sdp: Restore the original mp4 format h264 extradata if converted
[libav.git] / libavformat / sdp.c
1 /*
2 * copyright (c) 2007 Luca Abeni
3 *
4 * This file is part of Libav.
5 *
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.
10 *
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.
15 *
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
19 */
20
21 #include <string.h>
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"
29 #include "avformat.h"
30 #include "internal.h"
31 #include "avc.h"
32 #include "rtp.h"
33 #if CONFIG_NETWORK
34 #include "network.h"
35 #endif
36
37 #if CONFIG_RTP_MUXER
38 #define MAX_EXTRADATA_SIZE ((INT_MAX - 10) / 2)
39
40 struct sdp_session_level {
41 int sdp_version; /**< protocol version (currently 0) */
42 int id; /**< session ID */
43 int version; /**< session version */
44 int start_time; /**< session start time (NTP time, in seconds),
45 or 0 in case of permanent session */
46 int end_time; /**< session end time (NTP time, in seconds),
47 or 0 if the session is not bounded */
48 int ttl; /**< TTL, in case of multicast stream */
49 const char *user; /**< username of the session's creator */
50 const char *src_addr; /**< IP address of the machine from which the session was created */
51 const char *src_type; /**< address type of src_addr */
52 const char *dst_addr; /**< destination IP address (can be multicast) */
53 const char *dst_type; /**< destination IP address type */
54 const char *name; /**< session name (can be an empty string) */
55 };
56
57 static void sdp_write_address(char *buff, int size, const char *dest_addr,
58 const char *dest_type, int ttl)
59 {
60 if (dest_addr) {
61 if (!dest_type)
62 dest_type = "IP4";
63 if (ttl > 0 && !strcmp(dest_type, "IP4")) {
64 /* The TTL should only be specified for IPv4 multicast addresses,
65 * not for IPv6. */
66 av_strlcatf(buff, size, "c=IN %s %s/%d\r\n", dest_type, dest_addr, ttl);
67 } else {
68 av_strlcatf(buff, size, "c=IN %s %s\r\n", dest_type, dest_addr);
69 }
70 }
71 }
72
73 static void sdp_write_header(char *buff, int size, struct sdp_session_level *s)
74 {
75 av_strlcatf(buff, size, "v=%d\r\n"
76 "o=- %d %d IN %s %s\r\n"
77 "s=%s\r\n",
78 s->sdp_version,
79 s->id, s->version, s->src_type, s->src_addr,
80 s->name);
81 sdp_write_address(buff, size, s->dst_addr, s->dst_type, s->ttl);
82 av_strlcatf(buff, size, "t=%d %d\r\n"
83 "a=tool:libavformat " AV_STRINGIFY(LIBAVFORMAT_VERSION) "\r\n",
84 s->start_time, s->end_time);
85 }
86
87 #if CONFIG_NETWORK
88 static int resolve_destination(char *dest_addr, int size, char *type,
89 int type_size)
90 {
91 struct addrinfo hints, *ai;
92 int is_multicast;
93
94 av_strlcpy(type, "IP4", type_size);
95 if (!dest_addr[0])
96 return 0;
97
98 /* Resolve the destination, since it must be written
99 * as a numeric IP address in the SDP. */
100
101 memset(&hints, 0, sizeof(hints));
102 if (getaddrinfo(dest_addr, NULL, &hints, &ai))
103 return 0;
104 getnameinfo(ai->ai_addr, ai->ai_addrlen, dest_addr, size,
105 NULL, 0, NI_NUMERICHOST);
106 #ifdef AF_INET6
107 if (ai->ai_family == AF_INET6)
108 av_strlcpy(type, "IP6", type_size);
109 #endif
110 is_multicast = ff_is_multicast_address(ai->ai_addr);
111 freeaddrinfo(ai);
112 return is_multicast;
113 }
114 #else
115 static int resolve_destination(char *dest_addr, int size, char *type,
116 int type_size)
117 {
118 return 0;
119 }
120 #endif
121
122 static int sdp_get_address(char *dest_addr, int size, int *ttl, const char *url)
123 {
124 int port;
125 const char *p;
126 char proto[32];
127
128 av_url_split(proto, sizeof(proto), NULL, 0, dest_addr, size, &port, NULL, 0, url);
129
130 *ttl = 0;
131
132 if (strcmp(proto, "rtp")) {
133 /* The url isn't for the actual rtp sessions,
134 * don't parse out anything else than the destination.
