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
37 /* FIXME: we should make the base class (GstBaseParse) aware of the
38 * XING seek table somehow, so it can use it properly for things like
39 * accurate seeks. Currently it can only do a lookup via the convert function,
40 * but then doesn't know what the result represents exactly. One could either
41 * add a vfunc for index lookup, or just make mpegaudioparse populate the
42 * base class's index via the API provided.
50 #include "gstmpegaudioparse.h"
51 #include <gst/base/gstbytereader.h>
53 GST_DEBUG_CATEGORY_STATIC (mpeg_audio_parse_debug);
54 #define GST_CAT_DEFAULT mpeg_audio_parse_debug
56 #define MPEG_AUDIO_CHANNEL_MODE_UNKNOWN -1
57 #define MPEG_AUDIO_CHANNEL_MODE_STEREO 0
58 #define MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO 1
59 #define MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL 2
60 #define MPEG_AUDIO_CHANNEL_MODE_MONO 3
62 #define CRC_UNKNOWN -1
63 #define CRC_PROTECTED 0
64 #define CRC_NOT_PROTECTED 1
66 #define XING_FRAMES_FLAG 0x0001
67 #define XING_BYTES_FLAG 0x0002
68 #define XING_TOC_FLAG 0x0004
69 #define XING_VBR_SCALE_FLAG 0x0008
71 #define MIN_FRAME_SIZE 6
73 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
76 GST_STATIC_CAPS ("audio/mpeg, "
77 "mpegversion = (int) 1, "
78 "layer = (int) [ 1, 3 ], "
79 "rate = (int) [ 8000, 48000 ], channels = (int) [ 1, 2 ],"
80 "parsed=(boolean) true")
83 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
86 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) 1, parsed=(boolean)false")
89 static void gst_mpeg_audio_parse_finalize (GObject * object);
91 static gboolean gst_mpeg_audio_parse_start (GstBaseParse * parse);
92 static gboolean gst_mpeg_audio_parse_stop (GstBaseParse * parse);
93 static gboolean gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
94 GstBaseParseFrame * frame, guint * size, gint * skipsize);
95 static GstFlowReturn gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
96 GstBaseParseFrame * frame);
97 static GstFlowReturn gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
98 GstBaseParseFrame * frame);
99 static gboolean gst_mpeg_audio_parse_convert (GstBaseParse * parse,
100 GstFormat src_format, gint64 src_value,
101 GstFormat dest_format, gint64 * dest_value);
103 #define gst_mpeg_audio_parse_parent_class parent_class
104 G_DEFINE_TYPE (GstMpegAudioParse, gst_mpeg_audio_parse, GST_TYPE_BASE_PARSE);
106 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
107 (gst_mpeg_audio_channel_mode_get_type())
109 static const GEnumValue mpeg_audio_channel_mode[] = {
110 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
111 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
112 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
113 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
114 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
119 gst_mpeg_audio_channel_mode_get_type (void)
121 static GType mpeg_audio_channel_mode_type = 0;
123 if (!mpeg_audio_channel_mode_type) {
124 mpeg_audio_channel_mode_type =
125 g_enum_register_static ("GstMpegAudioChannelMode",
126 mpeg_audio_channel_mode);
128 return mpeg_audio_channel_mode_type;
132 gst_mpeg_audio_channel_mode_get_nick (gint mode)
135 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
136 if (mpeg_audio_channel_mode[i].value == mode)
137 return mpeg_audio_channel_mode[i].value_nick;
143 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
145 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
146 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
147 GObjectClass *object_class = G_OBJECT_CLASS (klass);
149 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
150 "MPEG1 audio stream parser");
152 object_class->finalize = gst_mpeg_audio_parse_finalize;
154 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
155 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
156 parse_class->check_valid_frame =
157 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
158 parse_class->parse_frame =
159 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
160 parse_class->pre_push_frame =
161 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
162 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
165 #define GST_TAG_CRC "has-crc"
166 #define GST_TAG_MODE "channel-mode"
168 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
169 "has crc", "Using CRC", NULL);
170 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
171 "channel mode", "MPEG audio channel mode", NULL);
