diff --git a/compiler-rt/lib/scudo/standalone/combined.h b/compiler-rt/lib/scudo/standalone/combined.h
index e4ba6ee7886f2..b013f03a73ae4 100644
--- a/compiler-rt/lib/scudo/standalone/combined.h
+++ b/compiler-rt/lib/scudo/standalone/combined.h
@@ -245,12 +245,14 @@ class Allocator {
   // - unlinking the local stats from the global ones (destroying the cache does
   //   the last two items).
   void commitBack(TSD<ThisT> *TSD) {
+    TSD->assertLocked(/*BypassCheck=*/true);
     Quarantine.drain(&TSD->getQuarantineCache(),
                      QuarantineCallback(*this, TSD->getCache()));
     TSD->getCache().destroy(&Stats);
   }
 
   void drainCache(TSD<ThisT> *TSD) {
+    TSD->assertLocked(/*BypassCheck=*/true);
     Quarantine.drainAndRecycle(&TSD->getQuarantineCache(),
                                QuarantineCallback(*this, TSD->getCache()));
     TSD->getCache().drain();
@@ -363,6 +365,7 @@ class Allocator {
       DCHECK_NE(ClassId, 0U);
       bool UnlockRequired;
       auto *TSD = TSDRegistry.getTSDAndLock(&UnlockRequired);
+      TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
       Block = TSD->getCache().allocate(ClassId);
       // If the allocation failed, the most likely reason with a 32-bit primary
       // is the region being full. In that event, retry in each successively
@@ -1147,6 +1150,7 @@ class Allocator {
       if (LIKELY(ClassId)) {
         bool UnlockRequired;
         auto *TSD = TSDRegistry.getTSDAndLock(&UnlockRequired);
+        TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
         const bool CacheDrained =
             TSD->getCache().deallocate(ClassId, BlockBegin);
         if (UnlockRequired)
@@ -1166,6 +1170,7 @@ class Allocator {
     } else {
       bool UnlockRequired;
       auto *TSD = TSDRegistry.getTSDAndLock(&UnlockRequired);
+      TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
       Quarantine.put(&TSD->getQuarantineCache(),
                      QuarantineCallback(*this, TSD->getCache()), Ptr, Size);
       if (UnlockRequired)
diff --git a/compiler-rt/lib/scudo/standalone/tests/combined_test.cpp b/compiler-rt/lib/scudo/standalone/tests/combined_test.cpp
index 1143aaab8371d..7958e9dabf60d 100644
--- a/compiler-rt/lib/scudo/standalone/tests/combined_test.cpp
+++ b/compiler-rt/lib/scudo/standalone/tests/combined_test.cpp
@@ -512,6 +512,7 @@ SCUDO_TYPED_TEST(ScudoCombinedTest, CacheDrain) NO_THREAD_SAFETY_ANALYSIS {
 
   bool UnlockRequired;
   auto *TSD = Allocator->getTSDRegistry()->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   EXPECT_TRUE(!TSD->getCache().isEmpty());
   TSD->getCache().drain();
   EXPECT_TRUE(TSD->getCache().isEmpty());
@@ -536,6 +537,7 @@ SCUDO_TYPED_TEST(ScudoCombinedTest, ForceCacheDrain) NO_THREAD_SAFETY_ANALYSIS {
 
   bool UnlockRequired;
   auto *TSD = Allocator->getTSDRegistry()->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   EXPECT_TRUE(TSD->getCache().isEmpty());
   EXPECT_EQ(TSD->getQuarantineCache().getSize(), 0U);
   EXPECT_TRUE(Allocator->getQuarantine()->isEmpty());
@@ -842,6 +844,7 @@ TEST(ScudoCombinedTest, BasicTrustyConfig) {
 
   bool UnlockRequired;
   auto *TSD = Allocator->getTSDRegistry()->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   TSD->getCache().drain();
 
   Allocator->releaseToOS(scudo::ReleaseToOS::Force);
diff --git a/compiler-rt/lib/scudo/standalone/tests/tsd_test.cpp b/compiler-rt/lib/scudo/standalone/tests/tsd_test.cpp
index 5953d759a69a9..fad8fcf900771 100644
--- a/compiler-rt/lib/scudo/standalone/tests/tsd_test.cpp
+++ b/compiler-rt/lib/scudo/standalone/tests/tsd_test.cpp
@@ -101,6 +101,7 @@ template <class AllocatorT> static void testRegistry() {
 
   bool UnlockRequired;
   auto TSD = Registry->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   EXPECT_NE(TSD, nullptr);
   EXPECT_EQ(TSD->getCache().Canary, 0U);
   if (UnlockRequired)
@@ -108,6 +109,7 @@ template <class AllocatorT> static void testRegistry() {
 