135 */
136 return 0;
137 }
138
139 p = strchr(url, '?');
140 if (p) {
141 char buff[64];
142
143 if (av_find_info_tag(buff, sizeof(buff), "ttl", p)) {
144 *ttl = strtol(buff, NULL, 10);
145 } else {
146 *ttl = 5;
147 }
148 }
149
150 return port;
151 }
152
153 #define MAX_PSET_SIZE 1024
154 static char *extradata2psets(AVCodecContext *c)
155 {
156 char *psets, *p;
157 const uint8_t *r;
158 const char *pset_string = "; sprop-parameter-sets=";
159 uint8_t *orig_extradata = NULL;
160 int orig_extradata_size = 0;
161
162 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
163 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
164
165 return NULL;
166 }
167 if (c->extradata[0] == 1) {
168 uint8_t *dummy_p;
169 int dummy_int;
170 AVBitStreamFilterContext *bsfc= av_bitstream_filter_init("h264_mp4toannexb");
171
172 if (!bsfc) {
173 av_log(c, AV_LOG_ERROR, "Cannot open the h264_mp4toannexb BSF!\n");
174
175 return NULL;
176 }
177
178 orig_extradata_size = c->extradata_size;
179 orig_extradata = av_mallocz(orig_extradata_size +
180 FF_INPUT_BUFFER_PADDING_SIZE);
181 if (!orig_extradata) {
182 av_bitstream_filter_close(bsfc);
183 return NULL;
184 }
185 memcpy(orig_extradata, c->extradata, orig_extradata_size);
186 av_bitstream_filter_filter(bsfc, c, NULL, &dummy_p, &dummy_int, NULL, 0, 0);
187 av_bitstream_filter_close(bsfc);
188 }
189
190 psets = av_mallocz(MAX_PSET_SIZE);
191 if (psets == NULL) {
192 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the parameter sets.\n");
193 av_free(orig_extradata);
194 return NULL;
195 }
196 memcpy(psets, pset_string, strlen(pset_string));
197 p = psets + strlen(pset_string);
198 r = ff_avc_find_startcode(c->extradata, c->extradata + c->extradata_size);
199 while (r < c->extradata + c->extradata_size) {
200 const uint8_t *r1;
201 uint8_t nal_type;
202
203 while (!*(r++));
204 nal_type = *r & 0x1f;
205 r1 = ff_avc_find_startcode(r, c->extradata + c->extradata_size);
206 if (nal_type != 7 && nal_type != 8) { /* Only output SPS and PPS */
207 r = r1;
208 continue;
209 }
210 if (p != (psets + strlen(pset_string))) {
211 *p = ',';
212 p++;
213 }
214 if (av_base64_encode(p, MAX_PSET_SIZE - (p - psets), r, r1 - r) == NULL) {
215 av_log(c, AV_LOG_ERROR, "Cannot Base64-encode %td %td!\n", MAX_PSET_SIZE - (p - psets), r1 - r);
216 av_free(psets);
217
218 return NULL;
219 }
220 p += strlen(p);
221 r = r1;
222 }
223 if (orig_extradata) {
224 av_free(c->extradata);
225 c->extradata = orig_extradata;
226 c->extradata_size = orig_extradata_size;
227 }
228
229 return psets;
230 }
231
232 static char *extradata2config(AVCodecContext *c)
233 {
234 char *config;
235
