Re: [PATCH] media: uvcvideo: Fix bounds for descriptor parsing

[Laurent Pinchart <laurent.pinchart@ideasonboard.com>]

On Mon, Sep 28, 2026 at 02:35:29PM +0200, Ricardo Ribalda wrote:
> On Mon, 28 Sept 2026 at 14:24, Laurent Pinchart
> <laurent.pinchart@ideasonboard.com> wrote:
> >
> > On Mon, Sep 28, 2026 at 02:10:26PM +0200, Ricardo Ribalda wrote:
> > > On Mon, 28 Sept 2026 at 14:02, Laurent Pinchart wrote:
> > > > On Fri, Sep 11, 2026 at 01:23:50PM +0000, Ricardo Ribalda wrote:
> > > > > uvc_parse_control() passes descriptor by descriptor to
> > > > > uvc_parse_standard_control() with the number of bytes remaining in the
> > > > > buffer, not the number of bytes of that descriptor.
> > > > >
> > > > > Because of this, malformed descriptors could leak over the next
> > > > > descriptor, leaving malformed data in our structures.
> > > >
> > > > What issue does this fix in practice ?
> > >
> > > Look into uvc_parse_standard_control().
> > > Let's take the UVC_VC_HEADER
> > >
> > > Imagine we have a malformed control at the beginning of the descriptor that has:
> > >
> > > buffer[0] = 3
> > > buffer[2] = UVC_VC_HEADER
> > > buffer[3] => Beggining of next control
> > >
> > > With the current code, buffer[3...X] is parsed as UVC_VC_HEADER.
> > >
> > > What does it fix in practice? It avoids that data outside the controls
> > > is parsed as part of the control.
> >
> > I understand that, but what does it fix in practice ? And have you
> > encountered a device that exhibits such a problem ?
> 
> I have not encountered such devices, but even if they do not exist
> they are easy to emulate and use for escalation.

That's what I'd like to understand, can this be used for any kind of
escalation ?

> We add minimal code and solve a family of bugs. Do you think that it
> is a better pattern that uvc_parse_vendor_control() and
> uvc_parse_vendor_control() have visibility beyond the control that
> they are parsing?

I'm concerned that some devices may stop working. The risk is likely
small though, but I'd like to understand what we get from this patch to
see if it's worth the risk.

> > > > > Change the code so we pass the actual length of the descriptor to the
> > > > > parser.
> > > > >
> > > > > Note that this makes the existing check more strict and some devices
> > > > > that are wrongly parsed today will not be probed now.
> > > > >
> > > > > Cc: stable@vger.kernel.org
> > > > > Fixes: c0efd232929c ("V4L/DVB (8145a): USB Video Class driver")
> > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > > > > ---
> > > > >  drivers/media/usb/uvc/uvc_driver.c | 7 +++++--
> > > > >  1 file changed, 5 insertions(+), 2 deletions(-)
> > > > >
> > > > > diff --git a/drivers/media/usb/uvc/uvc_driver.c b/drivers/media/usb/uvc/uvc_driver.c
> > > > > index e289cc71ba98..429f1ab19a2a 100644
> > > > > --- a/drivers/media/usb/uvc/uvc_driver.c
> > > > > +++ b/drivers/media/usb/uvc/uvc_driver.c
> > > > > @@ -1248,11 +1248,14 @@ static int uvc_parse_control(struct uvc_device *dev)
> > > > >        */
> > > > >
> > > > >       while (buflen > 2) {
> > > > > -             if (uvc_parse_vendor_control(dev, buffer, buflen) ||
> > > > > +             if (buflen < buffer[0] || buffer[0] < 3)
> > > > > +                     return -EINVAL;
> > > > > +
> > > > > +             if (uvc_parse_vendor_control(dev, buffer, buffer[0]) ||
> > > > >                   buffer[1] != USB_DT_CS_INTERFACE)
> > > > >                       goto next_descriptor;
> > > > >
> > > > > -             ret = uvc_parse_standard_control(dev, buffer, buflen);
> > > > > +             ret = uvc_parse_standard_control(dev, buffer, buffer[0]);
> > > > >               if (ret < 0)
> > > > >                       return ret;
> > > > >
> > > > >
> > > > > ---
> > > > > base-commit: 27953c044974baf7e24dee3e9342fe0103dea80c
> > > > > change-id: 20260911-uvc-ctrl-bound-9c1940f06fc7

-- 
Regards,

Laurent Pinchart