Red Hat Bugzilla – Bug 433872
softlockups at 1024p boot sequence
Last modified: 2011-04-25 15:13:24 EDT
Description of problem:
The point of this BZ is just the softlockups. I have included
the boot sequence with the threshold at the default 10secs.
Version-Release number of selected component (if applicable):
I always get at least 2, where, when, and other than cpu#0
which processor varies some.
Steps to Reproduce:
1. Simple boot of 5.1 with the 2.6.18-80 kernel.
Created attachment 295558 [details]
Is the softlockup related to staring autofs?
If the problem is harmless and un-avoidable due to the scale of the system,
should we suppress the message?
Also I'd like to know if upstream behaves different?
(In reply to comment #2)
> Is the softlockup related to staring autofs?
> If the problem is harmless and un-avoidable due to the scale of the system,
> should we suppress the message?
No -- we shouldn't. The issue is real and if it is a scaling issue it should be
resolved upstream and then backported into RHEL.
> Also I'd like to know if upstream behaves different?
1. I don't think it was specifically related to autofs; it was semi-random
after init started except that there usually was one just before the
2. It is certainly related to the scale of the system --- I cut the number
of cpus to 512 still w/ 4Tb of memory and the problem disappeared. NOTE
however that any area of the kernel that must be shared with all cpus
would have been on a node associated with one of the 512 CPUs and not
on the other 4 racks which would have had longer latency's.
3. The machine got shipped before I could get to this, grumble grumble.
Lack of customer interest.