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-prev] [day] [month] [year] [list]
Date:   Sun, 16 Jun 2019 12:07:44 +0800
From:   Xunlei Pang <xlpang@...ux.alibaba.com>
To:     Chris Down <chris@...isdown.name>
Cc:     Roman Gushchin <guro@...com>, Michal Hocko <mhocko@...nel.org>,
        Johannes Weiner <hannes@...xchg.org>,
        linux-kernel@...r.kernel.org, linux-mm@...ck.org
Subject: Re: [PATCH] psi: Don't account force reclaim as memory pressure

Hi Chris,

On 2019/6/15 PM 11:58, Chris Down wrote:
> Hi Xunlei,
> 
> Xunlei Pang writes:
>> There're several cases like resize and force_empty that don't
>> need to account to psi, otherwise is misleading.
> 
> I'm afraid I'm quite confused by this patch. Why do you think accounting
> for force reclaim in PSI is misleading? I completely expect that force
> reclaim should still be accounted for as memory pressure, can you
> present some reason why it shouldn't be?

We expect psi stands for negative factors to applications
which affect their response time, but force reclaims are
behaviours triggered on purpose like "/proc/sys/vm/drop_caches",
not the real negative pressure.

e.g. my module force reclaims the dead memcgs, there's no
application attached to it, and its memory(page caches) is
usually useless, force reclaiming them doesn't mean the
system or parent memcg is under memory pressure, while
actually the whole system or the parent memcg has plenty
of free memory. If the force reclaim causes further memory
pressure like hot page cache miss, then the workingset
refault psi will catch that.

Powered by blists - more mailing lists

Powered by Openwall GNU/*/Linux Powered by OpenVZ