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
 105     int sample_format = sample_fmt_lavc2native(codec->sample_fmt);
 106     if (sample_format == AF_FORMAT_UNKNOWN) {
 107 #ifndef CONFIG_LIBAVRESAMPLE
 108         if (av_sample_fmt_is_planar(codec->sample_fmt))
 109             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 110                    "The player has been compiled without libavresample "
 111                    "support,\nwhich is needed with this libavcodec decoder "
 112                    "version.\nCompile with libavresample enabled to make "
 113                    "audio decoding work!\n");
 114         else
 115             mp_msg(MSGT_DECAUDIO, MSGL_ERR, "Unsupported sample format\n");
 116         goto error;
 117 #else
 118         if (priv->avr && (priv->resample_fmt      != codec->sample_fmt ||
 119                           priv->resample_channels != codec->channels))
 120             avresample_free(&priv->avr);
 121
 122         if (!priv->avr) {
 123             int ret;
 124             uint8_t error[128];
 125             enum AVSampleFormat out_fmt =
 126                 av_get_packed_sample_fmt(codec->sample_fmt);
 127             uint64_t ch_layout = codec->channel_layout;
 128
 129             mp_msg(MSGT_DECAUDIO, MSGL_V,
 130                    "(Re)initializing libavresample format conversion...\n");
 131
 132             if (!ch_layout)
 133                 ch_layout = av_get_default_channel_layout(codec->channels);
 134
 135             /* if lavc format is planar, try just getting packed equivalent */
 136             sample_format = sample_fmt_lavc2native(out_fmt);
 137             if (sample_format == AF_FORMAT_UNKNOWN) {
 138                 /* fallback to s16 */
 139                 out_fmt = AV_SAMPLE_FMT_S16;
 140                 sample_format = AF_FORMAT_S16_NE;
 141             }
 142
 143             priv->avr = avresample_alloc_context();
 144             if (!priv->avr) {
 145                 mp_msg(MSGT_DECAUDIO, MSGL_FATAL, "Out of memory.\n");
 146                 abort();
 147             }
 148             av_opt_set_int(priv->avr, "in_channel_layout",  ch_layout, 0);
 149             av_opt_set_int(priv->avr, "out_channel_layout", ch_layout, 0);
 150             av_opt_set_int(priv->avr, "in_sample_rate",  codec->sample_rate, 0);
 151             av_opt_set_int(priv->avr, "out_sample_rate", codec->sample_rate, 0);
 152             av_opt_set_int(priv->avr, "in_sample_fmt",   codec->sample_fmt, 0);
 153             av_opt_set_int(priv->avr, "out_sample_fmt",  out_fmt, 0);
 154
 155             if ((ret = avresample_open(priv->avr)) < 0) {
 156                 av_strerror(ret, error, sizeof(error));
 157                 mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 158                        "Error opening libavresample: %s.\n", error);
 159                 goto error;
 160             }
 161             priv->resample_fmt      = codec->sample_fmt;
 162             priv->resample_channels = codec->channels;
 163             priv->out_fmt           = out_fmt;
 164             priv->unitsize          = av_get_bytes_per_sample(out_fmt) *
 165                                       codec->channels;
 166         } else
 167             sample_format = sh_audio->sample_format;
 168     } else if (priv->avr) {
 169         avresample_free(&priv->avr);
 170 #endif
 171     }
 172
 173     bool broken_srate        = false;
 174     int samplerate           = codec->sample_rate;
 175     int container_samplerate = sh_audio->container_out_samplerate;
 176     if (!container_samplerate && sh_audio->wf)
 177         container_samplerate = sh_audio->wf->nSamplesPerSec;
 178     if (codec->codec_id == CODEC_ID_AAC
 179         && samplerate == 2 * container_samplerate)
 180         broken_srate = true;
 181     else if (container_samplerate)
 182         samplerate = container_samplerate;
 183
 184     if (codec->channels != sh_audio->channels ||
 185         samplerate != sh_audio->samplerate ||
 186         sample_format != sh_audio->sample_format) {
 187         sh_audio->channels = codec->channels;
