Hi Niklas,
Thank you for the patch.
On Wed, Mar 18, 2020 at 10:30:51PM +0100, Niklas Söderlund wrote:
> If the v4l2-async notifier have once been complete and the video
> device(s) have been register, do not unregister them if one subdevice is
> unbound. Depending on which subdevice is unbound other parts of the
> pipeline could still be functional. For example if one of multiple
> sensors connected to a CSI-2 transmitter is unbound other sensors in
> that pipeline are still useable.
>
> This problem is extra critical for R-Car VIN which registers two
> independent CSI-2 receivers in the same media graph as they can both be
> used by the same dma-engines. If one of the CSI-2 receivers are unbound
> the other CSI-2 receiver pipeline is still fully functional.
I like this, but shouldn't you also move registration of the video nodes
to probe time then ?
> Signed-off-by: Niklas Söderlund <niklas.soderlund+renesas@xxxxxxxxxxxx>
> ---
> drivers/media/platform/rcar-vin/rcar-core.c | 5 -----
> 1 file changed, 5 deletions(-)
>
> diff --git a/drivers/media/platform/rcar-vin/rcar-core.c b/drivers/media/platform/rcar-vin/rcar-core.c
> index 21ce3de8168c3224..d51ffe75c34c97c5 100644
> --- a/drivers/media/platform/rcar-vin/rcar-core.c
> +++ b/drivers/media/platform/rcar-vin/rcar-core.c
> @@ -533,7 +533,6 @@ static int rvin_parallel_subdevice_attach(struct rvin_dev *vin,
>
> static void rvin_parallel_subdevice_detach(struct rvin_dev *vin)
> {
> - rvin_v4l2_unregister(vin);
> vin->parallel->subdev = NULL;
>
> if (!vin->info->use_mc) {
> @@ -758,10 +757,6 @@ static void rvin_group_notify_unbind(struct v4l2_async_notifier *notifier,
> struct rvin_dev *vin = v4l2_dev_to_vin(notifier->v4l2_dev);
> unsigned int i;
>
> - for (i = 0; i < RCAR_VIN_NUM; i++)
> - if (vin->group->vin[i])
> - rvin_v4l2_unregister(vin->group->vin[i]);
> -
> mutex_lock(&vin->group->lock);
>
> for (i = 0; i < RVIN_CSI_MAX; i++) {
--
Regards,
Laurent Pinchart
