[PATCH V2 for-next 0/8] Bug fixes & Code improvements in hip06 and hip08 RoCE driver

[PATCH V2 for-next 0/8] Bug fixes & Code improvements in hip06 and hip08 RoCE driver

[Wei Hu (Xavier)]

This patch-set introduces some bug fixes and code improvements
for hip06 and hip08 RoCE driver. It includes a patch for fixing
the assign algorithm of qp_attr->max_rd_atomic and
qp_attr->max_dest_rd_atomic, three patches for static check errors,
one for setting attr mask, one for returning RoCE device ah_attr
type when querying qp, one for unregistering inet addr, and the
last one for command queue delay processing in hip08 driver.

Lijun Ou (6):
  RDMA/hns: Modify the value with rd&dest_rd of qp_attr
  RDMA/hns: Factor out the code for checking sdb status into a new
    function
  RDMA/hns: Set mask for destination qp field of qp context assignment
  RDMA/hns: Set rdma_ah_attr type for querying qp
  RDMA/hns: Remove unnecessarily calling unregister_inetaddr_notifier
    function
  RDMA/hns: Replace usleep_range with udelay when checking command
    status

Wei Hu (Xavier) (2):
  RDMA/hns: Add return statement when kzalloc return NULL in
    hns_roce_v1_recreate_lp_qp
  RDMA/hns: Add return statement when checking error in
    hns_roce_v1_mr_free_work_fn

 drivers/infiniband/hw/hns/hns_roce_device.h |   3 +-
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c  | 131 +++++++++++++++++-----------
 drivers/infiniband/hw/hns/hns_roce_hw_v2.c  |   9 +-
 drivers/infiniband/hw/hns/hns_roce_main.c   |  18 ++--
 4 files changed, 96 insertions(+), 65 deletions(-)

-- 
1.9.1

[PATCH V2 for-next 1/8] RDMA/hns: Modify the value with rd&dest_rd of qp_attr

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

The value of max_rd_atomic and max_dest_rd_atomic in query_qp
are incorrect. It should be assigned by left shifting of
the bit in hip06 SoC.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 426f55a..6e9acfd 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -3484,10 +3484,10 @@ static int hns_roce_v1_q_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 			      QP_CONTEXT_QPC_BYTES_12_P_KEY_INDEX_S);
 	qp_attr->port_num = hr_qp->port + 1;
 	qp_attr->sq_draining = 0;
-	qp_attr->max_rd_atomic = roce_get_field(context->qpc_bytes_156,
+	qp_attr->max_rd_atomic = 1 << roce_get_field(context->qpc_bytes_156,
 				 QP_CONTEXT_QPC_BYTES_156_INITIATOR_DEPTH_M,
 				 QP_CONTEXT_QPC_BYTES_156_INITIATOR_DEPTH_S);
-	qp_attr->max_dest_rd_atomic = roce_get_field(context->qpc_bytes_32,
+	qp_attr->max_dest_rd_atomic = 1 << roce_get_field(context->qpc_bytes_32,
 				 QP_CONTEXT_QPC_BYTES_32_RESPONDER_RESOURCES_M,
 				 QP_CONTEXT_QPC_BYTES_32_RESPONDER_RESOURCES_S);
 	qp_attr->min_rnr_timer = (u8)(roce_get_field(context->qpc_bytes_24,
-- 
1.9.1

[PATCH V2 for-next 2/8] RDMA/hns: Factor out the code for checking sdb status into a new function

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

It mainly places the lines for checking send doorbell status
into a special functions. As a result, we can directly call it in
check_qp_db_process_status function and keep consistent indenting
style.

It fixes the warning from static checker:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:3562 check_qp_db_process_status()
warn: inconsistent indenting.

