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 static GstStaticPadTemplate src_template = GST_STATIC_PAD_TEMPLATE ("src",
74 GST_STATIC_CAPS ("audio/mpeg, "
75 "mpegversion = (int) 1, "
76 "layer = (int) [ 1, 3 ], "
77 "rate = (int) [ 8000, 48000 ], channels = (int) [ 1, 2 ],"
78 "parsed=(boolean) true")
81 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
84 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) 1, parsed=(boolean)false")
87 static void gst_mpeg_audio_parse_finalize (GObject * object);
89 static gboolean gst_mpeg_audio_parse_start (GstBaseParse * parse);
90 static gboolean gst_mpeg_audio_parse_stop (GstBaseParse * parse);
91 static gboolean gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
92 GstBaseParseFrame * frame, guint * size, gint * skipsize);
93 static GstFlowReturn gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
94 GstBaseParseFrame * frame);
95 static GstFlowReturn gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
96 GstBaseParseFrame * frame);
97 static gboolean gst_mpeg_audio_parse_convert (GstBaseParse * parse,
98 GstFormat src_format, gint64 src_value,
99 GstFormat dest_format, gint64 * dest_value);
101 GST_BOILERPLATE (GstMpegAudioParse, gst_mpeg_audio_parse, GstBaseParse,
102 GST_TYPE_BASE_PARSE);
104 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
105 (gst_mpeg_audio_channel_mode_get_type())
107 static const GEnumValue mpeg_audio_channel_mode[] = {
108 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
109 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
110 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
111 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
112 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
117 gst_mpeg_audio_channel_mode_get_type (void)
119 static GType mpeg_audio_channel_mode_type = 0;
121 if (!mpeg_audio_channel_mode_type) {
122 mpeg_audio_channel_mode_type =
123 g_enum_register_static ("GstMpegAudioChannelMode",
124 mpeg_audio_channel_mode);
126 return mpeg_audio_channel_mode_type;
130 gst_mpeg_audio_channel_mode_get_nick (gint mode)
133 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
134 if (mpeg_audio_channel_mode[i].value == mode)
135 return mpeg_audio_channel_mode[i].value_nick;
141 gst_mpeg_audio_parse_base_init (gpointer klass)
143 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
145 gst_element_class_add_pad_template (element_class,
146 gst_static_pad_template_get (&sink_template));
147 gst_element_class_add_pad_template (element_class,
148 gst_static_pad_template_get (&src_template));
150 gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
151 "Codec/Parser/Audio",
152 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
153 "Jan Schmidt <thaytan@mad.scientist.com>,"
154 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
158 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
160 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
161 GObjectClass *object_class = G_OBJECT_CLASS (klass);
163 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
164 "MPEG1 audio stream parser");
166 object_class->finalize = gst_mpeg_audio_parse_finalize;
168 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
169 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
170 parse_class->check_valid_frame =
171 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
172 parse_class->parse_frame =
173 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
174 parse_class->pre_push_frame =
175 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
176 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
179 #define GST_TAG_CRC "has-crc"
180 #define GST_TAG_MODE "channel-mode"
182 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
183 "has crc", "Using CRC", NULL);
184 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
185 "channel mode", "MPEG audio channel mode", NULL);
187 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
191 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
193 mp3parse->channels = -1;
195 mp3parse->sent_codec_tag = FALSE;
196 mp3parse->last_posted_crc = CRC_UNKNOWN;
197 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
199 mp3parse->hdr_bitrate = 0;
201 mp3parse->xing_flags = 0;
202 mp3parse->xing_bitrate = 0;
203 mp3parse->xing_frames = 0;
204 mp3parse->xing_total_time = 0;
205 mp3parse->xing_bytes = 0;
206 mp3parse->xing_vbr_scale = 0;
207 memset (mp3parse->xing_seek_table, 0, 100);
208 memset (mp3parse->xing_seek_table_inverse, 0, 256);
210 mp3parse->vbri_bitrate = 0;
211 mp3parse->vbri_frames = 0;
212 mp3parse->vbri_total_time = 0;
213 mp3parse->vbri_bytes = 0;
214 mp3parse->vbri_seek_points = 0;
