avformat/mov: Fix integer overflows related to sample_duration
Fixes: runtime error: signed integer overflow: -9166684017437101870 + -2495066639299164439 cannot be represented in type Fixes: Chromium bug 791349 Reported-by: Matt Wolenetz <wolenetz@google.com> Reviewed-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Jérôme Martinez [Wed, 7 Mar 2018 10:19:03 +0000 (11:19 +0100)]
avcodec/ffv1: support of more pix_fmt
Without direct support of such pix_fmt, content is padded to 16-bit
and it is not possible to know that the source file was with a smaller bit depth
so framemd5 is different
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
vc2enc: replace quantization LUT with a smaller division LUT
This commit replaces the huge and impractical LUT which converted coeffs
and a quantizer to bits to encode and instead uses a standard multiplication
and a shift to replace the division and then codes the values using the
regular golomb coding functions.
Found-by: Matt Wolenetz <wolenetz@google.com> Reviewed-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/oggparseogm: Fix undefined shift in ogm_packet()
Fixes: shift exponent 48 is too large for 32-bit type 'int' Fixes: Chromium bug 786793 Reported-by: Matt Wolenetz <wolenetz@google.com> Reviewed-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/avidec: Fix integer overflow in cum_len check
Fixes: signed integer overflow: 3775922176 * 4278190080 cannot be represented in type 'long' Fixes: Chromium bug 791237 Reported-by: Matt Wolenetz <wolenetz@google.com> Reviewed-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/oggparsetheora: Do not adjust AV_NOPTS_VALUE
Fixes: Chromium bug 795653 Fixes: signed integer overflow: 9223372036854775807 + 1 cannot be represented in type 'long' Reported-by: Matt Wolenetz <wolenetz@google.com> Reviewed-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Marton Balint [Mon, 5 Mar 2018 23:45:09 +0000 (00:45 +0100)]
avutil/parseutils: only accept full us duration, do not accept mss duration
Accepting 'u' suffix for a time specification is neither intuitive nor
consistent (now that we don't accept m). Also there was a bug in the code
accepting an extra 's' even after 'ms'.
Rodger Combs [Sat, 3 Feb 2018 02:50:13 +0000 (20:50 -0600)]
lavc/videotoolbox: fix threaded decoding
AVHWAccel.end_frame can run on a worker thread. The assumption of the
frame threading code is that the worker thread will change the AVFrame
image data, not the AVFrame fields. So the AVFrame fields are not synced
back to the main thread. But this breaks videotoolbox due to its special
requirements (everything else is fine). It actually wants to update
AVFrame fields.
The actual videotoolbox frame is now stored in the dummy AVBufferRef, so
it mimics what happens in non-videotoolbox cases. (Changing the
AVBufferRef contents is a bit like changing the image data.) The
post_process callback copies that reference to the proper AVFrame field.
Calvin Walton [Thu, 22 Feb 2018 19:10:03 +0000 (14:10 -0500)]
libavfilter/vf_fps: Rewrite using activate callback
The old version of the filter had a problem where it would queue up
all of the duplicate frames required to fill a timestamp gap in a
single call to filter_frame. In problematic files - I've hit this in
webcam streams with large gaps due to network issues - this will queue
up a potentially huge number of frames. (I've seen it trigger the Linux
OOM-killer on particularly large pts gaps.)
This revised version of the filter using the activate callback will
generate at most 1 frame each time it is called.
Haihao Xiang [Tue, 6 Feb 2018 08:17:17 +0000 (16:17 +0800)]
lavc/vaapi_encode: Don't return error if the underlying driver doesn't support B frames
The underlying driver need not support B frames - since they are enabled
by default for some codecs, it is better to disable them rather than
returning an error in this case. This makes the default settings usable
for low-power encoding on Intel platforms.
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com> Signed-off-by: Mark Thompson <sw@jkqxz.net>
Aman Gupta [Tue, 6 Mar 2018 21:07:32 +0000 (13:07 -0800)]
avcodec/mediacodecdec: add delay_flush option
The default behavior of the mediacodec decoder before this commit
was to delay flushes until all pending hardware frames were
returned to the decoder. This was useful for certain types of
applications, but was unexpected behavior for others.
The new default behavior with this commit is now to execute
flushes immediately to invalidate all pending frames. The old
behavior can be enabled by setting delay_flush=1.
With the new behavior, video players implementing seek can simply
call flush on the decoder without having to worry about whether
they have one or more mediacodec frames still buffered in their
rendering pipeline. Previously, all these frames had to be
explictly freed (or rendered) before the seek/flush would execute.
The new behavior matches the behavior of all other lavc decoders,
reducing the amount of special casing required when using the
mediacodec decoder.
Signed-off-by: Aman Gupta <aman@tmm1.net> Signed-off-by: Matthieu Bouron <matthieu.bouron@gmail.com>
avformat/utils: Fix integer overflow of fps_first/last_dts
Fixes: runtime error: signed integer overflow: 7738135736989908991 - -7898362169240453118 cannot be represented in type 'long' Fixes: Chromium bug 796778 Reported-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avformat/mov: Fix integer overflow in mov_get_stsc_samples()
Fixes: runtime error: signed integer overflow: 5 * -2147483647 cannot be represented in type 'int' Fixes: Chromium bug 817338 Reviewed-by: Matt Wolenetz <wolenetz@google.com> Reported-by: Matt Wolenetz <wolenetz@google.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Matt Wolenetz [Mon, 5 Mar 2018 23:59:18 +0000 (15:59 -0800)]
lavc/vorbisdec: Allow avcodec_open2 to call .close
If there is a decoder initialization failure detected in avcodec_open2
after .init is called, allow graceful decoder .close to prevent leaking
vorbis decoder allocations such as those from vorbis_parse_setup_*.
