3 files changed, 359 insertions, 207 deletions

diff --git a/drivers/crypto/hisilicon/sec2/sec.h b/drivers/crypto/hisilicon/sec2/sec.h
index 13e2d8d7be94..3598fa17beb2 100644
--- a/drivers/crypto/hisilicon/sec2/sec.h
+++ b/drivers/crypto/hisilicon/sec2/sec.h
@@ -11,6 +11,8 @@
 
 /* Algorithm resource per hardware SEC queue */
 struct sec_alg_res {
+	u8 *pbuf;
+	dma_addr_t pbuf_dma;
 	u8 *c_ivin;
 	dma_addr_t c_ivin_dma;
 	u8 *out_mac;
@@ -23,6 +25,8 @@ struct sec_cipher_req {
 	dma_addr_t c_in_dma;
 	struct hisi_acc_hw_sgl *c_out;
 	dma_addr_t c_out_dma;
+	u8 *c_ivin;
+	dma_addr_t c_ivin_dma;
 	struct skcipher_request *sk_req;
 	u32 c_len;
 	bool encrypt;
@@ -48,6 +52,7 @@ struct sec_req {
 
 	/* Status of the SEC request */
 	bool fake_busy;
+	bool use_pbuf;
 };
 
 /**
@@ -114,6 +119,7 @@ struct sec_ctx {
 	struct sec_qp_ctx *qp_ctx;
 	struct sec_dev *sec;
 	const struct sec_req_op *req_op;
+	struct hisi_qp **qps;
 
 	/* Half queues for encipher, and half for decipher */
 	u32 hlf_q_num;
@@ -128,6 +134,7 @@ struct sec_ctx {
 	atomic_t dec_qcyclic;
 
 	enum sec_alg_type alg_type;
+	bool pbuf_supported;
 	struct sec_cipher_ctx c_ctx;
 	struct sec_auth_ctx a_ctx;
 };
@@ -162,14 +169,15 @@ struct sec_debug {
 
 struct sec_dev {
 	struct hisi_qm qm;
-	struct list_head list;
 	struct sec_debug debug;
 	u32 ctx_q_num;
+	bool iommu_used;
 	u32 num_vfs;
 	unsigned long status;
 };
 
-struct sec_dev *sec_find_device(int node);
+void sec_destroy_qps(struct hisi_qp **qps, int qp_num);
+struct hisi_qp **sec_create_qps(void);
 int sec_register_to_crypto(void);
 void sec_unregister_from_crypto(void);
 #endif
diff --git a/drivers/crypto/hisilicon/sec2/sec_crypto.c b/drivers/crypto/hisilicon/sec2/sec_crypto.c
index a2cfcc9ccd94..7f1c6a31b82f 100644
--- a/drivers/crypto/hisilicon/sec2/sec_crypto.c
+++ b/drivers/crypto/hisilicon/sec2/sec_crypto.c
@@ -46,7 +46,21 @@
 #define SEC_CIPHER_AUTH		0xfe
 #define SEC_AUTH_CIPHER		0x1
 #define SEC_MAX_MAC_LEN		64
+#define SEC_MAX_AAD_LEN		65535
 #define SEC_TOTAL_MAC_SZ	(SEC_MAX_MAC_LEN * QM_Q_DEPTH)
+
+#define SEC_PBUF_SZ			512
+#define SEC_PBUF_IV_OFFSET		SEC_PBUF_SZ
+#define SEC_PBUF_MAC_OFFSET		(SEC_PBUF_SZ + SEC_IV_SIZE)
+#define SEC_PBUF_PKG		(SEC_PBUF_SZ + SEC_IV_SIZE +	\
+			SEC_MAX_MAC_LEN * 2)
+#define SEC_PBUF_NUM		(PAGE_SIZE / SEC_PBUF_PKG)
+#define SEC_PBUF_PAGE_NUM	(QM_Q_DEPTH / SEC_PBUF_NUM)
+#define SEC_PBUF_LEFT_SZ	(SEC_PBUF_PKG * (QM_Q_DEPTH -	\
+			SEC_PBUF_PAGE_NUM * SEC_PBUF_NUM))
+#define SEC_TOTAL_PBUF_SZ	(PAGE_SIZE * SEC_PBUF_PAGE_NUM +	\
+			SEC_PBUF_LEFT_SZ)
+
 #define SEC_SQE_LEN_RATE	4
 #define SEC_SQE_CFLAG		2
 #define SEC_SQE_AEAD_FLAG	3
@@ -110,12 +124,12 @@ static void sec_free_req_id(struct sec_req *req)
 	mutex_unlock(&qp_ctx->req_lock);
 }
 
-static int sec_aead_verify(struct sec_req *req, struct sec_qp_ctx *qp_ctx)
+static int sec_aead_verify(struct sec_req *req)
 {
 	struct aead_request *aead_req = req->aead_req.aead_req;
 	struct crypto_aead *tfm = crypto_aead_reqtfm(aead_req);
-	u8 *mac_out = qp_ctx->res[req->req_id].out_mac;
 	size_t authsize = crypto_aead_authsize(tfm);
+	u8 *mac_out = req->aead_req.out_mac;
 	u8 *mac = mac_out + SEC_MAX_MAC_LEN;
 	struct scatterlist *sgl = aead_req->src;
 	size_t sz;
@@ -163,7 +177,7 @@ static void sec_req_cb(struct hisi_qp *qp, void *resp)
 	}
 
 	if (ctx->alg_type == SEC_AEAD && !req->c_req.encrypt)
-		err = sec_aead_verify(req, qp_ctx);
+		err = sec_aead_verify(req);
 
 	atomic64_inc(&ctx->sec->debug.dfx.recv_cnt);
 
