[PATCH 0/4] crypto: hisilicon - Misc fixed on HPRE

[PATCH 0/4] crypto: hisilicon - Misc fixed on HPRE

[Zaibo Xu]

1.Bugfixed tfm leak with some other tiny bugfixed.
2.Fixed some tiny bugs and update some code style.
3.Adjust input parameter order of hpre_crt_para_get.
4.Add branch prediction macro on hot path with small performance gain(~1%).

This series is based on:
git://git.kernel.org/pub/scm/linux/kernel/git/herbert/cryptodev-2.6.git

Zaibo Xu (4):
  crypto: hisilicon - Bugfixed tfm leak
  crypto: hisilicon - Fixed some tiny bugs of HPRE
  crypto: hisilicon - adjust hpre_crt_para_get
  crypto: hisilicon - add branch prediction macro

 drivers/crypto/hisilicon/hpre/hpre_crypto.c | 141 ++++++++++++++--------------
 drivers/crypto/hisilicon/hpre/hpre_main.c   |  32 ++++---
 2 files changed, 86 insertions(+), 87 deletions(-)

-- 
2.8.1

[PATCH 1/4] crypto: hisilicon - Bugfixed tfm leak

[Zaibo Xu]

1.Fixed the bug of software tfm leakage.
2.Update HW error log message.

Signed-off-by: Zaibo Xu <xuzaibo@huawei.com>
---
 drivers/crypto/hisilicon/hpre/hpre_crypto.c |  7 ++++++-
 drivers/crypto/hisilicon/hpre/hpre_main.c   | 24 ++++++++++++------------
 2 files changed, 18 insertions(+), 13 deletions(-)

diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
index 98f037e..d8b0152 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_crypto.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
@@ -1043,6 +1043,7 @@ static unsigned int hpre_rsa_max_size(struct crypto_akcipher *tfm)
 static int hpre_rsa_init_tfm(struct crypto_akcipher *tfm)
 {
 	struct hpre_ctx *ctx = akcipher_tfm_ctx(tfm);
+	int ret;
 
 	ctx->rsa.soft_tfm = crypto_alloc_akcipher("rsa-generic", 0, 0);
 	if (IS_ERR(ctx->rsa.soft_tfm)) {
@@ -1050,7 +1051,11 @@ static int hpre_rsa_init_tfm(struct crypto_akcipher *tfm)
 		return PTR_ERR(ctx->rsa.soft_tfm);
 	}
 
-	return hpre_ctx_init(ctx);
+	ret = hpre_ctx_init(ctx);
+	if (ret)
+		crypto_free_akcipher(ctx->rsa.soft_tfm);
+
+	return ret;
 }
 
 static void hpre_rsa_exit_tfm(struct crypto_akcipher *tfm)
diff --git a/drivers/crypto/hisilicon/hpre/hpre_main.c b/drivers/crypto/hisilicon/hpre/hpre_main.c
index 711f5d1..753e43d 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_main.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_main.c
@@ -106,18 +106,18 @@ static const char * const hpre_debug_file_name[] = {
 };
 
