Re: [patch 2.6.15-mm4] sem2mutex: drivers/input/, #2

[Dmitry Torokhov <dtor_core@ameritech.net>]

On Saturday 14 January 2006 11:53, Arjan van de Ven wrote:
> On Sat, 2006-01-14 at 11:43 -0500, Dmitry Torokhov wrote:
> > On Saturday 14 January 2006 11:21, Faidon Liambotis wrote:
> > > Ingo Molnar wrote:
> > > > @@ -1371,11 +1373,11 @@ static ssize_t psmouse_attr_set_protocol
> > > >  			return -EIO;
> > > >  		}
> > > >  
> > > > -		up(&psmouse_sem);
> > > > +		mutex_unlock(&psmouse_mutex);
> > > >  		serio_unpin_driver(serio);
> > > >  		serio_unregister_child_port(serio);
> > > >  		serio_pin_driver_uninterruptible(serio);
> > > > -		down(&psmouse_sem);
> > > > +		mutex_lock(&psmouse_mutex);
> > > Isn't that supposed to be the other way around?
> > > 
> > 
> > No, that's correct. We temporarily dropping the lock to let the serio core
> > do its job and then reaqcuire it again to finish what we were doing. 
> 
> are you REALLY sure? It looks to me as if this function is called from
> sysfs.. so you don't know that you have the lock...
>

Yes, I am sure, the sysfs helper does acquire that lock. See 

	drivers/input/mouse/psmouse-base.c::psmouse_attr_set_helper()

-- 
Dmitry