1.. SPDX-License-Identifier: GPL-2.0
2
3.. _decoder:
4
5*************************************************
6Memory-to-Memory Stateful Video Decoder Interface
7*************************************************
8
9A stateful video decoder takes complete chunks of the bytestream (e.g. Annex-B
10H.264/HEVC stream, raw VP8/9 stream) and decodes them into raw video frames in
11display order. The decoder is expected not to require any additional information
12from the client to process these buffers.
13
14Performing software parsing, processing etc. of the stream in the driver in
15order to support this interface is strongly discouraged. In case such
16operations are needed, use of the Stateless Video Decoder Interface (in
17development) is strongly advised.
18
19Conventions and Notations Used in This Document
20===============================================
21
221. The general V4L2 API rules apply if not specified in this document
23   otherwise.
24
252. The meaning of words "must", "may", "should", etc. is as per `RFC
26   2119 <https://tools.ietf.org/html/rfc2119>`_.
27
283. All steps not marked "optional" are required.
29
304. :c:func:`VIDIOC_G_EXT_CTRLS` and :c:func:`VIDIOC_S_EXT_CTRLS` may be used
31   interchangeably with :c:func:`VIDIOC_G_CTRL` and :c:func:`VIDIOC_S_CTRL`,
32   unless specified otherwise.
33
345. Single-planar API (see :ref:`planar-apis`) and applicable structures may be
35   used interchangeably with multi-planar API, unless specified otherwise,
36   depending on decoder capabilities and following the general V4L2 guidelines.
37
386. i = [a..b]: sequence of integers from a to b, inclusive, i.e. i =
39   [0..2]: i = 0, 1, 2.
40
417. Given an ``OUTPUT`` buffer A, then A' represents a buffer on the ``CAPTURE``
42   queue containing data that resulted from processing buffer A.
43
44.. _decoder-glossary:
45
46Glossary
47========
48
49CAPTURE
50   the destination buffer queue; for decoders, the queue of buffers containing
51   decoded frames; for encoders, the queue of buffers containing an encoded
52   bytestream; ``V4L2_BUF_TYPE_VIDEO_CAPTURE`` or
53   ``V4L2_BUF_TYPE_VIDEO_CAPTURE_MPLANE``; data is captured from the hardware
54   into ``CAPTURE`` buffers.
55
56client
57   the application communicating with the decoder or encoder implementing
58   this interface.
59
60coded format
61   encoded/compressed video bytestream format (e.g. H.264, VP8, etc.); see
62   also: raw format.
63
64coded height
65   height for given coded resolution.
66
67coded resolution
68   stream resolution in pixels aligned to codec and hardware requirements;
69   typically visible resolution rounded up to full macroblocks;
70   see also: visible resolution.
71
72coded width
73   width for given coded resolution.
74
75decode order
76   the order in which frames are decoded; may differ from display order if the
77   coded format includes a feature of frame reordering; for decoders,
78   ``OUTPUT`` buffers must be queued by the client in decode order; for
79   encoders ``CAPTURE`` buffers must be returned by the encoder in decode order.
80
81destination
82   data resulting from the decode process; see ``CAPTURE``.
83
84display order
85   the order in which frames must be displayed; for encoders, ``OUTPUT``
86   buffers must be queued by the client in display order; for decoders,
87   ``CAPTURE`` buffers must be returned by the decoder in display order.
88
89DPB
90   Decoded Picture Buffer; an H.264/HEVC term for a buffer that stores a decoded
91   raw frame available for reference in further decoding steps.
92
93EOS
94   end of stream.
95
96IDR
97   Instantaneous Decoder Refresh; a type of a keyframe in an H.264/HEVC-encoded
98   stream, which clears the list of earlier reference frames (DPBs).
99
100keyframe
101   an encoded frame that does not reference frames decoded earlier, i.e.
102   can be decoded fully on its own.
103
104macroblock
105   a processing unit in image and video compression formats based on linear
106   block transforms (e.g. H.264, VP8, VP9); codec-specific, but for most of
107   popular codecs the size is 16x16 samples (pixels).
108
109OUTPUT
110   the source buffer queue; for decoders, the queue of buffers containing
111   an encoded bytestream; for encoders, the queue of buffers containing raw
112   frames; ``V4L2_BUF_TYPE_VIDEO_OUTPUT`` or
113   ``V4L2_BUF_TYPE_VIDEO_OUTPUT_MPLANE``; the hardware is fed with data
114   from ``OUTPUT`` buffers.
115
116PPS
117   Picture Parameter Set; a type of metadata entity in an H.264/HEVC bytestream.
118
119raw format
120   uncompressed format containing raw pixel data (e.g. YUV, RGB formats).
121
122resume point
123   a point in the bytestream from which decoding may start/continue, without
124   any previous state/data present, e.g.: a keyframe (VP8/VP9) or
125   SPS/PPS/IDR sequence (H.264/HEVC); a resume point is required to start decode
126   of a new stream, or to resume decoding after a seek.
127
128source
129   data fed to the decoder or encoder; see ``OUTPUT``.
130
131source height
132   height in pixels for given source resolution; relevant to encoders only.
133
134source resolution
135   resolution in pixels of source frames being source to the encoder and
136   subject to further cropping to the bounds of visible resolution; relevant to
137   encoders only.
138
139source width
140   width in pixels for given source resolution; relevant to encoders only.
141
142SPS
143   Sequence Parameter Set; a type of metadata entity in an H.264/HEVC bytestream.
144
145stream metadata
146   additional (non-visual) information contained inside encoded bytestream;
147   for example: coded resolution, visible resolution, codec profile.
148
149visible height
150   height for given visible resolution; display height.
