[PATCH v2 0/5] crypto: hisilicon - some misc bugfix for SEC engine

[PATCH v2 0/5] crypto: hisilicon - some misc bugfix for SEC engine

[Kai Ye]

some misc bugfix for SEC engine.

chages v1->v2:
	add rate limited for printing messages.

Kai Ye (5):
  crypto: hisilicon/sec - fixup icv checking enabled on Kunpeng 930
  crypto: hisilicon/sec - delete the print of fallback tfm application
    failure
  crypto: hisilicon/sec - fix the max length of AAD for the CCM mode
  crypto: hisilicon/sec - fix the CTR mode BD configuration
  crypto: hisilicon/sec - use the correct print format

 drivers/crypto/hisilicon/sec2/sec_crypto.c | 30 ++++++++++++++++++++----------
 drivers/crypto/hisilicon/sec2/sec_crypto.h |  6 ++++--
 2 files changed, 24 insertions(+), 12 deletions(-)

-- 
2.7.4

[PATCH v2 1/5] crypto: hisilicon/sec - fixup icv checking enabled on Kunpeng 930

[Kai Ye]

Fixup icv(integrity check value) checking enabled wrong on
Kunpeng 930

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_crypto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index 6a45bd23..bf93c98 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -1614,7 +1614,7 @@ static void sec_auth_bd_fill_ex_v3(struct sec_auth_ctx *ctx, int dir,
 		sqe3->auth_mac_key |= cpu_to_le32((u32)SEC_AUTH_TYPE1);
 		sqe3->huk_iv_seq &= SEC_CIPHER_AUTH_V3;
 	} else {
-		sqe3->auth_mac_key |= cpu_to_le32((u32)SEC_AUTH_TYPE1);
+		sqe3->auth_mac_key |= cpu_to_le32((u32)SEC_AUTH_TYPE2);
 		sqe3->huk_iv_seq |= SEC_AUTH_CIPHER_V3;
 	}
 	sqe3->a_len_key = cpu_to_le32(c_req->c_len + aq->assoclen);
-- 
2.7.4

[PATCH v2 2/5] crypto: hisilicon/sec - delete the print of fallback tfm application failure

[Kai Ye]

Modify the print of information that might lead to user misunderstanding.
Currently only XTS mode need the fallback tfm when using 192bit key.
Others algs not need soft fallback tfm. So others algs can return
directly.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_crypto.c | 15 +++++++--------
 1 file changed, 7 insertions(+), 8 deletions(-)

diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index bf93c98..890faf8 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -641,15 +641,15 @@ static int sec_skcipher_fbtfm_init(struct crypto_skcipher *tfm)
 	struct sec_cipher_ctx *c_ctx = &ctx->c_ctx;
 
 	c_ctx->fallback = false;
+
+	/* Currently, only XTS mode need fallback tfm when using 192bit key */
 	if (likely(strncmp(alg, "xts", SEC_XTS_NAME_SZ)))
 		return 0;
 
 	c_ctx->fbtfm = crypto_alloc_sync_skcipher(alg, 0,
 						  CRYPTO_ALG_NEED_FALLBACK);
-	if (IS_ERR(c_ctx->fbtfm)) {
-		pr_err("failed to alloc fallback tfm!\n");
-		return PTR_ERR(c_ctx->fbtfm);
-	}
+	if (IS_ERR(c_ctx->fbtfm))
+		c_ctx->fbtfm = NULL;
 
 	return 0;
 }
@@ -808,7 +808,7 @@ static int sec_skcipher_setkey(struct crypto_skcipher *tfm, const u8 *key,
 	}
 
 	memcpy(c_ctx->c_key, key, keylen);
-	if (c_ctx->fallback) {
+	if (c_ctx->fallback && c_ctx->fbtfm) {
 		ret = crypto_sync_skcipher_setkey(c_ctx->fbtfm, key, keylen);
 		if (ret) {
 			dev_err(dev, "failed to set fallback skcipher key!\n");
@@ -2032,13 +2032,12 @@ static int sec_skcipher_soft_crypto(struct sec_ctx *ctx,
 				    struct skcipher_request *sreq, bool encrypt)
 {
 	struct sec_cipher_ctx *c_ctx = &ctx->c_ctx;
+	SYNC_SKCIPHER_REQUEST_ON_STACK(subreq, c_ctx->fbtfm);
 	struct device *dev = ctx->dev;
 	int ret;
 
-	SYNC_SKCIPHER_REQUEST_ON_STACK(subreq, c_ctx->fbtfm);
-
 	if (!c_ctx->fbtfm) {
-		dev_err(dev, "failed to check fallback tfm\n");
+		dev_err_ratelimited(dev, "the soft tfm isn't supported in the current system.\n");
 		return -EINVAL;
 	}
 
-- 
2.7.4

[PATCH v2 3/5] crypto: hisilicon/sec - fix the max length of AAD for the CCM mode

[Kai Ye]

Fix the maximum length of AAD for the CCM mode due to the hardware limited.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_crypto.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index 890faf8..387fb3c 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -63,6 +63,7 @@
 #define SEC_AUTH_CIPHER		0x1
 #define SEC_MAX_MAC_LEN		64
 #define SEC_MAX_AAD_LEN		65535
+#define SEC_MAX_CCM_AAD_LEN	65279
 #define SEC_TOTAL_MAC_SZ	(SEC_MAX_MAC_LEN * QM_Q_DEPTH)
 
