On Wed, 1 Jul 2009, Valerie Bubb Fenwick wrote:

> On Wed, 1 Jul 2009, Krishna Yenduri wrote:
>
>> Valerie Bubb Fenwick wrote:
>>> On Tue, 30 Jun 2009, Krishna Yenduri wrote:
>>> 
>>> ...
>>>>> 
>>>>>> 
>>>>>> 
>>>>>> http://cr.opensolaris.org/~bubbva/fips-work.5/
>>>>> 
>>>>> KY-1   elfsign(1) command
>>>>> 
>>>>> ...
>>>> 
>>>> The specific case I had in mind is when elfsign is used with a SCA6000.
>>> 
>>> Hi Krishna -
>>> 
>>> I have access to a test machine with SCA 6000 on it that I can check this
>>> on (unfortunately, running the crypto tests unconfigured that, so I'll 
>>> have
>>> to see if I can get help to get it working again), but looking at the
>>> code, specifically in libelfsign's elfcertlib_settoken(), I'm not sure
>>> if we've tested this specific situation since we KMF'ized libelfsign.
>>> 
>>> This may be a problem. I'll let you know.
>> 
>> I am now fairly certain that this will be a problem since elfsign needs to
>> call libpkcs11/pkcs11_kernel routines to be able to use SCA 6000.
>
> Once the network was functioning again, I did verify that it was indeed
> a problem. While I can't find anyone actually using this flag, it is a good
> feature and has been backported to S10, so we shouldn't break it. :)
>
>> One solution could be to do this symbol interposing
>> in kcfd instead of doing it in libelfsign. This should be fine since all 
>> the
>> signature verification is done by kcfd.
>
> That's not a bad idea. Tony & I had been talking more about making
> changes to KMF, but that's another place we could maybe do this. I
> will try that as well today.

Actually, I just thought about that more, and it won't work, because
elfsign (the command) doesn't go through kcfd. So, making the changes
there would require rearhitecting of the elfsign command.

Valerie
-- 
Valerie Fenwick, http://blogs.sun.com/bubbva/ @bubbva
Solaris Security Technologies, Developer, Sun Microsystems, Inc.
17 Network Circle, Menlo Park, CA, 94025.

Reply via email to