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);
105 GST_BOILERPLATE (GstMpegAudioParse, gst_mpeg_audio_parse, GstBaseParse,
106 GST_TYPE_BASE_PARSE);
108 #define GST_TYPE_MPEG_AUDIO_CHANNEL_MODE \
109 (gst_mpeg_audio_channel_mode_get_type())
111 static const GEnumValue mpeg_audio_channel_mode[] = {
112 {MPEG_AUDIO_CHANNEL_MODE_UNKNOWN, "Unknown", "unknown"},
113 {MPEG_AUDIO_CHANNEL_MODE_MONO, "Mono", "mono"},
114 {MPEG_AUDIO_CHANNEL_MODE_DUAL_CHANNEL, "Dual Channel", "dual-channel"},
115 {MPEG_AUDIO_CHANNEL_MODE_JOINT_STEREO, "Joint Stereo", "joint-stereo"},
116 {MPEG_AUDIO_CHANNEL_MODE_STEREO, "Stereo", "stereo"},
121 gst_mpeg_audio_channel_mode_get_type (void)
123 static GType mpeg_audio_channel_mode_type = 0;
125 if (!mpeg_audio_channel_mode_type) {
126 mpeg_audio_channel_mode_type =
127 g_enum_register_static ("GstMpegAudioChannelMode",
128 mpeg_audio_channel_mode);
130 return mpeg_audio_channel_mode_type;
134 gst_mpeg_audio_channel_mode_get_nick (gint mode)
137 for (i = 0; i < G_N_ELEMENTS (mpeg_audio_channel_mode); i++) {
138 if (mpeg_audio_channel_mode[i].value == mode)
139 return mpeg_audio_channel_mode[i].value_nick;
145 gst_mpeg_audio_parse_base_init (gpointer klass)
147 GstElementClass *element_class = GST_ELEMENT_CLASS (klass);
149 gst_element_class_add_static_pad_template (element_class,
151 gst_element_class_add_static_pad_template (element_class, &src_template);
153 gst_element_class_set_details_simple (element_class, "MPEG1 Audio Parser",
154 "Codec/Parser/Audio",
155 "Parses and frames mpeg1 audio streams (levels 1-3), provides seek",
156 "Jan Schmidt <thaytan@mad.scientist.com>,"
157 "Mark Nauwelaerts <mark.nauwelaerts@collabora.co.uk>");
161 gst_mpeg_audio_parse_class_init (GstMpegAudioParseClass * klass)
163 GstBaseParseClass *parse_class = GST_BASE_PARSE_CLASS (klass);
164 GObjectClass *object_class = G_OBJECT_CLASS (klass);
166 GST_DEBUG_CATEGORY_INIT (mpeg_audio_parse_debug, "mpegaudioparse", 0,
167 "MPEG1 audio stream parser");
169 object_class->finalize = gst_mpeg_audio_parse_finalize;
171 parse_class->start = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_start);
172 parse_class->stop = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_stop);
173 parse_class->check_valid_frame =
174 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_check_valid_frame);
175 parse_class->parse_frame =
176 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_parse_frame);
177 parse_class->pre_push_frame =
178 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_pre_push_frame);
179 parse_class->convert = GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_convert);
180 parse_class->get_sink_caps =
181 GST_DEBUG_FUNCPTR (gst_mpeg_audio_parse_get_sink_caps);
184 #define GST_TAG_CRC "has-crc"
185 #define GST_TAG_MODE "channel-mode"
187 gst_tag_register (GST_TAG_CRC, GST_TAG_FLAG_META, G_TYPE_BOOLEAN,
188 "has crc", "Using CRC", NULL);
189 gst_tag_register (GST_TAG_MODE, GST_TAG_FLAG_ENCODED, G_TYPE_STRING,
190 "channel mode", "MPEG audio channel mode", NULL);
192 g_type_class_ref (GST_TYPE_MPEG_AUDIO_CHANNEL_MODE);
196 gst_mpeg_audio_parse_reset (GstMpegAudioParse * mp3parse)
198 mp3parse->channels = -1;
200 mp3parse->sent_codec_tag = FALSE;
201 mp3parse->last_posted_crc = CRC_UNKNOWN;
202 mp3parse->last_posted_channel_mode = MPEG_AUDIO_CHANNEL_MODE_UNKNOWN;
204 mp3parse->hdr_bitrate = 0;
206 mp3parse->xing_flags = 0;
207 mp3parse->xing_bitrate = 0;
208 mp3parse->xing_frames = 0;
209 mp3parse->xing_total_time = 0;
210 mp3parse->xing_bytes = 0;
211 mp3parse->xing_vbr_scale = 0;
212 memset (mp3parse->xing_seek_table, 0, 100);
213 memset (mp3parse->xing_seek_table_inverse, 0, 256);
215 mp3parse->vbri_bitrate = 0;
216 mp3parse->vbri_frames = 0;
217 mp3parse->vbri_total_time = 0;
