Fix (attempt) for crbug.com/459115.

Fix: Don't access MessagePortService in the UI thread.

WebContentsImpl::RouteMessageEvent is called in the UI thread, and it called
MessagePortService::UpdateMessagePort which assumed to be called in the IO
thread. Oops.

BUG=459115
Committed: https://crrev.com/b5fe7ad39469790b5b49271add893c9613c0ce66
Cr-Commit-Position: refs/heads/master@{#316912}

[Mike West, 2015-02-18 14:20:55]

mkwst@chromium.org changed reviewers:
+ creis@chromium.org, mkwst@chromium.org

[Mike West, 2015-02-18 14:20:56]

This LGTM. Moving the IPC should be safe.

I'd suggest asking creis@ to take a look, just in case there are site isolation
implications. Conveniently, he's also a //content OWNER. :)

https://codereview.chromium.org/940493002/diff/1/content/browser/web_contents...
File content/browser/web_contents/web_contents_impl.cc (right):

https://codereview.chromium.org/940493002/diff/1/content/browser/web_contents...
content/browser/web_contents/web_contents_impl.cc:4023: //
ViewMsg_PostMessageEvent after it's done with the updating.
Can you add something to this comment about the lifetime issue that led you to
this solution?

[marja, 2015-02-18 15:02:19]

New patchsets have been uploaded after l-g-t-m from mkwst@chromium.org

[marja, 2015-02-18 15:04:04]

creis, ptal

mkswst, thx for review

https://codereview.chromium.org/940493002/diff/1/content/browser/web_contents...
File content/browser/web_contents/web_contents_impl.cc (right):

https://codereview.chromium.org/940493002/diff/1/content/browser/web_contents...
content/browser/web_contents/web_contents_impl.cc:4023: //
ViewMsg_PostMessageEvent after it's done with the updating.
On 2015/02/18 14:20:55, Mike West wrote:
> Can you add something to this comment about the lifetime issue that led you to
> this solution?

Done.

[Mike West, 2015-02-18 15:07:59]

Could you also change the CL description to have a title like "Moving this IPC
so we don't do crazy cross-thread things that we shouldn't." instead of "Fix bug
XXX"? :)

[marja, 2015-02-18 15:09:43]

On 2015/02/18 15:07:59, Mike West wrote:
> Could you also change the CL description to have a title like "Moving this IPC
> so we don't do crazy cross-thread things that we shouldn't." instead of "Fix
bug
> XXX"? :)

Done

[Charlie Reis, 2015-02-18 19:55:35]

Thanks, LGTM.  (I don't see any issue with the thread hop, since postMessage is
already asynchronous.  Also, makes sense to send the IPC straight from the IO
thread if we need that hop.)

https://codereview.chromium.org/940493002/diff/20001/content/browser/web_cont...
File content/browser/web_contents/web_contents_impl.cc (right):

https://codereview.chromium.org/940493002/diff/20001/content/browser/web_cont...
content/browser/web_contents/web_contents_impl.cc:4029:
static_cast<RenderProcessHostImpl*>(GetRenderProcessHost())
This will need to be updated for OOPIF, but it's fine for now.  (Most of this
method will also need to be updated, which we're starting to track in
https://crbug.com/389721.)

When that time comes, we'll need to use the RenderProcessHost of the target
frame to get the filter, since filters are process-specific.

[marja, 2015-02-18 20:43:27]

The CQ bit was checked by marja@chromium.org

[marja, 2015-02-18 20:43:31]

The CQ bit was unchecked by marja@chromium.org

[marja, 2015-02-18 20:43:35]

The CQ bit was checked by marja@chromium.org

[commit-bot: I haz the power, 2015-02-18 20:44:00]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/940493002/20001

[commit-bot: I haz the power, 2015-02-18 23:09:00]

Committed patchset #2 (id:20001)

[commit-bot: I haz the power, 2015-02-18 23:09:46]

Patchset 2 (id:??) landed as
https://crrev.com/b5fe7ad39469790b5b49271add893c9613c0ce66
Cr-Commit-Position: refs/heads/master@{#316912}