FFmpeg: libavdevice/dshow.c Source File

00001 /*
00002  * Directshow capture interface
00003  * Copyright (c) 2010 Ramiro Polla
00004  *
00005  * This file is part of FFmpeg.
00006  *
00007  * FFmpeg is free software; you can redistribute it and/or
00008  * modify it under the terms of the GNU Lesser General Public
00009  * License as published by the Free Software Foundation; either
00010  * version 2.1 of the License, or (at your option) any later version.
00011  *
00012  * FFmpeg is distributed in the hope that it will be useful,
00013  * but WITHOUT ANY WARRANTY; without even the implied warranty of
00014  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
00015  * Lesser General Public License for more details.
00016  *
00017  * You should have received a copy of the GNU Lesser General Public
00018  * License along with FFmpeg; if not, write to the Free Software
00019  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
00020  */
00021 
00022 #include "libavutil/parseutils.h"
00023 #include "libavutil/pixdesc.h"
00024 #include "libavutil/opt.h"
00025 #include "libavformat/internal.h"
00026 #include "avdevice.h"
00027 #include "dshow_capture.h"
00028 
00029 struct dshow_ctx {
00030     const AVClass *class;
00031 
00032     IGraphBuilder *graph;
00033 
00034     char *device_name[2];
00035     int video_device_number;
00036     int audio_device_number;
00037 
00038     int   list_options;
00039     int   list_devices;
00040     int   audio_buffer_size;
00041 
00042     IBaseFilter *device_filter[2];
00043     IPin        *device_pin[2];
00044     libAVFilter *capture_filter[2];
00045     libAVPin    *capture_pin[2];
00046 
00047     HANDLE mutex;
00048     HANDLE event;
00049     AVPacketList *pktl;
00050 
00051     int64_t curbufsize;
00052     unsigned int video_frame_num;
00053 
00054     IMediaControl *control;
00055 
00056     enum PixelFormat pixel_format;
00057     enum AVCodecID video_codec_id;
00058     char *framerate;
00059 
00060     int requested_width;
00061     int requested_height;
00062     AVRational requested_framerate;
00063 
00064     int sample_rate;
00065     int sample_size;
00066     int channels;
00067 };
00068 
00069 static enum PixelFormat dshow_pixfmt(DWORD biCompression, WORD biBitCount)
00070 {
00071     switch(biCompression) {
00072     case MKTAG('U', 'Y', 'V', 'Y'):
00073         return PIX_FMT_UYVY422;
00074     case MKTAG('Y', 'U', 'Y', '2'):
00075         return PIX_FMT_YUYV422;
00076     case MKTAG('I', '4', '2', '0'):
00077         return PIX_FMT_YUV420P;
00078     case BI_BITFIELDS:
00079     case BI_RGB:
00080         switch(biBitCount) { /* 1-8 are untested */
00081             case 1:
00082                 return PIX_FMT_MONOWHITE;
00083             case 4:
00084                 return PIX_FMT_RGB4;
00085             case 8:
00086                 return PIX_FMT_RGB8;
00087             case 16:
00088                 return PIX_FMT_RGB555;
00089             case 24:
00090                 return PIX_FMT_BGR24;
00091             case 32:
00092                 return PIX_FMT_RGB32;
00093         }
00094     }
00095     return PIX_FMT_NONE;
00096 }
00097 
00098 static enum AVCodecID dshow_codecid(DWORD biCompression)
00099 {
00100     switch(biCompression) {
00101     case MKTAG('d', 'v', 's', 'd'):
00102         return AV_CODEC_ID_DVVIDEO;
00103     case MKTAG('M', 'J', 'P', 'G'):
00104     case MKTAG('m', 'j', 'p', 'g'):
00105         return AV_CODEC_ID_MJPEG;
00106     }
00107     return AV_CODEC_ID_NONE;
00108 }
00109 
00110 static int
00111 dshow_read_close(AVFormatContext *s)
00112 {
00113     struct dshow_ctx *ctx = s->priv_data;
00114     AVPacketList *pktl;
00115 
00116     if (ctx->control) {
00117         IMediaControl_Stop(ctx->control);
00118         IMediaControl_Release(ctx->control);
00119     }
00120 
00121     if (ctx->graph) {
00122         IEnumFilters *fenum;
00123         int r;
00124         r = IGraphBuilder_EnumFilters(ctx->graph, &fenum);
00125         if (r == S_OK) {
00126             IBaseFilter *f;
00127             IEnumFilters_Reset(fenum);
00128             while (IEnumFilters_Next(fenum, 1, &f, NULL) == S_OK) {
00129                 if (IGraphBuilder_RemoveFilter(ctx->graph, f) == S_OK)
00130                     IEnumFilters_Reset(fenum); /* When a filter is removed,
00131                                                 * the list must be reset. */
00132                 IBaseFilter_Release(f);
00133             }
00134             IEnumFilters_Release(fenum);
00135         }
00136         IGraphBuilder_Release(ctx->graph);
00137     }
00138 
00139     if (ctx->capture_pin[VideoDevice])
00140         libAVPin_Release(ctx->capture_pin[VideoDevice]);
00141     if (ctx->capture_pin[AudioDevice])
00142         libAVPin_Release(ctx->capture_pin[AudioDevice]);
00143     if (ctx->capture_filter[VideoDevice])
00144         libAVFilter_Release(ctx->capture_filter[VideoDevice]);
