ffmpeg/libavdevice/dshow.c

   1 /*
   2  * Directshow capture interface
   3  * Copyright (c) 2010 Ramiro Polla
   4  *
   5  * This file is part of FFmpeg.
   6  *
   7  * FFmpeg is free software; you can redistribute it and/or
   8  * modify it under the terms of the GNU Lesser General Public
   9  * License as published by the Free Software Foundation; either
  10  * version 2.1 of the License, or (at your option) any later version.
  11  *
  12  * FFmpeg is distributed in the hope that it will be useful,
  13  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  14  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  15  * Lesser General Public License for more details.
  16  *
  17  * You should have received a copy of the GNU Lesser General Public
  18  * License along with FFmpeg; if not, write to the Free Software
  19  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  20  */
  21
  22 #include "dshow_capture.h"
  23 #include "libavutil/parseutils.h"
  24 #include "libavutil/pixdesc.h"
  25 #include "libavutil/opt.h"
  26 #include "libavformat/internal.h"
  27 #include "libavformat/riff.h"
  28 #include "avdevice.h"
  29 #include "libavcodec/raw.h"
  30
  31 struct dshow_ctx {
  32     const AVClass *class;
  33
  34     IGraphBuilder *graph;
  35
  36     char *device_name[2];
  37     int video_device_number;
  38     int audio_device_number;
  39
  40     int   list_options;
  41     int   list_devices;
  42     int   audio_buffer_size;
  43
  44     IBaseFilter *device_filter[2];
  45     IPin        *device_pin[2];
  46     libAVFilter *capture_filter[2];
  47     libAVPin    *capture_pin[2];
  48
  49     HANDLE mutex;
  50     HANDLE event[2]; /* event[0] is set by DirectShow
  51                       * event[1] is set by callback() */
  52     AVPacketList *pktl;
  53
  54     int eof;
  55
  56     int64_t curbufsize[2];
  57     unsigned int video_frame_num;
  58
  59     IMediaControl *control;
  60     IMediaEvent *media_event;
  61
  62     enum AVPixelFormat pixel_format;
  63     enum AVCodecID video_codec_id;
  64     char *framerate;
  65
  66     int requested_width;
  67     int requested_height;
  68     AVRational requested_framerate;
  69
  70     int sample_rate;
  71     int sample_size;
  72     int channels;
  73 };
  74
  75 static enum AVPixelFormat dshow_pixfmt(DWORD biCompression, WORD biBitCount)
  76 {
  77     switch(biCompression) {
  78     case BI_BITFIELDS:
  79     case BI_RGB:
  80         switch(biBitCount) { /* 1-8 are untested */
  81             case 1:
  82                 return AV_PIX_FMT_MONOWHITE;
  83             case 4:
  84                 return AV_PIX_FMT_RGB4;
  85             case 8:
  86                 return AV_PIX_FMT_RGB8;
  87             case 16:
  88                 return AV_PIX_FMT_RGB555;
  89             case 24:
  90                 return AV_PIX_FMT_BGR24;
  91             case 32:
  92                 return AV_PIX_FMT_0RGB32;
  93         }
  94     }
  95     return avpriv_find_pix_fmt(avpriv_get_raw_pix_fmt_tags(), biCompression); // all others
  96 }
  97
  98 static int
  99 dshow_read_close(AVFormatContext *s)
 100 {
 101     struct dshow_ctx *ctx = s->priv_data;
 102     AVPacketList *pktl;
 103
 104     if (ctx->control) {
 105         IMediaControl_Stop(ctx->control);
 106         IMediaControl_Release(ctx->control);
 107     }
 108
 109     if (ctx->media_event)
 110         IMediaEvent_Release(ctx->media_event);
 111
 112     if (ctx->graph) {
 113         IEnumFilters *fenum;
 114         int r;
 115         r = IGraphBuilder_EnumFilters(ctx->graph, &fenum);
 116         if (r == S_OK) {
 117             IBaseFilter *f;
 118             IEnumFilters_Reset(fenum);
 119             while (IEnumFilters_Next(fenum, 1, &f, NULL) == S_OK) {
 120                 if (IGraphBuilder_RemoveFilter(ctx->graph, f) == S_OK)
 121                     IEnumFilters_Reset(fenum); /* When a filter is removed,
 122                                                 * the list must be reset. */
 123                 IBaseFilter_Release(f);
 124             }
 125             IEnumFilters_Release(fenum);
 126         }
 127         IGraphBuilder_Release(ctx->graph);
 128     }
 129
 130     if (ctx->capture_pin[VideoDevice])
 131         libAVPin_Release(ctx->capture_pin[VideoDevice]);
 132     if (ctx->capture_pin[AudioDevice])
 133         libAVPin_Release(ctx->capture_pin[AudioDevice]);
 134     if (ctx->capture_filter[VideoDevice])
 135         libAVFilter_Release(ctx->capture_filter[VideoDevice]);
 136     if (ctx->capture_filter[AudioDevice])
 137         libAVFilter_Release(ctx->capture_filter[AudioDevice]);
 138
 139     if (ctx->device_pin[VideoDevice])
 140         IPin_Release(ctx->device_pin[VideoDevice]);
 141     if (ctx->device_pin[AudioDevice])
 142         IPin_Release(ctx->device_pin[AudioDevice]);
 143     if (ctx->device_filter[VideoDevice])
 144         IBaseFilter_Release(ctx->device_filter[VideoDevice]);
 145     if (ctx->device_filter[AudioDevice])
 146         IBaseFilter_Release(ctx->device_filter[AudioDevice]);
 147
 148     if (ctx->device_name[0])
 149         av_free(ctx->device_name[0]);
 150     if (ctx->device_name[1])
 151         av_free(ctx->device_name[1]);
 152
 153     if(ctx->mutex)
 154         CloseHandle(ctx->mutex);
 155     if(ctx->event[0])
 156         CloseHandle(ctx->event[0]);
 157     if(ctx->event[1])
 158         CloseHandle(ctx->event[1]);
