Integrating the QuicStreamFactory into the network stack.

Integrating the QuicStreamFactory into the network stack.

Fix static initialized in quic_utils.cc

Reverted: 173321
Initially landed: 173311

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=173367

[Ryan Hamilton, 2012-11-19 18:43:03]

This adds a command line flag to make QUIC work.  I expect that landing this
will get noticed, but I could be wrong.

[willchan no longer on Chromium, 2012-11-22 03:04:53]

https://codereview.chromium.org/11416058/diff/5002/net/http/http_network_layer.h
File net/http/http_network_layer.h (right):

https://codereview.chromium.org/11416058/diff/5002/net/http/http_network_laye...
net/http/http_network_layer.h:52: static void EnableQuic(const std::string&
port);
Sorry, I'm trying to draw a hardline and eliminate all globals. We're passing
them as parameters now:
https://code.google.com/searchframe#OAMlx_jo-ck/src/chrome/browser/io_thread.....

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
File net/http/http_stream_factory_impl_job.cc (right):

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
net/http/http_stream_factory_impl_job.cc:645: return force_quic_port_ ==
origin_.port() && proxy_info_.is_direct();
FWIW, I read this as forcing all requests to go to a certain port for QUIC, not
forcing all http requests to a certain port to go over QUIC. Of course,
clarifying this would require a long ugly flag name. Perhaps it's worth it since
I do think we'l have the flag for awhile.

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
net/http/http_stream_factory_impl_job.cc:933: HostPortProxyPair(origin_,
proxy_server), net_log_, &quic_stream_,
How does cancellation work? Hm, everyone else just uses io_callback_ and it uses
base::Unretained(this). I need to page this code back into memory, but this all
seems reliant on Job never getting canceled.

I guess we never cancel jobs, even when the request goes away, since a new
request may come in that uses it. Except during shutdown since we must cancel
everything. This means shutdown order is probably very important, so I should
make sure that's done correctly for all this stuff.

[Ryan Hamilton, 2012-11-22 23:08:40]

https://codereview.chromium.org/11416058/diff/5002/net/http/http_network_layer.h
File net/http/http_network_layer.h (right):

https://codereview.chromium.org/11416058/diff/5002/net/http/http_network_laye...
net/http/http_network_layer.h:52: static void EnableQuic(const std::string&
port);
On 2012/11/22 03:04:53, willchan wrote:
> Sorry, I'm trying to draw a hardline and eliminate all globals. We're passing
> them as parameters now:
>
https://code.google.com/searchframe#OAMlx_jo-ck/src/chrome/browser/io_thread.....

Done.  I'd also be happy to audit the use of statics in net/ and replace then
with Params.

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
File net/http/http_stream_factory_impl_job.cc (right):

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
net/http/http_stream_factory_impl_job.cc:645: return force_quic_port_ ==
origin_.port() && proxy_info_.is_direct();
On 2012/11/22 03:04:53, willchan wrote:
> FWIW, I read this as forcing all requests to go to a certain port for QUIC,
not
> forcing all http requests to a certain port to go over QUIC. Of course,
> clarifying this would require a long ugly flag name. Perhaps it's worth it
since
> I do think we'l have the flag for awhile.

I hear ya.  Naming is hard.  What about --force-quic-for-requests-to-port=6121
(and a variable named force_quic_for_request_to_port)

Or --force-quic-when-origin-has-port ?

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
net/http/http_stream_factory_impl_job.cc:933: HostPortProxyPair(origin_,
proxy_server), net_log_, &quic_stream_,
On 2012/11/22 03:04:53, willchan wrote:
> How does cancellation work? Hm, everyone else just uses io_callback_ and it
uses
> base::Unretained(this). I need to page this code back into memory, but this
all
> seems reliant on Job never getting canceled.
> 
> I guess we never cancel jobs, even when the request goes away, since a new
> request may come in that uses it. Except during shutdown since we must cancel
> everything. This means shutdown order is probably very important, so I should
> make sure that's done correctly for all this stuff.

Yes, I think your reading is correct.  I *think* the shutdown ordering also
works, but I admit that this is an area full of subtlety.

[willchan no longer on Chromium, 2012-12-11 02:02:08]

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
File net/http/http_stream_factory_impl_job.cc (right):

https://codereview.chromium.org/11416058/diff/5002/net/http/http_stream_facto...
net/http/http_stream_factory_impl_job.cc:645: return force_quic_port_ ==
origin_.port() && proxy_info_.is_direct();
On 2012/11/22 23:08:40, Ryan Hamilton wrote:
> On 2012/11/22 03:04:53, willchan wrote:
> > FWIW, I read this as forcing all requests to go to a certain port for QUIC,
> not
> > forcing all http requests to a certain port to go over QUIC. Of course,
> > clarifying this would require a long ugly flag name. Perhaps it's worth it
> since
> > I do think we'l have the flag for awhile.
> 
> I hear ya.  Naming is hard.  What about --force-quic-for-requests-to-port=6121
> (and a variable named force_quic_for_request_to_port)
> 
> Or --force-quic-when-origin-has-port ?

I prefer the latter, but both are better.

https://codereview.chromium.org/11416058/diff/14001/net/http/http_network_ses...
File net/http/http_network_session.h (right):

https://codereview.chromium.org/11416058/diff/14001/net/http/http_network_ses...
net/http/http_network_session.h:180: QuicStreamFactory quic_stream_factory_;
This needs to be one above the SpdySessionPool. The reason is because you could
imagine chromium<=>quicproxy that has a stream over which we do a quic connect
to an endpoint, and then over that quic bytestream we do TLS-NPN=>SPDY.
Therefore, SPDY sits on top of quic, so we need to make sure the SPDY stuff gets
created after and destroyed before quic. Therefore, quic_stream_factory_ must be
before spdy_session_pool_.

https://codereview.chromium.org/11416058/diff/14001/net/http/http_stream_fact...
File net/http/http_stream_factory_impl_job.cc (right):

https://codereview.chromium.org/11416058/diff/14001/net/http/http_stream_fact...
net/http/http_stream_factory_impl_job.cc:932: return
session_->quic_stream_factory()->Create(
This is blocked on the other CL I guess.

[Ryan Hamilton, 2012-12-11 20:03:05]

https://codereview.chromium.org/11416058/diff/14001/net/http/http_network_ses...
File net/http/http_network_session.h (right):

https://codereview.chromium.org/11416058/diff/14001/net/http/http_network_ses...
net/http/http_network_session.h:180: QuicStreamFactory quic_stream_factory_;
On 2012/12/11 02:02:08, willchan wrote:
> This needs to be one above the SpdySessionPool. The reason is because you
could
> imagine chromium<=>quicproxy that has a stream over which we do a quic connect
> to an endpoint, and then over that quic bytestream we do TLS-NPN=>SPDY.
> Therefore, SPDY sits on top of quic, so we need to make sure the SPDY stuff
gets
> created after and destroyed before quic. Therefore, quic_stream_factory_ must
be
> before spdy_session_pool_.

Done.

https://codereview.chromium.org/11416058/diff/14001/net/http/http_stream_fact...
File net/http/http_stream_factory_impl_job.cc (right):

https://codereview.chromium.org/11416058/diff/14001/net/http/http_stream_fact...
net/http/http_stream_factory_impl_job.cc:932: return
session_->quic_stream_factory()->Create(
On 2012/12/11 02:02:08, willchan wrote:
> This is blocked on the other CL I guess.

Indeed.  (Though I've rebased to to integrate the changes in that CL).

[Ryan Hamilton, 2012-12-11 22:11:04]

Ok, rebased on top of the changes from the earlier CL.

[Ryan Hamilton, 2012-12-13 16:42:10]

ping

[willchan no longer on Chromium, 2012-12-13 19:49:59]

lgtm

[Ryan Hamilton, 2012-12-13 19:59:26]

thakis: can you review the chrome/ changes?

[Nico, 2012-12-13 20:09:59]

I don't like the flag name, but if Will likes it lgtm I guess.

https://codereview.chromium.org/11416058/diff/31001/chrome/browser/io_thread.h
File chrome/browser/io_thread.h (right):

https://codereview.chromium.org/11416058/diff/31001/chrome/browser/io_thread....
chrome/browser/io_thread.h:161: Optional<uint16>
force_quic_when_origin_has_port;
nit: The name suggests that this is a bool. Instead:
origin_port_to_force_quic_on_ or something.

https://codereview.chromium.org/11416058/diff/31001/chrome/common/chrome_swit...
File chrome/common/chrome_switches.h (right):

https://codereview.chromium.org/11416058/diff/31001/chrome/common/chrome_swit...
chrome/common/chrome_switches.h:200: extern const char
kForceQuicWhenOriginHasPort[];
Same here.

[Ryan Hamilton, 2012-12-14 02:45:46]

Thanks for the quic review!

https://codereview.chromium.org/11416058/diff/31001/chrome/browser/io_thread.h
File chrome/browser/io_thread.h (right):

https://codereview.chromium.org/11416058/diff/31001/chrome/browser/io_thread....
chrome/browser/io_thread.h:161: Optional<uint16>
force_quic_when_origin_has_port;
On 2012/12/13 20:10:00, Nico wrote:
> nit: The name suggests that this is a bool. Instead:
> origin_port_to_force_quic_on_ or something.

Done.

https://codereview.chromium.org/11416058/diff/31001/chrome/common/chrome_swit...
File chrome/common/chrome_switches.h (right):

https://codereview.chromium.org/11416058/diff/31001/chrome/common/chrome_swit...
chrome/common/chrome_switches.h:200: extern const char
kForceQuicWhenOriginHasPort[];
On 2012/12/13 20:10:00, Nico wrote:
> Same here.

Done.  And elsewhere.

[commit-bot: I haz the power, 2012-12-14 02:59:08]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/28002

[commit-bot: I haz the power, 2012-12-14 03:24:46]

Retried try job too often on ios_dbg_simulator for step(s) net_unittests

[commit-bot: I haz the power, 2012-12-14 03:55:34]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/35004

[commit-bot: I haz the power, 2012-12-14 04:40:12]

Sorry for I got bad news for ya.
Compile failed with a clobber build on win_rel.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...
Your code is likely broken or HEAD is junk. Please ensure your
code is not broken then alert the build sheriffs.
Look at the try server FAQ for more details.

[commit-bot: I haz the power, 2012-12-14 05:33:11]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/35004

[commit-bot: I haz the power, 2012-12-14 05:44:20]

Sorry for I got bad news for ya.
Compile failed with a clobber build on win_rel.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win_rel&nu...
Your code is likely broken or HEAD is junk. Please ensure your
code is not broken then alert the build sheriffs.
Look at the try server FAQ for more details.

[commit-bot: I haz the power, 2012-12-14 15:15:21]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/35004

[commit-bot: I haz the power, 2012-12-14 15:40:56]

Sorry for I got bad news for ya.
Compile failed with a clobber build on win.
http://build.chromium.org/p/tryserver.chromium/buildstatus?builder=win&number...
Your code is likely broken or HEAD is junk. Please ensure your
code is not broken then alert the build sheriffs.
Look at the try server FAQ for more details.

[commit-bot: I haz the power, 2012-12-14 23:49:52]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/43001

[commit-bot: I haz the power, 2012-12-15 17:14:01]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/52002

[commit-bot: I haz the power, 2012-12-15 20:17:35]

Change committed as 173311

[commit-bot: I haz the power, 2012-12-16 01:23:24]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/11416058/55002

[commit-bot: I haz the power, 2012-12-16 04:53:27]

Change committed as 173367