151
152visible resolution
153   stream resolution of the visible picture, in pixels, to be used for
154   display purposes; must be smaller or equal to coded resolution;
155   display resolution.
156
157visible width
158   width for given visible resolution; display width.
159
160State Machine
161=============
162
163.. kernel-render:: DOT
164   :alt: DOT digraph of decoder state machine
165   :caption: Decoder State Machine
166
167   digraph decoder_state_machine {
168       node [shape = doublecircle, label="Decoding"] Decoding;
169
170       node [shape = circle, label="Initialization"] Initialization;
171       node [shape = circle, label="Capture\nsetup"] CaptureSetup;
172       node [shape = circle, label="Dynamic\nResolution\nChange"] ResChange;
173       node [shape = circle, label="Stopped"] Stopped;
174       node [shape = circle, label="Drain"] Drain;
175       node [shape = circle, label="Seek"] Seek;
176       node [shape = circle, label="End of Stream"] EoS;
177
178       node [shape = point]; qi
179       qi -> Initialization [ label = "open()" ];
180
181       Initialization -> CaptureSetup [ label = "CAPTURE\nformat\nestablished" ];
182
183       CaptureSetup -> Stopped [ label = "CAPTURE\nbuffers\nready" ];
184
185       Decoding -> ResChange [ label = "Stream\nresolution\nchange" ];
186       Decoding -> Drain [ label = "V4L2_DEC_CMD_STOP" ];
187       Decoding -> EoS [ label = "EoS mark\nin the stream" ];
188       Decoding -> Seek [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
189       Decoding -> Stopped [ label = "VIDIOC_STREAMOFF(CAPTURE)" ];
190       Decoding -> Decoding;
191
192       ResChange -> CaptureSetup [ label = "CAPTURE\nformat\nestablished" ];
193       ResChange -> Seek [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
194
195       EoS -> Drain [ label = "Implicit\ndrain" ];
196
197       Drain -> Stopped [ label = "All CAPTURE\nbuffers dequeued\nor\nVIDIOC_STREAMOFF(CAPTURE)" ];
198       Drain -> Seek [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
199
200       Seek -> Decoding [ label = "VIDIOC_STREAMON(OUTPUT)" ];
201       Seek -> Initialization [ label = "VIDIOC_REQBUFS(OUTPUT, 0)" ];
202
203       Stopped -> Decoding [ label = "V4L2_DEC_CMD_START\nor\nVIDIOC_STREAMON(CAPTURE)" ];
204       Stopped -> Seek [ label = "VIDIOC_STREAMOFF(OUTPUT)" ];
205   }
206
207Querying Capabilities
208=====================
209
2101. To enumerate the set of coded formats supported by the decoder, the
211   client may call :c:func:`VIDIOC_ENUM_FMT` on ``OUTPUT``.
212
213   * The full set of supported formats will be returned, regardless of the
214     format set on ``CAPTURE``.
215   * Check the flags field of :c:type:`v4l2_fmtdesc` for more information
216     about the decoder's capabilities with respect to each coded format.
217     In particular whether or not the decoder has a full-fledged bytestream
218     parser and if the decoder supports dynamic resolution changes.
219
2202. To enumerate the set of supported raw formats, the client may call
221   :c:func:`VIDIOC_ENUM_FMT` on ``CAPTURE``.
222
223   * Only the formats supported for the format currently active on ``OUTPUT``
224     will be returned.
225
226   * In order to enumerate raw formats supported by a given coded format,
227     the client must first set that coded format on ``OUTPUT`` and then
228     enumerate formats on ``CAPTURE``.
229
2303. The client may use :c:func:`VIDIOC_ENUM_FRAMESIZES` to detect supported
231   resolutions for a given format, passing desired pixel format in
232   :c:type:`v4l2_frmsizeenum` ``pixel_format``.
233
234   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a coded pixel
235     format will include all possible coded resolutions supported by the
236     decoder for given coded pixel format.
237
238   * Values returned by :c:func:`VIDIOC_ENUM_FRAMESIZES` for a raw pixel format
239     will include all possible frame buffer resolutions supported by the
240     decoder for given raw pixel format and the coded format currently set on
241     ``OUTPUT``.
242
2434. Supported profiles and levels for the coded format currently set on
244   ``OUTPUT``, if applicable, may be queried using their respective controls
245   via :c:func:`VIDIOC_QUERYCTRL`.
246
247Initialization
248==============
249
2501. Set the coded format on ``OUTPUT`` via :c:func:`VIDIOC_S_FMT`.
251
252   * **Required fields:**
253
254     ``type``
255         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``.
256
257     ``pixelformat``
258         a coded pixel format.
259
260     ``width``, ``height``
261         coded resolution of the stream; required only if it cannot be parsed
262         from the stream for the given coded format; otherwise the decoder will
263         use this resolution as a placeholder resolution that will likely change
264         as soon as it can parse the actual coded resolution from the stream.
265
266     ``sizeimage``
267         desired size of ``OUTPUT`` buffers; the decoder may adjust it to
268         match hardware requirements.
269
270     other fields
271         follow standard semantics.
272
273   * **Return fields:**
274
275     ``sizeimage``
276         adjusted size of ``OUTPUT`` buffers.
277
278   * The ``CAPTURE`` format will be updated with an appropriate frame buffer
279     resolution instantly based on the width and height returned by
280     :c:func:`VIDIOC_S_FMT`.
281     However, for coded formats that include stream resolution information,
282     after the decoder is done parsing the information from the stream, it will
283     update the ``CAPTURE`` format with new values and signal a source change
284     event, regardless of whether they match the values set by the client or
285     not.