00145     if (ctx->capture_filter[AudioDevice])
00146         libAVFilter_Release(ctx->capture_filter[AudioDevice]);
00147 
00148     if (ctx->device_pin[VideoDevice])
00149         IPin_Release(ctx->device_pin[VideoDevice]);
00150     if (ctx->device_pin[AudioDevice])
00151         IPin_Release(ctx->device_pin[AudioDevice]);
00152     if (ctx->device_filter[VideoDevice])
00153         IBaseFilter_Release(ctx->device_filter[VideoDevice]);
00154     if (ctx->device_filter[AudioDevice])
00155         IBaseFilter_Release(ctx->device_filter[AudioDevice]);
00156 
00157     if (ctx->device_name[0])
00158         av_free(ctx->device_name[0]);
00159     if (ctx->device_name[1])
00160         av_free(ctx->device_name[1]);
00161 
00162     if(ctx->mutex)
00163         CloseHandle(ctx->mutex);
00164     if(ctx->event)
00165         CloseHandle(ctx->event);
00166 
00167     pktl = ctx->pktl;
00168     while (pktl) {
00169         AVPacketList *next = pktl->next;
00170         av_destruct_packet(&pktl->pkt);
00171         av_free(pktl);
00172         pktl = next;
00173     }
00174 
00175     return 0;
00176 }
00177 
00178 static char *dup_wchar_to_utf8(wchar_t *w)
00179 {
00180     char *s = NULL;
00181     int l = WideCharToMultiByte(CP_UTF8, 0, w, -1, 0, 0, 0, 0);
00182     s = av_malloc(l);
00183     if (s)
00184         WideCharToMultiByte(CP_UTF8, 0, w, -1, s, l, 0, 0);
00185     return s;
00186 }
00187 
00188 static int shall_we_drop(AVFormatContext *s)
00189 {
00190     struct dshow_ctx *ctx = s->priv_data;
00191     const uint8_t dropscore[] = {62, 75, 87, 100};
00192     const int ndropscores = FF_ARRAY_ELEMS(dropscore);
00193     unsigned int buffer_fullness = (ctx->curbufsize*100)/s->max_picture_buffer;
00194 
00195     if(dropscore[++ctx->video_frame_num%ndropscores] <= buffer_fullness) {
00196         av_log(s, AV_LOG_ERROR,
00197               "real-time buffer %d%% full! frame dropped!\n", buffer_fullness);
00198         return 1;
00199     }
00200 
00201     return 0;
00202 }
00203 
00204 static void
00205 callback(void *priv_data, int index, uint8_t *buf, int buf_size, int64_t time)
00206 {
00207     AVFormatContext *s = priv_data;
00208     struct dshow_ctx *ctx = s->priv_data;
00209     AVPacketList **ppktl, *pktl_next;
00210 
00211 //    dump_videohdr(s, vdhdr);
00212 
00213     WaitForSingleObject(ctx->mutex, INFINITE);
00214 
00215     if(shall_we_drop(s))
00216         goto fail;
00217 
00218     pktl_next = av_mallocz(sizeof(AVPacketList));
00219     if(!pktl_next)
00220         goto fail;
00221 
00222     if(av_new_packet(&pktl_next->pkt, buf_size) < 0) {
00223         av_free(pktl_next);
00224         goto fail;
00225     }
00226 
00227     pktl_next->pkt.stream_index = index;
00228     pktl_next->pkt.pts = time;
00229     memcpy(pktl_next->pkt.data, buf, buf_size);
00230 
00231     for(ppktl = &ctx->pktl ; *ppktl ; ppktl = &(*ppktl)->next);
00232     *ppktl = pktl_next;
00233 
00234     ctx->curbufsize += buf_size;
00235 
00236     SetEvent(ctx->event);
00237     ReleaseMutex(ctx->mutex);
00238 
00239     return;
00240 fail:
00241     ReleaseMutex(ctx->mutex);
00242     return;
00243 }
00244 
00251 static int
00252 dshow_cycle_devices(AVFormatContext *avctx, ICreateDevEnum *devenum,
00253                     enum dshowDeviceType devtype, IBaseFilter **pfilter)
00254 {
00255     struct dshow_ctx *ctx = avctx->priv_data;
00256     IBaseFilter *device_filter = NULL;
00257     IEnumMoniker *classenum = NULL;
00258     IMoniker *m = NULL;
00259     const char *device_name = ctx->device_name[devtype];
00260     int skip = (devtype == VideoDevice) ? ctx->video_device_number
00261                                         : ctx->audio_device_number;
00262     int r;
00263 
00264     const GUID *device_guid[2] = { &CLSID_VideoInputDeviceCategory,
00265                                    &CLSID_AudioInputDeviceCategory };
00266     const char *devtypename = (devtype == VideoDevice) ? "video" : "audio";
00267 
00268     r = ICreateDevEnum_CreateClassEnumerator(devenum, device_guid[devtype],
00269                                              (IEnumMoniker **) &classenum, 0);
00270     if (r != S_OK) {
00271         av_log(avctx, AV_LOG_ERROR, "Could not enumerate %s devices.\n",
00272                devtypename);
00273         return AVERROR(EIO);
00274     }
00275 
00276     while (!device_filter && IEnumMoniker_Next(classenum, 1, &m, NULL) == S_OK) {
00277         IPropertyBag *bag = NULL;
00278         char *buf = NULL;
00279         VARIANT var;
00280 
00281         r = IMoniker_BindToStorage(m, 0, 0, &IID_IPropertyBag, (void *) &bag);
00282         if (r != S_OK)
00283             goto fail1;
00284 
00285         var.vt = VT_BSTR;
00286         r = IPropertyBag_Read(bag, L"FriendlyName", &var, NULL);
00287         if (r != S_OK)
00288             goto fail1;
00289 
00290         buf = dup_wchar_to_utf8(var.bstrVal);
00291 
