3 * Copyright (c) 2002 Fabrice Bellard
5 * This file is part of Libav.
7 * Libav is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
10 * version 2.1 of the License, or (at your option) any later version.
12 * Libav is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
17 * You should have received a copy of the GNU Lesser General Public
18 * License along with Libav; if not, write to the Free Software
19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
25 #include "libavutil/mathematics.h"
26 #include "libavutil/random_seed.h"
27 #include "libavutil/opt.h"
31 static const AVOption options[] = {
32 FF_RTP_FLAG_OPTS(RTPMuxContext, flags),
33 { "payload_type", "Specify RTP payload type", offsetof(RTPMuxContext, payload_type), AV_OPT_TYPE_INT, {.i64 = -1 }, -1, 127, AV_OPT_FLAG_ENCODING_PARAM },
34 { "ssrc", "Stream identifier", offsetof(RTPMuxContext, ssrc), AV_OPT_TYPE_INT, { .i64 = 0 }, INT_MIN, INT_MAX, AV_OPT_FLAG_ENCODING_PARAM },
35 { "cname", "CNAME to include in RTCP SR packets", offsetof(RTPMuxContext, cname), AV_OPT_TYPE_STRING, { .str = NULL }, 0, 0, AV_OPT_FLAG_ENCODING_PARAM },
36 { "seq", "Starting sequence number", offsetof(RTPMuxContext, seq), AV_OPT_TYPE_INT, { .i64 = -1 }, -1, 65535, AV_OPT_FLAG_ENCODING_PARAM },
40 static const AVClass rtp_muxer_class = {
41 .class_name = "RTP muxer",
42 .item_name = av_default_item_name,
44 .version = LIBAVUTIL_VERSION_INT,
47 #define RTCP_SR_SIZE 28
49 static int is_supported(enum AVCodecID id)
52 case AV_CODEC_ID_H261:
53 case AV_CODEC_ID_H263:
54 case AV_CODEC_ID_H263P:
55 case AV_CODEC_ID_H264:
56 case AV_CODEC_ID_HEVC:
57 case AV_CODEC_ID_MPEG1VIDEO:
58 case AV_CODEC_ID_MPEG2VIDEO:
59 case AV_CODEC_ID_MPEG4:
63 case AV_CODEC_ID_PCM_ALAW:
64 case AV_CODEC_ID_PCM_MULAW:
65 case AV_CODEC_ID_PCM_S8:
66 case AV_CODEC_ID_PCM_S16BE:
67 case AV_CODEC_ID_PCM_S16LE:
68 case AV_CODEC_ID_PCM_U16BE:
69 case AV_CODEC_ID_PCM_U16LE:
70 case AV_CODEC_ID_PCM_U8:
71 case AV_CODEC_ID_MPEG2TS:
72 case AV_CODEC_ID_AMR_NB:
73 case AV_CODEC_ID_AMR_WB:
74 case AV_CODEC_ID_VORBIS:
75 case AV_CODEC_ID_THEORA:
77 case AV_CODEC_ID_ADPCM_G722:
78 case AV_CODEC_ID_ADPCM_G726:
79 case AV_CODEC_ID_ILBC:
80 case AV_CODEC_ID_MJPEG:
81 case AV_CODEC_ID_SPEEX:
82 case AV_CODEC_ID_OPUS:
89 static int rtp_write_header(AVFormatContext *s1)
91 RTPMuxContext *s = s1->priv_data;
92 int n, ret = AVERROR(EINVAL);
95 if (s1->nb_streams != 1) {
96 av_log(s1, AV_LOG_ERROR, "Only one stream supported in the RTP muxer\n");
97 return AVERROR(EINVAL);
100 if (!is_supported(st->codec->codec_id)) {
101 av_log(s1, AV_LOG_ERROR, "Unsupported codec %x\n", st->codec->codec_id);
106 if (s->payload_type < 0) {
107 /* Re-validate non-dynamic payload types */
108 if (st->id < RTP_PT_PRIVATE)
109 st->id = ff_rtp_get_payload_type(s1, st->codec, -1);
111 s->payload_type = st->id;
113 /* private option takes priority */
114 st->id = s->payload_type;
117 s->base_timestamp = av_get_random_seed();
118 s->timestamp = s->base_timestamp;
119 s->cur_timestamp = 0;
121 s->ssrc = av_get_random_seed();
123 s->first_rtcp_ntp_time = ff_ntp_time();
124 if (s1->start_time_realtime)
125 /* Round the NTP time to whole milliseconds. */
126 s->first_rtcp_ntp_time = (s1->start_time_realtime / 1000) * 1000 +
128 // Pick a random sequence start number, but in the lower end of the
129 // available range, so that any wraparound doesn't happen immediately.
