2 * Realmedia RTSP protocol (RDT) support.
3 * Copyright (c) 2007 Ronald S. Bultje
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
24 * @brief Realmedia RTSP protocol (RDT) support
25 * @author Ronald S. Bultje <rbultje@ronald.bitfreak.net>
29 #include "libavutil/avstring.h"
32 #include "libavutil/base64.h"
33 #include "libavutil/md5.h"
36 #include "libavcodec/get_bits.h"
38 struct RDTDemuxContext {
39 AVFormatContext *ic; /**< the containing (RTSP) demux context */
40 /** Each RDT stream-set (represented by one RTSPStream) can contain
41 * multiple streams (of the same content, but with possibly different
42 * codecs/bitrates). Each such stream is represented by one AVStream
43 * in the AVFormatContext, and this variable points to the offset in
44 * that array such that the first is the first stream of this set. */
46 int n_streams; /**< streams with identifical content in this set */
47 void *dynamic_protocol_context;
48 DynamicPayloadPacketHandlerProc parse_packet;
49 uint32_t prev_timestamp;
50 int prev_set_id, prev_stream_id;
54 ff_rdt_parse_open(AVFormatContext *ic, int first_stream_of_set_idx,
55 void *priv_data, RTPDynamicProtocolHandler *handler)
57 RDTDemuxContext *s = av_mallocz(sizeof(RDTDemuxContext));
62 s->streams = &ic->streams[first_stream_of_set_idx];
65 } while (first_stream_of_set_idx + s->n_streams < ic->nb_streams &&
66 s->streams[s->n_streams]->id == s->streams[0]->id);
68 s->prev_stream_id = -1;
69 s->prev_timestamp = -1;
70 s->parse_packet = handler ? handler->parse_packet : NULL;
71 s->dynamic_protocol_context = priv_data;
77 ff_rdt_parse_close(RDTDemuxContext *s)
82 struct PayloadContext {
83 AVFormatContext *rmctx;
87 unsigned int mlti_data_size;
88 char buffer[RTP_MAX_PACKET_LENGTH + FF_INPUT_BUFFER_PADDING_SIZE];
89 int audio_pkt_cnt; /**< remaining audio packets in rmdec */
93 ff_rdt_calc_response_and_checksum(char response[41], char chksum[9],
94 const char *challenge)
96 int ch_len = strlen (challenge), i;
97 unsigned char zres[16],
98 buf[64] = { 0xa1, 0xe9, 0x14, 0x9d, 0x0e, 0x6b, 0x3b, 0x59 };
99 #define XOR_TABLE_SIZE 37
100 const unsigned char xor_table[XOR_TABLE_SIZE] = {
101 0x05, 0x18, 0x74, 0xd0, 0x0d, 0x09, 0x02, 0x53,
102 0xc0, 0x01, 0x05, 0x05, 0x67, 0x03, 0x19, 0x70,
103 0x08, 0x27, 0x66, 0x10, 0x10, 0x72, 0x08, 0x09,
104 0x63, 0x11, 0x03, 0x71, 0x08, 0x08, 0x70, 0x02,
105 0x10, 0x57, 0x05, 0x18, 0x54 };
107 /* some (length) checks */
108 if (ch_len == 40) /* what a hack... */
110 else if (ch_len > 56)
112 memcpy(buf + 8, challenge, ch_len);
114 /* xor challenge bytewise with xor_table */
115 for (i = 0; i < XOR_TABLE_SIZE; i++)
116 buf[8 + i] ^= xor_table[i];
118 av_md5_sum(zres, buf, 64);
119 ff_data_to_hex(response, zres, 16, 1);
122 strcpy (response + 32, "01d0a8e3");
124 /* calculate checksum */
125 for (i = 0; i < 8; i++)
126 chksum[i] = response[i * 4];
131 rdt_load_mdpr (PayloadContext *rdt, AVStream *st, int rule_nr)
138 * Layout of the MLTI chunk:
140 * 2: number of streams
141 * Then for each stream ([number_of_streams] times):
143 * 2: number of mdpr chunks
144 * Then for each mdpr chunk ([number_of_mdpr_chunks] times):
147 * we skip MDPR chunks until we reach the one of the stream
148 * we're interested in, and forward that ([size]+[data]) to
149 * the RM demuxer to parse the stream-specific header data.
