diff mbox

[v6,05/16] IB/pvrdma: Add functions for Verbs support

Message ID 1c3455252b6f81588178a3e70089628aa92ee037.1475458407.git.aditr@vmware.com (mailing list archive)
State Accepted
Headers show

Commit Message

Adit Ranadive Oct. 3, 2016, 2:10 a.m. UTC
This patch adds Verbs-related PVRDMA specific definitions and implements
the remaining Verbs functions registered with the core RDMA stack.

Reviewed-by: Jorgen Hansen <jhansen@vmware.com>
Reviewed-by: George Zhang <georgezhang@vmware.com>
Reviewed-by: Aditya Sarwade <asarwade@vmware.com>
Reviewed-by: Bryan Tan <bryantan@vmware.com>
Signed-off-by: Adit Ranadive <aditr@vmware.com>
---
Changes v5->v6:
 - Structs/enums defined in pvrdma_ib_verbs.h (removed) are now in
   pvrdma_verbs.h.

Changes v4->v5:
 - Update include for headers in UAPI folder.
 - Removed setting any properties that are reported by device as 0.
 - Simplified modify_port.
 - PD should be allocated first in kernel then in device.
 - Update to pvrdma_cmd_post for creating/destroying PD, Query port/device.

Changes v3->v4:
 - Renamed priviledged -> privileged.
 - Added error numbers for command errors.
 - Removed unnecessary goto in modify_device.
 - Moved pd allocation to after command execution.
 - Removed an incorrect atomic_dec.
---
 drivers/infiniband/hw/pvrdma/pvrdma_verbs.c | 577 ++++++++++++++++++++++++++++
 drivers/infiniband/hw/pvrdma/pvrdma_verbs.h | 435 +++++++++++++++++++++
 2 files changed, 1012 insertions(+)
 create mode 100644 drivers/infiniband/hw/pvrdma/pvrdma_verbs.c
 create mode 100644 drivers/infiniband/hw/pvrdma/pvrdma_verbs.h
diff mbox

Patch

diff --git a/drivers/infiniband/hw/pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/pvrdma/pvrdma_verbs.c
new file mode 100644
index 0000000..a7aef93d
--- /dev/null
+++ b/drivers/infiniband/hw/pvrdma/pvrdma_verbs.c
@@ -0,0 +1,577 @@ 
+/*
+ * Copyright (c) 2012-2016 VMware, Inc.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of EITHER the GNU General Public License
+ * version 2 as published by the Free Software Foundation or the BSD
+ * 2-Clause License. This program is distributed in the hope that it
+ * will be useful, but WITHOUT ANY WARRANTY; WITHOUT EVEN THE IMPLIED
+ * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
+ * See the GNU General Public License version 2 for more details at
+ * http://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program available in the file COPYING in the main
+ * directory of this source tree.
+ *
+ * The BSD 2-Clause License
+ *
+ *     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.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
+ * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#include <asm/page.h>
+#include <linux/inet.h>
+#include <linux/io.h>
+#include <rdma/ib_addr.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_user_verbs.h>
+#include <rdma/pvrdma-abi.h>
+
+#include "pvrdma.h"
+
+/**
+ * pvrdma_query_device - query device
+ * @ibdev: the device to query
+ * @props: the device properties
+ * @uhw: user data
+ *
+ * @return: 0 on success, otherwise negative errno
+ */
+int pvrdma_query_device(struct ib_device *ibdev,
+			struct ib_device_attr *props,
+			struct ib_udata *uhw)
+{
+	struct pvrdma_dev *dev = to_vdev(ibdev);
+
+	if (uhw->inlen || uhw->outlen)
+		return -EINVAL;
+
+	memset(props, 0, sizeof(*props));
+
+	props->fw_ver = dev->dsr->caps.fw_ver;
+	props->sys_image_guid = dev->dsr->caps.sys_image_guid;
+	props->max_mr_size = dev->dsr->caps.max_mr_size;
+	props->page_size_cap = dev->dsr->caps.page_size_cap;
+	props->vendor_id = dev->dsr->caps.vendor_id;
+	props->vendor_part_id = dev->pdev->device;
+	props->hw_ver = dev->dsr->caps.hw_ver;
+	props->max_qp = dev->dsr->caps.max_qp;
+	props->max_qp_wr = dev->dsr->caps.max_qp_wr;
+	props->device_cap_flags = dev->dsr->caps.device_cap_flags;
+	props->max_sge = dev->dsr->caps.max_sge;
+	props->max_cq = dev->dsr->caps.max_cq;
+	props->max_cqe = dev->dsr->caps.max_cqe;
+	props->max_mr = dev->dsr->caps.max_mr;
+	props->max_pd = dev->dsr->caps.max_pd;
+	props->max_qp_rd_atom = dev->dsr->caps.max_qp_rd_atom;
+	props->max_qp_init_rd_atom = dev->dsr->caps.max_qp_init_rd_atom;
+	props->atomic_cap =
+		dev->dsr->caps.atomic_ops &
+		(PVRDMA_ATOMIC_OP_COMP_SWAP | PVRDMA_ATOMIC_OP_FETCH_ADD) ?