00292         if (pfilter) {
00293             if (strcmp(device_name, buf))
00294                 goto fail1;
00295 
00296             if (!skip--)
00297                 IMoniker_BindToObject(m, 0, 0, &IID_IBaseFilter, (void *) &device_filter);
00298         } else {
00299             av_log(avctx, AV_LOG_INFO, " \"%s\"\n", buf);
00300         }
00301 
00302 fail1:
00303         if (buf)
00304             av_free(buf);
00305         if (bag)
00306             IPropertyBag_Release(bag);
00307         IMoniker_Release(m);
00308     }
00309 
00310     IEnumMoniker_Release(classenum);
00311 
00312     if (pfilter) {
00313         if (!device_filter) {
00314             av_log(avctx, AV_LOG_ERROR, "Could not find %s device.\n",
00315                    devtypename);
00316             return AVERROR(EIO);
00317         }
00318         *pfilter = device_filter;
00319     }
00320 
00321     return 0;
00322 }
00323 
00330 static void
00331 dshow_cycle_formats(AVFormatContext *avctx, enum dshowDeviceType devtype,
00332                     IPin *pin, int *pformat_set)
00333 {
00334     struct dshow_ctx *ctx = avctx->priv_data;
00335     IAMStreamConfig *config = NULL;
00336     AM_MEDIA_TYPE *type = NULL;
00337     int format_set = 0;
00338     void *caps = NULL;
00339     int i, n, size;
00340 
00341     if (IPin_QueryInterface(pin, &IID_IAMStreamConfig, (void **) &config) != S_OK)
00342         return;
00343     if (IAMStreamConfig_GetNumberOfCapabilities(config, &n, &size) != S_OK)
00344         goto end;
00345 
00346     caps = av_malloc(size);
00347     if (!caps)
00348         goto end;
00349 
00350     for (i = 0; i < n && !format_set; i++) {
00351         IAMStreamConfig_GetStreamCaps(config, i, &type, (void *) caps);
00352 
00353 #if DSHOWDEBUG
00354         ff_print_AM_MEDIA_TYPE(type);
00355 #endif
00356 
00357         if (devtype == VideoDevice) {
00358             VIDEO_STREAM_CONFIG_CAPS *vcaps = caps;
00359             BITMAPINFOHEADER *bih;
00360             int64_t *fr;
00361 #if DSHOWDEBUG
00362             ff_print_VIDEO_STREAM_CONFIG_CAPS(vcaps);
00363 #endif
00364             if (IsEqualGUID(&type->formattype, &FORMAT_VideoInfo)) {
00365                 VIDEOINFOHEADER *v = (void *) type->pbFormat;
00366                 fr = &v->AvgTimePerFrame;
00367                 bih = &v->bmiHeader;
00368             } else if (IsEqualGUID(&type->formattype, &FORMAT_VideoInfo2)) {
00369                 VIDEOINFOHEADER2 *v = (void *) type->pbFormat;
00370                 fr = &v->AvgTimePerFrame;
00371                 bih = &v->bmiHeader;
00372             } else {
00373                 goto next;
00374             }
00375             if (!pformat_set) {
00376                 enum PixelFormat pix_fmt = dshow_pixfmt(bih->biCompression, bih->biBitCount);
00377                 if (pix_fmt == PIX_FMT_NONE) {
00378                     enum AVCodecID codec_id = dshow_codecid(bih->biCompression);
00379                     AVCodec *codec = avcodec_find_decoder(codec_id);
00380                     if (codec_id == AV_CODEC_ID_NONE || !codec) {
00381                         av_log(avctx, AV_LOG_INFO, "  unknown compression type");
00382                     } else {
00383                         av_log(avctx, AV_LOG_INFO, "  vcodec=%s", codec->name);
00384                     }
00385                 } else {
00386                     av_log(avctx, AV_LOG_INFO, "  pixel_format=%s", av_get_pix_fmt_name(pix_fmt));
00387                 }
00388                 av_log(avctx, AV_LOG_INFO, "  min s=%ldx%ld fps=%g max s=%ldx%ld fps=%g\n",
00389                        vcaps->MinOutputSize.cx, vcaps->MinOutputSize.cy,
00390                        1e7 / vcaps->MaxFrameInterval,
00391                        vcaps->MaxOutputSize.cx, vcaps->MaxOutputSize.cy,
00392                        1e7 / vcaps->MinFrameInterval);
00393                 continue;
00394             }
00395             if (ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO) {
00396                 if (ctx->video_codec_id != dshow_codecid(bih->biCompression))
00397                     goto next;
00398             }
00399             if (ctx->pixel_format != PIX_FMT_NONE &&
00400                 ctx->pixel_format != dshow_pixfmt(bih->biCompression, bih->biBitCount)) {
00401                 goto next;
00402             }
00403             if (ctx->framerate) {
00404                 int64_t framerate = ((int64_t) ctx->requested_framerate.den*10000000)
00405                                             /  ctx->requested_framerate.num;
00406                 if (framerate > vcaps->MaxFrameInterval ||
00407                     framerate < vcaps->MinFrameInterval)
00408                     goto next;
00409                 *fr = framerate;
00410             }
00411             if (ctx->requested_width && ctx->requested_height) {
00412                 if (ctx->requested_width  > vcaps->MaxOutputSize.cx ||
