MIME-Version: 1.0
References: <1638619795-71451-1-git-send-email-wang.yong12@zte.com.cn>
 <Yax01zjuzmNyyJK/@balbir-desktop> <CAOH5QeDhjyjAkS1bUju2cv67KFukUr0ov8uG+z3bM6Oa=iFrMA@mail.gmail.com>
 <Ya7uQingLC3fMJlt@balbir-desktop>
In-Reply-To: <Ya7uQingLC3fMJlt@balbir-desktop>
From:   yong w <yongw.pur@gmail.com>
Date:   Wed, 8 Dec 2021 00:50:18 +0800
Message-ID: <CAOH5QeC+0xDrgO+t3zwN4o48F9Q2CiTnzQDO78kuJLfyNJwoLA@mail.gmail.com>
Subject: Re: [PATCH v2 linux-next] delayacct: track delays from memory compact
To:     Balbir Singh <bsingharora@gmail.com>
Cc:     Andrew Morton <akpm@linux-foundation.org>,
        "Peter Zijlstra (Intel)" <peterz@infradead.org>, mingo@kernel.org,
        LKML <linux-kernel@vger.kernel.org>,
        Linux MM <linux-mm@kvack.org>, yang.yang29@zte.com.cn,
        wang.yong12@zte.com.cn
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

Balbir Singh <bsingharora@gmail.com> =E4=BA=8E2021=E5=B9=B412=E6=9C=887=E6=
=97=A5=E5=91=A8=E4=BA=8C 13:16=E5=86=99=E9=81=93=EF=BC=9A
>
> On Sun, Dec 05, 2021 at 07:08:02PM +0800, yong w wrote:
> > Balbir Singh <bsingharora@gmail.com> =E4=BA=8E2021=E5=B9=B412=E6=9C=885=
=E6=97=A5=E5=91=A8=E6=97=A5 16:17=E5=86=99=E9=81=93=EF=BC=9A
> > >
> > > On Sat, Dec 04, 2021 at 04:09:55AM -0800, yongw.pur@gmail.com wrote:
> > > > From: wangyong <wang.yong12@zte.com.cn>
> > > >
> > > > Delay accounting does not track the delay of memory compact.
> > > > When there is not enough free memory, tasks can spend
> > > > a amount of their time waiting for compact.
> > > >
> > > > To get the impact of tasks in direct memory compact, measure
> > > > the delay when allocating memory through memory compact.
> > > >
> > >
> > > Should we call this DIRECT_COMPACT and through documentation
> > > or name change imply that this won't work for kcompactd the
> > > kernel thread - based on my reading of the patches.
> > >
> > Using DIRECT_COMPACT is a little redundant=EF=BC=8Cbecause the
> > delayacct stats of delay accounting is specific to tasks, it has
> > nothing to do with kcompactd, which is similar to the RECLAIM field.
> >
>
> What would we expect when we call delayacct -p <pidof kcompactd>
> to be output?
If the slow path of memory allocation is invoked in the kcompacd process,
there may be delays being recorded.

> Don't feel to strongly, but it can be confusing that kcompactd
> has spent no time in compact'ing? Not that delayacct is used for
> kernel threads, but I am not sure if that use case exists today.
Yes, delayacct does not restrict the process of obtaining information=EF=BC=
=8C
but kcompactd is used for  compaction,  the compact delay of
kcompatd is not actually a delay.Maybe it can be added to the
document later to make it clearer.

Thanks for your reply!