 159
 160     pktl = ctx->pktl;
 161     while (pktl) {
 162         AVPacketList *next = pktl->next;
 163         av_destruct_packet(&pktl->pkt);
 164         av_free(pktl);
 165         pktl = next;
 166     }
 167
 168     CoUninitialize();
 169
 170     return 0;
 171 }
 172
 173 static char *dup_wchar_to_utf8(wchar_t *w)
 174 {
 175     char *s = NULL;
 176     int l = WideCharToMultiByte(CP_UTF8, 0, w, -1, 0, 0, 0, 0);
 177     s = av_malloc(l);
 178     if (s)
 179         WideCharToMultiByte(CP_UTF8, 0, w, -1, s, l, 0, 0);
 180     return s;
 181 }
 182
 183 static int shall_we_drop(AVFormatContext *s, int index, enum dshowDeviceType devtype)
 184 {
 185     struct dshow_ctx *ctx = s->priv_data;
 186     static const uint8_t dropscore[] = {62, 75, 87, 100};
 187     const int ndropscores = FF_ARRAY_ELEMS(dropscore);
 188     unsigned int buffer_fullness = (ctx->curbufsize[index]*100)/s->max_picture_buffer;
 189
 190     if(dropscore[++ctx->video_frame_num%ndropscores] <= buffer_fullness) {
 191         av_log(s, AV_LOG_ERROR,
 192               "real-time buffer[%s] too full (%d%% of size: %d)! frame dropped!\n", ctx->device_name[devtype], buffer_fullness, s->max_picture_buffer);
 193         return 1;
 194     }
 195
 196     return 0;
 197 }
 198
 199 static void
 200 callback(void *priv_data, int index, uint8_t *buf, int buf_size, int64_t time, enum dshowDeviceType devtype)
 201 {
 202     AVFormatContext *s = priv_data;
 203     struct dshow_ctx *ctx = s->priv_data;
 204     AVPacketList **ppktl, *pktl_next;
 205
 206 //    dump_videohdr(s, vdhdr);
 207
 208     WaitForSingleObject(ctx->mutex, INFINITE);
 209
 210     if(shall_we_drop(s, index, devtype))
 211         goto fail;
 212
 213     pktl_next = av_mallocz(sizeof(AVPacketList));
 214     if(!pktl_next)
 215         goto fail;
 216
 217     if(av_new_packet(&pktl_next->pkt, buf_size) < 0) {
 218         av_free(pktl_next);
 219         goto fail;
 220     }
 221
 222     pktl_next->pkt.stream_index = index;
 223     pktl_next->pkt.pts = time;
 224     memcpy(pktl_next->pkt.data, buf, buf_size);
 225
 226     for(ppktl = &ctx->pktl ; *ppktl ; ppktl = &(*ppktl)->next);
 227     *ppktl = pktl_next;
 228     ctx->curbufsize[index] += buf_size;
 229
 230     SetEvent(ctx->event[1]);
 231     ReleaseMutex(ctx->mutex);
 232
 233     return;
 234 fail:
 235     ReleaseMutex(ctx->mutex);
 236     return;
 237 }
 238
 239 /**
 240  * Cycle through available devices using the device enumerator devenum,
 241  * retrieve the device with type specified by devtype and return the
 242  * pointer to the object found in *pfilter.
 243  * If pfilter is NULL, list all device names.
 244  */
 245 static int
 246 dshow_cycle_devices(AVFormatContext *avctx, ICreateDevEnum *devenum,
 247                     enum dshowDeviceType devtype, IBaseFilter **pfilter)
 248 {
 249     struct dshow_ctx *ctx = avctx->priv_data;
 250     IBaseFilter *device_filter = NULL;
 251     IEnumMoniker *classenum = NULL;
 252     IMoniker *m = NULL;
 253     const char *device_name = ctx->device_name[devtype];
 254     int skip = (devtype == VideoDevice) ? ctx->video_device_number
 255                                         : ctx->audio_device_number;
 256     int r;
 257
 258     const GUID *device_guid[2] = { &CLSID_VideoInputDeviceCategory,
 259                                    &CLSID_AudioInputDeviceCategory };
 260     const char *devtypename = (devtype == VideoDevice) ? "video" : "audio";
 261
 262     r = ICreateDevEnum_CreateClassEnumerator(devenum, device_guid[devtype],
 263                                              (IEnumMoniker **) &classenum, 0);
 264     if (r != S_OK) {
 265         av_log(avctx, AV_LOG_ERROR, "Could not enumerate %s devices.\n",
 266                devtypename);
 267         return AVERROR(EIO);
 268     }
 269
 270     while (!device_filter && IEnumMoniker_Next(classenum, 1, &m, NULL) == S_OK) {
 271         IPropertyBag *bag = NULL;
 272         char *buf = NULL;
 273         VARIANT var;
 274
 275         r = IMoniker_BindToStorage(m, 0, 0, &IID_IPropertyBag, (void *) &bag);
 276         if (r != S_OK)
 277             goto fail1;
 278
 279         var.vt = VT_BSTR;
 280         r = IPropertyBag_Read(bag, L"FriendlyName", &var, NULL);
 281         if (r != S_OK)
 282             goto fail1;
 283
 284         buf = dup_wchar_to_utf8(var.bstrVal);
 285
 286         if (pfilter) {
 287             if (strcmp(device_name, buf))
 288                 goto fail1;
 289
 290             if (!skip--)
 291                 IMoniker_BindToObject(m, 0, 0, &IID_IBaseFilter, (void *) &device_filter);
 292         } else {
 293             av_log(avctx, AV_LOG_INFO, " \"%s\"\n", buf);
 294         }
 295
 296 fail1:
 297         if (buf)
 298             av_free(buf);
 299         if (bag)
 300             IPropertyBag_Release(bag);
 301         IMoniker_Release(m);
 302     }
 303
 304     IEnumMoniker_Release(classenum);
 305
 306     if (pfilter) {
 307         if (!device_filter) {
 308             av_log(avctx, AV_LOG_ERROR, "Could not find %s device.\n",
 309                    devtypename);
 310             return AVERROR(EIO);
 311         }
 312         *pfilter = device_filter;
 313     }
 314
 315     return 0;
 316 }
 317
 318 /**
 319  * Cycle through available formats using the specified pin,
 320  * try to set parameters specified through AVOptions and if successful
 321  * return 1 in *pformat_set.
