Fix Portal snapshot tracking to handle subtransactions properly.

Commit 84f5c2908 forgot to consider the possibility that
EnsurePortalSnapshotExists could run inside a subtransaction with
lifespan shorter than the Portal's.  In that case, the new active
snapshot would be popped at the end of the subtransaction, leaving
a dangling pointer in the Portal, with mayhem ensuing.

To fix, make sure the ActiveSnapshot stack entry is marked with
the same subtransaction nesting level as the associated Portal.
It's certainly safe to do so since we won't be here at all unless
the stack is empty; hence we can't create an out-of-order stack.

Let's also apply this logic in the case where PortalRunUtility
sets portalSnapshot, just to be sure that path can't cause similar
problems.  It's slightly less clear that that path can't create
an out-of-order stack, so add an assertion guarding it.

Report and patch by Bertrand Drouvot (with kibitzing by me).
Back-patch to v11, like the previous commit.

Discussion: https://postgr.es/m/ff82b8c5-77f4-3fe7-6028-fcf3303e82dd@amazon.com

diff --git a/src/backend/access/transam/xact.c b/src/backend/access/transam/xact.c
index be8ed0b2491551dd5033ed2bc89077a4116fef12..ad465d8eae2e101dc1b60dde5127d1d2b6b87843 100644 (file)
--- a/src/backend/access/transam/xact.c
+++ b/src/backend/access/transam/xact.c
@@ -4817,6 +4817,7 @@ CommitSubTransaction(void)
    AfterTriggerEndSubXact(true);
    AtSubCommit_Portals(s->subTransactionId,
                        s->parent->subTransactionId,
+                       s->parent->nestingLevel,
                        s->parent->curTransactionOwner);
    AtEOSubXact_LargeObject(true, s->subTransactionId,
                            s->parent->subTransactionId);

diff --git a/src/backend/tcop/pquery.c b/src/backend/tcop/pquery.c
index 0e0db5000b03ebff31e7335f643e366353e0d374..7b24098f370175edb0bf7ef6a068d37c9454c4aa 100644 (file)
--- a/src/backend/tcop/pquery.c
+++ b/src/backend/tcop/pquery.c
@@ -493,7 +493,9 @@ PortalStart(Portal portal, ParamListInfo params,
                 * We could remember the snapshot in portal->portalSnapshot,
                 * but presently there seems no need to, as this code path
                 * cannot be used for non-atomic execution.  Hence there can't
-                * be any commit/abort that might destroy the snapshot.
+                * be any commit/abort that might destroy the snapshot.  Since
+                * we don't do that, there's also no need to force a
+                * non-default nesting level for the snapshot.
                 */
 
                /*
@@ -1152,9 +1154,15 @@ PortalRunUtility(Portal portal, PlannedStmt *pstmt,
            snapshot = RegisterSnapshot(snapshot);
            portal->holdSnapshot = snapshot;
        }
-       /* In any case, make the snapshot active and remember it in portal */
-       PushActiveSnapshot(snapshot);
-       /* PushActiveSnapshot might have copied the snapshot */
+
+       /*
+        * In any case, make the snapshot active and remember it in portal.
+        * Because the portal now references the snapshot, we must tell
+        * snapmgr.c that the snapshot belongs to the portal's transaction
+        * level, else we risk portalSnapshot becoming a dangling pointer.
+        */
+       PushActiveSnapshotWithLevel(snapshot, portal->createLevel);
+       /* PushActiveSnapshotWithLevel might have copied the snapshot */
        portal->portalSnapshot = GetActiveSnapshot();
    }
    else
@@ -1810,8 +1818,13 @@ EnsurePortalSnapshotExists(void)
        elog(ERROR, "cannot execute SQL without an outer snapshot or portal");
    Assert(portal->portalSnapshot == NULL);
 
-   /* Create a new snapshot and make it active */
-   PushActiveSnapshot(GetTransactionSnapshot());
-   /* PushActiveSnapshot might have copied the snapshot */
+   /*
+    * Create a new snapshot, make it active, and remember it in portal.
+    * Because the portal now references the snapshot, we must tell snapmgr.c
+    * that the snapshot belongs to the portal's transaction level, else we
+    * risk portalSnapshot becoming a dangling pointer.
+    */
+   PushActiveSnapshotWithLevel(GetTransactionSnapshot(), portal->createLevel);
+   /* PushActiveSnapshotWithLevel might have copied the snapshot */
    portal->portalSnapshot = GetActiveSnapshot();
 }

