linux-kernel - Re: [PATCH 2/2] mm: Consider subtrees in memory.events

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

Message-ID: <20190130170658.GY50184@devbig004.ftw2.facebook.com>
Date:   Wed, 30 Jan 2019 09:06:58 -0800
From:   Tejun Heo <tj@...nel.org>
To:     Michal Hocko <mhocko@...nel.org>
Cc:     Johannes Weiner <hannes@...xchg.org>,
        Chris Down <chris@...isdown.name>,
        Andrew Morton <akpm@...ux-foundation.org>,
        Roman Gushchin <guro@...com>, Dennis Zhou <dennis@...nel.org>,
        linux-kernel@...r.kernel.org, cgroups@...r.kernel.org,
        linux-mm@...ck.org, kernel-team@...com
Subject: Re: [PATCH 2/2] mm: Consider subtrees in memory.events

Hello, Michal.

On Wed, Jan 30, 2019 at 05:50:58PM +0100, Michal Hocko wrote:
> > Yeah, cgroup.events and .stat files as some of the local stats would
> > be useful too, so if we don't flip memory.events we'll end up with sth
> > like cgroup.events.local, memory.events.tree and memory.stats.local,
> > which is gonna be hilarious.
> 
> Why cannot we simply have memory.events_tree and be done with it? Sure
> the file names are not goin to be consistent which is a minus but that
> ship has already sailed some time ago.

Because the overall cost of shitty interface will be way higher in the
longer term.  cgroup2 interface is far from perfect but is way better
than cgroup1 especially for the memory controller.  Why do you think
that is?

> > If you aren't willing to change your mind, the only option seems to be
> > introducing a mount option to gate the flip and additions of local
> > files.  Most likely, userspace will enable the option by default
> > everywhere, so the end result will be exactly the same but I guess
> > it'll better address your concern.
> 
> How does the consumer of the API learns about the mount type?

It's gonna be a mount flag exposed in /sys/kernel/cgroup/features.

Thanks.

-- 
tejun