[PATCH rdma-next 10/29] IB/rxe: Interface to ib_core

[Moni Shoua <monis-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>]

Implementation of verbs interface as declared in InfiniBand core device
(struct ib_device).

Signed-off-by: Kamal Heib <kamalh-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Amir Vadai <amirv-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Signed-off-by: Moni Shoua <monis-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
Reviewed-by: Haggai Eran <haggaie-VPRAkNaXOzVWk0Htik3J/w@public.gmane.org>
---
 drivers/infiniband/hw/rxe/rxe_verbs.c | 1352 +++++++++++++++++++++++++++++++++
 drivers/infiniband/hw/rxe/rxe_verbs.h |  486 ++++++++++++
 2 files changed, 1838 insertions(+)
 create mode 100644 drivers/infiniband/hw/rxe/rxe_verbs.c
 create mode 100644 drivers/infiniband/hw/rxe/rxe_verbs.h

diff --git a/drivers/infiniband/hw/rxe/rxe_verbs.c b/drivers/infiniband/hw/rxe/rxe_verbs.c
new file mode 100644
index 0000000..71302a3
--- /dev/null
+++ b/drivers/infiniband/hw/rxe/rxe_verbs.c
@@ -0,0 +1,1352 @@
+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
+ *
+ * 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 "rxe.h"
+#include "rxe_loc.h"
+#include "rxe_queue.h"
+
+static int rxe_query_device(struct ib_device *dev,
+			    struct ib_device_attr *attr,
+			    struct ib_udata *uhw)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+
+	if (uhw->inlen || uhw->outlen)
+		return -EINVAL;
+
+	*attr = rxe->attr;
+	return 0;
+}
+
+static void rxe_eth_speed_to_ib_speed(int speed, u8 *active_speed,
+				      u8 *active_width)
+{
+	if (speed <= 1000) {
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_SDR;
+	} else if (speed <= 10000) {
+		*active_width = IB_WIDTH_1X;
+		*active_speed = IB_SPEED_FDR10;
+	} else if (speed <= 20000) {
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_DDR;
+	} else if (speed <= 30000) {
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_QDR;
+	} else if (speed <= 40000) {
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_FDR10;
+	} else {
+		*active_width = IB_WIDTH_4X;
+		*active_speed = IB_SPEED_EDR;
+	}
+}
+
+static int rxe_query_port(struct ib_device *dev,
+			  u8 port_num, struct ib_port_attr *attr)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+	struct rxe_port *port;
+	u32 speed;
+
+	if (unlikely(port_num < 1 || port_num > rxe->num_ports)) {
+		pr_warn("invalid port_number %d\n", port_num);
+		goto err1;
+	}
+
+	port = &rxe->port[port_num - 1];
+
+	*attr = port->attr;
+
+	mutex_lock(&rxe->usdev_lock);
+	if (rxe->ndev->ethtool_ops->get_link_ksettings) {
+		struct ethtool_link_ksettings ks;
+
+		rxe->ndev->ethtool_ops->get_link_ksettings(rxe->ndev, &ks);
+		speed = ks.base.speed;
+	} else if (rxe->ndev->ethtool_ops->get_settings) {
+		struct ethtool_cmd cmd;
+
+		rxe->ndev->ethtool_ops->get_settings(rxe->ndev, &cmd);
+		speed = cmd.speed;
+	} else {
+		pr_warn("%s speed is unknown, defaulting to 1000\n", rxe->ndev->name);
+		speed = 1000;
+	}
+	rxe_eth_speed_to_ib_speed(speed, &attr->active_speed, &attr->active_width);
+	mutex_unlock(&rxe->usdev_lock);
+
+	return 0;
+
+err1:
+	return -EINVAL;
+}
+
+static int rxe_query_gid(struct ib_device *device,
+			 u8 port_num, int index, union ib_gid *gid)
+{
+	int ret;
+
+	if (index > RXE_PORT_GID_TBL_LEN)
+		return -EINVAL;
+
+	ret = ib_get_cached_gid(device, port_num, index, gid, NULL);
+	if (ret == -EAGAIN) {
+		memcpy(gid, &zgid, sizeof(*gid));
+		return 0;
+	}
+
+	return ret;
+}
+
+static int rxe_add_gid(struct ib_device *device, u8 port_num, unsigned int
+		       index, const union ib_gid *gid,
+		       const struct ib_gid_attr *attr, void **context)
+{
+	if (index >= RXE_PORT_GID_TBL_LEN)
+		return -EINVAL;
+	return 0;
+}
+
+static int rxe_del_gid(struct ib_device *device, u8 port_num, unsigned int
+		       index, void **context)
+{
+	if (index >= RXE_PORT_GID_TBL_LEN)
+		return -EINVAL;
+	return 0;
+}
+
+static struct net_device *rxe_get_netdev(struct ib_device *device,
+					 u8 port_num)
+{
+	struct rxe_dev *rxe = to_rdev(device);
+
+	if (rxe->ndev) {
+		dev_hold(rxe->ndev);
+		return rxe->ndev;
+	}
+
+	return NULL;
+}
+
+static int rxe_query_pkey(struct ib_device *device,
+			  u8 port_num, u16 index, u16 *pkey)
+{
+	struct rxe_dev *rxe = to_rdev(device);
+	struct rxe_port *port;
+
+	if (unlikely(port_num < 1 || port_num > rxe->num_ports)) {
+		dev_warn(device->dma_device, "invalid port_num = %d\n",
+			 port_num);
+		goto err1;
+	}
+
+	port = &rxe->port[port_num - 1];
+
+	if (unlikely(index >= port->attr.pkey_tbl_len)) {
+		dev_warn(device->dma_device, "invalid index = %d\n",
+			 index);
+		goto err1;
+	}
+
+	*pkey = port->pkey_tbl[index];
+	return 0;
+
+err1:
+	return -EINVAL;
+}
+
+static int rxe_modify_device(struct ib_device *dev,
+			     int mask, struct ib_device_modify *attr)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+
+	if (mask & IB_DEVICE_MODIFY_SYS_IMAGE_GUID)
+		rxe->attr.sys_image_guid = cpu_to_be64(attr->sys_image_guid);
+
+	if (mask & IB_DEVICE_MODIFY_NODE_DESC) {
+		memcpy(rxe->ib_dev.node_desc,
+		       attr->node_desc, sizeof(rxe->ib_dev.node_desc));
+	}
+
+	return 0;
+}
+
+static int rxe_modify_port(struct ib_device *dev,
+			   u8 port_num, int mask, struct ib_port_modify *attr)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+	struct rxe_port *port;
+
+	if (unlikely(port_num < 1 || port_num > rxe->num_ports)) {
+		pr_warn("invalid port_num = %d\n", port_num);
+		goto err1;
