Fix latent(?) race condition in LockReleaseAll.

tglsfdc · tglsfdc · commit 7747a76e2113 · 2013-11-28T12:17:49.000-05:00
We have for a long time checked the head pointer of each of the backend's
proclock lists and skipped acquiring the corresponding locktable partition
lock if the head pointer was NULL.  This was safe enough in the days when
proclock lists were changed only by the owning backend, but it is pretty
questionable now that the fast-path patch added cases where backends add
entries to other backends' proclock lists.  However, we don't really wish
to revert to locking each partition lock every time, because in simple
transactions that would add a lot of useless lock/unlock cycles on
already-heavily-contended LWLocks.  Fortunately, the only way that another
backend could be modifying our proclock list at this point would be if it
was promoting a formerly fast-path lock of ours; and any such lock must be
one that we'd decided not to delete in the previous loop over the locallock
table.  So it's okay if we miss seeing it in this loop; we'd just decide
not to delete it again.  However, once we've detected a non-empty list,
we'd better re-fetch the list head pointer after acquiring the partition
lock.  This guards against possibly fetching a corrupt-but-non-null pointer
if pointer fetch/store isn't atomic.  It's not clear if any practical
architectures are like that, but we've never assumed that before and don't
wish to start here.  In any case, the situation certainly deserves a code
comment.

While at it, refactor the partition traversal loop to use a for() construct
instead of a while() loop with goto's.

Back-patch, just in case the risk is real and not hypothetical.
diff --git a/src/backend/storage/lmgr/lock.c b/src/backend/storage/lmgr/lock.c
@@ -2099,19 +2099,39 @@ LockReleaseAll(LOCKMETHODID lockmethodid, bool allLocks)
 	{
 		LWLockId	partitionLock = FirstLockMgrLock + partition;
 		SHM_QUEUE  *procLocks = &(MyProc->myProcLocks[partition]);
+		PROCLOCK   *nextplock;
 
-		proclock = (PROCLOCK *) SHMQueueNext(procLocks, procLocks,
-											 offsetof(PROCLOCK, procLink));
-
-		if (!proclock)
+		/*
+		 * If the proclock list for this partition is empty, we can skip
+		 * acquiring the partition lock.  This optimization is trickier than
+		 * it looks, because another backend could be in process of adding
+		 * something to our proclock list due to promoting one of our
+		 * fast-path locks.  However, any such lock must be one that we
+		 * decided not to delete above, so it's okay to skip it again now;
+		 * we'd just decide not to delete it again.  We must, however, be
+		 * careful to re-fetch the list header once we've acquired the
+		 * partition lock, to be sure we have a valid, up-to-date pointer.
+		 * (There is probably no significant risk if pointer fetch/store is
+		 * atomic, but we don't wish to assume that.)
+		 *
+		 * XXX This argument assumes that the locallock table correctly
+		 * represents all of our fast-path locks.  While allLocks mode
+		 * guarantees to clean up all of our normal locks regardless of the
+		 * locallock situation, we lose that guarantee for fast-path locks.
+		 * This is not ideal.
+		 */
+		if (SHMQueueNext(procLocks, procLocks,
+						 offsetof(PROCLOCK, procLink)) == NULL)
 			continue;			/* needn't examine this partition */
 
 		LWLockAcquire(partitionLock, LW_EXCLUSIVE);
 
-		while (proclock)
+		for (proclock = (PROCLOCK *) SHMQueueNext(procLocks, procLocks,
+											   offsetof(PROCLOCK, procLink));
+			 proclock;
+			 proclock = nextplock)
 		{
 			bool		wakeupNeeded = false;
-			PROCLOCK   *nextplock;
 
 			/* Get link first, since we may unlink/delete this proclock */
 			nextplock = (PROCLOCK *)
@@ -2124,7 +2144,7 @@ LockReleaseAll(LOCKMETHODID lockmethodid, bool allLocks)
 
 			/* Ignore items that are not of the lockmethod to be removed */
 			if (LOCK_LOCKMETHOD(*lock) != lockmethodid)
-				goto next_item;
+				continue;
 
