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>
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.
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.
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., 59 Temple Place - Suite 330,
20
* Boston, MA 02111-1307, USA.
23
* SECTION:element-mpegaudioparse
24
* @short_description: MPEG audio parser
25
* @see_also: #GstAmrParse, #GstAACParse
27
* Parses and frames mpeg1 audio streams. Provides seeking.
30
* <title>Example launch line</title>
32
* gst-launch filesrc location=test.mp3 ! mpegaudioparse ! mad ! autoaudiosink
43
#include "gstmpegaudioparse.h"
44
#include <gst/base/gstbytereader.h>
46
GST_DEBUG_CATEGORY_STATIC (mpeg_audio_parse_debug);
47
#define GST_CAT_DEFAULT mpeg_audio_parse_debug
49
#define MPEG_AUDIO_CHANNEL_MODE_UNKNOWN -1
50
#define MPEG_AUDIO_CHANNEL_MODE_STEREO 0
51
#define MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO 1
52
#define MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL 2
53
#define MPEG_AUDIO_CHANNEL_MODE_MONO 3
55
#define CRC_UNKNOWN -1
56
#define CRC_PROTECTED 0
57
#define CRC_NOT_PROTECTED 1
59
#define XING_FRAMES_FLAG 0x0001
60
#define XING_BYTES_FLAG 0x0002
61
#define XING_TOC_FLAG 0x0004
62
#define XING_VBR_SCALE_FLAG 0x0008
64
static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
67
GST_STATIC_CAPS ("audio/mpeg, "
68
"mpegversion = (int) 1, "
69
"layer = (int) [ 1, 3 ], "
70
"rate = (int) [ 8000, 48000 ], channels = (int) [ 1, 2 ],"
71
"parsed=(boolean) true")
74
static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
77
GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) 1, parsed=(boolean)false")
80
static void gst_mpeg_audio_parse_finalize (GObject * object);
82
static gboolean gst_mpeg_audio_parse_start (GstBaseParse * parse);
83
static gboolean gst_mpeg_audio_parse_stop (GstBaseParse * parse);
84
static gboolean gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
85
GstBuffer * buffer, guint * size, gint * skipsize);
86
static GstFlowReturn gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
88
static GstFlowReturn gst_mpeg_audio_parse_pre_push_buffer (GstBaseParse * parse,
90
static gboolean gst_mpeg_audio_parse_convert (GstBaseParse * parse,
91
GstFormat src_format, gint64 src_value,
92
GstFormat dest_format, gint64 * dest_value);
94
GST_BOILERPLATE (GstMpegAudioParse, gst_mpeg_audio_parse, GstBaseParse,
97
#define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
98
(gst_mpeg_audio_channel_mode_get_type())
100
static const GEnumValue mpeg_audio_channel_mode[] = {
101
{MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
102
{MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
103
{MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
104
{MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
105
{MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
110
gst_mpeg_audio_channel_mode_get_type (void)
112
static GType mpeg_audio_channel_mode_type = 0;
114
if (!mpeg_audio_channel_mode_type) {
115
mpeg_audio_channel_mode_type =
116
g_enum_register_static ("GstMpegAudioChannelMode",
117
mpeg_audio_channel_mode);
119
return mpeg_audio_channel_mode_type;
123
gst_mpeg_audio_channel_mode_get_nick (gint mode)
126
for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
127
if (mpeg_audio_channel_mode[i].value == mode)
128
return mpeg_audio_channel_mode[i].value_nick;
134
gst_mpeg_audio_parse_base_init (gpointer klass)
136
GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
138
gst_element_class_add_pad_template (element_class,
139
gst_static_pad_template_get (&sink_template));
140
gst_element_class_add_pad_template (element_class,
141
gst_static_pad_template_get (&src_template));
143
gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
144
"Codec/Parser/Audio",
145
"Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
146
"Jan Schmidt <thaytan@mad.scientist.com>,"
147
"Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
151
gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
153
GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
154
GObjectClass *object_class = G_OBJECT_CLASS (klass);
156
GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
157
"MPEG1 audio stream parser");
159
object_class->finalize = gst_mpeg_audio_parse_finalize;
161
parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
162
parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
163
