RE: [EXT] [PATCH] cryptodev: add support for 25519 and 448 curves

["Kusztal, ArkadiuszX" <arkadiuszx.kusztal@intel.com>]

> -----Original Message-----
> From: Akhil Goyal <gakhil@marvell.com>
> Sent: Monday, May 16, 2022 8:58 PM
> To: Kusztal, ArkadiuszX <arkadiuszx.kusztal@intel.com>; dev@dpdk.org
> Cc: Zhang, Roy Fan <roy.fan.zhang@intel.com>
> Subject: RE: [EXT] [PATCH] cryptodev: add support for 25519 and 448 curves
> 
> > This commit adds support for following elliptic curves:
> > 1) Curve25519
> > 2) Curve448
> >
> > Signed-off-by: Arek Kusztal <arkadiuszx.kusztal@intel.com>
> > ---
> >  lib/cryptodev/rte_crypto_asym.h | 14 ++++++++++++--
> >  1 file changed, 12 insertions(+), 2 deletions(-)
> >
> > diff --git a/lib/cryptodev/rte_crypto_asym.h
> > b/lib/cryptodev/rte_crypto_asym.h index cd24d4b07b..775b2f6277 100644
> > --- a/lib/cryptodev/rte_crypto_asym.h
> > +++ b/lib/cryptodev/rte_crypto_asym.h
> > @@ -48,6 +48,8 @@ enum rte_crypto_ec_group {
> >  	RTE_CRYPTO_EC_GROUP_SECP256R1 = 23,
> >  	RTE_CRYPTO_EC_GROUP_SECP384R1 = 24,
> >  	RTE_CRYPTO_EC_GROUP_SECP521R1 = 25,
> > +	RTE_CRYPTO_EC_GROUP_CURVE25519 = 29,
> > +	RTE_CRYPTO_EC_GROUP_CURVE448 = 30,
> >  };
> >
> >  /**
> > @@ -180,9 +182,17 @@ typedef rte_crypto_param rte_crypto_uint;
> >   */
> >  struct rte_crypto_ec_point {
> >  	rte_crypto_param x;
> > -	/**< X coordinate */
> > +	/**<
> > +	 * X coordinate
> > +	 * For curve25519 and curve448 - little-endian integer
> > +	 * otherwise, big-endian integer
> > +	 */
> >  	rte_crypto_param y;
> > -	/**< Y coordinate */
> > +	/**<
> > +	 * Y coordinate
> > +	 * For curve25519 and curve448 - little-endian integer
> > +	 * otherwise, big-endian integer
> > +	 */
> Can you give reference of the document which specify this endianness?
[Arek] - sure, I may give rfc reference here, but if it will go into crypodev in this form I am not yet sure.
These curves could be used with DH, but cannot be used with ECDSA. Even with DH it may be that we will go with separate {dh_op, ecdh_op, x25519_op, x448_op} but this would make TLS group reference pointless, and we would not add Montgomery/Edwards curves at all as an enum.
> 
> And if it is implicit as per the protocol, do we need to add explicit comments
> here?