RFR: 8255557: Decouple GCM from CipherCore [v5]

[Valerie Peng]

On Mon, 24 May 2021 22:36:49 GMT, Anthony Scarpino <ascarpino at openjdk.org> wrote:

>> test/jdk/com/sun/crypto/provider/Cipher/AEAD/GCMShortBuffer.java line 34:
>> 
>>> 32: /*
>>> 33:  * @test
>>> 34:  * @summary Call decrypt doFinal() with different output values to see if the
>> 
>> Missing bug id?
>
> Here is my logic
> It's not a regression test to verify a specific bug id, it's just a general functionality test.  Therefore the bug id does not point to a previous failure.

Hmm, alright.

>> test/jdk/com/sun/crypto/provider/Cipher/AES/TestAESCipher.java line 90:
>> 
>>> 88:             i++;
>>> 89:         }
>>> 90:         return b;
>> 
>> The generated data seems too similar, all starts with 0 and increment. Maybe use i = len to start with? Or take some additional argument for starting value?
>
> I wanted consistent data because it's harder to solve crypto failures when you cannot see consistent pattern in the data.  For example, the GCM case where the ciphertext is correct, but the tag is wrong.  It is easier to know that the problem is in GHASH and not GCTR.
> I don't see any negatives to having similar data.

Might as well just hardcode the data then. Saves time to re-generate the similar bytes over and over. Not seeing much benefit of this method. Not a big deal if you intentionally use similar data.

-------------

PR: https://git.openjdk.java.net/jdk/pull/4072