+	}
+
+	port = &rxe->port[port_num - 1];
+
+	port->attr.port_cap_flags |= attr->set_port_cap_mask;
+	port->attr.port_cap_flags &= ~attr->clr_port_cap_mask;
+
+	if (mask & IB_PORT_RESET_QKEY_CNTR)
+		port->attr.qkey_viol_cntr = 0;
+
+	return 0;
+
+err1:
+	return -EINVAL;
+}
+
+static enum rdma_link_layer rxe_get_link_layer(struct ib_device *dev,
+					       u8 port_num)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+
+	return rxe->ifc_ops->link_layer(rxe, port_num);
+}
+
+static struct ib_ucontext *rxe_alloc_ucontext(struct ib_device *dev,
+					      struct ib_udata *udata)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+	struct rxe_ucontext *uc;
+
+	uc = rxe_alloc(&rxe->uc_pool);
+	return uc ? &uc->ibuc : ERR_PTR(-ENOMEM);
+}
+
+static int rxe_dealloc_ucontext(struct ib_ucontext *ibuc)
+{
+	struct rxe_ucontext *uc = to_ruc(ibuc);
+
+	rxe_drop_ref(uc);
+	return 0;
+}
+
+static int rxe_port_immutable(struct ib_device *dev, u8 port_num,
+			      struct ib_port_immutable *immutable)
+{
+	int err;
+	struct ib_port_attr attr;
+
+	err = rxe_query_port(dev, 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_UDP_ENCAP;
+	immutable->max_mad_size = IB_MGMT_MAD_SIZE;
+
+	return 0;
+}
+
+static struct ib_pd *rxe_alloc_pd(struct ib_device *dev,
+				  struct ib_ucontext *context,
+				  struct ib_udata *udata)
+{
+	struct rxe_dev *rxe = to_rdev(dev);
+	struct rxe_pd *pd;
+
+	pd = rxe_alloc(&rxe->pd_pool);
+	return pd ? &pd->ibpd : ERR_PTR(-ENOMEM);
+}
+
+static int rxe_dealloc_pd(struct ib_pd *ibpd)
+{
+	struct rxe_pd *pd = to_rpd(ibpd);
+
+	rxe_drop_ref(pd);
+	return 0;
+}
+
+static int rxe_init_av(struct rxe_dev *rxe, struct ib_ah_attr *attr,
+		       struct rxe_av *av)
+{
+	int err;
+	union ib_gid sgid;
+	struct ib_gid_attr sgid_attr;
+
+	err = ib_get_cached_gid(&rxe->ib_dev, attr->port_num,
+				attr->grh.sgid_index, &sgid,
+				&sgid_attr);
+	if (err) {
+		pr_err("Failed to query sgid. err = %d\n", err);
+		return err;
+	}
+
+	err = rxe_av_from_attr(rxe, attr->port_num, av, attr);
+	if (!err)
+		err = rxe_av_fill_ip_info(rxe, av, attr, &sgid_attr, &sgid);
+
+	if (sgid_attr.ndev)
+		dev_put(sgid_attr.ndev);
+	return err;
+}
+
+static struct ib_ah *rxe_create_ah(struct ib_pd *ibpd, struct ib_ah_attr *attr)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_ah *ah;
+
+	err = rxe_av_chk_attr(rxe, attr);
+	if (err)
+		goto err1;
+
+	ah = rxe_alloc(&rxe->ah_pool);
+	if (!ah) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	rxe_add_ref(pd);
+	ah->pd = pd;
+
+	err = rxe_init_av(rxe, attr, &ah->av);
+	if (err)
+		goto err2;
+
+	return &ah->ibah;
+
+err2:
+	rxe_drop_ref(pd);
+	rxe_drop_ref(ah);
+err1:
+	return ERR_PTR(err);
+}
+
+static int rxe_modify_ah(struct ib_ah *ibah, struct ib_ah_attr *attr)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibah->device);
+	struct rxe_ah *ah = to_rah(ibah);
+
+	err = rxe_av_chk_attr(rxe, attr);
+	if (err)
+		return err;
+
+	err = rxe_init_av(rxe, attr, &ah->av);
+	if (err)
+		return err;
+
+	return 0;
+}
+
+static int rxe_query_ah(struct ib_ah *ibah, struct ib_ah_attr *attr)
+{
+	struct rxe_dev *rxe = to_rdev(ibah->device);
+	struct rxe_ah *ah = to_rah(ibah);
+
+	rxe_av_to_attr(rxe, &ah->av, attr);
+	return 0;
+}
+
+static int rxe_destroy_ah(struct ib_ah *ibah)
+{
+	struct rxe_ah *ah = to_rah(ibah);
+
+	rxe_drop_ref(ah->pd);
+	rxe_drop_ref(ah);
+	return 0;
+}
+
+static int post_one_recv(struct rxe_rq *rq, struct ib_recv_wr *ibwr)
+{
+	int err;
+	int i;
+	u32 length;
+	struct rxe_recv_wqe *recv_wqe;
+	int num_sge = ibwr->num_sge;
+
+	if (unlikely(queue_full(rq->queue))) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	if (unlikely(num_sge > rq->max_sge)) {
+		err = -EINVAL;
+		goto err1;
+	}
+
+	length = 0;
+	for (i = 0; i < num_sge; i++)
+		length += ibwr->sg_list[i].length;
+
+	recv_wqe = producer_addr(rq->queue);
+	recv_wqe->wr_id = ibwr->wr_id;
+	recv_wqe->num_sge = num_sge;
+
+	memcpy(recv_wqe->dma.sge, ibwr->sg_list,
+	       num_sge * sizeof(struct ib_sge));
+
+	recv_wqe->dma.length		= length;
+	recv_wqe->dma.resid		= length;
+	recv_wqe->dma.num_sge		= num_sge;
+	recv_wqe->dma.cur_sge		= 0;
+	recv_wqe->dma.sge_offset	= 0;
+
+	/* make sure all changes to the work queue are written before we
+	 * update the producer pointer
+	 */
+	smp_wmb();
+
+	advance_producer(rq->queue);
+	return 0;
+
+err1:
+	return err;
+}
+
+static struct ib_srq *rxe_create_srq(struct ib_pd *ibpd,
+				     struct ib_srq_init_attr *init,
+				     struct ib_udata *udata)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_srq *srq;
+	struct ib_ucontext *context = udata ? ibpd->uobject->context : NULL;
+
+	err = rxe_srq_chk_attr(rxe, NULL, &init->attr, IB_SRQ_INIT_MASK);
+	if (err)
+		goto err1;
+
+	srq = rxe_alloc(&rxe->srq_pool);
+	if (!srq) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	rxe_add_index(srq);
+	rxe_add_ref(pd);
+	srq->pd = pd;
+
+	err = rxe_srq_from_init(rxe, srq, init, context, udata);
+	if (err)
+		goto err2;
+
+	return &srq->ibsrq;
+
+err2:
+	rxe_drop_ref(pd);
+	rxe_drop_index(srq);
+	rxe_drop_ref(srq);
+err1:
+	return ERR_PTR(err);
+}
+
+static int rxe_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
+			  enum ib_srq_attr_mask mask,
+			  struct ib_udata *udata)
+{
+	int err;
+	struct rxe_srq *srq = to_rsrq(ibsrq);
+	struct rxe_dev *rxe = to_rdev(ibsrq->device);
+
+	err = rxe_srq_chk_attr(rxe, srq, attr, mask);
+	if (err)
+		goto err1;
+
+	err = rxe_srq_from_attr(rxe, srq, attr, mask, udata);
