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_H263:
53 case AV_CODEC_ID_H263P:
54 case AV_CODEC_ID_H264:
55 case AV_CODEC_ID_HEVC:
56 case AV_CODEC_ID_MPEG1VIDEO:
57 case AV_CODEC_ID_MPEG2VIDEO:
58 case AV_CODEC_ID_MPEG4:
62 case AV_CODEC_ID_PCM_ALAW:
63 case AV_CODEC_ID_PCM_MULAW:
64 case AV_CODEC_ID_PCM_S8:
65 case AV_CODEC_ID_PCM_S16BE:
66 case AV_CODEC_ID_PCM_S16LE:
67 case AV_CODEC_ID_PCM_U16BE:
68 case AV_CODEC_ID_PCM_U16LE:
69 case AV_CODEC_ID_PCM_U8:
70 case AV_CODEC_ID_MPEG2TS:
71 case AV_CODEC_ID_AMR_NB:
72 case AV_CODEC_ID_AMR_WB:
73 case AV_CODEC_ID_VORBIS:
74 case AV_CODEC_ID_THEORA:
76 case AV_CODEC_ID_ADPCM_G722:
77 case AV_CODEC_ID_ADPCM_G726:
78 case AV_CODEC_ID_ILBC:
79 case AV_CODEC_ID_MJPEG:
80 case AV_CODEC_ID_SPEEX:
81 case AV_CODEC_ID_OPUS:
88 static int rtp_write_header(AVFormatContext *s1)
90 RTPMuxContext *s = s1->priv_data;
94 if (s1->nb_streams != 1) {
95 av_log(s1, AV_LOG_ERROR, "Only one stream supported in the RTP muxer\n");
96 return AVERROR(EINVAL);
99 if (!is_supported(st->codec->codec_id)) {
100 av_log(s1, AV_LOG_ERROR, "Unsupported codec %x\n", st->codec->codec_id);
105 if (s->payload_type < 0) {
106 /* Re-validate non-dynamic payload types */
107 if (st->id < RTP_PT_PRIVATE)
108 st->id = ff_rtp_get_payload_type(s1, st->codec, -1);
110 s->payload_type = st->id;
112 /* private option takes priority */
113 st->id = s->payload_type;
116 s->base_timestamp = av_get_random_seed();
117 s->timestamp = s->base_timestamp;
118 s->cur_timestamp = 0;
120 s->ssrc = av_get_random_seed();
122 s->first_rtcp_ntp_time = ff_ntp_time();
123 if (s1->start_time_realtime)
124 /* Round the NTP time to whole milliseconds. */
125 s->first_rtcp_ntp_time = (s1->start_time_realtime / 1000) * 1000 +
127 // Pick a random sequence start number, but in the lower end of the
128 // available range, so that any wraparound doesn't happen immediately.
129 // (Immediate wraparound would be an issue for SRTP.)
131 s->seq = av_get_random_seed() & 0x0fff;
133 s->seq &= 0xffff; // Use the given parameter, wrapped to the right interval
135 if (s1->packet_size) {
136 if (s1->pb->max_packet_size)
137 s1->packet_size = FFMIN(s1->packet_size,
138 s1->pb->max_packet_size);
140 s1->packet_size = s1->pb->max_packet_size;
141 if (s1->packet_size <= 12) {
142 av_log(s1, AV_LOG_ERROR, "Max packet size %d too low\n", s1->packet_size);
145 s->buf = av_malloc(s1->packet_size);
147 return AVERROR(ENOMEM);
149 s->max_payload_size = s1->packet_size - 12;
151 s->max_frames_per_packet = 0;
152 if (s1->max_delay > 0) {
153 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
154 int frame_size = av_get_audio_frame_duration(st->codec, 0);
156 frame_size = st->codec->frame_size;
157 if (frame_size == 0) {
158 av_log(s1, AV_LOG_ERROR, "Cannot respect max delay: frame size = 0\n");
160 s->max_frames_per_packet =
161 av_rescale_q_rnd(s1->max_delay,
163 (AVRational){ frame_size, st->codec->sample_rate },
167 if (st->codec->codec_type == AVMEDIA_TYPE_VIDEO) {
168 /* FIXME: We should round down here... */
