[PATCH for-next 0/8] RDMA/qedr: Add iWARP support for QL4xxxx

[PATCH for-next 0/8] RDMA/qedr: Add iWARP support for QL4xxxx

[Michal Kalderon]

Hi Doug,

This patch series adds iWARP support to our QL4xxxx networking adapters.

The first 6 patches of this series were previously sent as an RFC
(patches 13-18): https://www.spinics.net/lists/linux-rdma/msg51416.html
The qed part of the RFC has already been accepted as part of 4.13:
https://www.spinics.net/lists/netdev/msg443224.html

Changes from RFC:
-----------------
* Functions rdma_protocol_iwarp/roce, are used instead of IS_IWARP/IS_ROCE
macros when differentiating flows after device registration.
Pointed out by Leon
* Additional patch is added for user space iWARP support.
* Additional maintainer added for qedr for MAINTAINERS.

This series was tested and built against:
git://github.com/dledford/linux.git, branch: k.o/for-next

Michal Kalderon (8):
  RDMA/qedr: Add additional maintainer to MAINTAINERS file
  RDMA/qedr: Rename the qedr_cm file as a preparation for iWARP support
  RDMA/qedr: Add support for registering an iWARP device
  RDMA/qedr: Add iWARP support in existing verbs
  RDMA/qedr: Add support for read with invalidate, supported in iWARP
  RDMA/qedr: Add iWARP connection management qp related callbacks
  RDMA/qedr: Add iWARP connection management functions
  RDMA/qedr: Add support for iWARP in user space

 MAINTAINERS                                        |   1 +
 drivers/infiniband/hw/qedr/Makefile                |   2 +-
 drivers/infiniband/hw/qedr/main.c                  | 105 ++-
 drivers/infiniband/hw/qedr/qedr.h                  |  31 +-
 drivers/infiniband/hw/qedr/qedr_hsi_rdma.h         |   6 +-
 drivers/infiniband/hw/qedr/qedr_iw_cm.c            | 746 +++++++++++++++++++++
 drivers/infiniband/hw/qedr/qedr_iw_cm.h            |  49 ++
 .../hw/qedr/{qedr_cm.c => qedr_roce_cm.c}          |   2 +-
 .../hw/qedr/{qedr_cm.h => qedr_roce_cm.h}          |   0
 drivers/infiniband/hw/qedr/verbs.c                 | 350 +++++++---
 drivers/infiniband/hw/qedr/verbs.h                 |   2 +
 11 files changed, 1203 insertions(+), 91 deletions(-)
 create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.c
 create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.h
 rename drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c} (99%)
 rename drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h} (100%)

-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 1/8] RDMA/qedr: Add additional maintainer to MAINTAINERS file

[Michal Kalderon]

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 MAINTAINERS | 1 +
 1 file changed, 1 insertion(+)

diff --git a/MAINTAINERS b/MAINTAINERS
index f66488d..9d9e284 100644
--- a/MAINTAINERS
+++ b/MAINTAINERS
@@ -10830,6 +10830,7 @@ F:	drivers/net/ethernet/qlogic/qede/
 
 QLOGIC QL4xxx RDMA DRIVER
 M:	Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
+M:	Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
 M:	Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
 L:	linux-rdma-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
 S:	Supported
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 2/8] RDMA/qedr: Rename the qedr_cm file as a preparation for iWARP support

[Michal Kalderon]

The main differences between iWARP and RoCE lay in the communication
 management functions. These will be placed in separate files.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/Makefile                      | 2 +-
 drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c} | 2 +-
 drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h} | 0
 drivers/infiniband/hw/qedr/verbs.c                       | 2 +-
 4 files changed, 3 insertions(+), 3 deletions(-)
 rename drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c} (99%)
 rename drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h} (100%)

diff --git a/drivers/infiniband/hw/qedr/Makefile b/drivers/infiniband/hw/qedr/Makefile
index ba7067c..331a361 100644
--- a/drivers/infiniband/hw/qedr/Makefile
+++ b/drivers/infiniband/hw/qedr/Makefile
@@ -1,3 +1,3 @@
 obj-$(CONFIG_INFINIBAND_QEDR) := qedr.o
 
-qedr-y := main.o verbs.o qedr_cm.o
+qedr-y := main.o verbs.o qedr_roce_cm.o
diff --git a/drivers/infiniband/hw/qedr/qedr_cm.c b/drivers/infiniband/hw/qedr/qedr_roce_cm.c
similarity index 99%
rename from drivers/infiniband/hw/qedr/qedr_cm.c
rename to drivers/infiniband/hw/qedr/qedr_roce_cm.c
index 4689e80..c3c249b 100644
--- a/drivers/infiniband/hw/qedr/qedr_cm.c
+++ b/drivers/infiniband/hw/qedr/qedr_roce_cm.c
@@ -48,7 +48,7 @@
 #include "qedr.h"
 #include "verbs.h"
 #include <rdma/qedr-abi.h>
-#include "qedr_cm.h"
+#include "qedr_roce_cm.h"
 
 void qedr_inc_sw_gsi_cons(struct qedr_qp_hwq_info *info)
 {
diff --git a/drivers/infiniband/hw/qedr/qedr_cm.h b/drivers/infiniband/hw/qedr/qedr_roce_cm.h
similarity index 100%
rename from drivers/infiniband/hw/qedr/qedr_cm.h
rename to drivers/infiniband/hw/qedr/qedr_roce_cm.h
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 9ee2dce..03a24a6 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -49,7 +49,7 @@
 #include "qedr.h"
 #include "verbs.h"
 #include <rdma/qedr-abi.h>
-#include "qedr_cm.h"
+#include "qedr_roce_cm.h"
 
 #define DB_ADDR_SHIFT(addr)		((addr) << DB_PWM_ADDR_OFFSET_SHIFT)
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 3/8] RDMA/qedr: Add support for registering an iWARP device

[Michal Kalderon]

There are slight differences between iWARP and RoCE in the ibdev
registration. This patch handles the changes.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/main.c  | 84 +++++++++++++++++++++++++++++++++++---
 drivers/infiniband/hw/qedr/qedr.h  |  3 ++
 drivers/infiniband/hw/qedr/verbs.c | 34 +++++++--------
 drivers/infiniband/hw/qedr/verbs.h |  2 +
 4 files changed, 97 insertions(+), 26 deletions(-)

diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index 199b6ed..3a38a1a 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -33,6 +33,8 @@
 #include <rdma/ib_verbs.h>
 #include <rdma/ib_addr.h>
 #include <rdma/ib_user_verbs.h>
+#include <rdma/iw_cm.h>
+#include <rdma/ib_mad.h>
 #include <linux/netdevice.h>
 #include <linux/iommu.h>
 #include <linux/pci.h>
@@ -93,8 +95,75 @@ static struct net_device *qedr_get_netdev(struct ib_device *dev, u8 port_num)
 	return qdev->ndev;
 }
 
+int qedr_roce_port_immutable(struct ib_device *ibdev, u8 port_num,
+			     struct ib_port_immutable *immutable)
+{
+	struct ib_port_attr attr;
+	int err;
+
+	err = qedr_query_port(ibdev, port_num, &attr);
+	if (err)
+		return err;
+
+	immutable->pkey_tbl_len = attr.pkey_tbl_len;
+	immutable->gid_tbl_len = attr.gid_tbl_len;
+	immutable->core_cap_flags = RDMA_CORE_PORT_IBA_ROCE |
+	    RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
+	immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+
+	return 0;
+}
+
+int qedr_iw_port_immutable(struct ib_device *ibdev, u8 port_num,
+			   struct ib_port_immutable *immutable)
+{
+	struct ib_port_attr attr;
+	int err;
+
+	err = qedr_query_port(ibdev, port_num, &attr);
+	if (err)
+		return err;
+
+	immutable->pkey_tbl_len = 1;
+	immutable->gid_tbl_len = 1;
+	immutable->core_cap_flags = RDMA_CORE_PORT_IWARP;
+	immutable->max_mad_size = 0;
+
+	return 0;
+}
+
+int qedr_iw_register_device(struct qedr_dev *dev)
+{
+	dev->ibdev.node_type = RDMA_NODE_RNIC;
+	dev->ibdev.query_gid = qedr_iw_query_gid;
+
+	dev->ibdev.get_port_immutable = qedr_iw_port_immutable;
+
+	dev->ibdev.iwcm = kzalloc(sizeof(*dev->ibdev.iwcm), GFP_KERNEL);
+	if (!dev->ibdev.iwcm)
+		return -ENOMEM;
+
+	memcpy(dev->ibdev.iwcm->ifname,
+	       dev->ndev->name, sizeof(dev->ibdev.iwcm->ifname));
+
+	return 0;
+}
+
+void qedr_roce_register_device(struct qedr_dev *dev)
+{
+	dev->ibdev.node_type = RDMA_NODE_IB_CA;
+	dev->ibdev.query_gid = qedr_query_gid;
+
+	dev->ibdev.add_gid = qedr_add_gid;
+	dev->ibdev.del_gid = qedr_del_gid;
+
+	dev->ibdev.get_port_immutable = qedr_roce_port_immutable;
+}
+
 static int qedr_register_device(struct qedr_dev *dev)
 {
+	int rc;
+
 	strlcpy(dev->ibdev.name, "qedr%d", IB_DEVICE_NAME_MAX);
 
 	dev->ibdev.node_guid = dev->attr.node_guid;
@@ -122,18 +191,21 @@ static int qedr_register_device(struct qedr_dev *dev)
 				     QEDR_UVERBS(POST_SEND) |
 				     QEDR_UVERBS(POST_RECV);
 
+	if (IS_IWARP(dev)) {
+		rc = qedr_iw_register_device(dev);
+		if (rc)
+			return rc;
+	} else {
+		qedr_roce_register_device(dev);
+	}
+
 	dev->ibdev.phys_port_cnt = 1;
 	dev->ibdev.num_comp_vectors = dev->num_cnq;
-	dev->ibdev.node_type = RDMA_NODE_IB_CA;
 
 	dev->ibdev.query_device = qedr_query_device;
 	dev->ibdev.query_port = qedr_query_port;
 	dev->ibdev.modify_port = qedr_modify_port;
 
-	dev->ibdev.query_gid = qedr_query_gid;
-	dev->ibdev.add_gid = qedr_add_gid;
-	dev->ibdev.del_gid = qedr_del_gid;
-
 	dev->ibdev.alloc_ucontext = qedr_alloc_ucontext;
 	dev->ibdev.dealloc_ucontext = qedr_dealloc_ucontext;
 	dev->ibdev.mmap = qedr_mmap;
@@ -167,7 +239,7 @@ static int qedr_register_device(struct qedr_dev *dev)
 	dev->ibdev.post_recv = qedr_post_recv;
 
 	dev->ibdev.process_mad = qedr_process_mad;
-	dev->ibdev.get_port_immutable = qedr_port_immutable;
+
 	dev->ibdev.get_netdev = qedr_get_netdev;
 
 	dev->ibdev.dev.parent = &dev->pdev->dev;
diff --git a/drivers/infiniband/hw/qedr/qedr.h b/drivers/infiniband/hw/qedr/qedr.h
index b2bb42e..349e6ae 100644
--- a/drivers/infiniband/hw/qedr/qedr.h
+++ b/drivers/infiniband/hw/qedr/qedr.h
@@ -43,6 +43,8 @@
 