215 g_free (mp3parse->vbri_seek_table);
216 mp3parse->vbri_seek_table = NULL;
218 mp3parse->encoder_delay = 0;
219 mp3parse->encoder_padding = 0;
223 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse,
224 GstMpegAudioParseClass * klass)
226 gst_mpeg_audio_parse_reset (mp3parse);
230 gst_mpeg_audio_parse_finalize (GObject * object)
232 G_OBJECT_CLASS (parent_class)->finalize (object);
236 gst_mpeg_audio_parse_start (GstBaseParse * parse)
238 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
240 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), 1024);
241 GST_DEBUG_OBJECT (parse, "starting");
243 gst_mpeg_audio_parse_reset (mp3parse);
249 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
251 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
253 GST_DEBUG_OBJECT (parse, "stopping");
255 gst_mpeg_audio_parse_reset (mp3parse);
260 static const guint mp3types_bitrates[2][3][16] = {
262 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
263 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
264 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
267 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
268 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
269 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
273 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
274 {22050, 24000, 16000},
279 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
280 guint * put_version, guint * put_layer, guint * put_channels,
281 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
285 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
289 if (header & (1 << 20)) {
290 lsf = (header & (1 << 19)) ? 0 : 1;
297 version = 1 + lsf + mpg25;
299 layer = 4 - ((header >> 17) & 0x3);
301 crc = (header >> 16) & 0x1;
303 bitrate = (header >> 12) & 0xF;
304 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
305 /* The caller has ensured we have a valid header, so bitrate can't be
307 g_assert (bitrate != 0);
309 samplerate = (header >> 10) & 0x3;
310 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
312 padding = (header >> 9) & 0x1;
314 mode = (header >> 6) & 0x3;
315 channels = (mode == 3) ? 1 : 2;
319 length = 4 * ((bitrate * 12) / samplerate + padding);
322 length = (bitrate * 144) / samplerate + padding;
326 length = (bitrate * 144) / (samplerate << lsf) + padding;
330 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
332 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
333 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
334 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
337 *put_version = version;
341 *put_channels = channels;
343 *put_bitrate = bitrate;
345 *put_samplerate = samplerate;
354 /* Minimum number of consecutive, valid-looking frames to consider
356 #define MIN_RESYNC_FRAMES 3
358 /* Perform extended validation to check that subsequent headers match
359 * the first header given here in important characteristics, to avoid
360 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
361 * frames to match their major characteristics.
363 * If at_eos is set to TRUE, we just check that we don't find any invalid
364 * frames in whatever data is available, rather than requiring a full
365 * MIN_RESYNC_FRAMES of data.
367 * Returns TRUE if we've seen enough data to validate or reject the frame.
368 * If TRUE is returned, then *valid contains TRUE if it validated, or false
369 * if we decided it was false sync.
370 * If FALSE is returned, then *valid contains minimum needed data.
373 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
374 guint32 header, int bpf, gboolean at_eos, gint * valid)
379 int frames_found = 1;
382 available = GST_BUFFER_SIZE (buf);
383 data = GST_BUFFER_DATA (buf);
385 while (frames_found < MIN_RESYNC_FRAMES) {
386 /* Check if we have enough data for all these frames, plus the next
388 if (available < offset + 4) {
390 /* Running out of data at EOS is fine; just accept it */
399 next_header = GST_READ_UINT32_BE (data + offset);
400 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
401 offset, (unsigned int) header, (unsigned int) next_header, bpf);
403 /* mask the bits which are allowed to differ between frames */
404 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
405 (0x1 << 9) /* padding */ | \
406 (0xf << 4) /* mode|mode extension */ | \
407 (0xf)) /* copyright|emphasis */
409 if ((next_header & HDRMASK) != (header & HDRMASK)) {
410 /* If any of the unmasked bits don't match, then it's not valid */
411 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
412 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
413 (guint) header, (guint) header & HDRMASK, (guint) next_header,
