rtpdec: Move AAC depacketization code in rtpdec to a proper payload handler
[libav.git] / libavformat / rtpdec.c
1 /*
2 * RTP input format
3 * Copyright (c) 2002 Fabrice Bellard
4 *
5 * This file is part of FFmpeg.
6 *
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.
11 *
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.
16 *
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
20 */
21
22 /* needed for gethostname() */
23 #define _XOPEN_SOURCE 600
24
25 #include "libavcodec/get_bits.h"
26 #include "avformat.h"
27 #include "mpegts.h"
28
29 #include <unistd.h>
30 #include "network.h"
31
32 #include "rtpdec.h"
33 #include "rtpdec_amr.h"
34 #include "rtpdec_asf.h"
35 #include "rtpdec_h263.h"
36 #include "rtpdec_h264.h"
37 #include "rtpdec_mpeg4.h"
38 #include "rtpdec_xiph.h"
39
40 //#define DEBUG
41
42 /* TODO: - add RTCP statistics reporting (should be optional).
43
44 - add support for h263/mpeg4 packetized output : IDEA: send a
45 buffer to 'rtp_write_packet' contains all the packets for ONE
46 frame. Each packet should have a four byte header containing
47 the length in big endian format (same trick as
48 'url_open_dyn_packet_buf')
49 */
50
51 /* statistics functions */
52 RTPDynamicProtocolHandler *RTPFirstDynamicPayloadHandler= NULL;
53
54 void ff_register_dynamic_payload_handler(RTPDynamicProtocolHandler *handler)
55 {
56 handler->next= RTPFirstDynamicPayloadHandler;
57 RTPFirstDynamicPayloadHandler= handler;
58 }
59
60 void av_register_rtp_dynamic_payload_handlers(void)
61 {
62 ff_register_dynamic_payload_handler(&ff_mp4v_es_dynamic_handler);
63 ff_register_dynamic_payload_handler(&ff_mpeg4_generic_dynamic_handler);
64 ff_register_dynamic_payload_handler(&ff_amr_nb_dynamic_handler);
65 ff_register_dynamic_payload_handler(&ff_amr_wb_dynamic_handler);
66 ff_register_dynamic_payload_handler(&ff_h263_1998_dynamic_handler);
67 ff_register_dynamic_payload_handler(&ff_h263_2000_dynamic_handler);
68 ff_register_dynamic_payload_handler(&ff_h264_dynamic_handler);
69 ff_register_dynamic_payload_handler(&ff_vorbis_dynamic_handler);
70 ff_register_dynamic_payload_handler(&ff_theora_dynamic_handler);
71
72 ff_register_dynamic_payload_handler(&ff_ms_rtp_asf_pfv_handler);
73 ff_register_dynamic_payload_handler(&ff_ms_rtp_asf_pfa_handler);
74 }
75
76 static int rtcp_parse_packet(RTPDemuxContext *s, const unsigned char *buf, int len)
77 {
78 if (buf[1] != 200)
79 return -1;
80 s->last_rtcp_ntp_time = AV_RB64(buf + 8);
81 if (s->first_rtcp_ntp_time == AV_NOPTS_VALUE)
82 s->first_rtcp_ntp_time = s->last_rtcp_ntp_time;
83 s->last_rtcp_timestamp = AV_RB32(buf + 16);
84 return 0;
85 }
86
87 #define RTP_SEQ_MOD (1<<16)
88
89 /**
90 * called on parse open packet
91 */
92 static void rtp_init_statistics(RTPStatistics *s, uint16_t base_sequence) // called on parse open packet.
93 {
94 memset(s, 0, sizeof(RTPStatistics));
95 s->max_seq= base_sequence;
96 s->probation= 1;
97 }
98
99 /**
100 * called whenever there is a large jump in sequence numbers, or when they get out of probation...
101 */
102 static void rtp_init_sequence(RTPStatistics *s, uint16_t seq)
103 {
104 s->max_seq= seq;
105 s->cycles= 0;
106 s->base_seq= seq -1;
107 s->bad_seq= RTP_SEQ_MOD + 1;
108 s->received= 0;
109 s->expected_prior= 0;
110 s->received_prior= 0;
111 s->jitter= 0;
112 s->transit= 0;
113 }
114
115 /**
116 * returns 1 if we should handle this packet.