 322  * If pformat_set is NULL, list all pin capabilities.
 323  */
 324 static void
 325 dshow_cycle_formats(AVFormatContext *avctx, enum dshowDeviceType devtype,
 326                     IPin *pin, int *pformat_set)
 327 {
 328     struct dshow_ctx *ctx = avctx->priv_data;
 329     IAMStreamConfig *config = NULL;
 330     AM_MEDIA_TYPE *type = NULL;
 331     int format_set = 0;
 332     void *caps = NULL;
 333     int i, n, size;
 334
 335     if (IPin_QueryInterface(pin, &IID_IAMStreamConfig, (void **) &config) != S_OK)
 336         return;
 337     if (IAMStreamConfig_GetNumberOfCapabilities(config, &n, &size) != S_OK)
 338         goto end;
 339
 340     caps = av_malloc(size);
 341     if (!caps)
 342         goto end;
 343
 344     for (i = 0; i < n && !format_set; i++) {
 345         IAMStreamConfig_GetStreamCaps(config, i, &type, (void *) caps);
 346
 347 #if DSHOWDEBUG
 348         ff_print_AM_MEDIA_TYPE(type);
 349 #endif
 350
 351         if (devtype == VideoDevice) {
 352             VIDEO_STREAM_CONFIG_CAPS *vcaps = caps;
 353             BITMAPINFOHEADER *bih;
 354             int64_t *fr;
 355             const AVCodecTag *const tags[] = { avformat_get_riff_video_tags(), NULL };
 356 #if DSHOWDEBUG
 357             ff_print_VIDEO_STREAM_CONFIG_CAPS(vcaps);
 358 #endif
 359             if (IsEqualGUID(&type->formattype, &FORMAT_VideoInfo)) {
 360                 VIDEOINFOHEADER *v = (void *) type->pbFormat;
 361                 fr = &v->AvgTimePerFrame;
 362                 bih = &v->bmiHeader;
 363             } else if (IsEqualGUID(&type->formattype, &FORMAT_VideoInfo2)) {
 364                 VIDEOINFOHEADER2 *v = (void *) type->pbFormat;
 365                 fr = &v->AvgTimePerFrame;
 366                 bih = &v->bmiHeader;
 367             } else {
 368                 goto next;
 369             }
 370             if (!pformat_set) {
 371                 enum AVPixelFormat pix_fmt = dshow_pixfmt(bih->biCompression, bih->biBitCount);
 372                 if (pix_fmt == AV_PIX_FMT_NONE) {
 373                     enum AVCodecID codec_id = av_codec_get_id(tags, bih->biCompression);
 374                     AVCodec *codec = avcodec_find_decoder(codec_id);
 375                     if (codec_id == AV_CODEC_ID_NONE || !codec) {
 376                         av_log(avctx, AV_LOG_INFO, "  unknown compression type 0x%X", (int) bih->biCompression);
 377                     } else {
 378                         av_log(avctx, AV_LOG_INFO, "  vcodec=%s", codec->name);
 379                     }
 380                 } else {
 381                     av_log(avctx, AV_LOG_INFO, "  pixel_format=%s", av_get_pix_fmt_name(pix_fmt));
 382                 }
 383                 av_log(avctx, AV_LOG_INFO, "  min s=%ldx%ld fps=%g max s=%ldx%ld fps=%g\n",
 384                        vcaps->MinOutputSize.cx, vcaps->MinOutputSize.cy,
 385                        1e7 / vcaps->MaxFrameInterval,
 386                        vcaps->MaxOutputSize.cx, vcaps->MaxOutputSize.cy,
 387                        1e7 / vcaps->MinFrameInterval);
 388                 continue;
 389             }
 390             if (ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO) {
 391                 if (ctx->video_codec_id != av_codec_get_id(tags, bih->biCompression))
 392                     goto next;
 393             }
 394             if (ctx->pixel_format != AV_PIX_FMT_NONE &&
 395                 ctx->pixel_format != dshow_pixfmt(bih->biCompression, bih->biBitCount)) {
 396                 goto next;
 397             }
 398             if (ctx->framerate) {
 399                 int64_t framerate = ((int64_t) ctx->requested_framerate.den*10000000)
 400                                             /  ctx->requested_framerate.num;
 401                 if (framerate > vcaps->MaxFrameInterval ||
 402                     framerate < vcaps->MinFrameInterval)
 403                     goto next;
 404                 *fr = framerate;
 405             }
 406             if (ctx->requested_width && ctx->requested_height) {
 407                 if (ctx->requested_width  > vcaps->MaxOutputSize.cx ||
 408                     ctx->requested_width  < vcaps->MinOutputSize.cx ||
 409                     ctx->requested_height > vcaps->MaxOutputSize.cy ||
 410                     ctx->requested_height < vcaps->MinOutputSize.cy)
 411                     goto next;
 412                 bih->biWidth  = ctx->requested_width;
 413                 bih->biHeight = ctx->requested_height;
 414             }
 415         } else {
 416             AUDIO_STREAM_CONFIG_CAPS *acaps = caps;
 417             WAVEFORMATEX *fx;
 418 #if DSHOWDEBUG
 419             ff_print_AUDIO_STREAM_CONFIG_CAPS(acaps);
 420 #endif
 421             if (IsEqualGUID(&type->formattype, &FORMAT_WaveFormatEx)) {
 422                 fx = (void *) type->pbFormat;
 423             } else {
 424                 goto next;
 425             }
 426             if (!pformat_set) {
 427                 av_log(avctx, AV_LOG_INFO, "  min ch=%lu bits=%lu rate=%6lu max ch=%lu bits=%lu rate=%6lu\n",
 428                        acaps->MinimumChannels, acaps->MinimumBitsPerSample, acaps->MinimumSampleFrequency,
 429                        acaps->MaximumChannels, acaps->MaximumBitsPerSample, acaps->MaximumSampleFrequency);
 430                 continue;
 431             }
 432             if (ctx->sample_rate) {
