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 "mpegaudioversion = (int) [ 1, 3], "
80 "rate = (int) [ 8000, 48000 ], "
81 "channels = (int) [ 1, 2 ], " "parsed=(boolean) true")
84 static GstStaticPadTemplate sink_template = GST_STATIC_PAD_TEMPLATE ("sink",
87 GST_STATIC_CAPS ("audio/mpeg, mpegversion = (int) 1")
90 static void gst_mpeg_audio_parse_finalize (GObject * object);
92 static gboolean gst_mpeg_audio_parse_start (GstBaseParse * parse);
93 static gboolean gst_mpeg_audio_parse_stop (GstBaseParse * parse);
94 static gboolean gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
95 GstBaseParseFrame * frame, guint * size, gint * skipsize);
96 static GstFlowReturn gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
97 GstBaseParseFrame * frame);
98 static GstFlowReturn gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
99 GstBaseParseFrame * frame);
100 static gboolean gst_mpeg_audio_parse_convert (GstBaseParse * parse,
101 GstFormat src_format, gint64 src_value,
102 GstFormat dest_format, gint64 * dest_value);
103 static GstCaps *gst_mpeg_audio_parse_get_sink_caps (GstBaseParse * parse,
106 #define gst_mpeg_audio_parse_parent_class parent_class
107 G_DEFINE_TYPE (GstMpegAudioParse, gst_mpeg_audio_parse, GST_TYPE_BASE_PARSE);
109 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
110 (gst_mpeg_audio_channel_mode_get_type())
112 static const GEnumValue mpeg_audio_channel_mode[] = {
113 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
114 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
115 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
116 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
117 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
122 gst_mpeg_audio_channel_mode_get_type (void)
124 static GType mpeg_audio_channel_mode_type = 0;
126 if (!mpeg_audio_channel_mode_type) {
127 mpeg_audio_channel_mode_type =
128 g_enum_register_static ("GstMpegAudioChannelMode",
129 mpeg_audio_channel_mode);
131 return mpeg_audio_channel_mode_type;
135 gst_mpeg_audio_channel_mode_get_nick (gint mode)
138 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
139 if (mpeg_audio_channel_mode[i].value == mode)
140 return mpeg_audio_channel_mode[i].value_nick;
146 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
148 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
149 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
150 GObjectClass *object_class = G_OBJECT_CLASS (klass);
152 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
153 "MPEG1 audio stream parser");
155 object_class->finalize = gst_mpeg_audio_parse_finalize;
157 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
158 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
159 parse_class->check_valid_frame =
160 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
161 parse_class->parse_frame =
162 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
163 parse_class->pre_push_frame =
164 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
165 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
166 parse_class->get_sink_caps =
167 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_get_sink_caps);
170 #define GST_TAG_CRC "has-crc"
171 #define GST_TAG_MODE "channel-mode"
173 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
174 "has crc", "Using CRC", NULL);
175 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
176 "channel mode", "MPEG audio channel mode", NULL);
178 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
180 gst_element_class_add_pad_template (element_class,
181 gst_static_pad_template_get (&sink_template));
182 gst_element_class_add_pad_template (element_class,
183 gst_static_pad_template_get (&src_template));
185 gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
186 "Codec/Parser/Audio",
187 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
188 "Jan Schmidt <thaytan@mad.scientist.com>,"
189 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
193 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
195 mp3parse->channels = -1;
197 mp3parse->sent_codec_tag = FALSE;
198 mp3parse->last_posted_crc = CRC_UNKNOWN;
199 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
201 mp3parse->hdr_bitrate = 0;
203 mp3parse->xing_flags = 0;
204 mp3parse->xing_bitrate = 0;
205 mp3parse->xing_frames = 0;
206 mp3parse->xing_total_time = 0;
207 mp3parse->xing_bytes = 0;
208 mp3parse->xing_vbr_scale = 0;
209 memset (mp3parse->xing_seek_table, 0, 100);
210 memset (mp3parse->xing_seek_table_inverse, 0, 256);
212 mp3parse->vbri_bitrate = 0;
213 mp3parse->vbri_frames = 0;
214 mp3parse->vbri_total_time = 0;
215 mp3parse->vbri_bytes = 0;
216 mp3parse->vbri_seek_points = 0;
217 g_free (mp3parse->vbri_seek_table);
218 mp3parse->vbri_seek_table = NULL;
220 mp3parse->encoder_delay = 0;
221 mp3parse->encoder_padding = 0;
225 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse)
227 gst_mpeg_audio_parse_reset (mp3parse);
231 gst_mpeg_audio_parse_finalize (GObject * object)
233 G_OBJECT_CLASS (parent_class)->finalize (object);
237 gst_mpeg_audio_parse_start (GstBaseParse * parse)
239 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