218 mp3parse->vbri_bytes = 0;
219 mp3parse->vbri_seek_points = 0;
220 g_free (mp3parse->vbri_seek_table);
221 mp3parse->vbri_seek_table = NULL;
223 mp3parse->encoder_delay = 0;
224 mp3parse->encoder_padding = 0;
228 gst_mpeg_audio_parse_init (GstMpegAudioParse * mp3parse,
229 GstMpegAudioParseClass * klass)
231 gst_mpeg_audio_parse_reset (mp3parse);
235 gst_mpeg_audio_parse_finalize (GObject * object)
237 G_OBJECT_CLASS (parent_class)->finalize (object);
241 gst_mpeg_audio_parse_start (GstBaseParse * parse)
243 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
245 gst_base_parse_set_min_frame_size (GST_BASE_PARSE (mp3parse), MIN_FRAME_SIZE);
246 GST_DEBUG_OBJECT (parse, "starting");
248 gst_mpeg_audio_parse_reset (mp3parse);
254 gst_mpeg_audio_parse_stop (GstBaseParse * parse)
256 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
258 GST_DEBUG_OBJECT (parse, "stopping");
260 gst_mpeg_audio_parse_reset (mp3parse);
265 static const guint mp3types_bitrates[2][3][16] = {
267 {0, 32, 64, 96, 128, 160, 192, 224, 256, 288, 320, 352, 384, 416, 448,},
268 {0, 32, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320, 384,},
269 {0, 32, 40, 48, 56, 64, 80, 96, 112, 128, 160, 192, 224, 256, 320,}
272 {0, 32, 48, 56, 64, 80, 96, 112, 128, 144, 160, 176, 192, 224, 256,},
273 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,},
274 {0, 8, 16, 24, 32, 40, 48, 56, 64, 80, 96, 112, 128, 144, 160,}
278 static const guint mp3types_freqs[3][3] = { {44100, 48000, 32000},
279 {22050, 24000, 16000},
284 mp3_type_frame_length_from_header (GstMpegAudioParse * mp3parse, guint32 header,
285 guint * put_version, guint * put_layer, guint * put_channels,
286 guint * put_bitrate, guint * put_samplerate, guint * put_mode,
290 gulong mode, samplerate, bitrate, layer, channels, padding, crc;
294 if (header & (1 << 20)) {
295 lsf = (header & (1 << 19)) ? 0 : 1;
302 version = 1 + lsf + mpg25;
304 layer = 4 - ((header >> 17) & 0x3);
306 crc = (header >> 16) & 0x1;
308 bitrate = (header >> 12) & 0xF;
309 bitrate = mp3types_bitrates[lsf][layer - 1][bitrate] * 1000;
310 /* The caller has ensured we have a valid header, so bitrate can't be
312 g_assert (bitrate != 0);
314 samplerate = (header >> 10) & 0x3;
315 samplerate = mp3types_freqs[lsf + mpg25][samplerate];
317 padding = (header >> 9) & 0x1;
319 mode = (header >> 6) & 0x3;
320 channels = (mode == 3) ? 1 : 2;
324 length = 4 * ((bitrate * 12) / samplerate + padding);
327 length = (bitrate * 144) / samplerate + padding;
331 length = (bitrate * 144) / (samplerate << lsf) + padding;
335 GST_DEBUG_OBJECT (mp3parse, "Calculated mp3 frame length of %u bytes",
337 GST_DEBUG_OBJECT (mp3parse, "samplerate = %lu, bitrate = %lu, version = %lu, "
338 "layer = %lu, channels = %lu, mode = %s", samplerate, bitrate, version,
339 layer, channels, gst_mpeg_audio_channel_mode_get_nick (mode));
342 *put_version = version;
346 *put_channels = channels;
348 *put_bitrate = bitrate;
350 *put_samplerate = samplerate;
359 /* Minimum number of consecutive, valid-looking frames to consider
361 #define MIN_RESYNC_FRAMES 3
363 /* Perform extended validation to check that subsequent headers match
364 * the first header given here in important characteristics, to avoid
365 * false sync. We look for a minimum of MIN_RESYNC_FRAMES consecutive
366 * frames to match their major characteristics.
368 * If at_eos is set to TRUE, we just check that we don't find any invalid
369 * frames in whatever data is available, rather than requiring a full
370 * MIN_RESYNC_FRAMES of data.
372 * Returns TRUE if we've seen enough data to validate or reject the frame.
373 * If TRUE is returned, then *valid contains TRUE if it validated, or false
374 * if we decided it was false sync.
375 * If FALSE is returned, then *valid contains minimum needed data.
