Received-SPF: pass (google.com: domain of linux-kernel+bounces-55835-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:40f1:3f00::1 as permitted sender) client-ip=2604:1380:40f1:3f00::1;
Precedence: bulk
MIME-Version: 1.0
References: <20240206080024.2373490-1-yunlong.xing@unisoc.com> <ZcJj2dUUyAY_7XFS@slm.duckdns.org>
In-Reply-To: <ZcJj2dUUyAY_7XFS@slm.duckdns.org>
From: yunlong xing <yunlongxing23@gmail.com>
Date: Wed, 7 Feb 2024 10:06:56 +0800
Message-ID: <CA+3AYtQBci3WgTfhSA-8V34qWbnQ23V86rw7rGArGqStdpnVPw@mail.gmail.com>
Subject: Re: [PATCH] workqueue: Fix pool->nr_running type back to atomic
To: Tejun Heo <tj@kernel.org>
Cc: Yunlong Xing <yunlong.xing@unisoc.com>, jiangshanlai@gmail.com, 
	linux-kernel@vger.kernel.org, hongyu.jin@unisoc.com, zhiguo.niu@unisoc.com, 
	zhengxu.zhang@unisoc.com
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Feb 7, 2024 at 12:52=E2=80=AFAM Tejun Heo <tj@kernel.org> wrote:
>
> Hello,
>
> On Tue, Feb 06, 2024 at 04:00:24PM +0800, Yunlong Xing wrote:
> > In CPU-hotplug test, when plug the core, set_cpus_allowed_ptr() restori=
ng
> > the cpus_mask of the per-cpu worker may fail, the cpus_mask of the work=
er
> > remain wq_unbound_cpumask until the core hotpluged next time. so, worke=
rs
> > in the same per-cpu pool can run concurrently and change nr_running at =
the
> >  same time, atomic problem occur.
>
> How would set_cpus_allowed_ptr() fail? That should trigger WARN_ON, right=
?
> If set_cpus_allowed_ptr() fails, nr_running getting desynchronized is onl=
y a
> part of the problem. We will end up running per-cpu work items which must
> execute on the same CPU on foreign CPUs.

Hi tejun=EF=BC=8C

Yes, WARN_ON is triggered. The reason of  set_cpus_allowed_ptr() fail  stil=
l
needs to be further investigated.

I was originally planning to inquire workqueue whether or not allows the wo=
rker
that associated with one cpu running on the other cpus=EF=BC=9FFrom your re=
ply=EF=BC=8Cthe
answer of my question is not allow=EF=BC=8Cright=EF=BC=9F

>
> Thanks.
>
> --
> tejun