236 if (c->extradata_size > MAX_EXTRADATA_SIZE) {
237 av_log(c, AV_LOG_ERROR, "Too much extradata!\n");
238
239 return NULL;
240 }
241 config = av_malloc(10 + c->extradata_size * 2);
242 if (config == NULL) {
243 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
244 return NULL;
245 }
246 memcpy(config, "; config=", 9);
247 ff_data_to_hex(config + 9, c->extradata, c->extradata_size, 0);
248 config[9 + c->extradata_size * 2] = 0;
249
250 return config;
251 }
252
253 static char *xiph_extradata2config(AVCodecContext *c)
254 {
255 char *config, *encoded_config;
256 uint8_t *header_start[3];
257 int headers_len, header_len[3], config_len;
258 int first_header_size;
259
260 switch (c->codec_id) {
261 case CODEC_ID_THEORA:
262 first_header_size = 42;
263 break;
264 case CODEC_ID_VORBIS:
265 first_header_size = 30;
266 break;
267 default:
268 av_log(c, AV_LOG_ERROR, "Unsupported Xiph codec ID\n");
269 return NULL;
270 }
271
272 if (avpriv_split_xiph_headers(c->extradata, c->extradata_size,
273 first_header_size, header_start,
274 header_len) < 0) {
275 av_log(c, AV_LOG_ERROR, "Extradata corrupt.\n");
276 return NULL;
277 }
278
279 headers_len = header_len[0] + header_len[2];
280 config_len = 4 + // count
281 3 + // ident
282 2 + // packet size
283 1 + // header count
284 2 + // header size
285 headers_len; // and the rest
286
287 config = av_malloc(config_len);
288 if (!config)
289 goto xiph_fail;
290
291 encoded_config = av_malloc(AV_BASE64_SIZE(config_len));
292 if (!encoded_config) {
293 av_free(config);
294 goto xiph_fail;
295 }
296
297 config[0] = config[1] = config[2] = 0;
298 config[3] = 1;
299 config[4] = (RTP_XIPH_IDENT >> 16) & 0xff;
300 config[5] = (RTP_XIPH_IDENT >> 8) & 0xff;
301 config[6] = (RTP_XIPH_IDENT ) & 0xff;
302 config[7] = (headers_len >> 8) & 0xff;
303 config[8] = headers_len & 0xff;
304 config[9] = 2;
305 config[10] = header_len[0];
306 config[11] = 0; // size of comment header; nonexistent
307 memcpy(config + 12, header_start[0], header_len[0]);
308 memcpy(config + 12 + header_len[0], header_start[2], header_len[2]);
309
310 av_base64_encode(encoded_config, AV_BASE64_SIZE(config_len),
311 config, config_len);
312 av_free(config);
313
314 return encoded_config;
315
316 xiph_fail:
317 av_log(c, AV_LOG_ERROR,
318 "Not enough memory for configuration string\n");
319 return NULL;
320 }
321
322 static int latm_context2profilelevel(AVCodecContext *c)
323 {
324 /* MP4A-LATM
325 * The RTP payload format specification is described in RFC 3016
326 * The encoding specifications are provided in ISO/IEC 14496-3 */
327
328 int profile_level = 0x2B;
329
330 /* TODO: AAC Profile only supports AAC LC Object Type.
