Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]


Groups > linux.kernel > #1474647 > unrolled thread

[PATCH for-next 00/10] Add CM(Connection Manager) Support to HNS RoCe Driver

Started bySalil Mehta <salil.mehta@huawei.com>
First post2016-09-01 23:30 +0200
Last post2016-09-01 23:30 +0200
Articles 9 — 1 participant

Back to article view | Back to linux.kernel


Contents

  [PATCH for-next 00/10] Add CM(Connection Manager) Support to HNS RoCe Driver Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 10/10] IB/hns: Add node_guid definition to the bindings document Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 01/10] IB/hns: Register HNS RoCE Driver get_netdev() with IB Core Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 07/10] IB/hns: Change the logic for allocating uar registers Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 08/10] IB/hns: Fix the bug of rdma cm connecting on user mode Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 09/10] IB/hns: Fix two bugs for rdma cm connecting Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 04/10] IB/hns: Fix the value of device_cap_flags Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 05/10] IB/hns: Fix two possible bugs for rdma cm Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200
    [PATCH for-next 06/10] IB/hns: Add phy_port for computing GSI/QPN Salil Mehta <salil.mehta@huawei.com> - 2016-09-01 23:30 +0200

#1474647 — [PATCH for-next 00/10] Add CM(Connection Manager) Support to HNS RoCe Driver

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 00/10] Add CM(Connection Manager) Support to HNS RoCe Driver
Message-ID<scHIB-79t-15@gated-at.bofh.it>
This patch-set adds the CM(Connection Manager) support to the
HNS RoCE driver. Changes done are primarily to add support of
APIs in IB device and some fixes over the base driver to
support RDMA Connection Manager. This patch-set also updates
the Device binding document as new parameter node-guid was
added to the DT.

All of the patches have been authored by Oulijun and reviewed
by Huwei.

Lijun Ou (10):
  IB/hns: Register HNS RoCE Driver get_netdev() with IB Core
  IB/hns: Register add_gid and del_gid for GID Table management
  IB/hns: Add & initialize "node_guid" parameter for RDMA CM
  IB/hns: Fix the value of device_cap_flags
  IB/hns: Fix two possible bugs for rdma cm
  IB/hns: Add phy_port for computing GSI/QPN
  IB/hns: Change the logic for allocating uar registers
  IB/hns: Fix the bug of rdma cm connecting on user mode
  IB/hns: Fix two bugs for rdma cm connecting
  IB/hns: Add node_guid definition to the bindings document

 .../bindings/infiniband/hisilicon-hns-roce.txt     |    2 +
 drivers/infiniband/hw/hns/hns_roce_device.h        |    4 +-
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c         |   80 ++++++++------------
 drivers/infiniband/hw/hns/hns_roce_main.c          |   45 ++++++++++-
 drivers/infiniband/hw/hns/hns_roce_pd.c            |    4 +-
 drivers/infiniband/hw/hns/hns_roce_qp.c            |   13 ++--
 6 files changed, 89 insertions(+), 59 deletions(-)

-- 
1.7.9.5

[toc] | [next] | [standalone]


#1474649 — [PATCH for-next 10/10] IB/hns: Add node_guid definition to the bindings document

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 10/10] IB/hns: Add node_guid definition to the bindings document
Message-ID<scHIC-79t-27@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

This patch adds node_guid definition in bindings document.
The value of node_guid will be used during RDMA connection.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 .../bindings/infiniband/hisilicon-hns-roce.txt     |    2 ++
 1 file changed, 2 insertions(+)

