11 files changed, 328 insertions, 144 deletions

diff --git a/drivers/infiniband/hw/hns/hns_roce_ah.c b/drivers/infiniband/hw/hns/hns_roce_ah.c
index b4209b6aed..3e02c474f5 100644
--- a/drivers/infiniband/hw/hns/hns_roce_ah.c
+++ b/drivers/infiniband/hw/hns/hns_roce_ah.c
@@ -59,8 +59,10 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibah->device);
 	struct hns_roce_ib_create_ah_resp resp = {};
 	struct hns_roce_ah *ah = to_hr_ah(ibah);
-	int ret = 0;
-	u32 max_sl;
+	u8 tclass = get_tclass(grh);
+	u8 priority = 0;
+	u8 tc_mode = 0;
+	int ret;
 
 	if (hr_dev->pci_dev->revision == PCI_REVISION_ID_HIP08 && udata)
 		return -EOPNOTSUPP;
@@ -74,16 +76,23 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 	ah->av.hop_limit = grh->hop_limit;
 	ah->av.flowlabel = grh->flow_label;
 	ah->av.udp_sport = get_ah_udp_sport(ah_attr);
-	ah->av.tclass = get_tclass(grh);
-
-	ah->av.sl = rdma_ah_get_sl(ah_attr);
-	max_sl = min_t(u32, MAX_SERVICE_LEVEL, hr_dev->caps.sl_num - 1);
-	if (unlikely(ah->av.sl > max_sl)) {
-		ibdev_err_ratelimited(&hr_dev->ib_dev,
-				      "failed to set sl, sl (%u) shouldn't be larger than %u.\n",
-				      ah->av.sl, max_sl);
+	ah->av.tclass = tclass;
+
+	ret = hr_dev->hw->get_dscp(hr_dev, tclass, &tc_mode, &priority);
+	if (ret == -EOPNOTSUPP)
+		ret = 0;
+
+	if (ret && grh->sgid_attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+		return ret;
+
+	if (tc_mode == HNAE3_TC_MAP_MODE_DSCP &&
+	    grh->sgid_attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+		ah->av.sl = priority;
+	else
+		ah->av.sl = rdma_ah_get_sl(ah_attr);
+
+	if (!check_sl_valid(hr_dev, ah->av.sl))
 		return -EINVAL;
-	}
 
 	memcpy(ah->av.dgid, grh->dgid.raw, HNS_ROCE_GID_SIZE);
 	memcpy(ah->av.mac, ah_attr->roce.dmac, ETH_ALEN);
@@ -99,6 +108,8 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 	}
 
 	if (udata) {
+		resp.priority = ah->av.sl;
+		resp.tc_mode = tc_mode;
 		memcpy(resp.dmac, ah_attr->roce.dmac, ETH_ALEN);
 		ret = ib_copy_to_udata(udata, &resp,
 				       min(udata->outlen, sizeof(resp)));
diff --git a/drivers/infiniband/hw/hns/hns_roce_alloc.c b/drivers/infiniband/hw/hns/hns_roce_alloc.c
index 11a78ceae5..950c133d42 100644
--- a/drivers/infiniband/hw/hns/hns_roce_alloc.c
+++ b/drivers/infiniband/hw/hns/hns_roce_alloc.c
@@ -153,8 +153,7 @@ int hns_roce_get_kmem_bufs(struct hns_roce_dev *hr_dev, dma_addr_t *bufs,
 	return total;
 }
 
-int hns_roce_get_umem_bufs(struct hns_roce_dev *hr_dev, dma_addr_t *bufs,
-			   int buf_cnt, struct ib_umem *umem,
+int hns_roce_get_umem_bufs(dma_addr_t *bufs, int buf_cnt, struct ib_umem *umem,
 			   unsigned int page_shift)
 {
 	struct ib_block_iter biter;
diff --git a/drivers/infiniband/hw/hns/hns_roce_cq.c b/drivers/infiniband/hw/hns/hns_roce_cq.c
index 68e22f368d..56dc3908da 100644
--- a/drivers/infiniband/hw/hns/hns_roce_cq.c
+++ b/drivers/infiniband/hw/hns/hns_roce_cq.c
@@ -536,4 +536,5 @@ void hns_roce_cleanup_cq_table(struct hns_roce_dev *hr_dev)
 
 	for (i = 0; i < HNS_ROCE_CQ_BANK_NUM; i++)
 		ida_destroy(&hr_dev->cq_table.bank[i].ida);
+	mutex_destroy(&hr_dev->cq_table.bank_mutex);
 }
diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 0b47c6d688..7d5931872f 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -83,6 +83,7 @@
 #define MR_TYPE_DMA				0x03
 
 #define HNS_ROCE_FRMR_MAX_PA			512
+#define HNS_ROCE_FRMR_ALIGN_SIZE		128
 
 #define PKEY_ID					0xffff
 #define NODE_DESC_SIZE				64
@@ -91,6 +92,8 @@
 /* Configure to HW for PAGE_SIZE larger than 4KB */
 #define PG_SHIFT_OFFSET				(PAGE_SHIFT - 12)
 
+#define ATOMIC_WR_LEN				8
+
 #define HNS_ROCE_IDX_QUE_ENTRY_SZ		4
 #define SRQ_DB_REG				0x230
 
@@ -187,6 +190,9 @@ enum {
 #define HNS_HW_PAGE_SHIFT			12
 #define HNS_HW_PAGE_SIZE			(1 << HNS_HW_PAGE_SHIFT)
 
+#define HNS_HW_MAX_PAGE_SHIFT			27
+#define HNS_HW_MAX_PAGE_SIZE			(1 << HNS_HW_MAX_PAGE_SHIFT)
+
 struct hns_roce_uar {
 	u64		pfn;
 	unsigned long	index;
@@ -648,6 +654,8 @@ struct hns_roce_qp {
 	struct hns_user_mmap_entry *dwqe_mmap_entry;
 	u32			config;
 	enum hns_roce_cong_type	cong_type;
+	u8			tc_mode;
+	u8			priority;
 };
 
 struct hns_roce_ib_iboe {
@@ -713,6 +721,7 @@ struct hns_roce_eq {
 	int				shift;
 	int				event_type;
 	int				sub_type;
+	struct work_struct		work;
 };
 
