Merge branch 'roxie/webrtc-voice' of https://github.com/secondlife/viewer into roxie/webrtc-voice

9 files changed, 256 insertions, 42 deletions

diff --git a/indra/llwebrtc/llwebrtc.cpp b/indra/llwebrtc/llwebrtc.cpp
index bba1e99e2d..a1e125a8f2 100644
--- a/indra/llwebrtc/llwebrtc.cpp
+++ b/indra/llwebrtc/llwebrtc.cpp
@@ -157,7 +157,6 @@ LLWebRTCImpl::LLWebRTCImpl() :
 
 void LLWebRTCImpl::init()
 {
-    RTC_DCHECK(mPeerConnectionFactory);
     mPlayoutDevice   = 0;
     mRecordingDevice = 0;
     rtc::InitializeSSL();
@@ -222,12 +221,10 @@ void LLWebRTCImpl::init()
     mPeerCustomProcessor = new LLCustomProcessor;
     webrtc::AudioProcessingBuilder apb;
     apb.SetCapturePostProcessing(std::unique_ptr<webrtc::CustomProcessing>(mPeerCustomProcessor));
-    rtc::scoped_refptr<webrtc::AudioProcessing> apm = apb.Create();
+    mAudioProcessingModule = apb.Create();
 
-    // TODO: wire some of these to the primary interface and ultimately
-    // to the UI to allow user config.
     webrtc::AudioProcessing::Config apm_config;
-    apm_config.echo_canceller.enabled         = true;
+    apm_config.echo_canceller.enabled         = false;
     apm_config.echo_canceller.mobile_mode     = false;
     apm_config.gain_controller1.enabled       = true;
     apm_config.gain_controller1.mode          = webrtc::AudioProcessing::Config::GainController1::kAdaptiveAnalog;
@@ -250,8 +247,8 @@ void LLWebRTCImpl::init()
     processing_config.reverse_output_stream().set_num_channels(2);
     processing_config.reverse_output_stream().set_sample_rate_hz(48000);
 
-    apm->Initialize(processing_config);
-    apm->ApplyConfig(apm_config);
+    mAudioProcessingModule->Initialize(processing_config);
+    mAudioProcessingModule->ApplyConfig(apm_config);
 
     mPeerConnectionFactory = webrtc::CreatePeerConnectionFactory(mNetworkThread.get(),
                                                                  mWorkerThread.get(),
@@ -262,7 +259,7 @@ void LLWebRTCImpl::init()
                                                                  nullptr /* video_encoder_factory */,
                                                                  nullptr /* video_decoder_factory */,
                                                                  nullptr /* audio_mixer */,
-                                                                 apm);
+                                                                 mAudioProcessingModule);
 
     mWorkerThread->BlockingCall([this]() { mPeerDeviceModule->StartPlayout(); });
 }
@@ -318,6 +315,49 @@ void LLWebRTCImpl::setRecording(bool recording)
         });
 }
 