Fixes: 5f110ac4bed8 ("IB/hns: Fix for checkpatch.pl comment style)

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
Patch V2:
1.modify the fixes to Fixes, and put it before signed-off-by at
  Leon's comment. The related link as below:
    https://lkml.org/lkml/2017/9/28/437

Patch V1: Initial  Submit
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 95 ++++++++++++++++--------------
 1 file changed, 51 insertions(+), 44 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 6e9acfd..95f5c88 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -3532,6 +3532,53 @@ int hns_roce_v1_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 		hns_roce_v1_q_qp(ibqp, qp_attr, qp_attr_mask, qp_init_attr);
 }
 
+static void hns_roce_check_sdb_status(struct hns_roce_dev *hr_dev,
+				      u32 *old_send, u32 *old_retry,
+				      u32 *tsp_st, u32 *success_flags)
+{
+	u32 sdb_retry_cnt;
+	u32 sdb_send_ptr;
+	u32 cur_cnt, old_cnt;
+	u32 send_ptr;
+
+	sdb_send_ptr = roce_read(hr_dev, ROCEE_SDB_SEND_PTR_REG);
+	sdb_retry_cnt =	roce_read(hr_dev, ROCEE_SDB_RETRY_CNT_REG);
+	cur_cnt = roce_get_field(sdb_send_ptr,
+				 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
+				 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
+		  roce_get_field(sdb_retry_cnt,
+				 ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
+				 ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
+	if (!roce_get_bit(*tsp_st, ROCEE_CNT_CLR_CE_CNT_CLR_CE_S)) {
+		old_cnt = roce_get_field(*old_send,
+					 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
+					 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
+			  roce_get_field(*old_retry,
+					 ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
+					 ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
+		if (cur_cnt - old_cnt > SDB_ST_CMP_VAL)
+			*success_flags = 1;
+	} else {
+		old_cnt = roce_get_field(*old_send,
+					 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
+					 ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S);
+		if (cur_cnt - old_cnt > SDB_ST_CMP_VAL) {
+			*success_flags = 1;
+		} else {
+			send_ptr = roce_get_field(*old_send,
+					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
+					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
+				   roce_get_field(sdb_retry_cnt,
+					    ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
+					    ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
+			roce_set_field(*old_send,
+				       ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
+				       ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S,
+				       send_ptr);
+		}
+	}
+}
+
 static int check_qp_db_process_status(struct hns_roce_dev *hr_dev,
 				      struct hns_roce_qp *hr_qp,
 				      u32 sdb_issue_ptr,
@@ -3539,12 +3586,10 @@ static int check_qp_db_process_status(struct hns_roce_dev *hr_dev,
 				      u32 *wait_stage)
 {
 	struct device *dev = &hr_dev->pdev->dev;
-	u32 sdb_retry_cnt, old_retry;
 	u32 sdb_send_ptr, old_send;
 	u32 success_flags = 0;
-	u32 cur_cnt, old_cnt;
 	unsigned long end;
-	u32 send_ptr;
+	u32 old_retry;
 	u32 inv_cnt;
 	u32 tsp_st;
 
@@ -3602,47 +3647,9 @@ static int check_qp_db_process_status(struct hns_roce_dev *hr_dev,
 
 				msleep(HNS_ROCE_V1_CHECK_DB_SLEEP_MSECS);
 
-				sdb_send_ptr = roce_read(hr_dev,
-							ROCEE_SDB_SEND_PTR_REG);
-				sdb_retry_cnt =	roce_read(hr_dev,
-						       ROCEE_SDB_RETRY_CNT_REG);
-				cur_cnt = roce_get_field(sdb_send_ptr,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
-					roce_get_field(sdb_retry_cnt,
-					ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
-					ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
-				if (!roce_get_bit(tsp_st,
-					ROCEE_CNT_CLR_CE_CNT_CLR_CE_S)) {
-					old_cnt = roce_get_field(old_send,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
-					roce_get_field(old_retry,
-					ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
-					ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
-					if (cur_cnt - old_cnt > SDB_ST_CMP_VAL)
-						success_flags = 1;
-				} else {
-					old_cnt = roce_get_field(old_send,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
-					ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S);
-					if (cur_cnt - old_cnt >
-					    SDB_ST_CMP_VAL) {
-						success_flags = 1;
-					} else {
-						send_ptr =
-							roce_get_field(old_send,
-					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
-					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S) +
-					    roce_get_field(sdb_retry_cnt,
-					    ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_M,
-					    ROCEE_SDB_RETRY_CNT_SDB_RETRY_CT_S);
-					    roce_set_field(old_send,
-					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_M,
-					    ROCEE_SDB_SEND_PTR_SDB_SEND_PTR_S,
-						send_ptr);
-					}
-				}
+				hns_roce_check_sdb_status(hr_dev, &old_send,
+							  &old_retry, &tsp_st,
+							  &success_flags);
 			} while (!success_flags);
 		}
 
