You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.

674 lines
23 KiB

/*
* RTP output format
* Copyright (c) 2002 Fabrice Bellard
*
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
* License as published by the Free Software Foundation; either
* version 2.1 of the License, or (at your option) any later version.
*
* FFmpeg is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
*/
#include "avformat.h"
#include "mpegts.h"
#include "internal.h"
#include "libavutil/mathematics.h"
#include "libavutil/random_seed.h"
#include "libavutil/opt.h"
#include "rtpenc.h"
static const AVOption options[] = {
FF_RTP_FLAG_OPTS(RTPMuxContext, flags),
{ "payload_type", "Specify RTP payload type", offsetof(RTPMuxContext, payload_type), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 127, AV_OPT_FLAG_ENCODING_PARAM },
{ "ssrc", "Stream identifier", offsetof(RTPMuxContext, ssrc), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
{ "cname", "CNAME to include in RTCP SR packets", offsetof(RTPMuxContext, cname), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, AV_OPT_FLAG_ENCODING_PARAM },
{ "seq", "Starting sequence number", offsetof(RTPMuxContext, seq), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 65535, AV_OPT_FLAG_ENCODING_PARAM },
{ NULL },
};
static const AVClass rtp_muxer_class = {
.class_name = "RTP muxer",
.item_name = av_default_item_name,
.option = options,
.version = LIBAVUTIL_VERSION_INT,
};
#define RTCP_SR_SIZE 28
static int is_supported(enum AVCodecID id)
{
switch(id) {
case AV_CODEC_ID_DIRAC:
case AV_CODEC_ID_H261:
case AV_CODEC_ID_H263:
case AV_CODEC_ID_H263P:
case AV_CODEC_ID_H264:
case AV_CODEC_ID_HEVC:
case AV_CODEC_ID_MPEG1VIDEO:
case AV_CODEC_ID_MPEG2VIDEO:
case AV_CODEC_ID_MPEG4:
case AV_CODEC_ID_AAC:
case AV_CODEC_ID_MP2:
case AV_CODEC_ID_MP3:
case AV_CODEC_ID_PCM_ALAW:
case AV_CODEC_ID_PCM_MULAW:
case AV_CODEC_ID_PCM_S8:
case AV_CODEC_ID_PCM_S16BE:
case AV_CODEC_ID_PCM_S16LE:
case AV_CODEC_ID_PCM_S24BE:
case AV_CODEC_ID_PCM_U16BE:
case AV_CODEC_ID_PCM_U16LE:
case AV_CODEC_ID_PCM_U8:
case AV_CODEC_ID_MPEG2TS:
case AV_CODEC_ID_AMR_NB:
case AV_CODEC_ID_AMR_WB:
case AV_CODEC_ID_VORBIS:
case AV_CODEC_ID_THEORA:
case AV_CODEC_ID_VP8:
case AV_CODEC_ID_VP9:
case AV_CODEC_ID_ADPCM_G722:
case AV_CODEC_ID_ADPCM_G726:
case AV_CODEC_ID_ADPCM_G726LE:
case AV_CODEC_ID_ILBC:
case AV_CODEC_ID_MJPEG:
case AV_CODEC_ID_SPEEX:
case AV_CODEC_ID_OPUS:
case AV_CODEC_ID_RAWVIDEO:
case AV_CODEC_ID_BITPACKED:
return 1;
default:
return 0;
}
}
static int rtp_write_header(AVFormatContext *s1)
{
RTPMuxContext *s = s1->priv_data;
int n, ret = AVERROR(EINVAL);
AVStream *st;
if (s1->nb_streams != 1) {
av_log(s1, AV_LOG_ERROR, "Only one stream supported in the RTP muxer\n");
return AVERROR(EINVAL);
}
st = s1->streams[0];
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (!is_supported(st->codecpar->codec_id)) {
av_log(s1, AV_LOG_ERROR, "Unsupported codec %s\n", avcodec_get_name(st->codecpar->codec_id));
return -1;
}
if (s->payload_type < 0) {
/* Re-validate non-dynamic payload types */
if (st->id < RTP_PT_PRIVATE)
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
st->id = ff_rtp_get_payload_type(s1, st->codecpar, -1);
s->payload_type = st->id;
} else {
/* private option takes priority */
st->id = s->payload_type;
}
s->base_timestamp = av_get_random_seed();
s->timestamp = s->base_timestamp;
s->cur_timestamp = 0;
if (!s->ssrc)
s->ssrc = av_get_random_seed();
s->first_packet = 1;
s->first_rtcp_ntp_time = ff_ntp_time();
if (s1->start_time_realtime != 0 && s1->start_time_realtime != AV_NOPTS_VALUE)
/* Round the NTP time to whole milliseconds. */
s->first_rtcp_ntp_time = (s1->start_time_realtime / 1000) * 1000 +
NTP_OFFSET_US;
// Pick a random sequence start number, but in the lower end of the
// available range, so that any wraparound doesn't happen immediately.
