This wasn't an issue prior to 58404738, when the whole RTMP packet
was copied at once and the length of the individual embedded flv
packets only were validated by the flv demuxer.
Prior to this patch, this could lead to reads and writes out of bound.
Signed-off-by: Martin Storsjö <martin@martin.st>
If the embedded flv packets were incomplete and we aborted the
copying loop early, make sure the flv buffer is trimmed to
only contain full packets.
Signed-off-by: Martin Storsjö <martin@martin.st>
update_offset is also called from handle_metadata, where the
packet header sizes is already included in the size.
Previously this lead to flv_data/flv_size including 15 uninitialized
bytes at the end after each call to handle_metadata, making the
flv demuxer lose sync with the stream.
Also remove leftover copying in handle_metadata. This is a leftover
from the refactoring in 5840473. (Previously this final mempcy was
the one that copied all the packets at once, while this is done
within the loop right now.) After making sure flv_size is set to
the right size, this write was out of bounds.
Signed-off-by: Martin Storsjö <martin@martin.st>
This was overlooked in d872fb0f7 since I assumed all the realloc
issues in the rtmp code was fixed already.
Signed-off-by: Martin Storsjö <martin@martin.st>
The current magic numbers passed are values in seconds, while the
parameter itself should be passed over the wire in milliseconds.
This makes (some/all?) live streams from Red5 work correctly, that
previously returned StreamNotFound even with "-rtmp_live live". After
this commit, the default 'any' also works on these streams.
Signed-off-by: Martin Storsjö <martin@martin.st>
On (certain streams/setups at least on) Red5, the description string
actually is present, but empty. Therefore, first try loading the
description, but if not found or empty, load the code string instead.
The code string is quite understandable in most cases anyway (like
"NetStream.Play.StreamNotFound").
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure errors in setting stream parameters are passed
on to the caller. This avoids successfully opening files while
some parameters aren't filled in properly.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Fixes sync in some samples (e.g. bugs 7581 and 8374 in VLC).
Based on a commit by Matthieu Bouron <matthieu.bouron@gmail.com>
Reported-by: Jean-Baptiste Kempf <jb@videolan.org>
CC: libav-stable@libav.org
This avoids setting a negative number of frames, ending up with a
negative average frame rate.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
If a zero-length video packet is to be returned, just return
AVERROR(EAGAIN) and switch back to the audio stream.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids a division by zero for G726.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids a division by zero.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Even if the sample rate is valid, an invalid bitrate could
pass the mode combination test below.
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids divisions by zero later (and possibly assertions in
time base scaling), since an invalid rate_flag combined with an
invalid bitrate below could pass the mode combination test.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
When av_reallocp fails, the associated variables that keep track of
the number of elements in the array (and in some cases, the
separate number of allocated elements) need to be reset.
Not all of these might technically be needed, but it's better to
reset them if in doubt, to make sure variables don't end up
conflicting.
Signed-off-by: Martin Storsjö <martin@martin.st>
Also add options for specifying a certificate and key, which can
be used both when operating as client and as server.
Partially based on a patch by Peter Ross.
Signed-off-by: Martin Storsjö <martin@martin.st>
When passing a dict to the nested protocol, it will consume
the used options from it, so a separate copy needs to be used
when reopening the connection multiple times.
Signed-off-by: Martin Storsjö <martin@martin.st>
A file containing the trusted CA certificates needs to be
supplied via the ca_file AVOption, unless the TLS library
has got a system default file/database set up.
This doesn't check the hostname of the peer certificate with
openssl, which requires a non-trivial piece of code for
manually matching the desired hostname to the string provided
by the certificate, not provided as a library function.
That is, with openssl, this only validates that the received
certificate is signed with the right CA, but not that it is
the actual server we think we're talking to.
Verification is still disabled by default since we can't count
on a proper CA database existing at all times.
Signed-off-by: Martin Storsjö <martin@martin.st>
Fragmenting blindly to a certain duration isn't a good choice
if one should be able to switch between different qualities,
therefore default to keyframes instead.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure other sanity checks for conflicting options
can work properly, e.g. for the conflict between the faststart
flag when using the ismv mode.
Signed-off-by: Martin Storsjö <martin@martin.st>
Use update_offset() as done for rtmp audio, video and notifications and
read update and write the fields instead of replacing them in the rtmp
packet and then memcpying it to the output buffer.
