Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Date:   Mon, 22 May 2023 14:30:29 +0200
From:   Oleg Nesterov <oleg@redhat.com>
To:     Mike Christie <michael.christie@oracle.com>
Cc:     linux@leemhuis.info, nicolas.dichtel@6wind.com, axboe@kernel.dk,
        ebiederm@xmission.com, 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: [PATCH 3/3] fork, vhost: Use CLONE_THREAD to fix freezer/ps
 regression
Message-ID: <20230522123029.GA22159@redhat.com>
References: <20230522025124.5863-1-michael.christie@oracle.com>
 <20230522025124.5863-4-michael.christie@oracle.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <20230522025124.5863-4-michael.christie@oracle.com>
User-Agent: Mutt/1.5.24 (2015-08-30)
Precedence: bulk

Confused, please help...

On 05/21, Mike Christie wrote:
>
> --- a/drivers/vhost/vhost.c
> +++ b/drivers/vhost/vhost.c
> @@ -338,6 +338,7 @@ static int vhost_worker(void *data)
>  	struct vhost_worker *worker = data;
>  	struct vhost_work *work, *work_next;
>  	struct llist_node *node;
> +	bool dead = false;
>
>  	for (;;) {
>  		/* mb paired w/ kthread_stop */
> @@ -349,8 +350,22 @@ static int vhost_worker(void *data)
>  		}
>
>  		node = llist_del_all(&worker->work_list);
> -		if (!node)
> +		if (!node) {
>  			schedule();
> +			/*
> +			 * When we get a SIGKILL our release function will
> +			 * be called. That will stop new IOs from being queued
> +			 * and check for outstanding cmd responses. It will then
> +			 * call vhost_task_stop to tell us to return and exit.
> +			 */

But who will call the release function / vhost_task_stop() and when this
will happen after this thread gets SIGKILL ?

> +			if (!dead && signal_pending(current)) {
> +				struct ksignal ksig;
> +
> +				dead = get_signal(&ksig);
> +				if (dead)
> +					clear_thread_flag(TIF_SIGPENDING);

If you do clear_thread_flag(TIF_SIGPENDING), then why do we need 1/3 ?


Also. Suppose that vhost_worker() dequeues SIGKILL and clears TIF_SIGPENDING.

SIGSTOP, PTRACE_INTERRUPT, freezer can come and set TIF_SIGPENDING again.
In this case the main for (;;) loop will spin without sleeping until
vhost_task_should_stop() becomes true?

Oleg.