[v3,2/2] media: docs-rst: Document memory-to-memory video encoder interface

Commit Message

Tomasz Figa Jan. 24, 2019, 10:04 a.m. UTC

Due to complexity of the video encoding process, the V4L2 drivers of
stateful encoder hardware require specific sequences of V4L2 API calls
to be followed. These include capability enumeration, initialization,
encoding, encode parameters change, drain and reset.

Specifics of the above have been discussed during Media Workshops at
LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
Conference Europe 2014 in Düsseldorf. The de facto Codec API that
originated at those events was later implemented by the drivers we already
have merged in mainline, such as s5p-mfc or coda.

The only thing missing was the real specification included as a part of
Linux Media documentation. Fix it now and document the encoder part of
the Codec API.

Signed-off-by: Tomasz Figa <[e-mail]>
---
 Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
 Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
 Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
 Documentation/media/uapi/v4l/v4l2.rst         |   2 +
 .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
 5 files changed, 617 insertions(+), 15 deletions(-)
 create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst

Comments

Hans Verkuil Jan. 29, 2019, 1:52 p.m. UTC | #1

Hi Tomasz,

Some comments below. Nothing major, so I think a v4 should be ready to be
merged.

On 1/24/19 11:04 AM, Tomasz Figa wrote:
> Due to complexity of the video encoding process, the V4L2 drivers of
> stateful encoder hardware require specific sequences of V4L2 API calls
> to be followed. These include capability enumeration, initialization,
> encoding, encode parameters change, drain and reset.
> 
> Specifics of the above have been discussed during Media Workshops at
> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> originated at those events was later implemented by the drivers we already
> have merged in mainline, such as s5p-mfc or coda.
> 
> The only thing missing was the real specification included as a part of
> Linux Media documentation. Fix it now and document the encoder part of
> the Codec API.
> 
> Signed-off-by: Tomasz Figa <[e-mail]>
> ---
>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>  5 files changed, 617 insertions(+), 15 deletions(-)
>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> 
> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> new file mode 100644
> index 000000000000..fb8b05a132ee
> --- /dev/null
> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> @@ -0,0 +1,586 @@
> +.. -*- coding: utf-8; mode: rst -*-
> +
> +.. _encoder:
> +
> +*************************************************
> +Memory-to-memory Stateful Video Encoder Interface
> +*************************************************
> +
> +A stateful video encoder takes raw video frames in display order and encodes
> +them into a bitstream. It generates complete chunks of the bitstream, including
> +all metadata, headers, etc. The resulting bitstream does not require any
> +further post-processing by the client.
> +
> +Performing software stream processing, header generation etc. in the driver
> +in order to support this interface is strongly discouraged. In case such
> +operations are needed, use of the Stateless Video Encoder Interface (in
> +development) is strongly advised.
> +
> +Conventions and notation used in this document
> +==============================================
> +
> +1. The general V4L2 API rules apply if not specified in this document
> +   otherwise.
> +
> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> +
> +3. All steps not marked "optional" are required.
> +
> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> +   unless specified otherwise.
> +
> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> +   used interchangeably with multi-planar API, unless specified otherwise,
> +   depending on decoder capabilities and following the general V4L2 guidelines.
> +
> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> +   [0..2]: i = 0, 1, 2.
> +
> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> +   queue containing data that resulted from processing buffer A.
> +
> +Glossary
> +========
> +
> +Refer to :ref:`decoder-glossary`.
> +
> +State machine
> +=============
> +
> +.. kernel-render:: DOT
> +   :alt: DOT digraph of encoder state machine
> +   :caption: Encoder state machine
> +
> +   digraph encoder_state_machine {
> +       node [shape = doublecircle, label="Encoding"] Encoding;
> +
> +       node [shape = circle, label="Initialization"] Initialization;
> +       node [shape = circle, label="Stopped"] Stopped;
> +       node [shape = circle, label="Drain"] Drain;
> +       node [shape = circle, label="Reset"] Reset;
> +
> +       node [shape = point]; qi
> +       qi -> Initialization [ label = "open()" ];
> +
> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> +
> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> +       Encoding -> Encoding;
> +
> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +
> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> +
> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +   }
> +
> +Querying capabilities
> +=====================
> +
> +1. To enumerate the set of coded formats supported by the encoder, the
> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> +
> +   * The full set of supported formats will be returned, regardless of the
> +     format set on ``OUTPUT``.
> +
> +2. To enumerate the set of supported raw formats, the client may call
> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> +
> +   * Only the formats supported for the format currently active on ``CAPTURE``
> +     will be returned.
> +
> +   * In order to enumerate raw formats supported by a given coded format,
> +     the client must first set that coded format on ``CAPTURE`` and then
> +     enumerate the formats on ``OUTPUT``.
> +
> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> +   resolutions for a given format, passing desired pixel format in
> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> +     format will include all possible coded resolutions supported by the
> +     encoder for given coded pixel format.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> +     will include all possible frame buffer resolutions supported by the
> +     encoder for given raw pixel format and coded format currently set on
> +     ``CAPTURE``.
> +
> +4. Supported profiles and levels for the coded format currently set on
> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> +   via :c:func:`VIDIOC_QUERYCTRL`.
> +
> +5. Any additional encoder capabilities may be discovered by querying
> +   their respective controls.
> +
> +Initialization
> +==============
> +
> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> +
> +     ``pixelformat``
> +         the coded format to be produced
> +
> +     ``sizeimage``
> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> +         match hardware requirements
> +
> +     ``width``, ``height``
> +         ignored (always zero)
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``sizeimage``
> +         adjusted size of ``CAPTURE`` buffers
> +
> +   .. important::
> +
> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> +      format. The encoder will derive a new ``OUTPUT`` format from the
> +      ``CAPTURE`` format being set, including resolution, colorimetry
> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> +      must adjust it afterwards.

Hmm, "including resolution": if width and height are set to 0, what should the
OUTPUT resolution be? Up to the driver? I think this should be clarified since
at a first reading of this paragraph it appears to be contradictory.

> +
> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``pixelformat``
> +         raw format supported for the coded format currently selected on
> +         the ``CAPTURE`` queue.
> +
> +     other fields
> +         follow standard semantics
> +
> +3. Set the raw source format on the ``OUTPUT`` queue via
> +   :c:func:`VIDIOC_S_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``pixelformat``
> +         raw format of the source
> +
> +     ``width``, ``height``
> +         source resolution
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``width``, ``height``
> +         may be adjusted by encoder to match alignment requirements, as
> +         required by the currently selected formats

What if the width x height is larger than the maximum supported by the
selected coded format? This should probably mention that in that case the
width x height is reduced to the largest allowed value. Also mention that
this maximum is reported by VIDIOC_ENUM_FRAMESIZES.

> +
> +     other fields
> +         follow standard semantics
> +
> +   * Setting the source resolution will reset the selection rectangles to their
> +     default values, based on the new resolution, as described in the step 5

5 -> 4

Or just say: "as described in the next step."

> +     below.
> +
> +4. **Optional.** Set the visible resolution for the stream metadata via
> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``target``
> +         set to ``V4L2_SEL_TGT_CROP``
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> +         rectangle and may be subject to adjustment to match codec and
> +         hardware constraints
> +
> +   * **Return fields:**
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle adjusted by the encoder
> +
> +   * The following selection targets are supported on ``OUTPUT``:
> +
> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> +         equal to the full source frame, matching the active ``OUTPUT``
> +         format
> +
> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> +
> +     ``V4L2_SEL_TGT_CROP``
> +         rectangle within the source buffer to be encoded into the
> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> +
> +         .. note::
> +
> +            A common use case for this selection target is encoding a source
> +            video with a resolution that is not a multiple of a macroblock,
> +            e.g.  the common 1920x1080 resolution may require the source
> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> +            size. To avoid encoding the padding, the client needs to explicitly
> +            configure this selection target to 1920x1080.
> +
> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> +         maximum rectangle within the coded resolution, which the cropped
> +         source frame can be composed into; if the hardware does not support
> +         composition or scaling, then this is always equal to the rectangle of
> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> +         equal to a rectangle of width and height matching
> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE``
> +         rectangle within the coded frame, which the cropped source frame
> +         is to be composed into; defaults to
> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> +         additional compose/scaling capabilities; resulting stream will
> +         have this rectangle encoded as the visible rectangle in its
> +         metadata
> +
> +   .. warning::
> +
> +      The encoder may adjust the crop/compose rectangles to the nearest
> +      supported ones to meet codec and hardware requirements. The client needs
> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> +
> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> +         ``CAPTURE``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +          actual number of buffers allocated
> +
> +   .. warning::
> +
> +      The actual number of allocated buffers may differ from the ``count``
> +      given. The client must check the updated value of ``count`` after the
> +      call returns.
> +
> +   .. note::
> +
> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> +      control to get the minimum number of buffers required, and pass the
> +      obtained value plus the number of additional buffers needed in the
> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> +
> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> +   control over buffer allocation.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +         adjusted to the number of allocated buffers
> +
> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> +   encoding process starts when both queues start streaming.
> +
> +.. note::
> +
> +   If the client stops the ``CAPTURE`` queue during the encode process and then
> +   restarts it again, the encoder will begin generating a stream independent
> +   from the stream generated before the stop. The exact constraints depend
> +   on the coded format, but may include the following implications:
> +
> +   * encoded frames produced after the restart must not reference any
> +     frames produced before the stop, e.g. no long term references for
> +     H.264,
> +
> +   * any headers that must be included in a standalone stream must be
> +     produced again, e.g. SPS and PPS for H.264.
> +
> +Encoding
> +========
> +
> +This state is reached after the `Initialization` sequence finishes
> +successfully.  In this state, the client queues and dequeues buffers to both
> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> +standard semantics.
> +
> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> +format and may be affected by codec-specific extended controls, as stated
> +in the documentation of each format.
> +
> +Both queues operate independently, following standard behavior of V4L2 buffer
> +queues and memory-to-memory devices. In addition, the order of encoded frames
> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> +e.g. frame reordering.
> +
> +The client must not assume any direct relationship between ``CAPTURE`` and
> +``OUTPUT`` buffers and any specific timing of buffers becoming
> +available to dequeue. Specifically:
> +
> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> +  frame that preceded it in display, but succeeded it in the decode order),
> +
> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> +  ``CAPTURE`` later into encode process, and/or after processing further
> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> +  reordering is used,
> +
> +* buffers may become available on the ``CAPTURE`` queue without additional
> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> +  available at later time, due to specifics of the decoding process,
> +
> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> +  encoder needs to use the frame as a reference for encoding further frames.
> +
> +.. note::
> +
> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> +   originated from, the client can set the ``timestamp`` field of the
> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> +   will have their ``timestamp`` field set to the same value when dequeued.
> +
> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> +   one ``CAPTURE`` buffer, the following cases are defined:
> +
> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> +
> +   * the encoding order differs from the presentation order (i.e. the
> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> +     and thus monotonicity of the timestamps cannot be guaranteed.

Drop "and thus monotonicity of the timestamps cannot be guaranteed": this was never
guaranteed since timestamps are just copied from OUTPUT to CAPTURE.

This phrase doesn't occur in the decoder spec, so it should be removed here as well.

> +
> +.. note::
> +
> +   To let the client distinguish between frame types (keyframes, intermediate
> +   frames; the exact list of types depends on the coded format), the
> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> +   and their meanings.
> +
> +Encoding parameter changes
> +==========================
> +
> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> +parameters at any time. The availability of parameters is encoder-specific
> +and the client must query the encoder to find the set of available controls.
> +
> +The ability to change each parameter during encoding is encoder-specific, as
> +per the standard semantics of the V4L2 control interface. The client may
> +attempt to set a control during encoding and if the operation fails with the
> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> +configuration change to be allowed. To do this, it may follow the `Drain`
> +sequence to avoid losing the already queued/encoded frames.
> +
> +The timing of parameter updates is encoder-specific, as per the standard
> +semantics of the V4L2 control interface. If the client needs to apply the
> +parameters exactly at specific frame, using the Request API
> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> +
> +Drain
> +=====
> +
> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> +drain sequence described below. After the drain sequence ends, the client has
> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> +sequence was started.
> +
> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> +
> +   * **Required fields:**
> +
> +     ``cmd``
> +         set to ``V4L2_ENC_CMD_STOP``
> +
> +     ``flags``
> +         set to 0
> +
> +     ``pts``
> +         set to 0
> +
> +   .. warning::
> +
> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> +      queues are streaming. For compatibility reasons, the call to
> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> +      not streaming, but at the same time it will not initiate the `Drain`
> +      sequence and so the steps described below would not be applicable.
> +
> +2. Any ``OUTPUT`` buffers queued by the client before the
> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> +   normal. The client must continue to handle both queues independently,
> +   similarly to normal encode operation. This includes:
> +
> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> +
> +     .. warning::
> +
> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> +        as it does not contain an encoded frame.
> +
> +     .. note::
> +
> +        Any attempt to dequeue more buffers beyond the buffer marked with
> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> +        :c:func:`VIDIOC_DQBUF`.
> +
> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> +
> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> +
> +   .. note::
> +
> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> +      event when the last frame has been decoded and all frames are ready to be
> +      dequeued. It is deprecated behavior and the client must not rely on it.
> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> +
> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> +   until the client issues any of the following operations:
> +
> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> +     operation normally, with all the state from before the drain,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> +     and then resume encoding,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> +
> +.. note::
> +
> +   Once the drain sequence is initiated, the client needs to drive it to
> +   completion, as described by the steps above, unless it aborts the process by
> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> +   will fail with -EBUSY error code if attempted.
> +
> +   Although mandatory, the availability of encoder commands may be queried
> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> +
> +Reset
> +=====
> +
> +The client may want to request the encoder to reinitialize the encoding, so
> +that the following stream data becomes independent from the stream data
> +generated before. Depending on the coded format, that may imply that:
> +
> +* encoded frames produced after the restart must not reference any frames
> +  produced before the stop, e.g. no long term references for H.264,
> +
> +* any headers that must be included in a standalone stream must be produced
> +  again, e.g. SPS and PPS for H.264.
> +
> +This can be achieved by performing the reset sequence.
> +
> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
> +   and respective buffers are dequeued.
> +
> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
> +   will return all currently queued ``CAPTURE`` buffers to the client, without
> +   valid frame data.
> +
> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
> +   continue with regular encoding sequence. The encoded frames produced into
> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
> +   decoded without the need for frames encoded before the reset sequence,
> +   starting at the first ``OUTPUT`` buffer queued after issuing the
> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
> +
> +This sequence may be also used to change encoding parameters for encoders
> +without the ability to change the parameters on the fly.
> +
> +Commit points
> +=============
> +
> +Setting formats and allocating buffers triggers changes in the behavior of the
> +encoder.
> +
> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
> +   previously set format being preserved.
> +
> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
> +   supported for the current ``CAPTURE`` format.
> +
> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
> +   format that is not supported for the currently selected ``CAPTURE`` format
> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
> +   supported one.
> +
> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.

irrespectively -> irrespective

> +
> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
> +   change the format on the queue. Drivers will return the -EBUSY error code
> +   for any such format change attempt.
> +
> +To summarize, setting formats and allocation must always start with the
> +``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
> +set of supported formats for the ``OUTPUT`` queue.
> diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> index a0e06a88c872..82630ea89b16 100644
> --- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
> +++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> @@ -17,6 +17,7 @@ Video Memory-To-Memory Interface
>      :maxdepth: 1
>  
>      dev-decoder
> +    dev-encoder
>  
>  A V4L2 memory-to-memory device can compress, decompress, transform, or
>  otherwise convert video data from one format into another format, in memory.
> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> index caf14e440447..593d25b935bf 100644
> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> @@ -44,6 +44,11 @@ Single-planar format structure
>  	inside the stream, when fed to a stateful mem2mem decoder, the fields
>  	may be zero to rely on the decoder to detect the right values. For more
>  	details see :ref:`decoder` and format descriptions.
> +
> +	For compressed formats on the CAPTURE side of a stateful mem2mem
> +	encoder, the fields must be zero, since the coded size is expected to
> +	be calculated internally by the encoder itself, based on the OUTPUT
> +	side. For more details see :ref:`encoder` and format descriptions.
>      * - __u32
>        - ``pixelformat``
>        - The pixel format or type of compression, set by the application.
> diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
> index 97015b9b40b8..e236ea23481b 100644
> --- a/Documentation/media/uapi/v4l/v4l2.rst
> +++ b/Documentation/media/uapi/v4l/v4l2.rst
> @@ -63,6 +63,7 @@ Authors, in alphabetical order:
>  - Figa, Tomasz <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - H Schimek, Michael <[e-mail]>
>  
> @@ -75,6 +76,7 @@ Authors, in alphabetical order:
>  - Osciak, Pawel <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - Osciak, Pawel <[e-mail]>
>  
> diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> index c313ca8b8cb5..2b59a4e7f5ce 100644
> --- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> +++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> @@ -57,19 +57,23 @@ currently only used by the STOP command and contains one bit: If the
>  until the end of the current *Group Of Pictures*, otherwise it will stop
>  immediately.
>  
> -A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> -call sends an implicit START command to the encoder if it has not been
> -started yet. After a STOP command, :ref:`read() <func-read>` calls will read
> +After a STOP command, :ref:`read() <func-read>` calls will read
>  the remaining data buffered by the driver. When the buffer is empty,
>  :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
>  call will restart the encoder.
>  
> +A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> +call sends an implicit START command to the encoder if it has not been
> +started yet. Applies to both queues of mem2mem encoders.
> +
>  A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
>  call of a streaming file descriptor sends an implicit immediate STOP to
> -the encoder, and all buffered data is discarded.
> +the encoder, and all buffered data is discarded. Applies to both queues of
> +mem2mem encoders.
>  
>  These ioctls are optional, not all drivers may support them. They were
> -introduced in Linux 2.6.21.
> +introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
> +encoders (as further documented in :ref:`encoder`).
>  
>  
>  .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
> @@ -114,16 +118,20 @@ introduced in Linux 2.6.21.
>        - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
>  	is set, encoding will continue until the end of the current *Group
>  	Of Pictures*, otherwise encoding will stop immediately. When the
> -	encoder is already stopped, this command does nothing. mem2mem
> -	encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
> -	has been encoded and all frames are ready to be dequeued and will
> -	set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
> -	the capture queue to indicate there will be no new buffers
> -	produced to dequeue. This buffer may be empty, indicated by the
> -	driver setting the ``bytesused`` field to 0. Once the
> -	``V4L2_BUF_FLAG_LAST`` flag was set, the
> -	:ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
> -	but return an ``EPIPE`` error code.
> +	encoder is already stopped, this command does nothing.
> +
> +	A stateful mem2mem encoder will proceed with encoding the source
> +	buffers pending before the command is issued and then stop producing
> +	new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
> +	has been encoded and all frames are ready to be dequeued and will set
> +	the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
> +	capture queue to indicate there will be no new buffers produced to
> +	dequeue. This buffer may be empty, indicated by the driver setting the
> +	``bytesused`` field to 0. Once the buffer with the
> +	``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
> +	<VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
> +	error code. No flags or other arguments are accepted in case of mem2mem
> +	encoders.  See :ref:`encoder` for more details.
>      * - ``V4L2_ENC_CMD_PAUSE``
>        - 2
>        - Pause the encoder. When the encoder has not been started yet, the
> 

Regards,

	Hans

Hans Verkuil March 14, 2019, 1:57 p.m. UTC | #2

Hi Tomasz,

Some more comments...

On 1/29/19 2:52 PM, Hans Verkuil wrote:
> Hi Tomasz,
> 
> Some comments below. Nothing major, so I think a v4 should be ready to be
> merged.
> 
> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>> Due to complexity of the video encoding process, the V4L2 drivers of
>> stateful encoder hardware require specific sequences of V4L2 API calls
>> to be followed. These include capability enumeration, initialization,
>> encoding, encode parameters change, drain and reset.
>>
>> Specifics of the above have been discussed during Media Workshops at
>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>> originated at those events was later implemented by the drivers we already
>> have merged in mainline, such as s5p-mfc or coda.
>>
>> The only thing missing was the real specification included as a part of
>> Linux Media documentation. Fix it now and document the encoder part of
>> the Codec API.
>>
>> Signed-off-by: Tomasz Figa <[e-mail]>
>> ---
>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>
>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>> new file mode 100644
>> index 000000000000..fb8b05a132ee
>> --- /dev/null
>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>> @@ -0,0 +1,586 @@
>> +.. -*- coding: utf-8; mode: rst -*-
>> +
>> +.. _encoder:
>> +
>> +*************************************************
>> +Memory-to-memory Stateful Video Encoder Interface
>> +*************************************************
>> +
>> +A stateful video encoder takes raw video frames in display order and encodes
>> +them into a bitstream. It generates complete chunks of the bitstream, including
>> +all metadata, headers, etc. The resulting bitstream does not require any
>> +further post-processing by the client.
>> +
>> +Performing software stream processing, header generation etc. in the driver
>> +in order to support this interface is strongly discouraged. In case such
>> +operations are needed, use of the Stateless Video Encoder Interface (in
>> +development) is strongly advised.
>> +
>> +Conventions and notation used in this document
>> +==============================================
>> +
>> +1. The general V4L2 API rules apply if not specified in this document
>> +   otherwise.
>> +
>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
>> +
>> +3. All steps not marked "optional" are required.
>> +
>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
>> +   unless specified otherwise.
>> +
>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
>> +   used interchangeably with multi-planar API, unless specified otherwise,
>> +   depending on decoder capabilities and following the general V4L2 guidelines.
>> +
>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
>> +   [0..2]: i = 0, 1, 2.
>> +
>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
>> +   queue containing data that resulted from processing buffer A.
>> +
>> +Glossary
>> +========
>> +
>> +Refer to :ref:`decoder-glossary`.
>> +
>> +State machine
>> +=============
>> +
>> +.. kernel-render:: DOT
>> +   :alt: DOT digraph of encoder state machine
>> +   :caption: Encoder state machine
>> +
>> +   digraph encoder_state_machine {
>> +       node [shape = doublecircle, label="Encoding"] Encoding;
>> +
>> +       node [shape = circle, label="Initialization"] Initialization;
>> +       node [shape = circle, label="Stopped"] Stopped;
>> +       node [shape = circle, label="Drain"] Drain;
>> +       node [shape = circle, label="Reset"] Reset;
>> +
>> +       node [shape = point]; qi
>> +       qi -> Initialization [ label = "open()" ];
>> +
>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
>> +
>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
>> +       Encoding -> Encoding;
>> +
>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>> +
>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
>> +
>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>> +   }
>> +
>> +Querying capabilities
>> +=====================
>> +
>> +1. To enumerate the set of coded formats supported by the encoder, the
>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
>> +
>> +   * The full set of supported formats will be returned, regardless of the
>> +     format set on ``OUTPUT``.
>> +
>> +2. To enumerate the set of supported raw formats, the client may call
>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
>> +
>> +   * Only the formats supported for the format currently active on ``CAPTURE``
>> +     will be returned.
>> +
>> +   * In order to enumerate raw formats supported by a given coded format,
>> +     the client must first set that coded format on ``CAPTURE`` and then
>> +     enumerate the formats on ``OUTPUT``.
>> +
>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
>> +   resolutions for a given format, passing desired pixel format in
>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
>> +
>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
>> +     format will include all possible coded resolutions supported by the
>> +     encoder for given coded pixel format.
>> +
>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
>> +     will include all possible frame buffer resolutions supported by the
>> +     encoder for given raw pixel format and coded format currently set on
>> +     ``CAPTURE``.
>> +
>> +4. Supported profiles and levels for the coded format currently set on
>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
>> +   via :c:func:`VIDIOC_QUERYCTRL`.
>> +
>> +5. Any additional encoder capabilities may be discovered by querying
>> +   their respective controls.
>> +
>> +Initialization
>> +==============
>> +
>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>> +
>> +   * **Required fields:**
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>> +
>> +     ``pixelformat``
>> +         the coded format to be produced
>> +
>> +     ``sizeimage``
>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>> +         match hardware requirements
>> +
>> +     ``width``, ``height``
>> +         ignored (always zero)
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * **Return fields:**
>> +
>> +     ``sizeimage``
>> +         adjusted size of ``CAPTURE`` buffers
>> +
>> +   .. important::
>> +
>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>> +      must adjust it afterwards.
> 
> Hmm, "including resolution": if width and height are set to 0, what should the
> OUTPUT resolution be? Up to the driver? I think this should be clarified since
> at a first reading of this paragraph it appears to be contradictory.

I think the driver should just return the width and height of the OUTPUT
format. So the width and height that userspace specifies is just ignored
and replaced by the width and height of the OUTPUT format. After all, that's
what the bitstream will encode. Returning 0 for width and height would make
this a strange exception in V4L2 and I want to avoid that.

> 
>> +
>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
>> +
>> +   * **Required fields:**
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * **Return fields:**
>> +
>> +     ``pixelformat``
>> +         raw format supported for the coded format currently selected on
>> +         the ``CAPTURE`` queue.
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +3. Set the raw source format on the ``OUTPUT`` queue via
>> +   :c:func:`VIDIOC_S_FMT`.
>> +
>> +   * **Required fields:**
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>> +
>> +     ``pixelformat``
>> +         raw format of the source
>> +
>> +     ``width``, ``height``
>> +         source resolution
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * **Return fields:**
>> +
>> +     ``width``, ``height``
>> +         may be adjusted by encoder to match alignment requirements, as
>> +         required by the currently selected formats
> 
> What if the width x height is larger than the maximum supported by the
> selected coded format? This should probably mention that in that case the
> width x height is reduced to the largest allowed value. Also mention that
> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> 
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * Setting the source resolution will reset the selection rectangles to their
>> +     default values, based on the new resolution, as described in the step 5
> 
> 5 -> 4
> 
> Or just say: "as described in the next step."
> 
>> +     below.

It should also be made explicit that:

1) the crop rectangle will be set to the given width and height *before*
it is being adjusted by S_FMT.

Open question: should we support a compose rectangle for the CAPTURE that
is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
the adjusted width and height and the compose rectangle (read-only) contains
the visible width and height. It's not strictly necessary, but it is
symmetrical.

2) the CAPTURE format will be updated as well with the new OUTPUT width and
height. The CAPTURE sizeimage might change as well.

>> +
>> +4. **Optional.** Set the visible resolution for the stream metadata via
>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.

I think you should mention that this is only necessary if the crop rectangle
that is set when you set the format isn't what you want.

>> +
>> +   * **Required fields:**
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>> +
>> +     ``target``
>> +         set to ``V4L2_SEL_TGT_CROP``
>> +
>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
>> +         rectangle and may be subject to adjustment to match codec and
>> +         hardware constraints
>> +
>> +   * **Return fields:**
>> +
>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>> +         visible rectangle adjusted by the encoder
>> +
>> +   * The following selection targets are supported on ``OUTPUT``:
>> +
>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
>> +         equal to the full source frame, matching the active ``OUTPUT``
>> +         format
>> +
>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
>> +
>> +     ``V4L2_SEL_TGT_CROP``
>> +         rectangle within the source buffer to be encoded into the
>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
>> +
>> +         .. note::
>> +
>> +            A common use case for this selection target is encoding a source
>> +            video with a resolution that is not a multiple of a macroblock,
>> +            e.g.  the common 1920x1080 resolution may require the source
>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
>> +            size. To avoid encoding the padding, the client needs to explicitly
>> +            configure this selection target to 1920x1080.

This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).

>> +
>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
>> +         maximum rectangle within the coded resolution, which the cropped
>> +         source frame can be composed into; if the hardware does not support
>> +         composition or scaling, then this is always equal to the rectangle of
>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>> +
>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
>> +         equal to a rectangle of width and height matching
>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>> +
>> +     ``V4L2_SEL_TGT_COMPOSE``
>> +         rectangle within the coded frame, which the cropped source frame
>> +         is to be composed into; defaults to
>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
>> +         additional compose/scaling capabilities; resulting stream will
>> +         have this rectangle encoded as the visible rectangle in its
>> +         metadata

I think the compose targets for OUTPUT are only needed if the hardware can
actually do scaling and/or composition. Otherwise they can (must?) be
dropped.

>> +
>> +   .. warning::
>> +
>> +      The encoder may adjust the crop/compose rectangles to the nearest
>> +      supported ones to meet codec and hardware requirements. The client needs
>> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
>> +
>> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
>> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
>> +
>> +   * **Required fields:**
>> +
>> +     ``count``
>> +         requested number of buffers to allocate; greater than zero
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
>> +         ``CAPTURE``
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * **Return fields:**
>> +
>> +     ``count``
>> +          actual number of buffers allocated
>> +
>> +   .. warning::
>> +
>> +      The actual number of allocated buffers may differ from the ``count``
>> +      given. The client must check the updated value of ``count`` after the
>> +      call returns.
>> +
>> +   .. note::
>> +
>> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
>> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
>> +      control to get the minimum number of buffers required, and pass the
>> +      obtained value plus the number of additional buffers needed in the
>> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
>> +
>> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
>> +   control over buffer allocation.
>> +
>> +   * **Required fields:**
>> +
>> +     ``count``
>> +         requested number of buffers to allocate; greater than zero
>> +
>> +     ``type``
>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>> +
>> +     other fields
>> +         follow standard semantics
>> +
>> +   * **Return fields:**
>> +
>> +     ``count``
>> +         adjusted to the number of allocated buffers
>> +
>> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
>> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
>> +   encoding process starts when both queues start streaming.
>> +
>> +.. note::
>> +
>> +   If the client stops the ``CAPTURE`` queue during the encode process and then
>> +   restarts it again, the encoder will begin generating a stream independent
>> +   from the stream generated before the stop. The exact constraints depend
>> +   on the coded format, but may include the following implications:
>> +
>> +   * encoded frames produced after the restart must not reference any
>> +     frames produced before the stop, e.g. no long term references for
>> +     H.264,
>> +
>> +   * any headers that must be included in a standalone stream must be
>> +     produced again, e.g. SPS and PPS for H.264.
>> +
>> +Encoding
>> +========
>> +
>> +This state is reached after the `Initialization` sequence finishes
>> +successfully.  In this state, the client queues and dequeues buffers to both
>> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
>> +standard semantics.
>> +
>> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
>> +format and may be affected by codec-specific extended controls, as stated
>> +in the documentation of each format.
>> +
>> +Both queues operate independently, following standard behavior of V4L2 buffer
>> +queues and memory-to-memory devices. In addition, the order of encoded frames
>> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
>> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
>> +e.g. frame reordering.
>> +
>> +The client must not assume any direct relationship between ``CAPTURE`` and
>> +``OUTPUT`` buffers and any specific timing of buffers becoming
>> +available to dequeue. Specifically:
>> +
>> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
>> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
>> +  frame that preceded it in display, but succeeded it in the decode order),
>> +
>> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
>> +  ``CAPTURE`` later into encode process, and/or after processing further
>> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
>> +  reordering is used,
>> +
>> +* buffers may become available on the ``CAPTURE`` queue without additional
>> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
>> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
>> +  available at later time, due to specifics of the decoding process,
>> +
>> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
>> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
>> +  encoder needs to use the frame as a reference for encoding further frames.
>> +
>> +.. note::
>> +
>> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
>> +   originated from, the client can set the ``timestamp`` field of the
>> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
>> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
>> +   will have their ``timestamp`` field set to the same value when dequeued.
>> +
>> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
>> +   one ``CAPTURE`` buffer, the following cases are defined:
>> +
>> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
>> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
>> +
>> +   * the encoding order differs from the presentation order (i.e. the
>> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
>> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
>> +     and thus monotonicity of the timestamps cannot be guaranteed.
> 
> Drop "and thus monotonicity of the timestamps cannot be guaranteed": this was never
> guaranteed since timestamps are just copied from OUTPUT to CAPTURE.
> 
> This phrase doesn't occur in the decoder spec, so it should be removed here as well.
> 
>> +
>> +.. note::
>> +
>> +   To let the client distinguish between frame types (keyframes, intermediate
>> +   frames; the exact list of types depends on the coded format), the
>> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
>> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
>> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
>> +   and their meanings.
>> +
>> +Encoding parameter changes
>> +==========================
>> +
>> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
>> +parameters at any time. The availability of parameters is encoder-specific
>> +and the client must query the encoder to find the set of available controls.
>> +
>> +The ability to change each parameter during encoding is encoder-specific, as
>> +per the standard semantics of the V4L2 control interface. The client may
>> +attempt to set a control during encoding and if the operation fails with the
>> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
>> +configuration change to be allowed. To do this, it may follow the `Drain`
>> +sequence to avoid losing the already queued/encoded frames.
>> +
>> +The timing of parameter updates is encoder-specific, as per the standard
>> +semantics of the V4L2 control interface. If the client needs to apply the
>> +parameters exactly at specific frame, using the Request API
>> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
>> +
>> +Drain
>> +=====
>> +
>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>> +drain sequence described below. After the drain sequence ends, the client has
>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>> +sequence was started.
>> +
>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>> +
>> +   * **Required fields:**
>> +
>> +     ``cmd``
>> +         set to ``V4L2_ENC_CMD_STOP``
>> +
>> +     ``flags``
>> +         set to 0
>> +
>> +     ``pts``
>> +         set to 0
>> +
>> +   .. warning::
>> +
>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>> +      queues are streaming. For compatibility reasons, the call to
>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>> +      not streaming, but at the same time it will not initiate the `Drain`
>> +      sequence and so the steps described below would not be applicable.
>> +
>> +2. Any ``OUTPUT`` buffers queued by the client before the
>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>> +   normal. The client must continue to handle both queues independently,
>> +   similarly to normal encode operation. This includes:
>> +
>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>> +
>> +     .. warning::
>> +
>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>> +        as it does not contain an encoded frame.
>> +
>> +     .. note::
>> +
>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>> +        :c:func:`VIDIOC_DQBUF`.
>> +
>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>> +
>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>> +
>> +   .. note::
>> +
>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>> +      event when the last frame has been decoded and all frames are ready to be
>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>> +
>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>> +   until the client issues any of the following operations:
>> +
>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>> +     operation normally, with all the state from before the drain,
>> +
>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>> +     and then resume encoding,
>> +
>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>> +
>> +.. note::
>> +
>> +   Once the drain sequence is initiated, the client needs to drive it to
>> +   completion, as described by the steps above, unless it aborts the process by
>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>> +   will fail with -EBUSY error code if attempted.
>> +
>> +   Although mandatory, the availability of encoder commands may be queried
>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>> +
>> +Reset
>> +=====
>> +
>> +The client may want to request the encoder to reinitialize the encoding, so
>> +that the following stream data becomes independent from the stream data
>> +generated before. Depending on the coded format, that may imply that:
>> +
>> +* encoded frames produced after the restart must not reference any frames
>> +  produced before the stop, e.g. no long term references for H.264,
>> +
>> +* any headers that must be included in a standalone stream must be produced
>> +  again, e.g. SPS and PPS for H.264.
>> +
>> +This can be achieved by performing the reset sequence.
>> +
>> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
>> +   and respective buffers are dequeued.
>> +
>> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
>> +   will return all currently queued ``CAPTURE`` buffers to the client, without
>> +   valid frame data.
>> +
>> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
>> +   continue with regular encoding sequence. The encoded frames produced into
>> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
>> +   decoded without the need for frames encoded before the reset sequence,
>> +   starting at the first ``OUTPUT`` buffer queued after issuing the
>> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
>> +
>> +This sequence may be also used to change encoding parameters for encoders
>> +without the ability to change the parameters on the fly.
>> +
>> +Commit points
>> +=============
>> +
>> +Setting formats and allocating buffers triggers changes in the behavior of the
>> +encoder.
>> +
>> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
>> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
>> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
>> +   previously set format being preserved.
>> +
>> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
>> +   supported for the current ``CAPTURE`` format.
>> +
>> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
>> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
>> +   format that is not supported for the currently selected ``CAPTURE`` format
>> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
>> +   supported one.
>> +
>> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
>> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
> 
> irrespectively -> irrespective
> 
>> +
>> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
>> +   change the format on the queue. Drivers will return the -EBUSY error code

Change the format on which queue? CAPTURE or OUTPUT?

I think this should be more precise: changing the CAPTURE format will always
fail if CAPTURE buffers are already allocated, or if the OUTPUT is streaming,
or if changing the CAPTURE format will change the OUTPUT format and OUTPUT
buffers are already allocated.

Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
particular) and CAPTURE buffers were already allocated and are too small.

Regards,

	Hans

>> +   for any such format change attempt.
>> +
>> +To summarize, setting formats and allocation must always start with the
>> +``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
>> +set of supported formats for the ``OUTPUT`` queue.
>> diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
>> index a0e06a88c872..82630ea89b16 100644
>> --- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
>> +++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
>> @@ -17,6 +17,7 @@ Video Memory-To-Memory Interface
>>      :maxdepth: 1
>>  
>>      dev-decoder
>> +    dev-encoder
>>  
>>  A V4L2 memory-to-memory device can compress, decompress, transform, or
>>  otherwise convert video data from one format into another format, in memory.
>> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>> index caf14e440447..593d25b935bf 100644
>> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
>> @@ -44,6 +44,11 @@ Single-planar format structure
>>  	inside the stream, when fed to a stateful mem2mem decoder, the fields
>>  	may be zero to rely on the decoder to detect the right values. For more
>>  	details see :ref:`decoder` and format descriptions.
>> +
>> +	For compressed formats on the CAPTURE side of a stateful mem2mem
>> +	encoder, the fields must be zero, since the coded size is expected to
>> +	be calculated internally by the encoder itself, based on the OUTPUT
>> +	side. For more details see :ref:`encoder` and format descriptions.
>>      * - __u32
>>        - ``pixelformat``
>>        - The pixel format or type of compression, set by the application.
>> diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
>> index 97015b9b40b8..e236ea23481b 100644
>> --- a/Documentation/media/uapi/v4l/v4l2.rst
>> +++ b/Documentation/media/uapi/v4l/v4l2.rst
>> @@ -63,6 +63,7 @@ Authors, in alphabetical order:
>>  - Figa, Tomasz <[e-mail]>
>>  
>>    - Documented the memory-to-memory decoder interface.
>> +  - Documented the memory-to-memory encoder interface.
>>  
>>  - H Schimek, Michael <[e-mail]>
>>  
>> @@ -75,6 +76,7 @@ Authors, in alphabetical order:
>>  - Osciak, Pawel <[e-mail]>
>>  
>>    - Documented the memory-to-memory decoder interface.
>> +  - Documented the memory-to-memory encoder interface.
>>  
>>  - Osciak, Pawel <[e-mail]>
>>  
>> diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
>> index c313ca8b8cb5..2b59a4e7f5ce 100644
>> --- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
>> +++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
>> @@ -57,19 +57,23 @@ currently only used by the STOP command and contains one bit: If the
>>  until the end of the current *Group Of Pictures*, otherwise it will stop
>>  immediately.
>>  
>> -A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
>> -call sends an implicit START command to the encoder if it has not been
>> -started yet. After a STOP command, :ref:`read() <func-read>` calls will read
>> +After a STOP command, :ref:`read() <func-read>` calls will read
>>  the remaining data buffered by the driver. When the buffer is empty,
>>  :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
>>  call will restart the encoder.
>>  
>> +A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
>> +call sends an implicit START command to the encoder if it has not been
>> +started yet. Applies to both queues of mem2mem encoders.
>> +
>>  A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
>>  call of a streaming file descriptor sends an implicit immediate STOP to
>> -the encoder, and all buffered data is discarded.
>> +the encoder, and all buffered data is discarded. Applies to both queues of
>> +mem2mem encoders.
>>  
>>  These ioctls are optional, not all drivers may support them. They were
>> -introduced in Linux 2.6.21.
>> +introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
>> +encoders (as further documented in :ref:`encoder`).
>>  
>>  
>>  .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
>> @@ -114,16 +118,20 @@ introduced in Linux 2.6.21.
>>        - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
>>  	is set, encoding will continue until the end of the current *Group
>>  	Of Pictures*, otherwise encoding will stop immediately. When the
>> -	encoder is already stopped, this command does nothing. mem2mem
>> -	encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
>> -	has been encoded and all frames are ready to be dequeued and will
>> -	set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
>> -	the capture queue to indicate there will be no new buffers
>> -	produced to dequeue. This buffer may be empty, indicated by the
>> -	driver setting the ``bytesused`` field to 0. Once the
>> -	``V4L2_BUF_FLAG_LAST`` flag was set, the
>> -	:ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
>> -	but return an ``EPIPE`` error code.
>> +	encoder is already stopped, this command does nothing.
>> +
>> +	A stateful mem2mem encoder will proceed with encoding the source
>> +	buffers pending before the command is issued and then stop producing
>> +	new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
>> +	has been encoded and all frames are ready to be dequeued and will set
>> +	the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
>> +	capture queue to indicate there will be no new buffers produced to
>> +	dequeue. This buffer may be empty, indicated by the driver setting the
>> +	``bytesused`` field to 0. Once the buffer with the
>> +	``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
>> +	<VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
>> +	error code. No flags or other arguments are accepted in case of mem2mem
>> +	encoders.  See :ref:`encoder` for more details.
>>      * - ``V4L2_ENC_CMD_PAUSE``
>>        - 2
>>        - Pause the encoder. When the encoder has not been started yet, the
>>
> 
> Regards,
> 
> 	Hans
>

Hans Verkuil March 21, 2019, 10:10 a.m. UTC | #3

Hi Tomasz,

A few more comments:

On 1/24/19 11:04 AM, Tomasz Figa wrote:
> Due to complexity of the video encoding process, the V4L2 drivers of
> stateful encoder hardware require specific sequences of V4L2 API calls
> to be followed. These include capability enumeration, initialization,
> encoding, encode parameters change, drain and reset.
> 
> Specifics of the above have been discussed during Media Workshops at
> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> originated at those events was later implemented by the drivers we already
> have merged in mainline, such as s5p-mfc or coda.
> 
> The only thing missing was the real specification included as a part of
> Linux Media documentation. Fix it now and document the encoder part of
> the Codec API.
> 
> Signed-off-by: Tomasz Figa <[e-mail]>
> ---
>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>  5 files changed, 617 insertions(+), 15 deletions(-)
>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> 
> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> new file mode 100644
> index 000000000000..fb8b05a132ee
> --- /dev/null
> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> @@ -0,0 +1,586 @@
> +.. -*- coding: utf-8; mode: rst -*-
> +
> +.. _encoder:
> +
> +*************************************************
> +Memory-to-memory Stateful Video Encoder Interface
> +*************************************************
> +
> +A stateful video encoder takes raw video frames in display order and encodes
> +them into a bitstream. It generates complete chunks of the bitstream, including
> +all metadata, headers, etc. The resulting bitstream does not require any
> +further post-processing by the client.
> +
> +Performing software stream processing, header generation etc. in the driver
> +in order to support this interface is strongly discouraged. In case such
> +operations are needed, use of the Stateless Video Encoder Interface (in
> +development) is strongly advised.
> +
> +Conventions and notation used in this document
> +==============================================
> +
> +1. The general V4L2 API rules apply if not specified in this document
> +   otherwise.
> +
> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> +
> +3. All steps not marked "optional" are required.
> +
> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> +   unless specified otherwise.
> +
> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> +   used interchangeably with multi-planar API, unless specified otherwise,
> +   depending on decoder capabilities and following the general V4L2 guidelines.

decoder -> encoder

> +
> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> +   [0..2]: i = 0, 1, 2.
> +
> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> +   queue containing data that resulted from processing buffer A.
> +
> +Glossary
> +========
> +
> +Refer to :ref:`decoder-glossary`.
> +
> +State machine
> +=============
> +
> +.. kernel-render:: DOT
> +   :alt: DOT digraph of encoder state machine
> +   :caption: Encoder state machine
> +
> +   digraph encoder_state_machine {
> +       node [shape = doublecircle, label="Encoding"] Encoding;
> +
> +       node [shape = circle, label="Initialization"] Initialization;
> +       node [shape = circle, label="Stopped"] Stopped;
> +       node [shape = circle, label="Drain"] Drain;
> +       node [shape = circle, label="Reset"] Reset;
> +
> +       node [shape = point]; qi
> +       qi -> Initialization [ label = "open()" ];
> +
> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> +
> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> +       Encoding -> Encoding;
> +
> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +
> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> +
> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +   }
> +
> +Querying capabilities
> +=====================
> +
> +1. To enumerate the set of coded formats supported by the encoder, the
> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> +
> +   * The full set of supported formats will be returned, regardless of the
> +     format set on ``OUTPUT``.
> +
> +2. To enumerate the set of supported raw formats, the client may call
> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> +
> +   * Only the formats supported for the format currently active on ``CAPTURE``
> +     will be returned.
> +
> +   * In order to enumerate raw formats supported by a given coded format,
> +     the client must first set that coded format on ``CAPTURE`` and then
> +     enumerate the formats on ``OUTPUT``.
> +
> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> +   resolutions for a given format, passing desired pixel format in
> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> +     format will include all possible coded resolutions supported by the
> +     encoder for given coded pixel format.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> +     will include all possible frame buffer resolutions supported by the
> +     encoder for given raw pixel format and coded format currently set on
> +     ``CAPTURE``.
> +
> +4. Supported profiles and levels for the coded format currently set on
> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> +   via :c:func:`VIDIOC_QUERYCTRL`.
> +
> +5. Any additional encoder capabilities may be discovered by querying
> +   their respective controls.
> +
> +Initialization
> +==============
> +
> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> +
> +     ``pixelformat``
> +         the coded format to be produced
> +
> +     ``sizeimage``
> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> +         match hardware requirements
> +
> +     ``width``, ``height``
> +         ignored (always zero)
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``sizeimage``
> +         adjusted size of ``CAPTURE`` buffers
> +
> +   .. important::
> +
> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> +      format. The encoder will derive a new ``OUTPUT`` format from the
> +      ``CAPTURE`` format being set, including resolution, colorimetry
> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> +      must adjust it afterwards.
> +
> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``pixelformat``
> +         raw format supported for the coded format currently selected on
> +         the ``CAPTURE`` queue.
> +
> +     other fields
> +         follow standard semantics
> +
> +3. Set the raw source format on the ``OUTPUT`` queue via
> +   :c:func:`VIDIOC_S_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``pixelformat``
> +         raw format of the source
> +
> +     ``width``, ``height``
> +         source resolution
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``width``, ``height``
> +         may be adjusted by encoder to match alignment requirements, as
> +         required by the currently selected formats
> +
> +     other fields
> +         follow standard semantics
> +
> +   * Setting the source resolution will reset the selection rectangles to their
> +     default values, based on the new resolution, as described in the step 5
> +     below.
> +
> +4. **Optional.** Set the visible resolution for the stream metadata via
> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``target``
> +         set to ``V4L2_SEL_TGT_CROP``
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> +         rectangle and may be subject to adjustment to match codec and
> +         hardware constraints
> +
> +   * **Return fields:**
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle adjusted by the encoder
> +
> +   * The following selection targets are supported on ``OUTPUT``:
> +
> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> +         equal to the full source frame, matching the active ``OUTPUT``
> +         format
> +
> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> +
> +     ``V4L2_SEL_TGT_CROP``
> +         rectangle within the source buffer to be encoded into the
> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> +
> +         .. note::
> +
> +            A common use case for this selection target is encoding a source
> +            video with a resolution that is not a multiple of a macroblock,
> +            e.g.  the common 1920x1080 resolution may require the source
> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> +            size. To avoid encoding the padding, the client needs to explicitly
> +            configure this selection target to 1920x1080.
> +
> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> +         maximum rectangle within the coded resolution, which the cropped
> +         source frame can be composed into; if the hardware does not support
> +         composition or scaling, then this is always equal to the rectangle of
> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> +         equal to a rectangle of width and height matching
> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE``
> +         rectangle within the coded frame, which the cropped source frame
> +         is to be composed into; defaults to
> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> +         additional compose/scaling capabilities; resulting stream will
> +         have this rectangle encoded as the visible rectangle in its
> +         metadata

I would only support the COMPOSE targets if the hardware can actually do
scaling and/or composing. That is conform standard V4L2 behavior where
cropping/composing is only implemented if the hardware can actually do
this.

> +
> +   .. warning::
> +
> +      The encoder may adjust the crop/compose rectangles to the nearest
> +      supported ones to meet codec and hardware requirements. The client needs
> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> +
> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> +         ``CAPTURE``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +          actual number of buffers allocated
> +
> +   .. warning::
> +
> +      The actual number of allocated buffers may differ from the ``count``
> +      given. The client must check the updated value of ``count`` after the
> +      call returns.
> +
> +   .. note::
> +
> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> +      control to get the minimum number of buffers required, and pass the
> +      obtained value plus the number of additional buffers needed in the
> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> +
> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> +   control over buffer allocation.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +         adjusted to the number of allocated buffers
> +
> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> +   encoding process starts when both queues start streaming.
> +
> +.. note::
> +
> +   If the client stops the ``CAPTURE`` queue during the encode process and then
> +   restarts it again, the encoder will begin generating a stream independent
> +   from the stream generated before the stop. The exact constraints depend
> +   on the coded format, but may include the following implications:
> +
> +   * encoded frames produced after the restart must not reference any
> +     frames produced before the stop, e.g. no long term references for
> +     H.264,
> +
> +   * any headers that must be included in a standalone stream must be
> +     produced again, e.g. SPS and PPS for H.264.
> +
> +Encoding
> +========
> +
> +This state is reached after the `Initialization` sequence finishes
> +successfully.  In this state, the client queues and dequeues buffers to both
> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> +standard semantics.
> +
> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> +format and may be affected by codec-specific extended controls, as stated
> +in the documentation of each format.
> +
> +Both queues operate independently, following standard behavior of V4L2 buffer
> +queues and memory-to-memory devices. In addition, the order of encoded frames
> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> +e.g. frame reordering.
> +
> +The client must not assume any direct relationship between ``CAPTURE`` and
> +``OUTPUT`` buffers and any specific timing of buffers becoming
> +available to dequeue. Specifically:
> +
> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> +  frame that preceded it in display, but succeeded it in the decode order),
> +
> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> +  ``CAPTURE`` later into encode process, and/or after processing further
> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> +  reordering is used,
> +
> +* buffers may become available on the ``CAPTURE`` queue without additional
> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> +  available at later time, due to specifics of the decoding process,
> +
> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> +  encoder needs to use the frame as a reference for encoding further frames.
> +
> +.. note::
> +
> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> +   originated from, the client can set the ``timestamp`` field of the
> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> +   will have their ``timestamp`` field set to the same value when dequeued.
> +
> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> +   one ``CAPTURE`` buffer, the following cases are defined:
> +
> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> +
> +   * the encoding order differs from the presentation order (i.e. the
> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> +     and thus monotonicity of the timestamps cannot be guaranteed.
> +
> +.. note::
> +
> +   To let the client distinguish between frame types (keyframes, intermediate
> +   frames; the exact list of types depends on the coded format), the
> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> +   and their meanings.

I don't think we can require this since a capture buffer may contain multiple
encoded frames.

It would actually make more sense to return it in the output buffer, but I don't
know if a hardware encoder can actually provide that information.

Another use of these flags for an output buffer is to force a keyframe if for
example a scene change was detected.

My feeling is that we should drop this note. Forcing a keyframe by setting that
flag for the output buffer might actually be a useful thing to do for a stateful
encoder.

