From: peterz@infradead.org (Peter Zijlstra)
To: linux-arm-kernel@lists.infradead.org
Subject: [PATCH v3 01/12] sched: fix imbalance flag reset
Date: Wed, 9 Jul 2014 12:43:32 +0200 [thread overview]
Message-ID: <20140709104332.GS19379@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <53BCBD0E.2070609@linux.vnet.ibm.com>
On Wed, Jul 09, 2014 at 09:24:54AM +0530, Preeti U Murthy wrote:
> In the example that I mention above, t1 and t2 are on the rq of cpu0;
> while t1 is running on cpu0, t2 is on the rq but does not have cpu1 in
> its cpus allowed mask. So during load balance, cpu1 tries to pull t2,
> cannot do so, and hence LBF_ALL_PINNED flag is set and it jumps to
> out_balanced. Note that there are only two sched groups at this level of
> sched domain.one with cpu0 and the other with cpu1. In this scenario we
> do not try to do active load balancing, atleast thats what the code does
> now if LBF_ALL_PINNED flag is set.
I think Vince is right in saying that in this scenario ALL_PINNED won't
be set. move_tasks() will iterate cfs_rq::cfs_tasks, that list will also
include the current running task.
And can_migrate_task() only checks for current after the pinning bits.
> Continuing with the above explanation; when LBF_ALL_PINNED flag is
> set,and we jump to out_balanced, we clear the imbalance flag for the
> sched_group comprising of cpu0 and cpu1,although there is actually an
> imbalance. t2 could still be migrated to say cpu2/cpu3 (t2 has them in
> its cpus allowed mask) in another sched group when load balancing is
> done at the next sched domain level.
And this is where Vince is wrong; note how
update_sg_lb_stats()/sg_imbalance() uses group->sgc->imbalance, but
load_balance() sets: sd_parent->groups->sgc->imbalance, so explicitly
one level up.
So what we can do I suppose is clear 'group->sgc->imbalance' at
out_balanced.
In any case, the entirely of this group imbalance crap is just that,
crap. Its a terribly difficult situation and the current bits more or
less fudge around some of the common cases. Also see the comment near
sg_imbalanced(). Its not a solid and 'correct' anything. Its a bunch of
hacks trying to deal with hard cases.
A 'good' solution would be prohibitively expensive I fear.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: not available
URL: <http://lists.infradead.org/pipermail/linux-arm-kernel/attachments/20140709/28a56c08/attachment.sig>
WARNING: multiple messages have this Message-ID (diff)
From: Peter Zijlstra <peterz@infradead.org>
To: Preeti U Murthy <preeti@linux.vnet.ibm.com>
Cc: Vincent Guittot <vincent.guittot@linaro.org>,
Rik van Riel <riel@redhat.com>, Ingo Molnar <mingo@kernel.org>,
linux-kernel <linux-kernel@vger.kernel.org>,
Russell King - ARM Linux <linux@arm.linux.org.uk>,
LAK <linux-arm-kernel@lists.infradead.org>,
Morten Rasmussen <Morten.Rasmussen@arm.com>,
Mike Galbraith <efault@gmx.de>,
Nicolas Pitre <nicolas.pitre@linaro.org>,
"linaro-kernel@lists.linaro.org" <linaro-kernel@lists.linaro.org>,
Daniel Lezcano <daniel.lezcano@linaro.org>,
Dietmar Eggemann <dietmar.eggemann@arm.com>
Subject: Re: [PATCH v3 01/12] sched: fix imbalance flag reset
Date: Wed, 9 Jul 2014 12:43:32 +0200 [thread overview]
Message-ID: <20140709104332.GS19379@twins.programming.kicks-ass.net> (raw)
In-Reply-To: <53BCBD0E.2070609@linux.vnet.ibm.com>
[-- Attachment #1: Type: text/plain, Size: 1915 bytes --]
On Wed, Jul 09, 2014 at 09:24:54AM +0530, Preeti U Murthy wrote:
> In the example that I mention above, t1 and t2 are on the rq of cpu0;
> while t1 is running on cpu0, t2 is on the rq but does not have cpu1 in
> its cpus allowed mask. So during load balance, cpu1 tries to pull t2,
> cannot do so, and hence LBF_ALL_PINNED flag is set and it jumps to
> out_balanced. Note that there are only two sched groups at this level of
> sched domain.one with cpu0 and the other with cpu1. In this scenario we
> do not try to do active load balancing, atleast thats what the code does
> now if LBF_ALL_PINNED flag is set.