169 if (st->avg_frame_rate.num > 0 && st->avg_frame_rate.den > 0) {
170 s->max_frames_per_packet = av_rescale_q(s1->max_delay,
171 (AVRational){1, 1000000},
172 av_inv_q(st->avg_frame_rate));
174 s->max_frames_per_packet = 1;
178 avpriv_set_pts_info(st, 32, 1, 90000);
179 switch(st->codec->codec_id) {
180 case AV_CODEC_ID_MP2:
181 case AV_CODEC_ID_MP3:
182 s->buf_ptr = s->buf + 4;
184 case AV_CODEC_ID_MPEG1VIDEO:
185 case AV_CODEC_ID_MPEG2VIDEO:
187 case AV_CODEC_ID_MPEG2TS:
188 n = s->max_payload_size / TS_PACKET_SIZE;
191 s->max_payload_size = n * TS_PACKET_SIZE;
194 case AV_CODEC_ID_H264:
195 /* check for H.264 MP4 syntax */
196 if (st->codec->extradata_size > 4 && st->codec->extradata[0] == 1) {
197 s->nal_length_size = (st->codec->extradata[4] & 0x03) + 1;
200 case AV_CODEC_ID_HEVC:
201 /* Only check for the standardized hvcC version of extradata, keeping
202 * things simple and similar to the avcC/H264 case above, instead
203 * of trying to handle the pre-standardization versions (as in
204 * libavcodec/hevc.c). */
205 if (st->codec->extradata_size > 21 && st->codec->extradata[0] == 1) {
206 s->nal_length_size = (st->codec->extradata[21] & 0x03) + 1;
209 case AV_CODEC_ID_VORBIS:
210 case AV_CODEC_ID_THEORA:
211 if (!s->max_frames_per_packet) s->max_frames_per_packet = 15;
212 s->max_frames_per_packet = av_clip(s->max_frames_per_packet, 1, 15);
213 s->max_payload_size -= 6; // ident+frag+tdt/vdt+pkt_num+pkt_length
216 case AV_CODEC_ID_ADPCM_G722:
217 /* Due to a historical error, the clock rate for G722 in RTP is
218 * 8000, even if the sample rate is 16000. See RFC 3551. */
219 avpriv_set_pts_info(st, 32, 1, 8000);
221 case AV_CODEC_ID_OPUS:
222 if (st->codec->channels > 2) {
223 av_log(s1, AV_LOG_ERROR, "Multistream opus not supported in RTP\n");
226 /* The opus RTP RFC says that all opus streams should use 48000 Hz
227 * as clock rate, since all opus sample rates can be expressed in
228 * this clock rate, and sample rate changes on the fly are supported. */
229 avpriv_set_pts_info(st, 32, 1, 48000);
231 case AV_CODEC_ID_ILBC:
232 if (st->codec->block_align != 38 && st->codec->block_align != 50) {
233 av_log(s1, AV_LOG_ERROR, "Incorrect iLBC block size specified\n");
236 if (!s->max_frames_per_packet)
237 s->max_frames_per_packet = 1;
238 s->max_frames_per_packet = FFMIN(s->max_frames_per_packet,
239 s->max_payload_size / st->codec->block_align);
241 case AV_CODEC_ID_AMR_NB:
242 case AV_CODEC_ID_AMR_WB:
243 if (!s->max_frames_per_packet)
244 s->max_frames_per_packet = 12;
245 if (st->codec->codec_id == AV_CODEC_ID_AMR_NB)
249 /* max_header_toc_size + the largest AMR payload must fit */
250 if (1 + s->max_frames_per_packet + n > s->max_payload_size) {
251 av_log(s1, AV_LOG_ERROR, "RTP max payload size too small for AMR\n");
254 if (st->codec->channels != 1) {
255 av_log(s1, AV_LOG_ERROR, "Only mono is supported\n");
260 case AV_CODEC_ID_AAC:
265 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
266 avpriv_set_pts_info(st, 32, 1, st->codec->sample_rate);