173 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
175 gst_element_class_add_pad_template (element_class,
176 gst_static_pad_template_get (&sink_template));
177 gst_element_class_add_pad_template (element_class,
178 gst_static_pad_template_get (&src_template));
180 gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
181 "Codec/Parser/Audio",
182 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
183 "Jan Schmidt <thaytan@mad.scientist.com>,"
184 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
188 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
190 mp3parse->channels = -1;
192 mp3parse->sent_codec_tag = FALSE;
193 mp3parse->last_posted_crc = CRC_UNKNOWN;
194 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
196 mp3parse->hdr_bitrate = 0;
198 mp3parse->xing_flags = 0;
199 mp3parse->xing_bitrate = 0;
200 mp3parse->xing_frames = 0;
201 mp3parse->xing_total_time = 0;
202 mp3parse->xing_bytes = 0;
203 mp3parse->xing_vbr_scale = 0;
204 memset (mp3parse->xing_seek_table, 0, 100);
205 memset (mp3parse->xing_seek_table_inverse, 0, 256);
207 mp3parse->vbri_bitrate = 0;
208 mp3parse->vbri_frames = 0;
209 mp3parse->vbri_total_time = 0;
210 mp3parse->vbri_bytes = 0;
211 mp3parse->vbri_seek_points = 0;
212 g_free (mp3parse->vbri_seek_table);
213 mp3parse->vbri_seek_table = NULL;
215 mp3parse->encoder_delay = 0;
216 mp3parse->encoder_padding = 0;
220 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse)
222 gst_mpeg_audio_parse_reset (mp3parse);
226 gst_mpeg_audio_parse_finalize (GObject * object)
228 G_OBJECT_CLASS (parent_class)->finalize (object);
232 gst_mpeg_audio_parse_start (GstBaseParse * parse)
234 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
236 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), MIN_FRAME_SIZE);
237 GST_DEBUG_OBJECT (parse, "starting");
239 gst_mpeg_audio_parse_reset (mp3parse);
245 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
247 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
249 GST_DEBUG_OBJECT (parse, "stopping");
251 gst_mpeg_audio_parse_reset (mp3parse);
256 static const guint mp3types_bitrates[2][3][16] = {
258 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
259 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
260 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
263 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
264 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
265 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
269 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
270 {22050, 24000, 16000},
275 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
276 guint * put_version, guint * put_layer, guint * put_channels,
277 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
281 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
285 if (header & (1 << 20)) {
286 lsf = (header & (1 << 19)) ? 0 : 1;
293 version = 1 + lsf + mpg25;
295 layer = 4 - ((header >> 17) & 0x3);
297 crc = (header >> 16) & 0x1;
299 bitrate = (header >> 12) & 0xF;
300 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
301 /* The caller has ensured we have a valid header, so bitrate can't be
303 g_assert (bitrate != 0);
305 samplerate = (header >> 10) & 0x3;
306 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
308 padding = (header >> 9) & 0x1;
310 mode = (header >> 6) & 0x3;
311 channels = (mode == 3) ? 1 : 2;
315 length = 4 * ((bitrate * 12) / samplerate + padding);
318 length = (bitrate * 144) / samplerate + padding;
322 length = (bitrate * 144) / (samplerate << lsf) + padding;
326 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
328 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
329 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
330 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
333 *put_version = version;
337 *put_channels = channels;
339 *put_bitrate = bitrate;
341 *put_samplerate = samplerate;
350 /* Minimum number of consecutive, valid-looking frames to consider
352 #define MIN_RESYNC_FRAMES 3
354 /* Perform extended validation to check that subsequent headers match
355 * the first header given here in important characteristics, to avoid
356 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
357 * frames to match their major characteristics.
359 * If at_eos is set to TRUE, we just check that we don't find any invalid
360 * frames in whatever data is available, rather than requiring a full
361 * MIN_RESYNC_FRAMES of data.
363 * Returns TRUE if we've seen enough data to validate or reject the frame.
