Fix for rare race-condition-like failure: if a backend receives SIGUSR2

(notify/SI-overrun interrupt) while it is in process of doing proc_exit,
it is possible for Async_NotifyHandler() to try to start a transaction
when one is already running.  This leads to Asserts() or worse.  I think
it may only be possible to occur when frontend synchronization is lost
(ie, the elog(FATAL) in SocketBackend() fires), but that is a standard
occurrence after error during COPY.  In any case, I have seen this
failure occur during regression tests, so it is definitely possible.

diff --git a/src/backend/commands/async.c b/src/backend/commands/async.c
index 1945c751e18aa0f25c5f8d10ea3701be9d68c166..a3c27b8f810b1b5a0d6f65a9fa50c6f449a786c6 100644 (file)
--- a/src/backend/commands/async.c
+++ b/src/backend/commands/async.c
@@ -7,7 +7,7 @@
  * Portions Copyright (c) 1994, Regents of the University of California
  *
  * IDENTIFICATION
- *   $Header: /cvsroot/pgsql/src/backend/commands/async.c,v 1.90 2002/09/02 02:47:01 momjian Exp $
+ *   $Header: /cvsroot/pgsql/src/backend/commands/async.c,v 1.91 2002/09/16 01:24:41 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -593,6 +593,10 @@ Async_NotifyHandler(SIGNAL_ARGS)
     * ever turned on.
     */
 
+   /* Don't joggle the elbow of proc_exit */
+   if (proc_exit_inprogress)
+       return;
+
    if (notifyInterruptEnabled)
    {
        /*

diff --git a/src/backend/tcop/postgres.c b/src/backend/tcop/postgres.c
index 01d810088c8a3eabb4c674157fb52b21b3b080b8..cd5f76512a38648b26c3085432b1310277930a1b 100644 (file)
--- a/src/backend/tcop/postgres.c
+++ b/src/backend/tcop/postgres.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *   $Header: /cvsroot/pgsql/src/backend/tcop/postgres.c,v 1.291 2002/09/04 20:31:26 momjian Exp $
+ *   $Header: /cvsroot/pgsql/src/backend/tcop/postgres.c,v 1.292 2002/09/16 01:24:41 tgl Exp $
  *
  * NOTES
  *   this is the "main" module of the postgres backend and
@@ -1071,12 +1071,14 @@ ProcessInterrupts(void)
        ProcDiePending = false;
        QueryCancelPending = false;     /* ProcDie trumps QueryCancel */
        ImmediateInterruptOK = false;   /* not idle anymore */
+       DisableNotifyInterrupt();
        elog(FATAL, "This connection has been terminated by the administrator.");
    }
    if (QueryCancelPending)
    {
        QueryCancelPending = false;
        ImmediateInterruptOK = false;   /* not idle anymore */
+       DisableNotifyInterrupt();
        elog(ERROR, "Query was cancelled.");
    }
    /* If we get here, do nothing (probably, QueryCancelPending was reset) */
@@ -1689,7 +1691,7 @@ PostgresMain(int argc, char *argv[], const char *username)
    if (!IsUnderPostmaster)
    {
        puts("\nPOSTGRES backend interactive interface ");
-       puts("$Revision: 1.291 $ $Date: 2002/09/04 20:31:26 $\n");
+       puts("$Revision: 1.292 $ $Date: 2002/09/16 01:24:41 $\n");
    }
 
    /*
@@ -1736,6 +1738,7 @@ PostgresMain(int argc, char *argv[], const char *username)
        QueryCancelPending = false;
        InterruptHoldoffCount = 1;
        CritSectionCount = 0;   /* should be unnecessary, but... */
+       DisableNotifyInterrupt();
        debug_query_string = NULL;
 
        /*