276 return AVERROR(EINVAL);
279 /* send an rtcp sender report packet */
280 static void rtcp_send_sr(AVFormatContext *s1, int64_t ntp_time, int bye)
282 RTPMuxContext *s = s1->priv_data;
285 av_dlog(s1, "RTCP: %02x %"PRIx64" %x\n", s->payload_type, ntp_time, s->timestamp);
287 s->last_rtcp_ntp_time = ntp_time;
288 rtp_ts = av_rescale_q(ntp_time - s->first_rtcp_ntp_time, (AVRational){1, 1000000},
289 s1->streams[0]->time_base) + s->base_timestamp;
290 avio_w8(s1->pb, RTP_VERSION << 6);
291 avio_w8(s1->pb, RTCP_SR);
292 avio_wb16(s1->pb, 6); /* length in words - 1 */
293 avio_wb32(s1->pb, s->ssrc);
294 avio_wb64(s1->pb, NTP_TO_RTP_FORMAT(ntp_time));
295 avio_wb32(s1->pb, rtp_ts);
296 avio_wb32(s1->pb, s->packet_count);
297 avio_wb32(s1->pb, s->octet_count);
300 int len = FFMIN(strlen(s->cname), 255);
301 avio_w8(s1->pb, (RTP_VERSION << 6) + 1);
302 avio_w8(s1->pb, RTCP_SDES);
303 avio_wb16(s1->pb, (7 + len + 3) / 4); /* length in words - 1 */
305 avio_wb32(s1->pb, s->ssrc);
306 avio_w8(s1->pb, 0x01); /* CNAME */
307 avio_w8(s1->pb, len);
308 avio_write(s1->pb, s->cname, len);
309 avio_w8(s1->pb, 0); /* END */
310 for (len = (7 + len) % 4; len % 4; len++)
315 avio_w8(s1->pb, (RTP_VERSION << 6) | 1);
316 avio_w8(s1->pb, RTCP_BYE);
317 avio_wb16(s1->pb, 1); /* length in words - 1 */
318 avio_wb32(s1->pb, s->ssrc);
324 /* send an rtp packet. sequence number is incremented, but the caller
325 must update the timestamp itself */
326 void ff_rtp_send_data(AVFormatContext *s1, const uint8_t *buf1, int len, int m)
328 RTPMuxContext *s = s1->priv_data;
330 av_dlog(s1, "rtp_send_data size=%d\n", len);
332 /* build the RTP header */
333 avio_w8(s1->pb, RTP_VERSION << 6);
334 avio_w8(s1->pb, (s->payload_type & 0x7f) | ((m & 0x01) << 7));
335 avio_wb16(s1->pb, s->seq);
336 avio_wb32(s1->pb, s->timestamp);
337 avio_wb32(s1->pb, s->ssrc);
339 avio_write(s1->pb, buf1, len);
342 s->seq = (s->seq + 1) & 0xffff;
343 s->octet_count += len;
347 /* send an integer number of samples and compute time stamp and fill
348 the rtp send buffer before sending. */
349 static int rtp_send_samples(AVFormatContext *s1,
350 const uint8_t *buf1, int size, int sample_size_bits)
352 RTPMuxContext *s = s1->priv_data;
353 int len, max_packet_size, n;
354 /* Calculate the number of bytes to get samples aligned on a byte border */
355 int aligned_samples_size = sample_size_bits/av_gcd(sample_size_bits, 8);
357 max_packet_size = (s->max_payload_size / aligned_samples_size) * aligned_samples_size;
358 /* Not needed, but who knows. Don't check if samples aren't an even number of bytes. */
359 if ((sample_size_bits % 8) == 0 && ((8 * size) % sample_size_bits) != 0)
360 return AVERROR(EINVAL);
364 len = FFMIN(max_packet_size, size);
367 memcpy(s->buf_ptr, buf1, len);
371 s->timestamp = s->cur_timestamp + n * 8 / sample_size_bits;
372 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
373 n += (s->buf_ptr - s->buf);
378 static void rtp_send_mpegaudio(AVFormatContext *s1,
379 const uint8_t *buf1, int size)
381 RTPMuxContext *s = s1->priv_data;
