Due to a peculiarity in the ModR/M addressing encoding, the r12 and r13
registers sometimes requires an additional byte when used as a base register.

r14 and r15 doesn't have that issue, so prefer using them.

There's no point in emitting a rep prefix before ret on modern CPUs.

We overload the `call` instruction with a macro, but it would misbehave when
the macro argument wasn't a valid identifier. Fix it by explicitly checking
if the argument is an identifier.

Approved-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Signed-off-by: Paul B Mahol <onemda@gmail.com>

Android TV and FireOS hardware supports mpeg2 hardware decoding via
MediaCodec.
Signed-off-by: Matthieu Bouron <matthieu.bouron@gmail.com>

Fixes: Timeout
Fixes: 2115/clusterfuzz-testcase-minimized-6594111748440064

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/ac3dec_fixed: Fix multiple runtime error: signed integer overflow: -39271008 * 59 cannot be represented in type 'int'

Fixes: 2113/clusterfuzz-testcase-minimized-6510704959946752

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

Manojkumar is taking over MIPS maintenance.
Signed-off-by: Nedeljko Babic <nedeljko.babic@imgtec.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

The fate-aac-al_sbr_ps_04_ur test did not detect this mistake.

Fixes detection of some TV sample as 24.5 FPS. With the patch applied,
it's detected as 25 FPS.

This is enabled for mpegts only.

Switches temporary samples for processing to be stored in the encoder's
context, avoids memory leaks if any errors occur while encoding a frame.

Fixes CID1412026
Signed-off-by: Tyler Jones <tdjones879@gmail.com>
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com>

We have floor, ceil, and trunc. Let's add round.
Signed-off-by: Kevin Mark <kmark937@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

avcodec/mpeg4videodec: Fix runtime error: signed integer overflow: 53098 * 40448 cannot be represented in type 'int'

Fixes: 2106/clusterfuzz-testcase-minimized-6136503639998464

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

Fixes: 2100/clusterfuzz-testcase-minimized-4522961547558912

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/takdec: Fix multiple runtime error: signed integer overflow: 637072 * 4096 cannot be represented in type 'int'

Fixes: 2079/clusterfuzz-testcase-minimized-5345861779324928

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

vp9_diag_downleft_32x32_8bpp_c: 580.2
vp9_diag_downleft_32x32_8bpp_sse2: 75.6
vp9_diag_downleft_32x32_8bpp_ssse3: 73.7
vp9_diag_downleft_32x32_8bpp_avx: 72.7
vp9_diag_downleft_32x32_10bpp_c: 1101.2
vp9_diag_downleft_32x32_10bpp_sse2: 145.4
vp9_diag_downleft_32x32_10bpp_ssse3: 137.5
vp9_diag_downleft_32x32_10bpp_avx: 134.8
vp9_diag_downleft_32x32_10bpp_avx2: 94.0
vp9_diag_downleft_32x32_12bpp_c: 1108.5
vp9_diag_downleft_32x32_12bpp_sse2: 145.5
vp9_diag_downleft_32x32_12bpp_ssse3: 137.3
vp9_diag_downleft_32x32_12bpp_avx: 135.2
vp9_diag_downleft_32x32_12bpp_avx2: 94.0

~30% faster than avx implementation
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>

Forcing low_delay can be useful, even if not officially supported.

This test the demuxer discarding non ADTS frames at the beginning and
end of the input.

As a side effect, this commit also enables fate-adts-demux, which was
accidentally disabled in 324f0fbf.
Tested-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: James Almer <jamrial@gmail.com>

Fixes: out of array read
Fixes: 2097/clusterfuzz-testcase-minimized-5036861833609216

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

Fixes memleak
Fixes: 2076/clusterfuzz-testcase-minimized-6542640243802112

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/cavs: Fix runtime error: signed integer overflow: -12648062 * 256 cannot be represented in type 'int'

Fixes: 2067/clusterfuzz-testcase-minimized-5578430902960128

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

This API is more up-to-date, provides names compatible with filters
and external encoders, and is consistent with the other color
property variables.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

The encoder never actually signalled how many samples to skip.
Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

Switching the vorbis encoder to use a buffer queue for input frames allows
saving lookahead samples more easily and safely for psychoacoustic systems,
requiring less pointer arithmetic in the case of transient windows.