@@ -245,6 +259,50 @@ static void sec_free_mac_resource(struct device *dev, struct sec_alg_res *res)
 				  res->out_mac, res->out_mac_dma);
 }
 
+static void sec_free_pbuf_resource(struct device *dev, struct sec_alg_res *res)
+{
+	if (res->pbuf)
+		dma_free_coherent(dev, SEC_TOTAL_PBUF_SZ,
+				  res->pbuf, res->pbuf_dma);
+}
+
+/*
+ * To improve performance, pbuffer is used for
+ * small packets (< 512Bytes) as IOMMU translation using.
+ */
+static int sec_alloc_pbuf_resource(struct device *dev, struct sec_alg_res *res)
+{
+	int pbuf_page_offset;
+	int i, j, k;
+
+	res->pbuf = dma_alloc_coherent(dev, SEC_TOTAL_PBUF_SZ,
+				&res->pbuf_dma, GFP_KERNEL);
+	if (!res->pbuf)
+		return -ENOMEM;
+
+	/*
+	 * SEC_PBUF_PKG contains data pbuf, iv and
+	 * out_mac : <SEC_PBUF|SEC_IV|SEC_MAC>
+	 * Every PAGE contains six SEC_PBUF_PKG
+	 * The sec_qp_ctx contains QM_Q_DEPTH numbers of SEC_PBUF_PKG
+	 * So we need SEC_PBUF_PAGE_NUM numbers of PAGE
+	 * for the SEC_TOTAL_PBUF_SZ
+	 */
+	for (i = 0; i <= SEC_PBUF_PAGE_NUM; i++) {
+		pbuf_page_offset = PAGE_SIZE * i;
+		for (j = 0; j < SEC_PBUF_NUM; j++) {
+			k = i * SEC_PBUF_NUM + j;
+			if (k == QM_Q_DEPTH)
+				break;
+			res[k].pbuf = res->pbuf +
+				j * SEC_PBUF_PKG + pbuf_page_offset;
+			res[k].pbuf_dma = res->pbuf_dma +
+				j * SEC_PBUF_PKG + pbuf_page_offset;
+		}
+	}
+	return 0;
+}
+
 static int sec_alg_resource_alloc(struct sec_ctx *ctx,
 				  struct sec_qp_ctx *qp_ctx)
 {
@@ -259,11 +317,18 @@ static int sec_alg_resource_alloc(struct sec_ctx *ctx,
 	if (ctx->alg_type == SEC_AEAD) {
 		ret = sec_alloc_mac_resource(dev, res);
 		if (ret)
-			goto get_fail;
+			goto alloc_fail;
+	}
+	if (ctx->pbuf_supported) {
+		ret = sec_alloc_pbuf_resource(dev, res);
+		if (ret) {
+			dev_err(dev, "fail to alloc pbuf dma resource!\n");
+			goto alloc_fail;
+		}
 	}
 
 	return 0;
-get_fail:
+alloc_fail:
 	sec_free_civ_resource(dev, res);
 
 	return ret;
@@ -276,6 +341,8 @@ static void sec_alg_resource_free(struct sec_ctx *ctx,
 
 	sec_free_civ_resource(dev, qp_ctx->res);
 
+	if (ctx->pbuf_supported)
+		sec_free_pbuf_resource(dev, qp_ctx->res);
 	if (ctx->alg_type == SEC_AEAD)
 		sec_free_mac_resource(dev, qp_ctx->res);
 }
@@ -288,11 +355,8 @@ static int sec_create_qp_ctx(struct hisi_qm *qm, struct sec_ctx *ctx,
 	struct hisi_qp *qp;
 	int ret = -ENOMEM;
 
-	qp = hisi_qm_create_qp(qm, alg_type);
-	if (IS_ERR(qp))
-		return PTR_ERR(qp);
-
 	qp_ctx = &ctx->qp_ctx[qp_ctx_id];
+	qp = ctx->qps[qp_ctx_id];
 	qp->req_type = 0;
 	qp->qp_ctx = qp_ctx;
 	qp->req_cb = sec_req_cb;
@@ -335,7 +399,6 @@ err_free_c_in_pool:
 	hisi_acc_free_sgl_pool(dev, qp_ctx->c_in_pool);
 err_destroy_idr:
 	idr_destroy(&qp_ctx->req_idr);
-	hisi_qm_release_qp(qp);
 
 	return ret;
 }
@@ -352,7 +415,6 @@ static void sec_release_qp_ctx(struct sec_ctx *ctx,
 	hisi_acc_free_sgl_pool(dev, qp_ctx->c_in_pool);
 
 	idr_destroy(&qp_ctx->req_idr);
-	hisi_qm_release_qp(qp_ctx->qp);
 }
 
 static int sec_ctx_base_init(struct sec_ctx *ctx)
@@ -360,14 +422,18 @@ static int sec_ctx_base_init(struct sec_ctx *ctx)
 	struct sec_dev *sec;
 	int i, ret;
 
-	sec = sec_find_device(cpu_to_node(smp_processor_id()));
-	if (!sec) {
-		pr_err("Can not find proper Hisilicon SEC device!\n");
+	ctx->qps = sec_create_qps();
+	if (!ctx->qps) {
+		pr_err("Can not create sec qps!\n");
 		return -ENODEV;
 	}
+
+	sec = container_of(ctx->qps[0]->qm, struct sec_dev, qm);
 	ctx->sec = sec;
 	ctx->hlf_q_num = sec->ctx_q_num >> 1;
 
+	ctx->pbuf_supported = ctx->sec->iommu_used;
+
 	/* Half of queue depth is taken as fake requests limit in the queue. */
 	ctx->fake_req_limit = QM_Q_DEPTH >> 1;
 	ctx->qp_ctx = kcalloc(sec->ctx_q_num, sizeof(struct sec_qp_ctx),
@@ -386,6 +452,7 @@ err_sec_release_qp_ctx:
 	for (i = i - 1; i >= 0; i--)
 		sec_release_qp_ctx(ctx, &ctx->qp_ctx[i]);
 
