linux-kernel - Re: [patch 04/15] sched: throttle cfs_rq entities which exceed their local quota

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <BANLkTikBqOitKTKHq_FKcqvxkypU1g+zAw@mail.gmail.com>
Date:	Tue, 5 Apr 2011 16:15:27 -0700
From:	Paul Turner <pjt@...gle.com>
To:	Peter Zijlstra <a.p.zijlstra@...llo.nl>
Cc:	linux-kernel@...r.kernel.org,
	Bharata B Rao <bharata@...ux.vnet.ibm.com>,
	Dhaval Giani <dhaval.giani@...il.com>,
	Balbir Singh <balbir@...ux.vnet.ibm.com>,
	Vaidyanathan Srinivasan <svaidy@...ux.vnet.ibm.com>,
	Srivatsa Vaddagiri <vatsa@...ibm.com>,
	Kamalesh Babulal <kamalesh@...ux.vnet.ibm.com>,
	Ingo Molnar <mingo@...e.hu>,
	Pavel Emelyanov <xemul@...nvz.org>,
	Nikhil Rao <ncrao@...gle.com>
Subject: Re: [patch 04/15] sched: throttle cfs_rq entities which exceed their
 local quota

On Tue, Apr 5, 2011 at 6:28 AM, Peter Zijlstra <a.p.zijlstra@...llo.nl> wrote:
>
> On Tue, 2011-03-22 at 20:03 -0700, Paul Turner wrote:
>
> > @@ -1249,6 +1257,9 @@ entity_tick(struct cfs_rq *cfs_rq, struc
> >        */
> >       update_curr(cfs_rq);
> >
> > +     /* check that entity's usage is still within quota (if enabled) */
> > +     check_cfs_rq_quota(cfs_rq);
> > +
> >       /*
> >        * Update share accounting for long-running entities.
> >        */
>
> You already have a hook in update_curr() to account quota, why not also
> use that to trigger the reschedule? request_cfs_rq_quota() already has
> the information we failed to replenish the local quota.
>
> Then when you've gotten rid of check_cfs_rq_quota() there isn't a second
> user of within_bandwidth() and you can fold:
>

This actually what it looked like originally, but I broke it apart to
avoid a spurious need_resched coming from the put_path.

Looking again I realize we actually arbitrarily
clear_tsk_need_resched() on prev out of schedule() now so this isn't a
concern.

>
> > @@ -1230,6 +1233,9 @@ static void put_prev_entity(struct cfs_r
> >       if (prev->on_rq)
> >               update_curr(cfs_rq);
> >
> > +     if (!within_bandwidth(cfs_rq))
> > +             throttle_cfs_rq(cfs_rq);
> > +
> >       check_spread(cfs_rq, prev);
> >       if (prev->on_rq) {
> >               update_stats_wait_start(cfs_rq, prev);
>
> Into a single hook.
>
> > @@ -1447,10 +1544,15 @@ static void dequeue_task_fair(struct rq
> >         for_each_sched_entity(se) {
> >                 cfs_rq = cfs_rq_of(se);
> >                 dequeue_entity(cfs_rq, se, flags);
> > -
> > +               /* end evaluation on throttled cfs_rq */
> > +               if (cfs_rq_throttled(cfs_rq)) {
> > +                       se = NULL;
> > +                       break;
> > +               }
> >                 /* Don't dequeue parent if it has other entities besides us */
> >                 if (cfs_rq->load.weight)
> >                         break;
> > +               check_cfs_rq_quota(cfs_rq);
> >                 flags |= DEQUEUE_SLEEP;
> >         }
>
> dequeue_entity() calls update_curr(), so again, by folding
> check_cfs_rq_quota() into your update_curr() hook this becomes simpler.
>

Yes I preferred it as a single hook out of update_curr, will put it
back that way :)

> > +static inline int throttled_hierarchy(struct cfs_rq *cfs_rq)
> > +{
> > +       struct task_group *tg;
> > +       struct sched_entity *se;
> > +
> > +       if (cfs_rq_throttled(cfs_rq))
> > +               return 1;
> > +
> > +       tg = cfs_rq->tg;
> > +       se = tg->se[cpu_of(rq_of(cfs_rq))];
> > +       if (!se)
> > +               return 0;
> > +
> > +       for_each_sched_entity(se) {
> > +               if (cfs_rq_throttled(cfs_rq_of(se)))
> > +                       return 1;
> > +       }
> > +
> > +       return 0;
> > +}
>
> You can actually call for_each_sched_entity() with se==NULL, saves a few lines.

True enough, although this is subsequently subverted by throttle_count
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@...r.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/