 #define QEDR_NODE_DESC "QLogic 579xx RoCE HCA"
 #define DP_NAME(dev) ((dev)->ibdev.name)
+#define IS_IWARP(_dev) ((_dev)->rdma_type == QED_RDMA_TYPE_IWARP)
+#define IS_ROCE(_dev) ((_dev)->rdma_type == QED_RDMA_TYPE_ROCE)
 
 #define DP_DEBUG(dev, module, fmt, ...)					\
 	pr_debug("(%s) " module ": " fmt,				\
@@ -160,6 +162,7 @@ struct qedr_dev {
 	struct qedr_cq		*gsi_sqcq;
 	struct qedr_cq		*gsi_rqcq;
 	struct qedr_qp		*gsi_qp;
+	enum qed_rdma_type	rdma_type;
 
 	unsigned long enet_state;
 
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 03a24a6..b9245fc 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -70,6 +70,20 @@ int qedr_query_pkey(struct ib_device *ibdev, u8 port, u16 index, u16 *pkey)
 	return 0;
 }
 
+int qedr_iw_query_gid(struct ib_device *ibdev, u8 port,
+		      int index, union ib_gid *sgid)
+{
+	struct qedr_dev *dev = get_qedr_dev(ibdev);
+
+	memset(sgid->raw, 0, sizeof(sgid->raw));
+	ether_addr_copy(sgid->raw, dev->ndev->dev_addr);
+
+	DP_DEBUG(dev, QEDR_MSG_INIT, "QUERY sgid[%d]=%llx:%llx\n", index,
+		 sgid->global.interface_id, sgid->global.subnet_prefix);
+
+	return 0;
+}
+
 int qedr_query_gid(struct ib_device *ibdev, u8 port, int index,
 		   union ib_gid *sgid)
 {
@@ -3603,23 +3617,3 @@ int qedr_process_mad(struct ib_device *ibdev, int process_mad_flags,
 		 mad_hdr->method, mad_hdr->mgmt_class, mad_hdr->status);
 	return IB_MAD_RESULT_SUCCESS;
 }
-
-int qedr_port_immutable(struct ib_device *ibdev, u8 port_num,
-			struct ib_port_immutable *immutable)
-{
-	struct ib_port_attr attr;
-	int err;
-
-	immutable->core_cap_flags = RDMA_CORE_PORT_IBA_ROCE |
-				    RDMA_CORE_PORT_IBA_ROCE_UDP_ENCAP;
-
-	err = ib_query_port(ibdev, port_num, &attr);
-	if (err)
-		return err;
-
-	immutable->pkey_tbl_len = attr.pkey_tbl_len;
-	immutable->gid_tbl_len = attr.gid_tbl_len;
-	immutable->max_mad_size = IB_MGMT_MAD_SIZE;
-
-	return 0;
-}
diff --git a/drivers/infiniband/hw/qedr/verbs.h b/drivers/infiniband/hw/qedr/verbs.h
index 0f8ab49..1a94425 100644
--- a/drivers/infiniband/hw/qedr/verbs.h
+++ b/drivers/infiniband/hw/qedr/verbs.h
@@ -39,6 +39,8 @@ int qedr_modify_port(struct ib_device *, u8 port, int mask,
 		     struct ib_port_modify *props);
 
 int qedr_query_gid(struct ib_device *, u8 port, int index, union ib_gid *gid);
+int qedr_iw_query_gid(struct ib_device *ibdev, u8 port,
+		      int index, union ib_gid *gid);
 
 int qedr_query_pkey(struct ib_device *, u8 port, u16 index, u16 *pkey);
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 4/8] RDMA/qedr: Add iWARP support in existing verbs

[Michal Kalderon]

Make slight modifications to common RoCE/iWARP code.
Add additional doorbell for iWARP post_send.
iWARP QP pbl is allocated in qed and not in qedr.

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/qedr.h  |   3 +
 drivers/infiniband/hw/qedr/verbs.c | 171 +++++++++++++++++++++++++++++--------
 2 files changed, 139 insertions(+), 35 deletions(-)

diff --git a/drivers/infiniband/hw/qedr/qedr.h b/drivers/infiniband/hw/qedr/qedr.h
index 349e6ae..0ddac91 100644
--- a/drivers/infiniband/hw/qedr/qedr.h
+++ b/drivers/infiniband/hw/qedr/qedr.h
@@ -320,6 +320,9 @@ struct qedr_qp_hwq_info {
 	/* DB */
 	void __iomem *db;
 	union db_prod32 db_data;
+
+	void __iomem *iwarp_db2;
+	union db_prod32 iwarp_db2_data;
 };
 
 #define QEDR_INC_SW_IDX(p_info, index)					\
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index b9245fc..b40c815 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -277,8 +277,13 @@ int qedr_query_port(struct ib_device *ibdev, u8 port, struct ib_port_attr *attr)
 	attr->sm_lid = 0;
 	attr->sm_sl = 0;
 	attr->port_cap_flags = IB_PORT_IP_BASED_GIDS;
-	attr->gid_tbl_len = QEDR_MAX_SGID;
-	attr->pkey_tbl_len = QEDR_ROCE_PKEY_TABLE_LEN;
+	if (rdma_protocol_iwarp(&dev->ibdev, 1)) {
+		attr->gid_tbl_len = 1;
+		attr->pkey_tbl_len = 1;
+	} else {
+		attr->gid_tbl_len = QEDR_MAX_SGID;
+		attr->pkey_tbl_len = QEDR_ROCE_PKEY_TABLE_LEN;
+	}
 	attr->bad_pkey_cntr = rdma_port->pkey_bad_counter;
 	attr->qkey_viol_cntr = 0;
 	get_link_speed_and_width(rdma_port->link_speed,
@@ -1433,6 +1438,21 @@ static int qedr_create_user_qp(struct qedr_dev *dev,
 	return rc;
 }
 
+static void qedr_set_iwarp_db_info(struct qedr_dev *dev, struct qedr_qp *qp)
+{
+	qp->sq.db = dev->db_addr +
+	    DB_ADDR_SHIFT(DQ_PWM_OFFSET_XCM_RDMA_SQ_PROD);
+	qp->sq.db_data.data.icid = qp->icid;
+
+	qp->rq.db = dev->db_addr +
+		    DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_IWARP_RQ_PROD);
+	qp->rq.db_data.data.icid = qp->icid;
+	qp->rq.iwarp_db2 = dev->db_addr +
+			   DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_FLAGS);
+	qp->rq.iwarp_db2_data.data.icid = qp->icid;
+	qp->rq.iwarp_db2_data.data.value = DQ_TCM_IWARP_POST_RQ_CF_CMD;
+}
+
 static int
 qedr_roce_create_kernel_qp(struct qedr_dev *dev,
 			   struct qedr_qp *qp,
@@ -1479,8 +1499,71 @@ static int qedr_create_user_qp(struct qedr_dev *dev,
 	qp->icid = out_params.icid;
 
 	qedr_set_roce_db_info(dev, qp);
+	return rc;
+}
 
-	return 0;
+static int
+qedr_iwarp_create_kernel_qp(struct qedr_dev *dev,
+			    struct qedr_qp *qp,
+			    struct qed_rdma_create_qp_in_params *in_params,
+			    u32 n_sq_elems, u32 n_rq_elems)
+{
+	struct qed_rdma_create_qp_out_params out_params;
+	struct qed_chain_ext_pbl ext_pbl;
+	int rc;
+
+	in_params->sq_num_pages = QED_CHAIN_PAGE_CNT(n_sq_elems,
+						     QEDR_SQE_ELEMENT_SIZE,
+						     QED_CHAIN_MODE_PBL);
+	in_params->rq_num_pages = QED_CHAIN_PAGE_CNT(n_rq_elems,
+						     QEDR_RQE_ELEMENT_SIZE,
+						     QED_CHAIN_MODE_PBL);
+
+	qp->qed_qp = dev->ops->rdma_create_qp(dev->rdma_ctx,
+					      in_params, &out_params);
+
+	if (!qp->qed_qp)
+		return -EINVAL;
+
+	/* Now we allocate the chain */
+	ext_pbl.p_pbl_virt = out_params.sq_pbl_virt;
+	ext_pbl.p_pbl_phys = out_params.sq_pbl_phys;
+
+	rc = dev->ops->common->chain_alloc(dev->cdev,
+					   QED_CHAIN_USE_TO_PRODUCE,
+					   QED_CHAIN_MODE_PBL,
+					   QED_CHAIN_CNT_TYPE_U32,
+					   n_sq_elems,
+					   QEDR_SQE_ELEMENT_SIZE,
+					   &qp->sq.pbl, &ext_pbl);
+
+	if (rc)
+		goto err;
+
+	ext_pbl.p_pbl_virt = out_params.rq_pbl_virt;
+	ext_pbl.p_pbl_phys = out_params.rq_pbl_phys;
+
+	rc = dev->ops->common->chain_alloc(dev->cdev,
+					   QED_CHAIN_USE_TO_CONSUME_PRODUCE,
+					   QED_CHAIN_MODE_PBL,
+					   QED_CHAIN_CNT_TYPE_U32,
+					   n_rq_elems,
+					   QEDR_RQE_ELEMENT_SIZE,
+					   &qp->rq.pbl, &ext_pbl);
+
+	if (rc)
+		goto err;
+
+	qp->qp_id = out_params.qp_id;
+	qp->icid = out_params.icid;
+
+	qedr_set_iwarp_db_info(dev, qp);
+	return rc;
+
+err:
+	dev->ops->rdma_destroy_qp(dev->rdma_ctx, qp->qed_qp);
+
+	return rc;
 }
 
 static void qedr_cleanup_kernel(struct qedr_dev *dev, struct qedr_qp *qp)
@@ -1555,8 +1638,12 @@ static int qedr_create_kernel_qp(struct qedr_dev *dev,
 
 	n_rq_elems = qp->rq.max_wr * QEDR_MAX_RQE_ELEMENTS_PER_RQE;
 
-	rc = qedr_roce_create_kernel_qp(dev, qp, &in_params,
-					n_sq_elems, n_rq_elems);
+	if (rdma_protocol_iwarp(&dev->ibdev, 1))
+		rc = qedr_iwarp_create_kernel_qp(dev, qp, &in_params,
+						 n_sq_elems, n_rq_elems);
+	else
+		rc = qedr_roce_create_kernel_qp(dev, qp, &in_params,
+						n_sq_elems, n_rq_elems);
 	if (rc)
 		qedr_cleanup_kernel(dev, qp);
 
@@ -1703,10 +1790,13 @@ static int qedr_update_qp_state(struct qedr_dev *dev,
 			/* Update doorbell (in case post_recv was
 			 * done before move to RTR)
 			 */
-			wmb();
-			writel(qp->rq.db_data.raw, qp->rq.db);
-			/* Make sure write takes effect */
-			mmiowb();
+
+			if (rdma_protocol_roce(&dev->ibdev, 1)) {
+				wmb();
+				writel(qp->rq.db_data.raw, qp->rq.db);
+				/* Make sure write takes effect */
+				mmiowb();
+			}
 			break;
 		case QED_ROCE_QP_STATE_ERR:
 			break;
@@ -1800,16 +1890,18 @@ int qedr_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	else
 		new_qp_state = old_qp_state;
 