+	if (err)
+		goto err1;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+static int rxe_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr)
+{
+	struct rxe_srq *srq = to_rsrq(ibsrq);
+
+	if (srq->error)
+		return -EINVAL;
+
+	attr->max_wr = srq->rq.queue->buf->index_mask;
+	attr->max_sge = srq->rq.max_sge;
+	attr->srq_limit = srq->limit;
+	return 0;
+}
+
+static int rxe_destroy_srq(struct ib_srq *ibsrq)
+{
+	struct rxe_srq *srq = to_rsrq(ibsrq);
+
+	if (srq->rq.queue)
+		rxe_queue_cleanup(srq->rq.queue);
+
+	rxe_drop_ref(srq->pd);
+	rxe_drop_index(srq);
+	rxe_drop_ref(srq);
+
+	return 0;
+}
+
+static int rxe_post_srq_recv(struct ib_srq *ibsrq, struct ib_recv_wr *wr,
+			     struct ib_recv_wr **bad_wr)
+{
+	int err = 0;
+	unsigned long flags;
+	struct rxe_srq *srq = to_rsrq(ibsrq);
+
+	spin_lock_irqsave(&srq->rq.producer_lock, flags);
+
+	while (wr) {
+		err = post_one_recv(&srq->rq, wr);
+		if (unlikely(err))
+			break;
+		wr = wr->next;
+	}
+
+	spin_unlock_irqrestore(&srq->rq.producer_lock, flags);
+
+	if (err)
+		*bad_wr = wr;
+
+	return err;
+}
+
+static struct ib_qp *rxe_create_qp(struct ib_pd *ibpd,
+				   struct ib_qp_init_attr *init,
+				   struct ib_udata *udata)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_qp *qp;
+
+	err = rxe_qp_chk_init(rxe, init);
+	if (err)
+		goto err1;
+
+	qp = rxe_alloc(&rxe->qp_pool);
+	if (!qp) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	if (udata) {
+		if (udata->inlen) {
+			err = -EINVAL;
+			goto err1;
+		}
+		qp->is_user = 1;
+	}
+
+	rxe_add_index(qp);
+
+	err = rxe_qp_from_init(rxe, qp, pd, init, udata, ibpd);
+	if (err)
+		goto err2;
+
+	return &qp->ibqp;
+
+err2:
+	rxe_drop_index(qp);
+	rxe_drop_ref(qp);
+err1:
+	return ERR_PTR(err);
+}
+
+static int rxe_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+			 int mask, struct ib_udata *udata)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibqp->device);
+	struct rxe_qp *qp = to_rqp(ibqp);
+
+	err = rxe_qp_chk_attr(rxe, qp, attr, mask);
+	if (err)
+		goto err1;
+
+	err = rxe_qp_from_attr(qp, attr, mask, udata);
+	if (err)
+		goto err1;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+static int rxe_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+			int mask, struct ib_qp_init_attr *init)
+{
+	struct rxe_qp *qp = to_rqp(ibqp);
+
+	rxe_qp_to_init(qp, init);
+	rxe_qp_to_attr(qp, attr, mask);
+
+	return 0;
+}
+
+static int rxe_destroy_qp(struct ib_qp *ibqp)
+{
+	struct rxe_qp *qp = to_rqp(ibqp);
+
+	rxe_qp_destroy(qp);
+	rxe_drop_index(qp);
+	rxe_drop_ref(qp);
+	return 0;
+}
+
+static int validate_send_wr(struct rxe_qp *qp, struct ib_send_wr *ibwr,
+			    unsigned int mask, unsigned int length)
+{
+	int num_sge = ibwr->num_sge;
+	struct rxe_sq *sq = &qp->sq;
+
+	if (unlikely(num_sge > sq->max_sge))
+		goto err1;
+
+	if (unlikely(mask & WR_ATOMIC_MASK)) {
+		if (length < 8)
+			goto err1;
+
+		if (atomic_wr(ibwr)->remote_addr & 0x7)
+			goto err1;
+	}
+
+	if (unlikely((ibwr->send_flags & IB_SEND_INLINE) &&
+		     (length > sq->max_inline)))
+		goto err1;
+
+	return 0;
+
+err1:
+	return -EINVAL;
+}
+
+static void init_send_wr(struct rxe_qp *qp, struct rxe_send_wr *wr,
+			 struct ib_send_wr *ibwr)
+{
+	wr->wr_id = ibwr->wr_id;
+	wr->num_sge = ibwr->num_sge;
+	wr->opcode = ibwr->opcode;
+	wr->send_flags = ibwr->send_flags;
+
+	if (qp_type(qp) == IB_QPT_UD ||
+	    qp_type(qp) == IB_QPT_SMI ||
+	    qp_type(qp) == IB_QPT_GSI) {
+		wr->wr.ud.remote_qpn = ud_wr(ibwr)->remote_qpn;
+		wr->wr.ud.remote_qkey = ud_wr(ibwr)->remote_qkey;
+		if (qp_type(qp) == IB_QPT_GSI)
+			wr->wr.ud.pkey_index = ud_wr(ibwr)->pkey_index;
+		if (wr->opcode == IB_WR_SEND_WITH_IMM)
+			wr->ex.imm_data = ibwr->ex.imm_data;
+	} else {
+		switch (wr->opcode) {
+		case IB_WR_RDMA_WRITE_WITH_IMM:
+			wr->ex.imm_data = ibwr->ex.imm_data;
+		case IB_WR_RDMA_READ:
+		case IB_WR_RDMA_WRITE:
+			wr->wr.rdma.remote_addr = rdma_wr(ibwr)->remote_addr;
+			wr->wr.rdma.rkey	= rdma_wr(ibwr)->rkey;
+			break;
+		case IB_WR_SEND_WITH_IMM:
+			wr->ex.imm_data = ibwr->ex.imm_data;
+			break;
+		case IB_WR_SEND_WITH_INV:
+			wr->ex.invalidate_rkey = ibwr->ex.invalidate_rkey;
+			break;
+		case IB_WR_ATOMIC_CMP_AND_SWP:
+		case IB_WR_ATOMIC_FETCH_AND_ADD:
+			wr->wr.atomic.remote_addr =
+				atomic_wr(ibwr)->remote_addr;
+			wr->wr.atomic.compare_add =
+				atomic_wr(ibwr)->compare_add;
+			wr->wr.atomic.swap = atomic_wr(ibwr)->swap;
+			wr->wr.atomic.rkey = atomic_wr(ibwr)->rkey;
+			break;
+		default:
+			break;
+		}
+	}
+}
+
+static int init_send_wqe(struct rxe_qp *qp, struct ib_send_wr *ibwr,
+			 unsigned int mask, unsigned int length,
+			 struct rxe_send_wqe *wqe)
+{
+	int num_sge = ibwr->num_sge;
+	struct ib_sge *sge;
+	int i;
+	u8 *p;
+
+	init_send_wr(qp, &wqe->wr, ibwr);
+
+	if (qp_type(qp) == IB_QPT_UD ||
+	    qp_type(qp) == IB_QPT_SMI ||
+	    qp_type(qp) == IB_QPT_GSI)
+		memcpy(&wqe->av, &to_rah(ud_wr(ibwr)->ah)->av, sizeof(wqe->av));
+
+	if (unlikely(ibwr->send_flags & IB_SEND_INLINE)) {
+		p = wqe->dma.inline_data;
+
+		sge = ibwr->sg_list;
+		for (i = 0; i < num_sge; i++, sge++) {
+			if (qp->is_user && copy_from_user(p, (__user void *)
+					    (uintptr_t)sge->addr, sge->length))
+				return -EFAULT;
+
+			else if (!qp->is_user)
+				memcpy(p, (void *)(uintptr_t)sge->addr,
+				       sge->length);
+
+			p += sge->length;
+		}
+	} else
+		memcpy(wqe->dma.sge, ibwr->sg_list,
+		       num_sge * sizeof(struct ib_sge));
+
+	wqe->iova		= (mask & WR_ATOMIC_MASK) ?