diff --git a/Documentation/devicetree/bindings/infiniband/hisilicon-hns-roce.txt b/Documentation/devicetree/bindings/infiniband/hisilicon-hns-roce.txt
index f97993b..d3b273e 100644
--- a/Documentation/devicetree/bindings/infiniband/hisilicon-hns-roce.txt
+++ b/Documentation/devicetree/bindings/infiniband/hisilicon-hns-roce.txt
@@ -14,6 +14,7 @@ length of memory mapped region.
 representing a ethernet device.
 - dsaf-handle: phandle, specifies a reference to a node
 representing a dsaf device.
+- node_guid: a number that uniquely identifies a device or component
 - #address-cells: must be 2
 - #size-cells: must be 2
 Optional properties:
@@ -32,6 +33,7 @@ Example:
 			dma-coherent;
 			eth-handle = <&eth2 &eth3 &eth4 &eth5 &eth6 &eth7>;
 			dsaf-handle = <&soc0_dsa>;
+			node-guid = [00 9A CD 00 00 01 02 03];
 			#address-cells = <2>;
 			#size-cells = <2>;
 			interrupt-parent = <&mbigen_dsa>;
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474650 — [PATCH for-next 01/10] IB/hns: Register HNS RoCE Driver get_netdev() with IB Core

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 01/10] IB/hns: Register HNS RoCE Driver get_netdev() with IB Core
Message-ID<scHIC-79t-33@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

This patch adds get_netdev() function to the IB device. This shall be
used to fetch netdev corresponding to the port number. This function
would be called by IB core(Generic CM Agent) for example, when the
RDMA connection is being established.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_main.c |   20 ++++++++++++++++++++
 1 file changed, 20 insertions(+)

diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index f64f0dd..39e69c3 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -372,6 +372,25 @@ static int hns_roce_query_device(struct ib_device *ib_dev,
 	return 0;
 }
 
+static struct net_device *hns_roce_get_netdev(struct ib_device *ib_dev,
+					      u8 port_num)
+{
+	struct hns_roce_dev *hr_dev = to_hr_dev(ib_dev);
+	struct net_device *ndev;
+
+	if (port_num < 1 || port_num > hr_dev->caps.num_ports)
+		return NULL;
+
+	rcu_read_lock();
+
+	ndev = hr_dev->iboe.netdevs[port_num - 1];
+	if (ndev)
+		dev_hold(ndev);
+
+	rcu_read_unlock();
+	return ndev;
+}
+
 static int hns_roce_query_port(struct ib_device *ib_dev, u8 port_num,
 			       struct ib_port_attr *props)
 {
@@ -618,6 +637,7 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 	ib_dev->query_port		= hns_roce_query_port;
 	ib_dev->modify_port		= hns_roce_modify_port;
 	ib_dev->get_link_layer		= hns_roce_get_link_layer;
+	ib_dev->get_netdev		= hns_roce_get_netdev;
 	ib_dev->query_gid		= hns_roce_query_gid;
 	ib_dev->query_pkey		= hns_roce_query_pkey;
 	ib_dev->alloc_ucontext		= hns_roce_alloc_ucontext;
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474653 — [PATCH for-next 07/10] IB/hns: Change the logic for allocating uar registers

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 07/10] IB/hns: Change the logic for allocating uar registers
Message-ID<scHIC-79t-29@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

This patch mainly modifies the logic for allocating uar registers.
In HiP06 SoC, HW has 8 group of uar registers for kernel and
user space application. The uar index is assigned as follows:
    0   ------ for kernel
    1~7 ------ for user space application

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_pd.c |    4 +++-
 1 file changed, 3 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_pd.c b/drivers/infiniband/hw/hns/hns_roce_pd.c
index 16271b5..4109f74 100644
--- a/drivers/infiniband/hw/hns/hns_roce_pd.c
+++ b/drivers/infiniband/hw/hns/hns_roce_pd.c
@@ -117,7 +117,9 @@ int hns_roce_uar_alloc(struct hns_roce_dev *hr_dev, struct hns_roce_uar *uar)
 	if (ret == -1)
 		return -ENOMEM;
 
-	uar->index = (uar->index - 1) % hr_dev->caps.phy_num_uars + 1;
+	if (uar->index > 0)
+		uar->index = (uar->index - 1) %
+			     (hr_dev->caps.phy_num_uars - 1) + 1;
 
 	res = platform_get_resource(hr_dev->pdev, IORESOURCE_MEM, 0);
 	uar->pfn = ((res->start) >> PAGE_SHIFT) + uar->index;
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474656 — [PATCH for-next 08/10] IB/hns: Fix the bug of rdma cm connecting on user mode

