2 * copyright (c) 2007 Luca Abeni
4 * This file is part of Libav.
6 * Libav is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
11 * Libav is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with Libav; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
22 #include "libavutil/avstring.h"
23 #include "libavutil/base64.h"
24 #include "libavutil/dict.h"
25 #include "libavutil/parseutils.h"
26 #include "libavutil/opt.h"
27 #include "libavcodec/xiph.h"
28 #include "libavcodec/mpeg4audio.h"
39 #define MAX_EXTRADATA_SIZE ((INT_MAX - 10) / 2)
41 struct sdp_session_level {
42 int sdp_version; /**< protocol version (currently 0) */
43 int id; /**< session ID */
44 int version; /**< session version */
45 int start_time; /**< session start time (NTP time, in seconds),
46 or 0 in case of permanent session */
47 int end_time; /**< session end time (NTP time, in seconds),
48 or 0 if the session is not bounded */
49 int ttl; /**< TTL, in case of multicast stream */
50 const char *user; /**< username of the session's creator */
51 const char *src_addr; /**< IP address of the machine from which the session was created */
52 const char *src_type; /**< address type of src_addr */
53 const char *dst_addr; /**< destination IP address (can be multicast) */
54 const char *dst_type; /**< destination IP address type */
55 const char *name; /**< session name (can be an empty string) */
58 static void sdp_write_address(char *buff, int size, const char *dest_addr,
59 const char *dest_type, int ttl)
64 if (ttl > 0 && !strcmp(dest_type, "IP4")) {
65 /* The TTL should only be specified for IPv4 multicast addresses,
67 av_strlcatf(buff, size, "c=IN %s %s/%d\r\n", dest_type, dest_addr, ttl);
69 av_strlcatf(buff, size, "c=IN %s %s\r\n", dest_type, dest_addr);
74 static void sdp_write_header(char *buff, int size, struct sdp_session_level *s)
76 av_strlcatf(buff, size, "v=%d\r\n"
77 "o=- %d %d IN %s %s\r\n"
80 s->id, s->version, s->src_type, s->src_addr,
82 sdp_write_address(buff, size, s->dst_addr, s->dst_type, s->ttl);
83 av_strlcatf(buff, size, "t=%d %d\r\n"
84 "a=tool:libavformat " AV_STRINGIFY(LIBAVFORMAT_VERSION) "\r\n",
85 s->start_time, s->end_time);
89 static int resolve_destination(char *dest_addr, int size, char *type,
92 struct addrinfo hints = { 0 }, *ai;
95 av_strlcpy(type, "IP4", type_size);
99 /* Resolve the destination, since it must be written
100 * as a numeric IP address in the SDP. */
102 if (getaddrinfo(dest_addr, NULL, &hints, &ai))
104 getnameinfo(ai->ai_addr, ai->ai_addrlen, dest_addr, size,
105 NULL, 0, NI_NUMERICHOST);
107 if (ai->ai_family == AF_INET6)
108 av_strlcpy(type, "IP6", type_size);
110 is_multicast = ff_is_multicast_address(ai->ai_addr);
115 static int resolve_destination(char *dest_addr, int size, char *type,
122 static int sdp_get_address(char *dest_addr, int size, int *ttl, const char *url)
128 av_url_split(proto, sizeof(proto), NULL, 0, dest_addr, size, &port, NULL, 0, url);
132 if (strcmp(proto, "rtp") && strcmp(proto, "srtp")) {
133 /* The url isn't for the actual rtp sessions,
134 * don't parse out anything else than the destination.
