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;
Organization: Red Hat UK Ltd. Registered Address: Red Hat UK Ltd, Amberley
        Place, 107-111 Peascod Street, Windsor, Berkshire, SI4 1TE, United
        Kingdom.
        Registered in England and Wales under Company Registration No. 3798903
From:   David Howells <dhowells@redhat.com>
In-Reply-To: <20221216001958.1149-1-hdanton@sina.com>
References: <20221216001958.1149-1-hdanton@sina.com> <167112117887.152641.6194213035340041732.stgit@warthog.procyon.org.uk>
To:     Hillf Danton <hdanton@sina.com>
Cc:     dhowells@redhat.com, netdev@vger.kernel.org,
        Marc Dionne <marc.dionne@auristor.com>,
        linux-afs@lists.infradead.org, linux-kernel@vger.kernel.org
Subject: Re: [PATCH net 7/9] rxrpc: Fix I/O thread stop
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-ID: <430835.1671173161.1@warthog.procyon.org.uk>
Date:   Fri, 16 Dec 2022 06:46:01 +0000
Message-ID: <430836.1671173161@warthog.procyon.org.uk>
Precedence: bulk

Hillf Danton <hdanton@sina.com> wrote:

> > @@ -478,13 +479,14 @@ int rxrpc_io_thread(void *data)
> >  		}
> >  
> >  		set_current_state(TASK_INTERRUPTIBLE);
> > +		should_stop = kthread_should_stop();
> >  		if (!skb_queue_empty(&local->rx_queue) ||
> >  		    !list_empty(&local->call_attend_q)) {
> >  			__set_current_state(TASK_RUNNING);
> >  			continue;
> >  		}
> >  
> > -		if (kthread_should_stop())
> > +		if (should_stop)
> >  			break;
> >  		schedule();
> >  	}
> 
> At the second glance still fail to see the difference this change can
> make.

There is a window here:

  		if (!skb_queue_empty(&local->rx_queue) ...)
			continue;
	--->
		if (kthread_should_stop())
 			break;

in which an event can happen that should be attended to.  For instance the
AF_RXRPC socket being closed, aborting all its calls and stopping the kthread
by doing the final unuse on its rxrpc_local struct - in that order.  The
window can be expanded by an interrupt or softirq handler running.

So once we've observed that we've been asked to stop, we need to check if
there's more work to be done and, if so, do that work first.

So by doing:

		should_stop = kthread_should_stop();
  		if (!skb_queue_empty(&local->rx_queue) ...)
			continue;
		if (should_stop)
 			break;

we do all outstanding cleanup work before exiting the loop to stop the thread.

David