 struct hns_roce_eq_table {
@@ -926,8 +935,7 @@ struct hns_roce_hw {
 	int (*rereg_write_mtpt)(struct hns_roce_dev *hr_dev,
 				struct hns_roce_mr *mr, int flags,
 				void *mb_buf);
-	int (*frmr_write_mtpt)(struct hns_roce_dev *hr_dev, void *mb_buf,
-			       struct hns_roce_mr *mr);
+	int (*frmr_write_mtpt)(void *mb_buf, struct hns_roce_mr *mr);
 	int (*mw_write_mtpt)(void *mb_buf, struct hns_roce_mw *mw);
 	void (*write_cqc)(struct hns_roce_dev *hr_dev,
 			  struct hns_roce_cq *hr_cq, void *mb_buf, u64 *mtts,
@@ -953,6 +961,8 @@ struct hns_roce_hw {
 	int (*query_sccc)(struct hns_roce_dev *hr_dev, u32 qpn, void *buffer);
 	int (*query_hw_counter)(struct hns_roce_dev *hr_dev,
 				u64 *stats, u32 port, int *hw_counters);
+	int (*get_dscp)(struct hns_roce_dev *hr_dev, u8 dscp,
+			u8 *tc_mode, u8 *priority);
 	const struct ib_device_ops *hns_roce_dev_ops;
 	const struct ib_device_ops *hns_roce_dev_srq_ops;
 };
@@ -1231,7 +1241,7 @@ struct hns_roce_buf *hns_roce_buf_alloc(struct hns_roce_dev *hr_dev, u32 size,
 int hns_roce_get_kmem_bufs(struct hns_roce_dev *hr_dev, dma_addr_t *bufs,
 			   int buf_cnt, struct hns_roce_buf *buf,
 			   unsigned int page_shift);
-int hns_roce_get_umem_bufs(struct hns_roce_dev *hr_dev, dma_addr_t *bufs,
+int hns_roce_get_umem_bufs(dma_addr_t *bufs,
 			   int buf_cnt, struct ib_umem *umem,
 			   unsigned int page_shift);
 
@@ -1295,4 +1305,6 @@ struct hns_user_mmap_entry *
 hns_roce_user_mmap_entry_insert(struct ib_ucontext *ucontext, u64 address,
 				size_t length,
 				enum hns_roce_mmap_type mmap_type);
+bool check_sl_valid(struct hns_roce_dev *hr_dev, u8 sl);
+
 #endif /* _HNS_ROCE_DEVICE_H */
diff --git a/drivers/infiniband/hw/hns/hns_roce_hem.c b/drivers/infiniband/hw/hns/hns_roce_hem.c
index 658c522be7..02baa853a7 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hem.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hem.c
@@ -877,6 +877,7 @@ void hns_roce_cleanup_hem_table(struct hns_roce_dev *hr_dev,
 
 	if (hns_roce_check_whether_mhop(hr_dev, table->type)) {
 		hns_roce_cleanup_mhop_hem_table(hr_dev, table);
+		mutex_destroy(&table->mutex);
 		return;
 	}
 
@@ -891,6 +892,7 @@ void hns_roce_cleanup_hem_table(struct hns_roce_dev *hr_dev,
 			hns_roce_free_hem(hr_dev, table->hem[i]);
 		}
 
+	mutex_destroy(&table->mutex);
 	kfree(table->hem);
 }
 
@@ -986,15 +988,13 @@ static void hem_list_free_all(struct hns_roce_dev *hr_dev,
 	}
 }
 
-static void hem_list_link_bt(struct hns_roce_dev *hr_dev, void *base_addr,
-			     u64 table_addr)
+static void hem_list_link_bt(void *base_addr, u64 table_addr)
 {
 	*(u64 *)(base_addr) = table_addr;
 }
 
 /* assign L0 table address to hem from root bt */
-static void hem_list_assign_bt(struct hns_roce_dev *hr_dev,
-			       struct hns_roce_hem_item *hem, void *cpu_addr,
+static void hem_list_assign_bt(struct hns_roce_hem_item *hem, void *cpu_addr,
 			       u64 phy_addr)
 {
 	hem->addr = cpu_addr;
@@ -1163,8 +1163,7 @@ static int hem_list_alloc_mid_bt(struct hns_roce_dev *hr_dev,
 		if (level > 1) {
 			pre = hem_ptrs[level - 1];
 			step = (cur->start - pre->start) / step * BA_BYTE_LEN;
-			hem_list_link_bt(hr_dev, pre->addr + step,
-					 cur->dma_addr);
+			hem_list_link_bt(pre->addr + step, cur->dma_addr);
 		}
 	}
 
@@ -1222,7 +1221,7 @@ static int alloc_fake_root_bt(struct hns_roce_dev *hr_dev, void *cpu_base,
 	if (!hem)
 		return -ENOMEM;
 
-	hem_list_assign_bt(hr_dev, hem, cpu_base, phy_base);
+	hem_list_assign_bt(hem, cpu_base, phy_base);
 	list_add(&hem->list, branch_head);
 	list_add(&hem->sibling, leaf_head);
 
@@ -1245,7 +1244,7 @@ static int setup_middle_bt(struct hns_roce_dev *hr_dev, void *cpu_base,
 	/* if exist mid bt, link L1 to L0 */
 	list_for_each_entry_safe(hem, temp_hem, branch_head, list) {
 		offset = (hem->start - r->offset) / step * BA_BYTE_LEN;
-		hem_list_link_bt(hr_dev, cpu_base + offset, hem->dma_addr);
+		hem_list_link_bt(cpu_base + offset, hem->dma_addr);
 		total++;
 	}
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index 8800464c9a..621b057fb9 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -36,6 +36,7 @@
 #include <linux/iopoll.h>
 #include <linux/kernel.h>
 #include <linux/types.h>
+#include <linux/workqueue.h>
 #include <net/addrconf.h>
 #include <rdma/ib_addr.h>
 #include <rdma/ib_cache.h>
@@ -443,10 +444,6 @@ static int fill_ud_av(struct hns_roce_v2_ud_send_wqe *ud_sq_wqe,
 	hr_reg_write(ud_sq_wqe, UD_SEND_WQE_HOPLIMIT, ah->av.hop_limit);
 	hr_reg_write(ud_sq_wqe, UD_SEND_WQE_TCLASS, ah->av.tclass);
 	hr_reg_write(ud_sq_wqe, UD_SEND_WQE_FLOW_LABEL, ah->av.flowlabel);
-
-	if (WARN_ON(ah->av.sl > MAX_SERVICE_LEVEL))
-		return -EINVAL;
-
 	hr_reg_write(ud_sq_wqe, UD_SEND_WQE_SL, ah->av.sl);
 
 	ud_sq_wqe->sgid_index = ah->av.gid_index;
@@ -595,11 +592,16 @@ static inline int set_rc_wqe(struct hns_roce_qp *qp,
 		     (wr->send_flags & IB_SEND_SIGNALED) ? 1 : 0);
 
 	if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
-	    wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
+	    wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD) {
+		if (msg_len != ATOMIC_WR_LEN)
+			return -EINVAL;
 		set_atomic_seg(wr, rc_sq_wqe, valid_num_sge);
-	else if (wr->opcode != IB_WR_REG_MR)
+	} else if (wr->opcode != IB_WR_REG_MR) {
 		ret = set_rwqe_data_seg(&qp->ibqp, wr, rc_sq_wqe,
 					&curr_idx, valid_num_sge);
+		if (ret)
+			return ret;
+	}
 
 	/*
 	 * The pipeline can sequentially post all valid WQEs into WQ buffer,
@@ -1273,12 +1275,38 @@ static int hns_roce_cmd_err_convert_errno(u16 desc_ret)
 	return -EIO;
 }
 