+void LLWebRTCImpl::setAudioConfig(LLWebRTCDeviceInterface::AudioConfig config)
+{
+    webrtc::AudioProcessing::Config apm_config;
+    apm_config.echo_canceller.enabled         = config.mEchoCancellation;
+    apm_config.echo_canceller.mobile_mode     = false;
+    apm_config.gain_controller1.enabled       = true;
+    apm_config.gain_controller1.mode          = webrtc::AudioProcessing::Config::GainController1::kAdaptiveAnalog;
+    apm_config.gain_controller2.enabled       = true;
+    apm_config.high_pass_filter.enabled       = true;
+    apm_config.transient_suppression.enabled  = true;
+    apm_config.pipeline.multi_channel_render  = true;
+    apm_config.pipeline.multi_channel_capture = true;
+    apm_config.pipeline.multi_channel_capture = true;
+
+    switch (config.mNoiseSuppressionLevel)
+    {
+        case LLWebRTCDeviceInterface::AudioConfig::NOISE_SUPPRESSION_LEVEL_NONE:
+            apm_config.noise_suppression.enabled = false;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kLow;
+            break;
+        case LLWebRTCDeviceInterface::AudioConfig::NOISE_SUPPRESSION_LEVEL_LOW:
+            apm_config.noise_suppression.enabled = true;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kLow;
+            break;
+        case LLWebRTCDeviceInterface::AudioConfig::NOISE_SUPPRESSION_LEVEL_MODERATE:
+            apm_config.noise_suppression.enabled = true;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kModerate;
+            break;
+        case LLWebRTCDeviceInterface::AudioConfig::NOISE_SUPPRESSION_LEVEL_HIGH:
+            apm_config.noise_suppression.enabled = true;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kHigh;
+            break;
+        case LLWebRTCDeviceInterface::AudioConfig::NOISE_SUPPRESSION_LEVEL_VERY_HIGH:
+            apm_config.noise_suppression.enabled = true;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kVeryHigh;
+            break;
+        default:
+            apm_config.noise_suppression.enabled = false;
+            apm_config.noise_suppression.level = webrtc::AudioProcessing::Config::NoiseSuppression::kLow;
+    }
+    mAudioProcessingModule->ApplyConfig(apm_config);
+}
+
 void LLWebRTCImpl::refreshDevices()
 {
     mWorkerThread->PostTask([this]() { updateDevices(); });
@@ -619,32 +659,36 @@ void LLWebRTCPeerConnectionImpl::unsetSignalingObserver(LLWebRTCSignalingObserve
     }
 }
 
-// TODO: Add initialization structure through which
-// stun and turn servers may be passed in from
-// the sim or login.
-bool LLWebRTCPeerConnectionImpl::initializeConnection()
+bool LLWebRTCPeerConnectionImpl::initializeConnection(LLWebRTCPeerConnectionInterface::InitOptions options)
 {
     RTC_DCHECK(!mPeerConnection);
     mAnswerReceived = false;
 
     mWebRTCImpl->PostSignalingTask(
-        [this]()
+        [this, options]()
         {
+            std::vector<LLWebRTCPeerConnectionInterface::InitOptions::IceServers> servers = options.mServers;
+            if(servers.empty())
+            {
+                LLWebRTCPeerConnectionInterface::InitOptions::IceServers ice_servers;
+                ice_servers.mUrls.push_back("stun:stun.l.google.com:19302");
+                ice_servers.mUrls.push_back("stun1:stun.l.google.com:19302");
+                ice_servers.mUrls.push_back("stun2:stun.l.google.com:19302");
+                ice_servers.mUrls.push_back("stun3:stun.l.google.com:19302");
+                ice_servers.mUrls.push_back("stun4:stun.l.google.com:19302");
+            }
+
             webrtc::PeerConnectionInterface::RTCConfiguration config;
+            for (auto server : servers)
+            {
+                webrtc::PeerConnectionInterface::IceServer ice_server;
+                ice_server.urls = server.mUrls;
+                ice_server.username = server.mUserName;
+                ice_server.password = server.mPassword;
+                config.servers.push_back(ice_server);
+            }
+            
             config.sdp_semantics = webrtc::SdpSemantics::kUnifiedPlan;
-            webrtc::PeerConnectionInterface::IceServer server;
-            server.uri = "stun:roxie-turn.staging.secondlife.io:3478";
-            config.servers.push_back(server);
-            server.uri = "stun:stun.l.google.com:19302";
-            config.servers.push_back(server);
-            server.uri = "stun:stun1.l.google.com:19302";
-            config.servers.push_back(server);
-            server.uri = "stun:stun2.l.google.com:19302";
-            config.servers.push_back(server);
-            server.uri = "stun:stun3.l.google.com:19302";
-            config.servers.push_back(server);
-            server.uri = "stun:stun4.l.google.com:19302";
-            config.servers.push_back(server);
 