parse_class->check_valid_frame =
164
GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
165
parse_class->parse_frame =
166
GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
167
parse_class->pre_push_buffer =
168
GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_buffer);
169
parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
172
#define GST_TAG_CRC "has-crc"
173
#define GST_TAG_MODE "channel-mode"
175
gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
176
"has crc", "Using CRC", NULL);
177
gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
178
"channel mode", "MPEG audio channel mode", NULL);
180
g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
184
gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
186
mp3parse->channels = -1;
188
mp3parse->sent_codec_tag = FALSE;
189
mp3parse->last_posted_crc = CRC_UNKNOWN;
190
mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
192
mp3parse->xing_flags = 0;
193
mp3parse->xing_bitrate = 0;
194
mp3parse->xing_frames = 0;
195
mp3parse->xing_total_time = 0;
196
mp3parse->xing_bytes = 0;
197
mp3parse->xing_vbr_scale = 0;
198
memset (mp3parse->xing_seek_table, 0, 100);
199
memset (mp3parse->xing_seek_table_inverse, 0, 256);
201
mp3parse->vbri_bitrate = 0;
202
mp3parse->vbri_frames = 0;
203
mp3parse->vbri_total_time = 0;
204
mp3parse->vbri_bytes = 0;
205
mp3parse->vbri_seek_points = 0;
206
g_free (mp3parse->vbri_seek_table);
207
mp3parse->vbri_seek_table = NULL;
211
gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse,
212
GstMpegAudioParseClass * klass)
214
gst_mpeg_audio_parse_reset (mp3parse);
218
gst_mpeg_audio_parse_finalize (GObject * object)
220
G_OBJECT_CLASS (parent_class)->finalize (object);
224
gst_mpeg_audio_parse_start (GstBaseParse * parse)
226
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
228
gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), 1024);
229
GST_DEBUG_OBJECT (parse, "starting");
231
gst_mpeg_audio_parse_reset (mp3parse);
237
gst_mpeg_audio_parse_stop (GstBaseParse * parse)
239
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
241
GST_DEBUG_OBJECT (parse, "stopping");
243
gst_mpeg_audio_parse_reset (mp3parse);
248
static const guint mp3types_bitrates[2][3][16] = {
250
{0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
251
{0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
252
{0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
255
{0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
256
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
257
{0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
261
static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
262
{22050, 24000, 16000},
267
mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
268
guint * put_version, guint * put_layer, guint * put_channels,
269
guint * put_bitrate, guint * put_samplerate, guint * put_mode,
273
gulong mode, samplerate, bitrate, layer, channels, padding, crc;
277
if (header & (1 << 20)) {
278
lsf = (header & (1 << 19)) ? 0 : 1;
285
version = 1 + lsf + mpg25;
287
layer = 4 - ((header >> 17) & 0x3);
289
crc = (header >> 16) & 0x1;
291
bitrate = (header >> 12) & 0xF;
292
bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
293
/* The caller has ensured we have a valid header, so bitrate can't be
295
g_assert (bitrate != 0);
297
samplerate = (header >> 10) & 0x3;
298
samplerate = mp3types_freqs[lsf + mpg25][samplerate];
300
padding = (header >> 9) & 0x1;
302
mode = (header >> 6) & 0x3;
303
channels = (mode == 3) ? 1 : 2;
307
length = 4 * ((bitrate * 12) / samplerate + padding);
310
length = (bitrate * 144) / samplerate + padding;
314
length = (bitrate * 144) / (samplerate << lsf) + padding;
318
GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
320
GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
321
"layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
322
layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
325
*put_version = version;
329
*put_channels = channels;
331
*put_bitrate = bitrate;
333
*put_samplerate = samplerate;
342
/* Minimum number of consecutive, valid-looking frames to consider
344
#define MIN_RESYNC_FRAMES 3
346
/* Perform extended validation to check that subsequent headers match
347
* the first header given here in important characteristics, to avoid
348
* false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
349
* frames to match their major characteristics.