00413                     ctx->requested_width  < vcaps->MinOutputSize.cx ||
00414                     ctx->requested_height > vcaps->MaxOutputSize.cy ||
00415                     ctx->requested_height < vcaps->MinOutputSize.cy)
00416                     goto next;
00417                 bih->biWidth  = ctx->requested_width;
00418                 bih->biHeight = ctx->requested_height;
00419             }
00420         } else {
00421             AUDIO_STREAM_CONFIG_CAPS *acaps = caps;
00422             WAVEFORMATEX *fx;
00423 #if DSHOWDEBUG
00424             ff_print_AUDIO_STREAM_CONFIG_CAPS(acaps);
00425 #endif
00426             if (IsEqualGUID(&type->formattype, &FORMAT_WaveFormatEx)) {
00427                 fx = (void *) type->pbFormat;
00428             } else {
00429                 goto next;
00430             }
00431             if (!pformat_set) {
00432                 av_log(avctx, AV_LOG_INFO, "  min ch=%lu bits=%lu rate=%6lu max ch=%lu bits=%lu rate=%6lu\n",
00433                        acaps->MinimumChannels, acaps->MinimumBitsPerSample, acaps->MinimumSampleFrequency,
00434                        acaps->MaximumChannels, acaps->MaximumBitsPerSample, acaps->MaximumSampleFrequency);
00435                 continue;
00436             }
00437             if (ctx->sample_rate) {
00438                 if (ctx->sample_rate > acaps->MaximumSampleFrequency ||
00439                     ctx->sample_rate < acaps->MinimumSampleFrequency)
00440                     goto next;
00441                 fx->nSamplesPerSec = ctx->sample_rate;
00442             }
00443             if (ctx->sample_size) {
00444                 if (ctx->sample_size > acaps->MaximumBitsPerSample ||
00445                     ctx->sample_size < acaps->MinimumBitsPerSample)
00446                     goto next;
00447                 fx->wBitsPerSample = ctx->sample_size;
00448             }
00449             if (ctx->channels) {
00450                 if (ctx->channels > acaps->MaximumChannels ||
00451                     ctx->channels < acaps->MinimumChannels)
00452                     goto next;
00453                 fx->nChannels = ctx->channels;
00454             }
00455         }
00456         if (IAMStreamConfig_SetFormat(config, type) != S_OK)
00457             goto next;
00458         format_set = 1;
00459 next:
00460         if (type->pbFormat)
00461             CoTaskMemFree(type->pbFormat);
00462         CoTaskMemFree(type);
00463     }
00464 end:
00465     IAMStreamConfig_Release(config);
00466     if (caps)
00467         av_free(caps);
00468     if (pformat_set)
00469         *pformat_set = format_set;
00470 }
00471 
00476 static int
00477 dshow_set_audio_buffer_size(AVFormatContext *avctx, IPin *pin)
00478 {
00479     struct dshow_ctx *ctx = avctx->priv_data;
00480     IAMBufferNegotiation *buffer_negotiation = NULL;
00481     ALLOCATOR_PROPERTIES props = { -1, -1, -1, -1 };
00482     IAMStreamConfig *config = NULL;
00483     AM_MEDIA_TYPE *type = NULL;
00484     int ret = AVERROR(EIO);
00485 
00486     if (IPin_QueryInterface(pin, &IID_IAMStreamConfig, (void **) &config) != S_OK)
00487         goto end;
00488     if (IAMStreamConfig_GetFormat(config, &type) != S_OK)
00489         goto end;
00490     if (!IsEqualGUID(&type->formattype, &FORMAT_WaveFormatEx))
00491         goto end;
00492 
00493     props.cbBuffer = (((WAVEFORMATEX *) type->pbFormat)->nAvgBytesPerSec)
00494                    * ctx->audio_buffer_size / 1000;
00495 
00496     if (IPin_QueryInterface(pin, &IID_IAMBufferNegotiation, (void **) &buffer_negotiation) != S_OK)
00497         goto end;
00498     if (IAMBufferNegotiation_SuggestAllocatorProperties(buffer_negotiation, &props) != S_OK)
00499         goto end;
00500 
00501     ret = 0;
00502 
00503 end:
00504     if (buffer_negotiation)
00505         IAMBufferNegotiation_Release(buffer_negotiation);
00506     if (type) {
00507         if (type->pbFormat)
00508             CoTaskMemFree(type->pbFormat);
00509         CoTaskMemFree(type);
00510     }
00511     if (config)
00512         IAMStreamConfig_Release(config);
00513 
00514     return ret;
00515 }
00516 
00523 static int
00524 dshow_cycle_pins(AVFormatContext *avctx, enum dshowDeviceType devtype,
00525                  IBaseFilter *device_filter, IPin **ppin)
00526 {
00527     struct dshow_ctx *ctx = avctx->priv_data;
00528     IEnumPins *pins = 0;
00529     IPin *device_pin = NULL;
00530     IPin *pin;
00531     int r;
00532 
00533     const GUID *mediatype[2] = { &MEDIATYPE_Video, &MEDIATYPE_Audio };
00534     const char *devtypename = (devtype == VideoDevice) ? "video" : "audio";
00535 
00536     int set_format = (devtype == VideoDevice && (ctx->framerate ||
00537                                                 (ctx->requested_width && ctx->requested_height) ||
00538                                                  ctx->pixel_format != PIX_FMT_NONE ||
00539                                                  ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO))