 static const struct hpre_hw_error hpre_hw_errors[] = {
-	{ .int_msk = BIT(0), .msg = "hpre_ecc_1bitt_err" },
-	{ .int_msk = BIT(1), .msg = "hpre_ecc_2bit_err" },
-	{ .int_msk = BIT(2), .msg = "hpre_data_wr_err" },
-	{ .int_msk = BIT(3), .msg = "hpre_data_rd_err" },
-	{ .int_msk = BIT(4), .msg = "hpre_bd_rd_err" },
-	{ .int_msk = BIT(5), .msg = "hpre_ooo_2bit_ecc_err" },
-	{ .int_msk = BIT(6), .msg = "hpre_cltr1_htbt_tm_out_err" },
-	{ .int_msk = BIT(7), .msg = "hpre_cltr2_htbt_tm_out_err" },
-	{ .int_msk = BIT(8), .msg = "hpre_cltr3_htbt_tm_out_err" },
-	{ .int_msk = BIT(9), .msg = "hpre_cltr4_htbt_tm_out_err" },
-	{ .int_msk = GENMASK(15, 10), .msg = "hpre_ooo_rdrsp_err" },
-	{ .int_msk = GENMASK(21, 16), .msg = "hpre_ooo_wrrsp_err" },
+	{ .int_msk = BIT(0), .msg = "core_ecc_1bit_err_int_set" },
+	{ .int_msk = BIT(1), .msg = "core_ecc_2bit_err_int_set" },
+	{ .int_msk = BIT(2), .msg = "dat_wb_poison_int_set" },
+	{ .int_msk = BIT(3), .msg = "dat_rd_poison_int_set" },
+	{ .int_msk = BIT(4), .msg = "bd_rd_poison_int_set" },
+	{ .int_msk = BIT(5), .msg = "ooo_ecc_2bit_err_int_set" },
+	{ .int_msk = BIT(6), .msg = "cluster1_shb_timeout_int_set" },
+	{ .int_msk = BIT(7), .msg = "cluster2_shb_timeout_int_set" },
+	{ .int_msk = BIT(8), .msg = "cluster3_shb_timeout_int_set" },
+	{ .int_msk = BIT(9), .msg = "cluster4_shb_timeout_int_set" },
+	{ .int_msk = GENMASK(15, 10), .msg = "ooo_rdrsp_err_int_set" },
+	{ .int_msk = GENMASK(21, 16), .msg = "ooo_wrrsp_err_int_set" },
 	{ /* sentinel */ }
 };
 
-- 
2.8.1

[PATCH 2/4] crypto: hisilicon - Fixed some tiny bugs of HPRE

[Zaibo Xu]

1.Use memzero_explicit to clear key;
2.Fix some little endian writings;
3.Fix some other bugs and stuff of code style;

Signed-off-by: Zaibo Xu <xuzaibo@huawei.com>
---
 drivers/crypto/hisilicon/hpre/hpre_crypto.c | 62 +++++++++++++----------------
 drivers/crypto/hisilicon/hpre/hpre_main.c   |  8 ++--
 2 files changed, 32 insertions(+), 38 deletions(-)

diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
index d8b0152..76540a1 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_crypto.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
@@ -174,8 +174,8 @@ static struct hisi_qp *hpre_get_qp_and_start(void)
 }
 
 static int hpre_get_data_dma_addr(struct hpre_asym_request *hpre_req,
-			      struct scatterlist *data, unsigned int len,
-			      int is_src, dma_addr_t *tmp)
+				  struct scatterlist *data, unsigned int len,
+				  int is_src, dma_addr_t *tmp)
 {
 	struct hpre_ctx *ctx = hpre_req->ctx;
 	struct device *dev = HPRE_DEV(ctx);
@@ -199,8 +199,8 @@ static int hpre_get_data_dma_addr(struct hpre_asym_request *hpre_req,
 }
 
 static int hpre_prepare_dma_buf(struct hpre_asym_request *hpre_req,
-			       struct scatterlist *data, unsigned int len,
-			       int is_src, dma_addr_t *tmp)
+				struct scatterlist *data, unsigned int len,
+				int is_src, dma_addr_t *tmp)
 {
 	struct hpre_ctx *ctx = hpre_req->ctx;
 	struct device *dev = HPRE_DEV(ctx);
@@ -226,12 +226,12 @@ static int hpre_prepare_dma_buf(struct hpre_asym_request *hpre_req,
 }
 
 static int hpre_hw_data_init(struct hpre_asym_request *hpre_req,
-			 struct scatterlist *data, unsigned int len,
-			 int is_src, int is_dh)
+			     struct scatterlist *data, unsigned int len,
+			     int is_src, int is_dh)
 {
 	struct hpre_sqe *msg = &hpre_req->req;
 	struct hpre_ctx *ctx = hpre_req->ctx;
-	dma_addr_t tmp;
+	dma_addr_t tmp = 0;
 	int ret;
 