241 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), MIN_FRAME_SIZE);
242 GST_DEBUG_OBJECT (parse, "starting");
244 gst_mpeg_audio_parse_reset (mp3parse);
250 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
252 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
254 GST_DEBUG_OBJECT (parse, "stopping");
256 gst_mpeg_audio_parse_reset (mp3parse);
261 static const guint mp3types_bitrates[2][3][16] = {
263 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
264 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
265 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
268 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
269 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
270 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
274 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
275 {22050, 24000, 16000},
280 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
281 guint * put_version, guint * put_layer, guint * put_channels,
282 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
286 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
290 if (header & (1 << 20)) {
291 lsf = (header & (1 << 19)) ? 0 : 1;
298 version = 1 + lsf + mpg25;
300 layer = 4 - ((header >> 17) & 0x3);
302 crc = (header >> 16) & 0x1;
304 bitrate = (header >> 12) & 0xF;
305 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
306 /* The caller has ensured we have a valid header, so bitrate can't be
308 g_assert (bitrate != 0);
310 samplerate = (header >> 10) & 0x3;
311 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
313 padding = (header >> 9) & 0x1;
315 mode = (header >> 6) & 0x3;
316 channels = (mode == 3) ? 1 : 2;
320 length = 4 * ((bitrate * 12) / samplerate + padding);
323 length = (bitrate * 144) / samplerate + padding;
327 length = (bitrate * 144) / (samplerate << lsf) + padding;
331 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
333 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
334 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
335 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
338 *put_version = version;
342 *put_channels = channels;
344 *put_bitrate = bitrate;
346 *put_samplerate = samplerate;
355 /* Minimum number of consecutive, valid-looking frames to consider
357 #define MIN_RESYNC_FRAMES 3
359 /* Perform extended validation to check that subsequent headers match
360 * the first header given here in important characteristics, to avoid
361 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
362 * frames to match their major characteristics.
364 * If at_eos is set to TRUE, we just check that we don't find any invalid
365 * frames in whatever data is available, rather than requiring a full
366 * MIN_RESYNC_FRAMES of data.
368 * Returns TRUE if we've seen enough data to validate or reject the frame.
369 * If TRUE is returned, then *valid contains TRUE if it validated, or false
370 * if we decided it was false sync.
371 * If FALSE is returned, then *valid contains minimum needed data.
374 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
375 guint32 header, int bpf, gboolean at_eos, gint * valid)
380 int frames_found = 1;
383 gst_buffer_map (buf, &map, GST_MAP_READ);
385 while (frames_found < MIN_RESYNC_FRAMES) {
386 /* Check if we have enough data for all these frames, plus the next
388 if (map.size < offset + 4) {
390 /* Running out of data at EOS is fine; just accept it */
400 next_header = GST_READ_UINT32_BE (map.data + offset);
401 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
402 offset, (unsigned int) header, (unsigned int) next_header, bpf);
404 /* mask the bits which are allowed to differ between frames */
405 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
406 (0x1 << 9) /* padding */ | \
407 (0xf << 4) /* mode|mode extension */ | \
408 (0xf)) /* copyright|emphasis */
410 if ((next_header & HDRMASK) != (header & HDRMASK)) {
411 /* If any of the unmasked bits don't match, then it's not valid */
412 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
413 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
414 (guint) header, (guint) header & HDRMASK, (guint) next_header,
415 (guint) next_header & HDRMASK, bpf);
418 } else if ((((next_header >> 12) & 0xf) == 0) ||
419 (((next_header >> 12) & 0xf) == 0xf)) {
420 /* The essential parts were the same, but the bitrate held an
421 invalid value - also reject */
422 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
427 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
428 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
437 gst_buffer_unmap (buf, &map);
442 gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
445 GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
446 /* if it's not a valid sync */
447 if ((head & 0xffe00000) != 0xffe00000) {
448 GST_WARNING_OBJECT (mp3parse, "invalid sync");
451 /* if it's an invalid MPEG version */
452 if (((head >> 19) & 3) == 0x1) {
453 GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
457 /* if it's an invalid layer */
458 if (!((head >> 17) & 3)) {
459 GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
462 /* if it's an invalid bitrate */
463 if (((head >> 12) & 0xf) == 0x0) {
464 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx."