+	sec_destroy_qps(ctx->qps, sec->ctx_q_num);
 	kfree(ctx->qp_ctx);
 	return ret;
 }
@@ -397,6 +464,7 @@ static void sec_ctx_base_uninit(struct sec_ctx *ctx)
 	for (i = 0; i < ctx->sec->ctx_q_num; i++)
 		sec_release_qp_ctx(ctx, &ctx->qp_ctx[i]);
 
+	sec_destroy_qps(ctx->qps, ctx->sec->ctx_q_num);
 	kfree(ctx->qp_ctx);
 }
 
@@ -447,7 +515,6 @@ static int sec_skcipher_init(struct crypto_skcipher *tfm)
 	struct sec_ctx *ctx = crypto_skcipher_ctx(tfm);
 	int ret;
 
-	ctx = crypto_skcipher_ctx(tfm);
 	ctx->alg_type = SEC_SKCIPHER;
 	crypto_skcipher_set_reqsize(tfm, sizeof(struct sec_req));
 	ctx->c_ctx.ivsize = crypto_skcipher_ivsize(tfm);
@@ -591,11 +658,94 @@ GEN_SEC_SETKEY_FUNC(3des_cbc, SEC_CALG_3DES, SEC_CMODE_CBC)
 GEN_SEC_SETKEY_FUNC(sm4_xts, SEC_CALG_SM4, SEC_CMODE_XTS)
 GEN_SEC_SETKEY_FUNC(sm4_cbc, SEC_CALG_SM4, SEC_CMODE_CBC)
 
-static int sec_cipher_map(struct device *dev, struct sec_req *req,
+static int sec_cipher_pbuf_map(struct sec_ctx *ctx, struct sec_req *req,
+			struct scatterlist *src)
+{
+	struct aead_request *aead_req = req->aead_req.aead_req;
+	struct sec_cipher_req *c_req = &req->c_req;
+	struct sec_qp_ctx *qp_ctx = req->qp_ctx;
+	struct device *dev = SEC_CTX_DEV(ctx);
+	int copy_size, pbuf_length;
+	int req_id = req->req_id;
+
+	if (ctx->alg_type == SEC_AEAD)
+		copy_size = aead_req->cryptlen + aead_req->assoclen;
+	else
+		copy_size = c_req->c_len;
+
+	pbuf_length = sg_copy_to_buffer(src, sg_nents(src),
+				qp_ctx->res[req_id].pbuf,
+				copy_size);
+
+	if (unlikely(pbuf_length != copy_size)) {
+		dev_err(dev, "copy src data to pbuf error!\n");
+		return -EINVAL;
+	}
+
+	c_req->c_in_dma = qp_ctx->res[req_id].pbuf_dma;
+
+	if (!c_req->c_in_dma) {
+		dev_err(dev, "fail to set pbuffer address!\n");
+		return -ENOMEM;
+	}
+
+	c_req->c_out_dma = c_req->c_in_dma;
+
+	return 0;
+}
+
+static void sec_cipher_pbuf_unmap(struct sec_ctx *ctx, struct sec_req *req,
+			struct scatterlist *dst)
+{
+	struct aead_request *aead_req = req->aead_req.aead_req;
+	struct sec_cipher_req *c_req = &req->c_req;
+	struct sec_qp_ctx *qp_ctx = req->qp_ctx;
+	struct device *dev = SEC_CTX_DEV(ctx);
+	int copy_size, pbuf_length;
+	int req_id = req->req_id;
+
+	if (ctx->alg_type == SEC_AEAD)
+		copy_size = c_req->c_len + aead_req->assoclen;
+	else
+		copy_size = c_req->c_len;
+
+	pbuf_length = sg_copy_from_buffer(dst, sg_nents(dst),
+				qp_ctx->res[req_id].pbuf,
+				copy_size);
+
+	if (unlikely(pbuf_length != copy_size))
+		dev_err(dev, "copy pbuf data to dst error!\n");
+
+}
+
+static int sec_cipher_map(struct sec_ctx *ctx, struct sec_req *req,
 			  struct scatterlist *src, struct scatterlist *dst)
 {
 	struct sec_cipher_req *c_req = &req->c_req;
+	struct sec_aead_req *a_req = &req->aead_req;
 	struct sec_qp_ctx *qp_ctx = req->qp_ctx;
+	struct sec_alg_res *res = &qp_ctx->res[req->req_id];
+	struct device *dev = SEC_CTX_DEV(ctx);
+	int ret;
+
+	if (req->use_pbuf) {
+		ret = sec_cipher_pbuf_map(ctx, req, src);
+		c_req->c_ivin = res->pbuf + SEC_PBUF_IV_OFFSET;
+		c_req->c_ivin_dma = res->pbuf_dma + SEC_PBUF_IV_OFFSET;
+		if (ctx->alg_type == SEC_AEAD) {
+			a_req->out_mac = res->pbuf + SEC_PBUF_MAC_OFFSET;
+			a_req->out_mac_dma = res->pbuf_dma +
+					SEC_PBUF_MAC_OFFSET;
+		}
+
+		return ret;
+	}
+	c_req->c_ivin = res->c_ivin;
+	c_req->c_ivin_dma = res->c_ivin_dma;
+	if (ctx->alg_type == SEC_AEAD) {
+		a_req->out_mac = res->out_mac;
+		a_req->out_mac_dma = res->out_mac_dma;
+	}
 
 	c_req->c_in = hisi_acc_sg_buf_map_to_hw_sgl(dev, src,
 						    qp_ctx->c_in_pool,
@@ -626,29 +776,34 @@ static int sec_cipher_map(struct device *dev, struct sec_req *req,
 	return 0;
 }
 