364 * If TRUE is returned, then *valid contains TRUE if it validated, or false
365 * if we decided it was false sync.
366 * If FALSE is returned, then *valid contains minimum needed data.
369 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
370 guint32 header, int bpf, gboolean at_eos, gint * valid)
376 int frames_found = 1;
379 data = gst_buffer_map (buf, &available, NULL, GST_MAP_READ);
381 while (frames_found < MIN_RESYNC_FRAMES) {
382 /* Check if we have enough data for all these frames, plus the next
384 if (available < offset + 4) {
386 /* Running out of data at EOS is fine; just accept it */
396 next_header = GST_READ_UINT32_BE (data + offset);
397 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
398 offset, (unsigned int) header, (unsigned int) next_header, bpf);
400 /* mask the bits which are allowed to differ between frames */
401 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
402 (0x1 << 9) /* padding */ | \
403 (0xf << 4) /* mode|mode extension */ | \
404 (0xf)) /* copyright|emphasis */
406 if ((next_header & HDRMASK) != (header & HDRMASK)) {
407 /* If any of the unmasked bits don't match, then it's not valid */
408 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
409 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
410 (guint) header, (guint) header & HDRMASK, (guint) next_header,
411 (guint) next_header & HDRMASK, bpf);
414 } else if ((((next_header >> 12) & 0xf) == 0) ||
415 (((next_header >> 12) & 0xf) == 0xf)) {
416 /* The essential parts were the same, but the bitrate held an
417 invalid value - also reject */
418 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
423 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
424 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
433 gst_buffer_unmap (buf, data, available);
438 gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
441 GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
442 /* if it's not a valid sync */
443 if ((head & 0xffe00000) != 0xffe00000) {
444 GST_WARNING_OBJECT (mp3parse, "invalid sync");
447 /* if it's an invalid MPEG version */
448 if (((head >> 19) & 3) == 0x1) {
449 GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
453 /* if it's an invalid layer */
454 if (!((head >> 17) & 3)) {
455 GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
458 /* if it's an invalid bitrate */
459 if (((head >> 12) & 0xf) == 0x0) {
460 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx."
461 "Free format files are not supported yet", (head >> 12) & 0xf);
464 if (((head >> 12) & 0xf) == 0xf) {
465 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
468 /* if it's an invalid samplerate */
469 if (((head >> 10) & 0x3) == 0x3) {
470 GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
475 if ((head & 0x3) == 0x2) {
476 /* Ignore this as there are some files with emphasis 0x2 that can
477 * be played fine. See BGO #537235 */
478 GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
485 gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
486 GstBaseParseFrame * frame, guint * framesize, gint * skipsize)
488 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
489 GstBuffer *buf = frame->buffer;
490 GstByteReader reader;
492 gboolean lost_sync, draining, valid, caps_change;
494 guint bitrate, layer, rate, channels, version, mode, crc;
497 gboolean res = FALSE;
499 data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
500 if (G_UNLIKELY (bufsize < 6))
503 gst_byte_reader_init (&reader, data, bufsize);
505 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
508 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
510 /* didn't find anything that looks like a sync word, skip */
512 *skipsize = bufsize - 3;
516 /* possible frame header, but not at offset 0? skip bytes before sync */
522 /* make sure the values in the frame header look sane */
523 header = GST_READ_UINT32_BE (data);
524 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
529 GST_LOG_OBJECT (parse, "got frame");
531 bpf = mp3_type_frame_length_from_header (mp3parse, header,
532 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
535 if (channels != mp3parse->channels || rate != mp3parse->rate ||
536 layer != mp3parse->layer || version != mp3parse->version)
