From 71d1ace33c77914bcf0ee591f8fc57fd186cb18a Mon Sep 17 00:00:00 2001 From: Heikki Linnakangas Date: Mon, 5 Oct 2026 13:12:33 +0300 Subject: [PATCH] Don't sleep with vacuum_delay_point() while holding locks It's a bad idea to sleep while holding locks, because you might block another process that wants to acquire the same lock. Furthermore, it's not safe to call ProcessConfigFile() in a critical section or while holding locks. So avoid calling vacuum_delay_point() while holding locks, and also make vacuum_delay_point() return quickly without sleeping if it's called while interrupts cannot be processed. To find the vacuum_delay_point() calls that were made while holding locks, I added "Assert(INTERRUPTS_CAN_BE_PROCESSED())" in vacuum_delay_point() and ran the regression tests. I didn't include that Assert in this commit, because there might be more places that do that that are not covered by the regression tests, including extensions. In ginInsertCleanup(), there was already another vacuum_delay_point() call later in the loop, while not holding any locks, so just remove the other call that was made while holding the lock. In hashbucketcleanup(), move the vacuum_delay_point() call up the stack to its caller. The call in hashbucketcleanup() was not able to handle interrupts because it held a lock, and because there were no vacuum_delay_point() or CHECK_FOR_INTERRUPTS() calls in the caller's loop, the whole hash index vacuuming phase was uninterruptible by pending shutdown or query cancel. Now it can be interrupted between buckets. Unfortunately, hashbucketcleanup() has to process all the bucket's pages in one go without pausing; fixing that would require changing how hashbucketcleanup()'s lock chaining works. In acquire_sample_rows(), move the vacuum_delay_point() call in the loop to between pages, to a time where we're not holding the buffer lock. In master, also remove a misleading CHECK_FOR_INTERRUPTS() call from pgstat_write_statsfile(). It was a no-op because we're always holding a lock on the current dshash partition at that point. A call that does nothing is harmless, but it makes you think that the loop is interruptible when in reality it's not. Author: Kevin Rocker Author: Andrey Borodin Author: Mostafa Reported-by: Greg Burd Reported-by: Sergei Kornilov Reviewed-by: Kirill Reshke Reviewed-by: Neil Chen Tested-by: Greg Burd Discussion: https://postgr.es/m/19628-c2b17d358181a1ea@postgresql.org Discussion: https://postgr.es/m/492c6247-43d3-477b-8981-fb0c56767b38%40app.fastmail.com Backpatch-through: 14 --- src/backend/access/gin/ginfast.c | 6 +++--- src/backend/access/hash/hash.c | 5 +++-- src/backend/commands/analyze.c | 3 +-- src/backend/commands/vacuum.c | 8 ++++++++ 4 files changed, 15 insertions(+), 7 deletions(-) diff --git a/src/backend/access/gin/ginfast.c b/src/backend/access/gin/ginfast.c index 853553853a5..55b4c0af0f5 100644 --- a/src/backend/access/gin/ginfast.c +++ b/src/backend/access/gin/ginfast.c @@ -895,8 +895,6 @@ ginInsertCleanup(GinState *ginstate, bool must_empty_list, */ processPendingPage(&accum, &datums, page, FirstOffsetNumber); - vacuum_delay_point(); - /* * Is it time to flush memory to disk? Flush if we are at the end of * the pending list, or if we have a full row and memory is getting @@ -1002,10 +1000,12 @@ ginInsertCleanup(GinState *ginstate, bool must_empty_list, UnlockReleaseBuffer(buffer); } + /* call vacuum_delay_point while not holding any buffer lock */ + vacuum_delay_point(); + /* * Read next page in pending list */ - vacuum_delay_point(); buffer = ReadBuffer(index, blkno); LockBuffer(buffer, GIN_SHARE); page = BufferGetPage(buffer); diff --git a/src/backend/access/hash/hash.c b/src/backend/access/hash/hash.c index 01d06b7c328..14f4da546f7 100644 --- a/src/backend/access/hash/hash.c +++ b/src/backend/access/hash/hash.c @@ -503,6 +503,9 @@ loop_top: Page page; bool split_cleanup = false; + /* call vacuum_delay_point while not holding any buffer lock */ + vacuum_delay_point(); + /* Get address of bucket's start page */ bucket_blkno = BUCKET_TO_BLKNO(cachedmetap, cur_bucket); @@ -716,8 +719,6 @@ hashbucketcleanup(Relation rel, Bucket cur_bucket, Buffer bucket_buf, bool retain_pin = false; bool clear_dead_marking = false; - vacuum_delay_point(); - page = BufferGetPage(buf); opaque = HashPageGetOpaque(page); diff --git a/src/backend/commands/analyze.c b/src/backend/commands/analyze.c index c590a2adc35..9e2d56e7abf 100644 --- a/src/backend/commands/analyze.c +++ b/src/backend/commands/analyze.c @@ -1207,8 +1207,6 @@ acquire_sample_rows(Relation onerel, int elevel, /* Outer loop over blocks to sample */ while (table_scan_analyze_next_block(scan, stream)) { - vacuum_delay_point(); - while (table_scan_analyze_next_tuple(scan, OldestXmin, &liverows, &deadrows, slot)) { /* @@ -1256,6 +1254,7 @@ acquire_sample_rows(Relation onerel, int elevel, pgstat_progress_update_param(PROGRESS_ANALYZE_BLOCKS_DONE, ++blksdone); + vacuum_delay_point(); } read_stream_end(stream); diff --git a/src/backend/commands/vacuum.c b/src/backend/commands/vacuum.c index b6c4216d428..60da5aa859d 100644 --- a/src/backend/commands/vacuum.c +++ b/src/backend/commands/vacuum.c @@ -2391,6 +2391,14 @@ vacuum_delay_point(void) { double msec = 0; + /* + * If we're holding locks or holding interrupts for some other reason, + * don't sleep, because we don't want to hold locks any longer than + * necessary. + */ + if (!INTERRUPTS_CAN_BE_PROCESSED()) + return; + /* Always check for interrupts */ CHECK_FOR_INTERRUPTS();