The Musepack decoder uses static VLC tables to parse the bitstream.
There are 14 different quant tables VLCs and each of them has a varying
number of codes. The maximum number is 63, the average number is 25.3.
Up until now, the array containing the raw data was of type
uint16_t [7][2][64 * 2] (the 14 tables come in pairs of two, hence [7][2]
instead of [14]) and from this it follows that there were large gaps in
said array. This commit changes this by making it a continuous array
instead. Doing so saves about 2KB.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
The MobiClip decoder uses adjacent pixels for prediction; yet when
accessing the left pixel, it was forgotten to clip the x coordinate.
This results in an heap-buffer-overflow. It can e.g. be reproduced with
the sample from https://samples.ffmpeg.org/V-codecs/MOHD/crap.avi when
forcing the video decoder to mobiclip.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
If an error happens during init after an allocation has succeeded,
the already allocated data leaked up until now. Fix this by setting the
FF_CODEC_CAP_INIT_CLEANUP flag.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avcodec/indeo5: Fix memleaks upon allocation error
ff_ivi_init_planes() might error out after having allocated some arrays.
Set the FF_CODEC_CAP_INIT_CLEANUP flag in order to free these arrays in
this case.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
If allocating the tiles array for indeo 4/5 fails, the context is in an
inconsistent state, because the counter for the number of tiles is > 0.
This will lead to a segfault when freeing the tiles' substructures.
Fix this by setting the number of tiles to zero if the allocation was
unsuccessful.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
If an error happens during init after an allocation has succeeded,
the already allocated data leaked up until now. Fix this by setting the
FF_CODEC_CAP_INIT_CLEANUP flag.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avcodec/magicyuvenc: Fix memleak upon init failure
If an error happens during init after an allocation has succeeded,
the already allocated data leaked up until now. Fix this by setting the
FF_CODEC_CAP_INIT_CLEANUP flag.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
If one of several allocations the gif encoder performs in its init
function fails, the successful allocations leak. Fix this by adding the
FF_CODEC_CAP_INIT_CLEANUP flag.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avcodec/dsicinvideo: Remove redundant code for freeing
The dsicinvideo decoder already has the FF_CODEC_CAP_INIT_CLEANUP flag
set, so it is unnecessary to directly clean up some already allocated
buffers in case another one could not be allocated in the init function,
as all buffers will be freed anyway later in the decoder's close
function.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
James Almer [Mon, 17 Aug 2020 15:06:55 +0000 (12:06 -0300)]
avcodec/decode: use a packet list to store packet properties
Keeping only the latest packet fed to the decoder works only for decoders that
return a frame immediately after every consumed packet. Decoders that consume
several packets before they return a frame will fill said frame with properties
taken from the last consumed packet instead of the earliest.
James Almer [Mon, 17 Aug 2020 15:03:50 +0000 (12:03 -0300)]
avcodec/packet: move AVPacketList definition and function helpers over from libavformat
And replace the flags parameter with a function callback that can be used to
copy the contents of the packet (e.g, av_packet_ref and av_packet_copy_props).
The CineForm HD encoder attempts to allocate several buffers in its init
function; yet if only some of these allocations succeed, the
successfully allocated buffers leak. This is fixed by setting the
FF_CODEC_CAP_INIT_CLEANUP flag.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avutil/timecode: fix av_timecode_get_smpte_from_framenum with 50/60 fps
SMPTE 12M timecode can only count frames up to 39, because the tens-of-frames
value is stored in 2 bit. In order to resolve this 50/60 fps SMPTE timecode is
using the field bit (which is the same bit as the phase correction bit) to
signal the least significant bit of a 50/60 fps timecode. See SMPTE ST
12-1:2014 section 12.1.
Therefore we slightly change the format of the return value of
av_timecode_get_smpte_from_framenum and AV_FRAME_DATA_S12M_TIMECODE and start
using the previously unused Phase Correction bit as Field bit. (As the SMPTE
standard suggests)
We add 50/60 fps support to av_timecode_get_smpte_from_framenum by calling the
recently added av_timecode_get_smpte function in it which already handles this
properly.
This change affects the decklink indev and the DV and MXF muxers. MXF has no
fate test for 50/60fps content, DV does, therefore the changes.
MediaInfo (a recent version) confirms that half-frame timecode must be inserted
to DV. MXFInspect confirms valid timecode insertion to the System Item of MXF
files. For MXF, also see EBU R122.
Note that for DV the field flag is not used because in the HDV specs (SMPTE
370M) it is still defined as biphase mark polarity correction flag. So it
should not matter that the DV muxer overrides the field bit.
avformat/cdg: Fix integer overflow in duration computation
Fixes: signed integer overflow: 8398407 * 300 cannot be represented in type 'int' Fixes: 23914/clusterfuzz-testcase-minimized-ffmpeg_DEMUXER_fuzzer-4702539290509312 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
James Almer [Sat, 12 Sep 2020 20:39:45 +0000 (17:39 -0300)]
avcodec/av1dec: fix setting pix_fmt
Fill the array with the software pix_fmt and move the avctx->hwaccel
check back to the proper place.
Also remove the avoid probing flag to ensure an external av1 decoder
will not set a pix_fmt we don't want during format probing.
Fei Wang [Wed, 9 Sep 2020 03:39:55 +0000 (11:39 +0800)]
avcodec: add AV1 hardware accelerated decoder
This AV1 decoder is currently only used for hardware accelerated decoding.
It can be extended into a native decoder in the future, so set its name to
"av1" and temporarily give it the lowest priority in the codec list.