541 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
542 draining = GST_BASE_PARSE_DRAINING (parse);
544 if (!draining && (lost_sync || caps_change)) {
545 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
547 /* not enough data */
548 gst_base_parse_set_min_frame_size (parse, valid);
557 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
558 /* avoid caps jitter that we can't be sure of */
563 /* restore default minimum */
564 gst_base_parse_set_min_frame_size (parse, MIN_FRAME_SIZE);
570 gst_buffer_unmap (buf, data, bufsize);
575 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
578 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
579 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
580 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
581 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
582 gint offset_xing, offset_vbri;
584 gint64 upstream_total_bytes = 0;
585 guint32 read_id_xing = 0, read_id_vbri = 0;
586 guint8 *data, *origdata;
590 if (mp3parse->sent_codec_tag)
593 /* Check first frame for Xing info */
594 if (mp3parse->version == 1) { /* MPEG-1 file */
595 if (mp3parse->channels == 1)
599 } else { /* MPEG-2 header */
600 if (mp3parse->channels == 1)
606 /* The VBRI tag is always at offset 0x20 */
609 /* Skip the 4 bytes of the MP3 header too */
613 /* Check if we have enough data to read the Xing header */
614 origdata = data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
617 if (avail >= offset_xing + 4) {
618 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
620 if (avail >= offset_vbri + 4) {
621 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
624 /* obtain real upstream total bytes */
625 if (!gst_pad_query_peer_duration (GST_BASE_PARSE_SINK_PAD (mp3parse),
626 GST_FORMAT_BYTES, &upstream_total_bytes))
627 upstream_total_bytes = 0;
629 if (read_id_xing == xing_id || read_id_xing == info_id) {
631 guint bytes_needed = offset_xing + 8;
633 GstClockTime total_time;
635 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
637 /* Move data after Xing header */
638 data += offset_xing + 4;
640 /* Read 4 base bytes of flags, big-endian */
641 xing_flags = GST_READ_UINT32_BE (data);
643 if (xing_flags & XING_FRAMES_FLAG)
645 if (xing_flags & XING_BYTES_FLAG)
647 if (xing_flags & XING_TOC_FLAG)
649 if (xing_flags & XING_VBR_SCALE_FLAG)
651 if (avail < bytes_needed) {
652 GST_DEBUG_OBJECT (mp3parse,
653 "Not enough data to read Xing header (need %d)", bytes_needed);
657 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
658 mp3parse->xing_flags = xing_flags;
660 if (xing_flags & XING_FRAMES_FLAG) {
661 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
662 if (mp3parse->xing_frames == 0) {
663 GST_WARNING_OBJECT (mp3parse,
664 "Invalid number of frames in Xing header");
665 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
667 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
668 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
674 mp3parse->xing_frames = 0;
675 mp3parse->xing_total_time = 0;
678 if (xing_flags & XING_BYTES_FLAG) {
679 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
680 if (mp3parse->xing_bytes == 0) {
681 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
682 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
686 mp3parse->xing_bytes = 0;
689 /* If we know the upstream size and duration, compute the
690 * total bitrate, rounded up to the nearest kbit/sec */
691 if ((total_time = mp3parse->xing_total_time) &&
692 (total_bytes = mp3parse->xing_bytes)) {
693 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
694 8 * GST_SECOND, total_time);
695 mp3parse->xing_bitrate += 500;
696 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
699 if (xing_flags & XING_TOC_FLAG) {
701 guchar *table = mp3parse->xing_seek_table;
706 GST_DEBUG_OBJECT (mp3parse,
707 "Subtracting initial offset of %d bytes from Xing TOC", first);
709 /* xing seek table: percent time -> 1/256 bytepos */
710 for (i = 0; i < 100; i++) {
711 new = data[i] - first;
713 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
714 mp3parse->xing_flags &= ~XING_TOC_FLAG;
717 mp3parse->xing_seek_table[i] = old = new;
720 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
