Fix race condition introduced in r242200 where AwContentsIoThreadClient is queried before a subfram…

Fix race condition introduced in r242200 where AwContentsIoThreadClient is queried before a subframe's entry is added.

This happened because the entry was updated on the UI thread through WebContentsObserver::RenderFrameCreated. The race condition was that the IO thread could dispatch a resource request for that subframe before the map is updated.

The fix is to send a message from the renderer to the IO thread when a subframe is created so that the map is updated before any resouce requests arrive.

BUG=304341
R=mkosiba@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=244521

[mkosiba (inactive), 2014-01-13 16:47:14]

many thanks! LGTM

[jam, 2014-01-13 17:00:19]

On 2014/01/13 16:47:14, mkosiba wrote:
> many thanks! LGTM

np, thanks for finding that race

[jam, 2014-01-13 17:03:42]

TBR Justin for the IPC change. hmm why is there owners check for android_webview
IPC since there is no sandbox there (single-process)?

[jschuh, 2014-01-13 17:22:37]

On 2014/01/13 17:03:42, jam wrote:
> TBR Justin for the IPC change. hmm why is there owners check for
android_webview

lgtm

> IPC since there is no sandbox there (single-process)?

I believe the diff is generated by a script, and the script may not know that.

[jam, 2014-01-13 17:27:40]

On 2014/01/13 17:22:37, Justin Schuh wrote:
> On 2014/01/13 17:03:42, jam wrote:
> > TBR Justin for the IPC change. hmm why is there owners check for
> android_webview
> 
> lgtm
> 
> > IPC since there is no sandbox there (single-process)?
> 
> I believe the diff is generated by a script, and the script may not know that.

What diff do you mean?

[jschuh, 2014-01-13 17:35:40]

On 2014/01/13 17:27:40, jam wrote:
> On 2014/01/13 17:22:37, Justin Schuh wrote:
> > On 2014/01/13 17:03:42, jam wrote:
> > > TBR Justin for the IPC change. hmm why is there owners check for
> > android_webview
> > 
> > lgtm
> > 
> > > IPC since there is no sandbox there (single-process)?
> > 
> > I believe the diff is generated by a script, and the script may not know
that.
> 
> What diff do you mean?

Sorry, by "diff" I meant the CL when we add new owners. I think @tsepez has been
handling that, so I'll ask when he gets in.