Re: [Xen-devel] [PATCH 2/3] xen: Have schedulers revise initial placement

["Jan Beulich" <JBeulich@xxxxxxxx>]

>>> On 01.08.16 at 14:32, <dario.faggioli@xxxxxxxxxx> wrote:
> On Mon, 2016-08-01 at 04:40 -0600, Jan Beulich wrote:
>> > > > On 15.07.16 at 20:02, <george.dunlap@xxxxxxxxxx> wrote:
>> > 
>> > To solve this, when inserting a vcpu, always call the per-scheduler
>> > "pick" function to revise the initial placement.  This will
>> > automatically take all knowledge the scheduler has into account.
>> > 
>> > Signed-off-by: George Dunlap <george.dunlap@xxxxxxxxxx>
>>
>> Should this and patch 3 be backported?
>> 
> Yes, I think they're good backporting candidates.

Well, they appear to work fine on 4.7, but putting them onto 4.5
causes an early boot crash (BUG_ON( cpu != svc->vcpu->processor )
in __runq_insert()). Pulling in e59321d154 ("credit: remove cpu
argument to __runq_insert()") obviously makes that crash go
away, just to, a little later, hit the similar one close to the top of
csched_load_balance().

I'd really like to have those backported, but I have to ask one
of you to identify which prereq-s are needed on 4.6 and 4.5
(I'll revert them from 4.5 right away, but I'll wait for an osstest
flight to confirm the same issue exists on 4.6). On 4.5 I'd then
additionally depend on you to indicate whether sedf needs
some additional adjustment.

Jan


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