Re: [PATCH v3 01/15] [media] v4l: Document explicit synchronization behaviour

From: Hans Verkuil
Date: Mon Sep 11 2017 - 06:50:34 EST


On 09/07/2017 08:42 PM, Gustavo Padovan wrote:
> From: Gustavo Padovan <gustavo.padovan@xxxxxxxxxxxxx>
>
> Add section to VIDIOC_QBUF about it
>
> v2:
> - mention that fences are files (Hans)
> - rework for the new API
>
> Signed-off-by: Gustavo Padovan <gustavo.padovan@xxxxxxxxxxxxx>
> ---
> Documentation/media/uapi/v4l/vidioc-qbuf.rst | 31 ++++++++++++++++++++++++++++
> 1 file changed, 31 insertions(+)
>
> diff --git a/Documentation/media/uapi/v4l/vidioc-qbuf.rst b/Documentation/media/uapi/v4l/vidioc-qbuf.rst
> index 1f3612637200..fae0b1431672 100644
> --- a/Documentation/media/uapi/v4l/vidioc-qbuf.rst
> +++ b/Documentation/media/uapi/v4l/vidioc-qbuf.rst
> @@ -117,6 +117,37 @@ immediately with an ``EAGAIN`` error code when no buffer is available.
> The struct :c:type:`v4l2_buffer` structure is specified in
> :ref:`buffer`.
>
> +Explicit Synchronization
> +------------------------
> +
> +Explicit Synchronization allows us to control the synchronization of
> +shared buffers from userspace by passing fences to the kernel and/or
> +receiving them from it. Fences passed to the kernel are named in-fences and
> +the kernel should wait them to signal before using the buffer, i.e., queueing

wait them -> wait on them

(do you wait 'on' a fence or 'for' a fence? I think it's 'on' but I'm not 100% sure)

> +it to the driver. On the other side, the kernel can create out-fences for the
> +buffers it queues to the drivers, out-fences signal when the driver is

Start a new sentence here: ...drivers. Out-fences...

> +finished with buffer, that is the buffer is ready. The fence are represented

s/that is/i.e/

s/The fence/The fences/

> +by file and passed as file descriptor to userspace.

s/by file/as a file/
s/as file/as a file/

> +
> +The in-fences are communicated to the kernel at the ``VIDIOC_QBUF`` ioctl
> +using the ``V4L2_BUF_FLAG_IN_FENCE`` buffer
> +flags and the `fence_fd` field. If an in-fence needs to be passed to the kernel,
> +`fence_fd` should be set to the fence file descriptor number and the
> +``V4L2_BUF_FLAG_IN_FENCE`` should be set as well. Failure to set both will

s/Failure to set both/Setting one but not the other/

> +cause ``VIDIOC_QBUF`` to return with error.
> +
> +To get a out-fence back from V4L2 the ``V4L2_BUF_FLAG_OUT_FENCE`` flag should
> +be set to notify it that the next queued buffer should have a fence attached to
> +it. That means the out-fence may not be associated with the buffer in the
> +current ``VIDIOC_QBUF`` ioctl call because the ordering in which videobuf2 core
> +queues the buffers to the drivers can't be guaranteed. To become aware of the
> +of the next queued buffer and the out-fence attached to it the
> +``V4L2_EVENT_BUF_QUEUED`` event should be used. It will trigger an event
> +for every buffer queued to the V4L2 driver.

This makes no sense.

Setting this flag means IMHO that when *this* buffer is queued up to the driver,
then it should send the BUF_QUEUED event with an out fence.

I.e. it signals that userspace wants to have the out-fence. The requirement w.r.t.
ordering is that the BUF_QUEUED events have to be in order, but that is something
that the driver can ensure in the case it is doing internal re-ordering.

This requirement is something that needs to be documented here, BTW.

Anyway, the flag shouldn't refer to some 'next buffer', since that's very confusing.

> +
> +At streamoff the out-fences will either signal normally if the drivers wait

s/drivers wait/driver waits/

> +for the operations on the buffers to finish or signal with error if the
> +driver cancel the pending operations.

s/cancel/cancels/

Thinking with my evil hat on:

What happens if the application dequeues the buffer (VIDIOC_DQBUF) before
dequeuing the BUF_QUEUED event? Or if the application doesn't call VIDIOC_DQEVENT
at all? Should any pending BUF_QUEUED event with an out fence be removed from the
event queue if the application calls DQBUF on the corresponding buffer?

Regards,

Hans

>
> Return Value
> ============
>