465 "Free format files are not supported yet", (head >> 12) & 0xf);
468 if (((head >> 12) & 0xf) == 0xf) {
469 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
472 /* if it's an invalid samplerate */
473 if (((head >> 10) & 0x3) == 0x3) {
474 GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
479 if ((head & 0x3) == 0x2) {
480 /* Ignore this as there are some files with emphasis 0x2 that can
481 * be played fine. See BGO #537235 */
482 GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
489 gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
490 GstBaseParseFrame * frame, guint * framesize, gint * skipsize)
492 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
493 GstBuffer *buf = frame->buffer;
494 GstByteReader reader;
496 gboolean lost_sync, draining, valid, caps_change;
498 guint bitrate, layer, rate, channels, version, mode, crc;
500 gboolean res = FALSE;
502 gst_buffer_map (buf, &map, GST_MAP_READ);
503 if (G_UNLIKELY (map.size < 6))
506 gst_byte_reader_init (&reader, map.data, map.size);
508 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
511 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
513 /* didn't find anything that looks like a sync word, skip */
515 *skipsize = map.size - 3;
519 /* possible frame header, but not at offset 0? skip bytes before sync */
525 /* make sure the values in the frame header look sane */
526 header = GST_READ_UINT32_BE (map.data);
527 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
532 GST_LOG_OBJECT (parse, "got frame");
534 bpf = mp3_type_frame_length_from_header (mp3parse, header,
535 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
538 if (channels != mp3parse->channels || rate != mp3parse->rate ||
539 layer != mp3parse->layer || version != mp3parse->version)
544 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
545 draining = GST_BASE_PARSE_DRAINING (parse);
547 if (!draining && (lost_sync || caps_change)) {
548 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
550 /* not enough data */
551 gst_base_parse_set_min_frame_size (parse, valid);
560 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
561 /* avoid caps jitter that we can't be sure of */
566 /* restore default minimum */
567 gst_base_parse_set_min_frame_size (parse, MIN_FRAME_SIZE);
573 gst_buffer_unmap (buf, &map);
578 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
581 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
582 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
583 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
584 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
585 gint offset_xing, offset_vbri;
587 gint64 upstream_total_bytes = 0;
588 guint32 read_id_xing = 0, read_id_vbri = 0;
593 if (mp3parse->sent_codec_tag)
596 /* Check first frame for Xing info */
597 if (mp3parse->version == 1) { /* MPEG-1 file */
598 if (mp3parse->channels == 1)
602 } else { /* MPEG-2 header */
603 if (mp3parse->channels == 1)
609 /* The VBRI tag is always at offset 0x20 */
612 /* Skip the 4 bytes of the MP3 header too */
616 /* Check if we have enough data to read the Xing header */
617 gst_buffer_map (buf, &map, GST_MAP_READ);
621 if (avail >= offset_xing + 4) {
622 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
624 if (avail >= offset_vbri + 4) {
625 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
628 /* obtain real upstream total bytes */
629 if (!gst_pad_peer_query_duration (GST_BASE_PARSE_SINK_PAD (mp3parse),
630 GST_FORMAT_BYTES, &upstream_total_bytes))
631 upstream_total_bytes = 0;
633 if (read_id_xing == xing_id || read_id_xing == info_id) {
635 guint bytes_needed = offset_xing + 8;
637 GstClockTime total_time;
639 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
641 /* Move data after Xing header */
642 data += offset_xing + 4;
644 /* Read 4 base bytes of flags, big-endian */
645 xing_flags = GST_READ_UINT32_BE (data);
647 if (xing_flags & XING_FRAMES_FLAG)
649 if (xing_flags & XING_BYTES_FLAG)
651 if (xing_flags & XING_TOC_FLAG)
653 if (xing_flags & XING_VBR_SCALE_FLAG)
655 if (avail < bytes_needed) {
656 GST_DEBUG_OBJECT (mp3parse,
657 "Not enough data to read Xing header (need %d)", bytes_needed);
661 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
662 mp3parse->xing_flags = xing_flags;
