/** * @file llvoicewebrtc.cpp * @brief Implementation of LLWebRTCVoiceClient class which is the interface to the voice client process. * * $LicenseInfo:firstyear=2001&license=viewerlgpl$ * Second Life Viewer Source Code * Copyright (C) 2023, Linden Research, Inc. * * This library is free software; you can redistribute it and/or * modify it under the terms of the GNU Lesser General Public * License as published by the Free Software Foundation * version 2.1 of the License only. * * This library is distributed in the hope that it will be useful, * but WITHOUT ANY WARRANTY; without even the implied warranty of * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU * Lesser General Public License for more details. * * You should have received a copy of the GNU Lesser General Public * License along with this library; if not, write to the Free Software * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA * * Linden Research, Inc., 945 Battery Street, San Francisco, CA 94111 USA * $/LicenseInfo$ */ #include #include "llvoicewebrtc.h" #include "llsdutil.h" // Linden library includes #include "llavatarnamecache.h" #include "llvoavatarself.h" #include "llbufferstream.h" #include "llfile.h" #include "llmenugl.h" #ifdef LL_USESYSTEMLIBS # include "expat.h" #else # include "expat/expat.h" #endif #include "llcallbacklist.h" #include "llviewernetwork.h" // for gGridChoice #include "llbase64.h" #include "llviewercontrol.h" #include "llappviewer.h" // for gDisconnected, gDisableVoice #include "llprocess.h" // Viewer includes #include "llmutelist.h" // to check for muted avatars #include "llagent.h" #include "llcachename.h" #include "llimview.h" // for LLIMMgr #include "llworld.h" #include "llparcel.h" #include "llviewerparcelmgr.h" #include "llfirstuse.h" #include "llspeakers.h" #include "lltrans.h" #include "llrand.h" #include "llviewerwindow.h" #include "llviewercamera.h" #include "llversioninfo.h" #include "llviewernetwork.h" #include "llnotificationsutil.h" #include "llcorehttputil.h" #include "lleventfilter.h" #include "stringize.h" #include "llwebrtc.h" // for base64 decoding #include "apr_base64.h" #include "boost/json.hpp" const std::string WEBRTC_VOICE_SERVER_TYPE = "webrtc"; namespace { const F32 MAX_AUDIO_DIST = 50.0f; const F32 VOLUME_SCALE_WEBRTC = 0.01f; const F32 LEVEL_SCALE_WEBRTC = 0.008f; const F32 SPEAKING_AUDIO_LEVEL = 0.30; static const std::string REPORTED_VOICE_SERVER_TYPE = "Secondlife WebRTC Gateway"; // Don't send positional updates more frequently than this: const F32 UPDATE_THROTTLE_SECONDS = 0.1f; const F32 MAX_RETRY_WAIT_SECONDS = 10.0f; // Cosine of a "trivially" small angle const F32 FOUR_DEGREES = 4.0f * (F_PI / 180.0f); const F32 MINUSCULE_ANGLE_COS = (F32) cos(0.5f * FOUR_DEGREES); } // namespace /////////////////////////////////////////////////////////////////////////////////////////////// void LLVoiceWebRTCStats::reset() { mStartTime = -1.0f; mConnectCycles = 0; mConnectTime = -1.0f; mConnectAttempts = 0; mProvisionTime = -1.0f; mProvisionAttempts = 0; mEstablishTime = -1.0f; mEstablishAttempts = 0; } LLVoiceWebRTCStats::LLVoiceWebRTCStats() { reset(); } LLVoiceWebRTCStats::~LLVoiceWebRTCStats() { } void LLVoiceWebRTCStats::connectionAttemptStart() { if (!mConnectAttempts) { mStartTime = LLTimer::getTotalTime(); mConnectCycles++; } mConnectAttempts++; } void LLVoiceWebRTCStats::connectionAttemptEnd(bool success) { if ( success ) { mConnectTime = (LLTimer::getTotalTime() - mStartTime) / USEC_PER_SEC; } } void LLVoiceWebRTCStats::provisionAttemptStart() { if (!mProvisionAttempts) { mStartTime = LLTimer::getTotalTime(); } mProvisionAttempts++; } void LLVoiceWebRTCStats::provisionAttemptEnd(bool success) { if ( success ) { mProvisionTime = (LLTimer::getTotalTime() - mStartTime) / USEC_PER_SEC; } } void LLVoiceWebRTCStats::establishAttemptStart() { if (!mEstablishAttempts) { mStartTime = LLTimer::getTotalTime(); } mEstablishAttempts++; } void LLVoiceWebRTCStats::establishAttemptEnd(bool success) { if ( success ) { mEstablishTime = (LLTimer::getTotalTime() - mStartTime) / USEC_PER_SEC; } } LLSD LLVoiceWebRTCStats::read() { LLSD stats(LLSD::emptyMap()); stats["connect_cycles"] = LLSD::Integer(mConnectCycles); stats["connect_attempts"] = LLSD::Integer(mConnectAttempts); stats["connect_time"] = LLSD::Real(mConnectTime); stats["provision_attempts"] = LLSD::Integer(mProvisionAttempts); stats["provision_time"] = LLSD::Real(mProvisionTime); stats["establish_attempts"] = LLSD::Integer(mEstablishAttempts); stats["establish_time"] = LLSD::Real(mEstablishTime); return stats; } /////////////////////////////////////////////////////////////////////////////////////////////// bool LLWebRTCVoiceClient::sShuttingDown = false; LLWebRTCVoiceClient::LLWebRTCVoiceClient() : mHidden(false), mTuningMode(false), mTuningMicGain(0.0), mTuningSpeakerVolume(50), // Set to 50 so the user can hear themselves when he sets his mic volume mDevicesListUpdated(false), mSpatialCoordsDirty(false), mMuteMic(false), mEarLocation(0), mMicGain(0.0), mVoiceEnabled(false), mProcessChannels(false), mAvatarNameCacheConnection(), mIsInTuningMode(false), mIsProcessingChannels(false), mIsCoroutineActive(false), mWebRTCPump("WebRTCClientPump"), mWebRTCDeviceInterface(nullptr) { sShuttingDown = false; mSpeakerVolume = 0.0; mVoiceVersion.serverVersion = ""; mVoiceVersion.voiceServerType = REPORTED_VOICE_SERVER_TYPE; mVoiceVersion.internalVoiceServerType = WEBRTC_VOICE_SERVER_TYPE; mVoiceVersion.minorVersion = 0; mVoiceVersion.majorVersion = 2; mVoiceVersion.mBuildVersion = ""; } //--------------------------------------------------- LLWebRTCVoiceClient::~LLWebRTCVoiceClient() { } void LLWebRTCVoiceClient::cleanupSingleton() { if (mAvatarNameCacheConnection.connected()) { mAvatarNameCacheConnection.disconnect(); } sShuttingDown = true; if (mSession) { mSession->shutdownAllConnections(); } if (mNextSession) { mNextSession->shutdownAllConnections(); } cleanUp(); sessionState::clearSessions(); } //--------------------------------------------------- void LLWebRTCVoiceClient::init(LLPumpIO* pump) { // constructor will set up LLVoiceClient::getInstance() llwebrtc::init(this); mWebRTCDeviceInterface = llwebrtc::getDeviceInterface(); mWebRTCDeviceInterface->setDevicesObserver(this); mMainQueue = LL::WorkQueue::getInstance("mainloop"); refreshDeviceLists(); } void LLWebRTCVoiceClient::terminate() { if (sShuttingDown) { return; } mVoiceEnabled = false; llwebrtc::terminate(); sShuttingDown = true; } //--------------------------------------------------- void LLWebRTCVoiceClient::cleanUp() { mNextSession.reset(); mSession.reset(); mNeighboringRegions.clear(); sessionState::for_each(boost::bind(predShutdownSession, _1)); LL_DEBUGS("Voice") << "Exiting" << LL_ENDL; } void LLWebRTCVoiceClient::LogMessage(llwebrtc::LLWebRTCLogCallback::LogLevel level, const std::string& message) { switch (level) { case llwebrtc::LLWebRTCLogCallback::LOG_LEVEL_VERBOSE: LL_DEBUGS("Voice") << message << LL_ENDL; break; case llwebrtc::LLWebRTCLogCallback::LOG_LEVEL_INFO: LL_INFOS("Voice") << message << LL_ENDL; break; case llwebrtc::LLWebRTCLogCallback::LOG_LEVEL_WARNING: LL_WARNS("Voice") << message << LL_ENDL; break; case llwebrtc::LLWebRTCLogCallback::LOG_LEVEL_ERROR: // use WARN so that we don't crash on a webrtc error. // webrtc will force a crash on a fatal error. LL_WARNS("Voice") << message << LL_ENDL; break; default: break; } } // -------------------------------------------------- const LLVoiceVersionInfo& LLWebRTCVoiceClient::getVersion() { return mVoiceVersion; } //--------------------------------------------------- void LLWebRTCVoiceClient::updateSettings() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; setVoiceEnabled(LLVoiceClient::getInstance()->voiceEnabled()); static LLCachedControl sVoiceEarLocation(gSavedSettings, "VoiceEarLocation"); setEarLocation(sVoiceEarLocation); static LLCachedControl sInputDevice(gSavedSettings, "VoiceInputAudioDevice"); setCaptureDevice(sInputDevice); static LLCachedControl sOutputDevice(gSavedSettings, "VoiceOutputAudioDevice"); setRenderDevice(sOutputDevice); static LLCachedControl sMicLevel(gSavedSettings, "AudioLevelMic"); setMicGain(sMicLevel); llwebrtc::LLWebRTCDeviceInterface::AudioConfig config; static LLCachedControl sEchoCancellation(gSavedSettings, "VoiceEchoCancellation", true); config.mEchoCancellation = sEchoCancellation; static LLCachedControl sAGC(gSavedSettings, "VoiceAutomaticGainControl", true); config.mAGC = sAGC; static LLCachedControl sNoiseSuppressionLevel(gSavedSettings, "VoiceNoiseSuppressionLevel", llwebrtc::LLWebRTCDeviceInterface::AudioConfig::ENoiseSuppressionLevel::NOISE_SUPPRESSION_LEVEL_VERY_HIGH); config.mNoiseSuppressionLevel = (llwebrtc::LLWebRTCDeviceInterface::AudioConfig::ENoiseSuppressionLevel) (U32)sNoiseSuppressionLevel; mWebRTCDeviceInterface->setAudioConfig(config); } // Observers void LLWebRTCVoiceClient::addObserver(LLVoiceClientParticipantObserver *observer) { mParticipantObservers.insert(observer); } void LLWebRTCVoiceClient::removeObserver(LLVoiceClientParticipantObserver *observer) { mParticipantObservers.erase(observer); } void LLWebRTCVoiceClient::notifyParticipantObservers() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; for (observer_set_t::iterator it = mParticipantObservers.begin(); it != mParticipantObservers.end();) { LLVoiceClientParticipantObserver *observer = *it; observer->onParticipantsChanged(); // In case onParticipantsChanged() deleted an entry. it = mParticipantObservers.upper_bound(observer); } } void LLWebRTCVoiceClient::addObserver(LLVoiceClientStatusObserver *observer) { mStatusObservers.insert(observer); } void LLWebRTCVoiceClient::removeObserver(LLVoiceClientStatusObserver *observer) { mStatusObservers.erase(observer); } void LLWebRTCVoiceClient::notifyStatusObservers(LLVoiceClientStatusObserver::EStatusType status) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LL_DEBUGS("Voice") << "( " << LLVoiceClientStatusObserver::status2string(status) << " )" << " mSession=" << mSession << LL_ENDL; LL_DEBUGS("Voice") << " " << LLVoiceClientStatusObserver::status2string(status) << ", session channelInfo " << getAudioSessionChannelInfo() << ", proximal is " << inSpatialChannel() << LL_ENDL; mIsProcessingChannels = status == LLVoiceClientStatusObserver::STATUS_JOINED; LLSD channelInfo = getAudioSessionChannelInfo(); for (status_observer_set_t::iterator it = mStatusObservers.begin(); it != mStatusObservers.end();) { LLVoiceClientStatusObserver *observer = *it; observer->onChange(status, channelInfo, inSpatialChannel()); // In case onError() deleted an entry. it = mStatusObservers.upper_bound(observer); } // skipped