Re: [PATCH/RFC] core/nfsd: allow kernel threads to use task_work.

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.
Date: Tue, 28 Nov 2023 14:52:59 +0100	[thread overview]
Message-ID: <20231128135258.GB22743@redhat.com> (raw)
In-Reply-To: <20231128-arsch-halbieren-b2a95645de53@brauner>

On 11/28, Christian Brauner wrote:
>
> Should be simpler if you invert the logic?
>
> COMPLETELY UNTESTED

Agreed, this looks much better to me. But perhaps we can just add the new
PF_KTHREAD_XXX flag and change fput

	--- a/fs/file_table.c
	+++ b/fs/file_table.c
	@@ -445,7 +445,8 @@ void fput(struct file *file)
		if (atomic_long_dec_and_test(&file->f_count)) {
			struct task_struct *task = current;

	-		if (likely(!in_interrupt() && !(task->flags & PF_KTHREAD))) {
	+		if (likely(!in_interrupt() &&
	+		    task->flags & (PF_KTHREAD|PF_KTHREAD_XXX) != PF_KTHREAD) {
				init_task_work(&file->f_rcuhead, ____fput);
				if (!task_work_add(task, &file->f_rcuhead, TWA_RESUME))
					return;

?

Then nfsd() can simply set PF_KTHREAD_XXX. This looks even simpler to me.

Oleg.