Change the autovacuum launcher to use WaitLatch instead of a poll loop.

In pursuit of this (and with the expectation that WaitLatch will be needed
in more places), convert the latch field that was already added to PGPROC
for sync rep into a generic latch that is activated for all PGPROC-owning
processes, and change many of the standard backend signal handlers to set
that latch when a signal happens.  This will allow WaitLatch callers to be
wakened properly by these signals.

In passing, fix a whole bunch of signal handlers that had been hacked to do
things that might change errno, without adding the necessary save/restore
logic for errno.  Also make some minor fixes in unix_latch.c, and clean
up bizarre and unsafe scheme for disowning the process's latch.  Much of
this has to be back-patched into 9.1.

Peter Geoghegan, with additional work by Tom

Author: tglsfdc

src/backend/access/transam/xlog.c

@@ -9957,34 +9957,50 @@ startupproc_quickdie(SIGNAL_ARGS)
 static void
 StartupProcSigUsr1Handler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	latch_sigusr1_handler();
+
+	errno = save_errno;
 }
 
 /* SIGUSR2: set flag to finish recovery */
 static void
 StartupProcTriggerHandler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	promote_triggered = true;
 	WakeupRecovery();
+
+	errno = save_errno;
 }
 
 /* SIGHUP: set flag to re-read config file at next convenient time */
 static void
 StartupProcSigHupHandler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	got_SIGHUP = true;
 	WakeupRecovery();
+
+	errno = save_errno;
 }
 
 /* SIGTERM: set flag to abort redo and exit */
 static void
 StartupProcShutdownHandler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	if (in_restore_command)
 		proc_exit(1);
 	else
 		shutdown_requested = true;
 	WakeupRecovery();
+
+	errno = save_errno;
 }
 
 /* Handle SIGHUP and SIGTERM signals of startup process */

src/backend/port/unix_latch.c

@@ -138,9 +138,12 @@ DisownLatch(volatile Latch *latch)
  * function returns immediately.
  *
  * The 'timeout' is given in milliseconds. It must be >= 0 if WL_TIMEOUT flag
- * is given.  On some platforms, signals cause the timeout to be restarted,
- * so beware that the function can sleep for several times longer than the
- * specified timeout.
+ * is given.  On some platforms, signals do not interrupt the wait, or even
+ * cause the timeout to be restarted, so beware that the function can sleep
+ * for several times longer than the requested timeout.  However, this
+ * difficulty is not so great as it seems, because the signal handlers for any
+ * signals that the caller should respond to ought to be programmed to end the
+ * wait by calling SetLatch.  Ideally, the timeout parameter is vestigial.
  *
  * The latch must be owned by the current process, ie. it must be a
  * backend-local latch initialized with InitLatch, or a shared latch
@@ -261,6 +264,7 @@ WaitLatchOrSocket(volatile Latch *latch, int wakeEvents, pgsocket sock,
 		{
 			if (errno == EINTR)
 				continue;
+			waiting = false;
 			ereport(ERROR,
 					(errcode_for_socket_access(),
 					 errmsg("select() failed: %m")));
@@ -294,6 +298,10 @@ WaitLatchOrSocket(volatile Latch *latch, int wakeEvents, pgsocket sock,
  * Sets a latch and wakes up anyone waiting on it.
  *
  * This is cheap if the latch is already set, otherwise not so much.
+ *
+ * NB: when calling this in a signal handler, be sure to save and restore
+ * errno around it.  (That's standard practice in most signal handlers, of
+ * course, but we used to omit it in handlers that only set a flag.)
  */
 void
 SetLatch(volatile Latch *latch)
@@ -339,7 +347,10 @@ SetLatch(volatile Latch *latch)
 	if (owner_pid == 0)
 		return;
 	else if (owner_pid == MyProcPid)
-		sendSelfPipeByte();
+	{
+		if (waiting)
+			sendSelfPipeByte();
+	}
 	else
 		kill(owner_pid, SIGUSR1);
 }
@@ -371,7 +382,11 @@ ResetLatch(volatile Latch *latch)
  * SetLatch uses SIGUSR1 to wake up the process waiting on the latch.
  *
  * Wake up WaitLatch, if we're waiting.  (We might not be, since SIGUSR1 is
- * overloaded for multiple purposes.)
+ * overloaded for multiple purposes; or we might not have reached WaitLatch
+ * yet, in which case we don't need to fill the pipe either.)
+ *
+ * NB: when calling this in a signal handler, be sure to save and restore
+ * errno around it.
  */
 void
 latch_sigusr1_handler(void)
@@ -435,13 +450,19 @@ sendSelfPipeByte(void)
 	}
 }
 