Signed-off-by: Fei Wang <fei.w.wang@intel.com> Signed-off-by: James Almer <jamrial@gmail.com>
avcodec/mpc: Fix multiple numerical overflows in ff_mpc_dequantize_and_synth()
Fixes: -2.4187e+09 is outside the range of representable values of type 'int' Fixes: signed integer overflow: -14512205 + -2147483648 cannot be represented in type 'int' Fixes: 20492/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MPC7_fuzzer-5747263166480384 Fixes: 23528/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MPC7_fuzzer-5747263166480384 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/mobiclip: Check quantizer before table setup
Fixes: index -1 out of bounds for type 'const uint8_t [6][16]' Fixes: out of array read Fixes: shift exponent -21 is negative Fixes: 25422/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MOBICLIP_fuzzer-5748258226569216 Fixes: shift exponent 8039082 is too large for 32-bit type 'int' Fixes: 25430/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MOBICLIP_fuzzer-5698567770210304 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Ting Fu [Wed, 9 Sep 2020 01:52:18 +0000 (09:52 +0800)]
dnn/openvino: support run inference via GPU
for enabling OpenVINO GPU please:
1. install required OpenCL drivers, see: https://github.com/intel/compute-runtime/releases/tag/19.41.14441
2. build OpenVINO c lib with GPU enabled: use cmake config with: -DENABLE_CLDNN=ON
3. then make, and include the OpenVINO c lib in environment variables
detailed steps please refer: https://github.com/openvinotoolkit/openvino/blob/master/build-instruction.md
inference model with GPU please add: optioins=device=GPU
Signed-off-by: Ting Fu <ting.fu@intel.com> Signed-off-by: Guo, Yejun <yejun.guo@intel.com>
ff_formats_ref() takes a pointer to an AVFilterFormats and a pointer to
a pointer to an AVFilterFormats as arguments and adds the latter as an
owner to the list pointed to by the former; the latter is hereby always
the address of a list contained in an AVFilterFormatsConfig and can
therefore not be NULL. So remove the check for whether it is NULL; also
do the same for ff_channel_layouts_ref().
Also do the same for the unref functions where argument is never NULL
because it is always the address of an existing lvalue.
Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avfilter/formats: Don't typedef structure a second time
Before commit 2f76476549a01ae2c4ec2a440e4b14c5aba64869, avfilter.h
contained no typedef for AVFilterChannelLayouts; all references to it
were done using its struct tag. formats.h meanwhile contained the
definition of the struct and a typedef for it. Said commit now added a
typedef in avfilter.h, too, bringing it in line with AVFilterFormats;
yet this means that there are two typedefs for AVFilterChannelLayouts
(in contrast to AVFilterFormats which is only typedef'ed in avfilter.h).
The problem is that older versions of GCC don't like this and error out:
http://fate.ffmpeg.org/history.cgi?slot=x86_64-openbsd5.6-gcc4.2-conf2
is one of the FATE boxes that now fail to compile. So just remove the
typedef in formats.h.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
fftools, libavcodec, libavfilter: Add const to some AVCodec *
The user has no business modifying the underlying AVCodec.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
Reviewed-by: Nicolas George <george@nsup.org> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avfilter/vf_subtitles: Remove unnecessary initialization of AVPacket
av_read_frame() can handle uninitialized packets.
Reviewed-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
init_subtitles() sometimes returned directly upon error without cleaning
up after itself. The easiest way to trigger this is by using
picture-based subtitles; it is also possible to run into this in case of
missing decoders or allocation failures.
Furthermore, return the proper error code in case of missing decoder.
Reviewed-by: Nicolas George <george@nsup.org> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avfilter/lavfutils: Fix memleak when avformat_find_stream_info() fails
Reviewed-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avformat/utils: Avoid duplicating extradata from extract_extradata BSF
Instead move the extradata contained in packet side-data to its
destination. This is possible because the side data already has zeroed
padding.
Notice that the check for FF_MAX_EXTRADATA_SIZE has been dropped,
because said constant is from libavcodec/internal.h. If libavcodec
wanted to enforce this, it should do so in the extract_extradata BSF
instead.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
If a sequence header has already been found, it is certain that next
startcode (being disjoint from the sequence header startcode) can begin
at index four at the earliest.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avcodec/agm: Fix off by 1 error in decode_inter_plane()
Fixes: Regression since 1f2096945709a32315da740691b5716da55893c3 Found-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
The movie and amovie filters currently use two packets. One of the two,
pkt0, is the owner of the returned packet; it is also the destination
packet for av_read_frame(). The other one pkt is initially (i.e. after
av_read_frame()) a copy of pkt0; copy means that the contents of both
are absolutely the same: They both point to the same AVBufferRef and the
same side data. This violation of the refcounted packet API is only
possible because pkt is not considered to own its data. Only pkt0 is
ever unreferenced.
The reason for pkt's existence seems to be historic:
The API used for decoding audio (namely avcodec_decode_audio4()) could
consume frames partially, i.e. it could return multiple frames for one
packet and therefore it returned how much of the input buffer had been
consumed. The caller was then supposed to update the packet's data and
size pointer to reflect this and call avcodec_decode_audio4() again with
the updated packet to get the next frame.
But before the introduction of refcounted AVPackets where knowledge and
responsibility about what to free lies with the underlying AVBuffer such
a procedure required a spare packet (or one would need to record the
original data and size fields separately to restore them before freeing
the packet; notice that this code has been written when AVPackets still
had a destruct field). But these times are long gone, so just remove the
secondary AVPacket.
Reviewed-by: Nicolas George <george@nsup.org> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>