BUG=772699
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
James Almer [Wed, 7 Mar 2018 02:26:22 +0000 (23:26 -0300)]
avformat/isom: don't free extradata before calling ff_get_extradata()
ff_get_extradata() frees any existing extradata before allocating now,
and using av_free() here leaves a dangling pointer that will result in
a double free.
Zhong Li [Thu, 1 Mar 2018 05:20:19 +0000 (13:20 +0800)]
lavc/qsvenc: disable h264 look_ahead by default
Look_ahead can provide quality improvements, but would better disable it by default due to some reasons:
1. It is only available for some codecs (e.g. HEVC is not supported) on Intel
Haswell and plus platforms. Thus means it will be failed on some platforms.
2. It significantly increases encoding latency and memory consumption.
3. It may overwrite some other options such as CBR and CAVLC.
Signed-off-by: Zhong Li <zhong.li@intel.com> Signed-off-by: Mark Thompson <sw@jkqxz.net>
The commit which added those was pushed prematurely before anyone could object
to illogical suffixes like just m for milliseconds. Without this, we'd be locked
into never being able to implement the "m" suffix for minutes.
Right now, if someone configures ffmpeg with for example --enable-nvenc they will
get an error message complaining about missing cuda.
This is very confusing and already has lead people into installing the CUDA SDK,
even though it's not what they need.
This will make it complain about ffnvcodec instead.
Signed-off-by: Timo Rothenpieler <timo@rothenpieler.org>
Matt Wolenetz [Fri, 2 Mar 2018 23:12:41 +0000 (15:12 -0800)]
avformat/mov: Initialize a potential gap in ctts_data in mov_build_index
mov_read_ctts ignores ctts entries having count <= 0. Generally, the
aggregate of all ctts entries' count fields resulting from mov_read_ctts
can be less than the corresponding sample_count.
mov_build_index attempts to normalize any existing ctts_data counts to
be 1, to make a 1-1 mapping of a ctts_data entry to a sample.
That 1-1 mapping left a tail of uninitialized ctts_data entries when the
aggregate, normalized ctts_count < sample_count.
Even more generally, later usage of ctts_data may depend on the entire
ctts_allocated_size having been initialized.
This change memsets the entire allocation of the normalized ctts_data in
mov_build_index, to prevent use of uninitialized data later.
BUG=816787
Change-Id: I7fd7db255e3aeed076ee32c90cb2df211741c052
Reviewed-on: https://chromium-review.googlesource.com/947110 Reviewed-by: Xiaohan Wang <xhwang@chromium.org> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Jerome Borsboom [Sun, 25 Feb 2018 19:13:46 +0000 (20:13 +0100)]
avcodec/vaapi: mask unused bits in bitplane_present.value
Due to the union construct, unused bits in bitplane_present.value might
be uninitialized even when the used bits are all set to a value.
Masking the unused bits prevents spurious true values when all used
bits are unset, e.g. skipped pictures.
Philip Langdale [Sun, 25 Feb 2018 17:08:06 +0000 (09:08 -0800)]
avcodec/nvenc: Declare support for P016
nvenc doesn't support P016, but we have two problems today:
1) We declare support for YUV444P16 which nvenc also doesn't support.
We do this because it's the only pix_fmt we have that can
approximate nvenc's internal format that is YUV444P10 with data in
MSBs instead of LSBs. Because the declared format is a 16bit one,
it will be preferrentially chosen when encoding >10bit content,
but that content will normally be YUV420P12 or P016 which should
get mapped to P010 and not YUV444P10.
2) Transcoding P016 content with nvenc should be possible in a pure
hardware pipeline, and that can't be done if nvenc doesn't say it
accepts P016. By mapping it to P010, we can use it, albeit with
truncation. I have established that swscale doesn't know how to
dither to 10bits so we'd get truncation anyway, even if we tried
to do this 'properly'.
Philip Langdale [Fri, 2 Mar 2018 04:16:48 +0000 (20:16 -0800)]
swscale: Add p016 output support and generalise yuv420p1x to p010
To make the best use of existing code, I generalised the wrapper
that currently does yuv420p10 to p010 to support any mixture of
input and output sizes between 10 and 16 bits. This had the side
effect of yielding a working code path for all yuv420p1x formats
to p01x.
Xiaohan Wang [Sat, 24 Feb 2018 01:04:41 +0000 (17:04 -0800)]
ffmpeg: Fix memset size on ctts_data in mov_read_trun() (round 2)
The allocated size of sc->ctts_data is
(st->nb_index_entries + entries) * sizeof(*sc->ctts_data).
The size to memset at offset sc->ctts_data + sc->ctts_count should be
(st->nb_index_entries + entries - sc->ctts_count) *
sizeof(*sc->ctts_data))
The current code missed |entries| I believe, which was introduced in
https://patchwork.ffmpeg.org/patch/5541/.
However, after offline discussion, it seems the original code is much
more clear to read (before https://patchwork.ffmpeg.org/patch/5541/).
Hence this CL revert the memset logic to it's previous state by
remembering the |old_ctts_allocated_size|, and only memset the newly
allocated entries.
Marton Balint [Sat, 17 Feb 2018 18:31:33 +0000 (19:31 +0100)]
avformat/mxfdec: always use a stream which matches the first index table when seeking
Obviously this is still not perfect, but better then it was. Using the first
index table and mxf->current_edit_unit is still hardcoded in many places, so
this change has hopefully the less chance of breaking anything that works
now.
projects / ffmpeg / log