1 /* GStreamer MPEG audio parser
2 * Copyright (C) 2006-2007 Jan Schmidt <thaytan@mad.scientist.com>
3 * Copyright (C) 2010 Mark Nauwelaerts <mnauw users sf net>
4 * Copyright (C) 2010 Nokia Corporation. All rights reserved.
5 * Contact: Stefan Kost <stefan.kost@nokia.com>
6 *
7 * This library is free software; you can redistribute it and/or
8 * modify it under the terms of the GNU Library General Public
9 * License as published by the Free Software Foundation; either
10 * version 2 of the License, or (at your option) any later version.
11 *
12 * This library 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 * Library General Public License for more details.
16 *
17 * You should have received a copy of the GNU Library General Public
18 * License along with this library; if not, write to the
19 * Free Software Foundation, Inc., 51 Franklin St, Fifth Floor,
20 * Boston, MA 02110-1301, USA.
21 */
22 /**
23 * SECTION:element-mpegaudioparse
24 * @short_description: MPEG audio parser
25 * @see_also: #GstAmrParse, #GstAACParse
26 *
27 * Parses and frames mpeg1 audio streams. Provides seeking.
28 *
29 * <refsect2>
30 * <title>Example launch line</title>
31 * |[
32 * gst-launch-1.0 filesrc location=test.mp3 ! mpegaudioparse ! mpg123audiodec
33 * ! audioconvert ! audioresample ! autoaudiosink
34 * ]|
35 * </refsect2>
36 */
37
38 /* FIXME: we should make the base class (GstBaseParse) aware of the
39 * XING seek table somehow, so it can use it properly for things like
40 * accurate seeks. Currently it can only do a lookup via the convert function,
41 * but then doesn't know what the result represents exactly. One could either
42 * add a vfunc for index lookup, or just make mpegaudioparse populate the
43 * base class's index via the API provided.
44 */
45 #ifdef HAVE_CONFIG_H
46 #include "config.h"
47 #endif
48
49 #include <string.h>
50
51 #include "gstmpegaudioparse.h"
52 #include <gst/base/gstbytereader.h>
53 #include <gst/pbutils/pbutils.h>
54
55 GST_DEBUG_CATEGORY_STATIC (mpeg_audio_parse_debug);
56 #define GST_CAT_DEFAULT mpeg_audio_parse_debug
57
58 #define MPEG_AUDIO_CHANNEL_MODE_UNKNOWN -1
59 #define MPEG_AUDIO_CHANNEL_MODE_STEREO 0
60 #define MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO 1
61 #define MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL 2
62 #define MPEG_AUDIO_CHANNEL_MODE_MONO 3
63
64 #define CRC_UNKNOWN -1
65 #define CRC_PROTECTED 0
66 #define CRC_NOT_PROTECTED 1
67
68 #define XING_FRAMES_FLAG 0x0001
69 #define XING_BYTES_FLAG 0x0002
70 #define XING_TOC_FLAG 0x0004
71 #define XING_VBR_SCALE_FLAG 0x0008
72
73 #define MIN_FRAME_SIZE 6
74
75 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
76 GST_PAD_SRC,
77 GST_PAD_ALWAYS,
78 GST_STATIC_CAPS ("audio/mpeg, "
79 "mpegversion = (int) 1, "
80 "layer = (int) [ 1, 3 ], "
81 "mpegaudioversion = (int) [ 1, 3], "
82 "rate = (int) [ 8000, 48000 ], "
83 "channels = (int) [ 1, 2 ], " "parsed=(boolean) true")
84 );
85
86 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
87 GST_PAD_SINK,
88 GST_PAD_ALWAYS,
89 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) 1")
90 );
91
92 static void gst_mpeg_audio_parse_finalize (GObject * object);
93
94 static gboolean gst_mpeg_audio_parse_start (GstBaseParse * parse);
95 static gboolean gst_mpeg_audio_parse_stop (GstBaseParse * parse);
96 static GstFlowReturn gst_mpeg_audio_parse_handle_frame (GstBaseParse * parse,
97 GstBaseParseFrame * frame, gint * skipsize);
98 static GstFlowReturn gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
99 GstBaseParseFrame * frame);
100 static gboolean gst_mpeg_audio_parse_convert (GstBaseParse * parse,
101 GstFormat src_format, gint64 src_value,
102 GstFormat dest_format, gint64 * dest_value);
103 static GstCaps *gst_mpeg_audio_parse_get_sink_caps (GstBaseParse * parse,
104 GstCaps * filter);
105
106 static void gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse *
107 mp3parse, GstBuffer * buf);
108
109 #define gst_mpeg_audio_parse_parent_class parent_class
110 G_DEFINE_TYPE (GstMpegAudioParse, gst_mpeg_audio_parse, GST_TYPE_BASE_PARSE);
111
112 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
113 (gst_mpeg_audio_channel_mode_get_type())
114
115 static const GEnumValue mpeg_audio_channel_mode[] = {
116 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
117 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
118 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
119 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
120 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
121 {0, NULL, NULL},
122 };
123
124 static GType
gst_mpeg_audio_channel_mode_get_type(void)125 gst_mpeg_audio_channel_mode_get_type (void)
126 {
127 static GType mpeg_audio_channel_mode_type = 0;
128
129 if (!mpeg_audio_channel_mode_type) {
130 mpeg_audio_channel_mode_type =
131 g_enum_register_static ("GstMpegAudioChannelMode",
132 mpeg_audio_channel_mode);
133 }
134 return mpeg_audio_channel_mode_type;
135 }
136
137 static const gchar *
gst_mpeg_audio_channel_mode_get_nick(gint mode)138 gst_mpeg_audio_channel_mode_get_nick (gint mode)
139 {
140 guint i;
141 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
142 if (mpeg_audio_channel_mode[i].value == mode)
143 return mpeg_audio_channel_mode[i].value_nick;
144 }
145 return NULL;
146 }
147
148 static void
gst_mpeg_audio_parse_class_init(GstMpegAudioParseClass * klass)149 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
150 {
151 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
152 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
153 GObjectClass *object_class = G_OBJECT_CLASS (klass);
154
155 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
156 "MPEG1 audio stream parser");
157
158 object_class->finalize = gst_mpeg_audio_parse_finalize;
159
160 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
161 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
162 parse_class->handle_frame =
163 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_handle_frame);
164 parse_class->pre_push_frame =
165 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
166 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
167 parse_class->get_sink_caps =
168 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_get_sink_caps);
169
170 /* register tags */
