[vfio-users] Occasional freeze when 3 VMs with passthrough are running

[Nick Sukharev]

I am running a server with several virtual machines utilizing GPU
passthrough. I've been running it for a few months and recently started
getting occasional freezes. The host freezes after running for a few hours
or sometimes for 1-2 days. It doesn't seem to be related to VM reboots. I
am running stock Ubuntu 15.04 and radeon cards. I was able to capture the
stack using netconsole:
http://pastebin.com/85kPyj4i
A few days ago it was just priting messages such as this one without any
stack:
[23521.248008] NMI watchdog: BUG: soft lockup - CPU#7 stuck for 22s!
[qemu-system-x86:2492]
for CPUS 7,8,10 (I have Intel Core i7-5820K with 6 cores).
Would it be possible to figure something out based on this stack? It seems
to lead to the actual place where watchdog does its diagnostics but I see a
few kvm/vmx functions on the stack.

Another question: is kernel 4.2 broken with OVMF? I thought about upgrading
to 15.10 to see if the issue is still there.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/vfio-users/attachments/20151207/25fed79c/attachment.htm>