Test Mat's Win/ARM64 PR5 alone without VirtualThread changes

src/java.base/share/classes/java/util/concurrent/LinkedTransferQueue.java

@@ -590,6 +590,11 @@ final Object xfer(Object e, long ns) {
                 q = p.next;
                 if (p.isData != haveData && haveData != (m != null)) {
                     if (p.cmpExItem(m, e) == m) {
+                        // Full fence (StoreLoad) for Dekker with await() which
+                        // writes waiter then reads item. On ARM64, CAS
+                        // (ldaxr/stlxr) + plain load to a different field does
+                        // NOT provide StoreLoad ordering.
+                        VarHandle.fullFence();
                         Thread w = p.waiter; // matched complementary node
                         if (p != h && h == cmpExHead(h, (q == null) ? p : q))
                             h.next = h;     // advance head; self-link old

src/java.base/share/classes/java/util/concurrent/SynchronousQueue.java

@@ -177,6 +177,12 @@ final Object xferLifo(Object e, long ns) {
                     else if (p.cmpExItem(m, e) != m)
                         p = head;                  // missed; restart
                     else {                         // matched complementary node
+                        // Full fence (StoreLoad) for Dekker with await() which
+                        // writes waiter then reads item. On ARM64, CAS
+                        // (ldaxr/stlxr) + plain load to a different field does
+                        // NOT provide StoreLoad ordering.
+                        // GHA Run 04
+                        VarHandle.fullFence();
                         Thread w = p.waiter;
                         cmpExHead(p, p.next);
                         LockSupport.unpark(w);

src/java.base/share/classes/java/util/concurrent/locks/AbstractQueuedSynchronizer.java

@@ -782,6 +782,13 @@ final int acquire(Node node, int arg, boolean shared,
                 Thread.onSpinWait();
             } else if (node.status == 0) {
                 node.status = WAITING;          // enable signal and recheck
+                // Full fence (StoreLoad) to ensure WAITING status is visible
+                // before re-reading state in tryAcquire/tryAcquireShared
+                // (Dekker pattern with releaseShared/release which writes
+                // state then reads node.status in signalNext).
+                // On ARM64, volatile write (stlr) + volatile read (ldar) to
+                // different addresses does NOT provide StoreLoad ordering.
+                U.fullFence();
             } else {
                 spins = postSpins = (byte)((postSpins << 1) | 1);
                 try {
@@ -1097,6 +1104,13 @@ public final boolean tryAcquireNanos(int arg, long nanosTimeout)
      */
     public final boolean release(int arg) {
         if (tryRelease(arg)) {
+            // Full fence (StoreLoad) to ensure the state update from
+            // tryRelease is visible before reading node.status in signalNext
+            // (Dekker pattern: release writes state then reads status,
+            // acquire writes status then reads state).
+            // On ARM64, CAS (stlxr/release) + ldar to different addresses
+            // does NOT provide StoreLoad ordering.
+            U.fullFence();
             signalNext(head);
             return true;
         }
@@ -1184,6 +1198,8 @@ public final boolean tryAcquireSharedNanos(int arg, long nanosTimeout)
      */
     public final boolean releaseShared(int arg) {
         if (tryReleaseShared(arg)) {
+            // Full fence (StoreLoad) — see comment in release()
+            U.fullFence();
             signalNext(head);
             return true;
         }