-	if (!ib_modify_qp_is_ok
-	    (old_qp_state, new_qp_state, ibqp->qp_type, attr_mask,
-	     IB_LINK_LAYER_ETHERNET)) {
-		DP_ERR(dev,
-		       "modify qp: invalid attribute mask=0x%x specified for\n"
-		       "qpn=0x%x of type=0x%x old_qp_state=0x%x, new_qp_state=0x%x\n",
-		       attr_mask, qp->qp_id, ibqp->qp_type, old_qp_state,
-		       new_qp_state);
-		rc = -EINVAL;
-		goto err;
+	if (rdma_protocol_roce(&dev->ibdev, 1)) {
+		if (!ib_modify_qp_is_ok(old_qp_state, new_qp_state,
+					ibqp->qp_type, attr_mask,
+					IB_LINK_LAYER_ETHERNET)) {
+			DP_ERR(dev,
+			       "modify qp: invalid attribute mask=0x%x specified for\n"
+			       "qpn=0x%x of type=0x%x old_qp_state=0x%x, new_qp_state=0x%x\n",
+			       attr_mask, qp->qp_id, ibqp->qp_type,
+			       old_qp_state, new_qp_state);
+			rc = -EINVAL;
+			goto err;
+		}
 	}
 
 	/* Translate the masks... */
@@ -2125,15 +2217,17 @@ int qedr_destroy_qp(struct ib_qp *ibqp)
 	DP_DEBUG(dev, QEDR_MSG_QP, "destroy qp: destroying %p, qp type=%d\n",
 		 qp, qp->qp_type);
 
-	if ((qp->state != QED_ROCE_QP_STATE_RESET) &&
-	    (qp->state != QED_ROCE_QP_STATE_ERR) &&
-	    (qp->state != QED_ROCE_QP_STATE_INIT)) {
+	if (rdma_protocol_roce(&dev->ibdev, 1)) {
+		if ((qp->state != QED_ROCE_QP_STATE_RESET) &&
+		    (qp->state != QED_ROCE_QP_STATE_ERR) &&
+		    (qp->state != QED_ROCE_QP_STATE_INIT)) {
 
-		attr.qp_state = IB_QPS_ERR;
-		attr_mask |= IB_QP_STATE;
+			attr.qp_state = IB_QPS_ERR;
+			attr_mask |= IB_QP_STATE;
 
-		/* Change the QP state to ERROR */
-		qedr_modify_qp(ibqp, &attr, attr_mask, NULL);
+			/* Change the QP state to ERROR */
+			qedr_modify_qp(ibqp, &attr, attr_mask, NULL);
+		}
 	}
 
 	if (qp->qp_type == IB_QPT_GSI)
@@ -3028,15 +3122,17 @@ int qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 
 	spin_lock_irqsave(&qp->q_lock, flags);
 
-	if ((qp->state != QED_ROCE_QP_STATE_RTS) &&
-	    (qp->state != QED_ROCE_QP_STATE_ERR) &&
-	    (qp->state != QED_ROCE_QP_STATE_SQD)) {
-		spin_unlock_irqrestore(&qp->q_lock, flags);
-		*bad_wr = wr;
-		DP_DEBUG(dev, QEDR_MSG_CQ,
-			 "QP in wrong state! QP icid=0x%x state %d\n",
-			 qp->icid, qp->state);
-		return -EINVAL;
+	if (rdma_protocol_roce(&dev->ibdev, 1)) {
+		if ((qp->state != QED_ROCE_QP_STATE_RTS) &&
+		    (qp->state != QED_ROCE_QP_STATE_ERR) &&
+		    (qp->state != QED_ROCE_QP_STATE_SQD)) {
+			spin_unlock_irqrestore(&qp->q_lock, flags);
+			*bad_wr = wr;
+			DP_DEBUG(dev, QEDR_MSG_CQ,
+				 "QP in wrong state! QP icid=0x%x state %d\n",
+				 qp->icid, qp->state);
+			return -EINVAL;
+		}
 	}
 
 	while (wr) {
@@ -3156,6 +3252,11 @@ int qedr_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
 		/* Make sure write sticks */
 		mmiowb();
 
+		if (rdma_protocol_iwarp(&dev->ibdev, 1)) {
+			writel(qp->rq.iwarp_db2_data.raw, qp->rq.iwarp_db2);
+			mmiowb();	/* for second doorbell */
+		}
+
 		wr = wr->next;
 	}
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 5/8] RDMA/qedr: Add support for read with invalidate, supported in iWARP

[Michal Kalderon]

iWARP supports read with invalidate. There is an assumption
that read with invalidate will only be called on an iWARP device

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/qedr_hsi_rdma.h | 6 ++++--
 drivers/infiniband/hw/qedr/verbs.c         | 8 +++-----
 2 files changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h b/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h
index 5c98d20..b7587f1 100644
--- a/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h
+++ b/drivers/infiniband/hw/qedr/qedr_hsi_rdma.h
@@ -655,8 +655,10 @@ struct rdma_sq_rdma_wqe_1st {
 #define RDMA_SQ_RDMA_WQE_1ST_INLINE_FLG_SHIFT      4
 #define RDMA_SQ_RDMA_WQE_1ST_DIF_ON_HOST_FLG_MASK  0x1
 #define RDMA_SQ_RDMA_WQE_1ST_DIF_ON_HOST_FLG_SHIFT 5
-#define RDMA_SQ_RDMA_WQE_1ST_RESERVED0_MASK        0x3
-#define RDMA_SQ_RDMA_WQE_1ST_RESERVED0_SHIFT       6
+#define RDMA_SQ_RDMA_WQE_1ST_READ_INV_FLG_MASK     0x1
+#define RDMA_SQ_RDMA_WQE_1ST_READ_INV_FLG_SHIFT    6
+#define RDMA_SQ_RDMA_WQE_1ST_RESERVED0_MASK        0x1
+#define RDMA_SQ_RDMA_WQE_1ST_RESERVED0_SHIFT       7
 	u8 wqe_size;
 	u8 prev_wqe_size;
 };
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index b40c815..a439e0a 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -2848,6 +2848,7 @@ static enum ib_wc_opcode qedr_ib_to_wc_opcode(enum ib_wr_opcode opcode)
 	case IB_WR_SEND_WITH_INV:
 		return IB_WC_SEND;
 	case IB_WR_RDMA_READ:
+	case IB_WR_RDMA_READ_WITH_INV:
 		return IB_WC_RDMA_READ;
 	case IB_WR_ATOMIC_CMP_AND_SWP:
 		return IB_WC_COMP_SWAP;
@@ -3008,11 +3009,8 @@ static int __qedr_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
 		qp->wqe_wr_id[qp->sq.prod].bytes_len = rwqe->length;
 		break;
 	case IB_WR_RDMA_READ_WITH_INV:
-		DP_ERR(dev,
-		       "RDMA READ WITH INVALIDATE not supported\n");
-		*bad_wr = wr;
-		rc = -EINVAL;
-		break;
+		SET_FIELD2(wqe->flags, RDMA_SQ_RDMA_WQE_1ST_READ_INV_FLG, 1);
+		/* fallthrough... same is identical to RDMA READ */
 
 	case IB_WR_RDMA_READ:
 		wqe->req_type = RDMA_SQ_REQ_TYPE_RDMA_RD;
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 6/8] RDMA/qedr: Add iWARP connection management qp related callbacks

[Michal Kalderon]

This patch implements the following iWARP callbacks:
qp_add_ref
qp_rem_ref
get_qp

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/Makefile     |  2 +-
 drivers/infiniband/hw/qedr/main.c       | 10 ++++++
 drivers/infiniband/hw/qedr/qedr.h       |  5 ++-
 drivers/infiniband/hw/qedr/qedr_iw_cm.c | 57 +++++++++++++++++++++++++++++++++
 drivers/infiniband/hw/qedr/qedr_iw_cm.h | 37 +++++++++++++++++++++
 drivers/infiniband/hw/qedr/verbs.c      | 38 ++++++++++++++++++++--
 6 files changed, 145 insertions(+), 4 deletions(-)
 create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.c
 create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.h

diff --git a/drivers/infiniband/hw/qedr/Makefile b/drivers/infiniband/hw/qedr/Makefile
index 331a361..1c0bc4f 100644
--- a/drivers/infiniband/hw/qedr/Makefile
+++ b/drivers/infiniband/hw/qedr/Makefile
@@ -1,3 +1,3 @@
 obj-$(CONFIG_INFINIBAND_QEDR) := qedr.o
 
-qedr-y := main.o verbs.o qedr_roce_cm.o
+qedr-y := main.o verbs.o qedr_roce_cm.o qedr_iw_cm.o
diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index 3a38a1a..545cea9 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -39,12 +39,14 @@
 #include <linux/iommu.h>
 #include <linux/pci.h>
 #include <net/addrconf.h>
+#include <linux/idr.h>
 
 #include <linux/qed/qed_chain.h>
 #include <linux/qed/qed_if.h>
 #include "qedr.h"
 #include "verbs.h"
 #include <rdma/qedr-abi.h>
+#include "qedr_iw_cm.h"
 
 MODULE_DESCRIPTION("QLogic 40G/100G ROCE Driver");
 MODULE_AUTHOR("QLogic Corporation");
@@ -142,6 +144,9 @@ int qedr_iw_register_device(struct qedr_dev *dev)
 	dev->ibdev.iwcm = kzalloc(sizeof(*dev->ibdev.iwcm), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)
 		return -ENOMEM;
+	dev->ibdev.iwcm->add_ref = qedr_iw_qp_add_ref;
+	dev->ibdev.iwcm->rem_ref = qedr_iw_qp_rem_ref;
+	dev->ibdev.iwcm->get_qp = qedr_iw_get_qp;
 
 	memcpy(dev->ibdev.iwcm->ifname,
 	       dev->ndev->name, sizeof(dev->ibdev.iwcm->ifname));
@@ -314,6 +319,11 @@ static int qedr_alloc_resources(struct qedr_dev *dev)
 
 	spin_lock_init(&dev->sgid_lock);
 
+	if (IS_IWARP(dev)) {
+		spin_lock_init(&dev->idr_lock);
+		idr_init(&dev->qpidr);
+	}
+
 	/* Allocate Status blocks for CNQ */
 	dev->sb_array = kcalloc(dev->num_cnq, sizeof(*dev->sb_array),
 				GFP_KERNEL);
diff --git a/drivers/infiniband/hw/qedr/qedr.h b/drivers/infiniband/hw/qedr/qedr.h
index 0ddac91..ec361cb 100644
--- a/drivers/infiniband/hw/qedr/qedr.h
+++ b/drivers/infiniband/hw/qedr/qedr.h
@@ -33,6 +33,7 @@
 #define __QEDR_H__
 