171 #define GST_TAG_CRC "has-crc"
172 #define GST_TAG_MODE "channel-mode"
173
174 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
175 "has crc", "Using CRC", NULL);
176 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
177 "channel mode", "MPEG audio channel mode", NULL);
178
179 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
180
181 gst_element_class_add_static_pad_template (element_class, &sink_template);
182 gst_element_class_add_static_pad_template (element_class, &src_template);
183
184 gst_element_class_set_static_metadata (element_class, "MPEG1 Audio Parser",
185 "Codec/Parser/Audio",
186 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
187 "Jan Schmidt <thaytan@mad.scientist.com>,"
188 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
189 }
190
191 static void
gst_mpeg_audio_parse_reset(GstMpegAudioParse * mp3parse)192 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
193 {
194 mp3parse->channels = -1;
195 mp3parse->rate = -1;
196 mp3parse->sent_codec_tag = FALSE;
197 mp3parse->last_posted_crc = CRC_UNKNOWN;
198 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
199 mp3parse->freerate = 0;
200
201 mp3parse->hdr_bitrate = 0;
202
203 mp3parse->xing_flags = 0;
204 mp3parse->xing_bitrate = 0;
205 mp3parse->xing_frames = 0;
206 mp3parse->xing_total_time = 0;
207 mp3parse->xing_bytes = 0;
208 mp3parse->xing_vbr_scale = 0;
209 memset (mp3parse->xing_seek_table, 0, sizeof (mp3parse->xing_seek_table));
210 memset (mp3parse->xing_seek_table_inverse, 0,
211 sizeof (mp3parse->xing_seek_table_inverse));
212
213 mp3parse->vbri_bitrate = 0;
214 mp3parse->vbri_frames = 0;
215 mp3parse->vbri_total_time = 0;
216 mp3parse->vbri_bytes = 0;
217 mp3parse->vbri_seek_points = 0;
218 g_free (mp3parse->vbri_seek_table);
219 mp3parse->vbri_seek_table = NULL;
220
221 mp3parse->encoder_delay = 0;
222 mp3parse->encoder_padding = 0;
223 }
224
225 static void
gst_mpeg_audio_parse_init(GstMpegAudioParse * mp3parse)226 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse)
227 {
228 gst_mpeg_audio_parse_reset (mp3parse);
229 GST_PAD_SET_ACCEPT_INTERSECT (GST_BASE_PARSE_SINK_PAD (mp3parse));
230 GST_PAD_SET_ACCEPT_TEMPLATE (GST_BASE_PARSE_SINK_PAD (mp3parse));
231 }
232
233 static void
gst_mpeg_audio_parse_finalize(GObject * object)234 gst_mpeg_audio_parse_finalize (GObject * object)
235 {
236 G_OBJECT_CLASS (parent_class)->finalize (object);
237 }
238
239 static gboolean
gst_mpeg_audio_parse_start(GstBaseParse * parse)240 gst_mpeg_audio_parse_start (GstBaseParse * parse)
241 {
242 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
243
244 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), MIN_FRAME_SIZE);
245 GST_DEBUG_OBJECT (parse, "starting");
246
247 gst_mpeg_audio_parse_reset (mp3parse);
248
249 return TRUE;
250 }
251
252 static gboolean
gst_mpeg_audio_parse_stop(GstBaseParse * parse)253 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
254 {
255 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
256
257 GST_DEBUG_OBJECT (parse, "stopping");
258
259 gst_mpeg_audio_parse_reset (mp3parse);
260
261 return TRUE;
262 }
263
264 static const guint mp3types_bitrates[2][3][16] = {
265 {
266 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
267 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
268 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
269 },
270 {
271 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
272 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
273 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
274 },
275 };
276
277 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
278 {22050, 24000, 16000},
279 {11025, 12000, 8000}
280 };
281
282 static inline guint
mp3_type_frame_length_from_header(GstMpegAudioParse * mp3parse,guint32 header,guint * put_version,guint * put_layer,guint * put_channels,guint * put_bitrate,guint * put_samplerate,guint * put_mode,guint * put_crc)283 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
284 guint * put_version, guint * put_layer, guint * put_channels,
285 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
286 guint * put_crc)
287 {
288 guint length;
289 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
290 gulong version;
291 gint lsf, mpg25;
292
293 if (header & (1 << 20)) {
294 lsf = (header & (1 << 19)) ? 0 : 1;
295 mpg25 = 0;
296 } else {
297 lsf = 1;
298 mpg25 = 1;
299 }
300
301 version = 1 + lsf + mpg25;
302
303 layer = 4 - ((header >> 17) & 0x3);
304
305 crc = (header >> 16) & 0x1;
306
307 bitrate = (header >> 12) & 0xF;
308 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
309 if (!bitrate) {
310 GST_LOG_OBJECT (mp3parse, "using freeform bitrate");
311 bitrate = mp3parse->freerate;
312 }
313
314 samplerate = (header >> 10) & 0x3;
315 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
316
317 /* force 0 length if 0 bitrate */
318 padding = (bitrate > 0) ? (header >> 9) & 0x1 : 0;
319
320 mode = (header >> 6) & 0x3;
321 channels = (mode == 3) ? 1 : 2;
322
323 switch (layer) {
324 case 1:
325 length = 4 * ((bitrate * 12) / samplerate + padding);
326 break;
327 case 2:
328 length = (bitrate * 144) / samplerate + padding;
329 break;
330 default:
331 case 3:
332 length = (bitrate * 144) / (samplerate << lsf) + padding;
333 break;
334 }
335
336 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
337 length);
338 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
339 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
340 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
341
342 if (put_version)
343 *put_version = version;
344 if (put_layer)
345 *put_layer = layer;
346 if (put_channels)
347 *put_channels = channels;
348 if (put_bitrate)
349 *put_bitrate = bitrate;
350 if (put_samplerate)
351 *put_samplerate = samplerate;
352 if (put_mode)
353 *put_mode = mode;
354 if (put_crc)
355 *put_crc = crc;
356
357 return length;
358 }
359
360 /* Minimum number of consecutive, valid-looking frames to consider
361 * for resyncing */
362 #define MIN_RESYNC_FRAMES 3
363
364 /* Perform extended validation to check that subsequent headers match
365 * the first header given here in important characteristics, to avoid
366 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
367 * frames to match their major characteristics.
