linux-kernel - [PATCH v2] sched/fair: Clear target from cpus to scan in select_idle

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>] [day] [month] [year] [list]

Date:   Fri, 28 Jan 2022 14:39:06 +0800
From:   Yicong Yang <yangyicong@...ilicon.com>
To:     <mgorman@...hsingularity.net>, <mingo@...hat.com>,
        <peterz@...radead.org>, <juri.lelli@...hat.com>,
        <vincent.guittot@...aro.org>, <mgorman@...e.de>,
        <linux-kernel@...r.kernel.org>
CC:     <dietmar.eggemann@....com>, <rostedt@...dmis.org>,
        <bsegall@...gle.com>, <bristot@...hat.com>,
        <song.bao.hua@...ilicon.com>, <prime.zeng@...wei.com>,
        <yangyicong@...ilicon.com>, <linuxarm@...wei.com>,
        <21cnbao@...il.com>
Subject: [PATCH v2] sched/fair: Clear target from cpus to scan in select_idle_cpu

Commit 56498cfb045d noticed that "When select_idle_cpu starts scanning for
an idle CPU, it starts with a target CPU that has already been checked
by select_idle_sibling. This patch starts with the next CPU instead."
It only changed the scanning start cpu to target + 1 but still leave
the target in the scanning cpumask. The target still have a chance to be
checked in the last turn. Fix this by clear the target from the cpus
to scan.

Mel Gorman worries bit clear will introduce expensive cost so I
tested hackbench on a 2P Skylake server which shows no obvious
performance degradation.

hackbench-process-pipes
                           5.17-rc1                 patched
Amean     1        0.3332 (   0.00%)      0.3392 *  -1.80%*
Amean     4        0.9536 (   0.00%)      0.9192 *   3.61%*
Amean     7        1.3382 (   0.00%)      1.3226 *   1.17%*
Amean     12       2.1286 (   0.00%)      2.1132 *   0.72%*
Amean     21       3.9210 (   0.00%)      3.8996 *   0.55%*
Amean     30       6.9853 (   0.00%)      6.6562 *   4.71%*
Amean     48      12.5749 (   0.00%)     12.3100 *   2.11%*
Amean     79      17.2422 (   0.00%)     17.1231 *   0.69%*
Amean     110     21.8868 (   0.00%)     21.6772 *   0.96%*
Amean     141     26.9521 (   0.00%)     27.0702 *  -0.44%*
Amean     160     30.5096 (   0.00%)     30.3971 *   0.37%*

Fixes: 56498cfb045d ("sched/fair: Avoid a second scan of target in select_idle_cpu")
Signed-off-by: Yicong Yang <yangyicong@...ilicon.com>
---
Change since v1:
- add test in the commit
Link:https://lore.kernel.org/lkml/20211124085401.14411-1-yangyicong@hisilicon.com/

 kernel/sched/fair.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/kernel/sched/fair.c b/kernel/sched/fair.c
index 5146163bfabb..6b8bf86f5eff 100644
--- a/kernel/sched/fair.c
+++ b/kernel/sched/fair.c
@@ -6281,6 +6281,7 @@ static int select_idle_cpu(struct task_struct *p, struct sched_domain *sd, bool
 		return -1;
 
 	cpumask_and(cpus, sched_domain_span(sd), p->cpus_ptr);
+	cpumask_clear_cpu(target, cpus);
 
 	if (sched_feat(SIS_PROP) && !has_idle_core) {
 		u64 avg_cost, avg_idle, span_avg;
-- 
2.24.0