382 int len, count, max_packet_size;
384 max_packet_size = s->max_payload_size;
386 /* test if we must flush because not enough space */
387 len = (s->buf_ptr - s->buf);
388 if ((len + size) > max_packet_size) {
390 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 0);
391 s->buf_ptr = s->buf + 4;
394 if (s->buf_ptr == s->buf + 4) {
395 s->timestamp = s->cur_timestamp;
399 if (size > max_packet_size) {
400 /* big packet: fragment */
403 len = max_packet_size - 4;
406 /* build fragmented packet */
409 s->buf[2] = count >> 8;
411 memcpy(s->buf + 4, buf1, len);
412 ff_rtp_send_data(s1, s->buf, len + 4, 0);
418 if (s->buf_ptr == s->buf + 4) {
419 /* no fragmentation possible */
425 memcpy(s->buf_ptr, buf1, size);
430 static void rtp_send_raw(AVFormatContext *s1,
431 const uint8_t *buf1, int size)
433 RTPMuxContext *s = s1->priv_data;
434 int len, max_packet_size;
436 max_packet_size = s->max_payload_size;
439 len = max_packet_size;
443 s->timestamp = s->cur_timestamp;
444 ff_rtp_send_data(s1, buf1, len, (len == size));
451 /* NOTE: size is assumed to be an integer multiple of TS_PACKET_SIZE */
452 static void rtp_send_mpegts_raw(AVFormatContext *s1,
453 const uint8_t *buf1, int size)
455 RTPMuxContext *s = s1->priv_data;
458 s->timestamp = s->cur_timestamp;
459 while (size >= TS_PACKET_SIZE) {
460 len = s->max_payload_size - (s->buf_ptr - s->buf);
463 memcpy(s->buf_ptr, buf1, len);
468 out_len = s->buf_ptr - s->buf;
469 if (out_len >= s->max_payload_size) {
470 ff_rtp_send_data(s1, s->buf, out_len, 0);
476 static int rtp_send_ilbc(AVFormatContext *s1, const uint8_t *buf, int size)
478 RTPMuxContext *s = s1->priv_data;
479 AVStream *st = s1->streams[0];
480 int frame_duration = av_get_audio_frame_duration(st->codec, 0);
481 int frame_size = st->codec->block_align;
482 int frames = size / frame_size;
485 int n = FFMIN(s->max_frames_per_packet - s->num_frames, frames);
487 if (!s->num_frames) {
489 s->timestamp = s->cur_timestamp;
491 memcpy(s->buf_ptr, buf, n * frame_size);
494 s->buf_ptr += n * frame_size;
495 buf += n * frame_size;
496 s->cur_timestamp += n * frame_duration;
498 if (s->num_frames == s->max_frames_per_packet) {
499 ff_rtp_send_data(s1, s->buf, s->buf_ptr - s->buf, 1);
506 static int rtp_write_packet(AVFormatContext *s1, AVPacket *pkt)
508 RTPMuxContext *s = s1->priv_data;
509 AVStream *st = s1->streams[0];
513 av_dlog(s1, "%d: write len=%d\n", pkt->stream_index, size);
515 rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
517 if ((s->first_packet || ((rtcp_bytes >= RTCP_SR_SIZE) &&
518 (ff_ntp_time() - s->last_rtcp_ntp_time > 5000000))) &&
519 !(s->flags & FF_RTP_FLAG_SKIP_RTCP)) {
520 rtcp_send_sr(s1, ff_ntp_time(), 0);
521 s->last_octet_count = s->octet_count;
524 s->cur_timestamp = s->base_timestamp + pkt->pts;
526 switch(st->codec->codec_id) {
527 case AV_CODEC_ID_PCM_MULAW:
528 case AV_CODEC_ID_PCM_ALAW:
529 case AV_CODEC_ID_PCM_U8:
530 case AV_CODEC_ID_PCM_S8:
531 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
532 case AV_CODEC_ID_PCM_U16BE:
