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
  30 #include "talloc.h"
  31
  32 #include "config.h"
  33 #include "mp_msg.h"
  34 #include "options.h"
  35
  36 #include "ad_internal.h"
  37 #include "libaf/reorder_ch.h"
  38
  39 #include "mpbswap.h"
  40
  41 #ifdef CONFIG_LIBAVRESAMPLE
  42 #include <libavresample/avresample.h>
  43 #endif
  44
  45 static const ad_info_t info =
  46 {
  47     "libavcodec audio decoders",
  48     "ffmpeg",
  49     "",
  50     "",
  51     "",
  52     .print_name = "libavcodec",
  53 };
  54
  55 LIBAD_EXTERN(ffmpeg)
  56
  57 struct priv {
  58     AVCodecContext *avctx;
  59     AVFrame *avframe;
  60     char *output;
  61     int output_left;
  62     int unitsize;
  63     int previous_data_left;  // input demuxer packet data
  64
  65 #ifdef CONFIG_LIBAVRESAMPLE
  66     AVAudioResampleContext *avr;
  67     enum AVSampleFormat resample_fmt;
  68     enum AVSampleFormat out_fmt;
  69     int resample_channels;
  70     uint8_t *resample_buf;
  71     uint64_t resample_buf_size;
  72 #endif
  73 };
  74
  75 static int preinit(sh_audio_t *sh)
  76 {
  77     return 1;
  78 }
  79
  80 static const int sample_fmt_map[][2] = {
  81     { AV_SAMPLE_FMT_U8,  AF_FORMAT_U8 },
  82     { AV_SAMPLE_FMT_S16, AF_FORMAT_S16_NE },
  83     { AV_SAMPLE_FMT_S32, AF_FORMAT_S32_NE },
  84     { AV_SAMPLE_FMT_FLT, AF_FORMAT_FLOAT_NE },
  85 };
  86
  87 static int sample_fmt_lavc2native(enum AVSampleFormat sample_fmt)
  88 {
  89     for (int i = 0; i < FF_ARRAY_ELEMS(sample_fmt_map); i++)
  90         if (sample_fmt_map[i][0] == sample_fmt)
  91             return sample_fmt_map[i][1];
  92     return AF_FORMAT_UNKNOWN;
  93 }
  94
  95 /* Prefer playing audio with the samplerate given in container data
  96  * if available, but take number the number of channels and sample format
  97  * from the codec, since if the codec isn't using the correct values for
  98  * those everything breaks anyway.
  99  */
 100 static int setup_format(sh_audio_t *sh_audio)
 101 {
 102     struct priv *priv = sh_audio->context;
 103     AVCodecContext *codec = priv->avctx;
 104     int sample_format = sh_audio->sample_format;
 105
 106     sample_format = sample_fmt_lavc2native(codec->sample_fmt);
 107     if (sample_format == AF_FORMAT_UNKNOWN) {
 108 #ifndef CONFIG_LIBAVRESAMPLE
 109         mp_msg(MSGT_DECAUDIO, MSGL_FATAL, "Unsupported sample format\n");
 110         sample_format = AF_FORMAT_UNKNOWN;
 111 #else
 112         if (priv->avr && (priv->resample_fmt      != codec->sample_fmt ||
 113                           priv->resample_channels != codec->channels))
 114             avresample_free(&priv->avr);
 115
 116         if (!priv->avr) {
 117             int ret;
 118             uint8_t error[128];
 119             enum AVSampleFormat out_fmt =
 120                 av_get_packed_sample_fmt(codec->sample_fmt);
 121             uint64_t ch_layout = codec->channel_layout;
 122
 123             mp_msg(MSGT_DECAUDIO, MSGL_V,
 124                    "(Re)initializing libavresample format conversion...\n");
 125
 126             if (!ch_layout)
 127                 ch_layout = av_get_default_channel_layout(codec->channels);
 128
 129             /* if lavc format is planar, try just getting packed equivalent */
 130             sample_format = sample_fmt_lavc2native(out_fmt);
 131             if (sample_format == AF_FORMAT_UNKNOWN) {
 132                 /* fallback to s16 */
 133                 out_fmt = AV_SAMPLE_FMT_S16;
 134                 sample_format = AF_FORMAT_S16_NE;
 135             }
 136
 137             priv->avr = avresample_alloc_context();
 138             if (!priv->avr) {
