3 * Copyright (c) 2002 Fabrice Bellard
5 * This file is part of FFmpeg.
7 * FFmpeg 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 * FFmpeg 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 FFmpeg; 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;
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 %s\n", avcodec_get_name(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 != 0 && s1->start_time_realtime != AV_NOPTS_VALUE)
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 if (s1->flags & AVFMT_FLAG_BITEXACT) {
135 s->seq = av_get_random_seed() & 0x0fff;
137 s->seq &= 0xffff; // Use the given parameter, wrapped to the right interval
139 if (s1->packet_size) {
140 if (s1->pb->max_packet_size)
141 s1->packet_size = FFMIN(s1->packet_size,
142 s1->pb->max_packet_size);
144 s1->packet_size = s1->pb->max_packet_size;
145 if (s1->packet_size <= 12) {
146 av_log(s1, AV_LOG_ERROR, "Max packet size %d too low\n", s1->packet_size);
149 s->buf = av_malloc(s1->packet_size);
151 return AVERROR(ENOMEM);
153 s->max_payload_size = s1->packet_size - 12;
155 s->max_frames_per_packet = 0;
156 if (s1->max_delay > 0) {
157 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
158 int frame_size = av_get_audio_frame_duration(st->codec, 0);
160 frame_size = st->codec->frame_size;
161 if (frame_size == 0) {
162 av_log(s1, AV_LOG_ERROR, "Cannot respect max delay: frame size = 0\n");
164 s->max_frames_per_packet =
165 av_rescale_q_rnd(s1->max_delay,
167 (AVRational){ frame_size, st->codec->sample_rate },
171 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
172 /* FIXME: We should round down here... */
173 if (st->avg_frame_rate.num > 0 && st->avg_frame_rate.den > 0) {
174 s->max_frames_per_packet = av_rescale_q(s1->max_delay,
175 (AVRational){1, 1000000},
176 av_inv_q(st->avg_frame_rate));
178 s->max_frames_per_packet = 1;
182 avpriv_set_pts_info(st, 32, 1, 90000);
183 switch(st->codec->codec_id) {
184 case AV_CODEC_ID_MP2:
185 case AV_CODEC_ID_MP3:
186 s->buf_ptr = s->buf + 4;
188 case AV_CODEC_ID_MPEG1VIDEO:
189 case AV_CODEC_ID_MPEG2VIDEO:
191 case AV_CODEC_ID_MPEG2TS:
192 n = s->max_payload_size / TS_PACKET_SIZE;
195 s->max_payload_size = n * TS_PACKET_SIZE;
198 case AV_CODEC_ID_H264:
199 /* check for H.264 MP4 syntax */
200 if (st->codec->extradata_size > 4 && st->codec->extradata[0] == 1) {
201 s->nal_length_size = (st->codec->extradata[4] & 0x03) + 1;
204 case AV_CODEC_ID_HEVC:
205 /* Only check for the standardized hvcC version of extradata, keeping
206 * things simple and similar to the avcC/H264 case above, instead
207 * of trying to handle the pre-standardization versions (as in
208 * libavcodec/hevc.c). */
209 if (st->codec->extradata_size > 21 && st->codec->extradata[0] == 1) {
210 s->nal_length_size = (st->codec->extradata[21] & 0x03) + 1;
213 case AV_CODEC_ID_VORBIS:
214 case AV_CODEC_ID_THEORA:
215 if (!s->max_frames_per_packet) s->max_frames_per_packet = 15;
216 s->max_frames_per_packet = av_clip(s->max_frames_per_packet, 1, 15);
217 s->max_payload_size -= 6; // ident+frag+tdt/vdt+pkt_num+pkt_length
220 case AV_CODEC_ID_ADPCM_G722:
221 /* Due to a historical error, the clock rate for G722 in RTP is
222 * 8000, even if the sample rate is 16000. See RFC 3551. */
223 avpriv_set_pts_info(st, 32, 1, 8000);
225 case AV_CODEC_ID_OPUS:
226 if (st->codec->channels > 2) {
227 av_log(s1, AV_LOG_ERROR, "Multistream opus not supported in RTP\n");
230 /* The opus RTP RFC says that all opus streams should use 48000 Hz