 	/* when the data is dh's source, we should format it */
@@ -253,8 +253,9 @@ static int hpre_hw_data_init(struct hpre_asym_request *hpre_req,
 }
 
 static void hpre_hw_data_clr_all(struct hpre_ctx *ctx,
-			     struct hpre_asym_request *req,
-			     struct scatterlist *dst, struct scatterlist *src)
+				 struct hpre_asym_request *req,
+				 struct scatterlist *dst,
+				 struct scatterlist *src)
 {
 	struct device *dev = HPRE_DEV(ctx);
 	struct hpre_sqe *sqe = &req->req;
@@ -288,7 +289,7 @@ static void hpre_hw_data_clr_all(struct hpre_ctx *ctx,
 }
 
 static int hpre_alg_res_post_hf(struct hpre_ctx *ctx, struct hpre_sqe *sqe,
-			    void **kreq)
+				void **kreq)
 {
 	struct hpre_asym_request *req;
 	int err, id, done;
@@ -375,7 +376,7 @@ static void hpre_alg_cb(struct hisi_qp *qp, void *resp)
 	struct hpre_ctx *ctx = qp->qp_ctx;
 	struct hpre_sqe *sqe = resp;
 
-	ctx->req_list[sqe->tag]->cb(ctx, resp);
+	ctx->req_list[le16_to_cpu(sqe->tag)]->cb(ctx, resp);
 }
 
 static int hpre_ctx_init(struct hpre_ctx *ctx)
@@ -454,9 +455,6 @@ static int hpre_dh_compute_value(struct kpp_request *req)
 	int ctr = 0;
 	int ret;
 
-	if (!ctx)
-		return -EINVAL;
-
 	ret = hpre_msg_request_set(ctx, req, false);
 	if (ret)
 		return ret;
@@ -472,9 +470,9 @@ static int hpre_dh_compute_value(struct kpp_request *req)
 		goto clear_all;
 
 	if (ctx->crt_g2_mode && !req->src)
-		msg->dw0 |= HPRE_ALG_DH_G2;
+		msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH_G2);
 	else
-		msg->dw0 |= HPRE_ALG_DH;
+		msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) | HPRE_ALG_DH);
 	do {
 		ret = hisi_qp_send(ctx->qp, msg);
 	} while (ret == -EBUSY && ctr++ < HPRE_TRY_SEND_TIMES);
@@ -520,12 +518,12 @@ static int hpre_dh_set_params(struct hpre_ctx *ctx, struct dh *params)
 		return -EINVAL;
 
 	if (hpre_is_dh_params_length_valid(params->p_size <<
-		HPRE_BITS_2_BYTES_SHIFT))
+					   HPRE_BITS_2_BYTES_SHIFT))
 		return -EINVAL;
 
 	sz = ctx->key_sz = params->p_size;
 	ctx->dh.xa_p = dma_alloc_coherent(dev, sz << 1,
-				&ctx->dh.dma_xa_p, GFP_KERNEL);
+					  &ctx->dh.dma_xa_p, GFP_KERNEL);
 	if (!ctx->dh.xa_p)
 		return -ENOMEM;
 
@@ -559,13 +557,12 @@ static void hpre_dh_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all)
 		hisi_qm_stop_qp(ctx->qp);
 
 	if (ctx->dh.g) {
-		memset(ctx->dh.g, 0, sz);
 		dma_free_coherent(dev, sz, ctx->dh.g, ctx->dh.dma_g);
 		ctx->dh.g = NULL;
 	}
 
 	if (ctx->dh.xa_p) {
-		memset(ctx->dh.xa_p, 0, sz);
+		memzero_explicit(ctx->dh.xa_p, sz);
 		dma_free_coherent(dev, sz << 1, ctx->dh.xa_p,
 				  ctx->dh.dma_xa_p);
 		ctx->dh.xa_p = NULL;
@@ -661,9 +658,6 @@ static int hpre_rsa_enc(struct akcipher_request *req)
 	int ctr = 0;
 	int ret;
 
-	if (!ctx)
-		return -EINVAL;
-
 	/* For 512 and 1536 bits key size, use soft tfm instead */
 	if (ctx->key_sz == HPRE_RSA_512BITS_KSZ ||
 	    ctx->key_sz == HPRE_RSA_1536BITS_KSZ) {
@@ -680,7 +674,7 @@ static int hpre_rsa_enc(struct akcipher_request *req)
 	if (ret)
 		return ret;
 
-	msg->dw0 |= HPRE_ALG_NC_NCRT;
+	msg->dw0 |= cpu_to_le32(HPRE_ALG_NC_NCRT);
 	msg->key = cpu_to_le64((u64)ctx->rsa.dma_pubkey);
 