 			/*
 			 * In allLocks mode, force release of all locks even if locallock
@@ -2140,7 +2160,7 @@ LockReleaseAll(LOCKMETHODID lockmethodid, bool allLocks)
 			 * holdMask == 0 and are therefore recyclable
 			 */
 			if (proclock->releaseMask == 0 && proclock->holdMask != 0)
-				goto next_item;
+				continue;
 
 			PROCLOCK_PRINT("LockReleaseAll", proclock);
 			LOCK_PRINT("LockReleaseAll", lock, 0);
@@ -2169,9 +2189,6 @@ LockReleaseAll(LOCKMETHODID lockmethodid, bool allLocks)
 						lockMethodTable,
 						LockTagHashCode(&lock->tag),
 						wakeupNeeded);
-
-	next_item:
-			proclock = nextplock;
 		}						/* loop over PROCLOCKs within this partition */
 
 		LWLockRelease(partitionLock);
@@ -3143,19 +3160,27 @@ PostPrepare_Locks(TransactionId xid)
 	{
 		LWLockId	partitionLock = FirstLockMgrLock + partition;
 		SHM_QUEUE  *procLocks = &(MyProc->myProcLocks[partition]);
+		PROCLOCK   *nextplock;
 
-		proclock = (PROCLOCK *) SHMQueueNext(procLocks, procLocks,
-											 offsetof(PROCLOCK, procLink));
-
-		if (!proclock)
+		/*
+		 * If the proclock list for this partition is empty, we can skip
+		 * acquiring the partition lock.  This optimization is safer than the
+		 * situation in LockReleaseAll, because we got rid of any fast-path
+		 * locks during AtPrepare_Locks, so there cannot be any case where
+		 * another backend is adding something to our lists now.  For safety,
+		 * though, we code this the same way as in LockReleaseAll.
+		 */
+		if (SHMQueueNext(procLocks, procLocks,
+						 offsetof(PROCLOCK, procLink)) == NULL)
 			continue;			/* needn't examine this partition */
 
 		LWLockAcquire(partitionLock, LW_EXCLUSIVE);
 
-		while (proclock)
+		for (proclock = (PROCLOCK *) SHMQueueNext(procLocks, procLocks,
+											   offsetof(PROCLOCK, procLink));
+			 proclock;
+			 proclock = nextplock)
 		{
-			PROCLOCK   *nextplock;
-
 			/* Get link first, since we may unlink/relink this proclock */
 			nextplock = (PROCLOCK *)
 				SHMQueueNext(procLocks, &proclock->procLink,
@@ -3167,7 +3192,7 @@ PostPrepare_Locks(TransactionId xid)
 
 			/* Ignore VXID locks */
 			if (lock->tag.locktag_type == LOCKTAG_VIRTUALTRANSACTION)
-				goto next_item;
+				continue;
 
 			PROCLOCK_PRINT("PostPrepare_Locks", proclock);
 			LOCK_PRINT("PostPrepare_Locks", lock, 0);
@@ -3178,7 +3203,7 @@ PostPrepare_Locks(TransactionId xid)
 
 			/* Ignore it if nothing to release (must be a session lock) */
 			if (proclock->releaseMask == 0)
-				goto next_item;
+				continue;
 
 			/* Else we should be releasing all locks */
 			if (proclock->releaseMask != proclock->holdMask)
@@ -3220,9 +3245,6 @@ PostPrepare_Locks(TransactionId xid)
 								 &proclock->procLink);
 
 			PROCLOCK_PRINT("PostPrepare_Locks: updated", proclock);
-
-	next_item:
-			proclock = nextplock;
 		}						/* loop over PROCLOCKs within this partition */
 
 		LWLockRelease(partitionLock);
@@ -3919,7 +3941,7 @@ VirtualXactLockTableInsert(VirtualTransactionId vxid)
  *		unblocking waiters.
  */
 void
-VirtualXactLockTableCleanup()
+VirtualXactLockTableCleanup(void)
 {
 	bool		fastpath;
 	LocalTransactionId lxid;
