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);
104 #define gst_mpeg_audio_parse_parent_class parent_class
105 G_DEFINE_TYPE (GstMpegAudioParse, gst_mpeg_audio_parse, GST_TYPE_BASE_PARSE);
107 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
108 (gst_mpeg_audio_channel_mode_get_type())
110 static const GEnumValue mpeg_audio_channel_mode[] = {
111 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
112 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
113 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
114 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
115 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
120 gst_mpeg_audio_channel_mode_get_type (void)
122 static GType mpeg_audio_channel_mode_type = 0;
124 if (!mpeg_audio_channel_mode_type) {
125 mpeg_audio_channel_mode_type =
126 g_enum_register_static ("GstMpegAudioChannelMode",
127 mpeg_audio_channel_mode);
129 return mpeg_audio_channel_mode_type;
133 gst_mpeg_audio_channel_mode_get_nick (gint mode)
136 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
137 if (mpeg_audio_channel_mode[i].value == mode)
138 return mpeg_audio_channel_mode[i].value_nick;
144 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
146 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
147 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
148 GObjectClass *object_class = G_OBJECT_CLASS (klass);
150 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
151 "MPEG1 audio stream parser");
153 object_class->finalize = gst_mpeg_audio_parse_finalize;
155 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
156 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
157 parse_class->check_valid_frame =
158 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
159 parse_class->parse_frame =
160 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
161 parse_class->pre_push_frame =
162 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
163 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
166 #define GST_TAG_CRC "has-crc"
167 #define GST_TAG_MODE "channel-mode"
169 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
170 "has crc", "Using CRC", NULL);
171 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
172 "channel mode", "MPEG audio channel mode", NULL);
174 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
176 gst_element_class_add_pad_template (element_class,
177 gst_static_pad_template_get (&sink_template));
178 gst_element_class_add_pad_template (element_class,
179 gst_static_pad_template_get (&src_template));
181 gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
182 "Codec/Parser/Audio",
183 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
184 "Jan Schmidt <thaytan@mad.scientist.com>,"
185 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
189 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
191 mp3parse->channels = -1;
193 mp3parse->sent_codec_tag = FALSE;
194 mp3parse->last_posted_crc = CRC_UNKNOWN;
195 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
197 mp3parse->hdr_bitrate = 0;
199 mp3parse->xing_flags = 0;
200 mp3parse->xing_bitrate = 0;
201 mp3parse->xing_frames = 0;
202 mp3parse->xing_total_time = 0;
203 mp3parse->xing_bytes = 0;
204 mp3parse->xing_vbr_scale = 0;
205 memset (mp3parse->xing_seek_table, 0, 100);
206 memset (mp3parse->xing_seek_table_inverse, 0, 256);
208 mp3parse->vbri_bitrate = 0;
209 mp3parse->vbri_frames = 0;
210 mp3parse->vbri_total_time = 0;
211 mp3parse->vbri_bytes = 0;
212 mp3parse->vbri_seek_points = 0;
213 g_free (mp3parse->vbri_seek_table);
214 mp3parse->vbri_seek_table = NULL;
216 mp3parse->encoder_delay = 0;
217 mp3parse->encoder_padding = 0;
221 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse)
223 gst_mpeg_audio_parse_reset (mp3parse);
227 gst_mpeg_audio_parse_finalize (GObject * object)
229 G_OBJECT_CLASS (parent_class)->finalize (object);
233 gst_mpeg_audio_parse_start (GstBaseParse * parse)
235 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
237 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), MIN_FRAME_SIZE);
238 GST_DEBUG_OBJECT (parse, "starting");
240 gst_mpeg_audio_parse_reset (mp3parse);
246 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
248 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
250 GST_DEBUG_OBJECT (parse, "stopping");
252 gst_mpeg_audio_parse_reset (mp3parse);
257 static const guint mp3types_bitrates[2][3][16] = {
259 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
260 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
261 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
264 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
265 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
266 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
270 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
271 {22050, 24000, 16000},
276 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
277 guint * put_version, guint * put_layer, guint * put_channels,
278 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
282 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
286 if (header & (1 << 20)) {
287 lsf = (header & (1 << 19)) ? 0 : 1;
294 version = 1 + lsf + mpg25;
296 layer = 4 - ((header >> 17) & 0x3);
298 crc = (header >> 16) & 0x1;
300 bitrate = (header >> 12) & 0xF;
301 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
302 /* The caller has ensured we have a valid header, so bitrate can't be
304 g_assert (bitrate != 0);
306 samplerate = (header >> 10) & 0x3;
307 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
309 padding = (header >> 9) & 0x1;
311 mode = (header >> 6) & 0x3;
312 channels = (mode == 3) ? 1 : 2;
316 length = 4 * ((bitrate * 12) / samplerate + padding);
319 length = (bitrate * 144) / samplerate + padding;
323 length = (bitrate * 144) / (samplerate << lsf) + padding;
327 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
329 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
330 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
331 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
334 *put_version = version;
338 *put_channels = channels;
340 *put_bitrate = bitrate;
342 *put_samplerate = samplerate;
351 /* Minimum number of consecutive, valid-looking frames to consider
353 #define MIN_RESYNC_FRAMES 3
355 /* Perform extended validation to check that subsequent headers match
356 * the first header given here in important characteristics, to avoid
357 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
358 * frames to match their major characteristics.