And fix the AMF_DATA_TYPE_ARRAY parsing while at it.
A MIXEDARRAY type, as the ARRAY, store the number of elements in
an uint32 before the list. The ARRAY is strict and does not have
an OBJECT terminator, MIXEDARRAY behaves like an OBJECT type and
a different than stated number of element can be present.
Also make sure the existing length check can't overflow.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure that it doesn't try to free an AVBuffer belonging
to an earlier packet when we free the local packet at the end.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This avoids divisions by zero later.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Null buffers are useful for simulating writing to a real buffer
for the sake of measuring how many bytes are written.
Signed-off-by: Martin Storsjö <martin@martin.st>
ASF markers only have a start time, so we lose the chapter end times,
but that is ASF for you
Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
this was forgotten when we changed ASF to not output the preroll time
Signed-off-by: Vladimir Pantelic <vladoman@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Abort if it is invalid if strict error checking has been requested.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes the output fragments independent of their position in
the output stream, making the output work better when streamed.
QuickTime Player doesn't support fragmented mp4 without the base
data offset, though.
Signed-off-by: Martin Storsjö <martin@martin.st>
This is a bit more work, but avoids having to fill in
the data offset field afterwards instead of directly when
the rest of the trun atom is written.
This simplifies future cases where this field needs to be set to
something different.
Signed-off-by: Martin Storsjö <martin@martin.st>
A given packet won't always come in contiguously; sometimes
they may be broken up on chunk boundaries by packets of another
channel.
This support primarily involves tracking information about the
data that's been read, so the reader can pick up where it left
off for a given channel.
As a side effect, we no longer over-report the bytes read if
(toread = MIN(size, chunk_size)) == size
Signed-off-by: Martin Storsjö <martin@martin.st>
Since the number of channels is multiplied by 36 and assigned to
to a uint16_t, make sure this calculation didn't overflow. (In
certain cases the calculation could overflow leaving the
truncated block_align at 0, leading to divisions by zero later.)
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Some files have the duration set to -1 in the mdhd atom, more
or less legitimately. (We produce such files ourselves, for the
initial duration in fragmented mp4 files.)
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This more closely corresponds to the usage of the field.
Its usage here is unrelated to the channel ID.
Signed-off-by: Martin Storsjö <martin@martin.st>
Channel 4 is typically used by the Flash player to transmit
audio, channel 6 for video, and various stream-specific invokes
get sent over channel 8, which is designated the source channel.
This more closely matches the behavior of the Flash player,
including the transmission of play requests over channel 8.
Signed-off-by: Martin Storsjö <martin@martin.st>
Sending non-monotonic packets (e.g. when the audio and video
streams are monotonic within themselves but not muxed
monotonically) will lead to negative values the RTMP timestamp
field (where timestamps are transmitted only as deltas for each
channel), and this delta can end up being incorrectly written as
a large unsigned number.
Signed-off-by: Martin Storsjö <martin@martin.st>
Since 596e5d4783, this is not necessary anymore. It also allows to
actually disable the flushing, improving write performance (but
possibly giving worse latency in real-time streaming).
Signed-off-by: Martin Storsjö <martin@martin.st>
This is enabled by default and can be disabled with
"-fflags -flush_packets".
Inspired by a patch from Nicolas George <nicolas.george@normalesup.org>.
Signed-off-by: Martin Storsjö <martin@martin.st>
If we really want to support parameter changes, they need to be
signalled along with the AVPackets as parameter change side data,
not just changing the AVCodecContext parameters when a packet
is demuxed (since there may be other earlier packets yet undecoded).
Something similar was already done for the sample rate in 0883109b2,
but some parameters were left changeable.
This avoids having to recheck the channel count for validity for
each decoded frame in (ad)pcm decoders, unless the decoders
explicitly say that they accept parameter changes.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Limit the size to INT_MAX/2 (for simplicity) to be sure that
size + BYTES_PER_FRAME_RECORD won't overflow.
Also factorize other existing error return paths.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
Limit the size to INT_MAX/2 (for simplicity) to be sure that
size + FF_INPUT_BUFFER_PADDING_SIZE won't overflow.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
The seektable is required for filling in ape->frames[i].pos
further down.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes sure the faststart vs fragmentation check works as
intended when fragmentation is enabled due to using the ismv mode.