> +
> +Encoding parameter changes
> +==========================
> +
> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> +parameters at any time. The availability of parameters is encoder-specific
> +and the client must query the encoder to find the set of available controls.
> +
> +The ability to change each parameter during encoding is encoder-specific, as
> +per the standard semantics of the V4L2 control interface. The client may
> +attempt to set a control during encoding and if the operation fails with the
> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> +configuration change to be allowed. To do this, it may follow the `Drain`
> +sequence to avoid losing the already queued/encoded frames.
> +
> +The timing of parameter updates is encoder-specific, as per the standard
> +semantics of the V4L2 control interface. If the client needs to apply the
> +parameters exactly at specific frame, using the Request API
> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> +
> +Drain
> +=====
> +
> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> +drain sequence described below. After the drain sequence ends, the client has
> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> +sequence was started.
> +
> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> +
> +   * **Required fields:**
> +
> +     ``cmd``
> +         set to ``V4L2_ENC_CMD_STOP``
> +
> +     ``flags``
> +         set to 0
> +
> +     ``pts``
> +         set to 0
> +
> +   .. warning::
> +
> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> +      queues are streaming. For compatibility reasons, the call to
> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> +      not streaming, but at the same time it will not initiate the `Drain`
> +      sequence and so the steps described below would not be applicable.
> +
> +2. Any ``OUTPUT`` buffers queued by the client before the
> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> +   normal. The client must continue to handle both queues independently,
> +   similarly to normal encode operation. This includes:
> +
> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> +
> +     .. warning::
> +
> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> +        as it does not contain an encoded frame.
> +
> +     .. note::
> +
> +        Any attempt to dequeue more buffers beyond the buffer marked with
> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> +        :c:func:`VIDIOC_DQBUF`.
> +
> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> +
> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> +
> +   .. note::
> +
> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> +      event when the last frame has been decoded and all frames are ready to be

decoded -> encoded

> +      dequeued. It is deprecated behavior and the client must not rely on it.
> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> +
> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> +   until the client issues any of the following operations:
> +
> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> +     operation normally, with all the state from before the drain,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> +     and then resume encoding,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> +
> +.. note::
> +
> +   Once the drain sequence is initiated, the client needs to drive it to
> +   completion, as described by the steps above, unless it aborts the process by
> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> +   will fail with -EBUSY error code if attempted.
> +
> +   Although mandatory, the availability of encoder commands may be queried
> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> +
> +Reset
> +=====
> +
> +The client may want to request the encoder to reinitialize the encoding, so
> +that the following stream data becomes independent from the stream data
> +generated before. Depending on the coded format, that may imply that:
> +
> +* encoded frames produced after the restart must not reference any frames
> +  produced before the stop, e.g. no long term references for H.264,
> +
> +* any headers that must be included in a standalone stream must be produced
> +  again, e.g. SPS and PPS for H.264.
> +
> +This can be achieved by performing the reset sequence.
> +
> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
> +   and respective buffers are dequeued.
> +
> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
> +   will return all currently queued ``CAPTURE`` buffers to the client, without
> +   valid frame data.
> +
> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
> +   continue with regular encoding sequence. The encoded frames produced into
> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
> +   decoded without the need for frames encoded before the reset sequence,
> +   starting at the first ``OUTPUT`` buffer queued after issuing the
> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
> +
> +This sequence may be also used to change encoding parameters for encoders
> +without the ability to change the parameters on the fly.
> +
> +Commit points
> +=============
> +
> +Setting formats and allocating buffers triggers changes in the behavior of the
> +encoder.
> +
> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
> +   previously set format being preserved.
> +
> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
> +   supported for the current ``CAPTURE`` format.
> +
> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
> +   format that is not supported for the currently selected ``CAPTURE`` format
> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
> +   supported one.
> +
> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
> +
> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
> +   change the format on the queue. Drivers will return the -EBUSY error code
> +   for any such format change attempt.
> +
> +To summarize, setting formats and allocation must always start with the
> +``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
> +set of supported formats for the ``OUTPUT`` queue.
> diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> index a0e06a88c872..82630ea89b16 100644
> --- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
> +++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> @@ -17,6 +17,7 @@ Video Memory-To-Memory Interface
>      :maxdepth: 1
>  
>      dev-decoder
> +    dev-encoder
>  
>  A V4L2 memory-to-memory device can compress, decompress, transform, or
>  otherwise convert video data from one format into another format, in memory.
> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> index caf14e440447..593d25b935bf 100644
> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> @@ -44,6 +44,11 @@ Single-planar format structure
>  	inside the stream, when fed to a stateful mem2mem decoder, the fields
>  	may be zero to rely on the decoder to detect the right values. For more
>  	details see :ref:`decoder` and format descriptions.
> +
> +	For compressed formats on the CAPTURE side of a stateful mem2mem
> +	encoder, the fields must be zero, since the coded size is expected to
> +	be calculated internally by the encoder itself, based on the OUTPUT
> +	side. For more details see :ref:`encoder` and format descriptions.
>      * - __u32
>        - ``pixelformat``
>        - The pixel format or type of compression, set by the application.
> diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
> index 97015b9b40b8..e236ea23481b 100644
> --- a/Documentation/media/uapi/v4l/v4l2.rst
> +++ b/Documentation/media/uapi/v4l/v4l2.rst
> @@ -63,6 +63,7 @@ Authors, in alphabetical order:
>  - Figa, Tomasz <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - H Schimek, Michael <[e-mail]>
>  
> @@ -75,6 +76,7 @@ Authors, in alphabetical order:
>  - Osciak, Pawel <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - Osciak, Pawel <[e-mail]>
>  
> diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> index c313ca8b8cb5..2b59a4e7f5ce 100644
> --- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> +++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> @@ -57,19 +57,23 @@ currently only used by the STOP command and contains one bit: If the
>  until the end of the current *Group Of Pictures*, otherwise it will stop
>  immediately.
>  
> -A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> -call sends an implicit START command to the encoder if it has not been
> -started yet. After a STOP command, :ref:`read() <func-read>` calls will read
> +After a STOP command, :ref:`read() <func-read>` calls will read
>  the remaining data buffered by the driver. When the buffer is empty,
>  :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
>  call will restart the encoder.
>  
> +A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> +call sends an implicit START command to the encoder if it has not been
> +started yet. Applies to both queues of mem2mem encoders.
> +
>  A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
>  call of a streaming file descriptor sends an implicit immediate STOP to
> -the encoder, and all buffered data is discarded.
> +the encoder, and all buffered data is discarded. Applies to both queues of
> +mem2mem encoders.
>  
>  These ioctls are optional, not all drivers may support them. They were
> -introduced in Linux 2.6.21.
> +introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
> +encoders (as further documented in :ref:`encoder`).
>  
>  
>  .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
> @@ -114,16 +118,20 @@ introduced in Linux 2.6.21.
>        - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
>  	is set, encoding will continue until the end of the current *Group
>  	Of Pictures*, otherwise encoding will stop immediately. When the
> -	encoder is already stopped, this command does nothing. mem2mem
> -	encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
> -	has been encoded and all frames are ready to be dequeued and will
> -	set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
> -	the capture queue to indicate there will be no new buffers
> -	produced to dequeue. This buffer may be empty, indicated by the
> -	driver setting the ``bytesused`` field to 0. Once the
> -	``V4L2_BUF_FLAG_LAST`` flag was set, the
> -	:ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
> -	but return an ``EPIPE`` error code.
> +	encoder is already stopped, this command does nothing.
> +
> +	A stateful mem2mem encoder will proceed with encoding the source
> +	buffers pending before the command is issued and then stop producing
> +	new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
> +	has been encoded and all frames are ready to be dequeued and will set
> +	the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
> +	capture queue to indicate there will be no new buffers produced to
> +	dequeue. This buffer may be empty, indicated by the driver setting the
> +	``bytesused`` field to 0. Once the buffer with the
> +	``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
> +	<VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
> +	error code. No flags or other arguments are accepted in case of mem2mem
> +	encoders.  See :ref:`encoder` for more details.
>      * - ``V4L2_ENC_CMD_PAUSE``
>        - 2
>        - Pause the encoder. When the encoder has not been started yet, the
> 

Regards,

	Hans

Hans Verkuil March 25, 2019, 1:12 p.m. UTC | #4

Another comment found while creating compliance tests:

On 1/24/19 11:04 AM, Tomasz Figa wrote:
> +Drain
> +=====
> +
> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> +drain sequence described below. After the drain sequence ends, the client has
> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> +sequence was started.
> +
> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> +
> +   * **Required fields:**
> +
> +     ``cmd``
> +         set to ``V4L2_ENC_CMD_STOP``
> +
> +     ``flags``
> +         set to 0
> +
> +     ``pts``
> +         set to 0
> +
> +   .. warning::
> +
> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> +      queues are streaming. For compatibility reasons, the call to
> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> +      not streaming, but at the same time it will not initiate the `Drain`
> +      sequence and so the steps described below would not be applicable.
> +
> +2. Any ``OUTPUT`` buffers queued by the client before the
> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> +   normal. The client must continue to handle both queues independently,
> +   similarly to normal encode operation. This includes:
> +
> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> +
> +     .. warning::
> +
> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> +        as it does not contain an encoded frame.
> +
> +     .. note::
> +
> +        Any attempt to dequeue more buffers beyond the buffer marked with
> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> +        :c:func:`VIDIOC_DQBUF`.
> +
> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> +
> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> +
> +   .. note::
> +
> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> +      event when the last frame has been decoded and all frames are ready to be
> +      dequeued. It is deprecated behavior and the client must not rely on it.
> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> +
> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> +   until the client issues any of the following operations:
> +
> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> +     operation normally, with all the state from before the drain,

I assume that calling CMD_START when *not* draining will succeed but does nothing.

In other words: while draining is in progress START will return EBUSY. When draining
was finished, then START will resume the encoder. In all other cases it just returns
0 since the encoder is really already started.

> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> +     and then resume encoding,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> +
> +.. note::
> +
> +   Once the drain sequence is initiated, the client needs to drive it to
> +   completion, as described by the steps above, unless it aborts the process by
> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> +   will fail with -EBUSY error code if attempted.

I assume calling STOP again once the drain sequence completed just returns 0 and
doesn't do anything else (since we're already stopped).

> +
> +   Although mandatory, the availability of encoder commands may be queried
> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.

Some corner cases:

1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
   Now ENC_CMD_STOP is issued. What should happen?

   Proposal: the next time the applications queues a CAPTURE buffer it is returned
   at once as an empty buffer with FLAG_LAST set.

2) Both queues are streaming and buffers have been encoded, but currently no buffers
   are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?

   Proposal: the next time the applications queues a CAPTURE buffer it is returned
   at once as an empty buffer with FLAG_LAST set. This is consistent with the
   previous corner case.

3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
   is issued. What should happen?

   Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
   buffer with FLAG_LAST set.

4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
   Before the drain process completes STREAMOFF is called for either CAPTURE or
   OUTPUT queue. What should happen?

   Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
   returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
   will be used as input to the encoder.

   Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
   FLAG_LAST set.

Some of this might have to be documented, but these corner cases should certainly be
tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
with this.

Regards,

	Hans

Hans Verkuil March 25, 2019, 4:33 p.m. UTC | #5

On 3/25/19 2:12 PM, Hans Verkuil wrote:
> Another comment found while creating compliance tests:
> 
> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>> +Drain
>> +=====
>> +
>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>> +drain sequence described below. After the drain sequence ends, the client has
>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>> +sequence was started.
>> +
>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>> +
>> +   * **Required fields:**
>> +
>> +     ``cmd``
>> +         set to ``V4L2_ENC_CMD_STOP``
>> +
>> +     ``flags``
>> +         set to 0
>> +
>> +     ``pts``
>> +         set to 0
>> +
>> +   .. warning::
>> +
>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>> +      queues are streaming. For compatibility reasons, the call to
>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>> +      not streaming, but at the same time it will not initiate the `Drain`
>> +      sequence and so the steps described below would not be applicable.
>> +
>> +2. Any ``OUTPUT`` buffers queued by the client before the
>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>> +   normal. The client must continue to handle both queues independently,
>> +   similarly to normal encode operation. This includes:
>> +
>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>> +
>> +     .. warning::
>> +
>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>> +        as it does not contain an encoded frame.
>> +
>> +     .. note::
>> +
>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>> +        :c:func:`VIDIOC_DQBUF`.
>> +
>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>> +
>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>> +
>> +   .. note::
>> +
>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>> +      event when the last frame has been decoded and all frames are ready to be
>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>> +
>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>> +   until the client issues any of the following operations:
>> +
>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>> +     operation normally, with all the state from before the drain,
> 
> I assume that calling CMD_START when *not* draining will succeed but does nothing.
> 
> In other words: while draining is in progress START will return EBUSY. When draining
> was finished, then START will resume the encoder. In all other cases it just returns
> 0 since the encoder is really already started.
> 
>> +
>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>> +     and then resume encoding,
>> +
>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>> +
>> +.. note::
>> +
>> +   Once the drain sequence is initiated, the client needs to drive it to
>> +   completion, as described by the steps above, unless it aborts the process by
>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>> +   will fail with -EBUSY error code if attempted.
> 
> I assume calling STOP again once the drain sequence completed just returns 0 and
> doesn't do anything else (since we're already stopped).
> 
>> +
>> +   Although mandatory, the availability of encoder commands may be queried
>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> 
> Some corner cases:
> 
> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
>    Now ENC_CMD_STOP is issued. What should happen?
> 
>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>    at once as an empty buffer with FLAG_LAST set.
> 
> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> 
>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
>    previous corner case.
> 
> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
>    is issued. What should happen?
> 
>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
>    buffer with FLAG_LAST set.
> 
> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
>    Before the drain process completes STREAMOFF is called for either CAPTURE or
>    OUTPUT queue. What should happen?
> 
>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
>    will be used as input to the encoder.
> 
>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
>    FLAG_LAST set.

Note: I'm not quite sure about that last one yet. It's a bit tricky to implement.

I think some of these rules might have to be implemented in v4l2-mem2mem.c to
simplify drivers.

Regards,

	Hans

> 
> Some of this might have to be documented, but these corner cases should certainly be
> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> with this.
> 
> Regards,
> 
> 	Hans
>

Tomasz Figa April 5, 2019, 5:53 a.m. UTC | #6

On Tue, Jan 29, 2019 at 10:53 PM Hans Verkuil <[e-mail]> wrote:
>
> Hi Tomasz,
>
> Some comments below. Nothing major, so I think a v4 should be ready to be
> merged.
>
> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > Due to complexity of the video encoding process, the V4L2 drivers of
> > stateful encoder hardware require specific sequences of V4L2 API calls
> > to be followed. These include capability enumeration, initialization,
> > encoding, encode parameters change, drain and reset.
> >
> > Specifics of the above have been discussed during Media Workshops at
> > LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> > Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> > originated at those events was later implemented by the drivers we already
> > have merged in mainline, such as s5p-mfc or coda.
> >
> > The only thing missing was the real specification included as a part of
> > Linux Media documentation. Fix it now and document the encoder part of
> > the Codec API.
> >
> > Signed-off-by: Tomasz Figa <[e-mail]>
> > ---
> >  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >  5 files changed, 617 insertions(+), 15 deletions(-)
> >  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >
> > diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> > new file mode 100644
> > index 000000000000..fb8b05a132ee
> > --- /dev/null
> > +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> > @@ -0,0 +1,586 @@
> > +.. -*- coding: utf-8; mode: rst -*-
> > +
> > +.. _encoder:
> > +
> > +*************************************************
> > +Memory-to-memory Stateful Video Encoder Interface
> > +*************************************************
> > +
> > +A stateful video encoder takes raw video frames in display order and encodes
> > +them into a bitstream. It generates complete chunks of the bitstream, including
> > +all metadata, headers, etc. The resulting bitstream does not require any
> > +further post-processing by the client.
> > +
> > +Performing software stream processing, header generation etc. in the driver
> > +in order to support this interface is strongly discouraged. In case such
> > +operations are needed, use of the Stateless Video Encoder Interface (in
> > +development) is strongly advised.
> > +
> > +Conventions and notation used in this document
> > +==============================================
> > +
> > +1. The general V4L2 API rules apply if not specified in this document
> > +   otherwise.
> > +
> > +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> > +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> > +
> > +3. All steps not marked "optional" are required.
> > +
> > +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> > +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> > +   unless specified otherwise.
> > +
> > +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> > +   used interchangeably with multi-planar API, unless specified otherwise,
> > +   depending on decoder capabilities and following the general V4L2 guidelines.
> > +
> > +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> > +   [0..2]: i = 0, 1, 2.
> > +
> > +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> > +   queue containing data that resulted from processing buffer A.
> > +
> > +Glossary
> > +========
> > +
> > +Refer to :ref:`decoder-glossary`.
> > +
> > +State machine
> > +=============
> > +
> > +.. kernel-render:: DOT
> > +   :alt: DOT digraph of encoder state machine
> > +   :caption: Encoder state machine
> > +
> > +   digraph encoder_state_machine {
> > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > +
> > +       node [shape = circle, label="Initialization"] Initialization;
> > +       node [shape = circle, label="Stopped"] Stopped;
> > +       node [shape = circle, label="Drain"] Drain;
> > +       node [shape = circle, label="Reset"] Reset;
> > +
> > +       node [shape = point]; qi
> > +       qi -> Initialization [ label = "open()" ];
> > +
> > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > +
> > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > +       Encoding -> Encoding;
> > +
> > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> > +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +
> > +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> > +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> > +
> > +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> > +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +   }
> > +
> > +Querying capabilities
> > +=====================
> > +
> > +1. To enumerate the set of coded formats supported by the encoder, the
> > +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> > +
> > +   * The full set of supported formats will be returned, regardless of the
> > +     format set on ``OUTPUT``.
> > +
> > +2. To enumerate the set of supported raw formats, the client may call
> > +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> > +
> > +   * Only the formats supported for the format currently active on ``CAPTURE``
> > +     will be returned.
> > +
> > +   * In order to enumerate raw formats supported by a given coded format,
> > +     the client must first set that coded format on ``CAPTURE`` and then
> > +     enumerate the formats on ``OUTPUT``.
> > +
> > +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> > +   resolutions for a given format, passing desired pixel format in
> > +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> > +
> > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> > +     format will include all possible coded resolutions supported by the
> > +     encoder for given coded pixel format.
> > +
> > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> > +     will include all possible frame buffer resolutions supported by the
> > +     encoder for given raw pixel format and coded format currently set on
> > +     ``CAPTURE``.
> > +
> > +4. Supported profiles and levels for the coded format currently set on
> > +   ``CAPTURE``, if applicable, may be queried using their respective controls
> > +   via :c:func:`VIDIOC_QUERYCTRL`.
> > +
> > +5. Any additional encoder capabilities may be discovered by querying
> > +   their respective controls.
> > +
> > +Initialization
> > +==============
> > +
> > +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> > +
> > +     ``pixelformat``
> > +         the coded format to be produced
> > +
> > +     ``sizeimage``
> > +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> > +         match hardware requirements
> > +
> > +     ``width``, ``height``
> > +         ignored (always zero)
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``sizeimage``
> > +         adjusted size of ``CAPTURE`` buffers
> > +
> > +   .. important::
> > +
> > +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> > +      format. The encoder will derive a new ``OUTPUT`` format from the
> > +      ``CAPTURE`` format being set, including resolution, colorimetry
> > +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> > +      must adjust it afterwards.
>
> Hmm, "including resolution": if width and height are set to 0, what should the
> OUTPUT resolution be? Up to the driver? I think this should be clarified since
> at a first reading of this paragraph it appears to be contradictory.
>

Indeed, it's hard to derive some sensible resolution from 0...

The intention here is to prevent the application from making any
assumptions about the OUTPUT format, if it changes the CAPTURE format.
Then, it shouldn't actually matter what's the new OUTPUT format, since
the application needs to set it anyway.

Maybe let's just remove the mention of deriving and say something
along these lines?

"How the new ``OUTPUT`` format is determined is unspecified and the client must
 ensure it matches its needs afterwards."

> > +
> > +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> > +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``pixelformat``
> > +         raw format supported for the coded format currently selected on
> > +         the ``CAPTURE`` queue.
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +3. Set the raw source format on the ``OUTPUT`` queue via
> > +   :c:func:`VIDIOC_S_FMT`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     ``pixelformat``
> > +         raw format of the source
> > +
> > +     ``width``, ``height``
> > +         source resolution
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``width``, ``height``
> > +         may be adjusted by encoder to match alignment requirements, as
> > +         required by the currently selected formats
>
> What if the width x height is larger than the maximum supported by the
> selected coded format? This should probably mention that in that case the
> width x height is reduced to the largest allowed value. Also mention that
> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
>

Right, the description should mention minimums and maximums as well.

> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * Setting the source resolution will reset the selection rectangles to their
> > +     default values, based on the new resolution, as described in the step 5
>
> 5 -> 4
>
> Or just say: "as described in the next step."
>

Good idea.

> > +     below.
> > +
> > +4. **Optional.** Set the visible resolution for the stream metadata via
> > +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     ``target``
> > +         set to ``V4L2_SEL_TGT_CROP``
> > +
> > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> > +         rectangle and may be subject to adjustment to match codec and
> > +         hardware constraints
> > +
> > +   * **Return fields:**
> > +
> > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > +         visible rectangle adjusted by the encoder
> > +
> > +   * The following selection targets are supported on ``OUTPUT``:
> > +
> > +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> > +         equal to the full source frame, matching the active ``OUTPUT``
> > +         format
> > +
> > +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> > +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> > +
> > +     ``V4L2_SEL_TGT_CROP``
> > +         rectangle within the source buffer to be encoded into the
> > +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> > +
> > +         .. note::
> > +
> > +            A common use case for this selection target is encoding a source
> > +            video with a resolution that is not a multiple of a macroblock,
> > +            e.g.  the common 1920x1080 resolution may require the source
> > +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> > +            size. To avoid encoding the padding, the client needs to explicitly
> > +            configure this selection target to 1920x1080.
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> > +         maximum rectangle within the coded resolution, which the cropped
> > +         source frame can be composed into; if the hardware does not support
> > +         composition or scaling, then this is always equal to the rectangle of
> > +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> > +         equal to a rectangle of width and height matching
> > +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE``
> > +         rectangle within the coded frame, which the cropped source frame
> > +         is to be composed into; defaults to
> > +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> > +         additional compose/scaling capabilities; resulting stream will
> > +         have this rectangle encoded as the visible rectangle in its
> > +         metadata
> > +
> > +   .. warning::
> > +
> > +      The encoder may adjust the crop/compose rectangles to the nearest
> > +      supported ones to meet codec and hardware requirements. The client needs
> > +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> > +
> > +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> > +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> > +
> > +   * **Required fields:**
> > +
> > +     ``count``
> > +         requested number of buffers to allocate; greater than zero
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> > +         ``CAPTURE``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``count``
> > +          actual number of buffers allocated
> > +
> > +   .. warning::
> > +
> > +      The actual number of allocated buffers may differ from the ``count``
> > +      given. The client must check the updated value of ``count`` after the
> > +      call returns.
> > +
> > +   .. note::
> > +
> > +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> > +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> > +      control to get the minimum number of buffers required, and pass the
> > +      obtained value plus the number of additional buffers needed in the
> > +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> > +
> > +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> > +   control over buffer allocation.
> > +
> > +   * **Required fields:**
> > +
> > +     ``count``
> > +         requested number of buffers to allocate; greater than zero
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``count``
> > +         adjusted to the number of allocated buffers
> > +
> > +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> > +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> > +   encoding process starts when both queues start streaming.
> > +
> > +.. note::
> > +
> > +   If the client stops the ``CAPTURE`` queue during the encode process and then
> > +   restarts it again, the encoder will begin generating a stream independent
> > +   from the stream generated before the stop. The exact constraints depend
> > +   on the coded format, but may include the following implications:
> > +
> > +   * encoded frames produced after the restart must not reference any
> > +     frames produced before the stop, e.g. no long term references for
> > +     H.264,
> > +
> > +   * any headers that must be included in a standalone stream must be
> > +     produced again, e.g. SPS and PPS for H.264.
> > +
> > +Encoding
> > +========
> > +
> > +This state is reached after the `Initialization` sequence finishes
> > +successfully.  In this state, the client queues and dequeues buffers to both
> > +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> > +standard semantics.
> > +
> > +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> > +format and may be affected by codec-specific extended controls, as stated
> > +in the documentation of each format.
> > +
> > +Both queues operate independently, following standard behavior of V4L2 buffer
> > +queues and memory-to-memory devices. In addition, the order of encoded frames
> > +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> > +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> > +e.g. frame reordering.
> > +
> > +The client must not assume any direct relationship between ``CAPTURE`` and
> > +``OUTPUT`` buffers and any specific timing of buffers becoming
> > +available to dequeue. Specifically:
> > +
> > +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> > +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> > +  frame that preceded it in display, but succeeded it in the decode order),
> > +
> > +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> > +  ``CAPTURE`` later into encode process, and/or after processing further
> > +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> > +  reordering is used,
> > +
> > +* buffers may become available on the ``CAPTURE`` queue without additional
> > +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> > +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> > +  available at later time, due to specifics of the decoding process,
> > +
> > +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> > +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> > +  encoder needs to use the frame as a reference for encoding further frames.
> > +
> > +.. note::
> > +
> > +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> > +   originated from, the client can set the ``timestamp`` field of the
> > +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> > +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> > +   will have their ``timestamp`` field set to the same value when dequeued.
> > +
> > +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> > +   one ``CAPTURE`` buffer, the following cases are defined:
> > +
> > +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> > +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> > +
> > +   * the encoding order differs from the presentation order (i.e. the
> > +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> > +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> > +     and thus monotonicity of the timestamps cannot be guaranteed.
>
> Drop "and thus monotonicity of the timestamps cannot be guaranteed": this was never
> guaranteed since timestamps are just copied from OUTPUT to CAPTURE.
>
> This phrase doesn't occur in the decoder spec, so it should be removed here as well.
>

Indeed. I think I forgot to remove it from here.

> > +
> > +.. note::
> > +
> > +   To let the client distinguish between frame types (keyframes, intermediate
> > +   frames; the exact list of types depends on the coded format), the
> > +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> > +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> > +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> > +   and their meanings.
> > +
> > +Encoding parameter changes
> > +==========================
> > +
> > +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> > +parameters at any time. The availability of parameters is encoder-specific
> > +and the client must query the encoder to find the set of available controls.
> > +
> > +The ability to change each parameter during encoding is encoder-specific, as
> > +per the standard semantics of the V4L2 control interface. The client may
> > +attempt to set a control during encoding and if the operation fails with the
> > +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> > +configuration change to be allowed. To do this, it may follow the `Drain`
> > +sequence to avoid losing the already queued/encoded frames.
> > +
> > +The timing of parameter updates is encoder-specific, as per the standard
> > +semantics of the V4L2 control interface. If the client needs to apply the
> > +parameters exactly at specific frame, using the Request API
> > +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> > +
> > +Drain
> > +=====
> > +
> > +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> > +related ``CAPTURE`` buffers are given to the client, the client must follow the
> > +drain sequence described below. After the drain sequence ends, the client has
> > +received all encoded frames for all ``OUTPUT`` buffers queued before the
> > +sequence was started.
> > +
> > +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``cmd``
> > +         set to ``V4L2_ENC_CMD_STOP``
> > +
> > +     ``flags``
> > +         set to 0
> > +
> > +     ``pts``
> > +         set to 0
> > +
> > +   .. warning::
> > +
> > +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> > +      queues are streaming. For compatibility reasons, the call to
> > +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> > +      not streaming, but at the same time it will not initiate the `Drain`
> > +      sequence and so the steps described below would not be applicable.
> > +
> > +2. Any ``OUTPUT`` buffers queued by the client before the
> > +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> > +   normal. The client must continue to handle both queues independently,
> > +   similarly to normal encode operation. This includes:
> > +
> > +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> > +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> > +
> > +     .. warning::
> > +
> > +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> > +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> > +        as it does not contain an encoded frame.
> > +
> > +     .. note::
> > +
> > +        Any attempt to dequeue more buffers beyond the buffer marked with
> > +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> > +        :c:func:`VIDIOC_DQBUF`.
> > +
> > +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> > +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> > +
> > +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> > +
> > +   .. note::
> > +
> > +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> > +      event when the last frame has been decoded and all frames are ready to be
> > +      dequeued. It is deprecated behavior and the client must not rely on it.
> > +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> > +
> > +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> > +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> > +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> > +   until the client issues any of the following operations:
> > +
> > +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> > +     operation normally, with all the state from before the drain,
> > +
> > +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> > +     and then resume encoding,
> > +
> > +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> > +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> > +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> > +
> > +.. note::
> > +
> > +   Once the drain sequence is initiated, the client needs to drive it to
> > +   completion, as described by the steps above, unless it aborts the process by
> > +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> > +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> > +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> > +   will fail with -EBUSY error code if attempted.
> > +
> > +   Although mandatory, the availability of encoder commands may be queried
> > +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> > +
> > +Reset
> > +=====
> > +
> > +The client may want to request the encoder to reinitialize the encoding, so
> > +that the following stream data becomes independent from the stream data
> > +generated before. Depending on the coded format, that may imply that:
> > +
> > +* encoded frames produced after the restart must not reference any frames
> > +  produced before the stop, e.g. no long term references for H.264,
> > +
> > +* any headers that must be included in a standalone stream must be produced
> > +  again, e.g. SPS and PPS for H.264.
> > +
> > +This can be achieved by performing the reset sequence.
> > +
> > +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
> > +   and respective buffers are dequeued.
> > +
> > +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
> > +   will return all currently queued ``CAPTURE`` buffers to the client, without
> > +   valid frame data.
> > +
> > +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
> > +   continue with regular encoding sequence. The encoded frames produced into
> > +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
> > +   decoded without the need for frames encoded before the reset sequence,
> > +   starting at the first ``OUTPUT`` buffer queued after issuing the
> > +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
> > +
> > +This sequence may be also used to change encoding parameters for encoders
> > +without the ability to change the parameters on the fly.
> > +
> > +Commit points
> > +=============
> > +
> > +Setting formats and allocating buffers triggers changes in the behavior of the
> > +encoder.
> > +
> > +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
> > +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
> > +   that the ``OUTPUT`` format may be reset and the client must not rely on the
> > +   previously set format being preserved.
> > +
> > +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
> > +   supported for the current ``CAPTURE`` format.
> > +
> > +3. Setting the format on the ``OUTPUT`` queue does not change the list of
> > +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
> > +   format that is not supported for the currently selected ``CAPTURE`` format
> > +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
> > +   supported one.
> > +
> > +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
> > +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
>
> irrespectively -> irrespective
>

Ack.

Best regards,
Tomasz

Hans Verkuil April 5, 2019, 7:09 a.m. UTC | #7

On 4/5/19 7:53 AM, Tomasz Figa wrote:
> On Tue, Jan 29, 2019 at 10:53 PM Hans Verkuil <[e-mail]> wrote:
>>
>> Hi Tomasz,
>>
>> Some comments below. Nothing major, so I think a v4 should be ready to be
>> merged.
>>
>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>> Due to complexity of the video encoding process, the V4L2 drivers of
>>> stateful encoder hardware require specific sequences of V4L2 API calls
>>> to be followed. These include capability enumeration, initialization,
>>> encoding, encode parameters change, drain and reset.
>>>
>>> Specifics of the above have been discussed during Media Workshops at
>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>>> originated at those events was later implemented by the drivers we already
>>> have merged in mainline, such as s5p-mfc or coda.
>>>
>>> The only thing missing was the real specification included as a part of
>>> Linux Media documentation. Fix it now and document the encoder part of
>>> the Codec API.
>>>
>>> Signed-off-by: Tomasz Figa <[e-mail]>
>>> ---
>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>>
>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>>> new file mode 100644
>>> index 000000000000..fb8b05a132ee
>>> --- /dev/null
>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>>> @@ -0,0 +1,586 @@

<snip>

>>> +Initialization
>>> +==============
>>> +
>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>>> +
>>> +     ``pixelformat``
>>> +         the coded format to be produced
>>> +
>>> +     ``sizeimage``
>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>>> +         match hardware requirements
>>> +
>>> +     ``width``, ``height``
>>> +         ignored (always zero)
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``sizeimage``
>>> +         adjusted size of ``CAPTURE`` buffers
>>> +
>>> +   .. important::
>>> +
>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>>> +      must adjust it afterwards.
>>
>> Hmm, "including resolution": if width and height are set to 0, what should the
>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
>> at a first reading of this paragraph it appears to be contradictory.
>>
> 
> Indeed, it's hard to derive some sensible resolution from 0...
> 
> The intention here is to prevent the application from making any
> assumptions about the OUTPUT format, if it changes the CAPTURE format.
> Then, it shouldn't actually matter what's the new OUTPUT format, since
> the application needs to set it anyway.
> 
> Maybe let's just remove the mention of deriving and say something
> along these lines?
> 
> "How the new ``OUTPUT`` format is determined is unspecified and the client must
>  ensure it matches its needs afterwards."

I would replace "unspecified" with "driver specific" or possibly "up to the driver".

Regards,

	Hans

Tomasz Figa April 5, 2019, 8:12 a.m. UTC | #8

On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
>
> Hi Tomasz,
>
> Some more comments...
>
> On 1/29/19 2:52 PM, Hans Verkuil wrote:
> > Hi Tomasz,
> >
> > Some comments below. Nothing major, so I think a v4 should be ready to be
> > merged.
> >
> > On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >> Due to complexity of the video encoding process, the V4L2 drivers of
> >> stateful encoder hardware require specific sequences of V4L2 API calls
> >> to be followed. These include capability enumeration, initialization,
> >> encoding, encode parameters change, drain and reset.
> >>
> >> Specifics of the above have been discussed during Media Workshops at
> >> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> >> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> >> originated at those events was later implemented by the drivers we already
> >> have merged in mainline, such as s5p-mfc or coda.
> >>
> >> The only thing missing was the real specification included as a part of
> >> Linux Media documentation. Fix it now and document the encoder part of
> >> the Codec API.
> >>
> >> Signed-off-by: Tomasz Figa <[e-mail]>
> >> ---
> >>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >>  5 files changed, 617 insertions(+), 15 deletions(-)
> >>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >>
> >> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> >> new file mode 100644
> >> index 000000000000..fb8b05a132ee
> >> --- /dev/null
> >> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> >> @@ -0,0 +1,586 @@
> >> +.. -*- coding: utf-8; mode: rst -*-
> >> +
> >> +.. _encoder:
> >> +
> >> +*************************************************
> >> +Memory-to-memory Stateful Video Encoder Interface
> >> +*************************************************
> >> +
> >> +A stateful video encoder takes raw video frames in display order and encodes
> >> +them into a bitstream. It generates complete chunks of the bitstream, including
> >> +all metadata, headers, etc. The resulting bitstream does not require any
> >> +further post-processing by the client.
> >> +
> >> +Performing software stream processing, header generation etc. in the driver
> >> +in order to support this interface is strongly discouraged. In case such
> >> +operations are needed, use of the Stateless Video Encoder Interface (in
> >> +development) is strongly advised.
> >> +
> >> +Conventions and notation used in this document
> >> +==============================================
> >> +
> >> +1. The general V4L2 API rules apply if not specified in this document
> >> +   otherwise.
> >> +
> >> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> >> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> >> +
> >> +3. All steps not marked "optional" are required.
> >> +
> >> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> >> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> >> +   unless specified otherwise.
> >> +
> >> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> >> +   used interchangeably with multi-planar API, unless specified otherwise,
> >> +   depending on decoder capabilities and following the general V4L2 guidelines.
> >> +
> >> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> >> +   [0..2]: i = 0, 1, 2.
> >> +
> >> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> >> +   queue containing data that resulted from processing buffer A.
> >> +
> >> +Glossary
> >> +========
> >> +
> >> +Refer to :ref:`decoder-glossary`.
> >> +
> >> +State machine
> >> +=============
> >> +
> >> +.. kernel-render:: DOT
> >> +   :alt: DOT digraph of encoder state machine
> >> +   :caption: Encoder state machine
> >> +
> >> +   digraph encoder_state_machine {
> >> +       node [shape = doublecircle, label="Encoding"] Encoding;
> >> +
> >> +       node [shape = circle, label="Initialization"] Initialization;
> >> +       node [shape = circle, label="Stopped"] Stopped;
> >> +       node [shape = circle, label="Drain"] Drain;
> >> +       node [shape = circle, label="Reset"] Reset;
> >> +
> >> +       node [shape = point]; qi
> >> +       qi -> Initialization [ label = "open()" ];
> >> +
> >> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> >> +
> >> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> >> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> >> +       Encoding -> Encoding;
> >> +
> >> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> >> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >> +
> >> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> >> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> >> +
> >> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> >> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >> +   }
> >> +
> >> +Querying capabilities
> >> +=====================
> >> +
> >> +1. To enumerate the set of coded formats supported by the encoder, the
> >> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> >> +
> >> +   * The full set of supported formats will be returned, regardless of the
> >> +     format set on ``OUTPUT``.
> >> +
> >> +2. To enumerate the set of supported raw formats, the client may call
> >> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> >> +
> >> +   * Only the formats supported for the format currently active on ``CAPTURE``
> >> +     will be returned.
> >> +
> >> +   * In order to enumerate raw formats supported by a given coded format,
> >> +     the client must first set that coded format on ``CAPTURE`` and then
> >> +     enumerate the formats on ``OUTPUT``.
> >> +
> >> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> >> +   resolutions for a given format, passing desired pixel format in
> >> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> >> +
> >> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> >> +     format will include all possible coded resolutions supported by the
> >> +     encoder for given coded pixel format.
> >> +
> >> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> >> +     will include all possible frame buffer resolutions supported by the
> >> +     encoder for given raw pixel format and coded format currently set on
> >> +     ``CAPTURE``.
> >> +
> >> +4. Supported profiles and levels for the coded format currently set on
> >> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> >> +   via :c:func:`VIDIOC_QUERYCTRL`.
> >> +
> >> +5. Any additional encoder capabilities may be discovered by querying
> >> +   their respective controls.
> >> +
> >> +Initialization
> >> +==============
> >> +
> >> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> >> +
> >> +     ``pixelformat``
> >> +         the coded format to be produced
> >> +
> >> +     ``sizeimage``
> >> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> >> +         match hardware requirements
> >> +
> >> +     ``width``, ``height``
> >> +         ignored (always zero)
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``sizeimage``
> >> +         adjusted size of ``CAPTURE`` buffers
> >> +
> >> +   .. important::
> >> +
> >> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> >> +      format. The encoder will derive a new ``OUTPUT`` format from the
> >> +      ``CAPTURE`` format being set, including resolution, colorimetry
> >> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> >> +      must adjust it afterwards.
> >
> > Hmm, "including resolution": if width and height are set to 0, what should the
> > OUTPUT resolution be? Up to the driver? I think this should be clarified since
> > at a first reading of this paragraph it appears to be contradictory.
>
> I think the driver should just return the width and height of the OUTPUT
> format. So the width and height that userspace specifies is just ignored
> and replaced by the width and height of the OUTPUT format. After all, that's
> what the bitstream will encode. Returning 0 for width and height would make
> this a strange exception in V4L2 and I want to avoid that.
>

Hmm, however, the width and height of the OUTPUT format is not what's
actually encoded in the bitstream. The right selection rectangle
determines that.

In one of the previous versions I though we could put the codec
resolution as the width and height of the CAPTURE format, which would
be the resolution of the encoded image rounded up to full macroblocks
+/- some encoder-specific constraints. AFAIR there was some concern
about OUTPUT format changes triggering CAPTURE format changes, but to
be honest, I'm not sure if that's really a problem. I just decided to
drop that for the simplicity.

> >
> >> +
> >> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> >> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``pixelformat``
> >> +         raw format supported for the coded format currently selected on
> >> +         the ``CAPTURE`` queue.
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +3. Set the raw source format on the ``OUTPUT`` queue via
> >> +   :c:func:`VIDIOC_S_FMT`.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >> +
> >> +     ``pixelformat``
> >> +         raw format of the source
> >> +
> >> +     ``width``, ``height``
> >> +         source resolution
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``width``, ``height``
> >> +         may be adjusted by encoder to match alignment requirements, as
> >> +         required by the currently selected formats
> >
> > What if the width x height is larger than the maximum supported by the
> > selected coded format? This should probably mention that in that case the
> > width x height is reduced to the largest allowed value. Also mention that
> > this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> >
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * Setting the source resolution will reset the selection rectangles to their
> >> +     default values, based on the new resolution, as described in the step 5
> >
> > 5 -> 4
> >
> > Or just say: "as described in the next step."
> >
> >> +     below.
>
> It should also be made explicit that:
>
> 1) the crop rectangle will be set to the given width and height *before*
> it is being adjusted by S_FMT.
>

I don't think that's what we want here.

Defining the default rectangle to be exactly the same as the OUTPUT
resolution (after the adjustment) makes the semantics consistent - not
setting the crop rectangle gives you exactly the behavior as if there
was no cropping involved (or supported by the encoder).

> Open question: should we support a compose rectangle for the CAPTURE that
> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
> the adjusted width and height and the compose rectangle (read-only) contains
> the visible width and height. It's not strictly necessary, but it is
> symmetrical.

Wouldn't it rather be the CAPTURE crop rectangle that would be of the
same resolution of the OUTPUT compose rectangle? Then you could
actually have the CAPTURE compose rectangle for putting that into the
desired rectangle of the encoded stream, if the encoder supports that.
(I don't know any that does, so probably out of concern for now.)

>
> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
> height. The CAPTURE sizeimage might change as well.
>
> >> +
> >> +4. **Optional.** Set the visible resolution for the stream metadata via
> >> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
>
> I think you should mention that this is only necessary if the crop rectangle
> that is set when you set the format isn't what you want.
>

Ack.

> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >> +
> >> +     ``target``
> >> +         set to ``V4L2_SEL_TGT_CROP``
> >> +
> >> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> >> +         rectangle and may be subject to adjustment to match codec and
> >> +         hardware constraints
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >> +         visible rectangle adjusted by the encoder
> >> +
> >> +   * The following selection targets are supported on ``OUTPUT``:
> >> +
> >> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> >> +         equal to the full source frame, matching the active ``OUTPUT``
> >> +         format
> >> +
> >> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> >> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> >> +
> >> +     ``V4L2_SEL_TGT_CROP``
> >> +         rectangle within the source buffer to be encoded into the
> >> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> >> +
> >> +         .. note::
> >> +
> >> +            A common use case for this selection target is encoding a source
> >> +            video with a resolution that is not a multiple of a macroblock,
> >> +            e.g.  the common 1920x1080 resolution may require the source
> >> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> >> +            size. To avoid encoding the padding, the client needs to explicitly
> >> +            configure this selection target to 1920x1080.
>
> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
>

Sorry, which part exactly and what part of the proposal exactly? :)
(My comment above might be related, though.)

> >> +
> >> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> >> +         maximum rectangle within the coded resolution, which the cropped
> >> +         source frame can be composed into; if the hardware does not support
> >> +         composition or scaling, then this is always equal to the rectangle of
> >> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >> +
> >> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> >> +         equal to a rectangle of width and height matching
> >> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >> +
> >> +     ``V4L2_SEL_TGT_COMPOSE``
> >> +         rectangle within the coded frame, which the cropped source frame
> >> +         is to be composed into; defaults to
> >> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> >> +         additional compose/scaling capabilities; resulting stream will
> >> +         have this rectangle encoded as the visible rectangle in its
> >> +         metadata
>
> I think the compose targets for OUTPUT are only needed if the hardware can
> actually do scaling and/or composition. Otherwise they can (must?) be
> dropped.
>

Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
userspace to learn the target visible rectangle that's going to be
encoded in the stream metadata. If we omit it, we wouldn't have a way
that would be consistent between encoders that can do
scaling/composition and those that can't.

However, with your proposal of actually having selection rectangles
for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
would expose a varying set of rectangles, depending on the hardware
capability, while the CAPTURE queue would always expose its rectangle
with that information.

> >> +
> >> +   .. warning::
> >> +
> >> +      The encoder may adjust the crop/compose rectangles to the nearest
> >> +      supported ones to meet codec and hardware requirements. The client needs
> >> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> >> +
> >> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> >> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``count``
> >> +         requested number of buffers to allocate; greater than zero
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> >> +         ``CAPTURE``
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``count``
> >> +          actual number of buffers allocated
> >> +
> >> +   .. warning::
> >> +
> >> +      The actual number of allocated buffers may differ from the ``count``
> >> +      given. The client must check the updated value of ``count`` after the
> >> +      call returns.
> >> +
> >> +   .. note::
> >> +
> >> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> >> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> >> +      control to get the minimum number of buffers required, and pass the
> >> +      obtained value plus the number of additional buffers needed in the
> >> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> >> +
> >> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> >> +   control over buffer allocation.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``count``
> >> +         requested number of buffers to allocate; greater than zero
> >> +
> >> +     ``type``
> >> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >> +
> >> +     other fields
> >> +         follow standard semantics
> >> +
> >> +   * **Return fields:**
> >> +
> >> +     ``count``
> >> +         adjusted to the number of allocated buffers
> >> +
> >> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> >> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> >> +   encoding process starts when both queues start streaming.
> >> +
> >> +.. note::
> >> +
> >> +   If the client stops the ``CAPTURE`` queue during the encode process and then
> >> +   restarts it again, the encoder will begin generating a stream independent
> >> +   from the stream generated before the stop. The exact constraints depend
> >> +   on the coded format, but may include the following implications:
> >> +
> >> +   * encoded frames produced after the restart must not reference any
> >> +     frames produced before the stop, e.g. no long term references for
> >> +     H.264,
> >> +
> >> +   * any headers that must be included in a standalone stream must be
> >> +     produced again, e.g. SPS and PPS for H.264.
> >> +
> >> +Encoding
> >> +========
> >> +
> >> +This state is reached after the `Initialization` sequence finishes
> >> +successfully.  In this state, the client queues and dequeues buffers to both
> >> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> >> +standard semantics.
> >> +
> >> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> >> +format and may be affected by codec-specific extended controls, as stated
> >> +in the documentation of each format.
> >> +
> >> +Both queues operate independently, following standard behavior of V4L2 buffer
> >> +queues and memory-to-memory devices. In addition, the order of encoded frames
> >> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> >> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> >> +e.g. frame reordering.
> >> +
> >> +The client must not assume any direct relationship between ``CAPTURE`` and
> >> +``OUTPUT`` buffers and any specific timing of buffers becoming
> >> +available to dequeue. Specifically:
> >> +
> >> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> >> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> >> +  frame that preceded it in display, but succeeded it in the decode order),
> >> +
> >> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> >> +  ``CAPTURE`` later into encode process, and/or after processing further
> >> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> >> +  reordering is used,
> >> +
> >> +* buffers may become available on the ``CAPTURE`` queue without additional
> >> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> >> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> >> +  available at later time, due to specifics of the decoding process,
> >> +
> >> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> >> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> >> +  encoder needs to use the frame as a reference for encoding further frames.
> >> +
> >> +.. note::
> >> +
> >> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> >> +   originated from, the client can set the ``timestamp`` field of the
> >> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> >> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> >> +   will have their ``timestamp`` field set to the same value when dequeued.
> >> +
> >> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> >> +   one ``CAPTURE`` buffer, the following cases are defined:
> >> +
> >> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> >> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> >> +
> >> +   * the encoding order differs from the presentation order (i.e. the
> >> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> >> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> >> +     and thus monotonicity of the timestamps cannot be guaranteed.
> >
> > Drop "and thus monotonicity of the timestamps cannot be guaranteed": this was never
> > guaranteed since timestamps are just copied from OUTPUT to CAPTURE.
> >
> > This phrase doesn't occur in the decoder spec, so it should be removed here as well.
> >
> >> +
> >> +.. note::
> >> +
> >> +   To let the client distinguish between frame types (keyframes, intermediate
> >> +   frames; the exact list of types depends on the coded format), the
> >> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> >> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> >> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> >> +   and their meanings.
> >> +
> >> +Encoding parameter changes
> >> +==========================
> >> +
> >> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> >> +parameters at any time. The availability of parameters is encoder-specific
> >> +and the client must query the encoder to find the set of available controls.
> >> +
> >> +The ability to change each parameter during encoding is encoder-specific, as
> >> +per the standard semantics of the V4L2 control interface. The client may
> >> +attempt to set a control during encoding and if the operation fails with the
> >> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> >> +configuration change to be allowed. To do this, it may follow the `Drain`
> >> +sequence to avoid losing the already queued/encoded frames.
> >> +
> >> +The timing of parameter updates is encoder-specific, as per the standard
> >> +semantics of the V4L2 control interface. If the client needs to apply the
> >> +parameters exactly at specific frame, using the Request API
> >> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> >> +
> >> +Drain
> >> +=====
> >> +
> >> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> >> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> >> +drain sequence described below. After the drain sequence ends, the client has
> >> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> >> +sequence was started.
> >> +
> >> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``cmd``
> >> +         set to ``V4L2_ENC_CMD_STOP``
> >> +
> >> +     ``flags``
> >> +         set to 0
> >> +
> >> +     ``pts``
> >> +         set to 0
> >> +
> >> +   .. warning::
> >> +
> >> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> >> +      queues are streaming. For compatibility reasons, the call to
> >> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> >> +      not streaming, but at the same time it will not initiate the `Drain`
> >> +      sequence and so the steps described below would not be applicable.
> >> +
> >> +2. Any ``OUTPUT`` buffers queued by the client before the
> >> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> >> +   normal. The client must continue to handle both queues independently,
> >> +   similarly to normal encode operation. This includes:
> >> +
> >> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> >> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> >> +
> >> +     .. warning::
> >> +
> >> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> >> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> >> +        as it does not contain an encoded frame.
> >> +
> >> +     .. note::
> >> +
> >> +        Any attempt to dequeue more buffers beyond the buffer marked with
> >> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> >> +        :c:func:`VIDIOC_DQBUF`.
> >> +
> >> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> >> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> >> +
> >> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> >> +
> >> +   .. note::
> >> +
> >> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> >> +      event when the last frame has been decoded and all frames are ready to be
> >> +      dequeued. It is deprecated behavior and the client must not rely on it.
> >> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> >> +
> >> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> >> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> >> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> >> +   until the client issues any of the following operations:
> >> +
> >> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> >> +     operation normally, with all the state from before the drain,
> >> +
> >> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> >> +     and then resume encoding,
> >> +
> >> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> >> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> >> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> >> +
> >> +.. note::
> >> +
> >> +   Once the drain sequence is initiated, the client needs to drive it to
> >> +   completion, as described by the steps above, unless it aborts the process by
> >> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> >> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> >> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> >> +   will fail with -EBUSY error code if attempted.
> >> +
> >> +   Although mandatory, the availability of encoder commands may be queried
> >> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> >> +
> >> +Reset
> >> +=====
> >> +
> >> +The client may want to request the encoder to reinitialize the encoding, so
> >> +that the following stream data becomes independent from the stream data
> >> +generated before. Depending on the coded format, that may imply that:
> >> +
> >> +* encoded frames produced after the restart must not reference any frames
> >> +  produced before the stop, e.g. no long term references for H.264,
> >> +
> >> +* any headers that must be included in a standalone stream must be produced
> >> +  again, e.g. SPS and PPS for H.264.
> >> +
> >> +This can be achieved by performing the reset sequence.
> >> +
> >> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
> >> +   and respective buffers are dequeued.
> >> +
> >> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
> >> +   will return all currently queued ``CAPTURE`` buffers to the client, without
> >> +   valid frame data.
> >> +
> >> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
> >> +   continue with regular encoding sequence. The encoded frames produced into
> >> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
> >> +   decoded without the need for frames encoded before the reset sequence,
> >> +   starting at the first ``OUTPUT`` buffer queued after issuing the
> >> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
> >> +
> >> +This sequence may be also used to change encoding parameters for encoders
> >> +without the ability to change the parameters on the fly.
> >> +
> >> +Commit points
> >> +=============
> >> +
> >> +Setting formats and allocating buffers triggers changes in the behavior of the
> >> +encoder.
> >> +
> >> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
> >> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
> >> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
> >> +   previously set format being preserved.
> >> +
> >> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
> >> +   supported for the current ``CAPTURE`` format.
> >> +
> >> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
> >> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
> >> +   format that is not supported for the currently selected ``CAPTURE`` format
> >> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
> >> +   supported one.
> >> +
> >> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
> >> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
> >
> > irrespectively -> irrespective
> >
> >> +
> >> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
> >> +   change the format on the queue. Drivers will return the -EBUSY error code
>
> Change the format on which queue? CAPTURE or OUTPUT?

CAPTURE was intended here. One should be able to change the format on
OUTPUT freely.

>
> I think this should be more precise: changing the CAPTURE format will always
> fail if CAPTURE buffers are already allocated, or if the OUTPUT is streaming,
> or if changing the CAPTURE format will change the OUTPUT format and OUTPUT
> buffers are already allocated.

Indeed. I believe we can generalize this to:

5. While buffers are allocated on any of the ``OUTPUT`` or ``CAPTURE`` queues,
   the client must not change the format on the ``CAPTURE`` queue. Drivers will
   return the -EBUSY error code for any such format change attempt.

>
> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
> particular) and CAPTURE buffers were already allocated and are too small.

The OUTPUT format must not change the CAPTURE format by definition.
Otherwise we end up in a situation where we can't commit, because both
queue formats can affect each other. Any change to the OUTPUT format
that wouldn't work with the current CAPTURE format should be adjusted
by the driver to match the current CAPTURE format.

