diff mbox series

crypto: starfive - Pad adata with zeroes

Message ID 20231116021752.420680-1-jiajie.ho@starfivetech.com
State New
Headers
Series crypto: starfive - Pad adata with zeroes |

Commit Message

JiaJie Ho Nov. 16, 2023, 2:17 a.m. UTC 
  Ensure padding for adata is filled with zeroes. Additional bytes for
padding affects the ccm tag output even though input ad len has been
provided to the hardware.

Signed-off-by: Jia Jie Ho <jiajie.ho@starfivetech.com>
---
 drivers/crypto/starfive/jh7110-aes.c | 6 ++++--
 1 file changed, 4 insertions(+), 2 deletions(-)

Comments

Herbert Xu Nov. 16, 2023, 4:35 a.m. UTC | #1 
On Thu, Nov 16, 2023 at 10:17:52AM +0800, Jia Jie Ho wrote:
>
> diff --git a/drivers/crypto/starfive/jh7110-aes.c b/drivers/crypto/starfive/jh7110-aes.c
> index 9378e6682f0e..e0fe599f8192 100644
> --- a/drivers/crypto/starfive/jh7110-aes.c
> +++ b/drivers/crypto/starfive/jh7110-aes.c
> @@ -500,7 +500,7 @@ static int starfive_aes_prepare_req(struct skcipher_request *req,
>  	scatterwalk_start(&cryp->out_walk, rctx->out_sg);
>  
>  	if (cryp->assoclen) {
> -		rctx->adata = kzalloc(ALIGN(cryp->assoclen, AES_BLOCK_SIZE), GFP_KERNEL);
> +		rctx->adata = kzalloc(cryp->assoclen + AES_BLOCK_SIZE, GFP_KERNEL);

Please explain why you're changing the allocation size here.

This needs to go into the patch description.

Thanks,
JiaJie Ho Nov. 16, 2023, 5:39 a.m. UTC | #2 
On 16/11/2023 12:35 pm, Herbert Xu wrote:
> On Thu, Nov 16, 2023 at 10:17:52AM +0800, Jia Jie Ho wrote:
>>
>> diff --git a/drivers/crypto/starfive/jh7110-aes.c b/drivers/crypto/starfive/jh7110-aes.c
>> index 9378e6682f0e..e0fe599f8192 100644
>> --- a/drivers/crypto/starfive/jh7110-aes.c
>> +++ b/drivers/crypto/starfive/jh7110-aes.c
>> @@ -500,7 +500,7 @@ static int starfive_aes_prepare_req(struct skcipher_request *req,
>>  	scatterwalk_start(&cryp->out_walk, rctx->out_sg);
>>  
>>  	if (cryp->assoclen) {
>> -		rctx->adata = kzalloc(ALIGN(cryp->assoclen, AES_BLOCK_SIZE), GFP_KERNEL);
>> +		rctx->adata = kzalloc(cryp->assoclen + AES_BLOCK_SIZE, GFP_KERNEL);
> 
> Please explain why you're changing the allocation size here.
> 

Hi Herbert,

The hardware requires aad padded with zeroes up to 15 bytes in some cases.
This extra size and zeroing is meant for the padding and prevents driver 
accessing uninitialized memory region.

> This needs to go into the patch description.
> 

I'll update the v2 commit message if you're good with this implementation.
Thanks for reviewing this.

Jia Jie
Herbert Xu Nov. 17, 2023, 3:57 a.m. UTC | #3 
On Thu, Nov 16, 2023 at 01:39:57PM +0800, Jia Jie Ho wrote:
>
> I'll update the v2 commit message if you're good with this implementation.
> Thanks for reviewing this.

Thanks,
diff mbox series

Patch

diff --git a/drivers/crypto/starfive/jh7110-aes.c b/drivers/crypto/starfive/jh7110-aes.c
index 9378e6682f0e..e0fe599f8192 100644
--- a/drivers/crypto/starfive/jh7110-aes.c
+++ b/drivers/crypto/starfive/jh7110-aes.c
@@ -500,7 +500,7 @@  static int starfive_aes_prepare_req(struct skcipher_request *req,
 	scatterwalk_start(&cryp->out_walk, rctx->out_sg);
 
 	if (cryp->assoclen) {
-		rctx->adata = kzalloc(ALIGN(cryp->assoclen, AES_BLOCK_SIZE), GFP_KERNEL);
+		rctx->adata = kzalloc(cryp->assoclen + AES_BLOCK_SIZE, GFP_KERNEL);
 		if (!rctx->adata)
 			return dev_err_probe(cryp->dev, -ENOMEM,
 					     "Failed to alloc memory for adata");
@@ -569,7 +569,7 @@  static int starfive_aes_aead_do_one_req(struct crypto_engine *engine, void *areq
 	struct starfive_cryp_ctx *ctx =
 		crypto_aead_ctx(crypto_aead_reqtfm(req));
 	struct starfive_cryp_dev *cryp = ctx->cryp;
-	struct starfive_cryp_request_ctx *rctx = ctx->rctx;
+	struct starfive_cryp_request_ctx *rctx;
 	u32 block[AES_BLOCK_32];
 	u32 stat;
 	int err;
@@ -579,6 +579,8 @@  static int starfive_aes_aead_do_one_req(struct crypto_engine *engine, void *areq
 	if (err)
 		return err;
 
+	rctx = ctx->rctx;
+
 	if (!cryp->assoclen)
 		goto write_text;