664 if (xing_flags & XING_FRAMES_FLAG) {
665 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
666 if (mp3parse->xing_frames == 0) {
667 GST_WARNING_OBJECT (mp3parse,
668 "Invalid number of frames in Xing header");
669 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
671 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
672 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
678 mp3parse->xing_frames = 0;
679 mp3parse->xing_total_time = 0;
682 if (xing_flags & XING_BYTES_FLAG) {
683 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
684 if (mp3parse->xing_bytes == 0) {
685 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
686 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
690 mp3parse->xing_bytes = 0;
693 /* If we know the upstream size and duration, compute the
694 * total bitrate, rounded up to the nearest kbit/sec */
695 if ((total_time = mp3parse->xing_total_time) &&
696 (total_bytes = mp3parse->xing_bytes)) {
697 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
698 8 * GST_SECOND, total_time);
699 mp3parse->xing_bitrate += 500;
700 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
703 if (xing_flags & XING_TOC_FLAG) {
705 guchar *table = mp3parse->xing_seek_table;
710 GST_DEBUG_OBJECT (mp3parse,
711 "Subtracting initial offset of %d bytes from Xing TOC", first);
713 /* xing seek table: percent time -> 1/256 bytepos */
714 for (i = 0; i < 100; i++) {
715 new = data[i] - first;
717 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
718 mp3parse->xing_flags &= ~XING_TOC_FLAG;
721 mp3parse->xing_seek_table[i] = old = new;
724 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
725 for (i = 0; i < 256; i++) {
726 while (percent < 99 && table[percent + 1] <= i)
729 if (table[percent] == i) {
730 mp3parse->xing_seek_table_inverse[i] = percent * 100;
731 } else if (table[percent] < i && percent < 99) {
733 gint a = percent, b = percent + 1;
737 fx = (b - a) / (fb - fa) * (i - fa) + a;
738 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
739 } else if (percent == 99) {
741 gint a = percent, b = 100;
745 fx = (b - a) / (fb - fa) * (i - fa) + a;
746 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
752 memset (mp3parse->xing_seek_table, 0, 100);
753 memset (mp3parse->xing_seek_table_inverse, 0, 256);
756 if (xing_flags & XING_VBR_SCALE_FLAG) {
757 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
760 mp3parse->xing_vbr_scale = 0;
762 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
763 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
764 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
765 mp3parse->xing_vbr_scale);
767 /* check for truncated file */
768 if (upstream_total_bytes && mp3parse->xing_bytes &&
769 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
770 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
771 "invalidating Xing header duration and size");
772 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
773 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
776 /* Optional LAME tag? */
777 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
778 gchar lame_version[10] = { 0, };
780 guint32 encoder_delay, encoder_padding;
782 memcpy (lame_version, data, 9);
784 tag_rev = data[0] >> 4;
785 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
786 tag_rev, lame_version);
788 /* Skip all the information we're not interested in */
790 /* Encoder delay and end padding */
791 encoder_delay = GST_READ_UINT24_BE (data);
792 encoder_delay >>= 12;
793 encoder_padding = GST_READ_UINT24_BE (data);
794 encoder_padding &= 0x000fff;
796 mp3parse->encoder_delay = encoder_delay;
797 mp3parse->encoder_padding = encoder_padding;
799 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
800 encoder_delay, encoder_padding);
804 if (read_id_vbri == vbri_id) {
805 gint64 total_bytes, total_frames;
806 GstClockTime total_time;
807 guint16 nseek_points;
809 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
811 if (avail < offset_vbri + 26) {
812 GST_DEBUG_OBJECT (mp3parse,
813 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
817 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
819 /* Move data after VBRI header */
820 data += offset_vbri + 4;
822 if (GST_READ_UINT16_BE (data) != 0x0001) {
823 GST_WARNING_OBJECT (mp3parse,