Signed-off-by: Martin Storsjö <martin@martin.st>
This fixes warnings about making integers from pointers without
a cast, and avoids the theoretical case where the lower 32 bits of
the pointer would all be zero where the implicit cast wouldn't give
the right result.
Signed-off-by: Martin Storsjö <martin@martin.st>
This allows creation of frame accurate chapter marks from sources like
DVD and BD where the precise chapter location is not known until the
chapter mark has been reached during reading.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This should improve write performance quite significantly.
---
Tested with both writing a normal mp4, by using the faststart
feature and writing a fragmented mp4 file; all turn out with the
same md5sum as before.
Signed-off-by: Martin Storsjö <martin@martin.st>
Remove the header decoding for PCM audio from mpeg.c and the
20/24bit parts from pcm.c and merge them into a new decoder in
pcm-dvd.c.
The decoder has added support for samples that span multiple
packets and modified 20/24bit group decoding. Both is needed to
decode samples that have been generated with DVD-Lab Pro 2. The
decoding of 16bit PCM and two channel 24bit is identical to
before. No other samples are known to verify the correctness of
the encoding this software does.
The complete list of tested formats is
48kHz/16bit/2-8 channels
48kHz/24bit/2-5 channels
96kHz/16bit/2-4 channels
96kHz/24bit/2 channels
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
When streaming to limelight, the app name is either a full
"appname/subaccount" or "appname/_definst_". In the latter case,
the app name can be simplified into simply "appname", but the
authentication hashing assumes the /_definst_ still to be present.
Signed-off-by: Martin Storsjö <martin@martin.st>
If a client tries to read the file while it's being updated, the client
would get an incomplete manifest. Instead write to a separate temp file
and atomically rename it to replace the previous one.
Signed-off-by: Martin Storsjö <martin@martin.st>
The element was only being written when the value == 1. But the default
value of this element is 1, so this has no useful effect. This element
needs to be written when the value == 0.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
On failures in the write_trailer function, we could also ignore
the errors and try to finish the file despite these errors (which
would only leave an incomplete chapters track). It's probably better
to signal the error clearly to the caller though (and if this
function failed there's no guarantee that there's enough memory to
finish the trailer either).
Signed-off-by: Martin Storsjö <martin@martin.st>
QuickTime will play multiple audio tracks concurrently if this flag is
set for multiple audio tracks. And if no subtitle track has this flag
set, QuickTime will show no subtitles in the subtitle menu.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Faststart moves the moov atom to the beginning of the file and rewrites
the rest of the file after muxing is complete.
Signed-off-by: Martin Storsjö <martin@martin.st>
This allows creation of frame accurate chapter marks from sources
like DVD and BD where the precise chapter location is not known until
the chapter mark has been reached during reading.
Signed-off-by: Martin Storsjö <martin@martin.st>
Allow emitting the current cluster that is being written before
starting a new one, simplifying how to figure out where clusters
are positioned in the output stream (for live streaming).
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
Seeking in certain broken files would cause ogg_read_timestamp
to fail because ogg_packet would go into a state where all packets
of stream 1 would be discarded until the end of the stream.
Bug-Id: 553
CC: libav-stable@libav.org
Signed-off-by: Jan Gerber <j@v2v.cc>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
The mov/mp4 muxer has support for handling negative timestamps
via edit lists (which customarily is used for handling the 1-frame
delay due to B-frames as well).
Using the muxer's native way of handling it is better than using
the generic offsetting. The generic offsetting is a bit too
crude when e.g. the timebase of one track is 1/fps, where the
edit lists can handle it accurately.
Signed-off-by: Martin Storsjö <martin@martin.st>
The counter itself shouldn't be wrapped, since it is used for
determining end_pts for the next segment - only wrap the number
used for the segment file name.
Signed-off-by: Martin Storsjö <martin@martin.st>
The hls muxer itself doesn't have any direct (object file level)
dependencies on mpegtsenc.o, and including that object file
directly doesn't ensure that it is registered so that the muxer
actually is accessible.
Signed-off-by: Martin Storsjö <martin@martin.st>
IPPROTO_IPV6 is unrelated here (it's only used in udp.c for
multicast sockopts), check for support for the sockaddr_in6
struct itself.
