matroska.c

ffmpeg源码分析

static int
matroska_ebmlnum_sint (uint8_t  *data,
                       uint32_t  size,
                       int64_t  *num)
{
    uint64_t unum;
    int res;

    /* read as unsigned number first */
    if ((res = matroska_ebmlnum_uint(data, size, &unum)) < 0)
        return res;

    /* make signed (weird way) */
    if (unum == (uint64_t)-1)
        *num = INT64_MAX;
    else
        *num = unum - ((1LL << ((7 * res) - 1)) - 1);

    return res;
}

/*
 * Read an EBML header.
 * 0 is success, < 0 is failure.
 */

static int
ebml_read_header (MatroskaDemuxContext *matroska,
                  char                **doctype,
                  int                  *version)
{
    uint32_t id;
    int level_up, res = 0;

    /* default init */
    if (doctype)
        *doctype = NULL;
    if (version)
        *version = 1;

    if (!(id = ebml_peek_id(matroska, &level_up)) ||
        level_up != 0 || id != EBML_ID_HEADER) {
        av_log(matroska->ctx, AV_LOG_ERROR,
               "This is not an EBML file (id=0x%x/0x%x)\n", id, EBML_ID_HEADER);
        return AVERROR_INVALIDDATA;
    }
    if ((res = ebml_read_master(matroska, &id)) < 0)
        return res;

    while (res == 0) {
        if (!(id = ebml_peek_id(matroska, &level_up)))
            return AVERROR_IO;

        /* end-of-header */
        if (level_up)
            break;

        switch (id) {
            /* is our read version uptodate? */
            case EBML_ID_EBMLREADVERSION: {
                uint64_t num;

                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    return res;
                if (num > EBML_VERSION) {
                    av_log(matroska->ctx, AV_LOG_ERROR,
                           "EBML version %"PRIu64" (> %d) is not supported\n",
                           num, EBML_VERSION);
                    return AVERROR_INVALIDDATA;
                }
                break;
            }

            /* we only handle 8 byte lengths at max */
            case EBML_ID_EBMLMAXSIZELENGTH: {
                uint64_t num;

                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    return res;
                if (num > sizeof(uint64_t)) {
                    av_log(matroska->ctx, AV_LOG_ERROR,
                           "Integers of size %"PRIu64" (> %zd) not supported\n",
                           num, sizeof(uint64_t));
                    return AVERROR_INVALIDDATA;
                }
                break;
            }

            /* we handle 4 byte IDs at max */
            case EBML_ID_EBMLMAXIDLENGTH: {
                uint64_t num;

                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    return res;
                if (num > sizeof(uint32_t)) {
                    av_log(matroska->ctx, AV_LOG_ERROR,
                           "IDs of size %"PRIu64" (> %zu) not supported\n",
                            num, sizeof(uint32_t));
                    return AVERROR_INVALIDDATA;
                }
                break;
            }

            case EBML_ID_DOCTYPE: {
                char *text;

                if ((res = ebml_read_ascii(matroska, &id, &text)) < 0)
                    return res;
                if (doctype) {
                    if (*doctype)
                        av_free(*doctype);
                    *doctype = text;
                } else
                    av_free(text);
                break;
            }

            case EBML_ID_DOCTYPEREADVERSION: {
                uint64_t num;

                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    return res;
                if (version)
                    *version = num;
                break;
            }

            default:
                av_log(matroska->ctx, AV_LOG_INFO,
                       "Unknown data type 0x%x in EBML header", id);
                /* pass-through */

            case EBML_ID_VOID:
            /* we ignore these two, as they don't tell us anything we
             * care about */
            case EBML_ID_EBMLVERSION:
            case EBML_ID_DOCTYPEVERSION:
                res = ebml_read_skip (matroska);
                break;
        }
    }

    return 0;
}

/*
 * Put one packet in an application-supplied AVPacket struct.
 * Returns 0 on success or -1 on failure.
 */

static int
matroska_deliver_packet (MatroskaDemuxContext *matroska,
                         AVPacket             *pkt)
{
    if (matroska->num_packets > 0) {
        memcpy(pkt, matroska->packets[0], sizeof(AVPacket));
        av_free(matroska->packets[0]);
        if (matroska->num_packets > 1) {
            memmove(&matroska->packets[0], &matroska->packets[1],
                    (matroska->num_packets - 1) * sizeof(AVPacket *));
            matroska->packets =
                av_realloc(matroska->packets, (matroska->num_packets - 1) *
                           sizeof(AVPacket *));
        } else {
            av_free(matroska->packets);
            matroska->packets = NULL;
        }
        matroska->num_packets--;
        return 0;
    }

    return -1;
}

/*
 * Put a packet into our internal queue. Will be delivered to the
 * user/application during the next get_packet() call.
 */

static void
matroska_queue_packet (MatroskaDemuxContext *matroska,
                       AVPacket             *pkt)
{
    matroska->packets =
        av_realloc(matroska->packets, (matroska->num_packets + 1) *
                   sizeof(AVPacket *));
    matroska->packets[matroska->num_packets] = pkt;
    matroska->num_packets++;
}

/*
 * Autodetecting...
 */

static int
matroska_probe (AVProbeData *p)
{
    uint64_t total = 0;
    int len_mask = 0x80, size = 1, n = 1;
    uint8_t probe_data[] = { 'm', 'a', 't', 'r', 'o', 's', 'k', 'a' };

    if (p->buf_size < 5)
        return 0;

    /* ebml header? */
    if ((p->buf[0] << 24 | p->buf[1] << 16 |
         p->buf[2] << 8 | p->buf[3]) != EBML_ID_HEADER)
        return 0;