360 * If at_eos is set to TRUE, we just check that we don't find any invalid
361 * frames in whatever data is available, rather than requiring a full
362 * MIN_RESYNC_FRAMES of data.
364 * Returns TRUE if we've seen enough data to validate or reject the frame.
365 * If TRUE is returned, then *valid contains TRUE if it validated, or false
366 * if we decided it was false sync.
367 * If FALSE is returned, then *valid contains minimum needed data.
370 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
371 guint32 header, int bpf, gboolean at_eos, gint * valid)
377 int frames_found = 1;
380 data = gst_buffer_map (buf, &available, NULL, GST_MAP_READ);
382 while (frames_found < MIN_RESYNC_FRAMES) {
383 /* Check if we have enough data for all these frames, plus the next
385 if (available < offset + 4) {
387 /* Running out of data at EOS is fine; just accept it */
397 next_header = GST_READ_UINT32_BE (data + offset);
398 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
399 offset, (unsigned int) header, (unsigned int) next_header, bpf);
401 /* mask the bits which are allowed to differ between frames */
402 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
403 (0x1 << 9) /* padding */ | \
404 (0xf << 4) /* mode|mode extension */ | \
405 (0xf)) /* copyright|emphasis */
407 if ((next_header & HDRMASK) != (header & HDRMASK)) {
408 /* If any of the unmasked bits don't match, then it's not valid */
409 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
410 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
411 (guint) header, (guint) header & HDRMASK, (guint) next_header,
412 (guint) next_header & HDRMASK, bpf);
415 } else if ((((next_header >> 12) & 0xf) == 0) ||
416 (((next_header >> 12) & 0xf) == 0xf)) {
417 /* The essential parts were the same, but the bitrate held an
418 invalid value - also reject */
419 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
424 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
425 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
434 gst_buffer_unmap (buf, data, available);
439 gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
442 GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
443 /* if it's not a valid sync */
444 if ((head & 0xffe00000) != 0xffe00000) {
445 GST_WARNING_OBJECT (mp3parse, "invalid sync");
448 /* if it's an invalid MPEG version */
449 if (((head >> 19) & 3) == 0x1) {
450 GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
454 /* if it's an invalid layer */
455 if (!((head >> 17) & 3)) {
456 GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
459 /* if it's an invalid bitrate */
460 if (((head >> 12) & 0xf) == 0x0) {
461 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx."
