mpeg4_unpack_bframes: Avoid allocations and copies of packet structures
1. Since bd90a2ec, mpeg4_unpack_bframes caches whole packets instead of
just the pointer to the buffer and the buffer's size in order to be able
to make use of refcounting to avoid copying of data; this unfortunately
introduced copies of packet structures and side data (if existing),
although the only fields that are needed are the buffer-related ones
(data, size and buf). This can be changed without compromising the
advantages of refcounting by storing a reference to the buffer.
2. This change also makes it easy to use only one packet throughout
so that an allocation and free of an AVPacket structure per filtered
packet can be saved by switching to ff_bsf_get_packet_ref.
3. Furthermore, this commit also fixes a memleak introduced in bd90a2ec:
If a stored b_frame with side data was used for a later frame, the side
data would leak when the input frame's properties were copied into the
output frame.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 1092624416 * 2 cannot be represented in type 'int' Fixes: 18045/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_BINKAUDIO_RDFT_fuzzer-5718519492116480 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Reviewed-by: Paul B Mahol <onemda@gmail.com> Reviewed-by: Peter Ross <pross@xvid.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
If there is an error in mpeg_mux_init() (the write_header function of
the various MPEG-PS muxers), two things might happen:
1. Several fifos might leak. Instead of freeing them, the goto fail part
of the functions freed the private data of the AVStreams instead,
although this will be freed later in free_stream() anyway.
2. And if the function is exited via goto fail, it automatically
returned AVERROR(ENOMEM), although this is also used when the error is
not a memory allocation failure.
Both of these issues happened in ticket #8284 and have been fixed.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Reviewed-by: Paul B Mahol <onemda@gmail.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
vp9_superframe: Avoid allocations and copies of packet structures
Up until now, the vp9_superframe bsf used distinct packets for input and
output. But at no point in the bsf are the input and output packets used
at the same time (except during a call to av_packet_move_ref()), so that
one can avoid using two packets if one switches to ff_bsf_get_packet_ref
at the same time. This also saves one malloc+free of an AVPacket
structure per filtered packet.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com> Signed-off-by: James Almer <jamrial@gmail.com>
remove_doubling() returned a const float, but returning qualified types
is pointless anyway. This is all the more important given that GCC 4.2.1
(and maybe other compilers, too) seems to treat this as an error as several
FATE units that don't compile any more show.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@gmail.com>
avcodec/sbcdec: Fix integer overflows in sbc_synthesize_eight()
Fixes: signed integer overflow: 518484152 + 1868182638 cannot be represented in type 'int' Fixes: 17732/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_SBC_fuzzer-5663738132168704 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/adpcm: Check initial predictor for ADPCM_IMA_EA_EACS
Fixes: signed integer overflow: -2147483360 - 631 cannot be represented in type 'int' Fixes: 17701/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ADPCM_IMA_EA_EACS_fuzzer-5711517319692288 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 1008 is too large for 32-bit type 'int' Fixes: 17700/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_G723_1_fuzzer-5707633436131328 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/apedec: Fix integer overflow in predictor_update_3930()
Fixes: signed integer overflow: -69555262 * 31 cannot be represented in type 'int' Fixes: 17698/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_APE_fuzzer-5728970447781888 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: index -49 out of bounds for type 'int16_t [192]' Fixes: 17689/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ACELP_KELVIN_fuzzer-5756275014500352 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: left shift of negative value -12 Fixes: 17689/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ACELP_KELVIN_fuzzer-5756275014500352 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: left shift of negative value -30635 Fixes: 17689/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ACELP_KELVIN_fuzzer-5756275014500352 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/adpcm: Fix left shifts in AV_CODEC_ID_ADPCM_EA
Fixes: left shift of negative value -1 Fixes: 17683/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ADPCM_EA_R2_fuzzer-5111690013704192 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Andriy Gelman [Sun, 6 Oct 2019 05:49:49 +0000 (01:49 -0400)]
avformat/chromaprint: Fix writing raw fingerprint
The pointer fp after the call to chromaprint_get_raw_fingerpoint() points to an array
of uint32_t whereas the current code assumed just a char stream. Thus when writing the
raw fingerprint, the output would be truncated by a factor of 4.
For reference the declaration of the function from chromaprint.h is:
int chromaprint_get_raw_fingerprint(ChromaprintContext *ctx, uint32_t **fingerprint, int *size);