+					atomic_wr(ibwr)->remote_addr :
+					rdma_wr(ibwr)->remote_addr;
+	wqe->mask		= mask;
+	wqe->dma.length		= length;
+	wqe->dma.resid		= length;
+	wqe->dma.num_sge	= num_sge;
+	wqe->dma.cur_sge	= 0;
+	wqe->dma.sge_offset	= 0;
+	wqe->state		= wqe_state_posted;
+	wqe->ssn		= atomic_add_return(1, &qp->ssn);
+
+	return 0;
+}
+
+static int post_one_send(struct rxe_qp *qp, struct ib_send_wr *ibwr,
+			 unsigned mask, u32 length)
+{
+	int err;
+	struct rxe_sq *sq = &qp->sq;
+	struct rxe_send_wqe *send_wqe;
+	unsigned long flags;
+
+	err = validate_send_wr(qp, ibwr, mask, length);
+	if (err)
+		return err;
+
+	spin_lock_irqsave(&qp->sq.sq_lock, flags);
+
+	if (unlikely(queue_full(sq->queue))) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	send_wqe = producer_addr(sq->queue);
+
+	err = init_send_wqe(qp, ibwr, mask, length, send_wqe);
+	if (unlikely(err))
+		goto err1;
+
+	/*
+	 * make sure all changes to the work queue are
+	 * written before we update the producer pointer
+	 */
+	smp_wmb();
+
+	advance_producer(sq->queue);
+	spin_unlock_irqrestore(&qp->sq.sq_lock, flags);
+
+	return 0;
+
+err1:
+	spin_unlock_irqrestore(&qp->sq.sq_lock, flags);
+	return err;
+}
+
+static int rxe_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+			 struct ib_send_wr **bad_wr)
+{
+	int err = 0;
+	struct rxe_qp *qp = to_rqp(ibqp);
+	unsigned int mask;
+	unsigned int length = 0;
+	int i;
+	int must_sched;
+
+	if (unlikely(!qp->valid)) {
+		*bad_wr = wr;
+		return -EINVAL;
+	}
+
+	if (unlikely(qp->req.state < QP_STATE_READY)) {
+		*bad_wr = wr;
+		return -EINVAL;
+	}
+
+	while (wr) {
+		mask = wr_opcode_mask(wr->opcode, qp);
+		if (unlikely(!mask)) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			break;
+		}
+
+		if (unlikely((wr->send_flags & IB_SEND_INLINE) &&
+			     !(mask & WR_INLINE_MASK))) {
+			err = -EINVAL;
+			*bad_wr = wr;
+			break;
+		}
+
+		length = 0;
+		for (i = 0; i < wr->num_sge; i++)
+			length += wr->sg_list[i].length;
+
+		err = post_one_send(qp, wr, mask, length);
+
+		if (err) {
+			*bad_wr = wr;
+			break;
+		}
+		wr = wr->next;
+	}
+
+	/*
+	 * Must sched in case of GSI QP because ib_send_mad() hold irq lock,
+	 * and the requester call ip_local_out_sk() that takes spin_lock_bh.
+	 */
+	must_sched = (qp_type(qp) == IB_QPT_GSI) ||
+			(queue_count(qp->sq.queue) > 1);
+
+	rxe_run_task(&qp->req.task, must_sched);
+
+	return err;
+}
+
+static int rxe_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+			 struct ib_recv_wr **bad_wr)
+{
+	int err = 0;
+	struct rxe_qp *qp = to_rqp(ibqp);
+	struct rxe_rq *rq = &qp->rq;
+	unsigned long flags;
+
+	if (unlikely((qp_state(qp) < IB_QPS_INIT) || !qp->valid)) {
+		*bad_wr = wr;
+		err = -EINVAL;
+		goto err1;
+	}
+
+	if (unlikely(qp->srq)) {
+		*bad_wr = wr;
+		err = -EINVAL;
+		goto err1;
+	}
+
+	spin_lock_irqsave(&rq->producer_lock, flags);
+
+	while (wr) {
+		err = post_one_recv(rq, wr);
+		if (unlikely(err)) {
+			*bad_wr = wr;
+			break;
+		}
+		wr = wr->next;
+	}
+
+	spin_unlock_irqrestore(&rq->producer_lock, flags);
+
+err1:
+	return err;
+}
+
+static struct ib_cq *rxe_create_cq(struct ib_device *dev,
+				   const struct ib_cq_init_attr *attr,
+				   struct ib_ucontext *context,
+				   struct ib_udata *udata)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(dev);
+	struct rxe_cq *cq;
+
+	if (attr->flags)
+		return ERR_PTR(-EINVAL);
+
+	err = rxe_cq_chk_attr(rxe, NULL, attr->cqe, attr->comp_vector, udata);
+	if (err)
+		goto err1;
+
+	cq = rxe_alloc(&rxe->cq_pool);
+	if (!cq) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	err = rxe_cq_from_init(rxe, cq, attr->cqe, attr->comp_vector,
+			       context, udata);
+	if (err)
+		goto err2;
+
+	return &cq->ibcq;
+
+err2:
+	rxe_drop_ref(cq);
+err1:
+	return ERR_PTR(err);
+}
+
+static int rxe_destroy_cq(struct ib_cq *ibcq)
+{
+	struct rxe_cq *cq = to_rcq(ibcq);
+
+	rxe_drop_ref(cq);
+	return 0;
+}
+
+static int rxe_resize_cq(struct ib_cq *ibcq, int cqe, struct ib_udata *udata)
+{
+	int err;
+	struct rxe_cq *cq = to_rcq(ibcq);
+	struct rxe_dev *rxe = to_rdev(ibcq->device);
+
+	err = rxe_cq_chk_attr(rxe, cq, cqe, 0, udata);
+	if (err)
+		goto err1;
+
+	err = rxe_cq_resize_queue(cq, cqe, udata);
+	if (err)
+		goto err1;
+
+	return 0;
+
+err1:
+	return err;
+}
+
+static int rxe_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
+{
+	int i;
+	struct rxe_cq *cq = to_rcq(ibcq);
+	struct rxe_cqe *cqe;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cq->cq_lock, flags);
+	for (i = 0; i < num_entries; i++) {
+		cqe = queue_head(cq->queue);
+		if (!cqe)
+			break;
+
+		memcpy(wc++, &cqe->ibwc, sizeof(*wc));
+		advance_consumer(cq->queue);
+	}
+	spin_unlock_irqrestore(&cq->cq_lock, flags);
+
+	return i;
+}
+
+static int rxe_peek_cq(struct ib_cq *ibcq, int wc_cnt)
+{
+	struct rxe_cq *cq = to_rcq(ibcq);