231 * as clock rate, since all opus sample rates can be expressed in
232 * this clock rate, and sample rate changes on the fly are supported. */
233 avpriv_set_pts_info(st, 32, 1, 48000);
235 case AV_CODEC_ID_ILBC:
236 if (st->codec->block_align != 38 && st->codec->block_align != 50) {
237 av_log(s1, AV_LOG_ERROR, "Incorrect iLBC block size specified\n");
240 if (!s->max_frames_per_packet)
241 s->max_frames_per_packet = 1;
242 s->max_frames_per_packet = FFMIN(s->max_frames_per_packet,
243 s->max_payload_size / st->codec->block_align);
245 case AV_CODEC_ID_AMR_NB:
246 case AV_CODEC_ID_AMR_WB:
247 if (!s->max_frames_per_packet)
248 s->max_frames_per_packet = 12;
249 if (st->codec->codec_id == AV_CODEC_ID_AMR_NB)
253 /* max_header_toc_size + the largest AMR payload must fit */
254 if (1 + s->max_frames_per_packet + n > s->max_payload_size) {
255 av_log(s1, AV_LOG_ERROR, "RTP max payload size too small for AMR\n");
258 if (st->codec->channels != 1) {
259 av_log(s1, AV_LOG_ERROR, "Only mono is supported\n");
262 case AV_CODEC_ID_AAC:
266 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
267 avpriv_set_pts_info(st, 32, 1, st->codec->sample_rate);
277 return AVERROR(EINVAL);
280 /* send an rtcp sender report packet */
281 static void rtcp_send_sr(AVFormatContext *s1, int64_t ntp_time, int bye)
283 RTPMuxContext *s = s1->priv_data;
286 av_dlog(s1, "RTCP: %02x %"PRIx64" %x\n", s->payload_type, ntp_time, s->timestamp);
288 s->last_rtcp_ntp_time = ntp_time;
289 rtp_ts = av_rescale_q(ntp_time - s->first_rtcp_ntp_time, (AVRational){1, 1000000},
290 s1->streams[0]->time_base) + s->base_timestamp;
291 avio_w8(s1->pb, RTP_VERSION << 6);
292 avio_w8(s1->pb, RTCP_SR);
293 avio_wb16(s1->pb, 6); /* length in words - 1 */
294 avio_wb32(s1->pb, s->ssrc);
295 avio_wb64(s1->pb, NTP_TO_RTP_FORMAT(ntp_time));
296 avio_wb32(s1->pb, rtp_ts);
297 avio_wb32(s1->pb, s->packet_count);
298 avio_wb32(s1->pb, s->octet_count);
301 int len = FFMIN(strlen(s->cname), 255);
302 avio_w8(s1->pb, (RTP_VERSION << 6) + 1);
303 avio_w8(s1->pb, RTCP_SDES);
304 avio_wb16(s1->pb, (7 + len + 3) / 4); /* length in words - 1 */
306 avio_wb32(s1->pb, s->ssrc);
307 avio_w8(s1->pb, 0x01); /* CNAME */
308 avio_w8(s1->pb, len);
309 avio_write(s1->pb, s->cname, len);
310 avio_w8(s1->pb, 0); /* END */
311 for (len = (7 + len) % 4; len % 4; len++)
316 avio_w8(s1->pb, (RTP_VERSION << 6) | 1);
317 avio_w8(s1->pb, RTCP_BYE);
318 avio_wb16(s1->pb, 1); /* length in words - 1 */
319 avio_wb32(s1->pb, s->ssrc);
325 /* send an rtp packet. sequence number is incremented, but the caller
326 must update the timestamp itself */
327 void ff_rtp_send_data(AVFormatContext *s1, const uint8_t *buf1, int len, int m)
329 RTPMuxContext *s = s1->priv_data;
331 av_dlog(s1, "rtp_send_data size=%d\n", len);
333 /* build the RTP header */
334 avio_w8(s1->pb, RTP_VERSION << 6);
335 avio_w8(s1->pb, (s->payload_type & 0x7f) | ((m & 0x01) << 7));
336 avio_wb16(s1->pb, s->seq);
337 avio_wb32(s1->pb, s->timestamp);
338 avio_wb32(s1->pb, s->ssrc);
340 avio_write(s1->pb, buf1, len);
343 s->seq = (s->seq + 1) & 0xffff;
344 s->octet_count += len;
348 /* send an integer number of samples and compute time stamp and fill
349 the rtp send buffer before sending. */
350 static int rtp_send_samples(AVFormatContext *s1,
351 const uint8_t *buf1, int size, int sample_size_bits)
353 RTPMuxContext *s = s1->priv_data;
354 int len, max_packet_size, n;
355 /* Calculate the number of bytes to get samples aligned on a byte border */