 433                 if (ctx->sample_rate > acaps->MaximumSampleFrequency ||
 434                     ctx->sample_rate < acaps->MinimumSampleFrequency)
 435                     goto next;
 436                 fx->nSamplesPerSec = ctx->sample_rate;
 437             }
 438             if (ctx->sample_size) {
 439                 if (ctx->sample_size > acaps->MaximumBitsPerSample ||
 440                     ctx->sample_size < acaps->MinimumBitsPerSample)
 441                     goto next;
 442                 fx->wBitsPerSample = ctx->sample_size;
 443             }
 444             if (ctx->channels) {
 445                 if (ctx->channels > acaps->MaximumChannels ||
 446                     ctx->channels < acaps->MinimumChannels)
 447                     goto next;
 448                 fx->nChannels = ctx->channels;
 449             }
 450         }
 451         if (IAMStreamConfig_SetFormat(config, type) != S_OK)
 452             goto next;
 453         format_set = 1;
 454 next:
 455         if (type->pbFormat)
 456             CoTaskMemFree(type->pbFormat);
 457         CoTaskMemFree(type);
 458     }
 459 end:
 460     IAMStreamConfig_Release(config);
 461     if (caps)
 462         av_free(caps);
 463     if (pformat_set)
 464         *pformat_set = format_set;
 465 }
 466
 467 /**
 468  * Set audio device buffer size in milliseconds (which can directly impact
 469  * latency, depending on the device).
 470  */
 471 static int
 472 dshow_set_audio_buffer_size(AVFormatContext *avctx, IPin *pin)
 473 {
 474     struct dshow_ctx *ctx = avctx->priv_data;
 475     IAMBufferNegotiation *buffer_negotiation = NULL;
 476     ALLOCATOR_PROPERTIES props = { -1, -1, -1, -1 };
 477     IAMStreamConfig *config = NULL;
 478     AM_MEDIA_TYPE *type = NULL;
 479     int ret = AVERROR(EIO);
 480
 481     if (IPin_QueryInterface(pin, &IID_IAMStreamConfig, (void **) &config) != S_OK)
 482         goto end;
 483     if (IAMStreamConfig_GetFormat(config, &type) != S_OK)
 484         goto end;
 485     if (!IsEqualGUID(&type->formattype, &FORMAT_WaveFormatEx))
 486         goto end;
 487
 488     props.cbBuffer = (((WAVEFORMATEX *) type->pbFormat)->nAvgBytesPerSec)
 489                    * ctx->audio_buffer_size / 1000;
 490
 491     if (IPin_QueryInterface(pin, &IID_IAMBufferNegotiation, (void **) &buffer_negotiation) != S_OK)
 492         goto end;
 493     if (IAMBufferNegotiation_SuggestAllocatorProperties(buffer_negotiation, &props) != S_OK)
 494         goto end;
 495
 496     ret = 0;
 497
 498 end:
 499     if (buffer_negotiation)
 500         IAMBufferNegotiation_Release(buffer_negotiation);
 501     if (type) {
 502         if (type->pbFormat)
 503             CoTaskMemFree(type->pbFormat);
 504         CoTaskMemFree(type);
 505     }
 506     if (config)
 507         IAMStreamConfig_Release(config);
 508
 509     return ret;
 510 }
 511
 512 /**
 513  * Cycle through available pins using the device_filter device, of type
 514  * devtype, retrieve the first output pin and return the pointer to the
 515  * object found in *ppin.
 516  * If ppin is NULL, cycle through all pins listing audio/video capabilities.
 517  */
 518 static int
 519 dshow_cycle_pins(AVFormatContext *avctx, enum dshowDeviceType devtype,
 520                  IBaseFilter *device_filter, IPin **ppin)
 521 {
 522     struct dshow_ctx *ctx = avctx->priv_data;
 523     IEnumPins *pins = 0;
 524     IPin *device_pin = NULL;
 525     IPin *pin;
 526     int r;
 527
 528     const GUID *mediatype[2] = { &MEDIATYPE_Video, &MEDIATYPE_Audio };
 529     const char *devtypename = (devtype == VideoDevice) ? "video" : "audio";
 530
 531     int set_format = (devtype == VideoDevice && (ctx->framerate ||
 532                                                 (ctx->requested_width && ctx->requested_height) ||
 533                                                  ctx->pixel_format != AV_PIX_FMT_NONE ||
 534                                                  ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO))
 535                   || (devtype == AudioDevice && (ctx->channels || ctx->sample_rate));
 536     int format_set = 0;
 537
 538     r = IBaseFilter_EnumPins(device_filter, &pins);
 539     if (r != S_OK) {
 540         av_log(avctx, AV_LOG_ERROR, "Could not enumerate pins.\n");
 541         return AVERROR(EIO);
 542     }
 543
 544     if (!ppin) {
 545         av_log(avctx, AV_LOG_INFO, "DirectShow %s device options\n",
 546                devtypename);
 547     }
 548     while (!device_pin && IEnumPins_Next(pins, 1, &pin, NULL) == S_OK) {
 549         IKsPropertySet *p = NULL;
 550         IEnumMediaTypes *types = NULL;
 551         PIN_INFO info = {0};
 552         AM_MEDIA_TYPE *type;
 553         GUID category;
 554         DWORD r2;
 555
 556         IPin_QueryPinInfo(pin, &info);
 557         IBaseFilter_Release(info.pFilter);
 558
 559         if (info.dir != PINDIR_OUTPUT)
 560             goto next;
 561         if (IPin_QueryInterface(pin, &IID_IKsPropertySet, (void **) &p) != S_OK)
 562             goto next;
 563         if (IKsPropertySet_Get(p, &AMPROPSETID_Pin, AMPROPERTY_PIN_CATEGORY,
 564                                NULL, 0, &category, sizeof(GUID), &r2) != S_OK)
 565             goto next;
 566         if (!IsEqualGUID(&category, &PIN_CATEGORY_CAPTURE))
 567             goto next;
 568
 569         if (!ppin) {
 570             char *buf = dup_wchar_to_utf8(info.achName);
