Fix oversight in coding of _bt_start_vacuum: we can't assume that the LWLock

will be released by transaction abort before _bt_end_vacuum gets called.
If either of these "can't happen" errors actually happened, we'd freeze up
trying to acquire an already-held lock.  Latest word is that this does
not explain Martin Pitt's trouble report, but it still looks like a bug.

diff --git a/src/backend/access/nbtree/nbtutils.c b/src/backend/access/nbtree/nbtutils.c
index d453a93cafa61c2fc4bb259f81ab1220f83fb0a9..9c227d7f6c656650503b77b0fd1767ea3144633c 100644 (file)
--- a/src/backend/access/nbtree/nbtutils.c
+++ b/src/backend/access/nbtree/nbtutils.c
@@ -8,7 +8,7 @@
  *
  *
  * IDENTIFICATION
- *   $PostgreSQL: pgsql/src/backend/access/nbtree/nbtutils.c,v 1.82 2007/01/09 02:14:10 tgl Exp $
+ *   $PostgreSQL: pgsql/src/backend/access/nbtree/nbtutils.c,v 1.83 2007/03/30 00:12:59 tgl Exp $
  *
  *-------------------------------------------------------------------------
  */
@@ -1214,13 +1214,25 @@ _bt_start_vacuum(Relation rel)
        vac = &btvacinfo->vacuums[i];
        if (vac->relid.relId == rel->rd_lockInfo.lockRelId.relId &&
            vac->relid.dbId == rel->rd_lockInfo.lockRelId.dbId)
+       {
+           /*
+            * Unlike most places in the backend, we have to explicitly
+            * release our LWLock before throwing an error.  This is because
+            * we expect _bt_end_vacuum() to be called before transaction
+            * abort cleanup can run to release LWLocks.
+            */
+           LWLockRelease(BtreeVacuumLock);
            elog(ERROR, "multiple active vacuums for index \"%s\"",
                 RelationGetRelationName(rel));
+       }
    }
 
    /* OK, add an entry */
    if (btvacinfo->num_vacuums >= btvacinfo->max_vacuums)
+   {
+       LWLockRelease(BtreeVacuumLock);
        elog(ERROR, "out of btvacinfo slots");
+   }
    vac = &btvacinfo->vacuums[btvacinfo->num_vacuums];
    vac->relid = rel->rd_lockInfo.lockRelId;
    vac->cycleid = result;