aacdec.c

    for (i = 0; i < len; i++)
        dest[i] += gain * src[i];
}

/**
 * channel coupling transformation interface
 *
 * @param   apply_coupling_method   pointer to (in)dependent coupling function
 */
static void apply_channel_coupling(AACContext *ac, ChannelElement *cc,
                                   enum RawDataBlockType type, int elem_id,
                                   enum CouplingPoint coupling_point,
                                   void (*apply_coupling_method)(AACContext *ac, SingleChannelElement *target, ChannelElement *cce, int index))
{
    int i, c;

    for (i = 0; i < MAX_ELEM_ID; i++) {
        ChannelElement *cce = ac->che[TYPE_CCE][i];
        int index = 0;

        if (cce && cce->coup.coupling_point == coupling_point) {
            ChannelCoupling *coup = &cce->coup;

            for (c = 0; c <= coup->num_coupled; c++) {
                if (coup->type[c] == type && coup->id_select[c] == elem_id) {
                    if (coup->ch_select[c] != 1) {
                        apply_coupling_method(ac, &cc->ch[0], cce, index);
                        if (coup->ch_select[c] != 0)
                            index++;
                    }
                    if (coup->ch_select[c] != 2)
                        apply_coupling_method(ac, &cc->ch[1], cce, index++);
                } else
                    index += 1 + (coup->ch_select[c] == 3);
            }
        }
    }
}

/**
 * Convert spectral data to float samples, applying all supported tools as appropriate.
 */
static void spectral_to_sample(AACContext *ac)
{
    int i, type;
    for (type = 3; type >= 0; type--) {
        for (i = 0; i < MAX_ELEM_ID; i++) {
            ChannelElement *che = ac->che[type][i];
            if (che) {
                if (type <= TYPE_CPE)
                    apply_channel_coupling(ac, che, type, i, BEFORE_TNS, apply_dependent_coupling);
                if (ac->m4ac.object_type == AOT_AAC_LTP) {
                    if (che->ch[0].ics.predictor_present) {
                        if (che->ch[0].ics.ltp.present)
                            apply_ltp(ac, &che->ch[0]);
                        if (che->ch[1].ics.ltp.present && type == TYPE_CPE)
                            apply_ltp(ac, &che->ch[1]);
                    }
                }
                if (che->ch[0].tns.present)
                    apply_tns(che->ch[0].coeffs, &che->ch[0].tns, &che->ch[0].ics, 1);
                if (che->ch[1].tns.present)
                    apply_tns(che->ch[1].coeffs, &che->ch[1].tns, &che->ch[1].ics, 1);
                if (type <= TYPE_CPE)
                    apply_channel_coupling(ac, che, type, i, BETWEEN_TNS_AND_IMDCT, apply_dependent_coupling);
                if (type != TYPE_CCE || che->coup.coupling_point == AFTER_IMDCT) {
                    imdct_and_windowing(ac, &che->ch[0]);
                    if (ac->m4ac.object_type == AOT_AAC_LTP)
                        update_ltp(ac, &che->ch[0]);
                    if (type == TYPE_CPE) {
                        imdct_and_windowing(ac, &che->ch[1]);
                        if (ac->m4ac.object_type == AOT_AAC_LTP)
                            update_ltp(ac, &che->ch[1]);
                    }
                    if (ac->m4ac.sbr > 0) {
                        ff_sbr_apply(ac, &che->sbr, type, che->ch[0].ret, che->ch[1].ret);
                    }
                }
                if (type <= TYPE_CCE)
                    apply_channel_coupling(ac, che, type, i, AFTER_IMDCT, apply_independent_coupling);
            }
        }
    }
}

static int parse_adts_frame_header(AACContext *ac, GetBitContext *gb)
{
    int size;
    AACADTSHeaderInfo hdr_info;

    size = avpriv_aac_parse_header(gb, &hdr_info);
    if (size > 0) {
        if (hdr_info.chan_config) {
            enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
            memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
            ac->m4ac.chan_config = hdr_info.chan_config;
            if (set_default_channel_config(ac->avctx, new_che_pos, hdr_info.chan_config))
                return -7;
            if (output_configure(ac, ac->che_pos, new_che_pos, hdr_info.chan_config,
                                 FFMAX(ac->output_configured, OC_TRIAL_FRAME)))
                return -7;
        } else if (ac->output_configured != OC_LOCKED) {
            ac->m4ac.chan_config = 0;
            ac->output_configured = OC_NONE;
        }
        if (ac->output_configured != OC_LOCKED) {
            ac->m4ac.sbr = -1;
            ac->m4ac.ps  = -1;
            ac->m4ac.sample_rate     = hdr_info.sample_rate;
            ac->m4ac.sampling_index  = hdr_info.sampling_index;
            ac->m4ac.object_type     = hdr_info.object_type;
        }
        if (!ac->avctx->sample_rate)
            ac->avctx->sample_rate = hdr_info.sample_rate;
        if (hdr_info.num_aac_frames == 1) {
            if (!hdr_info.crc_absent)
                skip_bits(gb, 16);
        } else {
            av_log_missing_feature(ac->avctx, "More than one AAC RDB per ADTS frame is", 0);
            return -1;
        }
    }
    return size;
}