368 *
369 * If at_eos is set to TRUE, we just check that we don't find any invalid
370 * frames in whatever data is available, rather than requiring a full
371 * MIN_RESYNC_FRAMES of data.
372 *
373 * Returns TRUE if we've seen enough data to validate or reject the frame.
374 * If TRUE is returned, then *valid contains TRUE if it validated, or false
375 * if we decided it was false sync.
376 * If FALSE is returned, then *valid contains minimum needed data.
377 */
378 static gboolean
gst_mp3parse_validate_extended(GstMpegAudioParse * mp3parse,GstBuffer * buf,guint32 header,int bpf,gboolean at_eos,gint * valid)379 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
380 guint32 header, int bpf, gboolean at_eos, gint * valid)
381 {
382 guint32 next_header;
383 GstMapInfo map;
384 gboolean res = TRUE;
385 int frames_found = 1;
386 int offset = bpf;
387
388 gst_buffer_map (buf, &map, GST_MAP_READ);
389
390 while (frames_found < MIN_RESYNC_FRAMES) {
391 /* Check if we have enough data for all these frames, plus the next
392 frame header. */
393 if (map.size < offset + 4) {
394 if (at_eos) {
395 /* Running out of data at EOS is fine; just accept it */
396 *valid = TRUE;
397 goto cleanup;
398 } else {
399 *valid = offset + 4;
400 res = FALSE;
401 goto cleanup;
402 }
403 }
404
405 next_header = GST_READ_UINT32_BE (map.data + offset);
406 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
407 offset, (unsigned int) header, (unsigned int) next_header, bpf);
408
409 /* mask the bits which are allowed to differ between frames */
410 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
411 (0x1 << 9) /* padding */ | \
412 (0xf << 4) /* mode|mode extension */ | \
413 (0xf)) /* copyright|emphasis */
414
415 if ((next_header & HDRMASK) != (header & HDRMASK)) {
416 /* If any of the unmasked bits don't match, then it's not valid */
417 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
418 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
419 (guint) header, (guint) header & HDRMASK, (guint) next_header,
420 (guint) next_header & HDRMASK, bpf);
421 *valid = FALSE;
422 goto cleanup;
423 } else if (((next_header >> 12) & 0xf) == 0xf) {
424 /* The essential parts were the same, but the bitrate held an
425 invalid value - also reject */
426 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
427 *valid = FALSE;
428 goto cleanup;
429 }
430
431 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
432 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
433
434 /* if no bitrate, and no freeform rate known, then fail */
435 if (G_UNLIKELY (!bpf)) {
436 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate 0)");
437 *valid = FALSE;
438 goto cleanup;
439 }
440
441 offset += bpf;
442 frames_found++;
443 }
444
445 *valid = TRUE;
446
447 cleanup:
448 gst_buffer_unmap (buf, &map);
449 return res;
450 }
451
452 static gboolean
gst_mpeg_audio_parse_head_check(GstMpegAudioParse * mp3parse,unsigned long head)453 gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
454 unsigned long head)
455 {
456 GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
457 /* if it's not a valid sync */
458 if ((head & 0xffe00000) != 0xffe00000) {
459 GST_WARNING_OBJECT (mp3parse, "invalid sync");
460 return FALSE;
461 }
462 /* if it's an invalid MPEG version */
463 if (((head >> 19) & 3) == 0x1) {
464 GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
465 (head >> 19) & 3);
466 return FALSE;
467 }
468 /* if it's an invalid layer */
469 if (!((head >> 17) & 3)) {
470 GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
471 return FALSE;
472 }
473 /* if it's an invalid bitrate */
474 if (((head >> 12) & 0xf) == 0xf) {
475 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
476 return FALSE;
477 }
478 /* if it's an invalid samplerate */
479 if (((head >> 10) & 0x3) == 0x3) {
480 GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
481 (head >> 10) & 0x3);
482 return FALSE;
483 }
484
485 if ((head & 0x3) == 0x2) {
486 /* Ignore this as there are some files with emphasis 0x2 that can
487 * be played fine. See BGO #537235 */
488 GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
489 }
490
491 return TRUE;
492 }
493
494 /* Determines possible freeform frame rate/size by looking for next
495 * header with valid bitrate (0 or otherwise valid) (and sufficiently
496 * matching current header).
497 *
498 * Returns TRUE if we've found such one, and *rate then contains rate
499 * (or *rate contains 0 if decided no freeframe size could be determined).
500 * If not enough data, returns FALSE.