// (Immediate wraparound would be an issue for SRTP.)
if (s->seq < 0) {
if (s1->flags & AVFMT_FLAG_BITEXACT) {
s->seq = 0;
} else
s->seq = av_get_random_seed() & 0x0fff;
} else
s->seq &= 0xffff; // Use the given parameter, wrapped to the right interval
if (s1->packet_size) {
if (s1->pb->max_packet_size)
s1->packet_size = FFMIN(s1->packet_size,
s1->pb->max_packet_size);
} else
s1->packet_size = s1->pb->max_packet_size;
if (s1->packet_size <= 12) {
av_log(s1, AV_LOG_ERROR, "Max packet size %u too low\n", s1->packet_size);
return AVERROR(EIO);
}
s->buf = av_malloc(s1->packet_size);
if (!s->buf) {
return AVERROR(ENOMEM);
}
s->max_payload_size = s1->packet_size - 12;
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO) {
avpriv_set_pts_info(st, 32, 1, st->codecpar->sample_rate);
} else {
avpriv_set_pts_info(st, 32, 1, 90000);
}
s->buf_ptr = s->buf;
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
switch(st->codecpar->codec_id) {
case AV_CODEC_ID_MP2:
case AV_CODEC_ID_MP3:
s->buf_ptr = s->buf + 4;
avpriv_set_pts_info(st, 32, 1, 90000);
break;
case AV_CODEC_ID_MPEG1VIDEO:
case AV_CODEC_ID_MPEG2VIDEO:
break;
case AV_CODEC_ID_MPEG2TS:
n = s->max_payload_size / TS_PACKET_SIZE;
if (n < 1)
n = 1;
s->max_payload_size = n * TS_PACKET_SIZE;
break;
case AV_CODEC_ID_DIRAC:
if (s1->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
av_log(s, AV_LOG_ERROR,
"Packetizing VC-2 is experimental and does not use all values "
"of the specification "
"(even though most receivers may handle it just fine). "
"Please set -strict experimental in order to enable it.\n");
ret = AVERROR_EXPERIMENTAL;
goto fail;
}
break;
case AV_CODEC_ID_H261:
if (s1->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
av_log(s, AV_LOG_ERROR,
"Packetizing H.261 is experimental and produces incorrect "
"packetization for cases where GOBs don't fit into packets "
"(even though most receivers may handle it just fine). "
"Please set -f_strict experimental in order to enable it.\n");
ret = AVERROR_EXPERIMENTAL;
goto fail;
}
break;
case AV_CODEC_ID_H264:
/* check for H.264 MP4 syntax */
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (st->codecpar->extradata_size > 4 && st->codecpar->extradata[0] == 1) {
s->nal_length_size = (st->codecpar->extradata[4] & 0x03) + 1;
}
break;
case AV_CODEC_ID_HEVC:
/* Only check for the standardized hvcC version of extradata, keeping
* things simple and similar to the avcC/H.264 case above, instead
* of trying to handle the pre-standardization versions (as in
* libavcodec/hevc.c). */
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (st->codecpar->extradata_size > 21 && st->codecpar->extradata[0] == 1) {
s->nal_length_size = (st->codecpar->extradata[21] & 0x03) + 1;
}
break;
case AV_CODEC_ID_VP9:
if (s1->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
av_log(s, AV_LOG_ERROR,
"Packetizing VP9 is experimental and its specification is "
"still in draft state. "
"Please set -strict experimental in order to enable it.\n");
ret = AVERROR_EXPERIMENTAL;
goto fail;
}
break;
case AV_CODEC_ID_VORBIS:
case AV_CODEC_ID_THEORA:
s->max_frames_per_packet = 15;
break;
case AV_CODEC_ID_ADPCM_G722:
/* Due to a historical error, the clock rate for G722 in RTP is
* 8000, even if the sample rate is 16000. See RFC 3551. */
avpriv_set_pts_info(st, 32, 1, 8000);
break;
case AV_CODEC_ID_OPUS:
if (st->codecpar->ch_layout.nb_channels > 2) {