static int aac_decode_frame_int(AVCodecContext *avctx, void *data,
                                int *got_frame_ptr, GetBitContext *gb)
{
    AACContext *ac = avctx->priv_data;
    ChannelElement *che = NULL, *che_prev = NULL;
    enum RawDataBlockType elem_type, elem_type_prev = TYPE_END;
    int err, elem_id;
    int samples = 0, multiplier, audio_found = 0;

    if (show_bits(gb, 12) == 0xfff) {
        if (parse_adts_frame_header(ac, gb) < 0) {
            av_log(avctx, AV_LOG_ERROR, "Error decoding AAC frame header.\n");
            return -1;
        }
        if (ac->m4ac.sampling_index > 12) {
            av_log(ac->avctx, AV_LOG_ERROR, "invalid sampling rate index %d\n", ac->m4ac.sampling_index);
            return -1;
        }
    }

    ac->tags_mapped = 0;
    // parse
    while ((elem_type = get_bits(gb, 3)) != TYPE_END) {
        elem_id = get_bits(gb, 4);

        if (elem_type < TYPE_DSE) {
            if (!(che=get_che(ac, elem_type, elem_id))) {
                av_log(ac->avctx, AV_LOG_ERROR, "channel element %d.%d is not allocated\n",
                       elem_type, elem_id);
                return -1;
            }
            samples = 1024;
        }

        switch (elem_type) {

        case TYPE_SCE:
            err = decode_ics(ac, &che->ch[0], gb, 0, 0);
            audio_found = 1;
            break;

        case TYPE_CPE:
            err = decode_cpe(ac, gb, che);
            audio_found = 1;
            break;

        case TYPE_CCE:
            err = decode_cce(ac, gb, che);
            break;

        case TYPE_LFE:
            err = decode_ics(ac, &che->ch[0], gb, 0, 0);
            audio_found = 1;
            break;

        case TYPE_DSE:
            err = skip_data_stream_element(ac, gb);
            break;

        case TYPE_PCE: {
            enum ChannelPosition new_che_pos[4][MAX_ELEM_ID];
            memset(new_che_pos, 0, 4 * MAX_ELEM_ID * sizeof(new_che_pos[0][0]));
            if ((err = decode_pce(avctx, &ac->m4ac, new_che_pos, gb)))
                break;
            if (ac->output_configured > OC_TRIAL_PCE)
                av_log(avctx, AV_LOG_ERROR,
                       "Not evaluating a further program_config_element as this construct is dubious at best.\n");
            else
                err = output_configure(ac, ac->che_pos, new_che_pos, 0, OC_TRIAL_PCE);
            break;
        }

        case TYPE_FIL:
            if (elem_id == 15)
                elem_id += get_bits(gb, 8) - 1;
            if (get_bits_left(gb) < 8 * elem_id) {
                    av_log(avctx, AV_LOG_ERROR, overread_err);
                    return -1;
            }
            while (elem_id > 0)
                elem_id -= decode_extension_payload(ac, gb, elem_id, che_prev, elem_type_prev);
            err = 0; /* FIXME */
            break;

        default:
            err = -1; /* should not happen, but keeps compiler happy */
            break;
        }

        che_prev       = che;
        elem_type_prev = elem_type;

        if (err)
            return err;

        if (get_bits_left(gb) < 3) {
            av_log(avctx, AV_LOG_ERROR, overread_err);
            return -1;
        }
    }

    spectral_to_sample(ac);

    multiplier = (ac->m4ac.sbr == 1) ? ac->m4ac.ext_sample_rate > ac->m4ac.sample_rate : 0;
    samples <<= multiplier;
    if (ac->output_configured < OC_LOCKED) {
        avctx->sample_rate = ac->m4ac.sample_rate << multiplier;
        avctx->frame_size = samples;
    }

    if (samples) {
        /* get output buffer */
        ac->frame.nb_samples = samples;
        if ((err = avctx->get_buffer(avctx, &ac->frame)) < 0) {
            av_log(avctx, AV_LOG_ERROR, "get_buffer() failed\n");
            return err;
        }

        if (avctx->sample_fmt == AV_SAMPLE_FMT_FLT)
            ac->fmt_conv.float_interleave((float *)ac->frame.data[0],
                                          (const float **)ac->output_data,
                                          samples, avctx->channels);
        else
            ac->fmt_conv.float_to_int16_interleave((int16_t *)ac->frame.data[0],
                                                   (const float **)ac->output_data,
                                                   samples, avctx->channels);

