diff --git a/compiler-rt/lib/scudo/scudo_allocator.cpp b/compiler-rt/lib/scudo/scudo_allocator.cpp
--- a/compiler-rt/lib/scudo/scudo_allocator.cpp
+++ b/compiler-rt/lib/scudo/scudo_allocator.cpp
@@ -303,18 +303,23 @@
                  bool ForceZeroContents = false) {
     initThreadMaybe();
 
+    if (UNLIKELY(Alignment > MaxAlignment)) {
+      if (AllocatorMayReturnNull())
+        return nullptr;
+      reportAllocationAlignmentTooBig(Alignment, MaxAlignment);
+    }
+
 #ifdef GWP_ASAN_HOOKS
     if (UNLIKELY(GuardedAlloc.shouldSample())) {
-      if (void *Ptr = GuardedAlloc.allocate(Size))
+      if (void *Ptr = GuardedAlloc.allocate(Size, Alignment)) {
+        if (SCUDO_CAN_USE_HOOKS && &__sanitizer_malloc_hook)
+          __sanitizer_malloc_hook(Ptr, Size);
+
         return Ptr;
+      }
     }
 #endif // GWP_ASAN_HOOKS
 
-    if (UNLIKELY(Alignment > MaxAlignment)) {
-      if (AllocatorMayReturnNull())
-        return nullptr;
-      reportAllocationAlignmentTooBig(Alignment, MaxAlignment);
-    }
     if (UNLIKELY(Alignment < MinAlignment))
       Alignment = MinAlignment;
 
diff --git a/compiler-rt/lib/scudo/standalone/combined.h b/compiler-rt/lib/scudo/standalone/combined.h
--- a/compiler-rt/lib/scudo/standalone/combined.h
+++ b/compiler-rt/lib/scudo/standalone/combined.h
@@ -280,8 +280,12 @@
 
 #ifdef GWP_ASAN_HOOKS
     if (UNLIKELY(GuardedAlloc.shouldSample())) {
-      if (void *Ptr = GuardedAlloc.allocate(roundUpTo(Size, Alignment)))
+      if (void *Ptr = GuardedAlloc.allocate(Size, Alignment)) {
+        if (UNLIKELY(&__scudo_allocate_hook))
+          __scudo_allocate_hook(Ptr, Size);
+        Stats.add(StatAllocated, Size);
         return Ptr;
+      }
     }
 #endif // GWP_ASAN_HOOKS
 
@@ -480,18 +484,20 @@
     // being destroyed properly. Any other heap operation will do a full init.
     initThreadMaybe(/*MinimalInit=*/true);
 
+    if (UNLIKELY(&__scudo_deallocate_hook))
+      __scudo_deallocate_hook(Ptr);
+
+    if (UNLIKELY(!Ptr))
+      return;
+
 #ifdef GWP_ASAN_HOOKS
     if (UNLIKELY(GuardedAlloc.pointerIsMine(Ptr))) {
       GuardedAlloc.deallocate(Ptr);
+      Stats.add(StatFree, GuardedAlloc.getSize(Ptr));
       return;
     }
 #endif // GWP_ASAN_HOOKS
 
-    if (UNLIKELY(&__scudo_deallocate_hook))
-      __scudo_deallocate_hook(Ptr);
-
-    if (UNLIKELY(!Ptr))
-      return;
     if (UNLIKELY(!isAligned(reinterpret_cast<uptr>(Ptr), MinAlignment)))
       reportMisalignedPointer(AllocatorAction::Deallocating, Ptr);
 
@@ -547,6 +553,7 @@
       if (NewPtr)
         memcpy(NewPtr, OldPtr, (NewSize < OldSize) ? NewSize : OldSize);
       GuardedAlloc.deallocate(OldPtr);
+      Stats.add(StatFree, OldSize);
       return NewPtr;
     }
 #endif // GWP_ASAN_HOOKS
diff --git a/compiler-rt/lib/scudo/standalone/tests/wrappers_c_test.cpp b/compiler-rt/lib/scudo/standalone/tests/wrappers_c_test.cpp
--- a/compiler-rt/lib/scudo/standalone/tests/wrappers_c_test.cpp
+++ b/compiler-rt/lib/scudo/standalone/tests/wrappers_c_test.cpp
@@ -83,6 +83,18 @@
   EXPECT_EQ(errno, ENOMEM);
 }
 