 188         sh_audio->samplerate = samplerate;
 189         sh_audio->sample_format = sample_format;
 190         sh_audio->samplesize = af_fmt2bits(sh_audio->sample_format) / 8;
 191         if (broken_srate)
 192             mp_msg(MSGT_DECAUDIO, MSGL_WARN,
 193                    "Ignoring broken container sample rate for AAC with SBR\n");
 194         return 1;
 195     }
 196     return 0;
 197 error:
 198 #ifdef CONFIG_LIBAVRESAMPLE
 199     avresample_free(&priv->avr);
 200 #endif
 201     return -1;
 202 }
 203
 204 static int init(sh_audio_t *sh_audio)
 205 {
 206     struct MPOpts *opts = sh_audio->opts;
 207     AVCodecContext *lavc_context;
 208     AVCodec *lavc_codec;
 209
 210     if (sh_audio->codec->dll) {
 211         lavc_codec = avcodec_find_decoder_by_name(sh_audio->codec->dll);
 212         if (!lavc_codec) {
 213             mp_tmsg(MSGT_DECAUDIO, MSGL_ERR,
 214                     "Cannot find codec '%s' in libavcodec...\n",
 215                     sh_audio->codec->dll);
 216             return 0;
 217         }
 218     } else if (!sh_audio->libav_codec_id) {
 219         mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "No Libav codec ID known. "
 220                 "Generic lavc decoder is not applicable.\n");
 221         return 0;
 222     } else {
 223         lavc_codec = avcodec_find_decoder(sh_audio->libav_codec_id);
 224         if (!lavc_codec) {
 225             mp_tmsg(MSGT_DECAUDIO, MSGL_INFO, "Libavcodec has no decoder "
 226                    "for this codec\n");
 227             return 0;
 228         }
 229     }
 230
 231     sh_audio->codecname = lavc_codec->long_name;
 232     if (!sh_audio->codecname)
 233         sh_audio->codecname = lavc_codec->name;
 234
 235     struct priv *ctx = talloc_zero(NULL, struct priv);
 236     sh_audio->context = ctx;
 237     lavc_context = avcodec_alloc_context3(lavc_codec);
 238     ctx->avctx = lavc_context;
 239     ctx->avframe = avcodec_alloc_frame();
 240
 241     // Always try to set - option only exists for AC3 at the moment
 242     av_opt_set_double(lavc_context, "drc_scale", opts->drc_level,
 243                       AV_OPT_SEARCH_CHILDREN);
 244     lavc_context->sample_rate = sh_audio->samplerate;
 245     lavc_context->bit_rate = sh_audio->i_bps * 8;
 246     if (sh_audio->wf) {
 247         lavc_context->channels = sh_audio->wf->nChannels;
 248         lavc_context->sample_rate = sh_audio->wf->nSamplesPerSec;
 249         lavc_context->bit_rate = sh_audio->wf->nAvgBytesPerSec * 8;
 250         lavc_context->block_align = sh_audio->wf->nBlockAlign;
 251         lavc_context->bits_per_coded_sample = sh_audio->wf->wBitsPerSample;
 252     }
 253     lavc_context->request_channels = opts->audio_output_channels;
 254     lavc_context->codec_tag = sh_audio->format; //FOURCC
 255     lavc_context->codec_type = AVMEDIA_TYPE_AUDIO;
 256     lavc_context->codec_id = lavc_codec->id; // not sure if required, imho not --A'rpi
 257
 258     /* alloc extra data */
 259     if (sh_audio->wf && sh_audio->wf->cbSize > 0) {
 260         lavc_context->extradata = av_mallocz(sh_audio->wf->cbSize + FF_INPUT_BUFFER_PADDING_SIZE);
 261         lavc_context->extradata_size = sh_audio->wf->cbSize;
 262         memcpy(lavc_context->extradata, sh_audio->wf + 1,
 263                lavc_context->extradata_size);
 264     }
 265
 266     // for QDM2
 267     if (sh_audio->codecdata_len && sh_audio->codecdata &&
 268             !lavc_context->extradata) {
 269         lavc_context->extradata = av_malloc(sh_audio->codecdata_len +
 270                                             FF_INPUT_BUFFER_PADDING_SIZE);
 271         lavc_context->extradata_size = sh_audio->codecdata_len;
 272         memcpy(lavc_context->extradata, (char *)sh_audio->codecdata,
 273                lavc_context->extradata_size);
 274     }
 275
 276     /* open it */
 277     if (avcodec_open2(lavc_context, lavc_codec, NULL) < 0) {
 278         mp_tmsg(MSGT_DECAUDIO, MSGL_ERR, "Could not open codec.\n");
 279         uninit(sh_audio);
 280         return 0;
 281     }
 282     mp_msg(MSGT_DECAUDIO, MSGL_V, "INFO: libavcodec \"%s\" init OK!\n",
