3 files changed, 67 insertions, 17 deletions

diff --git a/src/backend/storage/lmgr/lwlock.c b/src/backend/storage/lmgr/lwlock.c
index d23ac62bf84..a62af27d22c 100644
--- a/src/backend/storage/lmgr/lwlock.c
+++ b/src/backend/storage/lmgr/lwlock.c
@@ -104,8 +104,8 @@ typedef struct lwlock_stats
 	int			spin_delay_count;
 }	lwlock_stats;
 
-static int	counts_for_pid = 0;
 static HTAB *lwlock_stats_htab;
+static lwlock_stats lwlock_stats_dummy;
 #endif
 
 #ifdef LOCK_DEBUG
@@ -142,21 +142,39 @@ static void
 init_lwlock_stats(void)
 {
 	HASHCTL		ctl;
+	static MemoryContext lwlock_stats_cxt = NULL;
+	static bool exit_registered = false;
 
-	if (lwlock_stats_htab != NULL)
-	{
-		hash_destroy(lwlock_stats_htab);
-		lwlock_stats_htab = NULL;
-	}
+	if (lwlock_stats_cxt != NULL)
+		MemoryContextDelete(lwlock_stats_cxt);
+
+	/*
+	 * The LWLock stats will be updated within a critical section, which
+	 * requires allocating new hash entries. Allocations within a critical
+	 * section are normally not allowed because running out of memory would
+	 * lead to a PANIC, but LWLOCK_STATS is debugging code that's not normally
+	 * turned on in production, so that's an acceptable risk. The hash entries
+	 * are small, so the risk of running out of memory is minimal in practice.
+	 */
+	lwlock_stats_cxt = AllocSetContextCreate(TopMemoryContext,
+											 "LWLock stats",
+											 ALLOCSET_DEFAULT_MINSIZE,
+											 ALLOCSET_DEFAULT_INITSIZE,
+											 ALLOCSET_DEFAULT_MAXSIZE);
+	MemoryContextAllowInCriticalSection(lwlock_stats_cxt, true);
 
 	MemSet(&ctl, 0, sizeof(ctl));
 	ctl.keysize = sizeof(lwlock_stats_key);
 	ctl.entrysize = sizeof(lwlock_stats);
 	ctl.hash = tag_hash;
+	ctl.hcxt = lwlock_stats_cxt;
 	lwlock_stats_htab = hash_create("lwlock stats", 16384, &ctl,
-									HASH_ELEM | HASH_FUNCTION);
-	counts_for_pid = MyProcPid;
-	on_shmem_exit(print_lwlock_stats, 0);
+									HASH_ELEM | HASH_FUNCTION | HASH_CONTEXT);
+	if (!exit_registered)
+	{
+		on_shmem_exit(print_lwlock_stats, 0);
+		exit_registered = true;
+	}
 }
 
 static void
@@ -190,9 +208,13 @@ get_lwlock_stats_entry(LWLock *lock)
 	lwlock_stats *lwstats;
 	bool		found;
 
-	/* Set up local count state first time through in a given process */
-	if (counts_for_pid != MyProcPid)
-		init_lwlock_stats();
+	/*
+	 * During shared memory initialization, the hash table doesn't exist yet.
+	 * Stats of that phase aren't very interesting, so just collect operations
+	 * on all locks in a single dummy entry.
+	 */
+	if (lwlock_stats_htab == NULL)
+		return &lwlock_stats_dummy;
 
 	/* Fetch or create the entry. */
 	key.tranche = lock->tranche;
@@ -361,6 +383,16 @@ CreateLWLocks(void)
 	LWLockRegisterTranche(0, &MainLWLockTranche);
 }
 
+/*
+ * InitLWLockAccess - initialize backend-local state needed to hold LWLocks
+ */
+void
+InitLWLockAccess(void)
+{
+#ifdef LWLOCK_STATS
+	init_lwlock_stats();
+#endif
+}
 