+TEST(ScudoWrappersCTest, SmallAlign) {
+  void *P;
+  for (size_t Size = 1; Size <= 0x10000; Size <<= 1) {
+    for (size_t Align = 1; Align <= 0x10000; Align <<= 1) {
+      for (size_t Count = 0; Count < 3; ++Count) {
+        P = memalign(Align, Size);
+        EXPECT_TRUE(reinterpret_cast<uintptr_t>(P) % Align == 0);
+      }
+    }
+  }
+}
+
 TEST(ScudoWrappersCTest, Memalign) {
   void *P;
   for (size_t I = FIRST_32_SECOND_64(2U, 3U); I <= 18U; I++) {
diff --git a/compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp b/compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
--- a/compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
+++ b/compiler-rt/lib/scudo/standalone/tests/wrappers_cpp_test.cpp
@@ -66,6 +66,10 @@
 };
 
 TEST(ScudoWrappersCppTest, New) {
+  if (getenv("SKIP_TYPE_MISMATCH")) {
+    printf("Skipped type mismatch tests.\n");
+    return;
+  }
   testCxxNew<bool>();
   testCxxNew<uint8_t>();
   testCxxNew<uint16_t>();
diff --git a/compiler-rt/test/scudo/standalone/CMakeLists.txt b/compiler-rt/test/scudo/standalone/CMakeLists.txt
--- a/compiler-rt/test/scudo/standalone/CMakeLists.txt
+++ b/compiler-rt/test/scudo/standalone/CMakeLists.txt
@@ -4,6 +4,12 @@
     ${CMAKE_CURRENT_BINARY_DIR}/unit/lit.site.cfg.py)
   list(APPEND SCUDO_STANDALONE_TEST_DEPS ScudoUnitTests)
   list(APPEND SCUDO_STANDALONE_TESTSUITES ${CMAKE_CURRENT_BINARY_DIR}/unit)
+  if (COMPILER_RT_HAS_GWP_ASAN)
+    configure_lit_site_cfg(
+      ${CMAKE_CURRENT_SOURCE_DIR}/unit/gwp_asan/lit.site.cfg.py.in
+      ${CMAKE_CURRENT_BINARY_DIR}/unit/gwp_asan/lit.site.cfg.py)
+    list(APPEND SCUDO_STANDALONE_TESTSUITES ${CMAKE_CURRENT_BINARY_DIR}/unit/gwp_asan)
+  endif()
 endif()
 
 add_lit_testsuite(check-scudo_standalone
diff --git a/compiler-rt/test/scudo/standalone/unit/gwp_asan/lit.site.cfg.py.in b/compiler-rt/test/scudo/standalone/unit/gwp_asan/lit.site.cfg.py.in
new file mode 100644
--- /dev/null
+++ b/compiler-rt/test/scudo/standalone/unit/gwp_asan/lit.site.cfg.py.in
@@ -0,0 +1,21 @@
+@LIT_SITE_CFG_IN_HEADER@
+
+# Load common config for all compiler-rt unit tests.
+lit_config.load_config(config, "@COMPILER_RT_BINARY_DIR@/unittests/lit.common.unit.configured")
+
+# Setup config name.
+config.name = 'ScudoStandalone-Unit-GwpAsanTorture'
+
+# Setup test source and exec root.
+# For unit tests, we define it as build directory with unit tests.
+config.test_exec_root = "@COMPILER_RT_BINARY_DIR@/lib/scudo/standalone/tests"
+config.test_source_root = config.test_exec_root
+
+# Enable GWP-ASan in 'torture' mode for scudo tests. Every allocation that can go to GWP-ASan,
+# should go to GWP-ASan. Reserves 409MiB of vaddr space for the guarded pool, and this should
+# be paged in on demand. If necessary (for 32-bit or places where kernel commits immediately),
+# this could possibly be reduced further.
+config.environment['GWP_ASAN_OPTIONS'] = 'SampleRate=1:MaxSimultaneousAllocations=100000'
+
+# GWP-ASan doesn't support malloc-type mismatch.
+config.environment['SKIP_TYPE_MISMATCH'] = '1'