00540                   || (devtype == AudioDevice && (ctx->channels || ctx->sample_rate));
00541     int format_set = 0;
00542 
00543     r = IBaseFilter_EnumPins(device_filter, &pins);
00544     if (r != S_OK) {
00545         av_log(avctx, AV_LOG_ERROR, "Could not enumerate pins.\n");
00546         return AVERROR(EIO);
00547     }
00548 
00549     if (!ppin) {
00550         av_log(avctx, AV_LOG_INFO, "DirectShow %s device options\n",
00551                devtypename);
00552     }
00553     while (!device_pin && IEnumPins_Next(pins, 1, &pin, NULL) == S_OK) {
00554         IKsPropertySet *p = NULL;
00555         IEnumMediaTypes *types = NULL;
00556         PIN_INFO info = {0};
00557         AM_MEDIA_TYPE *type;
00558         GUID category;
00559         DWORD r2;
00560 
00561         IPin_QueryPinInfo(pin, &info);
00562         IBaseFilter_Release(info.pFilter);
00563 
00564         if (info.dir != PINDIR_OUTPUT)
00565             goto next;
00566         if (IPin_QueryInterface(pin, &IID_IKsPropertySet, (void **) &p) != S_OK)
00567             goto next;
00568         if (IKsPropertySet_Get(p, &AMPROPSETID_Pin, AMPROPERTY_PIN_CATEGORY,
00569                                NULL, 0, &category, sizeof(GUID), &r2) != S_OK)
00570             goto next;
00571         if (!IsEqualGUID(&category, &PIN_CATEGORY_CAPTURE))
00572             goto next;
00573 
00574         if (!ppin) {
00575             char *buf = dup_wchar_to_utf8(info.achName);
00576             av_log(avctx, AV_LOG_INFO, " Pin \"%s\"\n", buf);
00577             av_free(buf);
00578             dshow_cycle_formats(avctx, devtype, pin, NULL);
00579             goto next;
00580         }
00581         if (set_format) {
00582             dshow_cycle_formats(avctx, devtype, pin, &format_set);
00583             if (!format_set) {
00584                 goto next;
00585             }
00586         }
00587         if (devtype == AudioDevice && ctx->audio_buffer_size) {
00588             if (dshow_set_audio_buffer_size(avctx, pin) < 0)
00589                 goto next;
00590         }
00591 
00592         if (IPin_EnumMediaTypes(pin, &types) != S_OK)
00593             goto next;
00594 
00595         IEnumMediaTypes_Reset(types);
00596         while (!device_pin && IEnumMediaTypes_Next(types, 1, &type, NULL) == S_OK) {
00597             if (IsEqualGUID(&type->majortype, mediatype[devtype])) {
00598                 device_pin = pin;
00599                 goto next;
00600             }
00601             CoTaskMemFree(type);
00602         }
00603 
00604 next:
00605         if (types)
00606             IEnumMediaTypes_Release(types);
00607         if (p)
00608             IKsPropertySet_Release(p);
00609         if (device_pin != pin)
00610             IPin_Release(pin);
00611     }
00612 
00613     IEnumPins_Release(pins);
00614 
00615     if (ppin) {
00616         if (set_format && !format_set) {
00617             av_log(avctx, AV_LOG_ERROR, "Could not set %s options\n", devtypename);
00618             return AVERROR(EIO);
00619         }
00620         if (!device_pin) {
00621             av_log(avctx, AV_LOG_ERROR,
00622                 "Could not find output pin from %s capture device.\n", devtypename);
00623             return AVERROR(EIO);
00624         }
00625         *ppin = device_pin;
00626     }
00627 
00628     return 0;
00629 }
00630 
00636 static int
00637 dshow_list_device_options(AVFormatContext *avctx, ICreateDevEnum *devenum,
00638                           enum dshowDeviceType devtype)
00639 {
00640     struct dshow_ctx *ctx = avctx->priv_data;
00641     IBaseFilter *device_filter = NULL;
00642     int r;
00643 
00644     if ((r = dshow_cycle_devices(avctx, devenum, devtype, &device_filter)) < 0)
00645         return r;
00646     ctx->device_filter[devtype] = device_filter;
00647     if ((r = dshow_cycle_pins(avctx, devtype, device_filter, NULL)) < 0)
00648         return r;
00649 
00650     return 0;
00651 }
00652 
00653 static int
00654 dshow_open_device(AVFormatContext *avctx, ICreateDevEnum *devenum,
00655                   enum dshowDeviceType devtype)
00656 {
00657     struct dshow_ctx *ctx = avctx->priv_data;
00658     IBaseFilter *device_filter = NULL;
00659     IGraphBuilder *graph = ctx->graph;
00660     IPin *device_pin = NULL;
00661     libAVPin *capture_pin = NULL;
00662     libAVFilter *capture_filter = NULL;
00663     int ret = AVERROR(EIO);
00664     int r;
00665 
00666     const wchar_t *filter_name[2] = { L"Audio capture filter", L"Video capture filter" };
00667 
00668     if ((r = dshow_cycle_devices(avctx, devenum, devtype, &device_filter)) < 0) {
00669         ret = r;
00670         goto error;
00671     }
00672 
00673     ctx->device_filter [devtype] = device_filter;
00674 
00675     r = IGraphBuilder_AddFilter(graph, device_filter, NULL);
00676     if (r != S_OK) {
00677         av_log(avctx, AV_LOG_ERROR, "Could not add device filter to graph.\n");
