[dpdk-dev] [PATCH] cryptodev: clarify wireless inputs in digest-encrypted cases

[Akhil Goyal]

> 
> Hi Akhil,
> 
> >
> > Hi Fiona,
> >
> > >
> > > Clarify constraints on fields specified in bits for wireless
> > > algorithms in digest-encrypted case.
> > >
> > > Signed-off-by: Fiona Trahe <fiona.trahe at intel.com>
> > > ---
> > >  lib/librte_cryptodev/rte_crypto_sym.h |    7 +++++++
> > >  1 files changed, 7 insertions(+), 0 deletions(-)
> > >
> > > diff --git a/lib/librte_cryptodev/rte_crypto_sym.h
> > > b/lib/librte_cryptodev/rte_crypto_sym.h
> > > index bc8da24..0204d37 100644
> > > --- a/lib/librte_cryptodev/rte_crypto_sym.h
> > > +++ b/lib/librte_cryptodev/rte_crypto_sym.h
> > > @@ -703,6 +703,13 @@ struct rte_crypto_sym_op {
> > >  					 * auth.data.length and is typically
> > >  					 * equal to auth.data.offset +
> > >  					 * auth.data.length + digest_length.
> > > +					 * - for wireless algorithms, i.e.
> > > +					 * SNOW 3G, KASUMI and ZUC, as the
> > > +					 * cipher.data.length,
> > > +					 * cipher.data.offset,
> > > +					 * auth.data.length and
> > > +					 * auth.data.offset are in bits, they
> > > +					 * must be 8-bit multiples.
> >
> > What is the meaning of 'they' here?
> > cipher.data.length, cipher.data.offset, auth.data.length and auth.data.offset
> are in bits
> > and may or may not be multiple of 8bits. This is documented in their comments.
> >
> [Fiona] This added note is an extra bullet under the explanation for Digest-
> encrypted cases.
> In these cases only those fields listed MUST be 8-bit multiples.
> Can you suggest a change that would make that clearer?

@@ -590,7 +590,9 @@ struct rte_crypto_sym_op {
                                          * For SNOW 3G @ RTE_CRYPTO_CIPHER_SNOW3G_UEA2,
                                          * KASUMI @ RTE_CRYPTO_CIPHER_KASUMI_F8
                                          * and ZUC @ RTE_CRYPTO_CIPHER_ZUC_EEA3,
-                                         * this field should be in bits.
+                                         * this field should be in bits and must be
+                                         * 8-bit multiples in cases of encrypted
+                                         * digest.
                                          */

Same change may be done in other 3 fields.

> 
> 
> >
> > >  					 *
> > >  					 * Note, that for security reasons, it
> > >  					 * is PMDs' responsibility to not
> > > --
> > > 1.7.0.7