Best regards,
Tomasz

>
> Regards,
>
>         Hans
>
> >> +   for any such format change attempt.
> >> +
> >> +To summarize, setting formats and allocation must always start with the
> >> +``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
> >> +set of supported formats for the ``OUTPUT`` queue.
> >> diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> >> index a0e06a88c872..82630ea89b16 100644
> >> --- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
> >> +++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> >> @@ -17,6 +17,7 @@ Video Memory-To-Memory Interface
> >>      :maxdepth: 1
> >>
> >>      dev-decoder
> >> +    dev-encoder
> >>
> >>  A V4L2 memory-to-memory device can compress, decompress, transform, or
> >>  otherwise convert video data from one format into another format, in memory.
> >> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> >> index caf14e440447..593d25b935bf 100644
> >> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> >> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> >> @@ -44,6 +44,11 @@ Single-planar format structure
> >>      inside the stream, when fed to a stateful mem2mem decoder, the fields
> >>      may be zero to rely on the decoder to detect the right values. For more
> >>      details see :ref:`decoder` and format descriptions.
> >> +
> >> +    For compressed formats on the CAPTURE side of a stateful mem2mem
> >> +    encoder, the fields must be zero, since the coded size is expected to
> >> +    be calculated internally by the encoder itself, based on the OUTPUT
> >> +    side. For more details see :ref:`encoder` and format descriptions.
> >>      * - __u32
> >>        - ``pixelformat``
> >>        - The pixel format or type of compression, set by the application.
> >> diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
> >> index 97015b9b40b8..e236ea23481b 100644
> >> --- a/Documentation/media/uapi/v4l/v4l2.rst
> >> +++ b/Documentation/media/uapi/v4l/v4l2.rst
> >> @@ -63,6 +63,7 @@ Authors, in alphabetical order:
> >>  - Figa, Tomasz <[e-mail]>
> >>
> >>    - Documented the memory-to-memory decoder interface.
> >> +  - Documented the memory-to-memory encoder interface.
> >>
> >>  - H Schimek, Michael <[e-mail]>
> >>
> >> @@ -75,6 +76,7 @@ Authors, in alphabetical order:
> >>  - Osciak, Pawel <[e-mail]>
> >>
> >>    - Documented the memory-to-memory decoder interface.
> >> +  - Documented the memory-to-memory encoder interface.
> >>
> >>  - Osciak, Pawel <[e-mail]>
> >>
> >> diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> >> index c313ca8b8cb5..2b59a4e7f5ce 100644
> >> --- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> >> +++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> >> @@ -57,19 +57,23 @@ currently only used by the STOP command and contains one bit: If the
> >>  until the end of the current *Group Of Pictures*, otherwise it will stop
> >>  immediately.
> >>
> >> -A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> >> -call sends an implicit START command to the encoder if it has not been
> >> -started yet. After a STOP command, :ref:`read() <func-read>` calls will read
> >> +After a STOP command, :ref:`read() <func-read>` calls will read
> >>  the remaining data buffered by the driver. When the buffer is empty,
> >>  :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
> >>  call will restart the encoder.
> >>
> >> +A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> >> +call sends an implicit START command to the encoder if it has not been
> >> +started yet. Applies to both queues of mem2mem encoders.
> >> +
> >>  A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
> >>  call of a streaming file descriptor sends an implicit immediate STOP to
> >> -the encoder, and all buffered data is discarded.
> >> +the encoder, and all buffered data is discarded. Applies to both queues of
> >> +mem2mem encoders.
> >>
> >>  These ioctls are optional, not all drivers may support them. They were
> >> -introduced in Linux 2.6.21.
> >> +introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
> >> +encoders (as further documented in :ref:`encoder`).
> >>
> >>
> >>  .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
> >> @@ -114,16 +118,20 @@ introduced in Linux 2.6.21.
> >>        - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
> >>      is set, encoding will continue until the end of the current *Group
> >>      Of Pictures*, otherwise encoding will stop immediately. When the
> >> -    encoder is already stopped, this command does nothing. mem2mem
> >> -    encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
> >> -    has been encoded and all frames are ready to be dequeued and will
> >> -    set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
> >> -    the capture queue to indicate there will be no new buffers
> >> -    produced to dequeue. This buffer may be empty, indicated by the
> >> -    driver setting the ``bytesused`` field to 0. Once the
> >> -    ``V4L2_BUF_FLAG_LAST`` flag was set, the
> >> -    :ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
> >> -    but return an ``EPIPE`` error code.
> >> +    encoder is already stopped, this command does nothing.
> >> +
> >> +    A stateful mem2mem encoder will proceed with encoding the source
> >> +    buffers pending before the command is issued and then stop producing
> >> +    new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
> >> +    has been encoded and all frames are ready to be dequeued and will set
> >> +    the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
> >> +    capture queue to indicate there will be no new buffers produced to
> >> +    dequeue. This buffer may be empty, indicated by the driver setting the
> >> +    ``bytesused`` field to 0. Once the buffer with the
> >> +    ``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
> >> +    <VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
> >> +    error code. No flags or other arguments are accepted in case of mem2mem
> >> +    encoders.  See :ref:`encoder` for more details.
> >>      * - ``V4L2_ENC_CMD_PAUSE``
> >>        - 2
> >>        - Pause the encoder. When the encoder has not been started yet, the
> >>
> >
> > Regards,
> >
> >       Hans
> >
>

Hans Verkuil April 5, 2019, 10:03 a.m. UTC | #9

On 4/5/19 10:12 AM, Tomasz Figa wrote:
> On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
>>
>> Hi Tomasz,
>>
>> Some more comments...
>>
>> On 1/29/19 2:52 PM, Hans Verkuil wrote:
>>> Hi Tomasz,
>>>
>>> Some comments below. Nothing major, so I think a v4 should be ready to be
>>> merged.
>>>
>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>>> Due to complexity of the video encoding process, the V4L2 drivers of
>>>> stateful encoder hardware require specific sequences of V4L2 API calls
>>>> to be followed. These include capability enumeration, initialization,
>>>> encoding, encode parameters change, drain and reset.
>>>>
>>>> Specifics of the above have been discussed during Media Workshops at
>>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>>>> originated at those events was later implemented by the drivers we already
>>>> have merged in mainline, such as s5p-mfc or coda.
>>>>
>>>> The only thing missing was the real specification included as a part of
>>>> Linux Media documentation. Fix it now and document the encoder part of
>>>> the Codec API.
>>>>
>>>> Signed-off-by: Tomasz Figa <[e-mail]>
>>>> ---
>>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>>>
>>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>> new file mode 100644
>>>> index 000000000000..fb8b05a132ee
>>>> --- /dev/null
>>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>> @@ -0,0 +1,586 @@
>>>> +.. -*- coding: utf-8; mode: rst -*-
>>>> +
>>>> +.. _encoder:
>>>> +
>>>> +*************************************************
>>>> +Memory-to-memory Stateful Video Encoder Interface
>>>> +*************************************************
>>>> +
>>>> +A stateful video encoder takes raw video frames in display order and encodes
>>>> +them into a bitstream. It generates complete chunks of the bitstream, including
>>>> +all metadata, headers, etc. The resulting bitstream does not require any
>>>> +further post-processing by the client.
>>>> +
>>>> +Performing software stream processing, header generation etc. in the driver
>>>> +in order to support this interface is strongly discouraged. In case such
>>>> +operations are needed, use of the Stateless Video Encoder Interface (in
>>>> +development) is strongly advised.
>>>> +
>>>> +Conventions and notation used in this document
>>>> +==============================================
>>>> +
>>>> +1. The general V4L2 API rules apply if not specified in this document
>>>> +   otherwise.
>>>> +
>>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
>>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
>>>> +
>>>> +3. All steps not marked "optional" are required.
>>>> +
>>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
>>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
>>>> +   unless specified otherwise.
>>>> +
>>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
>>>> +   used interchangeably with multi-planar API, unless specified otherwise,
>>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
>>>> +
>>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
>>>> +   [0..2]: i = 0, 1, 2.
>>>> +
>>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
>>>> +   queue containing data that resulted from processing buffer A.
>>>> +
>>>> +Glossary
>>>> +========
>>>> +
>>>> +Refer to :ref:`decoder-glossary`.
>>>> +
>>>> +State machine
>>>> +=============
>>>> +
>>>> +.. kernel-render:: DOT
>>>> +   :alt: DOT digraph of encoder state machine
>>>> +   :caption: Encoder state machine
>>>> +
>>>> +   digraph encoder_state_machine {
>>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
>>>> +
>>>> +       node [shape = circle, label="Initialization"] Initialization;
>>>> +       node [shape = circle, label="Stopped"] Stopped;
>>>> +       node [shape = circle, label="Drain"] Drain;
>>>> +       node [shape = circle, label="Reset"] Reset;
>>>> +
>>>> +       node [shape = point]; qi
>>>> +       qi -> Initialization [ label = "open()" ];
>>>> +
>>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
>>>> +
>>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
>>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
>>>> +       Encoding -> Encoding;
>>>> +
>>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
>>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>> +
>>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
>>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
>>>> +
>>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
>>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>> +   }
>>>> +
>>>> +Querying capabilities
>>>> +=====================
>>>> +
>>>> +1. To enumerate the set of coded formats supported by the encoder, the
>>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
>>>> +
>>>> +   * The full set of supported formats will be returned, regardless of the
>>>> +     format set on ``OUTPUT``.
>>>> +
>>>> +2. To enumerate the set of supported raw formats, the client may call
>>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
>>>> +
>>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
>>>> +     will be returned.
>>>> +
>>>> +   * In order to enumerate raw formats supported by a given coded format,
>>>> +     the client must first set that coded format on ``CAPTURE`` and then
>>>> +     enumerate the formats on ``OUTPUT``.
>>>> +
>>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
>>>> +   resolutions for a given format, passing desired pixel format in
>>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
>>>> +
>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
>>>> +     format will include all possible coded resolutions supported by the
>>>> +     encoder for given coded pixel format.
>>>> +
>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
>>>> +     will include all possible frame buffer resolutions supported by the
>>>> +     encoder for given raw pixel format and coded format currently set on
>>>> +     ``CAPTURE``.
>>>> +
>>>> +4. Supported profiles and levels for the coded format currently set on
>>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
>>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
>>>> +
>>>> +5. Any additional encoder capabilities may be discovered by querying
>>>> +   their respective controls.
>>>> +
>>>> +Initialization
>>>> +==============
>>>> +
>>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>>>> +
>>>> +     ``pixelformat``
>>>> +         the coded format to be produced
>>>> +
>>>> +     ``sizeimage``
>>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>>>> +         match hardware requirements
>>>> +
>>>> +     ``width``, ``height``
>>>> +         ignored (always zero)
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``sizeimage``
>>>> +         adjusted size of ``CAPTURE`` buffers
>>>> +
>>>> +   .. important::
>>>> +
>>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>>>> +      must adjust it afterwards.
>>>
>>> Hmm, "including resolution": if width and height are set to 0, what should the
>>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
>>> at a first reading of this paragraph it appears to be contradictory.
>>
>> I think the driver should just return the width and height of the OUTPUT
>> format. So the width and height that userspace specifies is just ignored
>> and replaced by the width and height of the OUTPUT format. After all, that's
>> what the bitstream will encode. Returning 0 for width and height would make
>> this a strange exception in V4L2 and I want to avoid that.
>>
> 
> Hmm, however, the width and height of the OUTPUT format is not what's
> actually encoded in the bitstream. The right selection rectangle
> determines that.
> 
> In one of the previous versions I though we could put the codec
> resolution as the width and height of the CAPTURE format, which would
> be the resolution of the encoded image rounded up to full macroblocks
> +/- some encoder-specific constraints. AFAIR there was some concern
> about OUTPUT format changes triggering CAPTURE format changes, but to
> be honest, I'm not sure if that's really a problem. I just decided to
> drop that for the simplicity.

I'm not sure what your point is.

The OUTPUT format has the coded resolution, so when you set the
CAPTURE format it can just copy the OUTPUT coded resolution unless the
chosen CAPTURE pixelformat can't handle that in which case both the
OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
or minimum the codec is capable of.

That said, I am fine with just leaving it up to the driver as suggested
before. Just as long as both the CAPTURE and OUTPUT formats remain valid
(i.e. width and height may never be out of range).

> 
>>>
>>>> +
>>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
>>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``pixelformat``
>>>> +         raw format supported for the coded format currently selected on
>>>> +         the ``CAPTURE`` queue.
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +3. Set the raw source format on the ``OUTPUT`` queue via
>>>> +   :c:func:`VIDIOC_S_FMT`.
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>> +
>>>> +     ``pixelformat``
>>>> +         raw format of the source
>>>> +
>>>> +     ``width``, ``height``
>>>> +         source resolution
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``width``, ``height``
>>>> +         may be adjusted by encoder to match alignment requirements, as
>>>> +         required by the currently selected formats
>>>
>>> What if the width x height is larger than the maximum supported by the
>>> selected coded format? This should probably mention that in that case the
>>> width x height is reduced to the largest allowed value. Also mention that
>>> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
>>>
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * Setting the source resolution will reset the selection rectangles to their
>>>> +     default values, based on the new resolution, as described in the step 5
>>>
>>> 5 -> 4
>>>
>>> Or just say: "as described in the next step."
>>>
>>>> +     below.
>>
>> It should also be made explicit that:
>>
>> 1) the crop rectangle will be set to the given width and height *before*
>> it is being adjusted by S_FMT.
>>
> 
> I don't think that's what we want here.
> 
> Defining the default rectangle to be exactly the same as the OUTPUT
> resolution (after the adjustment) makes the semantics consistent - not
> setting the crop rectangle gives you exactly the behavior as if there
> was no cropping involved (or supported by the encoder).

I think you are right. This seems to be what the coda driver does as well.
It is convenient to be able to just set a 1920x1080 format and have that
resolution be stored as the crop rectangle, since it avoids having to call
s_selection afterwards, but it is not really consistent with the way V4L2
works.

> 
>> Open question: should we support a compose rectangle for the CAPTURE that
>> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
>> the adjusted width and height and the compose rectangle (read-only) contains
>> the visible width and height. It's not strictly necessary, but it is
>> symmetrical.
> 
> Wouldn't it rather be the CAPTURE crop rectangle that would be of the
> same resolution of the OUTPUT compose rectangle? Then you could
> actually have the CAPTURE compose rectangle for putting that into the
> desired rectangle of the encoded stream, if the encoder supports that.
> (I don't know any that does, so probably out of concern for now.)

Yes, you are right.

But should we support this?

I actually think not for this initial version. It can be added later, I guess.

> 
>>
>> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
>> height. The CAPTURE sizeimage might change as well.
>>
>>>> +
>>>> +4. **Optional.** Set the visible resolution for the stream metadata via
>>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
>>
>> I think you should mention that this is only necessary if the crop rectangle
>> that is set when you set the format isn't what you want.
>>
> 
> Ack.
> 
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>> +
>>>> +     ``target``
>>>> +         set to ``V4L2_SEL_TGT_CROP``
>>>> +
>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
>>>> +         rectangle and may be subject to adjustment to match codec and
>>>> +         hardware constraints
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>> +         visible rectangle adjusted by the encoder
>>>> +
>>>> +   * The following selection targets are supported on ``OUTPUT``:
>>>> +
>>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>> +         equal to the full source frame, matching the active ``OUTPUT``
>>>> +         format
>>>> +
>>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>> +
>>>> +     ``V4L2_SEL_TGT_CROP``
>>>> +         rectangle within the source buffer to be encoded into the
>>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>> +
>>>> +         .. note::
>>>> +
>>>> +            A common use case for this selection target is encoding a source
>>>> +            video with a resolution that is not a multiple of a macroblock,
>>>> +            e.g.  the common 1920x1080 resolution may require the source
>>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
>>>> +            size. To avoid encoding the padding, the client needs to explicitly
>>>> +            configure this selection target to 1920x1080.
>>
>> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
>>
> 
> Sorry, which part exactly and what part of the proposal exactly? :)
> (My comment above might be related, though.)

Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
crop selection target, so this note remains valid.

> 
>>>> +
>>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
>>>> +         maximum rectangle within the coded resolution, which the cropped
>>>> +         source frame can be composed into; if the hardware does not support
>>>> +         composition or scaling, then this is always equal to the rectangle of
>>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>> +
>>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
>>>> +         equal to a rectangle of width and height matching
>>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>> +
>>>> +     ``V4L2_SEL_TGT_COMPOSE``
>>>> +         rectangle within the coded frame, which the cropped source frame
>>>> +         is to be composed into; defaults to
>>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
>>>> +         additional compose/scaling capabilities; resulting stream will
>>>> +         have this rectangle encoded as the visible rectangle in its
>>>> +         metadata
>>
>> I think the compose targets for OUTPUT are only needed if the hardware can
>> actually do scaling and/or composition. Otherwise they can (must?) be
>> dropped.
>>
> 
> Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
> userspace to learn the target visible rectangle that's going to be
> encoded in the stream metadata. If we omit it, we wouldn't have a way
> that would be consistent between encoders that can do
> scaling/composition and those that can't.

I'm not convinced about this. The standard API behavior is not to expose
functionality that the hardware can't do. So if scaling isn't possible on
the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.

I also believe it very unlikely that we'll see encoders capable of scaling
as it doesn't make much sense. I would prefer to drop this to simplify the
spec, and when we get encoders that can scale, then we can add support for
compose rectangles (and I'm sure we'll need to think about how that
influences the CAPTURE side as well).

For encoders without scaling it is the OUTPUT crop rectangle that defines
the visible rectangle.

> 
> However, with your proposal of actually having selection rectangles
> for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
> would expose a varying set of rectangles, depending on the hardware
> capability, while the CAPTURE queue would always expose its rectangle
> with that information.

I think we should keep it simple and only define selection rectangles
when really needed.

So encoders support CROP on the OUTPUT, and decoders support CAPTURE
COMPOSE (may be read-only). Nothing else.

Once support for scaling is needed (either on the encoder or decoder
side), then the spec should be enhanced. But I prefer to postpone that
until we actually have hardware that needs this.

> 
>>>> +
>>>> +   .. warning::
>>>> +
>>>> +      The encoder may adjust the crop/compose rectangles to the nearest
>>>> +      supported ones to meet codec and hardware requirements. The client needs
>>>> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
>>>> +
>>>> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
>>>> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``count``
>>>> +         requested number of buffers to allocate; greater than zero
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
>>>> +         ``CAPTURE``
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``count``
>>>> +          actual number of buffers allocated
>>>> +
>>>> +   .. warning::
>>>> +
>>>> +      The actual number of allocated buffers may differ from the ``count``
>>>> +      given. The client must check the updated value of ``count`` after the
>>>> +      call returns.
>>>> +
>>>> +   .. note::
>>>> +
>>>> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
>>>> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
>>>> +      control to get the minimum number of buffers required, and pass the
>>>> +      obtained value plus the number of additional buffers needed in the
>>>> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
>>>> +
>>>> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
>>>> +   control over buffer allocation.
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``count``
>>>> +         requested number of buffers to allocate; greater than zero
>>>> +
>>>> +     ``type``
>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>> +
>>>> +     other fields
>>>> +         follow standard semantics
>>>> +
>>>> +   * **Return fields:**
>>>> +
>>>> +     ``count``
>>>> +         adjusted to the number of allocated buffers
>>>> +
>>>> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
>>>> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
>>>> +   encoding process starts when both queues start streaming.
>>>> +
>>>> +.. note::
>>>> +
>>>> +   If the client stops the ``CAPTURE`` queue during the encode process and then
>>>> +   restarts it again, the encoder will begin generating a stream independent
>>>> +   from the stream generated before the stop. The exact constraints depend
>>>> +   on the coded format, but may include the following implications:
>>>> +
>>>> +   * encoded frames produced after the restart must not reference any
>>>> +     frames produced before the stop, e.g. no long term references for
>>>> +     H.264,
>>>> +
>>>> +   * any headers that must be included in a standalone stream must be
>>>> +     produced again, e.g. SPS and PPS for H.264.
>>>> +
>>>> +Encoding
>>>> +========
>>>> +
>>>> +This state is reached after the `Initialization` sequence finishes
>>>> +successfully.  In this state, the client queues and dequeues buffers to both
>>>> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
>>>> +standard semantics.
>>>> +
>>>> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
>>>> +format and may be affected by codec-specific extended controls, as stated
>>>> +in the documentation of each format.
>>>> +
>>>> +Both queues operate independently, following standard behavior of V4L2 buffer
>>>> +queues and memory-to-memory devices. In addition, the order of encoded frames
>>>> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
>>>> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
>>>> +e.g. frame reordering.
>>>> +
>>>> +The client must not assume any direct relationship between ``CAPTURE`` and
>>>> +``OUTPUT`` buffers and any specific timing of buffers becoming
>>>> +available to dequeue. Specifically:
>>>> +
>>>> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
>>>> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
>>>> +  frame that preceded it in display, but succeeded it in the decode order),
>>>> +
>>>> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
>>>> +  ``CAPTURE`` later into encode process, and/or after processing further
>>>> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
>>>> +  reordering is used,
>>>> +
>>>> +* buffers may become available on the ``CAPTURE`` queue without additional
>>>> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
>>>> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
>>>> +  available at later time, due to specifics of the decoding process,
>>>> +
>>>> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
>>>> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
>>>> +  encoder needs to use the frame as a reference for encoding further frames.
>>>> +
>>>> +.. note::
>>>> +
>>>> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
>>>> +   originated from, the client can set the ``timestamp`` field of the
>>>> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
>>>> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
>>>> +   will have their ``timestamp`` field set to the same value when dequeued.
>>>> +
>>>> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
>>>> +   one ``CAPTURE`` buffer, the following cases are defined:
>>>> +
>>>> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
>>>> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
>>>> +
>>>> +   * the encoding order differs from the presentation order (i.e. the
>>>> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
>>>> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
>>>> +     and thus monotonicity of the timestamps cannot be guaranteed.
>>>
>>> Drop "and thus monotonicity of the timestamps cannot be guaranteed": this was never
>>> guaranteed since timestamps are just copied from OUTPUT to CAPTURE.
>>>
>>> This phrase doesn't occur in the decoder spec, so it should be removed here as well.
>>>
>>>> +
>>>> +.. note::
>>>> +
>>>> +   To let the client distinguish between frame types (keyframes, intermediate
>>>> +   frames; the exact list of types depends on the coded format), the
>>>> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
>>>> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
>>>> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
>>>> +   and their meanings.
>>>> +
>>>> +Encoding parameter changes
>>>> +==========================
>>>> +
>>>> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
>>>> +parameters at any time. The availability of parameters is encoder-specific
>>>> +and the client must query the encoder to find the set of available controls.
>>>> +
>>>> +The ability to change each parameter during encoding is encoder-specific, as
>>>> +per the standard semantics of the V4L2 control interface. The client may
>>>> +attempt to set a control during encoding and if the operation fails with the
>>>> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
>>>> +configuration change to be allowed. To do this, it may follow the `Drain`
>>>> +sequence to avoid losing the already queued/encoded frames.
>>>> +
>>>> +The timing of parameter updates is encoder-specific, as per the standard
>>>> +semantics of the V4L2 control interface. If the client needs to apply the
>>>> +parameters exactly at specific frame, using the Request API
>>>> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
>>>> +
>>>> +Drain
>>>> +=====
>>>> +
>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>>>> +drain sequence described below. After the drain sequence ends, the client has
>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>>>> +sequence was started.
>>>> +
>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>>>> +
>>>> +   * **Required fields:**
>>>> +
>>>> +     ``cmd``
>>>> +         set to ``V4L2_ENC_CMD_STOP``
>>>> +
>>>> +     ``flags``
>>>> +         set to 0
>>>> +
>>>> +     ``pts``
>>>> +         set to 0
>>>> +
>>>> +   .. warning::
>>>> +
>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>>>> +      queues are streaming. For compatibility reasons, the call to
>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>>>> +      not streaming, but at the same time it will not initiate the `Drain`
>>>> +      sequence and so the steps described below would not be applicable.
>>>> +
>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>>>> +   normal. The client must continue to handle both queues independently,
>>>> +   similarly to normal encode operation. This includes:
>>>> +
>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>>>> +
>>>> +     .. warning::
>>>> +
>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>>>> +        as it does not contain an encoded frame.
>>>> +
>>>> +     .. note::
>>>> +
>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>>>> +        :c:func:`VIDIOC_DQBUF`.
>>>> +
>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>>>> +
>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>>>> +
>>>> +   .. note::
>>>> +
>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>>>> +      event when the last frame has been decoded and all frames are ready to be
>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>>>> +
>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>>>> +   until the client issues any of the following operations:
>>>> +
>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>>>> +     operation normally, with all the state from before the drain,
>>>> +
>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>>>> +     and then resume encoding,
>>>> +
>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>>>> +
>>>> +.. note::
>>>> +
>>>> +   Once the drain sequence is initiated, the client needs to drive it to
>>>> +   completion, as described by the steps above, unless it aborts the process by
>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>>>> +   will fail with -EBUSY error code if attempted.
>>>> +
>>>> +   Although mandatory, the availability of encoder commands may be queried
>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>>>> +
>>>> +Reset
>>>> +=====
>>>> +
>>>> +The client may want to request the encoder to reinitialize the encoding, so
>>>> +that the following stream data becomes independent from the stream data
>>>> +generated before. Depending on the coded format, that may imply that:
>>>> +
>>>> +* encoded frames produced after the restart must not reference any frames
>>>> +  produced before the stop, e.g. no long term references for H.264,
>>>> +
>>>> +* any headers that must be included in a standalone stream must be produced
>>>> +  again, e.g. SPS and PPS for H.264.
>>>> +
>>>> +This can be achieved by performing the reset sequence.
>>>> +
>>>> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
>>>> +   and respective buffers are dequeued.
>>>> +
>>>> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
>>>> +   will return all currently queued ``CAPTURE`` buffers to the client, without
>>>> +   valid frame data.
>>>> +
>>>> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
>>>> +   continue with regular encoding sequence. The encoded frames produced into
>>>> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
>>>> +   decoded without the need for frames encoded before the reset sequence,
>>>> +   starting at the first ``OUTPUT`` buffer queued after issuing the
>>>> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
>>>> +
>>>> +This sequence may be also used to change encoding parameters for encoders
>>>> +without the ability to change the parameters on the fly.
>>>> +
>>>> +Commit points
>>>> +=============
>>>> +
>>>> +Setting formats and allocating buffers triggers changes in the behavior of the
>>>> +encoder.
>>>> +
>>>> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
>>>> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
>>>> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
>>>> +   previously set format being preserved.
>>>> +
>>>> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
>>>> +   supported for the current ``CAPTURE`` format.
>>>> +
>>>> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
>>>> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
>>>> +   format that is not supported for the currently selected ``CAPTURE`` format
>>>> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
>>>> +   supported one.
>>>> +
>>>> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
>>>> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
>>>
>>> irrespectively -> irrespective
>>>
>>>> +
>>>> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
>>>> +   change the format on the queue. Drivers will return the -EBUSY error code
>>
>> Change the format on which queue? CAPTURE or OUTPUT?
> 
> CAPTURE was intended here. One should be able to change the format on
> OUTPUT freely.

Ack.

> 
>>
>> I think this should be more precise: changing the CAPTURE format will always
>> fail if CAPTURE buffers are already allocated, or if the OUTPUT is streaming,
>> or if changing the CAPTURE format will change the OUTPUT format and OUTPUT
>> buffers are already allocated.
> 
> Indeed. I believe we can generalize this to:
> 
> 5. While buffers are allocated on any of the ``OUTPUT`` or ``CAPTURE`` queues,
>    the client must not change the format on the ``CAPTURE`` queue. Drivers will
>    return the -EBUSY error code for any such format change attempt.

Sounds good.

> 
>>
>> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
>> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
>> particular) and CAPTURE buffers were already allocated and are too small.
> 
> The OUTPUT format must not change the CAPTURE format by definition.
> Otherwise we end up in a situation where we can't commit, because both
> queue formats can affect each other. Any change to the OUTPUT format
> that wouldn't work with the current CAPTURE format should be adjusted
> by the driver to match the current CAPTURE format.

But the CAPTURE format *does* depend on the OUTPUT format: if the output
resolution changes, then so does the CAPTURE resolution and esp. the
sizeimage value, since that is typically resolution dependent.

The coda driver does this as well: changing the output resolution
will update the capture resolution and sizeimage. The vicodec driver does the
same.

Setting the CAPTURE format basically just selects the codec to use, after
that you can set the OUTPUT format and read the updated CAPTURE format to
get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
change the OUTPUT format, unless the OUTPUT format is incompatible with the
newly selected codec.

Regards,

	Hans

Tomasz Figa April 8, 2019, 6:59 a.m. UTC | #10

On Thu, Mar 21, 2019 at 7:11 PM Hans Verkuil <[e-mail]> wrote:
>
> Hi Tomasz,
>
> A few more comments:
>
> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > Due to complexity of the video encoding process, the V4L2 drivers of
> > stateful encoder hardware require specific sequences of V4L2 API calls
> > to be followed. These include capability enumeration, initialization,
> > encoding, encode parameters change, drain and reset.
> >
> > Specifics of the above have been discussed during Media Workshops at
> > LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> > Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> > originated at those events was later implemented by the drivers we already
> > have merged in mainline, such as s5p-mfc or coda.
> >
> > The only thing missing was the real specification included as a part of
> > Linux Media documentation. Fix it now and document the encoder part of
> > the Codec API.
> >
> > Signed-off-by: Tomasz Figa <[e-mail]>
> > ---
> >  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >  5 files changed, 617 insertions(+), 15 deletions(-)
> >  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >
> > diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> > new file mode 100644
> > index 000000000000..fb8b05a132ee
> > --- /dev/null
> > +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> > @@ -0,0 +1,586 @@
> > +.. -*- coding: utf-8; mode: rst -*-
> > +
> > +.. _encoder:
> > +
> > +*************************************************
> > +Memory-to-memory Stateful Video Encoder Interface
> > +*************************************************
> > +
> > +A stateful video encoder takes raw video frames in display order and encodes
> > +them into a bitstream. It generates complete chunks of the bitstream, including
> > +all metadata, headers, etc. The resulting bitstream does not require any
> > +further post-processing by the client.
> > +
> > +Performing software stream processing, header generation etc. in the driver
> > +in order to support this interface is strongly discouraged. In case such
> > +operations are needed, use of the Stateless Video Encoder Interface (in
> > +development) is strongly advised.
> > +
> > +Conventions and notation used in this document
> > +==============================================
> > +
> > +1. The general V4L2 API rules apply if not specified in this document
> > +   otherwise.
> > +
> > +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> > +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> > +
> > +3. All steps not marked "optional" are required.
> > +
> > +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> > +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> > +   unless specified otherwise.
> > +
> > +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> > +   used interchangeably with multi-planar API, unless specified otherwise,
> > +   depending on decoder capabilities and following the general V4L2 guidelines.
>
> decoder -> encoder
>

Ack.

> > +
> > +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> > +   [0..2]: i = 0, 1, 2.
> > +
> > +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> > +   queue containing data that resulted from processing buffer A.
> > +
> > +Glossary
> > +========
> > +
> > +Refer to :ref:`decoder-glossary`.
> > +
> > +State machine
> > +=============
> > +
> > +.. kernel-render:: DOT
> > +   :alt: DOT digraph of encoder state machine
> > +   :caption: Encoder state machine
> > +
> > +   digraph encoder_state_machine {
> > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > +
> > +       node [shape = circle, label="Initialization"] Initialization;
> > +       node [shape = circle, label="Stopped"] Stopped;
> > +       node [shape = circle, label="Drain"] Drain;
> > +       node [shape = circle, label="Reset"] Reset;
> > +
> > +       node [shape = point]; qi
> > +       qi -> Initialization [ label = "open()" ];
> > +
> > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > +
> > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > +       Encoding -> Encoding;
> > +
> > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> > +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +
> > +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> > +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> > +
> > +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> > +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +   }
> > +
> > +Querying capabilities
> > +=====================
> > +
> > +1. To enumerate the set of coded formats supported by the encoder, the
> > +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> > +
> > +   * The full set of supported formats will be returned, regardless of the
> > +     format set on ``OUTPUT``.
> > +
> > +2. To enumerate the set of supported raw formats, the client may call
> > +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> > +
> > +   * Only the formats supported for the format currently active on ``CAPTURE``
> > +     will be returned.
> > +
> > +   * In order to enumerate raw formats supported by a given coded format,
> > +     the client must first set that coded format on ``CAPTURE`` and then
> > +     enumerate the formats on ``OUTPUT``.
> > +
> > +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> > +   resolutions for a given format, passing desired pixel format in
> > +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> > +
> > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> > +     format will include all possible coded resolutions supported by the
> > +     encoder for given coded pixel format.
> > +
> > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> > +     will include all possible frame buffer resolutions supported by the
> > +     encoder for given raw pixel format and coded format currently set on
> > +     ``CAPTURE``.
> > +
> > +4. Supported profiles and levels for the coded format currently set on
> > +   ``CAPTURE``, if applicable, may be queried using their respective controls
> > +   via :c:func:`VIDIOC_QUERYCTRL`.
> > +
> > +5. Any additional encoder capabilities may be discovered by querying
> > +   their respective controls.
> > +
> > +Initialization
> > +==============
> > +
> > +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> > +
> > +     ``pixelformat``
> > +         the coded format to be produced
> > +
> > +     ``sizeimage``
> > +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> > +         match hardware requirements
> > +
> > +     ``width``, ``height``
> > +         ignored (always zero)
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``sizeimage``
> > +         adjusted size of ``CAPTURE`` buffers
> > +
> > +   .. important::
> > +
> > +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> > +      format. The encoder will derive a new ``OUTPUT`` format from the
> > +      ``CAPTURE`` format being set, including resolution, colorimetry
> > +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> > +      must adjust it afterwards.
> > +
> > +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> > +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``pixelformat``
> > +         raw format supported for the coded format currently selected on
> > +         the ``CAPTURE`` queue.
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +3. Set the raw source format on the ``OUTPUT`` queue via
> > +   :c:func:`VIDIOC_S_FMT`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     ``pixelformat``
> > +         raw format of the source
> > +
> > +     ``width``, ``height``
> > +         source resolution
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``width``, ``height``
> > +         may be adjusted by encoder to match alignment requirements, as
> > +         required by the currently selected formats
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * Setting the source resolution will reset the selection rectangles to their
> > +     default values, based on the new resolution, as described in the step 5
> > +     below.
> > +
> > +4. **Optional.** Set the visible resolution for the stream metadata via
> > +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> > +
> > +   * **Required fields:**
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     ``target``
> > +         set to ``V4L2_SEL_TGT_CROP``
> > +
> > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> > +         rectangle and may be subject to adjustment to match codec and
> > +         hardware constraints
> > +
> > +   * **Return fields:**
> > +
> > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > +         visible rectangle adjusted by the encoder
> > +
> > +   * The following selection targets are supported on ``OUTPUT``:
> > +
> > +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> > +         equal to the full source frame, matching the active ``OUTPUT``
> > +         format
> > +
> > +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> > +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> > +
> > +     ``V4L2_SEL_TGT_CROP``
> > +         rectangle within the source buffer to be encoded into the
> > +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> > +
> > +         .. note::
> > +
> > +            A common use case for this selection target is encoding a source
> > +            video with a resolution that is not a multiple of a macroblock,
> > +            e.g.  the common 1920x1080 resolution may require the source
> > +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> > +            size. To avoid encoding the padding, the client needs to explicitly
> > +            configure this selection target to 1920x1080.
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> > +         maximum rectangle within the coded resolution, which the cropped
> > +         source frame can be composed into; if the hardware does not support
> > +         composition or scaling, then this is always equal to the rectangle of
> > +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> > +         equal to a rectangle of width and height matching
> > +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > +
> > +     ``V4L2_SEL_TGT_COMPOSE``
> > +         rectangle within the coded frame, which the cropped source frame
> > +         is to be composed into; defaults to
> > +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> > +         additional compose/scaling capabilities; resulting stream will
> > +         have this rectangle encoded as the visible rectangle in its
> > +         metadata
>
> I would only support the COMPOSE targets if the hardware can actually do
> scaling and/or composing. That is conform standard V4L2 behavior where
> cropping/composing is only implemented if the hardware can actually do
> this.
>

Please see my other reply to your earlier similar comment in this thread.

> > +
> > +   .. warning::
> > +
> > +      The encoder may adjust the crop/compose rectangles to the nearest
> > +      supported ones to meet codec and hardware requirements. The client needs
> > +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> > +
> > +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> > +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> > +
> > +   * **Required fields:**
> > +
> > +     ``count``
> > +         requested number of buffers to allocate; greater than zero
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> > +         ``CAPTURE``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``count``
> > +          actual number of buffers allocated
> > +
> > +   .. warning::
> > +
> > +      The actual number of allocated buffers may differ from the ``count``
> > +      given. The client must check the updated value of ``count`` after the
> > +      call returns.
> > +
> > +   .. note::
> > +
> > +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> > +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> > +      control to get the minimum number of buffers required, and pass the
> > +      obtained value plus the number of additional buffers needed in the
> > +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> > +
> > +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> > +   control over buffer allocation.
> > +
> > +   * **Required fields:**
> > +
> > +     ``count``
> > +         requested number of buffers to allocate; greater than zero
> > +
> > +     ``type``
> > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > +
> > +     other fields
> > +         follow standard semantics
> > +
> > +   * **Return fields:**
> > +
> > +     ``count``
> > +         adjusted to the number of allocated buffers
> > +
> > +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> > +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> > +   encoding process starts when both queues start streaming.
> > +
> > +.. note::
> > +
> > +   If the client stops the ``CAPTURE`` queue during the encode process and then
> > +   restarts it again, the encoder will begin generating a stream independent
> > +   from the stream generated before the stop. The exact constraints depend
> > +   on the coded format, but may include the following implications:
> > +
> > +   * encoded frames produced after the restart must not reference any
> > +     frames produced before the stop, e.g. no long term references for
> > +     H.264,
> > +
> > +   * any headers that must be included in a standalone stream must be
> > +     produced again, e.g. SPS and PPS for H.264.
> > +
> > +Encoding
> > +========
> > +
> > +This state is reached after the `Initialization` sequence finishes
> > +successfully.  In this state, the client queues and dequeues buffers to both
> > +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> > +standard semantics.
> > +
> > +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> > +format and may be affected by codec-specific extended controls, as stated
> > +in the documentation of each format.
> > +
> > +Both queues operate independently, following standard behavior of V4L2 buffer
> > +queues and memory-to-memory devices. In addition, the order of encoded frames
> > +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> > +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> > +e.g. frame reordering.
> > +
> > +The client must not assume any direct relationship between ``CAPTURE`` and
> > +``OUTPUT`` buffers and any specific timing of buffers becoming
> > +available to dequeue. Specifically:
> > +
> > +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> > +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> > +  frame that preceded it in display, but succeeded it in the decode order),
> > +
> > +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> > +  ``CAPTURE`` later into encode process, and/or after processing further
> > +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> > +  reordering is used,
> > +
> > +* buffers may become available on the ``CAPTURE`` queue without additional
> > +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> > +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> > +  available at later time, due to specifics of the decoding process,
> > +
> > +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> > +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> > +  encoder needs to use the frame as a reference for encoding further frames.
> > +
> > +.. note::
> > +
> > +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> > +   originated from, the client can set the ``timestamp`` field of the
> > +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> > +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> > +   will have their ``timestamp`` field set to the same value when dequeued.
> > +
> > +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> > +   one ``CAPTURE`` buffer, the following cases are defined:
> > +
> > +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> > +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> > +
> > +   * the encoding order differs from the presentation order (i.e. the
> > +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> > +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> > +     and thus monotonicity of the timestamps cannot be guaranteed.
> > +
> > +.. note::
> > +
> > +   To let the client distinguish between frame types (keyframes, intermediate
> > +   frames; the exact list of types depends on the coded format), the
> > +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> > +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> > +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> > +   and their meanings.
>
> I don't think we can require this since a capture buffer may contain multiple
> encoded frames.
>

I thought we required that only one encoded frame was in one CAPTURE
buffer. Real time use cases rely heavily on this frame type
information, so I can't imagine not requiring this.

> It would actually make more sense to return it in the output buffer, but I don't
> know if a hardware encoder can actually provide that information.
>

I believe all the already existing drivers provide the information
about the encoded frame type, but I don't think they provide the
information about what source frame it came from.

> Another use of these flags for an output buffer is to force a keyframe if for
> example a scene change was detected.
>
> My feeling is that we should drop this note. Forcing a keyframe by setting that
> flag for the output buffer might actually be a useful thing to do for a stateful
> encoder.
>

However, to force keyframe, one sets it in the OUTPUT buffer. Then, to
actually get the right CAPTURE buffer, one has to look for one with
this flag set.

> > +
> > +Encoding parameter changes
> > +==========================
> > +
> > +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> > +parameters at any time. The availability of parameters is encoder-specific
> > +and the client must query the encoder to find the set of available controls.
> > +
> > +The ability to change each parameter during encoding is encoder-specific, as
> > +per the standard semantics of the V4L2 control interface. The client may
> > +attempt to set a control during encoding and if the operation fails with the
> > +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> > +configuration change to be allowed. To do this, it may follow the `Drain`
> > +sequence to avoid losing the already queued/encoded frames.
> > +
> > +The timing of parameter updates is encoder-specific, as per the standard
> > +semantics of the V4L2 control interface. If the client needs to apply the
> > +parameters exactly at specific frame, using the Request API
> > +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> > +
> > +Drain
> > +=====
> > +
> > +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> > +related ``CAPTURE`` buffers are given to the client, the client must follow the
> > +drain sequence described below. After the drain sequence ends, the client has
> > +received all encoded frames for all ``OUTPUT`` buffers queued before the
> > +sequence was started.
> > +
> > +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``cmd``
> > +         set to ``V4L2_ENC_CMD_STOP``
> > +
> > +     ``flags``
> > +         set to 0
> > +
> > +     ``pts``
> > +         set to 0
> > +
> > +   .. warning::
> > +
> > +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> > +      queues are streaming. For compatibility reasons, the call to
> > +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> > +      not streaming, but at the same time it will not initiate the `Drain`
> > +      sequence and so the steps described below would not be applicable.
> > +
> > +2. Any ``OUTPUT`` buffers queued by the client before the
> > +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> > +   normal. The client must continue to handle both queues independently,
> > +   similarly to normal encode operation. This includes:
> > +
> > +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> > +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> > +
> > +     .. warning::
> > +
> > +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> > +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> > +        as it does not contain an encoded frame.
> > +
> > +     .. note::
> > +
> > +        Any attempt to dequeue more buffers beyond the buffer marked with
> > +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> > +        :c:func:`VIDIOC_DQBUF`.
> > +
> > +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> > +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> > +
> > +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> > +
> > +   .. note::
> > +
> > +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> > +      event when the last frame has been decoded and all frames are ready to be
>
> decoded -> encoded
>

Ack.

Best regards,
Tomasz

Hans Verkuil April 8, 2019, 7:43 a.m. UTC | #11

On 4/8/19 8:59 AM, Tomasz Figa wrote:
> On Thu, Mar 21, 2019 at 7:11 PM Hans Verkuil <[e-mail]> wrote:
>>
>> Hi Tomasz,
>>
>> A few more comments:
>>
>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>> Due to complexity of the video encoding process, the V4L2 drivers of
>>> stateful encoder hardware require specific sequences of V4L2 API calls
>>> to be followed. These include capability enumeration, initialization,
>>> encoding, encode parameters change, drain and reset.
>>>
>>> Specifics of the above have been discussed during Media Workshops at
>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>>> originated at those events was later implemented by the drivers we already
>>> have merged in mainline, such as s5p-mfc or coda.
>>>
>>> The only thing missing was the real specification included as a part of
>>> Linux Media documentation. Fix it now and document the encoder part of
>>> the Codec API.
>>>
>>> Signed-off-by: Tomasz Figa <[e-mail]>
>>> ---
>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>>
>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>>> new file mode 100644
>>> index 000000000000..fb8b05a132ee
>>> --- /dev/null
>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>>> @@ -0,0 +1,586 @@
>>> +.. -*- coding: utf-8; mode: rst -*-
>>> +
>>> +.. _encoder:
>>> +
>>> +*************************************************
>>> +Memory-to-memory Stateful Video Encoder Interface
>>> +*************************************************
>>> +
>>> +A stateful video encoder takes raw video frames in display order and encodes
>>> +them into a bitstream. It generates complete chunks of the bitstream, including
>>> +all metadata, headers, etc. The resulting bitstream does not require any
>>> +further post-processing by the client.
>>> +
>>> +Performing software stream processing, header generation etc. in the driver
>>> +in order to support this interface is strongly discouraged. In case such
>>> +operations are needed, use of the Stateless Video Encoder Interface (in
>>> +development) is strongly advised.
>>> +
>>> +Conventions and notation used in this document
>>> +==============================================
>>> +
>>> +1. The general V4L2 API rules apply if not specified in this document
>>> +   otherwise.
>>> +
>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
>>> +
>>> +3. All steps not marked "optional" are required.
>>> +
>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
>>> +   unless specified otherwise.
>>> +
>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
>>> +   used interchangeably with multi-planar API, unless specified otherwise,
>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
>>
>> decoder -> encoder
>>
> 
> Ack.
> 
>>> +
>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
>>> +   [0..2]: i = 0, 1, 2.
>>> +
>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
>>> +   queue containing data that resulted from processing buffer A.
>>> +
>>> +Glossary
>>> +========
>>> +
>>> +Refer to :ref:`decoder-glossary`.
>>> +
>>> +State machine
>>> +=============
>>> +
>>> +.. kernel-render:: DOT
>>> +   :alt: DOT digraph of encoder state machine
>>> +   :caption: Encoder state machine
>>> +
>>> +   digraph encoder_state_machine {
>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
>>> +
>>> +       node [shape = circle, label="Initialization"] Initialization;
>>> +       node [shape = circle, label="Stopped"] Stopped;
>>> +       node [shape = circle, label="Drain"] Drain;
>>> +       node [shape = circle, label="Reset"] Reset;
>>> +
>>> +       node [shape = point]; qi
>>> +       qi -> Initialization [ label = "open()" ];
>>> +
>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
>>> +
>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
>>> +       Encoding -> Encoding;
>>> +
>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>> +
>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
>>> +
>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>> +   }
>>> +
>>> +Querying capabilities
>>> +=====================
>>> +
>>> +1. To enumerate the set of coded formats supported by the encoder, the
>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
>>> +
>>> +   * The full set of supported formats will be returned, regardless of the
>>> +     format set on ``OUTPUT``.
>>> +
>>> +2. To enumerate the set of supported raw formats, the client may call
>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
>>> +
>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
>>> +     will be returned.
>>> +
>>> +   * In order to enumerate raw formats supported by a given coded format,
>>> +     the client must first set that coded format on ``CAPTURE`` and then
>>> +     enumerate the formats on ``OUTPUT``.
>>> +
>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
>>> +   resolutions for a given format, passing desired pixel format in
>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
>>> +
>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
>>> +     format will include all possible coded resolutions supported by the
>>> +     encoder for given coded pixel format.
>>> +
>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
>>> +     will include all possible frame buffer resolutions supported by the
>>> +     encoder for given raw pixel format and coded format currently set on
>>> +     ``CAPTURE``.
>>> +
>>> +4. Supported profiles and levels for the coded format currently set on
>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
>>> +
>>> +5. Any additional encoder capabilities may be discovered by querying
>>> +   their respective controls.
>>> +
>>> +Initialization
>>> +==============
>>> +
>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>>> +
>>> +     ``pixelformat``
>>> +         the coded format to be produced
>>> +
>>> +     ``sizeimage``
>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>>> +         match hardware requirements
>>> +
>>> +     ``width``, ``height``
>>> +         ignored (always zero)
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``sizeimage``
>>> +         adjusted size of ``CAPTURE`` buffers
>>> +
>>> +   .. important::
>>> +
>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>>> +      must adjust it afterwards.
>>> +
>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``pixelformat``
>>> +         raw format supported for the coded format currently selected on
>>> +         the ``CAPTURE`` queue.
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +3. Set the raw source format on the ``OUTPUT`` queue via
>>> +   :c:func:`VIDIOC_S_FMT`.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>> +
>>> +     ``pixelformat``
>>> +         raw format of the source
>>> +
>>> +     ``width``, ``height``
>>> +         source resolution
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``width``, ``height``
>>> +         may be adjusted by encoder to match alignment requirements, as
>>> +         required by the currently selected formats
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * Setting the source resolution will reset the selection rectangles to their
>>> +     default values, based on the new resolution, as described in the step 5
>>> +     below.
>>> +
>>> +4. **Optional.** Set the visible resolution for the stream metadata via
>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>> +
>>> +     ``target``
>>> +         set to ``V4L2_SEL_TGT_CROP``
>>> +
>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
>>> +         rectangle and may be subject to adjustment to match codec and
>>> +         hardware constraints
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>> +         visible rectangle adjusted by the encoder
>>> +
>>> +   * The following selection targets are supported on ``OUTPUT``:
>>> +
>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
>>> +         equal to the full source frame, matching the active ``OUTPUT``
>>> +         format
>>> +
>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
>>> +
>>> +     ``V4L2_SEL_TGT_CROP``
>>> +         rectangle within the source buffer to be encoded into the
>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
>>> +
>>> +         .. note::
>>> +
>>> +            A common use case for this selection target is encoding a source
>>> +            video with a resolution that is not a multiple of a macroblock,
>>> +            e.g.  the common 1920x1080 resolution may require the source
>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
>>> +            size. To avoid encoding the padding, the client needs to explicitly
>>> +            configure this selection target to 1920x1080.
>>> +
>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
>>> +         maximum rectangle within the coded resolution, which the cropped
>>> +         source frame can be composed into; if the hardware does not support
>>> +         composition or scaling, then this is always equal to the rectangle of
>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>> +
>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
>>> +         equal to a rectangle of width and height matching
>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>> +
>>> +     ``V4L2_SEL_TGT_COMPOSE``
>>> +         rectangle within the coded frame, which the cropped source frame
>>> +         is to be composed into; defaults to
>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
>>> +         additional compose/scaling capabilities; resulting stream will
>>> +         have this rectangle encoded as the visible rectangle in its
>>> +         metadata
>>
>> I would only support the COMPOSE targets if the hardware can actually do
>> scaling and/or composing. That is conform standard V4L2 behavior where
>> cropping/composing is only implemented if the hardware can actually do
>> this.
>>
> 
> Please see my other reply to your earlier similar comment in this thread.
> 
>>> +
>>> +   .. warning::
>>> +
>>> +      The encoder may adjust the crop/compose rectangles to the nearest
>>> +      supported ones to meet codec and hardware requirements. The client needs
>>> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
>>> +
>>> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
>>> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``count``
>>> +         requested number of buffers to allocate; greater than zero
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
>>> +         ``CAPTURE``
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``count``
>>> +          actual number of buffers allocated
>>> +
>>> +   .. warning::
>>> +
>>> +      The actual number of allocated buffers may differ from the ``count``
>>> +      given. The client must check the updated value of ``count`` after the
>>> +      call returns.
>>> +
>>> +   .. note::
>>> +
>>> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
>>> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
>>> +      control to get the minimum number of buffers required, and pass the
>>> +      obtained value plus the number of additional buffers needed in the
>>> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
>>> +
>>> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
>>> +   control over buffer allocation.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``count``
>>> +         requested number of buffers to allocate; greater than zero
>>> +
>>> +     ``type``
>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>> +
>>> +     other fields
>>> +         follow standard semantics
>>> +
>>> +   * **Return fields:**
>>> +
>>> +     ``count``
>>> +         adjusted to the number of allocated buffers
>>> +
>>> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
>>> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
>>> +   encoding process starts when both queues start streaming.
>>> +
>>> +.. note::
>>> +
>>> +   If the client stops the ``CAPTURE`` queue during the encode process and then
>>> +   restarts it again, the encoder will begin generating a stream independent
>>> +   from the stream generated before the stop. The exact constraints depend
>>> +   on the coded format, but may include the following implications:
>>> +
>>> +   * encoded frames produced after the restart must not reference any
>>> +     frames produced before the stop, e.g. no long term references for
>>> +     H.264,
>>> +
>>> +   * any headers that must be included in a standalone stream must be
>>> +     produced again, e.g. SPS and PPS for H.264.
>>> +
>>> +Encoding
>>> +========
>>> +
>>> +This state is reached after the `Initialization` sequence finishes
>>> +successfully.  In this state, the client queues and dequeues buffers to both
>>> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
>>> +standard semantics.
>>> +
>>> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
>>> +format and may be affected by codec-specific extended controls, as stated
>>> +in the documentation of each format.
>>> +
>>> +Both queues operate independently, following standard behavior of V4L2 buffer
>>> +queues and memory-to-memory devices. In addition, the order of encoded frames
>>> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
>>> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
>>> +e.g. frame reordering.
>>> +
>>> +The client must not assume any direct relationship between ``CAPTURE`` and
>>> +``OUTPUT`` buffers and any specific timing of buffers becoming
>>> +available to dequeue. Specifically:
>>> +
>>> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
>>> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
>>> +  frame that preceded it in display, but succeeded it in the decode order),
>>> +
>>> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
>>> +  ``CAPTURE`` later into encode process, and/or after processing further
>>> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
>>> +  reordering is used,
>>> +
>>> +* buffers may become available on the ``CAPTURE`` queue without additional
>>> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
>>> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
>>> +  available at later time, due to specifics of the decoding process,
>>> +
>>> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
>>> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
>>> +  encoder needs to use the frame as a reference for encoding further frames.
>>> +
>>> +.. note::
>>> +
>>> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
>>> +   originated from, the client can set the ``timestamp`` field of the
>>> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
>>> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
>>> +   will have their ``timestamp`` field set to the same value when dequeued.
>>> +
>>> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
>>> +   one ``CAPTURE`` buffer, the following cases are defined:
>>> +
>>> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
>>> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
>>> +
>>> +   * the encoding order differs from the presentation order (i.e. the
>>> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
>>> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
>>> +     and thus monotonicity of the timestamps cannot be guaranteed.
>>> +
>>> +.. note::
>>> +
>>> +   To let the client distinguish between frame types (keyframes, intermediate
>>> +   frames; the exact list of types depends on the coded format), the
>>> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
>>> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
>>> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
>>> +   and their meanings.
>>
>> I don't think we can require this since a capture buffer may contain multiple
>> encoded frames.
>>
> 
> I thought we required that only one encoded frame was in one CAPTURE
> buffer. Real time use cases rely heavily on this frame type
> information, so I can't imagine not requiring this.