 139                 mp_msg(MSGT_DECAUDIO, MSGL_FATAL, "Out of memory.\n");
 140                 abort();
 141             }
 142             av_opt_set_int(priv->avr, "in_channel_layout",  ch_layout, 0);
 143             av_opt_set_int(priv->avr, "out_channel_layout", ch_layout, 0);
 144             av_opt_set_int(priv->avr, "in_sample_rate",  codec->sample_rate, 0);
 145             av_opt_set_int(priv->avr, "out_sample_rate", codec->sample_rate, 0);
 146             av_opt_set_int(priv->avr, "in_sample_fmt",   codec->sample_fmt, 0);
 147             av_opt_set_int(priv->avr, "out_sample_fmt",  out_fmt, 0);
 148
 149             if ((ret = avresample_open(priv->avr)) < 0) {
 150                 av_strerror(ret, error, sizeof(error));
 151                 mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 152                        "Error opening libavresample: %s.\n", error);
 153             }
 154             priv->resample_fmt      = codec->sample_fmt;
 155             priv->resample_channels = codec->channels;
 156             priv->out_fmt           = out_fmt;
 157             priv->unitsize          = av_get_bytes_per_sample(out_fmt) *
 158                                       codec->channels;
 159         } else
 160             sample_format = sh_audio->sample_format;
 161     } else if (priv->avr) {
 162         avresample_free(&priv->avr);
 163 #endif
 164     }
 165
 166     bool broken_srate        = false;
 167     int samplerate           = codec->sample_rate;
 168     int container_samplerate = sh_audio->container_out_samplerate;
 169     if (!container_samplerate && sh_audio->wf)
 170         container_samplerate = sh_audio->wf->nSamplesPerSec;
 171     if (codec->codec_id == CODEC_ID_AAC
 172         && samplerate == 2 * container_samplerate)
 173         broken_srate = true;
 174     else if (container_samplerate)
 175         samplerate = container_samplerate;
 176
 177     if (codec->channels != sh_audio->channels ||
 178         samplerate != sh_audio->samplerate ||
 179         sample_format != sh_audio->sample_format) {
 180         sh_audio->channels = codec->channels;
 181         sh_audio->samplerate = samplerate;
 182         sh_audio->sample_format = sample_format;
 183         sh_audio->samplesize = af_fmt2bits(sh_audio->sample_format) / 8;
 184         if (broken_srate)
 185             mp_msg(MSGT_DECAUDIO, MSGL_WARN,
 186                    "Ignoring broken container sample rate for AAC with SBR\n");
 187         return 1;
 188     }
 189     return 0;
 190 }
 191
 192 static int init(sh_audio_t *sh_audio)
 193 {
 194     struct MPOpts *opts = sh_audio->opts;
 195     AVCodecContext *lavc_context;
 196     AVCodec *lavc_codec;
 197
 198     if (sh_audio->codec->dll) {
 199         lavc_codec = avcodec_find_decoder_by_name(sh_audio->codec->dll);
 200         if (!lavc_codec) {
 201             mp_tmsg(MSGT_DECAUDIO, MSGL_ERR,
 202                     "Cannot find codec '%s' in libavcodec...\n",
 203                     sh_audio->codec->dll);
 204             return 0;
 205         }
 206     } else if (!sh_audio->libav_codec_id) {
 207         mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "No Libav codec ID known. "
 208                 "Generic lavc decoder is not applicable.\n");
 209         return 0;
 210     } else {
 211         lavc_codec = avcodec_find_decoder(sh_audio->libav_codec_id);
 212         if (!lavc_codec) {
 213             mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "Libavcodec has no decoder "
 214                    "for this codec\n");
 215             return 0;
 216         }
 217     }
 218
 219     sh_audio->codecname = lavc_codec->long_name;
