Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Tue, 1 Mar 2022 18:57:03 +0100
From:   Michal =?iso-8859-1?Q?Koutn=FD?= <mkoutny@suse.com>
To:     Shakeel Butt <shakeelb@google.com>
Cc:     Michal Hocko <mhocko@suse.com>,
        Johannes Weiner <hannes@cmpxchg.org>,
        Roman Gushchin <roman.gushchin@linux.dev>,
        Ivan Babrou <ivan@cloudflare.com>,
        Andrew Morton <akpm@linux-foundation.org>,
        Cgroups <cgroups@vger.kernel.org>, Linux MM <linux-mm@kvack.org>,
        LKML <linux-kernel@vger.kernel.org>,
        Daniel Dao <dqminh@cloudflare.com>,
        stable <stable@vger.kernel.org>
Subject: Re: [PATCH] memcg: async flush memcg stats from perf sensitive
 codepaths
Message-ID: <20220301175703.GA10867@blackbody.suse.cz>
References: <20220226002412.113819-1-shakeelb@google.com>
 <Yh3h33W45+YaMo92@dhcp22.suse.cz>
 <CALvZod7aF9xRc+XvY7GPN7OnDyPitt1H6Q4yrwzAXTFzv1LzWQ@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CALvZod7aF9xRc+XvY7GPN7OnDyPitt1H6Q4yrwzAXTFzv1LzWQ@mail.gmail.com>
User-Agent: Mutt/1.10.1 (2018-07-13)
Precedence: bulk

Making decisions based on up to 2 s old information.

On Tue, Mar 01, 2022 at 09:21:12AM -0800, Shakeel Butt <shakeelb@google.com> wrote:
> Without flushing the worst that can happen in the refault path is
> false (or missed) activations of the refaulted page.

Yeah, this may under- or overestimate workingset size (when it's
changing), the result is likely only less efficient reclaim.

> For reclaim code, some heuristics (like deactivating active LRU or
> cache-trim) may act on old information.

Here, I'd be more careful whether such a delay cannot introduce some
unstable behavior (permanent oscillation in the worst case).

> Now, coming to your question, yes, we can remove the flushing from
> these performance critical codepaths as the stats at most will be 2
> second old due to periodic flush. 

Another aspect is that people will notice and report such a narrowly
located performance regression more easily than reduced/less predictable
reclaim behavior. (IMO the former is better, OTOH, it can also be
interpreted that noone notices (is able to notice).)

Michal