srs/trunk/src/app/srs_app_rtc_codec.cpp

//
// Copyright (c) 2013-2021 The SRS Authors
//
// SPDX-License-Identifier: MIT or MulanPSL-2.0
//

#include <srs_app_rtc_codec.hpp>

#include <srs_kernel_codec.hpp>
#include <srs_kernel_error.hpp>
#include <srs_kernel_log.hpp>

static const AVCodec* srs_find_decoder_by_id(SrsAudioCodecId id)
{
    if (id == SrsAudioCodecIdAAC) {
        return avcodec_find_decoder_by_name("aac");
    } else if (id == SrsAudioCodecIdOpus) {
        const AVCodec* codec = avcodec_find_decoder_by_name("libopus");
        if (!codec) {
            codec = avcodec_find_decoder_by_name("opus");
        }
        return codec;
    }
    return NULL;
}

static const AVCodec* srs_find_encoder_by_id(SrsAudioCodecId id)
{
    if (id == SrsAudioCodecIdAAC) {
        return avcodec_find_encoder_by_name("aac");
    } else if (id == SrsAudioCodecIdOpus) {
        const AVCodec* codec = avcodec_find_encoder_by_name("libopus");
        if (!codec) {
            codec = avcodec_find_encoder_by_name("opus");
        }
        return codec;
    }
    return NULL;
}

class SrsFFmpegLogHelper {
public:
    SrsFFmpegLogHelper() {
        av_log_set_callback(ffmpeg_log_callback);
        av_log_set_level(AV_LOG_TRACE);
    }

    static void ffmpeg_log_callback(void*, int level, const char* fmt, va_list vl) 
    {
        static char buf[4096] = {0};
        int nbytes = vsnprintf(buf, sizeof(buf), fmt, vl);
        if (nbytes > 0) {
            // Srs log is always start with new line, replcae '\n' to '\0', make log easy to read.
            if (buf[nbytes - 1] == '\n') {
                buf[nbytes - 1] = '\0';
            }
            switch (level) {
                case AV_LOG_PANIC:
                case AV_LOG_FATAL:
                case AV_LOG_ERROR:
                    srs_error("%s", buf);
                    break;
                case AV_LOG_WARNING:
                    srs_warn("%s", buf);
                    break;
                case AV_LOG_INFO:
                    srs_trace("%s", buf);
                    break;
                case AV_LOG_VERBOSE:
                case AV_LOG_DEBUG:
                case AV_LOG_TRACE:
                default:
                    srs_verbose("%s", buf);
                    break;
            }
        }
    }
};

// Register FFmpeg log callback funciton.
SrsFFmpegLogHelper _srs_ffmpeg_log_helper;

SrsAudioTranscoder::SrsAudioTranscoder()
{
    dec_ = NULL;
    dec_frame_ = NULL;
    dec_packet_ = NULL;
    enc_ = NULL;
    enc_frame_ = NULL;
    enc_packet_ = NULL;
    swr_ = NULL;
    swr_data_ = NULL;
    fifo_ = NULL;
    new_pkt_pts_ = AV_NOPTS_VALUE;
    next_out_pts_ = AV_NOPTS_VALUE;
}

SrsAudioTranscoder::~SrsAudioTranscoder()
{
    if (dec_) {
        avcodec_free_context(&dec_);
    }

    if (dec_frame_) {
        av_frame_free(&dec_frame_);
    }

    if (dec_packet_) {
        av_packet_free(&dec_packet_);
    }

    if (swr_) {
        swr_free(&swr_);
    }

    free_swr_samples();

    if (enc_) {
        avcodec_free_context(&enc_);
    }

    if (enc_frame_) {
        av_frame_free(&enc_frame_);
    }

    if (enc_packet_) {
        av_packet_free(&enc_packet_);
    }

    if (fifo_) {
        av_audio_fifo_free(fifo_);
        fifo_ = NULL;
    }
}

srs_error_t SrsAudioTranscoder::initialize(SrsAudioCodecId src_codec, SrsAudioCodecId dst_codec, int dst_channels, int dst_samplerate, int dst_bit_rate)
{
    srs_error_t err = srs_success;