+		IB_ATOMIC_HCA : IB_ATOMIC_NONE;
+	props->masked_atomic_cap = props->atomic_cap;
+	props->max_ah = dev->dsr->caps.max_ah;
+	props->max_pkeys = dev->dsr->caps.max_pkeys;
+	props->local_ca_ack_delay = dev->dsr->caps.local_ca_ack_delay;
+	if ((dev->dsr->caps.bmme_flags & PVRDMA_BMME_FLAG_LOCAL_INV) &&
+	    (dev->dsr->caps.bmme_flags & PVRDMA_BMME_FLAG_REMOTE_INV) &&
+	    (dev->dsr->caps.bmme_flags & PVRDMA_BMME_FLAG_FAST_REG_WR)) {
+		props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
+	}
+
+	return 0;
+}
+
+/**
+ * pvrdma_query_port - query device port attributes
+ * @ibdev: the device to query
+ * @port: the port number
+ * @props: the device properties
+ *
+ * @return: 0 on success, otherwise negative errno
+ */
+int pvrdma_query_port(struct ib_device *ibdev, u8 port,
+		      struct ib_port_attr *props)
+{
+	struct pvrdma_dev *dev = to_vdev(ibdev);
+	union pvrdma_cmd_req req;
+	union pvrdma_cmd_resp rsp;
+	struct pvrdma_cmd_query_port *cmd = &req.query_port;
+	struct pvrdma_cmd_query_port_resp *resp = &rsp.query_port_resp;
+	int err;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_QUERY_PORT;
+	cmd->port_num = port;
+
+	err = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_QUERY_PORT_RESP);
+	if (err < 0) {
+		dev_warn(&dev->pdev->dev,
+			 "could not query port, error: %d\n", err);
+		return err;
+	}
+
+	memset(props, 0, sizeof(*props));
+
+	props->state = pvrdma_port_state_to_ib(resp->attrs.state);
+	props->max_mtu = pvrdma_mtu_to_ib(resp->attrs.max_mtu);
+	props->active_mtu = pvrdma_mtu_to_ib(resp->attrs.active_mtu);
+	props->gid_tbl_len = resp->attrs.gid_tbl_len;
+	props->port_cap_flags =
+		pvrdma_port_cap_flags_to_ib(resp->attrs.port_cap_flags);
+	props->max_msg_sz = resp->attrs.max_msg_sz;
+	props->bad_pkey_cntr = resp->attrs.bad_pkey_cntr;
+	props->qkey_viol_cntr = resp->attrs.qkey_viol_cntr;
+	props->pkey_tbl_len = resp->attrs.pkey_tbl_len;
+	props->lid = resp->attrs.lid;
+	props->sm_lid = resp->attrs.sm_lid;
+	props->lmc = resp->attrs.lmc;
+	props->max_vl_num = resp->attrs.max_vl_num;
+	props->sm_sl = resp->attrs.sm_sl;
+	props->subnet_timeout = resp->attrs.subnet_timeout;
+	props->init_type_reply = resp->attrs.init_type_reply;
+	props->active_width = pvrdma_port_width_to_ib(resp->attrs.active_width);
+	props->active_speed = pvrdma_port_speed_to_ib(resp->attrs.active_speed);
+	props->phys_state = resp->attrs.phys_state;
+
+	return 0;
+}
+
+/**
+ * pvrdma_query_gid - query device gid
+ * @ibdev: the device to query
+ * @port: the port number
+ * @index: the index
+ * @gid: the device gid value
+ *
+ * @return: 0 on success, otherwise negative errno
+ */
+int pvrdma_query_gid(struct ib_device *ibdev, u8 port, int index,
+		     union ib_gid *gid)
+{
+	struct pvrdma_dev *dev = to_vdev(ibdev);
+
+	if (index >= dev->dsr->caps.gid_tbl_len)
+		return -EINVAL;
+
+	memcpy(gid, &dev->sgid_tbl[index], sizeof(union ib_gid));
