DPDK patches and discussions
 help / color / mirror / Atom feed
* [PATCH] app/crypto-perf: fix aad offset alignment
@ 2025-03-11  9:01 Shani Peretz
  2025-03-17 10:22 ` [EXTERNAL] " Akhil Goyal
  0 siblings, 1 reply; 2+ messages in thread
From: Shani Peretz @ 2025-03-11  9:01 UTC (permalink / raw)
  To: dev
  Cc: suanmingm, Shani Peretz, stable, Brian Dooley, Pablo de Lara,
	Akhil Goyal

AAD offset in AES-GCM crypto test was calculated by adding
16-byte alignment after the IV, which is only needed in AES-CCM.

The patch correct the AAD offset calculation in AES-GCM algorithm tests.

Fixes: 0b242422d385 ("app/crypto-perf: set AAD after the crypto operation")
Cc: stable@dpdk.org

Signed-off-by: Shani Peretz <shperetz@nvidia.com>
---
 app/test-crypto-perf/cperf_ops.c | 4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/app/test-crypto-perf/cperf_ops.c b/app/test-crypto-perf/cperf_ops.c
index 6d5f510220..f9be51e17f 100644
--- a/app/test-crypto-perf/cperf_ops.c
+++ b/app/test-crypto-perf/cperf_ops.c
@@ -688,7 +688,9 @@ cperf_set_ops_aead(struct rte_crypto_op **ops,
 	uint16_t i;
 	/* AAD is placed after the IV */
 	uint16_t aad_offset = iv_offset +
-			RTE_ALIGN_CEIL(test_vector->aead_iv.length, 16);
+			((options->aead_algo == RTE_CRYPTO_AEAD_AES_CCM) ?
+			RTE_ALIGN_CEIL(test_vector->aead_iv.length, 16) :
+			test_vector->aead_iv.length);
 
 	for (i = 0; i < nb_ops; i++) {
 		struct rte_crypto_sym_op *sym_op = ops[i]->sym;
-- 
2.25.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

* RE: [EXTERNAL] [PATCH] app/crypto-perf: fix aad offset alignment
  2025-03-11  9:01 [PATCH] app/crypto-perf: fix aad offset alignment Shani Peretz
@ 2025-03-17 10:22 ` Akhil Goyal
  0 siblings, 0 replies; 2+ messages in thread
From: Akhil Goyal @ 2025-03-17 10:22 UTC (permalink / raw)
  To: Shani Peretz, dev; +Cc: suanmingm, stable, Brian Dooley, Pablo de Lara

Hi,
> AAD offset in AES-GCM crypto test was calculated by adding
> 16-byte alignment after the IV, which is only needed in AES-CCM.

Agreed that CCM has a requirement for 16B alignment.
But for GCM, does it break any protocol? Can we not align to byte boundary for performance?
This is a performance application which mainly focus on getting the best throughput.
Did you check if it is having some performance degradation?

> 
> The patch correct the AAD offset calculation in AES-GCM algorithm tests.
> 
> Fixes: 0b242422d385 ("app/crypto-perf: set AAD after the crypto operation")
> Cc: stable@dpdk.org
> 
> Signed-off-by: Shani Peretz <shperetz@nvidia.com>
> ---
>  app/test-crypto-perf/cperf_ops.c | 4 +++-
>  1 file changed, 3 insertions(+), 1 deletion(-)
> 
> diff --git a/app/test-crypto-perf/cperf_ops.c b/app/test-crypto-perf/cperf_ops.c
> index 6d5f510220..f9be51e17f 100644
> --- a/app/test-crypto-perf/cperf_ops.c
> +++ b/app/test-crypto-perf/cperf_ops.c
> @@ -688,7 +688,9 @@ cperf_set_ops_aead(struct rte_crypto_op **ops,
>  	uint16_t i;
>  	/* AAD is placed after the IV */
>  	uint16_t aad_offset = iv_offset +
> -			RTE_ALIGN_CEIL(test_vector->aead_iv.length, 16);
> +			((options->aead_algo == RTE_CRYPTO_AEAD_AES_CCM)
> ?
> +			RTE_ALIGN_CEIL(test_vector->aead_iv.length, 16) :
> +			test_vector->aead_iv.length);
> 
>  	for (i = 0; i < nb_ops; i++) {
>  		struct rte_crypto_sym_op *sym_op = ops[i]->sym;
> --
> 2.25.1


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2025-03-17 10:22 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2025-03-11  9:01 [PATCH] app/crypto-perf: fix aad offset alignment Shani Peretz
2025-03-17 10:22 ` [EXTERNAL] " Akhil Goyal

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).