 220     if (!sh_audio->codecname)
 221         sh_audio->codecname = lavc_codec->name;
 222
 223     struct priv *ctx = talloc_zero(NULL, struct priv);
 224     sh_audio->context = ctx;
 225     lavc_context = avcodec_alloc_context3(lavc_codec);
 226     ctx->avctx = lavc_context;
 227     ctx->avframe = avcodec_alloc_frame();
 228
 229     // Always try to set - option only exists for AC3 at the moment
 230     av_opt_set_double(lavc_context, "drc_scale", opts->drc_level,
 231                       AV_OPT_SEARCH_CHILDREN);
 232     lavc_context->sample_rate = sh_audio->samplerate;
 233     lavc_context->bit_rate = sh_audio->i_bps * 8;
 234     if (sh_audio->wf) {
 235         lavc_context->channels = sh_audio->wf->nChannels;
 236         lavc_context->sample_rate = sh_audio->wf->nSamplesPerSec;
 237         lavc_context->bit_rate = sh_audio->wf->nAvgBytesPerSec * 8;
 238         lavc_context->block_align = sh_audio->wf->nBlockAlign;
 239         lavc_context->bits_per_coded_sample = sh_audio->wf->wBitsPerSample;
 240     }
 241     lavc_context->request_channels = opts->audio_output_channels;
 242     lavc_context->codec_tag = sh_audio->format; //FOURCC
 243     lavc_context->codec_type = AVMEDIA_TYPE_AUDIO;
 244     lavc_context->codec_id = lavc_codec->id; // not sure if required, imho not --A'rpi
 245
 246     /* alloc extra data */
 247     if (sh_audio->wf && sh_audio->wf->cbSize > 0) {
 248         lavc_context->extradata = av_mallocz(sh_audio->wf->cbSize + FF_INPUT_BUFFER_PADDING_SIZE);
 249         lavc_context->extradata_size = sh_audio->wf->cbSize;
 250         memcpy(lavc_context->extradata, sh_audio->wf + 1,
 251                lavc_context->extradata_size);
 252     }
 253
 254     // for QDM2
 255     if (sh_audio->codecdata_len && sh_audio->codecdata &&
 256             !lavc_context->extradata) {
 257         lavc_context->extradata = av_malloc(sh_audio->codecdata_len +
 258                                             FF_INPUT_BUFFER_PADDING_SIZE);
 259         lavc_context->extradata_size = sh_audio->codecdata_len;
 260         memcpy(lavc_context->extradata, (char *)sh_audio->codecdata,
 261                lavc_context->extradata_size);
 262     }
 263
 264     /* open it */
 265     if (avcodec_open2(lavc_context, lavc_codec, NULL) < 0) {
 266         mp_tmsg(MSGT_DECAUDIO, MSGL_ERR, "Could not open codec.\n");
 267         uninit(sh_audio);
 268         return 0;
 269     }
 270     mp_msg(MSGT_DECAUDIO, MSGL_V, "INFO: libavcodec \"%s\" init OK!\n",
 271            lavc_codec->name);
 272
 273     if (sh_audio->format == 0x3343414D) {
 274         // MACE 3:1
 275         sh_audio->ds->ss_div = 2 * 3; // 1 samples/packet
 276         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 277     } else if (sh_audio->format == 0x3643414D) {
 278         // MACE 6:1
 279         sh_audio->ds->ss_div = 2 * 6; // 1 samples/packet
 280         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 281     }
 282
 283     // Decode at least 1 byte:  (to get header filled)
 284     for (int tries = 0;;) {
 285         int x = decode_audio(sh_audio, sh_audio->a_buffer, 1,
 286                              sh_audio->a_buffer_size);
 287         if (x > 0) {
 288             sh_audio->a_buffer_len = x;
 289             break;
 290         }
 291         if (++tries >= 5) {
 292             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 293                    "ad_ffmpeg: initial decode failed\n");
 294             uninit(sh_audio);
 295             return 0;
 296         }
 297     }
 298
 299     sh_audio->i_bps = lavc_context->bit_rate / 8;
 300     if (sh_audio->wf && sh_audio->wf->nAvgBytesPerSec)
