[illumos-Developer] [REVIEW] 289 invalid padding when using java pkcs11 provider

[Jason King]

I thought I was able to invoke lint on those pieces after doing an
incremental nightly (from the original diff), however given some of
the wizardry of the ON build files, I might have failed while thinking
I did run it successfully.  I'll rerun a full nightly and see if it
spits anything out.

On Fri, Oct 8, 2010 at 7:38 PM, Robert Gordon <rbg at openrbg.com> wrote:
>
> On Oct 8, 2010, at 6:12 PM, Jason King wrote:
>
>> On Fri, Oct 8, 2010 at 1:38 PM, Garrett D'Amore <garrett at damore.org> wrote:
>>> On Fri, 2010-10-08 at 14:11 -0400, Gordon Ross wrote:
>>>> Oh, and the kmem_alloc needs another arg, I think :)
>>>> i.e. KM_SLEEP
>>>
>>> That's an excellent point.  The old code lacked the sleep parameter,
>>> which makes me thing this code has never even compiled for kernel use.
>>> (Which isn't surprising... there are no kernel consumers for DH -- the
>>> only related use would be IKE, but IKE runs in userland.)
>>>
>>> Since we know *nothing* about the context where this would possibly
>>> used, I think KM_NOSLEEP would be better.
>>>
>>>        - Garrett
>>
>> It looks like most of the ones that are used in the kernel have a
>> wrapper function that calls into the API to get the appropriate flag
>> to use.  I'd prefer to follow the example of the existing kernel
>> providers and just add an argument that is only used in the kernel.
>>
>> With that here is an updated webrev w/ the suggested changes:
>> http://cr.illumos.org/view/w041vj3b/
>
> Super-NIT:
>
> I'd suspect that dh_key_derive() in user-land is not lint clean (i've not tested that yet),
> however since flag is not used in user-land lint would be all fussy about that.. so the addition
> of /*ARGSUSED*/ would shut it up ;-)
>
> Robert.