130 // (Immediate wraparound would be an issue for SRTP.)
132 s->seq = av_get_random_seed() & 0x0fff;
134 s->seq &= 0xffff; // Use the given parameter, wrapped to the right interval
136 if (s1->packet_size) {
137 if (s1->pb->max_packet_size)
138 s1->packet_size = FFMIN(s1->packet_size,
139 s1->pb->max_packet_size);
141 s1->packet_size = s1->pb->max_packet_size;
142 if (s1->packet_size <= 12) {
143 av_log(s1, AV_LOG_ERROR, "Max packet size %d too low\n", s1->packet_size);
146 s->buf = av_malloc(s1->packet_size);
148 return AVERROR(ENOMEM);
150 s->max_payload_size = s1->packet_size - 12;
152 s->max_frames_per_packet = 0;
153 if (s1->max_delay > 0) {
154 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
155 int frame_size = av_get_audio_frame_duration(st->codec, 0);
157 frame_size = st->codec->frame_size;
158 if (frame_size == 0) {
159 av_log(s1, AV_LOG_ERROR, "Cannot respect max delay: frame size = 0\n");
161 s->max_frames_per_packet =
162 av_rescale_q_rnd(s1->max_delay,
164 (AVRational){ frame_size, st->codec->sample_rate },
168 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
169 /* FIXME: We should round down here... */
170 if (st->avg_frame_rate.num > 0 && st->avg_frame_rate.den > 0) {
171 s->max_frames_per_packet = av_rescale_q(s1->max_delay,
172 (AVRational){1, 1000000},
173 av_inv_q(st->avg_frame_rate));
175 s->max_frames_per_packet = 1;
179 avpriv_set_pts_info(st, 32, 1, 90000);
180 switch(st->codec->codec_id) {
181 case AV_CODEC_ID_MP2:
182 case AV_CODEC_ID_MP3:
183 s->buf_ptr = s->buf + 4;
185 case AV_CODEC_ID_MPEG1VIDEO:
186 case AV_CODEC_ID_MPEG2VIDEO:
188 case AV_CODEC_ID_MPEG2TS:
189 n = s->max_payload_size / TS_PACKET_SIZE;
192 s->max_payload_size = n * TS_PACKET_SIZE;
195 case AV_CODEC_ID_H261:
196 if (s1->strict_std_compliance > FF_COMPLIANCE_EXPERIMENTAL) {
197 av_log(s, AV_LOG_ERROR,
198 "Packetizing H261 is experimental and produces incorrect "
199 "packetization for cases where GOBs don't fit into packets "
200 "(even though most receivers may handle it just fine). "
201 "Please set -f_strict experimental in order to enable it.\n");
202 ret = AVERROR_EXPERIMENTAL;
206 case AV_CODEC_ID_H264:
207 /* check for H.264 MP4 syntax */
208 if (st->codec->extradata_size > 4 && st->codec->extradata[0] == 1) {
209 s->nal_length_size = (st->codec->extradata[4] & 0x03) + 1;
212 case AV_CODEC_ID_HEVC:
213 /* Only check for the standardized hvcC version of extradata, keeping
214 * things simple and similar to the avcC/H264 case above, instead
215 * of trying to handle the pre-standardization versions (as in
216 * libavcodec/hevc.c). */
217 if (st->codec->extradata_size > 21 && st->codec->extradata[0] == 1) {
218 s->nal_length_size = (st->codec->extradata[21] & 0x03) + 1;
221 case AV_CODEC_ID_VORBIS:
222 case AV_CODEC_ID_THEORA:
223 if (!s->max_frames_per_packet)
224 s->max_frames_per_packet = 15;
225 s->max_frames_per_packet = av_clip(s->max_frames_per_packet, 1, 15);
226 s->max_payload_size -= 6; // ident+frag+tdt/vdt+pkt_num+pkt_length
229 case AV_CODEC_ID_ADPCM_G722:
230 /* Due to a historical error, the clock rate for G722 in RTP is
231 * 8000, even if the sample rate is 16000. See RFC 3551. */
232 avpriv_set_pts_info(st, 32, 1, 8000);
234 case AV_CODEC_ID_OPUS:
235 if (st->codec->channels > 2) {
236 av_log(s1, AV_LOG_ERROR, "Multistream opus not supported in RTP\n");
239 /* The opus RTP RFC says that all opus streams should use 48000 Hz