        *(AVFrame *)data = ac->frame;
    }
    *got_frame_ptr = !!samples;

    if (ac->output_configured && audio_found)
        ac->output_configured = OC_LOCKED;

    return 0;
}

static int aac_decode_frame(AVCodecContext *avctx, void *data,
                            int *got_frame_ptr, AVPacket *avpkt)
{
    AACContext *ac = avctx->priv_data;
    const uint8_t *buf = avpkt->data;
    int buf_size = avpkt->size;
    GetBitContext gb;
    int buf_consumed;
    int buf_offset;
    int err;
    int new_extradata_size;
    const uint8_t *new_extradata = av_packet_get_side_data(avpkt,
                                       AV_PKT_DATA_NEW_EXTRADATA,
                                       &new_extradata_size);

    if (new_extradata) {
        av_free(avctx->extradata);
        avctx->extradata = av_mallocz(new_extradata_size +
                                      FF_INPUT_BUFFER_PADDING_SIZE);
        if (!avctx->extradata)
            return AVERROR(ENOMEM);
        avctx->extradata_size = new_extradata_size;
        memcpy(avctx->extradata, new_extradata, new_extradata_size);
        if (decode_audio_specific_config(ac, ac->avctx, &ac->m4ac,
                                         avctx->extradata,
                                         avctx->extradata_size*8, 1) < 0)
            return AVERROR_INVALIDDATA;
    }

    init_get_bits(&gb, buf, buf_size * 8);

    if ((err = aac_decode_frame_int(avctx, data, got_frame_ptr, &gb)) < 0)
        return err;

    buf_consumed = (get_bits_count(&gb) + 7) >> 3;
    for (buf_offset = buf_consumed; buf_offset < buf_size; buf_offset++)
        if (buf[buf_offset])
            break;

    return buf_size > buf_offset ? buf_consumed : buf_size;
}

static av_cold int aac_decode_close(AVCodecContext *avctx)
{
    AACContext *ac = avctx->priv_data;
    int i, type;

    for (i = 0; i < MAX_ELEM_ID; i++) {
        for (type = 0; type < 4; type++) {
            if (ac->che[type][i])
                ff_aac_sbr_ctx_close(&ac->che[type][i]->sbr);
            av_freep(&ac->che[type][i]);
        }
    }

    ff_mdct_end(&ac->mdct);
    ff_mdct_end(&ac->mdct_small);
    ff_mdct_end(&ac->mdct_ltp);
    return 0;
}


#define LOAS_SYNC_WORD   0x2b7       ///< 11 bits LOAS sync word

struct LATMContext {
    AACContext      aac_ctx;             ///< containing AACContext
    int             initialized;         ///< initilized after a valid extradata was seen

    // parser data
    int             audio_mux_version_A; ///< LATM syntax version
    int             frame_length_type;   ///< 0/1 variable/fixed frame length
    int             frame_length;        ///< frame length for fixed frame length
};

static inline uint32_t latm_get_value(GetBitContext *b)
{
    int length = get_bits(b, 2);

    return get_bits_long(b, (length+1)*8);
}

static int latm_decode_audio_specific_config(struct LATMContext *latmctx,
                                             GetBitContext *gb, int asclen)
{
    AACContext *ac        = &latmctx->aac_ctx;
    AVCodecContext *avctx = ac->avctx;
    MPEG4AudioConfig m4ac = {0};
    int config_start_bit  = get_bits_count(gb);
    int sync_extension    = 0;
    int bits_consumed, esize;

    if (asclen) {
        sync_extension = 1;
        asclen         = FFMIN(asclen, get_bits_left(gb));
    } else
        asclen         = get_bits_left(gb);

    if (config_start_bit % 8) {
        av_log_missing_feature(latmctx->aac_ctx.avctx, "audio specific "
                               "config not byte aligned.\n", 1);
        return AVERROR_INVALIDDATA;
    }
    bits_consumed = decode_audio_specific_config(NULL, avctx, &m4ac,
                                         gb->buffer + (config_start_bit / 8),
                                         asclen, sync_extension);

