[Xen-devel]: pci passthrough xhci host controller freeze: RCU detected CPU stalls

[Sander Eikelenboom <linux@xxxxxxxxxxxxxx>]

Hi Konrad/Jeremy,

I have finally got something on the dom0 console when the machine freezes (by 
enabling CONFIG_RCU_CPU_STALL_DETECTOR in the dom0 kernel config).
When it now freezes, after some time i get "RCU detected CPU stalls", it 
allways seems to be cpu 0 that stalls (although that's probably the vcpu 
instead of pcpu).
I have attached two photos with the complete error message. It doesn't appear 
in serial log, and i can't dump anything in the hypervisor, serial console 
seems to be frozen at that point :-(

The combination of the xhci passthrough together with a few iterations of a 
"make -j6" kernel compile seems to freeze it straight away, when using cpu 
pinning to set the domU with passthrough not to run on pcpu 0 it seems to last 
longer but also finally freezes.
So perhaps it's due to a lot of (msi)interrupts together with a lot of load 
that makes it freeze, an the freezes before where also due to some load 
generated by something ..

Versions:
- Xen-unstable xen_changeset Tue Sep 14 18:26:10 2010 +0100 22155:5393151a737b
- Dom0 kernel: xen/next-2.6.32 commit 0fe3a983f97ca2e1650c5d29f6a28a18222bd814
- DomU kernels:  a mix from 2.6.26 debian kernels, using konrad's pci-front-0.7 
tree for the xhci passthrough domain.

Any suggestions how to get more info around the point it freezes ?

--
Sander

Attached:
     - 2 photo's with the error.

SAM_0476.JPG
Description: JPEG image

SAM_0477.JPG
Description: JPEG image

_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxxxxxxxx
http://lists.xensource.com/xen-devel