Re: [PATCH v7 2/8] media: uvcvideo: Add support for per-device control mapping overrides

[Dan Carpenter]

On Fri, Jul 01, 2022 at 01:16:23AM +0200, Ricardo Ribalda wrote:
> >
> > "mapping" uninitialized if chain->dev->info->mappings is NULL.
> >
> > 10bdca4191d7a8 drivers/media/usb/uvc/uvc_ctrl.c   Ricardo Ribalda  2022-06-17  2457
> > 10bdca4191d7a8 drivers/media/usb/uvc/uvc_ctrl.c   Ricardo Ribalda  2022-06-17  2458     /* Process common mappings next. */
> > 10bdca4191d7a8 drivers/media/usb/uvc/uvc_ctrl.c   Ricardo Ribalda  2022-06-17 @2459     mend = mapping + ARRAY_SIZE(uvc_ctrl_mappings);
> >                                                                                                ^^^^^^^
> >
> > 10bdca4191d7a8 drivers/media/usb/uvc/uvc_ctrl.c   Ricardo Ribalda  2022-06-17  2460     for (mapping = uvc_ctrl_mappings; mapping < mend; ++mapping) {
> >
> > I'm surprised this made it through testing...  Anyway, the easier way to
> > to iterate through an array of structs is just say:
> 
> There is already a new version under review that fixes this:
> 
> https://patchwork.linuxtv.org/project/linux-media/patch/20220617235610.321917-3-ribalda@xxxxxxxxxxxx/ 
> 
> 
> >
> >         for (i = 0; i < ARRAY_SIZE(uvc_ctrl_mappings); i++) {
> >                 mapping = &uvc_ctrl_mappings[i];
> >
> 
> I agree, and that is what I normally used, but that driver iterate this way...

That's not really an excuse.  Imagine you were a lemming and Disney
employees were trying to chase you off a cliff!

> 
> Thanks again!
> 
> PS: Any idea why the test was triggered 14+ days after the patch was sent?

I don't know how kbuild works.  I run Smatch on linux-next so I would
have caught it eventually.  Plus the clang people would have caught it.

regards,
dan carpenter