-static void sec_cipher_unmap(struct device *dev, struct sec_cipher_req *req,
+static void sec_cipher_unmap(struct sec_ctx *ctx, struct sec_req *req,
 			     struct scatterlist *src, struct scatterlist *dst)
 {
-	if (dst != src)
-		hisi_acc_sg_buf_unmap(dev, src, req->c_in);
+	struct sec_cipher_req *c_req = &req->c_req;
+	struct device *dev = SEC_CTX_DEV(ctx);
+
+	if (req->use_pbuf) {
+		sec_cipher_pbuf_unmap(ctx, req, dst);
+	} else {
+		if (dst != src)
+			hisi_acc_sg_buf_unmap(dev, src, c_req->c_in);
 
-	hisi_acc_sg_buf_unmap(dev, dst, req->c_out);
+		hisi_acc_sg_buf_unmap(dev, dst, c_req->c_out);
+	}
 }
 
 static int sec_skcipher_sgl_map(struct sec_ctx *ctx, struct sec_req *req)
 {
 	struct skcipher_request *sq = req->c_req.sk_req;
 
-	return sec_cipher_map(SEC_CTX_DEV(ctx), req, sq->src, sq->dst);
+	return sec_cipher_map(ctx, req, sq->src, sq->dst);
 }
 
 static void sec_skcipher_sgl_unmap(struct sec_ctx *ctx, struct sec_req *req)
 {
-	struct device *dev = SEC_CTX_DEV(ctx);
-	struct sec_cipher_req *c_req = &req->c_req;
-	struct skcipher_request *sk_req = c_req->sk_req;
+	struct skcipher_request *sq = req->c_req.sk_req;
 
-	sec_cipher_unmap(dev, c_req, sk_req->src, sk_req->dst);
+	sec_cipher_unmap(ctx, req, sq->src, sq->dst);
 }
 
 static int sec_aead_aes_set_key(struct sec_cipher_ctx *c_ctx,
@@ -759,16 +914,14 @@ static int sec_aead_sgl_map(struct sec_ctx *ctx, struct sec_req *req)
 {
 	struct aead_request *aq = req->aead_req.aead_req;
 
-	return sec_cipher_map(SEC_CTX_DEV(ctx), req, aq->src, aq->dst);
+	return sec_cipher_map(ctx, req, aq->src, aq->dst);
 }
 
 static void sec_aead_sgl_unmap(struct sec_ctx *ctx, struct sec_req *req)
 {
-	struct device *dev = SEC_CTX_DEV(ctx);
-	struct sec_cipher_req *cq = &req->c_req;
 	struct aead_request *aq = req->aead_req.aead_req;
 
-	sec_cipher_unmap(dev, cq, aq->src, aq->dst);
+	sec_cipher_unmap(ctx, req, aq->src, aq->dst);
 }
 
 static int sec_request_transfer(struct sec_ctx *ctx, struct sec_req *req)
@@ -801,9 +954,9 @@ static void sec_request_untransfer(struct sec_ctx *ctx, struct sec_req *req)
 static void sec_skcipher_copy_iv(struct sec_ctx *ctx, struct sec_req *req)
 {
 	struct skcipher_request *sk_req = req->c_req.sk_req;
-	u8 *c_ivin = req->qp_ctx->res[req->req_id].c_ivin;
+	struct sec_cipher_req *c_req = &req->c_req;
 
-	memcpy(c_ivin, sk_req->iv, ctx->c_ctx.ivsize);
+	memcpy(c_req->c_ivin, sk_req->iv, ctx->c_ctx.ivsize);
 }
 
 static int sec_skcipher_bd_fill(struct sec_ctx *ctx, struct sec_req *req)
@@ -818,8 +971,7 @@ static int sec_skcipher_bd_fill(struct sec_ctx *ctx, struct sec_req *req)
 	memset(sec_sqe, 0, sizeof(struct sec_sqe));
 
 	sec_sqe->type2.c_key_addr = cpu_to_le64(c_ctx->c_key_dma);
-	sec_sqe->type2.c_ivin_addr =
-		cpu_to_le64(req->qp_ctx->res[req->req_id].c_ivin_dma);
+	sec_sqe->type2.c_ivin_addr = cpu_to_le64(c_req->c_ivin_dma);
 	sec_sqe->type2.data_src_addr = cpu_to_le64(c_req->c_in_dma);
 	sec_sqe->type2.data_dst_addr = cpu_to_le64(c_req->c_out_dma);
 
@@ -836,7 +988,10 @@ static int sec_skcipher_bd_fill(struct sec_ctx *ctx, struct sec_req *req)
 		cipher = SEC_CIPHER_DEC << SEC_CIPHER_OFFSET;
 	sec_sqe->type_cipher_auth = bd_type | cipher;
 
-	sa_type = SEC_SGL << SEC_SRC_SGL_OFFSET;
+	if (req->use_pbuf)
+		sa_type = SEC_PBUF << SEC_SRC_SGL_OFFSET;
+	else
+		sa_type = SEC_SGL << SEC_SRC_SGL_OFFSET;
 	scene = SEC_COMM_SCENE << SEC_SCENE_OFFSET;
 	if (c_req->c_in_dma != c_req->c_out_dma)
 		de = 0x1 << SEC_DE_OFFSET;
@@ -844,7 +999,10 @@ static int sec_skcipher_bd_fill(struct sec_ctx *ctx, struct sec_req *req)
 	sec_sqe->sds_sa_type = (de | scene | sa_type);
 