286
287   .. important::
288
289      Changing the ``OUTPUT`` format may change the currently set ``CAPTURE``
290      format. How the new ``CAPTURE`` format is determined is up to the decoder
291      and the client must ensure it matches its needs afterwards.
292
2932.  Allocate source (bytestream) buffers via :c:func:`VIDIOC_REQBUFS` on
294    ``OUTPUT``.
295
296    * **Required fields:**
297
298      ``count``
299          requested number of buffers to allocate; greater than zero.
300
301      ``type``
302          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``.
303
304      ``memory``
305          follows standard semantics.
306
307    * **Return fields:**
308
309      ``count``
310          the actual number of buffers allocated.
311
312    .. warning::
313
314       The actual number of allocated buffers may differ from the ``count``
315       given. The client must check the updated value of ``count`` after the
316       call returns.
317
318    Alternatively, :c:func:`VIDIOC_CREATE_BUFS` on the ``OUTPUT`` queue can be
319    used to have more control over buffer allocation.
320
321    * **Required fields:**
322
323      ``count``
324          requested number of buffers to allocate; greater than zero.
325
326      ``type``
327          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``.
328
329      ``memory``
330          follows standard semantics.
331
332      ``format``
333          follows standard semantics.
334
335    * **Return fields:**
336
337      ``count``
338          adjusted to the number of allocated buffers.
339
340    .. warning::
341
342       The actual number of allocated buffers may differ from the ``count``
343       given. The client must check the updated value of ``count`` after the
344       call returns.
345
3463.  Start streaming on the ``OUTPUT`` queue via :c:func:`VIDIOC_STREAMON`.
347
3484.  **This step only applies to coded formats that contain resolution information
349    in the stream.** Continue queuing/dequeuing bytestream buffers to/from the
350    ``OUTPUT`` queue via :c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`. The
351    buffers will be processed and returned to the client in order, until
352    required metadata to configure the ``CAPTURE`` queue are found. This is
353    indicated by the decoder sending a ``V4L2_EVENT_SOURCE_CHANGE`` event with
354    ``changes`` set to ``V4L2_EVENT_SRC_CH_RESOLUTION``.
355
356    * It is not an error if the first buffer does not contain enough data for
357      this to occur. Processing of the buffers will continue as long as more
358      data is needed.
359
360    * If data in a buffer that triggers the event is required to decode the
361      first frame, it will not be returned to the client, until the
362      initialization sequence completes and the frame is decoded.
363
364    * If the client has not set the coded resolution of the stream on its own,
365      calling :c:func:`VIDIOC_G_FMT`, :c:func:`VIDIOC_S_FMT`,
366      :c:func:`VIDIOC_TRY_FMT` or :c:func:`VIDIOC_REQBUFS` on the ``CAPTURE``
367      queue will not return the real values for the stream until a
368      ``V4L2_EVENT_SOURCE_CHANGE`` event with ``changes`` set to
369      ``V4L2_EVENT_SRC_CH_RESOLUTION`` is signaled.
370
371    .. important::
372
373       Any client query issued after the decoder queues the event will return
374       values applying to the just parsed stream, including queue formats,
375       selection rectangles and controls.
376
377    .. note::
378
379       A client capable of acquiring stream parameters from the bytestream on
380       its own may attempt to set the width and height of the ``OUTPUT`` format
381       to non-zero values matching the coded size of the stream, skip this step
382       and continue with the `Capture Setup` sequence. However, it must not
383       rely on any driver queries regarding stream parameters, such as
384       selection rectangles and controls, since the decoder has not parsed them
385       from the stream yet. If the values configured by the client do not match
386       those parsed by the decoder, a `Dynamic Resolution Change` will be
387       triggered to reconfigure them.
388
389    .. note::
390
391       No decoded frames are produced during this phase.
392
3935.  Continue with the `Capture Setup` sequence.
394
395Capture Setup
396=============
397
3981.  Call :c:func:`VIDIOC_G_FMT` on the ``CAPTURE`` queue to get format for the
399    destination buffers parsed/decoded from the bytestream.
400
401    * **Required fields:**
402
403      ``type``
404          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
405
406    * **Return fields:**
407
408      ``width``, ``height``
409          frame buffer resolution for the decoded frames.
410
411      ``pixelformat``
412          pixel format for decoded frames.
413
414      ``num_planes`` (for _MPLANE ``type`` only)
415          number of planes for pixelformat.
416
417      ``sizeimage``, ``bytesperline``
418          as per standard semantics; matching frame buffer format.
419
420    .. note::
421
422       The value of ``pixelformat`` may be any pixel format supported by the
423       decoder for the current stream. The decoder should choose a
424       preferred/optimal format for the default configuration. For example, a
425       YUV format may be preferred over an RGB format if an additional
426       conversion step would be required for the latter.
427
4282.  **Optional.** Acquire the visible resolution via
429    :c:func:`VIDIOC_G_SELECTION`.
430
431    * **Required fields:**
432
433      ``type``
434          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
435
436      ``target``
437          set to ``V4L2_SEL_TGT_COMPOSE``.
438
439    * **Return fields:**
440
441      ``r.left``, ``r.top``, ``r.width``, ``r.height``
442          the visible rectangle; it must fit within the frame buffer resolution
443          returned by :c:func:`VIDIOC_G_FMT` on ``CAPTURE``.
444
445    * The following selection targets are supported on ``CAPTURE``:
446
447      ``V4L2_SEL_TGT_CROP_BOUNDS``
448          corresponds to the coded resolution of the stream.
449
450      ``V4L2_SEL_TGT_CROP_DEFAULT``
451          the rectangle covering the part of the ``CAPTURE`` buffer that
452          contains meaningful picture data (visible area); width and height
453          will be equal to the visible resolution of the stream.
