This is a quick/temporary fix for a glibc issue which currently exists when running a CentOS6/RHEL6 Xen PV Guest on Xen 4.1.x (Ours was 4.1.2). It will hopefully be fixed officially in CentOS/RHEL 6.3.

Background:
I moved a customers VPS to a different node yesterday so we could increase the size of their VPS. Once it was moved, the PHP binary wouldn't execute at all and we had errors like this on the console:


php[3300] trap invalid opcode ip:7f9018688be0 sp:7fff950449e8 error:0 in ld-2.12.so[7f9018674000+20000]
Reason:

This is related to the AVC extension in newer CPU's such as the Xeon E3 series.

Fix for 64Bit Only (Disclaimer: At your risk, but works for me.):

I took the latest .srpm, added in a Patch (Source Here (http://www.mail-archive.com/debian-glibc@lists.debian.org/msg46859.html) - Thanks Guys) and created a new glibc RPM. Once installed, this fixed the issue.

Do the following:

wget http://smartservermanagement.com/files/glibc-2.12-1.47.el6.x86_64.rpm
rpm -ivh glibc-2.12-1.47.el6.x86_64.rpm --force
Then make sure you exclude glibc in /etc/yum.conf until CentOS 6.3 is out
Chris

Just got hit by this on a XenServer 6.0 E3 based hypervisor with CentOS 6.2. Cpanel installed then all hell broke loose with invalid opcodes, apache dead, all kinds of mess. Reverted back to CentOS 5.x and all is good.

Apparently it's still not fixed...

It should be in CentOS 6.3, in the meantime its not difficult to create a RPM with the Patch linked above.

yeah i came across this too, i tried building the linked patch into the latest SRPM's from centos but either i did it wrong or theres something else at play as well now.

anyone got any reasonably fool proof instructions for building SRPM's and adding patches?

just to update this, i dont think this is specifically a glibc issue as i succesfully applied the patch this time and still getting similar errors although with cpanel's perl instead of php

looks more like a xen issue with the AVX/AVC extension to us here with centos 6.2

thank you bro,, keep it up