On 10.02.2015 18:22, Marcus Meissner wrote:
> Hi Jussi,
>
> We were trying to use rfc4543(gcm(aes)) in the kernel for FIPS mode,
> but the testvectors seem to fail.
You probably need to add '.fips_allowed = 1,' in testmgr.c for "rfc4543(gcm(aes))" to enable algorithm in fips mode.
>
> Did you verify that they work? Are these the ones from Page 18 of
> https://tools.ietf.org/html/draft-mcgrew-gcm-test-01, as there the "plaintext"
> and "aaad" seem to be switched?
rfc4543() wrapper constructs the aad from '.assoc' and '.input'.
-Jussi
>
> Ciao, Marcus
>
On Tue, Feb 10, 2015 at 07:39:22PM +0200, Jussi Kivilinna wrote:
> On 10.02.2015 18:22, Marcus Meissner wrote:
> > Hi Jussi,
> >
> > We were trying to use rfc4543(gcm(aes)) in the kernel for FIPS mode,
> > but the testvectors seem to fail.
>
> You probably need to add '.fips_allowed = 1,' in testmgr.c for "rfc4543(gcm(aes))" to enable algorithm in fips mode.
>
> >
> > Did you verify that they work? Are these the ones from Page 18 of
> > https://tools.ietf.org/html/draft-mcgrew-gcm-test-01, as there the "plaintext"
> > and "aaad" seem to be switched?
>
> rfc4543() wrapper constructs the aad from '.assoc' and '.input'.
We tried above and still got -2 errors.
I think "ecb(cipher_null)" also needs to set the .fips_allowed flag, so that the rfc4543 setup can pull it in.
CIao, Marcus