 	/* Just set DST address type */
-	da_type = SEC_SGL << SEC_DST_SGL_OFFSET;
+	if (req->use_pbuf)
+		da_type = SEC_PBUF << SEC_DST_SGL_OFFSET;
+	else
+		da_type = SEC_SGL << SEC_DST_SGL_OFFSET;
 	sec_sqe->sdm_addr_type |= da_type;
 
 	sec_sqe->type2.clen_ivhlen |= cpu_to_le32(c_req->c_len);
@@ -904,9 +1062,9 @@ static void sec_skcipher_callback(struct sec_ctx *ctx, struct sec_req *req,
 static void sec_aead_copy_iv(struct sec_ctx *ctx, struct sec_req *req)
 {
 	struct aead_request *aead_req = req->aead_req.aead_req;
-	u8 *c_ivin = req->qp_ctx->res[req->req_id].c_ivin;
+	struct sec_cipher_req *c_req = &req->c_req;
 
-	memcpy(c_ivin, aead_req->iv, ctx->c_ctx.ivsize);
+	memcpy(c_req->c_ivin, aead_req->iv, ctx->c_ctx.ivsize);
 }
 
 static void sec_auth_bd_fill_ex(struct sec_auth_ctx *ctx, int dir,
@@ -939,8 +1097,7 @@ static void sec_auth_bd_fill_ex(struct sec_auth_ctx *ctx, int dir,
 
 	sec_sqe->type2.cipher_src_offset = cpu_to_le16((u16)aq->assoclen);
 
-	sec_sqe->type2.mac_addr =
-		cpu_to_le64(req->qp_ctx->res[req->req_id].out_mac_dma);
+	sec_sqe->type2.mac_addr = cpu_to_le64(a_req->out_mac_dma);
 }
 
 static int sec_aead_bd_fill(struct sec_ctx *ctx, struct sec_req *req)
@@ -964,6 +1121,7 @@ static void sec_aead_callback(struct sec_ctx *c, struct sec_req *req, int err)
 {
 	struct aead_request *a_req = req->aead_req.aead_req;
 	struct crypto_aead *tfm = crypto_aead_reqtfm(a_req);
+	struct sec_aead_req *aead_req = &req->aead_req;
 	struct sec_cipher_req *c_req = &req->c_req;
 	size_t authsize = crypto_aead_authsize(tfm);
 	struct sec_qp_ctx *qp_ctx = req->qp_ctx;
@@ -979,7 +1137,7 @@ static void sec_aead_callback(struct sec_ctx *c, struct sec_req *req, int err)
 		struct scatterlist *sgl = a_req->dst;
 
 		sz = sg_pcopy_from_buffer(sgl, sg_nents(sgl),
-					  qp_ctx->res[req->req_id].out_mac,
+					  aead_req->out_mac,
 					  authsize, a_req->cryptlen +
 					  a_req->assoclen);
 
@@ -1031,6 +1189,7 @@ static int sec_request_init(struct sec_ctx *ctx, struct sec_req *req)
 
 static int sec_process(struct sec_ctx *ctx, struct sec_req *req)
 {
+	struct sec_cipher_req *c_req = &req->c_req;
 	int ret;
 
 	ret = sec_request_init(ctx, req);
@@ -1057,12 +1216,10 @@ err_send_req:
 	/* As failing, restore the IV from user */
 	if (ctx->c_ctx.c_mode == SEC_CMODE_CBC && !req->c_req.encrypt) {
 		if (ctx->alg_type == SEC_SKCIPHER)
-			memcpy(req->c_req.sk_req->iv,
-			       req->qp_ctx->res[req->req_id].c_ivin,
+			memcpy(req->c_req.sk_req->iv, c_req->c_ivin,
 			       ctx->c_ctx.ivsize);
 		else
-			memcpy(req->aead_req.aead_req->iv,
-			       req->qp_ctx->res[req->req_id].c_ivin,
+			memcpy(req->aead_req.aead_req->iv, c_req->c_ivin,
 			       ctx->c_ctx.ivsize);
 	}
 
@@ -1208,6 +1365,12 @@ static int sec_skcipher_param_check(struct sec_ctx *ctx, struct sec_req *sreq)
 		return -EINVAL;
 	}
 	sreq->c_req.c_len = sk_req->cryptlen;
+
+	if (ctx->pbuf_supported && sk_req->cryptlen <= SEC_PBUF_SZ)
+		sreq->use_pbuf = true;
+	else
+		sreq->use_pbuf = false;
+
 	if (c_alg == SEC_CALG_3DES) {
 		if (unlikely(sk_req->cryptlen & (DES3_EDE_BLOCK_SIZE - 1))) {
 			dev_err(dev, "skcipher 3des input length error!\n");
@@ -1321,11 +1484,18 @@ static int sec_aead_param_check(struct sec_ctx *ctx, struct sec_req *sreq)
 	struct crypto_aead *tfm = crypto_aead_reqtfm(req);
 	size_t authsize = crypto_aead_authsize(tfm);
 
-	if (unlikely(!req->src || !req->dst || !req->cryptlen)) {
+	if (unlikely(!req->src || !req->dst || !req->cryptlen ||
+		req->assoclen > SEC_MAX_AAD_LEN)) {
 		dev_err(SEC_CTX_DEV(ctx), "aead input param error!\n");
 		return -EINVAL;
 	}
 