+
+	return 0;
+}
+
+/**
+ * pvrdma_query_pkey - query device port's P_Key table
+ * @ibdev: the device to query
+ * @port: the port number
+ * @index: the index
+ * @pkey: the device P_Key value
+ *
+ * @return: 0 on success, otherwise negative errno
+ */
+int pvrdma_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
+		      u16 *pkey)
+{
+	int err = 0;
+	union pvrdma_cmd_req req;
+	union pvrdma_cmd_resp rsp;
+	struct pvrdma_cmd_query_pkey *cmd = &req.query_pkey;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_QUERY_PKEY;
+	cmd->port_num = port;
+	cmd->index = index;
+
+	err = pvrdma_cmd_post(to_vdev(ibdev), &req, &rsp,
+			      PVRDMA_CMD_QUERY_PKEY_RESP);
+	if (err < 0) {
+		dev_warn(&to_vdev(ibdev)->pdev->dev,
+			 "could not query pkey, error: %d\n", err);
+		return err;
+	}
+
+	*pkey = rsp.query_pkey_resp.pkey;
+
+	return 0;
+}
+
+enum rdma_link_layer pvrdma_port_link_layer(struct ib_device *ibdev,
+					    u8 port)
+{
+	return IB_LINK_LAYER_ETHERNET;
+}
+
+int pvrdma_modify_device(struct ib_device *ibdev, int mask,
+			 struct ib_device_modify *props)
+{
+	unsigned long flags;
+
+	if (mask & ~(IB_DEVICE_MODIFY_SYS_IMAGE_GUID |
+		     IB_DEVICE_MODIFY_NODE_DESC)) {
+		dev_warn(&to_vdev(ibdev)->pdev->dev,
+			 "unsupported device modify mask %#x\n", mask);
+		return -EOPNOTSUPP;
+	}
+
+	if (mask & IB_DEVICE_MODIFY_NODE_DESC) {
+		spin_lock_irqsave(&to_vdev(ibdev)->desc_lock, flags);
+		memcpy(ibdev->node_desc, props->node_desc, 64);
+		spin_unlock_irqrestore(&to_vdev(ibdev)->desc_lock, flags);
+	}
+
+	if (mask & IB_DEVICE_MODIFY_SYS_IMAGE_GUID) {
+		mutex_lock(&to_vdev(ibdev)->port_mutex);
+		to_vdev(ibdev)->sys_image_guid =
+			cpu_to_be64(props->sys_image_guid);
+		mutex_unlock(&to_vdev(ibdev)->port_mutex);
+	}
+
+	return 0;
+}
+
+/**
+ * pvrdma_modify_port - modify device port attributes
+ * @ibdev: the device to modify
+ * @port: the port number
+ * @mask: attributes to modify
+ * @props: the device properties
+ *
+ * @return: 0 on success, otherwise negative errno
+ */
+int pvrdma_modify_port(struct ib_device *ibdev, u8 port, int mask,
+		       struct ib_port_modify *props)
+{
+	struct ib_port_attr attr;
+	struct pvrdma_dev *vdev = to_vdev(ibdev);
+	int ret;
+
+	if (mask & ~IB_PORT_SHUTDOWN) {
+		dev_warn(&vdev->pdev->dev,
+			 "unsupported port modify mask %#x\n", mask);
+		return -EOPNOTSUPP;
+	}
+
+	mutex_lock(&vdev->port_mutex);
+	ret = pvrdma_query_port(ibdev, port, &attr);
+	if (ret)
+		goto out;
+
+	vdev->port_cap_mask |= props->set_port_cap_mask;
+	vdev->port_cap_mask &= ~props->clr_port_cap_mask;
+
+	if (mask & IB_PORT_SHUTDOWN)
+		vdev->ib_active = false;
+
+out:
+	mutex_unlock(&vdev->port_mutex);
+	return ret;
+}
+
+/**
+ * pvrdma_alloc_ucontext - allocate ucontext
+ * @ibdev: the IB device
+ * @udata: user data
+ *
+ * @return: the ib_ucontext pointer on success, otherwise errno.