-- 
1.9.1

[PATCH V2 for-next 3/8] RDMA/hns: Add return statement when kzalloc return NULL in hns_roce_v1_recreate_lp_qp

[Wei Hu (Xavier)]

When lp_qp_work is NULL, it should be returned ENOMEM. This patch
mainly adds the error checking branch, modifies the return value of
the function named hns_roce_v1_set_mac that calling
hns_roce_v1_recreate_lp_qp.

Ihis patch fixes the smatch error as below:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:918 hns_roce_v1_recreate_lp_qp()
error: potential null dereference 'lp_qp_work'.  (kzalloc returns null)

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
Patch V2: Address comments by Leon:
	 https://lkml.org/lkml/2017/9/28/382

Patch V1: Initial Submit
---
 drivers/infiniband/hw/hns/hns_roce_device.h |  2 +-
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c  | 16 +++++++++++++---
 drivers/infiniband/hw/hns/hns_roce_hw_v2.c  |  6 ++++--
 drivers/infiniband/hw/hns/hns_roce_main.c   | 17 +++++++++++------
 4 files changed, 29 insertions(+), 12 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 4f43c91..4d9d5d7 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -581,7 +581,7 @@ struct hns_roce_hw {
 	int (*chk_mbox)(struct hns_roce_dev *hr_dev, unsigned long timeout);
 	void (*set_gid)(struct hns_roce_dev *hr_dev, u8 port, int gid_index,
 			union ib_gid *gid);
-	void (*set_mac)(struct hns_roce_dev *hr_dev, u8 phy_port, u8 *addr);
+	int (*set_mac)(struct hns_roce_dev *hr_dev, u8 phy_port, u8 *addr);
 	void (*set_mtu)(struct hns_roce_dev *hr_dev, u8 phy_port,
 			enum ib_mtu mtu);
 	int (*write_mtpt)(void *mb_buf, struct hns_roce_mr *mr,
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 95f5c88..6f309f7 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -912,6 +912,8 @@ static int hns_roce_v1_recreate_lp_qp(struct hns_roce_dev *hr_dev)
 
 	lp_qp_work = kzalloc(sizeof(struct hns_roce_recreate_lp_qp_work),
 			     GFP_KERNEL);
+	if (!lp_qp_work)
+		return -ENOMEM;
 
 	INIT_WORK(&(lp_qp_work->work), hns_roce_v1_recreate_lp_qp_work_fn);
 
@@ -1719,7 +1721,8 @@ void hns_roce_v1_set_gid(struct hns_roce_dev *hr_dev, u8 port, int gid_index,
 		       (HNS_ROCE_V1_GID_NUM * gid_idx));
 }
 
-void hns_roce_v1_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port, u8 *addr)
+static int hns_roce_v1_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port,
+			       u8 *addr)
 {
 	u32 reg_smac_l;
 	u16 reg_smac_h;
@@ -1732,8 +1735,13 @@ void hns_roce_v1_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port, u8 *addr)
 	 * because of smac not equal to dmac.
 	 * We Need to release and create reserved qp again.
 	 */
-	if (hr_dev->hw->dereg_mr && hns_roce_v1_recreate_lp_qp(hr_dev))
-		dev_warn(&hr_dev->pdev->dev, "recreate lp qp timeout!\n");
+	if (hr_dev->hw->dereg_mr) {
+		int ret;
+
+		ret = hns_roce_v1_recreate_lp_qp(hr_dev);
+		if (ret && ret != -ETIMEDOUT)
+			return ret;
+	}
 