139 p = strchr(url, '?');
143 if (av_find_info_tag(buff, sizeof(buff), "ttl", p)) {
144 *ttl = strtol(buff, NULL, 10);
153 #define MAX_PSET_SIZE 1024
154 static char *extradata2psets(AVCodecContext *c)
158 static const char pset_string[] = "; sprop-parameter-sets=";
159 static const char profile_string[] = "; profile-level-id=";
160 uint8_t *extradata = c->extradata;
161 int extradata_size = c->extradata_size;
162 uint8_t *tmpbuf = NULL;
163 const uint8_t *sps = NULL, *sps_end;
165 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
166 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
170 if (c->extradata[0] == 1) {
171 if (ff_avc_write_annexb_extradata(c->extradata, &extradata,
177 psets = av_mallocz(MAX_PSET_SIZE);
179 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the parameter sets.\n");
183 memcpy(psets, pset_string, strlen(pset_string));
184 p = psets + strlen(pset_string);
185 r = ff_avc_find_startcode(extradata, extradata + extradata_size);
186 while (r < extradata + extradata_size) {
191 nal_type = *r & 0x1f;
192 r1 = ff_avc_find_startcode(r, extradata + extradata_size);
193 if (nal_type != 7 && nal_type != 8) { /* Only output SPS and PPS */
197 if (p != (psets + strlen(pset_string))) {
205 if (!av_base64_encode(p, MAX_PSET_SIZE - (p - psets), r, r1 - r)) {
206 av_log(c, AV_LOG_ERROR, "Cannot Base64-encode %td %td!\n", MAX_PSET_SIZE - (p - psets), r1 - r);
215 if (sps && sps_end - sps >= 4) {
216 memcpy(p, profile_string, strlen(profile_string));
218 ff_data_to_hex(p, sps + 1, 3, 0);
226 static char *extradata2psets_hevc(AVCodecContext *c)
229 uint8_t *extradata = c->extradata;
230 int extradata_size = c->extradata_size;
231 uint8_t *tmpbuf = NULL;
232 int ps_pos[3] = { 0 };
233 static const char * const ps_names[3] = { "vps", "sps", "pps" };
234 int num_arrays, num_nalus;
237 // Convert to hvcc format. Since we need to group multiple NALUs of
238 // the same type, and we might need to convert from one format to the
239 // other anyway, we get away with a little less work by using the hvcc
241 if (c->extradata[0] != 1) {
243 if (avio_open_dyn_buf(&pb) < 0)
245 if (ff_isom_write_hvcc(pb, c->extradata, c->extradata_size, 0) < 0) {
246 avio_close_dyn_buf(pb, &tmpbuf);
249 extradata_size = avio_close_dyn_buf(pb, &extradata);
253 if (extradata_size < 23)
256 num_arrays = extradata[22];
258 for (i = 0; i < num_arrays; i++) {
259 int num_nalus, nalu_type;
260 if (pos + 3 > extradata_size)
262 nalu_type = extradata[pos] & 0x3f;
263 // Not including libavcodec/hevc.h to avoid confusion between
264 // NAL_* with the same name for both H264 and HEVC.
