6.14-stable review patch. If anyone has any objections, please let me know.
------------------
From: Akhil R akhilrajeev@nvidia.com
[ Upstream commit 1ddaff40c08abb926be5ba713c5efc412d0836c5 ]
Modifying the crypto_request turns out to be not the right way to handle the stale value issue with the IV. Though the IV is not used for AES ECB, it eventually get used in algorithms like LRW in the next step after AES ECB encryption/decryption. Setting req->iv to NULL breaks the implementation of such algorithms. Hence modify only the local reqctx to check for IV.
Fixes: bde558220866 ("crypto: tegra - Set IV to NULL explicitly for AES ECB") Signed-off-by: Akhil R akhilrajeev@nvidia.com Signed-off-by: Herbert Xu herbert@gondor.apana.org.au Signed-off-by: Sasha Levin sashal@kernel.org --- drivers/crypto/tegra/tegra-se-aes.c | 5 +---- 1 file changed, 1 insertion(+), 4 deletions(-)
diff --git a/drivers/crypto/tegra/tegra-se-aes.c b/drivers/crypto/tegra/tegra-se-aes.c index ca9d0cca1f748..0e07d0523291a 100644 --- a/drivers/crypto/tegra/tegra-se-aes.c +++ b/drivers/crypto/tegra/tegra-se-aes.c @@ -269,7 +269,7 @@ static int tegra_aes_do_one_req(struct crypto_engine *engine, void *areq) unsigned int cmdlen, key1_id, key2_id; int ret;
- rctx->iv = (u32 *)req->iv; + rctx->iv = (ctx->alg == SE_ALG_ECB) ? NULL : (u32 *)req->iv; rctx->len = req->cryptlen; key1_id = ctx->key1_id; key2_id = ctx->key2_id; @@ -498,9 +498,6 @@ static int tegra_aes_crypt(struct skcipher_request *req, bool encrypt) if (!req->cryptlen) return 0;
- if (ctx->alg == SE_ALG_ECB) - req->iv = NULL; - rctx->encrypt = encrypt;
return crypto_transfer_skcipher_request_to_engine(ctx->se->engine, req);