av_log(s1, AV_LOG_ERROR, "Multistream opus not supported in RTP\n");
goto fail;
}
/* The opus RTP RFC says that all opus streams should use 48000 Hz
* as clock rate, since all opus sample rates can be expressed in
* this clock rate, and sample rate changes on the fly are supported. */
avpriv_set_pts_info(st, 32, 1, 48000);
break;
case AV_CODEC_ID_ILBC:
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (st->codecpar->block_align != 38 && st->codecpar->block_align != 50) {
av_log(s1, AV_LOG_ERROR, "Incorrect iLBC block size specified\n");
goto fail;
}
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
s->max_frames_per_packet = s->max_payload_size / st->codecpar->block_align;
break;
case AV_CODEC_ID_AMR_NB:
case AV_CODEC_ID_AMR_WB:
s->max_frames_per_packet = 50;
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
if (st->codecpar->codec_id == AV_CODEC_ID_AMR_NB)
n = 31;
else
n = 61;
/* max_header_toc_size + the largest AMR payload must fit */
if (1 + s->max_frames_per_packet + n > s->max_payload_size) {
av_log(s1, AV_LOG_ERROR, "RTP max payload size too small for AMR\n");
goto fail;
}
if (st->codecpar->ch_layout.nb_channels != 1) {
av_log(s1, AV_LOG_ERROR, "Only mono is supported\n");
goto fail;
}
break;
case AV_CODEC_ID_AAC:
s->max_frames_per_packet = 50;
break;
default:
break;
}
return 0;
fail:
av_freep(&s->buf);
return ret;
}
/* send an rtcp sender report packet */
static void rtcp_send_sr(AVFormatContext *s1, int64_t ntp_time, int bye)
{
RTPMuxContext *s = s1->priv_data;
uint32_t rtp_ts;
av_log(s1, AV_LOG_TRACE, "RTCP: %02x %"PRIx64" %"PRIx32"\n", s->payload_type, ntp_time, s->timestamp);
s->last_rtcp_ntp_time = ntp_time;
rtp_ts = av_rescale_q(ntp_time - s->first_rtcp_ntp_time, (AVRational){1, 1000000},
s1->streams[0]->time_base) + s->base_timestamp;
avio_w8(s1->pb, RTP_VERSION << 6);
avio_w8(s1->pb, RTCP_SR);
avio_wb16(s1->pb, 6); /* length in words - 1 */
avio_wb32(s1->pb, s->ssrc);
avio_wb32(s1->pb, ntp_time / 1000000);
avio_wb32(s1->pb, ((ntp_time % 1000000) << 32) / 1000000);
avio_wb32(s1->pb, rtp_ts);
avio_wb32(s1->pb, s->packet_count);
avio_wb32(s1->pb, s->octet_count);
if (s->cname) {
int len = FFMIN(strlen(s->cname), 255);
avio_w8(s1->pb, (RTP_VERSION << 6) + 1);
avio_w8(s1->pb, RTCP_SDES);
avio_wb16(s1->pb, (7 + len + 3) / 4); /* length in words - 1 */
avio_wb32(s1->pb, s->ssrc);
avio_w8(s1->pb, 0x01); /* CNAME */
avio_w8(s1->pb, len);
avio_write(s1->pb, s->cname, len);
avio_w8(s1->pb, 0); /* END */
for (len = (7 + len) % 4; len % 4; len++)
avio_w8(s1->pb, 0);
}
if (bye) {
avio_w8(s1->pb, (RTP_VERSION << 6) | 1);
avio_w8(s1->pb, RTCP_BYE);
avio_wb16(s1->pb, 1); /* length in words - 1 */
avio_wb32(s1->pb, s->ssrc);
}
avio_flush(s1->pb);
}
/* send an rtp packet. sequence number is incremented, but the caller
must update the timestamp itself */
void ff_rtp_send_data(AVFormatContext *s1, const uint8_t *buf1, int len, int m)
{
RTPMuxContext *s = s1->priv_data;
av_log(s1, AV_LOG_TRACE, "rtp_send_data size=%d\n", len);
/* build the RTP header */
avio_w8(s1->pb, RTP_VERSION << 6);
avio_w8(s1->pb, (s->payload_type & 0x7f) | ((m & 0x01) << 7));
avio_wb16(s1->pb, s->seq);
avio_wb32(s1->pb, s->timestamp);
avio_wb32(s1->pb, s->ssrc);
avio_write(s1->pb, buf1, len);
avio_flush(s1->pb);
s->seq = (s->seq + 1) & 0xffff;
s->octet_count += len;
s->packet_count++;
}
/* send an integer number of samples and compute time stamp and fill