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 08/10] IB/hns: Fix the bug of rdma cm connecting on user mode
Message-ID<scHIC-79t-41@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

Fix bug of modify qp from init to init on user mode. Otherwise,
it will oops when rmda cm established.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c |   41 ++++++++++------------------
 1 file changed, 14 insertions(+), 27 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 581e542..206957b 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -2395,35 +2395,22 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 	if (cur_state == IB_QPS_INIT && new_state == IB_QPS_INIT) {
 		/* Memory barrier */
 		wmb();
-		if (hr_qp->ibqp.qp_type == IB_QPT_GSI) {
-			/* SW update GSI rq header */
-			reg_val = roce_read(hr_dev, ROCEE_QP1C_CFG3_0_REG +
-					    QP1C_CFGN_OFFSET * hr_qp->phy_port);
-			roce_set_field(reg_val,
-				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_M,
-				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_S,
-				       hr_qp->rq.head);
-			roce_write(hr_dev, ROCEE_QP1C_CFG3_0_REG +
-				   QP1C_CFGN_OFFSET * hr_qp->phy_port, reg_val);
-		} else {
-			rq_db.u32_4 = 0;
-			rq_db.u32_8 = 0;
 
-			roce_set_field(rq_db.u32_4, RQ_DOORBELL_U32_4_RQ_HEAD_M,
-				       RQ_DOORBELL_U32_4_RQ_HEAD_S,
-				       hr_qp->rq.head);
-			roce_set_field(rq_db.u32_8, RQ_DOORBELL_U32_8_QPN_M,
-				       RQ_DOORBELL_U32_8_QPN_S, hr_qp->qpn);
-			roce_set_field(rq_db.u32_8, RQ_DOORBELL_U32_8_CMD_M,
-				       RQ_DOORBELL_U32_8_CMD_S, 1);
-			roce_set_bit(rq_db.u32_8, RQ_DOORBELL_U32_8_HW_SYNC_S,
-				     1);
-
-			doorbell[0] = rq_db.u32_4;
-			doorbell[1] = rq_db.u32_8;
-
-			hns_roce_write64_k(doorbell, hr_qp->rq.db_reg_l);
+		roce_set_field(doorbell[0], RQ_DOORBELL_U32_4_RQ_HEAD_M,
+			       RQ_DOORBELL_U32_4_RQ_HEAD_S, hr_qp->rq.head);
+		roce_set_field(doorbell[1], RQ_DOORBELL_U32_8_QPN_M,
+			       RQ_DOORBELL_U32_8_QPN_S, hr_qp->qpn);
+		roce_set_field(doorbell[1], RQ_DOORBELL_U32_8_CMD_M,
+			       RQ_DOORBELL_U32_8_CMD_S, 1);
+		roce_set_bit(doorbell[1], RQ_DOORBELL_U32_8_HW_SYNC_S, 1);
+
+		if (ibqp->uobject) {
+			hr_qp->rq.db_reg_l = hr_dev->reg_base +
+				     ROCEE_DB_OTHERS_L_0_REG +
+				     DB_REG_OFFSET * hr_dev->priv_uar.index;
 		}
+
+		hns_roce_write64_k(doorbell, hr_qp->rq.db_reg_l);
 	}
 
 	hr_qp->state = new_state;
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474658 — [PATCH for-next 09/10] IB/hns: Fix two bugs for rdma cm connecting

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 09/10] IB/hns: Fix two bugs for rdma cm connecting
Message-ID<scHIC-79t-47@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

This patch mainly modify the value of HNS_ROCE_SL_SHIFT
and delete the lines for assigning for the field of
local_enable_e2e_credit in QP1C.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_device.h |    2 +-
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c  |    3 ---
 2 files changed, 1 insertion(+), 4 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index 833742b..066bb3e 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -62,7 +62,7 @@
 #define HNS_ROCE_AEQE_OF_VEC_NUM		1
 