351
* If at_eos is set to TRUE, we just check that we don't find any invalid
352
* frames in whatever data is available, rather than requiring a full
353
* MIN_RESYNC_FRAMES of data.
355
* Returns TRUE if we've seen enough data to validate or reject the frame.
356
* If TRUE is returned, then *valid contains TRUE if it validated, or false
357
* if we decided it was false sync.
358
* If FALSE is returned, then *valid contains minimum needed data.
361
gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
362
guint32 header, int bpf, gboolean at_eos, gint * valid)
367
int frames_found = 1;
370
available = GST_BUFFER_SIZE (buf);
371
data = GST_BUFFER_DATA (buf);
373
while (frames_found < MIN_RESYNC_FRAMES) {
374
/* Check if we have enough data for all these frames, plus the next
376
if (available < offset + 4) {
378
/* Running out of data at EOS is fine; just accept it */
387
next_header = GST_READ_UINT32_BE (data + offset);
388
GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
389
offset, (unsigned int) header, (unsigned int) next_header, bpf);
391
/* mask the bits which are allowed to differ between frames */
392
#define HDRMASK ~((0xF << 12) /* bitrate */ | \
393
(0x1 << 9) /* padding */ | \
394
(0xf << 4) /* mode|mode extension */ | \
395
(0xf)) /* copyright|emphasis */
397
if ((next_header & HDRMASK) != (header & HDRMASK)) {
398
/* If any of the unmasked bits don't match, then it's not valid */
399
GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
400
"(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
401
(guint) header, (guint) header & HDRMASK, (guint) next_header,
402
(guint) next_header & HDRMASK, bpf);
405
} else if ((((next_header >> 12) & 0xf) == 0) ||
406
(((next_header >> 12) & 0xf) == 0xf)) {
407
/* The essential parts were the same, but the bitrate held an
408
invalid value - also reject */
409
GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
414
bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
415
NULL, NULL, NULL, NULL, NULL, NULL, NULL);
426
gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
429
GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
430
/* if it's not a valid sync */
431
if ((head & 0xffe00000) != 0xffe00000) {
432
GST_WARNING_OBJECT (mp3parse, "invalid sync");
435
/* if it's an invalid MPEG version */
436
if (((head >> 19) & 3) == 0x1) {
437
GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
441
/* if it's an invalid layer */
442
if (!((head >> 17) & 3)) {
443
GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
446
/* if it's an invalid bitrate */
447
if (((head >> 12) & 0xf) == 0x0) {
448
GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx."
449
"Free format files are not supported yet", (head >> 12) & 0xf);
452
if (((head >> 12) & 0xf) == 0xf) {
453
GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
456
/* if it's an invalid samplerate */
457
if (((head >> 10) & 0x3) == 0x3) {
458
GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
463
if ((head & 0x3) == 0x2) {
464
/* Ignore this as there are some files with emphasis 0x2 that can
465
* be played fine. See BGO #537235 */
466
GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
473
gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse, GstBuffer * buf,
474
guint * framesize, gint * skipsize)
476
GstByteReader reader = GST_BYTE_READER_INIT_FROM_BUFFER (buf);
477
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
479
gboolean sync, drain, valid, caps_change;
481
guint bitrate, layer, rate, channels, version, mode, crc;
483
if (G_UNLIKELY (GST_BUFFER_SIZE (buf) < 6))
486
off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
487
0, GST_BUFFER_SIZE (buf));
489
GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
491
/* didn't find anything that looks like a sync word, skip */
493
*skipsize = GST_BUFFER_SIZE (buf) - 3;
497
/* possible frame header, but not at offset 0? skip bytes before sync */