 	ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0);
@@ -716,9 +710,6 @@ static int hpre_rsa_dec(struct akcipher_request *req)
 	int ctr = 0;
 	int ret;
 
-	if (!ctx)
-		return -EINVAL;
-
 	/* For 512 and 1536 bits key size, use soft tfm instead */
 	if (ctx->key_sz == HPRE_RSA_512BITS_KSZ ||
 	    ctx->key_sz == HPRE_RSA_1536BITS_KSZ) {
@@ -737,10 +728,12 @@ static int hpre_rsa_dec(struct akcipher_request *req)
 
 	if (ctx->crt_g2_mode) {
 		msg->key = cpu_to_le64((u64)ctx->rsa.dma_crt_prikey);
-		msg->dw0 |= HPRE_ALG_NC_CRT;
+		msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) |
+				       HPRE_ALG_NC_CRT);
 	} else {
 		msg->key = cpu_to_le64((u64)ctx->rsa.dma_prikey);
-		msg->dw0 |= HPRE_ALG_NC_NCRT;
+		msg->dw0 = cpu_to_le32(le32_to_cpu(msg->dw0) |
+				       HPRE_ALG_NC_NCRT);
 	}
 
 	ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0);
@@ -811,10 +804,8 @@ static int hpre_rsa_set_e(struct hpre_ctx *ctx, const char *value,
 
 	hpre_rsa_drop_leading_zeros(&ptr, &vlen);
 
-	if (!ctx->key_sz || !vlen || vlen > ctx->key_sz) {
-		ctx->rsa.pubkey = NULL;
+	if (!ctx->key_sz || !vlen || vlen > ctx->key_sz)
 		return -EINVAL;
-	}
 
 	memcpy(ctx->rsa.pubkey + ctx->key_sz - vlen, ptr, vlen);
 
@@ -899,7 +890,7 @@ static int hpre_rsa_setkey_crt(struct hpre_ctx *ctx, struct rsa_key *rsa_key)
 
 free_key:
 	offset = hlf_ksz * HPRE_CRT_PRMS;
-	memset(ctx->rsa.crt_prikey, 0, offset);
+	memzero_explicit(ctx->rsa.crt_prikey, offset);
 	dma_free_coherent(dev, hlf_ksz * HPRE_CRT_PRMS, ctx->rsa.crt_prikey,
 			  ctx->rsa.dma_crt_prikey);
 	ctx->rsa.crt_prikey = NULL;
@@ -924,14 +915,15 @@ static void hpre_rsa_clear_ctx(struct hpre_ctx *ctx, bool is_clear_all)
 	}
 
 	if (ctx->rsa.crt_prikey) {
-		memset(ctx->rsa.crt_prikey, 0, half_key_sz * HPRE_CRT_PRMS);
+		memzero_explicit(ctx->rsa.crt_prikey,
+				 half_key_sz * HPRE_CRT_PRMS);
 		dma_free_coherent(dev, half_key_sz * HPRE_CRT_PRMS,
 				  ctx->rsa.crt_prikey, ctx->rsa.dma_crt_prikey);
 		ctx->rsa.crt_prikey = NULL;
 	}
 
 	if (ctx->rsa.prikey) {
-		memset(ctx->rsa.prikey, 0, ctx->key_sz);
+		memzero_explicit(ctx->rsa.prikey, ctx->key_sz);
 		dma_free_coherent(dev, ctx->key_sz << 1, ctx->rsa.prikey,
 				  ctx->rsa.dma_prikey);
 		ctx->rsa.prikey = NULL;
diff --git a/drivers/crypto/hisilicon/hpre/hpre_main.c b/drivers/crypto/hisilicon/hpre/hpre_main.c
index 753e43d..401747d 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_main.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_main.c
@@ -73,7 +73,7 @@
 #define HPRE_DBGFS_VAL_MAX_LEN		20
 #define HPRE_PCI_DEVICE_ID		0xa258
 #define HPRE_PCI_VF_DEVICE_ID		0xa259
-#define HPRE_ADDR(qm, offset)		(qm->io_base + (offset))
+#define HPRE_ADDR(qm, offset)		((qm)->io_base + (offset))
 #define HPRE_QM_USR_CFG_MASK		0xfffffffe
 #define HPRE_QM_AXI_CFG_MASK		0xffff
 #define HPRE_QM_VFG_AX_MASK		0xff
@@ -490,7 +490,7 @@ static ssize_t hpre_ctrl_debug_read(struct file *filp, char __user *buf,
 		return -EINVAL;
 	}
 	spin_unlock_irq(&file->lock);
-	ret = sprintf(tbuf, "%u\n", val);
+	ret = snprintf(tbuf, HPRE_DBGFS_VAL_MAX_LEN, "%u\n", val);
 	return simple_read_from_buffer(buf, count, pos, tbuf, ret);
 }
 