 /* 4G/4K = 1M */
-#define HNS_ROCE_SL_SHIFT			29
+#define HNS_ROCE_SL_SHIFT			28
 #define HNS_ROCE_TCLASS_SHIFT			20
 #define HNS_ROCE_FLOW_LABLE_MASK		0xfffff
 
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 206957b..3feac8a 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -1734,9 +1734,6 @@ static int hns_roce_v1_m_sqp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		roce_set_bit(context->qp1c_bytes_16,
 			     QP1C_BYTES_16_SIGNALING_TYPE_S,
 			     hr_qp->sq_signal_bits);
-		roce_set_bit(context->qp1c_bytes_16,
-			     QP1C_BYTES_16_LOCAL_ENABLE_E2E_CREDIT_S,
-			     hr_qp->sq_signal_bits);
 		roce_set_bit(context->qp1c_bytes_16, QP1C_BYTES_16_RQ_BA_FLG_S,
 			     1);
 		roce_set_bit(context->qp1c_bytes_16, QP1C_BYTES_16_SQ_BA_FLG_S,
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474660 — [PATCH for-next 04/10] IB/hns: Fix the value of device_cap_flags

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 04/10] IB/hns: Fix the value of device_cap_flags
Message-ID<scHID-79t-55@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

In the latest IB core version, it has some known issues
with memory registration using the local_dma_lkey.
Thus RoCE don't expose support for it, and remove
device->local_dma_lkey which is introduced to working systems.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_main.c |    3 +--
 1 file changed, 1 insertion(+), 2 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_main.c b/drivers/infiniband/hw/hns/hns_roce_main.c
index ec27f0c..b54074d 100644
--- a/drivers/infiniband/hw/hns/hns_roce_main.c
+++ b/drivers/infiniband/hw/hns/hns_roce_main.c
@@ -368,8 +368,7 @@ static int hns_roce_query_device(struct ib_device *ib_dev,
 	props->max_qp = hr_dev->caps.num_qps;
 	props->max_qp_wr = hr_dev->caps.max_wqes;
 	props->device_cap_flags = IB_DEVICE_PORT_ACTIVE_EVENT |
-				  IB_DEVICE_RC_RNR_NAK_GEN |
-				  IB_DEVICE_LOCAL_DMA_LKEY;
+				  IB_DEVICE_RC_RNR_NAK_GEN;
 	props->max_sge = hr_dev->caps.max_sq_sg;
 	props->max_sge_rd = 1;
 	props->max_cq = hr_dev->caps.num_cqs;
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474661 — [PATCH for-next 05/10] IB/hns: Fix two possible bugs for rdma cm

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 05/10] IB/hns: Fix two possible bugs for rdma cm
Message-ID<scHID-79t-53@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

Fix the length of wqe that maybe lead to an error and
write the end bytes of QP1C into the register.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c |   10 ++++------
 1 file changed, 4 insertions(+), 6 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index 399f5de..aaea95c 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -205,8 +205,7 @@ int hns_roce_v1_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 				      (wr->send_flags & IB_SEND_FENCE ?
 				      (cpu_to_le32(HNS_ROCE_WQE_FENCE)) : 0);
 
-			wqe = (struct hns_roce_wqe_ctrl_seg *)wqe +
-			       sizeof(struct hns_roce_wqe_ctrl_seg);
+			wqe += sizeof(struct hns_roce_wqe_ctrl_seg);
 
 			switch (wr->opcode) {
 			case IB_WR_RDMA_READ:
@@ -235,8 +234,7 @@ int hns_roce_v1_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 				break;
 			}
 			ctrl->flag |= cpu_to_le32(ps_opcode);
-			wqe = (struct hns_roce_wqe_raddr_seg *)wqe +
-			       sizeof(struct hns_roce_wqe_raddr_seg);
+			wqe += sizeof(struct hns_roce_wqe_raddr_seg);
 