414 (guint) next_header & HDRMASK, bpf);
417 } else if ((((next_header >> 12) & 0xf) == 0) ||
418 (((next_header >> 12) & 0xf) == 0xf)) {
419 /* The essential parts were the same, but the bitrate held an
420 invalid value - also reject */
421 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
426 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
427 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
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 = GST_BYTE_READER_INIT_FROM_BUFFER (buf);
492 gboolean lost_sync, draining, valid, caps_change;
494 guint bitrate, layer, rate, channels, version, mode, crc;
496 if (G_UNLIKELY (GST_BUFFER_SIZE (buf) < 6))
499 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
500 0, GST_BUFFER_SIZE (buf));
502 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
504 /* didn't find anything that looks like a sync word, skip */
506 *skipsize = GST_BUFFER_SIZE (buf) - 3;
510 /* possible frame header, but not at offset 0? skip bytes before sync */
516 /* make sure the values in the frame header look sane */
517 header = GST_READ_UINT32_BE (GST_BUFFER_DATA (buf));
518 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
523 GST_LOG_OBJECT (parse, "got frame");
525 bpf = mp3_type_frame_length_from_header (mp3parse, header,
526 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
529 if (channels != mp3parse->channels || rate != mp3parse->rate ||
530 layer != mp3parse->layer || version != mp3parse->version)
535 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
536 draining = GST_BASE_PARSE_DRAINING (parse);
538 if (!draining && (lost_sync || caps_change)) {
539 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
541 /* not enough data */
542 gst_base_parse_set_min_frame_size (parse, valid);
551 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
552 /* avoid caps jitter that we can't be sure of */
562 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
565 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
566 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
567 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
568 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
569 gint offset_xing, offset_vbri;
571 gint64 upstream_total_bytes = 0;
572 GstFormat fmt = GST_FORMAT_BYTES;
573 guint32 read_id_xing = 0, read_id_vbri = 0;
577 if (mp3parse->sent_codec_tag)
580 /* Check first frame for Xing info */
581 if (mp3parse->version == 1) { /* MPEG-1 file */
582 if (mp3parse->channels == 1)
586 } else { /* MPEG-2 header */
587 if (mp3parse->channels == 1)
593 /* The VBRI tag is always at offset 0x20 */
596 /* Skip the 4 bytes of the MP3 header too */
600 /* Check if we have enough data to read the Xing header */
601 avail = GST_BUFFER_SIZE (buf);
602 data = GST_BUFFER_DATA (buf);
604 if (avail >= offset_xing + 4) {
605 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
607 if (avail >= offset_vbri + 4) {
608 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
611 /* obtain real upstream total bytes */
612 fmt = GST_FORMAT_BYTES;
613 if (!gst_pad_query_peer_duration (GST_BASE_PARSE_SINK_PAD (GST_BASE_PARSE
614 (mp3parse)), &fmt, &upstream_total_bytes))
615 upstream_total_bytes = 0;
617 if (read_id_xing == xing_id || read_id_xing == info_id) {
619 guint bytes_needed = offset_xing + 8;
621 GstClockTime total_time;
623 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
625 /* Move data after Xing header */
626 data += offset_xing + 4;
628 /* Read 4 base bytes of flags, big-endian */
629 xing_flags = GST_READ_UINT32_BE (data);
631 if (xing_flags & XING_FRAMES_FLAG)
633 if (xing_flags & XING_BYTES_FLAG)
635 if (xing_flags & XING_TOC_FLAG)
637 if (xing_flags & XING_VBR_SCALE_FLAG)
639 if (avail < bytes_needed) {
640 GST_DEBUG_OBJECT (mp3parse,
641 "Not enough data to read Xing header (need %d)", bytes_needed);
645 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
646 mp3parse->xing_flags = xing_flags;
648 if (xing_flags & XING_FRAMES_FLAG) {
649 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
650 if (mp3parse->xing_frames == 0) {
651 GST_WARNING_OBJECT (mp3parse,
652 "Invalid number of frames in Xing header");
653 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
655 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
656 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
662 mp3parse->xing_frames = 0;
663 mp3parse->xing_total_time = 0;
666 if (xing_flags & XING_BYTES_FLAG) {
667 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
668 if (mp3parse->xing_bytes == 0) {
669 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
670 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