 #include <linux/pci.h>
+#include <linux/idr.h>
 #include <rdma/ib_addr.h>
 #include <linux/qed/qed_if.h>
 #include <linux/qed/qed_chain.h>
@@ -163,7 +164,8 @@ struct qedr_dev {
 	struct qedr_cq		*gsi_rqcq;
 	struct qedr_qp		*gsi_qp;
 	enum qed_rdma_type	rdma_type;
-
+	spinlock_t		idr_lock; /* Protect qpidr data-structure */
+	struct idr		qpidr;
 	unsigned long enet_state;
 
 	u8 user_dpm_enabled;
@@ -400,6 +402,7 @@ struct qedr_qp {
 	/* Relevant to qps created from user space only (applications) */
 	struct qedr_userq usq;
 	struct qedr_userq urq;
+	atomic_t refcnt;
 };
 
 struct qedr_ah {
diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.c b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
new file mode 100644
index 0000000..8811dbc
--- /dev/null
+++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
@@ -0,0 +1,57 @@
+/* QLogic qedr NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include "qedr.h"
+void qedr_iw_qp_add_ref(struct ib_qp *ibqp)
+{
+	struct qedr_qp *qp = get_qedr_qp(ibqp);
+
+	atomic_inc(&qp->refcnt);
+}
+
+void qedr_iw_qp_rem_ref(struct ib_qp *ibqp)
+{
+	struct qedr_qp *qp = get_qedr_qp(ibqp);
+
+	if (atomic_dec_and_test(&qp->refcnt)) {
+		spin_lock_irq(&qp->dev->idr_lock);
+		idr_remove(&qp->dev->qpidr, qp->qp_id);
+		spin_unlock_irq(&qp->dev->idr_lock);
+		kfree(qp);
+	}
+}
+
+struct ib_qp *qedr_iw_get_qp(struct ib_device *ibdev, int qpn)
+{
+	struct qedr_dev *dev = get_qedr_dev(ibdev);
+
+	return idr_find(&dev->qpidr, qpn);
+}
diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.h b/drivers/infiniband/hw/qedr/qedr_iw_cm.h
new file mode 100644
index 0000000..ea6d17af
--- /dev/null
+++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.h
@@ -0,0 +1,37 @@
+/* QLogic qed NIC Driver
+ * Copyright (c) 2015-2017  QLogic Corporation
+ *
+ * This software is available to you under a choice of one of two
+ * licenses.  You may choose to be licensed under the terms of the GNU
+ * General Public License (GPL) Version 2, available from the file
+ * COPYING in the main directory of this source tree, or the
+ * OpenIB.org BSD license below:
+ *
+ *     Redistribution and use in source and binary forms, with or
+ *     without modification, are permitted provided that the following
+ *     conditions are met:
+ *
+ *      - Redistributions of source code must retain the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer.
+ *
+ *      - Redistributions in binary form must reproduce the above
+ *        copyright notice, this list of conditions and the following
+ *        disclaimer in the documentation and /or other materials
+ *        provided with the distribution.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ * SOFTWARE.
+ */
+#include <rdma/iw_cm.h>
+void qedr_iw_qp_add_ref(struct ib_qp *qp);
+
+void qedr_iw_qp_rem_ref(struct ib_qp *qp);
+
+struct ib_qp *qedr_iw_get_qp(struct ib_device *dev, int qpn);
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index a439e0a..bdce96b 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -1283,6 +1283,7 @@ static void qedr_set_common_qp_params(struct qedr_dev *dev,
 				      struct ib_qp_init_attr *attrs)
 {
 	spin_lock_init(&qp->q_lock);
+	atomic_set(&qp->refcnt, 1);
 	qp->pd = pd;
 	qp->qp_type = attrs->qp_type;
 	qp->max_inline_data = attrs->cap.max_inline_data;
@@ -1353,6 +1354,33 @@ static inline void qedr_qp_user_print(struct qedr_dev *dev, struct qedr_qp *qp)
 		 qp->usq.buf_len, qp->urq.buf_addr, qp->urq.buf_len);
 }
 
+static int qedr_idr_add(struct qedr_dev *dev, void *ptr, u32 id)
+{
+	int rc;
+
+	if (!rdma_protocol_iwarp(&dev->ibdev, 1))
+		return 0;
+
+	idr_preload(GFP_KERNEL);
+	spin_lock_irq(&dev->idr_lock);
+
+	rc = idr_alloc(&dev->qpidr, ptr, id, id + 1, GFP_ATOMIC);
+
+	spin_unlock_irq(&dev->idr_lock);
+	idr_preload_end();
+
+	return rc < 0 ? rc : 0;
+}
+
+static void qedr_idr_remove(struct qedr_dev *dev, u32 id)
+{
+	if (!rdma_protocol_iwarp(&dev->ibdev, 1))
+		return;
+
+	spin_lock_irq(&dev->idr_lock);
+	idr_remove(&dev->qpidr, id);
+	spin_unlock_irq(&dev->idr_lock);
+}
 static void qedr_cleanup_user(struct qedr_dev *dev, struct qedr_qp *qp)
 {
 	if (qp->usq.umem)
@@ -1703,6 +1731,10 @@ struct ib_qp *qedr_create_qp(struct ib_pd *ibpd,
 
 	qp->ibqp.qp_num = qp->qp_id;
 
+	rc = qedr_idr_add(dev, qp, qp->qp_id);
+	if (rc)
+		goto err;
+
 	return &qp->ibqp;
 
 err:
@@ -2235,8 +2267,10 @@ int qedr_destroy_qp(struct ib_qp *ibqp)
 
 	qedr_free_qp_resources(dev, qp);
 
-	kfree(qp);
-
+	if (atomic_dec_and_test(&qp->refcnt)) {
+		qedr_idr_remove(dev, qp->qp_id);
+		kfree(qp);
+	}
 	return rc;
 }
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 7/8] RDMA/qedr: Add iWARP connection management functions

[Michal Kalderon]

Implements the iWARP connection management functions:
connect, accept, create listener and destroy listener

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/main.c       |  11 +
 drivers/infiniband/hw/qedr/qedr.h       |  22 +-
 drivers/infiniband/hw/qedr/qedr_iw_cm.c | 689 ++++++++++++++++++++++++++++++++
 drivers/infiniband/hw/qedr/qedr_iw_cm.h |  12 +
 drivers/infiniband/hw/qedr/verbs.c      |  17 +
 5 files changed, 750 insertions(+), 1 deletion(-)

diff --git a/drivers/infiniband/hw/qedr/main.c b/drivers/infiniband/hw/qedr/main.c
index 545cea9..c3c60e1 100644
--- a/drivers/infiniband/hw/qedr/main.c
+++ b/drivers/infiniband/hw/qedr/main.c
@@ -144,6 +144,12 @@ int qedr_iw_register_device(struct qedr_dev *dev)
 	dev->ibdev.iwcm = kzalloc(sizeof(*dev->ibdev.iwcm), GFP_KERNEL);
 	if (!dev->ibdev.iwcm)
 		return -ENOMEM;
+
+	dev->ibdev.iwcm->connect = qedr_iw_connect;
+	dev->ibdev.iwcm->accept = qedr_iw_accept;
+	dev->ibdev.iwcm->reject = qedr_iw_reject;
+	dev->ibdev.iwcm->create_listen = qedr_iw_create_listen;
+	dev->ibdev.iwcm->destroy_listen = qedr_iw_destroy_listen;
 	dev->ibdev.iwcm->add_ref = qedr_iw_qp_add_ref;
 	dev->ibdev.iwcm->rem_ref = qedr_iw_qp_rem_ref;
 	dev->ibdev.iwcm->get_qp = qedr_iw_get_qp;
@@ -295,6 +301,9 @@ static void qedr_free_resources(struct qedr_dev *dev)
 {
 	int i;
 
+	if (IS_IWARP(dev))
+		destroy_workqueue(dev->iwarp_wq);
+
 	for (i = 0; i < dev->num_cnq; i++) {
 		qedr_free_mem_sb(dev, &dev->sb_array[i], dev->sb_start + i);
 		dev->ops->common->chain_free(dev->cdev, &dev->cnq_array[i].pbl);
@@ -322,6 +331,7 @@ static int qedr_alloc_resources(struct qedr_dev *dev)
 	if (IS_IWARP(dev)) {
 		spin_lock_init(&dev->idr_lock);
 		idr_init(&dev->qpidr);
+		dev->iwarp_wq = create_singlethread_workqueue("qedr_iwarpq");
 	}
 
 	/* Allocate Status blocks for CNQ */
@@ -799,6 +809,7 @@ static int qedr_init_hw(struct qedr_dev *dev)
 	in_params->events = &events;
 	in_params->cq_mode = QED_RDMA_CQ_MODE_32_BITS;
 	in_params->max_mtu = dev->ndev->mtu;
+	dev->iwarp_max_mtu = dev->ndev->mtu;
 	ether_addr_copy(&in_params->mac_addr[0], dev->ndev->dev_addr);
 
 	rc = dev->ops->rdma_init(dev->cdev, in_params);
diff --git a/drivers/infiniband/hw/qedr/qedr.h b/drivers/infiniband/hw/qedr/qedr.h
index ec361cb..61732ca 100644
--- a/drivers/infiniband/hw/qedr/qedr.h
+++ b/drivers/infiniband/hw/qedr/qedr.h
@@ -59,6 +59,7 @@
 #define QEDR_MSG_SQ   "  SQ"
 #define QEDR_MSG_QP   "  QP"
 #define QEDR_MSG_GSI  " GSI"
+#define QEDR_MSG_IWARP  " IW"
 
 #define QEDR_CQ_MAGIC_NUMBER	(0x11223344)
 
@@ -166,6 +167,9 @@ struct qedr_dev {
 	enum qed_rdma_type	rdma_type;
 	spinlock_t		idr_lock; /* Protect qpidr data-structure */
 	struct idr		qpidr;
+	struct workqueue_struct *iwarp_wq;
+	u16			iwarp_max_mtu;
+
 	unsigned long enet_state;
 
 	u8 user_dpm_enabled;
@@ -345,7 +349,7 @@ enum qedr_qp_err_bitmap {
 struct qedr_qp {
 	struct ib_qp ibqp;	/* must be first */
 	struct qedr_dev *dev;
-
+	struct qedr_iw_ep *ep;
 	struct qedr_qp_hwq_info sq;
 	struct qedr_qp_hwq_info rq;
 
@@ -403,6 +407,7 @@ struct qedr_qp {
 	struct qedr_userq usq;
 	struct qedr_userq urq;
 	atomic_t refcnt;
+	bool destroyed;
 };
 
 struct qedr_ah {
@@ -483,6 +488,21 @@ static inline int qedr_get_dmac(struct qedr_dev *dev,
 	return 0;
 }
 
