libmpcodecs/ad_ffmpeg.c

   1 /*
   2  * This file is part of MPlayer.
   3  *
   4  * MPlayer is free software; you can redistribute it and/or modify
   5  * it under the terms of the GNU General Public License as published by
   6  * the Free Software Foundation; either version 2 of the License, or
   7  * (at your option) any later version.
   8  *
   9  * MPlayer is distributed in the hope that it will be useful,
  10  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  11  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  12  * GNU General Public License for more details.
  13  *
  14  * You should have received a copy of the GNU General Public License along
  15  * with MPlayer; if not, write to the Free Software Foundation, Inc.,
  16  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
  17  */
  18
  19 #include <stdio.h>
  20 #include <stdlib.h>
  21 #include <unistd.h>
  22 #include <stdbool.h>
  23 #include <assert.h>
  24
  25 #include <libavcodec/avcodec.h>
  26 #include <libavutil/audioconvert.h>
  27 #include <libavutil/opt.h>
  28 #include <libavutil/samplefmt.h>
  29 #include <libavresample/avresample.h>
  30
  31 #include "talloc.h"
  32
  33 #include "config.h"
  34 #include "mp_msg.h"
  35 #include "options.h"
  36
  37 #include "ad_internal.h"
  38 #include "libaf/reorder_ch.h"
  39
  40 #include "mpbswap.h"
  41
  42 static const ad_info_t info =
  43 {
  44     "libavcodec audio decoders",
  45     "ffmpeg",
  46     "",
  47     "",
  48     "",
  49     .print_name = "libavcodec",
  50 };
  51
  52 LIBAD_EXTERN(ffmpeg)
  53
  54 struct priv {
  55     AVCodecContext *avctx;
  56     AVFrame *avframe;
  57     char *output;
  58     int output_left;
  59     int unitsize;
  60     int previous_data_left;  // input demuxer packet data
  61
  62     AVAudioResampleContext *avr;
  63     enum AVSampleFormat resample_fmt;
  64     enum AVSampleFormat out_fmt;
  65     int resample_channels;
  66     uint8_t *resample_buf;
  67     uint64_t resample_buf_size;
  68 };
  69
  70 static int preinit(sh_audio_t *sh)
  71 {
  72     return 1;
  73 }
  74
  75 static const int sample_fmt_map[][2] = {
  76     { AV_SAMPLE_FMT_U8,  AF_FORMAT_U8 },
  77     { AV_SAMPLE_FMT_S16, AF_FORMAT_S16_NE },
  78     { AV_SAMPLE_FMT_S32, AF_FORMAT_S32_NE },
  79     { AV_SAMPLE_FMT_FLT, AF_FORMAT_FLOAT_NE },
  80 };
  81
  82 static int sample_fmt_lavc2native(enum AVSampleFormat sample_fmt)
  83 {
  84     for (int i = 0; i < FF_ARRAY_ELEMS(sample_fmt_map); i++)
  85         if (sample_fmt_map[i][0] == sample_fmt)
  86             return sample_fmt_map[i][1];
  87     return AF_FORMAT_UNKNOWN;
  88 }
  89
  90 /* Prefer playing audio with the samplerate given in container data
  91  * if available, but take number the number of channels and sample format
  92  * from the codec, since if the codec isn't using the correct values for
  93  * those everything breaks anyway.
