Re: [PATCH v3 2/4] media: v4l2-common: add helper functions to call s_stream() callbacks

[Helen Koike <helen.koike@collabora.com>]

Hi Dafna,

On 5/21/20 9:03 AM, Dafna Hirschfeld wrote:
> Hi
> 
> On 15.04.20 03:30, Helen Koike wrote:
>> Add v4l2_pipeline_stream_{enable,disable} helper functions to iterate
>> through the subdevices in a given stream (i.e following links from sink
>> to source) and call .s_stream() callback.
>>
>> Add stream_count on the subdevice object for simultaneous streaming from
>> different video devices which shares subdevices.
>>
>> Prevent calling .s_stream(true) if it was already called previously by
>> another stream.
>>
>> Prevent calling .s_stream(false) from one stream when there are still
>> others active.
>>
>> If .s_stream(true) fails, call .s_stream(false) in the reverse order.
>>
>> Signed-off-by: Helen Koike <helen.koike@collabora.com>
>> ---
>>
>> Changes in v3:
>> - re-write helpers to use media walkers as in v1, but with
>> v4l2_pipeline_subdevs_get() to reverse the order we call s_stream(true)
>> in subdevices.
>> - rename size to max_size (and update docs) in v4l2_pipeline_subdevs_get() to
>> reflect the meaning of the argument.
>> - add if defined(CONFIG_MEDIA_CONTROLLER) around helpers
>>
>> Changes in v2:
>> - re-write helpers to not use media walkers
>>
>>   drivers/media/v4l2-core/v4l2-common.c | 125 ++++++++++++++++++++++++++
>>   include/media/v4l2-common.h           |  43 +++++++++
>>   include/media/v4l2-subdev.h           |   2 +
>>   3 files changed, 170 insertions(+)
>>
>> diff --git a/drivers/media/v4l2-core/v4l2-common.c b/drivers/media/v4l2-core/v4l2-common.c
>> index 9e8eb45a5b03c..2f991a1a57d7c 100644
>> --- a/drivers/media/v4l2-core/v4l2-common.c
>> +++ b/drivers/media/v4l2-core/v4l2-common.c
>> @@ -442,3 +442,128 @@ int v4l2_fill_pixfmt(struct v4l2_pix_format *pixfmt, u32 pixelformat,
>>       return 0;
>>   }
>>   EXPORT_SYMBOL_GPL(v4l2_fill_pixfmt);
>> +
>> +#if defined(CONFIG_MEDIA_CONTROLLER)
>> +
>> +/*
>> + * v4l2_pipeline_subdevs_get - Assemble a list of subdevices in a pipeline
>> + * @subdevs: the array to be filled.
>> + * @max_size: max number of elements that can fit in subdevs array.
>> + *
>> + * Walk from a video node, following links from sink to source and fill the
>> + * array with subdevices in the path.
>> + * The walker performs a depth-first traversal, which means that, in a topology
>> + * sd1->sd2->sd3->vdev, sd1 will be the first element placed in the array.
>> + *
>> + * Return the number of subdevices filled in the array.
>> + */
>> +static int v4l2_pipeline_subdevs_get(struct video_device *vdev,
>> +                     struct media_pipeline *pipe,
>> +                     struct v4l2_subdev **subdevs,
>> +                     unsigned int max_size)
>> +{
>> +    struct media_entity *entity = &vdev->entity;
>> +    struct media_device *mdev = entity->graph_obj.mdev;
>> +    int idx = 0;
>> +    int ret;
>> +
>> +    mutex_lock(&mdev->graph_mutex);
>> +
>> +    if (!pipe->streaming_count) {
>> +        ret = media_graph_walk_init(&pipe->graph, mdev);
>> +        if (ret) {
>> +            mutex_unlock(&mdev->graph_mutex);
>> +            return ret;
>> +        }
>> +    }
>> +
>> +    media_graph_walk_start(&pipe->graph, entity);
>> +
>> +    while ((entity = media_graph_walk_next_stream(&pipe->graph))) {
>> +        if (!is_media_entity_v4l2_subdev(entity))
>> +            continue;
>> +        if (WARN_ON(idx >= max_size)) {
>> +            mutex_unlock(&mdev->graph_mutex);
>> +            return -EINVAL;
