Re: [PATCH v5 1/2] media: uvcvideo: Refactor streamon/streamoff

[Laurent Pinchart]

Hi Ricardo,

Thank you for the patch.

On Tue, Dec 06, 2022 at 03:06:55PM +0100, Ricardo Ribalda wrote:
> Add a new variable to handle the streaming state and handle the
> streamoff errors, that were not handled before.
> 
> Suggested-by: Laurent Pinchart <laurent.pinchart@xxxxxxxxxxxxxxxx>
> Reviewed-by: Max Staudt <mstaudt@xxxxxxxxxxxx>
> Signed-off-by: Ricardo Ribalda <ribalda@xxxxxxxxxxxx>
> ---
>  drivers/media/usb/uvc/uvc_v4l2.c | 19 ++++++++++++++++---
>  drivers/media/usb/uvc/uvcvideo.h |  1 +
>  2 files changed, 17 insertions(+), 3 deletions(-)
> 
> diff --git a/drivers/media/usb/uvc/uvc_v4l2.c b/drivers/media/usb/uvc/uvc_v4l2.c
> index f4d4c33b6dfb..1389a87b8ae1 100644
> --- a/drivers/media/usb/uvc/uvc_v4l2.c
> +++ b/drivers/media/usb/uvc/uvc_v4l2.c
> @@ -840,13 +840,19 @@ static int uvc_ioctl_streamon(struct file *file, void *fh,
>  {
>  	struct uvc_fh *handle = fh;
>  	struct uvc_streaming *stream = handle->stream;
> -	int ret;
> +	int ret = -EBUSY;
>  
>  	if (!uvc_has_privileges(handle))
>  		return -EBUSY;
>  
>  	mutex_lock(&stream->mutex);
> +
> +	if (handle->is_streaming)
> +		goto unlock;

This isn't needed, uvc_queue_streamon() calls vb2_streamon(), which
returns an error if the queue is already streaming.

>  	ret = uvc_queue_streamon(&stream->queue, type);
> +	handle->is_streaming = !ret;

You can just turn this into

	if (!ret)
		handle->is_streaming = true;

and drop all other changes to this function.

> +
> +unlock:
>  	mutex_unlock(&stream->mutex);
>  
>  	return ret;
> @@ -857,15 +863,22 @@ static int uvc_ioctl_streamoff(struct file *file, void *fh,
>  {
>  	struct uvc_fh *handle = fh;
>  	struct uvc_streaming *stream = handle->stream;
> +	int ret = 0;
>  
>  	if (!uvc_has_privileges(handle))
>  		return -EBUSY;
>  
>  	mutex_lock(&stream->mutex);
> -	uvc_queue_streamoff(&stream->queue, type);
> +
> +	if (!handle->is_streaming)
> +		goto unlock;

More than unneeded, this is wrong. Calling VIDIOC_STREAMOFF on a queue
that is not streaming is a valid use case, it's the only way to release
buffers that have been queued to the device. vb2_core_streamoff()
handles this correctly. You should drop this check.

> +	ret = uvc_queue_streamoff(&stream->queue, type);
> +	handle->is_streaming = !!ret;

And turn this into

	if (!ret)
		handle->is_streaming = false;

and drop all other changes to this function.

> +
> +unlock:
>  	mutex_unlock(&stream->mutex);
>  
> -	return 0;
> +	return ret;
>  }
>  
>  static int uvc_ioctl_enum_input(struct file *file, void *fh,
> diff --git a/drivers/media/usb/uvc/uvcvideo.h b/drivers/media/usb/uvc/uvcvideo.h
> index df93db259312..0d9f335053b8 100644
> --- a/drivers/media/usb/uvc/uvcvideo.h
> +++ b/drivers/media/usb/uvc/uvcvideo.h
> @@ -584,6 +584,7 @@ struct uvc_fh {
>  	struct uvc_video_chain *chain;
>  	struct uvc_streaming *stream;
>  	enum uvc_handle_state state;
> +	bool is_streaming;
>  };
>  
>  struct uvc_driver {
> 

-- 
Regards,

Laurent Pinchart