+static u32 hns_roce_cmdq_tx_timeout(u16 opcode, u32 tx_timeout)
+{
+	static const struct hns_roce_cmdq_tx_timeout_map cmdq_tx_timeout[] = {
+		{HNS_ROCE_OPC_POST_MB, HNS_ROCE_OPC_POST_MB_TIMEOUT},
+	};
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(cmdq_tx_timeout); i++)
+		if (cmdq_tx_timeout[i].opcode == opcode)
+			return cmdq_tx_timeout[i].tx_timeout;
+
+	return tx_timeout;
+}
+
+static void hns_roce_wait_csq_done(struct hns_roce_dev *hr_dev, u16 opcode)
+{
+	struct hns_roce_v2_priv *priv = hr_dev->priv;
+	u32 tx_timeout = hns_roce_cmdq_tx_timeout(opcode, priv->cmq.tx_timeout);
+	u32 timeout = 0;
+
+	do {
+		if (hns_roce_cmq_csq_done(hr_dev))
+			break;
+		udelay(1);
+	} while (++timeout < tx_timeout);
+}
+
 static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev,
 			       struct hns_roce_cmq_desc *desc, int num)
 {
 	struct hns_roce_v2_priv *priv = hr_dev->priv;
 	struct hns_roce_v2_cmq_ring *csq = &priv->cmq.csq;
-	u32 timeout = 0;
 	u16 desc_ret;
 	u32 tail;
 	int ret;
@@ -1299,12 +1327,7 @@ static int __hns_roce_cmq_send(struct hns_roce_dev *hr_dev,
 
 	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CMDS_CNT]);
 
-	do {
-		if (hns_roce_cmq_csq_done(hr_dev))
-			break;
-		udelay(1);
-	} while (++timeout < priv->cmq.tx_timeout);
-
+	hns_roce_wait_csq_done(hr_dev, le16_to_cpu(desc->opcode));
 	if (hns_roce_cmq_csq_done(hr_dev)) {
 		ret = 0;
 		for (i = 0; i < num; i++) {
@@ -2461,14 +2484,16 @@ static int set_llm_cfg_to_hw(struct hns_roce_dev *hr_dev,
 static struct hns_roce_link_table *
 alloc_link_table_buf(struct hns_roce_dev *hr_dev)
 {
+	u16 total_sl = hr_dev->caps.sl_num * hr_dev->func_num;
 	struct hns_roce_v2_priv *priv = hr_dev->priv;
 	struct hns_roce_link_table *link_tbl;
 	u32 pg_shift, size, min_size;
 
 	link_tbl = &priv->ext_llm;
 	pg_shift = hr_dev->caps.llm_buf_pg_sz + PAGE_SHIFT;
-	size = hr_dev->caps.num_qps * HNS_ROCE_V2_EXT_LLM_ENTRY_SZ;
-	min_size = HNS_ROCE_EXT_LLM_MIN_PAGES(hr_dev->caps.sl_num) << pg_shift;
+	size = hr_dev->caps.num_qps * hr_dev->func_num *
+	       HNS_ROCE_V2_EXT_LLM_ENTRY_SZ;
+	min_size = HNS_ROCE_EXT_LLM_MIN_PAGES(total_sl) << pg_shift;
 
 	/* Alloc data table */
 	size = max(size, min_size);
@@ -2671,6 +2696,8 @@ static void free_mr_exit(struct hns_roce_dev *hr_dev)
 		kfree(free_mr->rsv_pd);
 		free_mr->rsv_pd = NULL;
 	}
+
+	mutex_destroy(&free_mr->mutex);
 }
 
 static int free_mr_alloc_res(struct hns_roce_dev *hr_dev)
@@ -2821,8 +2848,10 @@ static int free_mr_init(struct hns_roce_dev *hr_dev)
 	mutex_init(&free_mr->mutex);
 
 	ret = free_mr_alloc_res(hr_dev);
-	if (ret)
+	if (ret) {
+		mutex_destroy(&free_mr->mutex);
 		return ret;
+	}
 
 	ret = free_mr_modify_qp(hr_dev);
 	if (ret)
@@ -3208,13 +3237,14 @@ static int set_mtpt_pbl(struct hns_roce_dev *hr_dev,
 
 	/* Aligned to the hardware address access unit */
 	for (i = 0; i < ARRAY_SIZE(pages); i++)
-		pages[i] >>= 6;
+		pages[i] >>= MPT_PBL_BUF_ADDR_S;
 
 	pbl_ba = hns_roce_get_mtr_ba(&mr->pbl_mtr);
 
 	mpt_entry->pbl_size = cpu_to_le32(mr->npages);
-	mpt_entry->pbl_ba_l = cpu_to_le32(pbl_ba >> 3);
-	hr_reg_write(mpt_entry, MPT_PBL_BA_H, upper_32_bits(pbl_ba >> 3));
+	mpt_entry->pbl_ba_l = cpu_to_le32(pbl_ba >> MPT_PBL_BA_ADDR_S);
+	hr_reg_write(mpt_entry, MPT_PBL_BA_H,
+		     upper_32_bits(pbl_ba >> MPT_PBL_BA_ADDR_S));
 
 	mpt_entry->pa0_l = cpu_to_le32(lower_32_bits(pages[0]));
 	hr_reg_write(mpt_entry, MPT_PA0_H, upper_32_bits(pages[0]));
@@ -3307,8 +3337,7 @@ static int hns_roce_v2_rereg_write_mtpt(struct hns_roce_dev *hr_dev,
 	return ret;
 }
 
-static int hns_roce_v2_frmr_write_mtpt(struct hns_roce_dev *hr_dev,
-				       void *mb_buf, struct hns_roce_mr *mr)
+static int hns_roce_v2_frmr_write_mtpt(void *mb_buf, struct hns_roce_mr *mr)
 {
 	dma_addr_t pbl_ba = hns_roce_get_mtr_ba(&mr->pbl_mtr);
 	struct hns_roce_v2_mpt_entry *mpt_entry;
@@ -3335,8 +3364,10 @@ static int hns_roce_v2_frmr_write_mtpt(struct hns_roce_dev *hr_dev,
 
 	mpt_entry->pbl_size = cpu_to_le32(mr->npages);
 
-	mpt_entry->pbl_ba_l = cpu_to_le32(lower_32_bits(pbl_ba >> 3));
-	hr_reg_write(mpt_entry, MPT_PBL_BA_H, upper_32_bits(pbl_ba >> 3));
+	mpt_entry->pbl_ba_l = cpu_to_le32(lower_32_bits(pbl_ba >>
+							MPT_PBL_BA_ADDR_S));
+	hr_reg_write(mpt_entry, MPT_PBL_BA_H,
+		     upper_32_bits(pbl_ba >> MPT_PBL_BA_ADDR_S));
 