+struct qedr_iw_listener {
+	struct qedr_dev *dev;
+	struct iw_cm_id *cm_id;
+	int		backlog;
+	void		*qed_handle;
+};
+
+struct qedr_iw_ep {
+	struct qedr_dev	*dev;
+	struct iw_cm_id	*cm_id;
+	struct qedr_qp	*qp;
+	void		*qed_context;
+	u8		during_connect;
+};
+
 static inline
 struct qedr_ucontext *get_qedr_ucontext(struct ib_ucontext *ibucontext)
 {
diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.c b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
index 8811dbc..fe9b2b6 100644
--- a/drivers/infiniband/hw/qedr/qedr_iw_cm.c
+++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.c
@@ -29,7 +29,696 @@
  * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
  * SOFTWARE.
  */
+#include <net/ip.h>
+#include <net/ipv6.h>
+#include <net/udp.h>
+#include <net/addrconf.h>
+#include <net/route.h>
+#include <net/ip6_route.h>
+#include <net/flow.h>
 #include "qedr.h"
+#include "qedr_iw_cm.h"
+
+static inline void
+qedr_fill_sockaddr4(const struct qed_iwarp_cm_info *cm_info,
+		    struct iw_cm_event *event)
+{
+	struct sockaddr_in *laddr = (struct sockaddr_in *)&event->local_addr;
+	struct sockaddr_in *raddr = (struct sockaddr_in *)&event->remote_addr;
+
+	laddr->sin_family = AF_INET;
+	raddr->sin_family = AF_INET;
+
+	laddr->sin_port = htons(cm_info->local_port);
+	raddr->sin_port = htons(cm_info->remote_port);
+
+	laddr->sin_addr.s_addr = htonl(cm_info->local_ip[0]);
+	raddr->sin_addr.s_addr = htonl(cm_info->remote_ip[0]);
+}
+
+static inline void
+qedr_fill_sockaddr6(const struct qed_iwarp_cm_info *cm_info,
+		    struct iw_cm_event *event)
+{
+	struct sockaddr_in6 *laddr6 = (struct sockaddr_in6 *)&event->local_addr;
+	struct sockaddr_in6 *raddr6 =
+	    (struct sockaddr_in6 *)&event->remote_addr;
+	int i;
+
+	laddr6->sin6_family = AF_INET6;
+	raddr6->sin6_family = AF_INET6;
+
+	laddr6->sin6_port = htons(cm_info->local_port);
+	raddr6->sin6_port = htons(cm_info->remote_port);
+
+	for (i = 0; i < 4; i++) {
+		laddr6->sin6_addr.in6_u.u6_addr32[i] =
+		    htonl(cm_info->local_ip[i]);
+		raddr6->sin6_addr.in6_u.u6_addr32[i] =
+		    htonl(cm_info->remote_ip[i]);
+	}
+}
+
+void
+qedr_iw_mpa_request(void *context, struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_iw_listener *listener = (struct qedr_iw_listener *)context;
+	struct qedr_dev *dev = listener->dev;
+	struct iw_cm_event event;
+	struct qedr_iw_ep *ep;
+
+	ep = kzalloc(sizeof(*ep), GFP_ATOMIC);
+	if (!ep)
+		return;
+
+	ep->dev = dev;
+	ep->qed_context = params->ep_context;
+
+	memset(&event, 0, sizeof(event));
+	event.event = IW_CM_EVENT_CONNECT_REQUEST;
+	event.status = params->status;
+
+	if (params->cm_info->ip_version == QED_TCP_IPV4)
+		qedr_fill_sockaddr4(params->cm_info, &event);
+	else
+		qedr_fill_sockaddr6(params->cm_info, &event);
+
+	event.provider_data = (void *)ep;
+	event.private_data = (void *)params->cm_info->private_data;
+	event.private_data_len = (u8)params->cm_info->private_data_len;
+	event.ord = params->cm_info->ord;
+	event.ird = params->cm_info->ird;
+
+	listener->cm_id->event_handler(listener->cm_id, &event);
+}
+
+void
+qedr_iw_issue_event(void *context,
+		    struct qed_iwarp_cm_event_params *params,
+		    enum iw_cm_event_type event_type)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct iw_cm_event event;
+
+	memset(&event, 0, sizeof(event));
+	event.status = params->status;
+	event.event = event_type;
+
+	if (params->cm_info) {
+		event.ird = params->cm_info->ird;
+		event.ord = params->cm_info->ord;
+		event.private_data_len = params->cm_info->private_data_len;
+		event.private_data = (void *)params->cm_info->private_data;
+	}
+
+	if (ep->cm_id)
+		ep->cm_id->event_handler(ep->cm_id, &event);
+}
+
+void
+qedr_iw_close_event(void *context, struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+
+	if (ep->cm_id) {
+		qedr_iw_issue_event(context, params, IW_CM_EVENT_CLOSE);
+
+		ep->cm_id->rem_ref(ep->cm_id);
+		ep->cm_id = NULL;
+	}
+}
+
+void
+qedr_iw_qp_event(void *context,
+		 struct qed_iwarp_cm_event_params *params,
+		 enum ib_event_type ib_event, char *str)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct qedr_dev *dev = ep->dev;
+	struct ib_qp *ibqp = &ep->qp->ibqp;
+	struct ib_event event;
+
+	DP_NOTICE(dev, "QP error received: %s\n", str);
+
+	if (ibqp->event_handler) {
+		event.event = ib_event;
+		event.device = ibqp->device;
+		event.element.qp = ibqp;
+		ibqp->event_handler(&event, ibqp->qp_context);
+	}
+}
+
+struct qedr_discon_work {
+	struct work_struct		work;
+	struct qedr_iw_ep		*ep;
+	enum qed_iwarp_event_type	event;
+	int				status;
+};
+
+static void qedr_iw_disconnect_worker(struct work_struct *work)
+{
+	struct qedr_discon_work *dwork =
+	    container_of(work, struct qedr_discon_work, work);
+	struct qed_rdma_modify_qp_in_params qp_params = { 0 };
+	struct qedr_iw_ep *ep = dwork->ep;
+	struct qedr_dev *dev = ep->dev;
+	struct qedr_qp *qp = ep->qp;
+	struct iw_cm_event event;
+
+	if (qp->destroyed) {
+		kfree(dwork);
+		qedr_iw_qp_rem_ref(&qp->ibqp);
+		return;
+	}
+
+	memset(&event, 0, sizeof(event));
+	event.status = dwork->status;
+	event.event = IW_CM_EVENT_DISCONNECT;
+
+	/* Success means graceful disconnect was requested. modifying
+	 * to SQD is translated to graceful disconnect. O/w reset is sent
+	 */
+	if (dwork->status)
+		qp_params.new_state = QED_ROCE_QP_STATE_ERR;
+	else
+		qp_params.new_state = QED_ROCE_QP_STATE_SQD;
+
+	kfree(dwork);
+
+	if (ep->cm_id)
+		ep->cm_id->event_handler(ep->cm_id, &event);
+
+	SET_FIELD(qp_params.modify_flags,
+		  QED_RDMA_MODIFY_QP_VALID_NEW_STATE, 1);
+
+	dev->ops->rdma_modify_qp(dev->rdma_ctx, qp->qed_qp, &qp_params);
+
+	qedr_iw_qp_rem_ref(&qp->ibqp);
+}
+
+void
+qedr_iw_disconnect_event(void *context,
+			 struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_discon_work *work;
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct qedr_dev *dev = ep->dev;
+	struct qedr_qp *qp = ep->qp;
+
+	work = kzalloc(sizeof(*work), GFP_ATOMIC);
+	if (!work)
+		return;
+
+	qedr_iw_qp_add_ref(&qp->ibqp);
+	work->ep = ep;
+	work->event = params->event;
+	work->status = params->status;
+
+	INIT_WORK(&work->work, qedr_iw_disconnect_worker);
+	queue_work(dev->iwarp_wq, &work->work);
+}
+
+static void
+qedr_iw_passive_complete(void *context,
+			 struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct qedr_dev *dev = ep->dev;
+
+	/* We will only reach the following state if MPA_REJECT was called on
+	 * passive. In this case there will be no associated QP.
+	 */
+	if ((params->status == -ECONNREFUSED) && (!ep->qp)) {
+		DP_DEBUG(dev, QEDR_MSG_IWARP,
+			 "PASSIVE connection refused releasing ep...\n");
+		kfree(ep);
+		return;
+	}
+
+	qedr_iw_issue_event(context, params, IW_CM_EVENT_ESTABLISHED);
+
+	if (params->status < 0)
+		qedr_iw_close_event(context, params);
+}
+
+int
+qedr_iw_mpa_reply(void *context, struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct qedr_dev *dev = ep->dev;
+	struct qed_iwarp_send_rtr_in rtr_in;
+
+	rtr_in.ep_context = params->ep_context;
+
+	return dev->ops->iwarp_send_rtr(dev->rdma_ctx, &rtr_in);
+}
+
+int
+qedr_iw_event_handler(void *context, struct qed_iwarp_cm_event_params *params)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+	struct qedr_dev *dev = ep->dev;
+
+	switch (params->event) {
+	case QED_IWARP_EVENT_MPA_REQUEST:
+		qedr_iw_mpa_request(context, params);
+		break;
+	case QED_IWARP_EVENT_ACTIVE_MPA_REPLY:
+		qedr_iw_mpa_reply(context, params);
+		break;
+	case QED_IWARP_EVENT_PASSIVE_COMPLETE:
+		ep->during_connect = 0;
+		qedr_iw_passive_complete(context, params);
+		break;
+
+	case QED_IWARP_EVENT_ACTIVE_COMPLETE:
+		ep->during_connect = 0;
+		qedr_iw_issue_event(context,
+				    params,
+				    IW_CM_EVENT_CONNECT_REPLY);
+		if (params->status < 0) {
+			struct qedr_iw_ep *ep = (struct qedr_iw_ep *)context;
+
+			ep->cm_id->rem_ref(ep->cm_id);
+			ep->cm_id = NULL;
+		}
+		break;
+	case QED_IWARP_EVENT_DISCONNECT:
+		qedr_iw_disconnect_event(context, params);
+		break;
+	case QED_IWARP_EVENT_CLOSE:
+		ep->during_connect = 0;
+		qedr_iw_close_event(context, params);
+		break;
+	case QED_IWARP_EVENT_RQ_EMPTY:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_RQ_EMPTY");
+		break;
+	case QED_IWARP_EVENT_IRQ_FULL:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_IRQ_FULL");
+		break;
+	case QED_IWARP_EVENT_LLP_TIMEOUT:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_LLP_TIMEOUT");
+		break;
+	case QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_ACCESS_ERR,
+				 "QED_IWARP_EVENT_REMOTE_PROTECTION_ERROR");
+		break;
+	case QED_IWARP_EVENT_CQ_OVERFLOW:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_CQ_OVERFLOW");
+		break;
+	case QED_IWARP_EVENT_QP_CATASTROPHIC:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_QP_CATASTROPHIC");
+		break;
+	case QED_IWARP_EVENT_LOCAL_ACCESS_ERROR:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_ACCESS_ERR,
+				 "QED_IWARP_EVENT_LOCAL_ACCESS_ERROR");
+		break;
+	case QED_IWARP_EVENT_REMOTE_OPERATION_ERROR:
+		qedr_iw_qp_event(context, params, IB_EVENT_QP_FATAL,
+				 "QED_IWARP_EVENT_REMOTE_OPERATION_ERROR");
+		break;
+	case QED_IWARP_EVENT_TERMINATE_RECEIVED:
+		DP_NOTICE(dev, "Got terminate message\n");
+		break;
+	default:
+		DP_NOTICE(dev, "Unknown event received %d\n", params->event);
+		break;
+	};
+	return 0;
+}
+
+static u16 qedr_iw_get_vlan_ipv4(struct qedr_dev *dev, u32 *addr)
+{
+	struct net_device *ndev;
+	u16 vlan_id = 0;
+
+	ndev = ip_dev_find(&init_net, htonl(addr[0]));
+
+	if (ndev) {
+		vlan_id = rdma_vlan_dev_vlan_id(ndev);
+		dev_put(ndev);
+	}
+	if (vlan_id == 0xffff)
+		vlan_id = 0;
+	return vlan_id;
+}
+
+static u16 qedr_iw_get_vlan_ipv6(u32 *addr)