454
455      ``V4L2_SEL_TGT_CROP``
456          the rectangle within the coded resolution to be output to
457          ``CAPTURE``; defaults to ``V4L2_SEL_TGT_CROP_DEFAULT``; read-only on
458          hardware without additional compose/scaling capabilities.
459
460      ``V4L2_SEL_TGT_COMPOSE_BOUNDS``
461          the maximum rectangle within a ``CAPTURE`` buffer, which the cropped
462          frame can be composed into; equal to ``V4L2_SEL_TGT_CROP`` if the
463          hardware does not support compose/scaling.
464
465      ``V4L2_SEL_TGT_COMPOSE_DEFAULT``
466          equal to ``V4L2_SEL_TGT_CROP``.
467
468      ``V4L2_SEL_TGT_COMPOSE``
469          the rectangle inside a ``CAPTURE`` buffer into which the cropped
470          frame is written; defaults to ``V4L2_SEL_TGT_COMPOSE_DEFAULT``;
471          read-only on hardware without additional compose/scaling capabilities.
472
473      ``V4L2_SEL_TGT_COMPOSE_PADDED``
474          the rectangle inside a ``CAPTURE`` buffer which is overwritten by the
475          hardware; equal to ``V4L2_SEL_TGT_COMPOSE`` if the hardware does not
476          write padding pixels.
477
478    .. warning::
479
480       The values are guaranteed to be meaningful only after the decoder
481       successfully parses the stream metadata. The client must not rely on the
482       query before that happens.
483
4843.  **Optional.** Enumerate ``CAPTURE`` formats via :c:func:`VIDIOC_ENUM_FMT` on
485    the ``CAPTURE`` queue. Once the stream information is parsed and known, the
486    client may use this ioctl to discover which raw formats are supported for
487    given stream and select one of them via :c:func:`VIDIOC_S_FMT`.
488
489    .. important::
490
491       The decoder will return only formats supported for the currently
492       established coded format, as per the ``OUTPUT`` format and/or stream
493       metadata parsed in this initialization sequence, even if more formats
494       may be supported by the decoder in general. In other words, the set
495       returned will be a subset of the initial query mentioned in the
496       `Querying Capabilities` section.
497
498       For example, a decoder may support YUV and RGB formats for resolutions
499       1920x1088 and lower, but only YUV for higher resolutions (due to
500       hardware limitations). After parsing a resolution of 1920x1088 or lower,
501       :c:func:`VIDIOC_ENUM_FMT` may return a set of YUV and RGB pixel formats,
502       but after parsing resolution higher than 1920x1088, the decoder will not
503       return RGB, unsupported for this resolution.
504
505       However, subsequent resolution change event triggered after
506       discovering a resolution change within the same stream may switch
507       the stream into a lower resolution and :c:func:`VIDIOC_ENUM_FMT`
508       would return RGB formats again in that case.
509
5104.  **Optional.** Set the ``CAPTURE`` format via :c:func:`VIDIOC_S_FMT` on the
511    ``CAPTURE`` queue. The client may choose a different format than
512    selected/suggested by the decoder in :c:func:`VIDIOC_G_FMT`.
513
514    * **Required fields:**
515
516      ``type``
517          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
518
519      ``pixelformat``
520          a raw pixel format.
521
522      ``width``, ``height``
523         frame buffer resolution of the decoded stream; typically unchanged from
524	 what was returned with :c:func:`VIDIOC_G_FMT`, but it may be different
525	 if the hardware supports composition and/or scaling.
526
527   * Setting the ``CAPTURE`` format will reset the compose selection rectangles
528     to their default values, based on the new resolution, as described in the
529     previous step.
530
5315. **Optional.** Set the compose rectangle via :c:func:`VIDIOC_S_SELECTION` on
532   the ``CAPTURE`` queue if it is desired and if the decoder has compose and/or
533   scaling capabilities.
534
535   * **Required fields:**
536
537     ``type``
538         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
539
540     ``target``
541         set to ``V4L2_SEL_TGT_COMPOSE``.
542
543     ``r.left``, ``r.top``, ``r.width``, ``r.height``
544         the rectangle inside a ``CAPTURE`` buffer into which the cropped
545         frame is written; defaults to ``V4L2_SEL_TGT_COMPOSE_DEFAULT``;
546         read-only on hardware without additional compose/scaling capabilities.
547
548   * **Return fields:**
549
550     ``r.left``, ``r.top``, ``r.width``, ``r.height``
551         the visible rectangle; it must fit within the frame buffer resolution
552         returned by :c:func:`VIDIOC_G_FMT` on ``CAPTURE``.
553
554   .. warning::
555
556      The decoder may adjust the compose rectangle to the nearest
557      supported one to meet codec and hardware requirements. The client needs
558      to check the adjusted rectangle returned by :c:func:`VIDIOC_S_SELECTION`.
559
5606.  If all the following conditions are met, the client may resume the decoding
561    instantly:
562
563    * ``sizeimage`` of the new format (determined in previous steps) is less
564      than or equal to the size of currently allocated buffers,
565
566    * the number of buffers currently allocated is greater than or equal to the
567      minimum number of buffers acquired in previous steps. To fulfill this
568      requirement, the client may use :c:func:`VIDIOC_CREATE_BUFS` to add new
569      buffers.
570
571    In that case, the remaining steps do not apply and the client may resume
572    the decoding by one of the following actions:
573
574    * if the ``CAPTURE`` queue is streaming, call :c:func:`VIDIOC_DECODER_CMD`
575      with the ``V4L2_DEC_CMD_START`` command,
576
577    * if the ``CAPTURE`` queue is not streaming, call :c:func:`VIDIOC_STREAMON`
578      on the ``CAPTURE`` queue.
