*/
#include "avformat.h"
+#include "md5.h"
#include "riff.h"
#include "xiph.h"
#include "matroska.h"
typedef struct MatroskaMuxContext {
offset_t segment;
offset_t segment_offset;
+ offset_t segment_uid;
offset_t cluster;
offset_t cluster_pos; ///< file offset of the current cluster
uint64_t cluster_pts;
mkv_seekhead *main_seekhead;
mkv_seekhead *cluster_seekhead;
mkv_cues *cues;
+
+ struct AVMD5 *md5_ctx;
} MatroskaMuxContext;
+static int ebml_id_size(unsigned int id)
+{
+ return (av_log2(id+1)-1)/7+1;
+}
+
static void put_ebml_id(ByteIOContext *pb, unsigned int id)
{
- if (id >= 0x3fffff)
- put_byte(pb, id >> 24);
- if (id >= 0x7fff)
- put_byte(pb, id >> 16);
- if (id >= 0xff)
- put_byte(pb, id >> 8);
- put_byte(pb, id);
+ int i = ebml_id_size(id);
+ while (i--)
+ put_byte(pb, id >> (i*8));
}
-static int ebml_id_size(unsigned int id)
+/**
+ * Write an EBML size meaning "unknown size"
+ *
+ * @param bytes The number of bytes the size should occupy. Maximum of 8.
+ */
+static void put_ebml_size_unknown(ByteIOContext *pb, int bytes)
{
- return (av_log2(id+1)-1)/7+1;
+ uint64_t value = 0;
+ int i;
+
+ bytes = FFMIN(bytes, 8);
+ for (i = 0; i < bytes*7 + 1; i++)
+ value |= 1ULL << i;
+ for (i = bytes-1; i >= 0; i--)
+ put_byte(pb, value >> i*8);
+}
+
+/**
+ * Calculate how many bytes are needed to represent a given size in EBML
+ */
+static int ebml_size_bytes(uint64_t size)
+{
+ int bytes = 1;
+ while ((size+1) >> bytes*7) bytes++;
+ return bytes;
}
-// XXX: test this thoroughly and get rid of minbytes hack (currently needed to
-// use up all of the space reserved in start_ebml_master)
-static void put_ebml_size(ByteIOContext *pb, uint64_t size, int minbytes)
+/**
+ * Write a size in EBML variable length format.
+ *
+ * @param bytes The number of bytes that need to be used to write the size.
+ * If zero, any number of bytes can be used.
+ */
+static void put_ebml_size(ByteIOContext *pb, uint64_t size, int bytes)
{
- int bytes = minbytes;
+ int i, needed_bytes = ebml_size_bytes(size);
// sizes larger than this are currently undefined in EBML
// so write "unknown" size
- size = FFMIN(size, (1ULL<<56)-1);
+ if (size >= (1ULL<<56)-1) {
+ put_ebml_size_unknown(pb, 1);
+ return;
+ }
- while (size >> (bytes*7 + 7)) bytes++;
+ if (bytes == 0)
+ // don't care how many bytes are used, so use the min
+ bytes = needed_bytes;
+ else if (needed_bytes > bytes) {
+ // the bytes needed to write the given size would exceed the bytes
+ // that we need to use, so write unknown size. This shouldn't happen.
+ av_log(NULL, AV_LOG_WARNING, "Size of %llu needs %d bytes but only %d bytes reserved\n",
+ size, needed_bytes, bytes);
+ put_ebml_size_unknown(pb, bytes);
+ return;
+ }
- put_byte(pb, (0x80 >> bytes) | (size >> bytes*8));
- for (bytes -= 1; bytes >= 0; bytes--)
- put_byte(pb, size >> bytes*8);
+ size |= 1ULL << bytes*7;
+ for (i = bytes - 1; i >= 0; i--)
+ put_byte(pb, size >> i*8);
}
static void put_ebml_uint(ByteIOContext *pb, unsigned int elementid, uint64_t val)
{
- int bytes = 1;
+ int i, bytes = 1;
while (val >> bytes*8) bytes++;
put_ebml_id(pb, elementid);
put_ebml_size(pb, bytes, 0);
- for (bytes -= 1; bytes >= 0; bytes--)
- put_byte(pb, val >> bytes*8);
+ for (i = bytes - 1; i >= 0; i--)
+ put_byte(pb, val >> i*8);
}
static void put_ebml_float(ByteIOContext *pb, unsigned int elementid, double val)
{
- // XXX: single-precision floats?