501 */
502 static gboolean
gst_mp3parse_find_freerate(GstMpegAudioParse * mp3parse,GstMapInfo * map,guint32 header,gboolean at_eos,gint * _rate)503 gst_mp3parse_find_freerate (GstMpegAudioParse * mp3parse, GstMapInfo * map,
504 guint32 header, gboolean at_eos, gint * _rate)
505 {
506 guint32 next_header;
507 const guint8 *data;
508 guint available;
509 int offset = 4;
510 gulong samplerate, rate, layer, padding;
511 gboolean valid;
512 gint lsf, mpg25;
513
514 available = map->size;
515 data = map->data;
516
517 *_rate = 0;
518
519 /* pick apart header again partially */
520 if (header & (1 << 20)) {
521 lsf = (header & (1 << 19)) ? 0 : 1;
522 mpg25 = 0;
523 } else {
524 lsf = 1;
525 mpg25 = 1;
526 }
527 layer = 4 - ((header >> 17) & 0x3);
528 samplerate = (header >> 10) & 0x3;
529 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
530 padding = (header >> 9) & 0x1;
531
532 for (; offset < available; ++offset) {
533 /* Check if we have enough data for all these frames, plus the next
534 frame header. */
535 if (available < offset + 4) {
536 if (at_eos) {
537 /* Running out of data; failed to determine size */
538 return TRUE;
539 } else {
540 return FALSE;
541 }
542 }
543
544 valid = FALSE;
545 next_header = GST_READ_UINT32_BE (data + offset);
546 if ((next_header & 0xFFE00000) != 0xFFE00000)
547 goto next;
548
549 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X",
550 offset, (unsigned int) header, (unsigned int) next_header);
551
552 if ((next_header & HDRMASK) != (header & HDRMASK)) {
553 /* If any of the unmasked bits don't match, then it's not valid */
554 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
555 "(header=%08X (%08X), header2=%08X (%08X))",
556 (guint) header, (guint) header & HDRMASK, (guint) next_header,
557 (guint) next_header & HDRMASK);
558 goto next;
559 } else if (((next_header >> 12) & 0xf) == 0xf) {
560 /* The essential parts were the same, but the bitrate held an
561 invalid value - also reject */
562 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
563 goto next;
564 }
565
566 valid = TRUE;
567
568 next:
569 /* almost accept as free frame */
570 if (layer == 1) {
571 rate = samplerate * (offset - 4 * padding + 4) / 48000;
572 } else {
573 rate = samplerate * (offset - padding + 1) / (144 >> lsf) / 1000;
574 }
575
576 if (valid) {
577 GST_LOG_OBJECT (mp3parse, "calculated rate %lu", rate * 1000);
578 if (rate < 8 || (layer == 3 && rate > 640)) {
579 GST_DEBUG_OBJECT (mp3parse, "rate invalid");
580 if (rate < 8) {
581 /* maybe some hope */
582 continue;
583 } else {
584 GST_DEBUG_OBJECT (mp3parse, "aborting");
585 /* give up */
586 break;
587 }
588 }
589 *_rate = rate * 1000;
590 break;
591 } else {
592 /* avoid indefinite searching */
593 if (rate > 1000) {
594 GST_DEBUG_OBJECT (mp3parse, "exceeded sanity rate; aborting");
595 break;
596 }
597 }
598 }
599
600 return TRUE;
601 }
602
603 static GstFlowReturn
gst_mpeg_audio_parse_handle_frame(GstBaseParse * parse,GstBaseParseFrame * frame,gint * skipsize)604 gst_mpeg_audio_parse_handle_frame (GstBaseParse * parse,
605 GstBaseParseFrame * frame, gint * skipsize)
606 {
607 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
608 GstBuffer *buf = frame->buffer;
609 GstByteReader reader;
610 gint off, bpf = 0;
611 gboolean lost_sync, draining, valid, caps_change;
612 guint32 header;
613 guint bitrate, layer, rate, channels, version, mode, crc;
614 GstMapInfo map;
615 gboolean res = FALSE;
616
617 gst_buffer_map (buf, &map, GST_MAP_READ);
618 if (G_UNLIKELY (map.size < 6)) {
619 *skipsize = 1;
620 goto cleanup;
621 }
622
623 gst_byte_reader_init (&reader, map.data, map.size);
624
625 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
626 0, map.size);
627
628 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
629
630 /* didn't find anything that looks like a sync word, skip */
631 if (off < 0) {
632 *skipsize = map.size - 3;
633 goto cleanup;
634 }
635
636 /* possible frame header, but not at offset 0? skip bytes before sync */
637 if (off > 0) {
638 *skipsize = off;
639 goto cleanup;
640 }
641
642 /* make sure the values in the frame header look sane */
643 header = GST_READ_UINT32_BE (map.data);
644 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
645 *skipsize = 1;
646 goto cleanup;
647 }
648
649 GST_LOG_OBJECT (parse, "got frame");
650
651 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
652 draining = GST_BASE_PARSE_DRAINING (parse);
653
654 if (G_UNLIKELY (lost_sync))
655 mp3parse->freerate = 0;
656
657 bpf = mp3_type_frame_length_from_header (mp3parse, header,
658 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
659
660 if (channels != mp3parse->channels || rate != mp3parse->rate ||
661 layer != mp3parse->layer || version != mp3parse->version)
662 caps_change = TRUE;
663 else
664 caps_change = FALSE;
665
666 /* maybe free format */
667 if (bpf == 0) {
668 GST_LOG_OBJECT (mp3parse, "possibly free format");
669 if (lost_sync || mp3parse->freerate == 0) {
670 GST_DEBUG_OBJECT (mp3parse, "finding free format rate");
671 if (!gst_mp3parse_find_freerate (mp3parse, &map, header, draining,
672 &valid)) {
673 /* not enough data */
674 gst_base_parse_set_min_frame_size (parse, valid);
675 *skipsize = 0;
676 goto cleanup;
677 } else {
678 GST_DEBUG_OBJECT (parse, "determined freeform size %d", valid);
679 mp3parse->freerate = valid;
680 }
681 }
682 /* try again */
683 bpf = mp3_type_frame_length_from_header (mp3parse, header,
684 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
685 if (!bpf) {
686 /* did not come up with valid freeform length, reject after all */
687 *skipsize = 1;
688 goto cleanup;
689 }
690 }
691
692 if (!draining && (lost_sync || caps_change)) {
693 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
694 &valid)) {
695 /* not enough data */
696 gst_base_parse_set_min_frame_size (parse, valid);
697 *skipsize = 0;
698 goto cleanup;
699 } else {
700 if (!valid) {
701 *skipsize = off + 2;
702 goto cleanup;
703 }
704 }
705 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
706 /* avoid caps jitter that we can't be sure of */
707 *skipsize = off + 2;
708 goto cleanup;
709 }
710
711 /* restore default minimum */
712 gst_base_parse_set_min_frame_size (parse, MIN_FRAME_SIZE);
713
714 res = TRUE;
715
716 /* metadata handling */
717 if (G_UNLIKELY (caps_change)) {
718 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
719 "mpegversion", G_TYPE_INT, 1,
720 "mpegaudioversion", G_TYPE_INT, version,
721 "layer", G_TYPE_INT, layer,
722 "rate", G_TYPE_INT, rate,
723 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
724 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
725 gst_caps_unref (caps);
726
727 mp3parse->rate = rate;
728 mp3parse->channels = channels;
729 mp3parse->layer = layer;
730 mp3parse->version = version;
731
732 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
733 if (mp3parse->layer == 1)
734 mp3parse->spf = 384;
735 else if (mp3parse->layer == 2)
736 mp3parse->spf = 1152;
737 else if (mp3parse->version == 1) {
738 mp3parse->spf = 1152;
739 } else {
740 /* MPEG-2 or "2.5" */
741 mp3parse->spf = 576;
742 }
743
744 /* lead_in:
745 * We start pushing 9 frames earlier (29 frames for MPEG2) than
746 * segment start to be able to decode the first frame we want.
