Em Fri, 18 Mar 2016 16:20:19 +0200 Sakari Ailus <sakari.ailus@xxxxxxxxxxxxxxx> escreveu: > Shuah Khan wrote: > > On 03/18/2016 08:12 AM, Javier Martinez Canillas wrote: > >> Hello Shuah, > >> > >> On 03/18/2016 11:01 AM, Shuah Khan wrote: > >>> On 03/18/2016 07:05 AM, Mauro Carvalho Chehab wrote: > >>>> Now that media_device_unregister() also does a cleanup, rename it > >>>> to media_device_unregister_cleanup(). > >>>> > >>>> Signed-off-by: Mauro Carvalho Chehab <mchehab@xxxxxxxxxxxxxxx> > >>> > >>> I think adding cleanup is redundant. media_device_unregister() > >>> would imply that there has to be some cleanup releasing resources. > >>> I wouldn't make this change. > >>> > >> > >> Problem is that there is a media_device_init() and media_device_register(), > >> so having both unregister and cleanup in this function will make very clear > >> that a single function is the counter part of the previous two operations. > >> > > > > Yes. I realized that this change is motivated by the fact that there is > > the media_device_init() and we had the counterpart media_device_cleanup() > > as an exported function. I still think there is no need to make the change > > to add _cleanup() at the end of media_device_unregister(). It can be handled > > in API documentation that it does both. > > I think that's a bad idea. People will only read the documentation when > something doesn't work. In this case it's easy to miss that. After thinking about that, I guess the best is to use kref only if the media_device_*devres functions are used. With this, we don't need to touch at media_device_cleanup(). Just the patch to the ML: https://patchwork.linuxtv.org/patch/33533/ It was tested with HVR-950Q. Regards, Mauro