533 case AV_CODEC_ID_PCM_U16LE:
534 case AV_CODEC_ID_PCM_S16BE:
535 case AV_CODEC_ID_PCM_S16LE:
536 return rtp_send_samples(s1, pkt->data, size, 16 * st->codec->channels);
537 case AV_CODEC_ID_ADPCM_G722:
538 /* The actual sample size is half a byte per sample, but since the
539 * stream clock rate is 8000 Hz while the sample rate is 16000 Hz,
540 * the correct parameter for send_samples_bits is 8 bits per stream
542 return rtp_send_samples(s1, pkt->data, size, 8 * st->codec->channels);
543 case AV_CODEC_ID_ADPCM_G726:
544 return rtp_send_samples(s1, pkt->data, size,
545 st->codec->bits_per_coded_sample * st->codec->channels);
546 case AV_CODEC_ID_MP2:
547 case AV_CODEC_ID_MP3:
548 rtp_send_mpegaudio(s1, pkt->data, size);
550 case AV_CODEC_ID_MPEG1VIDEO:
551 case AV_CODEC_ID_MPEG2VIDEO:
552 ff_rtp_send_mpegvideo(s1, pkt->data, size);
554 case AV_CODEC_ID_AAC:
555 if (s->flags & FF_RTP_FLAG_MP4A_LATM)
556 ff_rtp_send_latm(s1, pkt->data, size);
558 ff_rtp_send_aac(s1, pkt->data, size);
560 case AV_CODEC_ID_AMR_NB:
561 case AV_CODEC_ID_AMR_WB:
562 ff_rtp_send_amr(s1, pkt->data, size);
564 case AV_CODEC_ID_MPEG2TS:
565 rtp_send_mpegts_raw(s1, pkt->data, size);
567 case AV_CODEC_ID_H264:
568 ff_rtp_send_h264(s1, pkt->data, size);
570 case AV_CODEC_ID_H263:
571 if (s->flags & FF_RTP_FLAG_RFC2190) {
572 int mb_info_size = 0;
573 const uint8_t *mb_info =
574 av_packet_get_side_data(pkt, AV_PKT_DATA_H263_MB_INFO,
576 ff_rtp_send_h263_rfc2190(s1, pkt->data, size, mb_info, mb_info_size);
580 case AV_CODEC_ID_H263P:
581 ff_rtp_send_h263(s1, pkt->data, size);
583 case AV_CODEC_ID_HEVC:
584 ff_rtp_send_hevc(s1, pkt->data, size);
586 case AV_CODEC_ID_VORBIS:
587 case AV_CODEC_ID_THEORA:
588 ff_rtp_send_xiph(s1, pkt->data, size);
590 case AV_CODEC_ID_VP8:
591 ff_rtp_send_vp8(s1, pkt->data, size);
593 case AV_CODEC_ID_ILBC:
594 rtp_send_ilbc(s1, pkt->data, size);
596 case AV_CODEC_ID_MJPEG:
597 ff_rtp_send_jpeg(s1, pkt->data, size);
599 case AV_CODEC_ID_OPUS:
600 if (size > s->max_payload_size) {
601 av_log(s1, AV_LOG_ERROR,
602 "Packet size %d too large for max RTP payload size %d\n",
603 size, s->max_payload_size);
604 return AVERROR(EINVAL);
606 /* Intentional fallthrough */
608 /* better than nothing : send the codec raw data */
609 rtp_send_raw(s1, pkt->data, size);
615 static int rtp_write_trailer(AVFormatContext *s1)
617 RTPMuxContext *s = s1->priv_data;
619 /* If the caller closes and recreates ->pb, this might actually
620 * be NULL here even if it was successfully allocated at the start. */
621 if (s1->pb && (s->flags & FF_RTP_FLAG_SEND_BYE))
622 rtcp_send_sr(s1, ff_ntp_time(), 1);
628 AVOutputFormat ff_rtp_muxer = {
630 .long_name = NULL_IF_CONFIG_SMALL("RTP output"),
631 .priv_data_size = sizeof(RTPMuxContext),
632 .audio_codec = AV_CODEC_ID_PCM_MULAW,
633 .video_codec = AV_CODEC_ID_MPEG4,
634 .write_header = rtp_write_header,
635 .write_packet = rtp_write_packet,
636 .write_trailer = rtp_write_trailer,
637 .priv_class = &rtp_muxer_class,
638 .flags = AVFMT_TS_NONSTRICT,