to avoid speak button blinking if (status != LLVoiceClientStatusObserver::STATUS_JOINING && status != LLVoiceClientStatusObserver::STATUS_LEFT_CHANNEL && status != LLVoiceClientStatusObserver::STATUS_VOICE_DISABLED) { bool voice_status = LLVoiceClient::getInstance()->voiceEnabled() && mIsProcessingChannels; gAgent.setVoiceConnected(voice_status); if (voice_status) { LLAppViewer::instance()->postToMainCoro([=]() { LLFirstUse::speak(true); }); } } } void LLWebRTCVoiceClient::addObserver(LLFriendObserver *observer) { } void LLWebRTCVoiceClient::removeObserver(LLFriendObserver *observer) { } //--------------------------------------------------- // Primary voice loop. // This voice loop is called every 100ms plus the time it // takes to process the various functions called in the loop // The loop does the following: // * gates whether we do channel processing depending on // whether we're running a WebRTC voice channel or // one from another voice provider. // * If in spatial voice, it determines whether we've changed // parcels, whether region/parcel voice settings have changed, // etc. and manages whether the voice channel needs to change. // * calls the state machines for the sessions to negotiate // connection to various voice channels. // * Sends updates to the voice server when this agent's // voice levels, or positions have changed. void LLWebRTCVoiceClient::voiceConnectionCoro() { LL_DEBUGS("Voice") << "starting" << LL_ENDL; mIsCoroutineActive = true; LLCoros::set_consuming(true); try { LLMuteList::getInstance()->addObserver(this); while (!sShuttingDown) { LL_PROFILE_ZONE_NAMED_CATEGORY_VOICE("voiceConnectionCoroLoop") // TODO: Doing some measurement and calculation here, // we could reduce the timeout to take into account the // time spent on the previous loop to have the loop // cycle at exactly 100ms, instead of 100ms + loop // execution time. // Could help with voice updates making for smoother // voice when we're busy. llcoro::suspendUntilTimeout(UPDATE_THROTTLE_SECONDS); if (sShuttingDown) return; // 'this' migh already be invalid bool voiceEnabled = mVoiceEnabled; if (!isAgentAvatarValid()) { continue; } LLViewerRegion *regionp = gAgent.getRegion(); if (!regionp) { continue; } if (!mProcessChannels) { // we've switched away from webrtc voice, so shut all channels down. // leave channel can be called again and again without adverse effects. // it merely tells channels to shut down if they're not already doing so. leaveChannel(false); } else if (inSpatialChannel()) { bool useEstateVoice = true; // add session for region or parcel voice. if (!regionp || regionp->getRegionID().isNull()) { // no region, no voice. continue; } voiceEnabled = voiceEnabled && regionp->isVoiceEnabled(); if (voiceEnabled) { LLParcel *parcel = LLViewerParcelMgr::getInstance()->getAgentParcel(); // check to see if parcel changed. if (parcel && parcel->getLocalID() != INVALID_PARCEL_ID) { // parcel voice if (!parcel->getParcelFlagAllowVoice()) { voiceEnabled = false; } else if (!parcel->getParcelFlagUseEstateVoiceChannel()) { // use the parcel-specific voice channel. S32 parcel_local_id = parcel->getLocalID(); std::string channelID = regionp->getRegionID().asString() + "-" + std::to_string(parcel->getLocalID()); useEstateVoice = false; if (!inOrJoiningChannel(channelID)) { startParcelSession(channelID, parcel_local_id); } } } if (voiceEnabled && useEstateVoice && !inEstateChannel()) { // estate voice startEstateSession(); } } if (!voiceEnabled) { // voice is disabled, so leave and disable PTT leaveChannel(true); } else { // we're in spatial voice, and voice is enabled, so determine positions in order // to send position updates. updatePosition(); } } LL::WorkQueue::postMaybe(mMainQueue, [=] { if (sShuttingDown) { return; } sessionState::processSessionStates(); if (mProcessChannels && voiceEnabled && !mHidden) { sendPositionUpdate(false); updateOwnVolume(); } }); } } catch (const LLCoros::Stop&) { LL_DEBUGS("LLWebRTCVoiceClient") << "Received a shutdown exception" << LL_ENDL; } catch (const LLContinueError&) { LOG_UNHANDLED_EXCEPTION("LLWebRTCVoiceClient"); } catch (...) { // Ideally for Windows need to log SEH exception instead or to set SEH // handlers but bugsplat shows local variables for windows, which should // be enough LL_WARNS("Voice") << "voiceConnectionStateMachine crashed" << LL_ENDL; throw; } cleanUp(); } // For spatial, determine which neighboring regions to connect to // for cross-region voice. void LLWebRTCVoiceClient::updateNeighboringRegions() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; static const std::vector neighbors {LLVector3d(0.0f, 1.0f, 0.0f), LLVector3d(0.707f, 0.707f, 0.0f), LLVector3d(1.0f, 0.0f, 0.0f), LLVector3d(0.707f, -0.707f, 0.0f), LLVector3d(0.0f, -1.0f, 0.0f), LLVector3d(-0.707f, -0.707f, 0.0f), LLVector3d(-1.0f, 0.0f, 0.0f), LLVector3d(-0.707f, 0.707f, 0.0f)}; // Estate voice requires connection to neighboring regions. mNeighboringRegions.clear(); // add current region. mNeighboringRegions.insert(gAgent.getRegion()->getRegionID()); // base off of speaker position as it'll move more slowly than camera position. // Once we have hysteresis, we may be able to track off of speaker and camera position at 50m // TODO: Add hysteresis so we don't flip-flop connections to neighbors LLVector3d speaker_pos = LLWebRTCVoiceClient::getInstance()->getSpeakerPosition(); for (auto &neighbor_pos : neighbors) { // include every region within 100m (2*MAX_AUDIO_DIST) to deal witht he fact that the camera // can stray 50m away from the avatar. LLViewerRegion *neighbor = LLWorld::instance().getRegionFromPosGlobal(speaker_pos + 2 * MAX_AUDIO_DIST * neighbor_pos); if (neighbor && !neighbor->getRegionID().isNull()) { mNeighboringRegions.insert(neighbor->getRegionID()); } } } //========================================================================= // shut down the current audio session to make room for the next one. void LLWebRTCVoiceClient::leaveAudioSession() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if(mSession) { LL_DEBUGS("Voice") << "leaving session: " << mSession->mChannelID << LL_ENDL; mSession->shutdownAllConnections(); } else { LL_WARNS("Voice") << "called with no active session" << LL_ENDL; } } //========================================================================= // Device Management void LLWebRTCVoiceClient::clearCaptureDevices() { LL_DEBUGS("Voice") << "called" << LL_ENDL; mCaptureDevices.clear(); } void LLWebRTCVoiceClient::addCaptureDevice(const LLVoiceDevice& device) { LL_INFOS("Voice") << "Voice Capture Device: '" << device.display_name << "' (" << device.full_name << ")" << LL_ENDL; mCaptureDevices.push_back(device); } LLVoiceDeviceList& LLWebRTCVoiceClient::getCaptureDevices() { return mCaptureDevices; } void LLWebRTCVoiceClient::setCaptureDevice(const std::string& name) { mWebRTCDeviceInterface->setCaptureDevice(name); } void LLWebRTCVoiceClient::setDevicesListUpdated(bool state) { mDevicesListUpdated = state; } // the singleton 'this' pointer will outlive the work queue. void LLWebRTCVoiceClient::OnDevicesChanged(const llwebrtc::LLWebRTCVoiceDeviceList& render_devices, const llwebrtc::LLWebRTCVoiceDeviceList& capture_devices) { LL::WorkQueue::postMaybe(mMainQueue, [=] { OnDevicesChangedImpl(render_devices, capture_devices); }); } void LLWebRTCVoiceClient::OnDevicesChangedImpl(const llwebrtc::LLWebRTCVoiceDeviceList &render_devices, const llwebrtc::LLWebRTCVoiceDeviceList &capture_devices) { if (sShuttingDown) { return; } LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; std::string inputDevice = gSavedSettings.getString("VoiceInputAudioDevice"); std::string outputDevice = gSavedSettings.getString("VoiceOutputAudioDevice"); LL_DEBUGS("Voice") << "Setting devices to-input: '" << inputDevice << "' output: '" << outputDevice << "'" << LL_ENDL; clearRenderDevices(); for (auto &device : render_devices) { addRenderDevice(LLVoiceDevice(device.mDisplayName, device.mID)); } setRenderDevice(outputDevice); clearCaptureDevices(); for (auto &device : capture_devices) { LL_DEBUGS("Voice") << "Checking capture device:'" << device.mID << "'" << LL_ENDL; addCaptureDevice(LLVoiceDevice(device.mDisplayName, device.mID)); } setCaptureDevice(inputDevice); setDevicesListUpdated(true); } void LLWebRTCVoiceClient::clearRenderDevices() { LL_DEBUGS("Voice") << "called" << LL_ENDL; mRenderDevices.clear(); } void LLWebRTCVoiceClient::addRenderDevice(const LLVoiceDevice& device) { LL_INFOS("Voice") << "Voice Render Device: '" << device.display_name << "' (" << device.full_name << ")" << LL_ENDL; mRenderDevices.push_back(device); } LLVoiceDeviceList& LLWebRTCVoiceClient::getRenderDevices() { return mRenderDevices; } void LLWebRTCVoiceClient::setRenderDevice(const std::string& name) { mWebRTCDeviceInterface->setRenderDevice(name); } void LLWebRTCVoiceClient::tuningStart() { if (!mIsInTuningMode) { mWebRTCDeviceInterface->setTuningMode(true); mIsInTuningMode = true; } } void LLWebRTCVoiceClient::tuningStop() { if (mIsInTuningMode) { mWebRTCDeviceInterface->setTuningMode(false); mIsInTuningMode = false; } } bool LLWebRTCVoiceClient::inTuningMode() { return mIsInTuningMode; } void LLWebRTCVoiceClient::tuningSetMicVolume(float volume) { mTuningMicGain = volume; } void LLWebRTCVoiceClient::tuningSetSpeakerVolume(float volume) { if (volume != mTuningSpeakerVolume) { mTuningSpeakerVolume = (int)volume; } } float LLWebRTCVoiceClient::getAudioLevel() { if (mIsInTuningMode) { return (1.0f - mWebRTCDeviceInterface->getTuningAudioLevel() * LEVEL_SCALE_WEBRTC) * mTuningMicGain / 2.1f; } else { return (1.0f - mWebRTCDeviceInterface->getPeerConnectionAudioLevel() * LEVEL_SCALE_WEBRTC) * mMicGain / 2.1f; } } float LLWebRTCVoiceClient::tuningGetEnergy(void) { return getAudioLevel(); } bool LLWebRTCVoiceClient::deviceSettingsAvailable() { bool result = true; if(mRenderDevices.empty() || mCaptureDevices.empty()) result = false; return result; } bool LLWebRTCVoiceClient::deviceSettingsUpdated() { bool updated = mDevicesListUpdated; mDevicesListUpdated = false; return updated; } void LLWebRTCVoiceClient::refreshDeviceLists(bool clearCurrentList) { if(clearCurrentList) { clearCaptureDevices(); clearRenderDevices(); } mWebRTCDeviceInterface->refreshDevices(); } void LLWebRTCVoiceClient::setHidden(bool hidden) { mHidden = hidden; if (inSpatialChannel()) { if (mHidden) { // get out of the channel entirely // mute the microphone. sessionState::for_each(boost::bind(predSetMuteMic, _1, true)); } else { // and put it back sessionState::for_each(boost::bind(predSetMuteMic, _1, mMuteMic)); updatePosition(); sendPositionUpdate(true); } } } ///////////////////////////// // session