356 int aligned_samples_size = sample_size_bits/av_gcd(sample_size_bits, 8);
358 max_packet_size = (s->max_payload_size / aligned_samples_size) * aligned_samples_size;
359 /* Not needed, but who knows. Don't check if samples aren't an even number of bytes. */
360 if ((sample_size_bits % 8) == 0 && ((8 * size) % sample_size_bits) != 0)
361 return AVERROR(EINVAL);
365 len = FFMIN(max_packet_size, size);
368 memcpy(s->buf_ptr, buf1, len);
372 s->timestamp = s->cur_timestamp + n * 8 / sample_size_bits;
373 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
374 n += (s->buf_ptr - s->buf);
379 static void rtp_send_mpegaudio(AVFormatContext *s1,
380 const uint8_t *buf1, int size)
382 RTPMuxContext *s = s1->priv_data;
383 int len, count, max_packet_size;
385 max_packet_size = s->max_payload_size;
387 /* test if we must flush because not enough space */
388 len = (s->buf_ptr - s->buf);
389 if ((len + size) > max_packet_size) {
391 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
392 s->buf_ptr = s->buf + 4;
395 if (s->buf_ptr == s->buf + 4) {
396 s->timestamp = s->cur_timestamp;
400 if (size > max_packet_size) {
401 /* big packet: fragment */
404 len = max_packet_size - 4;
407 /* build fragmented packet */
410 s->buf[2] = count >> 8;
412 memcpy(s->buf + 4, buf1, len);
413 ff_rtp_send_data(s1, s->buf, len + 4, 0);
419 if (s->buf_ptr == s->buf + 4) {
420 /* no fragmentation possible */
426 memcpy(s->buf_ptr, buf1, size);
431 static void rtp_send_raw(AVFormatContext *s1,
432 const uint8_t *buf1, int size)
434 RTPMuxContext *s = s1->priv_data;
435 int len, max_packet_size;
437 max_packet_size = s->max_payload_size;
440 len = max_packet_size;
444 s->timestamp = s->cur_timestamp;
445 ff_rtp_send_data(s1, buf1, len, (len == size));
452 /* NOTE: size is assumed to be an integer multiple of TS_PACKET_SIZE */
453 static void rtp_send_mpegts_raw(AVFormatContext *s1,
454 const uint8_t *buf1, int size)
456 RTPMuxContext *s = s1->priv_data;
459 s->timestamp = s->cur_timestamp;
460 while (size >= TS_PACKET_SIZE) {
461 len = s->max_payload_size - (s->buf_ptr - s->buf);
464 memcpy(s->buf_ptr, buf1, len);
469 out_len = s->buf_ptr - s->buf;
470 if (out_len >= s->max_payload_size) {
471 ff_rtp_send_data(s1, s->buf, out_len, 0);
477 static int rtp_send_ilbc(AVFormatContext *s1, const uint8_t *buf, int size)
479 RTPMuxContext *s = s1->priv_data;
480 AVStream *st = s1->streams[0];
481 int frame_duration = av_get_audio_frame_duration(st->codec, 0);
482 int frame_size = st->codec->block_align;
483 int frames = size / frame_size;
486 int n = FFMIN(s->max_frames_per_packet - s->num_frames, frames);
488 if (!s->num_frames) {
490 s->timestamp = s->cur_timestamp;
492 memcpy(s->buf_ptr, buf, n * frame_size);
495 s->buf_ptr += n * frame_size;
496 buf += n * frame_size;
497 s->cur_timestamp += n * frame_duration;
499 if (s->num_frames == s->max_frames_per_packet) {
500 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
507 static int rtp_write_packet(AVFormatContext *s1, AVPacket *pkt)
509 RTPMuxContext *s = s1->priv_data;
510 AVStream *st = s1->streams[0];
514 av_dlog(s1, "%d: write len=%d\n", pkt->stream_index, size);
516 rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
518 if ((s->first_packet || ((rtcp_bytes >= RTCP_SR_SIZE) &&
519 (ff_ntp_time() - s->last_rtcp_ntp_time > 5000000))) &&
520 !(s->flags & FF_RTP_FLAG_SKIP_RTCP)) {
521 rtcp_send_sr(s1, ff_ntp_time(), 0);
522 s->last_octet_count = s->octet_count;