+ */
+struct ib_ucontext *pvrdma_alloc_ucontext(struct ib_device *ibdev,
+					  struct ib_udata *udata)
+{
+	struct pvrdma_dev *vdev = to_vdev(ibdev);
+	struct pvrdma_ucontext *context;
+	union pvrdma_cmd_req req;
+	union pvrdma_cmd_resp rsp;
+	struct pvrdma_cmd_create_uc *cmd = &req.create_uc;
+	struct pvrdma_cmd_create_uc_resp *resp = &rsp.create_uc_resp;
+	struct pvrdma_alloc_ucontext_resp uresp;
+	int ret;
+	void *ptr;
+
+	if (!vdev->ib_active)
+		return ERR_PTR(-EAGAIN);
+
+	context = kmalloc(sizeof(*context), GFP_KERNEL);
+	if (!context)
+		return ERR_PTR(-ENOMEM);
+
+	context->dev = vdev;
+	ret = pvrdma_uar_alloc(vdev, &context->uar);
+	if (ret) {
+		kfree(context);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	/* get ctx_handle from host */
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->pfn = context->uar.pfn;
+	cmd->hdr.cmd = PVRDMA_CMD_CREATE_UC;
+	ret = pvrdma_cmd_post(vdev, &req, &rsp, PVRDMA_CMD_CREATE_UC_RESP);
+	if (ret < 0) {
+		dev_warn(&vdev->pdev->dev,
+			 "could not create ucontext, error: %d\n", ret);
+		ptr = ERR_PTR(ret);
+		goto err;
+	}
+
+	context->ctx_handle = resp->ctx_handle;
+
+	/* copy back to user */
+	uresp.qp_tab_size = vdev->dsr->caps.max_qp;
+	ret = ib_copy_to_udata(udata, &uresp, sizeof(uresp));
+	if (ret) {
+		pvrdma_uar_free(vdev, &context->uar);
+		context->ibucontext.device = ibdev;
+		pvrdma_dealloc_ucontext(&context->ibucontext);
+		return ERR_PTR(-EFAULT);
+	}
+
+	return &context->ibucontext;
+
+err:
+	pvrdma_uar_free(vdev, &context->uar);
+	kfree(context);
+	return ptr;
+}
+
+/**
+ * pvrdma_dealloc_ucontext - deallocate ucontext
+ * @ibcontext: the ucontext
+ *
+ * @return: 0 on success, otherwise errno.
+ */
+int pvrdma_dealloc_ucontext(struct ib_ucontext *ibcontext)
+{
+	struct pvrdma_ucontext *context = to_vucontext(ibcontext);
+	union pvrdma_cmd_req req;
+	struct pvrdma_cmd_destroy_uc *cmd = &req.destroy_uc;
+	int ret;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_DESTROY_UC;
+	cmd->ctx_handle = context->ctx_handle;
+
+	ret = pvrdma_cmd_post(context->dev, &req, NULL, 0);
+	if (ret < 0)
+		dev_warn(&context->dev->pdev->dev,
+			 "destroy ucontext failed, error: %d\n", ret);
+
+	/* Free the UAR even if the device command failed */
+	pvrdma_uar_free(to_vdev(ibcontext->device), &context->uar);
+	kfree(context);
+
+	return ret;
+}
+
+/**
+ * pvrdma_mmap - create mmap region
+ * @ibcontext: the user context
+ * @vma: the VMA
+ *
+ * @return: 0 on success, otherwise errno.