378 gst_mp3parse_validate_extended (GstMpegAudioParse * mp3parse, GstBuffer * buf,
379 guint32 header, int bpf, gboolean at_eos, gint * valid)
384 int frames_found = 1;
387 available = GST_BUFFER_SIZE (buf);
388 data = GST_BUFFER_DATA (buf);
390 while (frames_found < MIN_RESYNC_FRAMES) {
391 /* Check if we have enough data for all these frames, plus the next
393 if (available < offset + 4) {
395 /* Running out of data at EOS is fine; just accept it */
404 next_header = GST_READ_UINT32_BE (data + offset);
405 GST_DEBUG_OBJECT (mp3parse, "At %d: header=%08X, header2=%08X, bpf=%d",
406 offset, (unsigned int) header, (unsigned int) next_header, bpf);
408 /* mask the bits which are allowed to differ between frames */
409 #define HDRMASK ~((0xF << 12) /* bitrate */ | \
410 (0x1 << 9) /* padding */ | \
411 (0xf << 4) /* mode|mode extension */ | \
412 (0xf)) /* copyright|emphasis */
414 if ((next_header & HDRMASK) != (header & HDRMASK)) {
415 /* If any of the unmasked bits don't match, then it's not valid */
416 GST_DEBUG_OBJECT (mp3parse, "next header doesn't match "
417 "(header=%08X (%08X), header2=%08X (%08X), bpf=%d)",
418 (guint) header, (guint) header & HDRMASK, (guint) next_header,
419 (guint) next_header & HDRMASK, bpf);
422 } else if ((((next_header >> 12) & 0xf) == 0) ||
423 (((next_header >> 12) & 0xf) == 0xf)) {
424 /* The essential parts were the same, but the bitrate held an
425 invalid value - also reject */
426 GST_DEBUG_OBJECT (mp3parse, "next header invalid (bitrate)");
431 bpf = mp3_type_frame_length_from_header (mp3parse, next_header,
432 NULL, NULL, NULL, NULL, NULL, NULL, NULL);
443 gst_mpeg_audio_parse_head_check (GstMpegAudioParse * mp3parse,
446 GST_DEBUG_OBJECT (mp3parse, "checking mp3 header 0x%08lx", head);
447 /* if it's not a valid sync */
448 if ((head & 0xffe00000) != 0xffe00000) {
449 GST_WARNING_OBJECT (mp3parse, "invalid sync");
452 /* if it's an invalid MPEG version */
453 if (((head >> 19) & 3) == 0x1) {
454 GST_WARNING_OBJECT (mp3parse, "invalid MPEG version: 0x%lx",
458 /* if it's an invalid layer */
459 if (!((head >> 17) & 3)) {
460 GST_WARNING_OBJECT (mp3parse, "invalid layer: 0x%lx", (head >> 17) & 3);
463 /* if it's an invalid bitrate */
464 if (((head >> 12) & 0xf) == 0x0) {
465 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx."
466 "Free format files are not supported yet", (head >> 12) & 0xf);
469 if (((head >> 12) & 0xf) == 0xf) {
470 GST_WARNING_OBJECT (mp3parse, "invalid bitrate: 0x%lx", (head >> 12) & 0xf);
473 /* if it's an invalid samplerate */
474 if (((head >> 10) & 0x3) == 0x3) {
475 GST_WARNING_OBJECT (mp3parse, "invalid samplerate: 0x%lx",
480 if ((head & 0x3) == 0x2) {
481 /* Ignore this as there are some files with emphasis 0x2 that can
482 * be played fine. See BGO #537235 */
483 GST_WARNING_OBJECT (mp3parse, "invalid emphasis: 0x%lx", head & 0x3);
490 gst_mpeg_audio_parse_check_valid_frame (GstBaseParse * parse,
491 GstBaseParseFrame * frame, guint * framesize, gint * skipsize)
493 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
494 GstBuffer *buf = frame->buffer;
495 GstByteReader reader = GST_BYTE_READER_INIT_FROM_BUFFER (buf);
497 gboolean lost_sync, draining, valid, caps_change;
499 guint bitrate, layer, rate, channels, version, mode, crc;
501 if (G_UNLIKELY (GST_BUFFER_SIZE (buf) < 6))
504 off = gst_byte_reader_masked_scan_uint32 (&reader, 0xffe00000, 0xffe00000,
505 0, GST_BUFFER_SIZE (buf));
507 GST_LOG_OBJECT (parse, "possible sync at buffer offset %d", off);
509 /* didn't find anything that looks like a sync word, skip */
511 *skipsize = GST_BUFFER_SIZE (buf) - 3;
515 /* possible frame header, but not at offset 0? skip bytes before sync */
521 /* make sure the values in the frame header look sane */
522 header = GST_READ_UINT32_BE (GST_BUFFER_DATA (buf));
523 if (!gst_mpeg_audio_parse_head_check (mp3parse, header)) {
528 GST_LOG_OBJECT (parse, "got frame");
530 bpf = mp3_type_frame_length_from_header (mp3parse, header,
531 &version, &layer, &channels, &bitrate, &rate, &mode, &crc);
534 if (channels != mp3parse->channels || rate != mp3parse->rate ||
535 layer != mp3parse->layer || version != mp3parse->version)
540 lost_sync = GST_BASE_PARSE_LOST_SYNC (parse);
541 draining = GST_BASE_PARSE_DRAINING (parse);
543 if (!draining && (lost_sync || caps_change)) {