+	int count = queue_count(cq->queue);
+
+	return (count > wc_cnt) ? wc_cnt : count;
+}
+
+static int rxe_req_notify_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags)
+{
+	struct rxe_cq *cq = to_rcq(ibcq);
+
+	if (cq->notify != IB_CQ_NEXT_COMP)
+		cq->notify = flags & IB_CQ_SOLICITED_MASK;
+
+	return 0;
+}
+
+static struct ib_mr *rxe_get_dma_mr(struct ib_pd *ibpd, int access)
+{
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_mem *mr;
+	int err;
+
+	mr = rxe_alloc(&rxe->mr_pool);
+	if (!mr) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	rxe_add_index(mr);
+
+	rxe_add_ref(pd);
+
+	err = rxe_mem_init_dma(rxe, pd, access, mr);
+	if (err)
+		goto err2;
+
+	return &mr->ibmr;
+
+err2:
+	rxe_drop_ref(pd);
+	rxe_drop_index(mr);
+	rxe_drop_ref(mr);
+err1:
+	return ERR_PTR(err);
+}
+
+static struct ib_mr *rxe_reg_user_mr(struct ib_pd *ibpd,
+				     u64 start,
+				     u64 length,
+				     u64 iova,
+				     int access, struct ib_udata *udata)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_mem *mr;
+
+	mr = rxe_alloc(&rxe->mr_pool);
+	if (!mr) {
+		err = -ENOMEM;
+		goto err2;
+	}
+
+	rxe_add_index(mr);
+
+	rxe_add_ref(pd);
+
+	err = rxe_mem_init_user(rxe, pd, start, length, iova,
+				access, udata, mr);
+	if (err)
+		goto err3;
+
+	return &mr->ibmr;
+
+err3:
+	rxe_drop_ref(pd);
+	rxe_drop_index(mr);
+	rxe_drop_ref(mr);
+err2:
+	return ERR_PTR(err);
+}
+
+static int rxe_dereg_mr(struct ib_mr *ibmr)
+{
+	struct rxe_mem *mr = to_rmr(ibmr);
+
+	mr->state = RXE_MEM_STATE_ZOMBIE;
+	rxe_drop_ref(mr->pd);
+	rxe_drop_index(mr);
+	rxe_drop_ref(mr);
+	return 0;
+}
+
+static struct ib_mr *rxe_alloc_mr(struct ib_pd *ibpd,
+				  enum ib_mr_type mr_type,
+				  u32 max_num_sg)
+{
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_mem *mr;
+	int err;
+
+	if (mr_type != IB_MR_TYPE_MEM_REG)
+		return ERR_PTR(-EINVAL);
+
+	mr = rxe_alloc(&rxe->mr_pool);
+	if (!mr) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	rxe_add_index(mr);
+
+	rxe_add_ref(pd);
+
+	err = rxe_mem_init_fast(rxe, pd, max_num_sg, mr);
+	if (err)
+		goto err2;
+
+	return &mr->ibmr;
+
+err2:
+	rxe_drop_ref(pd);
+	rxe_drop_index(mr);
+	rxe_drop_ref(mr);
+err1:
+	return ERR_PTR(err);
+}
+
+static struct ib_fmr *rxe_alloc_fmr(struct ib_pd *ibpd,
+				    int access, struct ib_fmr_attr *attr)
+{
+	struct rxe_dev *rxe = to_rdev(ibpd->device);
+	struct rxe_pd *pd = to_rpd(ibpd);
+	struct rxe_mem *fmr;
+	int err;
+
+	fmr = rxe_alloc(&rxe->fmr_pool);
+	if (!fmr) {
+		err = -ENOMEM;
+		goto err1;
+	}
+
+	rxe_add_index(fmr);
+
+	rxe_add_ref(pd);
+
+	err = rxe_mem_init_fmr(rxe, pd, access, attr, fmr);
+	if (err)
+		goto err2;
+
+	return &fmr->ibfmr;
+
+err2:
+	rxe_drop_ref(pd);
+	rxe_drop_index(fmr);
+	rxe_drop_ref(fmr);
+err1:
+	return ERR_PTR(err);
+}
+
+static int rxe_map_phys_fmr(struct ib_fmr *ibfmr,
+			    u64 *page_list, int list_length, u64 iova)
+{
+	struct rxe_mem *fmr = to_rfmr(ibfmr);
+	struct rxe_dev *rxe = to_rdev(ibfmr->device);
+
+	return rxe_mem_map_pages(rxe, fmr, page_list, list_length, iova);
+}
+
+static int rxe_unmap_fmr(struct list_head *fmr_list)
+{
+	struct rxe_mem *fmr;
+
+	list_for_each_entry(fmr, fmr_list, ibfmr.list) {
+		if (fmr->state != RXE_MEM_STATE_VALID)
+			continue;
+
+		fmr->va = 0;
+		fmr->iova = 0;
+		fmr->length = 0;
+		fmr->num_buf = 0;
+		fmr->state = RXE_MEM_STATE_FREE;
+	}
+
+	return 0;
+}
+
+static int rxe_dealloc_fmr(struct ib_fmr *ibfmr)
+{
+	struct rxe_mem *fmr = to_rfmr(ibfmr);
+
+	fmr->state = RXE_MEM_STATE_ZOMBIE;
+	rxe_drop_ref(fmr->pd);
+	rxe_drop_index(fmr);
+	rxe_drop_ref(fmr);
+	return 0;
+}
+
+static int rxe_attach_mcast(struct ib_qp *ibqp, union ib_gid *mgid, u16 mlid)
+{
+	int err;
+	struct rxe_dev *rxe = to_rdev(ibqp->device);
+	struct rxe_qp *qp = to_rqp(ibqp);
+	struct rxe_mc_grp *grp;
+
+	/* takes a ref on grp if successful */
+	err = rxe_mcast_get_grp(rxe, mgid, &grp);
+	if (err)
+		return err;
+
+	err = rxe_mcast_add_grp_elem(rxe, qp, grp);
+
+	rxe_drop_ref(grp);
+	return err;
+}
+
+static int rxe_detach_mcast(struct ib_qp *ibqp, union ib_gid *mgid, u16 mlid)
+{
+	struct rxe_dev *rxe = to_rdev(ibqp->device);
+	struct rxe_qp *qp = to_rqp(ibqp);
+
+	return rxe_mcast_drop_grp_elem(rxe, qp, mgid);
+}
+
+static ssize_t rxe_show_parent(struct device *device,
+			       struct device_attribute *attr, char *buf)
+{
+	struct rxe_dev *rxe = container_of(device, struct rxe_dev,
+					   ib_dev.dev);
+	char *name;
+
+	name = rxe->ifc_ops->parent_name(rxe, 1);
+	return snprintf(buf, 16, "%s\n", name);
+}
+
+static DEVICE_ATTR(parent, S_IRUGO, rxe_show_parent, NULL);
+
+static struct device_attribute *rxe_dev_attributes[] = {
+	&dev_attr_parent,
+};
+
+int rxe_register_device(struct rxe_dev *rxe)
+{
+	int err;
+	int i;
+	struct ib_device *dev = &rxe->ib_dev;
+
+	strlcpy(dev->name, "rxe%d", IB_DEVICE_NAME_MAX);
