Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.18 as permitted sender) client-ip=23.128.96.18;
Date:   Mon, 8 Feb 2021 11:45:56 -0500
From:   Johannes Weiner <hannes@cmpxchg.org>
To:     Chengming Zhou <zhouchengming@bytedance.com>
Cc:     mingo@redhat.com, peterz@infradead.org, juri.lelli@redhat.com,
        vincent.guittot@linaro.org, rostedt@goodmis.org,
        dietmar.eggemann@arm.com, linux-kernel@vger.kernel.org
Subject: Re: [PATCH] psi: Remove the redundant psi_task_tick
Message-ID: <YCFqxEw0wJC6UGiE@cmpxchg.org>
References: <20210207115642.75620-1-zhouchengming@bytedance.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20210207115642.75620-1-zhouchengming@bytedance.com>
Precedence: bulk

On Sun, Feb 07, 2021 at 07:56:42PM +0800, Chengming Zhou wrote:
> From: zhouchengming <zhouchengming@bytedance.com>
> 
> When the current task in a cgroup is in_memstall, the corresponding psi_group
> is in PSI_MEM_FULL state

This is correct.

> so we can remove the redundant psi_task_tick from scheduler_tick to
> save this periodic cost.

But this isn't. It IS the task tick that makes it so. The state change
tracking counts MEMSTALL tasks and ONCPU tasks, but it doesn't check
whether the ONCPU task is the MEMSTALL one.

It would be possible to incorporate that into psi_task_switch(), but
you have to be careful with the branches because that path is often
hotter than the timer tick.

This patch by itself breaks page reclaim detection, so NAK.