Add a CRYPTO_BUFFER_POOL to SSLClientSocketImpl.

Add a CRYPTO_BUFFER_POOL to SSLClientSocketImpl.

In anticipation of using it for net::X509Certificate in the future, the
pool lives somewhere in net/cert. Until the rest of the work to remove
switch to CRYPTO_BUFFERs done, this will not be as effective as it could
be.

But, as part of that transition, we have already started using
CRYPTO_BUFFERs in parts of the stack. This allows that portion to be
deduplicated. The (more expensive) portion which isn't deduplicated will
be fixed by removing it altogether.

BUG=671420
Review-Url: https://codereview.chromium.org/2682413003
Cr-Commit-Position: refs/heads/master@{#449732}
Committed: https://chromium.googlesource.com/chromium/src/+/bf0fcf111ceab0a9143d75282ced0d4349670e61

[davidben, 2017-02-10 19:25:03]

The CQ bit was checked by davidben@chromium.org to run a CQ dry run

[commit-bot: I haz the power, 2017-02-10 19:26:02]

Dry run: CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[davidben, 2017-02-10 19:51:00]

davidben@chromium.org changed reviewers:
+ svaldez@chromium.org

[svaldez, 2017-02-10 19:52:39]

lgtm

[davidben, 2017-02-10 19:53:22]

The CQ bit was unchecked by davidben@chromium.org

[davidben, 2017-02-10 19:53:25]

The CQ bit was checked by davidben@chromium.org

[commit-bot: I haz the power, 2017-02-10 19:53:57]

CQ is trying da patch. Follow status at
 
https://chromium-cq-status.appspot.com/v2/patch-status/codereview.chromium.or...

[xunjieli, 2017-02-10 20:46:24]

This should have at least 200KiB improvement (see allocation under
CRYPTO_BUFFER_new) for every ~50 sockets.
I don't have a way to repeat experiment for before and after, so I am using
approximates here. In the future,we hope to use telemetry harness to repeat
experiment for before and after.

Here are the trace files:
https://drive.google.com/drive/folders/0Bw0C8-TgzJfsUlc2dXRfcUdmNUU?usp=sharing

[commit-bot: I haz the power, 2017-02-10 21:00:38]

CQ is committing da patch.

Bot data: {"patchset_id": 1, "attempt_start_ts": 1486756405623170, "parent_rev":
"97c42c445521b0f145afb0603bf7a38d4266e7dd", "commit_rev":
"bf0fcf111ceab0a9143d75282ced0d4349670e61"}

[davidben, 2017-02-10 21:00:40]

On 2017/02/10 20:46:24, xunjieli wrote:
> This should have at least 200KiB improvement (see allocation under
> CRYPTO_BUFFER_new) for every ~50 sockets.
> I don't have a way to repeat experiment for before and after, so I am using
> approximates here. In the future,we hope to use telemetry harness to repeat
> experiment for before and after.
> 
> Here are the trace files:
>
https://drive.google.com/drive/folders/0Bw0C8-TgzJfsUlc2dXRfcUdmNUU?usp=sharing

Cool. Although this will depend a lot on exactly what traffic patterns we see,
so it won't quite be every 50 sockets; this deduplication is if we see the same
certificate multiple times. (The other work to drop X509* altogether will be
more valuable since it will save memory on every certificate, whether it is a
duplicate or not.)

[commit-bot: I haz the power, 2017-02-10 21:01:15]

Description was changed from

==========
Add a CRYPTO_BUFFER_POOL to SSLClientSocketImpl.

In anticipation of using it for net::X509Certificate in the future, the
pool lives somewhere in net/cert. Until the rest of the work to remove
switch to CRYPTO_BUFFERs done, this will not be as effective as it could
be.

But, as part of that transition, we have already started using
CRYPTO_BUFFERs in parts of the stack. This allows that portion to be
deduplicated. The (more expensive) portion which isn't deduplicated will
be fixed by removing it altogether.

BUG=671420
==========

to

==========
Add a CRYPTO_BUFFER_POOL to SSLClientSocketImpl.

In anticipation of using it for net::X509Certificate in the future, the
pool lives somewhere in net/cert. Until the rest of the work to remove
switch to CRYPTO_BUFFERs done, this will not be as effective as it could
be.

But, as part of that transition, we have already started using
CRYPTO_BUFFERs in parts of the stack. This allows that portion to be
deduplicated. The (more expensive) portion which isn't deduplicated will
be fixed by removing it altogether.

BUG=671420

Review-Url: https://codereview.chromium.org/2682413003
Cr-Commit-Position: refs/heads/master@{#449732}
Committed:
https://chromium.googlesource.com/chromium/src/+/bf0fcf111ceab0a9143d75282ced...
==========

[commit-bot: I haz the power, 2017-02-10 21:01:16]

Committed patchset #1 (id:1) as
https://chromium.googlesource.com/chromium/src/+/bf0fcf111ceab0a9143d75282ced...