             config.set_min_port(60000);
             config.set_max_port(60100);
@@ -674,7 +718,7 @@ bool LLWebRTCPeerConnectionImpl::initializeConnection()
 
             cricket::AudioOptions audioOptions;
             audioOptions.auto_gain_control = true;
-            audioOptions.echo_cancellation = true;  // incompatible with opus stereo
+            audioOptions.echo_cancellation = false;  // incompatible with opus stereo
             audioOptions.noise_suppression = true;
 
             mLocalStream = mPeerConnectionFactory->CreateLocalMediaStream("SLStream");
diff --git a/indra/llwebrtc/llwebrtc.h b/indra/llwebrtc/llwebrtc.h
index be2e5cdf68..f3a33435c9 100644
--- a/indra/llwebrtc/llwebrtc.h
+++ b/indra/llwebrtc/llwebrtc.h
@@ -99,10 +99,32 @@ class LLWebRTCDevicesObserver
 // to enumerate, set, and get notifications of changes
 // for both capture (microphone) and render (speaker)
 // devices.
+
 class LLWebRTCDeviceInterface
 {
   public:
-
+    struct AudioConfig {
+
+        bool mAGC { true };
+
+        bool mEchoCancellation { true };
+
+        // TODO: The various levels of noise suppression are configured
+        // on the APM which would require setting config on the APM.
+        // We should pipe the various values through
+        // later.
+        typedef enum {
+            NOISE_SUPPRESSION_LEVEL_NONE = 0,
+            NOISE_SUPPRESSION_LEVEL_LOW,
+            NOISE_SUPPRESSION_LEVEL_MODERATE,
+            NOISE_SUPPRESSION_LEVEL_HIGH,
+            NOISE_SUPPRESSION_LEVEL_VERY_HIGH
+        } ENoiseSuppressionLevel;
+        ENoiseSuppressionLevel mNoiseSuppressionLevel { NOISE_SUPPRESSION_LEVEL_VERY_HIGH };
+    };
+
+    virtual void setAudioConfig(AudioConfig config) = 0;
+    
     // instructs webrtc to refresh the device list.
     virtual void refreshDevices() = 0;
 
@@ -194,19 +216,35 @@ class LLWebRTCSignalingObserver
     virtual void OnDataChannelReady(LLWebRTCDataInterface *data_interface) = 0;
 };
 
-
 // LLWebRTCPeerConnectionInterface representsd a connection to a peer,
 // in most cases a Secondlife WebRTC server.  This interface
 // allows for management of this peer connection.
 class LLWebRTCPeerConnectionInterface
 {
   public:
+    
+    struct InitOptions
+    {
+        // equivalent of PeerConnectionInterface::IceServer
+        struct IceServers {
+            // Valid formats are described in RFC7064 and RFC7065.
+            // Urls should containe dns hostnames (not IP addresses)
+            // as the TLS certificate policy is 'secure.'
+            // and we do not currentply support TLS extensions.
+            std::vector<std::string> mUrls;
+            std::string mUserName;
+            std::string mPassword;
+        };
+
+        std::vector<IceServers> mServers;
+    };
+
+    virtual bool initializeConnection(InitOptions options = InitOptions()) = 0;
+    virtual bool shutdownConnection() = 0;
 
     virtual void setSignalingObserver(LLWebRTCSignalingObserver* observer) = 0;
     virtual void unsetSignalingObserver(LLWebRTCSignalingObserver* observer) = 0;
 
-    virtual bool initializeConnection() = 0;
-    virtual bool shutdownConnection() = 0;
     virtual void AnswerAvailable(const std::string &sdp) = 0;
 };
 
diff --git a/indra/llwebrtc/llwebrtc_impl.h b/indra/llwebrtc/llwebrtc_impl.h
index 38810a29b5..32faf2516c 100644
--- a/indra/llwebrtc/llwebrtc_impl.h
+++ b/indra/llwebrtc/llwebrtc_impl.h
@@ -145,6 +145,8 @@ class LLWebRTCImpl : public LLWebRTCDeviceInterface, public webrtc::AudioDeviceS
     // LLWebRTCDeviceInterface
     //
 