462 "Free format files are not supported yet", (head >> 12) & 0xf);
465 if (((head >> 12) & 0xf) == 0xf) {
466 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
469 /* if it's an invalid samplerate */
470 if (((head >> 10) & 0x3) == 0x3) {
471 GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
476 if ((head & 0x3) == 0x2) {
477 /* Ignore this as there are some files with emphasis 0x2 that can
478 * be played fine. See BGO #537235 */
479 GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
486 gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
487 GstBaseParseFrame * frame, guint * framesize, gint * skipsize)
489 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
490 GstBuffer *buf = frame->buffer;
491 GstByteReader reader;
493 gboolean lost_sync, draining, valid, caps_change;
495 guint bitrate, layer, rate, channels, version, mode, crc;
498 gboolean res = FALSE;
500 data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
501 if (G_UNLIKELY (bufsize < 6))
504 gst_byte_reader_init (&reader, data, bufsize);
506 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
509 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
511 /* didn't find anything that looks like a sync word, skip */
513 *skipsize = bufsize - 3;
517 /* possible frame header, but not at offset 0? skip bytes before sync */
523 /* make sure the values in the frame header look sane */
524 header = GST_READ_UINT32_BE (data);
525 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
530 GST_LOG_OBJECT (parse, "got frame");
532 bpf = mp3_type_frame_length_from_header (mp3parse, header,
533 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
536 if (channels != mp3parse->channels || rate != mp3parse->rate ||
537 layer != mp3parse->layer || version != mp3parse->version)
542 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
543 draining = GST_BASE_PARSE_DRAINING (parse);
545 if (!draining && (lost_sync || caps_change)) {
546 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
548 /* not enough data */
549 gst_base_parse_set_min_frame_size (parse, valid);
558 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
559 /* avoid caps jitter that we can't be sure of */
564 /* restore default minimum */
565 gst_base_parse_set_min_frame_size (parse, MIN_FRAME_SIZE);
571 gst_buffer_unmap (buf, data, bufsize);
576 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
579 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
580 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
581 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
582 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
583 gint offset_xing, offset_vbri;
585 gint64 upstream_total_bytes = 0;
586 guint32 read_id_xing = 0, read_id_vbri = 0;
587 guint8 *data, *origdata;
591 if (mp3parse->sent_codec_tag)
594 /* Check first frame for Xing info */
595 if (mp3parse->version == 1) { /* MPEG-1 file */
596 if (mp3parse->channels == 1)
600 } else { /* MPEG-2 header */
601 if (mp3parse->channels == 1)
607 /* The VBRI tag is always at offset 0x20 */
610 /* Skip the 4 bytes of the MP3 header too */
614 /* Check if we have enough data to read the Xing header */
615 origdata = data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
618 if (avail >= offset_xing + 4) {
619 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
621 if (avail >= offset_vbri + 4) {
622 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
625 /* obtain real upstream total bytes */
626 if (!gst_pad_query_peer_duration (GST_BASE_PARSE_SINK_PAD (mp3parse),
627 GST_FORMAT_BYTES, &upstream_total_bytes))
628 upstream_total_bytes = 0;
630 if (read_id_xing == xing_id || read_id_xing == info_id) {
632 guint bytes_needed = offset_xing + 8;
634 GstClockTime total_time;
636 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
638 /* Move data after Xing header */
639 data += offset_xing + 4;
641 /* Read 4 base bytes of flags, big-endian */
642 xing_flags = GST_READ_UINT32_BE (data);
644 if (xing_flags & XING_FRAMES_FLAG)
646 if (xing_flags & XING_BYTES_FLAG)
648 if (xing_flags & XING_TOC_FLAG)
650 if (xing_flags & XING_VBR_SCALE_FLAG)
652 if (avail < bytes_needed) {
653 GST_DEBUG_OBJECT (mp3parse,
654 "Not enough data to read Xing header (need %d)", bytes_needed);
658 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
659 mp3parse->xing_flags = xing_flags;
661 if (xing_flags & XING_FRAMES_FLAG) {
662 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
663 if (mp3parse->xing_frames == 0) {
664 GST_WARNING_OBJECT (mp3parse,
665 "Invalid number of frames in Xing header");
666 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
668 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
669 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
675 mp3parse->xing_frames = 0;
676 mp3parse->xing_total_time = 0;