+	strlcpy(dev->node_desc, "rxe", sizeof(dev->node_desc));
+
+	dev->owner = THIS_MODULE;
+	dev->node_type = RDMA_NODE_IB_CA;
+	dev->phys_port_cnt = rxe->num_ports;
+	dev->num_comp_vectors = RXE_NUM_COMP_VECTORS;
+	dev->dma_device = rxe->ifc_ops->dma_device(rxe);
+	dev->local_dma_lkey = 0;
+	dev->node_guid = rxe->ifc_ops->node_guid(rxe);
+	dev->dma_ops = &rxe_dma_mapping_ops;
+
+	dev->uverbs_abi_ver = RXE_UVERBS_ABI_VERSION;
+	dev->uverbs_cmd_mask = BIT_ULL(IB_USER_VERBS_CMD_GET_CONTEXT)
+	    | BIT_ULL(IB_USER_VERBS_CMD_CREATE_COMP_CHANNEL)
+	    | BIT_ULL(IB_USER_VERBS_CMD_QUERY_DEVICE)
+	    | BIT_ULL(IB_USER_VERBS_CMD_QUERY_PORT)
+	    | BIT_ULL(IB_USER_VERBS_CMD_ALLOC_PD)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DEALLOC_PD)
+	    | BIT_ULL(IB_USER_VERBS_CMD_CREATE_SRQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_MODIFY_SRQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_QUERY_SRQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DESTROY_SRQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_POST_SRQ_RECV)
+	    | BIT_ULL(IB_USER_VERBS_CMD_CREATE_QP)
+	    | BIT_ULL(IB_USER_VERBS_CMD_MODIFY_QP)
+	    | BIT_ULL(IB_USER_VERBS_CMD_QUERY_QP)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DESTROY_QP)
+	    | BIT_ULL(IB_USER_VERBS_CMD_POST_SEND)
+	    | BIT_ULL(IB_USER_VERBS_CMD_POST_RECV)
+	    | BIT_ULL(IB_USER_VERBS_CMD_CREATE_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_RESIZE_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DESTROY_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_POLL_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_PEEK_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_REQ_NOTIFY_CQ)
+	    | BIT_ULL(IB_USER_VERBS_CMD_REG_MR)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DEREG_MR)
+	    | BIT_ULL(IB_USER_VERBS_CMD_CREATE_AH)
+	    | BIT_ULL(IB_USER_VERBS_CMD_MODIFY_AH)
+	    | BIT_ULL(IB_USER_VERBS_CMD_QUERY_AH)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DESTROY_AH)
+	    | BIT_ULL(IB_USER_VERBS_CMD_ATTACH_MCAST)
+	    | BIT_ULL(IB_USER_VERBS_CMD_DETACH_MCAST)
+	    ;
+
+	dev->query_device = rxe_query_device;
+	dev->modify_device = rxe_modify_device;
+	dev->query_port = rxe_query_port;
+	dev->modify_port = rxe_modify_port;
+	dev->get_link_layer = rxe_get_link_layer;
+	dev->query_gid = rxe_query_gid;
+	dev->get_netdev = rxe_get_netdev;
+	dev->add_gid = rxe_add_gid;
+	dev->del_gid = rxe_del_gid;
+	dev->query_pkey = rxe_query_pkey;
+	dev->alloc_ucontext = rxe_alloc_ucontext;
+	dev->dealloc_ucontext = rxe_dealloc_ucontext;
+	dev->mmap = rxe_mmap;
+	dev->get_port_immutable = rxe_port_immutable;
+	dev->alloc_pd = rxe_alloc_pd;
+	dev->dealloc_pd = rxe_dealloc_pd;
+	dev->create_ah = rxe_create_ah;
+	dev->modify_ah = rxe_modify_ah;
+	dev->query_ah = rxe_query_ah;
+	dev->destroy_ah = rxe_destroy_ah;
+	dev->create_srq = rxe_create_srq;
+	dev->modify_srq = rxe_modify_srq;
+	dev->query_srq = rxe_query_srq;
+	dev->destroy_srq = rxe_destroy_srq;
+	dev->post_srq_recv = rxe_post_srq_recv;
+	dev->create_qp = rxe_create_qp;
+	dev->modify_qp = rxe_modify_qp;
+	dev->query_qp = rxe_query_qp;
+	dev->destroy_qp = rxe_destroy_qp;
+	dev->post_send = rxe_post_send;
+	dev->post_recv = rxe_post_recv;
+	dev->create_cq = rxe_create_cq;
+	dev->destroy_cq = rxe_destroy_cq;
+	dev->resize_cq = rxe_resize_cq;
+	dev->poll_cq = rxe_poll_cq;
+	dev->peek_cq = rxe_peek_cq;
+	dev->req_notify_cq = rxe_req_notify_cq;
+	dev->get_dma_mr = rxe_get_dma_mr;
+	dev->reg_user_mr = rxe_reg_user_mr;
+	dev->dereg_mr = rxe_dereg_mr;
+	dev->alloc_mr = rxe_alloc_mr;
+	dev->alloc_fmr = rxe_alloc_fmr;
+	dev->map_phys_fmr = rxe_map_phys_fmr;
+	dev->unmap_fmr = rxe_unmap_fmr;
+	dev->dealloc_fmr = rxe_dealloc_fmr;
+	dev->attach_mcast = rxe_attach_mcast;
+	dev->detach_mcast = rxe_detach_mcast;
+
+	err = ib_register_device(dev, NULL);
+	if (err) {
+		pr_warn("rxe_register_device failed, err = %d\n", err);
+		goto err1;
+	}
+
+	for (i = 0; i < ARRAY_SIZE(rxe_dev_attributes); ++i) {
+		err = device_create_file(&dev->dev, rxe_dev_attributes[i]);
+		if (err) {
+			pr_warn("device_create_file failed, i = %d, err = %d\n",
+				i, err);
+			goto err2;
+		}
+	}
+
+	return 0;
+
+err2:
+	ib_unregister_device(dev);
+err1:
+	return err;
+}
+
+int rxe_unregister_device(struct rxe_dev *rxe)
+{
+	int i;
+	struct ib_device *dev = &rxe->ib_dev;
+
+	for (i = 0; i < ARRAY_SIZE(rxe_dev_attributes); ++i)
+		device_remove_file(&dev->dev, rxe_dev_attributes[i]);
+
+	ib_unregister_device(dev);
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/rxe/rxe_verbs.h b/drivers/infiniband/hw/rxe/rxe_verbs.h
new file mode 100644
index 0000000..5783de9
--- /dev/null
+++ b/drivers/infiniband/hw/rxe/rxe_verbs.h
@@ -0,0 +1,486 @@
+/*
+ * Copyright (c) 2016 Mellanox Technologies Ltd. All rights reserved.
