Add a V4L2 Request API hwaccel for VP8.
Co-developed-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Ezequiel Garcia <ezequiel@collabora.com>
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Co-developed-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add a V4L2 Request API hwaccel for HEVC, supporting both slice and
frame decoding modes.
Support for HEVC is enabled when Linux kernel headers declare the
control id V4L2_CID_STATELESS_HEVC_SPS, added in v6.0.
Co-developed-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Signed-off-by: Benjamin Gaignard <benjamin.gaignard@collabora.com>
Co-developed-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Co-developed-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add a V4L2 Request API hwaccel for H.264, supporting both slice and
frame decoding modes.
Support for H.264 is enabled when Linux kernel headers declare the
control id V4L2_CID_STATELESS_H264_DECODE_MODE, added in v5.11.
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Co-developed-by: Jonas Karlman <jonas@kwiboo.se>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
The V4L2_CID_STATELESS_H264_DECODE_PARAMS control require following:
- dec_ref_pic_marking_bit_size
Size in bits of the dec_ref_pic_marking() syntax element.
- pic_order_cnt_bit_size
Combined size in bits of the picture order count related syntax
elements: pic_order_cnt_lsb, delta_pic_order_cnt_bottom,
delta_pic_order_cnt0, and delta_pic_order_cnt1.
Save the bit sizes while parsing for later use in hwaccel, similar to
short/long_term_ref_pic_set_size in hevcdec.
Signed-off-by: Boris Brezillon <boris.brezillon@collabora.com>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add a V4L2 Request API hwaccel for MPEG2.
Support for MPEG2 is enabled when Linux kernel headers declare the
control id V4L2_CID_STATELESS_MPEG2_SEQUENCE, added in v5.14.
This also change v4l2_request hwaccel to use autodetect in configure.
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add common support for decoding using the V4L2 Request API.
Basic flow for decoding follow the kernel Memory-to-memory Stateless
Video Decoder Interface > Decoding [1].
A codec hwaccel typically handle decoding as follow:
In start_frame next OUTPUT buffer and its related request object is
picked from a circular queue and any codec specific CONTROLs is prepared.
In decode_slice the slice bitstream data is appended to the OUTPUT
buffer.
In end_frame a CAPTURE buffer tied to the AVFrame is queued, it will be
used as the decoding target by the driver / hw decoder. The prepared
codec specific CONTROLs get queued as part of the request object.
Finally the request object is submitted for decoding.
For slice based hw decoders only the request for the final slice of the
frame is submitted in end_frame, remaining is submitted in decode_slice.
[1] https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/dev-stateless-decoder.html#decoding
Co-developed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Co-developed-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Probe all media devices and its linked video devices to locate a video
device that support stateless decoding of the specific codec using the
V4L2 Request API.
When AVV4L2RequestDeviceContext.media_fd is a valid file descriptor only
video devices for the opened media device is probed.
E.g. using a -init_hw_device v4l2request:/dev/media1 parameter.
A video device is deemed capable when all tests pass, e.g. kernel driver
support the codec, FFmpeg v4l2-request code know about the buffer format
and kernel driver report that the frame size is supported.
A codec specific hook, post_probe, is supported to e.g. test for codec
specific limitations, PROFILE and LEVEL controls.
Basic flow for initialization follow the kernel Memory-to-memory
Stateless Video Decoder Interface > Initialization [1].
[1] https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/dev-stateless-decoder.html#initialization
Co-developed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Co-developed-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add common helpers for supporting V4L2 Request API hwaccels.
Basic flow for initialization follow the kernel Memory-to-memory
Stateless Video Decoder Interface > Initialization [1].
In init video devices is probed and when a capable device is found it is
initialized for decoding. Codec specific CONTROLs may be set to assist
kernel driver. A supported CAPTURE format is selected. A hw frame ctx is
created and CAPTURE buffers is allocated. OUTPUT buffers and request
objects for a circular queue is also allocated.
In frame_params a buffer pool is created and configured to allocate
CAPTURE buffers.
In flush all queued OUTPUT and CAPTURE buffers is dequeued.
In uninit any pending request is flushed, also video and media device is
closed. The media device is not closed when ownership of the media_fd
has been transfered to the hwdevice.
[1] https://www.kernel.org/doc/html/latest/userspace-api/media/v4l/dev-stateless-decoder.html#initialization
Co-developed-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Signed-off-by: Jernej Skrabec <jernej.skrabec@gmail.com>
Co-developed-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Alex Bee <knaerzche@gmail.com>
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Add a hwdevice type for V4L2 Request API with transfer_data_from support
for AV_PIX_FMT_DRM_PRIME, based on AV_HWDEVICE_TYPE_DRM.
AVV4L2RequestDeviceContext.media_fd can be set by the application or a
media device path can be supplied when hwdevice is created. When none
is supplied it default to -1 and hwaccel will auto-detect a media device
with a capable video device.
Signed-off-by: Jonas Karlman <jonas@kwiboo.se>
Fixes: Ticket5405
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
(cherry picked from commit a9c83e43f2)
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
The unaligned width test cases fail on i386; we have an assembly
function of rgb24toyv12 which is enabled only within
"#if ARCH_X86_32 && HAVE_7REGS", which seems to fail these new
test cases for unaligned widths.
As that assembly function has existed for a long time in that form,
the issue probably isn't very recent, thus skip testing these cases
for now.
Once the assembly function has been fixed, these test cases can
be readded.
Signed-off-by: Martin Storsjö <martin@martin.st>
(cherry picked from commit 157ce21939)
Signed-off-by: Martin Storsjö <martin@martin.st>
In case of errors in this function, the allocated context
was not properly freed in several cases.
Signed-off-by: Marton Balint <cus@passwd.hu>
(cherry picked from commit 8e76c99301)
In the early return when both draw_bars and signal_loss_action
options are used, the context allocated previously was not
properly freed.
Introduced in 9bcb86b0fa
Fixes CID1619296
Signed-off-by: Marton Balint <cus@passwd.hu>
(cherry picked from commit 7554346cb5)
An error from read_number() is non-fatal here and should not be
forwarded.
(cherry picked from commit 505fea34e5)
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This was broken in 2fb2cd5c79 and
74553f0026, making a simple make fate-mov fail
and most of the test not run with a simple make fate.
Signed-off-by: James Almer <jamrial@gmail.com>
(cherry picked from commit 8b4e32f30b)
Fixes: OOM and Timeout
Fixes: 71254/clusterfuzz-testcase-minimized-ffmpeg_SWR_fuzzer-5941896977907712
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: -1.80923e+19 is outside the range of representable values of type 'long'
Fixes: 71103/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_SNOW_fuzzer-6542773681979392
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: shift exponent 62 is too large for 32-bit type 'int'
Fixes: 71020/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VVC_fuzzer-6444916325023744
Fixes: 71285/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VVC_fuzzer-4761971281428480
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 1205469696 + 1901074655 cannot be represented in type 'int'
Fixes: 70773/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-5419594888577024
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 856827136 + 2123580416 cannot be represented in type 'int'
Fixes: 70772/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PRORES_KS_fuzzer-5180569961431040
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