That the CAPTURE buffer contains only one encoded frame is never stated
explicitly. I am not so sure I want that to be a hard requirement anyway
since the old ivtv MPEG encoder just produces a bitstream.

Perhaps this should be signaled with a flag in ENUM_FMT?

> 
>> It would actually make more sense to return it in the output buffer, but I don't
>> know if a hardware encoder can actually provide that information.
>>
> 
> I believe all the already existing drivers provide the information
> about the encoded frame type, but I don't think they provide the
> information about what source frame it came from.
> 
>> Another use of these flags for an output buffer is to force a keyframe if for
>> example a scene change was detected.
>>
>> My feeling is that we should drop this note. Forcing a keyframe by setting that
>> flag for the output buffer might actually be a useful thing to do for a stateful
>> encoder.
>>
> 
> However, to force keyframe, one sets it in the OUTPUT buffer. Then, to
> actually get the right CAPTURE buffer, one has to look for one with
> this flag set.

So *if* the driver stores only one encoded frame in a CAPTURE buffer, then we
can require that these flags have to be set for that CAPTURE buffer. Otherwise
they should be cleared since they cannot be associated with a specific buffer.

And I think it should be documented that you can set the KEYFRAME flag in the
OUTPUT buffer to force a keyframe (the driver may ignore this if it can't do
this for some reason).

Regards,

	Hans

> 
>>> +
>>> +Encoding parameter changes
>>> +==========================
>>> +
>>> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
>>> +parameters at any time. The availability of parameters is encoder-specific
>>> +and the client must query the encoder to find the set of available controls.
>>> +
>>> +The ability to change each parameter during encoding is encoder-specific, as
>>> +per the standard semantics of the V4L2 control interface. The client may
>>> +attempt to set a control during encoding and if the operation fails with the
>>> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
>>> +configuration change to be allowed. To do this, it may follow the `Drain`
>>> +sequence to avoid losing the already queued/encoded frames.
>>> +
>>> +The timing of parameter updates is encoder-specific, as per the standard
>>> +semantics of the V4L2 control interface. If the client needs to apply the
>>> +parameters exactly at specific frame, using the Request API
>>> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
>>> +
>>> +Drain
>>> +=====
>>> +
>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>>> +drain sequence described below. After the drain sequence ends, the client has
>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>>> +sequence was started.
>>> +
>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``cmd``
>>> +         set to ``V4L2_ENC_CMD_STOP``
>>> +
>>> +     ``flags``
>>> +         set to 0
>>> +
>>> +     ``pts``
>>> +         set to 0
>>> +
>>> +   .. warning::
>>> +
>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>>> +      queues are streaming. For compatibility reasons, the call to
>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>>> +      not streaming, but at the same time it will not initiate the `Drain`
>>> +      sequence and so the steps described below would not be applicable.
>>> +
>>> +2. Any ``OUTPUT`` buffers queued by the client before the
>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>>> +   normal. The client must continue to handle both queues independently,
>>> +   similarly to normal encode operation. This includes:
>>> +
>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>>> +
>>> +     .. warning::
>>> +
>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>>> +        as it does not contain an encoded frame.
>>> +
>>> +     .. note::
>>> +
>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>>> +        :c:func:`VIDIOC_DQBUF`.
>>> +
>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>>> +
>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>>> +
>>> +   .. note::
>>> +
>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>>> +      event when the last frame has been decoded and all frames are ready to be
>>
>> decoded -> encoded
>>
> 
> Ack.
> 
> Best regards,
> Tomasz
>

Tomasz Figa April 8, 2019, 8:36 a.m. UTC | #12

On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
>
> Another comment found while creating compliance tests:
>
> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > +Drain
> > +=====
> > +
> > +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> > +related ``CAPTURE`` buffers are given to the client, the client must follow the
> > +drain sequence described below. After the drain sequence ends, the client has
> > +received all encoded frames for all ``OUTPUT`` buffers queued before the
> > +sequence was started.
> > +
> > +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> > +
> > +   * **Required fields:**
> > +
> > +     ``cmd``
> > +         set to ``V4L2_ENC_CMD_STOP``
> > +
> > +     ``flags``
> > +         set to 0
> > +
> > +     ``pts``
> > +         set to 0
> > +
> > +   .. warning::
> > +
> > +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> > +      queues are streaming. For compatibility reasons, the call to
> > +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> > +      not streaming, but at the same time it will not initiate the `Drain`
> > +      sequence and so the steps described below would not be applicable.
> > +
> > +2. Any ``OUTPUT`` buffers queued by the client before the
> > +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> > +   normal. The client must continue to handle both queues independently,
> > +   similarly to normal encode operation. This includes:
> > +
> > +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> > +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> > +
> > +     .. warning::
> > +
> > +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> > +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> > +        as it does not contain an encoded frame.
> > +
> > +     .. note::
> > +
> > +        Any attempt to dequeue more buffers beyond the buffer marked with
> > +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> > +        :c:func:`VIDIOC_DQBUF`.
> > +
> > +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> > +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> > +
> > +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> > +
> > +   .. note::
> > +
> > +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> > +      event when the last frame has been decoded and all frames are ready to be
> > +      dequeued. It is deprecated behavior and the client must not rely on it.
> > +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> > +
> > +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> > +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> > +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> > +   until the client issues any of the following operations:
> > +
> > +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> > +     operation normally, with all the state from before the drain,
>
> I assume that calling CMD_START when *not* draining will succeed but does nothing.
>
> In other words: while draining is in progress START will return EBUSY. When draining
> was finished, then START will resume the encoder. In all other cases it just returns
> 0 since the encoder is really already started.
>

Yes, that was the intention and seems to be the closest to the
behavior described in the existing documentation.

> > +
> > +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> > +     and then resume encoding,
> > +
> > +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> > +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> > +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> > +
> > +.. note::
> > +
> > +   Once the drain sequence is initiated, the client needs to drive it to
> > +   completion, as described by the steps above, unless it aborts the process by
> > +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> > +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> > +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> > +   will fail with -EBUSY error code if attempted.
>
> I assume calling STOP again once the drain sequence completed just returns 0 and
> doesn't do anything else (since we're already stopped).
>

Right.

> > +
> > +   Although mandatory, the availability of encoder commands may be queried
> > +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>
> Some corner cases:
>
> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
>    Now ENC_CMD_STOP is issued. What should happen?
>
>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>    at once as an empty buffer with FLAG_LAST set.
>

SGTM. It's actually similar to a relatively common case where all
CAPTURE buffers have been dequeued and the application has to refill
the CAPTURE queue, but in the meantime a drain request needs to be
issued.

> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
>
>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
>    previous corner case.

Agreed.

>
> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
>    is issued. What should happen?
>
>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
>    buffer with FLAG_LAST set.

Generally agreed, but not sure if there is a reason to specifically
refer to the oldest buffer. (I'm personally for keeping the queues
ordered, though...)

>
> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
>    Before the drain process completes STREAMOFF is called for either CAPTURE or
>    OUTPUT queue. What should happen?
>
>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
>    will be used as input to the encoder.
>

Agreed.

>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
>    FLAG_LAST set.

Agreed.

>
> Some of this might have to be documented, but these corner cases should certainly be
> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> with this.

Agreed with just one minor comment. Thanks for checking with me!

Best regards,
Tomasz

Tomasz Figa April 8, 2019, 8:40 a.m. UTC | #13

On Tue, Mar 26, 2019 at 1:33 AM Hans Verkuil <[e-mail]> wrote:
>
> On 3/25/19 2:12 PM, Hans Verkuil wrote:
> > Another comment found while creating compliance tests:
> >
> > On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >> +Drain
> >> +=====
> >> +
> >> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> >> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> >> +drain sequence described below. After the drain sequence ends, the client has
> >> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> >> +sequence was started.
> >> +
> >> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> >> +
> >> +   * **Required fields:**
> >> +
> >> +     ``cmd``
> >> +         set to ``V4L2_ENC_CMD_STOP``
> >> +
> >> +     ``flags``
> >> +         set to 0
> >> +
> >> +     ``pts``
> >> +         set to 0
> >> +
> >> +   .. warning::
> >> +
> >> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> >> +      queues are streaming. For compatibility reasons, the call to
> >> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> >> +      not streaming, but at the same time it will not initiate the `Drain`
> >> +      sequence and so the steps described below would not be applicable.
> >> +
> >> +2. Any ``OUTPUT`` buffers queued by the client before the
> >> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> >> +   normal. The client must continue to handle both queues independently,
> >> +   similarly to normal encode operation. This includes:
> >> +
> >> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> >> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> >> +
> >> +     .. warning::
> >> +
> >> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> >> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> >> +        as it does not contain an encoded frame.
> >> +
> >> +     .. note::
> >> +
> >> +        Any attempt to dequeue more buffers beyond the buffer marked with
> >> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> >> +        :c:func:`VIDIOC_DQBUF`.
> >> +
> >> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> >> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> >> +
> >> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> >> +
> >> +   .. note::
> >> +
> >> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> >> +      event when the last frame has been decoded and all frames are ready to be
> >> +      dequeued. It is deprecated behavior and the client must not rely on it.
> >> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> >> +
> >> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> >> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> >> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> >> +   until the client issues any of the following operations:
> >> +
> >> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> >> +     operation normally, with all the state from before the drain,
> >
> > I assume that calling CMD_START when *not* draining will succeed but does nothing.
> >
> > In other words: while draining is in progress START will return EBUSY. When draining
> > was finished, then START will resume the encoder. In all other cases it just returns
> > 0 since the encoder is really already started.
> >
> >> +
> >> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> >> +     and then resume encoding,
> >> +
> >> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> >> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> >> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> >> +
> >> +.. note::
> >> +
> >> +   Once the drain sequence is initiated, the client needs to drive it to
> >> +   completion, as described by the steps above, unless it aborts the process by
> >> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> >> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> >> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> >> +   will fail with -EBUSY error code if attempted.
> >
> > I assume calling STOP again once the drain sequence completed just returns 0 and
> > doesn't do anything else (since we're already stopped).
> >
> >> +
> >> +   Although mandatory, the availability of encoder commands may be queried
> >> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> >
> > Some corner cases:
> >
> > 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
> >    Now ENC_CMD_STOP is issued. What should happen?
> >
> >    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >    at once as an empty buffer with FLAG_LAST set.
> >
> > 2) Both queues are streaming and buffers have been encoded, but currently no buffers
> >    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> >
> >    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >    at once as an empty buffer with FLAG_LAST set. This is consistent with the
> >    previous corner case.
> >
> > 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
> >    is issued. What should happen?
> >
> >    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
> >    buffer with FLAG_LAST set.
> >
> > 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
> >    Before the drain process completes STREAMOFF is called for either CAPTURE or
> >    OUTPUT queue. What should happen?
> >
> >    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
> >    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
> >    will be used as input to the encoder.
> >
> >    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
> >    FLAG_LAST set.
>
> Note: I'm not quite sure about that last one yet. It's a bit tricky to implement.
>
> I think some of these rules might have to be implemented in v4l2-mem2mem.c to
> simplify drivers.

Most of the drivers would implement drain as a fake buffer inserted
into the driver-internal queue of source buffers. They also normally
have to iterate over the remaining buffers in that queue when
.stop_streaming() is called. If they find the fake buffer there, they
should return an empty buffer. The problematic case is when there is
no buffer in the CAPTURE queue. It would require the driver to
remember that the drain ended and check for that when next CAPTURE
buffer is queued. I guess that would converge with cases 1) and 2),
though.

Best regards,
Tomasz

Hans Verkuil April 8, 2019, 8:43 a.m. UTC | #14

On 4/8/19 10:36 AM, Tomasz Figa wrote:
> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
>>
>> Another comment found while creating compliance tests:
>>
>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>> +Drain
>>> +=====
>>> +
>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>>> +drain sequence described below. After the drain sequence ends, the client has
>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>>> +sequence was started.
>>> +
>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>>> +
>>> +   * **Required fields:**
>>> +
>>> +     ``cmd``
>>> +         set to ``V4L2_ENC_CMD_STOP``
>>> +
>>> +     ``flags``
>>> +         set to 0
>>> +
>>> +     ``pts``
>>> +         set to 0
>>> +
>>> +   .. warning::
>>> +
>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>>> +      queues are streaming. For compatibility reasons, the call to
>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>>> +      not streaming, but at the same time it will not initiate the `Drain`
>>> +      sequence and so the steps described below would not be applicable.
>>> +
>>> +2. Any ``OUTPUT`` buffers queued by the client before the
>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>>> +   normal. The client must continue to handle both queues independently,
>>> +   similarly to normal encode operation. This includes:
>>> +
>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>>> +
>>> +     .. warning::
>>> +
>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>>> +        as it does not contain an encoded frame.
>>> +
>>> +     .. note::
>>> +
>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>>> +        :c:func:`VIDIOC_DQBUF`.
>>> +
>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>>> +
>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>>> +
>>> +   .. note::
>>> +
>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>>> +      event when the last frame has been decoded and all frames are ready to be
>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>>> +
>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>>> +   until the client issues any of the following operations:
>>> +
>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>>> +     operation normally, with all the state from before the drain,
>>
>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
>>
>> In other words: while draining is in progress START will return EBUSY. When draining
>> was finished, then START will resume the encoder. In all other cases it just returns
>> 0 since the encoder is really already started.
>>
> 
> Yes, that was the intention and seems to be the closest to the
> behavior described in the existing documentation.
> 
>>> +
>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>>> +     and then resume encoding,
>>> +
>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>>> +
>>> +.. note::
>>> +
>>> +   Once the drain sequence is initiated, the client needs to drive it to
>>> +   completion, as described by the steps above, unless it aborts the process by
>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>>> +   will fail with -EBUSY error code if attempted.
>>
>> I assume calling STOP again once the drain sequence completed just returns 0 and
>> doesn't do anything else (since we're already stopped).
>>
> 
> Right.
> 
>>> +
>>> +   Although mandatory, the availability of encoder commands may be queried
>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>>
>> Some corner cases:
>>
>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
>>    Now ENC_CMD_STOP is issued. What should happen?
>>
>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>    at once as an empty buffer with FLAG_LAST set.
>>
> 
> SGTM. It's actually similar to a relatively common case where all
> CAPTURE buffers have been dequeued and the application has to refill
> the CAPTURE queue, but in the meantime a drain request needs to be
> issued.
> 
>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
>>
>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
>>    previous corner case.
> 
> Agreed.
> 
>>
>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
>>    is issued. What should happen?
>>
>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
>>    buffer with FLAG_LAST set.
> 
> Generally agreed, but not sure if there is a reason to specifically
> refer to the oldest buffer. (I'm personally for keeping the queues
> ordered, though...)

Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
queued up for userspace to signal that the encoder has stopped." Or something along
those lines.

Regards,

	Hans

> 
>>
>> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
>>    Before the drain process completes STREAMOFF is called for either CAPTURE or
>>    OUTPUT queue. What should happen?
>>
>>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
>>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
>>    will be used as input to the encoder.
>>
> 
> Agreed.
> 
>>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
>>    FLAG_LAST set.
> 
> Agreed.
> 
>>
>> Some of this might have to be documented, but these corner cases should certainly be
>> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
>> with this.
> 
> Agreed with just one minor comment. Thanks for checking with me!
> 
> Best regards,
> Tomasz
>

Tomasz Figa April 8, 2019, 9:23 a.m. UTC | #15

On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
>
> On 4/5/19 10:12 AM, Tomasz Figa wrote:
> > On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
> >>
> >> Hi Tomasz,
> >>
> >> Some more comments...
> >>
> >> On 1/29/19 2:52 PM, Hans Verkuil wrote:
> >>> Hi Tomasz,
> >>>
> >>> Some comments below. Nothing major, so I think a v4 should be ready to be
> >>> merged.
> >>>
> >>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >>>> Due to complexity of the video encoding process, the V4L2 drivers of
> >>>> stateful encoder hardware require specific sequences of V4L2 API calls
> >>>> to be followed. These include capability enumeration, initialization,
> >>>> encoding, encode parameters change, drain and reset.
> >>>>
> >>>> Specifics of the above have been discussed during Media Workshops at
> >>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> >>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> >>>> originated at those events was later implemented by the drivers we already
> >>>> have merged in mainline, such as s5p-mfc or coda.
> >>>>
> >>>> The only thing missing was the real specification included as a part of
> >>>> Linux Media documentation. Fix it now and document the encoder part of
> >>>> the Codec API.
> >>>>
> >>>> Signed-off-by: Tomasz Figa <[e-mail]>
> >>>> ---
> >>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >>>>  5 files changed, 617 insertions(+), 15 deletions(-)
> >>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >>>>
> >>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>>> new file mode 100644
> >>>> index 000000000000..fb8b05a132ee
> >>>> --- /dev/null
> >>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>>> @@ -0,0 +1,586 @@
> >>>> +.. -*- coding: utf-8; mode: rst -*-
> >>>> +
> >>>> +.. _encoder:
> >>>> +
> >>>> +*************************************************
> >>>> +Memory-to-memory Stateful Video Encoder Interface
> >>>> +*************************************************
> >>>> +
> >>>> +A stateful video encoder takes raw video frames in display order and encodes
> >>>> +them into a bitstream. It generates complete chunks of the bitstream, including
> >>>> +all metadata, headers, etc. The resulting bitstream does not require any
> >>>> +further post-processing by the client.
> >>>> +
> >>>> +Performing software stream processing, header generation etc. in the driver
> >>>> +in order to support this interface is strongly discouraged. In case such
> >>>> +operations are needed, use of the Stateless Video Encoder Interface (in
> >>>> +development) is strongly advised.
> >>>> +
> >>>> +Conventions and notation used in this document
> >>>> +==============================================
> >>>> +
> >>>> +1. The general V4L2 API rules apply if not specified in this document
> >>>> +   otherwise.
> >>>> +
> >>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> >>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> >>>> +
> >>>> +3. All steps not marked "optional" are required.
> >>>> +
> >>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> >>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> >>>> +   unless specified otherwise.
> >>>> +
> >>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> >>>> +   used interchangeably with multi-planar API, unless specified otherwise,
> >>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
> >>>> +
> >>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> >>>> +   [0..2]: i = 0, 1, 2.
> >>>> +
> >>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> >>>> +   queue containing data that resulted from processing buffer A.
> >>>> +
> >>>> +Glossary
> >>>> +========
> >>>> +
> >>>> +Refer to :ref:`decoder-glossary`.
> >>>> +
> >>>> +State machine
> >>>> +=============
> >>>> +
> >>>> +.. kernel-render:: DOT
> >>>> +   :alt: DOT digraph of encoder state machine
> >>>> +   :caption: Encoder state machine
> >>>> +
> >>>> +   digraph encoder_state_machine {
> >>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
> >>>> +
> >>>> +       node [shape = circle, label="Initialization"] Initialization;
> >>>> +       node [shape = circle, label="Stopped"] Stopped;
> >>>> +       node [shape = circle, label="Drain"] Drain;
> >>>> +       node [shape = circle, label="Reset"] Reset;
> >>>> +
> >>>> +       node [shape = point]; qi
> >>>> +       qi -> Initialization [ label = "open()" ];
> >>>> +
> >>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> >>>> +
> >>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> >>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> >>>> +       Encoding -> Encoding;
> >>>> +
> >>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> >>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>> +
> >>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> >>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> >>>> +
> >>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> >>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>> +   }
> >>>> +
> >>>> +Querying capabilities
> >>>> +=====================
> >>>> +
> >>>> +1. To enumerate the set of coded formats supported by the encoder, the
> >>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> >>>> +
> >>>> +   * The full set of supported formats will be returned, regardless of the
> >>>> +     format set on ``OUTPUT``.
> >>>> +
> >>>> +2. To enumerate the set of supported raw formats, the client may call
> >>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> >>>> +
> >>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
> >>>> +     will be returned.
> >>>> +
> >>>> +   * In order to enumerate raw formats supported by a given coded format,
> >>>> +     the client must first set that coded format on ``CAPTURE`` and then
> >>>> +     enumerate the formats on ``OUTPUT``.
> >>>> +
> >>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> >>>> +   resolutions for a given format, passing desired pixel format in
> >>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> >>>> +
> >>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> >>>> +     format will include all possible coded resolutions supported by the
> >>>> +     encoder for given coded pixel format.
> >>>> +
> >>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> >>>> +     will include all possible frame buffer resolutions supported by the
> >>>> +     encoder for given raw pixel format and coded format currently set on
> >>>> +     ``CAPTURE``.
> >>>> +
> >>>> +4. Supported profiles and levels for the coded format currently set on
> >>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> >>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
> >>>> +
> >>>> +5. Any additional encoder capabilities may be discovered by querying
> >>>> +   their respective controls.
> >>>> +
> >>>> +Initialization
> >>>> +==============
> >>>> +
> >>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> >>>> +
> >>>> +   * **Required fields:**
> >>>> +
> >>>> +     ``type``
> >>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> >>>> +
> >>>> +     ``pixelformat``
> >>>> +         the coded format to be produced
> >>>> +
> >>>> +     ``sizeimage``
> >>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> >>>> +         match hardware requirements
> >>>> +
> >>>> +     ``width``, ``height``
> >>>> +         ignored (always zero)
> >>>> +
> >>>> +     other fields
> >>>> +         follow standard semantics
> >>>> +
> >>>> +   * **Return fields:**
> >>>> +
> >>>> +     ``sizeimage``
> >>>> +         adjusted size of ``CAPTURE`` buffers
> >>>> +
> >>>> +   .. important::
> >>>> +
> >>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> >>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
> >>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
> >>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> >>>> +      must adjust it afterwards.
> >>>
> >>> Hmm, "including resolution": if width and height are set to 0, what should the
> >>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
> >>> at a first reading of this paragraph it appears to be contradictory.
> >>
> >> I think the driver should just return the width and height of the OUTPUT
> >> format. So the width and height that userspace specifies is just ignored
> >> and replaced by the width and height of the OUTPUT format. After all, that's
> >> what the bitstream will encode. Returning 0 for width and height would make
> >> this a strange exception in V4L2 and I want to avoid that.
> >>
> >
> > Hmm, however, the width and height of the OUTPUT format is not what's
> > actually encoded in the bitstream. The right selection rectangle
> > determines that.
> >
> > In one of the previous versions I though we could put the codec

s/codec/coded/...

> > resolution as the width and height of the CAPTURE format, which would
> > be the resolution of the encoded image rounded up to full macroblocks
> > +/- some encoder-specific constraints. AFAIR there was some concern
> > about OUTPUT format changes triggering CAPTURE format changes, but to
> > be honest, I'm not sure if that's really a problem. I just decided to
> > drop that for the simplicity.
>
> I'm not sure what your point is.
>
> The OUTPUT format has the coded resolution,

That's not always true. The OUTPUT format is just the format of the
source frame buffers. In special cases where the source resolution is
nicely aligned, it would be the same as coded size, but the remaining
cases are valid as well.

> so when you set the
> CAPTURE format it can just copy the OUTPUT coded resolution unless the
> chosen CAPTURE pixelformat can't handle that in which case both the
> OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
> or minimum the codec is capable of.

As per my comment above, generally speaking, the encoder will derive
an appropriate coded format from the OUTPUT format, but also other
factors, like the crop rectangles and possibly some internal
constraints.

>
> That said, I am fine with just leaving it up to the driver as suggested
> before. Just as long as both the CAPTURE and OUTPUT formats remain valid
> (i.e. width and height may never be out of range).
>

Sounds good to me.

> >
> >>>
> >>>> +
> >>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> >>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> >>>> +
> >>>> +   * **Required fields:**
> >>>> +
> >>>> +     ``type``
> >>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>> +
> >>>> +     other fields
> >>>> +         follow standard semantics
> >>>> +
> >>>> +   * **Return fields:**
> >>>> +
> >>>> +     ``pixelformat``
> >>>> +         raw format supported for the coded format currently selected on
> >>>> +         the ``CAPTURE`` queue.
> >>>> +
> >>>> +     other fields
> >>>> +         follow standard semantics
> >>>> +
> >>>> +3. Set the raw source format on the ``OUTPUT`` queue via
> >>>> +   :c:func:`VIDIOC_S_FMT`.
> >>>> +
> >>>> +   * **Required fields:**
> >>>> +
> >>>> +     ``type``
> >>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>> +
> >>>> +     ``pixelformat``
> >>>> +         raw format of the source
> >>>> +
> >>>> +     ``width``, ``height``
> >>>> +         source resolution
> >>>> +
> >>>> +     other fields
> >>>> +         follow standard semantics
> >>>> +
> >>>> +   * **Return fields:**
> >>>> +
> >>>> +     ``width``, ``height``
> >>>> +         may be adjusted by encoder to match alignment requirements, as
> >>>> +         required by the currently selected formats
> >>>
> >>> What if the width x height is larger than the maximum supported by the
> >>> selected coded format? This should probably mention that in that case the
> >>> width x height is reduced to the largest allowed value. Also mention that
> >>> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> >>>
> >>>> +
> >>>> +     other fields
> >>>> +         follow standard semantics
> >>>> +
> >>>> +   * Setting the source resolution will reset the selection rectangles to their
> >>>> +     default values, based on the new resolution, as described in the step 5
> >>>
> >>> 5 -> 4
> >>>
> >>> Or just say: "as described in the next step."
> >>>
> >>>> +     below.
> >>
> >> It should also be made explicit that:
> >>
> >> 1) the crop rectangle will be set to the given width and height *before*
> >> it is being adjusted by S_FMT.
> >>
> >
> > I don't think that's what we want here.
> >
> > Defining the default rectangle to be exactly the same as the OUTPUT
> > resolution (after the adjustment) makes the semantics consistent - not
> > setting the crop rectangle gives you exactly the behavior as if there
> > was no cropping involved (or supported by the encoder).
>
> I think you are right. This seems to be what the coda driver does as well.
> It is convenient to be able to just set a 1920x1080 format and have that
> resolution be stored as the crop rectangle, since it avoids having to call
> s_selection afterwards, but it is not really consistent with the way V4L2
> works.
>
> >
> >> Open question: should we support a compose rectangle for the CAPTURE that
> >> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
> >> the adjusted width and height and the compose rectangle (read-only) contains
> >> the visible width and height. It's not strictly necessary, but it is
> >> symmetrical.
> >
> > Wouldn't it rather be the CAPTURE crop rectangle that would be of the
> > same resolution of the OUTPUT compose rectangle? Then you could
> > actually have the CAPTURE compose rectangle for putting that into the
> > desired rectangle of the encoded stream, if the encoder supports that.
> > (I don't know any that does, so probably out of concern for now.)
>
> Yes, you are right.
>
> But should we support this?
>
> I actually think not for this initial version. It can be added later, I guess.
>

I think it boils down on whether adding it later wouldn't
significantly complicate the application logic. It also relates to my
other comment somewhere below.

> >
> >>
> >> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
> >> height. The CAPTURE sizeimage might change as well.
> >>
> >>>> +
> >>>> +4. **Optional.** Set the visible resolution for the stream metadata via
> >>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> >>
> >> I think you should mention that this is only necessary if the crop rectangle
> >> that is set when you set the format isn't what you want.
> >>
> >
> > Ack.
> >
> >>>> +
> >>>> +   * **Required fields:**
> >>>> +
> >>>> +     ``type``
> >>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>> +
> >>>> +     ``target``
> >>>> +         set to ``V4L2_SEL_TGT_CROP``
> >>>> +
> >>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> >>>> +         rectangle and may be subject to adjustment to match codec and
> >>>> +         hardware constraints
> >>>> +
> >>>> +   * **Return fields:**
> >>>> +
> >>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>>> +         visible rectangle adjusted by the encoder
> >>>> +
> >>>> +   * The following selection targets are supported on ``OUTPUT``:
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>>> +         equal to the full source frame, matching the active ``OUTPUT``
> >>>> +         format
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_CROP``
> >>>> +         rectangle within the source buffer to be encoded into the
> >>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>>> +
> >>>> +         .. note::
> >>>> +
> >>>> +            A common use case for this selection target is encoding a source
> >>>> +            video with a resolution that is not a multiple of a macroblock,
> >>>> +            e.g.  the common 1920x1080 resolution may require the source
> >>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> >>>> +            size. To avoid encoding the padding, the client needs to explicitly
> >>>> +            configure this selection target to 1920x1080.
> >>
> >> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
> >>
> >
> > Sorry, which part exactly and what part of the proposal exactly? :)
> > (My comment above might be related, though.)
>
> Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
> crop selection target, so this note remains valid.
>

Ack.

> >
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> >>>> +         maximum rectangle within the coded resolution, which the cropped
> >>>> +         source frame can be composed into; if the hardware does not support
> >>>> +         composition or scaling, then this is always equal to the rectangle of
> >>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> >>>> +         equal to a rectangle of width and height matching
> >>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>>> +
> >>>> +     ``V4L2_SEL_TGT_COMPOSE``
> >>>> +         rectangle within the coded frame, which the cropped source frame
> >>>> +         is to be composed into; defaults to
> >>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> >>>> +         additional compose/scaling capabilities; resulting stream will
> >>>> +         have this rectangle encoded as the visible rectangle in its
> >>>> +         metadata
> >>
> >> I think the compose targets for OUTPUT are only needed if the hardware can
> >> actually do scaling and/or composition. Otherwise they can (must?) be
> >> dropped.
> >>
> >
> > Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
> > userspace to learn the target visible rectangle that's going to be
> > encoded in the stream metadata. If we omit it, we wouldn't have a way
> > that would be consistent between encoders that can do
> > scaling/composition and those that can't.
>
> I'm not convinced about this. The standard API behavior is not to expose
> functionality that the hardware can't do. So if scaling isn't possible on
> the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
>
> I also believe it very unlikely that we'll see encoders capable of scaling
> as it doesn't make much sense.

It does make a lot of sense - WebRTC requires 3 different sizes of the
stream to be encoded at the same time. However, unfortunately, I
haven't yet seen an encoder capable of doing so.

> I would prefer to drop this to simplify the
> spec, and when we get encoders that can scale, then we can add support for
> compose rectangles (and I'm sure we'll need to think about how that
> influences the CAPTURE side as well).
>
> For encoders without scaling it is the OUTPUT crop rectangle that defines
> the visible rectangle.
>
> >
> > However, with your proposal of actually having selection rectangles
> > for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
> > would expose a varying set of rectangles, depending on the hardware
> > capability, while the CAPTURE queue would always expose its rectangle
> > with that information.
>
> I think we should keep it simple and only define selection rectangles
> when really needed.
>
> So encoders support CROP on the OUTPUT, and decoders support CAPTURE
> COMPOSE (may be read-only). Nothing else.
>
> Once support for scaling is needed (either on the encoder or decoder
> side), then the spec should be enhanced. But I prefer to postpone that
> until we actually have hardware that needs this.
>

Okay, let's do it this way then. Actually, I don't even think there is
much value in exposing information internal to the bitstream metadata
like this, similarly to the coded size. My intention was to just
ensure that we can easily add scaling/composing functionality later.

I just removed the COMPOSE rectangles from my next draft.

[snip]
> >
> >>
> >> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
> >> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
> >> particular) and CAPTURE buffers were already allocated and are too small.
> >
> > The OUTPUT format must not change the CAPTURE format by definition.
> > Otherwise we end up in a situation where we can't commit, because both
> > queue formats can affect each other. Any change to the OUTPUT format
> > that wouldn't work with the current CAPTURE format should be adjusted
> > by the driver to match the current CAPTURE format.
>
> But the CAPTURE format *does* depend on the OUTPUT format: if the output
> resolution changes, then so does the CAPTURE resolution and esp. the
> sizeimage value, since that is typically resolution dependent.
>
> The coda driver does this as well: changing the output resolution
> will update the capture resolution and sizeimage. The vicodec driver does the
> same.
>
> Setting the CAPTURE format basically just selects the codec to use, after
> that you can set the OUTPUT format and read the updated CAPTURE format to
> get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
> change the OUTPUT format, unless the OUTPUT format is incompatible with the
> newly selected codec.

Let me think about it for a while.

Best regards,
Tomasz

Tomasz Figa April 8, 2019, 9:35 a.m. UTC | #16

On Mon, Apr 8, 2019 at 4:43 PM Hans Verkuil <[e-mail]> wrote:
>
> On 4/8/19 8:59 AM, Tomasz Figa wrote:
> > On Thu, Mar 21, 2019 at 7:11 PM Hans Verkuil <[e-mail]> wrote:
> >>
> >> Hi Tomasz,
> >>
> >> A few more comments:
> >>
> >> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >>> Due to complexity of the video encoding process, the V4L2 drivers of
> >>> stateful encoder hardware require specific sequences of V4L2 API calls
> >>> to be followed. These include capability enumeration, initialization,
> >>> encoding, encode parameters change, drain and reset.
> >>>
> >>> Specifics of the above have been discussed during Media Workshops at
> >>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> >>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> >>> originated at those events was later implemented by the drivers we already
> >>> have merged in mainline, such as s5p-mfc or coda.
> >>>
> >>> The only thing missing was the real specification included as a part of
> >>> Linux Media documentation. Fix it now and document the encoder part of
> >>> the Codec API.
> >>>
> >>> Signed-off-by: Tomasz Figa <[e-mail]>
> >>> ---
> >>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >>>  5 files changed, 617 insertions(+), 15 deletions(-)
> >>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >>>
> >>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>> new file mode 100644
> >>> index 000000000000..fb8b05a132ee
> >>> --- /dev/null
> >>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>> @@ -0,0 +1,586 @@
> >>> +.. -*- coding: utf-8; mode: rst -*-
> >>> +
> >>> +.. _encoder:
> >>> +
> >>> +*************************************************
> >>> +Memory-to-memory Stateful Video Encoder Interface
> >>> +*************************************************
> >>> +
> >>> +A stateful video encoder takes raw video frames in display order and encodes
> >>> +them into a bitstream. It generates complete chunks of the bitstream, including
> >>> +all metadata, headers, etc. The resulting bitstream does not require any
> >>> +further post-processing by the client.
> >>> +
> >>> +Performing software stream processing, header generation etc. in the driver
> >>> +in order to support this interface is strongly discouraged. In case such
> >>> +operations are needed, use of the Stateless Video Encoder Interface (in
> >>> +development) is strongly advised.
> >>> +
> >>> +Conventions and notation used in this document
> >>> +==============================================
> >>> +
> >>> +1. The general V4L2 API rules apply if not specified in this document
> >>> +   otherwise.
> >>> +
> >>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> >>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> >>> +
> >>> +3. All steps not marked "optional" are required.
> >>> +
> >>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> >>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> >>> +   unless specified otherwise.
> >>> +
> >>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> >>> +   used interchangeably with multi-planar API, unless specified otherwise,
> >>> +   depending on decoder capabilities and following the general V4L2 guidelines.
> >>
> >> decoder -> encoder
> >>
> >
> > Ack.
> >
> >>> +
> >>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> >>> +   [0..2]: i = 0, 1, 2.
> >>> +
> >>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> >>> +   queue containing data that resulted from processing buffer A.
> >>> +
> >>> +Glossary
> >>> +========
> >>> +
> >>> +Refer to :ref:`decoder-glossary`.
> >>> +
> >>> +State machine
> >>> +=============
> >>> +
> >>> +.. kernel-render:: DOT
> >>> +   :alt: DOT digraph of encoder state machine
> >>> +   :caption: Encoder state machine
> >>> +
> >>> +   digraph encoder_state_machine {
> >>> +       node [shape = doublecircle, label="Encoding"] Encoding;
> >>> +
> >>> +       node [shape = circle, label="Initialization"] Initialization;
> >>> +       node [shape = circle, label="Stopped"] Stopped;
> >>> +       node [shape = circle, label="Drain"] Drain;
> >>> +       node [shape = circle, label="Reset"] Reset;
> >>> +
> >>> +       node [shape = point]; qi
> >>> +       qi -> Initialization [ label = "open()" ];
> >>> +
> >>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> >>> +
> >>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> >>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> >>> +       Encoding -> Encoding;
> >>> +
> >>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> >>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>> +
> >>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> >>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> >>> +
> >>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> >>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>> +   }
> >>> +
> >>> +Querying capabilities
> >>> +=====================
> >>> +
> >>> +1. To enumerate the set of coded formats supported by the encoder, the
> >>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> >>> +
> >>> +   * The full set of supported formats will be returned, regardless of the
> >>> +     format set on ``OUTPUT``.
> >>> +
> >>> +2. To enumerate the set of supported raw formats, the client may call
> >>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> >>> +
> >>> +   * Only the formats supported for the format currently active on ``CAPTURE``
> >>> +     will be returned.
> >>> +
> >>> +   * In order to enumerate raw formats supported by a given coded format,
> >>> +     the client must first set that coded format on ``CAPTURE`` and then
> >>> +     enumerate the formats on ``OUTPUT``.
> >>> +
> >>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> >>> +   resolutions for a given format, passing desired pixel format in
> >>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> >>> +
> >>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> >>> +     format will include all possible coded resolutions supported by the
> >>> +     encoder for given coded pixel format.
> >>> +
> >>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> >>> +     will include all possible frame buffer resolutions supported by the
> >>> +     encoder for given raw pixel format and coded format currently set on
> >>> +     ``CAPTURE``.
> >>> +
> >>> +4. Supported profiles and levels for the coded format currently set on
> >>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> >>> +   via :c:func:`VIDIOC_QUERYCTRL`.
> >>> +
> >>> +5. Any additional encoder capabilities may be discovered by querying
> >>> +   their respective controls.
> >>> +
> >>> +Initialization
> >>> +==============
> >>> +
> >>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> >>> +
> >>> +     ``pixelformat``
> >>> +         the coded format to be produced
> >>> +
> >>> +     ``sizeimage``
> >>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> >>> +         match hardware requirements
> >>> +
> >>> +     ``width``, ``height``
> >>> +         ignored (always zero)
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``sizeimage``
> >>> +         adjusted size of ``CAPTURE`` buffers
> >>> +
> >>> +   .. important::
> >>> +
> >>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> >>> +      format. The encoder will derive a new ``OUTPUT`` format from the
> >>> +      ``CAPTURE`` format being set, including resolution, colorimetry
> >>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> >>> +      must adjust it afterwards.
> >>> +
> >>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> >>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``pixelformat``
> >>> +         raw format supported for the coded format currently selected on
> >>> +         the ``CAPTURE`` queue.
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +3. Set the raw source format on the ``OUTPUT`` queue via
> >>> +   :c:func:`VIDIOC_S_FMT`.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>> +
> >>> +     ``pixelformat``
> >>> +         raw format of the source
> >>> +
> >>> +     ``width``, ``height``
> >>> +         source resolution
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``width``, ``height``
> >>> +         may be adjusted by encoder to match alignment requirements, as
> >>> +         required by the currently selected formats
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * Setting the source resolution will reset the selection rectangles to their
> >>> +     default values, based on the new resolution, as described in the step 5
> >>> +     below.
> >>> +
> >>> +4. **Optional.** Set the visible resolution for the stream metadata via
> >>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>> +
> >>> +     ``target``
> >>> +         set to ``V4L2_SEL_TGT_CROP``
> >>> +
> >>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> >>> +         rectangle and may be subject to adjustment to match codec and
> >>> +         hardware constraints
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>> +         visible rectangle adjusted by the encoder
> >>> +
> >>> +   * The following selection targets are supported on ``OUTPUT``:
> >>> +
> >>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>> +         equal to the full source frame, matching the active ``OUTPUT``
> >>> +         format
> >>> +
> >>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>> +
> >>> +     ``V4L2_SEL_TGT_CROP``
> >>> +         rectangle within the source buffer to be encoded into the
> >>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>> +
> >>> +         .. note::
> >>> +
> >>> +            A common use case for this selection target is encoding a source
> >>> +            video with a resolution that is not a multiple of a macroblock,
> >>> +            e.g.  the common 1920x1080 resolution may require the source
> >>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> >>> +            size. To avoid encoding the padding, the client needs to explicitly
> >>> +            configure this selection target to 1920x1080.
> >>> +
> >>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> >>> +         maximum rectangle within the coded resolution, which the cropped
> >>> +         source frame can be composed into; if the hardware does not support
> >>> +         composition or scaling, then this is always equal to the rectangle of
> >>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>> +
> >>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> >>> +         equal to a rectangle of width and height matching
> >>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>> +
> >>> +     ``V4L2_SEL_TGT_COMPOSE``
> >>> +         rectangle within the coded frame, which the cropped source frame
> >>> +         is to be composed into; defaults to
> >>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> >>> +         additional compose/scaling capabilities; resulting stream will
> >>> +         have this rectangle encoded as the visible rectangle in its
> >>> +         metadata
> >>
> >> I would only support the COMPOSE targets if the hardware can actually do
> >> scaling and/or composing. That is conform standard V4L2 behavior where
> >> cropping/composing is only implemented if the hardware can actually do
> >> this.
> >>
> >
> > Please see my other reply to your earlier similar comment in this thread.
> >
> >>> +
> >>> +   .. warning::
> >>> +
> >>> +      The encoder may adjust the crop/compose rectangles to the nearest
> >>> +      supported ones to meet codec and hardware requirements. The client needs
> >>> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> >>> +
> >>> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> >>> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``count``
> >>> +         requested number of buffers to allocate; greater than zero
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> >>> +         ``CAPTURE``
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``count``
> >>> +          actual number of buffers allocated
> >>> +
> >>> +   .. warning::
> >>> +
> >>> +      The actual number of allocated buffers may differ from the ``count``
> >>> +      given. The client must check the updated value of ``count`` after the
> >>> +      call returns.
> >>> +
> >>> +   .. note::
> >>> +
> >>> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> >>> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> >>> +      control to get the minimum number of buffers required, and pass the
> >>> +      obtained value plus the number of additional buffers needed in the
> >>> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> >>> +
> >>> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> >>> +   control over buffer allocation.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``count``
> >>> +         requested number of buffers to allocate; greater than zero
> >>> +
> >>> +     ``type``
> >>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>> +
> >>> +     other fields
> >>> +         follow standard semantics
> >>> +
> >>> +   * **Return fields:**
> >>> +
> >>> +     ``count``
> >>> +         adjusted to the number of allocated buffers
> >>> +
> >>> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> >>> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> >>> +   encoding process starts when both queues start streaming.
> >>> +
> >>> +.. note::
> >>> +
> >>> +   If the client stops the ``CAPTURE`` queue during the encode process and then
> >>> +   restarts it again, the encoder will begin generating a stream independent
> >>> +   from the stream generated before the stop. The exact constraints depend
> >>> +   on the coded format, but may include the following implications:
> >>> +
> >>> +   * encoded frames produced after the restart must not reference any
> >>> +     frames produced before the stop, e.g. no long term references for
> >>> +     H.264,
> >>> +
> >>> +   * any headers that must be included in a standalone stream must be
> >>> +     produced again, e.g. SPS and PPS for H.264.
> >>> +
> >>> +Encoding
> >>> +========
> >>> +
> >>> +This state is reached after the `Initialization` sequence finishes
> >>> +successfully.  In this state, the client queues and dequeues buffers to both
> >>> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> >>> +standard semantics.
> >>> +
> >>> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> >>> +format and may be affected by codec-specific extended controls, as stated
> >>> +in the documentation of each format.
> >>> +
> >>> +Both queues operate independently, following standard behavior of V4L2 buffer
> >>> +queues and memory-to-memory devices. In addition, the order of encoded frames
> >>> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> >>> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> >>> +e.g. frame reordering.
> >>> +
> >>> +The client must not assume any direct relationship between ``CAPTURE`` and
> >>> +``OUTPUT`` buffers and any specific timing of buffers becoming
> >>> +available to dequeue. Specifically:
> >>> +
> >>> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> >>> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> >>> +  frame that preceded it in display, but succeeded it in the decode order),
> >>> +
> >>> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> >>> +  ``CAPTURE`` later into encode process, and/or after processing further
> >>> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> >>> +  reordering is used,
> >>> +
> >>> +* buffers may become available on the ``CAPTURE`` queue without additional
> >>> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> >>> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> >>> +  available at later time, due to specifics of the decoding process,
> >>> +
> >>> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> >>> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> >>> +  encoder needs to use the frame as a reference for encoding further frames.
> >>> +
> >>> +.. note::
> >>> +
> >>> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> >>> +   originated from, the client can set the ``timestamp`` field of the
> >>> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> >>> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> >>> +   will have their ``timestamp`` field set to the same value when dequeued.
> >>> +
> >>> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> >>> +   one ``CAPTURE`` buffer, the following cases are defined:
> >>> +
> >>> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> >>> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> >>> +
> >>> +   * the encoding order differs from the presentation order (i.e. the
> >>> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> >>> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> >>> +     and thus monotonicity of the timestamps cannot be guaranteed.
> >>> +
> >>> +.. note::
> >>> +
> >>> +   To let the client distinguish between frame types (keyframes, intermediate
> >>> +   frames; the exact list of types depends on the coded format), the
> >>> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> >>> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> >>> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> >>> +   and their meanings.
> >>
> >> I don't think we can require this since a capture buffer may contain multiple
> >> encoded frames.
> >>
> >
> > I thought we required that only one encoded frame was in one CAPTURE
> > buffer. Real time use cases rely heavily on this frame type
> > information, so I can't imagine not requiring this.
>
> That the CAPTURE buffer contains only one encoded frame is never stated
> explicitly. I am not so sure I want that to be a hard requirement anyway
> since the old ivtv MPEG encoder just produces a bitstream.
>
> Perhaps this should be signaled with a flag in ENUM_FMT?
>
> >
> >> It would actually make more sense to return it in the output buffer, but I don't
> >> know if a hardware encoder can actually provide that information.
> >>
> >
> > I believe all the already existing drivers provide the information
> > about the encoded frame type, but I don't think they provide the
> > information about what source frame it came from.
> >
> >> Another use of these flags for an output buffer is to force a keyframe if for
> >> example a scene change was detected.
> >>
> >> My feeling is that we should drop this note. Forcing a keyframe by setting that
> >> flag for the output buffer might actually be a useful thing to do for a stateful
> >> encoder.
> >>
> >
> > However, to force keyframe, one sets it in the OUTPUT buffer. Then, to
> > actually get the right CAPTURE buffer, one has to look for one with
> > this flag set.
>
> So *if* the driver stores only one encoded frame in a CAPTURE buffer, then we
> can require that these flags have to be set for that CAPTURE buffer. Otherwise
> they should be cleared since they cannot be associated with a specific buffer.

But then we don't know to which source frame it applies, while it's
usually quite important to force the key frame at the right frame,
e.g. scene change.

>
> And I think it should be documented that you can set the KEYFRAME flag in the
> OUTPUT buffer to force a keyframe (the driver may ignore this if it can't do
> this for some reason).

Indeed. Let me make sure it's included in the document.

