On Fri, 06 Jul 2007 10:46:41 +0100 "Richard W.M. Jones" wrote:
>> Did you find out what caused this in the end?
>
> I have no progress about this.
> I suspect that the cause is the type of cpumap, and I want to try to
> send the cpumap as list but I have no idea the way.
> So I can still not determine the reason.
I'm not sure how to reproduce this. How can I send the map over xend
instead of the HV?
I did fix a problem with virDomainPinVcpu & remote about two weeks ago.
It is rare case. This problem is occurred when the HV is failed.
So, to reproduce this, I changed the domainPinVcpu of HV to NULL.
---
625 #if 0
626 xenHypervisorPinVcpu, /* domainPinVcpu */
627 #else
628 NULL,
629 #endif
and executed it.
# ./virsh vcpupin 25 0 0
# virsh vcpuinfo 25
VCPU: 0
CPU: 1
State: blocked
CPU time: 261,1s
CPU Affinity: yy <---
VCPU: 1
CPU: 1
State: blocked
CPU time: 240,0s
CPU Affinity: yy
VCPU: 2
CPU: 1
State: blocked
CPU time: 245,2s
CPU Affinity: yy
Of course, the HV works fine.
# virsh vcpupin 25 0 0
# virsh vcpuinfo 25
VCPU: 0
CPU: 0
State: blocked
CPU time: 261,2s
CPU Affinity: y- <---
VCPU: 1
CPU: 0
State: blocked
CPU time: 240,0s
CPU Affinity: yy
VCPU: 2
CPU: 1
State: blocked
CPU time: 245,2s
CPU Affinity: yy
Thanks,
Saori