Enable swapping a frame back in to its parent process

content/browser/frame_host/frame_tree.cc

 // Copyright 2013 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/frame_host/frame_tree.h"
 
 #include <queue>
 
 #include "base/bind.h"
 #include "base/callback.h"
@@ skipping 41 matching lines @@
   return true;
 }
 
 bool CreateProxyForSiteInstance(FrameTreeNode* source_node,
                                 const scoped_refptr<SiteInstance>& instance,
                                 FrameTreeNode* node) {
   // Skip the node that initiated the creation.
   if (source_node == node)
     return true;
 
+  // If a given frame host has the same SiteInstance, no proxy is needed.
+  if (node->render_manager()->current_frame_host()->GetSiteInstance() ==

[Nate Chapin, 2014/09/30 21:44:47]

If we're swapping a LocalFrame in place of a RemoteFrame, there is at least one
LocalFrame in the target renderer. Don't create a proxy if one isn't needed.

[Charlie Reis, 2014/09/30 23:02:43]

Hmm, it feels wrong to be calling this function on nodes that already have a
frame in this SiteInstance, but maybe that's how it's written.  Nasko, do you
think there's a cleaner way to handle this?

[nasko, 2014/10/08 16:42:58]

CreateRenderFrameProxy already takes care of the case where there is a proxy for
this SiteInstance.
We shouldn't be creating a proxy for the SiteInstance of the current RFH. What
code path triggers that?

[Nate Chapin, 2014/10/08 20:30:05]

CreateRenderFrameProxy indeed correctly handles when a proxy exists for this
SiteInstance, but it starts by CHECKing that we never ask to create a proxy in
the same SiteInstance that we're navigating in. I guess that CHECK is no longer
valid?

[Charlie Reis, 2014/10/08 20:49:50]

I keep hitting that CHECK as well.  I think Nasko has a patch in progress to fix
it, blocked on some UaF issues.

[nasko, 2014/10/08 20:56:44]

I think the CHECK should be there and is valid. Charlie is correct that I have a
CL in place for that. See if patching it in will help:
https://codereview.chromium.org/600483003/

+      instance)
+    return true;
+
   node->render_manager()->CreateRenderFrameProxy(instance.get());
   return true;
 }
 
 }  // namespace
 
 FrameTree::FrameTree(Navigator* navigator,
                      RenderFrameHostDelegate* render_frame_delegate,
                      RenderViewHostDelegate* render_view_delegate,
                      RenderWidgetHostDelegate* render_widget_delegate,
@@ skipping 251 matching lines @@
         rvh->Shutdown();
         render_view_host_pending_shutdown_map_.erase(multi_iter);
       }
       break;
     }
     CHECK(render_view_host_found);
   }
 }
 
 }  // namespace content