        reinterpret_cast<u64 volatile *>(val_ptr), 0);
+#else
     v = __sync_fetch_and_add(
         const_cast<typename T::Type volatile *>(&a->val_dont_use), 0);
+#endif
   }
   return v;
 }
@@ -84,7 +141,14 @@ INLINE void atomic_store(volatile T *a, typename T::Type v, memory_order mo) {
     typename T::Type cmp = a->val_dont_use;
     typename T::Type cur;
     for (;;) {
+#if defined(_MIPS_SIM) && _MIPS_SIM == _ABIO32
+      typename T::Type volatile *val_ptr =
+          const_cast<typename T::Type volatile *>(&a->val_dont_use);
+      cur = __mips_sync_val_compare_and_swap<u64>(
+          reinterpret_cast<u64 volatile *>(val_ptr), (u64)cmp, (u64)v);
+#else
       cur = __sync_val_compare_and_swap(&a->val_dont_use, cmp, v);
+#endif
       if (cmp == v)
         break;
       cmp = cur;
diff --git a/lib/sanitizer_common/sanitizer_common.h b/lib/sanitizer_common/sanitizer_common.h
index 875a46009f49..560c53b6400e 100644
--- a/lib/sanitizer_common/sanitizer_common.h
+++ b/lib/sanitizer_common/sanitizer_common.h
@@ -95,7 +95,9 @@ void *MmapFixedOrDie(uptr fixed_addr, uptr size);
 void *MmapFixedNoAccess(uptr fixed_addr, uptr size, const char *name = nullptr);
 void *MmapNoAccess(uptr size);
 // Map aligned chunk of address space; size and alignment are powers of two.
-void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type);
+// Dies on all but out of memory errors, in the latter case returns nullptr.
+void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
+                                   const char *mem_type);
 // Disallow access to a memory range.  Use MmapFixedNoAccess to allocate an
 // unaccessible memory.
 bool MprotectNoAccess(uptr addr, uptr size);
@@ -808,8 +810,11 @@ INLINE void LogMessageOnPrintf(const char *str) {}
 #if SANITIZER_LINUX
 // Initialize Android logging. Any writes before this are silently lost.
 void AndroidLogInit();
+void SetAbortMessage(const char *);
 #else
 INLINE void AndroidLogInit() {}
+// FIXME: MacOS implementation could use CRSetCrashLogMessage.
+INLINE void SetAbortMessage(const char *) {}
 #endif
 
 #if SANITIZER_ANDROID
@@ -919,6 +924,10 @@ const s32 kReleaseToOSIntervalNever = -1;
 
 void CheckNoDeepBind(const char *filename, int flag);
 
+// Returns the requested amount of random data (up to 256 bytes) that can then
+// be used to seed a PRNG.
+bool GetRandom(void *buffer, uptr length);
+
 }  // namespace __sanitizer
 
 inline void *operator new(__sanitizer::operator_new_size_type size,
diff --git a/lib/sanitizer_common/sanitizer_linux.cc b/lib/sanitizer_common/sanitizer_linux.cc
index d31c49d694ae..a94a63c7f16d 100644
--- a/lib/sanitizer_common/sanitizer_linux.cc
+++ b/lib/sanitizer_common/sanitizer_linux.cc
@@ -1604,6 +1604,32 @@ uptr FindAvailableMemoryRange(uptr size, uptr alignment, uptr left_padding) {
   return 0;
 }
 
+bool GetRandom(void *buffer, uptr length) {
+  if (!buffer || !length || length > 256)
+    return false;
+#if defined(__NR_getrandom)
+  static atomic_uint8_t skip_getrandom_syscall;
+  if (!atomic_load_relaxed(&skip_getrandom_syscall)) {
+    // Up to 256 bytes, getrandom will not be interrupted.
+    uptr res = internal_syscall(SYSCALL(getrandom), buffer, length, 0);
+    int rverrno = 0;
+    if (internal_iserror(res, &rverrno) && rverrno == ENOSYS)
+      atomic_store_relaxed(&skip_getrandom_syscall, 1);
+    else if (res == length)
+      return true;
+  }
+#endif
+  uptr fd = internal_open("/dev/urandom", O_RDONLY);
+  if (internal_iserror(fd))
+    return false;
+  // internal_read deals with EINTR.
+  uptr res = internal_read(fd, buffer, length);
+  if (internal_iserror(res))
+    return false;
+  internal_close(fd);
+  return true;
+}
+
 } // namespace __sanitizer
 
 #endif // SANITIZER_FREEBSD || SANITIZER_LINUX
diff --git a/lib/sanitizer_common/sanitizer_linux_libcdep.cc b/lib/sanitizer_common/sanitizer_linux_libcdep.cc
index 894013ddd880..b9a48a1e496b 100644
--- a/lib/sanitizer_common/sanitizer_linux_libcdep.cc
+++ b/lib/sanitizer_common/sanitizer_linux_libcdep.cc
@@ -551,6 +551,13 @@ void LogMessageOnPrintf(const char *str) {
     WriteToSyslog(str);
 }
 
+#if SANITIZER_ANDROID && __ANDROID_API__ >= 21
+extern "C" void android_set_abort_message(const char *msg);
+void SetAbortMessage(const char *str) { android_set_abort_message(str); }
+#else
+void SetAbortMessage(const char *str) {}
+#endif
+
 #endif // SANITIZER_LINUX
 
 } // namespace __sanitizer
diff --git a/lib/sanitizer_common/sanitizer_mac.cc b/lib/sanitizer_common/sanitizer_mac.cc
index a788a091592f..b48238106dd9 100644
--- a/lib/sanitizer_common/sanitizer_mac.cc
+++ b/lib/sanitizer_common/sanitizer_mac.cc
@@ -923,6 +923,11 @@ void CheckNoDeepBind(const char *filename, int flag) {
   // Do nothing.
 }
 
+// FIXME: implement on this platform.
+bool GetRandom(void *buffer, uptr length) {
+  UNIMPLEMENTED();
+}
+
 }  // namespace __sanitizer
 
 #endif  // SANITIZER_MAC
diff --git a/lib/sanitizer_common/sanitizer_posix.cc b/lib/sanitizer_common/sanitizer_posix.cc
index 4184a84c73f8..87c5b9add5cf 100644
--- a/lib/sanitizer_common/sanitizer_posix.cc
+++ b/lib/sanitizer_common/sanitizer_posix.cc
@@ -164,11 +164,14 @@ void *MmapOrDieOnFatalError(uptr size, const char *mem_type) {
 // We want to map a chunk of address space aligned to 'alignment'.
 // We do it by maping a bit more and then unmaping redundant pieces.
 // We probably can do it with fewer syscalls in some OS-dependent way.
-void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type) {
+void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
+                                   const char *mem_type) {
   CHECK(IsPowerOfTwo(size));
   CHECK(IsPowerOfTwo(alignment));
   uptr map_size = size + alignment;
-  uptr map_res = (uptr)MmapOrDie(map_size, mem_type);
+  uptr map_res = (uptr)MmapOrDieOnFatalError(map_size, mem_type);
+  if (!map_res)
+    return nullptr;
   uptr map_end = map_res + map_size;
   uptr res = map_res;
   if (res & (alignment - 1))  // Not aligned.
diff --git a/lib/sanitizer_common/sanitizer_posix_libcdep.cc b/lib/sanitizer_common/sanitizer_posix_libcdep.cc
index 5b1d53698707..e113fb1093d4 100644
--- a/lib/sanitizer_common/sanitizer_posix_libcdep.cc
+++ b/lib/sanitizer_common/sanitizer_posix_libcdep.cc
@@ -189,25 +189,7 @@ void UnsetAlternateSignalStack() {
 