674 mp3parse->xing_bytes = 0;
677 /* If we know the upstream size and duration, compute the
678 * total bitrate, rounded up to the nearest kbit/sec */
679 if ((total_time = mp3parse->xing_total_time) &&
680 (total_bytes = mp3parse->xing_bytes)) {
681 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
682 8 * GST_SECOND, total_time);
683 mp3parse->xing_bitrate += 500;
684 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
687 if (xing_flags & XING_TOC_FLAG) {
689 guchar *table = mp3parse->xing_seek_table;
694 GST_DEBUG_OBJECT (mp3parse,
695 "Subtracting initial offset of %d bytes from Xing TOC", first);
697 /* xing seek table: percent time -> 1/256 bytepos */
698 for (i = 0; i < 100; i++) {
699 new = data[i] - first;
701 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
702 mp3parse->xing_flags &= ~XING_TOC_FLAG;
705 mp3parse->xing_seek_table[i] = old = new;
708 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
709 for (i = 0; i < 256; i++) {
710 while (percent < 99 && table[percent + 1] <= i)
713 if (table[percent] == i) {
714 mp3parse->xing_seek_table_inverse[i] = percent * 100;
715 } else if (table[percent] < i && percent < 99) {
717 gint a = percent, b = percent + 1;
721 fx = (b - a) / (fb - fa) * (i - fa) + a;
722 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
723 } else if (percent == 99) {
725 gint a = percent, b = 100;
729 fx = (b - a) / (fb - fa) * (i - fa) + a;
730 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
736 memset (mp3parse->xing_seek_table, 0, 100);
737 memset (mp3parse->xing_seek_table_inverse, 0, 256);
740 if (xing_flags & XING_VBR_SCALE_FLAG) {
741 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
744 mp3parse->xing_vbr_scale = 0;
746 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
747 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
748 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
749 mp3parse->xing_vbr_scale);
751 /* check for truncated file */
752 if (upstream_total_bytes && mp3parse->xing_bytes &&
753 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
754 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
755 "invalidating Xing header duration and size");
756 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
757 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
760 /* Optional LAME tag? */
761 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
762 gchar lame_version[10] = { 0, };
764 guint32 encoder_delay, encoder_padding;
766 memcpy (lame_version, data, 9);
768 tag_rev = data[0] >> 4;
769 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
770 tag_rev, lame_version);
772 /* Skip all the information we're not interested in */
774 /* Encoder delay and end padding */
775 encoder_delay = GST_READ_UINT24_BE (data);
776 encoder_delay >>= 12;
777 encoder_padding = GST_READ_UINT24_BE (data);
778 encoder_padding &= 0x000fff;
780 mp3parse->encoder_delay = encoder_delay;
781 mp3parse->encoder_padding = encoder_padding;
783 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
784 encoder_delay, encoder_padding);
788 if (read_id_vbri == vbri_id) {
789 gint64 total_bytes, total_frames;
790 GstClockTime total_time;
791 guint16 nseek_points;
793 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
795 if (avail < offset_vbri + 26) {
796 GST_DEBUG_OBJECT (mp3parse,
797 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
801 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
803 /* Move data after VBRI header */
804 data += offset_vbri + 4;
806 if (GST_READ_UINT16_BE (data) != 0x0001) {
807 GST_WARNING_OBJECT (mp3parse,
808 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
813 /* Skip encoder delay */
819 total_bytes = GST_READ_UINT32_BE (data);
820 if (total_bytes != 0)
821 mp3parse->vbri_bytes = total_bytes;
824 total_frames = GST_READ_UINT32_BE (data);
825 if (total_frames != 0) {
826 mp3parse->vbri_frames = total_frames;
827 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
828 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
832 /* If we know the upstream size and duration, compute the
833 * total bitrate, rounded up to the nearest kbit/sec */
834 if ((total_time = mp3parse->vbri_total_time) &&
835 (total_bytes = mp3parse->vbri_bytes)) {
836 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
837 8 * GST_SECOND, total_time);
838 mp3parse->vbri_bitrate += 500;
839 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
842 nseek_points = GST_READ_UINT16_BE (data);
845 if (nseek_points > 0) {
846 guint scale, seek_bytes, seek_frames;
849 mp3parse->vbri_seek_points = nseek_points;
851 scale = GST_READ_UINT16_BE (data);
854 seek_bytes = GST_READ_UINT16_BE (data);