Best regards,
Tomasz

Hans Verkuil April 8, 2019, 11:11 a.m. UTC | #17

On 4/8/19 11:23 AM, Tomasz Figa wrote:
> On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
>>
>> On 4/5/19 10:12 AM, Tomasz Figa wrote:
>>> On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
>>>>
>>>> Hi Tomasz,
>>>>
>>>> Some more comments...
>>>>
>>>> On 1/29/19 2:52 PM, Hans Verkuil wrote:
>>>>> Hi Tomasz,
>>>>>
>>>>> Some comments below. Nothing major, so I think a v4 should be ready to be
>>>>> merged.
>>>>>
>>>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>>>>> Due to complexity of the video encoding process, the V4L2 drivers of
>>>>>> stateful encoder hardware require specific sequences of V4L2 API calls
>>>>>> to be followed. These include capability enumeration, initialization,
>>>>>> encoding, encode parameters change, drain and reset.
>>>>>>
>>>>>> Specifics of the above have been discussed during Media Workshops at
>>>>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>>>>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>>>>>> originated at those events was later implemented by the drivers we already
>>>>>> have merged in mainline, such as s5p-mfc or coda.
>>>>>>
>>>>>> The only thing missing was the real specification included as a part of
>>>>>> Linux Media documentation. Fix it now and document the encoder part of
>>>>>> the Codec API.
>>>>>>
>>>>>> Signed-off-by: Tomasz Figa <[e-mail]>
>>>>>> ---
>>>>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>>>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>>>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>>>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>>>>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>>>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>>
>>>>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>> new file mode 100644
>>>>>> index 000000000000..fb8b05a132ee
>>>>>> --- /dev/null
>>>>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>> @@ -0,0 +1,586 @@
>>>>>> +.. -*- coding: utf-8; mode: rst -*-
>>>>>> +
>>>>>> +.. _encoder:
>>>>>> +
>>>>>> +*************************************************
>>>>>> +Memory-to-memory Stateful Video Encoder Interface
>>>>>> +*************************************************
>>>>>> +
>>>>>> +A stateful video encoder takes raw video frames in display order and encodes
>>>>>> +them into a bitstream. It generates complete chunks of the bitstream, including
>>>>>> +all metadata, headers, etc. The resulting bitstream does not require any
>>>>>> +further post-processing by the client.
>>>>>> +
>>>>>> +Performing software stream processing, header generation etc. in the driver
>>>>>> +in order to support this interface is strongly discouraged. In case such
>>>>>> +operations are needed, use of the Stateless Video Encoder Interface (in
>>>>>> +development) is strongly advised.
>>>>>> +
>>>>>> +Conventions and notation used in this document
>>>>>> +==============================================
>>>>>> +
>>>>>> +1. The general V4L2 API rules apply if not specified in this document
>>>>>> +   otherwise.
>>>>>> +
>>>>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
>>>>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
>>>>>> +
>>>>>> +3. All steps not marked "optional" are required.
>>>>>> +
>>>>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
>>>>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
>>>>>> +   unless specified otherwise.
>>>>>> +
>>>>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
>>>>>> +   used interchangeably with multi-planar API, unless specified otherwise,
>>>>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
>>>>>> +
>>>>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
>>>>>> +   [0..2]: i = 0, 1, 2.
>>>>>> +
>>>>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
>>>>>> +   queue containing data that resulted from processing buffer A.
>>>>>> +
>>>>>> +Glossary
>>>>>> +========
>>>>>> +
>>>>>> +Refer to :ref:`decoder-glossary`.
>>>>>> +
>>>>>> +State machine
>>>>>> +=============
>>>>>> +
>>>>>> +.. kernel-render:: DOT
>>>>>> +   :alt: DOT digraph of encoder state machine
>>>>>> +   :caption: Encoder state machine
>>>>>> +
>>>>>> +   digraph encoder_state_machine {
>>>>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
>>>>>> +
>>>>>> +       node [shape = circle, label="Initialization"] Initialization;
>>>>>> +       node [shape = circle, label="Stopped"] Stopped;
>>>>>> +       node [shape = circle, label="Drain"] Drain;
>>>>>> +       node [shape = circle, label="Reset"] Reset;
>>>>>> +
>>>>>> +       node [shape = point]; qi
>>>>>> +       qi -> Initialization [ label = "open()" ];
>>>>>> +
>>>>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
>>>>>> +
>>>>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
>>>>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
>>>>>> +       Encoding -> Encoding;
>>>>>> +
>>>>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>> +
>>>>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
>>>>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
>>>>>> +
>>>>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
>>>>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>> +   }
>>>>>> +
>>>>>> +Querying capabilities
>>>>>> +=====================
>>>>>> +
>>>>>> +1. To enumerate the set of coded formats supported by the encoder, the
>>>>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
>>>>>> +
>>>>>> +   * The full set of supported formats will be returned, regardless of the
>>>>>> +     format set on ``OUTPUT``.
>>>>>> +
>>>>>> +2. To enumerate the set of supported raw formats, the client may call
>>>>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
>>>>>> +
>>>>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
>>>>>> +     will be returned.
>>>>>> +
>>>>>> +   * In order to enumerate raw formats supported by a given coded format,
>>>>>> +     the client must first set that coded format on ``CAPTURE`` and then
>>>>>> +     enumerate the formats on ``OUTPUT``.
>>>>>> +
>>>>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
>>>>>> +   resolutions for a given format, passing desired pixel format in
>>>>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
>>>>>> +
>>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
>>>>>> +     format will include all possible coded resolutions supported by the
>>>>>> +     encoder for given coded pixel format.
>>>>>> +
>>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
>>>>>> +     will include all possible frame buffer resolutions supported by the
>>>>>> +     encoder for given raw pixel format and coded format currently set on
>>>>>> +     ``CAPTURE``.
>>>>>> +
>>>>>> +4. Supported profiles and levels for the coded format currently set on
>>>>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
>>>>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
>>>>>> +
>>>>>> +5. Any additional encoder capabilities may be discovered by querying
>>>>>> +   their respective controls.
>>>>>> +
>>>>>> +Initialization
>>>>>> +==============
>>>>>> +
>>>>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>>>>>> +
>>>>>> +   * **Required fields:**
>>>>>> +
>>>>>> +     ``type``
>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>>>>>> +
>>>>>> +     ``pixelformat``
>>>>>> +         the coded format to be produced
>>>>>> +
>>>>>> +     ``sizeimage``
>>>>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>>>>>> +         match hardware requirements
>>>>>> +
>>>>>> +     ``width``, ``height``
>>>>>> +         ignored (always zero)
>>>>>> +
>>>>>> +     other fields
>>>>>> +         follow standard semantics
>>>>>> +
>>>>>> +   * **Return fields:**
>>>>>> +
>>>>>> +     ``sizeimage``
>>>>>> +         adjusted size of ``CAPTURE`` buffers
>>>>>> +
>>>>>> +   .. important::
>>>>>> +
>>>>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>>>>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>>>>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>>>>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>>>>>> +      must adjust it afterwards.
>>>>>
>>>>> Hmm, "including resolution": if width and height are set to 0, what should the
>>>>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
>>>>> at a first reading of this paragraph it appears to be contradictory.
>>>>
>>>> I think the driver should just return the width and height of the OUTPUT
>>>> format. So the width and height that userspace specifies is just ignored
>>>> and replaced by the width and height of the OUTPUT format. After all, that's
>>>> what the bitstream will encode. Returning 0 for width and height would make
>>>> this a strange exception in V4L2 and I want to avoid that.
>>>>
>>>
>>> Hmm, however, the width and height of the OUTPUT format is not what's
>>> actually encoded in the bitstream. The right selection rectangle
>>> determines that.
>>>
>>> In one of the previous versions I though we could put the codec
> 
> s/codec/coded/...
> 
>>> resolution as the width and height of the CAPTURE format, which would
>>> be the resolution of the encoded image rounded up to full macroblocks
>>> +/- some encoder-specific constraints. AFAIR there was some concern
>>> about OUTPUT format changes triggering CAPTURE format changes, but to
>>> be honest, I'm not sure if that's really a problem. I just decided to
>>> drop that for the simplicity.
>>
>> I'm not sure what your point is.
>>
>> The OUTPUT format has the coded resolution,
> 
> That's not always true. The OUTPUT format is just the format of the
> source frame buffers. In special cases where the source resolution is
> nicely aligned, it would be the same as coded size, but the remaining
> cases are valid as well.
> 
>> so when you set the
>> CAPTURE format it can just copy the OUTPUT coded resolution unless the
>> chosen CAPTURE pixelformat can't handle that in which case both the
>> OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
>> or minimum the codec is capable of.
> 
> As per my comment above, generally speaking, the encoder will derive
> an appropriate coded format from the OUTPUT format, but also other
> factors, like the crop rectangles and possibly some internal
> constraints.
> 
>>
>> That said, I am fine with just leaving it up to the driver as suggested
>> before. Just as long as both the CAPTURE and OUTPUT formats remain valid
>> (i.e. width and height may never be out of range).
>>
> 
> Sounds good to me.
> 
>>>
>>>>>
>>>>>> +
>>>>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
>>>>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
>>>>>> +
>>>>>> +   * **Required fields:**
>>>>>> +
>>>>>> +     ``type``
>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>> +
>>>>>> +     other fields
>>>>>> +         follow standard semantics
>>>>>> +
>>>>>> +   * **Return fields:**
>>>>>> +
>>>>>> +     ``pixelformat``
>>>>>> +         raw format supported for the coded format currently selected on
>>>>>> +         the ``CAPTURE`` queue.
>>>>>> +
>>>>>> +     other fields
>>>>>> +         follow standard semantics
>>>>>> +
>>>>>> +3. Set the raw source format on the ``OUTPUT`` queue via
>>>>>> +   :c:func:`VIDIOC_S_FMT`.
>>>>>> +
>>>>>> +   * **Required fields:**
>>>>>> +
>>>>>> +     ``type``
>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>> +
>>>>>> +     ``pixelformat``
>>>>>> +         raw format of the source
>>>>>> +
>>>>>> +     ``width``, ``height``
>>>>>> +         source resolution
>>>>>> +
>>>>>> +     other fields
>>>>>> +         follow standard semantics
>>>>>> +
>>>>>> +   * **Return fields:**
>>>>>> +
>>>>>> +     ``width``, ``height``
>>>>>> +         may be adjusted by encoder to match alignment requirements, as
>>>>>> +         required by the currently selected formats
>>>>>
>>>>> What if the width x height is larger than the maximum supported by the
>>>>> selected coded format? This should probably mention that in that case the
>>>>> width x height is reduced to the largest allowed value. Also mention that
>>>>> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
>>>>>
>>>>>> +
>>>>>> +     other fields
>>>>>> +         follow standard semantics
>>>>>> +
>>>>>> +   * Setting the source resolution will reset the selection rectangles to their
>>>>>> +     default values, based on the new resolution, as described in the step 5
>>>>>
>>>>> 5 -> 4
>>>>>
>>>>> Or just say: "as described in the next step."
>>>>>
>>>>>> +     below.
>>>>
>>>> It should also be made explicit that:
>>>>
>>>> 1) the crop rectangle will be set to the given width and height *before*
>>>> it is being adjusted by S_FMT.
>>>>
>>>
>>> I don't think that's what we want here.
>>>
>>> Defining the default rectangle to be exactly the same as the OUTPUT
>>> resolution (after the adjustment) makes the semantics consistent - not
>>> setting the crop rectangle gives you exactly the behavior as if there
>>> was no cropping involved (or supported by the encoder).
>>
>> I think you are right. This seems to be what the coda driver does as well.
>> It is convenient to be able to just set a 1920x1080 format and have that
>> resolution be stored as the crop rectangle, since it avoids having to call
>> s_selection afterwards, but it is not really consistent with the way V4L2
>> works.
>>
>>>
>>>> Open question: should we support a compose rectangle for the CAPTURE that
>>>> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
>>>> the adjusted width and height and the compose rectangle (read-only) contains
>>>> the visible width and height. It's not strictly necessary, but it is
>>>> symmetrical.
>>>
>>> Wouldn't it rather be the CAPTURE crop rectangle that would be of the
>>> same resolution of the OUTPUT compose rectangle? Then you could
>>> actually have the CAPTURE compose rectangle for putting that into the
>>> desired rectangle of the encoded stream, if the encoder supports that.
>>> (I don't know any that does, so probably out of concern for now.)
>>
>> Yes, you are right.
>>
>> But should we support this?
>>
>> I actually think not for this initial version. It can be added later, I guess.
>>
> 
> I think it boils down on whether adding it later wouldn't
> significantly complicate the application logic. It also relates to my
> other comment somewhere below.
> 
>>>
>>>>
>>>> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
>>>> height. The CAPTURE sizeimage might change as well.
>>>>
>>>>>> +
>>>>>> +4. **Optional.** Set the visible resolution for the stream metadata via
>>>>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
>>>>
>>>> I think you should mention that this is only necessary if the crop rectangle
>>>> that is set when you set the format isn't what you want.
>>>>
>>>
>>> Ack.
>>>
>>>>>> +
>>>>>> +   * **Required fields:**
>>>>>> +
>>>>>> +     ``type``
>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>> +
>>>>>> +     ``target``
>>>>>> +         set to ``V4L2_SEL_TGT_CROP``
>>>>>> +
>>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
>>>>>> +         rectangle and may be subject to adjustment to match codec and
>>>>>> +         hardware constraints
>>>>>> +
>>>>>> +   * **Return fields:**
>>>>>> +
>>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>>>> +         visible rectangle adjusted by the encoder
>>>>>> +
>>>>>> +   * The following selection targets are supported on ``OUTPUT``:
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>>>> +         equal to the full source frame, matching the active ``OUTPUT``
>>>>>> +         format
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_CROP``
>>>>>> +         rectangle within the source buffer to be encoded into the
>>>>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>>>> +
>>>>>> +         .. note::
>>>>>> +
>>>>>> +            A common use case for this selection target is encoding a source
>>>>>> +            video with a resolution that is not a multiple of a macroblock,
>>>>>> +            e.g.  the common 1920x1080 resolution may require the source
>>>>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
>>>>>> +            size. To avoid encoding the padding, the client needs to explicitly
>>>>>> +            configure this selection target to 1920x1080.
>>>>
>>>> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
>>>>
>>>
>>> Sorry, which part exactly and what part of the proposal exactly? :)
>>> (My comment above might be related, though.)
>>
>> Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
>> crop selection target, so this note remains valid.
>>
> 
> Ack.
> 
>>>
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
>>>>>> +         maximum rectangle within the coded resolution, which the cropped
>>>>>> +         source frame can be composed into; if the hardware does not support
>>>>>> +         composition or scaling, then this is always equal to the rectangle of
>>>>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
>>>>>> +         equal to a rectangle of width and height matching
>>>>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>>>> +
>>>>>> +     ``V4L2_SEL_TGT_COMPOSE``
>>>>>> +         rectangle within the coded frame, which the cropped source frame
>>>>>> +         is to be composed into; defaults to
>>>>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
>>>>>> +         additional compose/scaling capabilities; resulting stream will
>>>>>> +         have this rectangle encoded as the visible rectangle in its
>>>>>> +         metadata
>>>>
>>>> I think the compose targets for OUTPUT are only needed if the hardware can
>>>> actually do scaling and/or composition. Otherwise they can (must?) be
>>>> dropped.
>>>>
>>>
>>> Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
>>> userspace to learn the target visible rectangle that's going to be
>>> encoded in the stream metadata. If we omit it, we wouldn't have a way
>>> that would be consistent between encoders that can do
>>> scaling/composition and those that can't.
>>
>> I'm not convinced about this. The standard API behavior is not to expose
>> functionality that the hardware can't do. So if scaling isn't possible on
>> the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
>>
>> I also believe it very unlikely that we'll see encoders capable of scaling
>> as it doesn't make much sense.
> 
> It does make a lot of sense - WebRTC requires 3 different sizes of the
> stream to be encoded at the same time. However, unfortunately, I
> haven't yet seen an encoder capable of doing so.
> 
>> I would prefer to drop this to simplify the
>> spec, and when we get encoders that can scale, then we can add support for
>> compose rectangles (and I'm sure we'll need to think about how that
>> influences the CAPTURE side as well).
>>
>> For encoders without scaling it is the OUTPUT crop rectangle that defines
>> the visible rectangle.
>>
>>>
>>> However, with your proposal of actually having selection rectangles
>>> for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
>>> would expose a varying set of rectangles, depending on the hardware
>>> capability, while the CAPTURE queue would always expose its rectangle
>>> with that information.
>>
>> I think we should keep it simple and only define selection rectangles
>> when really needed.
>>
>> So encoders support CROP on the OUTPUT, and decoders support CAPTURE
>> COMPOSE (may be read-only). Nothing else.
>>
>> Once support for scaling is needed (either on the encoder or decoder
>> side), then the spec should be enhanced. But I prefer to postpone that
>> until we actually have hardware that needs this.
>>
> 
> Okay, let's do it this way then. Actually, I don't even think there is
> much value in exposing information internal to the bitstream metadata
> like this, similarly to the coded size. My intention was to just
> ensure that we can easily add scaling/composing functionality later.
> 
> I just removed the COMPOSE rectangles from my next draft.

I don't think that supporting scaling will be a problem for the API as
such, since this is supported for standard video capture devices. It
just gets very complicated trying to describe how to configure all this.

So I prefer to avoid this until we need to.

> 
> [snip]
>>>
>>>>
>>>> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
>>>> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
>>>> particular) and CAPTURE buffers were already allocated and are too small.
>>>
>>> The OUTPUT format must not change the CAPTURE format by definition.
>>> Otherwise we end up in a situation where we can't commit, because both
>>> queue formats can affect each other. Any change to the OUTPUT format
>>> that wouldn't work with the current CAPTURE format should be adjusted
>>> by the driver to match the current CAPTURE format.
>>
>> But the CAPTURE format *does* depend on the OUTPUT format: if the output
>> resolution changes, then so does the CAPTURE resolution and esp. the
>> sizeimage value, since that is typically resolution dependent.
>>
>> The coda driver does this as well: changing the output resolution
>> will update the capture resolution and sizeimage. The vicodec driver does the
>> same.
>>
>> Setting the CAPTURE format basically just selects the codec to use, after
>> that you can set the OUTPUT format and read the updated CAPTURE format to
>> get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
>> change the OUTPUT format, unless the OUTPUT format is incompatible with the
>> newly selected codec.
> 
> Let me think about it for a while.

Sleep on it, always works well for me :-)

Regards,

	Hans

Tomasz Figa April 9, 2019, 7:11 a.m. UTC | #18

On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:
>
> On 4/8/19 10:36 AM, Tomasz Figa wrote:
> > On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
> >>
> >> Another comment found while creating compliance tests:
> >>
> >> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >>> +Drain
> >>> +=====
> >>> +
> >>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> >>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> >>> +drain sequence described below. After the drain sequence ends, the client has
> >>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> >>> +sequence was started.
> >>> +
> >>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> >>> +
> >>> +   * **Required fields:**
> >>> +
> >>> +     ``cmd``
> >>> +         set to ``V4L2_ENC_CMD_STOP``
> >>> +
> >>> +     ``flags``
> >>> +         set to 0
> >>> +
> >>> +     ``pts``
> >>> +         set to 0
> >>> +
> >>> +   .. warning::
> >>> +
> >>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> >>> +      queues are streaming. For compatibility reasons, the call to
> >>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> >>> +      not streaming, but at the same time it will not initiate the `Drain`
> >>> +      sequence and so the steps described below would not be applicable.
> >>> +
> >>> +2. Any ``OUTPUT`` buffers queued by the client before the
> >>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> >>> +   normal. The client must continue to handle both queues independently,
> >>> +   similarly to normal encode operation. This includes:
> >>> +
> >>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> >>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> >>> +
> >>> +     .. warning::
> >>> +
> >>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> >>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> >>> +        as it does not contain an encoded frame.
> >>> +
> >>> +     .. note::
> >>> +
> >>> +        Any attempt to dequeue more buffers beyond the buffer marked with
> >>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> >>> +        :c:func:`VIDIOC_DQBUF`.
> >>> +
> >>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> >>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> >>> +
> >>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> >>> +
> >>> +   .. note::
> >>> +
> >>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> >>> +      event when the last frame has been decoded and all frames are ready to be
> >>> +      dequeued. It is deprecated behavior and the client must not rely on it.
> >>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> >>> +
> >>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> >>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> >>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> >>> +   until the client issues any of the following operations:
> >>> +
> >>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> >>> +     operation normally, with all the state from before the drain,
> >>
> >> I assume that calling CMD_START when *not* draining will succeed but does nothing.
> >>
> >> In other words: while draining is in progress START will return EBUSY. When draining
> >> was finished, then START will resume the encoder. In all other cases it just returns
> >> 0 since the encoder is really already started.
> >>
> >
> > Yes, that was the intention and seems to be the closest to the
> > behavior described in the existing documentation.
> >
> >>> +
> >>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> >>> +     and then resume encoding,
> >>> +
> >>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> >>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> >>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> >>> +
> >>> +.. note::
> >>> +
> >>> +   Once the drain sequence is initiated, the client needs to drive it to
> >>> +   completion, as described by the steps above, unless it aborts the process by
> >>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> >>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> >>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> >>> +   will fail with -EBUSY error code if attempted.
> >>
> >> I assume calling STOP again once the drain sequence completed just returns 0 and
> >> doesn't do anything else (since we're already stopped).
> >>
> >
> > Right.
> >
> >>> +
> >>> +   Although mandatory, the availability of encoder commands may be queried
> >>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> >>
> >> Some corner cases:
> >>
> >> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
> >>    Now ENC_CMD_STOP is issued. What should happen?
> >>
> >>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >>    at once as an empty buffer with FLAG_LAST set.
> >>
> >
> > SGTM. It's actually similar to a relatively common case where all
> > CAPTURE buffers have been dequeued and the application has to refill
> > the CAPTURE queue, but in the meantime a drain request needs to be
> > issued.
> >
> >> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
> >>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> >>
> >>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
> >>    previous corner case.
> >
> > Agreed.
> >
> >>
> >> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
> >>    is issued. What should happen?
> >>
> >>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
> >>    buffer with FLAG_LAST set.
> >
> > Generally agreed, but not sure if there is a reason to specifically
> > refer to the oldest buffer. (I'm personally for keeping the queues
> > ordered, though...)
>
> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
> queued up for userspace to signal that the encoder has stopped." Or something along
> those lines.

I've added a note:

   For reference, handling of various corner cases is described below:

   * In case of no buffer in the ``OUTPUT`` queue at the time the
     ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
     immediately and the encoder returns an empty ``CAPTURE`` buffer with the
     ``V4L2_BUF_FLAG_LAST`` flag set.

   * In case of no buffer in the ``CAPTURE`` queue at the time the drain
     sequence completes, the next time the client queues a ``CAPTURE`` buffer
     it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
     flag set.

   * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
     middle of the drain sequence, the drain sequence is cancelled and all
     ``CAPTURE`` buffers are implicitly returned to the userpace.

   * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
     middle of the drain sequence, the drain sequence completes immediately and
     next ``CAPTURE`` buffer will be returned empty with the
     ``V4L2_BUF_FLAG_LAST`` flag set.

Slightly changed the split into cases to cover behaviors rather than
conditions. WDYT?

Best regards,
Tomasz

>
> Regards,
>
>         Hans
>
> >
> >>
> >> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
> >>    Before the drain process completes STREAMOFF is called for either CAPTURE or
> >>    OUTPUT queue. What should happen?
> >>
> >>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
> >>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
> >>    will be used as input to the encoder.
> >>
> >
> > Agreed.
> >
> >>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
> >>    FLAG_LAST set.
> >
> > Agreed.
> >
> >>
> >> Some of this might have to be documented, but these corner cases should certainly be
> >> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> >> with this.
> >
> > Agreed with just one minor comment. Thanks for checking with me!
> >
> > Best regards,
> > Tomasz
> >
>

Tomasz Figa April 9, 2019, 9:35 a.m. UTC | #19

On Mon, Apr 8, 2019 at 8:11 PM Hans Verkuil <[e-mail]> wrote:
>
> On 4/8/19 11:23 AM, Tomasz Figa wrote:
> > On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
> >>
> >> On 4/5/19 10:12 AM, Tomasz Figa wrote:
> >>> On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
> >>>>
> >>>> Hi Tomasz,
> >>>>
> >>>> Some more comments...
> >>>>
> >>>> On 1/29/19 2:52 PM, Hans Verkuil wrote:
> >>>>> Hi Tomasz,
> >>>>>
> >>>>> Some comments below. Nothing major, so I think a v4 should be ready to be
> >>>>> merged.
> >>>>>
> >>>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >>>>>> Due to complexity of the video encoding process, the V4L2 drivers of
> >>>>>> stateful encoder hardware require specific sequences of V4L2 API calls
> >>>>>> to be followed. These include capability enumeration, initialization,
> >>>>>> encoding, encode parameters change, drain and reset.
> >>>>>>
> >>>>>> Specifics of the above have been discussed during Media Workshops at
> >>>>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> >>>>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> >>>>>> originated at those events was later implemented by the drivers we already
> >>>>>> have merged in mainline, such as s5p-mfc or coda.
> >>>>>>
> >>>>>> The only thing missing was the real specification included as a part of
> >>>>>> Linux Media documentation. Fix it now and document the encoder part of
> >>>>>> the Codec API.
> >>>>>>
> >>>>>> Signed-off-by: Tomasz Figa <[e-mail]>
> >>>>>> ---
> >>>>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> >>>>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> >>>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> >>>>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> >>>>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> >>>>>>  5 files changed, 617 insertions(+), 15 deletions(-)
> >>>>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> >>>>>>
> >>>>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>>>>> new file mode 100644
> >>>>>> index 000000000000..fb8b05a132ee
> >>>>>> --- /dev/null
> >>>>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> >>>>>> @@ -0,0 +1,586 @@
> >>>>>> +.. -*- coding: utf-8; mode: rst -*-
> >>>>>> +
> >>>>>> +.. _encoder:
> >>>>>> +
> >>>>>> +*************************************************
> >>>>>> +Memory-to-memory Stateful Video Encoder Interface
> >>>>>> +*************************************************
> >>>>>> +
> >>>>>> +A stateful video encoder takes raw video frames in display order and encodes
> >>>>>> +them into a bitstream. It generates complete chunks of the bitstream, including
> >>>>>> +all metadata, headers, etc. The resulting bitstream does not require any
> >>>>>> +further post-processing by the client.
> >>>>>> +
> >>>>>> +Performing software stream processing, header generation etc. in the driver
> >>>>>> +in order to support this interface is strongly discouraged. In case such
> >>>>>> +operations are needed, use of the Stateless Video Encoder Interface (in
> >>>>>> +development) is strongly advised.
> >>>>>> +
> >>>>>> +Conventions and notation used in this document
> >>>>>> +==============================================
> >>>>>> +
> >>>>>> +1. The general V4L2 API rules apply if not specified in this document
> >>>>>> +   otherwise.
> >>>>>> +
> >>>>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> >>>>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> >>>>>> +
> >>>>>> +3. All steps not marked "optional" are required.
> >>>>>> +
> >>>>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> >>>>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> >>>>>> +   unless specified otherwise.
> >>>>>> +
> >>>>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> >>>>>> +   used interchangeably with multi-planar API, unless specified otherwise,
> >>>>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
> >>>>>> +
> >>>>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> >>>>>> +   [0..2]: i = 0, 1, 2.
> >>>>>> +
> >>>>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> >>>>>> +   queue containing data that resulted from processing buffer A.
> >>>>>> +
> >>>>>> +Glossary
> >>>>>> +========
> >>>>>> +
> >>>>>> +Refer to :ref:`decoder-glossary`.
> >>>>>> +
> >>>>>> +State machine
> >>>>>> +=============
> >>>>>> +
> >>>>>> +.. kernel-render:: DOT
> >>>>>> +   :alt: DOT digraph of encoder state machine
> >>>>>> +   :caption: Encoder state machine
> >>>>>> +
> >>>>>> +   digraph encoder_state_machine {
> >>>>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
> >>>>>> +
> >>>>>> +       node [shape = circle, label="Initialization"] Initialization;
> >>>>>> +       node [shape = circle, label="Stopped"] Stopped;
> >>>>>> +       node [shape = circle, label="Drain"] Drain;
> >>>>>> +       node [shape = circle, label="Reset"] Reset;
> >>>>>> +
> >>>>>> +       node [shape = point]; qi
> >>>>>> +       qi -> Initialization [ label = "open()" ];
> >>>>>> +
> >>>>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> >>>>>> +
> >>>>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> >>>>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> >>>>>> +       Encoding -> Encoding;
> >>>>>> +
> >>>>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> >>>>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>>>> +
> >>>>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> >>>>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> >>>>>> +
> >>>>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> >>>>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> >>>>>> +   }
> >>>>>> +
> >>>>>> +Querying capabilities
> >>>>>> +=====================
> >>>>>> +
> >>>>>> +1. To enumerate the set of coded formats supported by the encoder, the
> >>>>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> >>>>>> +
> >>>>>> +   * The full set of supported formats will be returned, regardless of the
> >>>>>> +     format set on ``OUTPUT``.
> >>>>>> +
> >>>>>> +2. To enumerate the set of supported raw formats, the client may call
> >>>>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> >>>>>> +
> >>>>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
> >>>>>> +     will be returned.
> >>>>>> +
> >>>>>> +   * In order to enumerate raw formats supported by a given coded format,
> >>>>>> +     the client must first set that coded format on ``CAPTURE`` and then
> >>>>>> +     enumerate the formats on ``OUTPUT``.
> >>>>>> +
> >>>>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> >>>>>> +   resolutions for a given format, passing desired pixel format in
> >>>>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> >>>>>> +
> >>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> >>>>>> +     format will include all possible coded resolutions supported by the
> >>>>>> +     encoder for given coded pixel format.
> >>>>>> +
> >>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> >>>>>> +     will include all possible frame buffer resolutions supported by the
> >>>>>> +     encoder for given raw pixel format and coded format currently set on
> >>>>>> +     ``CAPTURE``.
> >>>>>> +
> >>>>>> +4. Supported profiles and levels for the coded format currently set on
> >>>>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> >>>>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
> >>>>>> +
> >>>>>> +5. Any additional encoder capabilities may be discovered by querying
> >>>>>> +   their respective controls.
> >>>>>> +
> >>>>>> +Initialization
> >>>>>> +==============
> >>>>>> +
> >>>>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> >>>>>> +
> >>>>>> +   * **Required fields:**
> >>>>>> +
> >>>>>> +     ``type``
> >>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> >>>>>> +
> >>>>>> +     ``pixelformat``
> >>>>>> +         the coded format to be produced
> >>>>>> +
> >>>>>> +     ``sizeimage``
> >>>>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> >>>>>> +         match hardware requirements
> >>>>>> +
> >>>>>> +     ``width``, ``height``
> >>>>>> +         ignored (always zero)
> >>>>>> +
> >>>>>> +     other fields
> >>>>>> +         follow standard semantics
> >>>>>> +
> >>>>>> +   * **Return fields:**
> >>>>>> +
> >>>>>> +     ``sizeimage``
> >>>>>> +         adjusted size of ``CAPTURE`` buffers
> >>>>>> +
> >>>>>> +   .. important::
> >>>>>> +
> >>>>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> >>>>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
> >>>>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
> >>>>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> >>>>>> +      must adjust it afterwards.
> >>>>>
> >>>>> Hmm, "including resolution": if width and height are set to 0, what should the
> >>>>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
> >>>>> at a first reading of this paragraph it appears to be contradictory.
> >>>>
> >>>> I think the driver should just return the width and height of the OUTPUT
> >>>> format. So the width and height that userspace specifies is just ignored
> >>>> and replaced by the width and height of the OUTPUT format. After all, that's
> >>>> what the bitstream will encode. Returning 0 for width and height would make
> >>>> this a strange exception in V4L2 and I want to avoid that.
> >>>>
> >>>
> >>> Hmm, however, the width and height of the OUTPUT format is not what's
> >>> actually encoded in the bitstream. The right selection rectangle
> >>> determines that.
> >>>
> >>> In one of the previous versions I though we could put the codec
> >
> > s/codec/coded/...
> >
> >>> resolution as the width and height of the CAPTURE format, which would
> >>> be the resolution of the encoded image rounded up to full macroblocks
> >>> +/- some encoder-specific constraints. AFAIR there was some concern
> >>> about OUTPUT format changes triggering CAPTURE format changes, but to
> >>> be honest, I'm not sure if that's really a problem. I just decided to
> >>> drop that for the simplicity.
> >>
> >> I'm not sure what your point is.
> >>
> >> The OUTPUT format has the coded resolution,
> >
> > That's not always true. The OUTPUT format is just the format of the
> > source frame buffers. In special cases where the source resolution is
> > nicely aligned, it would be the same as coded size, but the remaining
> > cases are valid as well.
> >
> >> so when you set the
> >> CAPTURE format it can just copy the OUTPUT coded resolution unless the
> >> chosen CAPTURE pixelformat can't handle that in which case both the
> >> OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
> >> or minimum the codec is capable of.
> >
> > As per my comment above, generally speaking, the encoder will derive
> > an appropriate coded format from the OUTPUT format, but also other
> > factors, like the crop rectangles and possibly some internal
> > constraints.
> >
> >>
> >> That said, I am fine with just leaving it up to the driver as suggested
> >> before. Just as long as both the CAPTURE and OUTPUT formats remain valid
> >> (i.e. width and height may never be out of range).
> >>
> >
> > Sounds good to me.
> >
> >>>
> >>>>>
> >>>>>> +
> >>>>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> >>>>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> >>>>>> +
> >>>>>> +   * **Required fields:**
> >>>>>> +
> >>>>>> +     ``type``
> >>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>>>> +
> >>>>>> +     other fields
> >>>>>> +         follow standard semantics
> >>>>>> +
> >>>>>> +   * **Return fields:**
> >>>>>> +
> >>>>>> +     ``pixelformat``
> >>>>>> +         raw format supported for the coded format currently selected on
> >>>>>> +         the ``CAPTURE`` queue.
> >>>>>> +
> >>>>>> +     other fields
> >>>>>> +         follow standard semantics
> >>>>>> +
> >>>>>> +3. Set the raw source format on the ``OUTPUT`` queue via
> >>>>>> +   :c:func:`VIDIOC_S_FMT`.
> >>>>>> +
> >>>>>> +   * **Required fields:**
> >>>>>> +
> >>>>>> +     ``type``
> >>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>>>> +
> >>>>>> +     ``pixelformat``
> >>>>>> +         raw format of the source
> >>>>>> +
> >>>>>> +     ``width``, ``height``
> >>>>>> +         source resolution
> >>>>>> +
> >>>>>> +     other fields
> >>>>>> +         follow standard semantics
> >>>>>> +
> >>>>>> +   * **Return fields:**
> >>>>>> +
> >>>>>> +     ``width``, ``height``
> >>>>>> +         may be adjusted by encoder to match alignment requirements, as
> >>>>>> +         required by the currently selected formats
> >>>>>
> >>>>> What if the width x height is larger than the maximum supported by the
> >>>>> selected coded format? This should probably mention that in that case the
> >>>>> width x height is reduced to the largest allowed value. Also mention that
> >>>>> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> >>>>>
> >>>>>> +
> >>>>>> +     other fields
> >>>>>> +         follow standard semantics
> >>>>>> +
> >>>>>> +   * Setting the source resolution will reset the selection rectangles to their
> >>>>>> +     default values, based on the new resolution, as described in the step 5
> >>>>>
> >>>>> 5 -> 4
> >>>>>
> >>>>> Or just say: "as described in the next step."
> >>>>>
> >>>>>> +     below.
> >>>>
> >>>> It should also be made explicit that:
> >>>>
> >>>> 1) the crop rectangle will be set to the given width and height *before*
> >>>> it is being adjusted by S_FMT.
> >>>>
> >>>
> >>> I don't think that's what we want here.
> >>>
> >>> Defining the default rectangle to be exactly the same as the OUTPUT
> >>> resolution (after the adjustment) makes the semantics consistent - not
> >>> setting the crop rectangle gives you exactly the behavior as if there
> >>> was no cropping involved (or supported by the encoder).
> >>
> >> I think you are right. This seems to be what the coda driver does as well.
> >> It is convenient to be able to just set a 1920x1080 format and have that
> >> resolution be stored as the crop rectangle, since it avoids having to call
> >> s_selection afterwards, but it is not really consistent with the way V4L2
> >> works.
> >>
> >>>
> >>>> Open question: should we support a compose rectangle for the CAPTURE that
> >>>> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
> >>>> the adjusted width and height and the compose rectangle (read-only) contains
> >>>> the visible width and height. It's not strictly necessary, but it is
> >>>> symmetrical.
> >>>
> >>> Wouldn't it rather be the CAPTURE crop rectangle that would be of the
> >>> same resolution of the OUTPUT compose rectangle? Then you could
> >>> actually have the CAPTURE compose rectangle for putting that into the
> >>> desired rectangle of the encoded stream, if the encoder supports that.
> >>> (I don't know any that does, so probably out of concern for now.)
> >>
> >> Yes, you are right.
> >>
> >> But should we support this?
> >>
> >> I actually think not for this initial version. It can be added later, I guess.
> >>
> >
> > I think it boils down on whether adding it later wouldn't
> > significantly complicate the application logic. It also relates to my
> > other comment somewhere below.
> >
> >>>
> >>>>
> >>>> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
> >>>> height. The CAPTURE sizeimage might change as well.
> >>>>
> >>>>>> +
> >>>>>> +4. **Optional.** Set the visible resolution for the stream metadata via
> >>>>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> >>>>
> >>>> I think you should mention that this is only necessary if the crop rectangle
> >>>> that is set when you set the format isn't what you want.
> >>>>
> >>>
> >>> Ack.
> >>>
> >>>>>> +
> >>>>>> +   * **Required fields:**
> >>>>>> +
> >>>>>> +     ``type``
> >>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> >>>>>> +
> >>>>>> +     ``target``
> >>>>>> +         set to ``V4L2_SEL_TGT_CROP``
> >>>>>> +
> >>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>>>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> >>>>>> +         rectangle and may be subject to adjustment to match codec and
> >>>>>> +         hardware constraints
> >>>>>> +
> >>>>>> +   * **Return fields:**
> >>>>>> +
> >>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> >>>>>> +         visible rectangle adjusted by the encoder
> >>>>>> +
> >>>>>> +   * The following selection targets are supported on ``OUTPUT``:
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>>>>> +         equal to the full source frame, matching the active ``OUTPUT``
> >>>>>> +         format
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>>>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_CROP``
> >>>>>> +         rectangle within the source buffer to be encoded into the
> >>>>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> >>>>>> +
> >>>>>> +         .. note::
> >>>>>> +
> >>>>>> +            A common use case for this selection target is encoding a source
> >>>>>> +            video with a resolution that is not a multiple of a macroblock,
> >>>>>> +            e.g.  the common 1920x1080 resolution may require the source
> >>>>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> >>>>>> +            size. To avoid encoding the padding, the client needs to explicitly
> >>>>>> +            configure this selection target to 1920x1080.
> >>>>
> >>>> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
> >>>>
> >>>
> >>> Sorry, which part exactly and what part of the proposal exactly? :)
> >>> (My comment above might be related, though.)
> >>
> >> Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
> >> crop selection target, so this note remains valid.
> >>
> >
> > Ack.
> >
> >>>
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> >>>>>> +         maximum rectangle within the coded resolution, which the cropped
> >>>>>> +         source frame can be composed into; if the hardware does not support
> >>>>>> +         composition or scaling, then this is always equal to the rectangle of
> >>>>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> >>>>>> +         equal to a rectangle of width and height matching
> >>>>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> >>>>>> +
> >>>>>> +     ``V4L2_SEL_TGT_COMPOSE``
> >>>>>> +         rectangle within the coded frame, which the cropped source frame
> >>>>>> +         is to be composed into; defaults to
> >>>>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> >>>>>> +         additional compose/scaling capabilities; resulting stream will
> >>>>>> +         have this rectangle encoded as the visible rectangle in its
> >>>>>> +         metadata
> >>>>
> >>>> I think the compose targets for OUTPUT are only needed if the hardware can
> >>>> actually do scaling and/or composition. Otherwise they can (must?) be
> >>>> dropped.
> >>>>
> >>>
> >>> Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
> >>> userspace to learn the target visible rectangle that's going to be
> >>> encoded in the stream metadata. If we omit it, we wouldn't have a way
> >>> that would be consistent between encoders that can do
> >>> scaling/composition and those that can't.
> >>
> >> I'm not convinced about this. The standard API behavior is not to expose
> >> functionality that the hardware can't do. So if scaling isn't possible on
> >> the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
> >>
> >> I also believe it very unlikely that we'll see encoders capable of scaling
> >> as it doesn't make much sense.
> >
> > It does make a lot of sense - WebRTC requires 3 different sizes of the
> > stream to be encoded at the same time. However, unfortunately, I
> > haven't yet seen an encoder capable of doing so.
> >
> >> I would prefer to drop this to simplify the
> >> spec, and when we get encoders that can scale, then we can add support for
> >> compose rectangles (and I'm sure we'll need to think about how that
> >> influences the CAPTURE side as well).
> >>
> >> For encoders without scaling it is the OUTPUT crop rectangle that defines
> >> the visible rectangle.
> >>
> >>>
> >>> However, with your proposal of actually having selection rectangles
> >>> for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
> >>> would expose a varying set of rectangles, depending on the hardware
> >>> capability, while the CAPTURE queue would always expose its rectangle
> >>> with that information.
> >>
> >> I think we should keep it simple and only define selection rectangles
> >> when really needed.
> >>
> >> So encoders support CROP on the OUTPUT, and decoders support CAPTURE
> >> COMPOSE (may be read-only). Nothing else.
> >>
> >> Once support for scaling is needed (either on the encoder or decoder
> >> side), then the spec should be enhanced. But I prefer to postpone that
> >> until we actually have hardware that needs this.
> >>
> >
> > Okay, let's do it this way then. Actually, I don't even think there is
> > much value in exposing information internal to the bitstream metadata
> > like this, similarly to the coded size. My intention was to just
> > ensure that we can easily add scaling/composing functionality later.
> >
> > I just removed the COMPOSE rectangles from my next draft.
>
> I don't think that supporting scaling will be a problem for the API as
> such, since this is supported for standard video capture devices. It
> just gets very complicated trying to describe how to configure all this.
>
> So I prefer to avoid this until we need to.
>
> >
> > [snip]
> >>>
> >>>>
> >>>> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
> >>>> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
> >>>> particular) and CAPTURE buffers were already allocated and are too small.
> >>>
> >>> The OUTPUT format must not change the CAPTURE format by definition.
> >>> Otherwise we end up in a situation where we can't commit, because both
> >>> queue formats can affect each other. Any change to the OUTPUT format
> >>> that wouldn't work with the current CAPTURE format should be adjusted
> >>> by the driver to match the current CAPTURE format.
> >>
> >> But the CAPTURE format *does* depend on the OUTPUT format: if the output
> >> resolution changes, then so does the CAPTURE resolution and esp. the
> >> sizeimage value, since that is typically resolution dependent.
> >>
> >> The coda driver does this as well: changing the output resolution
> >> will update the capture resolution and sizeimage. The vicodec driver does the
> >> same.
> >>
> >> Setting the CAPTURE format basically just selects the codec to use, after
> >> that you can set the OUTPUT format and read the updated CAPTURE format to
> >> get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
> >> change the OUTPUT format, unless the OUTPUT format is incompatible with the
> >> newly selected codec.
> >
> > Let me think about it for a while.
>
> Sleep on it, always works well for me :-)

Okay, I think I'm not convinced.

I believe we decided to allow sizeimage to be specified by the
application, because it knows more about the stream it's going to
encode. Only setting the size to 0 would make the encoder fall back to
some simple internal heuristic.

Another thing is handling resolution changes. I believe that would
have to be handled by stopping the OUTPUT queue, changing the OUTPUT
format and starting the OUTPUT queue, all that without stopping the
CAPTURE queue. With the behavior you described it wouldn't work,
because the OUTPUT format couldn't be changed.

I'd suggest making OUTPUT format changes not change the CAPTURE sizeimage.

Best regards,
Tomasz

Hans Verkuil April 9, 2019, 9:37 a.m. UTC | #20

On 4/9/19 9:11 AM, Tomasz Figa wrote:
> On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:
>>
>> On 4/8/19 10:36 AM, Tomasz Figa wrote:
>>> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
>>>>
>>>> Another comment found while creating compliance tests:
>>>>
>>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>>>> +Drain
>>>>> +=====
>>>>> +
>>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>>>>> +drain sequence described below. After the drain sequence ends, the client has
>>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>>>>> +sequence was started.
>>>>> +
>>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>>>>> +
>>>>> +   * **Required fields:**
>>>>> +
>>>>> +     ``cmd``
>>>>> +         set to ``V4L2_ENC_CMD_STOP``
>>>>> +
>>>>> +     ``flags``
>>>>> +         set to 0
>>>>> +
>>>>> +     ``pts``
>>>>> +         set to 0
>>>>> +
>>>>> +   .. warning::
>>>>> +
>>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>>>>> +      queues are streaming. For compatibility reasons, the call to
>>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>>>>> +      not streaming, but at the same time it will not initiate the `Drain`
>>>>> +      sequence and so the steps described below would not be applicable.
>>>>> +
>>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
>>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>>>>> +   normal. The client must continue to handle both queues independently,
>>>>> +   similarly to normal encode operation. This includes:
>>>>> +
>>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>>>>> +
>>>>> +     .. warning::
>>>>> +
>>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>>>>> +        as it does not contain an encoded frame.
>>>>> +
>>>>> +     .. note::
>>>>> +
>>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>>>>> +        :c:func:`VIDIOC_DQBUF`.
>>>>> +
>>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>>>>> +
>>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>>>>> +
>>>>> +   .. note::
>>>>> +
>>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>>>>> +      event when the last frame has been decoded and all frames are ready to be
>>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>>>>> +
>>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>>>>> +   until the client issues any of the following operations:
>>>>> +
>>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>>>>> +     operation normally, with all the state from before the drain,
>>>>
>>>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
>>>>
>>>> In other words: while draining is in progress START will return EBUSY. When draining
>>>> was finished, then START will resume the encoder. In all other cases it just returns
>>>> 0 since the encoder is really already started.
>>>>
>>>
>>> Yes, that was the intention and seems to be the closest to the
>>> behavior described in the existing documentation.
>>>
>>>>> +
>>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>>>>> +     and then resume encoding,
>>>>> +
>>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>>>>> +
>>>>> +.. note::
>>>>> +
>>>>> +   Once the drain sequence is initiated, the client needs to drive it to
>>>>> +   completion, as described by the steps above, unless it aborts the process by
>>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>>>>> +   will fail with -EBUSY error code if attempted.
>>>>
>>>> I assume calling STOP again once the drain sequence completed just returns 0 and
>>>> doesn't do anything else (since we're already stopped).
>>>>
>>>
>>> Right.
>>>
>>>>> +
>>>>> +   Although mandatory, the availability of encoder commands may be queried
>>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>>>>
>>>> Some corner cases:
>>>>
>>>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
>>>>    Now ENC_CMD_STOP is issued. What should happen?
>>>>
>>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>>>    at once as an empty buffer with FLAG_LAST set.
>>>>
>>>
>>> SGTM. It's actually similar to a relatively common case where all
>>> CAPTURE buffers have been dequeued and the application has to refill
>>> the CAPTURE queue, but in the meantime a drain request needs to be
>>> issued.
>>>
>>>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
>>>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
>>>>
>>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
>>>>    previous corner case.
>>>
>>> Agreed.
>>>
>>>>
>>>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
>>>>    is issued. What should happen?
>>>>
>>>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
>>>>    buffer with FLAG_LAST set.
>>>
>>> Generally agreed, but not sure if there is a reason to specifically
>>> refer to the oldest buffer. (I'm personally for keeping the queues
>>> ordered, though...)
>>
>> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
>> queued up for userspace to signal that the encoder has stopped." Or something along
>> those lines.
> 
> I've added a note:
> 
>    For reference, handling of various corner cases is described below:
> 
>    * In case of no buffer in the ``OUTPUT`` queue at the time the
>      ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
>      immediately and the encoder returns an empty ``CAPTURE`` buffer with the
>      ``V4L2_BUF_FLAG_LAST`` flag set.
> 
>    * In case of no buffer in the ``CAPTURE`` queue at the time the drain
>      sequence completes, the next time the client queues a ``CAPTURE`` buffer
>      it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
>      flag set.
> 
>    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
>      middle of the drain sequence, the drain sequence is cancelled and all

cancelled -> canceled

>      ``CAPTURE`` buffers are implicitly returned to the userpace.

userpace -> userspace

> 
>    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
>      middle of the drain sequence, the drain sequence completes immediately and
>      next ``CAPTURE`` buffer will be returned empty with the
>      ``V4L2_BUF_FLAG_LAST`` flag set.
> 
> Slightly changed the split into cases to cover behaviors rather than
> conditions. WDYT?

Looks good (with those two typos fixed).

	Hans

> 
> Best regards,
> Tomasz
> 
>>
>> Regards,
>>
>>         Hans
>>
>>>
>>>>
>>>> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
>>>>    Before the drain process completes STREAMOFF is called for either CAPTURE or
>>>>    OUTPUT queue. What should happen?
>>>>
>>>>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
>>>>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
>>>>    will be used as input to the encoder.
>>>>
>>>
>>> Agreed.
>>>
>>>>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
>>>>    FLAG_LAST set.
>>>
>>> Agreed.
>>>
>>>>
>>>> Some of this might have to be documented, but these corner cases should certainly be
>>>> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
>>>> with this.
>>>
>>> Agreed with just one minor comment. Thanks for checking with me!
>>>
>>> Best regards,
>>> Tomasz
>>>
>>

Tomasz Figa April 9, 2019, 9:43 a.m. UTC | #21

On Tue, Apr 9, 2019 at 6:37 PM Hans Verkuil <[e-mail]> wrote:
>
> On 4/9/19 9:11 AM, Tomasz Figa wrote:
> > On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:
> >>
> >> On 4/8/19 10:36 AM, Tomasz Figa wrote:
> >>> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
> >>>>
> >>>> Another comment found while creating compliance tests:
> >>>>
> >>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> >>>>> +Drain
> >>>>> +=====
> >>>>> +
> >>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> >>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> >>>>> +drain sequence described below. After the drain sequence ends, the client has
> >>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> >>>>> +sequence was started.
> >>>>> +
> >>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> >>>>> +
> >>>>> +   * **Required fields:**
> >>>>> +
> >>>>> +     ``cmd``
> >>>>> +         set to ``V4L2_ENC_CMD_STOP``
> >>>>> +
> >>>>> +     ``flags``
> >>>>> +         set to 0
> >>>>> +
> >>>>> +     ``pts``
> >>>>> +         set to 0
> >>>>> +
> >>>>> +   .. warning::
> >>>>> +
> >>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> >>>>> +      queues are streaming. For compatibility reasons, the call to
> >>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> >>>>> +      not streaming, but at the same time it will not initiate the `Drain`
> >>>>> +      sequence and so the steps described below would not be applicable.
> >>>>> +
> >>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
> >>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> >>>>> +   normal. The client must continue to handle both queues independently,
> >>>>> +   similarly to normal encode operation. This includes:
> >>>>> +
> >>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> >>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> >>>>> +
> >>>>> +     .. warning::
> >>>>> +
> >>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> >>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> >>>>> +        as it does not contain an encoded frame.
> >>>>> +
> >>>>> +     .. note::
> >>>>> +
> >>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
> >>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> >>>>> +        :c:func:`VIDIOC_DQBUF`.
> >>>>> +
> >>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> >>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> >>>>> +
> >>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> >>>>> +
> >>>>> +   .. note::
> >>>>> +
> >>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> >>>>> +      event when the last frame has been decoded and all frames are ready to be
> >>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
> >>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> >>>>> +
> >>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> >>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> >>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> >>>>> +   until the client issues any of the following operations:
> >>>>> +
> >>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> >>>>> +     operation normally, with all the state from before the drain,
> >>>>
> >>>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
> >>>>
> >>>> In other words: while draining is in progress START will return EBUSY. When draining
> >>>> was finished, then START will resume the encoder. In all other cases it just returns
> >>>> 0 since the encoder is really already started.
> >>>>
> >>>
> >>> Yes, that was the intention and seems to be the closest to the
> >>> behavior described in the existing documentation.
> >>>
> >>>>> +
> >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> >>>>> +     and then resume encoding,
> >>>>> +
> >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> >>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> >>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> >>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> >>>>> +
> >>>>> +.. note::
> >>>>> +
> >>>>> +   Once the drain sequence is initiated, the client needs to drive it to
> >>>>> +   completion, as described by the steps above, unless it aborts the process by
> >>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> >>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> >>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> >>>>> +   will fail with -EBUSY error code if attempted.
> >>>>
> >>>> I assume calling STOP again once the drain sequence completed just returns 0 and
> >>>> doesn't do anything else (since we're already stopped).
> >>>>
> >>>
> >>> Right.
> >>>
> >>>>> +
> >>>>> +   Although mandatory, the availability of encoder commands may be queried
> >>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> >>>>
> >>>> Some corner cases:
> >>>>
> >>>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
> >>>>    Now ENC_CMD_STOP is issued. What should happen?
> >>>>
> >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >>>>    at once as an empty buffer with FLAG_LAST set.
> >>>>
> >>>
> >>> SGTM. It's actually similar to a relatively common case where all
> >>> CAPTURE buffers have been dequeued and the application has to refill
> >>> the CAPTURE queue, but in the meantime a drain request needs to be
> >>> issued.
> >>>
> >>>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
> >>>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> >>>>
> >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> >>>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
> >>>>    previous corner case.
> >>>
> >>> Agreed.
> >>>
> >>>>
> >>>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
> >>>>    is issued. What should happen?
> >>>>
> >>>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
> >>>>    buffer with FLAG_LAST set.
> >>>
> >>> Generally agreed, but not sure if there is a reason to specifically
> >>> refer to the oldest buffer. (I'm personally for keeping the queues
> >>> ordered, though...)
> >>
> >> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
> >> queued up for userspace to signal that the encoder has stopped." Or something along
> >> those lines.
> >
> > I've added a note:
> >
> >    For reference, handling of various corner cases is described below:
> >
> >    * In case of no buffer in the ``OUTPUT`` queue at the time the
> >      ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
> >      immediately and the encoder returns an empty ``CAPTURE`` buffer with the
> >      ``V4L2_BUF_FLAG_LAST`` flag set.
> >
> >    * In case of no buffer in the ``CAPTURE`` queue at the time the drain
> >      sequence completes, the next time the client queues a ``CAPTURE`` buffer
> >      it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
> >      flag set.
> >
> >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
> >      middle of the drain sequence, the drain sequence is cancelled and all
>
> cancelled -> canceled
>
> >      ``CAPTURE`` buffers are implicitly returned to the userpace.
>
> userpace -> userspace
>

Actually changed it to "client".

> >
> >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
> >      middle of the drain sequence, the drain sequence completes immediately and
> >      next ``CAPTURE`` buffer will be returned empty with the
> >      ``V4L2_BUF_FLAG_LAST`` flag set.
> >
> > Slightly changed the split into cases to cover behaviors rather than
> > conditions. WDYT?
>
> Looks good (with those two typos fixed).
>

Thanks.