544 if (!gst_mp3parse_validate_extended (mp3parse, buf, header, bpf, draining,
546 /* not enough data */
547 gst_base_parse_set_min_frame_size (parse, valid);
556 } else if (draining && lost_sync && caps_change && mp3parse->rate > 0) {
557 /* avoid caps jitter that we can't be sure of */
562 /* restore default minimum */
563 gst_base_parse_set_min_frame_size (parse, MIN_FRAME_SIZE);
570 gst_mpeg_audio_parse_handle_first_frame (GstMpegAudioParse * mp3parse,
573 const guint32 xing_id = 0x58696e67; /* 'Xing' in hex */
574 const guint32 info_id = 0x496e666f; /* 'Info' in hex - found in LAME CBR files */
575 const guint32 vbri_id = 0x56425249; /* 'VBRI' in hex */
576 const guint32 lame_id = 0x4c414d45; /* 'LAME' in hex */
577 gint offset_xing, offset_vbri;
579 gint64 upstream_total_bytes = 0;
580 GstFormat fmt = GST_FORMAT_BYTES;
581 guint32 read_id_xing = 0, read_id_vbri = 0;
585 if (mp3parse->sent_codec_tag)
588 /* Check first frame for Xing info */
589 if (mp3parse->version == 1) { /* MPEG-1 file */
590 if (mp3parse->channels == 1)
594 } else { /* MPEG-2 header */
595 if (mp3parse->channels == 1)
601 /* The VBRI tag is always at offset 0x20 */
604 /* Skip the 4 bytes of the MP3 header too */
608 /* Check if we have enough data to read the Xing header */
609 avail = GST_BUFFER_SIZE (buf);
610 data = GST_BUFFER_DATA (buf);
612 if (avail >= offset_xing + 4) {
613 read_id_xing = GST_READ_UINT32_BE (data + offset_xing);
615 if (avail >= offset_vbri + 4) {
616 read_id_vbri = GST_READ_UINT32_BE (data + offset_vbri);
619 /* obtain real upstream total bytes */
620 fmt = GST_FORMAT_BYTES;
621 if (!gst_pad_query_peer_duration (GST_BASE_PARSE_SINK_PAD (GST_BASE_PARSE
622 (mp3parse)), &fmt, &upstream_total_bytes))
623 upstream_total_bytes = 0;
625 if (read_id_xing == xing_id || read_id_xing == info_id) {
627 guint bytes_needed = offset_xing + 8;
629 GstClockTime total_time;
631 GST_DEBUG_OBJECT (mp3parse, "Found Xing header marker 0x%x", xing_id);
633 /* Move data after Xing header */
634 data += offset_xing + 4;
636 /* Read 4 base bytes of flags, big-endian */
637 xing_flags = GST_READ_UINT32_BE (data);
639 if (xing_flags & XING_FRAMES_FLAG)
641 if (xing_flags & XING_BYTES_FLAG)
643 if (xing_flags & XING_TOC_FLAG)
645 if (xing_flags & XING_VBR_SCALE_FLAG)
647 if (avail < bytes_needed) {
648 GST_DEBUG_OBJECT (mp3parse,
649 "Not enough data to read Xing header (need %d)", bytes_needed);
653 GST_DEBUG_OBJECT (mp3parse, "Reading Xing header");
654 mp3parse->xing_flags = xing_flags;
656 if (xing_flags & XING_FRAMES_FLAG) {
657 mp3parse->xing_frames = GST_READ_UINT32_BE (data);
658 if (mp3parse->xing_frames == 0) {
659 GST_WARNING_OBJECT (mp3parse,
660 "Invalid number of frames in Xing header");
661 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
663 mp3parse->xing_total_time = gst_util_uint64_scale (GST_SECOND,
664 (guint64) (mp3parse->xing_frames) * (mp3parse->spf),
670 mp3parse->xing_frames = 0;
671 mp3parse->xing_total_time = 0;
674 if (xing_flags & XING_BYTES_FLAG) {
675 mp3parse->xing_bytes = GST_READ_UINT32_BE (data);
676 if (mp3parse->xing_bytes == 0) {
677 GST_WARNING_OBJECT (mp3parse, "Invalid number of bytes in Xing header");
678 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
682 mp3parse->xing_bytes = 0;
685 /* If we know the upstream size and duration, compute the
686 * total bitrate, rounded up to the nearest kbit/sec */
687 if ((total_time = mp3parse->xing_total_time) &&
688 (total_bytes = mp3parse->xing_bytes)) {
689 mp3parse->xing_bitrate = gst_util_uint64_scale (total_bytes,
690 8 * GST_SECOND, total_time);
691 mp3parse->xing_bitrate += 500;
692 mp3parse->xing_bitrate -= mp3parse->xing_bitrate % 1000;
695 if (xing_flags & XING_TOC_FLAG) {
697 guchar *table = mp3parse->xing_seek_table;
702 GST_DEBUG_OBJECT (mp3parse,
703 "Subtracting initial offset of %d bytes from Xing TOC", first);
705 /* xing seek table: percent time -> 1/256 bytepos */
706 for (i = 0; i < 100; i++) {
707 new = data[i] - first;
709 GST_WARNING_OBJECT (mp3parse, "Skipping broken Xing TOC");
710 mp3parse->xing_flags &= ~XING_TOC_FLAG;
713 mp3parse->xing_seek_table[i] = old = new;
716 /* build inverse table: 1/256 bytepos -> 1/100 percent time */
717 for (i = 0; i < 256; i++) {
718 while (percent < 99 && table[percent + 1] <= i)