put_ebml_id(pb, elementid);
put_ebml_size(pb, 8, 0);
put_be64(pb, av_dbl2int(val));
put_ebml_binary(pb, elementid, str, strlen(str));
}
-// this reserves exactly the amount of space specified by size, which must be at least 2
+/**
+ * Writes a void element of a given size. Useful for reserving space in the file to be
+ * written to later.
+ *
+ * @param size The amount of space to reserve, which must be at least 2.
+ */
static void put_ebml_void(ByteIOContext *pb, uint64_t size)
{
offset_t currentpos = url_ftell(pb);
if (size < 10)
put_ebml_size(pb, size-1, 0);
else
- put_ebml_size(pb, size-9, 7);
+ put_ebml_size(pb, size-9, 8);
url_fseek(pb, currentpos + size, SEEK_SET);
}
put_ebml_id(pb, elementid);
// XXX: this always reserves the maximum needed space to store any size value
// we should be smarter (additional parameter for expected size?)
- put_ebml_size(pb, (1ULL<<56)-1, 0); // largest unknown size
+ put_ebml_size_unknown(pb, 8);
return url_ftell(pb);
}
offset_t pos = url_ftell(pb);
url_fseek(pb, start - 8, SEEK_SET);
- put_ebml_size(pb, pos - start, 7);
+ put_ebml_size(pb, pos - start, 8);
url_fseek(pb, pos, SEEK_SET);
}
-// initializes a mkv_seekhead element to be ready to index level 1 matroska elements
-// if numelements is greater than 0, it reserves enough space for that many elements
-// at the current file position and writes the seekhead there, otherwise the seekhead
-// will be appended to the file when end_mkv_seekhead() is called
+static void put_xiph_size(ByteIOContext *pb, int size)
+{
+ int i;
+ for (i = 0; i < size / 255; i++)
+ put_byte(pb, 255);
+ put_byte(pb, size % 255);
+}
+
+/**
+ * Initialize a mkv_seekhead element to be ready to index level 1 Matroska elements.
+ * If a maximum number of elements is specified, enough space will be reserved at
+ * the current file location to write a seek head of that size.
+ *
+ * @param segment_offset the absolute offset into the file that the segment begins
+ * @param numelements the maximum number of elements that will be indexed by this
+ * seek head, 0 if unlimited.
+ */
static mkv_seekhead * mkv_start_seekhead(ByteIOContext *pb, offset_t segment_offset, int numelements)
{
mkv_seekhead *new_seekhead = av_mallocz(sizeof(mkv_seekhead));
return 0;
}
-// returns the file offset where the seekhead was written and frees the seekhead
+/**
+ * Write the seek head to the file and free it. If a maximum number of elements was
+ * specified to mkv_start_seekhead(), the seek head will be written at the location
+ * reserved for it. Otherwise, it is written at the current location in the file.