diff --git a/src/backend/utils/mmgr/portalmem.c b/src/backend/utils/mmgr/portalmem.c
index a5d214d123261957f0f160be34a061fa023f7ac8..b7de468773edc760bb5d6fc556cc4ebe29b2b013 100644 (file)
--- a/src/backend/utils/mmgr/portalmem.c
+++ b/src/backend/utils/mmgr/portalmem.c
@@ -210,6 +210,7 @@ CreatePortal(const char *name, bool allowDup, bool dupSilent)
    portal->cleanup = PortalCleanup;
    portal->createSubid = GetCurrentSubTransactionId();
    portal->activeSubid = portal->createSubid;
+   portal->createLevel = GetCurrentTransactionNestLevel();
    portal->strategy = PORTAL_MULTI_QUERY;
    portal->cursorOptions = CURSOR_OPT_NO_SCROLL;
    portal->atStart = true;
@@ -655,6 +656,7 @@ HoldPortal(Portal portal)
     */
    portal->createSubid = InvalidSubTransactionId;
    portal->activeSubid = InvalidSubTransactionId;
+   portal->createLevel = 0;
 }
 
 /*
@@ -938,6 +940,7 @@ PortalErrorCleanup(void)
 void
 AtSubCommit_Portals(SubTransactionId mySubid,
                    SubTransactionId parentSubid,
+                   int parentLevel,
                    ResourceOwner parentXactOwner)
 {
    HASH_SEQ_STATUS status;
@@ -952,6 +955,7 @@ AtSubCommit_Portals(SubTransactionId mySubid,
        if (portal->createSubid == mySubid)
        {
            portal->createSubid = parentSubid;
+           portal->createLevel = parentLevel;
            if (portal->resowner)
                ResourceOwnerNewParent(portal->resowner, parentXactOwner);
        }

diff --git a/src/backend/utils/time/snapmgr.c b/src/backend/utils/time/snapmgr.c
index d07ca1b0b242f289db940a334cb8e4f22b4d5d74..f9cb2955ee9a02239eeb9cd286f8f6b3aeb42b4f 100644 (file)
--- a/src/backend/utils/time/snapmgr.c
+++ b/src/backend/utils/time/snapmgr.c
@@ -733,10 +733,25 @@ FreeSnapshot(Snapshot snapshot)
  */
 void
 PushActiveSnapshot(Snapshot snap)
+{
+   PushActiveSnapshotWithLevel(snap, GetCurrentTransactionNestLevel());
+}
+
+/*
+ * PushActiveSnapshotWithLevel
+ *     Set the given snapshot as the current active snapshot
+ *
+ * Same as PushActiveSnapshot except that caller can specify the
+ * transaction nesting level that "owns" the snapshot.  This level
+ * must not be deeper than the current top of the snapshot stack.
+ */
+void
+PushActiveSnapshotWithLevel(Snapshot snap, int snap_level)
 {
    ActiveSnapshotElt *newactive;
 
    Assert(snap != InvalidSnapshot);
+   Assert(ActiveSnapshot == NULL || snap_level >= ActiveSnapshot->as_level);
 
    newactive = MemoryContextAlloc(TopTransactionContext, sizeof(ActiveSnapshotElt));
 
@@ -750,7 +765,7 @@ PushActiveSnapshot(Snapshot snap)
        newactive->as_snap = snap;
 
    newactive->as_next = ActiveSnapshot;
-   newactive->as_level = GetCurrentTransactionNestLevel();
+   newactive->as_level = snap_level;
 
    newactive->as_snap->active_count++;
 

diff --git a/src/include/utils/portal.h b/src/include/utils/portal.h
index 364fae4b651bb6017cc2016c1308815fee2faaee..3c30ccbc6f851891c9b63e35138f398920094865 100644 (file)
--- a/src/include/utils/portal.h
+++ b/src/include/utils/portal.h
@@ -193,6 +193,8 @@ typedef struct PortalData
    TimestampTz creation_time;  /* time at which this portal was defined */
    bool        visible;        /* include this portal in pg_cursors? */
 
+   /* Stuff added at the end to avoid ABI break in stable branches: */
+
    /*
     * Outermost ActiveSnapshot for execution of the portal's queries.  For
     * all but a few utility commands, we require such a snapshot to exist.
@@ -200,6 +202,7 @@ typedef struct PortalData
     * and helps to reduce thrashing of the process's exposed xmin.
     */
    Snapshot    portalSnapshot; /* active snapshot, or NULL if none */
+   int         createLevel;    /* creating subxact's nesting level */
 }          PortalData;
 