 /*
  * LWLockAssign - assign a dynamically-allocated LWLock number
diff --git a/src/backend/storage/lmgr/proc.c b/src/backend/storage/lmgr/proc.c
index dfaf10e4b5c..ea88a241443 100644
--- a/src/backend/storage/lmgr/proc.c
+++ b/src/backend/storage/lmgr/proc.c
@@ -411,8 +411,9 @@ InitProcess(void)
 
 	/*
 	 * Now that we have a PGPROC, we could try to acquire locks, so initialize
-	 * the deadlock checker.
+	 * local state needed for LWLocks, and the deadlock checker.
 	 */
+	InitLWLockAccess();
 	InitDeadLockChecking();
 }
 
diff --git a/src/backend/storage/smgr/md.c b/src/backend/storage/smgr/md.c
index 3c1c81a7295..167d61c1afe 100644
--- a/src/backend/storage/smgr/md.c
+++ b/src/backend/storage/smgr/md.c
@@ -115,7 +115,7 @@ typedef struct _MdfdVec
 	struct _MdfdVec *mdfd_chain;	/* next segment, or NULL */
 } MdfdVec;
 
-static MemoryContext MdCxt;		/* context for all md.c allocations */
+static MemoryContext MdCxt;		/* context for all MdfdVec objects */
 
 
 /*
@@ -157,6 +157,7 @@ typedef struct
 
 static HTAB *pendingOpsTable = NULL;
 static List *pendingUnlinks = NIL;
+static MemoryContext pendingOpsCxt;		/* context for the above  */
 
 static CycleCtr mdsync_cycle_ctr = 0;
 static CycleCtr mdckpt_cycle_ctr = 0;
@@ -209,11 +210,27 @@ mdinit(void)
 	{
 		HASHCTL		hash_ctl;
 
+		/*
+		 * XXX: The checkpointer needs to add entries to the pending ops table
+		 * when absorbing fsync requests.  That is done within a critical
+		 * section, which isn't usually allowed, but we make an exception.
+		 * It means that there's a theoretical possibility that you run out of
+		 * memory while absorbing fsync requests, which leads to a PANIC.
+		 * Fortunately the hash table is small so that's unlikely to happen in
+		 * practice.
+		 */
+		pendingOpsCxt = AllocSetContextCreate(MdCxt,
+											  "Pending Ops Context",
+											  ALLOCSET_DEFAULT_MINSIZE,
+											  ALLOCSET_DEFAULT_INITSIZE,
+											  ALLOCSET_DEFAULT_MAXSIZE);
+		MemoryContextAllowInCriticalSection(pendingOpsCxt, true);
+
 		MemSet(&hash_ctl, 0, sizeof(hash_ctl));
 		hash_ctl.keysize = sizeof(RelFileNode);
 		hash_ctl.entrysize = sizeof(PendingOperationEntry);
 		hash_ctl.hash = tag_hash;
-		hash_ctl.hcxt = MdCxt;
+		hash_ctl.hcxt = pendingOpsCxt;
 		pendingOpsTable = hash_create("Pending Ops Table",
 									  100L,
 									  &hash_ctl,
@@ -1516,7 +1533,7 @@ RememberFsyncRequest(RelFileNode rnode, ForkNumber forknum, BlockNumber segno)
 	else if (segno == UNLINK_RELATION_REQUEST)
 	{
 		/* Unlink request: put it in the linked list */
-		MemoryContext oldcxt = MemoryContextSwitchTo(MdCxt);
+		MemoryContext oldcxt = MemoryContextSwitchTo(pendingOpsCxt);
 		PendingUnlinkEntry *entry;
 
 		/* PendingUnlinkEntry doesn't store forknum, since it's always MAIN */
@@ -1533,7 +1550,7 @@ RememberFsyncRequest(RelFileNode rnode, ForkNumber forknum, BlockNumber segno)
 	else
 	{
 		/* Normal case: enter a request to fsync this segment */
-		MemoryContext oldcxt = MemoryContextSwitchTo(MdCxt);
+		MemoryContext oldcxt = MemoryContextSwitchTo(pendingOpsCxt);
 		PendingOperationEntry *entry;
 		bool		found;