From: Qais Yousef <email@example.com>
To: Pavan Kondeti <firstname.lastname@example.org>
Cc: Ingo Molnar <email@example.com>,
Peter Zijlstra <firstname.lastname@example.org>,
Steven Rostedt <email@example.com>,
Dietmar Eggemann <firstname.lastname@example.org>,
Juri Lelli <email@example.com>,
Vincent Guittot <firstname.lastname@example.org>,
Ben Segall <email@example.com>, Mel Gorman <firstname.lastname@example.org>,
Subject: Re: [PATCH v2 5/6] sched/rt: Better manage pushing unfit tasks on wakeup
Date: Thu, 27 Feb 2020 10:29:58 +0000 [thread overview]
Message-ID: <email@example.com> (raw)
On 02/27/20 09:06, Pavan Kondeti wrote:
> On Wed, Feb 26, 2020 at 04:02:48PM +0000, Qais Yousef wrote:
> > On 02/25/20 09:25, Pavan Kondeti wrote:
> > > > I haven't been staring at this code for as long as you, but since we have
> > > > logic at wakeup to do a push, I think we need something here anyway for unfit
> > > > tasks.
> > > >
> > > > Fixing select_task_rq_rt() to better balance tasks will help a lot in general,
> > > > but if that was enough already then why do we need to consider a push at the
> > > > wakeup at all then?
> > > >
> > > > AFAIU, in SMP the whole push-pull mechanism is racy and we introduce redundancy
> > > > at taking the decision on various points to ensure we minimize this racy nature
> > > > of SMP systems. Anything could have happened between the time we called
> > > > select_task_rq_rt() and the wakeup, so we double check again before we finally
> > > > go and run. That's how I interpret it.
> > > >
> > > > I am open to hear about other alternatives first anyway. Your help has been
> > > > much appreciated so far.
> > > >
> > >
> > > The search inside find_lowest_rq() happens without any locks so I believe it
> > > is expected to have races like this. In fact there is a comment in the code
> > > saying "This test is optimistic, if we get it wrong the load-balancer
> > > will have to sort it out" in select_task_rq_rt(). However, the push logic
> > > as of today works only for overloaded case. In that sense, your patch fixes
> > > this race for b.L systems. At the same time, I feel like tracking nonfit tasks
> > > just to fix this race seems to be too much. I will leave this to Steve and
> > > others to take a decision.
> > I do think without this tasks can end up on the wrong CPU longer than they
> > should. Keep in mind that if a task is boosted to run on a big core, it still
> > have to compete with non-boosted tasks who can run on a any cpu. So this
> > opportunistic push might be necessary.
> > For 5.6 though, I'll send an updated series that removes the fitness check from
> > task_woken_rt() && switched_to_rt() and carry on with this discussion for 5.7.
> > >
> > > I thought of suggesting to remove the below check from select_task_rq_rt()
> > >
> > > p->prio < cpu_rq(target)->rt.highest_prio.curr
> > >
> > > which would then make the target CPU overloaded and the push logic would
> > > spread the tasks. That works for a b.L system too. However there seems to
> > > be a very good reason for doing this. see
> > > https://lore.kernel.org/patchwork/patch/539137/
> > >
> > > The fact that a CPU is part of lowest_mask but running a higher prio RT
> > > task means there is a race. Should we retry one more time to see if we find
> > > another CPU?
> > Isn't this what I did in v1?
> > https://firstname.lastname@example.org/
> Yes, that patch allows overloading the CPU When the priorities are same.
So I assume you're okay with this patch now?
> I think, We should also consider when a low prio task and high prio task
> are waking at the same time and high prio task winning the race.
You mean the bug I describe here?
That needs more serious thinking.
next prev parent reply other threads:[~2020-02-27 10:30 UTC|newest]
Thread overview: 21+ messages / expand[flat|nested] mbox.gz Atom feed top
2020-02-23 18:39 [PATCH v2 0/6] RT Capacity Awareness Fixes & Improvements Qais Yousef
2020-02-23 18:39 ` [PATCH v2 1/6] sched/rt: cpupri_find: implement fallback mechanism for !fit case Qais Yousef
2020-02-23 18:39 ` [PATCH v2 2/6] sched/rt: Re-instate old behavior in select_task_rq_rt Qais Yousef
2020-02-25 15:21 ` Dietmar Eggemann
2020-02-26 11:34 ` Qais Yousef
2020-02-23 18:39 ` [PATCH v2 3/6] sched/rt: Optimize cpupri_find on non-heterogenous systems Qais Yousef
2020-02-23 18:39 ` [PATCH v2 4/6] sched/rt: allow pulling unfitting task Qais Yousef
2020-02-23 18:40 ` [PATCH v2 5/6] sched/rt: Better manage pushing unfit tasks on wakeup Qais Yousef
2020-02-24 6:10 ` Pavan Kondeti
2020-02-24 12:11 ` Qais Yousef
2020-02-24 16:04 ` Pavan Kondeti
2020-02-24 17:41 ` Qais Yousef
2020-02-25 3:55 ` Pavan Kondeti
2020-02-26 16:02 ` Qais Yousef
2020-02-27 3:36 ` Pavan Kondeti
2020-02-27 10:29 ` Qais Yousef [this message]
2020-02-23 18:40 ` [PATCH v2 6/6] sched/rt: Remove unnecessary assignment in inc/dec_rt_migration Qais Yousef
2020-02-23 23:16 ` Dietmar Eggemann
2020-02-24 12:31 ` Qais Yousef
2020-02-24 13:03 ` Dietmar Eggemann
2020-02-24 13:47 ` Qais Yousef
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).