the rtp send buffer before sending. */
static int rtp_send_samples(AVFormatContext *s1,
const uint8_t *buf1, int size, int sample_size_bits)
{
RTPMuxContext *s = s1->priv_data;
int len, max_packet_size, n;
/* Calculate the number of bytes to get samples aligned on a byte border */
int aligned_samples_size = sample_size_bits/av_gcd(sample_size_bits, 8);
max_packet_size = (s->max_payload_size / aligned_samples_size) * aligned_samples_size;
/* Not needed, but who knows. Don't check if samples aren't an even number of bytes. */
if ((sample_size_bits % 8) == 0 && ((8 * size) % sample_size_bits) != 0)
return AVERROR(EINVAL);
n = 0;
while (size > 0) {
s->buf_ptr = s->buf;
len = FFMIN(max_packet_size, size);
/* copy data */
memcpy(s->buf_ptr, buf1, len);
s->buf_ptr += len;
buf1 += len;
size -= len;
s->timestamp = s->cur_timestamp + n * 8 / sample_size_bits;
ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
n += (s->buf_ptr - s->buf);
}
return 0;
}
static void rtp_send_mpegaudio(AVFormatContext *s1,
const uint8_t *buf1, int size)
{
RTPMuxContext *s = s1->priv_data;
int len, count, max_packet_size;
max_packet_size = s->max_payload_size;
/* test if we must flush because not enough space */
len = (s->buf_ptr - s->buf);
if ((len + size) > max_packet_size) {
if (len > 4) {
ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
s->buf_ptr = s->buf + 4;
}
}
if (s->buf_ptr == s->buf + 4) {
s->timestamp = s->cur_timestamp;
}
/* add the packet */
if (size > max_packet_size) {
/* big packet: fragment */
count = 0;
while (size > 0) {
len = max_packet_size - 4;
if (len > size)
len = size;
/* build fragmented packet */
s->buf[0] = 0;
s->buf[1] = 0;
s->buf[2] = count >> 8;
s->buf[3] = count;
memcpy(s->buf + 4, buf1, len);
ff_rtp_send_data(s1, s->buf, len + 4, 0);
size -= len;
buf1 += len;
count += len;
}
} else {
if (s->buf_ptr == s->buf + 4) {
/* no fragmentation possible */
s->buf[0] = 0;
s->buf[1] = 0;
s->buf[2] = 0;
s->buf[3] = 0;
}
memcpy(s->buf_ptr, buf1, size);
s->buf_ptr += size;
}
}
static void rtp_send_raw(AVFormatContext *s1,
const uint8_t *buf1, int size)
{
RTPMuxContext *s = s1->priv_data;
int len, max_packet_size;
max_packet_size = s->max_payload_size;
while (size > 0) {
len = max_packet_size;
if (len > size)
len = size;
s->timestamp = s->cur_timestamp;
ff_rtp_send_data(s1, buf1, len, (len == size));
buf1 += len;
size -= len;
}
}
/* NOTE: size is assumed to be an integer multiple of TS_PACKET_SIZE */
static void rtp_send_mpegts_raw(AVFormatContext *s1,
const uint8_t *buf1, int size)
{
RTPMuxContext *s = s1->priv_data;
int len, out_len;
s->timestamp = s->cur_timestamp;
while (size >= TS_PACKET_SIZE) {
len = s->max_payload_size - (s->buf_ptr - s->buf);
if (len > size)
len = size;
memcpy(s->buf_ptr, buf1, len);
buf1 += len;
size -= len;
s->buf_ptr += len;
out_len = s->buf_ptr - s->buf;
if (out_len >= s->max_payload_size) {
ff_rtp_send_data(s1, s->buf, out_len, 0);
s->buf_ptr = s->buf;
}
}
}
static int rtp_send_ilbc(AVFormatContext *s1, const uint8_t *buf, int size)
{
RTPMuxContext *s = s1->priv_data;
AVStream *st = s1->streams[0];
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
int frame_duration = av_get_audio_frame_duration2(st->codecpar, 0);
int frame_size = st->codecpar->block_align;
int frames = size / frame_size;
while (frames > 0) {
if (s->num_frames > 0 &&
av_compare_ts(s->cur_timestamp - s->timestamp, st->time_base,