 			dseg = wqe;
 			if (wr->send_flags & IB_SEND_INLINE && wr->num_sge) {
@@ -253,8 +251,7 @@ int hns_roce_v1_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 					memcpy(wqe, ((void *) (uintptr_t)
 					       wr->sg_list[i].addr),
 					       wr->sg_list[i].length);
-					wqe = (struct hns_roce_wqe_raddr_seg *)
-					       wqe + wr->sg_list[i].length;
+					wqe += wr->sg_list[i].length;
 				}
 				ctrl->flag |= HNS_ROCE_WQE_INLINE;
 			} else {
@@ -1795,6 +1792,7 @@ static int hns_roce_v1_m_sqp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		writel(context->qp1c_bytes_28, addr + 6);
 		writel(context->qp1c_bytes_32, addr + 7);
 		writel(context->cur_sq_wqe_ba_l, addr + 8);
+		writel(context->qp1c_bytes_40, addr + 9);
 	}
 
 	/* Modify QP1C status */
-- 
1.7.9.5

[toc] | [prev] | [next] | [standalone]


#1474664 — [PATCH for-next 06/10] IB/hns: Add phy_port for computing GSI/QPN

FromSalil Mehta <salil.mehta@huawei.com>
Date2016-09-01 23:30 +0200
Subject[PATCH for-next 06/10] IB/hns: Add phy_port for computing GSI/QPN
Message-ID<scHID-79t-65@gated-at.bofh.it>
In reply to#1474647
From: Lijun Ou <oulijun@huawei.com>

This patch mainly adds phy_port to HNS RoCE QP. This shall be
used in calculating the GSI QPN for the port.
Initally when RDMA is being established, all IB ports share a
QPN which later needs to be re-assigned to a particular GSI/QPN
and which is per-port.
This also fixes a bug in base driver where iboe port was being
used instead of phy_port at some places. This values might not
be same always.

Signed-off-by: Lijun Ou <oulijun@huawei.com>
Reviewed-by: Wei Hu <xavier.huwei@huawei.com>
Signed-off-by: Salil Mehta <salil.mehta@huawei.com>
---
 drivers/infiniband/hw/hns/hns_roce_device.h |    1 +
 drivers/infiniband/hw/hns/hns_roce_hw_v1.c  |   30 ++++++++++++++-------------
 drivers/infiniband/hw/hns/hns_roce_qp.c     |   13 ++++++------
 3 files changed, 23 insertions(+), 21 deletions(-)

diff --git a/drivers/infiniband/hw/hns/hns_roce_device.h b/drivers/infiniband/hw/hns/hns_roce_device.h
index e943b98..833742b 100644
--- a/drivers/infiniband/hw/hns/hns_roce_device.h
+++ b/drivers/infiniband/hw/hns/hns_roce_device.h
@@ -409,6 +409,7 @@ struct hns_roce_qp {
 	u32			buff_size;
 	struct mutex		mutex;
 	u8			port;
+	u8			phy_port;
 	u8			sl;
 	u8			resp_depth;
 	u8			state;
diff --git a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
index aaea95c..581e542 100644
--- a/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
+++ b/drivers/infiniband/hw/hns/hns_roce_hw_v1.c
@@ -162,7 +162,7 @@ int hns_roce_v1_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 			roce_set_field(ud_sq_wqe->u32_36,
 				       UD_SEND_WQE_U32_36_SGID_INDEX_M,
 				       UD_SEND_WQE_U32_36_SGID_INDEX_S,
-				       hns_get_gid_index(hr_dev, qp->port,
+				       hns_get_gid_index(hr_dev, qp->phy_port,
 							 ah->av.gid_index));
 
