Add waitable event for SyncMessageFilter init

Hold message in SyncMessageFilter before init

SyncMessageFilter::send will fail before init (OnFilterAdded)
on IO thread but provides no init signal externally. This can
cause races during start up.

Instead, before init, hold any messages in a queue and
flush the queue in init instead.

BUG=514929
Committed: https://crrev.com/9bd6368468d43c5eb586a108233a2cf4ab50c9ce
Cr-Commit-Position: refs/heads/master@{#340930}

[boliu, 2015-07-29 02:14:06]

boliu@chromium.org changed reviewers:
+ jam@chromium.org

[boliu, 2015-07-29 02:14:07]

jam: could you take a look and the ipc changes, and see if this direction make
sense at all? And if not, any advice on a better direction for?

Problem is described in linked bug. Essentially need a way to wait for init on
IO thread, so first sync ipc message is not dropped.

[piman, 2015-07-29 02:24:04]

On 2015/07/29 02:14:07, boliu wrote:
> jam: could you take a look and the ipc changes, and see if this direction make
> sense at all? And if not, any advice on a better direction for?
> 
> Problem is described in linked bug. Essentially need a way to wait for init on
> IO thread, so first sync ipc message is not dropped.

Can we leave GpuChannelHost out of it? Have IPCSyncMessageFilter queue pending
messages if no io thread / not connected yet, and have OnFilterAdded send them.
The existing WaitableEvent should take care of sync messages.

[boliu, 2015-07-29 02:28:06]

On 2015/07/29 02:24:04, piman (Very slow to review) wrote:
> On 2015/07/29 02:14:07, boliu wrote:
> > jam: could you take a look and the ipc changes, and see if this direction
make
> > sense at all? And if not, any advice on a better direction for?
> > 
> > Problem is described in linked bug. Essentially need a way to wait for init
on
> > IO thread, so first sync ipc message is not dropped.
> 
> Can we leave GpuChannelHost out of it? Have IPCSyncMessageFilter queue pending
> messages if no io thread / not connected yet, and have OnFilterAdded send
them.
> The existing WaitableEvent should take care of sync messages.

Hmm, yeah. That should work. Will try it tomorrow.

[boliu, 2015-07-29 17:13:17]

On 2015/07/29 02:28:06, boliu wrote:
> On 2015/07/29 02:24:04, piman (Very slow to review) wrote:
> > On 2015/07/29 02:14:07, boliu wrote:
> > > jam: could you take a look and the ipc changes, and see if this direction
> make
> > > sense at all? And if not, any advice on a better direction for?
> > > 
> > > Problem is described in linked bug. Essentially need a way to wait for
init
> on
> > > IO thread, so first sync ipc message is not dropped.
> > 
> > Can we leave GpuChannelHost out of it? Have IPCSyncMessageFilter queue
pending
> > messages if no io thread / not connected yet, and have OnFilterAdded send
> them.
> > The existing WaitableEvent should take care of sync messages.
> 
> Hmm, yeah. That should work. Will try it tomorrow.

Trybots are happy with PS3. PTAL

[jam, 2015-07-29 18:27:59]

lgtm

[boliu, 2015-07-29 18:29:02]

The CQ bit was checked by boliu@chromium.org

[commit-bot: I haz the power, 2015-07-29 18:30:15]

CQ is trying da patch. Follow status at
 https://chromium-cq-status.appspot.com/patch-status/1261253002/40001
View timeline at
 https://chromium-cq-status.appspot.com/patch-timeline/1261253002/40001

[commit-bot: I haz the power, 2015-07-29 18:38:16]

Committed patchset #3 (id:40001)

[commit-bot: I haz the power, 2015-07-29 18:39:22]

Patchset 3 (id:??) landed as
https://crrev.com/9bd6368468d43c5eb586a108233a2cf4ab50c9ce
Cr-Commit-Position: refs/heads/master@{#340930}