    /* length of header */
    total = p->buf[4];
    while (size <= 8 && !(total & len_mask)) {
        size++;
        len_mask >>= 1;
    }
    if (size > 8)
      return 0;
    total &= (len_mask - 1);
    while (n < size)
        total = (total << 8) | p->buf[4 + n++];

    /* does the probe data contain the whole header? */
    if (p->buf_size < 4 + size + total)
      return 0;

    /* the header must contain the document type 'matroska'. For now,
     * we don't parse the whole header but simply check for the
     * availability of that array of characters inside the header.
     * Not fully fool-proof, but good enough. */
    for (n = 4 + size; n < 4 + size + total - sizeof(probe_data); n++)
        if (!memcmp (&p->buf[n], probe_data, sizeof(probe_data)))
            return AVPROBE_SCORE_MAX;

    return 0;
}

/*
 * From here on, it's all XML-style DTD stuff... Needs no comments.
 */

static int
matroska_parse_info (MatroskaDemuxContext *matroska)
{
    int res = 0;
    uint32_t id;

    av_log(matroska->ctx, AV_LOG_DEBUG, "Parsing info...\n");

    while (res == 0) {
        if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
            res = AVERROR_IO;
            break;
        } else if (matroska->level_up) {
            matroska->level_up--;
            break;
        }

        switch (id) {
            /* cluster timecode */
            case MATROSKA_ID_TIMECODESCALE: {
                uint64_t num;
                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    break;
                matroska->time_scale = num;
                break;
            }

            case MATROSKA_ID_DURATION: {
                double num;
                if ((res = ebml_read_float(matroska, &id, &num)) < 0)
                    break;
                matroska->ctx->duration = num * matroska->time_scale * 1000 / AV_TIME_BASE;
                break;
            }

            case MATROSKA_ID_WRITINGAPP: {
                char *text;
                if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
                    break;
                matroska->writing_app = text;
                break;
            }

            case MATROSKA_ID_MUXINGAPP: {
                char *text;
                if ((res = ebml_read_utf8(matroska, &id, &text)) < 0)
                    break;
                matroska->muxing_app = text;
                break;
            }

            case MATROSKA_ID_DATEUTC: {
                int64_t time;
                if ((res = ebml_read_date(matroska, &id, &time)) < 0)
                    break;
                matroska->created = time;
                break;
            }

            default:
                av_log(matroska->ctx, AV_LOG_INFO,
                       "Unknown entry 0x%x in info header\n", id);
                /* fall-through */

            case EBML_ID_VOID:
                res = ebml_read_skip(matroska);
                break;
        }

        if (matroska->level_up) {
            matroska->level_up--;
            break;
        }
    }

    return res;
}

static int
matroska_add_stream (MatroskaDemuxContext *matroska)
{
    int res = 0;
    uint32_t id;
    MatroskaTrack *track;

    av_log(matroska->ctx, AV_LOG_DEBUG, "parsing track, adding stream..,\n");

    /* Allocate a generic track. As soon as we know its type we'll realloc. */
    track = av_mallocz(sizeof(MatroskaTrack));
    matroska->num_tracks++;

    /* start with the master */
    if ((res = ebml_read_master(matroska, &id)) < 0)
        return res;

    /* try reading the trackentry headers */
    while (res == 0) {
        if (!(id = ebml_peek_id(matroska, &matroska->level_up))) {
            res = AVERROR_IO;
            break;
        } else if (matroska->level_up > 0) {
            matroska->level_up--;
            break;
        }

        switch (id) {
            /* track number (unique stream ID) */
            case MATROSKA_ID_TRACKNUMBER: {
                uint64_t num;
                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    break;
                track->num = num;
                break;
            }

            /* track UID (unique identifier) */
            case MATROSKA_ID_TRACKUID: {
                uint64_t num;
                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    break;
                track->uid = num;
                break;
            }

            /* track type (video, audio, combined, subtitle, etc.) */
            case MATROSKA_ID_TRACKTYPE: {
                uint64_t num;
                if (track->type != 0) {
                    av_log(matroska->ctx, AV_LOG_INFO,
                           "More than one tracktype in an entry - skip\n");
                    break;
                }
                if ((res = ebml_read_uint(matroska, &id, &num)) < 0)
                    break;
                track->type = num;

                /* ok, so we're actually going to reallocate this thing */
                switch (track->type) {
                    case MATROSKA_TRACK_TYPE_VIDEO:
                        track = (MatroskaTrack *)
                            av_realloc(track, sizeof(MatroskaVideoTrack));
                        break;
                    case MATROSKA_TRACK_TYPE_AUDIO:
                        track = (MatroskaTrack *)
                            av_realloc(track, sizeof(MatroskaAudioTrack));
                        ((MatroskaAudioTrack *)track)->channels = 1;
                        ((MatroskaAudioTrack *)track)->samplerate = 8000;
                        break;
                    case MATROSKA_TRACK_TYPE_SUBTITLE:
                        track = (MatroskaTrack *)
                            av_realloc(track, sizeof(MatroskaSubtitleTrack));
                        break;
                    case MATROSKA_TRACK_TYPE_COMPLEX:
                    case MATROSKA_TRACK_TYPE_LOGO:
                    case MATROSKA_TRACK_TYPE_CONTROL:
                    default:
                        av_log(matroska->ctx, AV_LOG_INFO,
                               "Unknown or unsupported track type 0x%x\n",
                               track->type);
                        track->type = 0;
                        break;
                }
                matroska->tracks[matroska->num_tracks - 1] = track;
                break;
            }

            /* tracktype specific stuff for video */
            case MATROSKA_ID_TRACKVIDEO: {
                MatroskaVideoTrack *videotrack;
                if (track->type != MATROSKA_TRACK_TYPE_VIDEO) {
                    av_log(matroska->ctx, AV_LOG_INFO,
                           "video data in non-video track - ignoring\n");