331 * Different Object Types should implement different Profile Levels */
332
333 if (c->sample_rate <= 24000) {
334 if (c->channels <= 2)
335 profile_level = 0x28; // AAC Profile, Level 1
336 } else if (c->sample_rate <= 48000) {
337 if (c->channels <= 2) {
338 profile_level = 0x29; // AAC Profile, Level 2
339 } else if (c->channels <= 5) {
340 profile_level = 0x2A; // AAC Profile, Level 4
341 }
342 } else if (c->sample_rate <= 96000) {
343 if (c->channels <= 5) {
344 profile_level = 0x2B; // AAC Profile, Level 5
345 }
346 }
347
348 return profile_level;
349 }
350
351 static char *latm_context2config(AVCodecContext *c)
352 {
353 /* MP4A-LATM
354 * The RTP payload format specification is described in RFC 3016
355 * The encoding specifications are provided in ISO/IEC 14496-3 */
356
357 uint8_t config_byte[6];
358 int rate_index;
359 char *config;
360
361 for (rate_index = 0; rate_index < 16; rate_index++)
362 if (avpriv_mpeg4audio_sample_rates[rate_index] == c->sample_rate)
363 break;
364 if (rate_index == 16) {
365 av_log(c, AV_LOG_ERROR, "Unsupported sample rate\n");
366 return NULL;
367 }
368
369 config_byte[0] = 0x40;
370 config_byte[1] = 0;
371 config_byte[2] = 0x20 | rate_index;
372 config_byte[3] = c->channels << 4;
373 config_byte[4] = 0x3f;
374 config_byte[5] = 0xc0;
375
376 config = av_malloc(6*2+1);
377 if (!config) {
378 av_log(c, AV_LOG_ERROR, "Cannot allocate memory for the config info.\n");
379 return NULL;
380 }
381 ff_data_to_hex(config, config_byte, 6, 1);
382 config[12] = 0;
383
384 return config;
385 }
386
387 static char *sdp_write_media_attributes(char *buff, int size, AVCodecContext *c, int payload_type, AVFormatContext *fmt)
388 {
389 char *config = NULL;
390
391 switch (c->codec_id) {
392 case CODEC_ID_H264:
393 if (c->extradata_size) {
394 config = extradata2psets(c);
395 }
396 av_strlcatf(buff, size, "a=rtpmap:%d H264/90000\r\n"
397 "a=fmtp:%d packetization-mode=1%s\r\n",
398 payload_type,
399 payload_type, config ? config : "");
400 break;
401 case CODEC_ID_H263:
402 case CODEC_ID_H263P:
403 /* a=framesize is required by 3GPP TS 26.234 (PSS). It
404 * actually specifies the maximum video size, but we only know
405 * the current size. This is required for playback on Android
406 * stagefright and on Samsung bada. */
407 av_strlcatf(buff, size, "a=rtpmap:%d H263-2000/90000\r\n"
408 "a=framesize:%d %d-%d\r\n",
409 payload_type,
410 payload_type, c->width, c->height);
411 break;
412 case CODEC_ID_MPEG4:
413 if (c->extradata_size) {
414 config = extradata2config(c);
415 }
416 av_strlcatf(buff, size, "a=rtpmap:%d MP4V-ES/90000\r\n"
417 "a=fmtp:%d profile-level-id=1%s\r\n",
418 payload_type,
419 payload_type, config ? config : "");
420 break;
421 case CODEC_ID_AAC:
422 if (fmt && fmt->oformat->priv_class &&
423 av_opt_flag_is_set(fmt->priv_data, "rtpflags", "latm")) {
424 config = latm_context2config(c);
425 if (!config)
426 return NULL;
427 av_strlcatf(buff, size, "a=rtpmap:%d MP4A-LATM/%d/%d\r\n"
428 "a=fmtp:%d profile-level-id=%d;cpresent=0;config=%s\r\n",
429 payload_type, c->sample_rate, c->channels,
430 payload_type, latm_context2profilelevel(c), config);
431 } else {
432 if (c->extradata_size) {
433 config = extradata2config(c);
434 } else {
435 /* FIXME: maybe we can forge config information based on the
436 * codec parameters...