117 */
118 static int rtp_valid_packet_in_sequence(RTPStatistics *s, uint16_t seq)
119 {
120 uint16_t udelta= seq - s->max_seq;
121 const int MAX_DROPOUT= 3000;
122 const int MAX_MISORDER = 100;
123 const int MIN_SEQUENTIAL = 2;
124
125 /* source not valid until MIN_SEQUENTIAL packets with sequence seq. numbers have been received */
126 if(s->probation)
127 {
128 if(seq==s->max_seq + 1) {
129 s->probation--;
130 s->max_seq= seq;
131 if(s->probation==0) {
132 rtp_init_sequence(s, seq);
133 s->received++;
134 return 1;
135 }
136 } else {
137 s->probation= MIN_SEQUENTIAL - 1;
138 s->max_seq = seq;
139 }
140 } else if (udelta < MAX_DROPOUT) {
141 // in order, with permissible gap
142 if(seq < s->max_seq) {
143 //sequence number wrapped; count antother 64k cycles
144 s->cycles += RTP_SEQ_MOD;
145 }
146 s->max_seq= seq;
147 } else if (udelta <= RTP_SEQ_MOD - MAX_MISORDER) {
148 // sequence made a large jump...
149 if(seq==s->bad_seq) {
150 // two sequential packets-- assume that the other side restarted without telling us; just resync.
151 rtp_init_sequence(s, seq);
152 } else {
153 s->bad_seq= (seq + 1) & (RTP_SEQ_MOD-1);
154 return 0;
155 }
156 } else {
157 // duplicate or reordered packet...
158 }
159 s->received++;
160 return 1;
161 }
162
163 #if 0
164 /**
165 * This function is currently unused; without a valid local ntp time, I don't see how we could calculate the
166 * difference between the arrival and sent timestamp. As a result, the jitter and transit statistics values
167 * never change. I left this in in case someone else can see a way. (rdm)
168 */
169 static void rtcp_update_jitter(RTPStatistics *s, uint32_t sent_timestamp, uint32_t arrival_timestamp)
170 {
171 uint32_t transit= arrival_timestamp - sent_timestamp;
172 int d;
173 s->transit= transit;
174 d= FFABS(transit - s->transit);
175 s->jitter += d - ((s->jitter + 8)>>4);
176 }
177 #endif
178
179 int rtp_check_and_send_back_rr(RTPDemuxContext *s, int count)
180 {
181 ByteIOContext *pb;
182 uint8_t *buf;
183 int len;
184 int rtcp_bytes;
185 RTPStatistics *stats= &s->statistics;
186 uint32_t lost;
187 uint32_t extended_max;
188 uint32_t expected_interval;
189 uint32_t received_interval;
190 uint32_t lost_interval;
191 uint32_t expected;
192 uint32_t fraction;
193 uint64_t ntp_time= s->last_rtcp_ntp_time; // TODO: Get local ntp time?
194
195 if (!s->rtp_ctx || (count < 1))
196 return -1;
197
198 /* TODO: I think this is way too often; RFC 1889 has algorithm for this */
199 /* XXX: mpeg pts hardcoded. RTCP send every 0.5 seconds */
200 s->octet_count += count;
201 rtcp_bytes = ((s->octet_count - s->last_octet_count) * RTCP_TX_RATIO_NUM) /
202 RTCP_TX_RATIO_DEN;
203 rtcp_bytes /= 50; // mmu_man: that's enough for me... VLC sends much less btw !?
204 if (rtcp_bytes < 28)
205 return -1;
206 s->last_octet_count = s->octet_count;
207
208 if (url_open_dyn_buf(&pb) < 0)
209 return -1;
210
211 // Receiver Report
212 put_byte(pb, (RTP_VERSION << 6) + 1); /* 1 report block */
213 put_byte(pb, 201);
214 put_be16(pb, 7); /* length in words - 1 */
215 put_be32(pb, s->ssrc); // our own SSRC
216 put_be32(pb, s->ssrc); // XXX: should be the server's here!
217 // some placeholders we should really fill...