 571             av_log(avctx, AV_LOG_INFO, " Pin \"%s\"\n", buf);
 572             av_free(buf);
 573             dshow_cycle_formats(avctx, devtype, pin, NULL);
 574             goto next;
 575         }
 576         if (set_format) {
 577             dshow_cycle_formats(avctx, devtype, pin, &format_set);
 578             if (!format_set) {
 579                 goto next;
 580             }
 581         }
 582         if (devtype == AudioDevice && ctx->audio_buffer_size) {
 583             if (dshow_set_audio_buffer_size(avctx, pin) < 0) {
 584                 av_log(avctx, AV_LOG_ERROR, "unable to set audio buffer size %d to pin, using pin anyway...", ctx->audio_buffer_size);
 585             }
 586         }
 587
 588         if (IPin_EnumMediaTypes(pin, &types) != S_OK)
 589             goto next;
 590
 591         IEnumMediaTypes_Reset(types);
 592         while (!device_pin && IEnumMediaTypes_Next(types, 1, &type, NULL) == S_OK) {
 593             if (IsEqualGUID(&type->majortype, mediatype[devtype])) {
 594                 device_pin = pin;
 595                 goto next;
 596             }
 597             CoTaskMemFree(type);
 598         }
 599
 600 next:
 601         if (types)
 602             IEnumMediaTypes_Release(types);
 603         if (p)
 604             IKsPropertySet_Release(p);
 605         if (device_pin != pin)
 606             IPin_Release(pin);
 607     }
 608
 609     IEnumPins_Release(pins);
 610
 611     if (ppin) {
 612         if (set_format && !format_set) {
 613             av_log(avctx, AV_LOG_ERROR, "Could not set %s options\n", devtypename);
 614             return AVERROR(EIO);
 615         }
 616         if (!device_pin) {
 617             av_log(avctx, AV_LOG_ERROR,
 618                 "Could not find output pin from %s capture device.\n", devtypename);
 619             return AVERROR(EIO);
 620         }
 621         *ppin = device_pin;
 622     }
 623
 624     return 0;
 625 }
 626
 627 /**
 628  * List options for device with type devtype.
 629  *
 630  * @param devenum device enumerator used for accessing the device
 631  */
 632 static int
 633 dshow_list_device_options(AVFormatContext *avctx, ICreateDevEnum *devenum,
 634                           enum dshowDeviceType devtype)
 635 {
 636     struct dshow_ctx *ctx = avctx->priv_data;
 637     IBaseFilter *device_filter = NULL;
 638     int r;
 639
 640     if ((r = dshow_cycle_devices(avctx, devenum, devtype, &device_filter)) < 0)
 641         return r;
 642     ctx->device_filter[devtype] = device_filter;
 643     if ((r = dshow_cycle_pins(avctx, devtype, device_filter, NULL)) < 0)
 644         return r;
 645
 646     return 0;
 647 }
 648
 649 static int
 650 dshow_open_device(AVFormatContext *avctx, ICreateDevEnum *devenum,
 651                   enum dshowDeviceType devtype)
 652 {
 653     struct dshow_ctx *ctx = avctx->priv_data;
 654     IBaseFilter *device_filter = NULL;
 655     IGraphBuilder *graph = ctx->graph;
 656     IPin *device_pin = NULL;
 657     libAVPin *capture_pin = NULL;
 658     libAVFilter *capture_filter = NULL;
 659     int ret = AVERROR(EIO);
 660     int r;
 661
 662     const wchar_t *filter_name[2] = { L"Audio capture filter", L"Video capture filter" };
 663
 664     if ((r = dshow_cycle_devices(avctx, devenum, devtype, &device_filter)) < 0) {
 665         ret = r;
 666         goto error;
 667     }
 668
 669     ctx->device_filter [devtype] = device_filter;
 670
 671     r = IGraphBuilder_AddFilter(graph, device_filter, NULL);
 672     if (r != S_OK) {
 673         av_log(avctx, AV_LOG_ERROR, "Could not add device filter to graph.\n");
 674         goto error;
 675     }
 676
 677     if ((r = dshow_cycle_pins(avctx, devtype, device_filter, &device_pin)) < 0) {
 678         ret = r;
 679         goto error;
 680     }
 681     ctx->device_pin[devtype] = device_pin;
 682
 683     capture_filter = libAVFilter_Create(avctx, callback, devtype);
 684     if (!capture_filter) {
 685         av_log(avctx, AV_LOG_ERROR, "Could not create grabber filter.\n");
 686         goto error;
 687     }
 688     ctx->capture_filter[devtype] = capture_filter;
 689
 690     r = IGraphBuilder_AddFilter(graph, (IBaseFilter *) capture_filter,
 691                                 filter_name[devtype]);
 692     if (r != S_OK) {
 693         av_log(avctx, AV_LOG_ERROR, "Could not add capture filter to graph\n");
 694         goto error;
 695     }
 696
 697     libAVPin_AddRef(capture_filter->pin);
 698     capture_pin = capture_filter->pin;
 699     ctx->capture_pin[devtype] = capture_pin;
 700
 701     r = IGraphBuilder_ConnectDirect(graph, device_pin, (IPin *) capture_pin, NULL);
 702     if (r != S_OK) {
 703         av_log(avctx, AV_LOG_ERROR, "Could not connect pins\n");
 704         goto error;
 705     }
 706
 707     ret = 0;
 708
 709 error:
 710     return ret;
 711 }
 712
 713 static enum AVCodecID waveform_codec_id(enum AVSampleFormat sample_fmt)
 714 {
 715     switch (sample_fmt) {
 716     case AV_SAMPLE_FMT_U8:  return AV_CODEC_ID_PCM_U8;
 717     case AV_SAMPLE_FMT_S16: return AV_CODEC_ID_PCM_S16LE;
 718     case AV_SAMPLE_FMT_S32: return AV_CODEC_ID_PCM_S32LE;
 719     default:                return AV_CODEC_ID_NONE; /* Should never happen. */
 720     }
 721 }
 722
 723 static enum AVSampleFormat sample_fmt_bits_per_sample(int bits)
 724 {
 725     switch (bits) {
 726     case 8:  return AV_SAMPLE_FMT_U8;
