Re: [PATCH v2 5/8] media: v4l2: Add DELETE_BUF ioctl

[Tomasz Figa]

On Tue, Mar 21, 2023 at 11:28:52AM +0100, Benjamin Gaignard wrote:
> VIDIOC_DELETE_BUF ioctl allows to delete a buffer from a queue.
> 
> Signed-off-by: Benjamin Gaignard <benjamin.gaignard@xxxxxxxxxxxxx>
> ---
>  .../userspace-api/media/v4l/user-func.rst     |  1 +
>  .../media/v4l/vidioc-delete-buf.rst           | 51 ++++++++++++++++
>  .../media/common/videobuf2/videobuf2-core.c   | 59 ++++++++++++++++++-
>  .../media/common/videobuf2/videobuf2-v4l2.c   |  6 ++
>  drivers/media/v4l2-core/v4l2-dev.c            |  1 +
>  drivers/media/v4l2-core/v4l2-ioctl.c          | 10 ++++
>  include/media/v4l2-ioctl.h                    |  4 ++
>  include/media/videobuf2-core.h                |  9 +++
>  include/media/videobuf2-v4l2.h                | 11 ++++
>  include/uapi/linux/videodev2.h                |  1 +
>  10 files changed, 152 insertions(+), 1 deletion(-)
>  create mode 100644 Documentation/userspace-api/media/v4l/vidioc-delete-buf.rst
> 
> diff --git a/Documentation/userspace-api/media/v4l/user-func.rst b/Documentation/userspace-api/media/v4l/user-func.rst
> index 228c1521f190..93e0a6a117fc 100644
> --- a/Documentation/userspace-api/media/v4l/user-func.rst
> +++ b/Documentation/userspace-api/media/v4l/user-func.rst
> @@ -17,6 +17,7 @@ Function Reference
>      vidioc-dbg-g-chip-info
>      vidioc-dbg-g-register
>      vidioc-decoder-cmd
> +    vidioc-delete-buf
>      vidioc-dqevent
>      vidioc-dv-timings-cap
>      vidioc-encoder-cmd
> diff --git a/Documentation/userspace-api/media/v4l/vidioc-delete-buf.rst b/Documentation/userspace-api/media/v4l/vidioc-delete-buf.rst
> new file mode 100644
> index 000000000000..0e7ce58f91bc
> --- /dev/null
> +++ b/Documentation/userspace-api/media/v4l/vidioc-delete-buf.rst
> @@ -0,0 +1,51 @@
> +.. SPDX-License-Identifier: GFDL-1.1-no-invariants-or-later
> +.. c:namespace:: V4L
> +
> +.. _VIDIOC_DELETE_BUF:
> +
> +************************
> +ioctl VIDIOC_DELETE_BUF
> +************************
> +
> +Name
> +====
> +
> +VIDIOC_DELETE_BUF - Delete a buffer from a queue
> +
> +Synopsis
> +========
> +
> +.. c:macro:: VIDIOC_DELETE_BUF
> +
> +``int ioctl(int fd, VIDIOC_DELETE_BUF, struct v4l2_buffer *argp)``
> +
> +Arguments
> +=========
> +
> +``fd``
> +    File descriptor returned by :c:func:`open()`.
> +
> +``argp``
> +    Pointer to struct :c:type:`v4l2_buffer`.

Would struct v4l2_create_buffers make more sense here? With it, we could
actually make this ioctl VIDIOC_DELETE_BUF*S*, consistently with
VIDIOC_CREATE_BUF*S* and allow the user space to remove a block of buffers
the same way it created a block of buffers in the first place.

