linux-kernel - Re: [PATCH v3 06/10] sched/fair: Use the prefer

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 for Android: free password hash cracker in your pocket

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

Message-ID: <xhsmh1qmkr7pa.mognet@vschneid.remote.csb>
Date:   Mon, 20 Feb 2023 09:45:37 +0000
From:   Valentin Schneider <vschneid@...hat.com>
To:     hupu <921889738@...com>, peterz@...radead.org
Cc:     bristot@...hat.com, bsegall@...gle.com, dietmar.eggemann@....com,
        ionela.voinescu@....com, juri.lelli@...hat.com,
        len.brown@...el.com, linux-kernel@...r.kernel.org, mgorman@...e.de,
        rafael.j.wysocki@...el.com, ravi.v.shankar@...el.com,
        ricardo.neri-calderon@...ux.intel.com, ricardo.neri@...el.com,
        rostedt@...dmis.org, srinivas.pandruvada@...ux.intel.com,
        tim.c.chen@...el.com, tim.c.chen@...ux.intel.com,
        vincent.guittot@...aro.org, x86@...nel.org, hupu@...o.com
Subject: Re: [PATCH v3 06/10] sched/fair: Use the prefer_sibling flag of the
 current sched domain

On 20/02/23 15:22, hupu wrote:
> From: Peter Zijlstra <peterz@...radead.org>
>> Specifically, I'm thinking something in the degenerate area where it
>> looks if a given domain has equal depth children or so.
>
> By the way, in the last email you mentioned "the degenerate area". I can't understand what it means because of cultural differences. Does it refer to this scenario: sched_domain at the SMT layer is destroyed after an SMT thread goes offline, so mc_domain->child is NULL?
>

This references sched domain degeneration, cf. sd_parent_degenerate() and
sd_degenerate() in sched/topology.c

Sched domains are built following a layout described in
sched_domain_topology[], and depending on the actual CPU layout some
domains end up not being worth keeping (e.g. they only contain one CPU - we
need at least 2 to balance things), so we destroy (degenerate) them.

> hupu