Re: [PATCH -mm 3/4][AIO] - AIO completion signal notification

[Andrew Morton]

On Thu, 23 Nov 2006 10:47:55 +0100
Sébastien Dugué <sebastien.dugue@xxxxxxxx> wrote:

> On Thu, 23 Nov 2006 00:40:53 -0800
> Andrew Morton <akpm@xxxxxxxx> wrote:
> 
> > On Thu, 23 Nov 2006 09:28:05 +0100
> > Sébastien Dugué <sebastien.dugue@xxxxxxxx> wrote:
> > 
> > > > > +	target = good_sigevent(&event);
> > > > > +
> > > > > +	if (unlikely(!target || (target->flags & PF_EXITING)))
> > > > > +		goto out_unlock;
> > > > > +
> > > > > +	
> > > > > +
> > > > > +	if (notify->notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID)) {
> > > > > +		/*
> > > > > +		 * This reference will be dropped in really_put_req() when
> > > > > +		 * we're done with the request.
> > > > > +		 */
> > > > > +		get_task_struct(target);
> > > > > +	}
> > > > 
> > > > It worries me that this function can save away a task_struct* without
> > > > having taken a reference against it.
> > > > 
> > > 
> > >   OK. Does moving 'notify->target = target;' after the get_task_struct() will
> > > do, or am I missing something more subtle?
> > 
> > Well it's your code - you tell me ;)
> > 
> > It is unsafe (and rather pointless) to be saving the address of some structure
> > which can be freed at any time.
> 
>   Sorry, I expressed myself quite badly. What I wanted to know is whether you
> are worried with the task been freed between saving its pointer and getting a
> ref on it (which is trivial to fix) or you are thinking of something deeper.
> 

Look:

> +	notify->target = target;
> +
> +	if (notify->notify == (SIGEV_SIGNAL|SIGEV_THREAD_ID)) {
> +		/*
> +		 * This reference will be dropped in really_put_req() when
> +		 * we're done with the request.
> +		 */
> +		get_task_struct(target);

If that test fails, we've saved a pointer to the task_struct without having
taken a refreence on it.

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/