>         Hans
>
> >
> > Best regards,
> > Tomasz
> >
> >>
> >> Regards,
> >>
> >>         Hans
> >>
> >>>
> >>>>
> >>>> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
> >>>>    Before the drain process completes STREAMOFF is called for either CAPTURE or
> >>>>    OUTPUT queue. What should happen?
> >>>>
> >>>>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
> >>>>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
> >>>>    will be used as input to the encoder.
> >>>>
> >>>
> >>> Agreed.
> >>>
> >>>>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
> >>>>    FLAG_LAST set.
> >>>
> >>> Agreed.
> >>>
> >>>>
> >>>> Some of this might have to be documented, but these corner cases should certainly be
> >>>> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> >>>> with this.
> >>>
> >>> Agreed with just one minor comment. Thanks for checking with me!
> >>>
> >>> Best regards,
> >>> Tomasz
> >>>
> >>
>

Hans Verkuil April 10, 2019, 8:50 a.m. UTC | #22

On 4/9/19 11:35 AM, Tomasz Figa wrote:
> On Mon, Apr 8, 2019 at 8:11 PM Hans Verkuil <[e-mail]> wrote:
>>
>> On 4/8/19 11:23 AM, Tomasz Figa wrote:
>>> On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
>>>>
>>>> On 4/5/19 10:12 AM, Tomasz Figa wrote:
>>>>> On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
>>>>>>
>>>>>> Hi Tomasz,
>>>>>>
>>>>>> Some more comments...
>>>>>>
>>>>>> On 1/29/19 2:52 PM, Hans Verkuil wrote:
>>>>>>> Hi Tomasz,
>>>>>>>
>>>>>>> Some comments below. Nothing major, so I think a v4 should be ready to be
>>>>>>> merged.
>>>>>>>
>>>>>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>>>>>>> Due to complexity of the video encoding process, the V4L2 drivers of
>>>>>>>> stateful encoder hardware require specific sequences of V4L2 API calls
>>>>>>>> to be followed. These include capability enumeration, initialization,
>>>>>>>> encoding, encode parameters change, drain and reset.
>>>>>>>>
>>>>>>>> Specifics of the above have been discussed during Media Workshops at
>>>>>>>> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
>>>>>>>> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
>>>>>>>> originated at those events was later implemented by the drivers we already
>>>>>>>> have merged in mainline, such as s5p-mfc or coda.
>>>>>>>>
>>>>>>>> The only thing missing was the real specification included as a part of
>>>>>>>> Linux Media documentation. Fix it now and document the encoder part of
>>>>>>>> the Codec API.
>>>>>>>>
>>>>>>>> Signed-off-by: Tomasz Figa <[e-mail]>
>>>>>>>> ---
>>>>>>>>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>>>>>>>>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>>>>>>>>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>>>>>>>>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>>>>>>>>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>>>>>>>>  5 files changed, 617 insertions(+), 15 deletions(-)
>>>>>>>>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>>>>
>>>>>>>> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>>>> new file mode 100644
>>>>>>>> index 000000000000..fb8b05a132ee
>>>>>>>> --- /dev/null
>>>>>>>> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
>>>>>>>> @@ -0,0 +1,586 @@
>>>>>>>> +.. -*- coding: utf-8; mode: rst -*-
>>>>>>>> +
>>>>>>>> +.. _encoder:
>>>>>>>> +
>>>>>>>> +*************************************************
>>>>>>>> +Memory-to-memory Stateful Video Encoder Interface
>>>>>>>> +*************************************************
>>>>>>>> +
>>>>>>>> +A stateful video encoder takes raw video frames in display order and encodes
>>>>>>>> +them into a bitstream. It generates complete chunks of the bitstream, including
>>>>>>>> +all metadata, headers, etc. The resulting bitstream does not require any
>>>>>>>> +further post-processing by the client.
>>>>>>>> +
>>>>>>>> +Performing software stream processing, header generation etc. in the driver
>>>>>>>> +in order to support this interface is strongly discouraged. In case such
>>>>>>>> +operations are needed, use of the Stateless Video Encoder Interface (in
>>>>>>>> +development) is strongly advised.
>>>>>>>> +
>>>>>>>> +Conventions and notation used in this document
>>>>>>>> +==============================================
>>>>>>>> +
>>>>>>>> +1. The general V4L2 API rules apply if not specified in this document
>>>>>>>> +   otherwise.
>>>>>>>> +
>>>>>>>> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
>>>>>>>> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
>>>>>>>> +
>>>>>>>> +3. All steps not marked "optional" are required.
>>>>>>>> +
>>>>>>>> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
>>>>>>>> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
>>>>>>>> +   unless specified otherwise.
>>>>>>>> +
>>>>>>>> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
>>>>>>>> +   used interchangeably with multi-planar API, unless specified otherwise,
>>>>>>>> +   depending on decoder capabilities and following the general V4L2 guidelines.
>>>>>>>> +
>>>>>>>> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
>>>>>>>> +   [0..2]: i = 0, 1, 2.
>>>>>>>> +
>>>>>>>> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
>>>>>>>> +   queue containing data that resulted from processing buffer A.
>>>>>>>> +
>>>>>>>> +Glossary
>>>>>>>> +========
>>>>>>>> +
>>>>>>>> +Refer to :ref:`decoder-glossary`.
>>>>>>>> +
>>>>>>>> +State machine
>>>>>>>> +=============
>>>>>>>> +
>>>>>>>> +.. kernel-render:: DOT
>>>>>>>> +   :alt: DOT digraph of encoder state machine
>>>>>>>> +   :caption: Encoder state machine
>>>>>>>> +
>>>>>>>> +   digraph encoder_state_machine {
>>>>>>>> +       node [shape = doublecircle, label="Encoding"] Encoding;
>>>>>>>> +
>>>>>>>> +       node [shape = circle, label="Initialization"] Initialization;
>>>>>>>> +       node [shape = circle, label="Stopped"] Stopped;
>>>>>>>> +       node [shape = circle, label="Drain"] Drain;
>>>>>>>> +       node [shape = circle, label="Reset"] Reset;
>>>>>>>> +
>>>>>>>> +       node [shape = point]; qi
>>>>>>>> +       qi -> Initialization [ label = "open()" ];
>>>>>>>> +
>>>>>>>> +       Initialization -> Encoding [ label = "Both queues streaming" ];
>>>>>>>> +
>>>>>>>> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
>>>>>>>> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>>>> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
>>>>>>>> +       Encoding -> Encoding;
>>>>>>>> +
>>>>>>>> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>>>> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>>>> +
>>>>>>>> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
>>>>>>>> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
>>>>>>>> +
>>>>>>>> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
>>>>>>>> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
>>>>>>>> +   }
>>>>>>>> +
>>>>>>>> +Querying capabilities
>>>>>>>> +=====================
>>>>>>>> +
>>>>>>>> +1. To enumerate the set of coded formats supported by the encoder, the
>>>>>>>> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
>>>>>>>> +
>>>>>>>> +   * The full set of supported formats will be returned, regardless of the
>>>>>>>> +     format set on ``OUTPUT``.
>>>>>>>> +
>>>>>>>> +2. To enumerate the set of supported raw formats, the client may call
>>>>>>>> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
>>>>>>>> +
>>>>>>>> +   * Only the formats supported for the format currently active on ``CAPTURE``
>>>>>>>> +     will be returned.
>>>>>>>> +
>>>>>>>> +   * In order to enumerate raw formats supported by a given coded format,
>>>>>>>> +     the client must first set that coded format on ``CAPTURE`` and then
>>>>>>>> +     enumerate the formats on ``OUTPUT``.
>>>>>>>> +
>>>>>>>> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
>>>>>>>> +   resolutions for a given format, passing desired pixel format in
>>>>>>>> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
>>>>>>>> +
>>>>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
>>>>>>>> +     format will include all possible coded resolutions supported by the
>>>>>>>> +     encoder for given coded pixel format.
>>>>>>>> +
>>>>>>>> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
>>>>>>>> +     will include all possible frame buffer resolutions supported by the
>>>>>>>> +     encoder for given raw pixel format and coded format currently set on
>>>>>>>> +     ``CAPTURE``.
>>>>>>>> +
>>>>>>>> +4. Supported profiles and levels for the coded format currently set on
>>>>>>>> +   ``CAPTURE``, if applicable, may be queried using their respective controls
>>>>>>>> +   via :c:func:`VIDIOC_QUERYCTRL`.
>>>>>>>> +
>>>>>>>> +5. Any additional encoder capabilities may be discovered by querying
>>>>>>>> +   their respective controls.
>>>>>>>> +
>>>>>>>> +Initialization
>>>>>>>> +==============
>>>>>>>> +
>>>>>>>> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
>>>>>>>> +
>>>>>>>> +   * **Required fields:**
>>>>>>>> +
>>>>>>>> +     ``type``
>>>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
>>>>>>>> +
>>>>>>>> +     ``pixelformat``
>>>>>>>> +         the coded format to be produced
>>>>>>>> +
>>>>>>>> +     ``sizeimage``
>>>>>>>> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
>>>>>>>> +         match hardware requirements
>>>>>>>> +
>>>>>>>> +     ``width``, ``height``
>>>>>>>> +         ignored (always zero)
>>>>>>>> +
>>>>>>>> +     other fields
>>>>>>>> +         follow standard semantics
>>>>>>>> +
>>>>>>>> +   * **Return fields:**
>>>>>>>> +
>>>>>>>> +     ``sizeimage``
>>>>>>>> +         adjusted size of ``CAPTURE`` buffers
>>>>>>>> +
>>>>>>>> +   .. important::
>>>>>>>> +
>>>>>>>> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
>>>>>>>> +      format. The encoder will derive a new ``OUTPUT`` format from the
>>>>>>>> +      ``CAPTURE`` format being set, including resolution, colorimetry
>>>>>>>> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
>>>>>>>> +      must adjust it afterwards.
>>>>>>>
>>>>>>> Hmm, "including resolution": if width and height are set to 0, what should the
>>>>>>> OUTPUT resolution be? Up to the driver? I think this should be clarified since
>>>>>>> at a first reading of this paragraph it appears to be contradictory.
>>>>>>
>>>>>> I think the driver should just return the width and height of the OUTPUT
>>>>>> format. So the width and height that userspace specifies is just ignored
>>>>>> and replaced by the width and height of the OUTPUT format. After all, that's
>>>>>> what the bitstream will encode. Returning 0 for width and height would make
>>>>>> this a strange exception in V4L2 and I want to avoid that.
>>>>>>
>>>>>
>>>>> Hmm, however, the width and height of the OUTPUT format is not what's
>>>>> actually encoded in the bitstream. The right selection rectangle
>>>>> determines that.
>>>>>
>>>>> In one of the previous versions I though we could put the codec
>>>
>>> s/codec/coded/...
>>>
>>>>> resolution as the width and height of the CAPTURE format, which would
>>>>> be the resolution of the encoded image rounded up to full macroblocks
>>>>> +/- some encoder-specific constraints. AFAIR there was some concern
>>>>> about OUTPUT format changes triggering CAPTURE format changes, but to
>>>>> be honest, I'm not sure if that's really a problem. I just decided to
>>>>> drop that for the simplicity.
>>>>
>>>> I'm not sure what your point is.
>>>>
>>>> The OUTPUT format has the coded resolution,
>>>
>>> That's not always true. The OUTPUT format is just the format of the
>>> source frame buffers. In special cases where the source resolution is
>>> nicely aligned, it would be the same as coded size, but the remaining
>>> cases are valid as well.
>>>
>>>> so when you set the
>>>> CAPTURE format it can just copy the OUTPUT coded resolution unless the
>>>> chosen CAPTURE pixelformat can't handle that in which case both the
>>>> OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
>>>> or minimum the codec is capable of.
>>>
>>> As per my comment above, generally speaking, the encoder will derive
>>> an appropriate coded format from the OUTPUT format, but also other
>>> factors, like the crop rectangles and possibly some internal
>>> constraints.
>>>
>>>>
>>>> That said, I am fine with just leaving it up to the driver as suggested
>>>> before. Just as long as both the CAPTURE and OUTPUT formats remain valid
>>>> (i.e. width and height may never be out of range).
>>>>
>>>
>>> Sounds good to me.
>>>
>>>>>
>>>>>>>
>>>>>>>> +
>>>>>>>> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
>>>>>>>> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
>>>>>>>> +
>>>>>>>> +   * **Required fields:**
>>>>>>>> +
>>>>>>>> +     ``type``
>>>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>>>> +
>>>>>>>> +     other fields
>>>>>>>> +         follow standard semantics
>>>>>>>> +
>>>>>>>> +   * **Return fields:**
>>>>>>>> +
>>>>>>>> +     ``pixelformat``
>>>>>>>> +         raw format supported for the coded format currently selected on
>>>>>>>> +         the ``CAPTURE`` queue.
>>>>>>>> +
>>>>>>>> +     other fields
>>>>>>>> +         follow standard semantics
>>>>>>>> +
>>>>>>>> +3. Set the raw source format on the ``OUTPUT`` queue via
>>>>>>>> +   :c:func:`VIDIOC_S_FMT`.
>>>>>>>> +
>>>>>>>> +   * **Required fields:**
>>>>>>>> +
>>>>>>>> +     ``type``
>>>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>>>> +
>>>>>>>> +     ``pixelformat``
>>>>>>>> +         raw format of the source
>>>>>>>> +
>>>>>>>> +     ``width``, ``height``
>>>>>>>> +         source resolution
>>>>>>>> +
>>>>>>>> +     other fields
>>>>>>>> +         follow standard semantics
>>>>>>>> +
>>>>>>>> +   * **Return fields:**
>>>>>>>> +
>>>>>>>> +     ``width``, ``height``
>>>>>>>> +         may be adjusted by encoder to match alignment requirements, as
>>>>>>>> +         required by the currently selected formats
>>>>>>>
>>>>>>> What if the width x height is larger than the maximum supported by the
>>>>>>> selected coded format? This should probably mention that in that case the
>>>>>>> width x height is reduced to the largest allowed value. Also mention that
>>>>>>> this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
>>>>>>>
>>>>>>>> +
>>>>>>>> +     other fields
>>>>>>>> +         follow standard semantics
>>>>>>>> +
>>>>>>>> +   * Setting the source resolution will reset the selection rectangles to their
>>>>>>>> +     default values, based on the new resolution, as described in the step 5
>>>>>>>
>>>>>>> 5 -> 4
>>>>>>>
>>>>>>> Or just say: "as described in the next step."
>>>>>>>
>>>>>>>> +     below.
>>>>>>
>>>>>> It should also be made explicit that:
>>>>>>
>>>>>> 1) the crop rectangle will be set to the given width and height *before*
>>>>>> it is being adjusted by S_FMT.
>>>>>>
>>>>>
>>>>> I don't think that's what we want here.
>>>>>
>>>>> Defining the default rectangle to be exactly the same as the OUTPUT
>>>>> resolution (after the adjustment) makes the semantics consistent - not
>>>>> setting the crop rectangle gives you exactly the behavior as if there
>>>>> was no cropping involved (or supported by the encoder).
>>>>
>>>> I think you are right. This seems to be what the coda driver does as well.
>>>> It is convenient to be able to just set a 1920x1080 format and have that
>>>> resolution be stored as the crop rectangle, since it avoids having to call
>>>> s_selection afterwards, but it is not really consistent with the way V4L2
>>>> works.
>>>>
>>>>>
>>>>>> Open question: should we support a compose rectangle for the CAPTURE that
>>>>>> is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
>>>>>> the adjusted width and height and the compose rectangle (read-only) contains
>>>>>> the visible width and height. It's not strictly necessary, but it is
>>>>>> symmetrical.
>>>>>
>>>>> Wouldn't it rather be the CAPTURE crop rectangle that would be of the
>>>>> same resolution of the OUTPUT compose rectangle? Then you could
>>>>> actually have the CAPTURE compose rectangle for putting that into the
>>>>> desired rectangle of the encoded stream, if the encoder supports that.
>>>>> (I don't know any that does, so probably out of concern for now.)
>>>>
>>>> Yes, you are right.
>>>>
>>>> But should we support this?
>>>>
>>>> I actually think not for this initial version. It can be added later, I guess.
>>>>
>>>
>>> I think it boils down on whether adding it later wouldn't
>>> significantly complicate the application logic. It also relates to my
>>> other comment somewhere below.
>>>
>>>>>
>>>>>>
>>>>>> 2) the CAPTURE format will be updated as well with the new OUTPUT width and
>>>>>> height. The CAPTURE sizeimage might change as well.
>>>>>>
>>>>>>>> +
>>>>>>>> +4. **Optional.** Set the visible resolution for the stream metadata via
>>>>>>>> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
>>>>>>
>>>>>> I think you should mention that this is only necessary if the crop rectangle
>>>>>> that is set when you set the format isn't what you want.
>>>>>>
>>>>>
>>>>> Ack.
>>>>>
>>>>>>>> +
>>>>>>>> +   * **Required fields:**
>>>>>>>> +
>>>>>>>> +     ``type``
>>>>>>>> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
>>>>>>>> +
>>>>>>>> +     ``target``
>>>>>>>> +         set to ``V4L2_SEL_TGT_CROP``
>>>>>>>> +
>>>>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>>>>>> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
>>>>>>>> +         rectangle and may be subject to adjustment to match codec and
>>>>>>>> +         hardware constraints
>>>>>>>> +
>>>>>>>> +   * **Return fields:**
>>>>>>>> +
>>>>>>>> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
>>>>>>>> +         visible rectangle adjusted by the encoder
>>>>>>>> +
>>>>>>>> +   * The following selection targets are supported on ``OUTPUT``:
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>>>>>> +         equal to the full source frame, matching the active ``OUTPUT``
>>>>>>>> +         format
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>>>>>> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_CROP``
>>>>>>>> +         rectangle within the source buffer to be encoded into the
>>>>>>>> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
>>>>>>>> +
>>>>>>>> +         .. note::
>>>>>>>> +
>>>>>>>> +            A common use case for this selection target is encoding a source
>>>>>>>> +            video with a resolution that is not a multiple of a macroblock,
>>>>>>>> +            e.g.  the common 1920x1080 resolution may require the source
>>>>>>>> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
>>>>>>>> +            size. To avoid encoding the padding, the client needs to explicitly
>>>>>>>> +            configure this selection target to 1920x1080.
>>>>>>
>>>>>> This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
>>>>>>
>>>>>
>>>>> Sorry, which part exactly and what part of the proposal exactly? :)
>>>>> (My comment above might be related, though.)
>>>>
>>>> Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
>>>> crop selection target, so this note remains valid.
>>>>
>>>
>>> Ack.
>>>
>>>>>
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
>>>>>>>> +         maximum rectangle within the coded resolution, which the cropped
>>>>>>>> +         source frame can be composed into; if the hardware does not support
>>>>>>>> +         composition or scaling, then this is always equal to the rectangle of
>>>>>>>> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
>>>>>>>> +         equal to a rectangle of width and height matching
>>>>>>>> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
>>>>>>>> +
>>>>>>>> +     ``V4L2_SEL_TGT_COMPOSE``
>>>>>>>> +         rectangle within the coded frame, which the cropped source frame
>>>>>>>> +         is to be composed into; defaults to
>>>>>>>> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
>>>>>>>> +         additional compose/scaling capabilities; resulting stream will
>>>>>>>> +         have this rectangle encoded as the visible rectangle in its
>>>>>>>> +         metadata
>>>>>>
>>>>>> I think the compose targets for OUTPUT are only needed if the hardware can
>>>>>> actually do scaling and/or composition. Otherwise they can (must?) be
>>>>>> dropped.
>>>>>>
>>>>>
>>>>> Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
>>>>> userspace to learn the target visible rectangle that's going to be
>>>>> encoded in the stream metadata. If we omit it, we wouldn't have a way
>>>>> that would be consistent between encoders that can do
>>>>> scaling/composition and those that can't.
>>>>
>>>> I'm not convinced about this. The standard API behavior is not to expose
>>>> functionality that the hardware can't do. So if scaling isn't possible on
>>>> the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
>>>>
>>>> I also believe it very unlikely that we'll see encoders capable of scaling
>>>> as it doesn't make much sense.
>>>
>>> It does make a lot of sense - WebRTC requires 3 different sizes of the
>>> stream to be encoded at the same time. However, unfortunately, I
>>> haven't yet seen an encoder capable of doing so.
>>>
>>>> I would prefer to drop this to simplify the
>>>> spec, and when we get encoders that can scale, then we can add support for
>>>> compose rectangles (and I'm sure we'll need to think about how that
>>>> influences the CAPTURE side as well).
>>>>
>>>> For encoders without scaling it is the OUTPUT crop rectangle that defines
>>>> the visible rectangle.
>>>>
>>>>>
>>>>> However, with your proposal of actually having selection rectangles
>>>>> for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
>>>>> would expose a varying set of rectangles, depending on the hardware
>>>>> capability, while the CAPTURE queue would always expose its rectangle
>>>>> with that information.
>>>>
>>>> I think we should keep it simple and only define selection rectangles
>>>> when really needed.
>>>>
>>>> So encoders support CROP on the OUTPUT, and decoders support CAPTURE
>>>> COMPOSE (may be read-only). Nothing else.
>>>>
>>>> Once support for scaling is needed (either on the encoder or decoder
>>>> side), then the spec should be enhanced. But I prefer to postpone that
>>>> until we actually have hardware that needs this.
>>>>
>>>
>>> Okay, let's do it this way then. Actually, I don't even think there is
>>> much value in exposing information internal to the bitstream metadata
>>> like this, similarly to the coded size. My intention was to just
>>> ensure that we can easily add scaling/composing functionality later.
>>>
>>> I just removed the COMPOSE rectangles from my next draft.
>>
>> I don't think that supporting scaling will be a problem for the API as
>> such, since this is supported for standard video capture devices. It
>> just gets very complicated trying to describe how to configure all this.
>>
>> So I prefer to avoid this until we need to.
>>
>>>
>>> [snip]
>>>>>
>>>>>>
>>>>>> Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
>>>>>> or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
>>>>>> particular) and CAPTURE buffers were already allocated and are too small.
>>>>>
>>>>> The OUTPUT format must not change the CAPTURE format by definition.
>>>>> Otherwise we end up in a situation where we can't commit, because both
>>>>> queue formats can affect each other. Any change to the OUTPUT format
>>>>> that wouldn't work with the current CAPTURE format should be adjusted
>>>>> by the driver to match the current CAPTURE format.
>>>>
>>>> But the CAPTURE format *does* depend on the OUTPUT format: if the output
>>>> resolution changes, then so does the CAPTURE resolution and esp. the
>>>> sizeimage value, since that is typically resolution dependent.
>>>>
>>>> The coda driver does this as well: changing the output resolution
>>>> will update the capture resolution and sizeimage. The vicodec driver does the
>>>> same.
>>>>
>>>> Setting the CAPTURE format basically just selects the codec to use, after
>>>> that you can set the OUTPUT format and read the updated CAPTURE format to
>>>> get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
>>>> change the OUTPUT format, unless the OUTPUT format is incompatible with the
>>>> newly selected codec.
>>>
>>> Let me think about it for a while.
>>
>> Sleep on it, always works well for me :-)
> 
> Okay, I think I'm not convinced.
> 
> I believe we decided to allow sizeimage to be specified by the
> application, because it knows more about the stream it's going to
> encode. Only setting the size to 0 would make the encoder fall back to
> some simple internal heuristic.

Yes, that was the plan, but the patch stalled. I completely forgot
about this patch :-)

My last reply to "Re: [RFC PATCH] media/doc: Allow sizeimage to be set by
v4l clients" was March 14th.

Also, sizeimage must be at least the minimum size required for the given
CAPTURE width and height. So if it is less, then sizeimage will be set to that
minimum size.

> 
> Another thing is handling resolution changes. I believe that would
> have to be handled by stopping the OUTPUT queue, changing the OUTPUT
> format and starting the OUTPUT queue, all that without stopping the
> CAPTURE queue. With the behavior you described it wouldn't work,
> because the OUTPUT format couldn't be changed.
> 
> I'd suggest making OUTPUT format changes not change the CAPTURE sizeimage.

So OUTPUT format changes will still update the CAPTURE width and height?

It's kind of weird if you are encoding e.g. 1920x1080 but the CAPTURE format
says 1280x720. I'm not sure what is best.

What if the CAPTURE sizeimage is too small for the new OUTPUT resolution?
Should S_FMT(OUTPUT) fail with some error in that case?

Regards,

	Hans

> 
> Best regards,
> Tomasz
>

Nicolas Dufresne April 10, 2019, 4:05 p.m. UTC | #23

Le mercredi 10 avril 2019 à 10:50 +0200, Hans Verkuil a écrit :
> On 4/9/19 11:35 AM, Tomasz Figa wrote:
> > On Mon, Apr 8, 2019 at 8:11 PM Hans Verkuil <[e-mail]> wrote:
> > > On 4/8/19 11:23 AM, Tomasz Figa wrote:
> > > > On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
> > > > > On 4/5/19 10:12 AM, Tomasz Figa wrote:
> > > > > > On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
> > > > > > > Hi Tomasz,
> > > > > > > 
> > > > > > > Some more comments...
> > > > > > > 
> > > > > > > On 1/29/19 2:52 PM, Hans Verkuil wrote:
> > > > > > > > Hi Tomasz,
> > > > > > > > 
> > > > > > > > Some comments below. Nothing major, so I think a v4 should be ready to be
> > > > > > > > merged.
> > > > > > > > 
> > > > > > > > On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > > > > > > > > Due to complexity of the video encoding process, the V4L2 drivers of
> > > > > > > > > stateful encoder hardware require specific sequences of V4L2 API calls
> > > > > > > > > to be followed. These include capability enumeration, initialization,
> > > > > > > > > encoding, encode parameters change, drain and reset.
> > > > > > > > > 
> > > > > > > > > Specifics of the above have been discussed during Media Workshops at
> > > > > > > > > LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> > > > > > > > > Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> > > > > > > > > originated at those events was later implemented by the drivers we already
> > > > > > > > > have merged in mainline, such as s5p-mfc or coda.
> > > > > > > > > 
> > > > > > > > > The only thing missing was the real specification included as a part of
> > > > > > > > > Linux Media documentation. Fix it now and document the encoder part of
> > > > > > > > > the Codec API.
> > > > > > > > > 
> > > > > > > > > Signed-off-by: Tomasz Figa <[e-mail]>
> > > > > > > > > ---
> > > > > > > > >  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> > > > > > > > >  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> > > > > > > > >  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> > > > > > > > >  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> > > > > > > > >  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> > > > > > > > >  5 files changed, 617 insertions(+), 15 deletions(-)
> > > > > > > > >  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > 
> > > > > > > > > diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > new file mode 100644
> > > > > > > > > index 000000000000..fb8b05a132ee
> > > > > > > > > --- /dev/null
> > > > > > > > > +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > @@ -0,0 +1,586 @@
> > > > > > > > > +.. -*- coding: utf-8; mode: rst -*-
> > > > > > > > > +
> > > > > > > > > +.. _encoder:
> > > > > > > > > +
> > > > > > > > > +*************************************************
> > > > > > > > > +Memory-to-memory Stateful Video Encoder Interface
> > > > > > > > > +*************************************************
> > > > > > > > > +
> > > > > > > > > +A stateful video encoder takes raw video frames in display order and encodes
> > > > > > > > > +them into a bitstream. It generates complete chunks of the bitstream, including
> > > > > > > > > +all metadata, headers, etc. The resulting bitstream does not require any
> > > > > > > > > +further post-processing by the client.
> > > > > > > > > +
> > > > > > > > > +Performing software stream processing, header generation etc. in the driver
> > > > > > > > > +in order to support this interface is strongly discouraged. In case such
> > > > > > > > > +operations are needed, use of the Stateless Video Encoder Interface (in
> > > > > > > > > +development) is strongly advised.
> > > > > > > > > +
> > > > > > > > > +Conventions and notation used in this document
> > > > > > > > > +==============================================
> > > > > > > > > +
> > > > > > > > > +1. The general V4L2 API rules apply if not specified in this document
> > > > > > > > > +   otherwise.
> > > > > > > > > +
> > > > > > > > > +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> > > > > > > > > +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> > > > > > > > > +
> > > > > > > > > +3. All steps not marked "optional" are required.
> > > > > > > > > +
> > > > > > > > > +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> > > > > > > > > +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> > > > > > > > > +   unless specified otherwise.
> > > > > > > > > +
> > > > > > > > > +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> > > > > > > > > +   used interchangeably with multi-planar API, unless specified otherwise,
> > > > > > > > > +   depending on decoder capabilities and following the general V4L2 guidelines.
> > > > > > > > > +
> > > > > > > > > +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> > > > > > > > > +   [0..2]: i = 0, 1, 2.
> > > > > > > > > +
> > > > > > > > > +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> > > > > > > > > +   queue containing data that resulted from processing buffer A.
> > > > > > > > > +
> > > > > > > > > +Glossary
> > > > > > > > > +========
> > > > > > > > > +
> > > > > > > > > +Refer to :ref:`decoder-glossary`.
> > > > > > > > > +
> > > > > > > > > +State machine
> > > > > > > > > +=============
> > > > > > > > > +
> > > > > > > > > +.. kernel-render:: DOT
> > > > > > > > > +   :alt: DOT digraph of encoder state machine
> > > > > > > > > +   :caption: Encoder state machine
> > > > > > > > > +
> > > > > > > > > +   digraph encoder_state_machine {
> > > > > > > > > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > > > > > > > > +
> > > > > > > > > +       node [shape = circle, label="Initialization"] Initialization;
> > > > > > > > > +       node [shape = circle, label="Stopped"] Stopped;
> > > > > > > > > +       node [shape = circle, label="Drain"] Drain;
> > > > > > > > > +       node [shape = circle, label="Reset"] Reset;
> > > > > > > > > +
> > > > > > > > > +       node [shape = point]; qi
> > > > > > > > > +       qi -> Initialization [ label = "open()" ];
> > > > > > > > > +
> > > > > > > > > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > > > > > > > > +
> > > > > > > > > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > > > > > > > > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > > > > > > > > +       Encoding -> Encoding;
> > > > > > > > > +
> > > > > > > > > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > +
> > > > > > > > > +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> > > > > > > > > +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> > > > > > > > > +
> > > > > > > > > +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> > > > > > > > > +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > +   }
> > > > > > > > > +
> > > > > > > > > +Querying capabilities
> > > > > > > > > +=====================
> > > > > > > > > +
> > > > > > > > > +1. To enumerate the set of coded formats supported by the encoder, the
> > > > > > > > > +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> > > > > > > > > +
> > > > > > > > > +   * The full set of supported formats will be returned, regardless of the
> > > > > > > > > +     format set on ``OUTPUT``.
> > > > > > > > > +
> > > > > > > > > +2. To enumerate the set of supported raw formats, the client may call
> > > > > > > > > +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> > > > > > > > > +
> > > > > > > > > +   * Only the formats supported for the format currently active on ``CAPTURE``
> > > > > > > > > +     will be returned.
> > > > > > > > > +
> > > > > > > > > +   * In order to enumerate raw formats supported by a given coded format,
> > > > > > > > > +     the client must first set that coded format on ``CAPTURE`` and then
> > > > > > > > > +     enumerate the formats on ``OUTPUT``.
> > > > > > > > > +
> > > > > > > > > +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> > > > > > > > > +   resolutions for a given format, passing desired pixel format in
> > > > > > > > > +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> > > > > > > > > +
> > > > > > > > > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> > > > > > > > > +     format will include all possible coded resolutions supported by the
> > > > > > > > > +     encoder for given coded pixel format.
> > > > > > > > > +
> > > > > > > > > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> > > > > > > > > +     will include all possible frame buffer resolutions supported by the
> > > > > > > > > +     encoder for given raw pixel format and coded format currently set on
> > > > > > > > > +     ``CAPTURE``.
> > > > > > > > > +
> > > > > > > > > +4. Supported profiles and levels for the coded format currently set on
> > > > > > > > > +   ``CAPTURE``, if applicable, may be queried using their respective controls
> > > > > > > > > +   via :c:func:`VIDIOC_QUERYCTRL`.
> > > > > > > > > +
> > > > > > > > > +5. Any additional encoder capabilities may be discovered by querying
> > > > > > > > > +   their respective controls.
> > > > > > > > > +
> > > > > > > > > +Initialization
> > > > > > > > > +==============
> > > > > > > > > +
> > > > > > > > > +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> > > > > > > > > +
> > > > > > > > > +   * **Required fields:**
> > > > > > > > > +
> > > > > > > > > +     ``type``
> > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> > > > > > > > > +
> > > > > > > > > +     ``pixelformat``
> > > > > > > > > +         the coded format to be produced
> > > > > > > > > +
> > > > > > > > > +     ``sizeimage``
> > > > > > > > > +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> > > > > > > > > +         match hardware requirements
> > > > > > > > > +
> > > > > > > > > +     ``width``, ``height``
> > > > > > > > > +         ignored (always zero)
> > > > > > > > > +
> > > > > > > > > +     other fields
> > > > > > > > > +         follow standard semantics
> > > > > > > > > +
> > > > > > > > > +   * **Return fields:**
> > > > > > > > > +
> > > > > > > > > +     ``sizeimage``
> > > > > > > > > +         adjusted size of ``CAPTURE`` buffers
> > > > > > > > > +
> > > > > > > > > +   .. important::
> > > > > > > > > +
> > > > > > > > > +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> > > > > > > > > +      format. The encoder will derive a new ``OUTPUT`` format from the
> > > > > > > > > +      ``CAPTURE`` format being set, including resolution, colorimetry
> > > > > > > > > +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> > > > > > > > > +      must adjust it afterwards.
> > > > > > > > 
> > > > > > > > Hmm, "including resolution": if width and height are set to 0, what should the
> > > > > > > > OUTPUT resolution be? Up to the driver? I think this should be clarified since
> > > > > > > > at a first reading of this paragraph it appears to be contradictory.
> > > > > > > 
> > > > > > > I think the driver should just return the width and height of the OUTPUT
> > > > > > > format. So the width and height that userspace specifies is just ignored
> > > > > > > and replaced by the width and height of the OUTPUT format. After all, that's
> > > > > > > what the bitstream will encode. Returning 0 for width and height would make
> > > > > > > this a strange exception in V4L2 and I want to avoid that.
> > > > > > > 
> > > > > > 
> > > > > > Hmm, however, the width and height of the OUTPUT format is not what's
> > > > > > actually encoded in the bitstream. The right selection rectangle
> > > > > > determines that.
> > > > > > 
> > > > > > In one of the previous versions I though we could put the codec
> > > > 
> > > > s/codec/coded/...
> > > > 
> > > > > > resolution as the width and height of the CAPTURE format, which would
> > > > > > be the resolution of the encoded image rounded up to full macroblocks
> > > > > > +/- some encoder-specific constraints. AFAIR there was some concern
> > > > > > about OUTPUT format changes triggering CAPTURE format changes, but to
> > > > > > be honest, I'm not sure if that's really a problem. I just decided to
> > > > > > drop that for the simplicity.
> > > > > 
> > > > > I'm not sure what your point is.
> > > > > 
> > > > > The OUTPUT format has the coded resolution,
> > > > 
> > > > That's not always true. The OUTPUT format is just the format of the
> > > > source frame buffers. In special cases where the source resolution is
> > > > nicely aligned, it would be the same as coded size, but the remaining
> > > > cases are valid as well.
> > > > 
> > > > > so when you set the
> > > > > CAPTURE format it can just copy the OUTPUT coded resolution unless the
> > > > > chosen CAPTURE pixelformat can't handle that in which case both the
> > > > > OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
> > > > > or minimum the codec is capable of.
> > > > 
> > > > As per my comment above, generally speaking, the encoder will derive
> > > > an appropriate coded format from the OUTPUT format, but also other
> > > > factors, like the crop rectangles and possibly some internal
> > > > constraints.
> > > > 
> > > > > That said, I am fine with just leaving it up to the driver as suggested
> > > > > before. Just as long as both the CAPTURE and OUTPUT formats remain valid
> > > > > (i.e. width and height may never be out of range).
> > > > > 
> > > > 
> > > > Sounds good to me.
> > > > 
> > > > > > > > > +
> > > > > > > > > +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> > > > > > > > > +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> > > > > > > > > +
> > > > > > > > > +   * **Required fields:**
> > > > > > > > > +
> > > > > > > > > +     ``type``
> > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > +
> > > > > > > > > +     other fields
> > > > > > > > > +         follow standard semantics
> > > > > > > > > +
> > > > > > > > > +   * **Return fields:**
> > > > > > > > > +
> > > > > > > > > +     ``pixelformat``
> > > > > > > > > +         raw format supported for the coded format currently selected on
> > > > > > > > > +         the ``CAPTURE`` queue.
> > > > > > > > > +
> > > > > > > > > +     other fields
> > > > > > > > > +         follow standard semantics
> > > > > > > > > +
> > > > > > > > > +3. Set the raw source format on the ``OUTPUT`` queue via
> > > > > > > > > +   :c:func:`VIDIOC_S_FMT`.
> > > > > > > > > +
> > > > > > > > > +   * **Required fields:**
> > > > > > > > > +
> > > > > > > > > +     ``type``
> > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > +
> > > > > > > > > +     ``pixelformat``
> > > > > > > > > +         raw format of the source
> > > > > > > > > +
> > > > > > > > > +     ``width``, ``height``
> > > > > > > > > +         source resolution
> > > > > > > > > +
> > > > > > > > > +     other fields
> > > > > > > > > +         follow standard semantics
> > > > > > > > > +
> > > > > > > > > +   * **Return fields:**
> > > > > > > > > +
> > > > > > > > > +     ``width``, ``height``
> > > > > > > > > +         may be adjusted by encoder to match alignment requirements, as
> > > > > > > > > +         required by the currently selected formats
> > > > > > > > 
> > > > > > > > What if the width x height is larger than the maximum supported by the
> > > > > > > > selected coded format? This should probably mention that in that case the
> > > > > > > > width x height is reduced to the largest allowed value. Also mention that
> > > > > > > > this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> > > > > > > > 
> > > > > > > > > +
> > > > > > > > > +     other fields
> > > > > > > > > +         follow standard semantics
> > > > > > > > > +
> > > > > > > > > +   * Setting the source resolution will reset the selection rectangles to their
> > > > > > > > > +     default values, based on the new resolution, as described in the step 5
> > > > > > > > 
> > > > > > > > 5 -> 4
> > > > > > > > 
> > > > > > > > Or just say: "as described in the next step."
> > > > > > > > 
> > > > > > > > > +     below.
> > > > > > > 
> > > > > > > It should also be made explicit that:
> > > > > > > 
> > > > > > > 1) the crop rectangle will be set to the given width and height *before*
> > > > > > > it is being adjusted by S_FMT.
> > > > > > > 
> > > > > > 
> > > > > > I don't think that's what we want here.
> > > > > > 
> > > > > > Defining the default rectangle to be exactly the same as the OUTPUT
> > > > > > resolution (after the adjustment) makes the semantics consistent - not
> > > > > > setting the crop rectangle gives you exactly the behavior as if there
> > > > > > was no cropping involved (or supported by the encoder).
> > > > > 
> > > > > I think you are right. This seems to be what the coda driver does as well.
> > > > > It is convenient to be able to just set a 1920x1080 format and have that
> > > > > resolution be stored as the crop rectangle, since it avoids having to call
> > > > > s_selection afterwards, but it is not really consistent with the way V4L2
> > > > > works.
> > > > > 
> > > > > > > Open question: should we support a compose rectangle for the CAPTURE that
> > > > > > > is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
> > > > > > > the adjusted width and height and the compose rectangle (read-only) contains
> > > > > > > the visible width and height. It's not strictly necessary, but it is
> > > > > > > symmetrical.
> > > > > > 
> > > > > > Wouldn't it rather be the CAPTURE crop rectangle that would be of the
> > > > > > same resolution of the OUTPUT compose rectangle? Then you could
> > > > > > actually have the CAPTURE compose rectangle for putting that into the
> > > > > > desired rectangle of the encoded stream, if the encoder supports that.
> > > > > > (I don't know any that does, so probably out of concern for now.)
> > > > > 
> > > > > Yes, you are right.
> > > > > 
> > > > > But should we support this?
> > > > > 
> > > > > I actually think not for this initial version. It can be added later, I guess.
> > > > > 
> > > > 
> > > > I think it boils down on whether adding it later wouldn't
> > > > significantly complicate the application logic. It also relates to my
> > > > other comment somewhere below.
> > > > 
> > > > > > > 2) the CAPTURE format will be updated as well with the new OUTPUT width and
> > > > > > > height. The CAPTURE sizeimage might change as well.
> > > > > > > 
> > > > > > > > > +
> > > > > > > > > +4. **Optional.** Set the visible resolution for the stream metadata via
> > > > > > > > > +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> > > > > > > 
> > > > > > > I think you should mention that this is only necessary if the crop rectangle
> > > > > > > that is set when you set the format isn't what you want.
> > > > > > > 
> > > > > > 
> > > > > > Ack.
> > > > > > 
> > > > > > > > > +
> > > > > > > > > +   * **Required fields:**
> > > > > > > > > +
> > > > > > > > > +     ``type``
> > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > +
> > > > > > > > > +     ``target``
> > > > > > > > > +         set to ``V4L2_SEL_TGT_CROP``
> > > > > > > > > +
> > > > > > > > > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > > > > > > > > +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> > > > > > > > > +         rectangle and may be subject to adjustment to match codec and
> > > > > > > > > +         hardware constraints
> > > > > > > > > +
> > > > > > > > > +   * **Return fields:**
> > > > > > > > > +
> > > > > > > > > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > > > > > > > > +         visible rectangle adjusted by the encoder
> > > > > > > > > +
> > > > > > > > > +   * The following selection targets are supported on ``OUTPUT``:
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> > > > > > > > > +         equal to the full source frame, matching the active ``OUTPUT``
> > > > > > > > > +         format
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> > > > > > > > > +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_CROP``
> > > > > > > > > +         rectangle within the source buffer to be encoded into the
> > > > > > > > > +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> > > > > > > > > +
> > > > > > > > > +         .. note::
> > > > > > > > > +
> > > > > > > > > +            A common use case for this selection target is encoding a source
> > > > > > > > > +            video with a resolution that is not a multiple of a macroblock,
> > > > > > > > > +            e.g.  the common 1920x1080 resolution may require the source
> > > > > > > > > +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> > > > > > > > > +            size. To avoid encoding the padding, the client needs to explicitly
> > > > > > > > > +            configure this selection target to 1920x1080.
> > > > > > > 
> > > > > > > This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
> > > > > > > 
> > > > > > 
> > > > > > Sorry, which part exactly and what part of the proposal exactly? :)
> > > > > > (My comment above might be related, though.)
> > > > > 
> > > > > Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
> > > > > crop selection target, so this note remains valid.
> > > > > 
> > > > 
> > > > Ack.
> > > > 
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> > > > > > > > > +         maximum rectangle within the coded resolution, which the cropped
> > > > > > > > > +         source frame can be composed into; if the hardware does not support
> > > > > > > > > +         composition or scaling, then this is always equal to the rectangle of
> > > > > > > > > +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> > > > > > > > > +         equal to a rectangle of width and height matching
> > > > > > > > > +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > > > > > > > > +
> > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE``
> > > > > > > > > +         rectangle within the coded frame, which the cropped source frame
> > > > > > > > > +         is to be composed into; defaults to
> > > > > > > > > +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> > > > > > > > > +         additional compose/scaling capabilities; resulting stream will
> > > > > > > > > +         have this rectangle encoded as the visible rectangle in its
> > > > > > > > > +         metadata
> > > > > > > 
> > > > > > > I think the compose targets for OUTPUT are only needed if the hardware can
> > > > > > > actually do scaling and/or composition. Otherwise they can (must?) be
> > > > > > > dropped.
> > > > > > > 
> > > > > > 
> > > > > > Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
> > > > > > userspace to learn the target visible rectangle that's going to be
> > > > > > encoded in the stream metadata. If we omit it, we wouldn't have a way
> > > > > > that would be consistent between encoders that can do
> > > > > > scaling/composition and those that can't.
> > > > > 
> > > > > I'm not convinced about this. The standard API behavior is not to expose
> > > > > functionality that the hardware can't do. So if scaling isn't possible on
> > > > > the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
> > > > > 
> > > > > I also believe it very unlikely that we'll see encoders capable of scaling
> > > > > as it doesn't make much sense.
> > > > 
> > > > It does make a lot of sense - WebRTC requires 3 different sizes of the
> > > > stream to be encoded at the same time. However, unfortunately, I
> > > > haven't yet seen an encoder capable of doing so.
> > > > 
> > > > > I would prefer to drop this to simplify the
> > > > > spec, and when we get encoders that can scale, then we can add support for
> > > > > compose rectangles (and I'm sure we'll need to think about how that
> > > > > influences the CAPTURE side as well).
> > > > > 
> > > > > For encoders without scaling it is the OUTPUT crop rectangle that defines
> > > > > the visible rectangle.
> > > > > 
> > > > > > However, with your proposal of actually having selection rectangles
> > > > > > for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
> > > > > > would expose a varying set of rectangles, depending on the hardware
> > > > > > capability, while the CAPTURE queue would always expose its rectangle
> > > > > > with that information.
> > > > > 
> > > > > I think we should keep it simple and only define selection rectangles
> > > > > when really needed.
> > > > > 
> > > > > So encoders support CROP on the OUTPUT, and decoders support CAPTURE
> > > > > COMPOSE (may be read-only). Nothing else.
> > > > > 
> > > > > Once support for scaling is needed (either on the encoder or decoder
> > > > > side), then the spec should be enhanced. But I prefer to postpone that
> > > > > until we actually have hardware that needs this.
> > > > > 
> > > > 
> > > > Okay, let's do it this way then. Actually, I don't even think there is
> > > > much value in exposing information internal to the bitstream metadata
> > > > like this, similarly to the coded size. My intention was to just
> > > > ensure that we can easily add scaling/composing functionality later.
> > > > 
> > > > I just removed the COMPOSE rectangles from my next draft.
> > > 
> > > I don't think that supporting scaling will be a problem for the API as
> > > such, since this is supported for standard video capture devices. It
> > > just gets very complicated trying to describe how to configure all this.
> > > 
> > > So I prefer to avoid this until we need to.
> > > 
> > > > [snip]
> > > > > > > Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
> > > > > > > or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
> > > > > > > particular) and CAPTURE buffers were already allocated and are too small.
> > > > > > 
> > > > > > The OUTPUT format must not change the CAPTURE format by definition.
> > > > > > Otherwise we end up in a situation where we can't commit, because both
> > > > > > queue formats can affect each other. Any change to the OUTPUT format
> > > > > > that wouldn't work with the current CAPTURE format should be adjusted
> > > > > > by the driver to match the current CAPTURE format.
> > > > > 
> > > > > But the CAPTURE format *does* depend on the OUTPUT format: if the output
> > > > > resolution changes, then so does the CAPTURE resolution and esp. the
> > > > > sizeimage value, since that is typically resolution dependent.
> > > > > 
> > > > > The coda driver does this as well: changing the output resolution
> > > > > will update the capture resolution and sizeimage. The vicodec driver does the
> > > > > same.
> > > > > 
> > > > > Setting the CAPTURE format basically just selects the codec to use, after
> > > > > that you can set the OUTPUT format and read the updated CAPTURE format to
> > > > > get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
> > > > > change the OUTPUT format, unless the OUTPUT format is incompatible with the
> > > > > newly selected codec.
> > > > 
> > > > Let me think about it for a while.
> > > 
> > > Sleep on it, always works well for me :-)
> > 
> > Okay, I think I'm not convinced.
> > 
> > I believe we decided to allow sizeimage to be specified by the
> > application, because it knows more about the stream it's going to
> > encode. Only setting the size to 0 would make the encoder fall back to
> > some simple internal heuristic.
> 
> Yes, that was the plan, but the patch stalled. I completely forgot
> about this patch :-)
> 
> My last reply to "Re: [RFC PATCH] media/doc: Allow sizeimage to be set by
> v4l clients" was March 14th.
> 
> Also, sizeimage must be at least the minimum size required for the given
> CAPTURE width and height. So if it is less, then sizeimage will be set to that
> minimum size.
> 
> > Another thing is handling resolution changes. I believe that would
> > have to be handled by stopping the OUTPUT queue, changing the OUTPUT
> > format and starting the OUTPUT queue, all that without stopping the
> > CAPTURE queue. With the behavior you described it wouldn't work,
> > because the OUTPUT format couldn't be changed.
> > 
> > I'd suggest making OUTPUT format changes not change the CAPTURE sizeimage.
> 
> So OUTPUT format changes will still update the CAPTURE width and height?
> 
> It's kind of weird if you are encoding e.g. 1920x1080 but the CAPTURE format
> says 1280x720. I'm not sure what is best.
> 
> What if the CAPTURE sizeimage is too small for the new OUTPUT resolution?
> Should S_FMT(OUTPUT) fail with some error in that case?

Sounds like we need something similar to the SOURCE_CHANGE event
mechanism if we want to allow dynamic bitrate control which would
require re-allocation of the capture buffer queue. (Or any other
runtime control on our encoders, which is really expected to be
supported these days).

