Hi Niklas,

Thank you for the patch.

On Monday, 29 January 2018 18:34:31 EET Niklas Söderlund wrote:
> Add the ability to process media device link change request. Link
> enabling is a bit complicated on Gen3, whether or not it's possible to
> enable a link depends on what other links already are enabled. On Gen3
> the 8 VINs are split into two subgroup's (VIN0-3 and VIN4-7) and from a
> routing perspective these two groups are independent of each other.
> Each subgroup's routing is controlled by the subgroup VIN master
> instance (VIN0 and VIN4).
> 
> There are a limited number of possible route setups available for each
> subgroup and the configuration of each setup is dictated by the
> hardware. On H3 for example there are 6 possible route setups for each
> subgroup to choose from.
> 
> This leads to the media device link notification code being rather large
> since it will find the best routing configuration to try and accommodate
> as many links as possible. When it's not possible to enable a new link
> due to hardware constrains the link_notifier callback will return
> -EMLINK.
> 
> Signed-off-by: Niklas Söderlund <niklas.soderlund+rene...@ragnatech.se>
> ---
>  drivers/media/platform/rcar-vin/rcar-core.c | 129 +++++++++++++++++++++++++
>  1 file changed, 129 insertions(+)
> 
> diff --git a/drivers/media/platform/rcar-vin/rcar-core.c
> b/drivers/media/platform/rcar-vin/rcar-core.c index
> f08277a0dc11f477..7ceff0de40078580 100644
> --- a/drivers/media/platform/rcar-vin/rcar-core.c
> +++ b/drivers/media/platform/rcar-vin/rcar-core.c
> @@ -24,6 +24,7 @@
> 
>  #include <media/v4l2-async.h>
>  #include <media/v4l2-fwnode.h>
> +#include <media/v4l2-mc.h>
> 
>  #include "rcar-vin.h"
> 
> @@ -44,6 +45,133 @@
>   */
>  #define rvin_group_id_to_master(vin) ((vin) < 4 ? 0 : 4)
> 
> +/* ------------------------------------------------------------------------
> + * Media Controller link notification
> + */
> +
> +/* group lock should be held when calling this function */
> +static int rvin_group_entity_to_csi_id(struct rvin_group *group,
> +                                     struct media_entity *entity)
> +{
> +     struct v4l2_subdev *sd;
> +     int i;

unsigned int.

> +
> +     if (!is_media_entity_v4l2_subdev(entity))
> +             return -ENODEV;

Can this happen ?

> +     sd = media_entity_to_v4l2_subdev(entity);
> +
> +     for (i = 0; i < RVIN_CSI_MAX; i++)
> +             if (group->csi[i].subdev == sd)
> +                     return i;
> +
> +     return -ENODEV;
> +}
> +
> +static unsigned int rvin_group_get_mask(struct rvin_dev *vin,
> +                                     enum rvin_csi_id csi_id,
> +                                     unsigned char chan)
> +{
> +     const struct rvin_group_route *route;
> +     unsigned int mask = 0;
> +
> +     for (route = vin->info->routes; route->mask; route++) {

Please document the fact that the array needs to be terminated by an empty 
element in the kerneldoc for vin->info->routes.

> +             if (route->vin == vin->id &&
> +                 route->csi == csi_id &&
> +                 route->chan == chan) {
> +                     vin_dbg(vin, "Adding route: vin: %d csi: %d chan: %d\n",
> +                             route->vin, route->csi, route->chan);
> +                     mask |= route->mask;
> +             }
> +     }
> +
> +     return mask;
> +}
> +
> +static int rvin_group_link_notify(struct media_link *link, u32 flags,
> +                               unsigned int notification)
> +{
> +     struct rvin_group *group = container_of(link->graph_obj.mdev,
> +                                             struct rvin_group, mdev);
> +     unsigned int i, master_id, chan, mask_new, mask = ~0;

I'm sure you could spare a few more lines :-)

