[security-dev 01256]: Re: PING: [PATCH FOR REVIEW]: 6763530: Fix breakage of NSS-based Elliptic Curve Cryptography in OpenJDK6

Fri Sep 25 20:53:13 UTC 2009

Hi Vincent - 

The original code was not compliant with PKCS11.  It basically confused the X9.62 format for the public key with the ECPoint ASN1 format for the public key - the latter being the one needed in both files.  The changes described in the webrev only affect in/out from a PKCS11 module and shouldn't have any other side effects - I spent a bit of time looking around to ensure this was the case.  I may have missed something, but its unlikely.

The backwards compatibility issues would be with PKCS11 modules that mis-implemented the PKCS11 standard - I don't know of any that have this issue.  It would be relatively easy to deal with the decode of a broken returned key in P11Key, but since you would have to know a priori whether or not the PKCS11 module was broken or not, the fix for P11ECKeyFactory (where it currently has a broken encoding method) would need to depend on some configuration option from the PKCS11 class - and that's a much bigger change.

Mike

At 01:33 PM 9/24/2009, Vincent Ryan wrote:
>Hello Andrew,
>
>I'll need a little more time to come up to speed on this fix. I'm concerned that
>there may be interoperability or backwards compatibility issues.
>
>
>
>Andrew John Hughes wrote:
>> 2009/9/2 Andrew John Hughes <gnu_andrew at member.fsf.org>:
>>> 2009/9/2 Michael StJohns <mstjohns at comcast.net>:
>>>> At 09:38 PM 9/1/2009, Andrew John Hughes wrote:
>>>>> 2009/9/2 Michael StJohns <mstjohns at comcast.net>:
>>>>>> Ã‚ This appears to be related specifically to PKCS11.Ã‚  Specifically, PKCS11
>>>>>> v2.20 has some ambiguity of the representation of an EC point (which is
>>>>>> different in the text than an ASN1 ECPoint).
>>>>>>
>>>>>> This is being clarified in v2.30 with the unencoded point format (e.g.the
>>>>>> format described inÃ‚  X9.62, where the first octet indicates the encoding and
>>>>>> there are either N or 2N octets following)Ã‚  being the expected value, but
>>>>>> with PKCS11 providers allowed - legacy - to accept either.
>>>>>>
>>>>>> One of the reasons for going that way was how the JDK PKCS11 provider had
>>>>>> interpreted the issue and implemented its code.
>>>>>>
>>>>>> I don't support this fix - among other things, this fix only deals with 1/2
>>>>>> of the problem.Ã‚  The other half is related to encoding the value.Ã‚  Also,
>>>>>> changing the code at decodePoint seems further into the stack than needed
>>>>>> and may affect other uses of that method.
>>>>>>
>>>>> That's really too vague to be of much help in improving the patch.
>>>>> You seem to be saying little more than 'I don't like it'.
>>>> Sorry about that.  My point was that your patch didn't completely solve the problem and that the point at where you were fixing it could have some bad side effects for anyone calling decodePoint directly.
>>>>
>>>>
>>>>>> There's an existing JDK bug on this coming at it from a different direction
>>>>>> - 6763530 ... and there may be considerations at
>>>>>>
>>>>>> https://bugzilla.mozilla.org/show_bug.cgi?id=480280
>>>>>>
>>>>> It seems likely that's the NSS change that causes the current failure.
>>>>> The fix I submitted here is based on the way this is handle in NSS.
>>>>> In fact, the code is similar enough to suggest that one was developed
>>>> >from the other.
>>>>>> Ã‚ that should be looked at.
>>>>> The JDK bug is not really 'from a different direction', it's reporting
>>>>> exactly the same error but from a less trivial example (I get the same
>>>>> failure while trying to create an example key, while this seems to
>>>>> require specific hardware if I'm reading it correctly).
>>>> Not exactly.  You're using the NSS as a PKCS11 module - this problem would occur with any PKCS11 module that implements EC stuff.
>>>>