+ *
+ * @return the file offset where the seekhead was written
+ */
static offset_t mkv_write_seekhead(ByteIOContext *pb, mkv_seekhead *seekhead)
{
offset_t metaseek, seekentry, currentpos;
uint8_t *header_start[3];
int header_len[3];
int first_header_size;
- int j, k;
+ int j;
if (codec->codec_id == CODEC_ID_VORBIS)
first_header_size = 30;
codecprivate = start_ebml_master(pb, MATROSKA_ID_CODECPRIVATE);
put_byte(pb, 2); // number packets - 1
for (j = 0; j < 2; j++) {
- for (k = 0; k < header_len[j] / 255; k++)
- put_byte(pb, 255);
- put_byte(pb, header_len[j] % 255);
+ put_xiph_size(pb, header_len[j]);
}
for (j = 0; j < 3; j++)
put_buffer(pb, header_start[j], header_len[j]);
return 0;
}
+#define FLAC_STREAMINFO_SIZE 34
+
+static int put_flac_codecpriv(ByteIOContext *pb, AVCodecContext *codec)
+{
+ offset_t codecpriv = start_ebml_master(pb, MATROSKA_ID_CODECPRIVATE);
+
+ // if the extradata_size is greater than FLAC_STREAMINFO_SIZE,
+ // assume that it's in Matroska's format already
+ if (codec->extradata_size < FLAC_STREAMINFO_SIZE) {
+ av_log(codec, AV_LOG_ERROR, "Invalid FLAC extradata\n");
+ return -1;
+ } else if (codec->extradata_size == FLAC_STREAMINFO_SIZE) {
+ // only the streaminfo packet
+ put_byte(pb, 0);
+ put_xiph_size(pb, codec->extradata_size);
+ av_log(codec, AV_LOG_ERROR, "Only one packet\n");
+ }
+ put_buffer(pb, codec->extradata, codec->extradata_size);
+ end_ebml_master(pb, codecpriv);
+ return 0;
+}
+
static void get_aac_sample_rates(AVCodecContext *codec, int *sample_rate, int *output_sample_rate)
{
static const int aac_sample_rates[] = {
if (st->language[0])
put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, st->language);
+ else
+ put_ebml_string(pb, MATROSKA_ID_TRACKLANGUAGE, "und");
// look for a codec id string specific to mkv to use, if none are found, use AVI codes
for (j = 0; ff_mkv_codec_tags[j].id != CODEC_ID_NONE; j++) {
}
}
- // XXX: CodecPrivate for vorbis, theora, aac, native mpeg4, ...
if (native_id) {
if (codec->codec_id == CODEC_ID_VORBIS || codec->codec_id == CODEC_ID_THEORA) {
if (put_xiph_codecpriv(pb, codec) < 0)
return -1;
+ } else if (codec->codec_id == CODEC_ID_FLAC) {
+ if (put_flac_codecpriv(pb, codec) < 0)
+ return -1;
} else if (codec->extradata_size) {
put_ebml_binary(pb, MATROSKA_ID_CODECPRIVATE, codec->extradata, codec->extradata_size);
}
// XXX: interlace flag?
put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELWIDTH , codec->width);
put_ebml_uint (pb, MATROSKA_ID_VIDEOPIXELHEIGHT, codec->height);
- // XXX: display width/height
+ if (codec->sample_aspect_ratio.num) {
+ put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYWIDTH , codec->sample_aspect_ratio.num);
+ put_ebml_uint(pb, MATROSKA_ID_VIDEODISPLAYHEIGHT, codec->sample_aspect_ratio.den);
+ }
end_ebml_master(pb, subinfo);
break;
ByteIOContext *pb = &s->pb;
offset_t ebml_header, segment_info;
+ mkv->md5_ctx = av_mallocz(av_md5_size);
+ av_md5_init(mkv->md5_ctx);
+
ebml_header = start_ebml_master(pb, EBML_ID_HEADER);
put_ebml_uint (pb, EBML_ID_EBMLVERSION , 1);
put_ebml_uint (pb, EBML_ID_EBMLREADVERSION , 1);
if (strlen(s->title))
put_ebml_string(pb, MATROSKA_ID_TITLE, s->title);
if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
- put_ebml_string(pb, MATROSKA_ID_MUXINGAPP, LIBAVFORMAT_IDENT);
- // XXX: both are required; something better for writing app?