 727     case 16: return AV_SAMPLE_FMT_S16;
 728     case 32: return AV_SAMPLE_FMT_S32;
 729     default: return AV_SAMPLE_FMT_NONE; /* Should never happen. */
 730     }
 731 }
 732
 733 static int
 734 dshow_add_device(AVFormatContext *avctx,
 735                  enum dshowDeviceType devtype)
 736 {
 737     struct dshow_ctx *ctx = avctx->priv_data;
 738     AM_MEDIA_TYPE type;
 739     AVCodecContext *codec;
 740     AVStream *st;
 741     int ret = AVERROR(EIO);
 742
 743     st = avformat_new_stream(avctx, NULL);
 744     if (!st) {
 745         ret = AVERROR(ENOMEM);
 746         goto error;
 747     }
 748     st->id = devtype;
 749
 750     ctx->capture_filter[devtype]->stream_index = st->index;
 751
 752     libAVPin_ConnectionMediaType(ctx->capture_pin[devtype], &type);
 753
 754     codec = st->codec;
 755     if (devtype == VideoDevice) {
 756         BITMAPINFOHEADER *bih = NULL;
 757         AVRational time_base;
 758
 759         if (IsEqualGUID(&type.formattype, &FORMAT_VideoInfo)) {
 760             VIDEOINFOHEADER *v = (void *) type.pbFormat;
 761             time_base = (AVRational) { v->AvgTimePerFrame, 10000000 };
 762             bih = &v->bmiHeader;
 763         } else if (IsEqualGUID(&type.formattype, &FORMAT_VideoInfo2)) {
 764             VIDEOINFOHEADER2 *v = (void *) type.pbFormat;
 765             time_base = (AVRational) { v->AvgTimePerFrame, 10000000 };
 766             bih = &v->bmiHeader;
 767         }
 768         if (!bih) {
 769             av_log(avctx, AV_LOG_ERROR, "Could not get media type.\n");
 770             goto error;
 771         }
 772
 773         codec->time_base  = time_base;
 774         codec->codec_type = AVMEDIA_TYPE_VIDEO;
 775         codec->width      = bih->biWidth;
 776         codec->height     = bih->biHeight;
 777         codec->codec_tag  = bih->biCompression;
 778         codec->pix_fmt    = dshow_pixfmt(bih->biCompression, bih->biBitCount);
 779         if (bih->biCompression == MKTAG('H', 'D', 'Y', 'C')) {
 780             av_log(avctx, AV_LOG_DEBUG, "attempt to use full range for HDYC...\n");
 781             codec->color_range = AVCOL_RANGE_MPEG; // just in case it needs this...
 782         }
 783         if (codec->pix_fmt == AV_PIX_FMT_NONE) {
 784             const AVCodecTag *const tags[] = { avformat_get_riff_video_tags(), NULL };
 785             codec->codec_id = av_codec_get_id(tags, bih->biCompression);
 786             if (codec->codec_id == AV_CODEC_ID_NONE) {
 787                 av_log(avctx, AV_LOG_ERROR, "Unknown compression type. "
 788                                  "Please report type 0x%X.\n", (int) bih->biCompression);
 789                 return AVERROR_PATCHWELCOME;
 790             }
 791             codec->bits_per_coded_sample = bih->biBitCount;
 792         } else {
 793             codec->codec_id = AV_CODEC_ID_RAWVIDEO;
 794             if (bih->biCompression == BI_RGB || bih->biCompression == BI_BITFIELDS) {
 795                 codec->bits_per_coded_sample = bih->biBitCount;
 796                 codec->extradata = av_malloc(9 + FF_INPUT_BUFFER_PADDING_SIZE);
 797                 if (codec->extradata) {
 798                     codec->extradata_size = 9;
 799                     memcpy(codec->extradata, "BottomUp", 9);
 800                 }
 801             }
 802         }
 803     } else {
 804         WAVEFORMATEX *fx = NULL;
 805
 806         if (IsEqualGUID(&type.formattype, &FORMAT_WaveFormatEx)) {
 807             fx = (void *) type.pbFormat;
 808         }
 809         if (!fx) {
 810             av_log(avctx, AV_LOG_ERROR, "Could not get media type.\n");
 811             goto error;
 812         }
 813
 814         codec->codec_type  = AVMEDIA_TYPE_AUDIO;
 815         codec->sample_fmt  = sample_fmt_bits_per_sample(fx->wBitsPerSample);
 816         codec->codec_id    = waveform_codec_id(codec->sample_fmt);
 817         codec->sample_rate = fx->nSamplesPerSec;
 818         codec->channels    = fx->nChannels;
 819     }
 820
 821     avpriv_set_pts_info(st, 64, 1, 10000000);
 822
 823     ret = 0;
 824
 825 error:
 826     return ret;
 827 }
 828
 829 static int parse_device_name(AVFormatContext *avctx)
 830 {
 831     struct dshow_ctx *ctx = avctx->priv_data;
 832     char **device_name = ctx->device_name;
 833     char *name = av_strdup(avctx->filename);
 834     char *tmp = name;
 835     int ret = 1;
 836     char *type;
 837
 838     while ((type = strtok(tmp, "="))) {
 839         char *token = strtok(NULL, ":");
 840         tmp = NULL;
 841
 842         if        (!strcmp(type, "video")) {
 843             device_name[0] = token;
 844         } else if (!strcmp(type, "audio")) {
 845             device_name[1] = token;
 846         } else {
 847             device_name[0] = NULL;
 848             device_name[1] = NULL;
 849             break;
 850         }
 851     }
 852
 853     if (!device_name[0] && !device_name[1]) {
 854         ret = 0;
 855     } else {
 856         if (device_name[0])
 857             device_name[0] = av_strdup(device_name[0]);
 858         if (device_name[1])
 859             device_name[1] = av_strdup(device_name[1]);
 860     }
 861
 862     av_free(name);
 863     return ret;
 864 }
 865
 866 static int dshow_read_header(AVFormatContext *avctx)
 867 {
 868     struct dshow_ctx *ctx = avctx->priv_data;
 869     IGraphBuilder *graph = NULL;
