linux-kernel - Re: [PATCH] mm: vmscan: Replace shrinker

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

Message-ID: <YVQef1zSQ15JNnXP@dhcp22.suse.cz>
Date:   Wed, 29 Sep 2021 10:06:23 +0200
From:   Michal Hocko <mhocko@...e.com>
To:     Sultan Alsawaf <sultan@...neltoast.com>
Cc:     Andrew Morton <akpm@...ux-foundation.org>, linux-mm@...ck.org,
        linux-kernel@...r.kernel.org
Subject: Re: [PATCH] mm: vmscan: Replace shrinker_rwsem trylocks with SRCU
 protection

On Mon 27-09-21 09:53:59, Sultan Alsawaf wrote:
> On Mon, Sep 27, 2021 at 10:36:36AM +0200, Michal Hocko wrote:
> > Can you be more specific about those scenarios please?
> 
> I see frequent tmpfs mounts/unmounts in Android which result in lots of
> superblocks being created and destroyed, and each superblock has its own
> shrinker.

This is an important detail to mention in the changelog. Including some
details about the scale of the problem.

> > I have a vague recollection that this approach has been proposed in the
> > past. Have you checked previous attempts?
> 
> I wasn't aware that there were previous attempts. Is this [1] the previous
> attempt you're thinking of? It seems like the thread just died out. Was the main
> gripe that always enabling CONFIG_SRCU increased the kernel's size too much when
> tinyfication was desired?

There were more discussions. I have also found https://lore.kernel.org/lkml/1510609063-3327-1-git-send-email-penguin-kernel@I-love.SAKURA.ne.jp
My memory is quite dim and I do not have time to re-read those
discussions again right now. But if I remember correctly the last one
has died out because a deeper evaluation between the SRCU and rw
semaphores was required.

> Sultan
> 
> [1] https://lore.kernel.org/all/153365347929.19074.12509495712735843805.stgit@localhost.localdomain/

-- 
Michal Hocko
SUSE Labs