+    void setAudioConfig(LLWebRTCDeviceInterface::AudioConfig config = LLWebRTCDeviceInterface::AudioConfig()) override;
+
     void refreshDevices() override;
 
     void setDevicesObserver(LLWebRTCDevicesObserver *observer) override;
@@ -227,6 +229,8 @@ class LLWebRTCImpl : public LLWebRTCDeviceInterface, public webrtc::AudioDeviceS
 
     // The factory that allows creation of native webrtc PeerConnections.
     rtc::scoped_refptr<webrtc::PeerConnectionFactoryInterface> mPeerConnectionFactory;
+    
+    rtc::scoped_refptr<webrtc::AudioProcessing>                mAudioProcessingModule;
 
     // more native webrtc stuff
     std::unique_ptr<webrtc::TaskQueueFactory>                  mTaskQueueFactory;
@@ -278,11 +282,11 @@ class LLWebRTCPeerConnectionImpl : public LLWebRTCPeerConnectionInterface,
     //
     // LLWebRTCPeerConnection
     //
+    bool initializeConnection(InitOptions options = InitOptions()) override;
+    bool shutdownConnection() override;
 
     void setSignalingObserver(LLWebRTCSignalingObserver *observer) override;
     void unsetSignalingObserver(LLWebRTCSignalingObserver *observer) override;
-    bool initializeConnection() override;
-    bool shutdownConnection() override;
     void AnswerAvailable(const std::string &sdp) override;
 
     //
diff --git a/indra/newview/app_settings/settings.xml b/indra/newview/app_settings/settings.xml
index 5f7d1d8a21..0de6db0d65 100644
--- a/indra/newview/app_settings/settings.xml
+++ b/indra/newview/app_settings/settings.xml
@@ -15108,6 +15108,39 @@
       <key>Value</key>
       <integer>44125</integer>
     </map>
+    <key>VoiceEchoCancellation</key>
+    <map>
+      <key>Comment</key>
+      <string>Voice Echo Cancellation</string>
+      <key>Persist</key>
+      <integer>1</integer>
+      <key>Type</key>
+      <string>Boolean</string>
+      <key>Value</key>
+      <integer>1</integer>
+    </map>
+    <key>VoiceAutomaticGainControl</key>
+    <map>
+      <key>Comment</key>
+      <string>Voice Automatic Gain Control</string>
+      <key>Persist</key>
+      <integer>1</integer>
+      <key>Type</key>
+      <string>Boolean</string>
+      <key>Value</key>
+      <integer>1</integer>
+    </map>
+    <key>VoiceNoiseSuppressionLevel</key>
+    <map>
+      <key>Comment</key>
+      <string>Voice Noise Suppression Level</string>
+      <key>Persist</key>
+      <integer>1</integer>
+      <key>Type</key>
+      <string>U32</string>
+      <key>Value</key>
+      <integer>4</integer>
+    </map>
     <key>WarningsAsChat</key>
     <map>
       <key>Comment</key>
diff --git a/indra/newview/llviewercontrol.cpp b/indra/newview/llviewercontrol.cpp
index 7738cb904e..44173a8043 100644
--- a/indra/newview/llviewercontrol.cpp
+++ b/indra/newview/llviewercontrol.cpp
@@ -803,6 +803,9 @@ void settings_setup_listeners()
     setting_setup_signal_listener(gSavedSettings, "PushToTalkButton", handleVoiceClientPrefsChanged);
     setting_setup_signal_listener(gSavedSettings, "PushToTalkToggle", handleVoiceClientPrefsChanged);
     setting_setup_signal_listener(gSavedSettings, "VoiceEarLocation", handleVoiceClientPrefsChanged);
+    setting_setup_signal_listener(gSavedSettings, "VoiceEchoCancellation", handleVoiceClientPrefsChanged);
+    setting_setup_signal_listener(gSavedSettings, "VoiceAutomaticGainControl", handleVoiceClientPrefsChanged);
+    setting_setup_signal_listener(gSavedSettings, "VoiceNoiseSuppressionLevel", handleVoiceClientPrefsChanged);
     setting_setup_signal_listener(gSavedSettings, "VoiceInputAudioDevice", handleVoiceClientPrefsChanged);
     setting_setup_signal_listener(gSavedSettings, "VoiceOutputAudioDevice", handleVoiceClientPrefsChanged);
     setting_setup_signal_listener(gSavedSettings, "AudioLevelMic", handleVoiceClientPrefsChanged);
diff --git a/indra/newview/llvoicevivox.cpp b/indra/newview/llvoicevivox.cpp
index 29aba8ecba..f5c9e3aa98 100644
--- a/indra/newview/llvoicevivox.cpp
+++ b/indra/newview/llvoicevivox.cpp
@@ -1154,7 +1154,9 @@ bool LLVivoxVoiceClient::provisionVoiceAccount()
     do
     {
         LLVoiceVivoxStats::getInstance()->provisionAttemptStart();
-        result = httpAdapter->postAndSuspend(httpRequest, url, LLSD(), httpOpts);
+        LLSD body;
+        body["voice_server_type"] = "vivox";
+        result = httpAdapter->postAndSuspend(httpRequest, url, body, httpOpts);
 