Using fdsp improves readability and allows using architecture-specific
optimizations.
Signed-off-by: Tyler Jones <tdjones879@gmail.com>

Signed-off-by: Tyler Jones <tdjones879@gmail.com>

This reduces the attack surface of local file-system
information leaking.

It prevents the existing exploit leading to an information leak. As
well as similar hypothetical attacks.

Leaks of information from files and symlinks ending in common multimedia extensions
are still possible. But files with sensitive information like private keys and passwords
generally do not use common multimedia filename extensions.
It does not stop leaks via remote addresses in the LAN.

The existing exploit depends on a specific decoder as well.
It does appear though that the exploit should be possible with any decoder.
The problem is that as long as sensitive information gets into the decoder,
the output of the decoder becomes sensitive as well.
The only obvious solution is to prevent access to sensitive information. Or to
disable hls or possibly some of its feature. More complex solutions like
checking the path to limit access to only subdirectories of the hls path may
work as an alternative. But such solutions are fragile and tricky to implement
portably and would not stop every possible attack nor would they work with all
valid hls files.

Developers have expressed their dislike / objected to disabling hls by default as well
as disabling hls with local files. There also where objections against restricting
remote url file extensions. This here is a less robust but also lower
inconvenience solution.
It can be applied stand alone or together with other solutions.
limiting the check to local files was suggested by nevcairiel

This recommits the security fix without the author name joke which was
originally requested by Nicolas.

Found-by: Emil Lerner and Pavel Cheremushkin
Reported-by: Thierry Foucu <tfoucu@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Requested-by: Paul B Mahol <onemda@gmail.com>
This reverts commit caf7d617.

Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

~2% faster.

This reduces the attack surface of local file-system
information leaking.

It prevents the existing exploit leading to an information leak. As
well as similar hypothetical attacks.

Leaks of information from files and symlinks ending in common multimedia extensions
are still possible. But files with sensitive information like private keys and passwords
generally do not use common multimedia filename extensions.
It does not stop leaks via remote addresses in the LAN.

The existing exploit depends on a specific decoder as well.
It does appear though that the exploit should be possible with any decoder.
The problem is that as long as sensitive information gets into the decoder,
the output of the decoder becomes sensitive as well.
The only obvious solution is to prevent access to sensitive information. Or to
disable hls or possibly some of its feature. More complex solutions like
checking the path to limit access to only subdirectories of the hls path may
work as an alternative. But such solutions are fragile and tricky to implement
portably and would not stop every possible attack nor would they work with all
valid hls files.

Developers have expressed their dislike / objected to disabling hls by default as well
as disabling hls with local files. There also where objections against restricting
remote url file extensions. This here is a less robust but also lower
inconvenience solution.
It can be applied stand alone or together with other solutions.
limiting the check to local files was suggested by nevcairiel

Found-by: Emil Lerner and Pavel Cheremushkin
Reported-by: Thierry Foucu <tfoucu@google.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

This new FATE test for the scale2ref filter makes use of the recently
added scale2ref-specific variables to maintain the aspect ratio of a
test input.

Filtergraph explanation:
[main] has an AR of 4:3. [ref] has an AR of 16:9.
640 / 4 = 160. So the new width for [main] is 160.
160 / ((320 / 240) * (1 / 1)) = 160 / (4 / 3) = 120. So the new
height for [main] is 120.
160 / 120 = 4 / 3 so [main]'s aspect ratio has been maintained while
using [ref]'s width as a reference point.

[ref] is nullsink'd since it is left unchanged by scale2ref (and so
shouldn't need to be tested).

If we were to use "iw/4:-1" in place of "iw/4:ow/mdar":
640 / 4 = 160. So the new width for [main] would be 160.
360 / 4 = 90. So the new height for [main] would be 90.
160 / 90 = 16 / 9 so [main] now has the same aspect ratio as [ref]
which is probably what you do not want.

This is currently the only test for scale2ref.
Signed-off-by: Kevin Mark <kmark937@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

The RGB555 PACKBITSRGN case tries to read a palette, if such
palette is actually stored then it accesses a null pointer.
All 16bit samples i could find use DIRECTBITSRGN.

Fixes: 2065/clusterfuzz-testcase-minimized-6298930457346048

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>

This also increases the range of input values supported as well as
decreasing the operation dependencies in the main loop, improving
speed on modern CPUs.

Fixes part of: 2045/clusterfuzz-testcase-minimized-6751255865065472

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpegSigned-off-by: Michael Niedermayer <michael@niedermayer.cc>