 	return 0;
 }
@@ -3582,14 +3613,14 @@ static void hns_roce_v2_write_cqc(struct hns_roce_dev *hr_dev,
 		     to_hr_hw_page_shift(hr_cq->mtr.hem_cfg.ba_pg_shift));
 	hr_reg_write(cq_context, CQC_CQE_BUF_PG_SZ,
 		     to_hr_hw_page_shift(hr_cq->mtr.hem_cfg.buf_pg_shift));
-	hr_reg_write(cq_context, CQC_CQE_BA_L, dma_handle >> 3);
-	hr_reg_write(cq_context, CQC_CQE_BA_H, (dma_handle >> (32 + 3)));
+	hr_reg_write(cq_context, CQC_CQE_BA_L, dma_handle >> CQC_CQE_BA_L_S);
+	hr_reg_write(cq_context, CQC_CQE_BA_H, dma_handle >> CQC_CQE_BA_H_S);
 	hr_reg_write_bool(cq_context, CQC_DB_RECORD_EN,
 			  hr_cq->flags & HNS_ROCE_CQ_FLAG_RECORD_DB);
 	hr_reg_write(cq_context, CQC_CQE_DB_RECORD_ADDR_L,
 		     ((u32)hr_cq->db.dma) >> 1);
 	hr_reg_write(cq_context, CQC_CQE_DB_RECORD_ADDR_H,
-		     hr_cq->db.dma >> 32);
+		     hr_cq->db.dma >> CQC_CQE_DB_RECORD_ADDR_H_S);
 	hr_reg_write(cq_context, CQC_CQ_MAX_CNT,
 		     HNS_ROCE_V2_CQ_DEFAULT_BURST_NUM);
 	hr_reg_write(cq_context, CQC_CQ_PERIOD,
@@ -4218,8 +4249,7 @@ static void set_access_flags(struct hns_roce_qp *hr_qp,
 }
 
 static void set_qpc_wqe_cnt(struct hns_roce_qp *hr_qp,
-			    struct hns_roce_v2_qp_context *context,
-			    struct hns_roce_v2_qp_context *qpc_mask)
+			    struct hns_roce_v2_qp_context *context)
 {
 	hr_reg_write(context, QPC_SGE_SHIFT,
 		     to_hr_hem_entries_shift(hr_qp->sge.sge_cnt,
@@ -4241,7 +4271,6 @@ static inline int get_pdn(struct ib_pd *ib_pd)
 }
 
 static void modify_qp_reset_to_init(struct ib_qp *ibqp,
-				    const struct ib_qp_attr *attr,
 				    struct hns_roce_v2_qp_context *context,
 				    struct hns_roce_v2_qp_context *qpc_mask)
 {
@@ -4260,7 +4289,7 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 
 	hr_reg_write(context, QPC_RQWS, ilog2(hr_qp->rq.max_gs));
 
-	set_qpc_wqe_cnt(hr_qp, context, qpc_mask);
+	set_qpc_wqe_cnt(hr_qp, context);
 
 	/* No VLAN need to set 0xFFF */
 	hr_reg_write(context, QPC_VLAN_ID, 0xfff);
@@ -4301,7 +4330,6 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 }
 
 static void modify_qp_init_to_init(struct ib_qp *ibqp,
-				   const struct ib_qp_attr *attr,
 				   struct hns_roce_v2_qp_context *context,
 				   struct hns_roce_v2_qp_context *qpc_mask)
 {
@@ -4522,16 +4550,16 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 		return -EINVAL;
 	}
 
-	hr_reg_write(context, QPC_TRRL_BA_L, trrl_ba >> 4);
+	hr_reg_write(context, QPC_TRRL_BA_L, trrl_ba >> QPC_TRRL_BA_L_S);
 	hr_reg_clear(qpc_mask, QPC_TRRL_BA_L);
-	context->trrl_ba = cpu_to_le32(trrl_ba >> (16 + 4));
+	context->trrl_ba = cpu_to_le32(trrl_ba >> QPC_TRRL_BA_M_S);
 	qpc_mask->trrl_ba = 0;
-	hr_reg_write(context, QPC_TRRL_BA_H, trrl_ba >> (32 + 16 + 4));
+	hr_reg_write(context, QPC_TRRL_BA_H, trrl_ba >> QPC_TRRL_BA_H_S);
 	hr_reg_clear(qpc_mask, QPC_TRRL_BA_H);
 
-	context->irrl_ba = cpu_to_le32(irrl_ba >> 6);
+	context->irrl_ba = cpu_to_le32(irrl_ba >> QPC_IRRL_BA_L_S);
 	qpc_mask->irrl_ba = 0;
-	hr_reg_write(context, QPC_IRRL_BA_H, irrl_ba >> (32 + 6));
+	hr_reg_write(context, QPC_IRRL_BA_H, irrl_ba >> QPC_IRRL_BA_H_S);
 	hr_reg_clear(qpc_mask, QPC_IRRL_BA_H);
 
 	hr_reg_enable(context, QPC_RMT_E2E);
@@ -4593,8 +4621,9 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 	hr_reg_clear(qpc_mask, QPC_TRRL_HEAD_MAX);
 	hr_reg_clear(qpc_mask, QPC_TRRL_TAIL_MAX);
 
+#define MAX_LP_SGEN 3
 	/* rocee send 2^lp_sgen_ini segs every time */
-	hr_reg_write(context, QPC_LP_SGEN_INI, 3);
+	hr_reg_write(context, QPC_LP_SGEN_INI, MAX_LP_SGEN);
 	hr_reg_clear(qpc_mask, QPC_LP_SGEN_INI);
 
 	if (udata && ibqp->qp_type == IB_QPT_RC &&
@@ -4620,8 +4649,7 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 	return 0;
 }
 
-static int modify_qp_rtr_to_rts(struct ib_qp *ibqp,
-				const struct ib_qp_attr *attr, int attr_mask,
+static int modify_qp_rtr_to_rts(struct ib_qp *ibqp, int attr_mask,
 				struct hns_roce_v2_qp_context *context,
 				struct hns_roce_v2_qp_context *qpc_mask)
 {
@@ -4686,7 +4714,7 @@ static int get_dip_ctx_idx(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 	*tail = (*tail == hr_dev->caps.num_qps - 1) ? 0 : (*tail + 1);
 
 	list_for_each_entry(hr_dip, &hr_dev->dip_list, node) {
-		if (!memcmp(grh->dgid.raw, hr_dip->dgid, 16)) {
+		if (!memcmp(grh->dgid.raw, hr_dip->dgid, GID_LEN_V2)) {
 			*dip_idx = hr_dip->dip_idx;
 			goto out;
 		}
@@ -4829,6 +4857,69 @@ static int fill_cong_field(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 	return 0;
 }
 