I think Vince is right in saying that in this scenario ALL_PINNED won't
be set. move_tasks() will iterate cfs_rq::cfs_tasks, that list will also
include the current running task.
And can_migrate_task() only checks for current after the pinning bits.
> Continuing with the above explanation; when LBF_ALL_PINNED flag is
> set,and we jump to out_balanced, we clear the imbalance flag for the
> sched_group comprising of cpu0 and cpu1,although there is actually an
> imbalance. t2 could still be migrated to say cpu2/cpu3 (t2 has them in
> its cpus allowed mask) in another sched group when load balancing is
> done at the next sched domain level.
And this is where Vince is wrong; note how
update_sg_lb_stats()/sg_imbalance() uses group->sgc->imbalance, but
load_balance() sets: sd_parent->groups->sgc->imbalance, so explicitly
one level up.
So what we can do I suppose is clear 'group->sgc->imbalance' at
out_balanced.
In any case, the entirely of this group imbalance crap is just that,
crap. Its a terribly difficult situation and the current bits more or
less fudge around some of the common cases. Also see the comment near
sg_imbalanced(). Its not a solid and 'correct' anything. Its a bunch of
hacks trying to deal with hard cases.
A 'good' solution would be prohibitively expensive I fear.
[-- Attachment #2: Type: application/pgp-signature, Size: 836 bytes --]
next prev parent reply other threads:[~2014-07-09 10:43 UTC|newest]
Thread overview: 132+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-06-30 16:05 [PATCH v3 00/12] sched: consolidation of cpu_power Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 01/12] sched: fix imbalance flag reset Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-08 3:13 ` Preeti U Murthy
2014-07-08 3:13 ` Preeti U Murthy
2014-07-08 10:12 ` Vincent Guittot
2014-07-08 10:12 ` Vincent Guittot
2014-07-09 3:54 ` Preeti U Murthy
2014-07-09 3:54 ` Preeti U Murthy
2014-07-09 8:27 ` Vincent Guittot
2014-07-09 8:27 ` Vincent Guittot
2014-07-09 10:43 ` Peter Zijlstra [this message]
2014-07-09 10:43 ` Peter Zijlstra
2014-07-09 11:41 ` Preeti U Murthy
2014-07-09 11:41 ` Preeti U Murthy
2014-07-09 14:44 ` Peter Zijlstra
2014-07-09 14:44 ` Peter Zijlstra
2014-07-10 9:14 ` Vincent Guittot
2014-07-10 9:14 ` Vincent Guittot
2014-07-10 9:30 ` [PATCH v4 ] " Vincent Guittot
2014-07-10 9:30 ` Vincent Guittot
2014-07-10 10:57 ` Preeti U Murthy
2014-07-10 10:57 ` Preeti U Murthy
2014-07-10 11:04 ` [PATCH v3 01/12] " Preeti U Murthy
2014-07-10 11:04 ` Preeti U Murthy
2014-07-09 3:05 ` Rik van Riel
2014-07-09 3:05 ` Rik van Riel
2014-07-09 3:36 ` Rik van Riel
2014-07-09 3:36 ` Rik van Riel
2014-07-09 10:14 ` Peter Zijlstra
2014-07-09 10:14 ` Peter Zijlstra
2014-07-09 10:30 ` Vincent Guittot
2014-07-09 10:30 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 02/12] sched: remove a wake_affine condition Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:06 ` Rik van Riel
2014-07-09 3:06 ` Rik van Riel
2014-06-30 16:05 ` [PATCH v3 03/12] sched: fix avg_load computation Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:10 ` Rik van Riel
2014-07-09 3:10 ` Rik van Riel
2014-06-30 16:05 ` [PATCH v3 04/12] sched: Allow all archs to set the power_orig Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:11 ` Rik van Riel
2014-07-09 3:11 ` Rik van Riel
2014-07-09 10:57 ` Peter Zijlstra
2014-07-09 10:57 ` Peter Zijlstra
2014-07-10 13:42 ` Vincent Guittot
2014-07-10 13:42 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 05/12] ARM: topology: use new cpu_power interface Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:11 ` Rik van Riel