579
580    However, if the client intends to change the buffer set, to lower
581    memory usage or for any other reasons, it may be achieved by following
582    the steps below.
583
5847.  **If the** ``CAPTURE`` **queue is streaming,** keep queuing and dequeuing
585    buffers on the ``CAPTURE`` queue until a buffer marked with the
586    ``V4L2_BUF_FLAG_LAST`` flag is dequeued.
587
5888.  **If the** ``CAPTURE`` **queue is streaming,** call :c:func:`VIDIOC_STREAMOFF`
589    on the ``CAPTURE`` queue to stop streaming.
590
591    .. warning::
592
593       The ``OUTPUT`` queue must remain streaming. Calling
594       :c:func:`VIDIOC_STREAMOFF` on it would abort the sequence and trigger a
595       seek.
596
5979.  **If the** ``CAPTURE`` **queue has buffers allocated,** free the ``CAPTURE``
598    buffers using :c:func:`VIDIOC_REQBUFS`.
599
600    * **Required fields:**
601
602      ``count``
603          set to 0.
604
605      ``type``
606          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
607
608      ``memory``
609          follows standard semantics.
610
61110. Allocate ``CAPTURE`` buffers via :c:func:`VIDIOC_REQBUFS` on the
612    ``CAPTURE`` queue.
613
614    * **Required fields:**
615
616      ``count``
617          requested number of buffers to allocate; greater than zero.
618
619      ``type``
620          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
621
622      ``memory``
623          follows standard semantics.
624
625    * **Return fields:**
626
627      ``count``
628          actual number of buffers allocated.
629
630    .. warning::
631
632       The actual number of allocated buffers may differ from the ``count``
633       given. The client must check the updated value of ``count`` after the
634       call returns.
635
636    .. note::
637
638       To allocate more than the minimum number of buffers (for pipeline
639       depth), the client may query the ``V4L2_CID_MIN_BUFFERS_FOR_CAPTURE``
640       control to get the minimum number of buffers required, and pass the
641       obtained value plus the number of additional buffers needed in the
642       ``count`` field to :c:func:`VIDIOC_REQBUFS`.
643
644    Alternatively, :c:func:`VIDIOC_CREATE_BUFS` on the ``CAPTURE`` queue can be
645    used to have more control over buffer allocation. For example, by
646    allocating buffers larger than the current ``CAPTURE`` format, future
647    resolution changes can be accommodated.
648
649    * **Required fields:**
650
651      ``count``
652          requested number of buffers to allocate; greater than zero.
653
654      ``type``
655          a ``V4L2_BUF_TYPE_*`` enum appropriate for ``CAPTURE``.
656
657      ``memory``
658          follows standard semantics.
659
660      ``format``
661          a format representing the maximum framebuffer resolution to be
662          accommodated by newly allocated buffers.
663
664    * **Return fields:**
665
666      ``count``
667          adjusted to the number of allocated buffers.
668
669    .. warning::
670
671        The actual number of allocated buffers may differ from the ``count``
672        given. The client must check the updated value of ``count`` after the
673        call returns.
674
675    .. note::
676
677       To allocate buffers for a format different than parsed from the stream
678       metadata, the client must proceed as follows, before the metadata
679       parsing is initiated:
680
681       * set width and height of the ``OUTPUT`` format to desired coded resolution to
682         let the decoder configure the ``CAPTURE`` format appropriately,
683
684       * query the ``CAPTURE`` format using :c:func:`VIDIOC_G_FMT` and save it
685         until this step.
686
687       The format obtained in the query may be then used with
688       :c:func:`VIDIOC_CREATE_BUFS` in this step to allocate the buffers.
689
69011. Call :c:func:`VIDIOC_STREAMON` on the ``CAPTURE`` queue to start decoding
691    frames.
692
693Decoding
694========
695
696This state is reached after the `Capture Setup` sequence finishes successfully.
697In this state, the client queues and dequeues buffers to both queues via
698:c:func:`VIDIOC_QBUF` and :c:func:`VIDIOC_DQBUF`, following the standard
699semantics.
700
701The content of the source ``OUTPUT`` buffers depends on the active coded pixel
702format and may be affected by codec-specific extended controls, as stated in
703the documentation of each format.
704
705Both queues operate independently, following the standard behavior of V4L2
706buffer queues and memory-to-memory devices. In addition, the order of decoded
707frames dequeued from the ``CAPTURE`` queue may differ from the order of queuing
708coded frames to the ``OUTPUT`` queue, due to properties of the selected coded
709format, e.g. frame reordering.
710
711The client must not assume any direct relationship between ``CAPTURE``
712and ``OUTPUT`` buffers and any specific timing of buffers becoming
713available to dequeue. Specifically:
714
715* a buffer queued to ``OUTPUT`` may result in no buffers being produced
716  on ``CAPTURE`` (e.g. if it does not contain encoded data, or if only
717  metadata syntax structures are present in it),
718
719* a buffer queued to ``OUTPUT`` may result in more than one buffer produced
720  on ``CAPTURE`` (if the encoded data contained more than one frame, or if
721  returning a decoded frame allowed the decoder to return a frame that
722  preceded it in decode, but succeeded it in the display order),
723
724* a buffer queued to ``OUTPUT`` may result in a buffer being produced on
725  ``CAPTURE`` later into decode process, and/or after processing further
726  ``OUTPUT`` buffers, or be returned out of order, e.g. if display
727  reordering is used,
728
729* buffers may become available on the ``CAPTURE`` queue without additional
730  buffers queued to ``OUTPUT`` (e.g. during drain or ``EOS``), because of the
731  ``OUTPUT`` buffers queued in the past whose decoding results are only
732  available at later time, due to specifics of the decoding process.