> 
> Regards,
> 
> 	Hans
> 
> > Best regards,
> > Tomasz
> >

Tomasz Figa April 15, 2019, 8:56 a.m. UTC | #24

On Thu, Apr 11, 2019 at 1:05 AM Nicolas Dufresne <[e-mail]> wrote:
>
> Le mercredi 10 avril 2019 à 10:50 +0200, Hans Verkuil a écrit :
> > On 4/9/19 11:35 AM, Tomasz Figa wrote:
> > > On Mon, Apr 8, 2019 at 8:11 PM Hans Verkuil <[e-mail]> wrote:
> > > > On 4/8/19 11:23 AM, Tomasz Figa wrote:
> > > > > On Fri, Apr 5, 2019 at 7:03 PM Hans Verkuil <[e-mail]> wrote:
> > > > > > On 4/5/19 10:12 AM, Tomasz Figa wrote:
> > > > > > > On Thu, Mar 14, 2019 at 10:57 PM Hans Verkuil <[e-mail]> wrote:
> > > > > > > > Hi Tomasz,
> > > > > > > >
> > > > > > > > Some more comments...
> > > > > > > >
> > > > > > > > On 1/29/19 2:52 PM, Hans Verkuil wrote:
> > > > > > > > > Hi Tomasz,
> > > > > > > > >
> > > > > > > > > Some comments below. Nothing major, so I think a v4 should be ready to be
> > > > > > > > > merged.
> > > > > > > > >
> > > > > > > > > On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > > > > > > > > > Due to complexity of the video encoding process, the V4L2 drivers of
> > > > > > > > > > stateful encoder hardware require specific sequences of V4L2 API calls
> > > > > > > > > > to be followed. These include capability enumeration, initialization,
> > > > > > > > > > encoding, encode parameters change, drain and reset.
> > > > > > > > > >
> > > > > > > > > > Specifics of the above have been discussed during Media Workshops at
> > > > > > > > > > LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> > > > > > > > > > Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> > > > > > > > > > originated at those events was later implemented by the drivers we already
> > > > > > > > > > have merged in mainline, such as s5p-mfc or coda.
> > > > > > > > > >
> > > > > > > > > > The only thing missing was the real specification included as a part of
> > > > > > > > > > Linux Media documentation. Fix it now and document the encoder part of
> > > > > > > > > > the Codec API.
> > > > > > > > > >
> > > > > > > > > > Signed-off-by: Tomasz Figa <[e-mail]>
> > > > > > > > > > ---
> > > > > > > > > >  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
> > > > > > > > > >  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
> > > > > > > > > >  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
> > > > > > > > > >  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
> > > > > > > > > >  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
> > > > > > > > > >  5 files changed, 617 insertions(+), 15 deletions(-)
> > > > > > > > > >  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > >
> > > > > > > > > > diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > > new file mode 100644
> > > > > > > > > > index 000000000000..fb8b05a132ee
> > > > > > > > > > --- /dev/null
> > > > > > > > > > +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> > > > > > > > > > @@ -0,0 +1,586 @@
> > > > > > > > > > +.. -*- coding: utf-8; mode: rst -*-
> > > > > > > > > > +
> > > > > > > > > > +.. _encoder:
> > > > > > > > > > +
> > > > > > > > > > +*************************************************
> > > > > > > > > > +Memory-to-memory Stateful Video Encoder Interface
> > > > > > > > > > +*************************************************
> > > > > > > > > > +
> > > > > > > > > > +A stateful video encoder takes raw video frames in display order and encodes
> > > > > > > > > > +them into a bitstream. It generates complete chunks of the bitstream, including
> > > > > > > > > > +all metadata, headers, etc. The resulting bitstream does not require any
> > > > > > > > > > +further post-processing by the client.
> > > > > > > > > > +
> > > > > > > > > > +Performing software stream processing, header generation etc. in the driver
> > > > > > > > > > +in order to support this interface is strongly discouraged. In case such
> > > > > > > > > > +operations are needed, use of the Stateless Video Encoder Interface (in
> > > > > > > > > > +development) is strongly advised.
> > > > > > > > > > +
> > > > > > > > > > +Conventions and notation used in this document
> > > > > > > > > > +==============================================
> > > > > > > > > > +
> > > > > > > > > > +1. The general V4L2 API rules apply if not specified in this document
> > > > > > > > > > +   otherwise.
> > > > > > > > > > +
> > > > > > > > > > +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> > > > > > > > > > +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> > > > > > > > > > +
> > > > > > > > > > +3. All steps not marked "optional" are required.
> > > > > > > > > > +
> > > > > > > > > > +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> > > > > > > > > > +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> > > > > > > > > > +   unless specified otherwise.
> > > > > > > > > > +
> > > > > > > > > > +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> > > > > > > > > > +   used interchangeably with multi-planar API, unless specified otherwise,
> > > > > > > > > > +   depending on decoder capabilities and following the general V4L2 guidelines.
> > > > > > > > > > +
> > > > > > > > > > +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> > > > > > > > > > +   [0..2]: i = 0, 1, 2.
> > > > > > > > > > +
> > > > > > > > > > +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> > > > > > > > > > +   queue containing data that resulted from processing buffer A.
> > > > > > > > > > +
> > > > > > > > > > +Glossary
> > > > > > > > > > +========
> > > > > > > > > > +
> > > > > > > > > > +Refer to :ref:`decoder-glossary`.
> > > > > > > > > > +
> > > > > > > > > > +State machine
> > > > > > > > > > +=============
> > > > > > > > > > +
> > > > > > > > > > +.. kernel-render:: DOT
> > > > > > > > > > +   :alt: DOT digraph of encoder state machine
> > > > > > > > > > +   :caption: Encoder state machine
> > > > > > > > > > +
> > > > > > > > > > +   digraph encoder_state_machine {
> > > > > > > > > > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > > > > > > > > > +
> > > > > > > > > > +       node [shape = circle, label="Initialization"] Initialization;
> > > > > > > > > > +       node [shape = circle, label="Stopped"] Stopped;
> > > > > > > > > > +       node [shape = circle, label="Drain"] Drain;
> > > > > > > > > > +       node [shape = circle, label="Reset"] Reset;
> > > > > > > > > > +
> > > > > > > > > > +       node [shape = point]; qi
> > > > > > > > > > +       qi -> Initialization [ label = "open()" ];
> > > > > > > > > > +
> > > > > > > > > > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > > > > > > > > > +
> > > > > > > > > > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > > > > > > > > > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > > > > > > > > > +       Encoding -> Encoding;
> > > > > > > > > > +
> > > > > > > > > > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > > +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > > +
> > > > > > > > > > +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> > > > > > > > > > +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> > > > > > > > > > +
> > > > > > > > > > +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> > > > > > > > > > +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > > > > > > > +   }
> > > > > > > > > > +
> > > > > > > > > > +Querying capabilities
> > > > > > > > > > +=====================
> > > > > > > > > > +
> > > > > > > > > > +1. To enumerate the set of coded formats supported by the encoder, the
> > > > > > > > > > +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> > > > > > > > > > +
> > > > > > > > > > +   * The full set of supported formats will be returned, regardless of the
> > > > > > > > > > +     format set on ``OUTPUT``.
> > > > > > > > > > +
> > > > > > > > > > +2. To enumerate the set of supported raw formats, the client may call
> > > > > > > > > > +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> > > > > > > > > > +
> > > > > > > > > > +   * Only the formats supported for the format currently active on ``CAPTURE``
> > > > > > > > > > +     will be returned.
> > > > > > > > > > +
> > > > > > > > > > +   * In order to enumerate raw formats supported by a given coded format,
> > > > > > > > > > +     the client must first set that coded format on ``CAPTURE`` and then
> > > > > > > > > > +     enumerate the formats on ``OUTPUT``.
> > > > > > > > > > +
> > > > > > > > > > +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> > > > > > > > > > +   resolutions for a given format, passing desired pixel format in
> > > > > > > > > > +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> > > > > > > > > > +
> > > > > > > > > > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> > > > > > > > > > +     format will include all possible coded resolutions supported by the
> > > > > > > > > > +     encoder for given coded pixel format.
> > > > > > > > > > +
> > > > > > > > > > +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> > > > > > > > > > +     will include all possible frame buffer resolutions supported by the
> > > > > > > > > > +     encoder for given raw pixel format and coded format currently set on
> > > > > > > > > > +     ``CAPTURE``.
> > > > > > > > > > +
> > > > > > > > > > +4. Supported profiles and levels for the coded format currently set on
> > > > > > > > > > +   ``CAPTURE``, if applicable, may be queried using their respective controls
> > > > > > > > > > +   via :c:func:`VIDIOC_QUERYCTRL`.
> > > > > > > > > > +
> > > > > > > > > > +5. Any additional encoder capabilities may be discovered by querying
> > > > > > > > > > +   their respective controls.
> > > > > > > > > > +
> > > > > > > > > > +Initialization
> > > > > > > > > > +==============
> > > > > > > > > > +
> > > > > > > > > > +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> > > > > > > > > > +
> > > > > > > > > > +   * **Required fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``type``
> > > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> > > > > > > > > > +
> > > > > > > > > > +     ``pixelformat``
> > > > > > > > > > +         the coded format to be produced
> > > > > > > > > > +
> > > > > > > > > > +     ``sizeimage``
> > > > > > > > > > +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> > > > > > > > > > +         match hardware requirements
> > > > > > > > > > +
> > > > > > > > > > +     ``width``, ``height``
> > > > > > > > > > +         ignored (always zero)
> > > > > > > > > > +
> > > > > > > > > > +     other fields
> > > > > > > > > > +         follow standard semantics
> > > > > > > > > > +
> > > > > > > > > > +   * **Return fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``sizeimage``
> > > > > > > > > > +         adjusted size of ``CAPTURE`` buffers
> > > > > > > > > > +
> > > > > > > > > > +   .. important::
> > > > > > > > > > +
> > > > > > > > > > +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> > > > > > > > > > +      format. The encoder will derive a new ``OUTPUT`` format from the
> > > > > > > > > > +      ``CAPTURE`` format being set, including resolution, colorimetry
> > > > > > > > > > +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> > > > > > > > > > +      must adjust it afterwards.
> > > > > > > > >
> > > > > > > > > Hmm, "including resolution": if width and height are set to 0, what should the
> > > > > > > > > OUTPUT resolution be? Up to the driver? I think this should be clarified since
> > > > > > > > > at a first reading of this paragraph it appears to be contradictory.
> > > > > > > >
> > > > > > > > I think the driver should just return the width and height of the OUTPUT
> > > > > > > > format. So the width and height that userspace specifies is just ignored
> > > > > > > > and replaced by the width and height of the OUTPUT format. After all, that's
> > > > > > > > what the bitstream will encode. Returning 0 for width and height would make
> > > > > > > > this a strange exception in V4L2 and I want to avoid that.
> > > > > > > >
> > > > > > >
> > > > > > > Hmm, however, the width and height of the OUTPUT format is not what's
> > > > > > > actually encoded in the bitstream. The right selection rectangle
> > > > > > > determines that.
> > > > > > >
> > > > > > > In one of the previous versions I though we could put the codec
> > > > >
> > > > > s/codec/coded/...
> > > > >
> > > > > > > resolution as the width and height of the CAPTURE format, which would
> > > > > > > be the resolution of the encoded image rounded up to full macroblocks
> > > > > > > +/- some encoder-specific constraints. AFAIR there was some concern
> > > > > > > about OUTPUT format changes triggering CAPTURE format changes, but to
> > > > > > > be honest, I'm not sure if that's really a problem. I just decided to
> > > > > > > drop that for the simplicity.
> > > > > >
> > > > > > I'm not sure what your point is.
> > > > > >
> > > > > > The OUTPUT format has the coded resolution,
> > > > >
> > > > > That's not always true. The OUTPUT format is just the format of the
> > > > > source frame buffers. In special cases where the source resolution is
> > > > > nicely aligned, it would be the same as coded size, but the remaining
> > > > > cases are valid as well.
> > > > >
> > > > > > so when you set the
> > > > > > CAPTURE format it can just copy the OUTPUT coded resolution unless the
> > > > > > chosen CAPTURE pixelformat can't handle that in which case both the
> > > > > > OUTPUT and CAPTURE coded resolutions are clamped to whatever is the maximum
> > > > > > or minimum the codec is capable of.
> > > > >
> > > > > As per my comment above, generally speaking, the encoder will derive
> > > > > an appropriate coded format from the OUTPUT format, but also other
> > > > > factors, like the crop rectangles and possibly some internal
> > > > > constraints.
> > > > >
> > > > > > That said, I am fine with just leaving it up to the driver as suggested
> > > > > > before. Just as long as both the CAPTURE and OUTPUT formats remain valid
> > > > > > (i.e. width and height may never be out of range).
> > > > > >
> > > > >
> > > > > Sounds good to me.
> > > > >
> > > > > > > > > > +
> > > > > > > > > > +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> > > > > > > > > > +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> > > > > > > > > > +
> > > > > > > > > > +   * **Required fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``type``
> > > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > > +
> > > > > > > > > > +     other fields
> > > > > > > > > > +         follow standard semantics
> > > > > > > > > > +
> > > > > > > > > > +   * **Return fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``pixelformat``
> > > > > > > > > > +         raw format supported for the coded format currently selected on
> > > > > > > > > > +         the ``CAPTURE`` queue.
> > > > > > > > > > +
> > > > > > > > > > +     other fields
> > > > > > > > > > +         follow standard semantics
> > > > > > > > > > +
> > > > > > > > > > +3. Set the raw source format on the ``OUTPUT`` queue via
> > > > > > > > > > +   :c:func:`VIDIOC_S_FMT`.
> > > > > > > > > > +
> > > > > > > > > > +   * **Required fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``type``
> > > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > > +
> > > > > > > > > > +     ``pixelformat``
> > > > > > > > > > +         raw format of the source
> > > > > > > > > > +
> > > > > > > > > > +     ``width``, ``height``
> > > > > > > > > > +         source resolution
> > > > > > > > > > +
> > > > > > > > > > +     other fields
> > > > > > > > > > +         follow standard semantics
> > > > > > > > > > +
> > > > > > > > > > +   * **Return fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``width``, ``height``
> > > > > > > > > > +         may be adjusted by encoder to match alignment requirements, as
> > > > > > > > > > +         required by the currently selected formats
> > > > > > > > >
> > > > > > > > > What if the width x height is larger than the maximum supported by the
> > > > > > > > > selected coded format? This should probably mention that in that case the
> > > > > > > > > width x height is reduced to the largest allowed value. Also mention that
> > > > > > > > > this maximum is reported by VIDIOC_ENUM_FRAMESIZES.
> > > > > > > > >
> > > > > > > > > > +
> > > > > > > > > > +     other fields
> > > > > > > > > > +         follow standard semantics
> > > > > > > > > > +
> > > > > > > > > > +   * Setting the source resolution will reset the selection rectangles to their
> > > > > > > > > > +     default values, based on the new resolution, as described in the step 5
> > > > > > > > >
> > > > > > > > > 5 -> 4
> > > > > > > > >
> > > > > > > > > Or just say: "as described in the next step."
> > > > > > > > >
> > > > > > > > > > +     below.
> > > > > > > >
> > > > > > > > It should also be made explicit that:
> > > > > > > >
> > > > > > > > 1) the crop rectangle will be set to the given width and height *before*
> > > > > > > > it is being adjusted by S_FMT.
> > > > > > > >
> > > > > > >
> > > > > > > I don't think that's what we want here.
> > > > > > >
> > > > > > > Defining the default rectangle to be exactly the same as the OUTPUT
> > > > > > > resolution (after the adjustment) makes the semantics consistent - not
> > > > > > > setting the crop rectangle gives you exactly the behavior as if there
> > > > > > > was no cropping involved (or supported by the encoder).
> > > > > >
> > > > > > I think you are right. This seems to be what the coda driver does as well.
> > > > > > It is convenient to be able to just set a 1920x1080 format and have that
> > > > > > resolution be stored as the crop rectangle, since it avoids having to call
> > > > > > s_selection afterwards, but it is not really consistent with the way V4L2
> > > > > > works.
> > > > > >
> > > > > > > > Open question: should we support a compose rectangle for the CAPTURE that
> > > > > > > > is the same as the OUTPUT crop rectangle? I.e. the CAPTURE format contains
> > > > > > > > the adjusted width and height and the compose rectangle (read-only) contains
> > > > > > > > the visible width and height. It's not strictly necessary, but it is
> > > > > > > > symmetrical.
> > > > > > >
> > > > > > > Wouldn't it rather be the CAPTURE crop rectangle that would be of the
> > > > > > > same resolution of the OUTPUT compose rectangle? Then you could
> > > > > > > actually have the CAPTURE compose rectangle for putting that into the
> > > > > > > desired rectangle of the encoded stream, if the encoder supports that.
> > > > > > > (I don't know any that does, so probably out of concern for now.)
> > > > > >
> > > > > > Yes, you are right.
> > > > > >
> > > > > > But should we support this?
> > > > > >
> > > > > > I actually think not for this initial version. It can be added later, I guess.
> > > > > >
> > > > >
> > > > > I think it boils down on whether adding it later wouldn't
> > > > > significantly complicate the application logic. It also relates to my
> > > > > other comment somewhere below.
> > > > >
> > > > > > > > 2) the CAPTURE format will be updated as well with the new OUTPUT width and
> > > > > > > > height. The CAPTURE sizeimage might change as well.
> > > > > > > >
> > > > > > > > > > +
> > > > > > > > > > +4. **Optional.** Set the visible resolution for the stream metadata via
> > > > > > > > > > +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> > > > > > > >
> > > > > > > > I think you should mention that this is only necessary if the crop rectangle
> > > > > > > > that is set when you set the format isn't what you want.
> > > > > > > >
> > > > > > >
> > > > > > > Ack.
> > > > > > >
> > > > > > > > > > +
> > > > > > > > > > +   * **Required fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``type``
> > > > > > > > > > +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> > > > > > > > > > +
> > > > > > > > > > +     ``target``
> > > > > > > > > > +         set to ``V4L2_SEL_TGT_CROP``
> > > > > > > > > > +
> > > > > > > > > > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > > > > > > > > > +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> > > > > > > > > > +         rectangle and may be subject to adjustment to match codec and
> > > > > > > > > > +         hardware constraints
> > > > > > > > > > +
> > > > > > > > > > +   * **Return fields:**
> > > > > > > > > > +
> > > > > > > > > > +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> > > > > > > > > > +         visible rectangle adjusted by the encoder
> > > > > > > > > > +
> > > > > > > > > > +   * The following selection targets are supported on ``OUTPUT``:
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> > > > > > > > > > +         equal to the full source frame, matching the active ``OUTPUT``
> > > > > > > > > > +         format
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> > > > > > > > > > +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_CROP``
> > > > > > > > > > +         rectangle within the source buffer to be encoded into the
> > > > > > > > > > +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> > > > > > > > > > +
> > > > > > > > > > +         .. note::
> > > > > > > > > > +
> > > > > > > > > > +            A common use case for this selection target is encoding a source
> > > > > > > > > > +            video with a resolution that is not a multiple of a macroblock,
> > > > > > > > > > +            e.g.  the common 1920x1080 resolution may require the source
> > > > > > > > > > +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> > > > > > > > > > +            size. To avoid encoding the padding, the client needs to explicitly
> > > > > > > > > > +            configure this selection target to 1920x1080.
> > > > > > > >
> > > > > > > > This last sentence contradicts the proposed behavior of S_FMT(OUTPUT).
> > > > > > > >
> > > > > > >
> > > > > > > Sorry, which part exactly and what part of the proposal exactly? :)
> > > > > > > (My comment above might be related, though.)
> > > > > >
> > > > > > Ignore my comment. We go back to explicitly requiring userspace to set the OUTPUT
> > > > > > crop selection target, so this note remains valid.
> > > > > >
> > > > >
> > > > > Ack.
> > > > >
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> > > > > > > > > > +         maximum rectangle within the coded resolution, which the cropped
> > > > > > > > > > +         source frame can be composed into; if the hardware does not support
> > > > > > > > > > +         composition or scaling, then this is always equal to the rectangle of
> > > > > > > > > > +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> > > > > > > > > > +         equal to a rectangle of width and height matching
> > > > > > > > > > +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> > > > > > > > > > +
> > > > > > > > > > +     ``V4L2_SEL_TGT_COMPOSE``
> > > > > > > > > > +         rectangle within the coded frame, which the cropped source frame
> > > > > > > > > > +         is to be composed into; defaults to
> > > > > > > > > > +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> > > > > > > > > > +         additional compose/scaling capabilities; resulting stream will
> > > > > > > > > > +         have this rectangle encoded as the visible rectangle in its
> > > > > > > > > > +         metadata
> > > > > > > >
> > > > > > > > I think the compose targets for OUTPUT are only needed if the hardware can
> > > > > > > > actually do scaling and/or composition. Otherwise they can (must?) be
> > > > > > > > dropped.
> > > > > > > >
> > > > > > >
> > > > > > > Note that V4L2_SEL_TGT_COMPOSE is defined to be the way for the
> > > > > > > userspace to learn the target visible rectangle that's going to be
> > > > > > > encoded in the stream metadata. If we omit it, we wouldn't have a way
> > > > > > > that would be consistent between encoders that can do
> > > > > > > scaling/composition and those that can't.
> > > > > >
> > > > > > I'm not convinced about this. The standard API behavior is not to expose
> > > > > > functionality that the hardware can't do. So if scaling isn't possible on
> > > > > > the OUTPUT side, then it shouldn't expose OUTPUT compose rectangles.
> > > > > >
> > > > > > I also believe it very unlikely that we'll see encoders capable of scaling
> > > > > > as it doesn't make much sense.
> > > > >
> > > > > It does make a lot of sense - WebRTC requires 3 different sizes of the
> > > > > stream to be encoded at the same time. However, unfortunately, I
> > > > > haven't yet seen an encoder capable of doing so.
> > > > >
> > > > > > I would prefer to drop this to simplify the
> > > > > > spec, and when we get encoders that can scale, then we can add support for
> > > > > > compose rectangles (and I'm sure we'll need to think about how that
> > > > > > influences the CAPTURE side as well).
> > > > > >
> > > > > > For encoders without scaling it is the OUTPUT crop rectangle that defines
> > > > > > the visible rectangle.
> > > > > >
> > > > > > > However, with your proposal of actually having selection rectangles
> > > > > > > for the CAPTURE queue, it could be solved indeed. The OUTPUT queue
> > > > > > > would expose a varying set of rectangles, depending on the hardware
> > > > > > > capability, while the CAPTURE queue would always expose its rectangle
> > > > > > > with that information.
> > > > > >
> > > > > > I think we should keep it simple and only define selection rectangles
> > > > > > when really needed.
> > > > > >
> > > > > > So encoders support CROP on the OUTPUT, and decoders support CAPTURE
> > > > > > COMPOSE (may be read-only). Nothing else.
> > > > > >
> > > > > > Once support for scaling is needed (either on the encoder or decoder
> > > > > > side), then the spec should be enhanced. But I prefer to postpone that
> > > > > > until we actually have hardware that needs this.
> > > > > >
> > > > >
> > > > > Okay, let's do it this way then. Actually, I don't even think there is
> > > > > much value in exposing information internal to the bitstream metadata
> > > > > like this, similarly to the coded size. My intention was to just
> > > > > ensure that we can easily add scaling/composing functionality later.
> > > > >
> > > > > I just removed the COMPOSE rectangles from my next draft.
> > > >
> > > > I don't think that supporting scaling will be a problem for the API as
> > > > such, since this is supported for standard video capture devices. It
> > > > just gets very complicated trying to describe how to configure all this.
> > > >
> > > > So I prefer to avoid this until we need to.
> > > >
> > > > > [snip]
> > > > > > > > Changing the OUTPUT format will always fail if OUTPUT buffers are already allocated,
> > > > > > > > or if changing the OUTPUT format would change the CAPTURE format (sizeimage in
> > > > > > > > particular) and CAPTURE buffers were already allocated and are too small.
> > > > > > >
> > > > > > > The OUTPUT format must not change the CAPTURE format by definition.
> > > > > > > Otherwise we end up in a situation where we can't commit, because both
> > > > > > > queue formats can affect each other. Any change to the OUTPUT format
> > > > > > > that wouldn't work with the current CAPTURE format should be adjusted
> > > > > > > by the driver to match the current CAPTURE format.
> > > > > >
> > > > > > But the CAPTURE format *does* depend on the OUTPUT format: if the output
> > > > > > resolution changes, then so does the CAPTURE resolution and esp. the
> > > > > > sizeimage value, since that is typically resolution dependent.
> > > > > >
> > > > > > The coda driver does this as well: changing the output resolution
> > > > > > will update the capture resolution and sizeimage. The vicodec driver does the
> > > > > > same.
> > > > > >
> > > > > > Setting the CAPTURE format basically just selects the codec to use, after
> > > > > > that you can set the OUTPUT format and read the updated CAPTURE format to
> > > > > > get the new sizeimage value. In fact, setting the CAPTURE format shouldn't
> > > > > > change the OUTPUT format, unless the OUTPUT format is incompatible with the
> > > > > > newly selected codec.
> > > > >
> > > > > Let me think about it for a while.
> > > >
> > > > Sleep on it, always works well for me :-)
> > >
> > > Okay, I think I'm not convinced.
> > >
> > > I believe we decided to allow sizeimage to be specified by the
> > > application, because it knows more about the stream it's going to
> > > encode. Only setting the size to 0 would make the encoder fall back to
> > > some simple internal heuristic.
> >
> > Yes, that was the plan, but the patch stalled. I completely forgot
> > about this patch :-)
> >
> > My last reply to "Re: [RFC PATCH] media/doc: Allow sizeimage to be set by
> > v4l clients" was March 14th.
> >
> > Also, sizeimage must be at least the minimum size required for the given
> > CAPTURE width and height. So if it is less, then sizeimage will be set to that
> > minimum size.
> >
> > > Another thing is handling resolution changes. I believe that would
> > > have to be handled by stopping the OUTPUT queue, changing the OUTPUT
> > > format and starting the OUTPUT queue, all that without stopping the
> > > CAPTURE queue. With the behavior you described it wouldn't work,
> > > because the OUTPUT format couldn't be changed.
> > >
> > > I'd suggest making OUTPUT format changes not change the CAPTURE sizeimage.
> >
> > So OUTPUT format changes will still update the CAPTURE width and height?
> >
> > It's kind of weird if you are encoding e.g. 1920x1080 but the CAPTURE format
> > says 1280x720. I'm not sure what is best.
> >
> > What if the CAPTURE sizeimage is too small for the new OUTPUT resolution?
> > Should S_FMT(OUTPUT) fail with some error in that case?
>
> Sounds like we need something similar to the SOURCE_CHANGE event
> mechanism if we want to allow dynamic bitrate control which would
> require re-allocation of the capture buffer queue. (Or any other
> runtime control on our encoders, which is really expected to be
> supported these days).

Sounds like it. Or we could just assume that one needs to stop both
queues to do a resolution change, since most codes would anyway reset
the stream (e.g. send PPS/SPS, etc. for H.264) to change the
resolution. Not sure if that assumption always holds, though.

Best regards,
Tomasz

Nicolas Dufresne April 15, 2019, 12:30 p.m. UTC | #25

Le lundi 15 avril 2019 à 17:56 +0900, Tomasz Figa a écrit :
> > Sounds like we need something similar to the SOURCE_CHANGE event
> > mechanism if we want to allow dynamic bitrate control which would
> > require re-allocation of the capture buffer queue. (Or any other
> > runtime control on our encoders, which is really expected to be
> > supported these days).
> 
> Sounds like it. Or we could just assume that one needs to stop both
> queues to do a resolution change, since most codes would anyway reset
> the stream (e.g. send PPS/SPS, etc. for H.264) to change the
> resolution. Not sure if that assumption always holds, though.

I think for resolution/profile/level changes you have a good point, as
you said, we need to start a new stream (new header, new IDR). Maybe
then we should simply require the driver to allocate enough buffer to
support the highest bitrate for the selected resoltion/profile/level?
Though, this could create situation where we waste a lot of memory. On
the other side, if you need to reallocate your buffers on bitrate
change, it might create a visible freeze.

Nicolas

Michael Tretter April 30, 2019, 5:34 p.m. UTC | #26

On Thu, 24 Jan 2019 19:04:19 +0900, Tomasz Figa wrote:
> Due to complexity of the video encoding process, the V4L2 drivers of
> stateful encoder hardware require specific sequences of V4L2 API calls
> to be followed. These include capability enumeration, initialization,
> encoding, encode parameters change, drain and reset.
> 
> Specifics of the above have been discussed during Media Workshops at
> LinuxCon Europe 2012 in Barcelona and then later Embedded Linux
> Conference Europe 2014 in Düsseldorf. The de facto Codec API that
> originated at those events was later implemented by the drivers we already
> have merged in mainline, such as s5p-mfc or coda.
> 
> The only thing missing was the real specification included as a part of
> Linux Media documentation. Fix it now and document the encoder part of
> the Codec API.
> 
> Signed-off-by: Tomasz Figa <[e-mail]>
> ---
>  Documentation/media/uapi/v4l/dev-encoder.rst  | 586 ++++++++++++++++++
>  Documentation/media/uapi/v4l/dev-mem2mem.rst  |   1 +
>  Documentation/media/uapi/v4l/pixfmt-v4l2.rst  |   5 +
>  Documentation/media/uapi/v4l/v4l2.rst         |   2 +
>  .../media/uapi/v4l/vidioc-encoder-cmd.rst     |  38 +-
>  5 files changed, 617 insertions(+), 15 deletions(-)
>  create mode 100644 Documentation/media/uapi/v4l/dev-encoder.rst
> 
> diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
> new file mode 100644
> index 000000000000..fb8b05a132ee
> --- /dev/null
> +++ b/Documentation/media/uapi/v4l/dev-encoder.rst
> @@ -0,0 +1,586 @@
> +.. -*- coding: utf-8; mode: rst -*-
> +
> +.. _encoder:
> +
> +*************************************************
> +Memory-to-memory Stateful Video Encoder Interface
> +*************************************************
> +
> +A stateful video encoder takes raw video frames in display order and encodes
> +them into a bitstream. It generates complete chunks of the bitstream, including
> +all metadata, headers, etc. The resulting bitstream does not require any
> +further post-processing by the client.
> +
> +Performing software stream processing, header generation etc. in the driver
> +in order to support this interface is strongly discouraged. In case such
> +operations are needed, use of the Stateless Video Encoder Interface (in
> +development) is strongly advised.
> +
> +Conventions and notation used in this document
> +==============================================
> +
> +1. The general V4L2 API rules apply if not specified in this document
> +   otherwise.
> +
> +2. The meaning of words "must", "may", "should", etc. is as per `RFC
> +   2119 <https://tools.ietf.org/html/rfc2119>`_.
> +
> +3. All steps not marked "optional" are required.
> +
> +4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
> +   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
> +   unless specified otherwise.
> +
> +5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
> +   used interchangeably with multi-planar API, unless specified otherwise,
> +   depending on decoder capabilities and following the general V4L2 guidelines.
> +
> +6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
> +   [0..2]: i = 0, 1, 2.
> +
> +7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
> +   queue containing data that resulted from processing buffer A.
> +
> +Glossary
> +========
> +
> +Refer to :ref:`decoder-glossary`.
> +
> +State machine
> +=============
> +
> +.. kernel-render:: DOT
> +   :alt: DOT digraph of encoder state machine
> +   :caption: Encoder state machine
> +
> +   digraph encoder_state_machine {
> +       node [shape = doublecircle, label="Encoding"] Encoding;
> +
> +       node [shape = circle, label="Initialization"] Initialization;
> +       node [shape = circle, label="Stopped"] Stopped;
> +       node [shape = circle, label="Drain"] Drain;
> +       node [shape = circle, label="Reset"] Reset;
> +
> +       node [shape = point]; qi
> +       qi -> Initialization [ label = "open()" ];
> +
> +       Initialization -> Encoding [ label = "Both queues streaming" ];
> +
> +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> +       Encoding -> Encoding;
> +
> +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];

Shouldn't this be

	Drain -> Stopped [ label = "All OUTPUT\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(OUTPUT)" ];

? While draining, the encoder continues encoding until all source
buffers, i.e., buffers in the OUTPUT queue, are encoded or STREAMOFF
happens on the OUTPUT queue. At the same time, the client continues to
queue and dequeue buffers on the CAPTURE queue and there might be
buffers queued on the CAPTURE queue even if the driver returned the
buffer with the FLAG_LAST set and returns -EPIPE on further DQBUF
requests.

Michael

> +       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +
> +       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
> +       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
> +
> +       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
> +       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> +   }
> +
> +Querying capabilities
> +=====================
> +
> +1. To enumerate the set of coded formats supported by the encoder, the
> +   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
> +
> +   * The full set of supported formats will be returned, regardless of the
> +     format set on ``OUTPUT``.
> +
> +2. To enumerate the set of supported raw formats, the client may call
> +   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
> +
> +   * Only the formats supported for the format currently active on ``CAPTURE``
> +     will be returned.
> +
> +   * In order to enumerate raw formats supported by a given coded format,
> +     the client must first set that coded format on ``CAPTURE`` and then
> +     enumerate the formats on ``OUTPUT``.
> +
> +3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
> +   resolutions for a given format, passing desired pixel format in
> +   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
> +     format will include all possible coded resolutions supported by the
> +     encoder for given coded pixel format.
> +
> +   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
> +     will include all possible frame buffer resolutions supported by the
> +     encoder for given raw pixel format and coded format currently set on
> +     ``CAPTURE``.
> +
> +4. Supported profiles and levels for the coded format currently set on
> +   ``CAPTURE``, if applicable, may be queried using their respective controls
> +   via :c:func:`VIDIOC_QUERYCTRL`.
> +
> +5. Any additional encoder capabilities may be discovered by querying
> +   their respective controls.
> +
> +Initialization
> +==============
> +
> +1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
> +
> +     ``pixelformat``
> +         the coded format to be produced
> +
> +     ``sizeimage``
> +         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
> +         match hardware requirements
> +
> +     ``width``, ``height``
> +         ignored (always zero)
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``sizeimage``
> +         adjusted size of ``CAPTURE`` buffers
> +
> +   .. important::
> +
> +      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
> +      format. The encoder will derive a new ``OUTPUT`` format from the
> +      ``CAPTURE`` format being set, including resolution, colorimetry
> +      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
> +      must adjust it afterwards.
> +
> +2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
> +   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``pixelformat``
> +         raw format supported for the coded format currently selected on
> +         the ``CAPTURE`` queue.
> +
> +     other fields
> +         follow standard semantics
> +
> +3. Set the raw source format on the ``OUTPUT`` queue via
> +   :c:func:`VIDIOC_S_FMT`.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``pixelformat``
> +         raw format of the source
> +
> +     ``width``, ``height``
> +         source resolution
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``width``, ``height``
> +         may be adjusted by encoder to match alignment requirements, as
> +         required by the currently selected formats
> +
> +     other fields
> +         follow standard semantics
> +
> +   * Setting the source resolution will reset the selection rectangles to their
> +     default values, based on the new resolution, as described in the step 5
> +     below.
> +
> +4. **Optional.** Set the visible resolution for the stream metadata via
> +   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
> +
> +   * **Required fields:**
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     ``target``
> +         set to ``V4L2_SEL_TGT_CROP``
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
> +         rectangle and may be subject to adjustment to match codec and
> +         hardware constraints
> +
> +   * **Return fields:**
> +
> +     ``r.left``, ``r.top``, ``r.width``, ``r.height``
> +         visible rectangle adjusted by the encoder
> +
> +   * The following selection targets are supported on ``OUTPUT``:
> +
> +     ``V4L2_SEL_TGT_CROP_BOUNDS``
> +         equal to the full source frame, matching the active ``OUTPUT``
> +         format
> +
> +     ``V4L2_SEL_TGT_CROP_DEFAULT``
> +         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
> +
> +     ``V4L2_SEL_TGT_CROP``
> +         rectangle within the source buffer to be encoded into the
> +         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
> +
> +         .. note::
> +
> +            A common use case for this selection target is encoding a source
> +            video with a resolution that is not a multiple of a macroblock,
> +            e.g.  the common 1920x1080 resolution may require the source
> +            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
> +            size. To avoid encoding the padding, the client needs to explicitly
> +            configure this selection target to 1920x1080.
> +
> +     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
> +         maximum rectangle within the coded resolution, which the cropped
> +         source frame can be composed into; if the hardware does not support
> +         composition or scaling, then this is always equal to the rectangle of
> +         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
> +         equal to a rectangle of width and height matching
> +         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
> +
> +     ``V4L2_SEL_TGT_COMPOSE``
> +         rectangle within the coded frame, which the cropped source frame
> +         is to be composed into; defaults to
> +         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
> +         additional compose/scaling capabilities; resulting stream will
> +         have this rectangle encoded as the visible rectangle in its
> +         metadata
> +
> +   .. warning::
> +
> +      The encoder may adjust the crop/compose rectangles to the nearest
> +      supported ones to meet codec and hardware requirements. The client needs
> +      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
> +
> +5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
> +   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
> +         ``CAPTURE``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +          actual number of buffers allocated
> +
> +   .. warning::
> +
> +      The actual number of allocated buffers may differ from the ``count``
> +      given. The client must check the updated value of ``count`` after the
> +      call returns.
> +
> +   .. note::
> +
> +      To allocate more than the minimum number of OUTPUT buffers (for pipeline
> +      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
> +      control to get the minimum number of buffers required, and pass the
> +      obtained value plus the number of additional buffers needed in the
> +      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
> +
> +   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
> +   control over buffer allocation.
> +
> +   * **Required fields:**
> +
> +     ``count``
> +         requested number of buffers to allocate; greater than zero
> +
> +     ``type``
> +         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
> +
> +     other fields
> +         follow standard semantics
> +
> +   * **Return fields:**
> +
> +     ``count``
> +         adjusted to the number of allocated buffers
> +
> +6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
> +   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
> +   encoding process starts when both queues start streaming.
> +
> +.. note::
> +
> +   If the client stops the ``CAPTURE`` queue during the encode process and then
> +   restarts it again, the encoder will begin generating a stream independent
> +   from the stream generated before the stop. The exact constraints depend
> +   on the coded format, but may include the following implications:
> +
> +   * encoded frames produced after the restart must not reference any
> +     frames produced before the stop, e.g. no long term references for
> +     H.264,
> +
> +   * any headers that must be included in a standalone stream must be
> +     produced again, e.g. SPS and PPS for H.264.
> +
> +Encoding
> +========
> +
> +This state is reached after the `Initialization` sequence finishes
> +successfully.  In this state, the client queues and dequeues buffers to both
> +queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
> +standard semantics.
> +
> +The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
> +format and may be affected by codec-specific extended controls, as stated
> +in the documentation of each format.
> +
> +Both queues operate independently, following standard behavior of V4L2 buffer
> +queues and memory-to-memory devices. In addition, the order of encoded frames
> +dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
> +frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
> +e.g. frame reordering.
> +
> +The client must not assume any direct relationship between ``CAPTURE`` and
> +``OUTPUT`` buffers and any specific timing of buffers becoming
> +available to dequeue. Specifically:
> +
> +* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
> +  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
> +  frame that preceded it in display, but succeeded it in the decode order),
> +
> +* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
> +  ``CAPTURE`` later into encode process, and/or after processing further
> +  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
> +  reordering is used,
> +
> +* buffers may become available on the ``CAPTURE`` queue without additional
> +  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
> +  ``OUTPUT`` buffers queued in the past whose decoding results are only
> +  available at later time, due to specifics of the decoding process,
> +
> +* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
> +  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
> +  encoder needs to use the frame as a reference for encoding further frames.
> +
> +.. note::
> +
> +   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
> +   originated from, the client can set the ``timestamp`` field of the
> +   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
> +   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
> +   will have their ``timestamp`` field set to the same value when dequeued.
> +
> +   In addition to the straightforward case of one ``OUTPUT`` buffer producing
> +   one ``CAPTURE`` buffer, the following cases are defined:
> +
> +   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
> +     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
> +
> +   * the encoding order differs from the presentation order (i.e. the
> +     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
> +     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
> +     and thus monotonicity of the timestamps cannot be guaranteed.
> +
> +.. note::
> +
> +   To let the client distinguish between frame types (keyframes, intermediate
> +   frames; the exact list of types depends on the coded format), the
> +   ``CAPTURE`` buffers will have corresponding flag bits set in their
> +   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
> +   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
> +   and their meanings.
> +
> +Encoding parameter changes
> +==========================
> +
> +The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
> +parameters at any time. The availability of parameters is encoder-specific
> +and the client must query the encoder to find the set of available controls.
> +
> +The ability to change each parameter during encoding is encoder-specific, as
> +per the standard semantics of the V4L2 control interface. The client may
> +attempt to set a control during encoding and if the operation fails with the
> +-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
> +configuration change to be allowed. To do this, it may follow the `Drain`
> +sequence to avoid losing the already queued/encoded frames.
> +
> +The timing of parameter updates is encoder-specific, as per the standard
> +semantics of the V4L2 control interface. If the client needs to apply the
> +parameters exactly at specific frame, using the Request API
> +(:ref:`media-request-api`) should be considered, if supported by the encoder.
> +
> +Drain
> +=====
> +
> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> +drain sequence described below. After the drain sequence ends, the client has
> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> +sequence was started.
> +
> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> +
> +   * **Required fields:**
> +
> +     ``cmd``
> +         set to ``V4L2_ENC_CMD_STOP``
> +
> +     ``flags``
> +         set to 0
> +
> +     ``pts``
> +         set to 0
> +
> +   .. warning::
> +
> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> +      queues are streaming. For compatibility reasons, the call to
> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> +      not streaming, but at the same time it will not initiate the `Drain`
> +      sequence and so the steps described below would not be applicable.
> +
> +2. Any ``OUTPUT`` buffers queued by the client before the
> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> +   normal. The client must continue to handle both queues independently,
> +   similarly to normal encode operation. This includes:
> +
> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> +
> +     .. warning::
> +
> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> +        as it does not contain an encoded frame.
> +
> +     .. note::
> +
> +        Any attempt to dequeue more buffers beyond the buffer marked with
> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> +        :c:func:`VIDIOC_DQBUF`.
> +
> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> +
> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> +
> +   .. note::
> +
> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> +      event when the last frame has been decoded and all frames are ready to be
> +      dequeued. It is deprecated behavior and the client must not rely on it.
> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> +
> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> +   until the client issues any of the following operations:
> +
> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> +     operation normally, with all the state from before the drain,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> +     and then resume encoding,
> +
> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> +
> +.. note::
> +
> +   Once the drain sequence is initiated, the client needs to drive it to
> +   completion, as described by the steps above, unless it aborts the process by
> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> +   will fail with -EBUSY error code if attempted.
> +
> +   Although mandatory, the availability of encoder commands may be queried
> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> +
> +Reset
> +=====
> +
> +The client may want to request the encoder to reinitialize the encoding, so
> +that the following stream data becomes independent from the stream data
> +generated before. Depending on the coded format, that may imply that:
> +
> +* encoded frames produced after the restart must not reference any frames
> +  produced before the stop, e.g. no long term references for H.264,
> +
> +* any headers that must be included in a standalone stream must be produced
> +  again, e.g. SPS and PPS for H.264.
> +
> +This can be achieved by performing the reset sequence.
> +
> +1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
> +   and respective buffers are dequeued.
> +
> +2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
> +   will return all currently queued ``CAPTURE`` buffers to the client, without
> +   valid frame data.
> +
> +3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
> +   continue with regular encoding sequence. The encoded frames produced into
> +   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
> +   decoded without the need for frames encoded before the reset sequence,
> +   starting at the first ``OUTPUT`` buffer queued after issuing the
> +   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
> +
> +This sequence may be also used to change encoding parameters for encoders
> +without the ability to change the parameters on the fly.
> +
> +Commit points
> +=============
> +
> +Setting formats and allocating buffers triggers changes in the behavior of the
> +encoder.
> +
> +1. Setting the format on the ``CAPTURE`` queue may change the set of formats
> +   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
> +   that the ``OUTPUT`` format may be reset and the client must not rely on the
> +   previously set format being preserved.
> +
> +2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
> +   supported for the current ``CAPTURE`` format.
> +
> +3. Setting the format on the ``OUTPUT`` queue does not change the list of
> +   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
> +   format that is not supported for the currently selected ``CAPTURE`` format
> +   will result in the encoder adjusting the requested ``OUTPUT`` format to a
> +   supported one.
> +
> +4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
> +   supported coded formats, irrespectively of the current ``OUTPUT`` format.
> +
> +5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
> +   change the format on the queue. Drivers will return the -EBUSY error code
> +   for any such format change attempt.
> +
> +To summarize, setting formats and allocation must always start with the
> +``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
> +set of supported formats for the ``OUTPUT`` queue.
> diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> index a0e06a88c872..82630ea89b16 100644
> --- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
> +++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
> @@ -17,6 +17,7 @@ Video Memory-To-Memory Interface
>      :maxdepth: 1
>  
>      dev-decoder
> +    dev-encoder
>  
>  A V4L2 memory-to-memory device can compress, decompress, transform, or
>  otherwise convert video data from one format into another format, in memory.
> diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> index caf14e440447..593d25b935bf 100644
> --- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> +++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
> @@ -44,6 +44,11 @@ Single-planar format structure
>  	inside the stream, when fed to a stateful mem2mem decoder, the fields
>  	may be zero to rely on the decoder to detect the right values. For more
>  	details see :ref:`decoder` and format descriptions.
> +
> +	For compressed formats on the CAPTURE side of a stateful mem2mem
> +	encoder, the fields must be zero, since the coded size is expected to
> +	be calculated internally by the encoder itself, based on the OUTPUT
> +	side. For more details see :ref:`encoder` and format descriptions.
>      * - __u32
>        - ``pixelformat``
>        - The pixel format or type of compression, set by the application.
> diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
> index 97015b9b40b8..e236ea23481b 100644
> --- a/Documentation/media/uapi/v4l/v4l2.rst
> +++ b/Documentation/media/uapi/v4l/v4l2.rst
> @@ -63,6 +63,7 @@ Authors, in alphabetical order:
>  - Figa, Tomasz <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - H Schimek, Michael <[e-mail]>
>  
> @@ -75,6 +76,7 @@ Authors, in alphabetical order:
>  - Osciak, Pawel <[e-mail]>
>  
>    - Documented the memory-to-memory decoder interface.
> +  - Documented the memory-to-memory encoder interface.
>  
>  - Osciak, Pawel <[e-mail]>
>  
> diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> index c313ca8b8cb5..2b59a4e7f5ce 100644
> --- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> +++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
> @@ -57,19 +57,23 @@ currently only used by the STOP command and contains one bit: If the
>  until the end of the current *Group Of Pictures*, otherwise it will stop
>  immediately.
>  
> -A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> -call sends an implicit START command to the encoder if it has not been
> -started yet. After a STOP command, :ref:`read() <func-read>` calls will read
> +After a STOP command, :ref:`read() <func-read>` calls will read
>  the remaining data buffered by the driver. When the buffer is empty,
>  :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
>  call will restart the encoder.
>  
> +A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
> +call sends an implicit START command to the encoder if it has not been
> +started yet. Applies to both queues of mem2mem encoders.
> +
>  A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
>  call of a streaming file descriptor sends an implicit immediate STOP to
> -the encoder, and all buffered data is discarded.
> +the encoder, and all buffered data is discarded. Applies to both queues of
> +mem2mem encoders.
>  
>  These ioctls are optional, not all drivers may support them. They were
> -introduced in Linux 2.6.21.
> +introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
> +encoders (as further documented in :ref:`encoder`).
>  
>  
>  .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
> @@ -114,16 +118,20 @@ introduced in Linux 2.6.21.
>        - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
>  	is set, encoding will continue until the end of the current *Group
>  	Of Pictures*, otherwise encoding will stop immediately. When the
> -	encoder is already stopped, this command does nothing. mem2mem
> -	encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
> -	has been encoded and all frames are ready to be dequeued and will
> -	set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
> -	the capture queue to indicate there will be no new buffers
> -	produced to dequeue. This buffer may be empty, indicated by the
> -	driver setting the ``bytesused`` field to 0. Once the
> -	``V4L2_BUF_FLAG_LAST`` flag was set, the
> -	:ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
> -	but return an ``EPIPE`` error code.
> +	encoder is already stopped, this command does nothing.
> +
> +	A stateful mem2mem encoder will proceed with encoding the source
> +	buffers pending before the command is issued and then stop producing
> +	new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
> +	has been encoded and all frames are ready to be dequeued and will set
> +	the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
> +	capture queue to indicate there will be no new buffers produced to
> +	dequeue. This buffer may be empty, indicated by the driver setting the
> +	``bytesused`` field to 0. Once the buffer with the
> +	``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
> +	<VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
> +	error code. No flags or other arguments are accepted in case of mem2mem
> +	encoders.  See :ref:`encoder` for more details.
>      * - ``V4L2_ENC_CMD_PAUSE``
>        - 2
>        - Pause the encoder. When the encoder has not been started yet, the

Tomasz Figa May 14, 2019, 8:12 a.m. UTC | #27

Hi Michael,

On Tue, Apr 30, 2019 at 07:34:12PM +0200, Michael Tretter wrote:
> On Thu, 24 Jan 2019 19:04:19 +0900, Tomasz Figa wrote:

[snip]

> > +State machine
> > +=============
> > +
> > +.. kernel-render:: DOT
> > +   :alt: DOT digraph of encoder state machine
> > +   :caption: Encoder state machine
> > +
> > +   digraph encoder_state_machine {
> > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > +
> > +       node [shape = circle, label="Initialization"] Initialization;
> > +       node [shape = circle, label="Stopped"] Stopped;
> > +       node [shape = circle, label="Drain"] Drain;
> > +       node [shape = circle, label="Reset"] Reset;
> > +
> > +       node [shape = point]; qi
> > +       qi -> Initialization [ label = "open()" ];
> > +
> > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > +
> > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > +       Encoding -> Encoding;
> > +
> > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> 
> Shouldn't this be
> 
> 	Drain -> Stopped [ label = "All OUTPUT\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(OUTPUT)" ];
> 
> ? While draining, the encoder continues encoding until all source
> buffers, i.e., buffers in the OUTPUT queue, are encoded or STREAMOFF
> happens on the OUTPUT queue. At the same time, the client continues to
> queue and dequeue buffers on the CAPTURE queue and there might be
> buffers queued on the CAPTURE queue even if the driver returned the
> buffer with the FLAG_LAST set and returns -EPIPE on further DQBUF
> requests.
>

The STREAMOFF should be on OUTPUT indeed, because that immediately
removes any OUTPUT buffers from the queue, so there is nothing to be
encoded to wait for anymore.

The "All OUTPUT buffers dequeued" part is correct, though. The last
OUTPUT buffer in the flush sequence is considered encoded after the
application dequeues the corresponding CAPTURE buffer is dequeued and
that buffer is marked with the V4L2_BUF_FLAG_LAST flag.

Best regards,
Tomasz

Michael Tretter May 16, 2019, 8:37 a.m. UTC | #28

On Tue, 14 May 2019 17:12:04 +0900, Tomasz Figa wrote:
> Hi Michael,
> 
> On Tue, Apr 30, 2019 at 07:34:12PM +0200, Michael Tretter wrote:
> > On Thu, 24 Jan 2019 19:04:19 +0900, Tomasz Figa wrote:  
> 
> [snip]
> 
> > > +State machine
> > > +=============
> > > +
> > > +.. kernel-render:: DOT
> > > +   :alt: DOT digraph of encoder state machine
> > > +   :caption: Encoder state machine
> > > +
> > > +   digraph encoder_state_machine {
> > > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > > +
> > > +       node [shape = circle, label="Initialization"] Initialization;
> > > +       node [shape = circle, label="Stopped"] Stopped;
> > > +       node [shape = circle, label="Drain"] Drain;
> > > +       node [shape = circle, label="Reset"] Reset;
> > > +
> > > +       node [shape = point]; qi
> > > +       qi -> Initialization [ label = "open()" ];
> > > +
> > > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > > +
> > > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > > +       Encoding -> Encoding;
> > > +
> > > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];  
> > 
> > Shouldn't this be
> > 
> > 	Drain -> Stopped [ label = "All OUTPUT\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(OUTPUT)" ];
> > 
> > ? While draining, the encoder continues encoding until all source
> > buffers, i.e., buffers in the OUTPUT queue, are encoded or STREAMOFF
> > happens on the OUTPUT queue. At the same time, the client continues to
> > queue and dequeue buffers on the CAPTURE queue and there might be
> > buffers queued on the CAPTURE queue even if the driver returned the
> > buffer with the FLAG_LAST set and returns -EPIPE on further DQBUF
> > requests.
> >  
> 
> The STREAMOFF should be on OUTPUT indeed, because that immediately
> removes any OUTPUT buffers from the queue, so there is nothing to be
> encoded to wait for anymore.
> 
> The "All OUTPUT buffers dequeued" part is correct, though. The last
> OUTPUT buffer in the flush sequence is considered encoded after the
> application dequeues the corresponding CAPTURE buffer is dequeued and
> that buffer is marked with the V4L2_BUF_FLAG_LAST flag.

I understand. As the application continues to queue and dequeue buffers
on the CAPTURE queue until it received the last CAPTURE buffer and cannot
dequeue further CAPTURE buffers, "All CAPTURE buffers dequeued" is
correct. Thanks for the clarification.

