Horton

Re: Race condition in TransactionIdIsInProgress

Heikki Linnakangas <hlinnaka@iki.fi>

View thread View on postgr.es
From: Heikki Linnakangas <hlinnaka@iki.fi>
To: Andres Freund <andres@anarazel.de>
Cc: Robert Haas <robertmhaas@gmail.com>, Tom Lane <tgl@sss.pgh.pa.us>, Simon Riggs <simon.riggs@enterprisedb.com>, Konstantin Knizhnik <knizhnik@garret.ru>, PostgreSQL Hackers <pgsql-hackers@postgresql.org>
Date: 2022-06-25T09:18:51Z
Lists: pgsql-hackers

Attachments

On 24/06/2022 04:43, Andres Freund wrote:
> On 2022-06-23 22:03:27 +0300, Heikki Linnakangas wrote:
>> In summary, I think we should:
>> - commit and backpatch Simon's
>> just_remove_TransactionIdIsKnownCompleted_call.v1.patch
>> - fix pg_xact_status() to check TransactionIdIsInProgress()
>> - in v15, remove TransationIdIsKnownCompleted function altogether
>>
>> I'll try to get around to that in the next few days, unless someone beats me
>> to it.
> 
> Makes sense.

This is what I came up with for master. One difference from Simon's 
original patch is that I decided to not expose the 
TransactionIdIsKnownNotInProgress(), as there are no other callers of it 
in core, and it doesn't seem useful to extensions. I inlined it into the 
caller instead.

BTW, should we worry about XID wraparound with the cache? Could you have 
a backend sit idle for 2^32 transactions, without making any 
TransactionIdIsKnownNotInProgress() calls? That's not new with this 
patch, though, it could happen with the single-item cache in transam.c, too.

- Heikki

Commits

  1. Fix visibility check when XID is committed in CLOG but not in procarray.

  2. Fix possible HOT corruption when RECENTLY_DEAD changes to DEAD while pruning.

  3. Make TransactionIdIsInProgress check transam.c's single-item XID status cache