linux-kernel - [PATCH 0/2] Reduce stacking and overscheduling

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-next>] [day] [month] [year] [list]

Message-Id: <20211021145603.5313-1-mgorman@techsingularity.net>
Date:   Thu, 21 Oct 2021 15:56:01 +0100
From:   Mel Gorman <mgorman@...hsingularity.net>
To:     Peter Zijlstra <peterz@...radead.org>
Cc:     Ingo Molnar <mingo@...nel.org>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Valentin Schneider <valentin.schneider@....com>,
        Aubrey Li <aubrey.li@...ux.intel.com>,
        Barry Song <song.bao.hua@...ilicon.com>,
        Mike Galbraith <efault@....de>,
        Srikar Dronamraju <srikar@...ux.vnet.ibm.com>,
        LKML <linux-kernel@...r.kernel.org>,
        Mel Gorman <mgorman@...hsingularity.net>
Subject: [PATCH 0/2] Reduce stacking and overscheduling

These series tackles two problems. The first is that heavy wakers
can stack an excessive number of tasks on the same CPU. The
second is that tasks can overschedule when the task has not
reached its minimum preemption granularity.

The patches are independent but were discussed together in the thread
https://lore.kernel.org/r/20210920142614.4891-1-mgorman@techsingularity.net
so are presented together.

With both patches on a zen3 machine

hackbench-process-pipes
                          5.15.0-rc3             5.15.0-rc3
                             vanilla sched-scalewakegran-v3r2
Amean     1        0.3667 (   0.00%)      0.3823 (  -4.27%)
Amean     4        0.5343 (   0.00%)      0.4867 (   8.92%)
Amean     7        0.5300 (   0.00%)      0.5053 (   4.65%)
Amean     12       0.5737 (   0.00%)      0.5450 (   5.00%)
Amean     21       0.6727 (   0.00%)      0.6807 (  -1.19%)
Amean     30       0.8583 (   0.00%)      0.7107 *  17.20%*
Amean     48       1.3977 (   0.00%)      1.0447 *  25.26%*
Amean     79       1.9790 (   0.00%)      1.6033 *  18.98%*
Amean     110      2.8020 (   0.00%)      2.0763 *  25.90%*
Amean     141      3.6683 (   0.00%)      2.5313 *  31.00%*
Amean     172      4.6687 (   0.00%)      3.1163 *  33.25%*
Amean     203      5.2183 (   0.00%)      3.5560 *  31.86%*
Amean     234      6.1077 (   0.00%)      3.8913 *  36.29%*
Amean     265      7.1313 (   0.00%)      4.2293 *  40.69%*
Amean     296      7.7557 (   0.00%)      4.5357 *  41.52%*

                  5.15.0-rc3  5.15.0-rc3
                     vanilla sched-scalewakegran-v3r2
Duration User        2933.05     2034.17
Duration System     25652.83    17137.08
Duration Elapsed      162.50      120.25

-- 
2.31.1

Mel Gorman (2):
  sched/fair: Couple wakee flips with heavy wakers
  sched/fair: Increase wakeup_gran if current task has not executed the
    minimum granularity

 kernel/sched/fair.c     | 27 ++++++++++++++++++++++++---
 kernel/sched/features.h |  2 ++
 2 files changed, 26 insertions(+), 3 deletions(-)

-- 
2.31.1