+ * Copyright (c) 2015 System Fabric Works, Inc. All rights reserved.
+ *
+ * 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.
+ */
+
+#ifndef RXE_VERBS_H
+#define RXE_VERBS_H
+
+#include <linux/interrupt.h>
+#include <rdma/ib_rxe.h>
+#include "rxe_pool.h"
+#include "rxe_task.h"
+
+static inline int pkey_match(u16 key1, u16 key2)
+{
+	return (((key1 & 0x7fff) != 0) &&
+		((key1 & 0x7fff) == (key2 & 0x7fff)) &&
+		((key1 & 0x8000) || (key2 & 0x8000))) ? 1 : 0;
+}
+
+/* Return >0 if psn_a > psn_b
+ *	   0 if psn_a == psn_b
+ *	  <0 if psn_a < psn_b
+ */
+static inline int psn_compare(u32 psn_a, u32 psn_b)
+{
+	s32 diff;
+
+	diff = (psn_a - psn_b) << 8;
+	return diff;
+}
+
+struct rxe_ucontext {
+	struct rxe_pool_entry	pelem;
+	struct ib_ucontext	ibuc;
+};
+
+struct rxe_pd {
+	struct rxe_pool_entry	pelem;
+	struct ib_pd		ibpd;
+};
+
+struct rxe_ah {
+	struct rxe_pool_entry	pelem;
+	struct ib_ah		ibah;
+	struct rxe_pd		*pd;
+	struct rxe_av		av;
+};
+
+struct rxe_cqe {
+	union {
+		struct ib_wc		ibwc;
+		struct ib_uverbs_wc	uibwc;
+	};
+};
+
+struct rxe_cq {
+	struct rxe_pool_entry	pelem;
+	struct ib_cq		ibcq;
+	struct rxe_queue	*queue;
+	spinlock_t		cq_lock;
+	u8			notify;
+	int			is_user;
+	struct tasklet_struct	comp_task;
+};
+
+enum wqe_state {
+	wqe_state_posted,
+	wqe_state_processing,
+	wqe_state_pending,
+	wqe_state_done,
+	wqe_state_error,
+};
+
+struct rxe_sq {
+	int			max_wr;
+	int			max_sge;
+	int			max_inline;
+	spinlock_t		sq_lock; /* guard queue */
+	struct rxe_queue	*queue;
+};
+
+struct rxe_rq {
+	int			max_wr;
+	int			max_sge;
+	spinlock_t		producer_lock; /* guard queue producer */
+	spinlock_t		consumer_lock; /* guard queue consumer */
+	struct rxe_queue	*queue;
+};
+
+struct rxe_srq {
+	struct rxe_pool_entry	pelem;
+	struct ib_srq		ibsrq;
+	struct rxe_pd		*pd;
+	struct rxe_rq		rq;
+	u32			srq_num;
+
+	int			limit;
+	int			error;
+};
+
+enum rxe_qp_state {
+	QP_STATE_RESET,
+	QP_STATE_INIT,
+	QP_STATE_READY,
+	QP_STATE_DRAIN,		/* req only */
+	QP_STATE_DRAINED,	/* req only */
+	QP_STATE_ERROR
+};
+
+extern char *rxe_qp_state_name[];
+
+struct rxe_req_info {
+	enum rxe_qp_state	state;
+	int			wqe_index;
+	u32			psn;
+	int			opcode;
+	atomic_t		rd_atomic;
+	int			wait_fence;
+	int			need_rd_atomic;
+	int			wait_psn;
+	int			need_retry;
+	int			noack_pkts;
+	struct rxe_task		task;
+};
+
+struct rxe_comp_info {
+	u32			psn;
+	int			opcode;
+	int			timeout;
+	int			timeout_retry;
+	u32			retry_cnt;
+	u32			rnr_retry;
+	struct rxe_task		task;
+};
+
+enum rdatm_res_state {
+	rdatm_res_state_next,
+	rdatm_res_state_new,
+	rdatm_res_state_replay,
+};
+
+struct resp_res {
+	int			type;
+	u32			first_psn;
+	u32			last_psn;
+	u32			cur_psn;
+	enum rdatm_res_state	state;
+
+	union {
+		struct {
+			struct sk_buff	*skb;
+		} atomic;
+		struct {
+			struct rxe_mem	*mr;
+			u64		va_org;
+			u32		rkey;
+			u32		length;
+			u64		va;
+			u32		resid;
+		} read;
+	};
+};
+
+struct rxe_resp_info {
+	enum rxe_qp_state	state;
+	u32			msn;
+	u32			psn;
+	int			opcode;
+	int			drop_msg;
+	int			goto_error;
+	int			sent_psn_nak;
+	enum ib_wc_status	status;
+	u8			aeth_syndrome;
+
+	/* Receive only */
+	struct rxe_recv_wqe	*wqe;
+
+	/* RDMA read / atomic only */
+	u64			va;
+	struct rxe_mem		*mr;
+	u32			resid;
+	u32			rkey;
+	u64			atomic_orig;
+
+	/* SRQ only */
+	struct {
+		struct rxe_recv_wqe	wqe;
+		struct ib_sge		sge[RXE_MAX_SGE];
+	} srq_wqe;
+
+	/* Responder resources. It's a circular list where the oldest
+	 * resource is dropped first.
+	 */
+	struct resp_res		*resources;
+	unsigned int		res_head;
+	unsigned int		res_tail;
+	struct resp_res		*res;
+	struct rxe_task		task;
+};
+
+struct rxe_qp {
+	struct rxe_pool_entry	pelem;
+	struct ib_qp		ibqp;
+	struct ib_qp_attr	attr;
+	unsigned int		valid;
+	unsigned int		mtu;
+	int			is_user;
+
+	struct rxe_pd		*pd;
+	struct rxe_srq		*srq;
+	struct rxe_cq		*scq;
+	struct rxe_cq		*rcq;
+
+	enum ib_sig_type	sq_sig_type;
+
+	struct rxe_sq		sq;
+	struct rxe_rq		rq;
+
+	struct socket		*sk;
+
+	struct rxe_av		pri_av;
+	struct rxe_av		alt_av;
+
+	/* list of mcast groups qp has joined (for cleanup) */
+	struct list_head	grp_list;
+	spinlock_t		grp_lock; /* guard grp_list */
+
+	struct sk_buff_head	req_pkts;
+	struct sk_buff_head	resp_pkts;
+	struct sk_buff_head	send_pkts;
+
+	struct rxe_req_info	req;
+	struct rxe_comp_info	comp;
+	struct rxe_resp_info	resp;
+
+	atomic_t		ssn;
+	atomic_t		skb_out;
+	int			need_req_skb;
+
+	/* Timer for retranmitting packet when ACKs have been lost. RC
+	 * only. The requester sets it when it is not already
+	 * started. The responder resets it whenever an ack is
+	 * received.