 static void MaybeInstallSigaction(int signum,
                                   SignalHandlerType handler) {
-  switch (GetHandleSignalMode(signum)) {
-    case kHandleSignalNo:
-      return;
-    case kHandleSignalYes: {
-      struct sigaction sigact;
-      internal_memset(&sigact, 0, sizeof(sigact));
-      CHECK_EQ(0, internal_sigaction(signum, nullptr, &sigact));
-      if (sigact.sa_flags & SA_SIGINFO) {
-        if (sigact.sa_sigaction) return;
-      } else {
-        if (sigact.sa_handler != SIG_DFL && sigact.sa_handler != SIG_IGN &&
-            sigact.sa_handler != SIG_ERR)
-          return;
-      }
-      break;
-    }
-    case kHandleSignalExclusive:
-      break;
-  }
+  if (GetHandleSignalMode(signum) == kHandleSignalNo) return;
 
   struct sigaction sigact;
   internal_memset(&sigact, 0, sizeof(sigact));
diff --git a/lib/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cc b/lib/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cc
index d3c77b510d35..60ec7506c312 100644
--- a/lib/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cc
+++ b/lib/sanitizer_common/sanitizer_symbolizer_posix_libcdep.cc
@@ -495,7 +495,7 @@ static void ChooseSymbolizerTools(IntrusiveList<SymbolizerTool> *list,
     VReport(2, "Symbolizer is disabled.\n");
     return;
   }
-  if (IsReportingOOM()) {
+  if (IsAllocatorOutOfMemory()) {
     VReport(2, "Cannot use internal symbolizer: out of memory\n");
   } else if (SymbolizerTool *tool = InternalSymbolizer::get(allocator)) {
     VReport(2, "Using internal symbolizer.\n");
diff --git a/lib/sanitizer_common/sanitizer_win.cc b/lib/sanitizer_common/sanitizer_win.cc
index 506e7374a329..c6a146553412 100644
--- a/lib/sanitizer_common/sanitizer_win.cc
+++ b/lib/sanitizer_common/sanitizer_win.cc
@@ -131,18 +131,24 @@ void UnmapOrDie(void *addr, uptr size) {
   }
 }
 
+static void *ReturnNullptrOnOOMOrDie(uptr size, const char *mem_type,
+                                     const char *mmap_type) {
+  error_t last_error = GetLastError();
+  if (last_error == ERROR_NOT_ENOUGH_MEMORY)
+    return nullptr;
+  ReportMmapFailureAndDie(size, mem_type, mmap_type, last_error);
+}
+
 void *MmapOrDieOnFatalError(uptr size, const char *mem_type) {
   void *rv = VirtualAlloc(0, size, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
-  if (rv == 0) {
-    error_t last_error = GetLastError();
-    if (last_error != ERROR_NOT_ENOUGH_MEMORY)
-      ReportMmapFailureAndDie(size, mem_type, "allocate", last_error);
-  }
+  if (rv == 0)
+    return ReturnNullptrOnOOMOrDie(size, mem_type, "allocate");
   return rv;
 }
 
 // We want to map a chunk of address space aligned to 'alignment'.
-void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type) {
+void *MmapAlignedOrDieOnFatalError(uptr size, uptr alignment,
+                                   const char *mem_type) {
   CHECK(IsPowerOfTwo(size));
   CHECK(IsPowerOfTwo(alignment));
 
@@ -152,7 +158,7 @@ void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type) {
   uptr mapped_addr =
       (uptr)VirtualAlloc(0, size, MEM_RESERVE | MEM_COMMIT, PAGE_READWRITE);
   if (!mapped_addr)
-    ReportMmapFailureAndDie(size, mem_type, "allocate aligned", GetLastError());
+    return ReturnNullptrOnOOMOrDie(size, mem_type, "allocate aligned");
 
   // If we got it right on the first try, return. Otherwise, unmap it and go to
   // the slow path.
@@ -172,8 +178,7 @@ void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type) {
     mapped_addr =
         (uptr)VirtualAlloc(0, size + alignment, MEM_RESERVE, PAGE_NOACCESS);
     if (!mapped_addr)
-      ReportMmapFailureAndDie(size, mem_type, "allocate aligned",
-                              GetLastError());
+      return ReturnNullptrOnOOMOrDie(size, mem_type, "allocate aligned");
 
     // Find the aligned address.
     uptr aligned_addr = RoundUpTo(mapped_addr, alignment);
@@ -191,7 +196,7 @@ void *MmapAlignedOrDie(uptr size, uptr alignment, const char *mem_type) {
 
   // Fail if we can't make this work quickly.
   if (retries == kMaxRetries && mapped_addr == 0)
-    ReportMmapFailureAndDie(size, mem_type, "allocate aligned", GetLastError());
+    return ReturnNullptrOnOOMOrDie(size, mem_type, "allocate aligned");
 
   return (void *)mapped_addr;
 }
@@ -1002,6 +1007,11 @@ void CheckNoDeepBind(const char *filename, int flag) {
   // Do nothing.
 }
 
+// FIXME: implement on this platform.
+bool GetRandom(void *buffer, uptr length) {
+  UNIMPLEMENTED();
+}
+
 }  // namespace __sanitizer
 
 #endif  // _WIN32
diff --git a/lib/sanitizer_common/tests/sanitizer_allocator_test.cc b/lib/sanitizer_common/tests/sanitizer_allocator_test.cc
index b28159a2adaf..f256d8776d80 100644
--- a/lib/sanitizer_common/tests/sanitizer_allocator_test.cc
+++ b/lib/sanitizer_common/tests/sanitizer_allocator_test.cc
@@ -426,8 +426,8 @@ TEST(SanitizerCommon, SizeClassAllocator32MapUnmapCallback) {
 TEST(SanitizerCommon, LargeMmapAllocatorMapUnmapCallback) {
   TestMapUnmapCallback::map_count = 0;
   TestMapUnmapCallback::unmap_count = 0;
-  LargeMmapAllocator<TestMapUnmapCallback> a;
-  a.Init(/* may_return_null */ false);
+  LargeMmapAllocator<TestMapUnmapCallback, DieOnFailure> a;
+  a.Init();
   AllocatorStats stats;
   stats.Init();
   void *x = a.Allocate(&stats, 1 << 20, 1);
@@ -463,8 +463,8 @@ TEST(SanitizerCommon, SizeClassAllocator64Overflow) {
 #endif
 
 TEST(SanitizerCommon, LargeMmapAllocator) {
-  LargeMmapAllocator<> a;
-  a.Init(/* may_return_null */ false);
+  LargeMmapAllocator<NoOpMapUnmapCallback, DieOnFailure> a;
+  a.Init();
   AllocatorStats stats;
   stats.Init();
 
@@ -546,8 +546,9 @@ void TestCombinedAllocator() {
   typedef
       CombinedAllocator<PrimaryAllocator, AllocatorCache, SecondaryAllocator>
       Allocator;
+  SetAllocatorMayReturnNull(true);
   Allocator *a = new Allocator;
-  a->Init(/* may_return_null */ true, kReleaseToOSIntervalNever);
+  a->Init(kReleaseToOSIntervalNever);
   std::mt19937 r;
 
   AllocatorCache cache;
@@ -561,7 +562,7 @@ void TestCombinedAllocator() {
   EXPECT_EQ(a->Allocate(&cache, (uptr)-1 - 1023, 1024), (void*)0);
 
   // Set to false
-  a->SetMayReturnNull(false);
+  SetAllocatorMayReturnNull(false);
   EXPECT_DEATH(a->Allocate(&cache, -1, 1),
                "allocator is terminating the process");
 