721 if (table[percent] == i) {
722 mp3parse->xing_seek_table_inverse[i] = percent * 100;
723 } else if (table[percent] < i && percent < 99) {
725 gint a = percent, b = percent + 1;
729 fx = (b - a) / (fb - fa) * (i - fa) + a;
730 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
731 } else if (percent == 99) {
733 gint a = percent, b = 100;
737 fx = (b - a) / (fb - fa) * (i - fa) + a;
738 mp3parse->xing_seek_table_inverse[i] = (guint16) (fx * 100);
744 memset (mp3parse->xing_seek_table, 0, 100);
745 memset (mp3parse->xing_seek_table_inverse, 0, 256);
748 if (xing_flags & XING_VBR_SCALE_FLAG) {
749 mp3parse->xing_vbr_scale = GST_READ_UINT32_BE (data);
752 mp3parse->xing_vbr_scale = 0;
754 GST_DEBUG_OBJECT (mp3parse, "Xing header reported %u frames, time %"
755 GST_TIME_FORMAT ", %u bytes, vbr scale %u", mp3parse->xing_frames,
756 GST_TIME_ARGS (mp3parse->xing_total_time), mp3parse->xing_bytes,
757 mp3parse->xing_vbr_scale);
759 /* check for truncated file */
760 if (upstream_total_bytes && mp3parse->xing_bytes &&
761 mp3parse->xing_bytes * 0.8 > upstream_total_bytes) {
762 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
763 "invalidating Xing header duration and size");
764 mp3parse->xing_flags &= ~XING_BYTES_FLAG;
765 mp3parse->xing_flags &= ~XING_FRAMES_FLAG;
768 /* Optional LAME tag? */
769 if (avail - bytes_needed >= 36 && GST_READ_UINT32_BE (data) == lame_id) {
770 gchar lame_version[10] = { 0, };
772 guint32 encoder_delay, encoder_padding;
774 memcpy (lame_version, data, 9);
776 tag_rev = data[0] >> 4;
777 GST_DEBUG_OBJECT (mp3parse, "Found LAME tag revision %d created by '%s'",
778 tag_rev, lame_version);
780 /* Skip all the information we're not interested in */
782 /* Encoder delay and end padding */
783 encoder_delay = GST_READ_UINT24_BE (data);
784 encoder_delay >>= 12;
785 encoder_padding = GST_READ_UINT24_BE (data);
786 encoder_padding &= 0x000fff;
788 mp3parse->encoder_delay = encoder_delay;
789 mp3parse->encoder_padding = encoder_padding;
791 GST_DEBUG_OBJECT (mp3parse, "Encoder delay %u, encoder padding %u",
792 encoder_delay, encoder_padding);
796 if (read_id_vbri == vbri_id) {
797 gint64 total_bytes, total_frames;
798 GstClockTime total_time;
799 guint16 nseek_points;
801 GST_DEBUG_OBJECT (mp3parse, "Found VBRI header marker 0x%x", vbri_id);
803 if (avail < offset_vbri + 26) {
804 GST_DEBUG_OBJECT (mp3parse,
805 "Not enough data to read VBRI header (need %d)", offset_vbri + 26);
809 GST_DEBUG_OBJECT (mp3parse, "Reading VBRI header");
811 /* Move data after VBRI header */
812 data += offset_vbri + 4;
814 if (GST_READ_UINT16_BE (data) != 0x0001) {
815 GST_WARNING_OBJECT (mp3parse,
816 "Unsupported VBRI version 0x%x", GST_READ_UINT16_BE (data));
821 /* Skip encoder delay */
827 total_bytes = GST_READ_UINT32_BE (data);
828 if (total_bytes != 0)
829 mp3parse->vbri_bytes = total_bytes;
832 total_frames = GST_READ_UINT32_BE (data);
833 if (total_frames != 0) {
834 mp3parse->vbri_frames = total_frames;
835 mp3parse->vbri_total_time = gst_util_uint64_scale (GST_SECOND,
836 (guint64) (mp3parse->vbri_frames) * (mp3parse->spf), mp3parse->rate);
840 /* If we know the upstream size and duration, compute the
841 * total bitrate, rounded up to the nearest kbit/sec */
842 if ((total_time = mp3parse->vbri_total_time) &&
843 (total_bytes = mp3parse->vbri_bytes)) {
844 mp3parse->vbri_bitrate = gst_util_uint64_scale (total_bytes,
845 8 * GST_SECOND, total_time);
846 mp3parse->vbri_bitrate += 500;
847 mp3parse->vbri_bitrate -= mp3parse->vbri_bitrate % 1000;
850 nseek_points = GST_READ_UINT16_BE (data);
853 if (nseek_points > 0) {
854 guint scale, seek_bytes, seek_frames;
857 mp3parse->vbri_seek_points = nseek_points;
859 scale = GST_READ_UINT16_BE (data);
862 seek_bytes = GST_READ_UINT16_BE (data);
865 seek_frames = GST_READ_UINT16_BE (data);
867 if (scale == 0 || seek_bytes == 0 || seek_bytes > 4 || seek_frames == 0) {
868 GST_WARNING_OBJECT (mp3parse, "Unsupported VBRI seek table");
872 if (avail < offset_vbri + 26 + nseek_points * seek_bytes) {
873 GST_WARNING_OBJECT (mp3parse,
874 "Not enough data to read VBRI seek table (need %d)",
875 offset_vbri + 26 + nseek_points * seek_bytes);
879 if (seek_frames * nseek_points < total_frames - seek_frames ||
