On Tue, 2016-02-02 at 11:35 +0100, Juergen Gross wrote:
> On 02/02/16 10:53, Dario Faggioli wrote:
> >
> > In any case, since the lack of locking and lack of recalculation
> > look
> > like two pretty independent existing bugs to me, can we have
> > either:
> > a. two patches;
> > b. one patch b
On 02/02/16 10:53, Dario Faggioli wrote:
> On Fri, 2016-01-29 at 11:59 +0100, Juergen Gross wrote:
>> On 29/01/16 11:46, Jan Beulich wrote:
>> On 29.01.16 at 11:21, wrote:
--- a/xen/common/sched_credit.c
+++ b/xen/common/sched_credit.c
@@ -1086,12 +1086,19 @@ csched_dom_cntl(
>>
On Fri, 2016-01-29 at 11:59 +0100, Juergen Gross wrote:
> On 29/01/16 11:46, Jan Beulich wrote:
> > > > > On 29.01.16 at 11:21, wrote:
> > > --- a/xen/common/sched_credit.c
> > > +++ b/xen/common/sched_credit.c
> > > @@ -1086,12 +1086,19 @@ csched_dom_cntl(
> > > static inline void
> > > __csche
On Fri, Jan 29, 2016 at 11:21:48AM +0100, Juergen Gross wrote:
> When modifying the timeslice of the credit scheduler in a cpupool the
> cpupool global credit value (n_cpus * credits_per_tslice) isn't
> recalculated. This will lead to wrong scheduling decisions later.
>
> Do the recalculation when
On 29/01/16 11:46, Jan Beulich wrote:
On 29.01.16 at 11:21, wrote:
>> --- a/xen/common/sched_credit.c
>> +++ b/xen/common/sched_credit.c
>> @@ -1086,12 +1086,19 @@ csched_dom_cntl(
>> static inline void
>> __csched_set_tslice(struct csched_private *prv, unsigned timeslice)
>> {
>> +uns
>>> On 29.01.16 at 11:21, wrote:
> --- a/xen/common/sched_credit.c
> +++ b/xen/common/sched_credit.c
> @@ -1086,12 +1086,19 @@ csched_dom_cntl(
> static inline void
> __csched_set_tslice(struct csched_private *prv, unsigned timeslice)
> {
> +unsigned long flags;
> +
> +spin_lock_irqsave
When modifying the timeslice of the credit scheduler in a cpupool the
cpupool global credit value (n_cpus * credits_per_tslice) isn't
recalculated. This will lead to wrong scheduling decisions later.
Do the recalculation when updating the timeslice.
Signed-off-by: Juergen Gross
---
xen/common/s