+{
+	struct net_device *ndev = NULL;
+	struct in6_addr laddr6;
+	u16 vlan_id = 0;
+	int i;
+
+	if (!IS_ENABLED(CONFIG_IPV6))
+		return vlan_id;
+
+	for (i = 0; i < 4; i++)
+		laddr6.in6_u.u6_addr32[i] = htonl(addr[i]);
+
+	rcu_read_lock();
+	for_each_netdev_rcu(&init_net, ndev) {
+		if (ipv6_chk_addr(&init_net, &laddr6, ndev, 1)) {
+			vlan_id = rdma_vlan_dev_vlan_id(ndev);
+			break;
+		}
+	}
+
+	rcu_read_unlock();
+	if (vlan_id == 0xffff)
+		vlan_id = 0;
+
+	return vlan_id;
+}
+
+static int
+qedr_addr4_resolve(struct qedr_dev *dev,
+		   struct sockaddr_in *src_in,
+		   struct sockaddr_in *dst_in, u8 *dst_mac)
+{
+	__be32 src_ip = src_in->sin_addr.s_addr;
+	__be32 dst_ip = dst_in->sin_addr.s_addr;
+	struct neighbour *neigh = NULL;
+	struct rtable *rt = NULL;
+	int rc = 0;
+
+	rt = ip_route_output(&init_net, dst_ip, src_ip, 0, 0);
+	if (IS_ERR(rt)) {
+		DP_ERR(dev, "ip_route_output returned error\n");
+		return -EINVAL;
+	}
+
+	neigh = dst_neigh_lookup(&rt->dst, &dst_ip);
+
+	if (neigh) {
+		rcu_read_lock();
+		if (neigh->nud_state & NUD_VALID) {
+			ether_addr_copy(dst_mac, neigh->ha);
+			DP_DEBUG(dev, QEDR_MSG_QP, "mac_addr=[%pM]\n", dst_mac);
+		} else {
+			neigh_event_send(neigh, NULL);
+		}
+		rcu_read_unlock();
+		neigh_release(neigh);
+	}
+
+	ip_rt_put(rt);
+
+	return rc;
+}
+
+static int
+qedr_addr6_resolve(struct qedr_dev *dev,
+		   struct sockaddr_in6 *src_in,
+		   struct sockaddr_in6 *dst_in, u8 *dst_mac)
+{
+	struct neighbour *neigh = NULL;
+	struct dst_entry *dst;
+	struct flowi6 fl6;
+	int rc = 0;
+
+	memset(&fl6, 0, sizeof(fl6));
+	fl6.daddr = dst_in->sin6_addr;
+	fl6.saddr = src_in->sin6_addr;
+
+	dst = ip6_route_output(&init_net, NULL, &fl6);
+
+	if ((!dst) || dst->error) {
+		if (dst) {
+			dst_release(dst);
+			DP_ERR(dev,
+			       "ip6_route_output returned dst->error = %d\n",
+			       dst->error);
+		}
+		return -EINVAL;
+	}
+	neigh = dst_neigh_lookup(dst, &dst_in);
+
+	if (neigh) {
+		rcu_read_lock();
+		if (neigh->nud_state & NUD_VALID) {
+			ether_addr_copy(dst_mac, neigh->ha);
+			DP_DEBUG(dev, QEDR_MSG_QP, "mac_addr=[%pM]\n", dst_mac);
+		} else {
+			neigh_event_send(neigh, NULL);
+		}
+		rcu_read_unlock();
+		neigh_release(neigh);
+	}
+
+	dst_release(dst);
+
+	return rc;
+}
+
+int qedr_iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
+{
+	struct qedr_dev *dev = get_qedr_dev(cm_id->device);
+	struct qed_iwarp_connect_out out_params;
+	struct qed_iwarp_connect_in in_params;
+	struct qed_iwarp_cm_info *cm_info;
+	struct sockaddr_in6 *laddr6;
+	struct sockaddr_in6 *raddr6;
+	struct sockaddr_in *laddr;
+	struct sockaddr_in *raddr;
+	struct qedr_iw_ep *ep;
+	struct qedr_qp *qp;
+	int rc = 0;
+	int i;
+
+	qp = idr_find(&dev->qpidr, conn_param->qpn);
+
+	laddr = (struct sockaddr_in *)&cm_id->local_addr;
+	raddr = (struct sockaddr_in *)&cm_id->remote_addr;
+	laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr;
+	raddr6 = (struct sockaddr_in6 *)&cm_id->remote_addr;
+
+	DP_DEBUG(dev, QEDR_MSG_IWARP,
+		 "Connect source address: %pISpc, remote address: %pISpc\n",
+		 &cm_id->local_addr, &cm_id->remote_addr);
+
+	if (!laddr->sin_port || !raddr->sin_port)
+		return -EINVAL;
+
+	ep = kzalloc(sizeof(*ep), GFP_KERNEL);
+	if (!ep)
+		return -ENOMEM;
+
+	ep->dev = dev;
+	ep->qp = qp;
+	qp->ep = ep;
+	cm_id->add_ref(cm_id);
+	ep->cm_id = cm_id;
+
+	in_params.event_cb = qedr_iw_event_handler;
+	in_params.cb_context = ep;
+
+	cm_info = &in_params.cm_info;
+	memset(cm_info->local_ip, 0, sizeof(cm_info->local_ip));
+	memset(cm_info->remote_ip, 0, sizeof(cm_info->remote_ip));
+
+	if (cm_id->remote_addr.ss_family == AF_INET) {
+		cm_info->ip_version = QED_TCP_IPV4;
+
+		cm_info->remote_ip[0] = ntohl(raddr->sin_addr.s_addr);
+		cm_info->local_ip[0] = ntohl(laddr->sin_addr.s_addr);
+		cm_info->remote_port = ntohs(raddr->sin_port);
+		cm_info->local_port = ntohs(laddr->sin_port);
+		cm_info->vlan = qedr_iw_get_vlan_ipv4(dev, cm_info->local_ip);
+
+		rc = qedr_addr4_resolve(dev, laddr, raddr,
+					(u8 *)in_params.remote_mac_addr);
+
+		in_params.mss = dev->iwarp_max_mtu -
+		    (sizeof(struct iphdr) + sizeof(struct tcphdr));
+
+	} else {
+		in_params.cm_info.ip_version = QED_TCP_IPV6;
+
+		for (i = 0; i < 4; i++) {
+			cm_info->remote_ip[i] =
+			    ntohl(raddr6->sin6_addr.in6_u.u6_addr32[i]);
+			cm_info->local_ip[i] =
+			    ntohl(laddr6->sin6_addr.in6_u.u6_addr32[i]);
+		}
+
+		cm_info->local_port = ntohs(laddr6->sin6_port);
+		cm_info->remote_port = ntohs(raddr6->sin6_port);
+
+		in_params.mss = dev->iwarp_max_mtu -
+		    (sizeof(struct ipv6hdr) + sizeof(struct tcphdr));
+
+		cm_info->vlan = qedr_iw_get_vlan_ipv6(cm_info->local_ip);
+
+		rc = qedr_addr6_resolve(dev, laddr6, raddr6,
+					(u8 *)in_params.remote_mac_addr);
+	}
+	if (rc)
+		goto err;
+
+	DP_DEBUG(dev, QEDR_MSG_IWARP,
+		 "ord = %d ird=%d private_data=%p private_data_len=%d rq_psn=%d\n",
+		 conn_param->ord, conn_param->ird, conn_param->private_data,
+		 conn_param->private_data_len, qp->rq_psn);
+
+	cm_info->ord = conn_param->ord;
+	cm_info->ird = conn_param->ird;
+	cm_info->private_data = conn_param->private_data;
+	cm_info->private_data_len = conn_param->private_data_len;
+	in_params.qp = qp->qed_qp;
+	memcpy(in_params.local_mac_addr, dev->ndev->dev_addr, ETH_ALEN);
+
+	ep->during_connect = 1;
+	rc = dev->ops->iwarp_connect(dev->rdma_ctx, &in_params, &out_params);
+	if (rc)
+		goto err;
+
+	return rc;
+
+err:
+	cm_id->rem_ref(cm_id);
+	kfree(ep);
+	return rc;
+}
+
+int qedr_iw_create_listen(struct iw_cm_id *cm_id, int backlog)
+{
+	struct qedr_dev *dev = get_qedr_dev(cm_id->device);
+	struct qedr_iw_listener *listener;
+	struct qed_iwarp_listen_in iparams;
+	struct qed_iwarp_listen_out oparams;
+	struct sockaddr_in *laddr;
+	struct sockaddr_in6 *laddr6;
+	int rc;
+	int i;
+
+	laddr = (struct sockaddr_in *)&cm_id->local_addr;
+	laddr6 = (struct sockaddr_in6 *)&cm_id->local_addr;
+
+	DP_DEBUG(dev, QEDR_MSG_IWARP,
+		 "Create Listener address: %pISpc\n", &cm_id->local_addr);
+
+	listener = kzalloc(sizeof(*listener), GFP_KERNEL);
+	if (!listener)
+		return -ENOMEM;
+
+	listener->dev = dev;
+	cm_id->add_ref(cm_id);
+	listener->cm_id = cm_id;
+	listener->backlog = backlog;
+
+	iparams.cb_context = listener;
+	iparams.event_cb = qedr_iw_event_handler;
+	iparams.max_backlog = backlog;
+
+	if (cm_id->local_addr.ss_family == AF_INET) {
+		iparams.ip_version = QED_TCP_IPV4;
+		memset(iparams.ip_addr, 0, sizeof(iparams.ip_addr));
+
+		iparams.ip_addr[0] = ntohl(laddr->sin_addr.s_addr);
+		iparams.port = ntohs(laddr->sin_port);
+		iparams.vlan = qedr_iw_get_vlan_ipv4(dev, iparams.ip_addr);
+	} else {
+		iparams.ip_version = QED_TCP_IPV6;
+
+		for (i = 0; i < 4; i++) {
+			iparams.ip_addr[i] =
+			    ntohl(laddr6->sin6_addr.in6_u.u6_addr32[i]);
+		}
+
+		iparams.port = ntohs(laddr6->sin6_port);
+
+		iparams.vlan = qedr_iw_get_vlan_ipv6(iparams.ip_addr);
+	}
+	rc = dev->ops->iwarp_create_listen(dev->rdma_ctx, &iparams, &oparams);
+	if (rc)
+		goto err;
+
+	listener->qed_handle = oparams.handle;
+	cm_id->provider_data = listener;
+	return rc;
+
+err:
+	cm_id->rem_ref(cm_id);
+	kfree(listener);
+	return rc;
+}
+
+int qedr_iw_destroy_listen(struct iw_cm_id *cm_id)
+{
+	struct qedr_iw_listener *listener = cm_id->provider_data;
+	struct qedr_dev *dev = get_qedr_dev(cm_id->device);
+	int rc = 0;
+
+	if (listener->qed_handle)
+		rc = dev->ops->iwarp_destroy_listen(dev->rdma_ctx,
+						    listener->qed_handle);
+
+	cm_id->rem_ref(cm_id);
+	return rc;
+}
+
+int qedr_iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)cm_id->provider_data;
+	struct qedr_dev *dev = ep->dev;
+	struct qedr_qp *qp;
+	struct qed_iwarp_accept_in params;
+	int rc;
+
+	DP_DEBUG(dev, QEDR_MSG_IWARP, "Accept on qpid=%d\n", conn_param->qpn);
+
+	qp = idr_find(&dev->qpidr, conn_param->qpn);
+	if (!qp) {
+		DP_ERR(dev, "Invalid QP number %d\n", conn_param->qpn);
+		return -EINVAL;
+	}
+
+	ep->qp = qp;
+	qp->ep = ep;
+	cm_id->add_ref(cm_id);
+	ep->cm_id = cm_id;
+
+	params.ep_context = ep->qed_context;
+	params.cb_context = ep;
+	params.qp = ep->qp->qed_qp;
+	params.private_data = conn_param->private_data;
+	params.private_data_len = conn_param->private_data_len;
+	params.ird = conn_param->ird;
+	params.ord = conn_param->ord;
+
+	ep->during_connect = 1;
+	rc = dev->ops->iwarp_accept(dev->rdma_ctx, &params);
+	if (rc)
+		goto err;
+
+	return rc;
+err:
+	ep->during_connect = 0;
+	cm_id->rem_ref(cm_id);
+	return rc;
+}
+
+int qedr_iw_reject(struct iw_cm_id *cm_id, const void *pdata, u8 pdata_len)
+{
+	struct qedr_iw_ep *ep = (struct qedr_iw_ep *)cm_id->provider_data;
+	struct qedr_dev *dev = ep->dev;
+	struct qed_iwarp_reject_in params;
+
+	params.ep_context = ep->qed_context;
+	params.cb_context = ep;
+	params.private_data = pdata;
+	params.private_data_len = pdata_len;
+	ep->qp = NULL;
+
+	return dev->ops->iwarp_reject(dev->rdma_ctx, &params);
+}
+
 void qedr_iw_qp_add_ref(struct ib_qp *ibqp)
 {
 	struct qedr_qp *qp = get_qedr_qp(ibqp);
diff --git a/drivers/infiniband/hw/qedr/qedr_iw_cm.h b/drivers/infiniband/hw/qedr/qedr_iw_cm.h
index ea6d17af..08f4b10 100644
--- a/drivers/infiniband/hw/qedr/qedr_iw_cm.h
+++ b/drivers/infiniband/hw/qedr/qedr_iw_cm.h
@@ -30,6 +30,18 @@
  * SOFTWARE.
  */
 #include <rdma/iw_cm.h>
+
+int qedr_iw_connect(struct iw_cm_id *cm_id,
+		    struct iw_cm_conn_param *conn_param);
+
+int qedr_iw_create_listen(struct iw_cm_id *cm_id, int backlog);
+
+int qedr_iw_destroy_listen(struct iw_cm_id *cm_id);
+
+int qedr_iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param);
+
+int qedr_iw_reject(struct iw_cm_id *cm_id, const void *pdata, u8 pdata_len);
+
 void qedr_iw_qp_add_ref(struct ib_qp *qp);
 
