linux-kernel - [RFC] why can't dynamic isolation just like the static way

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

Date:   Tue, 11 Feb 2020 16:17:34 +0800
From:   王贇 <yun.wang@...ux.alibaba.com>
To:     Ingo Molnar <mingo@...hat.com>,
        Peter Zijlstra <peterz@...radead.org>,
        Juri Lelli <juri.lelli@...hat.com>,
        Vincent Guittot <vincent.guittot@...aro.org>,
        Dietmar Eggemann <dietmar.eggemann@....com>,
        Steven Rostedt <rostedt@...dmis.org>,
        Ben Segall <bsegall@...gle.com>, Mel Gorman <mgorman@...e.de>,
        "open list:SCHEDULER" <linux-kernel@...r.kernel.org>
Subject: [RFC] why can't dynamic isolation just like the static way

Hi, folks

We are dealing with isolcpus these days and try to do the isolation
dynamically.

The kernel doc lead us into the cpuset.sched_load_balance, it's fine
to achieve the dynamic isolation with it, however we got problem with
the systemd stuff.

It's keeping create cgroup with sched_load_balance enabled on default,
while the cpus are overlapped with the isolated ones, which lead into
sched domain rebuild and these cpus become non-isolated.

We're just looking forward an easy way to dynamic isolate some cpus,
just like the isolation parameter, but sched_load_balance forcing us
to dealing with the management of cgroups, we really don't get the
point in here...

Why do we have to mix the isolation with cgroups? Why not just provide
a proc entry to read cpumask and rebuild the domains?

Please let us know if there is any good reason to make the dynamic
isolation in that way, appreciated in advance :-)

Regards,
Michael Wang