@@ -873,8 +874,8 @@ TEST(SanitizerCommon, SizeClassAllocator32Iteration) {
 }
 
 TEST(SanitizerCommon, LargeMmapAllocatorIteration) {
-  LargeMmapAllocator<> a;
-  a.Init(/* may_return_null */ false);
+  LargeMmapAllocator<NoOpMapUnmapCallback, DieOnFailure> a;
+  a.Init();
   AllocatorStats stats;
   stats.Init();
 
@@ -900,8 +901,8 @@ TEST(SanitizerCommon, LargeMmapAllocatorIteration) {
 }
 
 TEST(SanitizerCommon, LargeMmapAllocatorBlockBegin) {
-  LargeMmapAllocator<> a;
-  a.Init(/* may_return_null */ false);
+  LargeMmapAllocator<NoOpMapUnmapCallback, DieOnFailure> a;
+  a.Init();
   AllocatorStats stats;
   stats.Init();
 
diff --git a/lib/sanitizer_common/tests/sanitizer_common_test.cc b/lib/sanitizer_common/tests/sanitizer_common_test.cc
index ebc885db7525..93a8794eeb8f 100644
--- a/lib/sanitizer_common/tests/sanitizer_common_test.cc
+++ b/lib/sanitizer_common/tests/sanitizer_common_test.cc
@@ -72,12 +72,12 @@ TEST(SanitizerCommon, SortTest) {
   EXPECT_TRUE(IsSorted(array, 2));
 }
 
-TEST(SanitizerCommon, MmapAlignedOrDie) {
+TEST(SanitizerCommon, MmapAlignedOrDieOnFatalError) {
   uptr PageSize = GetPageSizeCached();
   for (uptr size = 1; size <= 32; size *= 2) {
     for (uptr alignment = 1; alignment <= 32; alignment *= 2) {
       for (int iter = 0; iter < 100; iter++) {
-        uptr res = (uptr)MmapAlignedOrDie(
+        uptr res = (uptr)MmapAlignedOrDieOnFatalError(
             size * PageSize, alignment * PageSize, "MmapAlignedOrDieTest");
         EXPECT_EQ(0U, res % (alignment * PageSize));
         internal_memset((void*)res, 1, size * PageSize);
@@ -300,4 +300,21 @@ TEST(SanitizerCommon, InternalScopedString) {
   EXPECT_STREQ("012345678", str.data());
 }
 
+#if SANITIZER_LINUX
+TEST(SanitizerCommon, GetRandom) {
+  u8 buffer_1[32], buffer_2[32];
+  EXPECT_FALSE(GetRandom(nullptr, 32));
+  EXPECT_FALSE(GetRandom(buffer_1, 0));
+  EXPECT_FALSE(GetRandom(buffer_1, 512));
+  EXPECT_EQ(ARRAY_SIZE(buffer_1), ARRAY_SIZE(buffer_2));
+  for (uptr size = 4; size <= ARRAY_SIZE(buffer_1); size += 4) {
+    for (uptr i = 0; i < 100; i++) {
+      EXPECT_TRUE(GetRandom(buffer_1, size));
+      EXPECT_TRUE(GetRandom(buffer_2, size));
+      EXPECT_NE(internal_memcmp(buffer_1, buffer_2, size), 0);
+    }
+  }
+}
+#endif
+
 }  // namespace __sanitizer
diff --git a/lib/scudo/scudo_allocator.cpp b/lib/scudo/scudo_allocator.cpp
index ce69ddf55531..1d0db84a5aaf 100644
--- a/lib/scudo/scudo_allocator.cpp
+++ b/lib/scudo/scudo_allocator.cpp
@@ -273,6 +273,8 @@ struct ScudoAllocator {
   static const uptr MaxAllowedMallocSize =
       FIRST_32_SECOND_64(2UL << 30, 1ULL << 40);
 
+  typedef ReturnNullOrDieOnFailure FailureHandler;
+
   ScudoBackendAllocator BackendAllocator;
   ScudoQuarantine AllocatorQuarantine;
 
@@ -326,7 +328,8 @@ struct ScudoAllocator {
     DeallocationTypeMismatch = Options.DeallocationTypeMismatch;
     DeleteSizeMismatch = Options.DeleteSizeMismatch;
     ZeroContents = Options.ZeroContents;
-    BackendAllocator.Init(Options.MayReturnNull, Options.ReleaseToOSIntervalMs);
+    SetAllocatorMayReturnNull(Options.MayReturnNull);
+    BackendAllocator.Init(Options.ReleaseToOSIntervalMs);
     AllocatorQuarantine.Init(
         static_cast<uptr>(Options.QuarantineSizeMb) << 20,
         static_cast<uptr>(Options.ThreadLocalQuarantineSizeKb) << 10);
@@ -354,11 +357,11 @@ struct ScudoAllocator {
       dieWithMessage("ERROR: alignment is not a power of 2\n");
     }
     if (Alignment > MaxAlignment)
-      return BackendAllocator.ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
     if (Alignment < MinAlignment)
       Alignment = MinAlignment;
     if (Size >= MaxAllowedMallocSize)
-      return BackendAllocator.ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
     if (Size == 0)
       Size = 1;
 
@@ -366,7 +369,7 @@ struct ScudoAllocator {
     uptr AlignedSize = (Alignment > MinAlignment) ?
         NeededSize + (Alignment - AlignedChunkHeaderSize) : NeededSize;
     if (AlignedSize >= MaxAllowedMallocSize)
-      return BackendAllocator.ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
 
     // Primary and Secondary backed allocations have a different treatment. We
     // deal with alignment requirements of Primary serviced allocations here,
@@ -391,7 +394,7 @@ struct ScudoAllocator {
                                       AllocationAlignment, FromPrimary);
     }
     if (!Ptr)
-      return BackendAllocator.ReturnNullOrDieOnOOM();
+      return FailureHandler::OnOOM();
 
     // If requested, we will zero out the entire contents of the returned chunk.
     if ((ForceZeroContents || ZeroContents) && FromPrimary)
@@ -583,7 +586,7 @@ struct ScudoAllocator {
     initThreadMaybe();
     uptr Total = NMemB * Size;
     if (Size != 0 && Total / Size != NMemB)  // Overflow check
-      return BackendAllocator.ReturnNullOrDieOnBadRequest();
+      return FailureHandler::OnBadRequest();
     return allocate(Total, MinAlignment, FromMalloc, true);
   }
 
diff --git a/lib/scudo/scudo_allocator_combined.h b/lib/scudo/scudo_allocator_combined.h
index c978db55a9d9..21c45897b94e 100644
--- a/lib/scudo/scudo_allocator_combined.h
+++ b/lib/scudo/scudo_allocator_combined.h
@@ -23,11 +23,10 @@ template <class PrimaryAllocator, class AllocatorCache,
     class SecondaryAllocator>
 class ScudoCombinedAllocator {
  public:
-  void Init(bool AllocatorMayReturnNull, s32 ReleaseToOSIntervalMs) {
+  void Init(s32 ReleaseToOSIntervalMs) {
     Primary.Init(ReleaseToOSIntervalMs);
-    Secondary.Init(AllocatorMayReturnNull);
+    Secondary.Init();
     Stats.Init();
-    atomic_store_relaxed(&MayReturnNull, AllocatorMayReturnNull);
   }
 
   void *Allocate(AllocatorCache *Cache, uptr Size, uptr Alignment,
@@ -37,18 +36,6 @@ class ScudoCombinedAllocator {
     return Secondary.Allocate(&Stats, Size, Alignment);
   }
 