>> +        }
>> +        subdevs[idx++] = media_entity_to_v4l2_subdev(entity);
>> +    }
>> +
>> +    if (!pipe->streaming_count)
>> +        media_graph_walk_cleanup(&pipe->graph);
>> +
>> +    mutex_unlock(&mdev->graph_mutex);
>> +
>> +    return idx;
>> +}
>> +
>> +__must_check int v4l2_pipeline_stream_enable(struct video_device *vdev,
>> +                         struct media_pipeline *pipe)
>> +{
>> +    struct media_device *mdev = vdev->entity.graph_obj.mdev;
>> +    struct v4l2_subdev *subdevs[MEDIA_ENTITY_ENUM_MAX_DEPTH];
>> +    struct v4l2_subdev *sd;
>> +    int i, size, ret;
>> +
>> +    size = v4l2_pipeline_subdevs_get(vdev, pipe,
>> +                     subdevs, ARRAY_SIZE(subdevs));
>> +    if (size <= 0)
>> +        return size;
>> +
>> +    /* Call s_stream() in reverse order to enable sensors last */
>> +    for (i = size - 1; i >= 0; i--) {
>> +        sd = subdevs[i];
>> +        if (sd->stream_count++)
>> +            continue;
>> +        dev_dbg(mdev->dev,
>> +            "enabling stream for '%s'\n", sd->entity.name);
>> +        ret = v4l2_subdev_call(sd, video, s_stream, true);
>> +        if (ret && ret != -ENOIOCTLCMD) {
>> +            sd->stream_count = 0;
>> +            goto err_stream_disable;
>> +        }
>> +    }
>> +
>> +    return 0;
>> +
>> +err_stream_disable:
>> +    for (i++; i < size; i++) {
>> +        sd = subdevs[i];
>> +        if (--sd->stream_count)
>> +            continue;
>> +        dev_dbg(mdev->dev,
>> +            "disabling stream for '%s'\n", sd->entity.name);
>> +        v4l2_subdev_call(sd, video, s_stream, false);
>> +    };
>> +
>> +    return ret;
>> +}
>> +EXPORT_SYMBOL_GPL(v4l2_pipeline_stream_enable);
>> +
>> +void v4l2_pipeline_stream_disable(struct video_device *vdev,
>> +                  struct media_pipeline *pipe)
>> +{
>> +    struct media_device *mdev = vdev->entity.graph_obj.mdev;
>> +    struct v4l2_subdev *subdevs[MEDIA_ENTITY_ENUM_MAX_DEPTH];
>> +    unsigned int i;
>> +    int size;
>> +
>> +    size = v4l2_pipeline_subdevs_get(vdev, pipe,
>> +                     subdevs, ARRAY_SIZE(subdevs));
>> +    if (WARN_ON(size < 0))
>> +        return;
>> +
>> +    /* Call s_stream() in order to disable sensors first */
>> +    for (i = 0; i < size; i++) {
>> +        struct v4l2_subdev *sd = subdevs[i];
>> +
>> +        if (--sd->stream_count)
>> +            continue;
>> +        dev_dbg(mdev->dev,
>> +            "disabling stream for '%s'\n", sd->entity.name);
>> +        v4l2_subdev_call(sd, video, s_stream, false);
>> +    }
>> +}
>> +EXPORT_SYMBOL_GPL(v4l2_pipeline_stream_disable);
>> +
>> +#endif /* CONFIG_MEDIA_CONTROLLER */
>> diff --git a/include/media/v4l2-common.h b/include/media/v4l2-common.h
>> index 150ee16ebd811..dc46812120cdc 100644
>> --- a/include/media/v4l2-common.h
>> +++ b/include/media/v4l2-common.h
>> @@ -539,4 +539,47 @@ static inline void v4l2_buffer_set_timestamp(struct v4l2_buffer *buf,
>>       buf->timestamp.tv_usec = ts.tv_nsec / NSEC_PER_USEC;
>>   }
>>   +#if defined(CONFIG_MEDIA_CONTROLLER)
>> +
>> +/**
>> + * v4l2_pipeline_stream_enable - Call .s_stream(true) callbacks in the stream
>> + * @vdev: Starting video device.
>> + * @pipe: Pipeline this entity belongs to.
>> + *
>> + * Call .s_stream(true) callback in all the subdevices participating in the
>> + * stream, i.e. following links from sink to source.
>> + *
>> + * .s_stream(true) is also called from sink to source, i.e. in a topology
>> + * sd1->sd2->sd3->vdev, .s_stream(true) of sd3 is called first.
>> + *