240 * as clock rate, since all opus sample rates can be expressed in
241 * this clock rate, and sample rate changes on the fly are supported. */
242 avpriv_set_pts_info(st, 32, 1, 48000);
244 case AV_CODEC_ID_ILBC:
245 if (st->codec->block_align != 38 && st->codec->block_align != 50) {
246 av_log(s1, AV_LOG_ERROR, "Incorrect iLBC block size specified\n");
249 if (!s->max_frames_per_packet)
250 s->max_frames_per_packet = 1;
251 s->max_frames_per_packet = FFMIN(s->max_frames_per_packet,
252 s->max_payload_size / st->codec->block_align);
254 case AV_CODEC_ID_AMR_NB:
255 case AV_CODEC_ID_AMR_WB:
256 if (!s->max_frames_per_packet)
257 s->max_frames_per_packet = 12;
258 if (st->codec->codec_id == AV_CODEC_ID_AMR_NB)
262 /* max_header_toc_size + the largest AMR payload must fit */
263 if (1 + s->max_frames_per_packet + n > s->max_payload_size) {
264 av_log(s1, AV_LOG_ERROR, "RTP max payload size too small for AMR\n");
267 if (st->codec->channels != 1) {
268 av_log(s1, AV_LOG_ERROR, "Only mono is supported\n");
273 case AV_CODEC_ID_AAC:
275 if (!s->max_frames_per_packet)
276 s->max_frames_per_packet = 5;
280 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
281 avpriv_set_pts_info(st, 32, 1, st->codec->sample_rate);
294 /* send an rtcp sender report packet */
295 static void rtcp_send_sr(AVFormatContext *s1, int64_t ntp_time, int bye)
297 RTPMuxContext *s = s1->priv_data;
300 av_dlog(s1, "RTCP: %02x %"PRIx64" %x\n", s->payload_type, ntp_time, s->timestamp);
302 s->last_rtcp_ntp_time = ntp_time;
303 rtp_ts = av_rescale_q(ntp_time - s->first_rtcp_ntp_time, (AVRational){1, 1000000},
304 s1->streams[0]->time_base) + s->base_timestamp;
305 avio_w8(s1->pb, RTP_VERSION << 6);
306 avio_w8(s1->pb, RTCP_SR);
307 avio_wb16(s1->pb, 6); /* length in words - 1 */
308 avio_wb32(s1->pb, s->ssrc);
309 avio_wb64(s1->pb, NTP_TO_RTP_FORMAT(ntp_time));
310 avio_wb32(s1->pb, rtp_ts);
311 avio_wb32(s1->pb, s->packet_count);
312 avio_wb32(s1->pb, s->octet_count);
315 int len = FFMIN(strlen(s->cname), 255);
316 avio_w8(s1->pb, (RTP_VERSION << 6) + 1);
317 avio_w8(s1->pb, RTCP_SDES);
318 avio_wb16(s1->pb, (7 + len + 3) / 4); /* length in words - 1 */
320 avio_wb32(s1->pb, s->ssrc);
321 avio_w8(s1->pb, 0x01); /* CNAME */
322 avio_w8(s1->pb, len);
323 avio_write(s1->pb, s->cname, len);
324 avio_w8(s1->pb, 0); /* END */
325 for (len = (7 + len) % 4; len % 4; len++)
330 avio_w8(s1->pb, (RTP_VERSION << 6) | 1);
331 avio_w8(s1->pb, RTCP_BYE);
332 avio_wb16(s1->pb, 1); /* length in words - 1 */
333 avio_wb32(s1->pb, s->ssrc);
339 /* send an rtp packet. sequence number is incremented, but the caller
340 must update the timestamp itself */
341 void ff_rtp_send_data(AVFormatContext *s1, const uint8_t *buf1, int len, int m)
343 RTPMuxContext *s = s1->priv_data;
345 av_dlog(s1, "rtp_send_data size=%d\n", len);
347 /* build the RTP header */
348 avio_w8(s1->pb, RTP_VERSION << 6);
349 avio_w8(s1->pb, (s->payload_type & 0x7f) | ((m & 0x01) << 7));
350 avio_wb16(s1->pb, s->seq);
351 avio_wb32(s1->pb, s->timestamp);
352 avio_wb32(s1->pb, s->ssrc);
354 avio_write(s1->pb, buf1, len);
357 s->seq = (s->seq + 1) & 0xffff;
358 s->octet_count += len;
362 /* send an integer number of samples and compute time stamp and fill
363 the rtp send buffer before sending. */
364 static int rtp_send_samples(AVFormatContext *s1,
365 const uint8_t *buf1, int size, int sample_size_bits)
367 RTPMuxContext *s = s1->priv_data;
368 int len, max_packet_size, n;
