[dpdk-dev] [PATCH v1] crypto/qat: add support for 8 byte 3DES

[Trahe, Fiona]

Hi Marko,

> -----Original Message-----
> From: dev [mailto:dev-bounces at dpdk.org] On Behalf Of Kovacevic, Marko
> Sent: Tuesday, June 12, 2018 11:40 AM
> To: dev at dpdk.org
> Cc: De Lara Guarch, Pablo <pablo.de.lara.guarch at intel.com>; Jain, Deepak K
> <deepak.k.jain at intel.com>; Kovacevic, Marko <marko.kovacevic at intel.com>; Marko at dpdk.org
> Subject: [dpdk-dev] [PATCH v1] crypto/qat: add support for 8 byte 3DES
> 
> Added extra case to support 8 byte key size
> for 3DES CBC. Also changed capabilities to reflect
> the change.
> 
> Signed-off-by: Marko, Kovacevic <marko.kovacevic at intel.com>
> ---
>  drivers/crypto/qat/qat_adf/qat_algs.h            | 1 +
>  drivers/crypto/qat/qat_adf/qat_algs_build_desc.c | 1 +
>  drivers/crypto/qat/qat_crypto_capabilities.h     | 2 +-
>  test/test/test_cryptodev_des_test_vectors.h      | 6 ++++--
>  4 files changed, 7 insertions(+), 3 deletions(-)

This piece of QAT code should also be modified to copy K1 to K2 and K3:
if ((cdesc->qat_cipher_alg == ICP_QAT_HW_CIPHER_ALGO_3DES)
	&& (cipherkeylen == QAT_3DES_KEY_SZ_OPT2))
	/* K3 not provided so use K1 = K3*/
	memcpy(cdesc->cd_cur_ptr, cipherkey, padding_size);
else
	memset(cdesc->cd_cur_ptr, 0, padding_size);
	cdesc->cd_cur_ptr += padding_size;

I can see from the testcode added in the aesni_mb patch, that the test vector output
(ciphertext512_des) is the same as for plain DES. So as QAT would pad with 0s, this probably
explains why the tests are passing. But I can't see how the test can pass on aesni_mb
as K1 is being correctly copied there. I must be missing something?