s1->max_delay, AV_TIME_BASE_Q) >= 0) {
ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
s->num_frames = 0;
}
if (!s->num_frames) {
s->buf_ptr = s->buf;
s->timestamp = s->cur_timestamp;
}
memcpy(s->buf_ptr, buf, frame_size);
frames--;
s->num_frames++;
s->buf_ptr += frame_size;
buf += frame_size;
s->cur_timestamp += frame_duration;
if (s->num_frames == s->max_frames_per_packet) {
ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
s->num_frames = 0;
}
}
return 0;
}
static int rtp_write_packet(AVFormatContext *s1, AVPacket *pkt)
{
RTPMuxContext *s = s1->priv_data;
AVStream *st = s1->streams[0];
int rtcp_bytes;
int size= pkt->size;
av_log(s1, AV_LOG_TRACE, "%d: write len=%d\n", pkt->stream_index, size);
rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
RTCP_TX_RATIO_DEN;
if ((s->first_packet || ((rtcp_bytes >= RTCP_SR_SIZE) &&
(ff_ntp_time() - s->last_rtcp_ntp_time > 5000000))) &&
!(s->flags & FF_RTP_FLAG_SKIP_RTCP)) {
rtcp_send_sr(s1, ff_ntp_time(), 0);
s->last_octet_count = s->octet_count;
s->first_packet = 0;
}
s->cur_timestamp = s->base_timestamp + pkt->pts;
lavf: replace AVStream.codec with AVStream.codecpar Currently, AVStream contains an embedded AVCodecContext instance, which is used by demuxers to export stream parameters to the caller and by muxers to receive stream parameters from the caller. It is also used internally as the codec context that is passed to parsers. In addition, it is also widely used by the callers as the decoding (when demuxer) or encoding (when muxing) context, though this has been officially discouraged since Libav 11. There are multiple important problems with this approach: - the fields in AVCodecContext are in general one of * stream parameters * codec options * codec state However, it's not clear which ones are which. It is consequently unclear which fields are a demuxer allowed to set or a muxer allowed to read. This leads to erratic behaviour depending on whether decoding or encoding is being performed or not (and whether it uses the AVStream embedded codec context). - various synchronization issues arising from the fact that the same context is used by several different APIs (muxers/demuxers, parsers, bitstream filters and encoders/decoders) simultaneously, with there being no clear rules for who can modify what and the different processes being typically delayed with respect to each other. - avformat_find_stream_info() making it necessary to support opening and closing a single codec context multiple times, thus complicating the semantics of freeing various allocated objects in the codec context. Those problems are resolved by replacing the AVStream embedded codec context with a newly added AVCodecParameters instance, which stores only the stream parameters exported by the demuxers or read by the muxers.
11 years ago
switch(st->codecpar->codec_id) {
case AV_CODEC_ID_PCM_MULAW:
case AV_CODEC_ID_PCM_ALAW:
case AV_CODEC_ID_PCM_U8:
case AV_CODEC_ID_PCM_S8:
return rtp_send_samples(s1, pkt->data, size, 8 * st->codecpar->ch_layout.nb_channels);
case AV_CODEC_ID_PCM_U16BE:
case AV_CODEC_ID_PCM_U16LE:
case AV_CODEC_ID_PCM_S16BE:
case AV_CODEC_ID_PCM_S16LE:
return rtp_send_samples(s1, pkt->data, size, 16 * st->codecpar->ch_layout.nb_channels);
case AV_CODEC_ID_PCM_S24BE:
return rtp_send_samples(s1, pkt->data, size, 24 * st->codecpar->ch_layout.nb_channels);
case AV_CODEC_ID_ADPCM_G722:
