mov_try_read_block() allocates 1MB at least, which can be more than
enough. It was called when reading saiz box, which can appear
periodically inside fmp4. This consumes a lot of memory.
We can fix mov_try_read_block() by clamp 'block_size' with 'size'.
However, the function is harmful than helpful. It avoids allocating
large memory when the real data is small. Even in that case, if
allocating large memory directly failed, it's fine to return ENOMEM;
if allocating success and reading doesn't match the given size, it's
fine to free and return AVERROR_INVALIDDATA. In other cases, it's a
waste of CPU and memory.
So I decided to remove the function, and replace it by call
av_malloc() and avio_read() directly.
mov_read_saiz() and mov_read_pssh() need more check, but they don't
belong to this patch.
Fixes#7641 and #9243.
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
The old logic was trying to be excessively clever in "deducing" that the
user wanted to stretch/scale the image when ow/oh differed from iw/ih
aspect ratio. But this is almost surely unintended except in
pathological cases, and in those cases users should simply disable
normalize_sar and do all the stretching/scaling logic themselves. This
is especially important in multi-input mode, where the canvas may be
vastly different from the input dimensions of any stream. Also, passing
through input 0 SAR in multi-input mode is arbitrary and nearly useless,
so again force output SAR to 1:1 here.
Fixes: signed integer overflow: 256 * 668003712 cannot be represented in type 'int'
Fixes: 59819/clusterfuzz-testcase-minimized-ffmpeg_dem_MATROSKA_fuzzer-4674636538052608
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Parse through all NALUs in a packet, pull new ones when a complete AU could not
be assembled, or keep them around if an AU was assembled while data remained in
them.
Signed-off-by: James Almer <jamrial@gmail.com>
Default GOP size is now set by preset and tuning info. This GOP size is only overwriten if -g value is provided.
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
As per the spec:
VUID-VkVideoDecodeH264PictureInfoKHR-sliceCount-arraylength
sliceCount must be greater than 0
VUID-VkVideoDecodeH265PictureInfoKHR-sliceSegmentCount-arraylength
sliceSegmentCount must be greater than 0
This particularly happens with seeking in field-coded H264.
This commit scraps a bool to signal to recreate the session parameters,
but instead destroys them, forcing them to be recreated.
As this can happen between start_frame and end_frame, do this
at both places.
Move the slice offsets buffer to the thread decode context.
It isn't part of the resources for frame decoding, the driver
has to process and finish with it at submission time.
That way, it doesn't need to be alloc'd + freed on every frame.
This requires using the new AVHWFramesContext.opaque field, as
otherwise, the profile attached to the decoder will be freed
before the frames context, rendering the frames context useless.
But ensure the value returned by evc_read_nal_unit_length() fits in an int.
Should prevent integer overflows later in the code.
Signed-off-by: James Almer <jamrial@gmail.com>
Fixed-point AAC decoder currently does not produce the same output on
all platforms. Until that is fixed, silence the audio stream using the
volume filter.
Also, actually use the aac_fixed decoder as was the original intent.
Before the introduction of AV_CODEC_ID_TIMED_ID3 for timed_id3 metadata streams
in mpegts (commit 4a4437c0fb), AV_CODEC_ID_SMPTE_KLV
was the only existing codec for metadata.
It seems that this codec has a 5-bytes metadata header[1] that, for some reason,
was always skipped when decoding data packets.
However, when working with a AV_CODEC_ID_TIMED_ID3 streams, this results in the
5 first bytes of the payload being cut-off, which includes essential informations
such as the ID3 tag version.
This patch fixes the issue by keeping the 5-bytes skip only for AV_CODEC_ID_SMPTE_KLV
streams.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: James Almer <jamrial@gmail.com>
Fixes: out of array read
Fixes: 59828/clusterfuzz-testcase-minimized-ffmpeg_dem_JPEGXL_ANIM_fuzzer-5029813220671488
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Use the gcd of all input timebases to ensure PTS accuracy. For the
framerate, just pick the highest of all the inputs, under the assumption
that we will render frames with approximately this frequency. Of course,
this is not 100% accurate, in particular if the input frames are badly
misaligned. But this field is informational to begin with.
Importantly, it covers the "common" case of combining high FPS and low
FPS streams with aligned frames.
James Almer
Zhao Zhili
Michael Niedermayer
Niklas Haas
caleb
Ricardo Monteiro
Lynne
Dave Airlie
Anton Khirnov
Romain Beauxis
Leo Izen