linux-kernel - Re: [PATCH] media: uvcvideo: Filter hw errors while enumerating controls

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20241219104358.GA2510@pendragon.ideasonboard.com>
Date: Thu, 19 Dec 2024 12:43:58 +0200
From: Laurent Pinchart <laurent.pinchart@...asonboard.com>
To: Ricardo Ribalda <ribalda@...omium.org>
Cc: Hans Verkuil <hverkuil@...all.nl>, Hans de Goede <hdegoede@...hat.com>,
	Mauro Carvalho Chehab <mchehab@...nel.org>,
	linux-media@...r.kernel.org, linux-kernel@...r.kernel.org
Subject: Re: [PATCH] media: uvcvideo: Filter hw errors while enumerating
 controls

On Thu, Dec 19, 2024 at 09:18:30AM +0100, Ricardo Ribalda wrote:
> On Thu, 19 Dec 2024 at 00:28, Laurent Pinchart wrote:
> > On Wed, Dec 18, 2024 at 03:38:34PM +0100, Ricardo Ribalda wrote:
> > > On Fri, 13 Dec 2024 at 12:21, Ricardo Ribalda wrote:
> > > >
> > > > To implement VIDIOC_QUERYCTRL, we need to read from the hardware all the
> > > > values that were not cached previously. If that read fails, we used to
> > > > report back the error to the user.
> > > >
> > > > Unfortunately this does not play nice with userspace. When they are
> > > > enumerating the contols, the only expect an error when there are no
> > > > "next" control.
> > > >
> > > > This is probably a corner case, and could be handled in userspace, but
> > > > both v4l2-ctl and yavta fail to enumerate all the controls if we return
> > > > then -EIO during VIDIOC_QUERYCTRL. I suspect that there are tons of
> > > > userspace apps handling this wrongly as well.
> > >
> > > Actually it CANNOT be handled in userspace.
> > >
> > > If we return anything different than 0, the structure is not copied to
> > > userspace:
> >
> > That could be fixed, we do copy data back to userspace in case of
> > failure for some ioctls. I don't think that would be needed though, I
> > believe we can either mark controls as broken in the uvcvideo driver
> > through quirks, or in a dynamic fashion.
> 
> I'd rather not introduce more differences between uvc and the rest of
> the drivers.
> 
> the ctrl framework only seems to return -EINVAL or 0.

True, but that's likely because there's no driver today other than
uvcvideo that needs to interogate the hardware in a way that can fail to
implement QUERYCTRL. I wish we didn't have to, but UVC is special from
that point of view.

Anyway, as mentioned in a separate answer in this mail thread, we may
not need this if we hide the problematic controls from applications.

> > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/media/v4l2-core/v4l2-ioctl.c#n2929
> > > https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/drivers/media/v4l2-core/v4l2-ioctl.c#n3490

-- 
Regards,

Laurent Pinchart