Bug 712319

Summary: RHEL 6 Xen HVM guest with multiple vcpus crashes on Intel with 96 CPUs and 1T memory
Product: Red Hat Enterprise Linux 6 Reporter: Jinxin Zheng <jzheng>
Component: kernelAssignee: Red Hat Kernel Manager <kernel-mgr>
Status: CLOSED DUPLICATE QA Contact: Virtualization Bugs <virt-bugs>
Severity: high Docs Contact:
Priority: high    
Version: 6.1CC: drjones, leiwang, pcao, qwan
Target Milestone: rc   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: Bug Fix
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2011-06-10 12:37:56 UTC Type: ---
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Attachments:
Description Flags
guest domain config
none
guest console output
none
xend log
none
xm dmesg none

Description Jinxin Zheng 2011-06-10 09:48:56 UTC
Created attachment 504057 [details]
guest domain config

Description of problem:
On an Intel host with 96 CPUs and 1T mem, RHEL 6.1 Xen HVM fails to boot. It crashes a few seconds after it's created, endlessly outputing some like "BUG: soft lockup - CPU#1 stuck for 106s!".

Tested 64 bit and 32 bit guests, both crashed.

Version-Release number of selected component (if applicable):
guest: kernel-2.6.32-131.0.15.el6
host: kernel-xen-2.6.18-267.el5, xen-3.0.3-132.el5
CPU is Intel Xeon E7450, with 1T memory.

How reproducible:
Always

Steps to Reproduce:
1.
On the described host, xm create the HVM domain using the attached config.
2.
3.
  
Actual results:
Guest crash after a while.

Expected results:
Don't crash.

Additional info:
RHEL 6 PV guest does not crash in the same env.

Comment 1 Jinxin Zheng 2011-06-10 09:49:45 UTC
Created attachment 504058 [details]
guest console output

Comment 2 Jinxin Zheng 2011-06-10 09:50:08 UTC
Created attachment 504059 [details]
xend log

Comment 3 Jinxin Zheng 2011-06-10 09:50:27 UTC
Created attachment 504060 [details]
xm dmesg

Comment 6 Jinxin Zheng 2011-06-10 10:11:29 UTC
Oh, I forgot one thing: the HVM guest is given more than 1 vcpus. Single vcpu did not trigger the crash.

Comment 8 Andrew Jones 2011-06-10 11:22:31 UTC
Try clocksource=jiffies on the guest kernel command line.

Comment 9 Jinxin Zheng 2011-06-10 11:45:38 UTC
(In reply to comment #8)
> Try clocksource=jiffies on the guest kernel command line.

That's working!

It doesn't crash after adding clocksource=jiffies.

Comment 10 Andrew Jones 2011-06-10 12:37:56 UTC

*** This bug has been marked as a duplicate of bug 661211 ***