880 seek_frames * nseek_points > total_frames + seek_frames) {
881 GST_WARNING_OBJECT (mp3parse,
882 "VBRI seek table doesn't cover the complete file");
886 if (avail < offset_vbri + 26) {
887 GST_DEBUG_OBJECT (mp3parse,
888 "Not enough data to read VBRI header (need %d)",
889 offset_vbri + 26 + nseek_points * seek_bytes);
893 data = GST_BUFFER_DATA (buf);
894 data += offset_vbri + 26;
896 /* VBRI seek table: frame/seek_frames -> byte */
897 mp3parse->vbri_seek_table = g_new (guint32, nseek_points);
899 for (i = 0; i < nseek_points; i++) {
900 mp3parse->vbri_seek_table[i] = GST_READ_UINT32_BE (data) * scale;
902 } else if (seek_bytes == 3)
903 for (i = 0; i < nseek_points; i++) {
904 mp3parse->vbri_seek_table[i] = GST_READ_UINT24_BE (data) * scale;
906 } else if (seek_bytes == 2)
907 for (i = 0; i < nseek_points; i++) {
908 mp3parse->vbri_seek_table[i] = GST_READ_UINT16_BE (data) * scale;
910 } else /* seek_bytes == 1 */
911 for (i = 0; i < nseek_points; i++) {
912 mp3parse->vbri_seek_table[i] = GST_READ_UINT8 (data) * scale;
918 GST_DEBUG_OBJECT (mp3parse, "VBRI header reported %u frames, time %"
919 GST_TIME_FORMAT ", bytes %u", mp3parse->vbri_frames,
920 GST_TIME_ARGS (mp3parse->vbri_total_time), mp3parse->vbri_bytes);
922 /* check for truncated file */
923 if (upstream_total_bytes && mp3parse->vbri_bytes &&
924 mp3parse->vbri_bytes * 0.8 > upstream_total_bytes) {
925 GST_WARNING_OBJECT (mp3parse, "File appears to have been truncated; "
926 "invalidating VBRI header duration and size");
927 mp3parse->vbri_valid = FALSE;
929 mp3parse->vbri_valid = TRUE;
932 GST_DEBUG_OBJECT (mp3parse,
933 "Xing, LAME or VBRI header not found in first frame");
936 /* set duration if tables provided a valid one */
937 if (mp3parse->xing_flags & XING_FRAMES_FLAG) {
938 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
939 mp3parse->xing_total_time, 0);
941 if (mp3parse->vbri_total_time != 0 && mp3parse->vbri_valid) {
942 gst_base_parse_set_duration (GST_BASE_PARSE (mp3parse), GST_FORMAT_TIME,
943 mp3parse->vbri_total_time, 0);
946 /* tell baseclass how nicely we can seek, and a bitrate if one found */
947 /* FIXME: fill index with seek table */
949 seekable = GST_BASE_PARSE_SEEK_DEFAULT;
950 if ((mp3parse->xing_flags & XING_TOC_FLAG) && mp3parse->xing_bytes &&
951 mp3parse->xing_total_time)
952 seekable = GST_BASE_PARSE_SEEK_TABLE;
954 if (mp3parse->vbri_seek_table && mp3parse->vbri_bytes &&
955 mp3parse->vbri_total_time)
956 seekable = GST_BASE_PARSE_SEEK_TABLE;
959 if (mp3parse->xing_bitrate)
960 bitrate = mp3parse->xing_bitrate;
961 else if (mp3parse->vbri_bitrate)
962 bitrate = mp3parse->vbri_bitrate;
966 gst_base_parse_set_average_bitrate (GST_BASE_PARSE (mp3parse), bitrate);
970 gst_mpeg_audio_parse_parse_frame (GstBaseParse * parse,
971 GstBaseParseFrame * frame)
973 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
974 GstBuffer *buf = frame->buffer;
975 guint bitrate, layer, rate, channels, version, mode, crc;
977 g_return_val_if_fail (GST_BUFFER_SIZE (buf) >= 4, GST_FLOW_ERROR);
979 if (!mp3_type_frame_length_from_header (mp3parse,
980 GST_READ_UINT32_BE (GST_BUFFER_DATA (buf)),
981 &version, &layer, &channels, &bitrate, &rate, &mode, &crc))
984 if (G_UNLIKELY (channels != mp3parse->channels || rate != mp3parse->rate ||
985 layer != mp3parse->layer || version != mp3parse->version)) {
986 GstCaps *caps = gst_caps_new_simple ("audio/mpeg",
987 "mpegversion", G_TYPE_INT, 1,
988 "mpegaudioversion", G_TYPE_INT, version,
989 "layer", G_TYPE_INT, layer,
990 "rate", G_TYPE_INT, rate,
991 "channels", G_TYPE_INT, channels, "parsed", G_TYPE_BOOLEAN, TRUE, NULL);
992 gst_buffer_set_caps (buf, caps);
993 gst_pad_set_caps (GST_BASE_PARSE_SRC_PAD (parse), caps);
994 gst_caps_unref (caps);
996 mp3parse->rate = rate;
997 mp3parse->channels = channels;
998 mp3parse->layer = layer;
999 mp3parse->version = version;
1001 /* see http://www.codeproject.com/audio/MPEGAudioInfo.asp */
1002 if (mp3parse->layer == 1)
1003 mp3parse->spf = 384;
1004 else if (mp3parse->layer == 2)
1005 mp3parse->spf = 1152;
1006 else if (mp3parse->version == 1) {
1007 mp3parse->spf = 1152;
1009 /* MPEG-2 or "2.5" */
1010 mp3parse->spf = 576;
1014 * We start pushing 9 frames earlier (29 frames for MPEG2) than
1015 * segment start to be able to decode the first frame we want.