503
/* make sure the values in the frame header look sane */
504
header = GST_READ_UINT32_BE (GST_BUFFER_DATA (buf));
505
if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
510
GST_LOG_OBJECT (parse, "got frame");
512
bpf = mp3_type_frame_length_from_header (mp3parse, header,
513
&version, &layer, &channels, &bitrate, &rate, &mode, &crc);
516
if (channels != mp3parse->channels || rate != mp3parse->rate ||
517
layer != mp3parse->layer || version != mp3parse->version)
522
sync = gst_base_parse_get_sync (parse);
523
drain = gst_base_parse_get_drain (parse);
525
if (!drain && (!sync || caps_change)) {
526
if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, drain,
528
/* not enough data */
529
gst_base_parse_set_min_frame_size (parse, valid);
545
gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
548
const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
549
const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
550
const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
553
gint64 upstream_total_bytes = 0;
554
GstFormat fmt = GST_FORMAT_BYTES;
557
GstBaseParseSeekable seekable;
560
if (mp3parse->sent_codec_tag)
563
/* Check first frame for Xing info */
564
if (mp3parse->version == 1) { /* MPEG-1 file */
565
if (mp3parse->channels == 1)
569
} else { /* MPEG-2 header */
570
if (mp3parse->channels == 1)
575
/* Skip the 4 bytes of the MP3 header too */
578
/* Check if we have enough data to read the Xing header */
579
avail = GST_BUFFER_SIZE (buf);
580
data = GST_BUFFER_DATA (buf);
581
if (avail < offset + 8)
584
/* The header starts at the provided offset */
587
/* obtain real upstream total bytes */
588
fmt = GST_FORMAT_BYTES;
589
if (!gst_pad_query_peer_duration (GST_BASE_PARSE_SINK_PAD (GST_BASE_PARSE
590
(mp3parse)), &fmt, &upstream_total_bytes))
591
upstream_total_bytes = 0;
593
read_id = GST_READ_UINT32_BE (data);
594
if (read_id == xing_id || read_id == info_id) {
596
guint bytes_needed = offset + 8;
598
GstClockTime total_time;
600
GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
602
/* Read 4 base bytes of flags, big-endian */
603
xing_flags = GST_READ_UINT32_BE (data + 4);
604
if (xing_flags & XING_FRAMES_FLAG)
606
if (xing_flags & XING_BYTES_FLAG)
608
if (xing_flags & XING_TOC_FLAG)
610
if (xing_flags & XING_VBR_SCALE_FLAG)
612
if (avail < bytes_needed) {
613
GST_DEBUG_OBJECT (mp3parse,
614
"Not enough data to read Xing header (need %d)", bytes_needed);
618
GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
619
mp3parse->xing_flags = xing_flags;
621
data = GST_BUFFER_DATA (buf);
624
if (xing_flags & XING_FRAMES_FLAG) {
625
mp3parse->xing_frames = GST_READ_UINT32_BE (data);
626
if (mp3parse->xing_frames == 0) {
627
GST_WARNING_OBJECT (mp3parse,
628
"Invalid number of frames in Xing header");
629
mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
631
mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
632
(guint64) (mp3parse->xing_frames) * (mp3parse->spf),
638
mp3parse->xing_frames = 0;
639
mp3parse->xing_total_time = 0;
642
if (xing_flags & XING_BYTES_FLAG) {
643
mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
644
if (mp3parse->xing_bytes == 0) {
645
GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
646
mp3parse->xing_flags &= ~XING_BYTES_FLAG;
650
mp3parse->xing_bytes = 0;
653
/* If we know the upstream size and duration, compute the
654
* total bitrate, rounded up to the nearest kbit/sec */
655
if ((total_time = mp3parse->xing_total_time) &&
656
(total_bytes = mp3parse->xing_bytes)) {
657
mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
658
8 * GST_SECOND, total_time);
659
mp3parse->xing_bitrate += 500;
660
mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
663
if (xing_flags & XING_TOC_FLAG) {
665
guchar *table = mp3parse->xing_seek_table;
670
GST_DEBUG_OBJECT (mp3parse,
671
"Subtracting initial offset of %d bytes from Xing TOC", first);