 283            lavc_codec->name);
 284
 285     if (sh_audio->format == 0x3343414D) {
 286         // MACE 3:1
 287         sh_audio->ds->ss_div = 2 * 3; // 1 samples/packet
 288         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 289     } else if (sh_audio->format == 0x3643414D) {
 290         // MACE 6:1
 291         sh_audio->ds->ss_div = 2 * 6; // 1 samples/packet
 292         sh_audio->ds->ss_mul = 2 * sh_audio->wf->nChannels; // 1 byte*ch/packet
 293     }
 294
 295     // Decode at least 1 byte:  (to get header filled)
 296     for (int tries = 0;;) {
 297         int x = decode_audio(sh_audio, sh_audio->a_buffer, 1,
 298                              sh_audio->a_buffer_size);
 299         if (x > 0) {
 300             sh_audio->a_buffer_len = x;
 301             break;
 302         }
 303         if (++tries >= 5) {
 304             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 305                    "ad_ffmpeg: initial decode failed\n");
 306             uninit(sh_audio);
 307             return 0;
 308         }
 309     }
 310
 311     sh_audio->i_bps = lavc_context->bit_rate / 8;
 312     if (sh_audio->wf && sh_audio->wf->nAvgBytesPerSec)
 313         sh_audio->i_bps = sh_audio->wf->nAvgBytesPerSec;
 314
 315     return 1;
 316 }
 317
 318 static void uninit(sh_audio_t *sh)
 319 {
 320     sh->codecname = NULL;
 321     struct priv *ctx = sh->context;
 322     if (!ctx)
 323         return;
 324     AVCodecContext *lavc_context = ctx->avctx;
 325
 326     if (lavc_context) {
 327         if (avcodec_close(lavc_context) < 0)
 328             mp_tmsg(MSGT_DECVIDEO, MSGL_ERR, "Could not close codec.\n");
 329         av_freep(&lavc_context->extradata);
 330         av_freep(&lavc_context);
 331     }
 332 #ifdef CONFIG_LIBAVRESAMPLE
 333     avresample_free(&ctx->avr);
 334 #endif
 335 #if LIBAVCODEC_VERSION_INT >= (54 << 16 | 28 << 8)
 336     avcodec_free_frame(&ctx->avframe);
 337 #else
 338     av_free(ctx->avframe);
 339 #endif
 340     talloc_free(ctx);
 341     sh->context = NULL;
 342 }
 343
 344 static int control(sh_audio_t *sh, int cmd, void *arg, ...)
 345 {
 346     struct priv *ctx = sh->context;
 347     switch (cmd) {
 348     case ADCTRL_RESYNC_STREAM:
 349         avcodec_flush_buffers(ctx->avctx);
 350         ds_clear_parser(sh->ds);
 351         ctx->previous_data_left = 0;
 352         ctx->output_left = 0;
 353         return CONTROL_TRUE;
 354     }
 355     return CONTROL_UNKNOWN;
 356 }
 357
 358 static int decode_new_packet(struct sh_audio *sh)
 359 {
 360     struct priv *priv = sh->context;
 361     AVCodecContext *avctx = priv->avctx;
 362     double pts = MP_NOPTS_VALUE;
 363     int insize;
 364     bool packet_already_used = priv->previous_data_left;
 365     struct demux_packet *mpkt = ds_get_packet2(sh->ds,
 366                                                priv->previous_data_left);
 367     unsigned char *start;
 368     if (!mpkt) {
 369         assert(!priv->previous_data_left);
 370         start = NULL;
 371         insize = 0;
 372         ds_parse(sh->ds, &start, &insize, pts, 0);
 373         if (insize <= 0)
 374             return -1;  // error or EOF
 375     } else {
 376         assert(mpkt->len >= priv->previous_data_left);
 377         if (!priv->previous_data_left) {
 378             priv->previous_data_left = mpkt->len;
 379             pts = mpkt->pts;
 380         }
 381         insize = priv->previous_data_left;
 382         start = mpkt->buffer + mpkt->len - priv->previous_data_left;
 383         int consumed = ds_parse(sh->ds, &start, &insize, pts, 0);
 384         priv->previous_data_left -= consumed;
 385         priv->previous_data_left = FFMAX(priv->previous_data_left, 0);
 386     }
 387
 388     AVPacket pkt;
 389     av_init_packet(&pkt);
 390     pkt.data = start;
 391     pkt.size = insize;
 392     if (mpkt && mpkt->avpacket) {
 393         pkt.side_data = mpkt->avpacket->side_data;
 394         pkt.side_data_elems = mpkt->avpacket->side_data_elems;
 395     }
 396     if (pts != MP_NOPTS_VALUE && !packet_already_used) {
 397         sh->pts = pts;