+	 */
+	struct timer_list retrans_timer;
+	u64 qp_timeout_jiffies;
+
+	/* Timer for handling RNR NAKS. */
+	struct timer_list rnr_nak_timer;
+
+	spinlock_t		state_lock; /* guard requester and completer */
+};
+
+enum rxe_mem_state {
+	RXE_MEM_STATE_ZOMBIE,
+	RXE_MEM_STATE_INVALID,
+	RXE_MEM_STATE_FREE,
+	RXE_MEM_STATE_VALID,
+};
+
+enum rxe_mem_type {
+	RXE_MEM_TYPE_NONE,
+	RXE_MEM_TYPE_DMA,
+	RXE_MEM_TYPE_MR,
+	RXE_MEM_TYPE_FMR,
+	RXE_MEM_TYPE_MW,
+};
+
+#define RXE_BUF_PER_MAP		(PAGE_SIZE / sizeof(struct rxe_phys_buf))
+
+struct rxe_phys_buf {
+	u64      addr;
+	u64      size;
+};
+
+struct rxe_map {
+	struct rxe_phys_buf	buf[RXE_BUF_PER_MAP];
+};
+
+struct rxe_mem {
+	struct rxe_pool_entry	pelem;
+	union {
+		struct ib_mr		ibmr;
+		struct ib_fmr		ibfmr;
+		struct ib_mw		ibmw;
+	};
+
+	struct rxe_pd		*pd;
+	struct ib_umem		*umem;
+
+	u32			lkey;
+	u32			rkey;
+
+	enum rxe_mem_state	state;
+	enum rxe_mem_type	type;
+	u64			va;
+	u64			iova;
+	size_t			length;
+	u32			offset;
+	int			access;
+
+	int			page_shift;
+	int			page_mask;
+	int			map_shift;
+	int			map_mask;
+
+	u32			num_buf;
+
+	u32			max_buf;
+	u32			num_map;
+
+	struct rxe_map		**map;
+};
+
+struct rxe_mc_grp {
+	struct rxe_pool_entry	pelem;
+	spinlock_t		mcg_lock; /* guard group */
+	struct rxe_dev		*rxe;
+	struct list_head	qp_list;
+	union ib_gid		mgid;
+	int			num_qp;
+	u32			qkey;
+	u16			pkey;
+};
+
+struct rxe_mc_elem {
+	struct rxe_pool_entry	pelem;
+	struct list_head	qp_list;
+	struct list_head	grp_list;
+	struct rxe_qp		*qp;
+	struct rxe_mc_grp	*grp;
+};
+
+struct rxe_port {
+	struct ib_port_attr	attr;
+	u16			*pkey_tbl;
+	__be64			port_guid;
+	__be64			subnet_prefix;
+	spinlock_t		port_lock; /* guard port */
+	unsigned int		mtu_cap;
+	/* special QPs */
+	u32			qp_smi_index;
+	u32			qp_gsi_index;
+};
+
+/* callbacks from ib_rxe to network interface layer */
+struct rxe_ifc_ops {
+	void (*release)(struct rxe_dev *rxe);
+	__be64 (*node_guid)(struct rxe_dev *rxe);
+	__be64 (*port_guid)(struct rxe_dev *rxe, unsigned int port_num);
+	struct device *(*dma_device)(struct rxe_dev *rxe);
+	int (*mcast_add)(struct rxe_dev *rxe, union ib_gid *mgid);
+	int (*mcast_delete)(struct rxe_dev *rxe, union ib_gid *mgid);
+	int (*prepare)(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
+		       struct sk_buff *skb, u32 *crc);
+	int (*send)(struct rxe_dev *rxe, struct rxe_pkt_info *pkt,
+		    struct sk_buff *skb);
+	int (*loopback)(struct sk_buff *skb);
+	struct sk_buff *(*init_packet)(struct rxe_dev *rxe, struct rxe_av *av,
+				       int paylen, struct rxe_pkt_info *pkt);
+	char *(*parent_name)(struct rxe_dev *rxe, unsigned int port_num);
+	enum rdma_link_layer (*link_layer)(struct rxe_dev *rxe,
+					   unsigned int port_num);
+};
+
+struct rxe_dev {
+	struct ib_device	ib_dev;
+	struct ib_device_attr	attr;
+	int			max_ucontext;
+	int			max_inline_data;
+	struct kref		ref_cnt;
+	struct mutex	usdev_lock;
+
+	struct rxe_ifc_ops	*ifc_ops;
+
+	struct net_device	*ndev;
+
+	int			xmit_errors;
+
+	struct rxe_pool		uc_pool;
+	struct rxe_pool		pd_pool;
+	struct rxe_pool		ah_pool;
+	struct rxe_pool		srq_pool;
+	struct rxe_pool		qp_pool;
+	struct rxe_pool		cq_pool;
+	struct rxe_pool		mr_pool;
+	struct rxe_pool		mw_pool;
+	struct rxe_pool		fmr_pool;
+	struct rxe_pool		mc_grp_pool;
+	struct rxe_pool		mc_elem_pool;
+
+	spinlock_t		pending_lock; /* guard pending_mmaps */
+	struct list_head	pending_mmaps;
+
+	spinlock_t		mmap_offset_lock; /* guard mmap_offset */
+	int			mmap_offset;
+
+	u8			num_ports;
+	struct rxe_port		*port;
+};
+
+static inline struct rxe_dev *to_rdev(struct ib_device *dev)
+{
+	return dev ? container_of(dev, struct rxe_dev, ib_dev) : NULL;
+}
+
+static inline struct rxe_ucontext *to_ruc(struct ib_ucontext *uc)
+{
+	return uc ? container_of(uc, struct rxe_ucontext, ibuc) : NULL;
+}
+
+static inline struct rxe_pd *to_rpd(struct ib_pd *pd)
+{
+	return pd ? container_of(pd, struct rxe_pd, ibpd) : NULL;
+}
+
+static inline struct rxe_ah *to_rah(struct ib_ah *ah)
+{
+	return ah ? container_of(ah, struct rxe_ah, ibah) : NULL;
+}
+
+static inline struct rxe_srq *to_rsrq(struct ib_srq *srq)
+{
+	return srq ? container_of(srq, struct rxe_srq, ibsrq) : NULL;
+}
+
+static inline struct rxe_qp *to_rqp(struct ib_qp *qp)
+{
+	return qp ? container_of(qp, struct rxe_qp, ibqp) : NULL;
+}
+
+static inline struct rxe_cq *to_rcq(struct ib_cq *cq)
+{
+	return cq ? container_of(cq, struct rxe_cq, ibcq) : NULL;
+}
+
+static inline struct rxe_mem *to_rmr(struct ib_mr *mr)
+{
+	return mr ? container_of(mr, struct rxe_mem, ibmr) : NULL;
+}
+
+static inline struct rxe_mem *to_rfmr(struct ib_fmr *fmr)
+{
+	return fmr ? container_of(fmr, struct rxe_mem, ibfmr) : NULL;
+}
+
+static inline struct rxe_mem *to_rmw(struct ib_mw *mw)
+{
+	return mw ? container_of(mw, struct rxe_mem, ibmw) : NULL;
+}
+
+int rxe_register_device(struct rxe_dev *rxe);
+int rxe_unregister_device(struct rxe_dev *rxe);
+
+void rxe_mc_cleanup(void *arg);
+
+#endif /* RXE_VERBS_H */
-- 
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