    if ((err = init_dec(src_codec)) != srs_success) {
        return srs_error_wrap(err, "dec init codec:%d", src_codec);
    }

    if ((err = init_enc(dst_codec, dst_channels, dst_samplerate, dst_bit_rate)) != srs_success) {
        return srs_error_wrap(err, "enc init codec:%d, channels:%d, samplerate:%d, bitrate:%d",
            dst_codec, dst_channels, dst_samplerate, dst_bit_rate);
    }

    if ((err = init_fifo()) != srs_success) {
        return srs_error_wrap(err, "fifo init");
    }

    return err;
}

srs_error_t SrsAudioTranscoder::transcode(SrsAudioFrame *in_pkt, std::vector<SrsAudioFrame*>& out_pkts)
{
    srs_error_t err = srs_success;

    if ((err = decode_and_resample(in_pkt)) != srs_success) {
        return srs_error_wrap(err, "decode and resample");
    }

    if ((err = encode(out_pkts)) != srs_success) {
        return srs_error_wrap(err, "encode");
    }

    return err;
}

void SrsAudioTranscoder::free_frames(std::vector<SrsAudioFrame*>& frames)
{
    for (std::vector<SrsAudioFrame*>::iterator it = frames.begin(); it != frames.end(); ++it) {
        SrsAudioFrame* p = *it;

        for (int i = 0; i < p->nb_samples; i++) {
            char* pa = p->samples[i].bytes;
            srs_freepa(pa);
        }

        srs_freep(p);
    }
}

void SrsAudioTranscoder::aac_codec_header(uint8_t **data, int *len)
{
    //srs_assert(dst_codec == SrsAudioCodecIdAAC);
    *len = enc_->extradata_size;
    *data = enc_->extradata;
}

srs_error_t SrsAudioTranscoder::init_dec(SrsAudioCodecId src_codec)
{
    const AVCodec *codec = srs_find_decoder_by_id(src_codec);
    if (!codec) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Codec not found by %d", src_codec);
    }

    dec_ = avcodec_alloc_context3(codec);
    if (!dec_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio codec context");
    }

    if (avcodec_open2(dec_, codec, NULL) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not open codec");
    }
    
    dec_->channel_layout = av_get_default_channel_layout(dec_->channels);

    dec_frame_ = av_frame_alloc();
    if (!dec_frame_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio decode out frame");
    }

    dec_packet_ = av_packet_alloc();
    if (!dec_packet_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio decode in packet");
    }

    new_pkt_pts_ = AV_NOPTS_VALUE;
    return srs_success;
}

srs_error_t SrsAudioTranscoder::init_enc(SrsAudioCodecId dst_codec, int dst_channels, int dst_samplerate, int dst_bit_rate)
{
    const AVCodec *codec = srs_find_encoder_by_id(dst_codec);
    if (!codec) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Codec not found by %d", dst_codec);
    }

    enc_ = avcodec_alloc_context3(codec);
    if (!enc_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio codec context %d", dst_codec);
    }

    enc_->sample_rate = dst_samplerate;
    enc_->channels = dst_channels;
    enc_->channel_layout = av_get_default_channel_layout(dst_channels);
    enc_->bit_rate = dst_bit_rate;
    enc_->sample_fmt = codec->sample_fmts[0];
    enc_->time_base.num = 1; enc_->time_base.den = 1000; // {1, 1000}
    if (dst_codec == SrsAudioCodecIdOpus) {
        //TODO: for more level setting
        enc_->compression_level = 1;
        enc_->strict_std_compliance = FF_COMPLIANCE_EXPERIMENTAL;
    } else if (dst_codec == SrsAudioCodecIdAAC) {
        enc_->strict_std_compliance = FF_COMPLIANCE_EXPERIMENTAL;
    }