153 init_put_byte(&pb, rdt->mlti_data, rdt->mlti_data_size, 0,
154 NULL, NULL, NULL, NULL);
156 if (tag == MKTAG('M', 'L', 'T', 'I')) {
159 /* read index of MDPR chunk numbers */
161 if (rule_nr < 0 || rule_nr >= num)
163 url_fskip(&pb, rule_nr * 2);
164 chunk_nr = get_be16(&pb);
165 url_fskip(&pb, (num - 1 - rule_nr) * 2);
167 /* read MDPR chunks */
172 url_fskip(&pb, get_be32(&pb));
173 size = get_be32(&pb);
175 size = rdt->mlti_data_size;
176 url_fseek(&pb, 0, SEEK_SET);
178 if (ff_rm_read_mdpr_codecdata(rdt->rmctx, &pb, st, rdt->rmst[st->index], size) < 0)
185 * Actual data handling.
189 ff_rdt_parse_header(const uint8_t *buf, int len,
190 int *pset_id, int *pseq_no, int *pstream_id,
191 int *pis_keyframe, uint32_t *ptimestamp)
194 int consumed = 0, set_id, seq_no, stream_id, is_keyframe,
195 len_included, need_reliable;
198 /* skip status packets */
199 while (len >= 5 && buf[1] == 0xFF /* status packet */) {
202 if (!(buf[0] & 0x80))
203 return -1; /* not followed by a data packet */
205 pkt_len = AV_RB16(buf+3);
213 * Layout of the header (in bits):
215 * Flag indicating whether this header includes a length field;
216 * this can be used to concatenate multiple RDT packets in a
217 * single UDP/TCP data frame and is used to precede RDT data
218 * by stream status packets
220 * Flag indicating whether this header includes a "reliable
221 * sequence number"; these are apparently sequence numbers of
222 * data packets alone. For data packets, this flag is always
223 * set, according to the Real documentation [1]
225 * ID of a set of streams of identical content, possibly with
226 * different codecs or bitrates
228 * Flag set for certain streams deemed less tolerable for packet
231 * Packet sequence number; if >=0xFF00, this is a non-data packet
232 * containing stream status info, the second byte indicates the
233 * type of status packet (see wireshark docs / source code [2])
234 * if (len_included) {
237 * packet_len = remainder of UDP/TCP frame
240 * Back-to-Back flag; used for timing, set for one in every 10
241 * packets, according to the Real documentation [1]
243 * Slow-data flag; currently unused, according to Real docs [1]
245 * ID of the stream within this particular set of streams
247 * Non-keyframe flag (unset if packet belongs to a keyframe)
248 * 32: timestamp (PTS)
249 * if (set_id == 0x1F) {
250 * 16: set_id (extended set-of-streams ID; see set_id)
252 * if (need_reliable) {
253 * 16: reliable_seq_no
254 * Reliable sequence number (see need_reliable)
256 * if (stream_id == 0x3F) {
257 * 16: stream_id (extended stream ID; see stream_id)
259 * [1] https://protocol.helixcommunity.org/files/2005/devdocs/RDT_Feature_Level_20.txt
260 * [2] http://www.wireshark.org/docs/dfref/r/rdt.html and
261 * http://anonsvn.wireshark.org/viewvc/trunk/epan/dissectors/packet-rdt.c
263 init_get_bits(&gb, buf, len << 3);
264 len_included = get_bits1(&gb);
265 need_reliable = get_bits1(&gb);
266 set_id = get_bits(&gb, 5);
268 seq_no = get_bits(&gb, 16);
272 stream_id = get_bits(&gb, 5);
273 is_keyframe = !get_bits1(&gb);
274 timestamp = get_bits_long(&gb, 32);
276 set_id = get_bits(&gb, 16);
279 if (stream_id == 0x1f)
280 stream_id = get_bits(&gb, 16);
282 if (pset_id) *pset_id = set_id;
283 if (pseq_no) *pseq_no = seq_no;
284 if (pstream_id) *pstream_id = stream_id;
285 if (pis_keyframe) *pis_keyframe = is_keyframe;