265 if (nalu_type == 32) // VPS
267 else if (nalu_type == 33) // SPS
269 else if (nalu_type == 34) // PPS
271 num_nalus = AV_RB16(&extradata[pos + 1]);
273 for (j = 0; j < num_nalus; j++) {
275 if (pos + 2 > extradata_size)
277 len = AV_RB16(&extradata[pos]);
279 if (pos + len > extradata_size)
284 if (!ps_pos[0] || !ps_pos[1] || !ps_pos[2])
287 psets = av_mallocz(MAX_PSET_SIZE);
292 for (i = 0; i < 3; i++) {
296 av_strlcat(psets, "; ", MAX_PSET_SIZE);
297 av_strlcatf(psets, MAX_PSET_SIZE, "sprop-%s=", ps_names[i]);
299 // Skipping boundary checks in the input here; we've already traversed
300 // the whole hvcc structure above without issues
301 num_nalus = AV_RB16(&extradata[pos + 1]);
303 for (j = 0; j < num_nalus; j++) {
304 int len = AV_RB16(&extradata[pos]);
308 av_strlcat(psets, ",", MAX_PSET_SIZE);
309 strpos = strlen(psets);
310 if (!av_base64_encode(psets + strpos, MAX_PSET_SIZE - strpos,
311 &extradata[pos], len)) {
327 static char *extradata2config(AVCodecContext *c)
331 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
332 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
336 config = av_malloc(10 + c->extradata_size * 2);
338 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
341 memcpy(config, "; config=", 9);
342 ff_data_to_hex(config + 9, c->extradata, c->extradata_size, 0);
343 config[9 + c->extradata_size * 2] = 0;
348 static char *xiph_extradata2config(AVCodecContext *c)
350 char *config, *encoded_config;
351 uint8_t *header_start[3];
352 int headers_len, header_len[3], config_len;
353 int first_header_size;
355 switch (c->codec_id) {
356 case AV_CODEC_ID_THEORA:
357 first_header_size = 42;
359 case AV_CODEC_ID_VORBIS:
360 first_header_size = 30;
363 av_log(c, AV_LOG_ERROR, "Unsupported Xiph codec ID\n");
367 if (avpriv_split_xiph_headers(c->extradata, c->extradata_size,
368 first_header_size, header_start,
370 av_log(c, AV_LOG_ERROR, "Extradata corrupt.\n");
374 headers_len = header_len[0] + header_len[2];
375 config_len = 4 + // count
380 headers_len; // and the rest
382 config = av_malloc(config_len);
386 encoded_config = av_malloc(AV_BASE64_SIZE(config_len));
387 if (!encoded_config) {
392 config[0] = config[1] = config[2] = 0;
394 config[4] = (RTP_XIPH_IDENT >> 16) & 0xff;
395 config[5] = (RTP_XIPH_IDENT >> 8) & 0xff;
396 config[6] = (RTP_XIPH_IDENT ) & 0xff;
397 config[7] = (headers_len >> 8) & 0xff;
398 config[8] = headers_len & 0xff;
400 config[10] = header_len[0];
401 config[11] = 0; // size of comment header; nonexistent
402 memcpy(config + 12, header_start[0], header_len[0]);
403 memcpy(config + 12 + header_len[0], header_start[2], header_len[2]);
405 av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
409 return encoded_config;
412 av_log(c, AV_LOG_ERROR,
413 "Not enough memory for configuration string\n");
417 static int latm_context2profilelevel(AVCodecContext *c)
420 * The RTP payload format specification is described in RFC 3016
421 * The encoding specifications are provided in ISO/IEC 14496-3 */
423 int profile_level = 0x2B;
425 /* TODO: AAC Profile only supports AAC LC Object Type.
426 * Different Object Types should implement different Profile Levels */
428 if (c->sample_rate <= 24000) {
429 if (c->channels <= 2)
430 profile_level = 0x28; // AAC Profile, Level 1
431 } else if (c->sample_rate <= 48000) {
432 if (c->channels <= 2) {
433 profile_level = 0x29; // AAC Profile, Level 2
434 } else if (c->channels <= 5) {
