Re: [CFT][PATCH v3] fork, vhost: Use CLONE_THREAD to fix freezer/ps regression

From: Oleg Nesterov <oleg@redhat.com>
To: "Eric W. Biederman" <ebiederm@xmission.com>
Cc: Mike Christie <michael.christie@oracle.com>,
	linux@leemhuis.info, nicolas.dichtel@6wind.com, axboe@kernel.dk,
	torvalds@linux-foundation.org, linux-kernel@vger.kernel.org,
	virtualization@lists.linux-foundation.org, mst@redhat.com,
	sgarzare@redhat.com, jasowang@redhat.com, stefanha@redhat.com,
	brauner@kernel.org
Subject: Re: [CFT][PATCH v3] fork, vhost: Use CLONE_THREAD to fix freezer/ps regression
Date: Wed, 14 Jun 2023 19:08:04 +0200	[thread overview]
Message-ID: <20230614170803.GA18850@redhat.com> (raw)
In-Reply-To: <87o7llycau.fsf@email.froward.int.ebiederm.org>

On 06/11, Eric W. Biederman wrote:
>
> Oleg Nesterov <oleg@redhat.com> writes:
>
> > Could you spell to explain why this can't work (again, in this simple case) ?
> >
> > My current (and I know, very poor) understanding is that .release() should
> > roughly do the following:
> >
> > 	1. Ensure that vhost_work_queue() can't add the new callbacks
> >
> > 	2. Call vhost_dev_flush() to ensure that worker->work_list is empty
> >
> > 	3. Call vhost_task_stop()
>
> At least in the case of exec by the time the final fput happens
> from close_on_exec the task has already changed it's mm.

Of course you are right.

But can't resist, please note that I only meant "this simple case" which
doesn't include exec/etc.

Nevermind. As Mike explains there are more problems even in this particular
"simple" case, and I am not surprised.

Sorry for noise,

Oleg.