 void qedr_iw_qp_rem_ref(struct ib_qp *qp);
diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index bdce96b..7067249 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -2260,6 +2260,23 @@ int qedr_destroy_qp(struct ib_qp *ibqp)
 			/* Change the QP state to ERROR */
 			qedr_modify_qp(ibqp, &attr, attr_mask, NULL);
 		}
+	} else {
+		/* Wait for the connect/accept to complete */
+		if (qp->ep) {
+			int wait_count = 1;
+
+			while (qp->ep->during_connect) {
+				DP_DEBUG(dev, QEDR_MSG_QP,
+					 "Still in during connect/accept\n");
+
+				msleep(100);
+				if (wait_count++ > 200) {
+					DP_NOTICE(dev,
+						  "during connect timeout\n");
+					break;
+				}
+			}
+		}
 	}
 
 	if (qp->qp_type == IB_QPT_GSI)
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

[PATCH for-next 8/8] RDMA/qedr: Add support for iWARP in user space

[Michal Kalderon]

Pass the second doorbell offset to userspace in create_qp response.
Pbl allocation is different for RoCE and iWARP and requires
different handling. RoCE allocated the pbl and passes the pointer
to qed, where-as in iWARP, qed allocates the pbl and qedr
populates it after it returns

Signed-off-by: Michal Kalderon <Michal.Kalderon-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ram Amrani <Ram.Amrani-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
Signed-off-by: Ariel Elior <Ariel.Elior-YGCgFSpz5w/QT0dZR+AlfA@public.gmane.org>
---
 drivers/infiniband/hw/qedr/verbs.c | 80 ++++++++++++++++++++++++++++++--------
 1 file changed, 63 insertions(+), 17 deletions(-)

diff --git a/drivers/infiniband/hw/qedr/verbs.c b/drivers/infiniband/hw/qedr/verbs.c
index 7067249..4e14a55 100644
--- a/drivers/infiniband/hw/qedr/verbs.c
+++ b/drivers/infiniband/hw/qedr/verbs.c
@@ -789,7 +789,8 @@ static inline int qedr_init_user_queue(struct ib_ucontext *ib_ctx,
 				       struct qedr_dev *dev,
 				       struct qedr_userq *q,
 				       u64 buf_addr, size_t buf_len,
-				       int access, int dmasync)
+				       int access, int dmasync,
+				       int alloc_and_init)
 {
 	u32 fw_pages;
 	int rc;
@@ -810,19 +811,25 @@ static inline int qedr_init_user_queue(struct ib_ucontext *ib_ctx,
 	if (rc)
 		goto err0;
 
-	q->pbl_tbl = qedr_alloc_pbl_tbl(dev, &q->pbl_info, GFP_KERNEL);
-	if (IS_ERR(q->pbl_tbl)) {
-		rc = PTR_ERR(q->pbl_tbl);
-		goto err0;
-	}
-
+	if (alloc_and_init) {
+		q->pbl_tbl = qedr_alloc_pbl_tbl(dev, &q->pbl_info, GFP_KERNEL);
+		if (IS_ERR(q->pbl_tbl)) {
+			rc = PTR_ERR(q->pbl_tbl);
+			goto err0;
+		}
 		qedr_populate_pbls(dev, q->umem, q->pbl_tbl, &q->pbl_info,
 				   FW_PAGE_SHIFT);
+	} else {
+		q->pbl_tbl = kzalloc(sizeof(*q->pbl_tbl), GFP_KERNEL);
+		if (!q->pbl_tbl)
+			goto err0;
+	}
 
 	return 0;
 
 err0:
 	ib_umem_release(q->umem);
+	q->umem = NULL;
 
 	return rc;
 }
@@ -948,7 +955,8 @@ struct ib_cq *qedr_create_cq(struct ib_device *ibdev,
 		cq->cq_type = QEDR_CQ_TYPE_USER;
 
 		rc = qedr_init_user_queue(ib_ctx, dev, &cq->q, ureq.addr,
-					  ureq.len, IB_ACCESS_LOCAL_WRITE, 1);
+					  ureq.len, IB_ACCESS_LOCAL_WRITE,
+					  1, 1);
 		if (rc)
 			goto err0;
 
@@ -1241,18 +1249,34 @@ static int qedr_check_qp_attrs(struct ib_pd *ibpd, struct qedr_dev *dev,
 	return 0;
 }
 
-static void qedr_copy_rq_uresp(struct qedr_create_qp_uresp *uresp,
+static void qedr_copy_rq_uresp(struct qedr_dev *dev,
+			       struct qedr_create_qp_uresp *uresp,
 			       struct qedr_qp *qp)
 {
-	uresp->rq_db_offset = DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_ROCE_RQ_PROD);
+	/* iWARP requires two doorbells per RQ. */
+	if (rdma_protocol_iwarp(&dev->ibdev, 1)) {
+		uresp->rq_db_offset =
+		    DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_IWARP_RQ_PROD);
+		uresp->rq_db2_offset = DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_FLAGS);
+	} else {
+		uresp->rq_db_offset =
+		    DB_ADDR_SHIFT(DQ_PWM_OFFSET_TCM_ROCE_RQ_PROD);
+	}
+
 	uresp->rq_icid = qp->icid;
 }
 
-static void qedr_copy_sq_uresp(struct qedr_create_qp_uresp *uresp,
+static void qedr_copy_sq_uresp(struct qedr_dev *dev,
+			       struct qedr_create_qp_uresp *uresp,
 			       struct qedr_qp *qp)
 {
 	uresp->sq_db_offset = DB_ADDR_SHIFT(DQ_PWM_OFFSET_XCM_RDMA_SQ_PROD);
-	uresp->sq_icid = qp->icid + 1;
+
+	/* iWARP uses the same cid for rq and sq */
+	if (rdma_protocol_iwarp(&dev->ibdev, 1))
+		uresp->sq_icid = qp->icid;
+	else
+		uresp->sq_icid = qp->icid + 1;
 }
 
 static int qedr_copy_qp_uresp(struct qedr_dev *dev,
@@ -1262,8 +1286,8 @@ static int qedr_copy_qp_uresp(struct qedr_dev *dev,
 	int rc;
 
 	memset(&uresp, 0, sizeof(uresp));
-	qedr_copy_sq_uresp(&uresp, qp);
-	qedr_copy_rq_uresp(&uresp, qp);
+	qedr_copy_sq_uresp(dev, &uresp, qp);
+	qedr_copy_rq_uresp(dev, &uresp, qp);
 
 	uresp.atomic_supported = dev->atomic_cap != IB_ATOMIC_NONE;
 	uresp.qp_id = qp->qp_id;
@@ -1381,6 +1405,25 @@ static void qedr_idr_remove(struct qedr_dev *dev, u32 id)
 	idr_remove(&dev->qpidr, id);
 	spin_unlock_irq(&dev->idr_lock);
 }
+
+static inline void
+qedr_iwarp_populate_user_qp(struct qedr_dev *dev,
+			    struct qedr_qp *qp,
+			    struct qed_rdma_create_qp_out_params *out_params)
+{
+	qp->usq.pbl_tbl->va = out_params->sq_pbl_virt;
+	qp->usq.pbl_tbl->pa = out_params->sq_pbl_phys;
+
+	qedr_populate_pbls(dev, qp->usq.umem, qp->usq.pbl_tbl,
+			   &qp->usq.pbl_info, FW_PAGE_SHIFT);
+
+	qp->urq.pbl_tbl->va = out_params->rq_pbl_virt;
+	qp->urq.pbl_tbl->pa = out_params->rq_pbl_phys;
+
+	qedr_populate_pbls(dev, qp->urq.umem, qp->urq.pbl_tbl,
+			   &qp->urq.pbl_info, FW_PAGE_SHIFT);
+}
+
 static void qedr_cleanup_user(struct qedr_dev *dev, struct qedr_qp *qp)
 {
 	if (qp->usq.umem)
@@ -1404,6 +1447,7 @@ static int qedr_create_user_qp(struct qedr_dev *dev,
 	struct ib_ucontext *ib_ctx = NULL;
 	struct qedr_ucontext *ctx = NULL;
 	struct qedr_create_qp_ureq ureq;
+	int alloc_and_init = rdma_protocol_roce(&dev->ibdev, 1);
 	int rc = -EINVAL;
 
 	ib_ctx = ibpd->uobject->context;
@@ -1418,14 +1462,13 @@ static int qedr_create_user_qp(struct qedr_dev *dev,
 
 	/* SQ - read access only (0), dma sync not required (0) */
 	rc = qedr_init_user_queue(ib_ctx, dev, &qp->usq, ureq.sq_addr,
-				  ureq.sq_len, 0, 0);
+				  ureq.sq_len, 0, 0, alloc_and_init);
 	if (rc)
 		return rc;
 
 	/* RQ - read access only (0), dma sync not required (0) */
 	rc = qedr_init_user_queue(ib_ctx, dev, &qp->urq, ureq.rq_addr,
-				  ureq.rq_len, 0, 0);
-
+				  ureq.rq_len, 0, 0, alloc_and_init);
 	if (rc)
 		return rc;
 
@@ -1446,6 +1489,9 @@ static int qedr_create_user_qp(struct qedr_dev *dev,
 		goto err1;
 	}
 
+	if (rdma_protocol_iwarp(&dev->ibdev, 1))
+		qedr_iwarp_populate_user_qp(dev, qp, &out_params);
+
 	qp->qp_id = out_params.qp_id;
 	qp->icid = out_params.icid;
 
-- 
1.8.3.1

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH for-next 0/8] RDMA/qedr: Add iWARP support for QL4xxxx

[Doug Ledford]