> +     struct media_entity *entity;
> +     struct video_device *vdev;
> +     struct media_pad *csi_pad;
> +     struct rvin_dev *vin = NULL;
> +     int csi_id, ret;
> +
> +     ret = v4l2_pipeline_link_notify(link, flags, notification);
> +     if (ret)
> +             return ret;
> +
> +     /* Only care about link enablement for VIN nodes */
> +     if (!(flags & MEDIA_LNK_FL_ENABLED) ||
> +         !is_media_entity_v4l2_video_device(link->sink->entity))
> +             return 0;
> +
> +     /* If any entity are in use don't allow link changes */

s/are/is/

> +     media_device_for_each_entity(entity, &group->mdev)
> +             if (entity->use_count)
> +                     return -EBUSY;
> +
> +     mutex_lock(&group->lock);
> +
> +     /* Find VIN and its master for which the link */

The words make sense individually. Maybe you could try a different order ? :-)

> +     entity = link->sink->entity;
> +     vdev = media_entity_to_video_device(entity);

You can combine those two lines and remove the entity variable.

> +     for (i = 0; i < RCAR_VIN_NUM; i++) {
> +             if (group->vin[i] && &group->vin[i]->vdev == vdev) {
> +                     vin = group->vin[i];
> +                     master_id = rvin_group_id_to_master(vin->id);
> +                     break;
> +             }
> +     }
> +     if (WARN_ON(!vin || !group->vin[master_id])) {
> +             ret = -ENODEV;
> +             goto out;
> +     }
> +
> +     /* Build a mask for already enabled links */
> +     for (i = master_id; i < master_id + 4; i++) {
> +             if (!group->vin[i])
> +                     continue;
> +
> +             /* Get remote CSI-2, if any */
> +             csi_pad = media_entity_remote_pad(
> +                             &group->vin[i]->vdev.entity.pads[0]);
> +             if (!csi_pad)
> +                     continue;
> +
> +             csi_id = rvin_group_entity_to_csi_id(group, csi_pad->entity);
> +             chan = rvin_group_csi_pad_to_chan(csi_pad->index);
> +
> +             mask &= rvin_group_get_mask(group->vin[i], csi_id, chan);
> +     }
> +
> +     /* Add the new link to the existing mask and check if it works */
> +     csi_id = rvin_group_entity_to_csi_id(group, link->source->entity);
> +     chan = rvin_group_csi_pad_to_chan(link->source->index);
> +     mask_new = mask & rvin_group_get_mask(vin, csi_id, chan);
> +
> +     vin_dbg(vin, "Try link change mask: 0x%x new: 0x%x\n", mask, mask_new);
> +
> +     if (!mask_new) {
> +             ret = -EMLINK;
> +             goto out;
> +     }
> +
> +     /* New valid CHSEL found, set the new value */
> +     rvin_set_channel_routing(group->vin[master_id], __ffs(mask_new));
> +out:
> +     mutex_unlock(&group->lock);
> +
> +     return ret;

I assume this work, but without knowing what the mask field is, it's hard to 
comment. As stated in the review of a previous patch please document the mask 
field correctly, and please explain the algorithm at the beginning of this 
function.

> +}
> +
> +static const struct media_device_ops rvin_media_ops = {
> +     .link_notify = rvin_group_link_notify,
> +};
> +
>  /* ------------------------------------------------------------------------
>   * Gen3 CSI2 Group Allocator
>   */
> @@ -85,6 +213,7 @@ static int rvin_group_init(struct rvin_group *group,
> struct rvin_dev *vin) vin_dbg(vin, "found %u enabled VIN's in DT",
> group->count);
> 
>       mdev->dev = vin->dev;
> +     mdev->ops = &rvin_media_ops;
> 
>       match = of_match_node(vin->dev->driver->of_match_table,
>                             vin->dev->of_node);

-- 
Regards,

Laurent Pinchart

Reply via email to