 	p = (u32 *)(&addr[0]);
 	reg_smac_l = *p;
@@ -1748,6 +1756,8 @@ void hns_roce_v1_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port, u8 *addr)
 		       ROCEE_SMAC_H_ROCEE_SMAC_H_S, reg_smac_h);
 	roce_write(hr_dev, ROCEE_SMAC_H_0_REG + phy_port * PHY_PORT_OFFSET,
 		   val);
+
+	return 0;
 }
 
 void hns_roce_v1_set_mtu(struct hns_roce_dev *hr_dev, u8 phy_port,
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index 4171f73..e5fe2cd 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -1068,8 +1068,8 @@ static void hns_roce_v2_set_gid(struct hns_roce_dev *hr_dev, u8 port,
 	roce_write(hr_dev, ROCEE_VF_SGID_CFG4_REG + 0x20 * gid_index, val);
 }
 
-static void hns_roce_v2_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port,
-				u8 *addr)
+static int hns_roce_v2_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port,
+			       u8 *addr)
 {
 	u16 reg_smac_h;
 	u32 reg_smac_l;
@@ -1084,6 +1084,8 @@ static void hns_roce_v2_set_mac(struct hns_roce_dev *hr_dev, u8 phy_port,
 	roce_set_field(val, ROCEE_VF_SMAC_CFG1_VF_SMAC_H_M,
 		       ROCEE_VF_SMAC_CFG1_VF_SMAC_H_S, reg_smac_h);
 	roce_write(hr_dev, ROCEE_VF_SMAC_CFG1_REG + 0x08 * phy_port, val);
+
+	return 0;
 }
 
 static int hns_roce_v2_write_mtpt(void *mb_buf, struct hns_roce_mr *mr,
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 7a0c1e8..6f2d572 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -59,19 +59,19 @@ int hns_get_gid_index(struct hns_roce_dev *hr_dev, u8 port, int gid_index)
 }
 EXPORT_SYMBOL_GPL(hns_get_gid_index);
 
-static void hns_roce_set_mac(struct hns_roce_dev *hr_dev, u8 port, u8 *addr)
+static int hns_roce_set_mac(struct hns_roce_dev *hr_dev, u8 port, u8 *addr)
 {
 	u8 phy_port;
 	u32 i = 0;
 
 	if (!memcmp(hr_dev->dev_addr[port], addr, MAC_ADDR_OCTET_NUM))
-		return;
+		return 0;
 
 	for (i = 0; i < MAC_ADDR_OCTET_NUM; i++)
 		hr_dev->dev_addr[port][i] = addr[i];
 
 	phy_port = hr_dev->iboe.phy_port[port];
-	hr_dev->hw->set_mac(hr_dev, phy_port, addr);
+	return hr_dev->hw->set_mac(hr_dev, phy_port, addr);
 }
 
 static int hns_roce_add_gid(struct ib_device *device, u8 port_num,
@@ -119,6 +119,7 @@ static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
 {
 	struct device *dev = hr_dev->dev;
 	struct net_device *netdev;
+	int ret = 0;
 
 	netdev = hr_dev->iboe.netdevs[port];
 	if (!netdev) {
@@ -131,7 +132,7 @@ static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
 	case NETDEV_CHANGE:
 	case NETDEV_REGISTER:
 	case NETDEV_CHANGEADDR:
-		hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
+		ret = hns_roce_set_mac(hr_dev, port, netdev->dev_addr);
 		break;
 	case NETDEV_DOWN:
 		/*
@@ -143,7 +144,7 @@ static int handle_en_event(struct hns_roce_dev *hr_dev, u8 port,
 		break;
 	}
 