-  void *ReturnNullOrDieOnBadRequest() {
-    if (atomic_load_relaxed(&MayReturnNull))
-      return nullptr;
-    ReportAllocatorCannotReturnNull(false);
-  }
-
-  void *ReturnNullOrDieOnOOM() {
-    if (atomic_load_relaxed(&MayReturnNull))
-      return nullptr;
-    ReportAllocatorCannotReturnNull(true);
-  }
-
   void Deallocate(AllocatorCache *Cache, void *Ptr, bool FromPrimary) {
     if (FromPrimary)
       Cache->Deallocate(&Primary, Primary.GetSizeClass(Ptr), Ptr);
@@ -78,7 +65,6 @@ class ScudoCombinedAllocator {
   PrimaryAllocator Primary;
   SecondaryAllocator Secondary;
   AllocatorGlobalStats Stats;
-  atomic_uint8_t MayReturnNull;
 };
 
 #endif  // SCUDO_ALLOCATOR_COMBINED_H_
diff --git a/lib/scudo/scudo_allocator_secondary.h b/lib/scudo/scudo_allocator_secondary.h
index 2950909b547e..dbfb22565f9c 100644
--- a/lib/scudo/scudo_allocator_secondary.h
+++ b/lib/scudo/scudo_allocator_secondary.h
@@ -24,9 +24,8 @@
 class ScudoLargeMmapAllocator {
  public:
 
-  void Init(bool AllocatorMayReturnNull) {
+  void Init() {
     PageSize = GetPageSizeCached();
-    atomic_store_relaxed(&MayReturnNull, AllocatorMayReturnNull);
   }
 
   void *Allocate(AllocatorStats *Stats, uptr Size, uptr Alignment) {
@@ -42,7 +41,7 @@ class ScudoLargeMmapAllocator {
 
     uptr MapBeg = reinterpret_cast<uptr>(MmapNoAccess(MapSize));
     if (MapBeg == ~static_cast<uptr>(0))
-      return ReturnNullOrDieOnOOM();
+      return ReturnNullOrDieOnFailure::OnOOM();
     // A page-aligned pointer is assumed after that, so check it now.
     CHECK(IsAligned(MapBeg, PageSize));
     uptr MapEnd = MapBeg + MapSize;
@@ -96,12 +95,6 @@ class ScudoLargeMmapAllocator {
     return reinterpret_cast<void *>(Ptr);
   }
 
-  void *ReturnNullOrDieOnOOM() {
-    if (atomic_load_relaxed(&MayReturnNull))
-      return nullptr;
-    ReportAllocatorCannotReturnNull(true);
-  }
-
   void Deallocate(AllocatorStats *Stats, void *Ptr) {
     SecondaryHeader *Header = getHeader(Ptr);
     {
@@ -140,7 +133,6 @@ class ScudoLargeMmapAllocator {
   const uptr HeadersSize = SecondaryHeaderSize + AlignedChunkHeaderSize;
   uptr PageSize;
   SpinMutex StatsMutex;
-  atomic_uint8_t MayReturnNull;
 };
 
 #endif  // SCUDO_ALLOCATOR_SECONDARY_H_
diff --git a/lib/tsan/rtl/tsan_mman.cc b/lib/tsan/rtl/tsan_mman.cc
index 2dea24915e8a..fa0d0cafe9b6 100644
--- a/lib/tsan/rtl/tsan_mman.cc
+++ b/lib/tsan/rtl/tsan_mman.cc
@@ -112,9 +112,8 @@ ScopedGlobalProcessor::~ScopedGlobalProcessor() {
 }
 
 void InitializeAllocator() {
-  allocator()->Init(
-      common_flags()->allocator_may_return_null,
-      common_flags()->allocator_release_to_os_interval_ms);
+  SetAllocatorMayReturnNull(common_flags()->allocator_may_return_null);
+  allocator()->Init(common_flags()->allocator_release_to_os_interval_ms);
 }
 
 void InitializeAllocatorLate() {
@@ -151,7 +150,7 @@ static void SignalUnsafeCall(ThreadState *thr, uptr pc) {
 
 void *user_alloc(ThreadState *thr, uptr pc, uptr sz, uptr align, bool signal) {
   if ((sz >= (1ull << 40)) || (align >= (1ull << 40)))
-    return allocator()->ReturnNullOrDieOnBadRequest();
+    return Allocator::FailureHandler::OnBadRequest();
   void *p = allocator()->Allocate(&thr->proc()->alloc_cache, sz, align);
   if (p == 0)
     return 0;
@@ -164,7 +163,7 @@ void *user_alloc(ThreadState *thr, uptr pc, uptr sz, uptr align, bool signal) {
 
 void *user_calloc(ThreadState *thr, uptr pc, uptr size, uptr n) {
   if (CallocShouldReturnNullDueToOverflow(size, n))
-    return allocator()->ReturnNullOrDieOnBadRequest();
+    return Allocator::FailureHandler::OnBadRequest();
   void *p = user_alloc(thr, pc, n * size);
   if (p)
     internal_memset(p, 0, n * size);
diff --git a/lib/ubsan/ubsan_handlers.cc b/lib/ubsan/ubsan_handlers.cc
index 5dabbd8e08cc..185752719aff 100644
--- a/lib/ubsan/ubsan_handlers.cc
+++ b/lib/ubsan/ubsan_handlers.cc
@@ -473,9 +473,12 @@ void __ubsan::__ubsan_handle_function_type_mismatch_abort(
   Die();
 }
 
-static void handleNonNullReturn(NonNullReturnData *Data, ReportOptions Opts,
-                                bool IsAttr) {
-  SourceLocation Loc = Data->Loc.acquire();
+static void handleNonNullReturn(NonNullReturnData *Data, SourceLocation *LocPtr,
+                                ReportOptions Opts, bool IsAttr) {
+  if (!LocPtr)
+    UNREACHABLE("source location pointer is null!");
+
+  SourceLocation Loc = LocPtr->acquire();
   ErrorType ET = ErrorType::InvalidNullReturn;
 
   if (ignoreReport(Loc, Opts, ET))
@@ -491,25 +494,29 @@ static void handleNonNullReturn(NonNullReturnData *Data, ReportOptions Opts,
                    : "_Nonnull return type annotation");
 }
 
-void __ubsan::__ubsan_handle_nonnull_return(NonNullReturnData *Data) {
+void __ubsan::__ubsan_handle_nonnull_return_v1(NonNullReturnData *Data,
+                                               SourceLocation *LocPtr) {
   GET_REPORT_OPTIONS(false);
-  handleNonNullReturn(Data, Opts, true);
+  handleNonNullReturn(Data, LocPtr, Opts, true);
 }
 
-void __ubsan::__ubsan_handle_nonnull_return_abort(NonNullReturnData *Data) {
+void __ubsan::__ubsan_handle_nonnull_return_v1_abort(NonNullReturnData *Data,
+                                                     SourceLocation *LocPtr) {
   GET_REPORT_OPTIONS(true);
-  handleNonNullReturn(Data, Opts, true);
+  handleNonNullReturn(Data, LocPtr, Opts, true);
   Die();
 }
 
-void __ubsan::__ubsan_handle_nullability_return(NonNullReturnData *Data) {
+void __ubsan::__ubsan_handle_nullability_return_v1(NonNullReturnData *Data,
+                                                   SourceLocation *LocPtr) {
   GET_REPORT_OPTIONS(false);
-  handleNonNullReturn(Data, Opts, false);
+  handleNonNullReturn(Data, LocPtr, Opts, false);
 }
 