286 if (ptimestamp) *ptimestamp = timestamp;
288 return consumed + (get_bits_count(&gb) >> 3);
291 /**< return 0 on packet, no more left, 1 on packet, 1 on partial packet... */
293 rdt_parse_packet (AVFormatContext *ctx, PayloadContext *rdt, AVStream *st,
294 AVPacket *pkt, uint32_t *timestamp,
295 const uint8_t *buf, int len, int flags)
300 if (rdt->audio_pkt_cnt == 0) {
303 init_put_byte(&pb, buf, len, 0, NULL, NULL, NULL, NULL);
304 flags = (flags & RTP_FLAG_KEY) ? 2 : 0;
305 res = ff_rm_parse_packet (rdt->rmctx, &pb, st, rdt->rmst[st->index], len, pkt,
306 &seq, flags, *timestamp);
307 pos = url_ftell(&pb);
311 if (st->codec->codec_id == CODEC_ID_AAC) {
312 memcpy (rdt->buffer, buf + pos, len - pos);
313 rdt->rmctx->pb = av_alloc_put_byte (rdt->buffer, len - pos, 0,
314 NULL, NULL, NULL, NULL);
321 ff_rm_retrieve_cache (rdt->rmctx, rdt->rmctx->pb,
322 st, rdt->rmst[st->index], pkt);
323 if (rdt->audio_pkt_cnt == 0 &&
324 st->codec->codec_id == CODEC_ID_AAC)
325 av_freep(&rdt->rmctx->pb);
327 pkt->stream_index = st->index;
328 pkt->pts = *timestamp;
330 return rdt->audio_pkt_cnt > 0;
334 ff_rdt_parse_packet(RDTDemuxContext *s, AVPacket *pkt,
335 uint8_t **bufptr, int len)
337 uint8_t *buf = bufptr ? *bufptr : NULL;
338 int seq_no, flags = 0, stream_id, set_id, is_keyframe;
342 if (!s->parse_packet)
345 if (!buf && s->prev_stream_id != -1) {
346 /* return the next packets, if any */
347 timestamp= 0; ///< Should not be used if buf is NULL, but should be set to the timestamp of the packet returned....
348 rv= s->parse_packet(s->ic, s->dynamic_protocol_context,
349 s->streams[s->prev_stream_id],
350 pkt, ×tamp, NULL, 0, flags);
356 rv = ff_rdt_parse_header(buf, len, &set_id, &seq_no, &stream_id, &is_keyframe, ×tamp);
360 (set_id != s->prev_set_id || timestamp != s->prev_timestamp ||
361 stream_id != s->prev_stream_id)) {
362 flags |= RTP_FLAG_KEY;
363 s->prev_set_id = set_id;
364 s->prev_timestamp = timestamp;
366 s->prev_stream_id = stream_id;
370 if (s->prev_stream_id >= s->n_streams) {
371 s->prev_stream_id = -1;
375 rv = s->parse_packet(s->ic, s->dynamic_protocol_context,
376 s->streams[s->prev_stream_id],
377 pkt, ×tamp, buf, len, flags);
383 ff_rdt_subscribe_rule (char *cmd, int size,
384 int stream_nr, int rule_nr)
386 av_strlcatf(cmd, size, "stream=%d;rule=%d,stream=%d;rule=%d",
387 stream_nr, rule_nr * 2, stream_nr, rule_nr * 2 + 1);
390 static unsigned char *
391 rdt_parse_b64buf (unsigned int *target_len, const char *p)
393 unsigned char *target;
397 len -= 2; /* skip embracing " at start/end */
399 *target_len = len * 3 / 4;
400 target = av_mallocz(*target_len + FF_INPUT_BUFFER_PADDING_SIZE);
401 av_base64_decode(target, p, *target_len);
406 rdt_parse_sdp_line (AVFormatContext *s, int st_index,
407 PayloadContext *rdt, const char *line)
409 AVStream *stream = s->streams[st_index];
410 const char *p = line;
412 if (av_strstart(p, "OpaqueData:buffer;", &p)) {
413 rdt->mlti_data = rdt_parse_b64buf(&rdt->mlti_data_size, p);
414 } else if (av_strstart(p, "StartTime:integer;", &p))
415 stream->first_dts = atoi(p);
416 else if (av_strstart(p, "ASMRuleBook:string;", &p)) {
419 for (n = 0; n < s->nb_streams; n++)
420 if (s->streams[n]->id == stream->id) {
421 int count = s->streams[n]->index + 1;
422 if (first == -1) first = n;
423 if (rdt->nb_rmst < count) {
424 RMStream **rmst= av_realloc(rdt->rmst, count*sizeof(*rmst));