>> + * Calls to this function can be nested, in which case the same number of
>> + * v4l2_pipeline_stream_disable() calls will be required to disable streaming
>> + * through subdevices in the pipeline.
>> + * The  pipeline pointer must be identical for all nested calls to
>> + * v4l2_pipeline_stream_enable().
>> + */
>> +__must_check int v4l2_pipeline_stream_enable(struct video_device *vdev,
>> +                         struct media_pipeline *pipe);
>> +
>> +/**
>> + * v4l2_pipeline_stream_disable - Call .s_stream(false) callbacks in the stream
>> + * @vdev: Starting video device.
>> + * @pipe: Pipeline this entity belongs to.
>> + *
>> + * Call .s_stream(false) callback in all the subdevices participating in the
>> + * stream, i.e. following links from sink to source.
>> + *
>> + * s_stream(false) is called in a reverse order from
>> + * v4l2_pipeline_stream_enable(), i.e. in a topology sd1->sd2->sd3->vdev,
>> + * .s_stream(false) of sd1 is called first.
>> + *
>> + * If multiple calls to v4l2_pipeline_stream_enable() have been made, the same
>> + * number of calls to this function are required to disable streaming through
>> + * subdevices in the pipeline.
>> + */
>> +void v4l2_pipeline_stream_disable(struct video_device *vdev,
>> +                  struct media_pipeline *pipe);
>> +
>> +#endif /* CONFIG_MEDIA_CONTROLLER */
>> +
>>   #endif /* V4L2_COMMON_H_ */
>> diff --git a/include/media/v4l2-subdev.h b/include/media/v4l2-subdev.h
>> index a4848de598521..20f913a9f70c5 100644
>> --- a/include/media/v4l2-subdev.h
>> +++ b/include/media/v4l2-subdev.h
>> @@ -838,6 +838,7 @@ struct v4l2_subdev_platform_data {
>>    * @subdev_notifier: A sub-device notifier implicitly registered for the sub-
>>    *             device using v4l2_device_register_sensor_subdev().
>>    * @pdata: common part of subdevice platform data
>> + * @stream_count: Stream count for the subdevice.
>>    *
>>    * Each instance of a subdev driver should create this struct, either
>>    * stand-alone or embedded in a larger struct.
>> @@ -869,6 +870,7 @@ struct v4l2_subdev {
>>       struct v4l2_async_notifier *notifier;
>>       struct v4l2_async_notifier *subdev_notifier;
>>       struct v4l2_subdev_platform_data *pdata;
>> +    unsigned int stream_count;
> I wonder if it is worth it to add a new field to v4l2_subdev that is used only
> for the specific case, it seems that except of rkisp1 and vimc, the other drivers you pointed to that
> could use the helpers don't need this counter and they call s_stream without checking if it was already called.
> This counter is updated only by the new introduced functions and not when s_stream is called from other places.

Other drivers implement their own loop that navigates through the pipeline calling .s_stream().
They are very similar. The only difference from rkisp1/vimc is that they don't support simultaneous streaming from different
capture nodes.

Also, drivers usually don't handle errors very well, in cases of failures where .s_stream() should be called in reversed order.

Adding this counter + helpers allows a common generic implementation that can be used for both cases.
So we re-use a code that should be well tested in the core and avoid re-implementing it.

> Maybe we can add a helper that just return the next subdevice connected to the current entity as a source, and drivers can
> iterate it and can decide themselves if s_stream should be called or not

Patch 1/4 in the series provides this iterator :)
Unless if you think we could add helpers to make things easier.

Regards,
Helen

> 
> Thanks,
> Dafna
> 
>>   };
>>