437 */
438 av_log(c, AV_LOG_ERROR, "AAC with no global headers is currently not supported.\n");
439 return NULL;
440 }
441 if (config == NULL) {
442 return NULL;
443 }
444 av_strlcatf(buff, size, "a=rtpmap:%d MPEG4-GENERIC/%d/%d\r\n"
445 "a=fmtp:%d profile-level-id=1;"
446 "mode=AAC-hbr;sizelength=13;indexlength=3;"
447 "indexdeltalength=3%s\r\n",
448 payload_type, c->sample_rate, c->channels,
449 payload_type, config);
450 }
451 break;
452 case CODEC_ID_PCM_S16BE:
453 if (payload_type >= RTP_PT_PRIVATE)
454 av_strlcatf(buff, size, "a=rtpmap:%d L16/%d/%d\r\n",
455 payload_type,
456 c->sample_rate, c->channels);
457 break;
458 case CODEC_ID_PCM_MULAW:
459 if (payload_type >= RTP_PT_PRIVATE)
460 av_strlcatf(buff, size, "a=rtpmap:%d PCMU/%d/%d\r\n",
461 payload_type,
462 c->sample_rate, c->channels);
463 break;
464 case CODEC_ID_PCM_ALAW:
465 if (payload_type >= RTP_PT_PRIVATE)
466 av_strlcatf(buff, size, "a=rtpmap:%d PCMA/%d/%d\r\n",
467 payload_type,
468 c->sample_rate, c->channels);
469 break;
470 case CODEC_ID_AMR_NB:
471 av_strlcatf(buff, size, "a=rtpmap:%d AMR/%d/%d\r\n"
472 "a=fmtp:%d octet-align=1\r\n",
473 payload_type, c->sample_rate, c->channels,
474 payload_type);
475 break;
476 case CODEC_ID_AMR_WB:
477 av_strlcatf(buff, size, "a=rtpmap:%d AMR-WB/%d/%d\r\n"
478 "a=fmtp:%d octet-align=1\r\n",
479 payload_type, c->sample_rate, c->channels,
480 payload_type);
481 break;
482 case CODEC_ID_VORBIS:
483 if (c->extradata_size)
484 config = xiph_extradata2config(c);
485 else
486 av_log(c, AV_LOG_ERROR, "Vorbis configuration info missing\n");
487 if (!config)
488 return NULL;
489
490 av_strlcatf(buff, size, "a=rtpmap:%d vorbis/%d/%d\r\n"
491 "a=fmtp:%d configuration=%s\r\n",
492 payload_type, c->sample_rate, c->channels,
493 payload_type, config);
494 break;
495 case CODEC_ID_THEORA: {
496 const char *pix_fmt;
497 if (c->extradata_size)
498 config = xiph_extradata2config(c);
499 else
500 av_log(c, AV_LOG_ERROR, "Theora configuation info missing\n");
501 if (!config)
502 return NULL;
503
504 switch (c->pix_fmt) {
505 case PIX_FMT_YUV420P:
506 pix_fmt = "YCbCr-4:2:0";
507 break;
508 case PIX_FMT_YUV422P:
509 pix_fmt = "YCbCr-4:2:2";
510 break;
511 case PIX_FMT_YUV444P:
512 pix_fmt = "YCbCr-4:4:4";
513 break;
514 default:
515 av_log(c, AV_LOG_ERROR, "Unsupported pixel format.\n");
516 return NULL;
517 }
518
519 av_strlcatf(buff, size, "a=rtpmap:%d theora/90000\r\n"
520 "a=fmtp:%d delivery-method=inline; "
521 "width=%d; height=%d; sampling=%s; "
522 "configuration=%s\r\n",
523 payload_type, payload_type,
524 c->width, c->height, pix_fmt, config);
525 break;
526 }
527 case CODEC_ID_VP8:
528 av_strlcatf(buff, size, "a=rtpmap:%d VP8/90000\r\n",
529 payload_type);
530 break;
531 case CODEC_ID_ADPCM_G722:
532 if (payload_type >= RTP_PT_PRIVATE)
533 av_strlcatf(buff, size, "a=rtpmap:%d G722/%d/%d\r\n",
534 payload_type,
535 8000, c->channels);
536 break;
537 case CODEC_ID_ADPCM_G726: {
538 if (payload_type >= RTP_PT_PRIVATE)
539 av_strlcatf(buff, size, "a=rtpmap:%d G726-%d/%d\r\n",
540 payload_type,