 #define SEC_PBUF_SZ			512
@@ -2218,6 +2219,10 @@ static int sec_aead_spec_check(struct sec_ctx *ctx, struct sec_req *sreq)
 	}
 
 	if (c_mode == SEC_CMODE_CCM) {
+		if (unlikely(req->assoclen > SEC_MAX_CCM_AAD_LEN)) {
+			dev_err_ratelimited(dev, "CCM input aad parameter is too long!\n");
+			return -EINVAL;
+		}
 		ret = aead_iv_demension_check(req);
 		if (ret) {
 			dev_err(dev, "aead input iv param error!\n");
-- 
2.7.4

[PATCH v2 4/5] crypto: hisilicon/sec - fix the CTR mode BD configuration

[Kai Ye]

The CTR counter is 32bit rollover default on the BD.
But the NIST standard is 128bit rollover. it cause the
testing failed, so need to fix the BD configuration.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_crypto.c | 6 ++++++
 drivers/crypto/hisilicon/sec2/sec_crypto.h | 6 ++++--
 2 files changed, 10 insertions(+), 2 deletions(-)

diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index 387fb3c..5be1034 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -42,6 +42,8 @@
 #define SEC_DE_OFFSET_V3		9
 #define SEC_SCENE_OFFSET_V3	5
 #define SEC_CKEY_OFFSET_V3	13
+#define SEC_CTR_CNT_OFFSET	25
+#define SEC_CTR_CNT_ROLLOVER	2
 #define SEC_SRC_SGL_OFFSET_V3	11
 #define SEC_DST_SGL_OFFSET_V3	14
 #define SEC_CALG_OFFSET_V3	4
@@ -1301,6 +1303,10 @@ static int sec_skcipher_bd_fill_v3(struct sec_ctx *ctx, struct sec_req *req)
 		cipher = SEC_CIPHER_DEC;
 	sec_sqe3->c_icv_key |= cpu_to_le16(cipher);
 
+	/* Set the CTR counter mode is 128bit rollover */
+	sec_sqe3->auth_mac_key = cpu_to_le32((u32)SEC_CTR_CNT_ROLLOVER <<
+					SEC_CTR_CNT_OFFSET);
+
 	if (req->use_pbuf) {
 		bd_param |= SEC_PBUF << SEC_SRC_SGL_OFFSET_V3;
 		bd_param |= SEC_PBUF << SEC_DST_SGL_OFFSET_V3;
diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.h b/drivers/crypto/hisilicon/sec2/sec_crypto.h
index 9f71c35..5e039b5 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.h
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.h
@@ -354,8 +354,10 @@ struct sec_sqe3 {
 	 * akey_len: 9~14 bits
 	 * a_alg: 15~20 bits
 	 * key_sel: 21~24 bits
-	 * updata_key: 25 bits
-	 * reserved: 26~31 bits
+	 * ctr_count_mode/sm4_xts: 25~26 bits
+	 * sva_prefetch: 27 bits
+	 * key_wrap_num: 28~30 bits
+	 * update_key: 31 bits
 	 */
 	__le32 auth_mac_key;
 	__le32 salt;
-- 
2.7.4

[PATCH v2 5/5] crypto: hisilicon/sec - use the correct print format

[Kai Ye]

Use the correct print format. Printing an unsigned int value should
use %u instead of %d.

Signed-off-by: Kai Ye <yekai13@huawei.com>
---
 drivers/crypto/hisilicon/sec2/sec_crypto.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index 5be1034..3b698a2 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -240,7 +240,7 @@ static void sec_req_cb(struct hisi_qp *qp, void *resp)
 
 	if (unlikely(type != type_supported)) {
 		atomic64_inc(&dfx->err_bd_cnt);
-		pr_err("err bd type [%d]\n", type);
+		pr_err("err bd type [%u]\n", type);
 		return;
 	}
 
-- 
2.7.4

Re: [PATCH v2 2/5] crypto: hisilicon/sec - delete the print of fallback tfm application failure

[Herbert Xu]

On Fri, Aug 06, 2021 at 05:44:23PM +0800, Kai Ye wrote:
> Modify the print of information that might lead to user misunderstanding.
> Currently only XTS mode need the fallback tfm when using 192bit key.
> Others algs not need soft fallback tfm. So others algs can return
> directly.
> 
> Signed-off-by: Kai Ye <yekai13@huawei.com>
> ---
>  drivers/crypto/hisilicon/sec2/sec_crypto.c | 15 +++++++--------
>  1 file changed, 7 insertions(+), 8 deletions(-)

I still don't get the point of this change.  Fallbacks are typically
the software implementation, and they should never fail except when
you run out of memory.  So what do you gain by delaying the error
from allocation time to runtime?

Cheers,
-- 
Email: Herbert Xu <herbert@gondor.apana.org.au>
Home Page: http://gondor.apana.org.au/~herbert/
PGP Key: http://gondor.apana.org.au/~herbert/pubkey.txt