    if (bits_consumed < 0)
        return AVERROR_INVALIDDATA;

    if (ac->m4ac.sample_rate != m4ac.sample_rate ||
        ac->m4ac.chan_config != m4ac.chan_config) {

        av_log(avctx, AV_LOG_INFO, "audio config changed\n");
        latmctx->initialized = 0;

        esize = (bits_consumed+7) / 8;

        if (avctx->extradata_size < esize) {
            av_free(avctx->extradata);
            avctx->extradata = av_malloc(esize + FF_INPUT_BUFFER_PADDING_SIZE);
            if (!avctx->extradata)
                return AVERROR(ENOMEM);
        }

        avctx->extradata_size = esize;
        memcpy(avctx->extradata, gb->buffer + (config_start_bit/8), esize);
        memset(avctx->extradata+esize, 0, FF_INPUT_BUFFER_PADDING_SIZE);
    }
    skip_bits_long(gb, bits_consumed);

    return bits_consumed;
}

static int read_stream_mux_config(struct LATMContext *latmctx,
                                  GetBitContext *gb)
{
    int ret, audio_mux_version = get_bits(gb, 1);

    latmctx->audio_mux_version_A = 0;
    if (audio_mux_version)
        latmctx->audio_mux_version_A = get_bits(gb, 1);

    if (!latmctx->audio_mux_version_A) {

        if (audio_mux_version)
            latm_get_value(gb);                 // taraFullness

        skip_bits(gb, 1);                       // allStreamSameTimeFraming
        skip_bits(gb, 6);                       // numSubFrames
        // numPrograms
        if (get_bits(gb, 4)) {                  // numPrograms
            av_log_missing_feature(latmctx->aac_ctx.avctx,
                                   "multiple programs are not supported\n", 1);
            return AVERROR_PATCHWELCOME;
        }

        // for each program (which there is only on in DVB)

        // for each layer (which there is only on in DVB)
        if (get_bits(gb, 3)) {                   // numLayer
            av_log_missing_feature(latmctx->aac_ctx.avctx,
                                   "multiple layers are not supported\n", 1);
            return AVERROR_PATCHWELCOME;
        }

        // for all but first stream: use_same_config = get_bits(gb, 1);
        if (!audio_mux_version) {
            if ((ret = latm_decode_audio_specific_config(latmctx, gb, 0)) < 0)
                return ret;
        } else {
            int ascLen = latm_get_value(gb);
            if ((ret = latm_decode_audio_specific_config(latmctx, gb, ascLen)) < 0)
                return ret;
            ascLen -= ret;
            skip_bits_long(gb, ascLen);
        }

        latmctx->frame_length_type = get_bits(gb, 3);
        switch (latmctx->frame_length_type) {
        case 0:
            skip_bits(gb, 8);       // latmBufferFullness
            break;
        case 1:
            latmctx->frame_length = get_bits(gb, 9);
            break;
        case 3:
        case 4:
        case 5:
            skip_bits(gb, 6);       // CELP frame length table index
            break;
        case 6:
        case 7:
            skip_bits(gb, 1);       // HVXC frame length table index
            break;
        }

        if (get_bits(gb, 1)) {                  // other data
            if (audio_mux_version) {
                latm_get_value(gb);             // other_data_bits
            } else {
                int esc;
                do {
                    esc = get_bits(gb, 1);
                    skip_bits(gb, 8);
                } while (esc);
            }
        }

        if (get_bits(gb, 1))                     // crc present
            skip_bits(gb, 8);                    // config_crc
    }

    return 0;
}

static int read_payload_length_info(struct LATMContext *ctx, GetBitContext *gb)
{
    uint8_t tmp;

    if (ctx->frame_length_type == 0) {
        int mux_slot_length = 0;
        do {
            tmp = get_bits(gb, 8);
            mux_slot_length += tmp;
        } while (tmp == 255);
        return mux_slot_length;
    } else if (ctx->frame_length_type == 1) {
        return ctx->frame_length;
    } else if (ctx->frame_length_type == 3 ||
               ctx->frame_length_type == 5 ||
               ctx->frame_length_type == 7) {
        skip_bits(gb, 2);          // mux_slot_length_coded
    }
    return 0;
}

