Re: SSI can miss conflicts between index-only scans and heap writes

From: Rui Zhao <zhaorui126(at)gmail(dot)com>
To: pgsql-hackers <pgsql-hackers(at)lists(dot)postgresql(dot)org>, Peter Geoghegan <pg(at)bowt(dot)ie>
Subject: Re: SSI can miss conflicts between index-only scans and heap writes
Date: 2026-11-21 16:42:47
Message-ID: CAHWVJhFsagJw-H5ZXv_CJ=-fVTJA=tuoqQZz6rWuZbPQ_J7z2g@mail.gmail.com
Views: Whole Thread | Raw Message | Download mbox | Resend email
Thread:
Lists: pgsql-hackers

Hi Peter,

I'm copying you because I found this while checking when an index-only
scan takes the heap page predicate lock in your index prefetching patch.
The race also occurs on master without prefetching. Would you have a
look at the ordering in v4? At SERIALIZABLE, the scan rechecks the
all-visible bit after PredicateLockPage(). On the write side,
heap_delete() and heap_update() check for conflicts again after clearing
the bit, before they unlock the heap buffer.

Regards,
Rui

Attachment Content-Type Size
v4-0001-Fix-SSI-conflicts-between-index-only-scans-and-heap-writes.patch application/octet-stream 16.0 KB

In response to

Browse pgsql-hackers by date

  From Date Subject
Next Message Rui Zhao 2026-11-21 17:04:26 Fwd: Fix containment and B-tree ordering for cubes of different dimensions
Previous Message Andrey Borodin 2026-11-21 15:26:08 Re: Allow isolation test steps to be declared inline.