Re: [Xen-devel] [PATCH] SEDF: avoid gathering vCPU-s on pCPU0 [and 1 more messages]

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 05.03.13 at 16:54, Ian Jackson <Ian.Jackson@xxxxxxxxxxxxx> wrote:
> Jan Beulich writes ("Re: [Xen-devel] [PATCH] SEDF: avoid gathering vCPU-s on 
> pCPU0"):
>> So when a vCPU narrows its affinity (in the extreme to a mask
>> with just one bit set) and then widens it again, it would
>> nevertheless remain on the pCPU it was formerly restricted to.
>> When (perhaps much later) a second and then a third vCPU
>> do the same, they may all end up running on the same pCPU,
>> i.e. we'd get back to the problem addressed by the first half of
>> the fix here.
> 
> I don't understand why having every VCPU in the whole system running
> on a single PCPU would result in a complete hang.  Surely unless we
> have many VCPUs all piling on spinlocks there will not be that many
> runnable VCPUs ?
> 
> Or is this symptom due to the lack of PV spinlocks ?

The case that I observed here was all CPUs piling up in
stop_machine logic - that's effectively an open coded
spinning operation. But clearly non-ticket, non-pv spin locks
can result in similar live lock effects (as soon as the locking
rate is higher than what the single CPU can handle in terms
of executing locked region code).

Jan


_______________________________________________
Xen-devel mailing list
Xen-devel@xxxxxxxxxxxxx
http://lists.xen.org/xen-devel