Reland "Separate client and server pushed streams limits."

Reland "Separate client and server pushed streams limits."

https://codereview.chromium.org/367963003/

Fix use-after-free error

BUG=170544, 377538
R=jgraettinger@chromium.org

TEST=SpdySessionTest.PushedStreamShouldNotCountToClientConcurrencyLimit
SpdySessionTest.RejectPushedStreamExceedingConcurrencyLimit
SpdySessionTest.IgnoreReservedRemoteStreamsCount

Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=282889

[baranovich, 2014-07-08 11:16:30]

This relands https://codereview.chromium.org/367963003 which was reverted due to
use-after-free error in SpdySession::OnInitialResponseHeadersReceived()

First patchset is original CL, second adds early exit to
SpdySession::OnInitialResponseHeadersReceived() if
stream->OnInitialResponseHeadersReceived() returned error.

[Johnny, 2014-07-08 16:12:24]

lgtm.

FYI you can update the original CL with a fixing patch-set, and re-land it
there.

I'm assuming you've already done this, but just in case, you can produce a local
build instrumented with ASAN and run SPDY tests using that:

http://www.chromium.org/developers/testing/addresssanitizer

I'll usually run tests through ASAN just because the CQ unfortunately doesn't
catch these failures.

[baranovich, 2014-07-08 21:29:17]

On 2014/07/08 16:12:24, Johnny wrote:
> lgtm.
> 
> FYI you can update the original CL with a fixing patch-set, and re-land it
> there.
> 
> I'm assuming you've already done this, but just in case, you can produce a
local
> build instrumented with ASAN and run SPDY tests using that:
> 
> http://www.chromium.org/developers/testing/addresssanitizer
> 
> I'll usually run tests through ASAN just because the CQ unfortunately doesn't
> catch these failures.

Thanks for the tips!

While fixing these errors, found a bug in push stream counting.
|num_active_pushed_streams_| should be incremented always, since it is always
decremented in DestroyStream. Added a unit-test for it
(SpdySessionTest.CancelReservedStreamOnHeadersReceived).
Could you please take a look once more?

PS Perhaps, it would be good idea to create more sophisticated stream container
class which will encapsulate all that stream arithmetics and test it separately.
That would help to clean SpdySession code a little bit. What do you think?

[baranovich, 2014-07-13 21:50:55]

The CQ bit was checked by baranovich@yandex-team.ru

[commit-bot: I haz the power, 2014-07-13 21:51:59]

CQ is trying da patch. Follow status at

https://chromium-status.appspot.com/cq/baranovich@yandex-team.ru/371273003/40001

[commit-bot: I haz the power, 2014-07-14 02:21:59]

Change committed as 282889