  94  */
  95 static int setup_format(sh_audio_t *sh_audio)
  96 {
  97     struct priv *priv = sh_audio->context;
  98     AVCodecContext *codec = priv->avctx;
  99
 100     int sample_format = sample_fmt_lavc2native(codec->sample_fmt);
 101     if (sample_format == AF_FORMAT_UNKNOWN) {
 102         if (priv->avr && (priv->resample_fmt      != codec->sample_fmt ||
 103                           priv->resample_channels != codec->channels))
 104             avresample_free(&priv->avr);
 105
 106         if (!priv->avr) {
 107             int ret;
 108             uint8_t error[128];
 109             enum AVSampleFormat out_fmt =
 110                 av_get_packed_sample_fmt(codec->sample_fmt);
 111             uint64_t ch_layout = codec->channel_layout;
 112
 113             mp_msg(MSGT_DECAUDIO, MSGL_V,
 114                    "(Re)initializing libavresample format conversion...\n");
 115
 116             if (!ch_layout)
 117                 ch_layout = av_get_default_channel_layout(codec->channels);
 118
 119             /* if lavc format is planar, try just getting packed equivalent */
 120             sample_format = sample_fmt_lavc2native(out_fmt);
 121             if (sample_format == AF_FORMAT_UNKNOWN) {
 122                 /* fallback to s16 */
 123                 out_fmt = AV_SAMPLE_FMT_S16;
 124                 sample_format = AF_FORMAT_S16_NE;
 125             }
 126
 127             priv->avr = avresample_alloc_context();
 128             if (!priv->avr) {
 129                 mp_msg(MSGT_DECAUDIO, MSGL_FATAL, "Out of memory.\n");
 130                 abort();
 131             }
 132             av_opt_set_int(priv->avr, "in_channel_layout",  ch_layout, 0);
 133             av_opt_set_int(priv->avr, "out_channel_layout", ch_layout, 0);
 134             av_opt_set_int(priv->avr, "in_sample_rate",  codec->sample_rate, 0);
 135             av_opt_set_int(priv->avr, "out_sample_rate", codec->sample_rate, 0);
 136             av_opt_set_int(priv->avr, "in_sample_fmt",   codec->sample_fmt, 0);
 137             av_opt_set_int(priv->avr, "out_sample_fmt",  out_fmt, 0);
 138
 139             if ((ret = avresample_open(priv->avr)) < 0) {
 140                 av_strerror(ret, error, sizeof(error));
 141                 mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 142                        "Error opening libavresample: %s.\n", error);
 143                 goto error;
 144             }
 145             priv->resample_fmt      = codec->sample_fmt;
 146             priv->resample_channels = codec->channels;
 147             priv->out_fmt           = out_fmt;
 148             priv->unitsize          = av_get_bytes_per_sample(out_fmt) *
 149                                       codec->channels;
 150         } else
 151             sample_format = sh_audio->sample_format;
 152     } else if (priv->avr) {
 153         avresample_free(&priv->avr);
 154     }
 155
 156     bool broken_srate        = false;
 157     int samplerate           = codec->sample_rate;
 158     int container_samplerate = sh_audio->container_out_samplerate;
 159     if (!container_samplerate && sh_audio->wf)
 160         container_samplerate = sh_audio->wf->nSamplesPerSec;
 161     if (codec->codec_id == AV_CODEC_ID_AAC
 162         && samplerate == 2 * container_samplerate)
 163         broken_srate = true;
 164     else if (container_samplerate)
 165         samplerate = container_samplerate;
 166
 167     if (codec->channels != sh_audio->channels ||
 168         samplerate != sh_audio->samplerate ||
 169         sample_format != sh_audio->sample_format) {
 170         sh_audio->channels = codec->channels;
 171         sh_audio->samplerate = samplerate;
 172         sh_audio->sample_format = sample_format;
 173         sh_audio->samplesize = af_fmt2bits(sh_audio->sample_format) / 8;
 174         if (broken_srate)
 175             mp_msg(MSGT_DECAUDIO, MSGL_WARN,
 176                    "Ignoring broken container sample rate for AAC with SBR\n");