00678         goto error;
00679     }
00680 
00681     if ((r = dshow_cycle_pins(avctx, devtype, device_filter, &device_pin)) < 0) {
00682         ret = r;
00683         goto error;
00684     }
00685     ctx->device_pin[devtype] = device_pin;
00686 
00687     capture_filter = libAVFilter_Create(avctx, callback, devtype);
00688     if (!capture_filter) {
00689         av_log(avctx, AV_LOG_ERROR, "Could not create grabber filter.\n");
00690         goto error;
00691     }
00692     ctx->capture_filter[devtype] = capture_filter;
00693 
00694     r = IGraphBuilder_AddFilter(graph, (IBaseFilter *) capture_filter,
00695                                 filter_name[devtype]);
00696     if (r != S_OK) {
00697         av_log(avctx, AV_LOG_ERROR, "Could not add capture filter to graph\n");
00698         goto error;
00699     }
00700 
00701     libAVPin_AddRef(capture_filter->pin);
00702     capture_pin = capture_filter->pin;
00703     ctx->capture_pin[devtype] = capture_pin;
00704 
00705     r = IGraphBuilder_ConnectDirect(graph, device_pin, (IPin *) capture_pin, NULL);
00706     if (r != S_OK) {
00707         av_log(avctx, AV_LOG_ERROR, "Could not connect pins\n");
00708         goto error;
00709     }
00710 
00711     ret = 0;
00712 
00713 error:
00714     return ret;
00715 }
00716 
00717 static enum AVCodecID waveform_codec_id(enum AVSampleFormat sample_fmt)
00718 {
00719     switch (sample_fmt) {
00720     case AV_SAMPLE_FMT_U8:  return AV_CODEC_ID_PCM_U8;
00721     case AV_SAMPLE_FMT_S16: return AV_CODEC_ID_PCM_S16LE;
00722     case AV_SAMPLE_FMT_S32: return AV_CODEC_ID_PCM_S32LE;
00723     default:                return AV_CODEC_ID_NONE; /* Should never happen. */
00724     }
00725 }
00726 
00727 static enum AVSampleFormat sample_fmt_bits_per_sample(int bits)
00728 {
00729     switch (bits) {
00730     case 8:  return AV_SAMPLE_FMT_U8;
00731     case 16: return AV_SAMPLE_FMT_S16;
00732     case 32: return AV_SAMPLE_FMT_S32;
00733     default: return AV_SAMPLE_FMT_NONE; /* Should never happen. */
00734     }
00735 }
00736 
00737 static int
00738 dshow_add_device(AVFormatContext *avctx,
00739                  enum dshowDeviceType devtype)
00740 {
00741     struct dshow_ctx *ctx = avctx->priv_data;
00742     AM_MEDIA_TYPE type;
00743     AVCodecContext *codec;
00744     AVStream *st;
00745     int ret = AVERROR(EIO);
00746 
00747     st = avformat_new_stream(avctx, NULL);
00748     if (!st) {
00749         ret = AVERROR(ENOMEM);
00750         goto error;
00751     }
00752     st->id = devtype;
00753 
00754     ctx->capture_filter[devtype]->stream_index = st->index;
00755 
00756     libAVPin_ConnectionMediaType(ctx->capture_pin[devtype], &type);
00757 
00758     codec = st->codec;
00759     if (devtype == VideoDevice) {
00760         BITMAPINFOHEADER *bih = NULL;
00761         AVRational time_base;
00762 
00763         if (IsEqualGUID(&type.formattype, &FORMAT_VideoInfo)) {
00764             VIDEOINFOHEADER *v = (void *) type.pbFormat;
00765             time_base = (AVRational) { v->AvgTimePerFrame, 10000000 };
00766             bih = &v->bmiHeader;
00767         } else if (IsEqualGUID(&type.formattype, &FORMAT_VideoInfo2)) {
00768             VIDEOINFOHEADER2 *v = (void *) type.pbFormat;
00769             time_base = (AVRational) { v->AvgTimePerFrame, 10000000 };
00770             bih = &v->bmiHeader;
00771         }
00772         if (!bih) {
00773             av_log(avctx, AV_LOG_ERROR, "Could not get media type.\n");
00774             goto error;
00775         }
00776 
00777         codec->time_base  = time_base;
00778         codec->codec_type = AVMEDIA_TYPE_VIDEO;
00779         codec->width      = bih->biWidth;
00780         codec->height     = bih->biHeight;
00781         codec->pix_fmt    = dshow_pixfmt(bih->biCompression, bih->biBitCount);
00782         if (codec->pix_fmt == PIX_FMT_NONE) {
00783             codec->codec_id = dshow_codecid(bih->biCompression);
00784             if (codec->codec_id == AV_CODEC_ID_NONE) {
00785                 av_log(avctx, AV_LOG_ERROR, "Unknown compression type. "
00786                                  "Please report verbose (-v 9) debug information.\n");
00787                 dshow_read_close(avctx);
00788                 return AVERROR_PATCHWELCOME;
00789             }
00790             codec->bits_per_coded_sample = bih->biBitCount;
00791         } else {
00792             codec->codec_id = AV_CODEC_ID_RAWVIDEO;
00793             if (bih->biCompression == BI_RGB || bih->biCompression == BI_BITFIELDS) {
00794                 codec->bits_per_coded_sample = bih->biBitCount;
00795                 codec->extradata = av_malloc(9 + FF_INPUT_BUFFER_PADDING_SIZE);
00796                 if (codec->extradata) {
00797                     codec->extradata_size = 9;
00798                     memcpy(codec->extradata, "BottomUp", 9);
00799                 }
00800             }
00801         }