824 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
829 /* Skip encoder delay */
835 total_bytes = GST_READ_UINT32_BE (data);
836 if (total_bytes != 0)
837 mp3parse->vbri_bytes = total_bytes;
840 total_frames = GST_READ_UINT32_BE (data);
841 if (total_frames != 0) {
842 mp3parse->vbri_frames = total_frames;
843 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
844 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
848 /* If we know the upstream size and duration, compute the
849 * total bitrate, rounded up to the nearest kbit/sec */
850 if ((total_time = mp3parse->vbri_total_time) &&
851 (total_bytes = mp3parse->vbri_bytes)) {
852 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
853 8 * GST_SECOND, total_time);
854 mp3parse->vbri_bitrate += 500;
855 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
858 nseek_points = GST_READ_UINT16_BE (data);
861 if (nseek_points > 0) {
862 guint scale, seek_bytes, seek_frames;
865 mp3parse->vbri_seek_points = nseek_points;
867 scale = GST_READ_UINT16_BE (data);
870 seek_bytes = GST_READ_UINT16_BE (data);
873 seek_frames = GST_READ_UINT16_BE (data);
875 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
876 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
880 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
881 GST_WARNING_OBJECT (mp3parse,
882 "Not enough data to read VBRI seek table (need %d)",
883 offset_vbri + 26 + nseek_points * seek_bytes);
887 if (seek_frames * nseek_points < total_frames - seek_frames ||
888 seek_frames * nseek_points > total_frames + seek_frames) {
889 GST_WARNING_OBJECT (mp3parse,
890 "VBRI seek table doesn't cover the complete file");
894 if (avail < offset_vbri + 26) {
895 GST_DEBUG_OBJECT (mp3parse,
896 "Not enough data to read VBRI header (need %d)",
897 offset_vbri + 26 + nseek_points * seek_bytes);
902 data += offset_vbri + 26;
904 /* VBRI seek table: frame/seek_frames -> byte */
905 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
907 for (i = 0; i < nseek_points; i++) {
908 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
910 } else if (seek_bytes == 3)
911 for (i = 0; i < nseek_points; i++) {
912 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
914 } else if (seek_bytes == 2)
915 for (i = 0; i < nseek_points; i++) {
916 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
918 } else /* seek_bytes == 1 */
919 for (i = 0; i < nseek_points; i++) {
920 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
926 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
927 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
928 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
930 /* check for truncated file */
931 if (upstream_total_bytes && mp3parse->vbri_bytes &&
932 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
933 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
934 "invalidating VBRI header duration and size");
935 mp3parse->vbri_valid = FALSE;
937 mp3parse->vbri_valid = TRUE;
940 GST_DEBUG_OBJECT (mp3parse,
941 "Xing, LAME or VBRI header not found in first frame");
944 /* set duration if tables provided a valid one */
945 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
946 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
947 mp3parse->xing_total_time, 0);
949 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
950 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
951 mp3parse->vbri_total_time, 0);
954 /* tell baseclass how nicely we can seek, and a bitrate if one found */
955 /* FIXME: fill index with seek table */
957 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
958 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
959 mp3parse->xing_total_time)
960 seekable = GST_BASE_PARSE_SEEK_TABLE;
962 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
963 mp3parse->vbri_total_time)
964 seekable = GST_BASE_PARSE_SEEK_TABLE;
967 if (mp3parse->xing_bitrate)
968 bitrate = mp3parse->xing_bitrate;
969 else if (mp3parse->vbri_bitrate)
970 bitrate = mp3parse->vbri_bitrate;
974 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
977 gst_buffer_unmap (buf, &map);
981 gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
982 GstBaseParseFrame * frame)
984 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
985 GstBuffer *buf = frame->buffer;
