Hi Michael, On Tue, Dec 06, 2022 at 06:07:21PM +0100, Michael Grzeschik wrote: > On Mon, Dec 05, 2022 at 11:17:15PM +0200, Laurent Pinchart wrote: > >On Sun, Dec 04, 2022 at 09:29:16AM +0100, Greg KH wrote: > >> On Sat, Dec 03, 2022 at 11:26:14PM +0200, Laurent Pinchart wrote: > >> > On Sat, Sep 10, 2022 at 12:13:31AM +0200, Michael Grzeschik wrote: > >> > > This series improves the uvc video gadget by parsing the configfs > >> > > entries. With the configfs data, the userspace now is able to use simple > >> > > v4l2 api calls like enum and try_format to check for valid configurations > >> > > initially set by configfs. > >> > > >> > I've realized that this whole series got merged, despite my multiple > >> > attempts to explain why I think it's not a good idea. The UVC gadget > >> > requires userspace support, and there's no point in trying to move all > >> > these things to the kernel side. It only bloats the kernel, makes the > >> > code more complex, more difficult to maintain, and will make UVC 1.5 > >> > support more difficult. > >> > >> I can easily revert them, but I did not see any objections to them > >> originally and so I merged them as is the normal method :) > > > > I don't think a revert is needed. The issue I pointed out regarding the > > duplication of static const data can be solved on top. The API additions > > from this series are, in my opinion, not a good idea for the reasons I > > explained, but they don't hurt so much that we need to go nuclear on > > this. > > > > Michael, will you be addressing the static const data issue ? > > Yes. I will also move the uvc_fmts[] array and uvc_format_by_guid to its > own compile unit. Thank you. > I will go with drivers/media/usb/uvc.c > > While at it the headerfile will better also be moved from > include/media/v4l2-uvc.h to linux/usb/uvc.h. Works for me, especially for the GUIDs. For the structure and function prototype, I don't mind using include/media/, up to you. -- Regards, Laurent Pinchart