@@ -607,7 +607,9 @@ static int hpre_cluster_debugfs_init(struct hpre_debug *debug)
 	int i, ret;
 
 	for (i = 0; i < HPRE_CLUSTERS_NUM; i++) {
-		sprintf(buf, "cluster%d", i);
+		ret = snprintf(buf, HPRE_DBGFS_VAL_MAX_LEN, "cluster%d", i);
+		if (ret < 0)
+			return -EINVAL;
 		tmp_d = debugfs_create_dir(buf, debug->debug_root);
 
 		regset = devm_kzalloc(dev, sizeof(*regset), GFP_KERNEL);
-- 
2.8.1

[PATCH 3/4] crypto: hisilicon - adjust hpre_crt_para_get

[Zaibo Xu]

Reorder the input parameters of hpre_crt_para_get to
make it cleaner.

Signed-off-by: Zaibo Xu <xuzaibo@huawei.com>
---
 drivers/crypto/hisilicon/hpre/hpre_crypto.c | 28 ++++++++++++++--------------
 1 file changed, 14 insertions(+), 14 deletions(-)

diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
index 76540a1..954134d 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_crypto.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
@@ -827,17 +827,17 @@ static int hpre_rsa_set_d(struct hpre_ctx *ctx, const char *value,
 	return 0;
 }
 
-static int hpre_crt_para_get(char *para, const char *raw,
-			     unsigned int raw_sz, unsigned int para_size)
+static int hpre_crt_para_get(char *para, size_t para_sz,
+			     const char *raw, size_t raw_sz)
 {
 	const char *ptr = raw;
 	size_t len = raw_sz;
 
 	hpre_rsa_drop_leading_zeros(&ptr, &len);
-	if (!len || len > para_size)
+	if (!len || len > para_sz)
 		return -EINVAL;
 
-	memcpy(para + para_size - len, ptr, len);
+	memcpy(para + para_sz - len, ptr, len);
 
 	return 0;
 }
@@ -855,32 +855,32 @@ static int hpre_rsa_setkey_crt(struct hpre_ctx *ctx, struct rsa_key *rsa_key)
 	if (!ctx->rsa.crt_prikey)
 		return -ENOMEM;
 
-	ret = hpre_crt_para_get(ctx->rsa.crt_prikey, rsa_key->dq,
-				rsa_key->dq_sz, hlf_ksz);
+	ret = hpre_crt_para_get(ctx->rsa.crt_prikey, hlf_ksz,
+				rsa_key->dq, rsa_key->dq_sz);
 	if (ret)
 		goto free_key;
 
 	offset = hlf_ksz;
-	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, rsa_key->dp,
-				rsa_key->dp_sz, hlf_ksz);
+	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz,
+				rsa_key->dp, rsa_key->dp_sz);
 	if (ret)
 		goto free_key;
 
 	offset = hlf_ksz * HPRE_CRT_Q;
-	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset,
-				rsa_key->q, rsa_key->q_sz, hlf_ksz);
+	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz,
+				rsa_key->q, rsa_key->q_sz);
 	if (ret)
 		goto free_key;
 
 	offset = hlf_ksz * HPRE_CRT_P;
-	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset,
-				rsa_key->p, rsa_key->p_sz, hlf_ksz);
+	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz,
+				rsa_key->p, rsa_key->p_sz);
 	if (ret)
 		goto free_key;
 
 	offset = hlf_ksz * HPRE_CRT_INV;
-	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset,
-				rsa_key->qinv, rsa_key->qinv_sz, hlf_ksz);
+	ret = hpre_crt_para_get(ctx->rsa.crt_prikey + offset, hlf_ksz,
+				rsa_key->qinv, rsa_key->qinv_sz);
 	if (ret)
 		goto free_key;
 
-- 
2.8.1

[PATCH 4/4] crypto: hisilicon - add branch prediction macro

[Zaibo Xu]

This branch prediction macro on the hot path can improve
small performance(about 2%) according to the test.