+static int hns_roce_hw_v2_get_dscp(struct hns_roce_dev *hr_dev, u8 dscp,
+				   u8 *tc_mode, u8 *priority)
+{
+	struct hns_roce_v2_priv *priv = hr_dev->priv;
+	struct hnae3_handle *handle = priv->handle;
+	const struct hnae3_ae_ops *ops = handle->ae_algo->ops;
+
+	if (!ops->get_dscp_prio)
+		return -EOPNOTSUPP;
+
+	return ops->get_dscp_prio(handle, dscp, tc_mode, priority);
+}
+
+bool check_sl_valid(struct hns_roce_dev *hr_dev, u8 sl)
+{
+	u32 max_sl;
+
+	max_sl = min_t(u32, MAX_SERVICE_LEVEL, hr_dev->caps.sl_num - 1);
+	if (unlikely(sl > max_sl)) {
+		ibdev_err_ratelimited(&hr_dev->ib_dev,
+				      "failed to set SL(%u). Shouldn't be larger than %u.\n",
+				      sl, max_sl);
+		return false;
+	}
+
+	return true;
+}
+
+static int hns_roce_set_sl(struct ib_qp *ibqp,
+			   const struct ib_qp_attr *attr,
+			   struct hns_roce_v2_qp_context *context,
+			   struct hns_roce_v2_qp_context *qpc_mask)
+{
+	const struct ib_global_route *grh = rdma_ah_read_grh(&attr->ah_attr);
+	struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device);
+	struct hns_roce_qp *hr_qp = to_hr_qp(ibqp);
+	struct ib_device *ibdev = &hr_dev->ib_dev;
+	int ret;
+
+	ret = hns_roce_hw_v2_get_dscp(hr_dev, get_tclass(&attr->ah_attr.grh),
+				      &hr_qp->tc_mode, &hr_qp->priority);
+	if (ret && ret != -EOPNOTSUPP &&
+	    grh->sgid_attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) {
+		ibdev_err_ratelimited(ibdev,
+				      "failed to get dscp, ret = %d.\n", ret);
+		return ret;
+	}
+
+	if (hr_qp->tc_mode == HNAE3_TC_MAP_MODE_DSCP &&
+	    grh->sgid_attr->gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP)
+		hr_qp->sl = hr_qp->priority;
+	else
+		hr_qp->sl = rdma_ah_get_sl(&attr->ah_attr);
+
+	if (!check_sl_valid(hr_dev, hr_qp->sl))
+		return -EINVAL;
+
+	hr_reg_write(context, QPC_SL, hr_qp->sl);
+	hr_reg_clear(qpc_mask, QPC_SL);
+
+	return 0;
+}
+
 static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 				const struct ib_qp_attr *attr,
 				int attr_mask,
@@ -4844,25 +4935,18 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 	int is_roce_protocol;
 	u16 vlan_id = 0xffff;
 	bool is_udp = false;
-	u32 max_sl;
 	u8 ib_port;
 	u8 hr_port;
 	int ret;
 
-	max_sl = min_t(u32, MAX_SERVICE_LEVEL, hr_dev->caps.sl_num - 1);
-	if (unlikely(sl > max_sl)) {
-		ibdev_err_ratelimited(ibdev,
-				      "failed to fill QPC, sl (%u) shouldn't be larger than %u.\n",
-				      sl, max_sl);
-		return -EINVAL;
-	}
-
 	/*
 	 * If free_mr_en of qp is set, it means that this qp comes from
 	 * free mr. This qp will perform the loopback operation.
 	 * In the loopback scenario, only sl needs to be set.
 	 */
 	if (hr_qp->free_mr_en) {
+		if (!check_sl_valid(hr_dev, sl))
+			return -EINVAL;
 		hr_reg_write(context, QPC_SL, sl);
 		hr_reg_clear(qpc_mask, QPC_SL);
 		hr_qp->sl = sl;
@@ -4932,11 +5016,7 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 	memcpy(context->dgid, grh->dgid.raw, sizeof(grh->dgid.raw));
 	memset(qpc_mask->dgid, 0, sizeof(grh->dgid.raw));
 
-	hr_qp->sl = sl;
-	hr_reg_write(context, QPC_SL, hr_qp->sl);
-	hr_reg_clear(qpc_mask, QPC_SL);
-
-	return 0;
+	return  hns_roce_set_sl(ibqp, attr, context, qpc_mask);
 }
 
 static bool check_qp_state(enum ib_qp_state cur_state,
@@ -4983,15 +5063,14 @@ static int hns_roce_v2_set_abs_fields(struct ib_qp *ibqp,
 
 	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
 		memset(qpc_mask, 0, hr_dev->caps.qpc_sz);
-		modify_qp_reset_to_init(ibqp, attr, context, qpc_mask);
+		modify_qp_reset_to_init(ibqp, context, qpc_mask);
 	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_INIT) {
-		modify_qp_init_to_init(ibqp, attr, context, qpc_mask);
+		modify_qp_init_to_init(ibqp, context, qpc_mask);
 	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_RTR) {
 		ret = modify_qp_init_to_rtr(ibqp, attr, attr_mask, context,
 					    qpc_mask, udata);
 	} else if (cur_state == IB_QPS_RTR && new_state == IB_QPS_RTS) {
-		ret = modify_qp_rtr_to_rts(ibqp, attr, attr_mask, context,
-					   qpc_mask);
+		ret = modify_qp_rtr_to_rts(ibqp, attr_mask, context, qpc_mask);
 	}
 
 	return ret;
@@ -6085,33 +6164,11 @@ static struct hns_roce_ceqe *next_ceqe_sw_v2(struct hns_roce_eq *eq)
 		!!(eq->cons_index & eq->entries)) ? ceqe : NULL;
 }
 
-static irqreturn_t hns_roce_v2_ceq_int(struct hns_roce_dev *hr_dev,
-				       struct hns_roce_eq *eq)
+static irqreturn_t hns_roce_v2_ceq_int(struct hns_roce_eq *eq)
 {
-	struct hns_roce_ceqe *ceqe = next_ceqe_sw_v2(eq);
-	irqreturn_t ceqe_found = IRQ_NONE;
-	u32 cqn;
-
-	while (ceqe) {
-		/* Make sure we read CEQ entry after we have checked the
-		 * ownership bit
-		 */
-		dma_rmb();
+	queue_work(system_bh_wq, &eq->work);
 
-		cqn = hr_reg_read(ceqe, CEQE_CQN);
-
-		hns_roce_cq_completion(hr_dev, cqn);
-
-		++eq->cons_index;
-		ceqe_found = IRQ_HANDLED;
-		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CEQE_CNT]);
-
-		ceqe = next_ceqe_sw_v2(eq);
-	}
-
-	update_eq_db(eq);
-
-	return IRQ_RETVAL(ceqe_found);
+	return IRQ_HANDLED;
 }
 
 static irqreturn_t hns_roce_v2_msix_interrupt_eq(int irq, void *eq_ptr)
@@ -6122,7 +6179,7 @@ static irqreturn_t hns_roce_v2_msix_interrupt_eq(int irq, void *eq_ptr)
 
 	if (eq->type_flag == HNS_ROCE_CEQ)
 		/* Completion event interrupt */
-		int_work = hns_roce_v2_ceq_int(hr_dev, eq);
+		int_work = hns_roce_v2_ceq_int(eq);
 	else
 		/* Asynchronous event interrupt */
 		int_work = hns_roce_v2_aeq_int(hr_dev, eq);
@@ -6334,9 +6391,16 @@ static void hns_roce_v2_int_mask_enable(struct hns_roce_dev *hr_dev,
 	roce_write(hr_dev, ROCEE_VF_ABN_INT_CFG_REG, enable_flag);
 }
 
-static void hns_roce_v2_destroy_eqc(struct hns_roce_dev *hr_dev, u32 eqn)
+static void free_eq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq)
+{
+	hns_roce_mtr_destroy(hr_dev, &eq->mtr);
+}
+
+static void hns_roce_v2_destroy_eqc(struct hns_roce_dev *hr_dev,
+				    struct hns_roce_eq *eq)
 {
 	struct device *dev = hr_dev->dev;
+	int eqn = eq->eqn;
 	int ret;
 	u8 cmd;
 