721 for (i = 0; i < 256; i++) {
722 while (percent < 99 && table[percent + 1] <= i)
725 if (table[percent] == i) {
726 mp3parse->xing_seek_table_inverse[i] = percent * 100;
727 } else if (table[percent] < i && percent < 99) {
729 gint a = percent, b = percent + 1;
733 fx = (b - a) / (fb - fa) * (i - fa) + a;
734 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
735 } else if (percent == 99) {
737 gint a = percent, b = 100;
741 fx = (b - a) / (fb - fa) * (i - fa) + a;
742 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
748 memset (mp3parse->xing_seek_table, 0, 100);
749 memset (mp3parse->xing_seek_table_inverse, 0, 256);
752 if (xing_flags & XING_VBR_SCALE_FLAG) {
753 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
756 mp3parse->xing_vbr_scale = 0;
758 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
759 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
760 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
761 mp3parse->xing_vbr_scale);
763 /* check for truncated file */
764 if (upstream_total_bytes && mp3parse->xing_bytes &&
765 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
766 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
767 "invalidating Xing header duration and size");
768 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
769 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
772 /* Optional LAME tag? */
773 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
774 gchar lame_version[10] = { 0, };
776 guint32 encoder_delay, encoder_padding;
778 memcpy (lame_version, data, 9);
780 tag_rev = data[0] >> 4;
781 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
782 tag_rev, lame_version);
784 /* Skip all the information we're not interested in */
786 /* Encoder delay and end padding */
787 encoder_delay = GST_READ_UINT24_BE (data);
788 encoder_delay >>= 12;
789 encoder_padding = GST_READ_UINT24_BE (data);
790 encoder_padding &= 0x000fff;
792 mp3parse->encoder_delay = encoder_delay;
793 mp3parse->encoder_padding = encoder_padding;
795 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
796 encoder_delay, encoder_padding);
800 if (read_id_vbri == vbri_id) {
801 gint64 total_bytes, total_frames;
802 GstClockTime total_time;
803 guint16 nseek_points;
805 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
807 if (avail < offset_vbri + 26) {
808 GST_DEBUG_OBJECT (mp3parse,
809 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
813 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
815 /* Move data after VBRI header */
816 data += offset_vbri + 4;
818 if (GST_READ_UINT16_BE (data) != 0x0001) {
819 GST_WARNING_OBJECT (mp3parse,
820 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
825 /* Skip encoder delay */
831 total_bytes = GST_READ_UINT32_BE (data);
832 if (total_bytes != 0)
833 mp3parse->vbri_bytes = total_bytes;
836 total_frames = GST_READ_UINT32_BE (data);
837 if (total_frames != 0) {
838 mp3parse->vbri_frames = total_frames;
839 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
840 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
844 /* If we know the upstream size and duration, compute the
845 * total bitrate, rounded up to the nearest kbit/sec */
846 if ((total_time = mp3parse->vbri_total_time) &&
847 (total_bytes = mp3parse->vbri_bytes)) {
848 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
849 8 * GST_SECOND, total_time);
850 mp3parse->vbri_bitrate += 500;
851 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
854 nseek_points = GST_READ_UINT16_BE (data);
857 if (nseek_points > 0) {
858 guint scale, seek_bytes, seek_frames;
861 mp3parse->vbri_seek_points = nseek_points;
863 scale = GST_READ_UINT16_BE (data);
866 seek_bytes = GST_READ_UINT16_BE (data);
869 seek_frames = GST_READ_UINT16_BE (data);
871 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
872 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
876 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
877 GST_WARNING_OBJECT (mp3parse,
878 "Not enough data to read VBRI seek table (need %d)",
879 offset_vbri + 26 + nseek_points * seek_bytes);
883 if (seek_frames * nseek_points < total_frames - seek_frames ||
884 seek_frames * nseek_points > total_frames + seek_frames) {
