Kernel divide by zero panic in sched.c:update_sd_lb_stats

[Brandon Black]

host was an m2.2xlarge instance in us-east-1 (with EBS root and data volumes) running stock lucid and serving primarily as a MySQL server.

ProblemType: Bug
DistroRelease: Ubuntu 10.04
Package: linux-image-2.6.32-308-ec2 2.6.32-308.16
ProcVersionSignature: Ubuntu 2.6.32-308.16-ec2 2.6.32.15+drm33.5
Uname: Linux 2.6.32-308-ec2 x86_64
Architecture: amd64
Date: Sun Oct 31 09:58:32 2010
ProcEnviron:
 PATH=(custom, no user)
 LANG=en_US.UTF-8
 SHELL=/bin/bash
SourcePackage: linux-ec2

[Brandon Black (blblack)]

The part we crashed in is the code that calculates cpu scheduler stats, with a divide by zero. I've done some searching around on git.kernel.org and lxr.linux.no, and I don't see where any similar existing problem has been caught and fixed yet upstream through 2.6.32, but it could be non-obvious. Looking at linux_2.6.32-25.45.diff.gz from the relevant apt source for -virtual, I see a lot of patches to sched.c, some of which introduce new division operations. It wouldn't surprise me if the bug was in there somewhere.

[Brandon Black (blblack)]

This looks very likely to be related upstream: https://bugzilla.kernel.org/show_bug.cgi?id=16991

[Brandon Black (blblack)]

Ah, and from there I see a link to a different LP bug which has more details as well, should probably close this as a dupe of that one: https://bugs.launchpad.net/ubuntu/+source/linux-ec2/+bug/614853