>>>>
>>>>> Also see 6779460 which is mostly a duplicate of
>>>>>> 6763530.
>>>>>>
>>>>> The patch on 6779460 seems wrong.  It means that the method will
>>>>> return a DER-encoded value where it would either have returned an
>>>>> uncompressed value before or failed.
>>>> My point exactly as I mentioned in the comments.  :-)
>>>>
>>>>
>>>>>> It's probable that the fix I suggested at 6763530Ã‚  (in comments submitted 29
>>>>>> Nov 08) may be a better approach given the NSS fixes.Ã‚  I believe it will fix
>>>>>> the keytool problem noted in the original message.
>>>>>>
>>>>> Ok, I can see the logic in the fix and it would appear to work, though
>>>>> I haven't tested it.
>>>>> Given the patch was written nine months ago, why has it not been
>>>>> applied?  If it had, it would have saved me hours having to debug this
>>>>> same issue again.
>>>> Yup.  I did do a search for PKCS11 related bugs when I encountered the same problem and did find the original error.
>>>>
>>>>> Do you have an SCA with Sun? If so, I'll create a webrev based on your
>>>>> patch and we can finally get this fixed.  Without it, NSS support is
>>>>> completely broken in OpenJDK6 which makes me wonder why this is a low
>>>>> priority bug!
>>>> I do have an SCA on file.  Note that the recommendation from the NSS guys was to raise the priority.
>>>>
>>>> The reason I haven't submitted this is because I submitted a different EC fix  https://bugs.openjdk.java.net/show_bug.cgi?id=100048 per the documented process
>>>>  and was waiting on progress there before continuing.  I've got a number of EC and PKCS11 related fixes I'd like to submit, but I was trying for a worked example before proceeding.  And then I got busy with some other things...
>>>>
>>>> Mike
>>>>
>>>>
>>>>
>>>>
>>>>
>>>>>> Mike
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>>
>>>>>> At 04:39 PM 9/1/2009, Joe Darcy wrote:
>>>>>>
>>>>>> Andrew John Hughes wrote:
>>>>>>
>>>>>> 2009/8/28 Andrew John Hughes <gnu_andrew at member.fsf.org>:
>>>>>>
>>>>>> In OpenJDK6, the elliptic curve cryptography algorithms are available
>>>>>> if the PKCS11 provider is configured to point to NSS. See:
>>>>>>
>>>>>> http://blogs.sun.com/andreas/entry/the_java_pkcs_11_provider
>>>>>>
>>>>>> If NSS is configured as specified in this blog, keytool can be used to
>>>>>> generate a key as follows:
>>>>>>
>>>>>> Hello.
>>>>>>
>>>>>> Allowing keytool and friends to work in more cases if the provider is
>>>>>> capable seems fine to me.
>>>>>>
>>>>>> Security team, do you have concerns about this patch?
>>>>>>
>>>>>> Thanks,
>>>>>>
>>>>>> -Joe
>>>>>>
>>>>>
>>>>>
>>>>> --
>>>>> Andrew :-)
>>>>>
>>>>> Free Java Software Engineer
>>>>> Red Hat, Inc. (http://www.redhat.com)
>>>>>
>>>>> Support Free Java!
>>>>> Contribute to GNU Classpath and the OpenJDK
>>>>> http://www.gnu.org/software/classpath
>>>>> http://openjdk.java.net
>>>>>
>>>>> PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
>>>>> Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8
>>>>
>>>>
>>> Ok here is a new webrev:
>>>
>>> http://cr.openjdk.java.net/~andrew/6763530/webrev.02/
>>>
>>> with a slightly revised version of your change (you can't throw a
>>> PKCS11Exception which only takes a long ID from the native code, so I
>>> changed this to an IllegalArgumentException).
>>>
>>> Security team, does this look ok to push?
>>> --
>>> Andrew :-)
>>>
>>> Free Java Software Engineer
>>> Red Hat, Inc. (http://www.redhat.com)
>>>
>>> Support Free Java!
>>> Contribute to GNU Classpath and the OpenJDK
>>> http://www.gnu.org/software/classpath
>>> http://openjdk.java.net
>>>
>>> PGP Key: 94EFD9D8 (http://subkeys.pgp.net)
>>> Fingerprint: F8EF F1EA 401E 2E60 15FA  7927 142C 2591 94EF D9D8
>>>
>> 
>> Ping! Security developers, any thoughts on this patch:
>> 
>> http://cr.openjdk.java.net/~andrew/6763530/webrev.02/
>> 
>> Does it look ok to push?
>> 
>> Thanks,