8217579: TLS_EMPTY_RENEGOTIATION_INFO_SCSV is gone after 8211883

[Sean Mullan]

Updated webrev: 
http://cr.openjdk.java.net/~mullan/webrevs/8217579/webrev.01/

Comments inline ...

On 1/28/19 2:54 PM, Bernd Eckenfels wrote:
> Hello Sean,
> 
> Maybe you also want to change comment and name of the SUPPORTE_DDEFAULT 
> Array to „SUPPORTED_LIMITED“ since Unlimited is now Default?
> 
>      private final static String[] ENABLED_DEFAULT
> 
> ….
> 
>       // supported ciphersuites using default JCE policy jurisdiction files
> 
>       // AES/256 unavailable
> 
>       private final static String[] SUPPORTED_DEFAULT = {
> 
> 230 – remove „Default

Good point. I have renamed the *_UNLIMITED constants to *_DEFAULT and 
renamed the *_DEFAULT constants to *_LIMITED.

> Is the test already run with all available policies? With the new System 
> property it should be easy to run it with other/vm twice?

Good point. I have changed the test to use the crypto.policy security 
property to test the suites with the default and limited policies.

> Is Oracle considering pushing a emergency public update for this?

We are planning to backport it to all affected releases.

> The change Looks otherwise fine (I was first wondering if checking for a 
> _SVCS Family makes more sense but I guess that can be done once we have 
> more of those ciphers.

Ok, thanks for the review.

--Sean

> 
> Gruss
> 
> Bernd
> 
> -- 
> http://bernd.eckenfels.net
> 
> *Von: *Sean Mullan <mailto:sean.mullan at oracle.com>
> *Gesendet: *Montag, 28. Januar 2019 20:26
> *An: *security Dev OpenJDK <mailto:security-dev at openjdk.java.net>
> *Betreff: *RFR: 8217579: TLS_EMPTY_RENEGOTIATION_INFO_SCSV is gone after 
> 8211883
> 
> This fixes a regression introduced by the recent change to disable the
> 
> TLS NULL cipher suites [1]. This accidentally also disabled the
> 
> TLS_EMPTY_RENEGOTIATION_INFO_SCSV cipher suite because when the name is
> 
> decomposed by the algorithm constraints checking code it has NULL for
> 
> its different parts (key exchange, etc). But this cipher suite is not
> 
> negotiable and is only used for renegotiation purposes as defined in RFC
> 
> 5746. It should not have been disabled.
> 
> I also resurrected the CheckCipherSuites test which had an @ignore label
> 
> on it. This is a good test because it checks what the expected
> 
> enabled/supported suites should be, and will help catch issues like this
> 
> in the future.
> 
> webrev: http://cr.openjdk.java.net/~mullan/webrevs/8217579/webrev.00/
> 
> bug: https://bugs.openjdk.java.net/browse/JDK-8217579
> 
> Thanks,
> 
> Sean
> 
> [1] https://bugs.openjdk.java.net/browse/JDK-8211883
>