On Fri, 2015-10-23 at 15:42 +0200, Juergen Gross wrote: > On 10/21/2015 07:10 PM, Dario Faggioli wrote:
> > Also, all the operations done in schedule_cpu_switch() itself, > > depend > > either on per_cpu(scheduler) or on per_cpu(schedule_data) being > > updated > > properly, rather than on per_cpu(cpupool) (including the locking > > that > > you are mentioning above). > > > > What am I missing? > > Hmm, good question. I'm rather sure I had a problem related to > exactly > this topic in the early days of cpupools. Maybe the critical code has > been modified since then. Or my memory is wrong. > From a quick archeological investigation, some things certainly have changed. Still, I can't spot anything directly related to this, but it's quite possible that it's there and I'm missing it. > Or we both don't see > it now. ;-) > Yep! :-) > In case there is a problem it should show up doing a test which > concurrently does all of the following: > > - move a domain between two cpupools > - move a cpu between the two cpupools > - create and destroy a domain in one of the two cpupools > Ok, I'll arrange for this and report. > If the system is surviving this test for a couple of hours you are > fine > and then for the attached patch: > > Acked-by: Juergen Gross <jgr...@suse.com> > Thanks :-) Dario -- <<This happens because I choose it to happen!>> (Raistlin Majere) ----------------------------------------------------------------- Dario Faggioli, Ph.D, http://about.me/dario.faggioli Senior Software Engineer, Citrix Systems R&D Ltd., Cambridge (UK)
signature.asc
Description: This is a digitally signed message part
_______________________________________________ Xen-devel mailing list Xen-devel@lists.xen.org http://lists.xen.org/xen-devel
