Make SpeechRecognition per RenderFrame instead of per RenderView.

content/browser/speech/speech_recognition_dispatcher_host.cc

 // Copyright (c) 2012 The Chromium Authors. All rights reserved.
 // Use of this source code is governed by a BSD-style license that can be
 // found in the LICENSE file.
 
 #include "content/browser/speech/speech_recognition_dispatcher_host.h"
 
 #include "base/bind.h"
 #include "base/command_line.h"
 #include "base/lazy_instance.h"
 #include "content/browser/browser_plugin/browser_plugin_guest.h"
 #include "content/browser/child_process_security_policy_impl.h"
-#include "content/browser/renderer_host/render_view_host_impl.h"
 #include "content/browser/speech/speech_recognition_manager_impl.h"
 #include "content/browser/web_contents/web_contents_impl.h"
 #include "content/common/speech_recognition_messages.h"
+#include "content/public/browser/browser_context.h"
+#include "content/public/browser/browser_thread.h"
+#include "content/public/browser/render_frame_host.h"
 #include "content/public/browser/speech_recognition_manager_delegate.h"
 #include "content/public/browser/speech_recognition_session_config.h"
 #include "content/public/browser/speech_recognition_session_context.h"
-#include "content/public/common/content_switches.h"
+#include "content/public/common/child_process_host.h"
 
 namespace content {
 
+namespace {
+
+void SendMessageToFrameOnUI(int render_process_id,
+                            int render_frame_id,
+                            IPC::Message* message) {
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
+
+  RenderFrameHost* render_frame_host =
+      RenderFrameHost::FromID(render_process_id, render_frame_id);
+  if (render_frame_host)
+    render_frame_host->Send(message);
+}
+
+void SendMessageToFrame(int render_process_id,
+                        int render_frame_id,
+                        IPC::Message* message) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+
+  BrowserThread::PostTask(
+      BrowserThread::UI, FROM_HERE,
+      base::Bind(&SendMessageToFrameOnUI,
+                 render_process_id, render_frame_id, message));
+}
+
+}  // anonymous namespace
+
+using FrameDeletedCallback = base::Callback<void(int)>;
+using WebContentsDeletedCallback = base::Callback<void(WebContents*)>;
+
+class SpeechRecognitionDispatcherHost::FrameDeletedObserver
+    : public WebContentsObserver {
+ public:
+  FrameDeletedObserver(WebContents* web_contents,
+                       FrameDeletedCallback frame_deleted_cb,
+                       WebContentsDeletedCallback web_contents_deleted_cb)
+      : WebContentsObserver(web_contents) {
+  }
+
+  ~FrameDeletedObserver() override {
+  }
+
+ private:
+  void RenderFrameDeleted(RenderFrameHost* render_frame_host) override {
+    frame_deleted_cb_.Run(render_frame_host->GetRoutingID());
+  }
+
+  void WebContentsDestroyed() override {
+    web_contents_deleted_cb_.Run(web_contents());
+  }
+
+  const FrameDeletedCallback frame_deleted_cb_;
+  const WebContentsDeletedCallback web_contents_deleted_cb_;
+};
+
 SpeechRecognitionDispatcherHost::SpeechRecognitionDispatcherHost(
-    int render_process_id,
-    net::URLRequestContextGetter* context_getter)
+    int render_process_id)
     : BrowserMessageFilter(SpeechRecognitionMsgStart),
       render_process_id_(render_process_id),
-      context_getter_(context_getter),
       weak_factory_(this) {
   // Do not add any non-trivial initialization here, instead do it lazily when
   // required (e.g. see the method |SpeechRecognitionManager::GetInstance()|) or
   // add an Init() method.
 }
 
 SpeechRecognitionDispatcherHost::~SpeechRecognitionDispatcherHost() {
-  SpeechRecognitionManager::GetInstance()->AbortAllSessionsForRenderProcess(
-      render_process_id_);
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+  weak_factory_.InvalidateWeakPtrs();
+
+  STLDeleteContainerPairSecondPointers(frame_deleted_observers_.begin(),
+                                       frame_deleted_observers_.end());
 }
 
 base::WeakPtr<SpeechRecognitionDispatcherHost>
 SpeechRecognitionDispatcherHost::AsWeakPtr() {
   return weak_factory_.GetWeakPtr();
 }
 
 void SpeechRecognitionDispatcherHost::OnDestruct() const {
   BrowserThread::DeleteOnIOThread::Destruct(this);
 }
 