733
734.. note::
735
736   To allow matching decoded ``CAPTURE`` buffers with ``OUTPUT`` buffers they
737   originated from, the client can set the ``timestamp`` field of the
738   :c:type:`v4l2_buffer` struct when queuing an ``OUTPUT`` buffer. The
739   ``CAPTURE`` buffer(s), which resulted from decoding that ``OUTPUT`` buffer
740   will have their ``timestamp`` field set to the same value when dequeued.
741
742   In addition to the straightforward case of one ``OUTPUT`` buffer producing
743   one ``CAPTURE`` buffer, the following cases are defined:
744
745   * one ``OUTPUT`` buffer generates multiple ``CAPTURE`` buffers: the same
746     ``OUTPUT`` timestamp will be copied to multiple ``CAPTURE`` buffers.
747
748   * multiple ``OUTPUT`` buffers generate one ``CAPTURE`` buffer: timestamp of
749     the ``OUTPUT`` buffer queued first will be copied.
750
751   * the decoding order differs from the display order (i.e. the ``CAPTURE``
752     buffers are out-of-order compared to the ``OUTPUT`` buffers): ``CAPTURE``
753     timestamps will not retain the order of ``OUTPUT`` timestamps.
754
755.. note::
756
757   The backing memory of ``CAPTURE`` buffers that are used as reference frames
758   by the stream may be read by the hardware even after they are dequeued.
759   Consequently, the client should avoid writing into this memory while the
760   ``CAPTURE`` queue is streaming. Failure to observe this may result in
761   corruption of decoded frames.
762
763   Similarly, when using a memory type other than ``V4L2_MEMORY_MMAP``, the
764   client should make sure that each ``CAPTURE`` buffer is always queued with
765   the same backing memory for as long as the ``CAPTURE`` queue is streaming.
766   The reason for this is that V4L2 buffer indices can be used by drivers to
767   identify frames. Thus, if the backing memory of a reference frame is
768   submitted under a different buffer ID, the driver may misidentify it and
769   decode a new frame into it while it is still in use, resulting in corruption
770   of the following frames.
771
772During the decoding, the decoder may initiate one of the special sequences, as
773listed below. The sequences will result in the decoder returning all the
774``CAPTURE`` buffers that originated from all the ``OUTPUT`` buffers processed
775before the sequence started. Last of the buffers will have the
776``V4L2_BUF_FLAG_LAST`` flag set. To determine the sequence to follow, the client
777must check if there is any pending event and:
778
779* if a ``V4L2_EVENT_SOURCE_CHANGE`` event with ``changes`` set to
780  ``V4L2_EVENT_SRC_CH_RESOLUTION`` is pending, the `Dynamic Resolution
781  Change` sequence needs to be followed,
782
783* if a ``V4L2_EVENT_EOS`` event is pending, the `End of Stream` sequence needs
784  to be followed.
785
786Some of the sequences can be intermixed with each other and need to be handled
787as they happen. The exact operation is documented for each sequence.
788
789Should a decoding error occur, it will be reported to the client with the level
790of details depending on the decoder capabilities. Specifically:
791
792* the CAPTURE buffer that contains the results of the failed decode operation
793  will be returned with the V4L2_BUF_FLAG_ERROR flag set,
794
795* if the decoder is able to precisely report the OUTPUT buffer that triggered
796  the error, such buffer will be returned with the V4L2_BUF_FLAG_ERROR flag
797  set.
798
799In case of a fatal failure that does not allow the decoding to continue, any
800further operations on corresponding decoder file handle will return the -EIO
801error code. The client may close the file handle and open a new one, or
802alternatively reinitialize the instance by stopping streaming on both queues,
803releasing all buffers and performing the Initialization sequence again.
804
805Seek
806====
807
808Seek is controlled by the ``OUTPUT`` queue, as it is the source of coded data.
809The seek does not require any specific operation on the ``CAPTURE`` queue, but
810it may be affected as per normal decoder operation.
811
8121. Stop the ``OUTPUT`` queue to begin the seek sequence via
813   :c:func:`VIDIOC_STREAMOFF`.
814
815   * **Required fields:**
816
817     ``type``
818         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``.
819
820   * The decoder will drop all the pending ``OUTPUT`` buffers and they must be
821     treated as returned to the client (following standard semantics).
822
8232. Restart the ``OUTPUT`` queue via :c:func:`VIDIOC_STREAMON`.
824
825   * **Required fields:**
826
827     ``type``
828         a ``V4L2_BUF_TYPE_*`` enum appropriate for ``OUTPUT``.
829
830   * The decoder will start accepting new source bytestream buffers after the
831     call returns.
832
8333. Start queuing buffers containing coded data after the seek to the ``OUTPUT``
834   queue until a suitable resume point is found.
835
836   .. note::
837
838      There is no requirement to begin queuing coded data starting exactly
839      from a resume point (e.g. SPS or a keyframe). Any queued ``OUTPUT``
840      buffers will be processed and returned to the client until a suitable
841      resume point is found.  While looking for a resume point, the decoder
842      should not produce any decoded frames into ``CAPTURE`` buffers.
843
844      Some hardware is known to mishandle seeks to a non-resume point. Such an
845      operation may result in an unspecified number of corrupted decoded frames
846      being made available on the ``CAPTURE`` queue. Drivers must ensure that
847      no fatal decoding errors or crashes occur, and implement any necessary
848      handling and workarounds for hardware issues related to seek operations.
849
850   .. warning::
851
852      In case of the H.264/HEVC codec, the client must take care not to seek
853      over a change of SPS/PPS. Even though the target frame could be a
854      keyframe, the stale SPS/PPS inside decoder state would lead to undefined
855      results when decoding. Although the decoder must handle that case without
856      a crash or a fatal decode error, the client must not expect a sensible
857      decode output.
