Re: [PATCH] libxl: correct xl cpupool-numa-split with vcpu limited dom0

[Juergen Gross <jgross@suse.com>]

On 14/06/16 12:07, Ian Jackson wrote:
> Wei Liu writes ("Re: [Xen-devel] [PATCH] libxl: correct xl cpupool-numa-split with vcpu limited dom0"):
>> On Tue, Jun 14, 2016 at 11:01:50AM +0200, Dario Faggioli wrote:
>>> On Tue, 2016-06-14 at 06:30 +0200, Juergen Gross wrote:
>>>> When trying to use xl cpupool-numa-split and dom0 is limited to less
>>>> vcpus than one numa node the operation will fail.
>>>>
>>>> Correct this by allowing this configuration.
>>>>
>>>> Reported-by: Glenn Enright <glenn@rimuhosting.com>
>>>> Signed-off-by: Juergen Gross <jgross@suse.com>
>>>>
>>> Reviewed-by: Dario Faggioli <dario.faggioli@citrix.com>
>>
>> Acked-by: Wei Liu <wei.liu2@citrix.com>
>>
>> I think this is also a backport candidate?
> 
> Possibly.
> 
> I looked at the code for a minute or two, and perhaps I'm being dense
> this morning, but I wasn't able to see (from the code and the commit
> message and from the diff) precisely what misunderstanding the
> original author of the code had, and how this patch fixes it.

The problem arises if dom0 has less vcpus than a numa node. In this case
libxl_set_vcpuonline() will fail as the cpumap has more bits set than
the number of dom0's vcpus.

My patch will result in a call of libxl_set_vcpuonline() only in case
dom0 has more vcpus online than the number of cpus of node it is to be
restricted to.

> I don't want to backport (non-security) things unless they have a very
> low chance of regressions.  So perhaps someone could either explain it
> to me, or assert convincingly that they are sure it's right :-).


Juergen


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