+void SpeechRecognitionDispatcherHost::OnChannelClosing() {
+  weak_factory_.InvalidateWeakPtrs();
+}
+
+void SpeechRecognitionDispatcherHost::OverrideThreadForMessage(
+    const IPC::Message& message, BrowserThread::ID* thread) {
+  if (message.type() == SpeechRecognitionHostMsg_StartRequest::ID)
+    *thread = BrowserThread::UI;
+}
+
 bool SpeechRecognitionDispatcherHost::OnMessageReceived(
     const IPC::Message& message) {
   bool handled = true;
+
   IPC_BEGIN_MESSAGE_MAP(SpeechRecognitionDispatcherHost, message)
     IPC_MESSAGE_HANDLER(SpeechRecognitionHostMsg_StartRequest,
                         OnStartRequest)
     IPC_MESSAGE_HANDLER(SpeechRecognitionHostMsg_AbortRequest,
                         OnAbortRequest)
     IPC_MESSAGE_HANDLER(SpeechRecognitionHostMsg_StopCaptureRequest,
                         OnStopCaptureRequest)
     IPC_MESSAGE_HANDLER(SpeechRecognitionHostMsg_AbortAllRequests,
                         OnAbortAllRequests)
     IPC_MESSAGE_UNHANDLED(handled = false)
   IPC_END_MESSAGE_MAP()
+
   return handled;
 }
 
-void SpeechRecognitionDispatcherHost::OverrideThreadForMessage(
-    const IPC::Message& message,
-    BrowserThread::ID* thread) {
-  if (message.type() == SpeechRecognitionHostMsg_StartRequest::ID)
-    *thread = BrowserThread::UI;
-}
-
-void SpeechRecognitionDispatcherHost::OnChannelClosing() {
-  weak_factory_.InvalidateWeakPtrs();
-}
-
 void SpeechRecognitionDispatcherHost::OnStartRequest(
     const SpeechRecognitionHostMsg_StartRequest_Params& params) {
-  SpeechRecognitionHostMsg_StartRequest_Params input_params(params);
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
 
   // Check that the origin specified by the renderer process is one
   // that it is allowed to access.
   if (params.origin_url != "null" &&
       !ChildProcessSecurityPolicyImpl::GetInstance()->CanRequestURL(
           render_process_id_, GURL(params.origin_url))) {
     LOG(ERROR) << "SRDH::OnStartRequest, disallowed origin: "
                << params.origin_url;
     return;
   }
 
   int embedder_render_process_id = 0;
-  int embedder_render_view_id = MSG_ROUTING_NONE;
-  RenderViewHostImpl* render_view_host =
-      RenderViewHostImpl::FromID(render_process_id_, params.render_view_id);
-  if (!render_view_host) {
-    // RVH can be null if the tab was closed while continuous mode speech
-    // recognition was running. This seems to happen on mac.
-    LOG(WARNING) << "SRDH::OnStartRequest, RenderViewHost does not exist";
-    return;
-  }
-  WebContentsImpl* web_contents = static_cast<WebContentsImpl*>(
-      WebContents::FromRenderViewHost(render_view_host));
-  BrowserPluginGuest* guest = web_contents->GetBrowserPluginGuest();
+  int embedder_render_frame_id = MSG_ROUTING_NONE;
+
+  WebContents* web_contents = WebContents::FromRenderFrameHost(
+      RenderFrameHost::FromID(render_process_id_, params.render_frame_id));
+  DCHECK(web_contents);
+
+  BrowserPluginGuest* guest =
+      static_cast<WebContentsImpl*>(web_contents)->GetBrowserPluginGuest();
+
   if (guest) {
     // If the speech API request was from a guest, save the context of the
     // embedder since we will use it to decide permission.
     embedder_render_process_id =
         guest->embedder_web_contents()->GetRenderProcessHost()->GetID();
     DCHECK_NE(embedder_render_process_id, 0);
-    embedder_render_view_id =
-        guest->embedder_web_contents()->GetRenderViewHost()->GetRoutingID();
-    DCHECK_NE(embedder_render_view_id, MSG_ROUTING_NONE);
+    embedder_render_frame_id =
+        guest->embedder_web_contents()->GetMainFrame()->GetRoutingID();
+    DCHECK_NE(embedder_render_frame_id, MSG_ROUTING_NONE);
   }
 
-  // TODO(lazyboy): Check if filter_profanities should use |render_process_id|
-  // instead of |render_process_id_|.
   bool filter_profanities =
       SpeechRecognitionManagerImpl::GetInstance() &&
       SpeechRecognitionManagerImpl::GetInstance()->delegate() &&
       SpeechRecognitionManagerImpl::GetInstance()->delegate()->
           FilterProfanities(render_process_id_);
 