@@ -6347,12 +6411,9 @@ static void hns_roce_v2_destroy_eqc(struct hns_roce_dev *hr_dev, u32 eqn)
 
 	ret = hns_roce_destroy_hw_ctx(hr_dev, cmd, eqn & HNS_ROCE_V2_EQN_M);
 	if (ret)
-		dev_err(dev, "[mailbox cmd] destroy eqc(%u) failed.\n", eqn);
-}
+		dev_err(dev, "[mailbox cmd] destroy eqc(%d) failed.\n", eqn);
 
-static void free_eq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq)
-{
-	hns_roce_mtr_destroy(hr_dev, &eq->mtr);
+	free_eq_buf(hr_dev, eq);
 }
 
 static void init_eq_config(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq)
@@ -6490,6 +6551,34 @@ free_cmd_mbox:
 	return ret;
 }
 
+static void hns_roce_ceq_work(struct work_struct *work)
+{
+	struct hns_roce_eq *eq = from_work(eq, work, work);
+	struct hns_roce_ceqe *ceqe = next_ceqe_sw_v2(eq);
+	struct hns_roce_dev *hr_dev = eq->hr_dev;
+	int ceqe_num = 0;
+	u32 cqn;
+
+	while (ceqe && ceqe_num < hr_dev->caps.ceqe_depth) {
+		/* Make sure we read CEQ entry after we have checked the
+		 * ownership bit
+		 */
+		dma_rmb();
+
+		cqn = hr_reg_read(ceqe, CEQE_CQN);
+
+		hns_roce_cq_completion(hr_dev, cqn);
+
+		++eq->cons_index;
+		++ceqe_num;
+		atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_CEQE_CNT]);
+
+		ceqe = next_ceqe_sw_v2(eq);
+	}
+
+	update_eq_db(eq);
+}
+
 static int __hns_roce_request_irq(struct hns_roce_dev *hr_dev, int irq_num,
 				  int comp_num, int aeq_num, int other_num)
 {
@@ -6521,21 +6610,24 @@ static int __hns_roce_request_irq(struct hns_roce_dev *hr_dev, int irq_num,
 			 j - other_num - aeq_num);
 
 	for (j = 0; j < irq_num; j++) {
-		if (j < other_num)
+		if (j < other_num) {
 			ret = request_irq(hr_dev->irq[j],
 					  hns_roce_v2_msix_interrupt_abn,
 					  0, hr_dev->irq_names[j], hr_dev);
-
-		else if (j < (other_num + comp_num))
+		} else if (j < (other_num + comp_num)) {
+			INIT_WORK(&eq_table->eq[j - other_num].work,
+				  hns_roce_ceq_work);
 			ret = request_irq(eq_table->eq[j - other_num].irq,
 					  hns_roce_v2_msix_interrupt_eq,
 					  0, hr_dev->irq_names[j + aeq_num],
 					  &eq_table->eq[j - other_num]);
-		else
+		} else {
 			ret = request_irq(eq_table->eq[j - other_num].irq,
 					  hns_roce_v2_msix_interrupt_eq,
 					  0, hr_dev->irq_names[j - comp_num],
 					  &eq_table->eq[j - other_num]);
+		}
+
 		if (ret) {
 			dev_err(hr_dev->dev, "request irq error!\n");
 			goto err_request_failed;
@@ -6545,12 +6637,16 @@ static int __hns_roce_request_irq(struct hns_roce_dev *hr_dev, int irq_num,
 	return 0;
 
 err_request_failed:
-	for (j -= 1; j >= 0; j--)
-		if (j < other_num)
+	for (j -= 1; j >= 0; j--) {
+		if (j < other_num) {
 			free_irq(hr_dev->irq[j], hr_dev);
-		else
-			free_irq(eq_table->eq[j - other_num].irq,
-				 &eq_table->eq[j - other_num]);
+			continue;
+		}
+		free_irq(eq_table->eq[j - other_num].irq,
+			 &eq_table->eq[j - other_num]);
+		if (j < other_num + comp_num)
+			cancel_work_sync(&eq_table->eq[j - other_num].work);
+	}
 
 err_kzalloc_failed:
 	for (i -= 1; i >= 0; i--)
@@ -6571,8 +6667,11 @@ static void __hns_roce_free_irq(struct hns_roce_dev *hr_dev)
 	for (i = 0; i < hr_dev->caps.num_other_vectors; i++)
 		free_irq(hr_dev->irq[i], hr_dev);
 
-	for (i = 0; i < eq_num; i++)
+	for (i = 0; i < eq_num; i++) {
 		free_irq(hr_dev->eq_table.eq[i].irq, &hr_dev->eq_table.eq[i]);
+		if (i < hr_dev->caps.num_comp_vectors)
+			cancel_work_sync(&hr_dev->eq_table.eq[i].work);
+	}
 
 	for (i = 0; i < irq_num; i++)
 		kfree(hr_dev->irq_names[i]);
@@ -6661,7 +6760,7 @@ err_request_irq_fail:
 
 err_create_eq_fail:
 	for (i -= 1; i >= 0; i--)
-		free_eq_buf(hr_dev, &eq_table->eq[i]);
+		hns_roce_v2_destroy_eqc(hr_dev, &eq_table->eq[i]);
 	kfree(eq_table->eq);
 
 	return ret;
@@ -6681,11 +6780,8 @@ static void hns_roce_v2_cleanup_eq_table(struct hns_roce_dev *hr_dev)
 	__hns_roce_free_irq(hr_dev);
 	destroy_workqueue(hr_dev->irq_workq);
 
-	for (i = 0; i < eq_num; i++) {
-		hns_roce_v2_destroy_eqc(hr_dev, i);
-
-		free_eq_buf(hr_dev, &eq_table->eq[i]);
-	}
+	for (i = 0; i < eq_num; i++)
+		hns_roce_v2_destroy_eqc(hr_dev, &eq_table->eq[i]);
 
 	kfree(eq_table->eq);
 }
@@ -6736,6 +6832,7 @@ static const struct hns_roce_hw hns_roce_hw_v2 = {
 	.query_srqc = hns_roce_v2_query_srqc,
 	.query_sccc = hns_roce_v2_query_sccc,
 	.query_hw_counter = hns_roce_hw_v2_query_counter,
+	.get_dscp = hns_roce_hw_v2_get_dscp,
 	.hns_roce_dev_ops = &hns_roce_v2_dev_ops,
 	.hns_roce_dev_srq_ops = &hns_roce_v2_dev_srq_ops,
 };
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
index dfed6b4ddb..c65f68a14a 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.h
@@ -224,6 +224,12 @@ enum hns_roce_opcode_type {
 	HNS_SWITCH_PARAMETER_CFG			= 0x1033,
 };
 
+#define HNS_ROCE_OPC_POST_MB_TIMEOUT 35000
+struct hns_roce_cmdq_tx_timeout_map {
+	u16 opcode;
+	u32 tx_timeout;
+};
+
 enum {
 	TYPE_CRQ,
 	TYPE_CSQ,
@@ -276,6 +282,10 @@ struct hns_roce_v2_cq_context {
 	__le32 byte_64_se_cqe_idx;
 };
 