857 seek_frames = GST_READ_UINT16_BE (data);
859 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
860 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
864 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
865 GST_WARNING_OBJECT (mp3parse,
866 "Not enough data to read VBRI seek table (need %d)",
867 offset_vbri + 26 + nseek_points * seek_bytes);
871 if (seek_frames * nseek_points < total_frames - seek_frames ||
872 seek_frames * nseek_points > total_frames + seek_frames) {
873 GST_WARNING_OBJECT (mp3parse,
874 "VBRI seek table doesn't cover the complete file");
878 if (avail < offset_vbri + 26) {
879 GST_DEBUG_OBJECT (mp3parse,
880 "Not enough data to read VBRI header (need %d)",
881 offset_vbri + 26 + nseek_points * seek_bytes);
885 data = GST_BUFFER_DATA (buf);
886 data += offset_vbri + 26;
888 /* VBRI seek table: frame/seek_frames -> byte */
889 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
891 for (i = 0; i < nseek_points; i++) {
892 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
894 } else if (seek_bytes == 3)
895 for (i = 0; i < nseek_points; i++) {
896 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
898 } else if (seek_bytes == 2)
899 for (i = 0; i < nseek_points; i++) {
900 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
902 } else /* seek_bytes == 1 */
903 for (i = 0; i < nseek_points; i++) {
904 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
910 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
911 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
912 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
914 /* check for truncated file */
915 if (upstream_total_bytes && mp3parse->vbri_bytes &&
916 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
917 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
918 "invalidating VBRI header duration and size");
919 mp3parse->vbri_valid = FALSE;
921 mp3parse->vbri_valid = TRUE;
924 GST_DEBUG_OBJECT (mp3parse,
925 "Xing, LAME or VBRI header not found in first frame");
928 /* set duration if tables provided a valid one */
929 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
930 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
931 mp3parse->xing_total_time, 0);
933 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
934 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
935 mp3parse->vbri_total_time, 0);
938 /* tell baseclass how nicely we can seek, and a bitrate if one found */
939 /* FIXME: fill index with seek table */
941 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
942 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
943 mp3parse->xing_total_time)
944 seekable = GST_BASE_PARSE_SEEK_TABLE;
946 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
947 mp3parse->vbri_total_time)
948 seekable = GST_BASE_PARSE_SEEK_TABLE;
951 if (mp3parse->xing_bitrate)
952 bitrate = mp3parse->xing_bitrate;
953 else if (mp3parse->vbri_bitrate)
954 bitrate = mp3parse->vbri_bitrate;
958 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
962 gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
963 GstBaseParseFrame * frame)
965 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
966 GstBuffer *buf = frame->buffer;
967 guint bitrate, layer, rate, channels, version, mode, crc;
969 g_return_val_if_fail (GST_BUFFER_SIZE (buf) >= 4, GST_FLOW_ERROR);
971 if (!mp3_type_frame_length_from_header (mp3parse,
972 GST_READ_UINT32_BE (GST_BUFFER_DATA (buf)),
973 &version, &layer, &channels, &bitrate, &rate, &mode, &crc))
976 if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
977 layer != mp3parse->layer || version != mp3parse->version)) {
978 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
979 "mpegversion", G_TYPE_INT, 1,
980 "mpegaudioversion", G_TYPE_INT, version,
981 "layer", G_TYPE_INT, layer,
982 "rate", G_TYPE_INT, rate,
983 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
984 gst_buffer_set_caps (buf, caps);
985 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
986 gst_caps_unref (caps);
988 mp3parse->rate = rate;
989 mp3parse->channels = channels;
990 mp3parse->layer = layer;
991 mp3parse->version = version;
993 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
994 if (mp3parse->layer == 1)
996 else if (mp3parse->layer == 2)
997 mp3parse->spf = 1152;
998 else if (mp3parse->version == 1) {
999 mp3parse->spf = 1152;
1001 /* MPEG-2 or "2.5" */
1002 mp3parse->spf = 576;
1006 * We start pushing 9 frames earlier (29 frames for MPEG2) than
1007 * segment start to be able to decode the first frame we want.
1008 * 9 (29) frames are the theoretical maximum of frames that contain
1009 * data for the current frame (bit reservoir).