-  // TODO(miu): This is a hack to allow SpeechRecognition to operate with the
-  // MediaStreamManager, which partitions requests per RenderFrame, not per
-  // RenderView.  http://crbug.com/390749
-  const int params_render_frame_id = render_view_host ?
-      render_view_host->GetMainFrame()->GetRoutingID() : MSG_ROUTING_NONE;
-
-  BrowserThread::PostTask(
-      BrowserThread::IO,
-      FROM_HERE,
-      base::Bind(&SpeechRecognitionDispatcherHost::OnStartRequestOnIO,
-                 this,
-                 embedder_render_process_id,
-                 embedder_render_view_id,
-                 input_params,
-                 params_render_frame_id,
-                 filter_profanities));
-}
-
-void SpeechRecognitionDispatcherHost::OnStartRequestOnIO(
-    int embedder_render_process_id,
-    int embedder_render_view_id,
-    const SpeechRecognitionHostMsg_StartRequest_Params& params,
-    int params_render_frame_id,
-    bool filter_profanities) {
   SpeechRecognitionSessionContext context;
   context.context_name = params.origin_url;
   context.render_process_id = render_process_id_;
-  context.render_view_id = params.render_view_id;
-  context.render_frame_id = params_render_frame_id;
+  context.render_frame_id = params.render_frame_id;
   context.embedder_render_process_id = embedder_render_process_id;
-  context.embedder_render_view_id = embedder_render_view_id;
+  context.embedder_render_frame_id = embedder_render_frame_id;
   if (embedder_render_process_id)
-    context.guest_render_view_id = params.render_view_id;
+    context.guest_render_frame_id = params.render_frame_id;
   context.request_id = params.request_id;
 
+  BrowserThread::PostTask(
+      BrowserThread::IO, FROM_HERE,
+      base::Bind(&SpeechRecognitionDispatcherHost::StartSession,
+          base::Unretained(this),
+          params,
+          context,
+          base::Unretained(web_contents->GetBrowserContext()->
+              GetRequestContextForRenderProcess(render_process_id_)),
+          filter_profanities));
+}
+
+void SpeechRecognitionDispatcherHost::StartSession(
+    const SpeechRecognitionHostMsg_StartRequest_Params& params,
+    const SpeechRecognitionSessionContext& context,
+    net::URLRequestContextGetter* url_request_context_getter,
+    bool filter_profanities) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+
   SpeechRecognitionSessionConfig config;
   config.is_legacy_api = false;
   config.language = params.language;
   config.grammars = params.grammars;
   config.max_hypotheses = params.max_hypotheses;
   config.origin_url = params.origin_url;
   config.initial_context = context;
-  config.url_request_context_getter = context_getter_.get();
+  config.url_request_context_getter = url_request_context_getter;
   config.filter_profanities = filter_profanities;
   config.continuous = params.continuous;
   config.interim_results = params.interim_results;
   config.event_listener = AsWeakPtr();
 
   int session_id = SpeechRecognitionManager::GetInstance()->CreateSession(
       config);
   DCHECK_NE(session_id, SpeechRecognitionManager::kSessionIDInvalid);
   SpeechRecognitionManager::GetInstance()->StartSession(session_id);
 }
 
-void SpeechRecognitionDispatcherHost::OnAbortRequest(int render_view_id,
+void SpeechRecognitionDispatcherHost::OnAbortRequest(int render_frame_id,
                                                      int request_id) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+
   int session_id = SpeechRecognitionManager::GetInstance()->GetSession(
-      render_process_id_, render_view_id, request_id);
+      render_process_id_, render_frame_id, request_id);
 
   // The renderer might provide an invalid |request_id| if the session was not
   // started as expected, e.g., due to unsatisfied security requirements.
   if (session_id != SpeechRecognitionManager::kSessionIDInvalid)
     SpeechRecognitionManager::GetInstance()->AbortSession(session_id);
 }
 
-void SpeechRecognitionDispatcherHost::OnAbortAllRequests(int render_view_id) {
-  SpeechRecognitionManager::GetInstance()->AbortAllSessionsForRenderView(
-      render_process_id_, render_view_id);
+void SpeechRecognitionDispatcherHost::OnAbortAllRequests(int render_frame_id) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+
+  SpeechRecognitionManager::GetInstance()->AbortAllSessionsForRenderFrame(
+      render_process_id_, render_frame_id);
 }
 
 void SpeechRecognitionDispatcherHost::OnStopCaptureRequest(
-    int render_view_id, int request_id) {
+    int render_frame_id, int request_id) {
+  DCHECK_CURRENTLY_ON(BrowserThread::IO);
+
   int session_id = SpeechRecognitionManager::GetInstance()->GetSession(
-      render_process_id_, render_view_id, request_id);
+      render_process_id_, render_frame_id, request_id);
 