-void __ubsan::__ubsan_handle_nullability_return_abort(NonNullReturnData *Data) {
+void __ubsan::__ubsan_handle_nullability_return_v1_abort(
+    NonNullReturnData *Data, SourceLocation *LocPtr) {
   GET_REPORT_OPTIONS(true);
-  handleNonNullReturn(Data, Opts, false);
+  handleNonNullReturn(Data, LocPtr, Opts, false);
   Die();
 }
 
diff --git a/lib/ubsan/ubsan_handlers.h b/lib/ubsan/ubsan_handlers.h
index 483c18ced579..796321b81889 100644
--- a/lib/ubsan/ubsan_handlers.h
+++ b/lib/ubsan/ubsan_handlers.h
@@ -132,14 +132,13 @@ RECOVERABLE(function_type_mismatch,
             ValueHandle Val)
 
 struct NonNullReturnData {
-  SourceLocation Loc;
   SourceLocation AttrLoc;
 };
 
 /// \brief Handle returning null from function with the returns_nonnull
 /// attribute, or a return type annotated with _Nonnull.
-RECOVERABLE(nonnull_return, NonNullReturnData *Data)
-RECOVERABLE(nullability_return, NonNullReturnData *Data)
+RECOVERABLE(nonnull_return_v1, NonNullReturnData *Data, SourceLocation *Loc)
+RECOVERABLE(nullability_return_v1, NonNullReturnData *Data, SourceLocation *Loc)
 
 struct NonNullArgData {
   SourceLocation Loc;
diff --git a/lib/ubsan/ubsan_interface.inc b/lib/ubsan/ubsan_interface.inc
index 681476b3c91d..a69ca57cd7af 100644
--- a/lib/ubsan/ubsan_interface.inc
+++ b/lib/ubsan/ubsan_interface.inc
@@ -28,12 +28,12 @@ INTERFACE_FUNCTION(__ubsan_handle_negate_overflow)
 INTERFACE_FUNCTION(__ubsan_handle_negate_overflow_abort)
 INTERFACE_FUNCTION(__ubsan_handle_nonnull_arg)
 INTERFACE_FUNCTION(__ubsan_handle_nonnull_arg_abort)
-INTERFACE_FUNCTION(__ubsan_handle_nonnull_return)
-INTERFACE_FUNCTION(__ubsan_handle_nonnull_return_abort)
+INTERFACE_FUNCTION(__ubsan_handle_nonnull_return_v1)
+INTERFACE_FUNCTION(__ubsan_handle_nonnull_return_v1_abort)
 INTERFACE_FUNCTION(__ubsan_handle_nullability_arg)
 INTERFACE_FUNCTION(__ubsan_handle_nullability_arg_abort)
-INTERFACE_FUNCTION(__ubsan_handle_nullability_return)
-INTERFACE_FUNCTION(__ubsan_handle_nullability_return_abort)
+INTERFACE_FUNCTION(__ubsan_handle_nullability_return_v1)
+INTERFACE_FUNCTION(__ubsan_handle_nullability_return_v1_abort)
 INTERFACE_FUNCTION(__ubsan_handle_out_of_bounds)
 INTERFACE_FUNCTION(__ubsan_handle_out_of_bounds_abort)
 INTERFACE_FUNCTION(__ubsan_handle_pointer_overflow)
diff --git a/lib/xray/xray_interface.cc b/lib/xray/xray_interface.cc
index e912b6e478a3..694d34c0102b 100644
--- a/lib/xray/xray_interface.cc
+++ b/lib/xray/xray_interface.cc
@@ -312,7 +312,7 @@ __xray_unpatch_function(int32_t FuncId) XRAY_NEVER_INSTRUMENT {
   return patchFunction(FuncId, false);
 }
 
-int __xray_set_handler_arg1(void (*Handler)(int32_t, XRayEntryType, uint64_t)) {
+int __xray_set_handler_arg1(void (*entry)(int32_t, XRayEntryType, uint64_t)) {
   if (!__sanitizer::atomic_load(&XRayInitialized,
                                 __sanitizer::memory_order_acquire))
     return 0;
@@ -320,7 +320,7 @@ int __xray_set_handler_arg1(void (*Handler)(int32_t, XRayEntryType, uint64_t)) {
   // A relaxed write might not be visible even if the current thread gets
   // scheduled on a different CPU/NUMA node.  We need to wait for everyone to
   // have this handler installed for consistency of collected data across CPUs.
-  __sanitizer::atomic_store(&XRayArgLogger, reinterpret_cast<uint64_t>(Handler),
+  __sanitizer::atomic_store(&XRayArgLogger, reinterpret_cast<uint64_t>(entry),
                             __sanitizer::memory_order_release);
   return 1;
 }
diff --git a/test/asan/TestCases/Linux/allocator_oom_test.cc b/test/asan/TestCases/Linux/allocator_oom_test.cc
index c93e9fe21726..4c696f325803 100644
--- a/test/asan/TestCases/Linux/allocator_oom_test.cc
+++ b/test/asan/TestCases/Linux/allocator_oom_test.cc
@@ -28,6 +28,9 @@
 // RUN: %env_asan_opts=allocator_may_return_null=1     %run %t realloc-after-malloc 2>&1 \
 // RUN:   | FileCheck %s --check-prefixes=CHECK-MALLOC-REALLOC,CHECK-NULL
 
+// ASan shadow memory on s390 is too large for this test.
+// UNSUPPORTED: s390
+
 #include <stdlib.h>
 #include <string.h>
 #include <stdio.h>
diff --git a/test/asan/TestCases/Linux/preinstalled_signal.cc b/test/asan/TestCases/Linux/preinstalled_signal.cc
index 40dadf43dc4e..4d466c21f944 100644
--- a/test/asan/TestCases/Linux/preinstalled_signal.cc
+++ b/test/asan/TestCases/Linux/preinstalled_signal.cc
@@ -4,11 +4,13 @@
 // RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=2 not %run %t 2>&1 | FileCheck %s
 
 // RUN: %clangxx -std=c++11 -DTEST_INSTALL_SIG_HANDLER %s -o %t
-// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=1 not %run %t 2>&1 | FileCheck --check-prefix=CHECK-HANDLER %s
+// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=0 not %run %t 2>&1 | FileCheck %s --check-prefix=CHECK-HANDLER
+// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=1 not %run %t 2>&1 | FileCheck %s
 // RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=2 not %run %t 2>&1 | FileCheck %s
 
 // RUN: %clangxx -std=c++11 -DTEST_INSTALL_SIG_ACTION %s -o %t
-// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=1 not %run %t 2>&1 | FileCheck --check-prefix=CHECK-ACTION %s
+// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=0 not %run %t 2>&1 | FileCheck %s --check-prefix=CHECK-ACTION
+// RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=1 not %run %t 2>&1 | FileCheck %s
 // RUN: env LD_PRELOAD=%shared_libasan %env_asan_opts=handle_segv=2 not %run %t 2>&1 | FileCheck %s
 
 // REQUIRES: asan-dynamic-runtime
@@ -51,22 +53,22 @@ int InternalSigaction(int sig, KernelSigaction *act, KernelSigaction *oact) {
   return syscall(__NR_rt_sigaction, sig, act, oact, NSIG / 8);
 }
 
-struct KernelSigaction sigact = {};
+struct KernelSigaction pre_asan = {};
 
 static void Init() {
-  int res = InternalSigaction(SIGSEGV, nullptr, &sigact);
+  int res = InternalSigaction(SIGSEGV, nullptr, &pre_asan);
   assert(res >= 0);
-  assert(sigact.handler == SIG_DFL || sigact.handler == SIG_IGN);
+  assert(pre_asan.handler == SIG_DFL || pre_asan.handler == SIG_IGN);
 #if defined(TEST_INSTALL_SIG_HANDLER)
-  sigact = {};
-  sigact.handler = &SigHandler;
-  res = InternalSigaction(SIGSEGV, &sigact, nullptr);
+  pre_asan = {};
+  pre_asan.handler = &SigHandler;
+  res = InternalSigaction(SIGSEGV, &pre_asan, nullptr);
   assert(res >= 0);
 #elif defined(TEST_INSTALL_SIG_ACTION)
-  sigact = {};
-  sigact.flags = SA_SIGINFO | SA_NODEFER;
-  sigact.handler = (__sighandler_t)&SigAction;
-  res = InternalSigaction(SIGSEGV, &sigact, nullptr);
+  pre_asan = {};
+  pre_asan.flags = SA_SIGINFO | SA_NODEFER;
+  pre_asan.handler = (__sighandler_t)&SigAction;
+  res = InternalSigaction(SIGSEGV, &pre_asan, nullptr);
   assert(res >= 0);
 #endif
 }
@@ -74,21 +76,21 @@ static void Init() {
 __attribute__((section(".preinit_array"), used))
 void (*__local_test_preinit)(void) = Init;
 