1012 * Some mp3 streams have an offset in the timestamps, for which we have to
1013 * push the frame *after* the end position in order for the decoder to be
1014 * able to decode everything up until the segment.stop position. */
1015 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
1016 (version == 1) ? 10 : 30, 2);
1019 mp3parse->hdr_bitrate = bitrate;
1021 /* For first frame; check for seek tables and output a codec tag */
1022 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
1024 /* store some frame info for later processing */
1025 mp3parse->last_crc = crc;
1026 mp3parse->last_mode = mode;
1033 /* this really shouldn't ever happen */
1034 GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
1035 return GST_FLOW_ERROR;
1040 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1041 GstClockTime ts, gint64 * bytepos)
1044 GstClockTime total_time;
1046 /* If XING seek table exists use this for time->byte conversion */
1047 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1048 (total_bytes = mp3parse->xing_bytes) &&
1049 (total_time = mp3parse->xing_total_time)) {
1052 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1053 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1054 gint index = CLAMP (percent, 0, 99);
1056 fa = mp3parse->xing_seek_table[index];
1058 fb = mp3parse->xing_seek_table[index + 1];
1062 fx = fa + (fb - fa) * (percent - index);
1064 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1069 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1070 (total_time = mp3parse->vbri_total_time)) {
1072 gdouble a, b, fa, fb;
1074 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1075 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1077 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1078 mp3parse->vbri_seek_points));
1080 for (j = i; j >= 0; j--)
1081 fa += mp3parse->vbri_seek_table[j];
1083 if (i + 1 < mp3parse->vbri_seek_points) {
1084 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1085 mp3parse->vbri_seek_points));
1086 fb = fa + mp3parse->vbri_seek_table[i + 1];
1088 b = gst_guint64_to_gdouble (total_time);
1092 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1101 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1102 gint64 bytepos, GstClockTime * ts)
1105 GstClockTime total_time;
1107 /* If XING seek table exists use this for byte->time conversion */
1108 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1109 (total_bytes = mp3parse->xing_bytes) &&
1110 (total_time = mp3parse->xing_total_time)) {
1115 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1116 index = CLAMP (pos, 0, 255);
1117 fa = mp3parse->xing_seek_table_inverse[index];
1119 fb = mp3parse->xing_seek_table_inverse[index + 1];
1123 fx = fa + (fb - fa) * (pos - index);
1125 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1130 if (mp3parse->vbri_seek_table &&
1131 (total_bytes = mp3parse->vbri_bytes) &&
1132 (total_time = mp3parse->vbri_total_time)) {
1135 gdouble a, b, fa, fb;
1138 sum += mp3parse->vbri_seek_table[i];
1140 } while (i + 1 < mp3parse->vbri_seek_points
1141 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1144 a = gst_guint64_to_gdouble (sum);
1145 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1146 mp3parse->vbri_seek_points));
1148 if (i + 1 < mp3parse->vbri_seek_points) {
1149 b = a + mp3parse->vbri_seek_table[i + 1];
1150 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1151 mp3parse->vbri_seek_points));
1154 fb = gst_guint64_to_gdouble (total_time);
1157 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1166 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1167 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1169 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1170 gboolean res = FALSE;
1172 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1174 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1175 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1176 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1177 (GstClockTime *) dest_value);
1179 /* if no tables, fall back to default estimated rate based conversion */
1181 return gst_base_parse_convert_default (parse, src_format, src_value,
1182 dest_format, dest_value);
1187 static GstFlowReturn
1188 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1189 GstBaseParseFrame * frame)
1191 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1192 GstTagList *taglist;
1194 /* tag sending done late enough in hook to ensure pending events
1195 * have already been sent */
1197 if (!mp3parse->sent_codec_tag) {
1201 if (mp3parse->layer == 3) {
1202 codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1203 mp3parse->version, mp3parse->layer);
1205 codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1206 mp3parse->version, mp3parse->layer);
1208 taglist = gst_tag_list_new ();
1209 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1210 GST_TAG_AUDIO_CODEC, codec, NULL);
1211 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1212 mp3parse->vbri_bitrate == 0) {
1213 /* We don't have a VBR bitrate, so post the available bitrate as
1214 * nominal and let baseparse calculate the real bitrate */
1215 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1216 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1218 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1219 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1222 /* also signals the end of first-frame processing */
1223 mp3parse->sent_codec_tag = TRUE;
1226 /* we will create a taglist (if any of the parameters has changed)
1227 * to add the tags that changed */
1229 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1233 taglist = gst_tag_list_new ();
1235 mp3parse->last_posted_crc = mp3parse->last_crc;
1236 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1241 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1245 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1247 taglist = gst_tag_list_new ();
1249 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1251 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1252 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1255 /* if the taglist exists, we need to send it */
1257 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1258 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1261 /* usual clipping applies */
1262 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;