747 * 9 (29) frames are the theoretical maximum of frames that contain
748 * data for the current frame (bit reservoir).
749 *
750 * lead_out:
751 * Some mp3 streams have an offset in the timestamps, for which we have to
752 * push the frame *after* the end position in order for the decoder to be
753 * able to decode everything up until the segment.stop position. */
754 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
755 (version == 1) ? 10 : 30, 2);
756 }
757
758 mp3parse->hdr_bitrate = bitrate;
759
760 /* For first frame; check for seek tables and output a codec tag */
761 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
762
763 /* store some frame info for later processing */
764 mp3parse->last_crc = crc;
765 mp3parse->last_mode = mode;
766
767 cleanup:
768 gst_buffer_unmap (buf, &map);
769
770 if (res && bpf <= map.size) {
771 return gst_base_parse_finish_frame (parse, frame, bpf);
772 }
773
774 return GST_FLOW_OK;
775 }
776
777 static void
gst_mpeg_audio_parse_handle_first_frame(GstMpegAudioParse * mp3parse,GstBuffer * buf)778 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
779 GstBuffer * buf)
780 {
781 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
782 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
783 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
784 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
785 gint offset_xing, offset_vbri;
786 guint64 avail;
787 gint64 upstream_total_bytes = 0;
788 guint32 read_id_xing = 0, read_id_vbri = 0;
789 GstMapInfo map;
790 guint8 *data;
791 guint bitrate;
792
793 if (mp3parse->sent_codec_tag)
794 return;
795
796 /* Check first frame for Xing info */
797 if (mp3parse->version == 1) { /* MPEG-1 file */
798 if (mp3parse->channels == 1)
799 offset_xing = 0x11;
800 else
801 offset_xing = 0x20;
802 } else { /* MPEG-2 header */
803 if (mp3parse->channels == 1)
804 offset_xing = 0x09;
805 else
806 offset_xing = 0x11;
807 }
808
809 /* The VBRI tag is always at offset 0x20 */
810 offset_vbri = 0x20;
811
812 /* Skip the 4 bytes of the MP3 header too */
813 offset_xing += 4;
814 offset_vbri += 4;
815
816 /* Check if we have enough data to read the Xing header */
817 gst_buffer_map (buf, &map, GST_MAP_READ);
818 data = map.data;
819 avail = map.size;
820
821 if (avail >= offset_xing + 4) {
822 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
823 }
824 if (avail >= offset_vbri + 4) {
825 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
826 }
827
828 /* obtain real upstream total bytes */
829 if (!gst_pad_peer_query_duration (GST_BASE_PARSE_SINK_PAD (mp3parse),
830 GST_FORMAT_BYTES, &upstream_total_bytes))
831 upstream_total_bytes = 0;
832
833 if (read_id_xing == xing_id || read_id_xing == info_id) {
834 guint32 xing_flags;
835 guint bytes_needed = offset_xing + 8;
836 gint64 total_bytes;
837 GstClockTime total_time;
838
839 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
840
841 /* Move data after Xing header */
842 data += offset_xing + 4;
843
844 /* Read 4 base bytes of flags, big-endian */
845 xing_flags = GST_READ_UINT32_BE (data);
846 data += 4;
847 if (xing_flags & XING_FRAMES_FLAG)
848 bytes_needed += 4;
849 if (xing_flags & XING_BYTES_FLAG)
850 bytes_needed += 4;
851 if (xing_flags & XING_TOC_FLAG)
852 bytes_needed += 100;
853 if (xing_flags & XING_VBR_SCALE_FLAG)
854 bytes_needed += 4;
855 if (avail < bytes_needed) {
856 GST_DEBUG_OBJECT (mp3parse,
857 "Not enough data to read Xing header (need %d)", bytes_needed);
858 goto cleanup;
859 }
860
861 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
862 mp3parse->xing_flags = xing_flags;
863
864 if (xing_flags & XING_FRAMES_FLAG) {
865 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
866 if (mp3parse->xing_frames == 0) {
867 GST_WARNING_OBJECT (mp3parse,
868 "Invalid number of frames in Xing header");
869 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
870 } else {
871 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
872 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
873 mp3parse->rate);
874 }
875
876 data += 4;
877 } else {
878 mp3parse->xing_frames = 0;
879 mp3parse->xing_total_time = 0;
880 }
881
882 if (xing_flags & XING_BYTES_FLAG) {
883 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
884 if (mp3parse->xing_bytes == 0) {
885 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
886 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
887 }
888 data += 4;
889 } else {
890 mp3parse->xing_bytes = 0;
891 }
892
893 /* If we know the upstream size and duration, compute the
894 * total bitrate, rounded up to the nearest kbit/sec */
895 if ((total_time = mp3parse->xing_total_time) &&
896 (total_bytes = mp3parse->xing_bytes)) {
897 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
898 8 * GST_SECOND, total_time);
899 mp3parse->xing_bitrate += 500;
900 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
901 }
902
903 if (xing_flags & XING_TOC_FLAG) {
904 int i, percent = 0;
905 guchar *table = mp3parse->xing_seek_table;
906 guchar old = 0, new;
907 guint first;
908
909 first = data[0];
910 GST_DEBUG_OBJECT (mp3parse,
911 "Subtracting initial offset of %d bytes from Xing TOC", first);
912
913 /* xing seek table: percent time -> 1/256 bytepos */
914 for (i = 0; i < 100; i++) {
915 new = data[i] - first;
916 if (old > new) {
917 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
918 mp3parse->xing_flags &= ~XING_TOC_FLAG;
919 goto skip_toc;
920 }
921 mp3parse->xing_seek_table[i] = old = new;
922 }
923
924 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
925 for (i = 0; i < 256; i++) {
