linux-kernel - Re: [PATCH 02/16] mm: vmstat: use s32 for vm_node_stat_diff in struct per_cpu

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: <20191021011550.GA8869@castle>
Date:   Mon, 21 Oct 2019 01:15:57 +0000
From:   Roman Gushchin <guro@...com>
To:     Christopher Lameter <cl@...ux.com>
CC:     "linux-mm@...ck.org" <linux-mm@...ck.org>,
        Michal Hocko <mhocko@...nel.org>,
        Johannes Weiner <hannes@...xchg.org>,
        "linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
        Kernel Team <Kernel-team@...com>,
        Shakeel Butt <shakeelb@...gle.com>,
        Vladimir Davydov <vdavydov.dev@...il.com>,
        Waiman Long <longman@...hat.com>
Subject: Re: [PATCH 02/16] mm: vmstat: use s32 for vm_node_stat_diff in struct
 per_cpu_nodestat

On Sun, Oct 20, 2019 at 10:44:10PM +0000, Christopher Lameter wrote:
> On Thu, 17 Oct 2019, Roman Gushchin wrote:
> 
> > But if some counters are in bytes (and the next commit in the series
> > will convert slab counters to bytes), it's not gonna work:
> > value in bytes can easily exceed s8 without exceeding the threshold
> > converted to bytes. So to avoid overfilling per-cpu caches and breaking
> > vmstats correctness, let's use s32 instead.
> 
> This quardruples the cache footprint of the counters and likely has some
> influence on performance.

But otherwise slab allocation counters will be flushed too often, which
will be likely noticeable. I can do custom s32 buffers only for these counters,
but to me it seems like an unnecessary complication, unless we'll find
a clear regression.

Sp far I haven't noticed any regression on the set of workloads where I did test
the patchset, but if you know any benchmark or realistic test which can affected
by this check, I'll be happy to try.

Also, less-than-word-sized operations can be slower on some platforms.