> +
> +Description
> +===========
> +
> +Applications can optionally call the :ref:`VIDIOC_DELETE_BUF` ioctl to
> +delete a buffer from a queue.
> +
> +The struct :c:type:`v4l2_buffer` structure is specified in
> +:ref:`buffer`.
> +
> +Return Value
> +============
> +
> +On success 0 is returned, on error -1 and the ``errno`` variable is set
> +appropriately. The generic error codes are described at the
> +:ref:`Generic Error Codes <gen-errors>` chapter.
> +
> +EBUSY
> +    File I/O is in progress.
> +
> +EINVAL
> +    The buffer ``index`` doesn't exist in the queue.
> diff --git a/drivers/media/common/videobuf2/videobuf2-core.c b/drivers/media/common/videobuf2/videobuf2-core.c
> index 3c6ced360770..ec241d726fe6 100644
> --- a/drivers/media/common/videobuf2/videobuf2-core.c
> +++ b/drivers/media/common/videobuf2/videobuf2-core.c
> @@ -401,6 +401,24 @@ static void init_buffer_cache_hints(struct vb2_queue *q, struct vb2_buffer *vb)
>  		vb->skip_cache_sync_on_finish = 1;
>  }
>  
> +/*
> + * __vb2_queue_get_free_index() - find a free index in the queue for vb2 buffer.
> + *
> + * Returns an index for vb2 buffer.
> + */
> +static int __vb2_queue_get_free_index(struct vb2_queue *q)
> +{
> +	int i;
> +
> +	spin_lock(&q->bufs_lock);
> +	for (i = 0; i < q->max_num_bufs; i++)
> +		if (!q->bufs[i])
> +			break;
> +	spin_unlock(&q->bufs_lock);
> +
> +	return i;
> +}

Another reason to go with XArray, so that we don't have to open code
index reclaim.

> +
>  /*
>   * __vb2_queue_alloc() - allocate vb2 buffer structures and (for MMAP type)
>   * video buffer memory for all buffers/planes on the queue and initializes the
> @@ -427,7 +445,7 @@ static int __vb2_queue_alloc(struct vb2_queue *q, enum vb2_memory memory,
>  		vb->state = VB2_BUF_STATE_DEQUEUED;
>  		vb->vb2_queue = q;
>  		vb->num_planes = num_planes;
> -		vb->index = q->num_buffers + buffer;
> +		vb->index = __vb2_queue_get_free_index(q);
>  		vb->type = q->type;
>  		vb->memory = memory;
>  		init_buffer_cache_hints(q, vb);
> @@ -1570,6 +1588,45 @@ int vb2_core_prepare_buf(struct vb2_queue *q, unsigned int index, void *pb)
>  }
>  EXPORT_SYMBOL_GPL(vb2_core_prepare_buf);
>  
> +int vb2_core_delete_buf(struct vb2_queue *q, unsigned int index)
> +{
> +	struct vb2_buffer *vb;
> +
> +	vb = vb2_get_buffer(q, index);
> +	if (!vb) {
> +		dprintk(q, 1, "invalid buffer index %d\n", index);
> +		return -EINVAL;
> +	}
> +
> +	if (vb->state == VB2_BUF_STATE_QUEUED) {

How about other states? I'd probably only allow this when the state is
DEQUEUED for simplicity. Is there a need to allow deleting buffers in
other states?

Also, do we need to synchronize this with other contexts which could change
the buffer state?

> +		dprintk(q, 1, "can't delete queued buffer index %d\n", index);
> +		return -EINVAL;
> +	}
> +

Don't we also need to hold q->mmap_lock to prevent racing with an attempt
to mmap the buffer?

> +	if (vb && vb->planes[0].mem_priv)

nit: At this point it's not possible for vb to be NULL, since we already
ruled that out a few lines above.

> +		call_void_vb_qop(vb, buf_cleanup, vb);
> +
> +	/* Free MMAP buffers or release USERPTR buffers */
> +	if (q->memory == VB2_MEMORY_MMAP)
> +		__vb2_buf_mem_free(vb);
> +	else if (q->memory == VB2_MEMORY_DMABUF)
> +		__vb2_buf_dmabuf_put(vb);
> +	else
> +		__vb2_buf_userptr_put(vb);
> +
> +	vb2_queue_remove_buffer(q, vb);
> +	kfree(vb);
> +
> +	q->num_buffers--;
> +	if (!q->num_buffers) {
> +		q->memory = VB2_MEMORY_UNKNOWN;
> +		INIT_LIST_HEAD(&q->queued_list);
> +	}

Would it make sense to refactor __vb2_queue_free() instead to take a
range of buffer indexes rather than duplicating the code here?

Best regards,
Tomasz