926 while (percent < 99 && table[percent + 1] <= i)
927 percent++;
928
929 if (table[percent] == i) {
930 mp3parse->xing_seek_table_inverse[i] = percent * 100;
931 } else if (percent < 99 && table[percent]) {
932 gdouble fa, fb, fx;
933 gint a = percent, b = percent + 1;
934
935 fa = table[a];
936 fb = table[b];
937 fx = (b - a) / (fb - fa) * (i - fa) + a;
938 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
939 } else if (percent == 99) {
940 gdouble fa, fb, fx;
941 gint a = percent, b = 100;
942
943 fa = table[a];
944 fb = 256.0;
945 fx = (b - a) / (fb - fa) * (i - fa) + a;
946 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
947 }
948 }
949 skip_toc:
950 data += 100;
951 } else {
952 memset (mp3parse->xing_seek_table, 0, sizeof (mp3parse->xing_seek_table));
953 memset (mp3parse->xing_seek_table_inverse, 0,
954 sizeof (mp3parse->xing_seek_table_inverse));
955 }
956
957 if (xing_flags & XING_VBR_SCALE_FLAG) {
958 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
959 data += 4;
960 } else
961 mp3parse->xing_vbr_scale = 0;
962
963 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
964 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
965 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
966 mp3parse->xing_vbr_scale);
967
968 /* check for truncated file */
969 if (upstream_total_bytes && mp3parse->xing_bytes &&
970 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
971 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
972 "invalidating Xing header duration and size");
973 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
974 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
975 }
976
977 /* Optional LAME tag? */
978 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
979 gchar lame_version[10] = { 0, };
980 guint tag_rev;
981 guint32 encoder_delay, encoder_padding;
982
983 memcpy (lame_version, data, 9);
984 data += 9;
985 tag_rev = data[0] >> 4;
986 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
987 tag_rev, lame_version);
988
989 /* Skip all the information we're not interested in */
990 data += 12;
991 /* Encoder delay and end padding */
992 encoder_delay = GST_READ_UINT24_BE (data);
993 encoder_delay >>= 12;
994 encoder_padding = GST_READ_UINT24_BE (data);
995 encoder_padding &= 0x000fff;
996
997 mp3parse->encoder_delay = encoder_delay;
998 mp3parse->encoder_padding = encoder_padding;
999
1000 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
1001 encoder_delay, encoder_padding);
1002 }
1003 } else if (read_id_vbri == vbri_id) {
1004 gint64 total_bytes, total_frames;
1005 GstClockTime total_time;
1006 guint16 nseek_points;
1007
1008 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
1009
1010 if (avail < offset_vbri + 26) {
1011 GST_DEBUG_OBJECT (mp3parse,
1012 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
1013 goto cleanup;
1014 }
1015
1016 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
1017
1018 /* Move data after VBRI header */
1019 data += offset_vbri + 4;
1020
1021 if (GST_READ_UINT16_BE (data) != 0x0001) {
1022 GST_WARNING_OBJECT (mp3parse,
1023 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
1024 goto cleanup;
1025 }
1026 data += 2;
1027
1028 /* Skip encoder delay */
1029 data += 2;
1030
1031 /* Skip quality */
1032 data += 2;
1033
1034 total_bytes = GST_READ_UINT32_BE (data);
1035 if (total_bytes != 0)
1036 mp3parse->vbri_bytes = total_bytes;
1037 data += 4;
1038
1039 total_frames = GST_READ_UINT32_BE (data);
1040 if (total_frames != 0) {
1041 mp3parse->vbri_frames = total_frames;
1042 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
1043 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
1044 }
1045 data += 4;
1046
1047 /* If we know the upstream size and duration, compute the
1048 * total bitrate, rounded up to the nearest kbit/sec */
1049 if ((total_time = mp3parse->vbri_total_time) &&
1050 (total_bytes = mp3parse->vbri_bytes)) {
1051 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
1052 8 * GST_SECOND, total_time);
1053 mp3parse->vbri_bitrate += 500;
1054 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
1055 }
1056
1057 nseek_points = GST_READ_UINT16_BE (data);
1058 data += 2;
1059
1060 if (nseek_points > 0) {
1061 guint scale, seek_bytes, seek_frames;
1062 gint i;
1063
1064 mp3parse->vbri_seek_points = nseek_points;
1065
1066 scale = GST_READ_UINT16_BE (data);
1067 data += 2;
1068
1069 seek_bytes = GST_READ_UINT16_BE (data);
1070 data += 2;
1071
1072 seek_frames = GST_READ_UINT16_BE (data);
1073
1074 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
1075 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
1076 goto out_vbri;
1077 }
1078
1079 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
1080 GST_WARNING_OBJECT (mp3parse,
1081 "Not enough data to read VBRI seek table (need %d)",
1082 offset_vbri + 26 + nseek_points * seek_bytes);
1083 goto out_vbri;
1084 }
1085
1086 if (seek_frames * nseek_points < total_frames - seek_frames ||
1087 seek_frames * nseek_points > total_frames + seek_frames) {
1088 GST_WARNING_OBJECT (mp3parse,
1089 "VBRI seek table doesn't cover the complete file");
1090 goto out_vbri;
1091 }
1092
1093 data = map.data;
1094 data += offset_vbri + 26;
1095
1096 /* VBRI seek table: frame/seek_frames -> byte */
1097 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
1098 if (seek_bytes == 4)
1099 for (i = 0; i < nseek_points; i++) {
1100 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
1101 data += 4;
1102 } else if (seek_bytes == 3)
1103 for (i = 0; i < nseek_points; i++) {
1104 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