987 guint bitrate, layer, rate, channels, version, mode, crc;
989 gst_buffer_map (buf, &map, GST_MAP_READ);
990 if (G_UNLIKELY (map.size < 4))
993 if (!mp3_type_frame_length_from_header (mp3parse,
994 GST_READ_UINT32_BE (map.data),
995 &version, &layer, &channels, &bitrate, &rate, &mode, &crc))
998 if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
999 layer != mp3parse->layer || version != mp3parse->version)) {
1000 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
1001 "mpegversion", G_TYPE_INT, 1,
1002 "mpegaudioversion", G_TYPE_INT, version,
1003 "layer", G_TYPE_INT, layer,
1004 "rate", G_TYPE_INT, rate,
1005 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
1006 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
1007 gst_caps_unref (caps);
1009 mp3parse->rate = rate;
1010 mp3parse->channels = channels;
1011 mp3parse->layer = layer;
1012 mp3parse->version = version;
1014 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
1015 if (mp3parse->layer == 1)
1016 mp3parse->spf = 384;
1017 else if (mp3parse->layer == 2)
1018 mp3parse->spf = 1152;
1019 else if (mp3parse->version == 1) {
1020 mp3parse->spf = 1152;
1022 /* MPEG-2 or "2.5" */
1023 mp3parse->spf = 576;
1027 * We start pushing 9 frames earlier (29 frames for MPEG2) than
1028 * segment start to be able to decode the first frame we want.
1029 * 9 (29) frames are the theoretical maximum of frames that contain
1030 * data for the current frame (bit reservoir).
1033 * Some mp3 streams have an offset in the timestamps, for which we have to
1034 * push the frame *after* the end position in order for the decoder to be
1035 * able to decode everything up until the segment.stop position. */
1036 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
1037 (version == 1) ? 10 : 30, 2);
1040 mp3parse->hdr_bitrate = bitrate;
1042 /* For first frame; check for seek tables and output a codec tag */
1043 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
1045 /* store some frame info for later processing */
1046 mp3parse->last_crc = crc;
1047 mp3parse->last_mode = mode;
1049 gst_buffer_unmap (buf, &map);
1055 /* this really shouldn't ever happen */
1056 gst_buffer_unmap (buf, &map);
1057 GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
1058 return GST_FLOW_ERROR;
1063 gst_buffer_unmap (buf, &map);
1064 return GST_FLOW_ERROR;
1069 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1070 GstClockTime ts, gint64 * bytepos)
1073 GstClockTime total_time;
1075 /* If XING seek table exists use this for time->byte conversion */
1076 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1077 (total_bytes = mp3parse->xing_bytes) &&
1078 (total_time = mp3parse->xing_total_time)) {
1081 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1082 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1083 gint index = CLAMP (percent, 0, 99);
1085 fa = mp3parse->xing_seek_table[index];
1087 fb = mp3parse->xing_seek_table[index + 1];
1091 fx = fa + (fb - fa) * (percent - index);
1093 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1098 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1099 (total_time = mp3parse->vbri_total_time)) {
1101 gdouble a, b, fa, fb;
1103 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1104 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1106 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1107 mp3parse->vbri_seek_points));
1109 for (j = i; j >= 0; j--)
1110 fa += mp3parse->vbri_seek_table[j];
1112 if (i + 1 < mp3parse->vbri_seek_points) {
1113 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1114 mp3parse->vbri_seek_points));
1115 fb = fa + mp3parse->vbri_seek_table[i + 1];
1117 b = gst_guint64_to_gdouble (total_time);
1121 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1130 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1131 gint64 bytepos, GstClockTime * ts)
1134 GstClockTime total_time;
1136 /* If XING seek table exists use this for byte->time conversion */
1137 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1138 (total_bytes = mp3parse->xing_bytes) &&
1139 (total_time = mp3parse->xing_total_time)) {
1144 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1145 index = CLAMP (pos, 0, 255);
1146 fa = mp3parse->xing_seek_table_inverse[index];
1148 fb = mp3parse->xing_seek_table_inverse[index + 1];
