[<prev] [next>] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210203154210.246667-1-daniele.alessandrelli@linux.intel.com>
Date: Wed, 3 Feb 2021 15:42:10 +0000
From: Daniele Alessandrelli <daniele.alessandrelli@...ux.intel.com>
To: herbert@...dor.apana.org.au
Cc: daniele.alessandrelli@...el.com, linux-crypto@...r.kernel.org,
linux-kernel@...r.kernel.org
Subject: [PATCH] crypto: keembay-ocs-aes - Fix 'q' assignment during CCM B0 generation
From: Daniele Alessandrelli <daniele.alessandrelli@...el.com>
In ocs_aes_ccm_write_b0(), 'q' (the octet length of the binary
representation of the octet length of the payload) is set to 'iv[0]',
while it should be set to 'iv[0] & 0x7' (i.e., only the last 3
bits of iv[0] should be used), as documented in NIST Special Publication
800-38C:
https://nvlpubs.nist.gov/nistpubs/Legacy/SP/nistspecialpublication800-38c.pdf
In practice, this is not an issue, since 'iv[0]' is checked to be in the
range [1-7] by ocs_aes_validate_inputs(), but let's fix the assignment
anyway, in order to make the code more robust.
Signed-off-by: Daniele Alessandrelli <daniele.alessandrelli@...el.com>
---
drivers/crypto/keembay/ocs-aes.c | 6 +++---
1 file changed, 3 insertions(+), 3 deletions(-)
diff --git a/drivers/crypto/keembay/ocs-aes.c b/drivers/crypto/keembay/ocs-aes.c
index b85c89477afa..be9f32fc8f42 100644
--- a/drivers/crypto/keembay/ocs-aes.c
+++ b/drivers/crypto/keembay/ocs-aes.c
@@ -1080,15 +1080,15 @@ static int ocs_aes_ccm_write_b0(const struct ocs_aes_dev *aes_dev,
/*
* q is the octet length of Q.
* q can only be an element of {2, 3, 4, 5, 6, 7, 8} and is encoded as
- * q - 1 == iv[0]
+ * q - 1 == iv[0] & 0x7;
*/
b0[0] |= iv[0] & 0x7;
/*
* Copy the Nonce N from IV to B0; N is located in iv[1]..iv[15 - q]
* and must be copied to b0[1]..b0[15-q].
- * q == iv[0] + 1
+ * q == (iv[0] & 0x7) + 1
*/
- q = iv[0] + 1;
+ q = (iv[0] & 0x7) + 1;
for (i = 1; i <= 15 - q; i++)
b0[i] = iv[i];
/*
--
2.26.2
Powered by blists - more mailing lists