1105 data += 3;
1106 } else if (seek_bytes == 2)
1107 for (i = 0; i < nseek_points; i++) {
1108 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
1109 data += 2;
1110 } else /* seek_bytes == 1 */
1111 for (i = 0; i < nseek_points; i++) {
1112 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
1113 data += 1;
1114 }
1115 }
1116 out_vbri:
1117
1118 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
1119 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
1120 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
1121
1122 /* check for truncated file */
1123 if (upstream_total_bytes && mp3parse->vbri_bytes &&
1124 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
1125 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
1126 "invalidating VBRI header duration and size");
1127 mp3parse->vbri_valid = FALSE;
1128 } else {
1129 mp3parse->vbri_valid = TRUE;
1130 }
1131 } else {
1132 GST_DEBUG_OBJECT (mp3parse,
1133 "Xing, LAME or VBRI header not found in first frame");
1134 }
1135
1136 /* set duration if tables provided a valid one */
1137 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
1138 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
1139 mp3parse->xing_total_time, 0);
1140 }
1141 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
1142 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
1143 mp3parse->vbri_total_time, 0);
1144 }
1145
1146 /* tell baseclass how nicely we can seek, and a bitrate if one found */
1147 /* FIXME: fill index with seek table */
1148 #if 0
1149 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
1150 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
1151 mp3parse->xing_total_time)
1152 seekable = GST_BASE_PARSE_SEEK_TABLE;
1153
1154 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
1155 mp3parse->vbri_total_time)
1156 seekable = GST_BASE_PARSE_SEEK_TABLE;
1157 #endif
1158
1159 if (mp3parse->xing_bitrate)
1160 bitrate = mp3parse->xing_bitrate;
1161 else if (mp3parse->vbri_bitrate)
1162 bitrate = mp3parse->vbri_bitrate;
1163 else
1164 bitrate = 0;
1165
1166 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
1167
1168 cleanup:
1169 gst_buffer_unmap (buf, &map);
1170 }
1171
1172 static gboolean
gst_mpeg_audio_parse_time_to_bytepos(GstMpegAudioParse * mp3parse,GstClockTime ts,gint64 * bytepos)1173 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1174 GstClockTime ts, gint64 * bytepos)
1175 {
1176 gint64 total_bytes;
1177 GstClockTime total_time;
1178
1179 /* If XING seek table exists use this for time->byte conversion */
1180 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1181 (total_bytes = mp3parse->xing_bytes) &&
1182 (total_time = mp3parse->xing_total_time)) {
1183 gdouble fa, fb, fx;
1184 gdouble percent =
1185 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1186 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1187 gint index = CLAMP (percent, 0, 99);
1188
1189 fa = mp3parse->xing_seek_table[index];
1190 if (index < 99)
1191 fb = mp3parse->xing_seek_table[index + 1];
1192 else
1193 fb = 256.0;
1194
1195 fx = fa + (fb - fa) * (percent - index);
1196
1197 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1198
1199 return TRUE;
1200 }
1201
1202 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1203 (total_time = mp3parse->vbri_total_time)) {
1204 gint i, j;
1205 gdouble a, b, fa, fb;
1206
1207 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1208 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1209
1210 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1211 mp3parse->vbri_seek_points));
1212 fa = 0.0;
1213 for (j = i; j >= 0; j--)
1214 fa += mp3parse->vbri_seek_table[j];
1215
1216 if (i + 1 < mp3parse->vbri_seek_points) {
1217 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1218 mp3parse->vbri_seek_points));
1219 fb = fa + mp3parse->vbri_seek_table[i + 1];
1220 } else {
1221 b = gst_guint64_to_gdouble (total_time);
1222 fb = total_bytes;
1223 }
1224
1225 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1226
1227 return TRUE;
1228 }
1229
1230 return FALSE;
1231 }
1232
1233 static gboolean
gst_mpeg_audio_parse_bytepos_to_time(GstMpegAudioParse * mp3parse,gint64 bytepos,GstClockTime * ts)1234 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1235 gint64 bytepos, GstClockTime * ts)
1236 {
1237 gint64 total_bytes;
1238 GstClockTime total_time;
1239
1240 /* If XING seek table exists use this for byte->time conversion */
1241 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1242 (total_bytes = mp3parse->xing_bytes) &&
1243 (total_time = mp3parse->xing_total_time)) {
1244 gdouble fa, fb, fx;
1245 gdouble pos;
1246 gint index;
1247
1248 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1249 index = CLAMP (pos, 0, 255);
1250 fa = mp3parse->xing_seek_table_inverse[index];
1251 if (index < 255)
1252 fb = mp3parse->xing_seek_table_inverse[index + 1];
1253 else
1254 fb = 10000.0;
1255
1256 fx = fa + (fb - fa) * (pos - index);
1257
1258 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1259
1260 return TRUE;
1261 }
1262
1263 if (mp3parse->vbri_seek_table &&
1264 (total_bytes = mp3parse->vbri_bytes) &&
1265 (total_time = mp3parse->vbri_total_time)) {
1266 gint i = 0;
1267 guint64 sum = 0;
1268 gdouble a, b, fa, fb;
1269
1270 do {
1271 sum += mp3parse->vbri_seek_table[i];
1272 i++;
1273 } while (i + 1 < mp3parse->vbri_seek_points
1274 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1275 i--;
1276
1277 a = gst_guint64_to_gdouble (sum);
1278 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1279 mp3parse->vbri_seek_points));
1280