 177         return 1;
 178     }
 179     return 0;
 180 error:
 181     avresample_free(&priv->avr);
 182     return -1;
 183 }
 184
 185 static int init(sh_audio_t *sh_audio)
 186 {
 187     struct MPOpts *opts = sh_audio->opts;
 188     AVCodecContext *lavc_context;
 189     AVCodec *lavc_codec;
 190
 191     if (sh_audio->codec->dll) {
 192         lavc_codec = avcodec_find_decoder_by_name(sh_audio->codec->dll);
 193         if (!lavc_codec) {
 194             mp_tmsg(MSGT_DECAUDIO, MSGL_ERR,
 195                     "Cannot find codec '%s' in libavcodec...\n",
 196                     sh_audio->codec->dll);
 197             return 0;
 198         }
 199     } else if (!sh_audio->libav_codec_id) {
 200         mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "No Libav codec ID known. "
 201                 "Generic lavc decoder is not applicable.\n");
 202         return 0;
 203     } else {
 204         lavc_codec = avcodec_find_decoder(sh_audio->libav_codec_id);
 205         if (!lavc_codec) {
 206             mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "Libavcodec has no decoder "
 207                    "for this codec\n");
 208             return 0;
 209         }
 210     }
 211
 212     sh_audio->codecname = lavc_codec->long_name;
 213     if (!sh_audio->codecname)
 214         sh_audio->codecname = lavc_codec->name;
 215
 216     struct priv *ctx = talloc_zero(NULL, struct priv);
 217     sh_audio->context = ctx;
 218     lavc_context = avcodec_alloc_context3(lavc_codec);
 219     ctx->avctx = lavc_context;
 220     ctx->avframe = avcodec_alloc_frame();
 221
 222     // Always try to set - option only exists for AC3 at the moment
 223     av_opt_set_double(lavc_context, "drc_scale", opts->drc_level,
 224                       AV_OPT_SEARCH_CHILDREN);
 225     lavc_context->sample_rate = sh_audio->samplerate;
 226     lavc_context->bit_rate = sh_audio->i_bps * 8;
 227     if (sh_audio->wf) {
 228         lavc_context->channels = sh_audio->wf->nChannels;
 229         lavc_context->sample_rate = sh_audio->wf->nSamplesPerSec;
 230         lavc_context->bit_rate = sh_audio->wf->nAvgBytesPerSec * 8;
 231         lavc_context->block_align = sh_audio->wf->nBlockAlign;
 232         lavc_context->bits_per_coded_sample = sh_audio->wf->wBitsPerSample;
 233     }
 234     lavc_context->request_channels = opts->audio_output_channels;
 235     lavc_context->codec_tag = sh_audio->format; //FOURCC
 236     lavc_context->codec_type = AVMEDIA_TYPE_AUDIO;
 237     lavc_context->codec_id = lavc_codec->id; // not sure if required, imho not --A'rpi
 238
 239     /* alloc extra data */
 240     if (sh_audio->wf && sh_audio->wf->cbSize > 0) {
 241         lavc_context->extradata = av_mallocz(sh_audio->wf->cbSize + FF_INPUT_BUFFER_PADDING_SIZE);
 242         lavc_context->extradata_size = sh_audio->wf->cbSize;
 243         memcpy(lavc_context->extradata, sh_audio->wf + 1,
 244                lavc_context->extradata_size);
 245     }
 246
 247     // for QDM2
 248     if (sh_audio->codecdata_len && sh_audio->codecdata &&
 249             !lavc_context->extradata) {
 250         lavc_context->extradata = av_malloc(sh_audio->codecdata_len +
 251                                             FF_INPUT_BUFFER_PADDING_SIZE);
 252         lavc_context->extradata_size = sh_audio->codecdata_len;
 253         memcpy(lavc_context->extradata, (char *)sh_audio->codecdata,
 254                lavc_context->extradata_size);
 255     }
 256
 257     /* open it */
 258     if (avcodec_open2(lavc_context, lavc_codec, NULL) < 0) {
 259         mp_tmsg(MSGT_DECAUDIO, MSGL_ERR, "Could not open codec.\n");
 260         uninit(sh_audio);
 261         return 0;
 262     }
 263     mp_msg(MSGT_DECAUDIO, MSGL_V, "INFO: libavcodec \"%s\" init OK!\n",
 264            lavc_codec->name);
 265
 266     if (sh_audio->format == 0x3343414D) {
 267         // MACE 3:1
 268         sh_audio->ds->ss_div = 2 * 3; // 1 samples/packet