00802     } else {
00803         WAVEFORMATEX *fx = NULL;
00804 
00805         if (IsEqualGUID(&type.formattype, &FORMAT_WaveFormatEx)) {
00806             fx = (void *) type.pbFormat;
00807         }
00808         if (!fx) {
00809             av_log(avctx, AV_LOG_ERROR, "Could not get media type.\n");
00810             goto error;
00811         }
00812 
00813         codec->codec_type  = AVMEDIA_TYPE_AUDIO;
00814         codec->sample_fmt  = sample_fmt_bits_per_sample(fx->wBitsPerSample);
00815         codec->codec_id    = waveform_codec_id(codec->sample_fmt);
00816         codec->sample_rate = fx->nSamplesPerSec;
00817         codec->channels    = fx->nChannels;
00818     }
00819 
00820     avpriv_set_pts_info(st, 64, 1, 10000000);
00821 
00822     ret = 0;
00823 
00824 error:
00825     return ret;
00826 }
00827 
00828 static int parse_device_name(AVFormatContext *avctx)
00829 {
00830     struct dshow_ctx *ctx = avctx->priv_data;
00831     char **device_name = ctx->device_name;
00832     char *name = av_strdup(avctx->filename);
00833     char *tmp = name;
00834     int ret = 1;
00835     char *type;
00836 
00837     while ((type = strtok(tmp, "="))) {
00838         char *token = strtok(NULL, ":");
00839         tmp = NULL;
00840 
00841         if        (!strcmp(type, "video")) {
00842             device_name[0] = token;
00843         } else if (!strcmp(type, "audio")) {
00844             device_name[1] = token;
00845         } else {
00846             device_name[0] = NULL;
00847             device_name[1] = NULL;
00848             break;
00849         }
00850     }
00851 
00852     if (!device_name[0] && !device_name[1]) {
00853         ret = 0;
00854     } else {
00855         if (device_name[0])
00856             device_name[0] = av_strdup(device_name[0]);
00857         if (device_name[1])
00858             device_name[1] = av_strdup(device_name[1]);
00859     }
00860 
00861     av_free(name);
00862     return ret;
00863 }
00864 
00865 static int dshow_read_header(AVFormatContext *avctx)
00866 {
00867     struct dshow_ctx *ctx = avctx->priv_data;
00868     IGraphBuilder *graph = NULL;
00869     ICreateDevEnum *devenum = NULL;
00870     IMediaControl *control = NULL;
00871     int ret = AVERROR(EIO);
00872     int r;
00873 
00874     if (!ctx->list_devices && !parse_device_name(avctx)) {
00875         av_log(avctx, AV_LOG_ERROR, "Malformed dshow input string.\n");
00876         goto error;
00877     }
00878 
00879     ctx->video_codec_id = avctx->video_codec_id ? avctx->video_codec_id
00880                                                 : AV_CODEC_ID_RAWVIDEO;
00881     if (ctx->pixel_format != PIX_FMT_NONE) {
00882         if (ctx->video_codec_id != AV_CODEC_ID_RAWVIDEO) {
00883             av_log(avctx, AV_LOG_ERROR, "Pixel format may only be set when "
00884                               "video codec is not set or set to rawvideo\n");
00885             ret = AVERROR(EINVAL);
00886             goto error;
00887         }
00888     }
00889     if (ctx->framerate) {
00890         r = av_parse_video_rate(&ctx->requested_framerate, ctx->framerate);
00891         if (r < 0) {
00892             av_log(avctx, AV_LOG_ERROR, "Could not parse framerate '%s'.\n", ctx->framerate);
00893             goto error;
00894         }
00895     }
00896 
00897     CoInitialize(0);
00898 
00899     r = CoCreateInstance(&CLSID_FilterGraph, NULL, CLSCTX_INPROC_SERVER,
00900                          &IID_IGraphBuilder, (void **) &graph);
00901     if (r != S_OK) {
00902         av_log(avctx, AV_LOG_ERROR, "Could not create capture graph.\n");
00903         goto error;
00904     }
00905     ctx->graph = graph;
00906 
00907     r = CoCreateInstance(&CLSID_SystemDeviceEnum, NULL, CLSCTX_INPROC_SERVER,
00908                          &IID_ICreateDevEnum, (void **) &devenum);
00909     if (r != S_OK) {
00910         av_log(avctx, AV_LOG_ERROR, "Could not enumerate system devices.\n");
00911         goto error;
00912     }
00913 
00914     if (ctx->list_devices) {
00915         av_log(avctx, AV_LOG_INFO, "DirectShow video devices\n");
00916         dshow_cycle_devices(avctx, devenum, VideoDevice, NULL);
00917         av_log(avctx, AV_LOG_INFO, "DirectShow audio devices\n");
00918         dshow_cycle_devices(avctx, devenum, AudioDevice, NULL);
00919         ret = AVERROR_EXIT;
00920         goto error;
00921     }
00922     if (ctx->list_options) {
00923         if (ctx->device_name[VideoDevice])
00924             dshow_list_device_options(avctx, devenum, VideoDevice);
00925         if (ctx->device_name[AudioDevice])
00926             dshow_list_device_options(avctx, devenum, AudioDevice);
00927         ret = AVERROR_EXIT;
00928         goto error;
00929     }
00930 
00931     if (ctx->device_name[VideoDevice]) {
00932         ret = dshow_open_device(avctx, devenum, VideoDevice);
00933         if (ret < 0)
00934             goto error;
00935         ret = dshow_add_device(avctx, VideoDevice);