679 if (xing_flags & XING_BYTES_FLAG) {
680 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
681 if (mp3parse->xing_bytes == 0) {
682 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
683 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
687 mp3parse->xing_bytes = 0;
690 /* If we know the upstream size and duration, compute the
691 * total bitrate, rounded up to the nearest kbit/sec */
692 if ((total_time = mp3parse->xing_total_time) &&
693 (total_bytes = mp3parse->xing_bytes)) {
694 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
695 8 * GST_SECOND, total_time);
696 mp3parse->xing_bitrate += 500;
697 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
700 if (xing_flags & XING_TOC_FLAG) {
702 guchar *table = mp3parse->xing_seek_table;
707 GST_DEBUG_OBJECT (mp3parse,
708 "Subtracting initial offset of %d bytes from Xing TOC", first);
710 /* xing seek table: percent time -> 1/256 bytepos */
711 for (i = 0; i < 100; i++) {
712 new = data[i] - first;
714 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
715 mp3parse->xing_flags &= ~XING_TOC_FLAG;
718 mp3parse->xing_seek_table[i] = old = new;
721 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
722 for (i = 0; i < 256; i++) {
723 while (percent < 99 && table[percent + 1] <= i)
726 if (table[percent] == i) {
727 mp3parse->xing_seek_table_inverse[i] = percent * 100;
728 } else if (table[percent] < i && percent < 99) {
730 gint a = percent, b = percent + 1;
734 fx = (b - a) / (fb - fa) * (i - fa) + a;
735 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
736 } else if (percent == 99) {
738 gint a = percent, b = 100;
742 fx = (b - a) / (fb - fa) * (i - fa) + a;
743 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
749 memset (mp3parse->xing_seek_table, 0, 100);
750 memset (mp3parse->xing_seek_table_inverse, 0, 256);
753 if (xing_flags & XING_VBR_SCALE_FLAG) {
754 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
757 mp3parse->xing_vbr_scale = 0;
759 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
760 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
761 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
762 mp3parse->xing_vbr_scale);
764 /* check for truncated file */
765 if (upstream_total_bytes && mp3parse->xing_bytes &&
766 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
767 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
768 "invalidating Xing header duration and size");
769 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
770 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
773 /* Optional LAME tag? */
774 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
775 gchar lame_version[10] = { 0, };
777 guint32 encoder_delay, encoder_padding;
779 memcpy (lame_version, data, 9);
781 tag_rev = data[0] >> 4;
782 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
783 tag_rev, lame_version);
785 /* Skip all the information we're not interested in */
787 /* Encoder delay and end padding */
788 encoder_delay = GST_READ_UINT24_BE (data);
789 encoder_delay >>= 12;
790 encoder_padding = GST_READ_UINT24_BE (data);
791 encoder_padding &= 0x000fff;
793 mp3parse->encoder_delay = encoder_delay;
794 mp3parse->encoder_padding = encoder_padding;
796 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
797 encoder_delay, encoder_padding);
801 if (read_id_vbri == vbri_id) {
802 gint64 total_bytes, total_frames;
803 GstClockTime total_time;
804 guint16 nseek_points;
806 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
808 if (avail < offset_vbri + 26) {
809 GST_DEBUG_OBJECT (mp3parse,
810 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
814 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
816 /* Move data after VBRI header */
817 data += offset_vbri + 4;
819 if (GST_READ_UINT16_BE (data) != 0x0001) {
820 GST_WARNING_OBJECT (mp3parse,
821 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
826 /* Skip encoder delay */
832 total_bytes = GST_READ_UINT32_BE (data);
833 if (total_bytes != 0)
834 mp3parse->vbri_bytes = total_bytes;
837 total_frames = GST_READ_UINT32_BE (data);
838 if (total_frames != 0) {
839 mp3parse->vbri_frames = total_frames;
840 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
841 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
845 /* If we know the upstream size and duration, compute the
846 * total bitrate, rounded up to the nearest kbit/sec */
847 if ((total_time = mp3parse->vbri_total_time) &&
848 (total_bytes = mp3parse->vbri_bytes)) {
849 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
850 8 * GST_SECOND, total_time);
851 mp3parse->vbri_bitrate += 500;
852 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
855 nseek_points = GST_READ_UINT16_BE (data);
858 if (nseek_points > 0) {
859 guint scale, seek_bytes, seek_frames;