885 GST_WARNING_OBJECT (mp3parse,
886 "VBRI seek table doesn't cover the complete file");
890 if (avail < offset_vbri + 26) {
891 GST_DEBUG_OBJECT (mp3parse,
892 "Not enough data to read VBRI header (need %d)",
893 offset_vbri + 26 + nseek_points * seek_bytes);
898 data += offset_vbri + 26;
900 /* VBRI seek table: frame/seek_frames -> byte */
901 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
903 for (i = 0; i < nseek_points; i++) {
904 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
906 } else if (seek_bytes == 3)
907 for (i = 0; i < nseek_points; i++) {
908 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
910 } else if (seek_bytes == 2)
911 for (i = 0; i < nseek_points; i++) {
912 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
914 } else /* seek_bytes == 1 */
915 for (i = 0; i < nseek_points; i++) {
916 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
922 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
923 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
924 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
926 /* check for truncated file */
927 if (upstream_total_bytes && mp3parse->vbri_bytes &&
928 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
929 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
930 "invalidating VBRI header duration and size");
931 mp3parse->vbri_valid = FALSE;
933 mp3parse->vbri_valid = TRUE;
936 GST_DEBUG_OBJECT (mp3parse,
937 "Xing, LAME or VBRI header not found in first frame");
940 /* set duration if tables provided a valid one */
941 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
942 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
943 mp3parse->xing_total_time, 0);
945 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
946 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
947 mp3parse->vbri_total_time, 0);
950 /* tell baseclass how nicely we can seek, and a bitrate if one found */
951 /* FIXME: fill index with seek table */
953 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
954 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
955 mp3parse->xing_total_time)
956 seekable = GST_BASE_PARSE_SEEK_TABLE;
958 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
959 mp3parse->vbri_total_time)
960 seekable = GST_BASE_PARSE_SEEK_TABLE;
963 if (mp3parse->xing_bitrate)
964 bitrate = mp3parse->xing_bitrate;
965 else if (mp3parse->vbri_bitrate)
966 bitrate = mp3parse->vbri_bitrate;
970 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
973 gst_buffer_unmap (buf, origdata, bufsize);
977 gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
978 GstBaseParseFrame * frame)
980 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
981 GstBuffer *buf = frame->buffer;
984 guint bitrate, layer, rate, channels, version, mode, crc;
986 data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
987 if (G_UNLIKELY (bufsize < 4))
990 if (!mp3_type_frame_length_from_header (mp3parse,
991 GST_READ_UINT32_BE (data),
992 &version, &layer, &channels, &bitrate, &rate, &mode, &crc))
995 if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
996 layer != mp3parse->layer || version != mp3parse->version)) {
997 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
998 "mpegversion", G_TYPE_INT, 1,
999 "mpegaudioversion", G_TYPE_INT, version,
1000 "layer", G_TYPE_INT, layer,
1001 "rate", G_TYPE_INT, rate,
1002 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
1003 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
1004 gst_caps_unref (caps);
1006 mp3parse->rate = rate;
1007 mp3parse->channels = channels;
1008 mp3parse->layer = layer;
1009 mp3parse->version = version;
1011 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
1012 if (mp3parse->layer == 1)
1013 mp3parse->spf = 384;
1014 else if (mp3parse->layer == 2)
1015 mp3parse->spf = 1152;
1016 else if (mp3parse->version == 1) {
1017 mp3parse->spf = 1152;
1019 /* MPEG-2 or "2.5" */
1020 mp3parse->spf = 576;
1024 * We start pushing 9 frames earlier (29 frames for MPEG2) than
1025 * segment start to be able to decode the first frame we want.
1026 * 9 (29) frames are the theoretical maximum of frames that contain
1027 * data for the current frame (bit reservoir).