369 /* Calculate the number of bytes to get samples aligned on a byte border */
370 int aligned_samples_size = sample_size_bits/av_gcd(sample_size_bits, 8);
372 max_packet_size = (s->max_payload_size / aligned_samples_size) * aligned_samples_size;
373 /* Not needed, but who knows. Don't check if samples aren't an even number of bytes. */
374 if ((sample_size_bits % 8) == 0 && ((8 * size) % sample_size_bits) != 0)
375 return AVERROR(EINVAL);
379 len = FFMIN(max_packet_size, size);
382 memcpy(s->buf_ptr, buf1, len);
386 s->timestamp = s->cur_timestamp + n * 8 / sample_size_bits;
387 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
388 n += (s->buf_ptr - s->buf);
393 static void rtp_send_mpegaudio(AVFormatContext *s1,
394 const uint8_t *buf1, int size)
396 RTPMuxContext *s = s1->priv_data;
397 int len, count, max_packet_size;
399 max_packet_size = s->max_payload_size;
401 /* test if we must flush because not enough space */
402 len = (s->buf_ptr - s->buf);
403 if ((len + size) > max_packet_size) {
405 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
406 s->buf_ptr = s->buf + 4;
409 if (s->buf_ptr == s->buf + 4) {
410 s->timestamp = s->cur_timestamp;
414 if (size > max_packet_size) {
415 /* big packet: fragment */
418 len = max_packet_size - 4;
421 /* build fragmented packet */
424 s->buf[2] = count >> 8;
426 memcpy(s->buf + 4, buf1, len);
427 ff_rtp_send_data(s1, s->buf, len + 4, 0);
433 if (s->buf_ptr == s->buf + 4) {
434 /* no fragmentation possible */
440 memcpy(s->buf_ptr, buf1, size);
445 static void rtp_send_raw(AVFormatContext *s1,
446 const uint8_t *buf1, int size)
448 RTPMuxContext *s = s1->priv_data;
449 int len, max_packet_size;
451 max_packet_size = s->max_payload_size;
454 len = max_packet_size;
458 s->timestamp = s->cur_timestamp;
459 ff_rtp_send_data(s1, buf1, len, (len == size));
466 /* NOTE: size is assumed to be an integer multiple of TS_PACKET_SIZE */
467 static void rtp_send_mpegts_raw(AVFormatContext *s1,
468 const uint8_t *buf1, int size)
470 RTPMuxContext *s = s1->priv_data;
473 s->timestamp = s->cur_timestamp;
474 while (size >= TS_PACKET_SIZE) {
475 len = s->max_payload_size - (s->buf_ptr - s->buf);
478 memcpy(s->buf_ptr, buf1, len);
483 out_len = s->buf_ptr - s->buf;
484 if (out_len >= s->max_payload_size) {
485 ff_rtp_send_data(s1, s->buf, out_len, 0);
491 static int rtp_send_ilbc(AVFormatContext *s1, const uint8_t *buf, int size)
493 RTPMuxContext *s = s1->priv_data;
494 AVStream *st = s1->streams[0];
495 int frame_duration = av_get_audio_frame_duration(st->codec, 0);
496 int frame_size = st->codec->block_align;
497 int frames = size / frame_size;
500 int n = FFMIN(s->max_frames_per_packet - s->num_frames, frames);
502 if (!s->num_frames) {
504 s->timestamp = s->cur_timestamp;
506 memcpy(s->buf_ptr, buf, n * frame_size);
509 s->buf_ptr += n * frame_size;
510 buf += n * frame_size;
511 s->cur_timestamp += n * frame_duration;
513 if (s->num_frames == s->max_frames_per_packet) {
514 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
521 static int rtp_write_packet(AVFormatContext *s1, AVPacket *pkt)
523 RTPMuxContext *s = s1->priv_data;
524 AVStream *st = s1->streams[0];
528 av_dlog(s1, "%d: write len=%d\n", pkt->stream_index, size);
530 rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
532 if ((s->first_packet || ((rtcp_bytes >= RTCP_SR_SIZE) &&
533 (ff_ntp_time() - s->last_rtcp_ntp_time > 5000000))) &&
534 !(s->flags & FF_RTP_FLAG_SKIP_RTCP)) {
