Re: [PATCH] Input: wacom_w8001 - Ignore bogus idx values in interrupt

From: Dmitry Torokhov
Date: Mon May 23 2016 - 12:52:36 EST


On Sun, May 22, 2016 at 10:21:45PM -0700, Ping Cheng wrote:
> Hi Chris,
>
> On Sun, May 22, 2016 at 6:42 PM, Chris J Arges
> <christopherarges@xxxxxxxxx> wrote:
> > I've noticed crashes when using my x60t using a coreboot bios. When using
> > the pen I can produce a crash simply by tapping a few times. This
> > generates an event which has an idx of 0xc. This in turn crashes the
> > machine because the array access is greater than W8001_MAX_LENGTH. This
> > patch checks for bogus values and filters them in order to prevent crashes.
>
> Thank you for submitting a patch in addition to reporting the issue.
>
> > Signed-off-by: Chris J Arges <christopherarges@xxxxxxxxx>
> > ---
> > drivers/input/touchscreen/wacom_w8001.c | 9 +++++++++
> > 1 file changed, 9 insertions(+)
> >
> > diff --git a/drivers/input/touchscreen/wacom_w8001.c b/drivers/input/touchscreen/wacom_w8001.c
> > index bab3c6a..c858200 100644
> > --- a/drivers/input/touchscreen/wacom_w8001.c
> > +++ b/drivers/input/touchscreen/wacom_w8001.c
> > @@ -283,6 +283,15 @@ static irqreturn_t w8001_interrupt(struct serio *serio,
> > unsigned char tmp;
> >
> > w8001->data[w8001->idx] = data;
> > +
> > + /* ignore bogus idx values */
> > + if (w8001->idx >= W8001_MAX_LENGTH) {
> > + pr_info("w8001: ignored interrupt: data 0x%02x idx %d\n", data,
> > + w8001->idx);
> > + w8001->idx = 0;
> > + return IRQ_HANDLED;
> > + }
> > +
>
> I don't have an x60t system to test with. I wonder if your system
> supports two finger touch or not. We at least have a bug in the code
> since W8001_MAX_LENGTH should be 13 instead of 11. How come no one had
> encountered that issue before?
>
> I'm going to email a patch to the list. Please test it and let us know
> your result. Maybe we still need your patch if your device doesn't
> support two finger touch or the idx=0xc can't be fixed by
> W8001_MAX_LENGTH=13.

Just so we are clear this version of the patch is buggy as we check the
index only after [potentially] writing past the array bounds of
w8001->data[].

Thanks.

--
Dmitry