Re: [dpdk-stable] [dpdk-dev] [PATCH] cryptodev: fix SHA-1 digest enum comment

[Anoob Joseph <anoobj@marvell.com>]

Hi Fiona,

Please see inline.

Thanks,
Anoob

> -----Original Message-----
> From: Trahe, Fiona <fiona.trahe@intel.com>
> Sent: Tuesday, May 19, 2020 3:59 PM
> To: Dybkowski, AdamX <adamx.dybkowski@intel.com>; Anoob Joseph
> <anoobj@marvell.com>; dev@dpdk.org; akhil.goyal@nxp.com
> Cc: stable@dpdk.org; Trahe, Fiona <fiona.trahe@intel.com>
> Subject: [EXT] RE: [dpdk-dev] [PATCH] cryptodev: fix SHA-1 digest enum
> comment
> 
> External Email
> 
> ----------------------------------------------------------------------
> Hi Adam, Anoob,
> 
> 
> > > > >
> > > > > > -----Original Message-----
> > > > > > From: dev <dev-bounces@dpdk.org> On Behalf Of Adam Dybkowski
> > > > > > Sent: Monday, May 18, 2020 7:54 PM
> > > > > > To: dev@dpdk.org; fiona.trahe@intel.com; akhil.goyal@nxp.com
> > > > > > Cc: Adam Dybkowski <adamx.dybkowski@intel.com>;
> > > > > > stable@dpdk.org
> > > > > > Subject: [dpdk-dev] [PATCH] cryptodev: fix SHA-1 digest enum
> > > > > > comment
> > > > > >
> > > > > > This patch fixes improper SHA-1 digest size in the enum comment.
> > > > > >
> > > > > > Fixes: 1bd407fac80b ("cryptodev: extract symmetric
> > > > > > operations")
> > > > > > Cc: stable@dpdk.org
> > > > > >
> > > > > > Signed-off-by: Adam Dybkowski <adamx.dybkowski@intel.com>
> > > > > > ---
> > > > > >  lib/librte_cryptodev/rte_crypto_sym.h | 4 ++--
> > > > > >  1 file changed, 2 insertions(+), 2 deletions(-)
> > > > > >
> > > > > > diff --git a/lib/librte_cryptodev/rte_crypto_sym.h
> > > > > > b/lib/librte_cryptodev/rte_crypto_sym.h
> > > > > > index d9585ecd6..9cea4e5f0 100644
> > > > > > --- a/lib/librte_cryptodev/rte_crypto_sym.h
> > > > > > +++ b/lib/librte_cryptodev/rte_crypto_sym.h
> > > > > > @@ -269,9 +269,9 @@ enum rte_crypto_auth_algorithm {
> > > > > >  	/**< HMAC using MD5 algorithm */
> > > > > >
> > > > > >  	RTE_CRYPTO_AUTH_SHA1,
> > > > > > -	/**< 128 bit SHA algorithm. */
> > > > > > +	/**< 160 bit SHA algorithm. */
> > > > > >  	RTE_CRYPTO_AUTH_SHA1_HMAC,
> > > > > > -	/**< HMAC using 128 bit SHA algorithm. */
> > > > > > +	/**< HMAC using 160 bit SHA algorithm. */
> > > > >
> > > > > [Anoob] This raises one interesting question. Is HMAC-SHA-1-96
> > > > > (https://urldefense.proofpoint.com/v2/url?u=https-3A__tools.ietf
> > > > > .org
> > > > > _h
> > > > >
> > > >
> > > tml_rfc2404&d=DwIFAg&c=nKjWec2b6R0mOyPaz7xtfQ&r=jPfB8rwwviRSxyL
> > > Ws2
> > > > n6B-
> > > >
> > > WYLn1v9SyTMrT5EQqh2TU&m=pgqB2BfFosCF_7l1SdoodgXcipf7G1ofht68ZK
> > > Mt
> > > > oW0&s=hxzUN2QfRJGaR7NQ7VKGm5oQvOZez6Z70mqOyg7gKTY&e= )
> > > treated as a
> > > > separate algorithm or is it a digest size variant of
> > > > RTE_CRYPTO_AUTH_SHA1_HMAC? I assume the later.
> > > >
> > > > [Adam] DPDK doesn't have separate enums for such variations.
> > > > According to RFC 2104, chapter "5. Truncated output", the
> > > > calculation of
> > > > HMAC-SHA-1-96 is done using ordinary HMAC-SHA-1, but the
> > > > calculation result is then truncated to 96 bits, or any other
> > > > digest size provided in auth
> > > xform.
> > >
> > > [Anoob] So it is allowed to support digest sizes 12 & 20, right? Can
> > > you update the above comment in that case?
> >
> > [Adam] This depends on a particular PMD and its capabilities so we
> > cannot write here in this public header file. For example, have a look in the
> capabilities structure of QAT PMD:
> >
> > 	{	/* SHA1 */						\
> > 		.op = RTE_CRYPTO_OP_TYPE_SYMMETRIC,
> 	\
> > 		{.sym = {						\
> > 			.xform_type = RTE_CRYPTO_SYM_XFORM_AUTH,
> 	\
> > 			{.auth = {					\
> > 				.algo = RTE_CRYPTO_AUTH_SHA1,
> 	\
> > 				.block_size = 64,			\
> > 				.key_size = {				\
> > 					.min = 0,			\
> > 					.max = 0,			\
> > 					.increment = 0			\
> > 				},					\
> > 				.digest_size = {			\
> > 					.min = 1,			\
> > 					.max = 20,			\
> > 					.increment = 1			\
> > 				},					\
> > 				.iv_size = { 0 }			\
> > 			}, }						\
> > 		}, }							\
> > 	},								\
> >
> > It shows QAT PMD is able to truncate the output digest to any size from 1 to
> 20 bytes (160 bits).
> >
> > This can be different in other PMDs, for example OpenSSL and CCP PMDs
> > allow only 20-byte digests here while Marvell MVSAM allows 12-20 bytes.
> >
> > Adam
> [Fiona] True. But as 96 bits is a standard variant described in the spec, how
> about adding
> 
> /**< HMAC using 160 bit SHA algorithm.
> HMAC-SHA-1-96 can be generated by setting digest_size to 12 */
> 
> This doesn't imply that all PMDs must support that - same as for any other
> feature, the capabilities can vary.

[Anoob] Your suggestion sounds ok to me.

One confusion though.

For SHA224 etc, the RFC specifies fixed digest size (like 224 bits or 28 bytes). So the comment in the spec/library is accurate, but QAT has digest size of 1-28 bytes. Is that correct? The comment in the spec says, "224 bit SHA algorithm" but if other digest sizes are supported, that comment becomes invalid, right?