 301         sh_audio->i_bps = sh_audio->wf->nAvgBytesPerSec;
 302
 303     return 1;
 304 }
 305
 306 static void uninit(sh_audio_t *sh)
 307 {
 308     sh->codecname = NULL;
 309     struct priv *ctx = sh->context;
 310     if (!ctx)
 311         return;
 312     AVCodecContext *lavc_context = ctx->avctx;
 313
 314     if (lavc_context) {
 315         if (avcodec_close(lavc_context) < 0)
 316             mp_tmsg(MSGT_DECVIDEO, MSGL_ERR, "Could not close codec.\n");
 317         av_freep(&lavc_context->extradata);
 318         av_freep(&lavc_context);
 319     }
 320 #ifdef CONFIG_LIBAVRESAMPLE
 321     avresample_free(&ctx->avr);
 322 #endif
 323 #if LIBAVCODEC_VERSION_INT >= (54 << 16 | 28 << 8)
 324     avcodec_free_frame(&ctx->avframe);
 325 #else
 326     av_free(ctx->avframe);
 327 #endif
 328     talloc_free(ctx);
 329     sh->context = NULL;
 330 }
 331
 332 static int control(sh_audio_t *sh, int cmd, void *arg, ...)
 333 {
 334     struct priv *ctx = sh->context;
 335     switch (cmd) {
 336     case ADCTRL_RESYNC_STREAM:
 337         avcodec_flush_buffers(ctx->avctx);
 338         ds_clear_parser(sh->ds);
 339         ctx->previous_data_left = 0;
 340         ctx->output_left = 0;
 341         return CONTROL_TRUE;
 342     }
 343     return CONTROL_UNKNOWN;
 344 }
 345
 346 static int decode_new_packet(struct sh_audio *sh)
 347 {
 348     struct priv *priv = sh->context;
 349     AVCodecContext *avctx = priv->avctx;
 350     double pts = MP_NOPTS_VALUE;
 351     int insize;
 352     bool packet_already_used = priv->previous_data_left;
 353     struct demux_packet *mpkt = ds_get_packet2(sh->ds,
 354                                                priv->previous_data_left);
 355     unsigned char *start;
 356     if (!mpkt) {
 357         assert(!priv->previous_data_left);
 358         start = NULL;
 359         insize = 0;
 360         ds_parse(sh->ds, &start, &insize, pts, 0);
 361         if (insize <= 0)
 362             return -1;  // error or EOF
 363     } else {
 364         assert(mpkt->len >= priv->previous_data_left);
 365         if (!priv->previous_data_left) {
 366             priv->previous_data_left = mpkt->len;
 367             pts = mpkt->pts;
 368         }
 369         insize = priv->previous_data_left;
 370         start = mpkt->buffer + mpkt->len - priv->previous_data_left;
 371         int consumed = ds_parse(sh->ds, &start, &insize, pts, 0);
 372         priv->previous_data_left -= consumed;
 373         priv->previous_data_left = FFMAX(priv->previous_data_left, 0);
 374     }
 375
 376     AVPacket pkt;
 377     av_init_packet(&pkt);
 378     pkt.data = start;
 379     pkt.size = insize;
 380     if (mpkt && mpkt->avpacket) {
 381         pkt.side_data = mpkt->avpacket->side_data;
 382         pkt.side_data_elems = mpkt->avpacket->side_data_elems;
 383     }
 384     if (pts != MP_NOPTS_VALUE && !packet_already_used) {
 385         sh->pts = pts;
 386         sh->pts_bytes = 0;
 387     }
 388     int got_frame = 0;
 389     int ret = avcodec_decode_audio4(avctx, priv->avframe, &got_frame, &pkt);
 390     // LATM may need many packets to find mux info
 391     if (ret == AVERROR(EAGAIN))
 392         return 0;
 393     if (ret < 0) {
 394         mp_msg(MSGT_DECAUDIO, MSGL_V, "lavc_audio: error\n");
 395         return -1;
 396     }
 397     // The "insize >= ret" test is sanity check against decoder overreads
 398     if (!sh->parser && insize >= ret)
 399         priv->previous_data_left = insize - ret;
 400     if (!got_frame)
 401         return 0;
 402
 403     setup_format(sh);
 404
 405 #ifdef CONFIG_LIBAVRESAMPLE
 406     if (priv->avr) {
 407         int ret;