Signed-off-by: Martin Storsjö <martin@martin.st>
An SDP description normally only contains the target IP address
and port for the packets. This means that we don't really have
any clue where to send the RTCP RR packets - previously they're
sent to the destination IP written in the SDP (at the same port),
which rarely is the actual peer. And if the source for the packets
is on a different port than the destination, it's never correct.
With a new option, we can choose to send the packets to the
address that the latest packet on each socket arrived from.
---
Some may even argue that this should be the default - perhaps,
but I'd rather keep it optional at first. Additionally, I'm not
sure if sending RTCP RR directly back to the source is
desireable for e.g. multicast.
Signed-off-by: Martin Storsjö <martin@martin.st>
If we've received packets on the same socket before, the return
packets are sent to that address. If we've only received packets
on the other socket, try to guess the source port for the other
one assuming the basic +1/-1 logic.
Signed-off-by: Martin Storsjö <martin@martin.st>
Move the sources documentation up below the marker for deprecated
otpions. Also mention the new block parameter, that was added
in 749722209.
Signed-off-by: Martin Storsjö <martin@martin.st>
It is possible to have an initial broken header and then valid packets.
Reported-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
CC: libav-stable@libav.org
Add one copy of the function into each of the libraries, similarly
to what we do for log2_tab. When using static libs, only one
copy of the file_open.o object file gets included, while when
using shared libraries, each of them get a copy of its own.
This fixes DLL builds with a statically linked C runtime, where
each DLL effectively has got its own instance of the C runtime,
where file descriptors can't be shared across runtimes.
On systems not using msvcrt, the function is not duplicated.
Signed-off-by: Martin Storsjö <martin@martin.st>
This supports non-Linux systems (SOCK_CLOEXEC is non-standard) and
older Linux kernels to the extent possible.
Signed-off-by: Martin Storsjö <martin@martin.st>
When libavformat was changed to use the new avpriv_open function
in 51eb213d00, this silently bypassed the existing wrapper for
win32. Move the win32 wrapper into libavutil/file.c to make sure
it gets called everywhere (not just in the libavformat case).
This makes sure that non-ascii file names gets opened properly
(where file names internally are stored as utf8, but they get
converted to wchar_t and opened with _wsopen).
Signed-off-by: Martin Storsjö <martin@martin.st>
This provides at least some protection against potential accidental
corruption of AVIO buffer workspace.
Signed-off-by: Martin Storsjö <martin@martin.st>
It's only relevant for the RTSP demuxer. Similarly, the custom_io
flag is only present in the SDP demuxer options list.
Signed-off-by: Martin Storsjö <martin@martin.st>
Also clear the AVIOContext handle after freeing, to avoid
possible dangling pointers if the later call fails.
Signed-off-by: Martin Storsjö <martin@martin.st>
This lowers the level of warnings printed if trying to connect
to a host name that provides both v6 and v4 addresses but the
service only is available on the v4 address (often occurring for
'localhost', with servers that aren't v6-aware).
Signed-off-by: Martin Storsjö <martin@martin.st>
The common case of the pointer having increased by one packet (which results
in no change to the modulus) can be detected with a 64-bit subtraction,
which is far cheaper than a division on many platforms.
Before After
Mean StdDev Mean StdDev Change
Divisions 248.3 8.8 51.5 7.4 +381.7%
Overall 2773.2 25.6 2372.5 43.1 +16.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
When a stream contains a single program, there's no point in doing a
PID -> program lookup. Normally the one and only program isn't disabled,
so no packets should be discarded.
Before After
Mean StdDev Mean StdDev Change
discard_pid() 73.8 9.4 20.2 1.5 +264.8%
Overall 2300.8 28.0 2253.1 20.6 +2.1%
Signed-off-by: Martin Storsjö <martin@martin.st>
This was being performed to ensure that a complete packet was held in
contiguous memory, prior to parsing the packet. However, the source buffer
is typically large enough that the packet was already contiguous, so it is
beneficial to return the packet by reference in most cases.
Before After
Mean StdDev Mean StdDev Change
memcpy 720.7 32.7 649.8 25.1 +10.9%
Overall 2372.7 46.1 2291.7 21.8 +3.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
As long as there is enough contiguous data in the avio buffer,
just return a pointer to it instead of copying it to the caller
provided buffer.
Signed-off-by: Martin Storsjö <martin@martin.st>