1016 * 9 (29) frames are the theoretical maximum of frames that contain
1017 * data for the current frame (bit reservoir).
1020 * Some mp3 streams have an offset in the timestamps, for which we have to
1021 * push the frame *after* the end position in order for the decoder to be
1022 * able to decode everything up until the segment.stop position. */
1023 gst_base_parse_set_frame_rate (parse, mp3parse->rate, mp3parse->spf,
1024 (version == 1) ? 10 : 30, 2);
1027 mp3parse->hdr_bitrate = bitrate;
1029 /* For first frame; check for seek tables and output a codec tag */
1030 gst_mpeg_audio_parse_handle_first_frame (mp3parse, buf);
1032 /* store some frame info for later processing */
1033 mp3parse->last_crc = crc;
1034 mp3parse->last_mode = mode;
1041 /* this really shouldn't ever happen */
1042 GST_ELEMENT_ERROR (parse, STREAM, DECODE, (NULL), (NULL));
1043 return GST_FLOW_ERROR;
1048 gst_mpeg_audio_parse_time_to_bytepos (GstMpegAudioParse * mp3parse,
1049 GstClockTime ts, gint64 * bytepos)
1052 GstClockTime total_time;
1054 /* If XING seek table exists use this for time->byte conversion */
1055 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1056 (total_bytes = mp3parse->xing_bytes) &&
1057 (total_time = mp3parse->xing_total_time)) {
1060 CLAMP ((100.0 * gst_util_guint64_to_gdouble (ts)) /
1061 gst_util_guint64_to_gdouble (total_time), 0.0, 100.0);
1062 gint index = CLAMP (percent, 0, 99);
1064 fa = mp3parse->xing_seek_table[index];
1066 fb = mp3parse->xing_seek_table[index + 1];
1070 fx = fa + (fb - fa) * (percent - index);
1072 *bytepos = (1.0 / 256.0) * fx * total_bytes;
1077 if (mp3parse->vbri_seek_table && (total_bytes = mp3parse->vbri_bytes) &&
1078 (total_time = mp3parse->vbri_total_time)) {
1080 gdouble a, b, fa, fb;
1082 i = gst_util_uint64_scale (ts, mp3parse->vbri_seek_points - 1, total_time);
1083 i = CLAMP (i, 0, mp3parse->vbri_seek_points - 1);
1085 a = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1086 mp3parse->vbri_seek_points));
1088 for (j = i; j >= 0; j--)
1089 fa += mp3parse->vbri_seek_table[j];
1091 if (i + 1 < mp3parse->vbri_seek_points) {
1092 b = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1093 mp3parse->vbri_seek_points));
1094 fb = fa + mp3parse->vbri_seek_table[i + 1];
1096 b = gst_guint64_to_gdouble (total_time);
1100 *bytepos = fa + ((fb - fa) / (b - a)) * (gst_guint64_to_gdouble (ts) - a);
1109 gst_mpeg_audio_parse_bytepos_to_time (GstMpegAudioParse * mp3parse,
1110 gint64 bytepos, GstClockTime * ts)
1113 GstClockTime total_time;
1115 /* If XING seek table exists use this for byte->time conversion */
1116 if ((mp3parse->xing_flags & XING_TOC_FLAG) &&
1117 (total_bytes = mp3parse->xing_bytes) &&
1118 (total_time = mp3parse->xing_total_time)) {
1123 pos = CLAMP ((bytepos * 256.0) / total_bytes, 0.0, 256.0);
1124 index = CLAMP (pos, 0, 255);
1125 fa = mp3parse->xing_seek_table_inverse[index];
1127 fb = mp3parse->xing_seek_table_inverse[index + 1];
1131 fx = fa + (fb - fa) * (pos - index);
1133 *ts = (1.0 / 10000.0) * fx * gst_util_guint64_to_gdouble (total_time);
1138 if (mp3parse->vbri_seek_table &&
1139 (total_bytes = mp3parse->vbri_bytes) &&
1140 (total_time = mp3parse->vbri_total_time)) {
1143 gdouble a, b, fa, fb;
1146 sum += mp3parse->vbri_seek_table[i];
1148 } while (i + 1 < mp3parse->vbri_seek_points
1149 && sum + mp3parse->vbri_seek_table[i] < bytepos);
1152 a = gst_guint64_to_gdouble (sum);
1153 fa = gst_guint64_to_gdouble (gst_util_uint64_scale (i, total_time,
1154 mp3parse->vbri_seek_points));
1156 if (i + 1 < mp3parse->vbri_seek_points) {
1157 b = a + mp3parse->vbri_seek_table[i + 1];
1158 fb = gst_guint64_to_gdouble (gst_util_uint64_scale (i + 1, total_time,
1159 mp3parse->vbri_seek_points));
