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:   Thu, 7 May 2020 16:04:38 -0700
From:   Andrew Morton <akpm@linux-foundation.org>
To:     "Guilherme G. Piccoli" <gpiccoli@canonical.com>
Cc:     linux-mm@kvack.org, linux-kernel@vger.kernel.org,
        gavin.guo@canonical.com, kernel@gpiccoli.net,
        Mel Gorman <mgorman@techsingularity.net>
Subject: Re: [PATCH] mm, compaction: Indicate when compaction is manually
 triggered by sysctl
Message-Id: <20200507160438.ed336a1e00c23c6863d75ae5@linux-foundation.org>
In-Reply-To: <20200507215946.22589-1-gpiccoli@canonical.com>
References: <20200507215946.22589-1-gpiccoli@canonical.com>
Mime-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

On Thu,  7 May 2020 18:59:46 -0300 "Guilherme G. Piccoli" <gpiccoli@canonical.com> wrote:

> Currently we have no way to determine if compaction was triggered
> by sysctl write, but this is an interesting information to have,
> specially in systems with high uptime that presents lots of
> fragmented memory. There's no statistic indicating if compaction
> was triggered manually or ran by Linux itself, the vmstat numbers
> cannot tell the user this information.

Could add it to vmstat?

> This patch adds a very simple message to kernel log when compaction
> is requested through a write to sysctl file, and also it accumulates
> the number of previously manual compaction executions. It follows
> the approach used by drop_caches.

Userspace could write to /dev/kmsg when it decides to trigger
compaction?  Although using the kernel log seems a fairly lame way for
userspace to record its own actions...