    // TODO: FIXME: Show detail error.
    if (avcodec_open2(enc_, codec, NULL) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not open codec");
    }

    enc_frame_ = av_frame_alloc();
    if (!enc_frame_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio encode in frame");
    }

    enc_frame_->format = enc_->sample_fmt;
    enc_frame_->nb_samples = enc_->frame_size;
    enc_frame_->channel_layout = enc_->channel_layout;

    if (av_frame_get_buffer(enc_frame_, 0) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not get audio frame buffer");
    }

    enc_packet_ = av_packet_alloc();
    if (!enc_packet_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate audio encode out packet");
    }

    next_out_pts_ = AV_NOPTS_VALUE;
    return srs_success;
}

srs_error_t SrsAudioTranscoder::init_swr(AVCodecContext* decoder)
{
    swr_ = swr_alloc_set_opts(NULL, enc_->channel_layout, enc_->sample_fmt, enc_->sample_rate,
        decoder->channel_layout, decoder->sample_fmt, decoder->sample_rate, 0, NULL);
    if (!swr_) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "alloc swr");
    }

    int error;
    char err_buf[AV_ERROR_MAX_STRING_SIZE] = {0};
    if ((error = swr_init(swr_)) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "open swr(%d:%s)", error,
            av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
    }

    /* Allocate as many pointers as there are audio channels.
    * Each pointer will later point to the audio samples of the corresponding
    * channels (although it may be NULL for interleaved formats).
    */
    if (!(swr_data_ = (uint8_t **)calloc(enc_->channels, sizeof(*swr_data_)))) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "alloc swr buffer");
    }

    /* Allocate memory for the samples of all channels in one consecutive
    * block for convenience. */
    if ((error = av_samples_alloc(swr_data_, NULL, enc_->channels, enc_->frame_size, enc_->sample_fmt, 0)) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "alloc swr buffer(%d:%s)", error,
            av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
    }

    return srs_success;
}

srs_error_t SrsAudioTranscoder::init_fifo()
{
    if (!(fifo_ = av_audio_fifo_alloc(enc_->sample_fmt, enc_->channels, 1))) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not allocate FIFO");
    }
    return srs_success;
}

srs_error_t SrsAudioTranscoder::decode_and_resample(SrsAudioFrame *pkt)
{
    srs_error_t err = srs_success;

    dec_packet_->data = (uint8_t *)pkt->samples[0].bytes;
    dec_packet_->size = pkt->samples[0].size;

    // Ignore empty packet, see https://github.com/ossrs/srs/pull/2757#discussion_r759797651
    if (!dec_packet_->data || !dec_packet_->size){
        return err;
    }

    char err_buf[AV_ERROR_MAX_STRING_SIZE] = {0};
    int error = avcodec_send_packet(dec_, dec_packet_);
    if (error < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "submit to dec(%d,%s)", error,
            av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
    }

    new_pkt_pts_ = pkt->dts + pkt->cts;
    while (error >= 0) {
        error = avcodec_receive_frame(dec_, dec_frame_);
        if (error == AVERROR(EAGAIN) || error == AVERROR_EOF) {
            return err;
        } else if (error < 0) {
            return srs_error_new(ERROR_RTC_RTP_MUXER, "Error during decoding(%d,%s)", error,
                av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
        }

        // Decoder is OK now, try to init swr if not initialized.
        if (!swr_ && (err = init_swr(dec_)) != srs_success) {
            return srs_error_wrap(err, "resample init");
        }

        int in_samples = dec_frame_->nb_samples;
        const uint8_t **in_data = (const uint8_t**)dec_frame_->extended_data;
        do {
            /* Convert the samples using the resampler. */
            int frame_size = swr_convert(swr_, swr_data_, enc_->frame_size, in_data, in_samples);
            if ((error = frame_size) < 0) {
                return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not convert input samples(%d,%s)", error,
                    av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
            }