 269         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 270     } else if (sh_audio->format == 0x3643414D) {
 271         // MACE 6:1
 272         sh_audio->ds->ss_div = 2 * 6; // 1 samples/packet
 273         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 274     }
 275
 276     // Decode at least 1 byte:  (to get header filled)
 277     for (int tries = 0;;) {
 278         int x = decode_audio(sh_audio, sh_audio->a_buffer, 1,
 279                              sh_audio->a_buffer_size);
 280         if (x > 0) {
 281             sh_audio->a_buffer_len = x;
 282             break;
 283         }
 284         if (++tries >= 5) {
 285             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 286                    "ad_ffmpeg: initial decode failed\n");
 287             uninit(sh_audio);
 288             return 0;
 289         }
 290     }
 291
 292     sh_audio->i_bps = lavc_context->bit_rate / 8;
 293     if (sh_audio->wf && sh_audio->wf->nAvgBytesPerSec)
 294         sh_audio->i_bps = sh_audio->wf->nAvgBytesPerSec;
 295
 296     return 1;
 297 }
 298
 299 static void uninit(sh_audio_t *sh)
 300 {
 301     sh->codecname = NULL;
 302     struct priv *ctx = sh->context;
 303     if (!ctx)
 304         return;
 305     AVCodecContext *lavc_context = ctx->avctx;
 306
 307     if (lavc_context) {
 308         if (avcodec_close(lavc_context) < 0)
 309             mp_tmsg(MSGT_DECVIDEO, MSGL_ERR, "Could not close codec.\n");
 310         av_freep(&lavc_context->extradata);
 311         av_freep(&lavc_context);
 312     }
 313     avresample_free(&ctx->avr);
 314 #if LIBAVCODEC_VERSION_INT >= (54 << 16 | 28 << 8)
 315     avcodec_free_frame(&ctx->avframe);
 316 #else
 317     av_free(ctx->avframe);
 318 #endif
 319     talloc_free(ctx);
 320     sh->context = NULL;
 321 }
 322
 323 static int control(sh_audio_t *sh, int cmd, void *arg, ...)
 324 {
 325     struct priv *ctx = sh->context;
 326     switch (cmd) {
 327     case ADCTRL_RESYNC_STREAM:
 328         avcodec_flush_buffers(ctx->avctx);
 329         ds_clear_parser(sh->ds);
 330         ctx->previous_data_left = 0;
 331         ctx->output_left = 0;
 332         return CONTROL_TRUE;
 333     }
 334     return CONTROL_UNKNOWN;
 335 }
 336
 337 static int decode_new_packet(struct sh_audio *sh)
 338 {
 339     struct priv *priv = sh->context;
 340     AVCodecContext *avctx = priv->avctx;
 341     double pts = MP_NOPTS_VALUE;
 342     int insize;
 343     bool packet_already_used = priv->previous_data_left;
 344     bool eof = false;
 345     struct demux_packet *mpkt = ds_get_packet2(sh->ds,
 346                                                priv->previous_data_left);
 347     unsigned char *start;
 348     if (!mpkt) {
 349         assert(!priv->previous_data_left);
 350         start = NULL;
 351         insize = 0;
 352         ds_parse(sh->ds, &start, &insize, pts, 0);
 353         if (insize <= 0)
 354             eof = true;
 355     } else {
 356         assert(mpkt->len >= priv->previous_data_left);
 357         if (!priv->previous_data_left) {
 358             priv->previous_data_left = mpkt->len;
 359             pts = mpkt->pts;
 360         }
 361         insize = priv->previous_data_left;
 362         start = mpkt->buffer + mpkt->len - priv->previous_data_left;
 363         int consumed = ds_parse(sh->ds, &start, &insize, pts, 0);
 364         priv->previous_data_left -= consumed;
 365         priv->previous_data_left = FFMAX(priv->previous_data_left, 0);
 366     }
 367
 368     AVPacket pkt;
 369     av_init_packet(&pkt);
 370     pkt.data = start;
 371     pkt.size = insize;
 372     if (mpkt && mpkt->avpacket) {
 373         pkt.side_data = mpkt->avpacket->side_data;
 374         pkt.side_data_elems = mpkt->avpacket->side_data_elems;
 375     }
 376     if (pts != MP_NOPTS_VALUE && !packet_already_used) {
 377         sh->pts = pts;
 378         sh->pts_bytes = 0;
 379     }
 380     int got_frame = 0;
 381     int ret = avcodec_decode_audio4(avctx, priv->avframe, &got_frame, &pkt);
