Skip to content

Commit 4da24f1

Browse files
committed
Fix two bugs in setting the vm bit of empty pages.
Use a critical section when setting the all-visible flag on an empty page, and WAL-logging it. log_newpage_buffer() contains an assertion that it must be called inside a critical section, and it's the right thing to do when modifying a buffer anyway. Also, the page should be marked dirty before calling log_newpage_buffer(), per the comment in log_newpage_buffer() and src/backend/access/transam/README. Patch by Andres Freund, in response to my report. Backpatch to 9.2, like the patch that introduced these bugs (a6370fd).
1 parent b89cede commit 4da24f1

File tree

1 file changed

+6
-1
lines changed

1 file changed

+6
-1
lines changed

src/backend/commands/vacuumlazy.c

Lines changed: 6 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -650,6 +650,11 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
650650
/* empty pages are always all-visible */
651651
if (!PageIsAllVisible(page))
652652
{
653+
START_CRIT_SECTION();
654+
655+
/* mark buffer dirty before writing a WAL record */
656+
MarkBufferDirty(buf);
657+
653658
/*
654659
* It's possible that another backend has extended the heap,
655660
* initialized the page, and then failed to WAL-log the page
@@ -669,9 +674,9 @@ lazy_scan_heap(Relation onerel, LVRelStats *vacrelstats,
669674
log_newpage_buffer(buf);
670675

671676
PageSetAllVisible(page);
672-
MarkBufferDirty(buf);
673677
visibilitymap_set(onerel, blkno, InvalidXLogRecPtr, vmbuffer,
674678
InvalidTransactionId);
679+
END_CRIT_SECTION();
675680
}
676681

677682
UnlockReleaseBuffer(buf);

0 commit comments

Comments
 (0)