862 mp3parse->vbri_seek_points = nseek_points;
864 scale = GST_READ_UINT16_BE (data);
867 seek_bytes = GST_READ_UINT16_BE (data);
870 seek_frames = GST_READ_UINT16_BE (data);
872 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
873 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
877 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
878 GST_WARNING_OBJECT (mp3parse,
879 "Not enough data to read VBRI seek table (need %d)",
880 offset_vbri + 26 + nseek_points * seek_bytes);
884 if (seek_frames * nseek_points < total_frames - seek_frames ||
885 seek_frames * nseek_points > total_frames + seek_frames) {
886 GST_WARNING_OBJECT (mp3parse,
887 "VBRI seek table doesn't cover the complete file");
891 if (avail < offset_vbri + 26) {
892 GST_DEBUG_OBJECT (mp3parse,
893 "Not enough data to read VBRI header (need %d)",
894 offset_vbri + 26 + nseek_points * seek_bytes);
899 data += offset_vbri + 26;
901 /* VBRI seek table: frame/seek_frames -> byte */
902 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
904 for (i = 0; i < nseek_points; i++) {
905 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
907 } else if (seek_bytes == 3)
908 for (i = 0; i < nseek_points; i++) {
909 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
911 } else if (seek_bytes == 2)
912 for (i = 0; i < nseek_points; i++) {
913 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
915 } else /* seek_bytes == 1 */
916 for (i = 0; i < nseek_points; i++) {
917 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
923 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
924 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
925 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
927 /* check for truncated file */
928 if (upstream_total_bytes && mp3parse->vbri_bytes &&
929 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
930 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
931 "invalidating VBRI header duration and size");
932 mp3parse->vbri_valid = FALSE;
934 mp3parse->vbri_valid = TRUE;
937 GST_DEBUG_OBJECT (mp3parse,
938 "Xing, LAME or VBRI header not found in first frame");
941 /* set duration if tables provided a valid one */
942 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
943 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
944 mp3parse->xing_total_time, 0);
946 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
947 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
948 mp3parse->vbri_total_time, 0);
951 /* tell baseclass how nicely we can seek, and a bitrate if one found */
952 /* FIXME: fill index with seek table */
954 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
955 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
956 mp3parse->xing_total_time)
957 seekable = GST_BASE_PARSE_SEEK_TABLE;
959 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
960 mp3parse->vbri_total_time)
961 seekable = GST_BASE_PARSE_SEEK_TABLE;
964 if (mp3parse->xing_bitrate)
965 bitrate = mp3parse->xing_bitrate;
966 else if (mp3parse->vbri_bitrate)
967 bitrate = mp3parse->vbri_bitrate;
971 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
974 gst_buffer_unmap (buf, origdata, bufsize);
978 gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
979 GstBaseParseFrame * frame)
981 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
982 GstBuffer *buf = frame->buffer;
985 guint bitrate, layer, rate, channels, version, mode, crc;
987 data = gst_buffer_map (buf, &bufsize, NULL, GST_MAP_READ);
988 if (G_UNLIKELY (bufsize < 4))
991 if (!mp3_type_frame_length_from_header (mp3parse,
992 GST_READ_UINT32_BE (data),
993 &version, &layer, &channels, &bitrate, &rate, &mode, &crc))
996 if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
997 layer != mp3parse->layer || version != mp3parse->version)) {
998 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
999 "mpegversion", G_TYPE_INT, 1,
1000 "mpegaudioversion", G_TYPE_INT, version,
1001 "layer", G_TYPE_INT, layer,
1002 "rate", G_TYPE_INT, rate,
1003 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
1004 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
1005 gst_caps_unref (caps);
1007 mp3parse->rate = rate;
1008 mp3parse->channels = channels;
1009 mp3parse->layer = layer;
1010 mp3parse->version = version;
1012 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
1013 if (mp3parse->layer == 1)
1014 mp3parse->spf = 384;
1015 else if (mp3parse->layer == 2)
1016 mp3parse->spf = 1152;
1017 else if (mp3parse->version == 1) {
1018 mp3parse->spf = 1152;
1020 /* MPEG-2 or "2.5" */
1021 mp3parse->spf = 576;
1025 * We start pushing 9 frames earlier (29 frames for MPEG2) than
1026 * segment start to be able to decode the first frame we want.
1027 * 9 (29) frames are the theoretical maximum of frames that contain
1028 * data for the current frame (bit reservoir).