541 c->bits_per_coded_sample*8,
542 c->sample_rate);
543 break;
544 }
545 default:
546 /* Nothing special to do here... */
547 break;
548 }
549
550 av_free(config);
551
552 return buff;
553 }
554
555 void ff_sdp_write_media(char *buff, int size, AVCodecContext *c, const char *dest_addr, const char *dest_type, int port, int ttl, AVFormatContext *fmt)
556 {
557 const char *type;
558 int payload_type;
559
560 payload_type = ff_rtp_get_payload_type(fmt, c);
561
562 switch (c->codec_type) {
563 case AVMEDIA_TYPE_VIDEO : type = "video" ; break;
564 case AVMEDIA_TYPE_AUDIO : type = "audio" ; break;
565 case AVMEDIA_TYPE_SUBTITLE: type = "text" ; break;
566 default : type = "application"; break;
567 }
568
569 av_strlcatf(buff, size, "m=%s %d RTP/AVP %d\r\n", type, port, payload_type);
570 sdp_write_address(buff, size, dest_addr, dest_type, ttl);
571 if (c->bit_rate) {
572 av_strlcatf(buff, size, "b=AS:%d\r\n", c->bit_rate / 1000);
573 }
574
575 sdp_write_media_attributes(buff, size, c, payload_type, fmt);
576 }
577
578 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
579 {
580 AVDictionaryEntry *title = av_dict_get(ac[0]->metadata, "title", NULL, 0);
581 struct sdp_session_level s;
582 int i, j, port, ttl, is_multicast;
583 char dst[32], dst_type[5];
584
585 memset(buf, 0, size);
586 memset(&s, 0, sizeof(struct sdp_session_level));
587 s.user = "-";
588 s.src_addr = "127.0.0.1"; /* FIXME: Properly set this */
589 s.src_type = "IP4";
590 s.name = title ? title->value : "No Name";
591
592 port = 0;
593 ttl = 0;
594 if (n_files == 1) {
595 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[0]->filename);
596 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
597 sizeof(dst_type));
598 if (!is_multicast)
599 ttl = 0;
600 if (dst[0]) {
601 s.dst_addr = dst;
602 s.dst_type = dst_type;
603 s.ttl = ttl;
604 if (!strcmp(dst_type, "IP6")) {
605 s.src_addr = "::1";
606 s.src_type = "IP6";
607 }
608 }
609 }
610 sdp_write_header(buf, size, &s);
611
612 dst[0] = 0;
613 for (i = 0; i < n_files; i++) {
614 if (n_files != 1) {
615 port = sdp_get_address(dst, sizeof(dst), &ttl, ac[i]->filename);
616 is_multicast = resolve_destination(dst, sizeof(dst), dst_type,
617 sizeof(dst_type));
618 if (!is_multicast)
619 ttl = 0;
620 }
621 for (j = 0; j < ac[i]->nb_streams; j++) {
622 ff_sdp_write_media(buf, size,
623 ac[i]->streams[j]->codec, dst[0] ? dst : NULL,
624 dst_type, (port > 0) ? port + j * 2 : 0, ttl,
625 ac[i]);
626 if (port <= 0) {
627 av_strlcatf(buf, size,
628 "a=control:streamid=%d\r\n", i + j);
629 }
630 }
631 }
632
633 return 0;
634 }
635 #else
636 int av_sdp_create(AVFormatContext *ac[], int n_files, char *buf, int size)
637 {
638 return AVERROR(ENOSYS);
639 }
640
641 void ff_sdp_write_media(char *buff, int size, AVCodecContext *c, const char *dest_addr, const char *dest_type, int port, int ttl, AVFormatContext *fmt)
642 {
643 }
644 #endif
645
646 #if FF_API_SDP_CREATE
647 int avf_sdp_create(AVFormatContext *ac[], int n_files, char *buff, int size)
648 {
649 return av_sdp_create(ac, n_files, buff, size);
650 }
651 #endif