218 // RFC 1889/p64
219 extended_max= stats->cycles + stats->max_seq;
220 expected= extended_max - stats->base_seq + 1;
221 lost= expected - stats->received;
222 lost= FFMIN(lost, 0xffffff); // clamp it since it's only 24 bits...
223 expected_interval= expected - stats->expected_prior;
224 stats->expected_prior= expected;
225 received_interval= stats->received - stats->received_prior;
226 stats->received_prior= stats->received;
227 lost_interval= expected_interval - received_interval;
228 if (expected_interval==0 || lost_interval<=0) fraction= 0;
229 else fraction = (lost_interval<<8)/expected_interval;
230
231 fraction= (fraction<<24) | lost;
232
233 put_be32(pb, fraction); /* 8 bits of fraction, 24 bits of total packets lost */
234 put_be32(pb, extended_max); /* max sequence received */
235 put_be32(pb, stats->jitter>>4); /* jitter */
236
237 if(s->last_rtcp_ntp_time==AV_NOPTS_VALUE)
238 {
239 put_be32(pb, 0); /* last SR timestamp */
240 put_be32(pb, 0); /* delay since last SR */
241 } else {
242 uint32_t middle_32_bits= s->last_rtcp_ntp_time>>16; // this is valid, right? do we need to handle 64 bit values special?
243 uint32_t delay_since_last= ntp_time - s->last_rtcp_ntp_time;
244
245 put_be32(pb, middle_32_bits); /* last SR timestamp */
246 put_be32(pb, delay_since_last); /* delay since last SR */
247 }
248
249 // CNAME
250 put_byte(pb, (RTP_VERSION << 6) + 1); /* 1 report block */
251 put_byte(pb, 202);
252 len = strlen(s->hostname);
253 put_be16(pb, (6 + len + 3) / 4); /* length in words - 1 */
254 put_be32(pb, s->ssrc);
255 put_byte(pb, 0x01);
256 put_byte(pb, len);
257 put_buffer(pb, s->hostname, len);
258 // padding
259 for (len = (6 + len) % 4; len % 4; len++) {
260 put_byte(pb, 0);
261 }
262
263 put_flush_packet(pb);
264 len = url_close_dyn_buf(pb, &buf);
265 if ((len > 0) && buf) {
266 int result;
267 dprintf(s->ic, "sending %d bytes of RR\n", len);
268 result= url_write(s->rtp_ctx, buf, len);
269 dprintf(s->ic, "result from url_write: %d\n", result);
270 av_free(buf);
271 }
272 return 0;
273 }
274
275 void rtp_send_punch_packets(URLContext* rtp_handle)
276 {
277 ByteIOContext *pb;
278 uint8_t *buf;
279 int len;
280
281 /* Send a small RTP packet */
282 if (url_open_dyn_buf(&pb) < 0)
283 return;
284
285 put_byte(pb, (RTP_VERSION << 6));
286 put_byte(pb, 0); /* Payload type */
287 put_be16(pb, 0); /* Seq */
288 put_be32(pb, 0); /* Timestamp */
289 put_be32(pb, 0); /* SSRC */
290
291 put_flush_packet(pb);
292 len = url_close_dyn_buf(pb, &buf);
293 if ((len > 0) && buf)
294 url_write(rtp_handle, buf, len);
295 av_free(buf);
296
297 /* Send a minimal RTCP RR */
298 if (url_open_dyn_buf(&pb) < 0)
299 return;
300
301 put_byte(pb, (RTP_VERSION << 6));
302 put_byte(pb, 201); /* receiver report */
303 put_be16(pb, 1); /* length in words - 1 */
304 put_be32(pb, 0); /* our own SSRC */
305
306 put_flush_packet(pb);
307 len = url_close_dyn_buf(pb, &buf);
308 if ((len > 0) && buf)
309 url_write(rtp_handle, buf, len);
310 av_free(buf);
311 }
312
313
314 /**
315 * open a new RTP parse context for stream 'st'. 'st' can be NULL for
316 * MPEG2TS streams to indicate that they should be demuxed inside the
317 * rtp demux (otherwise CODEC_ID_MPEG2TS packets are returned)
318 * TODO: change this to not take rtp_payload data, and use the new dynamic payload system.
