/* This Source Code Form is subject to the terms of the Mozilla Public * License, v. 2.0. If a copy of the MPL was not distributed with this file, * You can obtain one at http://mozilla.org/MPL/2.0/. */ #include "MediaEngineWebRTC.h" #include "Layers.h" #include "ImageTypes.h" #include "ImageContainer.h" #include "mtransport/runnable_utils.h" namespace mozilla { #ifdef PR_LOGGING extern PRLogModuleInfo* GetMediaManagerLog(); #define LOG(msg) PR_LOG(GetMediaManagerLog(), PR_LOG_DEBUG, msg) #define LOGFRAME(msg) PR_LOG(GetMediaManagerLog(), 6, msg) #else #define LOG(msg) #define LOGFRAME(msg) #endif /** * Webrtc video source. */ NS_IMPL_THREADSAFE_ISUPPORTS1(MediaEngineWebRTCVideoSource, nsIRunnable) // ViEExternalRenderer Callback. int MediaEngineWebRTCVideoSource::FrameSizeChange( unsigned int w, unsigned int h, unsigned int streams) { mWidth = w; mHeight = h; LOG(("Video FrameSizeChange: %ux%u", w, h)); return 0; } // ViEExternalRenderer Callback. Process every incoming frame here. int MediaEngineWebRTCVideoSource::DeliverFrame( unsigned char* buffer, int size, uint32_t time_stamp, int64_t render_time) { // mInSnapshotMode can only be set before the camera is turned on and // the renderer is started, so this amounts to a 1-shot if (mInSnapshotMode) { // Set the condition variable to false and notify Snapshot(). MonitorAutoLock lock(mMonitor); mInSnapshotMode = false; lock.Notify(); return 0; } // Check for proper state. if (mState != kStarted) { LOG(("DeliverFrame: video not started")); return 0; } MOZ_ASSERT(mWidth*mHeight*3/2 == size); if (mWidth*mHeight*3/2 != size) { return 0; } // Create a video frame and append it to the track. ImageFormat format = PLANAR_YCBCR; nsRefPtr image = mImageContainer->CreateImage(&format, 1); layers::PlanarYCbCrImage* videoImage = static_cast(image.get()); uint8_t* frame = static_cast (buffer); const uint8_t lumaBpp = 8; const uint8_t chromaBpp = 4; layers::PlanarYCbCrImage::Data data; data.mYChannel = frame; data.mYSize = gfxIntSize(mWidth, mHeight); data.mYStride = mWidth * lumaBpp/ 8; data.mCbCrStride = mWidth * chromaBpp / 8; data.mCbChannel = frame + mHeight * data.mYStride; data.mCrChannel = data.mCbChannel + mHeight * data.mCbCrStride / 2; data.mCbCrSize = gfxIntSize(mWidth/ 2, mHeight/ 2); data.mPicX = 0; data.mPicY = 0; data.mPicSize = gfxIntSize(mWidth, mHeight); data.mStereoMode = STEREO_MODE_MONO; videoImage->SetData(data); #ifdef DEBUG static uint32_t frame_num = 0; LOGFRAME(("frame %d (%dx%d); timestamp %u, render_time %lu", frame_num++, mWidth, mHeight, time_stamp, render_time)); #endif // we don't touch anything in 'this' until here (except for snapshot, // which has it's own lock) MonitorAutoLock lock(mMonitor); // implicitly releases last image mImage = image.forget(); return 0; } // Called if the graph thinks it's running out of buffered video; repeat // the last frame for whatever minimum period it think it needs. Note that // this means that no *real* frame can be inserted during this period. void MediaEngineWebRTCVideoSource::NotifyPull(MediaStreamGraph* aGraph, SourceMediaStream *aSource, TrackID aID, StreamTime aDesiredTime, TrackTicks &aLastEndTime) { VideoSegment segment; MonitorAutoLock lock(mMonitor); if (mState != kStarted) return; // Note: we're not giving up mImage here nsRefPtr image = mImage; TrackTicks target = TimeToTicksRoundUp(USECS_PER_S, aDesiredTime); TrackTicks delta = target - aLastEndTime; LOGFRAME(("NotifyPull, desired = %ld, target = %ld, delta = %ld %s", (int64_t) aDesiredTime, (int64_t) target, (int64_t) delta, image ? "" : "")); // Bug 846188 We may want to limit incoming frames to the requested frame rate // mFps - if you want 30FPS, and the camera gives you 60FPS, this could // cause issues. // We may want to signal if the actual frame rate is below mMinFPS - // cameras often don't return the requested frame rate especially in low // light; we should consider surfacing this so that we can switch to a // lower resolution (which may up the frame rate) // Don't append if we've already provided a frame that supposedly goes past the current aDesiredTime // Doing so means a negative delta and thus messes up handling of the graph if (delta > 0) { // NULL images are allowed if (image) { segment.AppendFrame(image.forget(), delta, gfxIntSize(mWidth, mHeight)); } else { segment.AppendFrame(nullptr, delta, gfxIntSize(0,0)); } // This can fail if either a) we haven't added the track yet, or b) // we've removed or finished the track. if (aSource->AppendToTrack(aID, &(segment))) { aLastEndTime = target; } } } void MediaEngineWebRTCVideoSource::ChooseCapability(const MediaEnginePrefs &aPrefs) { int num = mViECapture->NumberOfCapabilities(mUniqueId, KMaxUniqueIdLength); LOG(("ChooseCapability: prefs: %dx%d @%d-%dfps", aPrefs.mWidth, aPrefs.mHeight, aPrefs.mFPS, aPrefs.mMinFPS)); if (num <= 0) { // Set to default values mCapability.width = aPrefs.mWidth; mCapability.height = aPrefs.mHeight; mCapability.maxFPS = MediaEngine::DEFAULT_VIDEO_FPS; // Mac doesn't support capabilities. return; } // Default is closest to available capability but equal to or below; // otherwise closest above. Since we handle the num=0 case above and // take the first entry always, we can never exit uninitialized. webrtc::CaptureCapability cap; bool higher = true; for (int i = 0; i < num; i++) { mViECapture->GetCaptureCapability(mUniqueId, KMaxUniqueIdLength, i, cap); if (higher) { if (i == 0 || (mCapability.width > cap.width && mCapability.height > cap.height)) { // closer than the current choice mCapability = cap; // FIXME: expose expected capture delay? } if (cap.width <= (uint32_t) aPrefs.mWidth && cap.height <= (uint32_t) aPrefs.mHeight) { higher = false; } } else { if (cap.width > (uint32_t) aPrefs.mWidth || cap.height > (uint32_t) aPrefs.mHeight || cap.maxFPS < (uint32_t) aPrefs.mMinFPS) { continue; } if (mCapability.width < cap.width && mCapability.height < cap.height) { mCapability = cap; // FIXME: expose expected capture delay? } } } LOG(("chose cap %dx%d @%dfps", mCapability.width, mCapability.height, mCapability.maxFPS)); } void MediaEngineWebRTCVideoSource::GetName(nsAString& aName) { // mDeviceName is UTF8 CopyUTF8toUTF16(mDeviceName, aName); } void MediaEngineWebRTCVideoSource::GetUUID(nsAString& aUUID) { // mUniqueId is UTF8 CopyUTF8toUTF16(mUniqueId, aUUID); } nsresult MediaEngineWebRTCVideoSource::Allocate(const MediaEnginePrefs &aPrefs) { LOG((__FUNCTION__)); if (mState == kReleased && mInitDone) { // Note: if shared, we don't allow a later opener to affect the resolution. // (This may change depending on spec changes for Constraints/settings) ChooseCapability(aPrefs); if (mViECapture->AllocateCaptureDevice(mUniqueId, KMaxUniqueIdLength, mCaptureIndex)) { return NS_ERROR_FAILURE; } mState = kAllocated; LOG(("Video device %d allocated", mCaptureIndex)); } else if (mSources.IsEmpty()) { LOG(("Video device %d reallocated", mCaptureIndex)); } else { LOG(("Video device %d allocated shared", mCaptureIndex)); } return NS_OK; } nsresult MediaEngineWebRTCVideoSource::Deallocate() { LOG((__FUNCTION__)); if (mSources.IsEmpty()) { if (mState != kStopped && mState != kAllocated) { return NS_ERROR_FAILURE; } #ifdef XP_MACOSX // Bug 829907 - on mac, in shutdown, the mainthread stops processing // 'native' events, and the QTKit code uses events to the main native CFRunLoop // in order to provide thread safety. In order to avoid this locking us up, // release the ViE capture device synchronously on MainThread (so the native // event isn't needed). // XXX Note if MainThread Dispatch()es NS_DISPATCH_SYNC to us we can deadlock. // XXX It might be nice to only do this if we're in shutdown... Hard to be // sure when that is though. // Thread safety: a) we call this synchronously, and don't use ViECapture from // another thread anywhere else, b) ViEInputManager::DestroyCaptureDevice() grabs // an exclusive object lock and deletes it in a critical section, so all in all // this should be safe threadwise. NS_DispatchToMainThread(WrapRunnable(mViECapture, &webrtc::ViECapture::ReleaseCaptureDevice, mCaptureIndex), NS_DISPATCH_SYNC); #else mViECapture->ReleaseCaptureDevice(mCaptureIndex); #endif mState = kReleased; LOG(("Video device %d deallocated", mCaptureIndex)); } else { LOG(("Video device %d deallocated but still in use", mCaptureIndex)); } return NS_OK; } nsresult MediaEngineWebRTCVideoSource::Start(SourceMediaStream* aStream, TrackID aID) { LOG((__FUNCTION__)); int error = 0; if (!mInitDone || !aStream) { return NS_ERROR_FAILURE; } mSources.AppendElement(aStream); aStream->AddTrack(aID, USECS_PER_S, 0, new VideoSegment()); aStream->AdvanceKnownTracksTime(STREAM_TIME_MAX); if (mState == kStarted) { return NS_OK; } mState = kStarted; mImageContainer = layers::LayerManager::CreateImageContainer(); error = mViERender->AddRenderer(mCaptureIndex, webrtc::kVideoI420, (webrtc::ExternalRenderer*)this); if (error == -1) { return NS_ERROR_FAILURE; } error = mViERender->StartRender(mCaptureIndex); if (error == -1) { return NS_ERROR_FAILURE; } if (mViECapture->StartCapture(mCaptureIndex, mCapability) < 0) { return NS_ERROR_FAILURE; } return NS_OK; } nsresult MediaEngineWebRTCVideoSource::Stop(SourceMediaStream *aSource, TrackID aID) { LOG((__FUNCTION__)); if (!mSources.RemoveElement(aSource)) { // Already stopped - this is allowed return NS_OK; } if (!mSources.IsEmpty()) { return NS_OK; } if (mState != kStarted) { return NS_ERROR_FAILURE; } { MonitorAutoLock lock(mMonitor); mState = kStopped; aSource->EndTrack(aID); // Drop any cached image so we don't start with a stale image on next // usage mImage = nullptr; } mViERender->StopRender(mCaptureIndex); mViERender->RemoveRenderer(mCaptureIndex); mViECapture->StopCapture(mCaptureIndex); return NS_OK; } nsresult MediaEngineWebRTCVideoSource::Snapshot(uint32_t aDuration, nsIDOMFile** aFile) { /** * To get a Snapshot we do the following: * - Set a condition variable (mInSnapshotMode) to true * - Attach the external renderer and start the camera * - Wait for the condition variable to change to false * * Starting the camera has the effect of invoking DeliverFrame() when * the first frame arrives from the camera. We only need one frame for * GetCaptureDeviceSnapshot to work, so we immediately set the condition * variable to false and notify this method. * * This causes the current thread to continue (PR_CondWaitVar will