-	return 0;
+	return ret;
 }
 
 static int hns_roce_netdev_event(struct notifier_block *self,
@@ -172,13 +173,17 @@ static int hns_roce_netdev_event(struct notifier_block *self,
 
 static int hns_roce_setup_mtu_mac(struct hns_roce_dev *hr_dev)
 {
+	int ret;
 	u8 i;
 
 	for (i = 0; i < hr_dev->caps.num_ports; i++) {
 		if (hr_dev->hw->set_mtu)
 			hr_dev->hw->set_mtu(hr_dev, hr_dev->iboe.phy_port[i],
 					    hr_dev->caps.max_mtu);
-		hns_roce_set_mac(hr_dev, i, hr_dev->iboe.netdevs[i]->dev_addr);
+		ret = hns_roce_set_mac(hr_dev, i,
+				       hr_dev->iboe.netdevs[i]->dev_addr);
+		if (ret)
+			return ret;
 	}
 
 	return 0;
-- 
1.9.1

Re: [PATCH V2 for-next 3/8] RDMA/hns: Add return statement when kzalloc return NULL in hns_roce_v1_recreate_lp_qp

[Doug Ledford]

On Fri, 2017-09-29 at 23:10 +0800, Wei Hu (Xavier) wrote:
> When lp_qp_work is NULL, it should be returned ENOMEM. This patch
> mainly adds the error checking branch, modifies the return value of
> the function named hns_roce_v1_set_mac that calling
> hns_roce_v1_recreate_lp_qp.
> 
> Ihis patch fixes the smatch error as below:
  ^^^^
Typo, but I can fix it, no need to respin.


-- 
Doug Ledford <dledford@redhat.com>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

[PATCH V2 for-next 4/8] RDMA/hns: Set mask for destination qp field of qp context assignment

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

When only set IB_QP_DEST_QPN flag for attr_mask, the operation of
assigning the dest_qp_num for dest_qp field of qp context is valid.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 6f309f7..10d828c 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -2888,10 +2888,11 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 			       QP_CONTEXT_QPC_BYTES_32_RESPONDER_RESOURCES_S,
 			       ilog2((unsigned int)attr->max_dest_rd_atomic));
 
-		roce_set_field(context->qpc_bytes_36,
-			       QP_CONTEXT_QPC_BYTES_36_DEST_QP_M,
-			       QP_CONTEXT_QPC_BYTES_36_DEST_QP_S,
-			       attr->dest_qp_num);
+		if (attr_mask & IB_QP_DEST_QPN)
+			roce_set_field(context->qpc_bytes_36,
+				       QP_CONTEXT_QPC_BYTES_36_DEST_QP_M,
+				       QP_CONTEXT_QPC_BYTES_36_DEST_QP_S,
+				       attr->dest_qp_num);
 
 		/* Configure GID index */
 		port_num = rdma_ah_get_port_num(&attr->ah_attr);
-- 
1.9.1

Re: [PATCH V2 for-next 4/8] RDMA/hns: Set mask for destination qp field of qp context assignment

[Doug Ledford]

On Fri, 2017-09-29 at 23:10 +0800, Wei Hu (Xavier) wrote:
> From: Lijun Ou <oulijun@huawei.com>
> 
> When only set IB_QP_DEST_QPN flag for attr_mask, the operation of
> assigning the dest_qp_num for dest_qp field of qp context is valid.

This commit message reads poorly.  I reworded it.

-- 
Doug Ledford <dledford@redhat.com>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

[PATCH V2 for-next 5/8] RDMA/hns: Set rdma_ah_attr type for querying qp

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

When querying qp, It needs to return RoCE device ah_attr type
that may be specific to RoCE devices.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 2 ++
 drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 1 +
 2 files changed, 3 insertions(+)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 10d828c..d82b4de 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -3359,6 +3359,7 @@ static int hns_roce_v1_q_sqp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 	qp_attr->path_mtu	= IB_MTU_256;
 	qp_attr->path_mig_state	= IB_MIG_ARMED;
 	qp_attr->qkey		= QKEY_VAL;
+	qp_attr->ah_attr.type   = RDMA_AH_ATTR_TYPE_ROCE;
 	qp_attr->rq_psn		= 0;
 	qp_attr->sq_psn		= 0;
 	qp_attr->dest_qp_num	= 1;
@@ -3440,6 +3441,7 @@ static int hns_roce_v1_q_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 					       QP_CONTEXT_QPC_BYTES_48_MTU_M,
 					       QP_CONTEXT_QPC_BYTES_48_MTU_S);
 	qp_attr->path_mig_state = IB_MIG_ARMED;
+	qp_attr->ah_attr.type   = RDMA_AH_ATTR_TYPE_ROCE;
 	if (hr_qp->ibqp.qp_type == IB_QPT_UD)
 		qp_attr->qkey = QKEY_VAL;
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index e5fe2cd..2def203 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -2831,6 +2831,7 @@ static int hns_roce_v2_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 							V2_QPC_BYTE_24_MTU_M,
 							V2_QPC_BYTE_24_MTU_S);
 	qp_attr->path_mig_state = IB_MIG_ARMED;
+	qp_attr->ah_attr.type   = RDMA_AH_ATTR_TYPE_ROCE;
 	if (hr_qp->ibqp.qp_type == IB_QPT_UD)
 		qp_attr->qkey = V2_QKEY_VAL;
 