 870     ICreateDevEnum *devenum = NULL;
 871     IMediaControl *control = NULL;
 872     IMediaEvent *media_event = NULL;
 873     HANDLE media_event_handle;
 874     HANDLE proc;
 875     int ret = AVERROR(EIO);
 876     int r;
 877
 878     CoInitialize(0);
 879
 880     if (!ctx->list_devices && !parse_device_name(avctx)) {
 881         av_log(avctx, AV_LOG_ERROR, "Malformed dshow input string.\n");
 882         goto error;
 883     }
 884
 885     ctx->video_codec_id = avctx->video_codec_id ? avctx->video_codec_id
 886                                                 : AV_CODEC_ID_RAWVIDEO;
 887     if (ctx->pixel_format != AV_PIX_FMT_NONE) {
 888         if (ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO) {
 889             av_log(avctx, AV_LOG_ERROR, "Pixel format may only be set when "
 890                               "video codec is not set or set to rawvideo\n");
 891             ret = AVERROR(EINVAL);
 892             goto error;
 893         }
 894     }
 895     if (ctx->framerate) {
 896         r = av_parse_video_rate(&ctx->requested_framerate, ctx->framerate);
 897         if (r < 0) {
 898             av_log(avctx, AV_LOG_ERROR, "Could not parse framerate '%s'.\n", ctx->framerate);
 899             goto error;
 900         }
 901     }
 902
 903     r = CoCreateInstance(&CLSID_FilterGraph, NULL, CLSCTX_INPROC_SERVER,
 904                          &IID_IGraphBuilder, (void **) &graph);
 905     if (r != S_OK) {
 906         av_log(avctx, AV_LOG_ERROR, "Could not create capture graph.\n");
 907         goto error;
 908     }
 909     ctx->graph = graph;
 910
 911     r = CoCreateInstance(&CLSID_SystemDeviceEnum, NULL, CLSCTX_INPROC_SERVER,
 912                          &IID_ICreateDevEnum, (void **) &devenum);
 913     if (r != S_OK) {
 914         av_log(avctx, AV_LOG_ERROR, "Could not enumerate system devices.\n");
 915         goto error;
 916     }
 917
 918     if (ctx->list_devices) {
 919         av_log(avctx, AV_LOG_INFO, "DirectShow video devices\n");
 920         dshow_cycle_devices(avctx, devenum, VideoDevice, NULL);
 921         av_log(avctx, AV_LOG_INFO, "DirectShow audio devices\n");
 922         dshow_cycle_devices(avctx, devenum, AudioDevice, NULL);
 923         ret = AVERROR_EXIT;
 924         goto error;
 925     }
 926     if (ctx->list_options) {
 927         if (ctx->device_name[VideoDevice])
 928             dshow_list_device_options(avctx, devenum, VideoDevice);
 929         if (ctx->device_name[AudioDevice])
 930             dshow_list_device_options(avctx, devenum, AudioDevice);
 931         ret = AVERROR_EXIT;
 932         goto error;
 933     }
 934
 935     if (ctx->device_name[VideoDevice]) {
 936         if ((r = dshow_open_device(avctx, devenum, VideoDevice)) < 0 ||
 937             (r = dshow_add_device(avctx, VideoDevice)) < 0) {
 938             ret = r;
 939             goto error;
 940         }
 941     }
 942     if (ctx->device_name[AudioDevice]) {
 943         if ((r = dshow_open_device(avctx, devenum, AudioDevice)) < 0 ||
 944             (r = dshow_add_device(avctx, AudioDevice)) < 0) {
 945             ret = r;
 946             goto error;
 947         }
 948     }
 949     ctx->curbufsize[0] = 0;
 950     ctx->curbufsize[1] = 0;
 951     ctx->mutex = CreateMutex(NULL, 0, NULL);
 952     if (!ctx->mutex) {
 953         av_log(avctx, AV_LOG_ERROR, "Could not create Mutex\n");
 954         goto error;
 955     }
 956     ctx->event[1] = CreateEvent(NULL, 1, 0, NULL);
 957     if (!ctx->event[1]) {
 958         av_log(avctx, AV_LOG_ERROR, "Could not create Event\n");
 959         goto error;
 960     }
 961
 962     r = IGraphBuilder_QueryInterface(graph, &IID_IMediaControl, (void **) &control);
 963     if (r != S_OK) {
 964         av_log(avctx, AV_LOG_ERROR, "Could not get media control.\n");
 965         goto error;
 966     }
 967     ctx->control = control;
 968
 969     r = IGraphBuilder_QueryInterface(graph, &IID_IMediaEvent, (void **) &media_event);
 970     if (r != S_OK) {
 971         av_log(avctx, AV_LOG_ERROR, "Could not get media event.\n");
 972         goto error;
 973     }
 974     ctx->media_event = media_event;
 975
 976     r = IMediaEvent_GetEventHandle(media_event, (void *) &media_event_handle);
 977     if (r != S_OK) {
 978         av_log(avctx, AV_LOG_ERROR, "Could not get media event handle.\n");
 979         goto error;
 980     }
 981     proc = GetCurrentProcess();
 982     r = DuplicateHandle(proc, media_event_handle, proc, &ctx->event[0],
 983                         0, 0, DUPLICATE_SAME_ACCESS);
 984     if (!r) {
 985         av_log(avctx, AV_LOG_ERROR, "Could not duplicate media event handle.\n");
 986         goto error;
 987     }
 988
 989     r = IMediaControl_Run(control);
 990     if (r == S_FALSE) {
 991         OAFilterState pfs;
 992         r = IMediaControl_GetState(control, 0, &pfs);
 993     }
 994     if (r != S_OK) {
 995         av_log(avctx, AV_LOG_ERROR, "Could not run filter\n");
 996         goto error;
 997     }
 998
 999     ret = 0;
1000
1001 error:
1002
1003     if (devenum)
1004         ICreateDevEnum_Release(devenum);
1005
1006     if (ret < 0)
1007         dshow_read_close(avctx);
1008
1009     return ret;
1010 }
1011
1012 /**
1013  * Checks media events from DirectShow and returns -1 on error or EOF. Also
1014  * purges all events that might be in the event queue to stop the trigger
1015  * of event notification.