858
859      If the hardware can detect such corrupted decoded frames, then
860      corresponding buffers will be returned to the client with the
861      V4L2_BUF_FLAG_ERROR set. See the `Decoding` section for further
862      description of decode error reporting.
863
8644. After a resume point is found, the decoder will start returning ``CAPTURE``
865   buffers containing decoded frames.
866
867.. important::
868
869   A seek may result in the `Dynamic Resolution Change` sequence being
870   initiated, due to the seek target having decoding parameters different from
871   the part of the stream decoded before the seek. The sequence must be handled
872   as per normal decoder operation.
873
874.. warning::
875
876   It is not specified when the ``CAPTURE`` queue starts producing buffers
877   containing decoded data from the ``OUTPUT`` buffers queued after the seek,
878   as it operates independently from the ``OUTPUT`` queue.
879
880   The decoder may return a number of remaining ``CAPTURE`` buffers containing
881   decoded frames originating from the ``OUTPUT`` buffers queued before the
882   seek sequence is performed.
883
884   The ``VIDIOC_STREAMOFF`` operation discards any remaining queued
885   ``OUTPUT`` buffers, which means that not all of the ``OUTPUT`` buffers
886   queued before the seek sequence may have matching ``CAPTURE`` buffers
887   produced.  For example, given the sequence of operations on the
888   ``OUTPUT`` queue:
889
890     QBUF(A), QBUF(B), STREAMOFF(), STREAMON(), QBUF(G), QBUF(H),
891
892   any of the following results on the ``CAPTURE`` queue is allowed:
893
894     {A', B', G', H'}, {A', G', H'}, {G', H'}.
895
896   To determine the CAPTURE buffer containing the first decoded frame after the
897   seek, the client may observe the timestamps to match the CAPTURE and OUTPUT
898   buffers or use V4L2_DEC_CMD_STOP and V4L2_DEC_CMD_START to drain the
899   decoder.
900
901.. note::
902
903   To achieve instantaneous seek, the client may restart streaming on the
904   ``CAPTURE`` queue too to discard decoded, but not yet dequeued buffers.
905
906Dynamic Resolution Change
907=========================
908
909Streams that include resolution metadata in the bytestream may require switching
910to a different resolution during the decoding.
911
912.. note::
913
914   Not all decoders can detect resolution changes. Those that do set the
915   ``V4L2_FMT_FLAG_DYN_RESOLUTION`` flag for the coded format when
916   :c:func:`VIDIOC_ENUM_FMT` is called.
917
918The sequence starts when the decoder detects a coded frame with one or more of
919the following parameters different from those previously established (and
920reflected by corresponding queries):
921
922* coded resolution (``OUTPUT`` width and height),
923
924* visible resolution (selection rectangles),
925
926* the minimum number of buffers needed for decoding,
927
928* bit-depth of the bitstream has been changed.
929
930Whenever that happens, the decoder must proceed as follows:
931
9321.  After encountering a resolution change in the stream, the decoder sends a
933    ``V4L2_EVENT_SOURCE_CHANGE`` event with ``changes`` set to
934    ``V4L2_EVENT_SRC_CH_RESOLUTION``.
935
936    .. important::
937
938       Any client query issued after the decoder queues the event will return
939       values applying to the stream after the resolution change, including
940       queue formats, selection rectangles and controls.
941
9422.  The decoder will then process and decode all remaining buffers from before
943    the resolution change point.
944
945    * The last buffer from before the change must be marked with the
946      ``V4L2_BUF_FLAG_LAST`` flag, similarly to the `Drain` sequence above.
947
948    .. warning::
949
950       The last buffer may be empty (with :c:type:`v4l2_buffer` ``bytesused``
951       = 0) and in that case it must be ignored by the client, as it does not
952       contain a decoded frame.
953
954    .. note::
955
956       Any attempt to dequeue more ``CAPTURE`` buffers beyond the buffer marked
957       with ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
958       :c:func:`VIDIOC_DQBUF`.
959
960The client must continue the sequence as described below to continue the
961decoding process.
962
9631.  Dequeue the source change event.
964
965    .. important::
966
967       A source change triggers an implicit decoder drain, similar to the
968       explicit `Drain` sequence. The decoder is stopped after it completes.
969       The decoding process must be resumed with either a pair of calls to
970       :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
971       ``CAPTURE`` queue, or a call to :c:func:`VIDIOC_DECODER_CMD` with the
972       ``V4L2_DEC_CMD_START`` command.
973
9742.  Continue with the `Capture Setup` sequence.
975
976.. note::
977
978   During the resolution change sequence, the ``OUTPUT`` queue must remain
979   streaming. Calling :c:func:`VIDIOC_STREAMOFF` on the ``OUTPUT`` queue would
980   abort the sequence and initiate a seek.
981
982   In principle, the ``OUTPUT`` queue operates separately from the ``CAPTURE``
983   queue and this remains true for the duration of the entire resolution change
984   sequence as well.
985
986   The client should, for best performance and simplicity, keep queuing/dequeuing
987   buffers to/from the ``OUTPUT`` queue even while processing this sequence.
988
989Drain
990=====
991
992To ensure that all queued ``OUTPUT`` buffers have been processed and related
993``CAPTURE`` buffers are given to the client, the client must follow the drain
994sequence described below. After the drain sequence ends, the client has
995received all decoded frames for all ``OUTPUT`` buffers queued before the
996sequence was started.
997
9981. Begin drain by issuing :c:func:`VIDIOC_DECODER_CMD`.
999
1000   * **Required fields:**
1001
1002     ``cmd``
1003         set to ``V4L2_DEC_CMD_STOP``.
