On Tue, Oct 24, 2023 at 11:41:57AM +0300, Tomi Valkeinen wrote: > On 24/10/2023 09:55, Eugen Hristev wrote: > > On 10/24/23 00:40, Laurent Pinchart wrote: > >> Hello, > > > > Hello Laurent, > > > >> > >> This series is the result of me getting bothered by the following note > >> in the documentation of the v4l2_subdev_pad_config structure: > >> > >> * Note: This struct is also used in active state, and the 'try' > >> prefix is > >> * historical and to be removed. > >> > >> So I decided to drop the prefix. > >> > >> Patches 1/7 to 6/7 replace direct usage of the fields in drivers with > >> the corresponding accessor functions. There was a relatively large > >> number of them in sensor drivers (in 6/7), but more worryingly, the > >> atmel-isi (1/7), microchip-isc (2/7) and tegra-video (5/7) should really > >> not have messed up with creating a v4l2_subdev_pad_config structure > >> manually. I urge the maintainers of those drivers to address the issue. > > > > Could you hint a bit about how the issue should be addressed ? > > Instead of creating a `v4l2_subdev_pad_config`, one should use > > v4l2_subdev_lock_and_get_active_state() ? Is this the right way to do it ? > > I had a quick look at atmel-isi. If I understand it right, it does not > expose the subdevs to the userspace, and 'isi->entity.subdev' refers to > the sensor. > > In that case I think using v4l2_subdev_call_state_active() and > v4l2_subdev_call_state_try() should usually work. > > If there are cases where the same try state needs to be used for > multiple calls, then the state management has to be done manually with > __v4l2_subdev_state_alloc() and __v4l2_subdev_state_free() (e.g. > drivers/media/platform/renesas/rcar-vin/rcar-v4l2.c). And for the microchip-isc driver, my understanding is that it's an MC-centric driver. It should therefore not call the sensor's .enum_frame_size() operation, which would remove the stack-allocated state in the isc_validate() function. -- Regards, Laurent Pinchart