/* The actual sample size is half a byte per sample, but since the
* stream clock rate is 8000 Hz while the sample rate is 16000 Hz,
* the correct parameter for send_samples_bits is 8 bits per stream
* clock. */
return rtp_send_samples(s1, pkt->data, size, 8 * st->codecpar->ch_layout.nb_channels);
case AV_CODEC_ID_ADPCM_G726:
case AV_CODEC_ID_ADPCM_G726LE:
return rtp_send_samples(s1, pkt->data, size,
st->codecpar->bits_per_coded_sample * st->codecpar->ch_layout.nb_channels);
case AV_CODEC_ID_MP2:
case AV_CODEC_ID_MP3:
rtp_send_mpegaudio(s1, pkt->data, size);
break;
case AV_CODEC_ID_MPEG1VIDEO:
case AV_CODEC_ID_MPEG2VIDEO:
ff_rtp_send_mpegvideo(s1, pkt->data, size);
break;
case AV_CODEC_ID_AAC:
if (s->flags & FF_RTP_FLAG_MP4A_LATM)
ff_rtp_send_latm(s1, pkt->data, size);
else
ff_rtp_send_aac(s1, pkt->data, size);
break;
case AV_CODEC_ID_AMR_NB:
case AV_CODEC_ID_AMR_WB:
ff_rtp_send_amr(s1, pkt->data, size);
break;
case AV_CODEC_ID_MPEG2TS:
rtp_send_mpegts_raw(s1, pkt->data, size);
break;
case AV_CODEC_ID_DIRAC:
ff_rtp_send_vc2hq(s1, pkt->data, size, st->codecpar->field_order != AV_FIELD_PROGRESSIVE ? 1 : 0);
break;
case AV_CODEC_ID_H264:
ff_rtp_send_h264_hevc(s1, pkt->data, size);
break;
case AV_CODEC_ID_H261:
ff_rtp_send_h261(s1, pkt->data, size);
break;
case AV_CODEC_ID_H263:
if (s->flags & FF_RTP_FLAG_RFC2190) {
size_t mb_info_size;
const uint8_t *mb_info =
av_packet_get_side_data(pkt, AV_PKT_DATA_H263_MB_INFO,
&mb_info_size);
ff_rtp_send_h263_rfc2190(s1, pkt->data, size, mb_info, mb_info_size);
break;
}
/* Fallthrough */
case AV_CODEC_ID_H263P:
ff_rtp_send_h263(s1, pkt->data, size);
break;
case AV_CODEC_ID_HEVC:
ff_rtp_send_h264_hevc(s1, pkt->data, size);
break;
case AV_CODEC_ID_VORBIS:
case AV_CODEC_ID_THEORA:
ff_rtp_send_xiph(s1, pkt->data, size);
break;
case AV_CODEC_ID_VP8:
ff_rtp_send_vp8(s1, pkt->data, size);
break;
case AV_CODEC_ID_VP9:
ff_rtp_send_vp9(s1, pkt->data, size);
break;
case AV_CODEC_ID_ILBC:
rtp_send_ilbc(s1, pkt->data, size);
break;
case AV_CODEC_ID_MJPEG:
ff_rtp_send_jpeg(s1, pkt->data, size);
break;
case AV_CODEC_ID_BITPACKED:
case AV_CODEC_ID_RAWVIDEO: {
int interlaced = st->codecpar->field_order != AV_FIELD_PROGRESSIVE;
ff_rtp_send_raw_rfc4175(s1, pkt->data, size, interlaced, 0);
if (interlaced)
ff_rtp_send_raw_rfc4175(s1, pkt->data, size, interlaced, 1);
break;
}
case AV_CODEC_ID_OPUS:
if (size > s->max_payload_size) {
av_log(s1, AV_LOG_ERROR,
"Packet size %d too large for max RTP payload size %d\n",
size, s->max_payload_size);
return AVERROR(EINVAL);
}
/* Intentional fallthrough */
default:
/* better than nothing : send the codec raw data */
rtp_send_raw(s1, pkt->data, size);
break;
}
return 0;
}
static int rtp_write_trailer(AVFormatContext *s1)
{
RTPMuxContext *s = s1->priv_data;
/* If the caller closes and recreates ->pb, this might actually
* be NULL here even if it was successfully allocated at the start. */
if (s1->pb && (s->flags & FF_RTP_FLAG_SEND_BYE))
rtcp_send_sr(s1, ff_ntp_time(), 1);
av_freep(&s->buf);
return 0;
}
const AVOutputFormat ff_rtp_muxer = {
.name = "rtp",
.long_name = NULL_IF_CONFIG_SMALL("RTP output"),
.priv_data_size = sizeof(RTPMuxContext),
.audio_codec = AV_CODEC_ID_PCM_MULAW,
.video_codec = AV_CODEC_ID_MPEG4,
.write_header = rtp_write_header,
.write_packet = rtp_write_packet,
.write_trailer = rtp_write_trailer,
.priv_class = &rtp_muxer_class,
.flags = AVFMT_TS_NONSTRICT,
};