1031 * Some mp3 streams have an offset in the timestamps, for which we have to
1032 * push the frame *after* the end position in order for the decoder to be
1033 * able to decode everything up until the segment.stop position. */
1034 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
1035 (version == 1) ? 10 : 30, 2);
1038 mp3parse->hdr_bitrate = bitrate;
1040 /* For first frame; check for seek tables and output a codec tag */
1041 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
1043 /* store some frame info for later processing */
1044 mp3parse->last_crc = crc;
1045 mp3parse->last_mode = mode;
1047 gst_buffer_unmap (buf, data, bufsize);
1053 /* this really shouldn't ever happen */
1054 gst_buffer_unmap (buf, data, bufsize);
1055 GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
1056 return GST_FLOW_ERROR;
1061 gst_buffer_unmap (buf, data, bufsize);
1062 return GST_FLOW_ERROR;
1067 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1068 GstClockTime ts, gint64 * bytepos)
1071 GstClockTime total_time;
1073 /* If XING seek table exists use this for time->byte conversion */
1074 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1075 (total_bytes = mp3parse->xing_bytes) &&
1076 (total_time = mp3parse->xing_total_time)) {
1079 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1080 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1081 gint index = CLAMP (percent, 0, 99);
1083 fa = mp3parse->xing_seek_table[index];
1085 fb = mp3parse->xing_seek_table[index + 1];
1089 fx = fa + (fb - fa) * (percent - index);
1091 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1096 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1097 (total_time = mp3parse->vbri_total_time)) {
1099 gdouble a, b, fa, fb;
1101 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1102 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1104 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1105 mp3parse->vbri_seek_points));
1107 for (j = i; j >= 0; j--)
1108 fa += mp3parse->vbri_seek_table[j];
1110 if (i + 1 < mp3parse->vbri_seek_points) {
1111 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1112 mp3parse->vbri_seek_points));
1113 fb = fa + mp3parse->vbri_seek_table[i + 1];
1115 b = gst_guint64_to_gdouble (total_time);
1119 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1128 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1129 gint64 bytepos, GstClockTime * ts)
1132 GstClockTime total_time;
1134 /* If XING seek table exists use this for byte->time conversion */
1135 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1136 (total_bytes = mp3parse->xing_bytes) &&
1137 (total_time = mp3parse->xing_total_time)) {
1142 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1143 index = CLAMP (pos, 0, 255);
1144 fa = mp3parse->xing_seek_table_inverse[index];
1146 fb = mp3parse->xing_seek_table_inverse[index + 1];
1150 fx = fa + (fb - fa) * (pos - index);
1152 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1157 if (mp3parse->vbri_seek_table &&
1158 (total_bytes = mp3parse->vbri_bytes) &&
1159 (total_time = mp3parse->vbri_total_time)) {
1162 gdouble a, b, fa, fb;
1165 sum += mp3parse->vbri_seek_table[i];
1167 } while (i + 1 < mp3parse->vbri_seek_points
1168 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1171 a = gst_guint64_to_gdouble (sum);
1172 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1173 mp3parse->vbri_seek_points));
1175 if (i + 1 < mp3parse->vbri_seek_points) {
1176 b = a + mp3parse->vbri_seek_table[i + 1];
1177 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1178 mp3parse->vbri_seek_points));
1181 fb = gst_guint64_to_gdouble (total_time);
1184 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1193 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1194 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1196 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1197 gboolean res = FALSE;
1199 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1201 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1202 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1203 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1204 (GstClockTime *) dest_value);
1206 /* if no tables, fall back to default estimated rate based conversion */
1208 return gst_base_parse_convert_default (parse, src_format, src_value,
1209 dest_format, dest_value);
1214 static GstFlowReturn
1215 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1216 GstBaseParseFrame * frame)
1218 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1219 GstTagList *taglist;
1221 /* tag sending done late enough in hook to ensure pending events
1222 * have already been sent */
1224 if (!mp3parse->sent_codec_tag) {
1228 if (mp3parse->layer == 3) {
1229 codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1230 mp3parse->version, mp3parse->layer);
1232 codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1233 mp3parse->version, mp3parse->layer);
1235 taglist = gst_tag_list_new (GST_TAG_AUDIO_CODEC, codec, NULL);
1236 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1237 mp3parse->vbri_bitrate == 0) {
1238 /* We don't have a VBR bitrate, so post the available bitrate as
1239 * nominal and let baseparse calculate the real bitrate */
1240 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1241 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1243 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1244 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1247 /* also signals the end of first-frame processing */
1248 mp3parse->sent_codec_tag = TRUE;
1251 /* we will create a taglist (if any of the parameters has changed)
1252 * to add the tags that changed */
1254 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1258 taglist = gst_tag_list_new_empty ();
1260 mp3parse->last_posted_crc = mp3parse->last_crc;
1261 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1266 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1270 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1272 taglist = gst_tag_list_new_empty ();
1274 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1276 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1277 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1280 /* if the taglist exists, we need to send it */
1282 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1283 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1286 /* usual clipping applies */
1287 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;