426 return AVERROR(ENOMEM);
427 memset(rmst + rdt->nb_rmst, 0,
428 (count - rdt->nb_rmst) * sizeof(*rmst));
430 rdt->nb_rmst = count;
432 rdt->rmst[s->streams[n]->index] = ff_rm_alloc_rmstream();
433 rdt_load_mdpr(rdt, s->streams[n], (n - first) * 2);
435 if (s->streams[n]->codec->codec_id == CODEC_ID_AAC)
436 s->streams[n]->codec->frame_size = 1; // FIXME
444 real_parse_asm_rule(AVStream *st, const char *p, const char *end)
447 /* can be either averagebandwidth= or AverageBandwidth= */
448 if (sscanf(p, " %*1[Aa]verage%*1[Bb]andwidth=%d", &st->codec->bit_rate) == 1)
450 if (!(p = strchr(p, ',')) || p > end)
457 add_dstream(AVFormatContext *s, AVStream *orig_st)
461 if (!(st = av_new_stream(s, orig_st->id)))
463 st->codec->codec_type = orig_st->codec->codec_type;
464 st->first_dts = orig_st->first_dts;
470 real_parse_asm_rulebook(AVFormatContext *s, AVStream *orig_st,
474 int n_rules = 0, odd = 0;
478 * The ASMRuleBook contains a list of comma-separated strings per rule,
479 * and each rule is separated by a ;. The last one also has a ; at the
480 * end so we can use it as delimiter.
481 * Every rule occurs twice, once for when the RTSP packet header marker
482 * is set and once for if it isn't. We only read the first because we
483 * don't care much (that's what the "odd" variable is for).
484 * Each rule contains a set of one or more statements, optionally
485 * preceeded by a single condition. If there's a condition, the rule
486 * starts with a '#'. Multiple conditions are merged between brackets,
487 * so there are never multiple conditions spread out over separate
488 * statements. Generally, these conditions are bitrate limits (min/max)
489 * for multi-bitrate streams.
493 if (!(end = strchr(p, ';')))
495 if (!odd && end != p) {
497 st = add_dstream(s, orig_st);
502 real_parse_asm_rule(st, p, end);
511 ff_real_parse_sdp_a_line (AVFormatContext *s, int stream_index,
514 const char *p = line;
516 if (av_strstart(p, "ASMRuleBook:string;", &p))
517 real_parse_asm_rulebook(s, s->streams[stream_index], p);
520 static PayloadContext *
521 rdt_new_context (void)
523 PayloadContext *rdt = av_mallocz(sizeof(PayloadContext));
525 av_open_input_stream(&rdt->rmctx, NULL, "", &ff_rdt_demuxer, NULL);
531 rdt_free_context (PayloadContext *rdt)
535 for (i = 0; i < rdt->nb_rmst; i++)
537 ff_rm_free_rmstream(rdt->rmst[i]);
538 av_freep(&rdt->rmst[i]);
541 av_close_input_stream(rdt->rmctx);
542 av_freep(&rdt->mlti_data);
543 av_freep(&rdt->rmst);
547 #define RDT_HANDLER(n, s, t) \
548 static RTPDynamicProtocolHandler ff_rdt_ ## n ## _handler = { \
551 .codec_id = CODEC_ID_NONE, \
552 .parse_sdp_a_line = rdt_parse_sdp_line, \
553 .open = rdt_new_context, \
554 .close = rdt_free_context, \
555 .parse_packet = rdt_parse_packet \
558 RDT_HANDLER(live_video, "x-pn-multirate-realvideo-live", AVMEDIA_TYPE_VIDEO);
559 RDT_HANDLER(live_audio, "x-pn-multirate-realaudio-live", AVMEDIA_TYPE_AUDIO);
560 RDT_HANDLER(video, "x-pn-realvideo", AVMEDIA_TYPE_VIDEO);
561 RDT_HANDLER(audio, "x-pn-realaudio", AVMEDIA_TYPE_AUDIO);
563 void av_register_rdt_dynamic_payload_handlers(void)
565 ff_register_dynamic_payload_handler(&ff_rdt_video_handler);
566 ff_register_dynamic_payload_handler(&ff_rdt_audio_handler);
567 ff_register_dynamic_payload_handler(&ff_rdt_live_video_handler);
568 ff_register_dynamic_payload_handler(&ff_rdt_live_audio_handler);