Received-SPF: pass (google.com: domain of linux-nfs+bounces-133-linux.lists.archive=gmail.com@vger.kernel.org designates 2604:1380:45e3:2400::1 as permitted sender) client-ip=2604:1380:45e3:2400::1;
Date: Tue, 28 Nov 2023 17:59:45 +0100
From: Oleg Nesterov <oleg@redhat.com>
To: Christian Brauner <brauner@kernel.org>
Cc: NeilBrown <neilb@suse.de>, Al Viro <viro@zeniv.linux.org.uk>,
	Jens Axboe <axboe@kernel.dk>, Chuck Lever <chuck.lever@oracle.com>,
	Jeff Layton <jlayton@kernel.org>, Ingo Molnar <mingo@redhat.com>,
	Peter Zijlstra <peterz@infradead.org>,
	Juri Lelli <juri.lelli@redhat.com>,
	Vincent Guittot <vincent.guittot@linaro.org>,
	linux-fsdevel@vger.kernel.org, linux-kernel@vger.kernel.org,
	linux-nfs@vger.kernel.org
Subject: Re: [PATCH/RFC] core/nfsd: allow kernel threads to use task_work.
Message-ID: <20231128165945.GD22743@redhat.com>
References: <170112272125.7109.6245462722883333440@noble.neil.brown.name>
 <20231128-arsch-halbieren-b2a95645de53@brauner>
 <20231128135258.GB22743@redhat.com>
 <20231128-elastisch-freuden-f9de91041218@brauner>
Precedence: bulk
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20231128-elastisch-freuden-f9de91041218@brauner>
User-Agent: Mutt/1.5.24 (2015-08-30)

On 11/28, Christian Brauner wrote:
>
> Yeah, I had played with that as well. Only reason I didn't do it was to
> avoid a PF_* flag. If that's preferable it might be worth to just add
> PF_TASK_WORK and decouple this from PF_KTHREAD.

OK, I won't insist.

But,

> +       /*
> +        * By default only non-kernel threads can use task work. Kernel
> +        * threads that manage task work explicitly can add that flag in
> +        * their kthread callback.
> +        */
> +       if (!args->kthread)
> +               p->flags |= PF_TASK_WORK;

The comment and the name of the new flag look a bit misleading to me...

kthreads can use task_work's. You can create a kthread which does
task_work_run() from time to time and use task_work_add() on it,
nothing wrong with that.

Probably nobody does this right now (I didn't try to check), but please
note irq_thread()->task_work_add(on_exit_work).

Oleg.