673
/* xing seek table: percent time -> 1/256 bytepos */
674
for (i = 0; i < 100; i++) {
675
new = data[i] - first;
677
GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
678
mp3parse->xing_flags &= ~XING_TOC_FLAG;
681
mp3parse->xing_seek_table[i] = old = new;
684
/* build inverse table: 1/256 bytepos -> 1/100 percent time */
685
for (i = 0; i < 256; i++) {
686
while (percent < 99 && table[percent + 1] <= i)
689
if (table[percent] == i) {
690
mp3parse->xing_seek_table_inverse[i] = percent * 100;
691
} else if (table[percent] < i && percent < 99) {
693
gint a = percent, b = percent + 1;
697
fx = (b - a) / (fb - fa) * (i - fa) + a;
698
mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
699
} else if (percent == 99) {
701
gint a = percent, b = 100;
705
fx = (b - a) / (fb - fa) * (i - fa) + a;
706
mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
712
memset (mp3parse->xing_seek_table, 0, 100);
713
memset (mp3parse->xing_seek_table_inverse, 0, 256);
716
if (xing_flags & XING_VBR_SCALE_FLAG) {
717
mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
719
mp3parse->xing_vbr_scale = 0;
721
GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
722
GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
723
GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
724
mp3parse->xing_vbr_scale);
726
/* check for truncated file */
727
if (upstream_total_bytes && mp3parse->xing_bytes &&
728
mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
729
GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
730
"invalidating Xing header duration and size");
731
mp3parse->xing_flags &= ~XING_BYTES_FLAG;
732
mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
734
} else if (read_id == vbri_id) {
735
gint64 total_bytes, total_frames;
736
GstClockTime total_time;
737
guint16 nseek_points;
739
GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
740
if (avail < offset + 26) {
741
GST_DEBUG_OBJECT (mp3parse,
742
"Not enough data to read VBRI header (need %d)", offset + 26);
746
GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
747
data = GST_BUFFER_DATA (buf);
750
if (GST_READ_UINT16_BE (data) != 0x0001) {
751
GST_WARNING_OBJECT (mp3parse,
752
"Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
757
/* Skip encoder delay */
763
total_bytes = GST_READ_UINT32_BE (data);
764
if (total_bytes != 0)
765
mp3parse->vbri_bytes = total_bytes;
768
total_frames = GST_READ_UINT32_BE (data);
769
if (total_frames != 0) {
770
mp3parse->vbri_frames = total_frames;
771
mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
772
(guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
776
/* If we know the upstream size and duration, compute the
777
* total bitrate, rounded up to the nearest kbit/sec */
778
if ((total_time = mp3parse->vbri_total_time) &&
779
(total_bytes = mp3parse->vbri_bytes)) {
780
mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
781
8 * GST_SECOND, total_time);
782
mp3parse->vbri_bitrate += 500;
783
mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
786
nseek_points = GST_READ_UINT16_BE (data);
789
if (nseek_points > 0) {
790
guint scale, seek_bytes, seek_frames;
793
mp3parse->vbri_seek_points = nseek_points;
795
scale = GST_READ_UINT16_BE (data);
798
seek_bytes = GST_READ_UINT16_BE (data);
801
seek_frames = GST_READ_UINT16_BE (data);
803
if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
804
GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
808
if (avail < offset + 26 + nseek_points * seek_bytes) {
809
GST_WARNING_OBJECT (mp3parse,
810
"Not enough data to read VBRI seek table (need %d)",
811
offset + 26 + nseek_points * seek_bytes);
815
if (seek_frames * nseek_points < total_frames - seek_frames ||
816
seek_frames * nseek_points > total_frames + seek_frames) {
817
GST_WARNING_OBJECT (mp3parse,
818
"VBRI seek table doesn't cover the complete file");
822
if (avail < offset + 26) {
823