            in_data = NULL; in_samples = 0;
            if ((err = add_samples_to_fifo(swr_data_, frame_size)) != srs_success) {
                return srs_error_wrap(err, "write samples");
            }
        } while (swr_get_out_samples(swr_, in_samples) >= enc_->frame_size);
    }

    return err;
}

srs_error_t SrsAudioTranscoder::encode(std::vector<SrsAudioFrame*> &pkts)
{
    char err_buf[AV_ERROR_MAX_STRING_SIZE] = {0};

    if (next_out_pts_ == AV_NOPTS_VALUE) {
        next_out_pts_ = new_pkt_pts_;
    } else {
        int64_t diff = llabs(new_pkt_pts_ - next_out_pts_);
        if (diff > 1000) {
            srs_trace("time diff to large=%lld, next out=%lld, new pkt=%lld, set to new pkt",
                diff, next_out_pts_, new_pkt_pts_);
            next_out_pts_ = new_pkt_pts_;
        }
    }

    int frame_cnt = 0;
    while (av_audio_fifo_size(fifo_) >= enc_->frame_size) {
        /* Read as many samples from the FIFO buffer as required to fill the frame.
        * The samples are stored in the frame temporarily. */
        if (av_audio_fifo_read(fifo_, (void **)enc_frame_->data, enc_->frame_size) < enc_->frame_size) {
            return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not read data from FIFO");
        }
        /* send the frame for encoding */
        enc_frame_->pts = next_out_pts_ + av_rescale(enc_->frame_size * frame_cnt, 1000, enc_->sample_rate);
        ++frame_cnt;
        int error = avcodec_send_frame(enc_, enc_frame_);
        if (error < 0) {
            return srs_error_new(ERROR_RTC_RTP_MUXER, "Error sending the frame to the encoder(%d,%s)", error,
                av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
        }

        av_init_packet(enc_packet_);
        enc_packet_->data = NULL;
        enc_packet_->size = 0;
        /* read all the available output packets (in general there may be any
        * number of them */
        while (error >= 0) {
            error = avcodec_receive_packet(enc_, enc_packet_);
            if (error == AVERROR(EAGAIN) || error == AVERROR_EOF) {
                break;
            } else if (error < 0) {
                free_frames(pkts);
                return srs_error_new(ERROR_RTC_RTP_MUXER, "Error during decoding(%d,%s)", error,
                    av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
            }

            SrsAudioFrame *out_frame = new SrsAudioFrame;
            char *buf = new char[enc_packet_->size];
            memcpy(buf, enc_packet_->data, enc_packet_->size);
            out_frame->add_sample(buf, enc_packet_->size);
            out_frame->dts = enc_packet_->dts;
            out_frame->cts = enc_packet_->pts - enc_packet_->dts;
            pkts.push_back(out_frame);
        }
    }

    next_out_pts_ += av_rescale(enc_->frame_size * frame_cnt, 1000, enc_->sample_rate);

    return srs_success;
}

srs_error_t  SrsAudioTranscoder::add_samples_to_fifo(uint8_t **samples, int frame_size)
{
    char err_buf[AV_ERROR_MAX_STRING_SIZE] = {0};

    int error;

    /* Make the FIFO as large as it needs to be to hold both,
     * the old and the new samples. */
    if ((error = av_audio_fifo_realloc(fifo_, av_audio_fifo_size(fifo_) + frame_size)) < 0) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not reallocate FIFO(%d,%s)", error,
            av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
    }

    /* Store the new samples in the FIFO buffer. */
    if ((error = av_audio_fifo_write(fifo_, (void **)samples, frame_size)) < frame_size) {
        return srs_error_new(ERROR_RTC_RTP_MUXER, "Could not write data to FIFO(%d,%s)", error,
            av_make_error_string(err_buf, AV_ERROR_MAX_STRING_SIZE, error));
    }

    return srs_success;
}

void SrsAudioTranscoder::free_swr_samples()
{
    if (swr_data_) {
        av_freep(&swr_data_[0]);
        free(swr_data_);
        swr_data_ = NULL;
    }
}