1152 fx = fa + (fb - fa) * (pos - index);
1154 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1159 if (mp3parse->vbri_seek_table &&
1160 (total_bytes = mp3parse->vbri_bytes) &&
1161 (total_time = mp3parse->vbri_total_time)) {
1164 gdouble a, b, fa, fb;
1167 sum += mp3parse->vbri_seek_table[i];
1169 } while (i + 1 < mp3parse->vbri_seek_points
1170 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1173 a = gst_guint64_to_gdouble (sum);
1174 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1175 mp3parse->vbri_seek_points));
1177 if (i + 1 < mp3parse->vbri_seek_points) {
1178 b = a + mp3parse->vbri_seek_table[i + 1];
1179 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1180 mp3parse->vbri_seek_points));
1183 fb = gst_guint64_to_gdouble (total_time);
1186 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1195 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1196 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1198 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1199 gboolean res = FALSE;
1201 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1203 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1204 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1205 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1206 (GstClockTime *) dest_value);
1208 /* if no tables, fall back to default estimated rate based conversion */
1210 return gst_base_parse_convert_default (parse, src_format, src_value,
1211 dest_format, dest_value);
1216 static GstFlowReturn
1217 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1218 GstBaseParseFrame * frame)
1220 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1221 GstTagList *taglist;
1223 /* tag sending done late enough in hook to ensure pending events
1224 * have already been sent */
1226 if (!mp3parse->sent_codec_tag) {
1230 if (mp3parse->layer == 3) {
1231 codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1232 mp3parse->version, mp3parse->layer);
1234 codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1235 mp3parse->version, mp3parse->layer);
1237 taglist = gst_tag_list_new (GST_TAG_AUDIO_CODEC, codec, NULL);
1238 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1239 mp3parse->vbri_bitrate == 0) {
1240 /* We don't have a VBR bitrate, so post the available bitrate as
1241 * nominal and let baseparse calculate the real bitrate */
1242 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1243 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1245 gst_pad_push_event (GST_BASE_PARSE_SRC_PAD (mp3parse),
1246 gst_event_new_tag (taglist));
1249 /* also signals the end of first-frame processing */
1250 mp3parse->sent_codec_tag = TRUE;
1253 /* we will create a taglist (if any of the parameters has changed)
1254 * to add the tags that changed */
1256 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1260 taglist = gst_tag_list_new_empty ();
1262 mp3parse->last_posted_crc = mp3parse->last_crc;
1263 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1268 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1272 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1274 taglist = gst_tag_list_new_empty ();
1276 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1278 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1279 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1282 /* if the taglist exists, we need to send it */
1284 gst_pad_push_event (GST_BASE_PARSE_SRC_PAD (mp3parse),
1285 gst_event_new_tag (taglist));
1288 /* usual clipping applies */
1289 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;
1295 gst_mpeg_audio_parse_get_sink_caps (GstBaseParse * parse, GstCaps * filter)
1300 /* FIXME: handle filter caps */
1302 peercaps = gst_pad_get_allowed_caps (GST_BASE_PARSE_SRC_PAD (parse));
1306 /* Remove the parsed field */
1307 peercaps = gst_caps_make_writable (peercaps);
1308 n = gst_caps_get_size (peercaps);
1309 for (i = 0; i < n; i++) {
1310 GstStructure *s = gst_caps_get_structure (peercaps, i);
1312 gst_structure_remove_field (s, "parsed");
1316 gst_caps_intersect_full (peercaps,
1317 gst_pad_get_pad_template_caps (GST_BASE_PARSE_SRC_PAD (parse)),
1318 GST_CAPS_INTERSECT_FIRST);
1319 gst_caps_unref (peercaps);
1322 gst_caps_copy (gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD