[SPDY] Avoid leaking bytes from the session flow control send window

[SPDY] Avoid leaking bytes from the session flow control send window

Add a ConsumeSource parameter to SpdyBuffer::ConsumeCallback. Use
it to detect when a DATA frame to be written is dropped before it
is written.

Put all the flow control functions together and clean them up a bit.

BUG=176592
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=194851

[akalin, 2013-04-18 01:18:54]

+rch for review

[Ryan Hamilton, 2013-04-18 03:44:37]

lgtm

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.cc#n...
net/spdy/spdy_session.cc:2206: // an error and will be closing the session.
If we're going to tear down the session, is there any harm in Increasing the
send window?

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h
File net/spdy/spdy_session.h (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h#ne...
net/spdy/spdy_session.h:662: void OnWriteBufferConsume(int32 payload_size,
nit: OnWriteBufferConsumed? (trailing "d")

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h#ne...
net/spdy/spdy_session.h:690: void OnReadBufferConsume(size_t consume_size,
nit: OnReadBufferConsumed?

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session_spdy...
File net/spdy/spdy_session_spdy3_unittest.cc (left):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session_spdy...
net/spdy/spdy_session_spdy3_unittest.cc:2108: TEST_F(SpdySessionSpdy3Test,
IncreaseRecvWindowSize) {
Is this test no longer needed?

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.cc
File net/spdy/spdy_stream.cc (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.cc#ne...
net/spdy/spdy_stream.cc:268: // an error and will be closing the stream.
Well, if the stream is going to be town down, there's no harm in increasing the
window right?

That being said, Why don't we need to increase the window size when the source
is CONSUME?  (Does that happen somewhere else?)  Oh, right.  That should happen
when the peer sends a WINDOW_UPDATE.  Can you add a comment?  In hindsight, this
is obvious but I was confused.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h
File net/spdy/spdy_stream.h (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h#new...
net/spdy/spdy_stream.h:168: void OnWriteBufferConsume(int32 payload_size,
ditto

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h#new...
net/spdy/spdy_stream.h:198: void OnReadBufferConsume(size_t consume_size,
ditto

[akalin, 2013-04-18 06:18:42]

Committing after trybots

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.cc
File net/spdy/spdy_session.cc (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.cc#n...
net/spdy/spdy_session.cc:2206: // an error and will be closing the session.
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> If we're going to tear down the session, is there any harm in Increasing the
> send window?

I guess not.  Done, added comment.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h
File net/spdy/spdy_session.h (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h#ne...
net/spdy/spdy_session.h:662: void OnWriteBufferConsume(int32 payload_size,
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> nit: OnWriteBufferConsumed? (trailing "d")

Done.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session.h#ne...
net/spdy/spdy_session.h:690: void OnReadBufferConsume(size_t consume_size,
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> nit: OnReadBufferConsumed?

Done.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session_spdy...
File net/spdy/spdy_session_spdy3_unittest.cc (left):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_session_spdy...
net/spdy/spdy_session_spdy3_unittest.cc:2108: TEST_F(SpdySessionSpdy3Test,
IncreaseRecvWindowSize) {
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> Is this test no longer needed?

That's right.  I changed IncreaseRecvWindowSize so that it must not be called
with plain SPDY/3 (as opposed to doing nothing). This test was testing the
do-nothing part, so it's not needed anymore. (A death test isn't worth it.)

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.cc
File net/spdy/spdy_stream.cc (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.cc#ne...
net/spdy/spdy_stream.cc:268: // an error and will be closing the stream.
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> Well, if the stream is going to be town down, there's no harm in increasing
the
> window right?
> 
> That being said, Why don't we need to increase the window size when the source
> is CONSUME?  (Does that happen somewhere else?)  Oh, right.  That should
happen
> when the peer sends a WINDOW_UPDATE.  Can you add a comment?  In hindsight,
this
> is obvious but I was confused.

Done.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h
File net/spdy/spdy_stream.h (right):

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h#new...
net/spdy/spdy_stream.h:168: void OnWriteBufferConsume(int32 payload_size,
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> ditto

Done.

https://codereview.chromium.org/14188025/diff/7001/net/spdy/spdy_stream.h#new...
net/spdy/spdy_stream.h:198: void OnReadBufferConsume(size_t consume_size,
On 2013/04/18 03:44:37, Ryan Hamilton wrote:
> ditto

Done.

[commit-bot: I haz the power, 2013-04-18 06:25:21]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/[email protected]/14188025/20001

[commit-bot: I haz the power, 2013-04-18 08:32:00]

Change committed as 194851