-bool ShouldAsanInstallHandlers() {
+bool ExpectUserHandler() {
 #if defined(TEST_INSTALL_SIG_HANDLER) || defined(TEST_INSTALL_SIG_ACTION)
-  return !strcmp(getenv("ASAN_OPTIONS"), "handle_segv=2");
+  return !strcmp(getenv("ASAN_OPTIONS"), "handle_segv=0");
 #endif
-  return true;
+  return false;
 }
 
 int main(int argc, char *argv[]) {
-  KernelSigaction sigact_asan = {};
-  InternalSigaction(SIGSEGV, nullptr, &sigact_asan);
+  KernelSigaction post_asan = {};
+  InternalSigaction(SIGSEGV, nullptr, &post_asan);
 
-  assert(sigact_asan.handler != SIG_DFL);
-  assert(sigact_asan.handler != SIG_IGN);
-  assert(ShouldAsanInstallHandlers() ==
-         (sigact_asan.handler != sigact.handler));
+  assert(post_asan.handler != SIG_DFL);
+  assert(post_asan.handler != SIG_IGN);
+  assert(ExpectUserHandler() ==
+         (post_asan.handler == pre_asan.handler));
 
   raise(SIGSEGV);
   printf("%s\n", handler);
diff --git a/test/asan/TestCases/Windows/oom.cc b/test/asan/TestCases/Windows/oom.cc
index 59cc7ed0e9d1..71a9c2a759a9 100644
--- a/test/asan/TestCases/Windows/oom.cc
+++ b/test/asan/TestCases/Windows/oom.cc
@@ -8,5 +8,5 @@ int main() {
   while (true) {
     void *ptr = malloc(200 * 1024 * 1024);  // 200MB
   }
-// CHECK: failed to allocate
+// CHECK: allocator is terminating the process instead of returning 0
 }
diff --git a/test/asan/TestCases/pr33372.cc b/test/asan/TestCases/pr33372.cc
new file mode 100644
index 000000000000..a4b606e025c9
--- /dev/null
+++ b/test/asan/TestCases/pr33372.cc
@@ -0,0 +1,39 @@
+// RUN: %clangxx_asan -O0 -std=c++11 %s -o %t && %run %t 2>&1 | FileCheck %s
+// RUN: %clangxx_asan -O1 -std=c++11 %s -o %t && %run %t 2>&1 | FileCheck %s
+// RUN: %clangxx_asan -O2 -std=c++11 %s -o %t && %run %t 2>&1 | FileCheck %s
+
+// Test that we do not detect false buffer overflows cased by optimization when
+// when local variable replaced by a smaller global constant.
+// https://bugs.llvm.org/show_bug.cgi?id=33372
+
+#include <stdio.h>
+#include <string.h>
+
+struct A { int x, y, z; };
+struct B { A a; /*gap*/ long b; };
+B *bb;
+
+void test1() {
+  A a1 = {1, 1, 2};
+  B b1 = {a1, 6};
+  bb = new B(b1);
+}
+
+const char KKK[] = {1, 1, 2};
+char bbb[100000];
+
+void test2() {
+  char cc[sizeof(bbb)];
+  memcpy(cc, KKK , sizeof(KKK));
+  memcpy(bbb, cc, sizeof(bbb));
+}
+
+int main(int argc, char *argv[]) {
+  test1();
+  test2();
+  printf("PASSED");
+  return 0;
+}
+
+// CHECK-NOT: ERROR: AddressSanitizer
+// CHECK: PASSED
diff --git a/test/asan/lit.cfg b/test/asan/lit.cfg
index 063c33b02697..e25dd297aa3b 100644
--- a/test/asan/lit.cfg
+++ b/test/asan/lit.cfg
@@ -114,6 +114,7 @@ asan_lit_source_dir = get_required_attr(config, "asan_lit_source_dir")
 if config.android == "1":
   config.available_features.add('android')
   compile_wrapper = os.path.join(asan_lit_source_dir, "android_commands", "android_compile.py") + " "
+  config.compile_wrapper = compile_wrapper
 else:
   config.available_features.add('not-android')
 
diff --git a/test/esan/TestCases/workingset-midreport.cpp b/test/esan/TestCases/workingset-midreport.cpp
index 2c29cf48ccf7..38c376554380 100644
--- a/test/esan/TestCases/workingset-midreport.cpp
+++ b/test/esan/TestCases/workingset-midreport.cpp
@@ -4,6 +4,9 @@
 // RUN: %clang -O0 %s -o %t 2>&1
 // RUN: %run %t 2>&1 | FileCheck %s --check-prefix=CHECK-NO-ESAN
 
+// FIXME: Re-enable once PR33590 is fixed.
+// UNSUPPORTED: x86_64
+
 #include <sanitizer/esan_interface.h>
 #include <sched.h>
 #include <stdio.h>
diff --git a/test/esan/TestCases/workingset-samples.cpp b/test/esan/TestCases/workingset-samples.cpp
index 7db2110b924f..d97b62ba4e8a 100644
--- a/test/esan/TestCases/workingset-samples.cpp
+++ b/test/esan/TestCases/workingset-samples.cpp
@@ -1,6 +1,9 @@
 // RUN: %clang_esan_wset -O0 %s -o %t 2>&1
 // RUN: %run %t 2>&1 | FileCheck %s
 
+// FIXME: Re-enable once PR33590 is fixed.
+// UNSUPPORTED: x86_64
+
 #include <sanitizer/esan_interface.h>
 #include <sched.h>
 #include <stdlib.h>
diff --git a/test/esan/TestCases/workingset-simple.cpp b/test/esan/TestCases/workingset-simple.cpp
index c8a2d52e7b55..f1ac2ecfe132 100644
--- a/test/esan/TestCases/workingset-simple.cpp
+++ b/test/esan/TestCases/workingset-simple.cpp
@@ -1,6 +1,9 @@
 // RUN: %clang_esan_wset -O0 %s -o %t 2>&1
 // RUN: %run %t 2>&1 | FileCheck %s
 