+	if (ctx->pbuf_supported && (req->cryptlen + req->assoclen) <=
+		SEC_PBUF_SZ)
+		sreq->use_pbuf = true;
+	else
+		sreq->use_pbuf = false;
+
 	/* Support AES only */
 	if (unlikely(c_alg != SEC_CALG_AES)) {
 		dev_err(SEC_CTX_DEV(ctx), "aead crypto alg error!\n");
diff --git a/drivers/crypto/hisilicon/sec2/sec_main.c b/drivers/crypto/hisilicon/sec2/sec_main.c
index 2bbaf1e2dae7..1f54ebe164b6 100644
--- a/drivers/crypto/hisilicon/sec2/sec_main.c
+++ b/drivers/crypto/hisilicon/sec2/sec_main.c
@@ -7,6 +7,7 @@
 #include <linux/debugfs.h>
 #include <linux/init.h>
 #include <linux/io.h>
+#include <linux/iommu.h>
 #include <linux/kernel.h>
 #include <linux/module.h>
 #include <linux/pci.h>
@@ -89,8 +90,7 @@ struct sec_hw_error {
 
 static const char sec_name[] = "hisi_sec2";
 static struct dentry *sec_debugfs_root;
-static LIST_HEAD(sec_list);
-static DEFINE_MUTEX(sec_list_lock);
+static struct hisi_qm_list sec_devices;
 
 static const struct sec_hw_error sec_hw_errors[] = {
 	{.int_msk = BIT(0), .msg = "sec_axi_rresp_err_rint"},
@@ -105,37 +105,6 @@ static const struct sec_hw_error sec_hw_errors[] = {
 	{ /* sentinel */ }
 };
 
-struct sec_dev *sec_find_device(int node)
-{
-#define SEC_NUMA_MAX_DISTANCE	100
-	int min_distance = SEC_NUMA_MAX_DISTANCE;
-	int dev_node = 0, free_qp_num = 0;
-	struct sec_dev *sec, *ret = NULL;
-	struct hisi_qm *qm;
-	struct device *dev;
-
-	mutex_lock(&sec_list_lock);
-	list_for_each_entry(sec, &sec_list, list) {
-		qm = &sec->qm;
-		dev = &qm->pdev->dev;
-#ifdef CONFIG_NUMA
-		dev_node = dev->numa_node;
-		if (dev_node < 0)
-			dev_node = 0;
-#endif
-		if (node_distance(dev_node, node) < min_distance) {
-			free_qp_num = hisi_qm_get_free_qp_num(qm);
-			if (free_qp_num >= sec->ctx_q_num) {
-				ret = sec;
-				min_distance = node_distance(dev_node, node);
-			}
-		}
-	}
-	mutex_unlock(&sec_list_lock);
-
-	return ret;
-}
-
 static const char * const sec_dbg_file_name[] = {
 	[SEC_CURRENT_QM] = "current_qm",
 	[SEC_CLEAR_ENABLE] = "clear_enable",
@@ -238,6 +207,32 @@ static u32 ctx_q_num = SEC_CTX_Q_NUM_DEF;
 module_param_cb(ctx_q_num, &sec_ctx_q_num_ops, &ctx_q_num, 0444);
 MODULE_PARM_DESC(ctx_q_num, "Queue num in ctx (24 default, 2, 4, ..., 32)");
 
+void sec_destroy_qps(struct hisi_qp **qps, int qp_num)
+{
+	hisi_qm_free_qps(qps, qp_num);
+	kfree(qps);
+}
+
+struct hisi_qp **sec_create_qps(void)
+{
+	int node = cpu_to_node(smp_processor_id());
+	u32 ctx_num = ctx_q_num;
+	struct hisi_qp **qps;
+	int ret;
+
+	qps = kcalloc(ctx_num, sizeof(struct hisi_qp *), GFP_KERNEL);
+	if (!qps)
+		return NULL;
+
+	ret = hisi_qm_alloc_qps_node(&sec_devices, ctx_num, 0, node, qps);
+	if (!ret)
+		return qps;
+
+	kfree(qps);
+	return NULL;
+}
+
+
 static const struct pci_device_id sec_dev_ids[] = {
 	{ PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, SEC_PF_PCI_DEVICE_ID) },
 	{ PCI_DEVICE(PCI_VENDOR_ID_HUAWEI, SEC_VF_PCI_DEVICE_ID) },
@@ -245,20 +240,6 @@ static const struct pci_device_id sec_dev_ids[] = {
 };
 MODULE_DEVICE_TABLE(pci, sec_dev_ids);
 
-static inline void sec_add_to_list(struct sec_dev *sec)
-{
-	mutex_lock(&sec_list_lock);
-	list_add_tail(&sec->list, &sec_list);
-	mutex_unlock(&sec_list_lock);
-}
-
-static inline void sec_remove_from_list(struct sec_dev *sec)
-{
-	mutex_lock(&sec_list_lock);
-	list_del(&sec->list);
-	mutex_unlock(&sec_list_lock);
-}
-
 static u8 sec_get_endian(struct sec_dev *sec)
 {
 	struct hisi_qm *qm = &sec->qm;
@@ -384,9 +365,8 @@ static void sec_debug_regs_clear(struct hisi_qm *qm)
 	hisi_qm_debug_regs_clear(qm);
 }
 
-static void sec_hw_error_enable(struct sec_dev *sec)
+static void sec_hw_error_enable(struct hisi_qm *qm)
 {
-	struct hisi_qm *qm = &sec->qm;
 	u32 val;
 
 	if (qm->ver == QM_HW_V1) {
@@ -414,9 +394,8 @@ static void sec_hw_error_enable(struct sec_dev *sec)
 	writel(val, qm->io_base + SEC_CONTROL_REG);
 }
 
-static void sec_hw_error_disable(struct sec_dev *sec)
+static void sec_hw_error_disable(struct hisi_qm *qm)
 {
-	struct hisi_qm *qm = &sec->qm;
 	u32 val;
 
 	val = readl(qm->io_base + SEC_CONTROL_REG);
@@ -435,27 +414,6 @@ static void sec_hw_error_disable(struct sec_dev *sec)
 	writel(val, qm->io_base + SEC_CONTROL_REG);
 }
 
