Received-SPF: pass (google.com: domain of linux-kernel+bounces-85047-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20240201092559.910982-1-yukuai1@huaweicloud.com>
 <20240201092559.910982-5-yukuai1@huaweicloud.com> <d233fc29-e3ab-4761-9368-c203efc0466e@redhat.com>
 <40ac6914-7c1f-00b7-f480-25c9786482fc@huaweicloud.com>
In-Reply-To: <40ac6914-7c1f-00b7-f480-25c9786482fc@huaweicloud.com>
From: Xiao Ni <xni@redhat.com>
Date: Wed, 28 Feb 2024 20:57:16 +0800
Message-ID: <CALTww29JwP+_1vfiodjy3YCze9pQ92JRGhCNVygLpm3k0gVJAA@mail.gmail.com>
Subject: Re: [PATCH v5 04/14] md: don't register sync_thread for reshape directly
To: Yu Kuai <yukuai1@huaweicloud.com>
Cc: mpatocka@redhat.com, heinzm@redhat.com, blazej.kucman@linux.intel.com, 
	agk@redhat.com, snitzer@kernel.org, dm-devel@lists.linux.dev, song@kernel.org, 
	neilb@suse.de, shli@fb.com, akpm@osdl.org, linux-kernel@vger.kernel.org, 
	linux-raid@vger.kernel.org, yi.zhang@huawei.com, yangerkun@huawei.com, 
	"yukuai (C)" <yukuai3@huawei.com>
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

On Wed, Feb 28, 2024 at 8:44=E2=80=AFPM Yu Kuai <yukuai1@huaweicloud.com> w=
rote:
>
> Hi,
>
> =E5=9C=A8 2024/02/28 20:07, Xiao Ni =E5=86=99=E9=81=93:
> > I have a question here. Is it the reason sync_thread can't run
> > md_do_sync because kthread_should_stop, so it doesn't have the chance t=
o
> > set MD_RECOVERY_DONE? Why creating sync thread in md_check_recovery
> > doesn't have this problem? Could you explain more about this?
>
> raid10_run() only register sync_thread, without calling
> md_wakeup_thread() to set the bit 'THREAD_WAKEUP', md_do_sync() will not
> be executed.

I c. The user is responsible to wake up the thread. If raid10 wakes up
the thread in the right way, we don't need to move register reshape
thread to md_check_recovery, right?

>
> raid5 defines 'pers->start' hence md_start() will call
> md_wakeup_thread().
>
> md_start_sync() will always call md_wakeup_thread() hence there is no
> such problem.
>
> BTW, this patch fix the same problem as you mentioned in your other
> thread:
>
> diff --git a/drivers/md/md.c b/drivers/md/md.c
> index 2266358d8074..54790261254d 100644
> --- a/drivers/md/md.c
> +++ b/drivers/md/md.c
> @@ -4904,6 +4904,7 @@ static void stop_sync_thread(struct mddev *mddev,
> bool locked, bool check_seq)
>          * never happen
>          */
>         md_wakeup_thread_directly(mddev->sync_thread);
> +       md_wakeup_thread(mddev->sync_thread);
>         if (work_pending(&mddev->sync_work))
>                 flush_work(&mddev->sync_work);

The first patch of my patch set has this already. Maybe it's the
reason that my patch01 can fix this similar problem.

>
> However, I think the one to register sync_thread is responsible to wake
> it up.

Agree, the user that registers thread should wake it up. So start/stop
sync thread apis are common. And they can be called by many users.

Best Regards
Xiao
>
> Thanks,
> Kuai
>