control messages. // // these are called by the sessions to report // status for a given channel. By filtering // on channel and region, these functions // can send various notifications to // other parts of the viewer, as well as // managing housekeeping // A connection to a channel was successfully established, // so shut down the current session and move on to the next // if one is available. // if the current session is the one that was established, // notify the observers. void LLWebRTCVoiceClient::OnConnectionEstablished(const std::string &channelID, const LLUUID ®ionID) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (gAgent.getRegion()->getRegionID() == regionID) { if (mNextSession && mNextSession->mChannelID == channelID) { if (mSession) { mSession->shutdownAllConnections(); } mSession = mNextSession; mNextSession.reset(); } if (mSession) { // Add ourselves as a participant. mSession->addParticipant(gAgentID, gAgent.getRegion()->getRegionID()); } // The current session was established. if (mSession && mSession->mChannelID == channelID) { LLWebRTCVoiceClient::getInstance()->notifyStatusObservers(LLVoiceClientStatusObserver::STATUS_LOGGED_IN); // only set status to joined if asked to. This will happen in the case where we're not // doing an ad-hoc based p2p session. Those sessions expect a STATUS_JOINED when the peer // has, in fact, joined, which we detect elsewhere. if (!mSession->mNotifyOnFirstJoin) { LLWebRTCVoiceClient::getInstance()->notifyStatusObservers(LLVoiceClientStatusObserver::STATUS_JOINED); } } } } void LLWebRTCVoiceClient::OnConnectionShutDown(const std::string &channelID, const LLUUID ®ionID) { if (mSession && (mSession->mChannelID == channelID)) { if (gAgent.getRegion()->getRegionID() == regionID) { if (mSession && mSession->mChannelID == channelID) { LL_DEBUGS("Voice") << "Main WebRTC Connection Shut Down." << LL_ENDL; } } mSession->removeAllParticipants(regionID); } } void LLWebRTCVoiceClient::OnConnectionFailure(const std::string &channelID, const LLUUID ®ionID, LLVoiceClientStatusObserver::EStatusType status_type) { LL_DEBUGS("Voice") << "A connection failed. channel:" << channelID << LL_ENDL; if (gAgent.getRegion()->getRegionID() == regionID) { if (mNextSession && mNextSession->mChannelID == channelID) { LLWebRTCVoiceClient::getInstance()->notifyStatusObservers(status_type); } else if (mSession && mSession->mChannelID == channelID) { LLWebRTCVoiceClient::getInstance()->notifyStatusObservers(status_type); } } } // ----------------------------------------------------------- // positional functionality. void LLWebRTCVoiceClient::setEarLocation(S32 loc) { if (mEarLocation != loc) { LL_DEBUGS("Voice") << "Setting mEarLocation to " << loc << LL_ENDL; mEarLocation = loc; mSpatialCoordsDirty = true; } } void LLWebRTCVoiceClient::updatePosition(void) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LLViewerRegion *region = gAgent.getRegion(); if (region && isAgentAvatarValid()) { // get the avatar position. LLVector3d avatar_pos = gAgentAvatarp->getPositionGlobal(); LLQuaternion avatar_qrot = gAgentAvatarp->getRootJoint()->getWorldRotation(); avatar_pos += LLVector3d(0.f, 0.f, 1.f); // bump it up to head height LLVector3d earPosition; LLQuaternion earRot; switch (mEarLocation) { case earLocCamera: default: earPosition = region->getPosGlobalFromRegion(LLViewerCamera::getInstance()->getOrigin()); earRot = LLViewerCamera::getInstance()->getQuaternion(); break; case earLocAvatar: earPosition = mAvatarPosition; earRot = mAvatarRot; break; case earLocMixed: earPosition = mAvatarPosition; earRot = LLViewerCamera::getInstance()->getQuaternion(); break; } setListenerPosition(earPosition, // position LLVector3::zero, // velocity earRot); // rotation matrix setAvatarPosition(avatar_pos, // position LLVector3::zero, // velocity avatar_qrot); // rotation matrix enforceTether(); updateNeighboringRegions(); // update own region id to be the region id avatar is currently in. LLWebRTCVoiceClient::participantStatePtr_t participant = findParticipantByID("Estate", gAgentID); if(participant) { participant->mRegion = gAgent.getRegion()->getRegionID(); } } } void LLWebRTCVoiceClient::setListenerPosition(const LLVector3d &position, const LLVector3 &velocity, const LLQuaternion &rot) { mListenerRequestedPosition = position; if (mListenerVelocity != velocity) { mListenerVelocity = velocity; mSpatialCoordsDirty = true; } if (mListenerRot != rot) { mListenerRot = rot; mSpatialCoordsDirty = true; } } void LLWebRTCVoiceClient::setAvatarPosition(const LLVector3d &position, const LLVector3 &velocity, const LLQuaternion &rot) { if (dist_vec_squared(mAvatarPosition, position) > 0.01) { mAvatarPosition = position; mSpatialCoordsDirty = true; } if (mAvatarVelocity != velocity) { mAvatarVelocity = velocity; mSpatialCoordsDirty = true; } // If the two rotations are not exactly equal test their dot product // to get the cos of the angle between them. // If it is too small, don't update. F32 rot_cos_diff = llabs(dot(mAvatarRot, rot)); if ((mAvatarRot != rot) && (rot_cos_diff < MINUSCULE_ANGLE_COS)) { mAvatarRot = rot; mSpatialCoordsDirty = true; } } // The listener (camera) must be within 50m of the // avatar. Enforce it on the client. // This will also be enforced on the voice server // based on position sent from the simulator to the // voice server. void LLWebRTCVoiceClient::enforceTether() { LLVector3d tethered = mListenerRequestedPosition; // constrain 'tethered' to within 50m of mAvatarPosition. { LLVector3d camera_offset = mListenerRequestedPosition - mAvatarPosition; F32 camera_distance = (F32) camera_offset.magVec(); if (camera_distance > MAX_AUDIO_DIST) { tethered = mAvatarPosition + (MAX_AUDIO_DIST / camera_distance) * camera_offset; } } if (dist_vec_squared(mListenerPosition, tethered) > 0.01) { mListenerPosition = tethered; mSpatialCoordsDirty = true; } } // We send our position via a WebRTC data channel to the WebRTC // server for fine-grained, low latency updates. On the server, // these updates will be 'tethered' to the actual position of the avatar. // Those updates are higher latency, however. // This mechanism gives low latency spatial updates and server-enforced // prevention of 'evesdropping' by sending camera updates beyond the // standard 50m void LLWebRTCVoiceClient::sendPositionUpdate(bool force) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; std::string spatial_data; if (mSpatialCoordsDirty || force) { boost::json::object spatial; spatial["sp"] = { {"x", (int) (mAvatarPosition[0] * 100)}, {"y", (int) (mAvatarPosition[1] * 100)}, {"z", (int) (mAvatarPosition[2] * 100)} }; spatial["sh"] = { {"x", (int) (mAvatarRot[0] * 100)}, {"y", (int) (mAvatarRot[1] * 100)}, {"z", (int) (mAvatarRot[2] * 100)}, {"w", (int) (mAvatarRot[3] * 100)} }; spatial["lp"] = { {"x", (int) (mListenerPosition[0] * 100)}, {"y", (int) (mListenerPosition[1] * 100)}, {"z", (int) (mListenerPosition[2] * 100)} }; spatial["lh"] = { {"x", (int) (mListenerRot[0] * 100)}, {"y", (int) (mListenerRot[1] * 100)}, {"z", (int) (mListenerRot[2] * 100)}, {"w", (int) (mListenerRot[3] * 100)}}; mSpatialCoordsDirty = false; spatial_data = boost::json::serialize(spatial); sessionState::for_each(boost::bind(predSendData, _1, spatial_data)); } } // Update our own volume on our participant, so it'll show up // in the UI. This is done on all sessions, so switching // sessions retains consistent volume levels. void LLWebRTCVoiceClient::updateOwnVolume() { F32 audio_level = 0.0; if (!mMuteMic && !mTuningMode) { audio_level = getAudioLevel(); } sessionState::for_each(boost::bind(predUpdateOwnVolume, _1, audio_level)); } //////////////////////////////////// // Managing list of participants // Provider-level participant management bool LLWebRTCVoiceClient::isParticipantAvatar(const LLUUID &id) { // WebRTC participants are always SL avatars. return true; } void LLWebRTCVoiceClient::getParticipantList(std::set &participants) { if (mProcessChannels && mSession) { for (participantUUIDMap::iterator iter = mSession->mParticipantsByUUID.begin(); iter != mSession->mParticipantsByUUID.end(); iter++) { participants.insert(iter->first); } } } bool LLWebRTCVoiceClient::isParticipant(const LLUUID &speaker_id) { if (mProcessChannels && mSession) { return (mSession->mParticipantsByUUID.find(speaker_id) != mSession->mParticipantsByUUID.end()); } return false; } // protected provider-level participant management. LLWebRTCVoiceClient::participantStatePtr_t LLWebRTCVoiceClient::findParticipantByID(const std::string &channelID, const LLUUID &id) { participantStatePtr_t result; LLWebRTCVoiceClient::sessionState::ptr_t session = sessionState::matchSessionByChannelID(channelID); if (session) { result = session->findParticipantByID(id); } return result; } LLWebRTCVoiceClient::participantStatePtr_t LLWebRTCVoiceClient::addParticipantByID(const std::string &channelID, const LLUUID &id, const LLUUID& region) { participantStatePtr_t result; LLWebRTCVoiceClient::sessionState::ptr_t session = sessionState::matchSessionByChannelID(channelID); if (session) { result = session->addParticipant(id, region); if (session->mNotifyOnFirstJoin && (id != gAgentID)) { notifyStatusObservers(LLVoiceClientStatusObserver::STATUS_JOINED); } } return result; } void LLWebRTCVoiceClient::removeParticipantByID(const std::string &channelID, const LLUUID &id, const LLUUID& region) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; participantStatePtr_t result; LLWebRTCVoiceClient::sessionState::ptr_t session = sessionState::matchSessionByChannelID(channelID); if (session) { participantStatePtr_t participant = session->findParticipantByID(id); if (participant && (participant->mRegion == region)) { session->removeParticipant(participant); } } } // participantState level participant management LLWebRTCVoiceClient::participantState::participantState(const LLUUID& agent_id, const LLUUID& region) : mURI(agent_id.asString()), mAvatarID(agent_id), mIsSpeaking(false), mIsModeratorMuted(false), mLevel(0.f), mVolume(LLVoiceClient::VOLUME_DEFAULT), mRegion(region) { } LLWebRTCVoiceClient::participantStatePtr_t LLWebRTCVoiceClient::sessionState::addParticipant(const LLUUID& agent_id, const LLUUID& region) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; participantStatePtr_t result; participantUUIDMap::iterator iter = mParticipantsByUUID.find(agent_id); if (iter != mParticipantsByUUID.end()) { result = iter->second; result->mRegion = region; } if (!result) { // participant isn't already in one list or the other. result.reset(new participantState(agent_id, region)); mParticipantsByUUID.insert(participantUUIDMap::value_type(agent_id, result)); result->mAvatarID = agent_id; } LLWebRTCVoiceClient::getInstance()->lookupName(agent_id); LLSpeakerVolumeStorage::getInstance()->getSpeakerVolume(result->mAvatarID, result->mVolume); if (!LLWebRTCVoiceClient::sShuttingDown) { LLWebRTCVoiceClient::getInstance()->notifyParticipantObservers(); } LL_DEBUGS("Voice") << "Participant \"" << result->mURI << "\" added." << LL_ENDL; return result; } // session-level participant management LLWebRTCVoiceClient::participantStatePtr_t LLWebRTCVoiceClient::sessionState::findParticipantByID(const LLUUID& id) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; participantStatePtr_t result; participantUUIDMap::iterator iter = mParticipantsByUUID.find(id); if(iter != mParticipantsByUUID.end()) { result = iter->second; } return result; } void LLWebRTCVoiceClient::sessionState::removeParticipant(const LLWebRTCVoiceClient::participantStatePtr_t &participant) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (participant) { LLUUID participantID = participant->mAvatarID; participantUUIDMap::iterator iter = mParticipantsByUUID.find(participant->mAvatarID); LL_DEBUGS("Voice") << "participant \"" << participant->mURI << "\" (" << participantID << ") removed." << LL_ENDL; if (iter == mParticipantsByUUID.end()) { LL_WARNS("Voice") << "Internal error: participant ID " << participantID << " not in UUID map" << LL_ENDL; } else { mParticipantsByUUID.erase(iter); if (!LLWebRTCVoiceClient::sShuttingDown) { LLWebRTCVoiceClient::getInstance()->notifyParticipantObservers(); } } if (mHangupOnLastLeave && (participantID != gAgentID) && (mParticipantsByUUID.size() <= 1) && LLWebRTCVoiceClient::instanceExists()) { LLWebRTCVoiceClient::getInstance()->notifyStatusObservers(LLVoiceClientStatusObserver::STATUS_LEFT_CHANNEL); } } } void LLWebRTCVoiceClient::sessionState::removeAllParticipants(const LLUUID ®ion) { std::vector participantsToRemove; for (auto& participantEntry : mParticipantsByUUID) { if (region.isNull() || (participantEntry.second->mRegion == region)) { participantsToRemove.push_back(participantEntry.second); } } for (auto& participant : participantsToRemove) { removeParticipant(participant); } } // Initiated the various types of sessions. bool LLWebRTCVoiceClient::startEstateSession() { leaveChannel(false); mNextSession = addSession("Estate", sessionState::ptr_t(new estateSessionState())); return true; } bool LLWebRTCVoiceClient::startParcelSession(const std::string &channelID, S32 parcelID) { leaveChannel(false); mNextSession = addSession(channelID, sessionState::ptr_t(new parcelSessionState(channelID, parcelID))); return true; } bool LLWebRTCVoiceClient::startAdHocSession(const LLSD& channelInfo, bool notify_on_first_join, bool hangup_on_last_leave) { leaveChannel(false); LL_WARNS("Voice") << "Start AdHoc Session " << channelInfo << LL_ENDL; std::string channelID = channelInfo["channel_uri"]; std::string credentials = channelInfo["channel_credentials"]; mNextSession = addSession(channelID, sessionState::ptr_t(new adhocSessionState(channelID, credentials, notify_on_first_join, hangup_on_last_leave))); return true; } bool LLWebRTCVoiceClient::isVoiceWorking() const { // webrtc is working if the coroutine is active in the case of // webrtc. WebRTC doesn't need to connect to a secondary process // or a login server to become active. return mIsCoroutineActive; } // Returns true if calling back the session URI after the session has closed is possible. // Currently this will be false only for PSTN P2P calls. bool LLWebRTCVoiceClient::isSessionCallBackPossible(const LLUUID &session_id) { sessionStatePtr_t session(findP2PSession(session_id)); return session && session->isCallbackPossible(); } // Channel Management bool LLWebRTCVoiceClient::setSpatialChannel(const LLSD &channelInfo) { LL_INFOS("Voice") << "SetSpatialChannel " << channelInfo << LL_ENDL; LLViewerRegion *regionp = gAgent.getRegion(); if (!regionp) { return false; } LLParcel *parcel = LLViewerParcelMgr::getInstance()->getAgentParcel(); // we don't really have credentials for a spatial channel in webrtc, // it's all handled by the sim. if (channelInfo.isMap() && channelInfo.has("channel_uri")) { bool allow_voice = !channelInfo["channel_uri"].asString().empty(); if (parcel) { parcel->setParcelFlag(PF_ALLOW_VOICE_CHAT, allow_voice); parcel->setParcelFlag(PF_USE_ESTATE_VOICE_CHAN, channelInfo["channel_uri"].asUUID() == regionp->getRegionID()); } else { regionp->setRegionFlag(REGION_FLAGS_ALLOW_VOICE, allow_voice); } } return true; } void LLWebRTCVoiceClient::leaveNonSpatialChannel() { LL_DEBUGS("Voice") << "Request to leave non-spatial channel." << LL_ENDL; // make sure we're not simply rejoining the current session deleteSession(mNextSession); leaveChannel(true); } // determine whether we're processing channels, or whether // another voice provider is. void LLWebRTCVoiceClient::processChannels(bool process) { mProcessChannels = process; } bool LLWebRTCVoiceClient::inProximalChannel() { return inSpatialChannel(); } bool LLWebRTCVoiceClient::inOrJoiningChannel(const std::string& channelID) { return (mSession && mSession->mChannelID == channelID) || (mNextSession && mNextSession->mChannelID == channelID); } bool LLWebRTCVoiceClient::inEstateChannel() { return (mSession && mSession->isEstate()) || (mNextSession && mNextSession->isEstate()); } bool LLWebRTCVoiceClient::inSpatialChannel() { bool result = true; if (mNextSession) { result = mNextSession->isSpatial(); } else if(mSession) { result = mSession->isSpatial(); } return result; } // retrieves information used to negotiate p2p, adhoc, and group // channels LLSD LLWebRTCVoiceClient::getAudioSessionChannelInfo() { LLSD result; if (mSession) { result["voice_server_type"] = WEBRTC_VOICE_SERVER_TYPE; result["channel_uri"] = mSession->mChannelID; } return result; } void LLWebRTCVoiceClient::leaveChannel(bool stopTalking) { if (mSession) { deleteSession(mSession); } if (mNextSession) { deleteSession(mNextSession); } // If voice was on, turn it off if (stopTalking && LLVoiceClient::getInstance()->getUserPTTState()) { LLVoiceClient::getInstance()->setUserPTTState(false); } } bool LLWebRTCVoiceClient::isCurrentChannel(const LLSD &channelInfo) { if (!mProcessChannels || (channelInfo["voice_server_type"].asString() != WEBRTC_VOICE_SERVER_TYPE)) { return false; } sessionStatePtr_t session = mSession; if (!session) { session = mNextSession; } if (session) { if (!channelInfo["session_handle"].asString().empty()) { return session->mHandle == channelInfo["session_handle"].asString(); } return channelInfo["channel_uri"].asString() == session->mChannelID; } return false; } bool LLWebRTCVoiceClient::compareChannels(const LLSD &channelInfo1, const LLSD &channelInfo2) { return (channelInfo1["voice_server_type"] == WEBRTC_VOICE_SERVER_TYPE) && (channelInfo1["voice_server_type"] == channelInfo2["voice_server_type"]) && (channelInfo1["sip_uri"] == channelInfo2["sip_uri"]); } //---------------------------------------------- // Audio muting, volume, gain, etc. // we're muting the mic, so tell each session such void LLWebRTCVoiceClient::setMuteMic(bool muted) { mMuteMic = muted; // when you're hidden, your mic is always muted. if (!mHidden) { sessionState::for_each(boost::bind(predSetMuteMic, _1, muted)); } } void LLWebRTCVoiceClient::predSetMuteMic(const LLWebRTCVoiceClient::sessionStatePtr_t &session, bool muted) { participantStatePtr_t participant = session->findParticipantByID(gAgentID); if (participant) { participant->mLevel = 0.0; } session->setMuteMic(muted); } void LLWebRTCVoiceClient::setVoiceVolume(F32 volume) { if (volume != mSpeakerVolume) { { mSpeakerVolume = volume; } sessionState::for_each(boost::bind(predSetSpeakerVolume, _1, volume)); } } void LLWebRTCVoiceClient::predSetSpeakerVolume(const LLWebRTCVoiceClient::sessionStatePtr_t &session, F32 volume) { session->setSpeakerVolume(volume); } void LLWebRTCVoiceClient::setMicGain(F32 gain) { if (gain != mMicGain) { mMicGain = gain; mWebRTCDeviceInterface->setPeerConnectionGain(gain); } } void LLWebRTCVoiceClient::setVoiceEnabled(bool enabled) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LL_DEBUGS("Voice") << "( " << (enabled ? "enabled" : "disabled") << " )" << " was "<< (mVoiceEnabled ? "enabled" : "disabled") << " coro "<< (mIsCoroutineActive ? "active" : "inactive") << LL_ENDL; if (enabled != mVoiceEnabled) { // TODO: Refactor this so we don't call into LLVoiceChannel, but simply // use the status observer mVoiceEnabled = enabled; LLVoiceClientStatusObserver::EStatusType status; if (enabled) { LL_DEBUGS("Voice") << "enabling" << LL_ENDL; LLVoiceChannel::getCurrentVoiceChannel()->activate(); status = LLVoiceClientStatusObserver::STATUS_VOICE_ENABLED; mSpatialCoordsDirty = true; updatePosition(); if (!mIsCoroutineActive) { LLCoros::instance().launch("LLWebRTCVoiceClient::voiceConnectionCoro", boost::bind(&LLWebRTCVoiceClient::voiceConnectionCoro, LLWebRTCVoiceClient::getInstance())); } else { LL_DEBUGS("Voice") << "coro should be active.. not launching" << LL_ENDL; } } else { // Turning voice off looses your current channel -- this makes sure the UI isn't out of sync when you re-enable it. LLVoiceChannel::getCurrentVoiceChannel()->deactivate(); gAgent.setVoiceConnected(false); status = LLVoiceClientStatusObserver::STATUS_VOICE_DISABLED; cleanUp(); } notifyStatusObservers(status); } else { LL_DEBUGS("Voice") << " no-op" << LL_ENDL; } } ///////////////////////////// // Accessors for data related to nearby speakers std::string LLWebRTCVoiceClient::getDisplayName(const LLUUID& id) { std::string result; if (mProcessChannels && mSession) { participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant) { result = participant->mDisplayName; } } return result; } bool LLWebRTCVoiceClient::getIsSpeaking(const LLUUID& id) { bool result = false; if (mProcessChannels && mSession) { participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant) { result = participant->mIsSpeaking; } } return result; } // TODO: Need to pull muted status from the webrtc server bool LLWebRTCVoiceClient::getIsModeratorMuted(const LLUUID& id) { bool result = false; if (mProcessChannels && mSession) { participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant) { result = participant->mIsModeratorMuted; } } return result; } F32 LLWebRTCVoiceClient::getCurrentPower(const LLUUID &id) { F32 result = 0.0; if (!mProcessChannels || !mSession) { return result; } participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant) { if (participant->mIsSpeaking) { result = participant->mLevel; } } return result; } // External accessors. F32 LLWebRTCVoiceClient::getUserVolume(const