         if (sShuttingDown)
         {
diff --git a/indra/newview/llvoicewebrtc.cpp b/indra/newview/llvoicewebrtc.cpp
index c196e16684..f3d3460022 100644
--- a/indra/newview/llvoicewebrtc.cpp
+++ b/indra/newview/llvoicewebrtc.cpp
@@ -24,6 +24,7 @@
  * $/LicenseInfo$
  */
 #include <algorithm>
+#include <format>
 #include "llvoicewebrtc.h"
 
 #include "llsdutil.h"
@@ -279,7 +280,7 @@ void LLWebRTCVoiceClient::cleanUp()
     LL_DEBUGS("Voice") << "Exiting" << LL_ENDL;
 }
 
-//---------------------------------------------------
+// --------------------------------------------------
 
 const LLVoiceVersionInfo& LLWebRTCVoiceClient::getVersion()
 {
@@ -301,6 +302,13 @@ void LLWebRTCVoiceClient::updateSettings()
     setRenderDevice(outputDevice);
     F32 mic_level = gSavedSettings.getF32("AudioLevelMic");
     setMicGain(mic_level);
+    
+    llwebrtc::LLWebRTCDeviceInterface::AudioConfig config;
+    config.mEchoCancellation = gSavedSettings.getBOOL("VoiceEchoCancellation");
+    config.mAGC = gSavedSettings.getBOOL("VoiceAutomaticGainControl");
+    config.mNoiseSuppressionLevel = (llwebrtc::LLWebRTCDeviceInterface::AudioConfig::ENoiseSuppressionLevel)gSavedSettings.getU32("VoiceNoiseSuppressionLevel");
+    mWebRTCDeviceInterface->setAudioConfig(config);
+    
 }
 
 // Observers
@@ -2515,6 +2523,28 @@ void LLVoiceWebRTCConnection::OnVoiceConnectionRequestSuccess(const LLSD &result
     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()
@@ -2535,10 +2565,11 @@ bool LLVoiceWebRTCConnection::connectionStateMachine()
             }
             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())
+            if (!mWebRTCPeerConnectionInterface->initializeConnection(getConnectionOptions()))
             {
                 setVoiceConnectionState(VOICE_STATE_SESSION_RETRY);
             }
diff --git a/indra/newview/llvoicewebrtc.h b/indra/newview/llvoicewebrtc.h
index 4fe0e756a7..aa3298ec1b 100644
--- a/indra/newview/llvoicewebrtc.h
+++ b/indra/newview/llvoicewebrtc.h
@@ -274,7 +274,7 @@ public:
         participantStatePtr_t findParticipantByID(const LLUUID& id);
 