2014-07-09 3:11 ` Rik van Riel
2014-07-09 7:49 ` Amit Kucheria
2014-07-09 7:49 ` Amit Kucheria
2014-07-09 10:09 ` Vincent Guittot
2014-07-09 10:09 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 06/12] sched: add per rq cpu_power_orig Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:11 ` Rik van Riel
2014-07-09 3:11 ` Rik van Riel
2014-07-09 7:50 ` Amit Kucheria
2014-07-09 7:50 ` Amit Kucheria
2014-06-30 16:05 ` [PATCH v3 07/12] sched: test the cpu's capacity in wake affine Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-09 3:12 ` Rik van Riel
2014-07-09 3:12 ` Rik van Riel
2014-07-10 11:06 ` Peter Zijlstra
2014-07-10 11:06 ` Peter Zijlstra
2014-07-10 13:58 ` Vincent Guittot
2014-07-10 13:58 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 08/12] sched: move cfs task on a CPU with higher capacity Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-10 11:18 ` Peter Zijlstra
2014-07-10 11:18 ` Peter Zijlstra
2014-07-10 14:03 ` Vincent Guittot
2014-07-10 14:03 ` Vincent Guittot
2014-07-11 14:51 ` Peter Zijlstra
2014-07-11 14:51 ` Peter Zijlstra
2014-07-11 15:17 ` Vincent Guittot
2014-07-11 15:17 ` Vincent Guittot
2014-07-14 13:51 ` Peter Zijlstra
2014-07-14 13:51 ` Peter Zijlstra
2014-07-15 9:21 ` Vincent Guittot
2014-07-15 9:21 ` Vincent Guittot
2014-07-10 11:24 ` Peter Zijlstra
2014-07-10 11:24 ` Peter Zijlstra
2014-07-10 13:59 ` Vincent Guittot
2014-07-10 13:59 ` Vincent Guittot
2014-07-10 11:31 ` Peter Zijlstra
2014-07-10 11:31 ` Peter Zijlstra
2014-06-30 16:05 ` [PATCH v3 09/12] Revert "sched: Put rq's sched_avg under CONFIG_FAIR_GROUP_SCHED" Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-07-10 13:16 ` Peter Zijlstra
2014-07-10 13:16 ` Peter Zijlstra
2014-07-11 7:51 ` Vincent Guittot
2014-07-11 7:51 ` Vincent Guittot
2014-07-11 15:13 ` Peter Zijlstra
2014-07-11 15:13 ` Peter Zijlstra
2014-07-11 17:39 ` Vincent Guittot
2014-07-11 17:39 ` Vincent Guittot
2014-07-11 20:12 ` Peter Zijlstra
2014-07-11 20:12 ` Peter Zijlstra
2014-07-14 12:55 ` Morten Rasmussen
2014-07-14 12:55 ` Morten Rasmussen
2014-07-14 13:20 ` Peter Zijlstra
2014-07-14 13:20 ` Peter Zijlstra
2014-07-14 14:04 ` Morten Rasmussen
2014-07-14 14:04 ` Morten Rasmussen
2014-07-14 16:22 ` Peter Zijlstra
2014-07-14 16:22 ` Peter Zijlstra
2014-07-15 9:20 ` Vincent Guittot
2014-07-15 9:20 ` Vincent Guittot
2014-07-14 17:54 ` Dietmar Eggemann
2014-07-14 17:54 ` Dietmar Eggemann
2014-07-18 1:27 ` Yuyang Du
2014-07-18 1:27 ` Yuyang Du
2014-07-11 16:13 ` Morten Rasmussen
2014-07-11 16:13 ` Morten Rasmussen
2014-07-15 9:27 ` Vincent Guittot
2014-07-15 9:27 ` Vincent Guittot
2014-07-15 9:32 ` Morten Rasmussen
2014-07-15 9:32 ` Morten Rasmussen
2014-07-15 9:53 ` Vincent Guittot
2014-07-15 9:53 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 10/12] sched: get CPU's utilization statistic Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 11/12] sched: replace capacity_factor by utilization Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
2014-06-30 16:05 ` [PATCH v3 12/12] sched: add SD_PREFER_SIBLING for SMT level Vincent Guittot
2014-06-30 16:05 ` Vincent Guittot
Reply instructions:
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:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20140709104332.GS19379@twins.programming.kicks-ass.net \
--to=peterz@infradead.org \
--cc=linux-arm-kernel@lists.infradead.org \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* 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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.