 408         uint64_t needed_size = av_samples_get_buffer_size(NULL, priv->resample_channels,
 409                                                           priv->avframe->nb_samples,
 410                                                           priv->resample_fmt, 0);
 411         if (needed_size > priv->resample_buf_size) {
 412             priv->resample_buf = talloc_realloc(priv, priv->resample_buf,
 413                                                 uint8_t, needed_size);
 414             priv->resample_buf_size = needed_size;
 415         }
 416
 417         ret = avresample_convert(priv->avr, &priv->resample_buf,
 418                                  priv->resample_buf_size, priv->avframe->nb_samples,
 419                                  priv->avframe->extended_data, priv->avframe->linesize[0],
 420                                  priv->avframe->nb_samples);
 421         if (ret < 0) {
 422             uint8_t error[128];
 423             av_strerror(ret, error, sizeof(error));
 424             mp_msg(MSGT_DECAUDIO, MSGL_ERR, "Error during sample format conversion: %s.\n",
 425                    error);
 426             return -1;
 427         }
 428
 429         assert(ret == priv->avframe->nb_samples);
 430
 431         priv->output = priv->resample_buf;
 432         priv->output_left = priv->unitsize * ret;
 433     } else {
 434 #else
 435     /* An error is reported later from output format checking, but make
 436      * sure we don't crash by overreading first plane. */
 437     if (!av_sample_fmt_is_planar(avctx->sample_fmt) || avctx->channels == 1) {
 438 #endif
 439         uint64_t unitsize = (uint64_t)av_get_bytes_per_sample(avctx->sample_fmt) *
 440                             avctx->channels;
 441         if (unitsize > 100000)
 442             abort();
 443         priv->unitsize = unitsize;
 444         uint64_t output_left = unitsize * priv->avframe->nb_samples;
 445         if (output_left > 500000000)
 446             abort();
 447         priv->output_left = output_left;
 448         priv->output = priv->avframe->data[0];
 449     }
 450
 451     mp_dbg(MSGT_DECAUDIO, MSGL_DBG2, "Decoded %d -> %d  \n", insize,
 452            priv->output_left);
 453     return 0;
 454 }
 455
 456
 457 static int decode_audio(sh_audio_t *sh_audio, unsigned char *buf, int minlen,
 458                         int maxlen)
 459 {
 460     struct priv *priv = sh_audio->context;
 461     AVCodecContext *avctx = priv->avctx;
 462
 463     int len = -1;
 464     while (len < minlen) {
 465         if (!priv->output_left) {
 466             if (decode_new_packet(sh_audio) < 0)
 467                 break;
 468             continue;
 469         }
 470         int size = (minlen - len + priv->unitsize - 1);
 471         size -= size % priv->unitsize;
 472         size = FFMIN(size, priv->output_left);
 473         if (size > maxlen)
 474             abort();
 475         memcpy(buf, priv->output, size);
 476         priv->output += size;
 477         priv->output_left -= size;
 478         if (avctx->channels >= 5) {
 479             int samplesize = av_get_bytes_per_sample(avctx->sample_fmt);
 480             reorder_channel_nch(buf, AF_CHANNEL_LAYOUT_LAVC_DEFAULT,
 481                                 AF_CHANNEL_LAYOUT_MPLAYER_DEFAULT,
 482                                 avctx->channels,
 483                                 size / samplesize, samplesize);
 484         }
 485         if (len < 0)
 486             len = size;
 487         else
 488             len += size;
 489         buf += size;
 490         maxlen -= size;
 491         sh_audio->pts_bytes += size;
 492     }
 493     return len;
 494 }