Re: [PATCH v4 1/2] media: uvcvideo: Support partial control reads

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

On Wed, Nov 27, 2024 at 10:35:54AM +0100, Ricardo Ribalda wrote:
> On Wed, 27 Nov 2024 at 10:14, Laurent Pinchart wrote:
> > On Wed, Nov 27, 2024 at 09:58:21AM +0100, Ricardo Ribalda wrote:
> > > On Wed, 27 Nov 2024 at 09:34, Laurent Pinchart wrote:
> > > > On Tue, Nov 26, 2024 at 07:12:53PM +0100, Ricardo Ribalda wrote:
> > > > > On Tue, 26 Nov 2024 at 19:06, Laurent Pinchart wrote:
> > > > > > On Wed, Nov 20, 2024 at 03:26:19PM +0000, Ricardo Ribalda wrote:
> > > > > > > Some cameras, like the ELMO MX-P3, do not return all the bytes
> > > > > > > requested from a control if it can fit in less bytes.
> > > > > > > Eg: Returning 0xab instead of 0x00ab.
> > > > > > > usb 3-9: Failed to query (GET_DEF) UVC control 3 on unit 2: 1 (exp. 2).
> > > > > > >
> > > > > > > Extend the returned value from the camera and return it.
> > > > > > >
> > > > > > > Cc: stable@vger.kernel.org
> > > > > > > Fixes: a763b9fb58be ("media: uvcvideo: Do not return positive errors in uvc_query_ctrl()")
> > > > > > > Reviewed-by: Hans de Goede <hdegoede@redhat.com>
> > > > > > > Signed-off-by: Ricardo Ribalda <ribalda@chromium.org>
> > > > > > > ---
> > > > > > >  drivers/media/usb/uvc/uvc_video.c | 16 ++++++++++++++++
> > > > > > >  1 file changed, 16 insertions(+)
> > > > > > >
> > > > > > > diff --git a/drivers/media/usb/uvc/uvc_video.c b/drivers/media/usb/uvc/uvc_video.c
> > > > > > > index cd9c29532fb0..482c4ceceaac 100644
> > > > > > > --- a/drivers/media/usb/uvc/uvc_video.c
> > > > > > > +++ b/drivers/media/usb/uvc/uvc_video.c
> > > > > > > @@ -79,6 +79,22 @@ int uvc_query_ctrl(struct uvc_device *dev, u8 query, u8 unit,
> > > > > > >       if (likely(ret == size))
> > > > > > >               return 0;
> > > > > > >
> > > > > > > +     /*
> > > > > > > +      * In UVC the data is usually represented in little-endian.
> > > > > >
> > > > > > I had a comment about this in the previous version, did you ignore it on
> > > > > > purpose because you disagreed, or was it an oversight ?
> > > > >
> > > > > I rephrased the comment. I added "usually" to make it clear that it
> > > > > might not be the case for all the data types. Like composed or xu.
> > > >
> > > > Ah, that's what you meant by "usually". I read it as "usually in
> > > > little-endian, but could be big-endian too", which confused me.
> > > >
> > > > Data types that are not integers will not work nicely with the
> > > > workaround below. How do you envision that being handled ? Do you
> > > > consider that the device will return too few bytes only for integer data
> > > > types, or that affected devices don't have controls that use compound
> > > > data types ? I don't see what else we could do so I'd be fine with such
> > > > a heuristic for this workaround, but it needs to be clearly explained.
> > >
> > > Non integer datatypes might work if the last part of the data is
> > > expected to be zero.
> > > I do not think that we can find a heuristic that can work for all the cases.
> > >
> > > For years we have ignored partial reads and it has never been an
> > > issue. I vote for not adding any heuristics, the logging should help
> > > identify future issues (if there is any).
> >
> > What you're doing below is already a heuristic :-) I don't think the
> > code needs to be changed, but I'd like this comment to explain why we
> > consider that the heuristic in this patch is fine, to help the person
> > (possibly you or me) who will read this code in a year and wonder what's
> > going on.
> 
> What about:
> 
> * Some devices return shorter USB control packets than expected if the
> * returned value can fit in less bytes. Zero all the bytes that the
> * device has not written.
> *
> * This quirk is applied to all datatypes, even to non little-endian integers
> * or composite values. We exclude UVC_GET_INFO from the quirk.
> * UVC_GET_LEN does not need to be excluded because its size is
> * always 1.

For the second paragraph you could write

 * This quirk is applied to all controls, regardless of their data type. Most
 * controls are little-endian integers, in which case the missing bytes become 0
 * MSBs. For other data types, a different heuristic could be implemented if a
 * device is found needing it.
 *
 * We exclude UVC_GET_INFO from the quirk. UVC_GET_LEN does not need to be
 * excluded because its size is always 1.

> > > > > I also r/package/packet/
> > > > >
> > > > > Did I miss another comment?
> > > > >
> > > > > > > +      * Some devices return shorter USB control packets that expected if the
> > > > > > > +      * returned value can fit in less bytes. Zero all the bytes that the
> > > > > > > +      * device have not written.
> > > > > >
> > > > > > s/have/has/
> > > > > >
> > > > > > And if you meant to start a new paragraph here, a blank line is missing.
> > > > > > Otherwise, no need to break the line before 80 columns.
> > > > >
> > > > > The patch is already in the uvc tree. How do you want to handle this?
> > > >
> > > > The branch shared between Hans and me can be rebased, it's a staging
> > > > area.
> > >
> > > I will send a new version, fixing the typo. and the missing new line.
> > > I will also remove the sentence
> > > `* In UVC the data is usually represented in little-endian.`
> > > It is confusing.
> > >
> > > > > > > +      * We exclude UVC_GET_INFO from the quirk. UVC_GET_LEN does not need to
> > > > > > > +      * be excluded because its size is always 1.
> > > > > > > +      */
> > > > > > > +     if (ret > 0 && query != UVC_GET_INFO) {
> > > > > > > +             memset(data + ret, 0, size - ret);
> > > > > > > +             dev_warn_once(&dev->udev->dev,
> > > > > > > +                           "UVC non compliance: %s control %u on unit %u returned %d bytes when we expected %u.\n",
> > > > > > > +                           uvc_query_name(query), cs, unit, ret, size);
> > > > > > > +             return 0;
> > > > > > > +     }
> > > > > > > +
> > > > > > >       if (ret != -EPIPE) {
> > > > > > >               dev_err(&dev->udev->dev,
> > > > > > >                       "Failed to query (%s) UVC control %u on unit %u: %d (exp. %u).\n",

-- 
Regards,

Laurent Pinchart