-/* Read all available data from the self-pipe */
+/*
+ * Read all available data from the self-pipe
+ *
+ * Note: this is only called when waiting = true.  If it fails and doesn't
+ * return, it must reset that flag first (though ideally, this will never
+ * happen).
+ */
 static void
 drainSelfPipe(void)
 {
 	/*
 	 * There shouldn't normally be more than one byte in the pipe, or maybe a
-	 * few more if multiple processes run SetLatch at the same instant.
+	 * few bytes if multiple processes run SetLatch at the same instant.
 	 */
 	char		buf[16];
 	int			rc;
@@ -456,9 +477,21 @@ drainSelfPipe(void)
 			else if (errno == EINTR)
 				continue;		/* retry */
 			else
+			{
+				waiting = false;
 				elog(ERROR, "read() on self-pipe failed: %m");
+			}
 		}
 		else if (rc == 0)
+		{
+			waiting = false;
 			elog(ERROR, "unexpected EOF on self-pipe");
+		}
+		else if (rc < sizeof(buf))
+		{
+			/* we successfully drained the pipe; no need to read() again */
+			break;
+		}
+		/* else buffer wasn't big enough, so read again */
 	}
 }

src/backend/postmaster/autovacuum.c

@@ -84,6 +84,7 @@
 #include "postmaster/postmaster.h"
 #include "storage/bufmgr.h"
 #include "storage/ipc.h"
+#include "storage/latch.h"
 #include "storage/pmsignal.h"
 #include "storage/proc.h"
 #include "storage/procsignal.h"
@@ -553,11 +554,11 @@ AutoVacLauncherMain(int argc, char *argv[])
 		Dlelem	   *elem;
 
 		/*
-		 * Emergency bailout if postmaster has died.  This is to avoid the
-		 * necessity for manual cleanup of all postmaster children.
+		 * This loop is a bit different from the normal use of WaitLatch,
+		 * because we'd like to sleep before the first launch of a child
+		 * process.  So it's WaitLatch, then ResetLatch, then check for
+		 * wakening conditions.
 		 */
-		if (!PostmasterIsAlive())
-			proc_exit(1);
 
 		launcher_determine_sleep((AutoVacuumShmem->av_freeWorkers != NULL),
 								 false, &nap);
@@ -566,42 +567,24 @@ AutoVacLauncherMain(int argc, char *argv[])
 		EnableCatchupInterrupt();
 
 		/*
-		 * Sleep for a while according to schedule.
-		 *
-		 * On some platforms, signals won't interrupt the sleep.  To ensure we
-		 * respond reasonably promptly when someone signals us, break down the
-		 * sleep into 1-second increments, and check for interrupts after each
-		 * nap.
+		 * Wait until naptime expires or we get some type of signal (all the
+		 * signal handlers will wake us by calling SetLatch).
 		 */
-		while (nap.tv_sec > 0 || nap.tv_usec > 0)
-		{
-			uint32		sleeptime;
-
-			if (nap.tv_sec > 0)
-			{
-				sleeptime = 1000000;
-				nap.tv_sec--;
-			}
-			else
-			{
-				sleeptime = nap.tv_usec;
-				nap.tv_usec = 0;
-			}
-			pg_usleep(sleeptime);
-
-			/*
-			 * Emergency bailout if postmaster has died.  This is to avoid the
-			 * necessity for manual cleanup of all postmaster children.
-			 */
-			if (!PostmasterIsAlive())
-				proc_exit(1);
+		WaitLatch(&MyProc->procLatch,
+				  WL_LATCH_SET | WL_TIMEOUT | WL_POSTMASTER_DEATH,
+				  (nap.tv_sec * 1000L) + (nap.tv_usec / 1000L));
 
-			if (got_SIGTERM || got_SIGHUP || got_SIGUSR2)
-				break;
-		}
+		ResetLatch(&MyProc->procLatch);
 
 		DisableCatchupInterrupt();
 
+		/*
+		 * Emergency bailout if postmaster has died.  This is to avoid the
+		 * necessity for manual cleanup of all postmaster children.
+		 */
+		if (!PostmasterIsAlive())
+			proc_exit(1);
+
 		/* the normal shutdown case */
 		if (got_SIGTERM)
 			break;
@@ -1321,21 +1304,39 @@ AutoVacWorkerFailed(void)
 static void
 avl_sighup_handler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	got_SIGHUP = true;
+	if (MyProc)
+		SetLatch(&MyProc->procLatch);
+
+	errno = save_errno;
 }
 
 /* SIGUSR2: a worker is up and running, or just finished, or failed to fork */
 static void
 avl_sigusr2_handler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	got_SIGUSR2 = true;
+	if (MyProc)
+		SetLatch(&MyProc->procLatch);
+
+	errno = save_errno;
 }
 