 398         sh->pts_bytes = 0;
 399     }
 400     int got_frame = 0;
 401     int ret = avcodec_decode_audio4(avctx, priv->avframe, &got_frame, &pkt);
 402     // LATM may need many packets to find mux info
 403     if (ret == AVERROR(EAGAIN))
 404         return 0;
 405     if (ret < 0) {
 406         mp_msg(MSGT_DECAUDIO, MSGL_V, "lavc_audio: error\n");
 407         return -1;
 408     }
 409     // The "insize >= ret" test is sanity check against decoder overreads
 410     if (!sh->parser && insize >= ret)
 411         priv->previous_data_left = insize - ret;
 412     if (!got_frame)
 413         return 0;
 414
 415     int format_result = setup_format(sh);
 416     if (format_result < 0)
 417         return format_result;
 418
 419 #ifdef CONFIG_LIBAVRESAMPLE
 420     if (priv->avr) {
 421         int ret;
 422         uint64_t needed_size = av_samples_get_buffer_size(
 423                 NULL, priv->resample_channels, priv->avframe->nb_samples,
 424                 priv->resample_fmt, 0);
 425         if (needed_size > priv->resample_buf_size) {
 426             priv->resample_buf = talloc_realloc(priv, priv->resample_buf,
 427                                                 uint8_t, needed_size);
 428             priv->resample_buf_size = needed_size;
 429         }
 430
 431         ret = avresample_convert(priv->avr, &priv->resample_buf,
 432                 priv->resample_buf_size, priv->avframe->nb_samples,
 433                 priv->avframe->extended_data, priv->avframe->linesize[0],
 434                 priv->avframe->nb_samples);
 435         if (ret < 0) {
 436             uint8_t error[128];
 437             av_strerror(ret, error, sizeof(error));
 438             mp_msg(MSGT_DECAUDIO, MSGL_ERR,
 439                    "Error during sample format conversion: %s.\n", error);
 440             return -1;
 441         }
 442
 443         assert(ret == priv->avframe->nb_samples);
 444
 445         priv->output = priv->resample_buf;
 446         priv->output_left = priv->unitsize * ret;
 447     } else
 448 #endif
 449     {
 450         uint64_t unitsize = av_get_bytes_per_sample(avctx->sample_fmt) *
 451                             (uint64_t)avctx->channels;
 452         if (unitsize > 100000)
 453             abort();
 454         priv->unitsize = unitsize;
 455         uint64_t output_left = unitsize * priv->avframe->nb_samples;
 456         if (output_left > 500000000)
 457             abort();
 458         priv->output_left = output_left;
 459         priv->output = priv->avframe->data[0];
 460     }
 461
 462     mp_dbg(MSGT_DECAUDIO, MSGL_DBG2, "Decoded %d -> %d  \n", insize,
 463            priv->output_left);
 464     return format_result;
 465 }
 466
 467
 468 static int decode_audio(sh_audio_t *sh_audio, unsigned char *buf, int minlen,
 469                         int maxlen)
 470 {
 471     struct priv *priv = sh_audio->context;
 472     AVCodecContext *avctx = priv->avctx;
 473
 474     int len = -1;
 475     while (len < minlen) {
 476         if (!priv->output_left) {
 477             if (decode_new_packet(sh_audio) != 0)
 478                 break;
 479             continue;
 480         }
 481         int size = (minlen - len + priv->unitsize - 1);
 482         size -= size % priv->unitsize;
 483         size = FFMIN(size, priv->output_left);
 484         if (size > maxlen)
 485             abort();
 486         memcpy(buf, priv->output, size);
 487         priv->output += size;
 488         priv->output_left -= size;
 489         if (avctx->channels >= 5) {
 490             int samplesize = av_get_bytes_per_sample(avctx->sample_fmt);
 491             reorder_channel_nch(buf, AF_CHANNEL_LAYOUT_LAVC_DEFAULT,
 492                                 AF_CHANNEL_LAYOUT_MPLAYER_DEFAULT,
 493                                 avctx->channels,
 494                                 size / samplesize, samplesize);
 495         }
 496         if (len < 0)
 497             len = size;
 498         else
 499             len += size;
 500         buf += size;
 501         maxlen -= size;
 502         sh_audio->pts_bytes += size;
 503     }
 504     return len;
 505 }