   // The renderer might provide an invalid |request_id| if the session was not
   // started as expected, e.g., due to unsatisfied security requirements.
   if (session_id != SpeechRecognitionManager::kSessionIDInvalid) {
     SpeechRecognitionManager::GetInstance()->StopAudioCaptureForSession(
         session_id);
   }
 }
 
+void SpeechRecognitionDispatcherHost::CreateObserverFor(
+    WebContents* web_contents) {
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
+
+  if (frame_deleted_observers_.find(web_contents) !=
+      frame_deleted_observers_.end()) {
+    return;
+  }
+
+  frame_deleted_observers_.emplace(
+      web_contents,
+      new FrameDeletedObserver(
+          web_contents,
+          base::Bind(&SpeechRecognitionDispatcherHost::AbortAllRequestsFromUI,
+                     AsWeakPtr()),
+          base::Bind(&SpeechRecognitionDispatcherHost::RemoveObserverFor,
+                     AsWeakPtr())));
+}
+
+void SpeechRecognitionDispatcherHost::RemoveObserverFor(
+    WebContents* web_contents) {
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
+
+  FrameDeletedObservers::iterator it =
+      frame_deleted_observers_.find(web_contents);
+  if (it == frame_deleted_observers_.end())
+    return;
+
+  delete it->second;
+  frame_deleted_observers_.erase(web_contents);
+}
+
+void SpeechRecognitionDispatcherHost::AbortAllRequestsFromUI(
+    int render_frame_id) {
+  DCHECK_CURRENTLY_ON(BrowserThread::UI);
+
+  BrowserThread::PostTask(
+    BrowserThread::IO, FROM_HERE,
+    base::Bind(&SpeechRecognitionDispatcherHost::OnAbortAllRequests,
+               this,
+               render_frame_id));
+}
+
 // -------- SpeechRecognitionEventListener interface implementation -----------
 
 void SpeechRecognitionDispatcherHost::OnRecognitionStart(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_Started(context.render_view_id,
-                                        context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_Started(context.render_frame_id,
+                                                      context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnAudioStart(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_AudioStarted(context.render_view_id,
-                                             context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_AudioStarted(
+                         context.render_frame_id, context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnSoundStart(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_SoundStarted(context.render_view_id,
-                                             context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_SoundStarted(
+                         context.render_frame_id, context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnSoundEnd(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_SoundEnded(context.render_view_id,
-                                           context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_SoundEnded(
+                         context.render_frame_id, context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnAudioEnd(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_AudioEnded(context.render_view_id,
-                                           context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_AudioEnded(
+                         context.render_frame_id, context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnRecognitionEnd(int session_id) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_Ended(context.render_view_id,
-                                      context.request_id));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_Ended(
+                         context.render_frame_id, context.request_id));
 }
 
 void SpeechRecognitionDispatcherHost::OnRecognitionResults(
     int session_id,
     const SpeechRecognitionResults& results) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_ResultRetrieved(context.render_view_id,
-                                                context.request_id,
-                                                results));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_ResultRetrieved(
+                         context.render_frame_id, context.request_id, results));
 }
 
 void SpeechRecognitionDispatcherHost::OnRecognitionError(
     int session_id,
     const SpeechRecognitionError& error) {
   const SpeechRecognitionSessionContext& context =
       SpeechRecognitionManager::GetInstance()->GetSessionContext(session_id);
-  Send(new SpeechRecognitionMsg_ErrorOccurred(context.render_view_id,
-                                              context.request_id,
-                                              error));
+
+  SendMessageToFrame(context.render_process_id, context.render_frame_id,
+                     new SpeechRecognitionMsg_ErrorOccurred(
+                         context.render_frame_id, context.request_id, error));
 }
 
 // The events below are currently not used by speech JS APIs implementation.
 void SpeechRecognitionDispatcherHost::OnAudioLevelsChange(int session_id,
                                                           float volume,
                                                           float noise_volume) {
 }
 
 void SpeechRecognitionDispatcherHost::OnEnvironmentEstimationComplete(
     int session_id) {
 }
 
 }  // namespace content