1016  */
1017 static int dshow_check_event_queue(IMediaEvent *media_event)
1018 {
1019     LONG_PTR p1, p2;
1020     long code;
1021     int ret = 0;
1022
1023     while (IMediaEvent_GetEvent(media_event, &code, &p1, &p2, 0) != E_ABORT) {
1024         if (code == EC_COMPLETE || code == EC_DEVICE_LOST || code == EC_ERRORABORT)
1025             ret = -1;
1026         IMediaEvent_FreeEventParams(media_event, code, p1, p2);
1027     }
1028
1029     return ret;
1030 }
1031
1032 static int dshow_read_packet(AVFormatContext *s, AVPacket *pkt)
1033 {
1034     struct dshow_ctx *ctx = s->priv_data;
1035     AVPacketList *pktl = NULL;
1036
1037     while (!ctx->eof && !pktl) {
1038         WaitForSingleObject(ctx->mutex, INFINITE);
1039         pktl = ctx->pktl;
1040         if (pktl) {
1041             *pkt = pktl->pkt;
1042             ctx->pktl = ctx->pktl->next;
1043             av_free(pktl);
1044             ctx->curbufsize[pkt->stream_index] -= pkt->size;
1045         }
1046         ResetEvent(ctx->event[1]);
1047         ReleaseMutex(ctx->mutex);
1048         if (!pktl) {
1049             if (dshow_check_event_queue(ctx->media_event) < 0) {
1050                 ctx->eof = 1;
1051             } else if (s->flags & AVFMT_FLAG_NONBLOCK) {
1052                 return AVERROR(EAGAIN);
1053             } else {
1054                 WaitForMultipleObjects(2, ctx->event, 0, INFINITE);
1055             }
1056         }
1057     }
1058
1059     return ctx->eof ? AVERROR(EIO) : pkt->size;
1060 }
1061
1062 #define OFFSET(x) offsetof(struct dshow_ctx, x)
1063 #define DEC AV_OPT_FLAG_DECODING_PARAM
1064 static const AVOption options[] = {
1065     { "video_size", "set video size given a string such as 640x480 or hd720.", OFFSET(requested_width), AV_OPT_TYPE_IMAGE_SIZE, {.str = NULL}, 0, 0, DEC },
1066     { "pixel_format", "set video pixel format", OFFSET(pixel_format), AV_OPT_TYPE_PIXEL_FMT, {.i64 = AV_PIX_FMT_NONE}, -1, INT_MAX, DEC },
1067     { "framerate", "set video frame rate", OFFSET(framerate), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, DEC },
1068     { "sample_rate", "set audio sample rate", OFFSET(sample_rate), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
1069     { "sample_size", "set audio sample size", OFFSET(sample_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 16, DEC },
1070     { "channels", "set number of audio channels, such as 1 or 2", OFFSET(channels), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
1071     { "list_devices", "list available devices", OFFSET(list_devices), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, DEC, "list_devices" },
1072     { "true", "", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, DEC, "list_devices" },
1073     { "false", "", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, DEC, "list_devices" },
1074     { "list_options", "list available options for specified device", OFFSET(list_options), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, DEC, "list_options" },
1075     { "true", "", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, DEC, "list_options" },
1076     { "false", "", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, DEC, "list_options" },
1077     { "video_device_number", "set video device number for devices with same name (starts at 0)", OFFSET(video_device_number), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
1078     { "audio_device_number", "set audio device number for devices with same name (starts at 0)", OFFSET(audio_device_number), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
1079     { "audio_buffer_size", "set audio device buffer latency size in milliseconds (default is the device's default)", OFFSET(audio_buffer_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
1080     { NULL },
1081 };
1082
1083 static const AVClass dshow_class = {
1084     .class_name = "dshow indev",
1085     .item_name  = av_default_item_name,
1086     .option     = options,
1087     .version    = LIBAVUTIL_VERSION_INT,
1088     .category   = AV_CLASS_CATEGORY_DEVICE_VIDEO_INPUT,
1089 };
1090
1091 AVInputFormat ff_dshow_demuxer = {
1092     .name           = "dshow",
1093     .long_name      = NULL_IF_CONFIG_SMALL("DirectShow capture"),
1094     .priv_data_size = sizeof(struct dshow_ctx),
1095     .read_header    = dshow_read_header,
1096     .read_packet    = dshow_read_packet,
1097     .read_close     = dshow_read_close,
1098     .flags          = AVFMT_NOFILE,
1099     .priv_class     = &dshow_class,
1100 };