319 */
320 RTPDemuxContext *rtp_parse_open(AVFormatContext *s1, AVStream *st, URLContext *rtpc, int payload_type, RTPPayloadData *rtp_payload_data)
321 {
322 RTPDemuxContext *s;
323
324 s = av_mallocz(sizeof(RTPDemuxContext));
325 if (!s)
326 return NULL;
327 s->payload_type = payload_type;
328 s->last_rtcp_ntp_time = AV_NOPTS_VALUE;
329 s->first_rtcp_ntp_time = AV_NOPTS_VALUE;
330 s->ic = s1;
331 s->st = st;
332 s->rtp_payload_data = rtp_payload_data;
333 rtp_init_statistics(&s->statistics, 0); // do we know the initial sequence from sdp?
334 if (!strcmp(ff_rtp_enc_name(payload_type), "MP2T")) {
335 s->ts = ff_mpegts_parse_open(s->ic);
336 if (s->ts == NULL) {
337 av_free(s);
338 return NULL;
339 }
340 } else {
341 av_set_pts_info(st, 32, 1, 90000);
342 switch(st->codec->codec_id) {
343 case CODEC_ID_MPEG1VIDEO:
344 case CODEC_ID_MPEG2VIDEO:
345 case CODEC_ID_MP2:
346 case CODEC_ID_MP3:
347 case CODEC_ID_MPEG4:
348 case CODEC_ID_H263:
349 case CODEC_ID_H264:
350 st->need_parsing = AVSTREAM_PARSE_FULL;
351 break;
352 default:
353 if (st->codec->codec_type == AVMEDIA_TYPE_AUDIO) {
354 av_set_pts_info(st, 32, 1, st->codec->sample_rate);
355 }
356 break;
357 }
358 }
359 // needed to send back RTCP RR in RTSP sessions
360 s->rtp_ctx = rtpc;
361 gethostname(s->hostname, sizeof(s->hostname));
362 return s;
363 }
364
365 void
366 rtp_parse_set_dynamic_protocol(RTPDemuxContext *s, PayloadContext *ctx,
367 RTPDynamicProtocolHandler *handler)
368 {
369 s->dynamic_protocol_context = ctx;
370 s->parse_packet = handler->parse_packet;
371 }
372
373 /**
374 * This was the second switch in rtp_parse packet. Normalizes time, if required, sets stream_index, etc.
375 */
376 static void finalize_packet(RTPDemuxContext *s, AVPacket *pkt, uint32_t timestamp)
377 {
378 if (s->last_rtcp_ntp_time != AV_NOPTS_VALUE) {
379 int64_t addend;
380 int delta_timestamp;
381
382 /* compute pts from timestamp with received ntp_time */
383 delta_timestamp = timestamp - s->last_rtcp_timestamp;
384 /* convert to the PTS timebase */
385 addend = av_rescale(s->last_rtcp_ntp_time - s->first_rtcp_ntp_time, s->st->time_base.den, (uint64_t)s->st->time_base.num << 32);
386 pkt->pts = s->range_start_offset + addend + delta_timestamp;
387 }
388 }
389
390 /**
391 * Parse an RTP or RTCP packet directly sent as a buffer.
392 * @param s RTP parse context.
393 * @param pkt returned packet
394 * @param buf input buffer or NULL to read the next packets
395 * @param len buffer len
396 * @return 0 if a packet is returned, 1 if a packet is returned and more can follow
397 * (use buf as NULL to read the next). -1 if no packet (error or no more packet).
398 */
399 int rtp_parse_packet(RTPDemuxContext *s, AVPacket *pkt,
400 const uint8_t *buf, int len)
401 {
402 unsigned int ssrc, h;
403 int payload_type, seq, ret, flags = 0;
404 AVStream *st;
405 uint32_t timestamp;
406 int rv= 0;
407
408 if (!buf) {
409 /* return the next packets, if any */
410 if(s->st && s->parse_packet) {
411 timestamp= 0; ///< Should not be used if buf is NULL, but should be set to the timestamp of the packet returned....
