Code Review Requests for 7196382 and 8010134

[Xuelei Fan]

src/share/classes/sun/security/pkcs11/P11KeyPairGenerator.java
==============================================================
--------
It is instinctive and easy to maintain to code if we have the comment:
+        // RSA, DH, and DSA
 104     keySize = 1024;


--------
I'm not sure, just a double check. Do we really allow DSA  keysize is
not a multiple of 64 bits when it is bigger than or equal to 1024?

 240     } else if (algorithm.equals("DSA")) {
 241         if ((ks < 1024) && ((ks & 0x3f) != 0)) {
 242             throw new InvalidAlgorithmParameterException
 243                 ("DSA key must be a multiple of 64 bits");

--------
If the mechanism info is not specified in the underlying device
(minKeySize/maxKeySize is -1), do we still want to reserve the minimal
length checking (112 for EC, 512 for non-EC)? I think we may still want
the constraints. For example:

 226         } else {
+                // The key size is too small.
+                if (algorithm.equals("EC")) {
+                    // keysize < 112, throws exception
+                } else {
+                    // keysizse < 512, throws exception
+                }


Xuelei

On 7/10/2013 8:33 AM, Valerie (Yu-Ching) Peng wrote:
> Xuelei,
> 
> I got side-tracked w/ other bugs before coming back to this one.
> 
> I have updated the webrev to incorporate your review comments, including
> 1) auto-adjust the default keysize when it's out of range
> 2) apply our own range checking besides the native limit checking
> 
> I also noticed that the DH parameter generator from SunJCE provider
> relies on DSA, so I checked its keysize checking to match the DSA keysize.
> 
> The webrev is updated at:
> http://cr.openjdk.java.net/~valeriep/7196382/webrev.02/
> Thanks,
> Valerie
> 
> On 04/27/13 01:58, Xuelei Fan wrote:
>> I like this update.
>>
>> On 4/27/2013 7:29 AM, Valerie (Yu-Ching) Peng wrote:
>>> Xuelei,
>>>
>>> I have updated the webrev for 7196382 so it uses the key size range info
>>> from the underlying PKCS library for key size checking:
>>> http://cr.openjdk.java.net/~valeriep/7196382/webrev.01/
>>>
>> 107  //TBD: auto-adjust default keysize in case it's out-of-range?
>>
>> I think it's nice to enable this following block. The key size will be
>> checked in initialize(), so I think it is a little bit reasonable to
>> select a proper default key size instead of throwing an exception later.
>>
>>
>> 209 private void checkKeySize(int ks, RSAKeyGenParameterSpec params)
>>
>> I think when minKeySize is -1, we need to consider the default key size
>> limit (EC 112, RSA/DH/DSA 512).  In this update, it seems that if
>> minKeySize is -1, we can generate small keys. I don't think it is the
>> intended design.  It is similar when maxKeySize is -1.
>>
>> I was wondering that if minKeySize is -1 (or less than the default
>> hard-coded key size), or maxKeySize is -1 (or greater than the
>> hard-coded default key size), we may be able to reset them to the
>> default hard-coded sizes.
>>
>> Thanks,
>> Xuelei
>>
>>> Thanks,
>>> Valerie
>>>
>>> On 04/25/13 10:59, Valerie (Yu-Ching) Peng wrote:
>>>> Xuelei,
>>>>
>>>> Thanks for the review and comments.
>>>>
>>>> Supposedly, we don't have to have default parameters for all valid key
>>>> sizes.
>>>> The pre-generated default parameters are for the most-commonly used
>>>> keysizes.
>>>> As for the rest of supported key sizes, the needed parameters will be
>>>> generated at runtime upon request.
>>>>
>>>> Well, I don't quite like the current approach of hardcoding ranges
>>>> inside the checkKeySize(...) method.
>>>> There is a way to query the supported keysize ranges from the PKCS11
>>>> library and I think that should be the values that we base the key
>>>> size check on, plus any additional algorithm-specific check (e.g.
>>>> multiples of 64 bits) that can't be expressed through the ranges. I am
>>>> still testing out the changes. Will post an updated webrev for 7196382
>>>> once I am done testing...
>>>>
>>>> Thanks!
>>>> Valerie
>>>>
>>>> On 04/18/13 21:45, Xuelei Fan wrote:
>>>>> On 4/19/2013 10:43 AM, Valerie (Yu-Ching) Peng wrote:
>>>>>> Xuelei,
>>>>>>
>>>>>> Do you have time to review the following two fixes?
>>>>>> 7196382: PKCS11 provider should support 2048-bit DH
>>>>>> 8010134: A finalizer in sun.security.pkcs11.wrapper.PKCS11 perhaps
>>>>>> should be protected
>>>>>>
>>>>>> The first one removes the hardcoded limit of 1024 for DH and the
>>>>>> second
>>>>>> one is making the finalize() method protected.
>>>>>>
>>>>>> Webrevs:
>>>>>> http://cr.openjdk.java.net/~valeriep/7196382/webrev.00/
>>>>> Looks fine.
>>>>>
>>>>> Do we plan to support DH keys bwteen 1024 and 2048 with default (null)
>>>>> parameters, for example 1536, in PKCS11 provider?  Recently, I run
>>>>> into
>>>>> a case that uses DH public keys of 1536 bits. I was wondering we may
>>>>> also want to support more.
>>>>>
>>>>>> http://cr.openjdk.java.net/~valeriep/8010134/webrev.00/
>>>>> Looks fine.
>>>>>
>>>>> Xuelei
>>>>>
>>>>>> Thanks!
>>>>>> Valerie
>