 382     // LATM may need many packets to find mux info
 383     if (ret == AVERROR(EAGAIN))
 384         return 0;
 385     if (ret < 0) {
 386         mp_msg(MSGT_DECAUDIO, MSGL_V, "lavc_audio: error\n");
 387         return -1;
 388     }
 389     // The "insize >= ret" test is sanity check against decoder overreads
 390     if (!sh->parser && insize >= ret)
 391         priv->previous_data_left = insize - ret;
 392     if (!got_frame)
 393         return eof ? -1 : 0;
 394
 395     int format_result = setup_format(sh);
 396     if (format_result < 0)
 397         return format_result;
 398
 399     if (priv->avr) {
 400         int ret;
 401         uint64_t needed_size = av_samples_get_buffer_size(
 402                 NULL, priv->resample_channels, priv->avframe->nb_samples,
 403                 priv->resample_fmt, 0);
 404         if (needed_size > priv->resample_buf_size) {
 405             priv->resample_buf = talloc_realloc(priv, priv->resample_buf,
 406                                                 uint8_t, needed_size);
 407             priv->resample_buf_size = needed_size;
 408         }
 409
 410         ret = avresample_convert(priv->avr, &priv->resample_buf,
 411                 priv->resample_buf_size, priv->avframe->nb_samples,
 412                 priv->avframe->extended_data, priv->avframe->linesize[0],
 413                 priv->avframe->nb_samples);
 414         if (ret < 0) {
 415             uint8_t error[128];
 416             av_strerror(ret, error, sizeof(error));
 417             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 418                    "Error during sample format conversion: %s.\n", error);
 419             return -1;
 420         }
 421
 422         assert(ret == priv->avframe->nb_samples);
 423
 424         priv->output = priv->resample_buf;
 425         priv->output_left = priv->unitsize * ret;
 426     } else {
 427         uint64_t unitsize = av_get_bytes_per_sample(avctx->sample_fmt) *
 428                             (uint64_t)avctx->channels;
 429         if (unitsize > 100000)
 430             abort();
 431         priv->unitsize = unitsize;
 432         uint64_t output_left = unitsize * priv->avframe->nb_samples;
 433         if (output_left > 500000000)
 434             abort();
 435         priv->output_left = output_left;
 436         priv->output = priv->avframe->data[0];
 437     }
 438
 439     mp_dbg(MSGT_DECAUDIO, MSGL_DBG2, "Decoded %d -> %d  \n", insize,
 440            priv->output_left);
 441     return format_result;
 442 }
 443
 444
 445 static int decode_audio(sh_audio_t *sh_audio, unsigned char *buf, int minlen,
 446                         int maxlen)
 447 {
 448     struct priv *priv = sh_audio->context;
 449     AVCodecContext *avctx = priv->avctx;
 450
 451     int len = -1;
 452     while (len < minlen) {
 453         if (!priv->output_left) {
 454             if (decode_new_packet(sh_audio) != 0)
 455                 break;
 456             continue;
 457         }
 458         int size = (minlen - len + priv->unitsize - 1);
 459         size -= size % priv->unitsize;
 460         size = FFMIN(size, priv->output_left);
 461         if (size > maxlen)
 462             abort();
 463         memcpy(buf, priv->output, size);
 464         priv->output += size;
 465         priv->output_left -= size;
 466         if (avctx->channels >= 5) {
 467             int samplesize = av_get_bytes_per_sample(avctx->sample_fmt);
 468             reorder_channel_nch(buf, AF_CHANNEL_LAYOUT_LAVC_DEFAULT,
 469                                 AF_CHANNEL_LAYOUT_MPLAYER_DEFAULT,
 470                                 avctx->channels,
 471                                 size / samplesize, samplesize);
 472         }
 473         if (len < 0)
 474             len = size;
 475         else
 476             len += size;
 477         buf += size;
 478         maxlen -= size;
 479         sh_audio->pts_bytes += size;
 480     }
 481     return len;
 482 }