   Registry->initThreadMaybe(Allocator.get(), /*MinimalInit=*/false);
   TSD = Registry->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   EXPECT_NE(TSD, nullptr);
   EXPECT_EQ(TSD->getCache().Canary, 0U);
   memset(&TSD->getCache(), 0x42, sizeof(TSD->getCache()));
@@ -137,6 +139,7 @@ template <typename AllocatorT> static void stressCache(AllocatorT *Allocator) {
   Registry->initThreadMaybe(Allocator, /*MinimalInit=*/false);
   bool UnlockRequired;
   auto TSD = Registry->getTSDAndLock(&UnlockRequired);
+  TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
   EXPECT_NE(TSD, nullptr);
   // For an exclusive TSD, the cache should be empty. We cannot guarantee the
   // same for a shared TSD.
@@ -195,6 +198,7 @@ static void stressSharedRegistry(MockAllocator<SharedCaches> *Allocator) {
   bool UnlockRequired;
   for (scudo::uptr I = 0; I < 4096U; I++) {
     auto TSD = Registry->getTSDAndLock(&UnlockRequired);
+    TSD->assertLocked(/*BypassCheck=*/!UnlockRequired);
     EXPECT_NE(TSD, nullptr);
     Set.insert(reinterpret_cast<void *>(TSD));
     if (UnlockRequired)
diff --git a/compiler-rt/lib/scudo/standalone/tsd.h b/compiler-rt/lib/scudo/standalone/tsd.h
index f4fa545de5e04..b2108a01900bc 100644
--- a/compiler-rt/lib/scudo/standalone/tsd.h
+++ b/compiler-rt/lib/scudo/standalone/tsd.h
@@ -53,8 +53,14 @@ template <class Allocator> struct alignas(SCUDO_CACHE_LINE_SIZE) TSD {
   inline void unlock() NO_THREAD_SAFETY_ANALYSIS { Mutex.unlock(); }
   inline uptr getPrecedence() { return atomic_load_relaxed(&Precedence); }
 
-  void commitBack(Allocator *Instance) ASSERT_CAPABILITY(Mutex) {
-    Instance->commitBack(this);
+  void commitBack(Allocator *Instance) { Instance->commitBack(this); }
+
+  // As the comments attached to `getCache()`, the TSD doesn't always need to be
+  // locked. In that case, we would only skip the check before we have all TSDs
+  // locked in all paths.
+  void assertLocked(bool BypassCheck) ASSERT_CAPABILITY(Mutex) {
+    if (SCUDO_DEBUG && !BypassCheck)
+      Mutex.assertHeld();
   }
 
   // Ideally, we may want to assert that all the operations on
@@ -66,11 +72,8 @@ template <class Allocator> struct alignas(SCUDO_CACHE_LINE_SIZE) TSD {
   // TODO(chiahungduan): Ideally, we want to do `Mutex.assertHeld` but acquiring
   // TSD doesn't always require holding the lock. Add this assertion while the
   // lock is always acquired.
-  typename Allocator::CacheT &getCache() ASSERT_CAPABILITY(Mutex) {
-    return Cache;
-  }
-  typename Allocator::QuarantineCacheT &getQuarantineCache()
-      ASSERT_CAPABILITY(Mutex) {
+  typename Allocator::CacheT &getCache() REQUIRES(Mutex) { return Cache; }
+  typename Allocator::QuarantineCacheT &getQuarantineCache() REQUIRES(Mutex) {
     return QuarantineCache;
   }
 
diff --git a/compiler-rt/lib/scudo/standalone/tsd_shared.h b/compiler-rt/lib/scudo/standalone/tsd_shared.h
index dcb0948ad78fa..1bca578ee14be 100644
--- a/compiler-rt/lib/scudo/standalone/tsd_shared.h
+++ b/compiler-rt/lib/scudo/standalone/tsd_shared.h
@@ -120,6 +120,11 @@ struct TSDRegistrySharedT {
                 TSDsArraySize);
     for (uptr I = 0; I < NumberOfTSDs; ++I) {
       TSDs[I].lock();
+      // Theoretically, we want to mark TSD::lock()/TSD::unlock() with proper
+      // thread annotations. However, given the TSD is only locked on shared
+      // path, do the assertion in a separate path to avoid confusing the
+      // analyzer.
+      TSDs[I].assertLocked(/*BypassCheck=*/true);
       Str->append("  Shared TSD[%zu]:\n", I);
       TSDs[I].getCache().getStats(Str);
       TSDs[I].unlock();