Michael

> 
> Best regards,
> Tomasz
>

Tomasz Figa May 16, 2019, 8:48 a.m. UTC | #29

On Thu, May 16, 2019 at 5:37 PM Michael Tretter
<[e-mail]> wrote:
>
> On Tue, 14 May 2019 17:12:04 +0900, Tomasz Figa wrote:
> > Hi Michael,
> >
> > On Tue, Apr 30, 2019 at 07:34:12PM +0200, Michael Tretter wrote:
> > > On Thu, 24 Jan 2019 19:04:19 +0900, Tomasz Figa wrote:
> >
> > [snip]
> >
> > > > +State machine
> > > > +=============
> > > > +
> > > > +.. kernel-render:: DOT
> > > > +   :alt: DOT digraph of encoder state machine
> > > > +   :caption: Encoder state machine
> > > > +
> > > > +   digraph encoder_state_machine {
> > > > +       node [shape = doublecircle, label="Encoding"] Encoding;
> > > > +
> > > > +       node [shape = circle, label="Initialization"] Initialization;
> > > > +       node [shape = circle, label="Stopped"] Stopped;
> > > > +       node [shape = circle, label="Drain"] Drain;
> > > > +       node [shape = circle, label="Reset"] Reset;
> > > > +
> > > > +       node [shape = point]; qi
> > > > +       qi -> Initialization [ label = "open()" ];
> > > > +
> > > > +       Initialization -> Encoding [ label = "Both queues streaming" ];
> > > > +
> > > > +       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
> > > > +       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
> > > > +       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
> > > > +       Encoding -> Encoding;
> > > > +
> > > > +       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
> > >
> > > Shouldn't this be
> > >
> > >     Drain -> Stopped [ label = "All OUTPUT\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(OUTPUT)" ];
> > >
> > > ? While draining, the encoder continues encoding until all source
> > > buffers, i.e., buffers in the OUTPUT queue, are encoded or STREAMOFF
> > > happens on the OUTPUT queue. At the same time, the client continues to
> > > queue and dequeue buffers on the CAPTURE queue and there might be
> > > buffers queued on the CAPTURE queue even if the driver returned the
> > > buffer with the FLAG_LAST set and returns -EPIPE on further DQBUF
> > > requests.
> > >
> >
> > The STREAMOFF should be on OUTPUT indeed, because that immediately
> > removes any OUTPUT buffers from the queue, so there is nothing to be
> > encoded to wait for anymore.
> >
> > The "All OUTPUT buffers dequeued" part is correct, though. The last

I just realized that I made a typo here, sorry. It obviously should be
"All CAPTURE buffers dequeued", as per the existing spec draft and
rest of the sentence. Fortunately you seem to have figured that out.
:)

> > OUTPUT buffer in the flush sequence is considered encoded after the
> > application dequeues the corresponding CAPTURE buffer is dequeued and
> > that buffer is marked with the V4L2_BUF_FLAG_LAST flag.
>
> I understand. As the application continues to queue and dequeue buffers
> on the CAPTURE queue until it received the last CAPTURE buffer and cannot
> dequeue further CAPTURE buffers, "All CAPTURE buffers dequeued" is
> correct. Thanks for the clarification.
>
> Michael
>
> >
> > Best regards,
> > Tomasz
> >

Tomasz Figa May 22, 2019, 8:43 a.m. UTC | #30

On Tue, Apr 9, 2019 at 6:43 PM Tomasz Figa <[e-mail]> wrote:
>
> On Tue, Apr 9, 2019 at 6:37 PM Hans Verkuil <[e-mail]> wrote:
> >
> > On 4/9/19 9:11 AM, Tomasz Figa wrote:
> > > On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:
> > >>
> > >> On 4/8/19 10:36 AM, Tomasz Figa wrote:
> > >>> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
> > >>>>
> > >>>> Another comment found while creating compliance tests:
> > >>>>
> > >>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
> > >>>>> +Drain
> > >>>>> +=====
> > >>>>> +
> > >>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> > >>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> > >>>>> +drain sequence described below. After the drain sequence ends, the client has
> > >>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> > >>>>> +sequence was started.
> > >>>>> +
> > >>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> > >>>>> +
> > >>>>> +   * **Required fields:**
> > >>>>> +
> > >>>>> +     ``cmd``
> > >>>>> +         set to ``V4L2_ENC_CMD_STOP``
> > >>>>> +
> > >>>>> +     ``flags``
> > >>>>> +         set to 0
> > >>>>> +
> > >>>>> +     ``pts``
> > >>>>> +         set to 0
> > >>>>> +
> > >>>>> +   .. warning::
> > >>>>> +
> > >>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> > >>>>> +      queues are streaming. For compatibility reasons, the call to
> > >>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> > >>>>> +      not streaming, but at the same time it will not initiate the `Drain`
> > >>>>> +      sequence and so the steps described below would not be applicable.
> > >>>>> +
> > >>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
> > >>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> > >>>>> +   normal. The client must continue to handle both queues independently,
> > >>>>> +   similarly to normal encode operation. This includes:
> > >>>>> +
> > >>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> > >>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> > >>>>> +
> > >>>>> +     .. warning::
> > >>>>> +
> > >>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> > >>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> > >>>>> +        as it does not contain an encoded frame.
> > >>>>> +
> > >>>>> +     .. note::
> > >>>>> +
> > >>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
> > >>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> > >>>>> +        :c:func:`VIDIOC_DQBUF`.
> > >>>>> +
> > >>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> > >>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> > >>>>> +
> > >>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> > >>>>> +
> > >>>>> +   .. note::
> > >>>>> +
> > >>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> > >>>>> +      event when the last frame has been decoded and all frames are ready to be
> > >>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
> > >>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> > >>>>> +
> > >>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> > >>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> > >>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> > >>>>> +   until the client issues any of the following operations:
> > >>>>> +
> > >>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> > >>>>> +     operation normally, with all the state from before the drain,
> > >>>>
> > >>>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
> > >>>>
> > >>>> In other words: while draining is in progress START will return EBUSY. When draining
> > >>>> was finished, then START will resume the encoder. In all other cases it just returns
> > >>>> 0 since the encoder is really already started.
> > >>>>
> > >>>
> > >>> Yes, that was the intention and seems to be the closest to the
> > >>> behavior described in the existing documentation.
> > >>>
> > >>>>> +
> > >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > >>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> > >>>>> +     and then resume encoding,
> > >>>>> +
> > >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > >>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> > >>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> > >>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> > >>>>> +
> > >>>>> +.. note::
> > >>>>> +
> > >>>>> +   Once the drain sequence is initiated, the client needs to drive it to
> > >>>>> +   completion, as described by the steps above, unless it aborts the process by
> > >>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> > >>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> > >>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> > >>>>> +   will fail with -EBUSY error code if attempted.
> > >>>>
> > >>>> I assume calling STOP again once the drain sequence completed just returns 0 and
> > >>>> doesn't do anything else (since we're already stopped).
> > >>>>
> > >>>
> > >>> Right.
> > >>>
> > >>>>> +
> > >>>>> +   Although mandatory, the availability of encoder commands may be queried
> > >>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
> > >>>>
> > >>>> Some corner cases:
> > >>>>
> > >>>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
> > >>>>    Now ENC_CMD_STOP is issued. What should happen?
> > >>>>
> > >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> > >>>>    at once as an empty buffer with FLAG_LAST set.
> > >>>>
> > >>>
> > >>> SGTM. It's actually similar to a relatively common case where all
> > >>> CAPTURE buffers have been dequeued and the application has to refill
> > >>> the CAPTURE queue, but in the meantime a drain request needs to be
> > >>> issued.
> > >>>
> > >>>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
> > >>>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> > >>>>
> > >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> > >>>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
> > >>>>    previous corner case.
> > >>>
> > >>> Agreed.
> > >>>
> > >>>>
> > >>>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
> > >>>>    is issued. What should happen?
> > >>>>
> > >>>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
> > >>>>    buffer with FLAG_LAST set.
> > >>>
> > >>> Generally agreed, but not sure if there is a reason to specifically
> > >>> refer to the oldest buffer. (I'm personally for keeping the queues
> > >>> ordered, though...)
> > >>
> > >> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
> > >> queued up for userspace to signal that the encoder has stopped." Or something along
> > >> those lines.
> > >
> > > I've added a note:
> > >
> > >    For reference, handling of various corner cases is described below:
> > >
> > >    * In case of no buffer in the ``OUTPUT`` queue at the time the
> > >      ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
> > >      immediately and the encoder returns an empty ``CAPTURE`` buffer with the
> > >      ``V4L2_BUF_FLAG_LAST`` flag set.
> > >
> > >    * In case of no buffer in the ``CAPTURE`` queue at the time the drain
> > >      sequence completes, the next time the client queues a ``CAPTURE`` buffer
> > >      it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
> > >      flag set.
> > >
> > >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
> > >      middle of the drain sequence, the drain sequence is cancelled and all
> >
> > cancelled -> canceled
> >
> > >      ``CAPTURE`` buffers are implicitly returned to the userpace.
> >
> > userpace -> userspace
> >
>
> Actually changed it to "client".
>
> > >
> > >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
> > >      middle of the drain sequence, the drain sequence completes immediately and
> > >      next ``CAPTURE`` buffer will be returned empty with the
> > >      ``V4L2_BUF_FLAG_LAST`` flag set.
> > >
> > > Slightly changed the split into cases to cover behaviors rather than
> > > conditions. WDYT?
> >
> > Looks good (with those two typos fixed).
> >
>
> Thanks.

Hmm, we actually looked into implementing this in mtk-vcodec and
handling of this corner case gets quite complicated. When stopping the
streaming on OUTPUT, you may not have any available CAPTURE buffer, so
you need to keep some extra state in the driver and check it in vb2
.buf_queue for CAPTURE to return the first buffer and complete the
drain.

The general handling of drain would have to look like this:
 - VIDIOC_DECODER_CMD must check if a drain isn't already in progress
and also whether the queues are streaming,
 - STREAMOFF(CAPTURE) needs to cancel any pending drain,
 - STREAMOFF(OUTPUT) needs to return a CAPTURE buffer with LAST set if
there is one or postpone it until a buffer is queued,
 - QBUF(CAPTURE) must return the buffer instantly if such return was
postponed by STREAMOFF(OUTPUT).
 - DQBUF(CAPTURE) of a buffer with the LAST flag set would finish the sequence.

Sounds like we definitely need some generic code to handle this...

>
> >         Hans
> >
> > >
> > > Best regards,
> > > Tomasz
> > >
> > >>
> > >> Regards,
> > >>
> > >>         Hans
> > >>
> > >>>
> > >>>>
> > >>>> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
> > >>>>    Before the drain process completes STREAMOFF is called for either CAPTURE or
> > >>>>    OUTPUT queue. What should happen?
> > >>>>
> > >>>>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
> > >>>>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
> > >>>>    will be used as input to the encoder.
> > >>>>
> > >>>
> > >>> Agreed.
> > >>>
> > >>>>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
> > >>>>    FLAG_LAST set.
> > >>>
> > >>> Agreed.
> > >>>
> > >>>>
> > >>>> Some of this might have to be documented, but these corner cases should certainly be
> > >>>> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> > >>>> with this.
> > >>>
> > >>> Agreed with just one minor comment. Thanks for checking with me!
> > >>>
> > >>> Best regards,
> > >>> Tomasz
> > >>>
> > >>
> >

Hans Verkuil May 22, 2019, 8:51 a.m. UTC | #31

On 5/22/19 10:43 AM, Tomasz Figa wrote:
> On Tue, Apr 9, 2019 at 6:43 PM Tomasz Figa <[e-mail]> wrote:
>>
>> On Tue, Apr 9, 2019 at 6:37 PM Hans Verkuil <[e-mail]> wrote:
>>>
>>> On 4/9/19 9:11 AM, Tomasz Figa wrote:
>>>> On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:
>>>>>
>>>>> On 4/8/19 10:36 AM, Tomasz Figa wrote:
>>>>>> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:
>>>>>>>
>>>>>>> Another comment found while creating compliance tests:
>>>>>>>
>>>>>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:
>>>>>>>> +Drain
>>>>>>>> +=====
>>>>>>>> +
>>>>>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
>>>>>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
>>>>>>>> +drain sequence described below. After the drain sequence ends, the client has
>>>>>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
>>>>>>>> +sequence was started.
>>>>>>>> +
>>>>>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
>>>>>>>> +
>>>>>>>> +   * **Required fields:**
>>>>>>>> +
>>>>>>>> +     ``cmd``
>>>>>>>> +         set to ``V4L2_ENC_CMD_STOP``
>>>>>>>> +
>>>>>>>> +     ``flags``
>>>>>>>> +         set to 0
>>>>>>>> +
>>>>>>>> +     ``pts``
>>>>>>>> +         set to 0
>>>>>>>> +
>>>>>>>> +   .. warning::
>>>>>>>> +
>>>>>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
>>>>>>>> +      queues are streaming. For compatibility reasons, the call to
>>>>>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
>>>>>>>> +      not streaming, but at the same time it will not initiate the `Drain`
>>>>>>>> +      sequence and so the steps described below would not be applicable.
>>>>>>>> +
>>>>>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
>>>>>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
>>>>>>>> +   normal. The client must continue to handle both queues independently,
>>>>>>>> +   similarly to normal encode operation. This includes:
>>>>>>>> +
>>>>>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
>>>>>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
>>>>>>>> +
>>>>>>>> +     .. warning::
>>>>>>>> +
>>>>>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
>>>>>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
>>>>>>>> +        as it does not contain an encoded frame.
>>>>>>>> +
>>>>>>>> +     .. note::
>>>>>>>> +
>>>>>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
>>>>>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
>>>>>>>> +        :c:func:`VIDIOC_DQBUF`.
>>>>>>>> +
>>>>>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
>>>>>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
>>>>>>>> +
>>>>>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
>>>>>>>> +
>>>>>>>> +   .. note::
>>>>>>>> +
>>>>>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
>>>>>>>> +      event when the last frame has been decoded and all frames are ready to be
>>>>>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
>>>>>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
>>>>>>>> +
>>>>>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
>>>>>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
>>>>>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
>>>>>>>> +   until the client issues any of the following operations:
>>>>>>>> +
>>>>>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
>>>>>>>> +     operation normally, with all the state from before the drain,
>>>>>>>
>>>>>>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
>>>>>>>
>>>>>>> In other words: while draining is in progress START will return EBUSY. When draining
>>>>>>> was finished, then START will resume the encoder. In all other cases it just returns
>>>>>>> 0 since the encoder is really already started.
>>>>>>>
>>>>>>
>>>>>> Yes, that was the intention and seems to be the closest to the
>>>>>> behavior described in the existing documentation.
>>>>>>
>>>>>>>> +
>>>>>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>>>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
>>>>>>>> +     and then resume encoding,
>>>>>>>> +
>>>>>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
>>>>>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
>>>>>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
>>>>>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
>>>>>>>> +
>>>>>>>> +.. note::
>>>>>>>> +
>>>>>>>> +   Once the drain sequence is initiated, the client needs to drive it to
>>>>>>>> +   completion, as described by the steps above, unless it aborts the process by
>>>>>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
>>>>>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
>>>>>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
>>>>>>>> +   will fail with -EBUSY error code if attempted.
>>>>>>>
>>>>>>> I assume calling STOP again once the drain sequence completed just returns 0 and
>>>>>>> doesn't do anything else (since we're already stopped).
>>>>>>>
>>>>>>
>>>>>> Right.
>>>>>>
>>>>>>>> +
>>>>>>>> +   Although mandatory, the availability of encoder commands may be queried
>>>>>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
>>>>>>>
>>>>>>> Some corner cases:
>>>>>>>
>>>>>>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
>>>>>>>    Now ENC_CMD_STOP is issued. What should happen?
>>>>>>>
>>>>>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>>>>>>    at once as an empty buffer with FLAG_LAST set.
>>>>>>>
>>>>>>
>>>>>> SGTM. It's actually similar to a relatively common case where all
>>>>>> CAPTURE buffers have been dequeued and the application has to refill
>>>>>> the CAPTURE queue, but in the meantime a drain request needs to be
>>>>>> issued.
>>>>>>
>>>>>>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
>>>>>>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
>>>>>>>
>>>>>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
>>>>>>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
>>>>>>>    previous corner case.
>>>>>>
>>>>>> Agreed.
>>>>>>
>>>>>>>
>>>>>>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
>>>>>>>    is issued. What should happen?
>>>>>>>
>>>>>>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
>>>>>>>    buffer with FLAG_LAST set.
>>>>>>
>>>>>> Generally agreed, but not sure if there is a reason to specifically
>>>>>> refer to the oldest buffer. (I'm personally for keeping the queues
>>>>>> ordered, though...)
>>>>>
>>>>> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
>>>>> queued up for userspace to signal that the encoder has stopped." Or something along
>>>>> those lines.
>>>>
>>>> I've added a note:
>>>>
>>>>    For reference, handling of various corner cases is described below:
>>>>
>>>>    * In case of no buffer in the ``OUTPUT`` queue at the time the
>>>>      ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
>>>>      immediately and the encoder returns an empty ``CAPTURE`` buffer with the
>>>>      ``V4L2_BUF_FLAG_LAST`` flag set.
>>>>
>>>>    * In case of no buffer in the ``CAPTURE`` queue at the time the drain
>>>>      sequence completes, the next time the client queues a ``CAPTURE`` buffer
>>>>      it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
>>>>      flag set.
>>>>
>>>>    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
>>>>      middle of the drain sequence, the drain sequence is cancelled and all
>>>
>>> cancelled -> canceled
>>>
>>>>      ``CAPTURE`` buffers are implicitly returned to the userpace.
>>>
>>> userpace -> userspace
>>>
>>
>> Actually changed it to "client".
>>
>>>>
>>>>    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
>>>>      middle of the drain sequence, the drain sequence completes immediately and
>>>>      next ``CAPTURE`` buffer will be returned empty with the
>>>>      ``V4L2_BUF_FLAG_LAST`` flag set.
>>>>
>>>> Slightly changed the split into cases to cover behaviors rather than
>>>> conditions. WDYT?
>>>
>>> Looks good (with those two typos fixed).
>>>
>>
>> Thanks.
> 
> Hmm, we actually looked into implementing this in mtk-vcodec and
> handling of this corner case gets quite complicated. When stopping the
> streaming on OUTPUT, you may not have any available CAPTURE buffer, so
> you need to keep some extra state in the driver and check it in vb2
> .buf_queue for CAPTURE to return the first buffer and complete the
> drain.
> 
> The general handling of drain would have to look like this:
>  - VIDIOC_DECODER_CMD must check if a drain isn't already in progress
> and also whether the queues are streaming,
>  - STREAMOFF(CAPTURE) needs to cancel any pending drain,
>  - STREAMOFF(OUTPUT) needs to return a CAPTURE buffer with LAST set if
> there is one or postpone it until a buffer is queued,
>  - QBUF(CAPTURE) must return the buffer instantly if such return was
> postponed by STREAMOFF(OUTPUT).
>  - DQBUF(CAPTURE) of a buffer with the LAST flag set would finish the sequence.
> 
> Sounds like we definitely need some generic code to handle this...

Yeah, I realized the same thing while implementing this for vicodec.

I started on adding generic support to v4l2-mem2mem.c, but then decided
to wait until we have agreement on this spec. From the point of view of
the application this all makes sense and it is easy to implement, but for
a driver (or v4l2-mem2mem.c) this is definitely more complex.

Regards,

	Hans

Michael Tretter May 22, 2019, 2:16 p.m. UTC | #32

On Wed, 22 May 2019 17:43:49 +0900, Tomasz Figa wrote:
> On Tue, Apr 9, 2019 at 6:43 PM Tomasz Figa <[e-mail]> wrote:
> >
> > On Tue, Apr 9, 2019 at 6:37 PM Hans Verkuil <[e-mail]> wrote:  
> > >
> > > On 4/9/19 9:11 AM, Tomasz Figa wrote:  
> > > > On Mon, Apr 8, 2019 at 5:43 PM Hans Verkuil <[e-mail]> wrote:  
> > > >>
> > > >> On 4/8/19 10:36 AM, Tomasz Figa wrote:  
> > > >>> On Mon, Mar 25, 2019 at 10:12 PM Hans Verkuil <[e-mail]> wrote:  
> > > >>>>
> > > >>>> Another comment found while creating compliance tests:
> > > >>>>
> > > >>>> On 1/24/19 11:04 AM, Tomasz Figa wrote:  
> > > >>>>> +Drain
> > > >>>>> +=====
> > > >>>>> +
> > > >>>>> +To ensure that all the queued ``OUTPUT`` buffers have been processed and the
> > > >>>>> +related ``CAPTURE`` buffers are given to the client, the client must follow the
> > > >>>>> +drain sequence described below. After the drain sequence ends, the client has
> > > >>>>> +received all encoded frames for all ``OUTPUT`` buffers queued before the
> > > >>>>> +sequence was started.
> > > >>>>> +
> > > >>>>> +1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
> > > >>>>> +
> > > >>>>> +   * **Required fields:**
> > > >>>>> +
> > > >>>>> +     ``cmd``
> > > >>>>> +         set to ``V4L2_ENC_CMD_STOP``
> > > >>>>> +
> > > >>>>> +     ``flags``
> > > >>>>> +         set to 0
> > > >>>>> +
> > > >>>>> +     ``pts``
> > > >>>>> +         set to 0
> > > >>>>> +
> > > >>>>> +   .. warning::
> > > >>>>> +
> > > >>>>> +      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
> > > >>>>> +      queues are streaming. For compatibility reasons, the call to
> > > >>>>> +      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
> > > >>>>> +      not streaming, but at the same time it will not initiate the `Drain`
> > > >>>>> +      sequence and so the steps described below would not be applicable.
> > > >>>>> +
> > > >>>>> +2. Any ``OUTPUT`` buffers queued by the client before the
> > > >>>>> +   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
> > > >>>>> +   normal. The client must continue to handle both queues independently,
> > > >>>>> +   similarly to normal encode operation. This includes:
> > > >>>>> +
> > > >>>>> +   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
> > > >>>>> +     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
> > > >>>>> +
> > > >>>>> +     .. warning::
> > > >>>>> +
> > > >>>>> +        The last buffer may be empty (with :c:type:`v4l2_buffer`
> > > >>>>> +        ``bytesused`` = 0) and in that case it must be ignored by the client,
> > > >>>>> +        as it does not contain an encoded frame.
> > > >>>>> +
> > > >>>>> +     .. note::
> > > >>>>> +
> > > >>>>> +        Any attempt to dequeue more buffers beyond the buffer marked with
> > > >>>>> +        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
> > > >>>>> +        :c:func:`VIDIOC_DQBUF`.
> > > >>>>> +
> > > >>>>> +   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
> > > >>>>> +     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
> > > >>>>> +
> > > >>>>> +   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
> > > >>>>> +
> > > >>>>> +   .. note::
> > > >>>>> +
> > > >>>>> +      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
> > > >>>>> +      event when the last frame has been decoded and all frames are ready to be
> > > >>>>> +      dequeued. It is deprecated behavior and the client must not rely on it.
> > > >>>>> +      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
> > > >>>>> +
> > > >>>>> +3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
> > > >>>>> +   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
> > > >>>>> +   and it will accept, but not process any newly queued ``OUTPUT`` buffers
> > > >>>>> +   until the client issues any of the following operations:
> > > >>>>> +
> > > >>>>> +   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
> > > >>>>> +     operation normally, with all the state from before the drain,  
> > > >>>>
> > > >>>> I assume that calling CMD_START when *not* draining will succeed but does nothing.
> > > >>>>
> > > >>>> In other words: while draining is in progress START will return EBUSY. When draining
> > > >>>> was finished, then START will resume the encoder. In all other cases it just returns
> > > >>>> 0 since the encoder is really already started.
> > > >>>>  
> > > >>>
> > > >>> Yes, that was the intention and seems to be the closest to the
> > > >>> behavior described in the existing documentation.
> > > >>>  
> > > >>>>> +
> > > >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > > >>>>> +     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
> > > >>>>> +     and then resume encoding,
> > > >>>>> +
> > > >>>>> +   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
> > > >>>>> +     ``OUTPUT`` queue - the encoder will resume operation normally, however any
> > > >>>>> +     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
> > > >>>>> +     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
> > > >>>>> +
> > > >>>>> +.. note::
> > > >>>>> +
> > > >>>>> +   Once the drain sequence is initiated, the client needs to drive it to
> > > >>>>> +   completion, as described by the steps above, unless it aborts the process by
> > > >>>>> +   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
> > > >>>>> +   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
> > > >>>>> +   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
> > > >>>>> +   will fail with -EBUSY error code if attempted.  
> > > >>>>
> > > >>>> I assume calling STOP again once the drain sequence completed just returns 0 and
> > > >>>> doesn't do anything else (since we're already stopped).
> > > >>>>  
> > > >>>
> > > >>> Right.
> > > >>>  
> > > >>>>> +
> > > >>>>> +   Although mandatory, the availability of encoder commands may be queried
> > > >>>>> +   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.  
> > > >>>>
> > > >>>> Some corner cases:
> > > >>>>
> > > >>>> 1) No buffers are queued on either vb2_queue, but STREAMON is called for both queues.
> > > >>>>    Now ENC_CMD_STOP is issued. What should happen?
> > > >>>>
> > > >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> > > >>>>    at once as an empty buffer with FLAG_LAST set.
> > > >>>>  
> > > >>>
> > > >>> SGTM. It's actually similar to a relatively common case where all
> > > >>> CAPTURE buffers have been dequeued and the application has to refill
> > > >>> the CAPTURE queue, but in the meantime a drain request needs to be
> > > >>> issued.
> > > >>>  
> > > >>>> 2) Both queues are streaming and buffers have been encoded, but currently no buffers
> > > >>>>    are queued on either vb2_queue. Now ENC_CMD_STOP is issued. What should happen?
> > > >>>>
> > > >>>>    Proposal: the next time the applications queues a CAPTURE buffer it is returned
> > > >>>>    at once as an empty buffer with FLAG_LAST set. This is consistent with the
> > > >>>>    previous corner case.  
> > > >>>
> > > >>> Agreed.
> > > >>>  
> > > >>>>
> > > >>>> 3) The CAPTURE queue contains buffers, the OUTPUT queue does not. Now ENC_CMD_STOP
> > > >>>>    is issued. What should happen?
> > > >>>>
> > > >>>>    Proposal: the oldest CAPTURE buffer in the ready queue is returned as an empty
> > > >>>>    buffer with FLAG_LAST set.  
> > > >>>
> > > >>> Generally agreed, but not sure if there is a reason to specifically
> > > >>> refer to the oldest buffer. (I'm personally for keeping the queues
> > > >>> ordered, though...)  
> > > >>
> > > >> Feel free to rephrase. Perhaps: "an empty CAPTURE buffer with FLAG_LAST set should be
> > > >> queued up for userspace to signal that the encoder has stopped." Or something along
> > > >> those lines.  
> > > >
> > > > I've added a note:
> > > >
> > > >    For reference, handling of various corner cases is described below:
> > > >
> > > >    * In case of no buffer in the ``OUTPUT`` queue at the time the
> > > >      ``V4L2_ENC_CMD_STOP`` command was issued, the drain sequence completes
> > > >      immediately and the encoder returns an empty ``CAPTURE`` buffer with the
> > > >      ``V4L2_BUF_FLAG_LAST`` flag set.
> > > >
> > > >    * In case of no buffer in the ``CAPTURE`` queue at the time the drain
> > > >      sequence completes, the next time the client queues a ``CAPTURE`` buffer
> > > >      it is returned at once as an empty buffer with the ``V4L2_BUF_FLAG_LAST``
> > > >      flag set.
> > > >
> > > >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``CAPTURE`` queue in the
> > > >      middle of the drain sequence, the drain sequence is cancelled and all  
> > >
> > > cancelled -> canceled
> > >  
> > > >      ``CAPTURE`` buffers are implicitly returned to the userpace.  
> > >
> > > userpace -> userspace
> > >  
> >
> > Actually changed it to "client".
> >  
> > > >
> > > >    * If :c:func:`VIDIOC_STREAMOFF` is called on the ``OUTPUT`` queue in the
> > > >      middle of the drain sequence, the drain sequence completes immediately and
> > > >      next ``CAPTURE`` buffer will be returned empty with the
> > > >      ``V4L2_BUF_FLAG_LAST`` flag set.
> > > >
> > > > Slightly changed the split into cases to cover behaviors rather than
> > > > conditions. WDYT?  
> > >
> > > Looks good (with those two typos fixed).
> > >  
> >
> > Thanks.  
> 
> Hmm, we actually looked into implementing this in mtk-vcodec and
> handling of this corner case gets quite complicated. When stopping the
> streaming on OUTPUT, you may not have any available CAPTURE buffer, so
> you need to keep some extra state in the driver and check it in vb2
> .buf_queue for CAPTURE to return the first buffer and complete the
> drain.
> 
> The general handling of drain would have to look like this:
>  - VIDIOC_DECODER_CMD must check if a drain isn't already in progress
> and also whether the queues are streaming,
>  - STREAMOFF(CAPTURE) needs to cancel any pending drain,
>  - STREAMOFF(OUTPUT) needs to return a CAPTURE buffer with LAST set if
> there is one or postpone it until a buffer is queued,
>  - QBUF(CAPTURE) must return the buffer instantly if such return was
> postponed by STREAMOFF(OUTPUT).
>  - DQBUF(CAPTURE) of a buffer with the LAST flag set would finish the sequence.
> 
> Sounds like we definitely need some generic code to handle this...

I implemented this as well for allegro. I ended up exactly with an
additional state (ALLEGRO_STATE_WAIT_FOR_BUFFER) to handle this corner
case. The implementation was tricky and I would highly appreciate
generic code that is used by other drivers as well.

Michael


> 
> >  
> > >         Hans
> > >  
> > > >
> > > > Best regards,
> > > > Tomasz
> > > >  .
> > > >>
> > > >> Regards,
> > > >>
> > > >>         Hans
> > > >>  
> > > >>>  
> > > >>>>
> > > >>>> 4) Both queues have queued buffers. ENC_CMD_STOP is issued to start the drain process.
> > > >>>>    Before the drain process completes STREAMOFF is called for either CAPTURE or
> > > >>>>    OUTPUT queue. What should happen?
> > > >>>>
> > > >>>>    Proposal for STREAMOFF(CAPTURE): aborts the drain process and all CAPTURE buffers are
> > > >>>>    returned to userspace. If encoding is restarted, then any remaining OUTPUT buffers
> > > >>>>    will be used as input to the encoder.
> > > >>>>  
> > > >>>
> > > >>> Agreed.
> > > >>>  
> > > >>>>    Proposal for STREAMOFF(OUTPUT): the next capture buffer will be empty and have
> > > >>>>    FLAG_LAST set.  
> > > >>>
> > > >>> Agreed.
> > > >>>  
> > > >>>>
> > > >>>> Some of this might have to be documented, but these corner cases should certainly be
> > > >>>> tested by v4l2-compliance. Before I write those tests I'd like to know if you agree
> > > >>>> with this.  
> > > >>>
> > > >>> Agreed with just one minor comment. Thanks for checking with me!
> > > >>>
> > > >>> Best regards,
> > > >>> Tomasz
> > > >>>  
> > > >>  
> > >  
>

Patch

diff --git a/Documentation/media/uapi/v4l/dev-encoder.rst b/Documentation/media/uapi/v4l/dev-encoder.rst
new file mode 100644
index 000000000000..fb8b05a132ee
--- /dev/null
+++ b/Documentation/media/uapi/v4l/dev-encoder.rst
@@ -0,0 +1,586 @@ 
+.. -*- coding: utf-8; mode: rst -*-
+
+.. _encoder:
+
+*************************************************
+Memory-to-memory Stateful Video Encoder Interface
+*************************************************
+
+A stateful video encoder takes raw video frames in display order and encodes
+them into a bitstream. It generates complete chunks of the bitstream, including
+all metadata, headers, etc. The resulting bitstream does not require any
+further post-processing by the client.
+
+Performing software stream processing, header generation etc. in the driver
+in order to support this interface is strongly discouraged. In case such
+operations are needed, use of the Stateless Video Encoder Interface (in
+development) is strongly advised.
+
+Conventions and notation used in this document
+==============================================
+
+1. The general V4L2 API rules apply if not specified in this document
+   otherwise.
+
+2. The meaning of words "must", "may", "should", etc. is as per `RFC
+   2119 <https://tools.ietf.org/html/rfc2119>`_.
+
+3. All steps not marked "optional" are required.
+
+4. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
+   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
+   unless specified otherwise.
+
+5. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
+   used interchangeably with multi-planar API, unless specified otherwise,
+   depending on decoder capabilities and following the general V4L2 guidelines.
+
+6. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
+   [0..2]: i = 0, 1, 2.
+
+7. Given an ``OUTPUT`` buffer A, then A’ represents a buffer on the ``CAPTURE``
+   queue containing data that resulted from processing buffer A.
+
+Glossary
+========
+
+Refer to :ref:`decoder-glossary`.
+
+State machine
+=============
+
+.. kernel-render:: DOT
+   :alt: DOT digraph of encoder state machine
+   :caption: Encoder state machine
+
+   digraph encoder_state_machine {
+       node [shape = doublecircle, label="Encoding"] Encoding;
+
+       node [shape = circle, label="Initialization"] Initialization;
+       node [shape = circle, label="Stopped"] Stopped;
+       node [shape = circle, label="Drain"] Drain;
+       node [shape = circle, label="Reset"] Reset;
+
+       node [shape = point]; qi
+       qi -> Initialization [ label = "open()" ];
+
+       Initialization -> Encoding [ label = "Both queues streaming" ];
+
+       Encoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
+       Encoding -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
+       Encoding -> Stopped [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
+       Encoding -> Encoding;
+
+       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
+       Drain -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
+
+       Reset -> Encoding [ label = "VIDIOC_STREAMON(CAPTURE)" ];
+       Reset -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
+
+       Stopped -> Encoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(OUTPUT)" ];
+       Stopped -> Reset [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
+   }
+
+Querying capabilities
+=====================
+
+1. To enumerate the set of coded formats supported by the encoder, the
+   client may call :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
+
+   * The full set of supported formats will be returned, regardless of the
+     format set on ``OUTPUT``.
+
+2. To enumerate the set of supported raw formats, the client may call
+   :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
+
+   * Only the formats supported for the format currently active on ``CAPTURE``
+     will be returned.
+
+   * In order to enumerate raw formats supported by a given coded format,
+     the client must first set that coded format on ``CAPTURE`` and then
+     enumerate the formats on ``OUTPUT``.
+
+3. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
+   resolutions for a given format, passing desired pixel format in
+   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
+
+   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
+     format will include all possible coded resolutions supported by the
+     encoder for given coded pixel format.
+
+   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
+     will include all possible frame buffer resolutions supported by the
+     encoder for given raw pixel format and coded format currently set on
+     ``CAPTURE``.
+
+4. Supported profiles and levels for the coded format currently set on
+   ``CAPTURE``, if applicable, may be queried using their respective controls
+   via :c:func:`VIDIOC_QUERYCTRL`.
+
+5. Any additional encoder capabilities may be discovered by querying
+   their respective controls.
+
+Initialization
+==============
+
+1. Set the coded format on the ``CAPTURE`` queue via :c:func:`VIDIOC_S_FMT`
+
+   * **Required fields:**
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``
+
+     ``pixelformat``
+         the coded format to be produced
+
+     ``sizeimage``
+         desired size of ``CAPTURE`` buffers; the encoder may adjust it to
+         match hardware requirements
+
+     ``width``, ``height``
+         ignored (always zero)
+
+     other fields
+         follow standard semantics
+
+   * **Return fields:**
+
+     ``sizeimage``
+         adjusted size of ``CAPTURE`` buffers
+
+   .. important::
+
+      Changing the ``CAPTURE`` format may change the currently set ``OUTPUT``
+      format. The encoder will derive a new ``OUTPUT`` format from the
+      ``CAPTURE`` format being set, including resolution, colorimetry
+      parameters, etc. If the client needs a specific ``OUTPUT`` format, it
+      must adjust it afterwards.
+
+2. **Optional.** Enumerate supported ``OUTPUT`` formats (raw formats for
+   source) for the selected coded format via :c:func:`VIDIOC_ENUM_FMT`.
+
+   * **Required fields:**
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
+
+     other fields
+         follow standard semantics
+
+   * **Return fields:**
+
+     ``pixelformat``
+         raw format supported for the coded format currently selected on
+         the ``CAPTURE`` queue.
+
+     other fields
+         follow standard semantics
+
+3. Set the raw source format on the ``OUTPUT`` queue via
+   :c:func:`VIDIOC_S_FMT`.
+
+   * **Required fields:**
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
+
+     ``pixelformat``
+         raw format of the source
+
+     ``width``, ``height``
+         source resolution
+
+     other fields
+         follow standard semantics
+
+   * **Return fields:**
+
+     ``width``, ``height``
+         may be adjusted by encoder to match alignment requirements, as
+         required by the currently selected formats
+
+     other fields
+         follow standard semantics
+
+   * Setting the source resolution will reset the selection rectangles to their
+     default values, based on the new resolution, as described in the step 5
+     below.
+
+4. **Optional.** Set the visible resolution for the stream metadata via
+   :c:func:`VIDIOC_S_SELECTION` on the ``OUTPUT`` queue.
+
+   * **Required fields:**
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
+
+     ``target``
+         set to ``V4L2_SEL_TGT_CROP``
+
+     ``r.left``, ``r.top``, ``r.width``, ``r.height``
+         visible rectangle; this must fit within the `V4L2_SEL_TGT_CROP_BOUNDS`
+         rectangle and may be subject to adjustment to match codec and
+         hardware constraints
+
+   * **Return fields:**
+
+     ``r.left``, ``r.top``, ``r.width``, ``r.height``
+         visible rectangle adjusted by the encoder
+
+   * The following selection targets are supported on ``OUTPUT``:
+
+     ``V4L2_SEL_TGT_CROP_BOUNDS``
+         equal to the full source frame, matching the active ``OUTPUT``
+         format
+
+     ``V4L2_SEL_TGT_CROP_DEFAULT``
+         equal to ``V4L2_SEL_TGT_CROP_BOUNDS``
+
+     ``V4L2_SEL_TGT_CROP``
+         rectangle within the source buffer to be encoded into the
+         ``CAPTURE`` stream; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``
+
+         .. note::
+
+            A common use case for this selection target is encoding a source
+            video with a resolution that is not a multiple of a macroblock,
+            e.g.  the common 1920x1080 resolution may require the source
+            buffers to be aligned to 1920x1088 for codecs with 16x16 macroblock
+            size. To avoid encoding the padding, the client needs to explicitly
+            configure this selection target to 1920x1080.
+
+     ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
+         maximum rectangle within the coded resolution, which the cropped
+         source frame can be composed into; if the hardware does not support
+         composition or scaling, then this is always equal to the rectangle of
+         width and height matching ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
+
+     ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
+         equal to a rectangle of width and height matching
+         ``V4L2_SEL_TGT_CROP`` and located at (0, 0)
+
+     ``V4L2_SEL_TGT_COMPOSE``
+         rectangle within the coded frame, which the cropped source frame
+         is to be composed into; defaults to
+         ``V4L2_SEL_TGT_COMPOSE_DEFAULT``; read-only on hardware without
+         additional compose/scaling capabilities; resulting stream will
+         have this rectangle encoded as the visible rectangle in its
+         metadata
+
+   .. warning::
+
+      The encoder may adjust the crop/compose rectangles to the nearest
+      supported ones to meet codec and hardware requirements. The client needs
+      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
+
+5. Allocate buffers for both ``OUTPUT`` and ``CAPTURE`` via
+   :c:func:`VIDIOC_REQBUFS`. This may be performed in any order.
+
+   * **Required fields:**
+
+     ``count``
+         requested number of buffers to allocate; greater than zero
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT`` or
+         ``CAPTURE``
+
+     other fields
+         follow standard semantics
+
+   * **Return fields:**
+
+     ``count``
+          actual number of buffers allocated
+
+   .. warning::
+
+      The actual number of allocated buffers may differ from the ``count``
+      given. The client must check the updated value of ``count`` after the
+      call returns.
+
+   .. note::
+
+      To allocate more than the minimum number of OUTPUT buffers (for pipeline
+      depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_OUTPUT``
+      control to get the minimum number of buffers required, and pass the
+      obtained value plus the number of additional buffers needed in the
+      ``count`` field to :c:func:`VIDIOC_REQBUFS`.
+
+   Alternatively, :c:func:`VIDIOC_CREATE_BUFS` can be used to have more
+   control over buffer allocation.
+
+   * **Required fields:**
+
+     ``count``
+         requested number of buffers to allocate; greater than zero
+
+     ``type``
+         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``
+
+     other fields
+         follow standard semantics
+
+   * **Return fields:**
+
+     ``count``
+         adjusted to the number of allocated buffers
+
+6. Begin streaming on both ``OUTPUT`` and ``CAPTURE`` queues via
+   :c:func:`VIDIOC_STREAMON`. This may be performed in any order. The actual
+   encoding process starts when both queues start streaming.
+
+.. note::
+
+   If the client stops the ``CAPTURE`` queue during the encode process and then
+   restarts it again, the encoder will begin generating a stream independent
+   from the stream generated before the stop. The exact constraints depend
+   on the coded format, but may include the following implications:
+
+   * encoded frames produced after the restart must not reference any
+     frames produced before the stop, e.g. no long term references for
+     H.264,
+
+   * any headers that must be included in a standalone stream must be
+     produced again, e.g. SPS and PPS for H.264.
+
+Encoding
+========
+
+This state is reached after the `Initialization` sequence finishes
+successfully.  In this state, the client queues and dequeues buffers to both
+queues via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the
+standard semantics.
+
+The contents of encoded ``CAPTURE`` buffers depend on the active coded pixel
+format and may be affected by codec-specific extended controls, as stated
+in the documentation of each format.
+
+Both queues operate independently, following standard behavior of V4L2 buffer
+queues and memory-to-memory devices. In addition, the order of encoded frames
+dequeued from the ``CAPTURE`` queue may differ from the order of queuing raw
+frames to the ``OUTPUT`` queue, due to properties of the selected coded format,
+e.g. frame reordering.
+
+The client must not assume any direct relationship between ``CAPTURE`` and
+``OUTPUT`` buffers and any specific timing of buffers becoming
+available to dequeue. Specifically:
+
+* a buffer queued to ``OUTPUT`` may result in more than 1 buffer produced on
+  ``CAPTURE`` (if returning an encoded frame allowed the encoder to return a
+  frame that preceded it in display, but succeeded it in the decode order),
+
+* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
+  ``CAPTURE`` later into encode process, and/or after processing further
+  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
+  reordering is used,
+
+* buffers may become available on the ``CAPTURE`` queue without additional
+  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
+  ``OUTPUT`` buffers queued in the past whose decoding results are only
+  available at later time, due to specifics of the decoding process,
+
+* buffers queued to ``OUTPUT`` may not become available to dequeue instantly
+  after being encoded into a corresponding ``CATPURE`` buffer, e.g. if the
+  encoder needs to use the frame as a reference for encoding further frames.
+
+.. note::
+
+   To allow matching encoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
+   originated from, the client can set the ``timestamp`` field of the
+   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
+   ``CAPTURE`` buffer(s), which resulted from encoding that ``OUTPUT`` buffer
+   will have their ``timestamp`` field set to the same value when dequeued.
+
+   In addition to the straightforward case of one ``OUTPUT`` buffer producing
+   one ``CAPTURE`` buffer, the following cases are defined:
+
+   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
+     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers,
+
+   * the encoding order differs from the presentation order (i.e. the
+     ``CAPTURE`` buffers are out-of-order compared to the ``OUTPUT`` buffers):
+     ``CAPTURE`` timestamps will not retain the order of ``OUTPUT`` timestamps
+     and thus monotonicity of the timestamps cannot be guaranteed.
+
+.. note::
+
+   To let the client distinguish between frame types (keyframes, intermediate
+   frames; the exact list of types depends on the coded format), the
+   ``CAPTURE`` buffers will have corresponding flag bits set in their
+   :c:type:`v4l2_buffer` struct when dequeued. See the documentation of
+   :c:type:`v4l2_buffer` and each coded pixel format for exact list of flags
+   and their meanings.
+
+Encoding parameter changes
+==========================
+
+The client is allowed to use :c:func:`VIDIOC_S_CTRL` to change encoder
+parameters at any time. The availability of parameters is encoder-specific
+and the client must query the encoder to find the set of available controls.
+
+The ability to change each parameter during encoding is encoder-specific, as
+per the standard semantics of the V4L2 control interface. The client may
+attempt to set a control during encoding and if the operation fails with the
+-EBUSY error code, the ``CAPTURE`` queue needs to be stopped for the
+configuration change to be allowed. To do this, it may follow the `Drain`
+sequence to avoid losing the already queued/encoded frames.
+
+The timing of parameter updates is encoder-specific, as per the standard
+semantics of the V4L2 control interface. If the client needs to apply the
+parameters exactly at specific frame, using the Request API
+(:ref:`media-request-api`) should be considered, if supported by the encoder.
+
+Drain
+=====
+
+To ensure that all the queued ``OUTPUT`` buffers have been processed and the
+related ``CAPTURE`` buffers are given to the client, the client must follow the
+drain sequence described below. After the drain sequence ends, the client has
+received all encoded frames for all ``OUTPUT`` buffers queued before the
+sequence was started.
+
+1. Begin the drain sequence by issuing :c:func:`VIDIOC_ENCODER_CMD`.
+
+   * **Required fields:**
+
+     ``cmd``
+         set to ``V4L2_ENC_CMD_STOP``
+
+     ``flags``
+         set to 0
+
+     ``pts``
+         set to 0
+
+   .. warning::
+
+      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
+      queues are streaming. For compatibility reasons, the call to
+      :c:func:`VIDIOC_ENCODER_CMD` will not fail even if any of the queues is
+      not streaming, but at the same time it will not initiate the `Drain`
+      sequence and so the steps described below would not be applicable.
+
+2. Any ``OUTPUT`` buffers queued by the client before the
+   :c:func:`VIDIOC_ENCODER_CMD` was issued will be processed and encoded as
+   normal. The client must continue to handle both queues independently,
+   similarly to normal encode operation. This includes:
+
+   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
+     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
+
+     .. warning::
+
+        The last buffer may be empty (with :c:type:`v4l2_buffer`
+        ``bytesused`` = 0) and in that case it must be ignored by the client,
+        as it does not contain an encoded frame.
+
+     .. note::
+
+        Any attempt to dequeue more buffers beyond the buffer marked with
+        ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
+        :c:func:`VIDIOC_DQBUF`.
+
+   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
+     before the ``V4L2_ENC_CMD_STOP`` command are dequeued,
+
+   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribes to it.
+
+   .. note::
+
+      For backwards compatibility, the encoder will signal a ``V4L2_EVENT_EOS``
+      event when the last frame has been decoded and all frames are ready to be
+      dequeued. It is deprecated behavior and the client must not rely on it.
+      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
+
+3. Once all ``OUTPUT`` buffers queued before the ``V4L2_ENC_CMD_STOP`` call are
+   dequeued and the last ``CAPTURE`` buffer is dequeued, the encoder is stopped
+   and it will accept, but not process any newly queued ``OUTPUT`` buffers
+   until the client issues any of the following operations:
+
+   * ``V4L2_ENC_CMD_START`` - the encoder will not be reset and will resume
+     operation normally, with all the state from before the drain,
+
+   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
+     ``CAPTURE`` queue - the encoder will be reset (see the `Reset` sequence)
+     and then resume encoding,
+
+   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
+     ``OUTPUT`` queue - the encoder will resume operation normally, however any
+     source frames queued to the ``OUTPUT`` queue between ``V4L2_ENC_CMD_STOP``
+     and :c:func:`VIDIOC_STREAMOFF` will be discarded.
+
+.. note::
+
+   Once the drain sequence is initiated, the client needs to drive it to
+   completion, as described by the steps above, unless it aborts the process by
+   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
+   queues.  The client is not allowed to issue ``V4L2_ENC_CMD_START`` or
+   ``V4L2_ENC_CMD_STOP`` again while the drain sequence is in progress and they
+   will fail with -EBUSY error code if attempted.
+
+   Although mandatory, the availability of encoder commands may be queried
+   using :c:func:`VIDIOC_TRY_ENCODER_CMD`.
+
+Reset
+=====
+
+The client may want to request the encoder to reinitialize the encoding, so
+that the following stream data becomes independent from the stream data
+generated before. Depending on the coded format, that may imply that:
+
+* encoded frames produced after the restart must not reference any frames
+  produced before the stop, e.g. no long term references for H.264,
+
+* any headers that must be included in a standalone stream must be produced
+  again, e.g. SPS and PPS for H.264.
+
+This can be achieved by performing the reset sequence.
+
+1. Perform the `Drain` sequence to ensure all the in-flight encoding finishes
+   and respective buffers are dequeued.
+
+2. Stop streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMOFF`. This
+   will return all currently queued ``CAPTURE`` buffers to the client, without
+   valid frame data.
+
+3. Start streaming on the ``CAPTURE`` queue via :c:func:`VIDIOC_STREAMON` and
+   continue with regular encoding sequence. The encoded frames produced into
+   ``CAPTURE`` buffers from now on will contain a standalone stream that can be
+   decoded without the need for frames encoded before the reset sequence,
+   starting at the first ``OUTPUT`` buffer queued after issuing the
+   `V4L2_ENC_CMD_STOP` of the `Drain` sequence.
+
+This sequence may be also used to change encoding parameters for encoders
+without the ability to change the parameters on the fly.
+
+Commit points
+=============
+
+Setting formats and allocating buffers triggers changes in the behavior of the
+encoder.
+
+1. Setting the format on the ``CAPTURE`` queue may change the set of formats
+   supported/advertised on the ``OUTPUT`` queue. In particular, it also means
+   that the ``OUTPUT`` format may be reset and the client must not rely on the
+   previously set format being preserved.
+
+2. Enumerating formats on the ``OUTPUT`` queue always returns only formats
+   supported for the current ``CAPTURE`` format.
+
+3. Setting the format on the ``OUTPUT`` queue does not change the list of
+   formats available on the ``CAPTURE`` queue. An attempt to set the ``OUTPUT``
+   format that is not supported for the currently selected ``CAPTURE`` format
+   will result in the encoder adjusting the requested ``OUTPUT`` format to a
+   supported one.
+
+4. Enumerating formats on the ``CAPTURE`` queue always returns the full set of
+   supported coded formats, irrespectively of the current ``OUTPUT`` format.
+
+5. While buffers are allocated on the ``CAPTURE`` queue, the client must not
+   change the format on the queue. Drivers will return the -EBUSY error code
+   for any such format change attempt.
+
+To summarize, setting formats and allocation must always start with the
+``CAPTURE`` queue and the ``CAPTURE`` queue is the master that governs the
+set of supported formats for the ``OUTPUT`` queue.
diff --git a/Documentation/media/uapi/v4l/dev-mem2mem.rst b/Documentation/media/uapi/v4l/dev-mem2mem.rst
index a0e06a88c872..82630ea89b16 100644
--- a/Documentation/media/uapi/v4l/dev-mem2mem.rst
+++ b/Documentation/media/uapi/v4l/dev-mem2mem.rst
@@ -17,6 +17,7 @@  Video Memory-To-Memory Interface
     :maxdepth: 1
 
     dev-decoder
+    dev-encoder
 
 A V4L2 memory-to-memory device can compress, decompress, transform, or
 otherwise convert video data from one format into another format, in memory.
diff --git a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
index caf14e440447..593d25b935bf 100644
--- a/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
+++ b/Documentation/media/uapi/v4l/pixfmt-v4l2.rst
@@ -44,6 +44,11 @@  Single-planar format structure
 	inside the stream, when fed to a stateful mem2mem decoder, the fields
 	may be zero to rely on the decoder to detect the right values. For more
 	details see :ref:`decoder` and format descriptions.
+
+	For compressed formats on the CAPTURE side of a stateful mem2mem
+	encoder, the fields must be zero, since the coded size is expected to
+	be calculated internally by the encoder itself, based on the OUTPUT
+	side. For more details see :ref:`encoder` and format descriptions.
     * - __u32
       - ``pixelformat``
       - The pixel format or type of compression, set by the application.
diff --git a/Documentation/media/uapi/v4l/v4l2.rst b/Documentation/media/uapi/v4l/v4l2.rst
index 97015b9b40b8..e236ea23481b 100644
--- a/Documentation/media/uapi/v4l/v4l2.rst
+++ b/Documentation/media/uapi/v4l/v4l2.rst
@@ -63,6 +63,7 @@  Authors, in alphabetical order:
 - Figa, Tomasz <[e-mail]>
 
   - Documented the memory-to-memory decoder interface.
+  - Documented the memory-to-memory encoder interface.
 
 - H Schimek, Michael <[e-mail]>
 
@@ -75,6 +76,7 @@  Authors, in alphabetical order:
 - Osciak, Pawel <[e-mail]>
 
   - Documented the memory-to-memory decoder interface.
+  - Documented the memory-to-memory encoder interface.
 
 - Osciak, Pawel <[e-mail]>
 
diff --git a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
index c313ca8b8cb5..2b59a4e7f5ce 100644
--- a/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
+++ b/Documentation/media/uapi/v4l/vidioc-encoder-cmd.rst
@@ -57,19 +57,23 @@  currently only used by the STOP command and contains one bit: If the
 until the end of the current *Group Of Pictures*, otherwise it will stop
 immediately.
 
-A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
-call sends an implicit START command to the encoder if it has not been
-started yet. After a STOP command, :ref:`read() <func-read>` calls will read
+After a STOP command, :ref:`read() <func-read>` calls will read
 the remaining data buffered by the driver. When the buffer is empty,
 :ref:`read() <func-read>` will return zero and the next :ref:`read() <func-read>`
 call will restart the encoder.
 
+A :ref:`read() <func-read>` or :ref:`VIDIOC_STREAMON <VIDIOC_STREAMON>`
+call sends an implicit START command to the encoder if it has not been
+started yet. Applies to both queues of mem2mem encoders.
+
 A :ref:`close() <func-close>` or :ref:`VIDIOC_STREAMOFF <VIDIOC_STREAMON>`
 call of a streaming file descriptor sends an implicit immediate STOP to
-the encoder, and all buffered data is discarded.
+the encoder, and all buffered data is discarded. Applies to both queues of
+mem2mem encoders.
 
 These ioctls are optional, not all drivers may support them. They were
-introduced in Linux 2.6.21.
+introduced in Linux 2.6.21. They are, however, mandatory for stateful mem2mem
+encoders (as further documented in :ref:`encoder`).
 
 
 .. tabularcolumns:: |p{4.4cm}|p{4.4cm}|p{8.7cm}|
@@ -114,16 +118,20 @@  introduced in Linux 2.6.21.
       - Stop the encoder. When the ``V4L2_ENC_CMD_STOP_AT_GOP_END`` flag
 	is set, encoding will continue until the end of the current *Group
 	Of Pictures*, otherwise encoding will stop immediately. When the
-	encoder is already stopped, this command does nothing. mem2mem
-	encoders will send a ``V4L2_EVENT_EOS`` event when the last frame
-	has been encoded and all frames are ready to be dequeued and will
-	set the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of
-	the capture queue to indicate there will be no new buffers
-	produced to dequeue. This buffer may be empty, indicated by the
-	driver setting the ``bytesused`` field to 0. Once the
-	``V4L2_BUF_FLAG_LAST`` flag was set, the
-	:ref:`VIDIOC_DQBUF <VIDIOC_QBUF>` ioctl will not block anymore,
-	but return an ``EPIPE`` error code.
+	encoder is already stopped, this command does nothing.
+
+	A stateful mem2mem encoder will proceed with encoding the source
+	buffers pending before the command is issued and then stop producing
+	new frames. It will send a ``V4L2_EVENT_EOS`` event when the last frame
+	has been encoded and all frames are ready to be dequeued and will set
+	the ``V4L2_BUF_FLAG_LAST`` buffer flag on the last buffer of the
+	capture queue to indicate there will be no new buffers produced to
+	dequeue. This buffer may be empty, indicated by the driver setting the
+	``bytesused`` field to 0. Once the buffer with the
+	``V4L2_BUF_FLAG_LAST`` flag set was dequeued, the :ref:`VIDIOC_DQBUF
+	<VIDIOC_QBUF>` ioctl will not block anymore, but return an ``EPIPE``
+	error code. No flags or other arguments are accepted in case of mem2mem
+	encoders.  See :ref:`encoder` for more details.
     * - ``V4L2_ENC_CMD_PAUSE``
       - 2
       - Pause the encoder. When the encoder has not been started yet, the