[BlobAsync] Fixed race between IPC messages and IO task queue

[BlobAsync] Fixed race between IPC messages and IO task queue

The message responses to our blob start message are dispatched on the IO
thread but they're not added to the task queue. So a race can happen
where OnMemoryRequest is called before we store the consolidation in our
map.

I moved the StartBuildingBlob message - which signals the browser to
start requesting memory from the renderer - to our IO task which stores
our blob data. This eliminates the race case with a small sacrifice of
construction speed.

In the future, I'd like to be able to still dispatch both IPC messages
right away to keep the fast path, and correctly handle the race case.

BUG=612358
Committed: https://crrev.com/65a8fb95a8c71a7be8515f5e1c8125b04a477716
Cr-Commit-Position: refs/heads/master@{#394549}

[dmurph, 2016-05-18 19:56:26]

dmurph@chromium.org changed reviewers:
+ michaeln@chromium.org

[dmurph, 2016-05-18 19:56:26]

Hey, can you PTAL? This is killing docs.

[michaeln, 2016-05-18 20:00:15]

lgtm!

[dmurph, 2016-05-18 20:05:31]

The CQ bit was checked by dmurph@chromium.org

[commit-bot: I haz the power, 2016-05-18 20:06:03]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1988293002/1
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1988293002/1

[commit-bot: I haz the power, 2016-05-18 21:14:08]

Committed patchset #1 (id:1)

[commit-bot: I haz the power, 2016-05-18 21:16:22]

Description was changed from

==========
[BlobAsync] Fixed race between IPC messages and IO task queue

The message responses to our blob start message are dispatched on the IO
thread but they're not added to the task queue. So a race can happen
where OnMemoryRequest is called before we store the consolidation in our
map.

I moved the StartBuildingBlob message - which signals the browser to
start requesting memory from the renderer - to our IO task which stores
our blob data. This eliminates the race case with a small sacrifice of
construction speed.

In the future, I'd like to be able to still dispatch both IPC messages
right away to keep the fast path, and correctly handle the race case.

BUG=612358
==========

to

==========
[BlobAsync] Fixed race between IPC messages and IO task queue

The message responses to our blob start message are dispatched on the IO
thread but they're not added to the task queue. So a race can happen
where OnMemoryRequest is called before we store the consolidation in our
map.

I moved the StartBuildingBlob message - which signals the browser to
start requesting memory from the renderer - to our IO task which stores
our blob data. This eliminates the race case with a small sacrifice of
construction speed.

In the future, I'd like to be able to still dispatch both IPC messages
right away to keep the fast path, and correctly handle the race case.

BUG=612358

Committed: https://crrev.com/65a8fb95a8c71a7be8515f5e1c8125b04a477716
Cr-Commit-Position: refs/heads/master@{#394549}
==========

[commit-bot: I haz the power, 2016-05-18 21:16:23]

Patchset 1 (id:??) landed as
https://crrev.com/65a8fb95a8c71a7be8515f5e1c8125b04a477716
Cr-Commit-Position: refs/heads/master@{#394549}

[Sami, 2016-05-19 11:40:47]

A revert of this CL (patchset #1 id:1) has been created in
https://codereview.chromium.org/1997623002/ by skyostil@chromium.org.

The reason for reverting is: The oortonline benchmark started failing to load
blobs around the time this landed, so I'll do a speculative revert. Failure log:

https://build.chromium.org/p/chromium.perf/builders/Mac%20Retina%20Perf%20%28....