         static ptr_t matchSessionByChannelID(const std::string& channel_id);
-
+        
         void shutdownAllConnections();
         void revive();
 
diff --git a/indra/newview/skins/default/xui/en/panel_preferences_sound.xml b/indra/newview/skins/default/xui/en/panel_preferences_sound.xml
index ab2e9c72f3..3dba5d060b 100644
--- a/indra/newview/skins/default/xui/en/panel_preferences_sound.xml
+++ b/indra/newview/skins/default/xui/en/panel_preferences_sound.xml
@@ -340,7 +340,7 @@
    follows="left|top"
    top_delta="-6"
    layout="topleft"
-   left_pad="20"
+   left_pad="10"
    width="360"
    height="40"
    name="media_ear_location">
@@ -422,7 +422,7 @@
    control_name="VoiceEarLocation"
    follows="left|top"
    layout="topleft"
-   left_pad="20"
+   left_pad="10"
    top_delta="-6"
    width="360"
    height="40"
@@ -455,11 +455,21 @@
    top_pad="10" 
    width="237"/>
   <check_box
+   control_name="VoiceEchoCancellation"
+   height="15"
+   tool_tip="Check to enable voice echo cancellation"
+   label="Echo Cancellation"
+   layout="topleft"
+   left="260"
+   name="enable_echo_cancellation"
+   top_pad="-15"
+   width="200"/>
+  <check_box
    follows="top|left"
    enabled_control="EnableVoiceChat"
    control_name="PushToTalkToggle"
    height="15"
-   label="Toggle speak on/off when I press button in toolbar"
+   label="Toggle speak on/off with toolbar button"
    layout="topleft"
    left="20"
    name="push_to_talk_toggle_check"
@@ -467,6 +477,16 @@
    tool_tip="When in toggle mode, press and release the trigger key ONCE to switch your microphone on or off. When not in toggle mode, the microphone broadcasts your voice only while the trigger is being held down."
    top_pad="5"/>
   <check_box
+   control_name="VoiceAutomaticGainControl"
+   height="15"
+   tool_tip="Check to enable automatic gain control"
+   label="Automatic Gain Control"
+   layout="topleft"
+   name="voice_automatic_gain_control"
+   left="260"
+   top_pad="-15"
+   width="200"/>
+  <check_box
    name="gesture_audio_play_btn"
    control_name="EnableGestureSounds"
    disabled_control="MuteAudio"
@@ -477,6 +497,45 @@
    label="Play sounds from gestures"
    top_pad="5"
    left="20"/>
+  <text
+   layout="topleft"
+   height="15"
+   left="260"
+   top_pad="-12"
+   width="100"
+   name="noise_suppression_label">
+    Noise Suppression
+  </text>
+  <combo_box
+      control_name="VoiceNoiseSuppressionLevel"
+      enabled_control="AudioStreamingMedia"
+      layout="topleft"
+      height="23"
+      left_pad="10"
+      top_pad="-18"
+      name="noise_suppression_combo"
+      width="80">
+  <item
+      label="Off"
+      name="noise_suppression_none"
+      value="0"/>
+  <item
+      label="Low"
+      name="noise_suppression_low"
+      value="1"/>
+  <item
+      label="Moderate"
+      name="noise_suppression_moderate"
+      value="2"/>
+  <item
+      label="High"
+      name="noise_suppression_high"
+      value="3"/>
+  <item
+      label="Max"
+      name="noise_suppression_max"
+      value="4"/>
+  </combo_box>
   <button
    control_name="ShowDeviceSettings"
    follows="left|top"
@@ -485,9 +544,9 @@
    label="Voice Input/Output devices"
    layout="topleft"
    left="20"
-   top_pad="9"
+   top_pad="0"
    name="device_settings_btn"
-   width="230">
+   width="200">
   </button>
     <panel
     layout="topleft"