LLUUID& id) { // Minimum volume will be returned for users with voice disabled F32 result = LLVoiceClient::VOLUME_MIN; if (mSession) { participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant) { result = participant->mVolume; } } return result; } void LLWebRTCVoiceClient::setUserVolume(const LLUUID& id, F32 volume) { F32 clamped_volume = llclamp(volume, LLVoiceClient::VOLUME_MIN, LLVoiceClient::VOLUME_MAX); if(mSession) { participantStatePtr_t participant(mSession->findParticipantByID(id)); if (participant && (participant->mAvatarID != gAgentID)) { if (!is_approx_equal(volume, LLVoiceClient::VOLUME_DEFAULT)) { // Store this volume setting for future sessions if it has been // changed from the default LLSpeakerVolumeStorage::getInstance()->storeSpeakerVolume(id, volume); } else { // Remove stored volume setting if it is returned to the default LLSpeakerVolumeStorage::getInstance()->removeSpeakerVolume(id); } participant->mVolume = clamped_volume; } } sessionState::for_each(boost::bind(predSetUserVolume, _1, id, clamped_volume)); } // set volume level (gain level) for another user. void LLWebRTCVoiceClient::predSetUserVolume(const LLWebRTCVoiceClient::sessionStatePtr_t &session, const LLUUID &id, F32 volume) { session->setUserVolume(id, volume); } //////////////////////// ///LLMuteListObserver /// void LLWebRTCVoiceClient::onChange() { } void LLWebRTCVoiceClient::onChangeDetailed(const LLMute& mute) { if (mute.mType == LLMute::AGENT) { bool muted = ((mute.mFlags & LLMute::flagVoiceChat) == 0); sessionState::for_each(boost::bind(predSetUserMute, _1, mute.mID, muted)); } } void LLWebRTCVoiceClient::predSetUserMute(const LLWebRTCVoiceClient::sessionStatePtr_t &session, const LLUUID &id, bool mute) { session->setUserMute(id, mute); } //------------------------------------------------------------------------ // Sessions std::map LLWebRTCVoiceClient::sessionState::sSessions; LLWebRTCVoiceClient::sessionState::sessionState() : mHangupOnLastLeave(false), mNotifyOnFirstJoin(false), mMuted(false), mSpeakerVolume(1.0), mShuttingDown(false) { } // ------------------------------------------------------------------ // Predicates, for calls to all sessions void LLWebRTCVoiceClient::predUpdateOwnVolume(const LLWebRTCVoiceClient::sessionStatePtr_t &session, F32 audio_level) { participantStatePtr_t participant = session->findParticipantByID(gAgentID); if (participant) { participant->mLevel = audio_level; // TODO: Add VAD for our own voice. participant->mIsSpeaking = audio_level > SPEAKING_AUDIO_LEVEL; } } void LLWebRTCVoiceClient::predSendData(const LLWebRTCVoiceClient::sessionStatePtr_t &session, const std::string &spatial_data) { if (session->isSpatial() && !spatial_data.empty()) { session->sendData(spatial_data); } } void LLWebRTCVoiceClient::sessionState::sendData(const std::string &data) { for (auto &connection : mWebRTCConnections) { connection->sendData(data); } } void LLWebRTCVoiceClient::sessionState::setMuteMic(bool muted) { mMuted = muted; for (auto &connection : mWebRTCConnections) { connection->setMuteMic(muted); } } void LLWebRTCVoiceClient::sessionState::setSpeakerVolume(F32 volume) { mSpeakerVolume = volume; for (auto &connection : mWebRTCConnections) { connection->setSpeakerVolume(volume); } } void LLWebRTCVoiceClient::sessionState::setUserVolume(const LLUUID &id, F32 volume) { if (mParticipantsByUUID.find(id) == mParticipantsByUUID.end()) { return; } for (auto &connection : mWebRTCConnections) { connection->setUserVolume(id, volume); } } void LLWebRTCVoiceClient::sessionState::setUserMute(const LLUUID &id, bool mute) { if (mParticipantsByUUID.find(id) == mParticipantsByUUID.end()) { return; } for (auto &connection : mWebRTCConnections) { connection->setUserMute(id, mute); } } /*static*/ void LLWebRTCVoiceClient::sessionState::addSession( const std::string & channelID, LLWebRTCVoiceClient::sessionState::ptr_t& session) { sSessions[channelID] = session; } LLWebRTCVoiceClient::sessionState::~sessionState() { LL_DEBUGS("Voice") << "Destroying session CHANNEL=" << mChannelID << LL_ENDL; if (!mShuttingDown) { shutdownAllConnections(); } mWebRTCConnections.clear(); removeAllParticipants(); } /*static*/ LLWebRTCVoiceClient::sessionState::ptr_t LLWebRTCVoiceClient::sessionState::matchSessionByChannelID(const std::string& channel_id) { sessionStatePtr_t result; // *TODO: My kingdom for a lambda! std::map::iterator it = sSessions.find(channel_id); if (it != sSessions.end()) { result = (*it).second; } return result; } void LLWebRTCVoiceClient::sessionState::for_each(sessionFunc_t func) { std::for_each(sSessions.begin(), sSessions.end(), boost::bind(for_eachPredicate, _1, func)); } void LLWebRTCVoiceClient::sessionState::reapEmptySessions() { std::map::iterator iter; for (iter = sSessions.begin(); iter != sSessions.end();) { if (iter->second->isEmpty()) { iter = sSessions.erase(iter); } else { ++iter; } } } /*static*/ void LLWebRTCVoiceClient::sessionState::for_eachPredicate(const std::pair &a, sessionFunc_t func) { ptr_t aLock(a.second.lock()); if (aLock) func(aLock); else { LL_WARNS("Voice") << "Stale handle in session map!" << LL_ENDL; } } LLWebRTCVoiceClient::sessionStatePtr_t LLWebRTCVoiceClient::addSession(const std::string &channel_id, sessionState::ptr_t session) { sessionStatePtr_t existingSession = sessionState::matchSessionByChannelID(channel_id); if (!existingSession) { // No existing session found. LL_DEBUGS("Voice") << "adding new session with channel: " << channel_id << LL_ENDL; session->setMuteMic(mMuteMic); session->setSpeakerVolume(mSpeakerVolume); sessionState::addSession(channel_id, session); return session; } else { // Found an existing session LL_DEBUGS("Voice") << "Attempting to add already-existing session " << channel_id << LL_ENDL; existingSession->revive(); return existingSession; } } void LLWebRTCVoiceClient::sessionState::clearSessions() { sSessions.clear(); } LLWebRTCVoiceClient::sessionStatePtr_t LLWebRTCVoiceClient::findP2PSession(const LLUUID &agent_id) { sessionStatePtr_t result = sessionState::matchSessionByChannelID(agent_id.asString()); if (result && !result->isSpatial()) { return result; } result.reset(); return result; } void LLWebRTCVoiceClient::sessionState::shutdownAllConnections() { mShuttingDown = true; for (auto &&connection : mWebRTCConnections) { connection->shutDown(); } } // in case we drop into a session (spatial, etc.) right after // telling the session to shut down, revive it so it reconnects. void LLWebRTCVoiceClient::sessionState::revive() { mShuttingDown = false; } //========================================================================= // the following are methods to support the coroutine implementation of the // voice connection and processing. They should only be called in the context // of a coroutine. // // void LLWebRTCVoiceClient::sessionState::processSessionStates() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; auto iter = sSessions.begin(); while (iter != sSessions.end()) { if (!iter->second->processConnectionStates() && iter->second->mShuttingDown) { // if the connections associated with a session are gone, // and this session is shutting down, remove it. iter = sSessions.erase(iter); } else { iter++; } } } // process the states on each connection associated with a session. bool LLWebRTCVoiceClient::sessionState::processConnectionStates() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; std::list::iterator iter = mWebRTCConnections.begin(); while (iter != mWebRTCConnections.end()) { if (!iter->get()->connectionStateMachine()) { // if the state machine returns false, the connection is shut down // so delete it. iter = mWebRTCConnections.erase(iter); } else { ++iter; } } return !mWebRTCConnections.empty(); } // processing of spatial voice connection states requires special handling. // as neighboring regions need to be started up or shut down depending // on our location. bool LLWebRTCVoiceClient::estateSessionState::processConnectionStates() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (!mShuttingDown) { // Estate voice requires connection to neighboring regions. std::set neighbor_ids = LLWebRTCVoiceClient::getInstance()->getNeighboringRegions(); for (auto &connection : mWebRTCConnections) { std::shared_ptr spatialConnection = std::static_pointer_cast(connection); LLUUID regionID = spatialConnection.get()->getRegionID(); if (neighbor_ids.find(regionID) == neighbor_ids.end()) { // shut down connections to neighbors that are too far away. spatialConnection.get()->shutDown(); } if (!spatialConnection.get()->isShuttingDown()) { neighbor_ids.erase(regionID); } } // add new connections for new neighbors for (auto &neighbor : neighbor_ids) { connectionPtr_t connection(new LLVoiceWebRTCSpatialConnection(neighbor, INVALID_PARCEL_ID, mChannelID)); mWebRTCConnections.push_back(connection); connection->setMuteMic(mMuted); connection->setSpeakerVolume(mSpeakerVolume); } } return LLWebRTCVoiceClient::sessionState::processConnectionStates(); } // Various session state constructors. LLWebRTCVoiceClient::estateSessionState::estateSessionState() { mHangupOnLastLeave = false; mNotifyOnFirstJoin = false; mChannelID = "Estate"; LLUUID region_id = gAgent.getRegion()->getRegionID(); mWebRTCConnections.emplace_back(new LLVoiceWebRTCSpatialConnection(region_id, INVALID_PARCEL_ID, "Estate")); } LLWebRTCVoiceClient::parcelSessionState::parcelSessionState(const std::string &channelID, S32 parcel_local_id) { mHangupOnLastLeave = false; mNotifyOnFirstJoin = false; LLUUID region_id = gAgent.getRegion()->getRegionID(); mChannelID = channelID; mWebRTCConnections.emplace_back(new LLVoiceWebRTCSpatialConnection(region_id, parcel_local_id, channelID)); } LLWebRTCVoiceClient::adhocSessionState::adhocSessionState(const std::string &channelID, const std::string &credentials, bool notify_on_first_join, bool hangup_on_last_leave) : mCredentials(credentials) { mHangupOnLastLeave = hangup_on_last_leave; mNotifyOnFirstJoin = notify_on_first_join; LLUUID region_id = gAgent.getRegion()->getRegionID(); mChannelID = channelID; mWebRTCConnections.emplace_back(new LLVoiceWebRTCAdHocConnection(region_id, channelID, credentials)); } void LLWebRTCVoiceClient::predShutdownSession(const LLWebRTCVoiceClient::sessionStatePtr_t& session) { session->shutdownAllConnections(); } void LLWebRTCVoiceClient::deleteSession(const sessionStatePtr_t &session) { if (!session) { return; } // At this point, the session should be unhooked from all lists and all state should be consistent. session->shutdownAllConnections(); // If this is the current audio session, clean up the pointer which will soon be dangling. bool deleteAudioSession = mSession == session; bool deleteNextAudioSession = mNextSession == session; if (deleteAudioSession) { mSession.reset(); } // ditto for the next audio session if (deleteNextAudioSession) { mNextSession.reset(); } } // Name resolution void LLWebRTCVoiceClient::lookupName(const LLUUID &id) { if (mAvatarNameCacheConnection.connected()) { mAvatarNameCacheConnection.disconnect(); } mAvatarNameCacheConnection = LLAvatarNameCache::get(id, boost::bind(&LLWebRTCVoiceClient::onAvatarNameCache, this, _1, _2)); } void LLWebRTCVoiceClient::onAvatarNameCache(const LLUUID& agent_id, const LLAvatarName& av_name) { mAvatarNameCacheConnection.disconnect(); std::string display_name = av_name.getDisplayName(); avatarNameResolved(agent_id, display_name); } void LLWebRTCVoiceClient::predAvatarNameResolution(const LLWebRTCVoiceClient::sessionStatePtr_t &session, LLUUID id, std::string name) { participantStatePtr_t participant(session->findParticipantByID(id)); if (participant) { // Found -- fill in the name participant->mDisplayName = name; // and post a "participants updated" message to listeners later. LLWebRTCVoiceClient::getInstance()->notifyParticipantObservers(); } } void LLWebRTCVoiceClient::avatarNameResolved(const LLUUID &id, const std::string &name) { sessionState::for_each(boost::bind(predAvatarNameResolution, _1, id, name)); } // Leftover from vivox PTSN std::string LLWebRTCVoiceClient::sipURIFromID(const LLUUID& id) const { return id.asString(); } LLSD LLWebRTCVoiceClient::getP2PChannelInfoTemplate(const LLUUID& id) const { return LLSD(); } ///////////////////////////// // LLVoiceWebRTCConnection // These connections manage state transitions, negotiating webrtc connections, // and other such things for a single connection to a Secondlife WebRTC server. // Multiple of these connections may be active at once, in the case of // cross-region voice, or when a new connection is being created before the old // has a chance to shut down. LLVoiceWebRTCConnection::LLVoiceWebRTCConnection(const LLUUID ®ionID, const std::string &channelID) : mWebRTCAudioInterface(nullptr), mWebRTCDataInterface(nullptr), mVoiceConnectionState(VOICE_STATE_START_SESSION), mCurrentStatus(LLVoiceClientStatusObserver::STATUS_VOICE_ENABLED), mMuted(true), mShutDown(false), mIceCompleted(false), mSpeakerVolume(0.0), mOutstandingRequests(0), mChannelID(channelID), mRegionID(regionID), mPrimary(true), mRetryWaitPeriod(0) { // retries wait a short period...randomize it so // all clients don't try to reconnect at once. mRetryWaitSecs = (F32)((F32) rand() / (RAND_MAX)) + 0.5f; mWebRTCPeerConnectionInterface = llwebrtc::newPeerConnection(); mWebRTCPeerConnectionInterface->setSignalingObserver(this); mMainQueue = LL::WorkQueue::getInstance("mainloop"); } LLVoiceWebRTCConnection::~LLVoiceWebRTCConnection() { if (LLWebRTCVoiceClient::isShuttingDown()) { // peer connection and observers will be cleaned up // by llwebrtc::terminate() on shutdown. return; } mWebRTCPeerConnectionInterface->unsetSignalingObserver(this); llwebrtc::freePeerConnection(mWebRTCPeerConnectionInterface); } // ICE (Interactive Connectivity Establishment) // When WebRTC tries to negotiate a connection to the Secondlife WebRTC Server, // the negotiation will result in a few updates about the best path // to which to connect. // The Secondlife servers are configured for ICE trickling, where, after a session is partially // negotiated, updates about the best connectivity paths may trickle in. These need to be // sent to the Secondlife WebRTC server via the simulator so that both sides have a clear // view of the network environment. // callback from llwebrtc void LLVoiceWebRTCConnection::OnIceGatheringState(llwebrtc::LLWebRTCSignalingObserver::EIceGatheringState state) { LL::WorkQueue::postMaybe(mMainQueue, [=] { LL_DEBUGS("Voice") << "Ice Gathering voice account. " << state << LL_ENDL; switch (state) { case llwebrtc::LLWebRTCSignalingObserver::EIceGatheringState::ICE_GATHERING_COMPLETE: { mIceCompleted = true; break; } case llwebrtc::LLWebRTCSignalingObserver::EIceGatheringState::ICE_GATHERING_NEW: { mIceCompleted = false; } default: break; } }); } // callback from llwebrtc void LLVoiceWebRTCConnection::OnIceCandidate(const llwebrtc::LLWebRTCIceCandidate& candidate) { LL::WorkQueue::postMaybe(mMainQueue, [=] { mIceCandidates.push_back(candidate); }); } void LLVoiceWebRTCConnection::processIceUpdates() { mOutstandingRequests++; LLCoros::getInstance()->launch("LLVoiceWebRTCConnection::processIceUpdatesCoro", boost::bind(&LLVoiceWebRTCConnection::processIceUpdatesCoro, this->shared_from_this())); } // Ice candidates may be streamed in before or after the SDP offer is available (see below) // This function determines whether candidates are available to send to the Secondlife WebRTC // server via the simulator. If so, and there are no more candidates, this code // will make the cap call to the server sending up the ICE candidates. void LLVoiceWebRTCConnection::processIceUpdatesCoro(connectionPtr_t connection) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (connection->mShutDown || LLWebRTCVoiceClient::isShuttingDown()) { connection->mOutstandingRequests--; return; } bool iceCompleted = false; LLSD body; if (!connection->mIceCandidates.empty() || connection->mIceCompleted) { LLViewerRegion *regionp = LLWorld::instance().getRegionFromID(connection->mRegionID); if (!regionp || !regionp->capabilitiesReceived()) { LL_DEBUGS("Voice") << "no capabilities for ice gathering; waiting " << LL_ENDL; connection->mOutstandingRequests--; return; } std::string url = regionp->getCapability("VoiceSignalingRequest"); if (url.empty()) { connection->mOutstandingRequests--; return; } LL_DEBUGS("Voice") << "region ready to complete voice signaling; url=" << url << LL_ENDL; if (!connection->mIceCandidates.empty()) { LLSD candidates = LLSD::emptyArray(); for (auto &ice_candidate : connection->mIceCandidates) { LLSD body_candidate; body_candidate["sdpMid"] = ice_candidate.mSdpMid; body_candidate["sdpMLineIndex"] = ice_candidate.mMLineIndex; body_candidate["candidate"] = ice_candidate.mCandidate; candidates.append(body_candidate); } body["candidates"] = candidates; connection->mIceCandidates.clear(); } else if (connection->mIceCompleted) { LLSD body_candidate; body_candidate["completed"] = true; body["candidate"] = body_candidate; iceCompleted = connection->mIceCompleted; connection->mIceCompleted = false; } body["viewer_session"] = connection->mViewerSession; body["voice_server_type"] = WEBRTC_VOICE_SERVER_TYPE; LLCoreHttpUtil::HttpCoroutineAdapter::ptr_t httpAdapter( new LLCoreHttpUtil::HttpCoroutineAdapter("LLVoiceWebRTCAdHocConnection::processIceUpdatesCoro", LLCore::HttpRequest::DEFAULT_POLICY_ID)); LLCore::HttpRequest::ptr_t httpRequest(new LLCore::HttpRequest); LLCore::HttpOptions::ptr_t httpOpts(new LLCore::HttpOptions); httpOpts->setWantHeaders(true); LLSD result = httpAdapter->postAndSuspend(httpRequest, url, body, httpOpts); if (LLWebRTCVoiceClient::isShuttingDown()) { connection->mOutstandingRequests--; return; } LLSD httpResults = result[LLCoreHttpUtil::HttpCoroutineAdapter::HTTP_RESULTS]; LLCore::HttpStatus status = LLCoreHttpUtil::HttpCoroutineAdapter::getStatusFromLLSD(httpResults); if (!status) { // couldn't trickle the candidates, so restart the session. connection->setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); } } connection->mOutstandingRequests--; } // An 'Offer' comes in the form of a SDP (Session Description Protocol) // which contains all sorts of info about the session, from network paths // to the type of session (audio, video) to characteristics (the encoder type.) // This SDP also serves as the 'ticket' to the server, security-wise. // The Offer is retrieved from the WebRTC library on the client, // and is passed to the simulator via a CAP, which then passes // it on to the Secondlife WebRTC server. // // The LLWebRTCVoiceConnection object will not be deleted // before the webrtc connection itself is shut down, so // we shouldn't be getting this callback on a nonexistant // this pointer. // callback from llwebrtc void LLVoiceWebRTCConnection::OnOfferAvailable(const std::string &sdp) { LL::WorkQueue::postMaybe(mMainQueue, [=] { if (mShutDown) { return; } LL_DEBUGS("Voice") << "On Offer Available." << LL_ENDL; mChannelSDP = sdp; if (mVoiceConnectionState == VOICE_STATE_WAIT_FOR_SESSION_START) { mVoiceConnectionState = VOICE_STATE_REQUEST_CONNECTION; } }); } // // The LLWebRTCVoiceConnection object will not be deleted // before the webrtc connection itself is shut down, so // we shouldn't be getting this callback on a nonexistant // this pointer. // nor should audio_interface be invalid if the LLWebRTCVoiceConnection // is shut down. // callback from llwebrtc void LLVoiceWebRTCConnection::OnAudioEstablished(llwebrtc::LLWebRTCAudioInterface* audio_interface) { LL::WorkQueue::postMaybe(mMainQueue, [=] { if (mShutDown) { return; } LL_DEBUGS("Voice") << "On AudioEstablished." << LL_ENDL; mWebRTCAudioInterface = audio_interface; setVoiceConnectionState(VOICE_STATE_SESSION_ESTABLISHED); }); } // // The LLWebRTCVoiceConnection object will not be deleted // before the webrtc connection itself is shut down, so // we shouldn't be getting this callback on a nonexistant // this pointer. // callback from llwebrtc void LLVoiceWebRTCConnection::OnRenegotiationNeeded() { LL::WorkQueue::postMaybe(mMainQueue, [=] { LL_DEBUGS("Voice") << "Voice channel requires renegotiation." << LL_ENDL; if (!mShutDown) { setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); } mCurrentStatus = LLVoiceClientStatusObserver::ERROR_UNKNOWN; }); } // callback from llwebrtc void LLVoiceWebRTCConnection::OnPeerConnectionClosed() { LL::WorkQueue::postMaybe(mMainQueue, [=] { LL_DEBUGS("Voice") << "Peer connection has closed." << LL_ENDL; if (mVoiceConnectionState == VOICE_STATE_WAIT_FOR_CLOSE) { setVoiceConnectionState(VOICE_STATE_CLOSED); mOutstandingRequests--; } else if (LLWebRTCVoiceClient::isShuttingDown()) { // disconnect was initialized by llwebrtc::terminate() instead of connectionStateMachine LL_INFOS("Voice") << "Peer connection has closed, but state is " << mVoiceConnectionState << LL_ENDL; setVoiceConnectionState(VOICE_STATE_CLOSED); } }); } void LLVoiceWebRTCConnection::setMuteMic(bool muted) { mMuted = muted; if (mWebRTCAudioInterface) { mWebRTCAudioInterface->setMute(muted); } } void LLVoiceWebRTCConnection::setSpeakerVolume(F32 volume) { mSpeakerVolume = volume; if (mWebRTCAudioInterface) { mWebRTCAudioInterface->setReceiveVolume(volume); } } void LLVoiceWebRTCConnection::setUserVolume(const LLUUID& id, F32 volume) { boost::json::object root = {{"ug", {id.asString(), (uint32_t) (volume * 200)}}}; std::string json_data = boost::json::serialize(root); if (mWebRTCDataInterface) { mWebRTCDataInterface->sendData(json_data, false); } } void LLVoiceWebRTCConnection::setUserMute(const LLUUID& id, bool mute) { boost::json::object root = {{"m", {id.asString(), mute}}}; std::string json_data = boost::json::serialize(root); if (mWebRTCDataInterface) { mWebRTCDataInterface->sendData(json_data, false); } } // Send data to the Secondlife WebRTC server via the webrtc // data channel. void LLVoiceWebRTCConnection::sendData(const std::string &data) { if (getVoiceConnectionState() == VOICE_STATE_SESSION_UP && mWebRTCDataInterface) { mWebRTCDataInterface->sendData(data, false); } } // Tell the simulator that we're shutting down a voice connection. // The simulator will pass this on to the Secondlife WebRTC server. void LLVoiceWebRTCConnection::breakVoiceConnectionCoro(connectionPtr_t connection) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LL_DEBUGS("Voice") << "Disconnecting voice." << LL_ENDL; if (connection->mWebRTCDataInterface) { connection->mWebRTCDataInterface->unsetDataObserver(connection.get()); connection->mWebRTCDataInterface = nullptr; } connection->mWebRTCAudioInterface = nullptr; LLViewerRegion *regionp = LLWorld::instance().getRegionFromID(connection->mRegionID); if (!regionp || !regionp->capabilitiesReceived()) { LL_DEBUGS("Voice") << "no capabilities for voice provisioning; waiting " << LL_ENDL; connection->setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); connection->mOutstandingRequests--; return; } std::string url = regionp->getCapability("ProvisionVoiceAccountRequest"); if (url.empty()) { connection->setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); connection->mOutstandingRequests--; return; } LL_DEBUGS("Voice") << "region ready for voice break; url=" << url << LL_ENDL; LLVoiceWebRTCStats::getInstance()->provisionAttemptStart(); LLSD body; body["logout"] = true; body["viewer_session"] = connection->mViewerSession; body["voice_server_type"] = WEBRTC_VOICE_SERVER_TYPE; LLCoreHttpUtil::HttpCoroutineAdapter::ptr_t httpAdapter( new LLCoreHttpUtil::HttpCoroutineAdapter("LLVoiceWebRTCAdHocConnection::breakVoiceConnection", LLCore::HttpRequest::DEFAULT_POLICY_ID)); LLCore::HttpRequest::ptr_t httpRequest(new LLCore::HttpRequest); LLCore::HttpOptions::ptr_t httpOpts(new LLCore::HttpOptions); httpOpts->setWantHeaders(true); // tell the server to shut down the connection as a courtesy. // shutdownConnection will drop the WebRTC connection which will // also shut things down. LLSD result = httpAdapter->postAndSuspend(httpRequest, url, body, httpOpts); connection->mOutstandingRequests--; if (connection->getVoiceConnectionState() == VOICE_STATE_WAIT_FOR_EXIT) { connection->setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); } } // Tell the simulator to tell the Secondlife WebRTC server that we want a voice // connection. The SDP is sent up as part of this, and the simulator will respond // with an 'answer' which is in the form of another SDP. The webrtc library // will use the offer and answer to negotiate the session. void LLVoiceWebRTCSpatialConnection::requestVoiceConnection() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LLViewerRegion *regionp = LLWorld::instance().getRegionFromID(mRegionID); LL_DEBUGS("Voice") << "Requesting voice connection." << LL_ENDL; if (!regionp || !regionp->capabilitiesReceived()) { LL_DEBUGS("Voice") << "no capabilities for voice provisioning; waiting " << LL_ENDL; // try again. setVoiceConnectionState(VOICE_STATE_REQUEST_CONNECTION); mOutstandingRequests--; return; } std::string url = regionp->getCapability("ProvisionVoiceAccountRequest"); if (url.empty()) { setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); mOutstandingRequests--; return; } LL_DEBUGS("Voice") << "region ready for voice provisioning; url=" << url << LL_ENDL; LLVoiceWebRTCStats::getInstance()->provisionAttemptStart(); LLSD body; LLSD jsep; jsep["type"] = "offer"; jsep["sdp"] = mChannelSDP; body["jsep"] = jsep; if (mParcelLocalID != INVALID_PARCEL_ID) { body["parcel_local_id"] = mParcelLocalID; } body["channel_type"] = "local"; body["voice_server_type"] = WEBRTC_VOICE_SERVER_TYPE; LLCoreHttpUtil::HttpCoroutineAdapter::ptr_t httpAdapter( new LLCoreHttpUtil::HttpCoroutineAdapter("LLVoiceWebRTCAdHocConnection::requestVoiceConnection", LLCore::HttpRequest::DEFAULT_POLICY_ID)); LLCore::HttpRequest::ptr_t httpRequest(new LLCore::HttpRequest); LLCore::HttpOptions::ptr_t httpOpts(new LLCore::HttpOptions); httpOpts->setWantHeaders(true); LLSD result = httpAdapter->postAndSuspend(httpRequest, url, body, httpOpts); LLSD httpResults = result[LLCoreHttpUtil::HttpCoroutineAdapter::HTTP_RESULTS]; LLCore::HttpStatus status = LLCoreHttpUtil::HttpCoroutineAdapter::getStatusFromLLSD(httpResults); if (status) { OnVoiceConnectionRequestSuccess(result); } else { switch (status.getType()) { case HTTP_CONFLICT: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_CHANNEL_FULL; break; case HTTP_UNAUTHORIZED: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_CHANNEL_LOCKED; break; default: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_UNKNOWN; break; } setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); } mOutstandingRequests--; } void LLVoiceWebRTCConnection::OnVoiceConnectionRequestSuccess(const LLSD &result) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (LLWebRTCVoiceClient::isShuttingDown()) { return; } LLVoiceWebRTCStats::getInstance()->provisionAttemptEnd(true); if (result.has("viewer_session") && result.has("jsep") && result["jsep"].has("type") && result["jsep"]["type"] == "answer" && result["jsep"].has("sdp")) { mRemoteChannelSDP = result["jsep"]["sdp"].asString(); mViewerSession = result["viewer_session"]; } else { LL_WARNS("Voice") << "Invalid voice provision request result:" << result << LL_ENDL; setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); return; } LL_DEBUGS("Voice") << "ProvisionVoiceAccountRequest response" << " channel sdp " << mRemoteChannelSDP << LL_ENDL; mWebRTCPeerConnectionInterface->AnswerAvailable(mRemoteChannelSDP); } static llwebrtc::LLWebRTCPeerConnectionInterface::InitOptions getConnectionOptions() { llwebrtc::LLWebRTCPeerConnectionInterface::InitOptions options; llwebrtc::LLWebRTCPeerConnectionInterface::InitOptions::IceServers servers; // TODO: Pull these from login std::string grid = LLGridManager::getInstance()->getGridLoginID(); std::transform(grid.begin(), grid.end(), grid.begin(), [](unsigned char c){ return std::tolower(c); }); int num_servers = 2; if (grid == "agni") { num_servers = 3; } for (int i=1; i <= num_servers; i++) { servers.mUrls.push_back(llformat("stun:stun%d.%s.secondlife.io:3478", i, grid.c_str())); } options.mServers.push_back(servers); return options; } // Primary state machine for negotiating a single voice connection to the // Secondlife WebRTC server. bool LLVoiceWebRTCConnection::connectionStateMachine() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (!mShutDown) { processIceUpdates(); } switch (getVoiceConnectionState()) { case VOICE_STATE_START_SESSION: { LL_PROFILE_ZONE_NAMED_CATEGORY_VOICE("VOICE_STATE_START_SESSION") if (mShutDown) { setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); break; } mIceCompleted = false; setVoiceConnectionState(VOICE_STATE_WAIT_FOR_SESSION_START); // tell the webrtc library that we want a connection. The library will // respond with an offer on a separate thread, which will cause // the session state to change. if (!mWebRTCPeerConnectionInterface->initializeConnection(getConnectionOptions())) { setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); } break; } case VOICE_STATE_WAIT_FOR_SESSION_START: { if (mShutDown) { setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); } break; } case VOICE_STATE_REQUEST_CONNECTION: if (mShutDown) { setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); break; } // Ask the sim to ask the Secondlife WebRTC server for a connection to // a given voice channel. On completion, we'll move on to the // VOICE_STATE_SESSION_ESTABLISHED via a callback on a webrtc thread. setVoiceConnectionState(VOICE_STATE_CONNECTION_WAIT); mOutstandingRequests++; LLCoros::getInstance()->launch("LLVoiceWebRTCConnection::requestVoiceConnectionCoro", boost::bind(&LLVoiceWebRTCConnection::requestVoiceConnectionCoro, this->shared_from_this())); break; case VOICE_STATE_CONNECTION_WAIT: if (mShutDown) { setVoiceConnectionState(VOICE_STATE_DISCONNECT); } break; case VOICE_STATE_SESSION_ESTABLISHED: { if (mShutDown) { setVoiceConnectionState(VOICE_STATE_DISCONNECT); break; } // update the peer connection with the various characteristics of // this connection. mWebRTCAudioInterface->setMute(mMuted); mWebRTCAudioInterface->setReceiveVolume(mSpeakerVolume); LLWebRTCVoiceClient::getInstance()->OnConnectionEstablished(mChannelID, mRegionID); setVoiceConnectionState(VOICE_STATE_WAIT_FOR_DATA_CHANNEL); break; } case VOICE_STATE_WAIT_FOR_DATA_CHANNEL: { if (mShutDown) { setVoiceConnectionState(VOICE_STATE_DISCONNECT); break; } if (mWebRTCDataInterface) // the interface will be set when the session is negotiated. { sendJoin(); // tell the Secondlife WebRTC server that we're here via the data channel. setVoiceConnectionState(VOICE_STATE_SESSION_UP); if (isSpatial()) { LLWebRTCVoiceClient::getInstance()->updatePosition(); LLWebRTCVoiceClient::getInstance()->sendPositionUpdate(true); } } break; } case VOICE_STATE_SESSION_UP: { mRetryWaitPeriod = 0; mRetryWaitSecs = (F32)((F32)rand() / (RAND_MAX)) + 0.5f; // we'll stay here as long as the session remains up. if (mShutDown) { setVoiceConnectionState(VOICE_STATE_DISCONNECT); } else { if (isSpatial() && gAgent.getRegion()) { bool primary = (mRegionID == gAgent.getRegion()->getRegionID()); if (primary != mPrimary) { mPrimary = primary; sendJoin(); } } } break; } case VOICE_STATE_SESSION_RETRY: // only retry ever 'n' seconds if (mRetryWaitPeriod++ * UPDATE_THROTTLE_SECONDS > mRetryWaitSecs) { // something went wrong, so notify that the connection has failed. LLWebRTCVoiceClient::getInstance()->OnConnectionFailure(mChannelID, mRegionID, mCurrentStatus); setVoiceConnectionState(VOICE_STATE_DISCONNECT); mRetryWaitPeriod = 0; if (mRetryWaitSecs < MAX_RETRY_WAIT_SECONDS) { // back off the retry period, and do it by a small random // bit so all clients don't reconnect at once. mRetryWaitSecs += (F32)((F32) rand() / (RAND_MAX)) + 0.5f; mRetryWaitPeriod = 0; } } break; case VOICE_STATE_DISCONNECT: if (!LLWebRTCVoiceClient::isShuttingDown()) { mOutstandingRequests++; setVoiceConnectionState(VOICE_STATE_WAIT_FOR_EXIT); LLCoros::instance().launch("LLVoiceWebRTCConnection::breakVoiceConnectionCoro", boost::bind(&LLVoiceWebRTCConnection::breakVoiceConnectionCoro, this->shared_from_this())); } else { // llwebrtc::terminate() is already shuting down the connection. setVoiceConnectionState(VOICE_STATE_WAIT_FOR_CLOSE); } break; case VOICE_STATE_WAIT_FOR_EXIT: break; case VOICE_STATE_SESSION_EXIT: { setVoiceConnectionState(VOICE_STATE_WAIT_FOR_CLOSE); mOutstandingRequests++; if (!LLWebRTCVoiceClient::isShuttingDown()) { mWebRTCPeerConnectionInterface->shutdownConnection(); } // else was already posted by llwebrtc::terminate(). break; case VOICE_STATE_WAIT_FOR_CLOSE: break; case VOICE_STATE_CLOSED: if (!mShutDown) { mVoiceConnectionState = VOICE_STATE_START_SESSION; } else { // if we still have outstanding http or webrtc calls, wait for them to // complete so we don't delete objects while they still may be used. if (mOutstandingRequests <= 0) { LLWebRTCVoiceClient::getInstance()->OnConnectionShutDown(mChannelID, mRegionID); return false; } } break; } default: { LL_WARNS("Voice") << "Unknown voice control state " << getVoiceConnectionState() << LL_ENDL; return false; } } return true; } // Data has been received on the webrtc data channel // incoming data will be a json structure (if it's not binary.) We may pack // binary for size reasons. Most of the keys in the json objects are // single or double characters for size reasons. // The primary element is: // An object where each key is an agent id. (in the future, we may allow // integer indices into an agentid list, populated on join commands. For size. // Each key will point to a json object with keys identifying what's updated. // 'p' - audio source power (level/volume) (int8 as int) // 'j' - object of join data (currently only a boolean 'p' marking a primary participant) // 'l' - boolean, always true if exists. // 'v' - boolean - voice activity has been detected. // llwebrtc callback void LLVoiceWebRTCConnection::OnDataReceived(const std::string& data, bool binary) { LL::WorkQueue::postMaybe(mMainQueue, [=] { LLVoiceWebRTCConnection::OnDataReceivedImpl(data, binary); }); } // // The LLWebRTCVoiceConnection object will not be deleted // before the webrtc connection itself is shut down, so // we shouldn't be getting this callback on a nonexistant // this pointer. void LLVoiceWebRTCConnection::OnDataReceivedImpl(const std::string &data, bool binary) { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; if (mShutDown) { return; } if (binary) { LL_WARNS("Voice") << "Binary data received from data channel." << LL_ENDL; return; } boost::system::error_code ec; boost::json::value voice_data_parsed = boost::json::parse(data, ec); if (!ec) // don't collect comments { if (!voice_data_parsed.is_object()) { LL_WARNS("Voice") << "Expected object from data channel:" << data << LL_ENDL; return; } boost::json::object voice_data = voice_data_parsed.as_object(); bool new_participant = false; boost::json::object mute; boost::json::object user_gain; for (auto &participant_elem : voice_data) { boost::json::string participant_id(participant_elem.key()); LLUUID agent_id(participant_id.c_str()); if (agent_id.isNull()) { // probably a test client. continue; } if (!participant_elem.value().is_object()) { continue; } boost::json::object participant_obj = participant_elem.value().as_object(); LLWebRTCVoiceClient::participantStatePtr_t participant = LLWebRTCVoiceClient::getInstance()->findParticipantByID(mChannelID, agent_id); bool joined = false; bool primary = false; // we ignore any 'joins' reported about participants // that come from voice servers that aren't their primary // voice server. This will happen with cross-region voice // where a participant on a neighboring region may be // connected to multiple servers. We don't want to // add new identical participants from all of those servers. if (participant_obj.contains("j") && participant_obj["j"].is_object()) { // a new participant has announced that they're joining. joined = true; if (participant_elem.value().as_object()["j"].as_object().contains("p") && participant_elem.value().as_object()["j"].as_object()["p"].is_bool()) { primary = participant_elem.value().as_object()["j"].as_object()["p"].as_bool(); } // track incoming participants that are muted so we can mute their connections (or set their volume) bool isMuted = LLMuteList::getInstance()->isMuted(agent_id, LLMute::flagVoiceChat); if (isMuted) { mute[participant_id] = true; } F32 volume; if(LLSpeakerVolumeStorage::getInstance()->getSpeakerVolume(agent_id, volume)) { user_gain[participant_id] = (uint32_t)(volume * 200); } } new_participant |= joined; if (!participant && joined && (primary || !isSpatial())) { participant = LLWebRTCVoiceClient::getInstance()->addParticipantByID(mChannelID, agent_id, mRegionID); } if (participant) { if (participant_obj.contains("l") && participant_obj["l"].is_bool() && participant_obj["l"].as_bool()) { // an existing participant is leaving. if (agent_id != gAgentID) { LLWebRTCVoiceClient::getInstance()->removeParticipantByID(mChannelID, agent_id, mRegionID); } } else { // we got a 'power' update. if (participant_obj.contains("p") && participant_obj["p"].is_number()) { // server sends up power as an integer which is level * 128 to save // character count. participant->mLevel = (F32)participant_obj["p"].as_int64()/128.0f; } if (participant_obj.contains("v") && participant_obj["v"].is_bool()) { participant->mIsSpeaking = participant_obj["v"].as_bool(); } if (participant_obj.contains("m") && participant_obj["m"].is_bool()) { participant->mIsModeratorMuted = participant_obj["m"].as_bool(); } } } } // tell the simulator to set the mute and volume data for this // participant, if there are any updates. boost::json::object root; if (mute.size() > 0) { root["m"] = mute; } if (user_gain.size() > 0) { root["ug"] = user_gain; } if (root.size() > 0) { std::string json_data = boost::json::serialize(root); mWebRTCDataInterface->sendData(json_data, false); } } } // // The LLWebRTCVoiceConnection object will not be deleted // before the webrtc connection itself is shut down, so // we shouldn't be getting this callback on a nonexistant // this pointer. // nor should data_interface be invalid if the LLWebRTCVoiceConnection // is shut down. // llwebrtc callback void LLVoiceWebRTCConnection::OnDataChannelReady(llwebrtc::LLWebRTCDataInterface *data_interface) { LL::WorkQueue::postMaybe(mMainQueue, [=] { if (mShutDown) { return; } if (data_interface) { mWebRTCDataInterface = data_interface; mWebRTCDataInterface->setDataObserver(this); } }); } // tell the Secondlife WebRTC server that // we're joining and whether we're // joining a server associated with the // the region we currently occupy or not (primary) // The WebRTC voice server will pass this info // to peers. void LLVoiceWebRTCConnection::sendJoin() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; boost::json::object root; boost::json::object join_obj; if (mPrimary) { join_obj["p"] = true; } root["j"] = join_obj; std::string json_data = boost::json::serialize(root); mWebRTCDataInterface->sendData(json_data, false); } ///////////////////////////// // WebRTC Spatial Connection LLVoiceWebRTCSpatialConnection::LLVoiceWebRTCSpatialConnection(const LLUUID ®ionID, S32 parcelLocalID, const std::string &channelID) : LLVoiceWebRTCConnection(regionID, channelID), mParcelLocalID(parcelLocalID) { if (gAgent.getRegion()) { mPrimary = (regionID == gAgent.getRegion()->getRegionID()); } } LLVoiceWebRTCSpatialConnection::~LLVoiceWebRTCSpatialConnection() { } void LLVoiceWebRTCSpatialConnection::setMuteMic(bool muted) { if (mMuted != muted) { mMuted = muted; if (mWebRTCAudioInterface) { LLViewerRegion *regionp = gAgent.getRegion(); if (regionp && mRegionID == regionp->getRegionID()) { mWebRTCAudioInterface->setMute(muted); } else { // Always mute this agent with respect to neighboring regions. // Peers don't want to hear this agent from multiple regions // as that'll echo. mWebRTCAudioInterface->setMute(true); } } } } ///////////////////////////// // WebRTC Spatial Connection LLVoiceWebRTCAdHocConnection::LLVoiceWebRTCAdHocConnection(const LLUUID ®ionID, const std::string& channelID, const std::string& credentials) : LLVoiceWebRTCConnection(regionID, channelID), mCredentials(credentials) { } LLVoiceWebRTCAdHocConnection::~LLVoiceWebRTCAdHocConnection() { } // Add-hoc connections require a different channel type // as they go to a different set of Secondlife WebRTC servers. // They also require credentials for the given channels. // So, we have a separate requestVoiceConnection call. void LLVoiceWebRTCAdHocConnection::requestVoiceConnection() { LL_PROFILE_ZONE_SCOPED_CATEGORY_VOICE; LLViewerRegion *regionp = LLWorld::instance().getRegionFromID(mRegionID); LL_DEBUGS("Voice") << "Requesting voice connection." << LL_ENDL; if (!regionp || !regionp->capabilitiesReceived()) { LL_DEBUGS("Voice") << "no capabilities for voice provisioning; retrying " << LL_ENDL; // try again. setVoiceConnectionState(VOICE_STATE_REQUEST_CONNECTION); mOutstandingRequests--; return; } std::string url = regionp->getCapability("ProvisionVoiceAccountRequest"); if (url.empty()) { setVoiceConnectionState(VOICE_STATE_SESSION_RETRY); mOutstandingRequests--; return; } LLVoiceWebRTCStats::getInstance()->provisionAttemptStart(); LLSD body; LLSD jsep; jsep["type"] = "offer"; { jsep["sdp"] = mChannelSDP; } body["jsep"] = jsep; body["credentials"] = mCredentials; body["channel"] = mChannelID; body["channel_type"] = "multiagent"; body["voice_server_type"] = WEBRTC_VOICE_SERVER_TYPE; LLCoreHttpUtil::HttpCoroutineAdapter::ptr_t httpAdapter( new LLCoreHttpUtil::HttpCoroutineAdapter("LLVoiceWebRTCAdHocConnection::requestVoiceConnection", LLCore::HttpRequest::DEFAULT_POLICY_ID)); LLCore::HttpRequest::ptr_t httpRequest(new LLCore::HttpRequest); LLCore::HttpOptions::ptr_t httpOpts(new LLCore::HttpOptions); httpOpts->setWantHeaders(true); LLSD result = httpAdapter->postAndSuspend(httpRequest, url, body, httpOpts); LLSD httpResults = result[LLCoreHttpUtil::HttpCoroutineAdapter::HTTP_RESULTS]; LLCore::HttpStatus status = LLCoreHttpUtil::HttpCoroutineAdapter::getStatusFromLLSD(httpResults); if (!status) { switch (status.getType()) { case HTTP_CONFLICT: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_CHANNEL_FULL; break; case HTTP_UNAUTHORIZED: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_CHANNEL_LOCKED; break; default: mCurrentStatus = LLVoiceClientStatusObserver::ERROR_UNKNOWN; break; } setVoiceConnectionState(VOICE_STATE_SESSION_EXIT); } else { OnVoiceConnectionRequestSuccess(result); } mOutstandingRequests--; }