static int read_audio_mux_element(struct LATMContext *latmctx,
                                  GetBitContext *gb)
{
    int err;
    uint8_t use_same_mux = get_bits(gb, 1);
    if (!use_same_mux) {
        if ((err = read_stream_mux_config(latmctx, gb)) < 0)
            return err;
    } else if (!latmctx->aac_ctx.avctx->extradata) {
        av_log(latmctx->aac_ctx.avctx, AV_LOG_DEBUG,
               "no decoder config found\n");
        return AVERROR(EAGAIN);
    }
    if (latmctx->audio_mux_version_A == 0) {
        int mux_slot_length_bytes = read_payload_length_info(latmctx, gb);
        if (mux_slot_length_bytes * 8 > get_bits_left(gb)) {
            av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR, "incomplete frame\n");
            return AVERROR_INVALIDDATA;
        } else if (mux_slot_length_bytes * 8 + 256 < get_bits_left(gb)) {
            av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
                   "frame length mismatch %d << %d\n",
                   mux_slot_length_bytes * 8, get_bits_left(gb));
            return AVERROR_INVALIDDATA;
        }
    }
    return 0;
}


static int latm_decode_frame(AVCodecContext *avctx, void *out,
                             int *got_frame_ptr, AVPacket *avpkt)
{
    struct LATMContext *latmctx = avctx->priv_data;
    int                 muxlength, err;
    GetBitContext       gb;

    init_get_bits(&gb, avpkt->data, avpkt->size * 8);

    // check for LOAS sync word
    if (get_bits(&gb, 11) != LOAS_SYNC_WORD)
        return AVERROR_INVALIDDATA;

    muxlength = get_bits(&gb, 13) + 3;
    // not enough data, the parser should have sorted this
    if (muxlength > avpkt->size)
        return AVERROR_INVALIDDATA;

    if ((err = read_audio_mux_element(latmctx, &gb)) < 0)
        return err;

    if (!latmctx->initialized) {
        if (!avctx->extradata) {
            *got_frame_ptr = 0;
            return avpkt->size;
        } else {
            if ((err = decode_audio_specific_config(
                    &latmctx->aac_ctx, avctx, &latmctx->aac_ctx.m4ac,
                    avctx->extradata, avctx->extradata_size*8, 1)) < 0)
                return err;
            latmctx->initialized = 1;
        }
    }

    if (show_bits(&gb, 12) == 0xfff) {
        av_log(latmctx->aac_ctx.avctx, AV_LOG_ERROR,
               "ADTS header detected, probably as result of configuration "
               "misparsing\n");
        return AVERROR_INVALIDDATA;
    }

    if ((err = aac_decode_frame_int(avctx, out, got_frame_ptr, &gb)) < 0)
        return err;

    return muxlength;
}

av_cold static int latm_decode_init(AVCodecContext *avctx)
{
    struct LATMContext *latmctx = avctx->priv_data;
    int ret = aac_decode_init(avctx);

    if (avctx->extradata_size > 0)
        latmctx->initialized = !ret;

    return ret;
}


AVCodec ff_aac_decoder = {
    .name           = "aac",
    .type           = AVMEDIA_TYPE_AUDIO,
    .id             = CODEC_ID_AAC,
    .priv_data_size = sizeof(AACContext),
    .init           = aac_decode_init,
    .close          = aac_decode_close,
    .decode         = aac_decode_frame,
    .long_name = NULL_IF_CONFIG_SMALL("Advanced Audio Coding"),
    .sample_fmts = (const enum AVSampleFormat[]) {
        AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
    },
    .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
    .channel_layouts = aac_channel_layout,
};

/*
    Note: This decoder filter is intended to decode LATM streams transferred
    in MPEG transport streams which only contain one program.
    To do a more complex LATM demuxing a separate LATM demuxer should be used.
*/
AVCodec ff_aac_latm_decoder = {
    .name = "aac_latm",
    .type = AVMEDIA_TYPE_AUDIO,
    .id   = CODEC_ID_AAC_LATM,
    .priv_data_size = sizeof(struct LATMContext),
    .init   = latm_decode_init,
    .close  = aac_decode_close,
    .decode = latm_decode_frame,
    .long_name = NULL_IF_CONFIG_SMALL("AAC LATM (Advanced Audio Codec LATM syntax)"),
    .sample_fmts = (const enum AVSampleFormat[]) {
        AV_SAMPLE_FMT_FLT, AV_SAMPLE_FMT_S16, AV_SAMPLE_FMT_NONE
    },
    .capabilities = CODEC_CAP_CHANNEL_CONF | CODEC_CAP_DR1,
    .channel_layouts = aac_channel_layout,
};