525 s->cur_timestamp = s->base_timestamp + pkt->pts;
527 switch(st->codec->codec_id) {
528 case AV_CODEC_ID_PCM_MULAW:
529 case AV_CODEC_ID_PCM_ALAW:
530 case AV_CODEC_ID_PCM_U8:
531 case AV_CODEC_ID_PCM_S8:
532 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
533 case AV_CODEC_ID_PCM_U16BE:
534 case AV_CODEC_ID_PCM_U16LE:
535 case AV_CODEC_ID_PCM_S16BE:
536 case AV_CODEC_ID_PCM_S16LE:
537 return rtp_send_samples(s1, pkt->data, size, 16 * st->codec->channels);
538 case AV_CODEC_ID_ADPCM_G722:
539 /* The actual sample size is half a byte per sample, but since the
540 * stream clock rate is 8000 Hz while the sample rate is 16000 Hz,
541 * the correct parameter for send_samples_bits is 8 bits per stream
543 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
544 case AV_CODEC_ID_ADPCM_G726:
545 return rtp_send_samples(s1, pkt->data, size,
546 st->codec->bits_per_coded_sample * st->codec->channels);
547 case AV_CODEC_ID_MP2:
548 case AV_CODEC_ID_MP3:
549 rtp_send_mpegaudio(s1, pkt->data, size);
551 case AV_CODEC_ID_MPEG1VIDEO:
552 case AV_CODEC_ID_MPEG2VIDEO:
553 ff_rtp_send_mpegvideo(s1, pkt->data, size);
555 case AV_CODEC_ID_AAC:
556 if (s->flags & FF_RTP_FLAG_MP4A_LATM)
557 ff_rtp_send_latm(s1, pkt->data, size);
559 ff_rtp_send_aac(s1, pkt->data, size);
561 case AV_CODEC_ID_AMR_NB:
562 case AV_CODEC_ID_AMR_WB:
563 ff_rtp_send_amr(s1, pkt->data, size);
565 case AV_CODEC_ID_MPEG2TS:
566 rtp_send_mpegts_raw(s1, pkt->data, size);
568 case AV_CODEC_ID_H264:
569 ff_rtp_send_h264(s1, pkt->data, size);
571 case AV_CODEC_ID_H261:
572 ff_rtp_send_h261(s1, pkt->data, size);
574 case AV_CODEC_ID_H263:
575 if (s->flags & FF_RTP_FLAG_RFC2190) {
576 int mb_info_size = 0;
577 const uint8_t *mb_info =
578 av_packet_get_side_data(pkt, AV_PKT_DATA_H263_MB_INFO,
581 av_log(s1, AV_LOG_ERROR, "failed to allocate side data\n");
582 return AVERROR(ENOMEM);
584 ff_rtp_send_h263_rfc2190(s1, pkt->data, size, mb_info, mb_info_size);
588 case AV_CODEC_ID_H263P:
589 ff_rtp_send_h263(s1, pkt->data, size);
591 case AV_CODEC_ID_HEVC:
592 ff_rtp_send_hevc(s1, pkt->data, size);
594 case AV_CODEC_ID_VORBIS:
595 case AV_CODEC_ID_THEORA:
596 ff_rtp_send_xiph(s1, pkt->data, size);
598 case AV_CODEC_ID_VP8:
599 ff_rtp_send_vp8(s1, pkt->data, size);
601 case AV_CODEC_ID_ILBC:
602 rtp_send_ilbc(s1, pkt->data, size);
604 case AV_CODEC_ID_MJPEG:
605 ff_rtp_send_jpeg(s1, pkt->data, size);
607 case AV_CODEC_ID_OPUS:
608 if (size > s->max_payload_size) {
609 av_log(s1, AV_LOG_ERROR,
610 "Packet size %d too large for max RTP payload size %d\n",
611 size, s->max_payload_size);
612 return AVERROR(EINVAL);
614 /* Intentional fallthrough */
616 /* better than nothing : send the codec raw data */
617 rtp_send_raw(s1, pkt->data, size);
623 static int rtp_write_trailer(AVFormatContext *s1)
625 RTPMuxContext *s = s1->priv_data;
627 /* If the caller closes and recreates ->pb, this might actually
628 * be NULL here even if it was successfully allocated at the start. */
629 if (s1->pb && (s->flags & FF_RTP_FLAG_SEND_BYE))
630 rtcp_send_sr(s1, ff_ntp_time(), 1);
636 AVOutputFormat ff_rtp_muxer = {
638 .long_name = NULL_IF_CONFIG_SMALL("RTP output"),
639 .priv_data_size = sizeof(RTPMuxContext),
640 .audio_codec = AV_CODEC_ID_PCM_MULAW,
641 .video_codec = AV_CODEC_ID_MPEG4,
642 .write_header = rtp_write_header,
643 .write_packet = rtp_write_packet,
644 .write_trailer = rtp_write_trailer,
645 .priv_class = &rtp_muxer_class,
646 .flags = AVFMT_TS_NONSTRICT,