Horton

Re: BUG #17949: Adding an index introduces serialisation anomalies.

Thomas Munro <thomas.munro@gmail.com>

View thread View on postgr.es
From: Thomas Munro <thomas.munro@gmail.com>
To: Dmitry Dolgov <9erthalion6@gmail.com>
Cc: artem.anisimov.255@gmail.com, pgsql-bugs@lists.postgresql.org
Date: 2023-06-19T03:23:31Z
Lists: pgsql-bugs 
On Mon, Jun 19, 2023 at 2:29 PM Thomas Munro <thomas.munro@gmail.com> wrote:
> S2:  _bt_search(&buf)

> S1:  INSERT ...

> S2:    PredicateLockRelation(...);

> My point is that S2 won't ever scan S1's tuples, so it won't pass S1's
> xid to CheckForSerializableConflictOut()

[completing that sentence a little more] ... so this was our only
chance to detect that S2 read an object that S1 wrote.  But the
_bt_search() and PredicateLockXXX() calls are not atomic and not
rechecked, so a write between them is invisible to the algorithm.  I'm
not sure about this, but it's the idea I'm exploring...

Commits

  1. Fix race in SSI interaction with bitmap heap scan.

  2. Fix race in SSI interaction with gin fast path.

  3. Fix race in SSI interaction with empty btrees.

  4. Re-think predicate locking on GIN indexes.