Signed-off-by: Zaibo Xu <xuzaibo@huawei.com>
---
 drivers/crypto/hisilicon/hpre/hpre_crypto.c | 44 ++++++++++++++---------------
 1 file changed, 22 insertions(+), 22 deletions(-)

diff --git a/drivers/crypto/hisilicon/hpre/hpre_crypto.c b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
index 954134d..5d400d6 100644
--- a/drivers/crypto/hisilicon/hpre/hpre_crypto.c
+++ b/drivers/crypto/hisilicon/hpre/hpre_crypto.c
@@ -123,7 +123,7 @@ static int hpre_add_req_to_ctx(struct hpre_asym_request *hpre_req)
 
 	ctx = hpre_req->ctx;
 	id = hpre_alloc_req_id(ctx);
-	if (id < 0)
+	if (unlikely(id < 0))
 		return -EINVAL;
 
 	ctx->req_list[id] = hpre_req;
@@ -190,7 +190,7 @@ static int hpre_get_data_dma_addr(struct hpre_asym_request *hpre_req,
 	}
 	*tmp = dma_map_single(dev, sg_virt(data),
 			      len, dma_dir);
-	if (dma_mapping_error(dev, *tmp)) {
+	if (unlikely(dma_mapping_error(dev, *tmp))) {
 		dev_err(dev, "dma map data err!\n");
 		return -ENOMEM;
 	}
@@ -208,11 +208,11 @@ static int hpre_prepare_dma_buf(struct hpre_asym_request *hpre_req,
 	int shift;
 
 	shift = ctx->key_sz - len;
-	if (shift < 0)
+	if (unlikely(shift < 0))
 		return -EINVAL;
 
 	ptr = dma_alloc_coherent(dev, ctx->key_sz, tmp, GFP_KERNEL);
-	if (!ptr)
+	if (unlikely(!ptr))
 		return -ENOMEM;
 
 	if (is_src) {
@@ -241,7 +241,7 @@ static int hpre_hw_data_init(struct hpre_asym_request *hpre_req,
 	else
 		ret = hpre_prepare_dma_buf(hpre_req, data, len,
 					  is_src, &tmp);
-	if (ret)
+	if (unlikely(ret))
 		return ret;
 
 	if (is_src)
@@ -262,7 +262,7 @@ static void hpre_hw_data_clr_all(struct hpre_ctx *ctx,
 	dma_addr_t tmp;
 
 	tmp = le64_to_cpu(sqe->in);
-	if (!tmp)
+	if (unlikely(!tmp))
 		return;
 
 	if (src) {
@@ -275,7 +275,7 @@ static void hpre_hw_data_clr_all(struct hpre_ctx *ctx,
 	}
 
 	tmp = le64_to_cpu(sqe->out);
-	if (!tmp)
+	if (unlikely(!tmp))
 		return;
 
 	if (req->dst) {
@@ -309,7 +309,7 @@ static int hpre_alg_res_post_hf(struct hpre_ctx *ctx, struct hpre_sqe *sqe,
 	done = (le32_to_cpu(sqe->dw0) >> HPRE_SQE_DONE_SHIFT) &
 		HREE_SQE_DONE_MASK;
 
-	if (err == HPRE_NO_HW_ERR &&  done == HPRE_HW_TASK_DONE)
+	if (likely(err == HPRE_NO_HW_ERR && done == HPRE_HW_TASK_DONE))
 		return  0;
 
 	return -EINVAL;
@@ -456,17 +456,17 @@ static int hpre_dh_compute_value(struct kpp_request *req)
 	int ret;
 
 	ret = hpre_msg_request_set(ctx, req, false);
-	if (ret)
+	if (unlikely(ret))
 		return ret;
 
 	if (req->src) {
 		ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 1);
-		if (ret)
+		if (unlikely(ret))
 			goto clear_all;
 	}
 