1030 * Some mp3 streams have an offset in the timestamps, for which we have to
1031 * push the frame *after* the end position in order for the decoder to be
1032 * able to decode everything up until the segment.stop position. */
1033 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
1034 (version == 1) ? 10 : 30, 2);
1037 mp3parse->hdr_bitrate = bitrate;
1039 /* For first frame; check for seek tables and output a codec tag */
1040 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
1042 /* store some frame info for later processing */
1043 mp3parse->last_crc = crc;
1044 mp3parse->last_mode = mode;
1046 gst_buffer_unmap (buf, data, bufsize);
1052 /* this really shouldn't ever happen */
1053 gst_buffer_unmap (buf, data, bufsize);
1054 GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
1055 return GST_FLOW_ERROR;
1060 gst_buffer_unmap (buf, data, bufsize);
1061 return GST_FLOW_ERROR;
1066 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1067 GstClockTime ts, gint64 * bytepos)
1070 GstClockTime total_time;
1072 /* If XING seek table exists use this for time->byte conversion */
1073 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1074 (total_bytes = mp3parse->xing_bytes) &&
1075 (total_time = mp3parse->xing_total_time)) {
1078 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1079 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1080 gint index = CLAMP (percent, 0, 99);
1082 fa = mp3parse->xing_seek_table[index];
1084 fb = mp3parse->xing_seek_table[index + 1];
1088 fx = fa + (fb - fa) * (percent - index);
1090 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1095 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1096 (total_time = mp3parse->vbri_total_time)) {
1098 gdouble a, b, fa, fb;
1100 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1101 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1103 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1104 mp3parse->vbri_seek_points));
1106 for (j = i; j >= 0; j--)
1107 fa += mp3parse->vbri_seek_table[j];
1109 if (i + 1 < mp3parse->vbri_seek_points) {
1110 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1111 mp3parse->vbri_seek_points));
1112 fb = fa + mp3parse->vbri_seek_table[i + 1];
1114 b = gst_guint64_to_gdouble (total_time);
1118 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1127 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1128 gint64 bytepos, GstClockTime * ts)
1131 GstClockTime total_time;
1133 /* If XING seek table exists use this for byte->time conversion */
1134 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1135 (total_bytes = mp3parse->xing_bytes) &&
1136 (total_time = mp3parse->xing_total_time)) {
1141 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1142 index = CLAMP (pos, 0, 255);
1143 fa = mp3parse->xing_seek_table_inverse[index];
1145 fb = mp3parse->xing_seek_table_inverse[index + 1];
1149 fx = fa + (fb - fa) * (pos - index);
1151 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1156 if (mp3parse->vbri_seek_table &&
1157 (total_bytes = mp3parse->vbri_bytes) &&
1158 (total_time = mp3parse->vbri_total_time)) {
1161 gdouble a, b, fa, fb;
1164 sum += mp3parse->vbri_seek_table[i];
1166 } while (i + 1 < mp3parse->vbri_seek_points
1167 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1170 a = gst_guint64_to_gdouble (sum);
1171 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1172 mp3parse->vbri_seek_points));
1174 if (i + 1 < mp3parse->vbri_seek_points) {
1175 b = a + mp3parse->vbri_seek_table[i + 1];
1176 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1177 mp3parse->vbri_seek_points));
1180 fb = gst_guint64_to_gdouble (total_time);
1183 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1192 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1193 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1195 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1196 gboolean res = FALSE;
1198 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1200 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1201 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1202 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1203 (GstClockTime *) dest_value);
1205 /* if no tables, fall back to default estimated rate based conversion */
1207 return gst_base_parse_convert_default (parse, src_format, src_value,
1208 dest_format, dest_value);
1213 static GstFlowReturn
1214 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1215 GstBaseParseFrame * frame)
1217 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1218 GstTagList *taglist;
1220 /* tag sending done late enough in hook to ensure pending events
1221 * have already been sent */
1223 if (!mp3parse->sent_codec_tag) {
1227 if (mp3parse->layer == 3) {
1228 codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1229 mp3parse->version, mp3parse->layer);
1231 codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1232 mp3parse->version, mp3parse->layer);
1234 taglist = gst_tag_list_new ();
1235 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1236 GST_TAG_AUDIO_CODEC, codec, NULL);
1237 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1238 mp3parse->vbri_bitrate == 0) {
1239 /* We don't have a VBR bitrate, so post the available bitrate as
1240 * nominal and let baseparse calculate the real bitrate */
1241 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1242 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1244 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1245 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1248 /* also signals the end of first-frame processing */
1249 mp3parse->sent_codec_tag = TRUE;
1252 /* we will create a taglist (if any of the parameters has changed)
1253 * to add the tags that changed */
1255 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1259 taglist = gst_tag_list_new ();
1261 mp3parse->last_posted_crc = mp3parse->last_crc;
1262 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1267 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1271 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1273 taglist = gst_tag_list_new ();
1275 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1277 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1278 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1281 /* if the taglist exists, we need to send it */
1283 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1284 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1287 /* usual clipping applies */
1288 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;