1004
1005     ``flags``
1006         set to 0.
1007
1008     ``pts``
1009         set to 0.
1010
1011   .. warning::
1012
1013      The sequence can be only initiated if both ``OUTPUT`` and ``CAPTURE``
1014      queues are streaming. For compatibility reasons, the call to
1015      :c:func:`VIDIOC_DECODER_CMD` will not fail even if any of the queues is
1016      not streaming, but at the same time it will not initiate the `Drain`
1017      sequence and so the steps described below would not be applicable.
1018
10192. Any ``OUTPUT`` buffers queued by the client before the
1020   :c:func:`VIDIOC_DECODER_CMD` was issued will be processed and decoded as
1021   normal. The client must continue to handle both queues independently,
1022   similarly to normal decode operation. This includes:
1023
1024   * handling any operations triggered as a result of processing those buffers,
1025     such as the `Dynamic Resolution Change` sequence, before continuing with
1026     the drain sequence,
1027
1028   * queuing and dequeuing ``CAPTURE`` buffers, until a buffer marked with the
1029     ``V4L2_BUF_FLAG_LAST`` flag is dequeued,
1030
1031     .. warning::
1032
1033        The last buffer may be empty (with :c:type:`v4l2_buffer`
1034        ``bytesused`` = 0) and in that case it must be ignored by the client,
1035        as it does not contain a decoded frame.
1036
1037     .. note::
1038
1039        Any attempt to dequeue more ``CAPTURE`` buffers beyond the buffer
1040        marked with ``V4L2_BUF_FLAG_LAST`` will result in a -EPIPE error from
1041        :c:func:`VIDIOC_DQBUF`.
1042
1043   * dequeuing processed ``OUTPUT`` buffers, until all the buffers queued
1044     before the ``V4L2_DEC_CMD_STOP`` command are dequeued,
1045
1046   * dequeuing the ``V4L2_EVENT_EOS`` event, if the client subscribed to it.
1047
1048   .. note::
1049
1050      For backwards compatibility, the decoder will signal a ``V4L2_EVENT_EOS``
1051      event when the last frame has been decoded and all frames are ready to be
1052      dequeued. It is a deprecated behavior and the client must not rely on it.
1053      The ``V4L2_BUF_FLAG_LAST`` buffer flag should be used instead.
1054
10553. Once all the ``OUTPUT`` buffers queued before the ``V4L2_DEC_CMD_STOP`` call
1056   are dequeued and the last ``CAPTURE`` buffer is dequeued, the decoder is
1057   stopped and it will accept, but not process, any newly queued ``OUTPUT``
1058   buffers until the client issues any of the following operations:
1059
1060   * ``V4L2_DEC_CMD_START`` - the decoder will not be reset and will resume
1061     operation normally, with all the state from before the drain,
1062
1063   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
1064     ``CAPTURE`` queue - the decoder will resume the operation normally,
1065     however any ``CAPTURE`` buffers still in the queue will be returned to the
1066     client,
1067
1068   * a pair of :c:func:`VIDIOC_STREAMOFF` and :c:func:`VIDIOC_STREAMON` on the
1069     ``OUTPUT`` queue - any pending source buffers will be returned to the
1070     client and the `Seek` sequence will be triggered.
1071
1072.. note::
1073
1074   Once the drain sequence is initiated, the client needs to drive it to
1075   completion, as described by the steps above, unless it aborts the process by
1076   issuing :c:func:`VIDIOC_STREAMOFF` on any of the ``OUTPUT`` or ``CAPTURE``
1077   queues.  The client is not allowed to issue ``V4L2_DEC_CMD_START`` or
1078   ``V4L2_DEC_CMD_STOP`` again while the drain sequence is in progress and they
1079   will fail with -EBUSY error code if attempted.
1080
1081   Although not mandatory, the availability of decoder commands may be queried
1082   using :c:func:`VIDIOC_TRY_DECODER_CMD`.
1083
1084End of Stream
1085=============
1086
1087If the decoder encounters an end of stream marking in the stream, the decoder
1088will initiate the `Drain` sequence, which the client must handle as described
1089above, skipping the initial :c:func:`VIDIOC_DECODER_CMD`.
1090
1091Commit Points
1092=============
1093
1094Setting formats and allocating buffers trigger changes in the behavior of the
1095decoder.
1096
10971. Setting the format on the ``OUTPUT`` queue may change the set of formats
1098   supported/advertised on the ``CAPTURE`` queue. In particular, it also means
1099   that the ``CAPTURE`` format may be reset and the client must not rely on the
1100   previously set format being preserved.
1101
11022. Enumerating formats on the ``CAPTURE`` queue always returns only formats
1103   supported for the current ``OUTPUT`` format.
1104
11053. Setting the format on the ``CAPTURE`` queue does not change the list of
1106   formats available on the ``OUTPUT`` queue. An attempt to set a ``CAPTURE``
1107   format that is not supported for the currently selected ``OUTPUT`` format
1108   will result in the decoder adjusting the requested ``CAPTURE`` format to a
1109   supported one.
1110
11114. Enumerating formats on the ``OUTPUT`` queue always returns the full set of
1112   supported coded formats, irrespectively of the current ``CAPTURE`` format.
1113
11145. While buffers are allocated on any of the ``OUTPUT`` or ``CAPTURE`` queues,
1115   the client must not change the format on the ``OUTPUT`` queue. Drivers will
1116   return the -EBUSY error code for any such format change attempt.
1117
1118To summarize, setting formats and allocation must always start with the
1119``OUTPUT`` queue and the ``OUTPUT`` queue is the master that governs the
1120set of supported formats for the ``CAPTURE`` queue.
1121