 /*
@@ -217,6 +220,7 @@ extern void AtCleanup_Portals(void);
 extern void PortalErrorCleanup(void);
 extern void AtSubCommit_Portals(SubTransactionId mySubid,
                                SubTransactionId parentSubid,
+                               int parentLevel,
                                ResourceOwner parentXactOwner);
 extern void AtSubAbort_Portals(SubTransactionId mySubid,
                               SubTransactionId parentSubid,

diff --git a/src/include/utils/snapmgr.h b/src/include/utils/snapmgr.h
index 62351a06bdf4dd9fd518d73c56cce058fd073cec..adb47dcd22c9293f202303abdfebdb882c36edc8 100644 (file)
--- a/src/include/utils/snapmgr.h
+++ b/src/include/utils/snapmgr.h
@@ -111,6 +111,7 @@ extern void InvalidateCatalogSnapshot(void);
 extern void InvalidateCatalogSnapshotConditionally(void);
 
 extern void PushActiveSnapshot(Snapshot snapshot);
+extern void PushActiveSnapshotWithLevel(Snapshot snapshot, int snap_level);
 extern void PushCopiedSnapshot(Snapshot snapshot);
 extern void UpdateActiveSnapshotCommandId(void);
 extern void PopActiveSnapshot(void);

diff --git a/src/pl/plpgsql/src/expected/plpgsql_transaction.out b/src/pl/plpgsql/src/expected/plpgsql_transaction.out
index 4beead7ad5df55507ad7aa8d5dd7081584789108..7302a5fa1d6a61fc8f65a452349255c56de92fcc 100644 (file)
--- a/src/pl/plpgsql/src/expected/plpgsql_transaction.out
+++ b/src/pl/plpgsql/src/expected/plpgsql_transaction.out
@@ -430,6 +430,34 @@ SELECT * FROM test1;
 ---+---
 (0 rows)
 
+-- test commit/rollback inside exception handler, too
+TRUNCATE test1;
+DO LANGUAGE plpgsql $$
+BEGIN
+    FOR i IN 1..10 LOOP
+      BEGIN
+        INSERT INTO test1 VALUES (i, 'good');
+        INSERT INTO test1 VALUES (i/0, 'bad');
+      EXCEPTION
+        WHEN division_by_zero THEN
+            INSERT INTO test1 VALUES (i, 'exception');
+            IF (i % 3) > 0 THEN COMMIT; ELSE ROLLBACK; END IF;
+      END;
+    END LOOP;
+END;
+$$;
+SELECT * FROM test1;
+ a  |     b     
+----+-----------
+  1 | exception
+  2 | exception
+  4 | exception
+  5 | exception
+  7 | exception
+  8 | exception
+ 10 | exception
+(7 rows)
+
 -- detoast result of simple expression after commit
 CREATE TEMP TABLE test4(f1 text);
 ALTER TABLE test4 ALTER COLUMN f1 SET STORAGE EXTERNAL; -- disable compression

diff --git a/src/pl/plpgsql/src/sql/plpgsql_transaction.sql b/src/pl/plpgsql/src/sql/plpgsql_transaction.sql
index a219e725575ea44b254d57b85ae7a87583e8b069..521e7b0f669e2dc679a9ca76a650792f6d0e40a7 100644 (file)
--- a/src/pl/plpgsql/src/sql/plpgsql_transaction.sql
+++ b/src/pl/plpgsql/src/sql/plpgsql_transaction.sql
@@ -354,6 +354,27 @@ $$;
 SELECT * FROM test1;
 
 
+-- test commit/rollback inside exception handler, too
+TRUNCATE test1;
+
+DO LANGUAGE plpgsql $$
+BEGIN
+    FOR i IN 1..10 LOOP
+      BEGIN
+        INSERT INTO test1 VALUES (i, 'good');
+        INSERT INTO test1 VALUES (i/0, 'bad');
+      EXCEPTION
+        WHEN division_by_zero THEN
+            INSERT INTO test1 VALUES (i, 'exception');
+            IF (i % 3) > 0 THEN COMMIT; ELSE ROLLBACK; END IF;
+      END;
+    END LOOP;
+END;
+$$;
+
+SELECT * FROM test1;
+
+
 -- detoast result of simple expression after commit
 CREATE TEMP TABLE test4(f1 text);
 ALTER TABLE test4 ALTER COLUMN f1 SET STORAGE EXTERNAL; -- disable compression