GST_DEBUG_OBJECT (mp3parse,
824
"Not enough data to read VBRI header (need %d)",
825
offset + 26 + nseek_points * seek_bytes);
829
data = GST_BUFFER_DATA (buf);
832
/* VBRI seek table: frame/seek_frames -> byte */
833
mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
835
for (i = 0; i < nseek_points; i++) {
836
mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
838
} else if (seek_bytes == 3)
839
for (i = 0; i < nseek_points; i++) {
840
mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
842
} else if (seek_bytes == 2)
843
for (i = 0; i < nseek_points; i++) {
844
mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
846
} else /* seek_bytes == 1 */
847
for (i = 0; i < nseek_points; i++) {
848
mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
854
GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
855
GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
856
GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
858
/* check for truncated file */
859
if (upstream_total_bytes && mp3parse->vbri_bytes &&
860
mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
861
GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
862
"invalidating VBRI header duration and size");
863
mp3parse->vbri_valid = FALSE;
865
mp3parse->vbri_valid = TRUE;
868
GST_DEBUG_OBJECT (mp3parse,
869
"Xing, LAME or VBRI header not found in first frame");
872
/* set duration if tables provided a valid one */
873
if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
874
gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
875
mp3parse->xing_total_time, 0);
877
if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
878
gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
879
mp3parse->vbri_total_time, 0);
882
/* tell baseclass how nicely we can seek, and a bitrate if one found */
883
seekable = GST_BASE_PARSE_SEEK_DEFAULT;
884
if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
885
mp3parse->xing_total_time)
886
seekable = GST_BASE_PARSE_SEEK_TABLE;
888
if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
889
mp3parse->vbri_total_time)
890
seekable = GST_BASE_PARSE_SEEK_TABLE;
892
if (mp3parse->xing_bitrate)
893
bitrate = mp3parse->xing_bitrate;
894
else if (mp3parse->vbri_bitrate)
895
bitrate = mp3parse->vbri_bitrate;
899
gst_base_parse_set_seek (GST_BASE_PARSE (mp3parse), seekable, bitrate);
903
gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse, GstBuffer * buf)
905
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
906
guint bitrate, layer, rate, channels, version, mode, crc;
908
g_return_val_if_fail (GST_BUFFER_SIZE (buf) >= 4, GST_FLOW_ERROR);
910
if (!mp3_type_frame_length_from_header (mp3parse,
911
GST_READ_UINT32_BE (GST_BUFFER_DATA (buf)),
912
&version, &layer, &channels, &bitrate, &rate, &mode, &crc))
915
if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
916
layer != mp3parse->layer || version != mp3parse->version)) {
917
GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
918
"mpegversion", G_TYPE_INT, 1,
919
"mpegaudioversion", G_TYPE_INT, version,
920
"layer", G_TYPE_INT, layer,
921
"rate", G_TYPE_INT, rate,
922
"channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
923
gst_buffer_set_caps (buf, caps);
924
gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
925
gst_caps_unref (caps);
927
mp3parse->rate = rate;
928
mp3parse->channels = channels;
929
mp3parse->layer = layer;
930
mp3parse->version = version;
932
/* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
933
if (mp3parse->layer == 1)
935
else if (mp3parse->layer == 2)
936
mp3parse->spf = 1152;
937
else if (mp3parse->version == 1) {
938
mp3parse->spf = 1152;
940
/* MPEG-2 or "2.5" */
945
* We start pushing 9 frames earlier (29 frames for MPEG2) than
946
* segment start to be able to decode the first frame we want.
947
* 9 (29) frames are the theoretical maximum of frames that contain
948
* data for the current frame (bit reservoir).