+ */
+int pvrdma_mmap(struct ib_ucontext *ibcontext, struct vm_area_struct *vma)
+{
+	struct pvrdma_ucontext *context = to_vucontext(ibcontext);
+	unsigned long start = vma->vm_start;
+	unsigned long size = vma->vm_end - vma->vm_start;
+	unsigned long offset = vma->vm_pgoff << PAGE_SHIFT;
+
+	dev_dbg(&context->dev->pdev->dev, "create mmap region\n");
+
+	if ((size != PAGE_SIZE) || (offset & ~PAGE_MASK)) {
+		dev_warn(&context->dev->pdev->dev,
+			 "invalid params for mmap region\n");
+		return -EINVAL;
+	}
+
+	/* Map UAR to kernel space, VM_LOCKED? */
+	vma->vm_flags |= VM_DONTCOPY | VM_DONTEXPAND;
+	vma->vm_page_prot = pgprot_noncached(vma->vm_page_prot);
+	if (io_remap_pfn_range(vma, start, context->uar.pfn, size,
+			       vma->vm_page_prot))
+		return -EAGAIN;
+
+	return 0;
+}
+
+/**
+ * pvrdma_alloc_pd - allocate protection domain
+ * @ibdev: the IB device
+ * @context: user context
+ * @udata: user data
+ *
+ * @return: the ib_pd protection domain pointer on success, otherwise errno.
+ */
+struct ib_pd *pvrdma_alloc_pd(struct ib_device *ibdev,
+			      struct ib_ucontext *context,
+			      struct ib_udata *udata)
+{
+	struct pvrdma_pd *pd;
+	struct pvrdma_dev *dev = to_vdev(ibdev);
+	union pvrdma_cmd_req req;
+	union pvrdma_cmd_resp rsp;
+	struct pvrdma_cmd_create_pd *cmd = &req.create_pd;
+	struct pvrdma_cmd_create_pd_resp *resp = &rsp.create_pd_resp;
+	int ret;
+	void *ptr;
+
+	/* Check allowed max pds */
+	if (!atomic_add_unless(&dev->num_pds, 1, dev->dsr->caps.max_pd))
+		return ERR_PTR(-ENOMEM);
+
+	pd = kmalloc(sizeof(*pd), GFP_KERNEL);
+	if (!pd) {
+		ptr = ERR_PTR(-ENOMEM);
+		goto err;
+	}
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_CREATE_PD;
+	cmd->ctx_handle = (context) ? to_vucontext(context)->ctx_handle : 0;
+	ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_CREATE_PD_RESP);
+	if (ret < 0) {
+		dev_warn(&dev->pdev->dev,
+			 "failed to allocate protection domain, error: %d\n",
+			 ret);
+		ptr = ERR_PTR(ret);
+		goto freepd;
+	}
+
+	pd->privileged = !context;
+	pd->pd_handle = resp->pd_handle;
+	pd->pdn = resp->pd_handle;
+
+	if (context) {
+		if (ib_copy_to_udata(udata, &pd->pdn, sizeof(__u32))) {
+			dev_warn(&dev->pdev->dev,
+				 "failed to copy back protection domain\n");
+			pvrdma_dealloc_pd(&pd->ibpd);
+			return ERR_PTR(-EFAULT);
+		}
+	}
+
+	/* u32 pd handle */
+	return &pd->ibpd;
+
+freepd:
+	kfree(pd);
+err:
+	atomic_dec(&dev->num_pds);
+	return ptr;
+}
+
+/**
+ * pvrdma_dealloc_pd - deallocate protection domain
+ * @pd: the protection domain to be released
+ *
+ * @return: 0 on success, otherwise errno.
+ */
+int pvrdma_dealloc_pd(struct ib_pd *pd)
+{
+	struct pvrdma_dev *dev = to_vdev(pd->device);
+	union pvrdma_cmd_req req;
+	struct pvrdma_cmd_destroy_pd *cmd = &req.destroy_pd;
+	int ret;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_DESTROY_PD;
+	cmd->pd_handle = to_vpd(pd)->pd_handle;
+
+	ret = pvrdma_cmd_post(dev, &req, NULL, 0);
+	if (ret)
+		dev_warn(&dev->pdev->dev,
+			 "could not dealloc protection domain, error: %d\n",
+			 ret);
+
+	kfree(to_vpd(pd));
+	atomic_dec(&dev->num_pds);
+
+	return 0;
+}
+
+/**
+ * pvrdma_create_ah - create an address handle
+ * @pd: the protection domain
+ * @ah_attr: the attributes of the AH
+ *
+ * @return: the ib_ah pointer on success, otherwise errno.
