| Lists: | pgsql-hackers |
|---|
| From: | Dan Ports <drkp(at)csail(dot)mit(dot)edu> |
|---|---|
| To: | pgsql-hackers(at)postgresql(dot)org |
| Cc: | Kevin Grittner <Kevin(dot)Grittner(at)wicourts(dot)gov> |
| Subject: | patch: fix SSI finished list corruption |
| Date: | 2012-01-07 00:15:25 |
| Message-ID: | 20120107001524.GK11222@csail.mit.edu |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-hackers |
There's a corner case in the SSI cleanup code that isn't handled
correctly. It can arise when running workloads that are comprised
mostly (but not 100%) of READ ONLY transactions, and can corrupt the
finished SERIALIZABLEXACT list, potentially causing a segfault. The
attached patch fixes it.
Specifically, when the only remaining active transactions are READ
ONLY, we do a "partial cleanup" of committed transactions because
certain types of conflicts aren't possible anymore. For committed r/w
transactions, we release the SIREAD locks but keep the
SERIALIZABLEXACT. However, for committed r/o transactions, we can go
further and release the SERIALIZABLEXACT too. The problem was with the
latter case: we were returning the SERIALIZABLEXACT to the free list
without removing it from the finished list.
The only real change in the patch is the SHMQueueDelete line, but I
also reworked some of the surrounding code to make it obvious that r/o
and r/w transactions are handled differently -- the existing code felt
a bit too clever.
Dan
--
Dan R. K. Ports MIT CSAIL http://drkp.net/
| Attachment | Content-Type | Size |
|---|---|---|
| ssi-partial-cleanup.patch | text/x-diff | 0 bytes |
| From: | Heikki Linnakangas <heikki(dot)linnakangas(at)enterprisedb(dot)com> |
|---|---|
| To: | Dan Ports <drkp(at)csail(dot)mit(dot)edu> |
| Cc: | pgsql-hackers(at)postgresql(dot)org, Kevin Grittner <Kevin(dot)Grittner(at)wicourts(dot)gov> |
| Subject: | Re: patch: fix SSI finished list corruption |
| Date: | 2012-01-18 16:01:56 |
| Message-ID: | 4F16ECF4.7070304@enterprisedb.com |
| Views: | Raw Message | Whole Thread | Download mbox | Resend email |
| Lists: | pgsql-hackers |
On 07.01.2012 02:15, Dan Ports wrote:
> There's a corner case in the SSI cleanup code that isn't handled
> correctly. It can arise when running workloads that are comprised
> mostly (but not 100%) of READ ONLY transactions, and can corrupt the
> finished SERIALIZABLEXACT list, potentially causing a segfault. The
> attached patch fixes it.
>
> Specifically, when the only remaining active transactions are READ
> ONLY, we do a "partial cleanup" of committed transactions because
> certain types of conflicts aren't possible anymore. For committed r/w
> transactions, we release the SIREAD locks but keep the
> SERIALIZABLEXACT. However, for committed r/o transactions, we can go
> further and release the SERIALIZABLEXACT too. The problem was with the
> latter case: we were returning the SERIALIZABLEXACT to the free list
> without removing it from the finished list.
>
> The only real change in the patch is the SHMQueueDelete line, but I
> also reworked some of the surrounding code to make it obvious that r/o
> and r/w transactions are handled differently -- the existing code felt
> a bit too clever.
Thanks, committed!
--
Heikki Linnakangas
EnterpriseDB http://www.enterprisedb.com