-static void sec_hw_error_init(struct sec_dev *sec)
-{
-	if (sec->qm.fun_type == QM_HW_VF)
-		return;
-
-	hisi_qm_hw_error_init(&sec->qm, QM_BASE_CE,
-			      QM_BASE_NFE | QM_ACC_DO_TASK_TIMEOUT
-			      | QM_ACC_WB_NOT_READY_TIMEOUT, 0,
-			      QM_DB_RANDOM_INVALID);
-	sec_hw_error_enable(sec);
-}
-
-static void sec_hw_error_uninit(struct sec_dev *sec)
-{
-	if (sec->qm.fun_type == QM_HW_VF)
-		return;
-
-	sec_hw_error_disable(sec);
-	writel(GENMASK(12, 0), sec->qm.io_base + SEC_QM_ABNORMAL_INT_MASK);
-}
-
 static u32 sec_current_qm_read(struct sec_debug_file *file)
 {
 	struct hisi_qm *qm = file->qm;
@@ -695,6 +653,51 @@ static void sec_debugfs_exit(struct sec_dev *sec)
 	debugfs_remove_recursive(sec->qm.debug.debug_root);
 }
 
+static void sec_log_hw_error(struct hisi_qm *qm, u32 err_sts)
+{
+	const struct sec_hw_error *errs = sec_hw_errors;
+	struct device *dev = &qm->pdev->dev;
+	u32 err_val;
+
+	while (errs->msg) {
+		if (errs->int_msk & err_sts) {
+			dev_err(dev, "%s [error status=0x%x] found\n",
+				errs->msg, errs->int_msk);
+
+			if (SEC_CORE_INT_STATUS_M_ECC & errs->int_msk) {
+				err_val = readl(qm->io_base +
+						SEC_CORE_SRAM_ECC_ERR_INFO);
+				dev_err(dev, "multi ecc sram num=0x%x\n",
+					SEC_ECC_NUM(err_val));
+				dev_err(dev, "multi ecc sram addr=0x%x\n",
+					SEC_ECC_ADDR(err_val));
+			}
+		}
+		errs++;
+	}
+
+	writel(err_sts, qm->io_base + SEC_CORE_INT_SOURCE);
+}
+
+static u32 sec_get_hw_err_status(struct hisi_qm *qm)
+{
+	return readl(qm->io_base + SEC_CORE_INT_STATUS);
+}
+
+static const struct hisi_qm_err_ini sec_err_ini = {
+	.hw_err_enable		= sec_hw_error_enable,
+	.hw_err_disable		= sec_hw_error_disable,
+	.get_dev_hw_err_status	= sec_get_hw_err_status,
+	.log_dev_hw_err		= sec_log_hw_error,
+	.err_info		= {
+		.ce			= QM_BASE_CE,
+		.nfe			= QM_BASE_NFE | QM_ACC_DO_TASK_TIMEOUT |
+					  QM_ACC_WB_NOT_READY_TIMEOUT,
+		.fe			= 0,
+		.msi			= QM_DB_RANDOM_INVALID,
+	}
+};
+
 static int sec_pf_probe_init(struct sec_dev *sec)
 {
 	struct hisi_qm *qm = &sec->qm;
@@ -713,11 +716,13 @@ static int sec_pf_probe_init(struct sec_dev *sec)
 		return -EINVAL;
 	}
 
+	qm->err_ini = &sec_err_ini;
+
 	ret = sec_set_user_domain_and_cache(sec);
 	if (ret)
 		return ret;
 
-	sec_hw_error_init(sec);
+	hisi_qm_dev_err_init(qm);
 	sec_debug_regs_clear(qm);
 
 	return 0;
@@ -750,12 +755,30 @@ static void sec_qm_uninit(struct hisi_qm *qm)
 
 static int sec_probe_init(struct hisi_qm *qm, struct sec_dev *sec)
 {
+	int ret;
+
+	/*
+	 * WQ_HIGHPRI: SEC request must be low delayed,
+	 * so need a high priority workqueue.
+	 * WQ_UNBOUND: SEC task is likely with long
+	 * running CPU intensive workloads.
+	 */
+	qm->wq = alloc_workqueue("%s", WQ_HIGHPRI |
+		WQ_MEM_RECLAIM | WQ_UNBOUND, num_online_cpus(),
+		pci_name(qm->pdev));
+	if (!qm->wq) {
+		pci_err(qm->pdev, "fail to alloc workqueue\n");
+		return -ENOMEM;
+	}
+
 	if (qm->fun_type == QM_HW_PF) {
 		qm->qp_base = SEC_PF_DEF_Q_BASE;
 		qm->qp_num = pf_q_num;
 		qm->debug.curr_qm_qp_num = pf_q_num;
 
-		return sec_pf_probe_init(sec);
+		ret = sec_pf_probe_init(sec);
+		if (ret)
+			goto err_probe_uninit;
 	} else if (qm->fun_type == QM_HW_VF) {
 		/*
 		 * have no way to get qm configure in VM in v1 hardware,
@@ -768,18 +791,43 @@ static int sec_probe_init(struct hisi_qm *qm, struct sec_dev *sec)
 			qm->qp_num = SEC_QUEUE_NUM_V1 - SEC_PF_DEF_Q_NUM;
 		} else if (qm->ver == QM_HW_V2) {
 			/* v2 starts to support get vft by mailbox */
-			return hisi_qm_get_vft(qm, &qm->qp_base, &qm->qp_num);
+			ret = hisi_qm_get_vft(qm, &qm->qp_base, &qm->qp_num);
+			if (ret)
+				goto err_probe_uninit;
 		}
 	} else {
-		return -ENODEV;
+		ret = -ENODEV;
+		goto err_probe_uninit;
 	}
 