+// FIXME: Re-enable once PR33590 is fixed.
+// UNSUPPORTED: x86_64
+
 #include <stdlib.h>
 #include <string.h>
 #include <sys/mman.h>
diff --git a/test/lsan/TestCases/Darwin/dispatch.mm b/test/lsan/TestCases/Darwin/dispatch.mm
new file mode 100644
index 000000000000..606cc9e1cb1e
--- /dev/null
+++ b/test/lsan/TestCases/Darwin/dispatch.mm
@@ -0,0 +1,59 @@
+// Test for threads spawned with wqthread_start
+// RUN: LSAN_BASE="report_objects=1:use_stacks=0:use_registers=0"
+// RUN: %clangxx_lsan %s -DDISPATCH_ASYNC -o %t-async -framework Foundation
+// RUN: %clangxx_lsan %s -DDISPATCH_SYNC -o %t-sync -framework Foundation
+// RUN: %env_lsan_opts=$LSAN_BASE not %run %t-async 2>&1 | FileCheck %s
+// RUN: %env_lsan_opts=$LSAN_BASE not %run %t-sync 2>&1 | FileCheck %s
+
+#include <dispatch/dispatch.h>
+#include <pthread.h>
+#include <stdlib.h>
+
+#include "sanitizer_common/print_address.h"
+
+bool done = false;
+
+void worker_do_leak(int size) {
+  void *p = malloc(size);
+  print_address("Test alloc: ", 1, p);
+  done = true;
+}
+
+#if DISPATCH_ASYNC
+// Tests for the Grand Central Dispatch. See
+// http://developer.apple.com/library/mac/#documentation/Performance/Reference/GCD_libdispatch_Ref/Reference/reference.html
+// for the reference.
+void TestGCDDispatch() {
+  dispatch_queue_t queue = dispatch_get_global_queue(0, 0);
+  dispatch_block_t block = ^{
+    worker_do_leak(1337);
+  };
+  // dispatch_async() runs the task on a worker thread that does not go through
+  // pthread_create(). We need to verify that LeakSanitizer notices that the
+  // thread has started.
+  dispatch_async(queue, block);
+  while (!done)
+    pthread_yield_np();
+}
+#elif DISPATCH_SYNC
+void TestGCDDispatch() {
+  dispatch_queue_t queue = dispatch_get_global_queue(2, 0);
+  dispatch_block_t block = ^{
+    worker_do_leak(1337);
+  };
+  // dispatch_sync() runs the task on a worker thread that does not go through
+  // pthread_create(). We need to verify that LeakSanitizer notices that the
+  // thread has started.
+  dispatch_sync(queue, block);
+}
+#endif
+
+int main() {
+  TestGCDDispatch();
+  return 0;
+}
+
+// CHECK: Test alloc: [[addr:0x[0-9,a-f]+]]
+// CHECK: LeakSanitizer: detected memory leaks
+// CHECK: [[addr]] (1337 bytes)
+// CHECK: SUMMARY: {{(Leak|Address)}}Sanitizer:
diff --git a/test/lsan/TestCases/Darwin/lit.local.cfg b/test/lsan/TestCases/Darwin/lit.local.cfg
new file mode 100644
index 000000000000..a85dfcd24c08
--- /dev/null
+++ b/test/lsan/TestCases/Darwin/lit.local.cfg
@@ -0,0 +1,9 @@
+def getRoot(config):
+  if not config.parent:
+    return config
+  return getRoot(config.parent)
+
+root = getRoot(config)
+
+if root.host_os not in ['Darwin']:
+  config.unsupported = True
diff --git a/test/lsan/lit.common.cfg b/test/lsan/lit.common.cfg
index 309e8f27be66..610b1b1ad958 100644
--- a/test/lsan/lit.common.cfg
+++ b/test/lsan/lit.common.cfg
@@ -77,4 +77,4 @@ if not (supported_linux or supported_darwin):
 if re.search('mthumb', config.target_cflags) is not None:
   config.unsupported = True
 
-config.suffixes = ['.c', '.cc', '.cpp']
+config.suffixes = ['.c', '.cc', '.cpp', '.mm']
diff --git a/test/profile/Linux/counter_promo_for.c b/test/profile/Linux/counter_promo_for.c
new file mode 100644
index 000000000000..3139646699d5
--- /dev/null
+++ b/test/profile/Linux/counter_promo_for.c
@@ -0,0 +1,59 @@
+// RUN: rm -fr %t.promo.prof
+// RUN: rm -fr %t.nopromo.prof
+// RUN: %clang_pgogen=%t.promo.prof/ -o %t.promo.gen -O2 %s
+// RUN: %clang_pgogen=%t.promo.prof/ -o %t.promo.gen.ll -emit-llvm -S -O2 %s
+// RUN: cat %t.promo.gen.ll | FileCheck --check-prefix=PROMO %s
+// RUN: %run %t.promo.gen
+// RUN: llvm-profdata merge -o %t.promo.profdata %t.promo.prof/
+// RUN: llvm-profdata show --counts --all-functions %t.promo.profdata  > %t.promo.dump
+// RUN: %clang_pgogen=%t.nopromo.prof/ -mllvm -do-counter-promotion=false -o %t.nopromo.gen -O2 %s
+// RUN: %clang_pgogen=%t.nopromo.prof/ -mllvm -do-counter-promotion=false -o %t.nopromo.gen.ll -emit-llvm -S -O2 %s
+// RUN: cat %t.nopromo.gen.ll | FileCheck --check-prefix=NOPROMO %s
+// RUN: %run %t.nopromo.gen
+// RUN: llvm-profdata merge -o %t.nopromo.profdata %t.nopromo.prof/
+// RUN: llvm-profdata show --counts --all-functions %t.nopromo.profdata  > %t.nopromo.dump
+// RUN: diff %t.promo.profdata %t.nopromo.profdata
+
+int g;
+__attribute__((noinline)) void bar(int i) { g += i; }
+
+__attribute__((noinline)) void foo(int n, int N) {
+// PROMO-LABEL: @foo
+// PROMO: load{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 2){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 2){{.*}}
+// PROMO: load{{.*}}@__profc_foo{{.*}} 3){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 3){{.*}}
+//
+// NOPROMO-LABEL: @foo
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 2){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 2){{.*}}
+  int i;
+  for (i = 0; i < N; i++) {
+    if (i < n + 1)
+      bar(1);
+    else if (i < n - 1)
+      bar(2);
+    else
+      bar(3);
+  }
+}
+
+int main() {
+  foo(10, 20); 
+  return 0;
+}
diff --git a/test/profile/Linux/counter_promo_while.c b/test/profile/Linux/counter_promo_while.c
new file mode 100644
index 000000000000..b4d4e7aa20fe
--- /dev/null
+++ b/test/profile/Linux/counter_promo_while.c
@@ -0,0 +1,55 @@
+// RUN: rm -fr %t.promo.prof
+// RUN: rm -fr %t.nopromo.prof
+// RUN: %clang_pgogen=%t.promo.prof/ -o %t.promo.gen -O2 %s
+// RUN: %clang_pgogen=%t.promo.prof/ -o %t.promo.gen.ll -emit-llvm -S -O2 %s
+// RUN: cat %t.promo.gen.ll | FileCheck --check-prefix=PROMO %s
+// RUN: %run %t.promo.gen
+// RUN: llvm-profdata merge -o %t.promo.profdata %t.promo.prof/
+// RUN: llvm-profdata show --counts --all-functions %t.promo.profdata  > %t.promo.dump
+// RUN: %clang_pgogen=%t.nopromo.prof/ -mllvm -do-counter-promotion=false -o %t.nopromo.gen -O2 %s
+// RUN: %clang_pgogen=%t.nopromo.prof/ -mllvm -do-counter-promotion=false -o %t.nopromo.gen.ll -emit-llvm -S -O2 %s
+// RUN: cat %t.nopromo.gen.ll | FileCheck --check-prefix=NOPROMO %s
+// RUN: %run %t.nopromo.gen
+// RUN: llvm-profdata merge -o %t.nopromo.profdata %t.nopromo.prof/
+// RUN: llvm-profdata show --counts --all-functions %t.nopromo.profdata  > %t.nopromo.dump
+// RUN: diff %t.promo.profdata %t.nopromo.profdata
+int g;
+__attribute__((noinline)) void bar(int i) { g += i; }
+__attribute__((noinline)) void foo(int n, int N) {
+// PROMO-LABEL: @foo
+// PROMO: load{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// PROMO-NEXT: load{{.*}}@__profc_foo{{.*}} 2){{.*}}
+// PROMO-NEXT: add
+// PROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 2){{.*}}
+//
+// NOPROMO-LABEL: @foo
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 0){{.*}}
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 1){{.*}}
+// NOPROMO: load{{.*}}@__profc_foo{{.*}} 2){{.*}}
+// NOPROMO-NEXT: add
+// NOPROMO-NEXT: store{{.*}}@__profc_foo{{.*}} 2){{.*}}
+  int i = 0;
+  while (i < N) {
+    if (i < n + 1)
+      bar(1);
+    else if (i < n - 1)
+      bar(2);
+    else
+      bar(3);
+    i++;
+  }
+}
+
+int main() {
+  foo(10, 20);
+  return 0;
+}
diff --git a/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter.cc b/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter.cc
index 1a4ad1f066d9..b7246ebf2751 100644
--- a/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter.cc
+++ b/test/sanitizer_common/TestCases/sanitizer_coverage_inline8bit_counter.cc
@@ -1,7 +1,7 @@
 // Tests -fsanitize-coverage=inline-8bit-counters
 //
 // REQUIRES: has_sancovcc,stable-runtime
-// UNSUPPORTED: i386-darwin, x86_64-darwin, x86_64h-darwin
+// UNSUPPORTED: i386-darwin
 //
 // RUN: %clangxx -O0 %s -fsanitize-coverage=inline-8bit-counters 2>&1
 