-- 
1.9.1

[PATCH V2 for-next 6/8] RDMA/hns: Add return statement when checking error in hns_roce_v1_mr_free_work_fn

[Wei Hu (Xavier)]

After the loop in hns_roce_v1_mr_free_work_fn function, it is possible that
the local variable named hr_qp is NULL, the operation "hr_qp->qpn" will
result in the exception. As a result, we add return statement when checking
error.

This patch fixes the smatch error as below:
drivers/infiniband/hw/hns/hns_roce_hw_v1.c:1009 hns_roce_v1_mr_free_work_fn()
error: we previously assumed 'hr_qp' could be null

Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c | 5 +++++
 1 file changed, 5 insertions(+)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index d82b4de..15e0f2e 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -1004,6 +1004,11 @@ static void hns_roce_v1_mr_free_work_fn(struct work_struct *work)
 		}
 	}
 
+	if (!ne) {
+		dev_err(dev, "Reseved loop qp is absent!\n");
+		goto free_work;
+	}
+
 	do {
 		ret = hns_roce_v1_poll_cq(&mr_free_cq->ib_cq, ne, wc);
 		if (ret < 0) {
-- 
1.9.1

[PATCH V2 for-next 7/8] RDMA/hns: Remove unnecessarily calling unregister_inetaddr_notifier function

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

When the driver doesn't call register_inetaddr_notifier function, it need
not call unregister_inetaddr_notifier to unregister inet addr. This patch
fixes it.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_device.h | 1 -
 drivers/infiniband/hw/hns/hns_roce_main.c   | 1 -
 2 files changed, 2 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 4d9d5d7..b314ac0 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -474,7 +474,6 @@ struct hns_roce_ib_iboe {
 	spinlock_t		lock;
 	struct net_device      *netdevs[HNS_ROCE_MAX_PORTS];
 	struct notifier_block	nb;
-	struct notifier_block	nb_inet;
 	u8			phy_port[HNS_ROCE_MAX_PORTS];
 };
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index 6f2d572..3dcb2df 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -424,7 +424,6 @@ static void hns_roce_unregister_device(struct hns_roce_dev *hr_dev)
 {
 	struct hns_roce_ib_iboe *iboe = &hr_dev->iboe;
 
-	unregister_inetaddr_notifier(&iboe->nb_inet);
 	unregister_netdevice_notifier(&iboe->nb);
 	ib_unregister_device(&hr_dev->ib_dev);
 }
-- 
1.9.1

[PATCH V2 for-next 8/8] RDMA/hns: Replace usleep_range with udelay when checking command status

[Wei Hu (Xavier)]

From: Lijun Ou <oulijun@huawei.com>

It replaces usleep_range with udelay to avoid using usleep_range function
in spin_lock_bh spin region, because it probably cause calltrace.