 			roce_set_field(ud_sq_wqe->u32_40,
@@ -282,7 +282,7 @@ out:
 			       SQ_DOORBELL_U32_4_SQ_HEAD_S,
 			      (qp->sq.head & ((qp->sq.wqe_cnt << 1) - 1)));
 		roce_set_field(sq_db.u32_4, SQ_DOORBELL_U32_4_PORT_M,
-			       SQ_DOORBELL_U32_4_PORT_S, qp->port);
+			       SQ_DOORBELL_U32_4_PORT_S, qp->phy_port);
 		roce_set_field(sq_db.u32_8, SQ_DOORBELL_U32_8_QPN_M,
 			       SQ_DOORBELL_U32_8_QPN_S, qp->doorbell_qpn);
 		roce_set_bit(sq_db.u32_8, SQ_DOORBELL_HW_SYNC_S, 1);
@@ -362,14 +362,14 @@ out:
 			/* SW update GSI rq header */
 			reg_val = roce_read(to_hr_dev(ibqp->device),
 					    ROCEE_QP1C_CFG3_0_REG +
-					    QP1C_CFGN_OFFSET * hr_qp->port);
+					    QP1C_CFGN_OFFSET * hr_qp->phy_port);
 			roce_set_field(reg_val,
 				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_M,
 				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_S,
 				       hr_qp->rq.head);
 			roce_write(to_hr_dev(ibqp->device),
 				   ROCEE_QP1C_CFG3_0_REG +
-				   QP1C_CFGN_OFFSET * hr_qp->port, reg_val);
+				   QP1C_CFGN_OFFSET * hr_qp->phy_port, reg_val);
 		} else {
 			rq_db.u32_4 = 0;
 			rq_db.u32_8 = 0;
@@ -1730,7 +1730,7 @@ static int hns_roce_v1_m_sqp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		roce_set_field(context->qp1c_bytes_16, QP1C_BYTES_16_RQ_HEAD_M,
 			       QP1C_BYTES_16_RQ_HEAD_S, hr_qp->rq.head);
 		roce_set_field(context->qp1c_bytes_16, QP1C_BYTES_16_PORT_NUM_M,
-			       QP1C_BYTES_16_PORT_NUM_S, hr_qp->port);
+			       QP1C_BYTES_16_PORT_NUM_S, hr_qp->phy_port);
 		roce_set_bit(context->qp1c_bytes_16,
 			     QP1C_BYTES_16_SIGNALING_TYPE_S,
 			     hr_qp->sq_signal_bits);
@@ -1781,7 +1781,7 @@ static int hns_roce_v1_m_sqp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 
 		/* Copy context to QP1C register */
 		addr = (u32 *)(hr_dev->reg_base + ROCEE_QP1C_CFG0_0_REG +
-			hr_qp->port * sizeof(*context));
+			hr_qp->phy_port * sizeof(*context));
 
 		writel(context->qp1c_bytes_4, addr);
 		writel(context->sq_rq_bt_l, addr + 1);
@@ -1797,11 +1797,11 @@ static int hns_roce_v1_m_sqp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 
 	/* Modify QP1C status */
 	reg_val = roce_read(hr_dev, ROCEE_QP1C_CFG0_0_REG +
-			    hr_qp->port * sizeof(*context));
+			    hr_qp->phy_port * sizeof(*context));
 	roce_set_field(reg_val, ROCEE_QP1C_CFG0_0_ROCEE_QP1C_QP_ST_M,
 		       ROCEE_QP1C_CFG0_0_ROCEE_QP1C_QP_ST_S, new_state);
 	roce_write(hr_dev, ROCEE_QP1C_CFG0_0_REG +
-		    hr_qp->port * sizeof(*context), reg_val);
+		    hr_qp->phy_port * sizeof(*context), reg_val);
 
