2015-11-10 04:46:29

by Al Viro

[permalink] [raw]
Subject: Re: [PATCH 1/4] nbd: Remove signal usage

On Thu, Oct 29, 2015 at 04:42:37PM +0100, Markus Pargmann wrote:
> del_timer_sync(&nbd->timeout_timer);
> +
> +out:
> + spin_unlock_irq(&nbd->sock_lock);

... and in its callback we have this:

> @@ -148,17 +155,15 @@ static void nbd_xmit_timeout(unsigned long arg)
>
> nbd->disconnect = true;
>
> - spin_lock_irqsave(&nbd->tasks_lock, flags);
> + spin_lock_irqsave(&nbd->sock_lock, flags);

* CPU 1 enters sock_shutdown() and grabs ->sock_lock.
* on CPU2 the timer hits and we enter the callback, where we spin on that
spinlock.
* in the meanwhile, CPU1 calls del_timer_sync()

Deadlock...


2015-11-10 10:23:23

by Markus Pargmann

[permalink] [raw]
Subject: Re: [PATCH 1/4] nbd: Remove signal usage

Hi,

On Tuesday 10 November 2015 04:46:18 Al Viro wrote:
> On Thu, Oct 29, 2015 at 04:42:37PM +0100, Markus Pargmann wrote:
> > del_timer_sync(&nbd->timeout_timer);
> > +
> > +out:
> > + spin_unlock_irq(&nbd->sock_lock);
>
> ... and in its callback we have this:
>
> > @@ -148,17 +155,15 @@ static void nbd_xmit_timeout(unsigned long arg)
> >
> > nbd->disconnect = true;
> >
> > - spin_lock_irqsave(&nbd->tasks_lock, flags);
> > + spin_lock_irqsave(&nbd->sock_lock, flags);
>
> * CPU 1 enters sock_shutdown() and grabs ->sock_lock.
> * on CPU2 the timer hits and we enter the callback, where we spin on that
> spinlock.
> * in the meanwhile, CPU1 calls del_timer_sync()
>
> Deadlock...

Thank you. Yes that locking block in sock_shutdown is to large. And probably
the del_timer_sync() isn't necessary, we can just use del_timer().

It may even be possible to remove the sock_lock completely. Will look into this
and post a v2.

Thanks,

Markus

--
Pengutronix e.K. | |
Industrial Linux Solutions | http://www.pengutronix.de/ |
Peiner Str. 6-8, 31137 Hildesheim, Germany | Phone: +49-5121-206917-0 |
Amtsgericht Hildesheim, HRA 2686 | Fax: +49-5121-206917-5555 |


Attachments:
signature.asc (819.00 B)
This is a digitally signed message part.