 /* SIGTERM: time to die */
 static void
 avl_sigterm_handler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	got_SIGTERM = true;
+	if (MyProc)
+		SetLatch(&MyProc->procLatch);
+
+	errno = save_errno;
 }
 
 

src/backend/postmaster/pgarch.c

@@ -288,45 +288,57 @@ pgarch_exit(SIGNAL_ARGS)
 static void
 ArchSigHupHandler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	/* set flag to re-read config file at next convenient time */
 	got_SIGHUP = true;
-	/* let the waiting loop iterate */
 	SetLatch(&mainloop_latch);
+
+	errno = save_errno;
 }
 
 /* SIGTERM signal handler for archiver process */
 static void
 ArchSigTermHandler(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	/*
 	 * The postmaster never sends us SIGTERM, so we assume that this means
 	 * that init is trying to shut down the whole system.  If we hang around
 	 * too long we'll get SIGKILL'd.  Set flag to prevent starting any more
 	 * archive commands.
 	 */
 	got_SIGTERM = true;
-	/* let the waiting loop iterate */
 	SetLatch(&mainloop_latch);
+
+	errno = save_errno;
 }
 
 /* SIGUSR1 signal handler for archiver process */
 static void
 pgarch_waken(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	/* set flag that there is work to be done */
 	wakened = true;
-	/* let the waiting loop iterate */
 	SetLatch(&mainloop_latch);
+
+	errno = save_errno;
 }
 
 /* SIGUSR2 signal handler for archiver process */
 static void
 pgarch_waken_stop(SIGNAL_ARGS)
 {
+	int			save_errno = errno;
+
 	/* set flag to do a final cycle and shut down afterwards */
 	ready_to_stop = true;
-	/* let the waiting loop iterate */
 	SetLatch(&mainloop_latch);
+
+	errno = save_errno;
 }
 
 /*

src/backend/replication/syncrep.c

@@ -111,9 +111,6 @@ SyncRepWaitForLSN(XLogRecPtr XactCommitLSN)
 	Assert(SHMQueueIsDetached(&(MyProc->syncRepLinks)));
 	Assert(WalSndCtl != NULL);
 
-	/* Reset the latch before adding ourselves to the queue. */
-	ResetLatch(&MyProc->waitLatch);
-
 	LWLockAcquire(SyncRepLock, LW_EXCLUSIVE);
 	Assert(MyProc->syncRepState == SYNC_REP_NOT_WAITING);
 
@@ -167,7 +164,7 @@ SyncRepWaitForLSN(XLogRecPtr XactCommitLSN)
 		int			syncRepState;
 
 		/* Must reset the latch before testing state. */
-		ResetLatch(&MyProc->waitLatch);
+		ResetLatch(&MyProc->procLatch);
 
 		/*
 		 * Try checking the state without the lock first.  There's no
@@ -247,11 +244,10 @@ SyncRepWaitForLSN(XLogRecPtr XactCommitLSN)
 		}
 
 		/*
-		 * Wait on latch for up to 60 seconds. This allows us to check for
-		 * cancel/die signal or postmaster death regularly while waiting. Note
-		 * that timeout here does not necessarily release from loop.
+		 * Wait on latch.  Any condition that should wake us up will set
+		 * the latch, so no need for timeout.
 		 */
-		WaitLatch(&MyProc->waitLatch, WL_LATCH_SET | WL_TIMEOUT, 60000L);
+		WaitLatch(&MyProc->procLatch, WL_LATCH_SET | WL_POSTMASTER_DEATH, -1);
 	}
 
 	/*
@@ -322,16 +318,14 @@ SyncRepCancelWait(void)
 }
 
 void
-SyncRepCleanupAtProcExit(int code, Datum arg)
+SyncRepCleanupAtProcExit(void)
 {
 	if (!SHMQueueIsDetached(&(MyProc->syncRepLinks)))
 	{
 		LWLockAcquire(SyncRepLock, LW_EXCLUSIVE);
 		SHMQueueDelete(&(MyProc->syncRepLinks));
 		LWLockRelease(SyncRepLock);
 	}
-
-	DisownLatch(&MyProc->waitLatch);
 }
 
 /*
@@ -567,9 +561,7 @@ SyncRepWakeQueue(bool all)
 		/*
 		 * Wake only when we have set state and removed from queue.
 		 */
-		Assert(SHMQueueIsDetached(&(thisproc->syncRepLinks)));
-		Assert(thisproc->syncRepState == SYNC_REP_WAIT_COMPLETE);
-		SetLatch(&(thisproc->waitLatch));
+		SetLatch(&(thisproc->procLatch));
 
 		numprocs++;
 	}