BUG: scheduling while atomic: insmod/1428/0x00000002
Modules linked in: hns-roce-hw-v2(+) hns_roce rdma_ucm rdma_cm iw_cm ib_uverbs ib_cm ib_core
CPU: 0 PID: 1428 Comm: insmod Not tainted 4.12.0-rc1-00677-g252e8fd-dirty #43
Hardware name: (null) (DT)
Call trace:
[<ffff000008089d20>] dump_backtrace+0x0/0x274
[<ffff00000808a068>] show_stack+0x20/0x28
[<ffff00000844ea58>] dump_stack+0x94/0xb4
[<ffff0000080f975c>] __schedule_bug+0x68/0x84
[<ffff000008a988d4>] __schedule+0x5fc/0x70c
[<ffff000008a98a24>] schedule+0x40/0xa4
[<ffff000008a9c6f0>] schedule_hrtimeout_range_clock+0x98/0xfc
[<ffff000008a9c788>] schedule_hrtimeout_range+0x34/0x40
[<ffff000008a9c098>] usleep_range+0x6c/0x80
[<ffff000000b9ae68>] hns_roce_cmd_send+0xe4/0x264 [hns-roce-hw-v2]
[<ffff000000b9b748>] hns_roce_cmd_query_hw_info+0x40/0x60 [hns-roce-hw-v2]
[<ffff000000b9b790>] hns_roce_v2_profile+0x28/0x668 [hns-roce-hw-v2]
[<ffff000000b6b1f4>] hns_roce_init+0x6c/0x948 [hns-roce-hw-v2]

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Signed-off-by: Wei Hu (Xavier) <xavier.huwei@huawei.com>
Signed-off-by: Shaobo Xu <xushaobo2@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v2.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
index 2def203..7be53021 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v2.c
@@ -589,7 +589,7 @@ int hns_roce_cmq_send(struct hns_roce_dev *hr_dev,
 		do {
 			if (hns_roce_cmq_csq_done(hr_dev))
 				break;
-			usleep_range(1000, 2000);
+			udelay(1);
 			timeout++;
 		} while (timeout < priv->cmq.tx_timeout);
 	}
-- 
1.9.1

Re: [PATCH V2 for-next 0/8] Bug fixes & Code improvements in hip06 and hip08 RoCE driver

[Doug Ledford]

On Fri, 2017-09-29 at 23:10 +0800, Wei Hu (Xavier) wrote:
> This patch-set introduces some bug fixes and code improvements
> for hip06 and hip08 RoCE driver. It includes a patch for fixing
> the assign algorithm of qp_attr->max_rd_atomic and
> qp_attr->max_dest_rd_atomic, three patches for static check errors,
> one for setting attr mask, one for returning RoCE device ah_attr
> type when querying qp, one for unregistering inet addr, and the
> last one for command queue delay processing in hip08 driver.
> 
> Lijun Ou (6):
>   RDMA/hns: Modify the value with rd&dest_rd of qp_attr
>   RDMA/hns: Factor out the code for checking sdb status into a new
>     function
>   RDMA/hns: Set mask for destination qp field of qp context
> assignment
>   RDMA/hns: Set rdma_ah_attr type for querying qp
>   RDMA/hns: Remove unnecessarily calling unregister_inetaddr_notifier
>     function
>   RDMA/hns: Replace usleep_range with udelay when checking command
>     status
> 
> Wei Hu (Xavier) (2):
>   RDMA/hns: Add return statement when kzalloc return NULL in
>     hns_roce_v1_recreate_lp_qp
>   RDMA/hns: Add return statement when checking error in
>     hns_roce_v1_mr_free_work_fn
> 
>  drivers/infiniband/hw/hns/hns_roce_device.h |   3 +-
>  drivers/infiniband/hw/hns/hns_roce_hw_v1.c  | 131 +++++++++++++++++-
> ----------
>  drivers/infiniband/hw/hns/hns_roce_hw_v2.c  |   9 +-
>  drivers/infiniband/hw/hns/hns_roce_main.c   |  18 ++--
>  4 files changed, 96 insertions(+), 65 deletions(-)

Hi Wei,

I've taken this series into for-next.  However, a lot of the commit
subjects needed fixing up (and a few of the commit logs as well for
general readability).  Please don't use such long subjects as they
interrupt the output of git log --oneline when using the kernel
standard 12 digit hex output for commit hashes.

-- 
Doug Ledford <dledford@redhat.com>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD