commit ada61f05b370f588c06371c4977be0267a800ce4
author Tero Jääskö <tero.jaasko@arm.com> Mon Jan 18 17:53:48 2021 +0200
committer Tero Jääskö <tero.jaasko@arm.com> Tue Jan 19 18:43:02 2021 +0200
tree 4419bf2b58cea2516b25970984bcbcfb1e9d1a9a
parent 0241f81cbc059d0b3b1541db19212142476c8c5e

AES: masked config: remove 176 bytes of zeroing on each AES finalize round

On MBEDTLS_AES_128_BIT_MASKED+FI_COUNTERMEASURES config the rk_mask array
was unconditionally zeroed on each dummy finalization round even though
it will be zeroed after the loop too. Remove the repeated zeroing to increase
overall transfer speed on DTLS session by 17..20% on a Cortex-M4.

Signed-off-by: Tero Jääskö <tero.jaasko@arm.com>

library/aes.c

diff --git a/library/aes.c b/library/aes.c
index 061b6fd..72cfc10 100644
--- a/library/aes.c
+++ b/library/aes.c
@@ -1560,8 +1560,6 @@
         {
             flow_control++;
         }
-        //Cleanup the masked key
-        mbedtls_platform_memset( rk_masked, 0, sizeof(rk_masked) );
 #else
         aes_fround_final( aes_data_ptr->rk_ptr,
             &aes_data_ptr->xy_values[0],