435 profile_level = 0x2A; // AAC Profile, Level 4
437 } else if (c->sample_rate <= 96000) {
438 if (c->channels <= 5) {
439 profile_level = 0x2B; // AAC Profile, Level 5
443 return profile_level;
446 static char *latm_context2config(AVCodecContext *c)
449 * The RTP payload format specification is described in RFC 3016
450 * The encoding specifications are provided in ISO/IEC 14496-3 */
452 uint8_t config_byte[6];
456 for (rate_index = 0; rate_index < 16; rate_index++)
457 if (avpriv_mpeg4audio_sample_rates[rate_index] == c->sample_rate)
459 if (rate_index == 16) {
460 av_log(c, AV_LOG_ERROR, "Unsupported sample rate\n");
464 config_byte[0] = 0x40;
466 config_byte[2] = 0x20 | rate_index;
467 config_byte[3] = c->channels << 4;
468 config_byte[4] = 0x3f;
469 config_byte[5] = 0xc0;
471 config = av_malloc(6*2+1);
473 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
476 ff_data_to_hex(config, config_byte, 6, 1);
482 static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type, AVFormatContext *fmt)
486 switch (c->codec_id) {
487 case AV_CODEC_ID_H264: {
489 if (fmt && fmt->oformat && fmt->oformat->priv_class &&
490 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "h264_mode0"))
492 if (c->extradata_size) {
493 config = extradata2psets(c);
495 av_strlcatf(buff, size, "a=rtpmap:%d H264/90000\r\n"
496 "a=fmtp:%d packetization-mode=%d%s\r\n",
498 payload_type, mode, config ? config : "");
501 case AV_CODEC_ID_H261:
503 const char *pic_fmt = NULL;
504 /* only QCIF and CIF are specified as supported in RFC 4587 */
505 if (c->width == 176 && c->height == 144)
507 else if (c->width == 352 && c->height == 288)
509 if (payload_type >= RTP_PT_PRIVATE)
510 av_strlcatf(buff, size, "a=rtpmap:%d H261/90000\r\n", payload_type);
512 av_strlcatf(buff, size, "a=fmtp:%d %s\r\n", payload_type, pic_fmt);
515 case AV_CODEC_ID_H263:
516 case AV_CODEC_ID_H263P:
517 /* a=framesize is required by 3GPP TS 26.234 (PSS). It
518 * actually specifies the maximum video size, but we only know
519 * the current size. This is required for playback on Android
520 * stagefright and on Samsung bada. */
521 if (!fmt || !fmt->oformat->priv_class ||
522 !av_opt_flag_is_set(fmt->priv_data, "rtpflags", "rfc2190") ||
523 c->codec_id == AV_CODEC_ID_H263P)
524 av_strlcatf(buff, size, "a=rtpmap:%d H263-2000/90000\r\n"
525 "a=framesize:%d %d-%d\r\n",
527 payload_type, c->width, c->height);
529 case AV_CODEC_ID_HEVC:
530 if (c->extradata_size)
531 config = extradata2psets_hevc(c);
532 av_strlcatf(buff, size, "a=rtpmap:%d H265/90000\r\n", payload_type);
534 av_strlcatf(buff, size, "a=fmtp:%d %s\r\n",
535 payload_type, config);
537 case AV_CODEC_ID_MPEG4:
538 if (c->extradata_size) {
539 config = extradata2config(c);
541 av_strlcatf(buff, size, "a=rtpmap:%d MP4V-ES/90000\r\n"
542 "a=fmtp:%d profile-level-id=1%s\r\n",
544 payload_type, config ? config : "");
546 case AV_CODEC_ID_AAC:
547 if (fmt && fmt->oformat->priv_class &&
548 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "latm")) {
549 config = latm_context2config(c);
552 av_strlcatf(buff, size, "a=rtpmap:%d MP4A-LATM/%d/%d\r\n"
553 "a=fmtp:%d profile-level-id=%d;cpresent=0;config=%s\r\n",
554 payload_type, c->sample_rate, c->channels,
555 payload_type, latm_context2profilelevel(c), config);
557 if (c->extradata_size) {
558 config = extradata2config(c);
560 /* FIXME: maybe we can forge config information based on the
561 * codec parameters...