return), * at which point we can grab a snapshot, convert it to a file and * return from this function after cleaning up the temporary stream object * and caling Stop() on the media source. */ *aFile = nullptr; if (!mInitDone || mState != kAllocated) { return NS_ERROR_FAILURE; } { MonitorAutoLock lock(mMonitor); mInSnapshotMode = true; } // Start the rendering (equivalent to calling Start(), but without a track). int error = 0; if (!mInitDone || mState != kAllocated) { return NS_ERROR_FAILURE; } error = mViERender->AddRenderer(mCaptureIndex, webrtc::kVideoI420, (webrtc::ExternalRenderer*)this); if (error == -1) { return NS_ERROR_FAILURE; } error = mViERender->StartRender(mCaptureIndex); if (error == -1) { return NS_ERROR_FAILURE; } if (mViECapture->StartCapture(mCaptureIndex, mCapability) < 0) { return NS_ERROR_FAILURE; } // Wait for the condition variable, will be set in DeliverFrame. // We use a while loop, because even if Wait() returns, it's not // guaranteed that the condition variable changed. // FIX: we need need a way to cancel this and to bail if it appears to not be working // Perhaps a maximum time, though some cameras can take seconds to start. 10 seconds? { MonitorAutoLock lock(mMonitor); while (mInSnapshotMode) { lock.Wait(); } } // If we get here, DeliverFrame received at least one frame. webrtc::ViEFile* vieFile = webrtc::ViEFile::GetInterface(mVideoEngine); if (!vieFile) { return NS_ERROR_FAILURE; } // Create a temporary file on the main thread and put the snapshot in it. // See Run() in MediaEngineWebRTCVideo.h (sets mSnapshotPath). NS_DispatchToMainThread(this, NS_DISPATCH_SYNC); if (!mSnapshotPath) { return NS_ERROR_FAILURE; } NS_ConvertUTF16toUTF8 path(*mSnapshotPath); if (vieFile->GetCaptureDeviceSnapshot(mCaptureIndex, path.get()) < 0) { delete mSnapshotPath; mSnapshotPath = NULL; return NS_ERROR_FAILURE; } // Stop the camera. mViERender->StopRender(mCaptureIndex); mViERender->RemoveRenderer(mCaptureIndex); nsCOMPtr file; nsresult rv = NS_NewLocalFile(*mSnapshotPath, false, getter_AddRefs(file)); delete mSnapshotPath; mSnapshotPath = NULL; NS_ENSURE_SUCCESS(rv, rv); NS_ADDREF(*aFile = new nsDOMFileFile(file)); return NS_OK; } /** * Initialization and Shutdown functions for the video source, called by the * constructor and destructor respectively. */ void MediaEngineWebRTCVideoSource::Init() { mDeviceName[0] = '\0'; // paranoia mUniqueId[0] = '\0'; // fix compile warning for these being unused. (remove once used) (void) mFps; (void) mMinFps; LOG((__FUNCTION__)); if (mVideoEngine == NULL) { return; } mViEBase = webrtc::ViEBase::GetInterface(mVideoEngine); if (mViEBase == NULL) { return; } // Get interfaces for capture, render for now mViECapture = webrtc::ViECapture::GetInterface(mVideoEngine); mViERender = webrtc::ViERender::GetInterface(mVideoEngine); if (mViECapture == NULL || mViERender == NULL) { return; } if (mViECapture->GetCaptureDevice(mCaptureIndex, mDeviceName, sizeof(mDeviceName), mUniqueId, sizeof(mUniqueId))) { return; } mInitDone = true; } void MediaEngineWebRTCVideoSource::Shutdown() { LOG((__FUNCTION__)); if (!mInitDone) { return; } if (mState == kStarted) { while (!mSources.IsEmpty()) { Stop(mSources[0], kVideoTrack); // XXX change to support multiple tracks } MOZ_ASSERT(mState == kStopped); } if (mState == kAllocated || mState == kStopped) { Deallocate(); } mViECapture->Release(); mViERender->Release(); mViEBase->Release(); mState = kReleased; mInitDone = false; } }