Make Platform::queryStorageUsageAndQuota work from worker threads

Make Platform::queryStorageUsageAndQuota work from worker threads

For renderer/worker to browser side:
- Always use thread-local QuotaDispatcher
- Make QuotaDispatcher use ThreadSafeSender to send IPC msgs
- Keep ipc_thread_id:request_id mapping in QuotaMessageFilter

For browser to renderer/worker side:
- Add QuotaMessageFilter to receive quota IPC msgs and
to relay the msgs to the correct thread's QuotaDispatcher

BUG=259660
TEST=no behavioral changes yet (fast/workers/*storagequota* after blink side changes)
R=jam@chromium.org, michaeln@chromium.org

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=214162

[kinuko, 2013-07-23 00:51:45]

Slightly not sure who to ask review for this... jam@, would you be able to take
a look at this one?

[jam, 2013-07-23 16:41:55]

some comments:

-it seems that we don't need to send the thread id across in ICPs. we already
send a request id. can we use that as a key in a map that's store in the child
process side (i.e. in QuotaMessageFilter)?

-if quota_dispatcher_host.cc is servicing IPCs from worker processes, then
having it in content/browser/renderer_host is a bit confusing. can you move it
to content/browser/quota_dispatcher_host.cc?

[kinuko, 2013-07-24 11:25:27]

On 2013/07/23 16:41:55, jam wrote:
> some comments:
> 
> -it seems that we don't need to send the thread id across in ICPs. we already
> send a request id. can we use that as a key in a map that's store in the child
> process side (i.e. in QuotaMessageFilter)?

We can, but since sending and receiving might happen on different threads we
need to protect the map with lock and the map needs to be shared with
thread-safe-refcounted etc. I found sending thread ID over IPC might be easier
(also I found IDB is doing the same, which motivated me to go with this design)

> -if quota_dispatcher_host.cc is servicing IPCs from worker processes, then
> having it in content/browser/renderer_host is a bit confusing. can you move it
> to content/browser/quota_dispatcher_host.cc?

Makes sense, will do.

[kinuko, 2013-07-25 06:29:28]

On 2013/07/24 11:25:27, kinuko wrote:
> On 2013/07/23 16:41:55, jam wrote: 
> > -it seems that we don't need to send the thread id across in ICPs. we
already
> > send a request id. can we use that as a key in a map that's store in the
child
> > process side (i.e. in QuotaMessageFilter)?
> 
> We can, but since sending and receiving might happen on different threads we
> need to protect the map with lock and the map needs to be shared with
> thread-safe-refcounted etc. I found sending thread ID over IPC might be easier
> (also I found IDB is doing the same, which motivated me to go with this
design)

Ok, I changed my mind... I introduced locked map in QuotaMessageFilter, so now
all changes are in the child process side.
Can you take another look?

> > -if quota_dispatcher_host.cc is servicing IPCs from worker processes, then
> > having it in content/browser/renderer_host is a bit confusing. can you move
it
> > to content/browser/quota_dispatcher_host.cc?
> 
> Makes sense, will do.

Done.

[kinuko, 2013-07-26 02:23:33]

Michael, maybe you can review this one too?  This one's more straightforwardly
following IDB model (except for locked map).

[jam, 2013-07-26 04:57:30]

lgtm

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_dispat...
File content/child/quota_dispatcher.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_dispat...
content/child/quota_dispatcher.cc:57: return
WorkerTaskRunner::Instance()->CurrentWorkerId();
nit: adding this function for 2 calls before doesn't seem worth it...

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_messag...
File content/child/quota_message_filter.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_messag...
content/child/quota_message_filter.cc:46: ->OnMessageReceived(msg);
nit: the "->" needs to be on the previous line...

https://codereview.chromium.org/20015002/diff/2001/content/renderer/renderer_...
File content/renderer/renderer_webkitplatformsupport_impl.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/renderer/renderer_...
content/renderer/renderer_webkitplatformsupport_impl.cc:1167: storage_partition,
nit: need to indent again here now

https://codereview.chromium.org/20015002/diff/20001/content/child/quota_messa...
File content/child/quota_message_filter.cc (right):

https://codereview.chromium.org/20015002/diff/20001/content/child/quota_messa...
content/child/quota_message_filter.cc:60: ->OnMessageReceived(msg);
nit: normally lines don't start with "->"

[michaeln, 2013-07-26 20:27:29]

lgtm2

[kinuko, 2013-07-29 04:31:55]

Thanks!

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_dispat...
File content/child/quota_dispatcher.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_dispat...
content/child/quota_dispatcher.cc:57: return
WorkerTaskRunner::Instance()->CurrentWorkerId();
On 2013/07/26 04:57:30, jam wrote:
> nit: adding this function for 2 calls before doesn't seem worth it...

Done.

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_messag...
File content/child/quota_message_filter.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/child/quota_messag...
content/child/quota_message_filter.cc:46: ->OnMessageReceived(msg);
On 2013/07/26 04:57:30, jam wrote:
> nit: the "->" needs to be on the previous line...

Done.

https://codereview.chromium.org/20015002/diff/2001/content/renderer/renderer_...
File content/renderer/renderer_webkitplatformsupport_impl.cc (right):

https://codereview.chromium.org/20015002/diff/2001/content/renderer/renderer_...
content/renderer/renderer_webkitplatformsupport_impl.cc:1167: storage_partition,
On 2013/07/26 04:57:30, jam wrote:
> nit: need to indent again here now

Done.

https://codereview.chromium.org/20015002/diff/20001/content/child/quota_messa...
File content/child/quota_message_filter.cc (right):

https://codereview.chromium.org/20015002/diff/20001/content/child/quota_messa...
content/child/quota_message_filter.cc:60: ->OnMessageReceived(msg);
On 2013/07/26 04:57:31, jam wrote:
> nit: normally lines don't start with "->"

Done.

[kinuko, 2013-07-29 06:22:15]

Committed patchset #3 manually as r214162 (presubmit successful).