1281 if (i + 1 < mp3parse->vbri_seek_points) {
1282 b = a + mp3parse->vbri_seek_table[i + 1];
1283 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1284 mp3parse->vbri_seek_points));
1285 } else {
1286 b = total_bytes;
1287 fb = gst_guint64_to_gdouble (total_time);
1288 }
1289
1290 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1291
1292 return TRUE;
1293 }
1294
1295 return FALSE;
1296 }
1297
1298 static gboolean
gst_mpeg_audio_parse_convert(GstBaseParse * parse,GstFormat src_format,gint64 src_value,GstFormat dest_format,gint64 * dest_value)1299 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1300 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1301 {
1302 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1303 gboolean res = FALSE;
1304
1305 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1306 res =
1307 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1308 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1309 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1310 (GstClockTime *) dest_value);
1311
1312 /* if no tables, fall back to default estimated rate based conversion */
1313 if (!res)
1314 return gst_base_parse_convert_default (parse, src_format, src_value,
1315 dest_format, dest_value);
1316
1317 return res;
1318 }
1319
1320 static GstFlowReturn
gst_mpeg_audio_parse_pre_push_frame(GstBaseParse * parse,GstBaseParseFrame * frame)1321 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1322 GstBaseParseFrame * frame)
1323 {
1324 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1325 GstTagList *taglist = NULL;
1326
1327 /* we will create a taglist (if any of the parameters has changed)
1328 * to add the tags that changed */
1329 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1330 gboolean using_crc;
1331
1332 if (!taglist)
1333 taglist = gst_tag_list_new_empty ();
1334
1335 mp3parse->last_posted_crc = mp3parse->last_crc;
1336 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1337 using_crc = TRUE;
1338 } else {
1339 using_crc = FALSE;
1340 }
1341 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1342 using_crc, NULL);
1343 }
1344
1345 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1346 if (!taglist)
1347 taglist = gst_tag_list_new_empty ();
1348
1349 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1350
1351 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1352 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1353 }
1354
1355 /* tag sending done late enough in hook to ensure pending events
1356 * have already been sent */
1357 if (taglist != NULL || !mp3parse->sent_codec_tag) {
1358 GstCaps *caps;
1359
1360 if (taglist == NULL)
1361 taglist = gst_tag_list_new_empty ();
1362
1363 /* codec tag */
1364 caps = gst_pad_get_current_caps (GST_BASE_PARSE_SRC_PAD (parse));
1365 if (G_UNLIKELY (caps == NULL)) {
1366 gst_tag_list_unref (taglist);
1367
1368 if (GST_PAD_IS_FLUSHING (GST_BASE_PARSE_SRC_PAD (parse))) {
1369 GST_INFO_OBJECT (parse, "Src pad is flushing");
1370 return GST_FLOW_FLUSHING;
1371 } else {
1372 GST_INFO_OBJECT (parse, "Src pad is not negotiated!");
1373 return GST_FLOW_NOT_NEGOTIATED;
1374 }
1375 }
1376 gst_pb_utils_add_codec_description_to_tag_list (taglist,
1377 GST_TAG_AUDIO_CODEC, caps);
1378 gst_caps_unref (caps);
1379
1380 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1381 mp3parse->vbri_bitrate == 0) {
1382 /* We don't have a VBR bitrate, so post the available bitrate as
1383 * nominal and let baseparse calculate the real bitrate */
1384 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1385 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1386 }
1387
1388 /* also signals the end of first-frame processing */
1389 mp3parse->sent_codec_tag = TRUE;
1390 }
1391
1392 /* if the taglist exists, we need to update it so it gets sent out */
1393 if (taglist) {
1394 gst_base_parse_merge_tags (parse, taglist, GST_TAG_MERGE_REPLACE);
1395 gst_tag_list_unref (taglist);
1396 }
1397
1398 /* usual clipping applies */
1399 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;
1400
1401 return GST_FLOW_OK;
1402 }
1403
1404 static void
remove_fields(GstCaps * caps)1405 remove_fields (GstCaps * caps)
1406 {
1407 guint i, n;
1408
1409 n = gst_caps_get_size (caps);
1410 for (i = 0; i < n; i++) {
1411 GstStructure *s = gst_caps_get_structure (caps, i);
1412
1413 gst_structure_remove_field (s, "parsed");
1414 }
1415 }
1416
1417 static GstCaps *
gst_mpeg_audio_parse_get_sink_caps(GstBaseParse * parse,GstCaps * filter)1418 gst_mpeg_audio_parse_get_sink_caps (GstBaseParse * parse, GstCaps * filter)
1419 {
1420 GstCaps *peercaps, *templ;
1421 GstCaps *res;
1422
1423 templ = gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD (parse));
1424 if (filter) {
1425 GstCaps *fcopy = gst_caps_copy (filter);
1426 /* Remove the fields we convert */
1427 remove_fields (fcopy);
1428 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), fcopy);
1429 gst_caps_unref (fcopy);
1430 } else
1431 peercaps = gst_pad_peer_query_caps (GST_BASE_PARSE_SRC_PAD (parse), NULL);
1432
1433 if (peercaps) {
1434 /* Remove the parsed field */
1435 peercaps = gst_caps_make_writable (peercaps);
1436 remove_fields (peercaps);
1437
1438 res = gst_caps_intersect_full (peercaps, templ, GST_CAPS_INTERSECT_FIRST);
1439 gst_caps_unref (peercaps);
1440 gst_caps_unref (templ);
1441 } else {
1442 res = templ;
1443 }
1444
1445 if (filter) {
1446 GstCaps *intersection;
1447
1448 intersection =
1449 gst_caps_intersect_full (filter, res, GST_CAPS_INTERSECT_FIRST);
1450 gst_caps_unref (res);
1451 res = intersection;
1452 }
1453
1454 return res;
1455 }
1456