Re: [PATCH v2] gpio: cdev: sanitize the label before requesting the interrupt

From: Kent Gibson
Date: Fri Mar 22 2024 - 08:04:45 EST


On Fri, Mar 22, 2024 at 12:52:29PM +0100, Bartosz Golaszewski wrote:
> On Fri, Mar 22, 2024 at 12:18 PM Kent Gibson <warthog618@xxxxxxxxx> wrote:
> >
> > On Fri, Mar 22, 2024 at 10:02:08AM +0100, Bartosz Golaszewski wrote:
> > > From: Bartosz Golaszewski <bartosz.golaszewski@xxxxxxxxxx>
> > >
> > > @@ -2198,12 +2216,18 @@ static int lineevent_create(struct gpio_device *gdev, void __user *ip)
> > > if (ret)
> > > goto out_free_le;
> > >
> > > + label = make_irq_label(le->label);
> > > + if (!label) {
> > > + ret = -ENOMEM;
> > > + goto out_free_le;
> > > + }
> > > +
> > > /* Request a thread to read the events */
> > > ret = request_threaded_irq(irq,
> > > lineevent_irq_handler,
> > > lineevent_irq_thread,
> > > irqflags,
> > > - le->label,
> > > + label,
> > > le);
> > > if (ret)
> > > goto out_free_le;
> >
> > Leaks label if the request_threaded_irq() fails.
> >
>
> Ah, dammit, I didn't catch the fact that irq_free() will not return
> the label address if the request failed.
>

You caught it in edge_detector_setup() - you call free_irq_label() if
request_threaded_irq() fails there.

> Nice catch.

The give away here is that the goto targets are the same.
Bit irritated that I didn't catch it earlier - got distracted by the
return code problem.

Anyway, third time lucky.

Cheers,
Kent.