535 rtcp_send_sr(s1, ff_ntp_time(), 0);
536 s->last_octet_count = s->octet_count;
539 s->cur_timestamp = s->base_timestamp + pkt->pts;
541 switch(st->codec->codec_id) {
542 case AV_CODEC_ID_PCM_MULAW:
543 case AV_CODEC_ID_PCM_ALAW:
544 case AV_CODEC_ID_PCM_U8:
545 case AV_CODEC_ID_PCM_S8:
546 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
547 case AV_CODEC_ID_PCM_U16BE:
548 case AV_CODEC_ID_PCM_U16LE:
549 case AV_CODEC_ID_PCM_S16BE:
550 case AV_CODEC_ID_PCM_S16LE:
551 return rtp_send_samples(s1, pkt->data, size, 16 * st->codec->channels);
552 case AV_CODEC_ID_ADPCM_G722:
553 /* The actual sample size is half a byte per sample, but since the
554 * stream clock rate is 8000 Hz while the sample rate is 16000 Hz,
555 * the correct parameter for send_samples_bits is 8 bits per stream
557 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
558 case AV_CODEC_ID_ADPCM_G726:
559 return rtp_send_samples(s1, pkt->data, size,
560 st->codec->bits_per_coded_sample * st->codec->channels);
561 case AV_CODEC_ID_MP2:
562 case AV_CODEC_ID_MP3:
563 rtp_send_mpegaudio(s1, pkt->data, size);
565 case AV_CODEC_ID_MPEG1VIDEO:
566 case AV_CODEC_ID_MPEG2VIDEO:
567 ff_rtp_send_mpegvideo(s1, pkt->data, size);
569 case AV_CODEC_ID_AAC:
570 if (s->flags & FF_RTP_FLAG_MP4A_LATM)
571 ff_rtp_send_latm(s1, pkt->data, size);
573 ff_rtp_send_aac(s1, pkt->data, size);
575 case AV_CODEC_ID_AMR_NB:
576 case AV_CODEC_ID_AMR_WB:
577 ff_rtp_send_amr(s1, pkt->data, size);
579 case AV_CODEC_ID_MPEG2TS:
580 rtp_send_mpegts_raw(s1, pkt->data, size);
582 case AV_CODEC_ID_H264:
583 ff_rtp_send_h264_hevc(s1, pkt->data, size);
585 case AV_CODEC_ID_H261:
586 ff_rtp_send_h261(s1, pkt->data, size);
588 case AV_CODEC_ID_H263:
589 if (s->flags & FF_RTP_FLAG_RFC2190) {
590 int mb_info_size = 0;
591 const uint8_t *mb_info =
592 av_packet_get_side_data(pkt, AV_PKT_DATA_H263_MB_INFO,
594 ff_rtp_send_h263_rfc2190(s1, pkt->data, size, mb_info, mb_info_size);
598 case AV_CODEC_ID_H263P:
599 ff_rtp_send_h263(s1, pkt->data, size);
601 case AV_CODEC_ID_HEVC:
602 ff_rtp_send_h264_hevc(s1, pkt->data, size);
604 case AV_CODEC_ID_VORBIS:
605 case AV_CODEC_ID_THEORA:
606 ff_rtp_send_xiph(s1, pkt->data, size);
608 case AV_CODEC_ID_VP8:
609 ff_rtp_send_vp8(s1, pkt->data, size);
611 case AV_CODEC_ID_ILBC:
612 rtp_send_ilbc(s1, pkt->data, size);
614 case AV_CODEC_ID_MJPEG:
615 ff_rtp_send_jpeg(s1, pkt->data, size);
617 case AV_CODEC_ID_OPUS:
618 if (size > s->max_payload_size) {
619 av_log(s1, AV_LOG_ERROR,
620 "Packet size %d too large for max RTP payload size %d\n",
621 size, s->max_payload_size);
622 return AVERROR(EINVAL);
624 /* Intentional fallthrough */
626 /* better than nothing : send the codec raw data */
627 rtp_send_raw(s1, pkt->data, size);
633 static int rtp_write_trailer(AVFormatContext *s1)
635 RTPMuxContext *s = s1->priv_data;
637 /* If the caller closes and recreates ->pb, this might actually
638 * be NULL here even if it was successfully allocated at the start. */
639 if (s1->pb && (s->flags & FF_RTP_FLAG_SEND_BYE))
640 rtcp_send_sr(s1, ff_ntp_time(), 1);
646 AVOutputFormat ff_rtp_muxer = {
648 .long_name = NULL_IF_CONFIG_SMALL("RTP output"),
649 .priv_data_size = sizeof(RTPMuxContext),
650 .audio_codec = AV_CODEC_ID_PCM_MULAW,
651 .video_codec = AV_CODEC_ID_MPEG4,
652 .write_header = rtp_write_header,
653 .write_packet = rtp_write_packet,
654 .write_trailer = rtp_write_trailer,
655 .priv_class = &rtp_muxer_class,
656 .flags = AVFMT_TS_NONSTRICT,