+ put_ebml_string(pb, MATROSKA_ID_MUXINGAPP , LIBAVFORMAT_IDENT);
put_ebml_string(pb, MATROSKA_ID_WRITINGAPP, LIBAVFORMAT_IDENT);
+
+ // reserve space to write the segment UID later
+ mkv->segment_uid = url_ftell(pb);
+ put_ebml_void(pb, 19);
}
- // XXX: segment UID
+
// reserve space for the duration
mkv->duration = 0;
mkv->duration_offset = url_ftell(pb);
MatroskaMuxContext *mkv = s->priv_data;
ByteIOContext *pb = &s->pb;
+ av_log(s, AV_LOG_DEBUG, "Writing block at offset %llu, size %d, pts %lld, dts %lld, duration %d, flags %d\n",
+ url_ftell(pb), pkt->size, pkt->pts, pkt->dts, pkt->duration, flags);
put_ebml_id(pb, blockid);
put_ebml_size(pb, pkt->size + 4, 0);
put_byte(pb, 0x80 | (pkt->stream_index + 1)); // this assumes stream_index is less than 126
{
MatroskaMuxContext *mkv = s->priv_data;
ByteIOContext *pb = &s->pb;
+ AVCodecContext *codec = s->streams[pkt->stream_index]->codec;
int keyframe = !!(pkt->flags & PKT_FLAG_KEY);
// start a new cluster every 5 MB or 5 sec
if (url_ftell(pb) > mkv->cluster + 5*1024*1024 || pkt->pts > mkv->cluster_pts + 5000) {
+ av_log(s, AV_LOG_DEBUG, "Starting new cluster at offset %llu bytes, pts %llu\n", url_ftell(pb), pkt->pts);
end_ebml_master(pb, mkv->cluster);
if (mkv_add_seekhead_entry(mkv->cluster_seekhead, MATROSKA_ID_CLUSTER, url_ftell(pb)) < 0)
mkv->cluster = start_ebml_master(pb, MATROSKA_ID_CLUSTER);
put_ebml_uint(pb, MATROSKA_ID_CLUSTERTIMECODE, pkt->pts);
mkv->cluster_pts = pkt->pts;
+ av_md5_update(mkv->md5_ctx, pkt->data, FFMIN(200, pkt->size));
}
- if (s->streams[pkt->stream_index]->codec->codec_type != CODEC_TYPE_SUBTITLE) {
- mkv_write_block(s, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
+ if (codec->codec_type != CODEC_TYPE_SUBTITLE) {
+ mkv_write_block(s, MATROSKA_ID_SIMPLEBLOCK, pkt, keyframe << 7);
} else {
offset_t blockgroup = start_ebml_master(pb, MATROSKA_ID_BLOCKGROUP);
mkv_write_block(s, MATROSKA_ID_BLOCK, pkt, 0);
end_ebml_master(pb, blockgroup);
}
- if (s->streams[pkt->stream_index]->codec->codec_type == CODEC_TYPE_VIDEO && keyframe) {
+ if (codec->codec_type == CODEC_TYPE_VIDEO && keyframe) {
if (mkv_add_cuepoint(mkv->cues, pkt, mkv->cluster_pos) < 0)
return -1;
}
mkv_write_seekhead(pb, mkv->main_seekhead);
// update the duration
+ av_log(s, AV_LOG_DEBUG, "end duration = %llu\n", mkv->duration);
currentpos = url_ftell(pb);
url_fseek(pb, mkv->duration_offset, SEEK_SET);
put_ebml_float(pb, MATROSKA_ID_DURATION, mkv->duration);
+
+ // write the md5sum of some frames as the segment UID
+ if (!(s->streams[0]->codec->flags & CODEC_FLAG_BITEXACT)) {
+ uint8_t segment_uid[16];
+ av_md5_final(mkv->md5_ctx, segment_uid);
+ url_fseek(pb, mkv->segment_uid, SEEK_SET);
+ put_ebml_binary(pb, MATROSKA_ID_SEGMENTUID, segment_uid, 16);
+ }
url_fseek(pb, currentpos, SEEK_SET);
end_ebml_master(pb, mkv->segment);
+ av_free(mkv->md5_ctx);
return 0;
}
mkv_write_packet,
mkv_write_trailer,
.codec_tag = (const AVCodecTag*[]){codec_bmp_tags, codec_wav_tags, 0},
+ .subtitle_codec = CODEC_ID_TEXT,
+};
+
+AVOutputFormat matroska_audio_muxer = {
+ "matroska",
+ "Matroska File Format",
+ "audio/x-matroska",
+ "mka",
+ sizeof(MatroskaMuxContext),
+ CODEC_ID_MP2,
+ CODEC_ID_NONE,
+ mkv_write_header,
+ mkv_write_packet,
+ mkv_write_trailer,
+ .codec_tag = (const AVCodecTag*[]){codec_wav_tags, 0},
};