 	return 0;
+err_probe_uninit:
+	destroy_workqueue(qm->wq);
+	return ret;
 }
 
-static void sec_probe_uninit(struct sec_dev *sec)
+static void sec_probe_uninit(struct hisi_qm *qm)
 {
-	sec_hw_error_uninit(sec);
+	hisi_qm_dev_err_uninit(qm);
+
+	destroy_workqueue(qm->wq);
+}
+
+static void sec_iommu_used_check(struct sec_dev *sec)
+{
+	struct iommu_domain *domain;
+	struct device *dev = &sec->qm.pdev->dev;
+
+	domain = iommu_get_domain_for_dev(dev);
+
+	/* Check if iommu is used */
+	sec->iommu_used = false;
+	if (domain) {
+		if (domain->type & __IOMMU_DOMAIN_PAGING)
+			sec->iommu_used = true;
+		dev_info(dev, "SMMU Opened, the iommu type = %u\n",
+			domain->type);
+	}
 }
 
 static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id)
@@ -795,6 +843,7 @@ static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	pci_set_drvdata(pdev, sec);
 
 	sec->ctx_q_num = ctx_q_num;
+	sec_iommu_used_check(sec);
 
 	qm = &sec->qm;
 
@@ -820,7 +869,7 @@ static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	if (ret)
 		pci_warn(pdev, "Failed to init debugfs!\n");
 
-	sec_add_to_list(sec);
+	hisi_qm_add_to_list(qm, &sec_devices);
 
 	ret = sec_register_to_crypto();
 	if (ret < 0) {
@@ -831,12 +880,12 @@ static int sec_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 	return 0;
 
 err_remove_from_list:
-	sec_remove_from_list(sec);
+	hisi_qm_del_from_list(qm, &sec_devices);
 	sec_debugfs_exit(sec);
 	hisi_qm_stop(qm);
 
 err_probe_uninit:
-	sec_probe_uninit(sec);
+	sec_probe_uninit(qm);
 
 err_qm_uninit:
 	sec_qm_uninit(qm);
@@ -955,7 +1004,7 @@ static void sec_remove(struct pci_dev *pdev)
 
 	sec_unregister_from_crypto();
 
-	sec_remove_from_list(sec);
+	hisi_qm_del_from_list(qm, &sec_devices);
 
 	if (qm->fun_type == QM_HW_PF && sec->num_vfs)
 		(void)sec_sriov_disable(pdev);
@@ -967,89 +1016,13 @@ static void sec_remove(struct pci_dev *pdev)
 	if (qm->fun_type == QM_HW_PF)
 		sec_debug_regs_clear(qm);
 
-	sec_probe_uninit(sec);
+	sec_probe_uninit(qm);
 
 	sec_qm_uninit(qm);
 }
 
-static void sec_log_hw_error(struct sec_dev *sec, u32 err_sts)
-{
-	const struct sec_hw_error *errs = sec_hw_errors;
-	struct device *dev = &sec->qm.pdev->dev;
-	u32 err_val;
-
-	while (errs->msg) {
-		if (errs->int_msk & err_sts) {
-			dev_err(dev, "%s [error status=0x%x] found\n",
-				errs->msg, errs->int_msk);
-
-			if (SEC_CORE_INT_STATUS_M_ECC & err_sts) {
-				err_val = readl(sec->qm.io_base +
-						SEC_CORE_SRAM_ECC_ERR_INFO);
-				dev_err(dev, "multi ecc sram num=0x%x\n",
-					SEC_ECC_NUM(err_val));
-				dev_err(dev, "multi ecc sram addr=0x%x\n",
-					SEC_ECC_ADDR(err_val));
-			}
-		}
-		errs++;
-	}
-}
-
-static pci_ers_result_t sec_hw_error_handle(struct sec_dev *sec)
-{
-	u32 err_sts;
-
-	/* read err sts */
-	err_sts = readl(sec->qm.io_base + SEC_CORE_INT_STATUS);
-	if (err_sts) {
-		sec_log_hw_error(sec, err_sts);
-
-		/* clear error interrupts */
-		writel(err_sts, sec->qm.io_base + SEC_CORE_INT_SOURCE);
-
-		return PCI_ERS_RESULT_NEED_RESET;
-	}
-
-	return PCI_ERS_RESULT_RECOVERED;
-}
-
-static pci_ers_result_t sec_process_hw_error(struct pci_dev *pdev)
-{
-	struct sec_dev *sec = pci_get_drvdata(pdev);
-	pci_ers_result_t qm_ret, sec_ret;
-
-	if (!sec) {
-		pci_err(pdev, "Can't recover error during device init\n");
-		return PCI_ERS_RESULT_NONE;
-	}
-
-	/* log qm error */
-	qm_ret = hisi_qm_hw_error_handle(&sec->qm);
-
-	/* log sec error */
-	sec_ret = sec_hw_error_handle(sec);
-
-	return (qm_ret == PCI_ERS_RESULT_NEED_RESET ||
-		sec_ret == PCI_ERS_RESULT_NEED_RESET) ?
-		PCI_ERS_RESULT_NEED_RESET : PCI_ERS_RESULT_RECOVERED;
-}
-
-static pci_ers_result_t sec_error_detected(struct pci_dev *pdev,
-					   pci_channel_state_t state)
-{
-	if (pdev->is_virtfn)
-		return PCI_ERS_RESULT_NONE;
-
-	pci_info(pdev, "PCI error detected, state(=%d)!!\n", state);
-	if (state == pci_channel_io_perm_failure)
-		return PCI_ERS_RESULT_DISCONNECT;
-
-	return sec_process_hw_error(pdev);
-}
-
 static const struct pci_error_handlers sec_err_handler = {
-	.error_detected = sec_error_detected,
+	.error_detected = hisi_qm_dev_err_detected,
 };
 
 static struct pci_driver sec_pci_driver = {
@@ -1078,6 +1051,7 @@ static int __init sec_init(void)
 {
 	int ret;
 
+	hisi_qm_init_list(&sec_devices);
 	sec_register_debugfs();
 
 	ret = pci_register_driver(&sec_pci_driver);