Add support for built-in certificate pins, and add a pin list suitable for

Add support for built-in certificate pins, and add a pin list suitable for
Google properties, in discussion with the central security team. Use said pin
list for an initial sample domain.
Also factor out a couple of functions to eliminate duplicated code.

TEST=TransportSecurityStateTest.BuiltinCertPins
Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=82923

[Chris Evans, 2011-04-22 18:23:54]

And so continues the experiment. This is being done in conjunction with
discussions with the group managing the server-side certs, and is hence very
conservative.
A quick note on the "mandatory" concept. It's needed for the next CL, and I
don't want to edit the entire HSTS list twice. (In a nutshell, we'll need to pin
certs whilst not requiring https on a given domain. This is because
https://secured.example.com/ might <script src> something from
https://www.google.com/jquery/jq.js or similar. We need to pin the subresource
cert but cannot enforce https on the whole origin.)

[abarth-chromium, 2011-04-22 18:26:59]

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
File net/base/transport_security_state.cc (right):

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
net/base/transport_security_state.cc:349: }
Should we NOTREACHED() here?

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
net/base/transport_security_state.cc:493: bool mandatory;
mandatory is a confusing name.  Maybe require_tls ?

[Chris Evans, 2011-04-22 21:47:55]

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
File net/base/transport_security_state.cc (right):

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
net/base/transport_security_state.cc:349: }
On 2011/04/22 18:26:59, abarth wrote:
> Should we NOTREACHED() here?

Probably not. The original client doesn't want it. I asked agl@ the same
question, but he made the point that an older browser might open the
TransportSecurity of a newer profile, and encounter e.g. some sha256 hashes.

If you think it's worth it, we could return a bool and have IsPreloadedSTS()
DCHECK() the result.

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
net/base/transport_security_state.cc:493: bool mandatory;
On 2011/04/22 18:26:59, abarth wrote:
> mandatory is a confusing name.  Maybe require_tls ?

Done.

Yes, poor name. I went for "https_required" if that's ok?

[abarth-chromium, 2011-04-22 21:56:19]

> Probably not. The original client doesn't want it. I asked agl@ the same
> question, but he made the point that an older browser might open the
> TransportSecurity of a newer profile, and encounter e.g. some sha256 hashes.

Good point.

> If you think it's worth it, we could return a bool and have IsPreloadedSTS()
> DCHECK() the result.

That might be a good idea.

http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...

> Yes, poor name. I went for "https_required" if that's ok?

Yep.  Thanks.

[Chris Evans, 2011-04-22 22:27:28]

On 2011/04/22 21:56:19, abarth wrote:
> > Probably not. The original client doesn't want it. I asked agl@ the same
> > question, but he made the point that an older browser might open the
> > TransportSecurity of a newer profile, and encounter e.g. some sha256 hashes.
> 
> Good point.
> 
> > If you think it's worth it, we could return a bool and have IsPreloadedSTS()
> > DCHECK() the result.
> 
> That might be a good idea.

Done.

> 
>
http://codereview.chromium.org/6894026/diff/1/net/base/transport_security_sta...
> 
> > Yes, poor name. I went for "https_required" if that's ok?
> 
> Yep.  Thanks.

[agl, 2011-04-25 19:39:46]

LGTM

http://codereview.chromium.org/6894026/diff/8/net/base/transport_security_sta...
File net/base/transport_security_state.cc (right):

http://codereview.chromium.org/6894026/diff/8/net/base/transport_security_sta...
net/base/transport_security_state.cc:551: static const char*
kCertPKHashVerisignClass3 =
I'm conflicted about parsing these strings all the time, but it does have the
advantage that it matches the format everywhere else. I think it's fine for now.

http://codereview.chromium.org/6894026/diff/8/net/base/transport_security_sta...
net/base/transport_security_state.cc:551: static const char*
kCertPKHashVerisignClass3 =
Need to change all of these from "const char*" to "const char foo[]"

http://codereview.chromium.org/6894026/diff/8/net/base/transport_security_sta...
net/base/transport_security_state.cc:638: &ret))
I think you should have { } around this if body because the if is multiline
itself.

http://codereview.chromium.org/6894026/diff/8/net/base/transport_security_sta...
net/base/transport_security_state.cc:641: HasPreload(kPreloadedSNISTS,
kNumPreloadedSNISTS, canonicalized_host, i,
ditto.