diff --git a/test/scudo/random_shuffle.cpp b/test/scudo/random_shuffle.cpp
index 41e67ded67e9..05a432615017 100644
--- a/test/scudo/random_shuffle.cpp
+++ b/test/scudo/random_shuffle.cpp
@@ -7,7 +7,7 @@
 // RUN: %run %t 10000 > %T/random_shuffle_tmp_dir/out2
 // RUN: not diff %T/random_shuffle_tmp_dir/out?
 // RUN: rm -rf %T/random_shuffle_tmp_dir
-// UNSUPPORTED: i386-linux,i686-linux,arm-linux,armhf-linux,aarch64-linux
+// UNSUPPORTED: i386-linux,i686-linux,arm-linux,armhf-linux,aarch64-linux,mips-linux,mipsel-linux,mips64-linux,mips64el-linux
 
 // Tests that the allocator shuffles the chunks before returning to the user.
 
diff --git a/test/ubsan/TestCases/Misc/nonnull.cpp b/test/ubsan/TestCases/Misc/nonnull.cpp
index c3ab49c11df7..d5cd2bf763b7 100644
--- a/test/ubsan/TestCases/Misc/nonnull.cpp
+++ b/test/ubsan/TestCases/Misc/nonnull.cpp
@@ -1,15 +1,42 @@
-// RUN: %clangxx -fsanitize=returns-nonnull-attribute %s -O3 -o %t
-// RUN: %run %t foo
+// RUN: %clangxx -fsanitize=returns-nonnull-attribute -w %s -O3 -o %t
+// RUN: %run %t foo 2>&1 | count 0
 // RUN: %run %t 2>&1 | FileCheck %s
+// RUN: %clangxx -fsanitize=returns-nonnull-attribute -fno-sanitize-recover=returns-nonnull-attribute -w %s -O3 -o %t.abort
+// RUN: not %run %t.abort &> /dev/null
 
 __attribute__((returns_nonnull)) char *foo(char *a);
 
 char *foo(char *a) {
+  // CHECK: nonnull.cpp:[[@LINE+2]]:3: runtime error: null pointer returned from function declared to never return null
+  // CHECK-NEXT: nonnull.cpp:[[@LINE-4]]:16: note: returns_nonnull attribute specified here
   return a;
-  // CHECK: nonnull.cpp:[[@LINE+2]]:1: runtime error: null pointer returned from function declared to never return null
-  // CHECK-NEXT: nonnull.cpp:[[@LINE-5]]:16: note: returns_nonnull attribute specified here
+}
+
+__attribute__((returns_nonnull)) char *bar(int x, char *a) {
+  if (x > 10) {
+    // CHECK: nonnull.cpp:[[@LINE+2]]:5: runtime error: null pointer returned from function declared to never return null
+    // CHECK-NEXT: nonnull.cpp:[[@LINE-3]]:16: note: returns_nonnull attribute specified here
+    return a;
+  } else {
+    // CHECK: nonnull.cpp:[[@LINE+2]]:5: runtime error: null pointer returned from function declared to never return null
+    // CHECK-NEXT: nonnull.cpp:[[@LINE-7]]:16: note: returns_nonnull attribute specified here
+    return a;
+  }
 }
 
 int main(int argc, char **argv) {
-  return foo(argv[1]) == 0;
+  char *a = argv[1];
+
+  foo(a);
+
+  bar(20, a);
+
+  // We expect to see a runtime error the first time we cover the "else"...
+  bar(5, a);
+
+  // ... but not a second time.
+  // CHECK-NOT: runtime error
+  bar(5, a);
+
+  return 0;
 }
diff --git a/test/ubsan/TestCases/Misc/nullability.c b/test/ubsan/TestCases/Misc/nullability.c
index 583b8ec12583..a6ddf0e2315b 100644
--- a/test/ubsan/TestCases/Misc/nullability.c
+++ b/test/ubsan/TestCases/Misc/nullability.c
@@ -2,7 +2,7 @@
 // RUN: %run %t foo 2>&1 | count 0
 // RUN: %run %t 2>&1 | FileCheck %s
 
-// CHECK: nullability.c:[[@LINE+2]]:51: runtime error: null pointer returned from function declared to never return null
+// CHECK: nullability.c:[[@LINE+2]]:41: runtime error: null pointer returned from function declared to never return null
 // CHECK-NEXT: nullability.c:[[@LINE+1]]:6: note: _Nonnull return type annotation specified here
 int *_Nonnull nonnull_retval1(int *p) { return p; }
 
diff --git a/test/xray/TestCases/Linux/arg1-arg0-logging.cc b/test/xray/TestCases/Linux/arg1-arg0-logging.cc
new file mode 100644
index 000000000000..e7730bfa6172
--- /dev/null
+++ b/test/xray/TestCases/Linux/arg1-arg0-logging.cc
@@ -0,0 +1,39 @@
+// Allow having both the no-arg and arg1 logging implementation live together,
+// and be called in the correct cases.
+//
+// RUN: rm arg0-arg1-logging-* || true
+// RUN: %clangxx_xray -std=c++11 %s -o %t
+// RUN: XRAY_OPTIONS="patch_premain=true verbosity=1 xray_logfile_base=arg0-arg1-logging-" %run %t
+//
+// TODO: Support these in ARM and PPC
+// XFAIL: arm || aarch64 || mips
+// UNSUPPORTED: powerpc64le
+
+#include "xray/xray_interface.h"
+#include <cassert>
+#include <cstdio>
+
+using namespace std;
+
+bool arg0loggercalled = false;
+void arg0logger(int32_t, XRayEntryType) { arg0loggercalled = true; }
+
+[[clang::xray_always_instrument]] void arg0fn() { printf("hello, arg0!\n"); }
+
+bool arg1loggercalled = false;
+void arg1logger(int32_t, XRayEntryType, uint64_t) { arg1loggercalled = true; }
+
+[[ clang::xray_always_instrument, clang::xray_log_args(1) ]] void
+arg1fn(uint64_t arg1) {
+  printf("hello, arg1!\n");
+}
+
+int main(int argc, char *argv[]) {
+  __xray_set_handler(arg0logger);
+  __xray_set_handler_arg1(arg1logger);
+  arg0fn();
+  arg1fn(0xcafef00d);
+  __xray_remove_handler_arg1();
+  __xray_remove_handler();
+  assert(arg0loggercalled && arg1loggercalled);
+}