+#define CQC_CQE_BA_L_S 3
+#define CQC_CQE_BA_H_S (32 + CQC_CQE_BA_L_S)
+#define CQC_CQE_DB_RECORD_ADDR_H_S 32
+
 #define HNS_ROCE_V2_CQ_DEFAULT_BURST_NUM 0x0
 #define HNS_ROCE_V2_CQ_DEFAULT_INTERVAL	0x0
 
@@ -447,6 +457,12 @@ struct hns_roce_v2_qp_context {
 	struct hns_roce_v2_qp_context_ex ext;
 };
 
+#define QPC_TRRL_BA_L_S 4
+#define QPC_TRRL_BA_M_S (16 + QPC_TRRL_BA_L_S)
+#define QPC_TRRL_BA_H_S (32 + QPC_TRRL_BA_M_S)
+#define QPC_IRRL_BA_L_S 6
+#define QPC_IRRL_BA_H_S (32 + QPC_IRRL_BA_L_S)
+
 #define QPC_FIELD_LOC(h, l) FIELD_LOC(struct hns_roce_v2_qp_context, h, l)
 
 #define QPC_TST QPC_FIELD_LOC(2, 0)
@@ -716,6 +732,9 @@ struct hns_roce_v2_mpt_entry {
 	__le32	byte_64_buf_pa1;
 };
 
+#define MPT_PBL_BUF_ADDR_S 6
+#define MPT_PBL_BA_ADDR_S 3
+
 #define MPT_FIELD_LOC(h, l) FIELD_LOC(struct hns_roce_v2_mpt_entry, h, l)
 
 #define MPT_ST MPT_FIELD_LOC(1, 0)
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index d202258368..4cb0af7335 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -429,6 +429,9 @@ static int hns_roce_alloc_ucontext(struct ib_ucontext *uctx,
 	return 0;
 
 error_fail_copy_to_udata:
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_CQ_RECORD_DB ||
+	    hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_QP_RECORD_DB)
+		mutex_destroy(&context->page_mutex);
 	hns_roce_dealloc_uar_entry(context);
 
 error_fail_uar_entry:
@@ -445,6 +448,10 @@ static void hns_roce_dealloc_ucontext(struct ib_ucontext *ibcontext)
 	struct hns_roce_ucontext *context = to_hr_ucontext(ibcontext);
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibcontext->device);
 
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_CQ_RECORD_DB ||
+	    hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_QP_RECORD_DB)
+		mutex_destroy(&context->page_mutex);
+
 	hns_roce_dealloc_uar_entry(context);
 
 	ida_free(&hr_dev->uar_ida.ida, (int)context->uar.logic_idx);
@@ -933,6 +940,15 @@ err_unmap_dmpt:
 	return ret;
 }
 
+static void hns_roce_teardown_hca(struct hns_roce_dev *hr_dev)
+{
+	hns_roce_cleanup_bitmap(hr_dev);
+
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_CQ_RECORD_DB ||
+	    hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_QP_RECORD_DB)
+		mutex_destroy(&hr_dev->pgdir_mutex);
+}
+
 /**
  * hns_roce_setup_hca - setup host channel adapter
  * @hr_dev: pointer to hns roce device
@@ -981,6 +997,10 @@ static int hns_roce_setup_hca(struct hns_roce_dev *hr_dev)
 
 err_uar_table_free:
 	ida_destroy(&hr_dev->uar_ida.ida);
+	if (hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_CQ_RECORD_DB ||
+	    hr_dev->caps.flags & HNS_ROCE_CAP_FLAG_QP_RECORD_DB)
+		mutex_destroy(&hr_dev->pgdir_mutex);
+
 	return ret;
 }
 
@@ -1126,7 +1146,7 @@ error_failed_register_device:
 		hr_dev->hw->hw_exit(hr_dev);
 
 error_failed_engine_init:
-	hns_roce_cleanup_bitmap(hr_dev);
+	hns_roce_teardown_hca(hr_dev);
 
 error_failed_setup_hca:
 	hns_roce_cleanup_hem(hr_dev);
@@ -1156,7 +1176,7 @@ void hns_roce_exit(struct hns_roce_dev *hr_dev)
 
 	if (hr_dev->hw->hw_exit)
 		hr_dev->hw->hw_exit(hr_dev);
-	hns_roce_cleanup_bitmap(hr_dev);
+	hns_roce_teardown_hca(hr_dev);
 	hns_roce_cleanup_hem(hr_dev);
 
 	if (hr_dev->cmd_mod)
diff --git a/drivers/infiniband/hw/hns/hns_roce_mr.c b/drivers/infiniband/hw/hns/hns_roce_mr.c
index 80c050d7d0..846da8c78b 100644
--- a/drivers/infiniband/hw/hns/hns_roce_mr.c
+++ b/drivers/infiniband/hw/hns/hns_roce_mr.c
@@ -162,7 +162,7 @@ static int hns_roce_mr_enable(struct hns_roce_dev *hr_dev,
 	if (mr->type != MR_TYPE_FRMR)
 		ret = hr_dev->hw->write_mtpt(hr_dev, mailbox->buf, mr);
 	else
-		ret = hr_dev->hw->frmr_write_mtpt(hr_dev, mailbox->buf, mr);
+		ret = hr_dev->hw->frmr_write_mtpt(mailbox->buf, mr);
 	if (ret) {
 		dev_err(dev, "failed to write mtpt, ret = %d.\n", ret);
 		goto err_page;
@@ -443,6 +443,11 @@ int hns_roce_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
 	struct hns_roce_mtr *mtr = &mr->pbl_mtr;
 	int ret, sg_num = 0;
 
+	if (!IS_ALIGNED(*sg_offset, HNS_ROCE_FRMR_ALIGN_SIZE) ||
+	    ibmr->page_size < HNS_HW_PAGE_SIZE ||
+	    ibmr->page_size > HNS_HW_MAX_PAGE_SIZE)
+		return sg_num;
+
 	mr->npages = 0;
 	mr->page_list = kvcalloc(mr->pbl_mtr.hem_cfg.buf_pg_count,
 				 sizeof(dma_addr_t), GFP_KERNEL);
@@ -755,7 +760,7 @@ static int mtr_map_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr)
 		return -ENOMEM;
 
 	if (mtr->umem)
-		npage = hns_roce_get_umem_bufs(hr_dev, pages, page_count,
+		npage = hns_roce_get_umem_bufs(pages, page_count,
 					       mtr->umem, page_shift);
 	else
 		npage = hns_roce_get_kmem_bufs(hr_dev, pages, page_count,
diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
index f35a66325d..1de384ce4d 100644
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@@ -410,7 +410,8 @@ static void free_qpn(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp)
 
 	bankid = get_qp_bankid(hr_qp->qpn);
 
-	ida_free(&hr_dev->qp_table.bank[bankid].ida, hr_qp->qpn >> 3);
+	ida_free(&hr_dev->qp_table.bank[bankid].ida,
+		 hr_qp->qpn / HNS_ROCE_QP_BANK_NUM);
 
 	mutex_lock(&hr_dev->qp_table.bank_mutex);
 	hr_dev->qp_table.bank[bankid].inuse--;
@@ -531,13 +532,15 @@ static unsigned int get_sge_num_from_max_inl_data(bool is_ud_or_gsi,
 {
 	unsigned int inline_sge;
 