563 av_log(c, AV_LOG_ERROR, "AAC with no global headers is currently not supported.\n");
569 av_strlcatf(buff, size, "a=rtpmap:%d MPEG4-GENERIC/%d/%d\r\n"
570 "a=fmtp:%d profile-level-id=1;"
571 "mode=AAC-hbr;sizelength=13;indexlength=3;"
572 "indexdeltalength=3%s\r\n",
573 payload_type, c->sample_rate, c->channels,
574 payload_type, config);
577 case AV_CODEC_ID_PCM_S16BE:
578 if (payload_type >= RTP_PT_PRIVATE)
579 av_strlcatf(buff, size, "a=rtpmap:%d L16/%d/%d\r\n",
581 c->sample_rate, c->channels);
583 case AV_CODEC_ID_PCM_MULAW:
584 if (payload_type >= RTP_PT_PRIVATE)
585 av_strlcatf(buff, size, "a=rtpmap:%d PCMU/%d/%d\r\n",
587 c->sample_rate, c->channels);
589 case AV_CODEC_ID_PCM_ALAW:
590 if (payload_type >= RTP_PT_PRIVATE)
591 av_strlcatf(buff, size, "a=rtpmap:%d PCMA/%d/%d\r\n",
593 c->sample_rate, c->channels);
595 case AV_CODEC_ID_AMR_NB:
596 av_strlcatf(buff, size, "a=rtpmap:%d AMR/%d/%d\r\n"
597 "a=fmtp:%d octet-align=1\r\n",
598 payload_type, c->sample_rate, c->channels,
601 case AV_CODEC_ID_AMR_WB:
602 av_strlcatf(buff, size, "a=rtpmap:%d AMR-WB/%d/%d\r\n"
603 "a=fmtp:%d octet-align=1\r\n",
604 payload_type, c->sample_rate, c->channels,
607 case AV_CODEC_ID_VORBIS:
608 if (c->extradata_size)
609 config = xiph_extradata2config(c);
611 av_log(c, AV_LOG_ERROR, "Vorbis configuration info missing\n");
615 av_strlcatf(buff, size, "a=rtpmap:%d vorbis/%d/%d\r\n"
616 "a=fmtp:%d configuration=%s\r\n",
617 payload_type, c->sample_rate, c->channels,
618 payload_type, config);
620 case AV_CODEC_ID_THEORA: {
622 switch (c->pix_fmt) {
623 case AV_PIX_FMT_YUV420P:
624 pix_fmt = "YCbCr-4:2:0";
626 case AV_PIX_FMT_YUV422P:
627 pix_fmt = "YCbCr-4:2:2";
629 case AV_PIX_FMT_YUV444P:
630 pix_fmt = "YCbCr-4:4:4";
633 av_log(c, AV_LOG_ERROR, "Unsupported pixel format.\n");
637 if (c->extradata_size)
638 config = xiph_extradata2config(c);
640 av_log(c, AV_LOG_ERROR, "Theora configuation info missing\n");
644 av_strlcatf(buff, size, "a=rtpmap:%d theora/90000\r\n"
645 "a=fmtp:%d delivery-method=inline; "
646 "width=%d; height=%d; sampling=%s; "
647 "configuration=%s\r\n",
648 payload_type, payload_type,
649 c->width, c->height, pix_fmt, config);
652 case AV_CODEC_ID_VP8:
653 av_strlcatf(buff, size, "a=rtpmap:%d VP8/90000\r\n",
656 case AV_CODEC_ID_MJPEG:
657 if (payload_type >= RTP_PT_PRIVATE)
658 av_strlcatf(buff, size, "a=rtpmap:%d JPEG/90000\r\n",
661 case AV_CODEC_ID_ADPCM_G722:
662 if (payload_type >= RTP_PT_PRIVATE)
663 av_strlcatf(buff, size, "a=rtpmap:%d G722/%d/%d\r\n",
667 case AV_CODEC_ID_ADPCM_G726: {
668 if (payload_type >= RTP_PT_PRIVATE)
669 av_strlcatf(buff, size, "a=rtpmap:%d G726-%d/%d\r\n",
671 c->bits_per_coded_sample*8,
675 case AV_CODEC_ID_ILBC:
676 av_strlcatf(buff, size, "a=rtpmap:%d iLBC/%d\r\n"
677 "a=fmtp:%d mode=%d\r\n",
678 payload_type, c->sample_rate,
679 payload_type, c->block_align == 38 ? 20 : 30);
681 case AV_CODEC_ID_SPEEX:
682 av_strlcatf(buff, size, "a=rtpmap:%d speex/%d\r\n",
683 payload_type, c->sample_rate);
685 case AV_CODEC_ID_OPUS:
686 /* The opus RTP draft says that all opus streams MUST be declared
687 as stereo, to avoid negotiation failures. The actual number of
688 channels can change on a packet-by-packet basis. The number of
689 channels a receiver prefers to receive or a sender plans to send
690 can be declared via fmtp parameters (both default to mono), but
691 receivers MUST be able to receive and process stereo packets. */
692 av_strlcatf(buff, size, "a=rtpmap:%d opus/48000/2\r\n",
694 if (c->channels == 2) {
695 av_strlcatf(buff, size, "a=fmtp:%d sprop-stereo:1\r\n",
700 /* Nothing special to do here... */
709 void ff_sdp_write_media(char *buff, int size, AVStream *st, int idx,
710 const char *dest_addr, const char *dest_type,
711 int port, int ttl, AVFormatContext *fmt)
713 AVCodecContext *c = st->codec;
717 payload_type = ff_rtp_get_payload_type(fmt, c, idx);
719 switch (c->codec_type) {
720 case AVMEDIA_TYPE_VIDEO : type = "video" ; break;
721 case AVMEDIA_TYPE_AUDIO : type = "audio" ; break;
722 case AVMEDIA_TYPE_SUBTITLE: type = "text" ; break;
723 default : type = "application"; break;
726 av_strlcatf(buff, size, "m=%s %d RTP/AVP %d\r\n", type, port, payload_type);
727 sdp_write_address(buff, size, dest_addr, dest_type, ttl);
729 av_strlcatf(buff, size, "b=AS:%d\r\n", c->bit_rate / 1000);
732 sdp_write_media_attributes(buff, size, c, payload_type, fmt);
735 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
737 AVDictionaryEntry *title = av_dict_get(ac[0]->metadata, "title", NULL, 0);
738 struct sdp_session_level s = { 0 };
739 int i, j, port, ttl, is_multicast, index = 0;
740 char dst[32], dst_type[5];
742 memset(buf, 0, size);
744 s.src_addr = "127.0.0.1"; /* FIXME: Properly set this */
746 s.name = title ? title->value : "No Name";
751 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[0]->filename);
752 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
758 s.dst_type = dst_type;
760 if (!strcmp(dst_type, "IP6")) {
766 sdp_write_header(buf, size, &s);
769 for (i = 0; i < n_files; i++) {
771 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[i]->filename);
772 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
777 for (j = 0; j < ac[i]->nb_streams; j++) {
778 ff_sdp_write_media(buf, size, ac[i]->streams[j], index++,
779 dst[0] ? dst : NULL, dst_type,
780 (port > 0) ? port + j * 2 : 0,
783 av_strlcatf(buf, size,
784 "a=control:streamid=%d\r\n", i + j);
786 if (ac[i]->pb && ac[i]->pb->av_class) {
787 uint8_t *crypto_suite = NULL, *crypto_params = NULL;
788 av_opt_get(ac[i]->pb, "srtp_out_suite", AV_OPT_SEARCH_CHILDREN,
790 av_opt_get(ac[i]->pb, "srtp_out_params", AV_OPT_SEARCH_CHILDREN,
792 if (crypto_suite && crypto_suite[0])
793 av_strlcatf(buf, size,
794 "a=crypto:1 %s inline:%s\r\n",
795 crypto_suite, crypto_params);
796 av_free(crypto_suite);
797 av_free(crypto_params);
805 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
807 return AVERROR(ENOSYS);
810 void ff_sdp_write_media(char *buff, int size, AVStream *st, int idx,
811 const char *dest_addr, const char *dest_type,
812 int port, int ttl, AVFormatContext *fmt)