412 rv= s->parse_packet(s->ic, s->dynamic_protocol_context,
413 s->st, pkt, &timestamp, NULL, 0, flags);
414 finalize_packet(s, pkt, timestamp);
415 return rv;
416 } else {
417 // TODO: Move to a dynamic packet handler (like above)
418 if (s->read_buf_index >= s->read_buf_size)
419 return -1;
420 ret = ff_mpegts_parse_packet(s->ts, pkt, s->buf + s->read_buf_index,
421 s->read_buf_size - s->read_buf_index);
422 if (ret < 0)
423 return -1;
424 s->read_buf_index += ret;
425 if (s->read_buf_index < s->read_buf_size)
426 return 1;
427 else
428 return 0;
429 }
430 }
431
432 if (len < 12)
433 return -1;
434
435 if ((buf[0] & 0xc0) != (RTP_VERSION << 6))
436 return -1;
437 if (buf[1] >= 200 && buf[1] <= 204) {
438 rtcp_parse_packet(s, buf, len);
439 return -1;
440 }
441 payload_type = buf[1] & 0x7f;
442 if (buf[1] & 0x80)
443 flags |= RTP_FLAG_MARKER;
444 seq = AV_RB16(buf + 2);
445 timestamp = AV_RB32(buf + 4);
446 ssrc = AV_RB32(buf + 8);
447 /* store the ssrc in the RTPDemuxContext */
448 s->ssrc = ssrc;
449
450 /* NOTE: we can handle only one payload type */
451 if (s->payload_type != payload_type)
452 return -1;
453
454 st = s->st;
455 // only do something with this if all the rtp checks pass...
456 if(!rtp_valid_packet_in_sequence(&s->statistics, seq))
457 {
458 av_log(st?st->codec:NULL, AV_LOG_ERROR, "RTP: PT=%02x: bad cseq %04x expected=%04x\n",
459 payload_type, seq, ((s->seq + 1) & 0xffff));
460 return -1;
461 }
462
463 s->seq = seq;
464 len -= 12;
465 buf += 12;
466
467 if (!st) {
468 /* specific MPEG2TS demux support */
469 ret = ff_mpegts_parse_packet(s->ts, pkt, buf, len);
470 if (ret < 0)
471 return -1;
472 if (ret < len) {
473 s->read_buf_size = len - ret;
474 memcpy(s->buf, buf + ret, s->read_buf_size);
475 s->read_buf_index = 0;
476 return 1;
477 }
478 return 0;
479 } else if (s->parse_packet) {
480 rv = s->parse_packet(s->ic, s->dynamic_protocol_context,
481 s->st, pkt, &timestamp, buf, len, flags);
482 } else {
483 // at this point, the RTP header has been stripped; This is ASSUMING that there is only 1 CSRC, which in't wise.
484 switch(st->codec->codec_id) {
485 case CODEC_ID_MP2:
486 case CODEC_ID_MP3:
487 /* better than nothing: skip mpeg audio RTP header */
488 if (len <= 4)
489 return -1;
490 h = AV_RB32(buf);
491 len -= 4;
492 buf += 4;
493 av_new_packet(pkt, len);
494 memcpy(pkt->data, buf, len);
495 break;
496 case CODEC_ID_MPEG1VIDEO:
497 case CODEC_ID_MPEG2VIDEO:
498 /* better than nothing: skip mpeg video RTP header */
499 if (len <= 4)
500 return -1;
501 h = AV_RB32(buf);
502 buf += 4;
503 len -= 4;
504 if (h & (1 << 26)) {
505 /* mpeg2 */
506 if (len <= 4)
507 return -1;
508 buf += 4;
509 len -= 4;
510 }
511 av_new_packet(pkt, len);
512 memcpy(pkt->data, buf, len);
513 break;
514 default:
515 av_new_packet(pkt, len);
516 memcpy(pkt->data, buf, len);
517 break;
518 }
519
520 pkt->stream_index = st->index;
521 }
522
523 // now perform timestamp things....
524 finalize_packet(s, pkt, timestamp);
525
526 return rv;
527 }
528
529 void rtp_parse_close(RTPDemuxContext *s)
530 {
531 // TODO: fold this into the protocol specific data fields.
532 av_free(s->rtp_payload_data->mode);
533 av_free(s->rtp_payload_data->au_headers);
534 if (!strcmp(ff_rtp_enc_name(s->payload_type), "MP2T")) {
535 ff_mpegts_parse_close(s->ts);
536 }
537 av_free(s);
538 }