951
* Some mp3 streams have an offset in the timestamps, for which we have to
952
* push the frame *after* the end position in order for the decoder to be
953
* able to decode everything up until the segment.stop position. */
954
gst_base_parse_set_frame_props (parse, mp3parse->rate, mp3parse->spf,
955
(version == 1) ? 10 : 30, 2);
958
/* For first frame; check for seek tables and output a codec tag */
959
gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
961
/* store some frame info for later processing */
962
mp3parse->last_crc = crc;
963
mp3parse->last_mode = mode;
970
/* this really shouldn't ever happen */
971
GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
972
return GST_FLOW_ERROR;
977
gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
978
GstClockTime ts, gint64 * bytepos)
981
GstClockTime total_time;
983
/* If XING seek table exists use this for time->byte conversion */
984
if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
985
(total_bytes = mp3parse->xing_bytes) &&
986
(total_time = mp3parse->xing_total_time)) {
989
CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
990
gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
991
gint index = CLAMP (percent, 0, 99);
993
fa = mp3parse->xing_seek_table[index];
995
fb = mp3parse->xing_seek_table[index + 1];
999
fx = fa + (fb - fa) * (percent - index);
1001
*bytepos = (1.0 / 256.0) * fx * total_bytes;
1006
if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1007
(total_time = mp3parse->vbri_total_time)) {
1009
gdouble a, b, fa, fb;
1011
i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1012
i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1014
a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1015
mp3parse->vbri_seek_points));
1017
for (j = i; j >= 0; j--)
1018
fa += mp3parse->vbri_seek_table[j];
1020
if (i + 1 < mp3parse->vbri_seek_points) {
1021
b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1022
mp3parse->vbri_seek_points));
1023
fb = fa + mp3parse->vbri_seek_table[i + 1];
1025
b = gst_guint64_to_gdouble (total_time);
1029
*bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1038
gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1039
gint64 bytepos, GstClockTime * ts)
1042
GstClockTime total_time;
1044
/* If XING seek table exists use this for byte->time conversion */
1045
if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1046
(total_bytes = mp3parse->xing_bytes) &&
1047
(total_time = mp3parse->xing_total_time)) {
1052
pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1053
index = CLAMP (pos, 0, 255);
1054
fa = mp3parse->xing_seek_table_inverse[index];
1056
fb = mp3parse->xing_seek_table_inverse[index + 1];
1060
fx = fa + (fb - fa) * (pos - index);
1062
*ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1067
if (mp3parse->vbri_seek_table &&
1068
(total_bytes = mp3parse->vbri_bytes) &&
1069
(total_time = mp3parse->vbri_total_time)) {
1072
gdouble a, b, fa, fb;
1075
sum += mp3parse->vbri_seek_table[i];
1077
} while (i + 1 < mp3parse->vbri_seek_points
1078
&& sum + mp3parse->vbri_seek_table[i] < bytepos);
1081
a = gst_guint64_to_gdouble (sum);
1082
fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1083
mp3parse->vbri_seek_points));
1085
if (i + 1 < mp3parse->vbri_seek_points) {
1086
b = a + mp3parse->vbri_seek_table[i + 1];
1087
fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1088
mp3parse->vbri_seek_points));
1091
fb = gst_guint64_to_gdouble (total_time);
1094
*ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1103
gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1104
gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1106
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1107
gboolean res = FALSE;
1109
if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1111
gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1112
else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1113
res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1114
(GstClockTime *) dest_value);
1116
/* if no tables, fall back to default estimated rate based conversion */
1118
return gst_base_parse_convert_default (parse, src_format, src_value,
1119
dest_format, dest_value);
1124
static GstFlowReturn
1125
gst_mpeg_audio_parse_pre_push_buffer (GstBaseParse * parse, GstBuffer * buf)
1127
GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1128
GstTagList *taglist;
1130
/* tag sending done late enough in hook to ensure pending events
1131
* have already been sent */
1133
if (!mp3parse->sent_codec_tag) {
1137
if (mp3parse->layer == 3) {
1138
codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1139
mp3parse->version, mp3parse->layer);
1141
codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1142
mp3parse->version, mp3parse->layer);
1144
taglist = gst_tag_list_new ();
1145
gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1146
GST_TAG_AUDIO_CODEC, codec, NULL);
1147
gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1148
GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1151
/* also signals the end of first-frame processing */
1152
mp3parse->sent_codec_tag = TRUE;
1155
/* we will create a taglist (if any of the parameters has changed)
1156
* to add the tags that changed */
1158
if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1162
taglist = gst_tag_list_new ();
1164
mp3parse->last_posted_crc = mp3parse->last_crc;
1165
if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1170
gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1174
if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1176
taglist = gst_tag_list_new ();
1178
mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1180
gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1181
gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1184
/* if the taglist exists, we need to send it */
1186
gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1187
GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1190
/* usual clipping applies */
1191
return GST_BASE_PARSE_FLOW_CLIP;
added
removed
gst/audioparsers/gstmpegaudioparse.c