00936         if (ret < 0)
00937             goto error;
00938     }
00939     if (ctx->device_name[AudioDevice]) {
00940         ret = dshow_open_device(avctx, devenum, AudioDevice);
00941         if (ret < 0)
00942             goto error;
00943         ret = dshow_add_device(avctx, AudioDevice);
00944         if (ret < 0)
00945             goto error;
00946     }
00947 
00948     ctx->mutex = CreateMutex(NULL, 0, NULL);
00949     if (!ctx->mutex) {
00950         av_log(avctx, AV_LOG_ERROR, "Could not create Mutex\n");
00951         goto error;
00952     }
00953     ctx->event = CreateEvent(NULL, 1, 0, NULL);
00954     if (!ctx->event) {
00955         av_log(avctx, AV_LOG_ERROR, "Could not create Event\n");
00956         goto error;
00957     }
00958 
00959     r = IGraphBuilder_QueryInterface(graph, &IID_IMediaControl, (void **) &control);
00960     if (r != S_OK) {
00961         av_log(avctx, AV_LOG_ERROR, "Could not get media control.\n");
00962         goto error;
00963     }
00964     ctx->control = control;
00965 
00966     r = IMediaControl_Run(control);
00967     if (r == S_FALSE) {
00968         OAFilterState pfs;
00969         r = IMediaControl_GetState(control, 0, &pfs);
00970     }
00971     if (r != S_OK) {
00972         av_log(avctx, AV_LOG_ERROR, "Could not run filter\n");
00973         goto error;
00974     }
00975 
00976     ret = 0;
00977 
00978 error:
00979 
00980     if (ret < 0)
00981         dshow_read_close(avctx);
00982 
00983     if (devenum)
00984         ICreateDevEnum_Release(devenum);
00985 
00986     return ret;
00987 }
00988 
00989 static int dshow_read_packet(AVFormatContext *s, AVPacket *pkt)
00990 {
00991     struct dshow_ctx *ctx = s->priv_data;
00992     AVPacketList *pktl = NULL;
00993 
00994     while (!pktl) {
00995         WaitForSingleObject(ctx->mutex, INFINITE);
00996         pktl = ctx->pktl;
00997         if (pktl) {
00998             *pkt = pktl->pkt;
00999             ctx->pktl = ctx->pktl->next;
01000             av_free(pktl);
01001             ctx->curbufsize -= pkt->size;
01002         }
01003         ResetEvent(ctx->event);
01004         ReleaseMutex(ctx->mutex);
01005         if (!pktl) {
01006             if (s->flags & AVFMT_FLAG_NONBLOCK) {
01007                 return AVERROR(EAGAIN);
01008             } else {
01009                 WaitForSingleObject(ctx->event, INFINITE);
01010             }
01011         }
01012     }
01013 
01014     return pkt->size;
01015 }
01016 
01017 #define OFFSET(x) offsetof(struct dshow_ctx, x)
01018 #define DEC AV_OPT_FLAG_DECODING_PARAM
01019 static const AVOption options[] = {
01020     { "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 },
01021     { "pixel_format", "set video pixel format", OFFSET(pixel_format), AV_OPT_TYPE_PIXEL_FMT, {.str = NULL}, 0, 0, DEC },
01022     { "framerate", "set video frame rate", OFFSET(framerate), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, DEC },
01023     { "sample_rate", "set audio sample rate", OFFSET(sample_rate), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
01024     { "sample_size", "set audio sample size", OFFSET(sample_size), AV_OPT_TYPE_INT, {.i64 = 0}, 0, 16, DEC },
01025     { "channels", "set number of audio channels, such as 1 or 2", OFFSET(channels), AV_OPT_TYPE_INT, {.i64 = 0}, 0, INT_MAX, DEC },
01026     { "list_devices", "list available devices", OFFSET(list_devices), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, DEC, "list_devices" },
01027     { "true", "", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, DEC, "list_devices" },
01028     { "false", "", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, DEC, "list_devices" },
01029     { "list_options", "list available options for specified device", OFFSET(list_options), AV_OPT_TYPE_INT, {.i64=0}, 0, 1, DEC, "list_options" },
01030     { "true", "", 0, AV_OPT_TYPE_CONST, {.i64=1}, 0, 0, DEC, "list_options" },
01031     { "false", "", 0, AV_OPT_TYPE_CONST, {.i64=0}, 0, 0, DEC, "list_options" },
01032     { "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 },
01033     { "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 },
01034     { "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 },
01035     { NULL },
01036 };
01037 
01038 static const AVClass dshow_class = {
01039     .class_name = "DirectShow indev",
01040     .item_name  = av_default_item_name,
01041     .option     = options,
01042     .version    = LIBAVUTIL_VERSION_INT,
01043 };
01044 
01045 AVInputFormat ff_dshow_demuxer = {
01046     .name           = "dshow",
01047     .long_name      = NULL_IF_CONFIG_SMALL("DirectShow capture"),
01048     .priv_data_size = sizeof(struct dshow_ctx),
01049     .read_header    = dshow_read_header,
01050     .read_packet    = dshow_read_packet,
01051     .read_close     = dshow_read_close,
01052     .flags          = AVFMT_NOFILE,
01053     .priv_class     = &dshow_class,
01054 };