 	hr_qp->state = new_state;
 	if (new_state == IB_QPS_RESET) {
@@ -2184,7 +2184,7 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		roce_set_field(context->qpc_bytes_156,
 			       QP_CONTEXT_QPC_BYTES_156_PORT_NUM_M,
 			       QP_CONTEXT_QPC_BYTES_156_PORT_NUM_S,
-			       hr_qp->port);
+			       hr_qp->phy_port);
 		roce_set_field(context->qpc_bytes_156,
 			       QP_CONTEXT_QPC_BYTES_156_SL_M,
 			       QP_CONTEXT_QPC_BYTES_156_SL_S, attr->ah_attr.sl);
@@ -2290,7 +2290,7 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		roce_set_field(context->qpc_bytes_156,
 			       QP_CONTEXT_QPC_BYTES_156_PORT_NUM_M,
 			       QP_CONTEXT_QPC_BYTES_156_PORT_NUM_S,
-			       hr_qp->port);
+			       hr_qp->phy_port);
 		roce_set_field(context->qpc_bytes_156,
 			       QP_CONTEXT_QPC_BYTES_156_SL_M,
 			       QP_CONTEXT_QPC_BYTES_156_SL_S, attr->ah_attr.sl);
@@ -2398,13 +2398,13 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 		if (hr_qp->ibqp.qp_type == IB_QPT_GSI) {
 			/* SW update GSI rq header */
 			reg_val = roce_read(hr_dev, ROCEE_QP1C_CFG3_0_REG +
-					    QP1C_CFGN_OFFSET * hr_qp->port);
+					    QP1C_CFGN_OFFSET * hr_qp->phy_port);
 			roce_set_field(reg_val,
 				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_M,
 				       ROCEE_QP1C_CFG3_0_ROCEE_QP1C_RQ_HEAD_S,
 				       hr_qp->rq.head);
 			roce_write(hr_dev, ROCEE_QP1C_CFG3_0_REG +
-				    QP1C_CFGN_OFFSET * hr_qp->port, reg_val);
+				   QP1C_CFGN_OFFSET * hr_qp->phy_port, reg_val);
 		} else {
 			rq_db.u32_4 = 0;
 			rq_db.u32_8 = 0;
@@ -2430,8 +2430,10 @@ static int hns_roce_v1_m_qp(struct ib_qp *ibqp, const struct ib_qp_attr *attr,
 
 	if (attr_mask & IB_QP_MAX_DEST_RD_ATOMIC)
 		hr_qp->resp_depth = attr->max_dest_rd_atomic;
-	if (attr_mask & IB_QP_PORT)
-		hr_qp->port = (attr->port_num - 1);
+	if (attr_mask & IB_QP_PORT) {
+		hr_qp->port = attr->port_num - 1;
+		hr_qp->phy_port = hr_dev->iboe.phy_port[hr_qp->port];
+	}
 
 	if (new_state == IB_QPS_RESET && !ibqp->uobject) {
 		hns_roce_v1_cq_clean(to_hr_cq(ibqp->recv_cq), hr_qp->qpn,
diff --git a/drivers/infiniband/hw/hns/hns_roce_qp.c b/drivers/infiniband/hw/hns/hns_roce_qp.c
index 645c18d..089da7f 100644
--- a/drivers/infiniband/hw/hns/hns_roce_qp.c
+++ b/drivers/infiniband/hw/hns/hns_roce_qp.c
@@ -617,21 +617,20 @@ struct ib_qp *hns_roce_create_qp(struct ib_pd *pd,
 			return ERR_PTR(-ENOMEM);
 
 		hr_qp = &hr_sqp->hr_qp;
+		hr_qp->port = init_attr->port_num - 1;
+		hr_qp->phy_port = hr_dev->iboe.phy_port[hr_qp->port];
+		hr_qp->ibqp.qp_num = hr_dev->caps.sqp_start +
+				     HNS_ROCE_MAX_PORTS +
+				     hr_dev->iboe.phy_port[hr_qp->port];
 
 		ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata,
-						hr_dev->caps.sqp_start +
-						hr_dev->caps.num_ports +
-						init_attr->port_num - 1, hr_qp);
+						hr_qp->ibqp.qp_num, hr_qp);
 		if (ret) {
 			dev_err(dev, "Create GSI QP failed!\n");
 			kfree(hr_sqp);
 			return ERR_PTR(ret);
 		}
 
-		hr_qp->port = (init_attr->port_num - 1);
-		hr_qp->ibqp.qp_num = hr_dev->caps.sqp_start +
-				     hr_dev->caps.num_ports +
-				     init_attr->port_num - 1;
 		break;
 	}
 	default:{
-- 
1.7.9.5

[toc] | [prev] | [standalone]


Back to top | Article view | linux.kernel


csiph-web