Bug 159694

Summary:	PostgreSQL 7.3 doesn't have spinlock support on x86_64
Product:	Red Hat Enterprise Linux 3	Reporter:	Tom Lane <tgl>
Component:	rh-postgresql	Assignee:	Tom Lane <tgl>
Status:	CLOSED ERRATA	QA Contact:
Severity:	medium	Docs Contact:
Priority:	medium
Version:	3.0	CC:	hhorak, thomas
Target Milestone:	---
Target Release:	---
Hardware:	x86_64
OS:	Linux
URL:	http://archives.postgresql.org/pgsql-performance/2005-06/msg00125.php
Whiteboard:
Fixed In Version:	RHBA-2005-684	Doc Type:	Bug Fix
Doc Text:		Story Points:	---
Clone Of:		Environment:
Last Closed:	2005-09-28 16:30:16 UTC	Type:	---
Regression:	---	Mount Type:	---
Documentation:	---	CRM:
Verified Versions:		Category:	---
oVirt Team:	---	RHEL 7.3 requirements from Atomic Host:
Cloudforms Team:	---	Target Upstream Version:
Embargoed:
Bug Depends On:
Bug Blocks:	156320

Description Tom Lane 2005-06-07 04:14:33 UTC

Description of problem:
For lack of any built-in assembly code support for spinlocks on x86_64, our RHEL3 PostgreSQL build is 
falling back to using SysV IPC semaphore kernel calls for operations that ought to take a couple of 
machine instructions.  This is a pretty horrid performance loss, and can even result in database startup 
failure if the number of semaphores needed exceeds what the kernel is configured to allow (see 
referenced URL for an example)

Version-Release number of selected component (if applicable):
rh-postgresql-7.3.10-1

How reproducible:
100%

Additional info:
Need to backport assembly-code spinlock support from later Postgres versions.  Also check to see if 
any other platforms supported by RHEL3 have similar problems.

Comment 3 Tom Lane 2005-07-20 16:25:52 UTC

It turns out that ppc64 spinlock support is missing too.

On investigation, the best bet seems to be to update s_lock.h and s_lock.c to
exactly match the PostgreSQL 7.4 sources; we know that code works from both
community testing and our own usage in RHEL4.  We could fix the immediate
problems with a slightly smaller patch, but a partial backport seems to me more
likely to create new issues than avoid them.

Comment 5 Tom Lane 2005-07-28 16:40:32 UTC

My previous comment that ppc64 spinlock support is needed was incorrect; there
was already a patch in the RPMs that added ppc64 support.  The ppc64 code in 7.4
is marginally more efficient, though.

Comment 6 Red Hat Bugzilla 2005-09-28 16:30:17 UTC

An advisory has been issued which should help the problem
described in this bug report. This report is therefore being
closed with a resolution of ERRATA. For more information
on the solution and/or where to find the updated files,
please follow the link below. You may reopen this bug report
if the solution does not work for you.

http://rhn.redhat.com/errata/RHBA-2005-684.html