On Wed, 2017-07-26 at 14:41 +0300, Michal Kalderon wrote:
> Hi Doug,
> 
> This patch series adds iWARP support to our QL4xxxx networking
> adapters.
> 
> The first 6 patches of this series were previously sent as an RFC
> (patches 13-18): https://www.spinics.net/lists/linux-rdma/msg51416.ht
> ml
> The qed part of the RFC has already been accepted as part of 4.13:
> https://www.spinics.net/lists/netdev/msg443224.html
> 
> Changes from RFC:
> -----------------
> * Functions rdma_protocol_iwarp/roce, are used instead of
> IS_IWARP/IS_ROCE
> macros when differentiating flows after device registration.
> Pointed out by Leon
> * Additional patch is added for user space iWARP support.
> * Additional maintainer added for qedr for MAINTAINERS.
> 
> This series was tested and built against:
> git://github.com/dledford/linux.git, branch: k.o/for-next
> 
> Michal Kalderon (8):
>   RDMA/qedr: Add additional maintainer to MAINTAINERS file
>   RDMA/qedr: Rename the qedr_cm file as a preparation for iWARP
> support
>   RDMA/qedr: Add support for registering an iWARP device
>   RDMA/qedr: Add iWARP support in existing verbs
>   RDMA/qedr: Add support for read with invalidate, supported in iWARP
>   RDMA/qedr: Add iWARP connection management qp related callbacks
>   RDMA/qedr: Add iWARP connection management functions
>   RDMA/qedr: Add support for iWARP in user space
> 
>  MAINTAINERS                                        |   1 +
>  drivers/infiniband/hw/qedr/Makefile                |   2 +-
>  drivers/infiniband/hw/qedr/main.c                  | 105 ++-
>  drivers/infiniband/hw/qedr/qedr.h                  |  31 +-
>  drivers/infiniband/hw/qedr/qedr_hsi_rdma.h         |   6 +-
>  drivers/infiniband/hw/qedr/qedr_iw_cm.c            | 746
> +++++++++++++++++++++
>  drivers/infiniband/hw/qedr/qedr_iw_cm.h            |  49 ++
>  .../hw/qedr/{qedr_cm.c => qedr_roce_cm.c}          |   2 +-
>  .../hw/qedr/{qedr_cm.h => qedr_roce_cm.h}          |   0
>  drivers/infiniband/hw/qedr/verbs.c                 | 350 +++++++---
>  drivers/infiniband/hw/qedr/verbs.h                 |   2 +
>  11 files changed, 1203 insertions(+), 91 deletions(-)
>  create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.c
>  create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.h
>  rename drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c}
> (99%)
>  rename drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h}
> (100%)

I've pulled this into a branch and pushed that branch to github.  It's
the qedr branch.  There was some fuzz in taking the patches just due to
code float since you posted this.  Please review my branch and make
sure the final result is to your satisfaction.

-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG KeyID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD

--
To unsubscribe from this list: send the line "unsubscribe linux-rdma" in
the body of a message to majordomo-u79uwXL29TY76Z2rM5mHXA@public.gmane.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

RE: [PATCH for-next 0/8] RDMA/qedr: Add iWARP support for QL4xxxx

[Kalderon, Michal]

> From: Doug Ledford [mailto:dledford@redhat.com]
> Sent: Friday, August 18, 2017 7:33 PM
> 
> On Wed, 2017-07-26 at 14:41 +0300, Michal Kalderon wrote:
> > Hi Doug,
> >
> > This patch series adds iWARP support to our QL4xxxx networking
> > adapters.
> >
> > The first 6 patches of this series were previously sent as an RFC
> > (patches 13-18): https://www.spinics.net/lists/linux-rdma/msg51416.ht
> > ml
> > The qed part of the RFC has already been accepted as part of 4.13:
> > https://www.spinics.net/lists/netdev/msg443224.html
> >
> > Changes from RFC:
> > -----------------
> > * Functions rdma_protocol_iwarp/roce, are used instead of
> > IS_IWARP/IS_ROCE macros when differentiating flows after device
> > registration.
> > Pointed out by Leon
> > * Additional patch is added for user space iWARP support.
> > * Additional maintainer added for qedr for MAINTAINERS.
> >
> > This series was tested and built against:
> > git://github.com/dledford/linux.git, branch: k.o/for-next
> >
> > Michal Kalderon (8):
> >   RDMA/qedr: Add additional maintainer to MAINTAINERS file
> >   RDMA/qedr: Rename the qedr_cm file as a preparation for iWARP
> > support
> >   RDMA/qedr: Add support for registering an iWARP device
> >   RDMA/qedr: Add iWARP support in existing verbs
> >   RDMA/qedr: Add support for read with invalidate, supported in iWARP
> >   RDMA/qedr: Add iWARP connection management qp related callbacks
> >   RDMA/qedr: Add iWARP connection management functions
> >   RDMA/qedr: Add support for iWARP in user space
> >
> >  MAINTAINERS                                        |   1 +
> >  drivers/infiniband/hw/qedr/Makefile                |   2 +-
> >  drivers/infiniband/hw/qedr/main.c                  | 105 ++-
> >  drivers/infiniband/hw/qedr/qedr.h                  |  31 +-
> >  drivers/infiniband/hw/qedr/qedr_hsi_rdma.h         |   6 +-
> >  drivers/infiniband/hw/qedr/qedr_iw_cm.c            | 746
> > +++++++++++++++++++++
> >  drivers/infiniband/hw/qedr/qedr_iw_cm.h            |  49 ++
> >  .../hw/qedr/{qedr_cm.c => qedr_roce_cm.c}          |   2 +-
> >  .../hw/qedr/{qedr_cm.h => qedr_roce_cm.h}          |   0
> >  drivers/infiniband/hw/qedr/verbs.c                 | 350 +++++++---
> >  drivers/infiniband/hw/qedr/verbs.h                 |   2 +
> >  11 files changed, 1203 insertions(+), 91 deletions(-)  create mode
> > 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.c
> >  create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.h
> >  rename drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c}
> > (99%)
> >  rename drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h}
> > (100%)
> 
> I've pulled this into a branch and pushed that branch to github.  It's the qedr
> branch.  There was some fuzz in taking the patches just due to code float
> since you posted this.  Please review my branch and make sure the final
> result is to your satisfaction.

Thanks Doug! I looked at k.o/for-next-merged branch ( didn't see a qedr branch ) 
Qedr Files in this branch look good and as expected.
Could you please clarify which kernel version the  for-next-merged branch is intended for ? 4.14? 
I have additional iWARP patches for net-next, would like to know which kernel version to aim them for so that they
Don't make it in before this patch series. 

Thanks!
Michal

Re: [PATCH for-next 0/8] RDMA/qedr: Add iWARP support for QL4xxxx

[Doug Ledford]

[-- Attachment #1.1: Type: text/plain, Size: 4180 bytes --]

On 8/21/2017 10:29 AM, Kalderon, Michal wrote:
>> From: Doug Ledford [mailto:dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org]
>> Sent: Friday, August 18, 2017 7:33 PM
>>
>> On Wed, 2017-07-26 at 14:41 +0300, Michal Kalderon wrote:
>>> Hi Doug,
>>>
>>> This patch series adds iWARP support to our QL4xxxx networking
>>> adapters.
>>>
>>> The first 6 patches of this series were previously sent as an RFC
>>> (patches 13-18): https://www.spinics.net/lists/linux-rdma/msg51416.ht
>>> ml
>>> The qed part of the RFC has already been accepted as part of 4.13:
>>> https://www.spinics.net/lists/netdev/msg443224.html
>>>
>>> Changes from RFC:
>>> -----------------
>>> * Functions rdma_protocol_iwarp/roce, are used instead of
>>> IS_IWARP/IS_ROCE macros when differentiating flows after device
>>> registration.
>>> Pointed out by Leon
>>> * Additional patch is added for user space iWARP support.
>>> * Additional maintainer added for qedr for MAINTAINERS.
>>>
>>> This series was tested and built against:
>>> git://github.com/dledford/linux.git, branch: k.o/for-next
>>>
>>> Michal Kalderon (8):
>>>   RDMA/qedr: Add additional maintainer to MAINTAINERS file
>>>   RDMA/qedr: Rename the qedr_cm file as a preparation for iWARP
>>> support
>>>   RDMA/qedr: Add support for registering an iWARP device
>>>   RDMA/qedr: Add iWARP support in existing verbs
>>>   RDMA/qedr: Add support for read with invalidate, supported in iWARP
>>>   RDMA/qedr: Add iWARP connection management qp related callbacks
>>>   RDMA/qedr: Add iWARP connection management functions
>>>   RDMA/qedr: Add support for iWARP in user space
>>>
>>>  MAINTAINERS                                        |   1 +
>>>  drivers/infiniband/hw/qedr/Makefile                |   2 +-
>>>  drivers/infiniband/hw/qedr/main.c                  | 105 ++-
>>>  drivers/infiniband/hw/qedr/qedr.h                  |  31 +-
>>>  drivers/infiniband/hw/qedr/qedr_hsi_rdma.h         |   6 +-
>>>  drivers/infiniband/hw/qedr/qedr_iw_cm.c            | 746
>>> +++++++++++++++++++++
>>>  drivers/infiniband/hw/qedr/qedr_iw_cm.h            |  49 ++
>>>  .../hw/qedr/{qedr_cm.c => qedr_roce_cm.c}          |   2 +-
>>>  .../hw/qedr/{qedr_cm.h => qedr_roce_cm.h}          |   0
>>>  drivers/infiniband/hw/qedr/verbs.c                 | 350 +++++++---
>>>  drivers/infiniband/hw/qedr/verbs.h                 |   2 +
>>>  11 files changed, 1203 insertions(+), 91 deletions(-)  create mode
>>> 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.c
>>>  create mode 100644 drivers/infiniband/hw/qedr/qedr_iw_cm.h
>>>  rename drivers/infiniband/hw/qedr/{qedr_cm.c => qedr_roce_cm.c}
>>> (99%)
>>>  rename drivers/infiniband/hw/qedr/{qedr_cm.h => qedr_roce_cm.h}
>>> (100%)
>>
>> I've pulled this into a branch and pushed that branch to github.  It's the qedr
>> branch.  There was some fuzz in taking the patches just due to code float
>> since you posted this.  Please review my branch and make sure the final
>> result is to your satisfaction.
> 
> Thanks Doug! I looked at k.o/for-next-merged branch ( didn't see a qedr branch ) 
> Qedr Files in this branch look good and as expected.

Yeah.  That happens.  I had them in their own qedr branch, but when I
merged it to net-next-based, it was a straight fast-forward so no merge
commit was generated.  Then when I deleted the qedr branch, it's as
though it never existed.

> Could you please clarify which kernel version the  for-next-merged branch is intended for ? 4.14?

It is intended for 4.15.  I have it as an early check on the future
submission.

> I have additional iWARP patches for net-next, would like to know which kernel version to aim them for so that they
> Don't make it in before this patch series. 

Understood.  Think of anything based on net-next as a for-next-next
branch.  When I open for-next in the next submission cycle, I'll
ultimately create a branch on -rc2 of the next kernel, and merge this
branch in.


-- 
Doug Ledford <dledford-H+wXaHxf7aLQT0dZR+AlfA@public.gmane.org>
    GPG Key ID: B826A3330E572FDD
    Key fingerprint = AE6B 1BDA 122B 23B4 265B  1274 B826 A333 0E57 2FDD


[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 884 bytes --]