1162 fb = gst_guint64_to_gdouble (total_time);
1165 *ts = gst_gdouble_to_guint64 (fa + ((fb - fa) / (b - a)) * (bytepos - a));
1174 gst_mpeg_audio_parse_convert (GstBaseParse * parse, GstFormat src_format,
1175 gint64 src_value, GstFormat dest_format, gint64 * dest_value)
1177 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1178 gboolean res = FALSE;
1180 if (src_format == GST_FORMAT_TIME && dest_format == GST_FORMAT_BYTES)
1182 gst_mpeg_audio_parse_time_to_bytepos (mp3parse, src_value, dest_value);
1183 else if (src_format == GST_FORMAT_BYTES && dest_format == GST_FORMAT_TIME)
1184 res = gst_mpeg_audio_parse_bytepos_to_time (mp3parse, src_value,
1185 (GstClockTime *) dest_value);
1187 /* if no tables, fall back to default estimated rate based conversion */
1189 return gst_base_parse_convert_default (parse, src_format, src_value,
1190 dest_format, dest_value);
1195 static GstFlowReturn
1196 gst_mpeg_audio_parse_pre_push_frame (GstBaseParse * parse,
1197 GstBaseParseFrame * frame)
1199 GstMpegAudioParse *mp3parse = GST_MPEG_AUDIO_PARSE (parse);
1200 GstTagList *taglist;
1202 /* tag sending done late enough in hook to ensure pending events
1203 * have already been sent */
1205 if (!mp3parse->sent_codec_tag) {
1209 if (mp3parse->layer == 3) {
1210 codec = g_strdup_printf ("MPEG %d Audio, Layer %d (MP3)",
1211 mp3parse->version, mp3parse->layer);
1213 codec = g_strdup_printf ("MPEG %d Audio, Layer %d",
1214 mp3parse->version, mp3parse->layer);
1216 taglist = gst_tag_list_new ();
1217 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1218 GST_TAG_AUDIO_CODEC, codec, NULL);
1219 if (mp3parse->hdr_bitrate > 0 && mp3parse->xing_bitrate == 0 &&
1220 mp3parse->vbri_bitrate == 0) {
1221 /* We don't have a VBR bitrate, so post the available bitrate as
1222 * nominal and let baseparse calculate the real bitrate */
1223 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE,
1224 GST_TAG_NOMINAL_BITRATE, mp3parse->hdr_bitrate, NULL);
1226 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1227 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1230 /* also signals the end of first-frame processing */
1231 mp3parse->sent_codec_tag = TRUE;
1234 /* we will create a taglist (if any of the parameters has changed)
1235 * to add the tags that changed */
1237 if (mp3parse->last_posted_crc != mp3parse->last_crc) {
1241 taglist = gst_tag_list_new ();
1243 mp3parse->last_posted_crc = mp3parse->last_crc;
1244 if (mp3parse->last_posted_crc == CRC_PROTECTED) {
1249 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_CRC,
1253 if (mp3parse->last_posted_channel_mode != mp3parse->last_mode) {
1255 taglist = gst_tag_list_new ();
1257 mp3parse->last_posted_channel_mode = mp3parse->last_mode;
1259 gst_tag_list_add (taglist, GST_TAG_MERGE_REPLACE, GST_TAG_MODE,
1260 gst_mpeg_audio_channel_mode_get_nick (mp3parse->last_mode), NULL);
1263 /* if the taglist exists, we need to send it */
1265 gst_element_found_tags_for_pad (GST_ELEMENT (mp3parse),
1266 GST_BASE_PARSE_SRC_PAD (mp3parse), taglist);
1269 /* usual clipping applies */
1270 frame->flags |= GST_BASE_PARSE_FRAME_FLAG_CLIP;
1276 gst_mpeg_audio_parse_get_sink_caps (GstBaseParse * parse)
1281 peercaps = gst_pad_get_allowed_caps (GST_BASE_PARSE_SRC_PAD (parse));
1285 /* Remove the parsed field */
1286 peercaps = gst_caps_make_writable (peercaps);
1287 n = gst_caps_get_size (peercaps);
1288 for (i = 0; i < n; i++) {
1289 GstStructure *s = gst_caps_get_structure (peercaps, i);
1291 gst_structure_remove_field (s, "parsed");
1295 gst_caps_intersect_full (peercaps,
1296 gst_pad_get_pad_template_caps (GST_BASE_PARSE_SRC_PAD (parse)),
1297 GST_CAPS_INTERSECT_FIRST);
1298 gst_caps_unref (peercaps);
1301 gst_caps_copy (gst_pad_get_pad_template_caps (GST_BASE_PARSE_SINK_PAD