-	inline_sge = roundup_pow_of_two(max_inline_data) / HNS_ROCE_SGE_SIZE;
+	if (!max_inline_data)
+		return 0;
 
 	/*
 	 * if max_inline_data less than
 	 * HNS_ROCE_SGE_IN_WQE * HNS_ROCE_SGE_SIZE,
 	 * In addition to ud's mode, no need to extend sge.
 	 */
+	inline_sge = roundup_pow_of_two(max_inline_data) / HNS_ROCE_SGE_SIZE;
 	if (!is_ud_or_gsi && inline_sge <= HNS_ROCE_SGE_IN_WQE)
 		inline_sge = 0;
 
@@ -1117,7 +1120,6 @@ static int set_qp_param(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp,
 }
 
 static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
-				     struct ib_pd *ib_pd,
 				     struct ib_qp_init_attr *init_attr,
 				     struct ib_udata *udata,
 				     struct hns_roce_qp *hr_qp)
@@ -1140,7 +1142,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 	ret = set_qp_param(hr_dev, hr_qp, init_attr, udata, &ucmd);
 	if (ret) {
 		ibdev_err(ibdev, "failed to set QP param, ret = %d.\n", ret);
-		return ret;
+		goto err_out;
 	}
 
 	if (!udata) {
@@ -1148,7 +1150,7 @@ static int hns_roce_create_qp_common(struct hns_roce_dev *hr_dev,
 		if (ret) {
 			ibdev_err(ibdev, "failed to alloc wrid, ret = %d.\n",
 				  ret);
-			return ret;
+			goto err_out;
 		}
 	}
 
@@ -1219,6 +1221,8 @@ err_qpn:
 	free_qp_buf(hr_dev, hr_qp);
 err_buf:
 	free_kernel_wrid(hr_qp);
+err_out:
+	mutex_destroy(&hr_qp->mutex);
 	return ret;
 }
 
@@ -1234,6 +1238,7 @@ void hns_roce_qp_destroy(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp,
 	free_qp_buf(hr_dev, hr_qp);
 	free_kernel_wrid(hr_qp);
 	free_qp_db(hr_dev, hr_qp, udata);
+	mutex_destroy(&hr_qp->mutex);
 }
 
 static int check_qp_type(struct hns_roce_dev *hr_dev, enum ib_qp_type type,
@@ -1271,7 +1276,6 @@ int hns_roce_create_qp(struct ib_qp *qp, struct ib_qp_init_attr *init_attr,
 	struct ib_device *ibdev = qp->device;
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibdev);
 	struct hns_roce_qp *hr_qp = to_hr_qp(qp);
-	struct ib_pd *pd = qp->pd;
 	int ret;
 
 	ret = check_qp_type(hr_dev, init_attr->qp_type, !!udata);
@@ -1286,7 +1290,7 @@ int hns_roce_create_qp(struct ib_qp *qp, struct ib_qp_init_attr *init_attr,
 		hr_qp->phy_port = hr_dev->iboe.phy_port[hr_qp->port];
 	}
 
-	ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata, hr_qp);
+	ret = hns_roce_create_qp_common(hr_dev, init_attr, udata, hr_qp);
 	if (ret)
 		ibdev_err(ibdev, "create QP type 0x%x failed(%d)\n",
 			  init_attr->qp_type, ret);
@@ -1386,6 +1390,7 @@ int hns_roce_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 		       int attr_mask, struct ib_udata *udata)
 {
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device);
+	struct hns_roce_ib_modify_qp_resp resp = {};
 	struct hns_roce_qp *hr_qp = to_hr_qp(ibqp);
 	enum ib_qp_state cur_state, new_state;
 	int ret = -EINVAL;
@@ -1427,6 +1432,18 @@ int hns_roce_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 	ret = hr_dev->hw->modify_qp(ibqp, attr, attr_mask, cur_state,
 				    new_state, udata);
+	if (ret)
+		goto out;
+
+	if (udata && udata->outlen) {
+		resp.tc_mode = hr_qp->tc_mode;
+		resp.priority = hr_qp->sl;
+		ret = ib_copy_to_udata(udata, &resp,
+				       min(udata->outlen, sizeof(resp)));
+		if (ret)
+			ibdev_err_ratelimited(&hr_dev->ib_dev,
+					      "failed to copy modify qp resp.\n");
+	}
 
 out:
 	mutex_unlock(&hr_qp->mutex);
@@ -1561,5 +1578,7 @@ void hns_roce_cleanup_qp_table(struct hns_roce_dev *hr_dev)
 
 	for (i = 0; i < HNS_ROCE_QP_BANK_NUM; i++)
 		ida_destroy(&hr_dev->qp_table.bank[i].ida);
+	mutex_destroy(&hr_dev->qp_table.bank_mutex);
+	mutex_destroy(&hr_dev->qp_table.scc_mutex);
 	kfree(hr_dev->qp_table.idx_table.spare_idx);
 }
diff --git a/drivers/infiniband/hw/hns/hns_roce_srq.c b/drivers/infiniband/hw/hns/hns_roce_srq.c
index 8f48c6723e..c9b8233f4b 100644
--- a/drivers/infiniband/hw/hns/hns_roce_srq.c
+++ b/drivers/infiniband/hw/hns/hns_roce_srq.c
@@ -250,7 +250,7 @@ static void free_srq_wqe_buf(struct hns_roce_dev *hr_dev,
 	hns_roce_mtr_destroy(hr_dev, &srq->buf_mtr);
 }
 
-static int alloc_srq_wrid(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq)
+static int alloc_srq_wrid(struct hns_roce_srq *srq)
 {
 	srq->wrid = kvmalloc_array(srq->wqe_cnt, sizeof(u64), GFP_KERNEL);
 	if (!srq->wrid)
@@ -297,7 +297,7 @@ static int set_srq_basic_param(struct hns_roce_srq *srq,
 
 	max_sge = proc_srq_sge(hr_dev, srq, !!udata);
 	if (attr->max_wr > hr_dev->caps.max_srq_wrs ||
-	    attr->max_sge > max_sge) {
+	    attr->max_sge > max_sge || !attr->max_sge) {
 		ibdev_err(&hr_dev->ib_dev,
 			  "invalid SRQ attr, depth = %u, sge = %u.\n",
 			  attr->max_wr, attr->max_sge);
@@ -366,7 +366,7 @@ static int alloc_srq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_srq *srq,
 		goto err_idx;
 
 	if (!udata) {
-		ret = alloc_srq_wrid(hr_dev, srq);
+		ret = alloc_srq_wrid(srq);
 		if (ret)
 			goto err_wqe_buf;
 	}
@@ -518,6 +518,7 @@ err_srq_db:
 err_srq_buf:
 	free_srq_buf(hr_dev, srq);
 err_out:
+	mutex_destroy(&srq->mutex);
 	atomic64_inc(&hr_dev->dfx_cnt[HNS_ROCE_DFX_SRQ_CREATE_ERR_CNT]);
 
 	return ret;
@@ -532,6 +533,7 @@ int hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
 	free_srqn(hr_dev, srq);
 	free_srq_db(hr_dev, srq, udata);
 	free_srq_buf(hr_dev, srq);
+	mutex_destroy(&srq->mutex);
 	return 0;
 }