 	ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 1);
-	if (ret)
+	if (unlikely(ret))
 		goto clear_all;
 
 	if (ctx->crt_g2_mode && !req->src)
@@ -478,7 +478,7 @@ static int hpre_dh_compute_value(struct kpp_request *req)
 	} while (ret == -EBUSY && ctr++ < HPRE_TRY_SEND_TIMES);
 
 	/* success */
-	if (!ret)
+	if (likely(!ret))
 		return -EINPROGRESS;
 
 clear_all:
@@ -667,22 +667,22 @@ static int hpre_rsa_enc(struct akcipher_request *req)
 		return ret;
 	}
 
-	if (!ctx->rsa.pubkey)
+	if (unlikely(!ctx->rsa.pubkey))
 		return -EINVAL;
 
 	ret = hpre_msg_request_set(ctx, req, true);
-	if (ret)
+	if (unlikely(ret))
 		return ret;
 
 	msg->dw0 |= cpu_to_le32(HPRE_ALG_NC_NCRT);
 	msg->key = cpu_to_le64((u64)ctx->rsa.dma_pubkey);
 
 	ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0);
-	if (ret)
+	if (unlikely(ret))
 		goto clear_all;
 
 	ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0);
-	if (ret)
+	if (unlikely(ret))
 		goto clear_all;
 
 	do {
@@ -690,7 +690,7 @@ static int hpre_rsa_enc(struct akcipher_request *req)
 	} while (ret == -EBUSY && ctr++ < HPRE_TRY_SEND_TIMES);
 
 	/* success */
-	if (!ret)
+	if (likely(!ret))
 		return -EINPROGRESS;
 
 clear_all:
@@ -719,11 +719,11 @@ static int hpre_rsa_dec(struct akcipher_request *req)
 		return ret;
 	}
 
-	if (!ctx->rsa.prikey)
+	if (unlikely(!ctx->rsa.prikey))
 		return -EINVAL;
 
 	ret = hpre_msg_request_set(ctx, req, true);
-	if (ret)
+	if (unlikely(ret))
 		return ret;
 
 	if (ctx->crt_g2_mode) {
@@ -737,11 +737,11 @@ static int hpre_rsa_dec(struct akcipher_request *req)
 	}
 
 	ret = hpre_hw_data_init(hpre_req, req->src, req->src_len, 1, 0);
-	if (ret)
+	if (unlikely(ret))
 		goto clear_all;
 
 	ret = hpre_hw_data_init(hpre_req, req->dst, req->dst_len, 0, 0);
-	if (ret)
+	if (unlikely(ret))
 		goto clear_all;
 
 	do {
@@ -749,7 +749,7 @@ static int hpre_rsa_dec(struct akcipher_request *req)
 	} while (ret == -EBUSY && ctr++ < HPRE_TRY_SEND_TIMES);
 
 	/* success */
-	if (!ret)
+	if (likely(!ret))
 		return -EINPROGRESS;
 
 clear_all:
-- 
2.8.1

Re: [PATCH 0/4] crypto: hisilicon - Misc fixed on HPRE

[Herbert Xu]

On Sat, Jan 11, 2020 at 04:58:14PM +0800, Zaibo Xu wrote:
> 1.Bugfixed tfm leak with some other tiny bugfixed.
> 2.Fixed some tiny bugs and update some code style.
> 3.Adjust input parameter order of hpre_crt_para_get.
> 4.Add branch prediction macro on hot path with small performance gain(~1%).
> 
> This series is based on:
> git://git.kernel.org/pub/scm/linux/kernel/git/herbert/cryptodev-2.6.git
> 
> Zaibo Xu (4):
>   crypto: hisilicon - Bugfixed tfm leak
>   crypto: hisilicon - Fixed some tiny bugs of HPRE
>   crypto: hisilicon - adjust hpre_crt_para_get
>   crypto: hisilicon - add branch prediction macro
> 
>  drivers/crypto/hisilicon/hpre/hpre_crypto.c | 141 ++++++++++++++--------------
>  drivers/crypto/hisilicon/hpre/hpre_main.c   |  32 ++++---
>  2 files changed, 86 insertions(+), 87 deletions(-)

All applied.  Thanks.
-- 
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