+ */
+struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr)
+{
+	struct pvrdma_dev *dev = to_vdev(pd->device);
+	struct pvrdma_ah *ah;
+	enum rdma_link_layer ll;
+
+	if (!(ah_attr->ah_flags & IB_AH_GRH))
+		return ERR_PTR(-EINVAL);
+
+	ll = rdma_port_get_link_layer(pd->device, ah_attr->port_num);
+
+	if (ll != IB_LINK_LAYER_ETHERNET ||
+	    rdma_is_multicast_addr((struct in6_addr *)ah_attr->grh.dgid.raw))
+		return ERR_PTR(-EINVAL);
+
+	if (!atomic_add_unless(&dev->num_ahs, 1, dev->dsr->caps.max_ah))
+		return ERR_PTR(-ENOMEM);
+
+	ah = kzalloc(sizeof(*ah), GFP_KERNEL);
+	if (!ah) {
+		atomic_dec(&dev->num_ahs);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	ah->av.port_pd = to_vpd(pd)->pd_handle | (ah_attr->port_num << 24);
+	ah->av.src_path_bits = ah_attr->src_path_bits;
+	ah->av.src_path_bits |= 0x80;
+	ah->av.gid_index = ah_attr->grh.sgid_index;
+	ah->av.hop_limit = ah_attr->grh.hop_limit;
+	ah->av.sl_tclass_flowlabel = (ah_attr->grh.traffic_class << 20) |
+				      ah_attr->grh.flow_label;
+	memcpy(ah->av.dgid, ah_attr->grh.dgid.raw, 16);
+	memcpy(ah->av.dmac, ah_attr->dmac, 6);
+
+	ah->ibah.device = pd->device;
+	ah->ibah.pd = pd;
+	ah->ibah.uobject = NULL;
+
+	return &ah->ibah;
+}
+
+/**
+ * pvrdma_destroy_ah - destroy an address handle
+ * @ah: the address handle to destroyed
+ *
+ * @return: 0 on success.
+ */
+int pvrdma_destroy_ah(struct ib_ah *ah)
+{
+	struct pvrdma_dev *dev = to_vdev(ah->device);
+
+	kfree(to_vah(ah));
+	atomic_dec(&dev->num_ahs);
+
+	return 0;
+}
diff --git a/drivers/infiniband/hw/pvrdma/pvrdma_verbs.h b/drivers/infiniband/hw/pvrdma/pvrdma_verbs.h
new file mode 100644
index 0000000..25ce9fd
--- /dev/null
+++ b/drivers/infiniband/hw/pvrdma/pvrdma_verbs.h
@@ -0,0 +1,435 @@ 
+/*
+ * Copyright (c) 2012-2016 VMware, Inc.  All rights reserved.
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of EITHER the GNU General Public License
+ * version 2 as published by the Free Software Foundation or the BSD
+ * 2-Clause License. This program is distributed in the hope that it
+ * will be useful, but WITHOUT ANY WARRANTY; WITHOUT EVEN THE IMPLIED
+ * WARRANTY OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE.
+ * See the GNU General Public License version 2 for more details at
+ * http://www.gnu.org/licenses/old-licenses/gpl-2.0.en.html.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program available in the file COPYING in the main
+ * directory of this source tree.
+ *
+ * The BSD 2-Clause License
+ *
+ *     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.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
+ * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
+ * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
+ * FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE
+ * COPYRIGHT HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT,
+ * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES
+ * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR
+ * SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
+ * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
+ * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
+ * OF THE POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef __PVRDMA_VERBS_H__
+#define __PVRDMA_VERBS_H__
+
+#include <linux/types.h>
+
+union pvrdma_gid {
+	u8	raw[16];
+	struct {
+		__be64	subnet_prefix;
+		__be64	interface_id;
+	} global;
+};
+
+enum pvrdma_link_layer {
+	PVRDMA_LINK_LAYER_UNSPECIFIED,
+	PVRDMA_LINK_LAYER_INFINIBAND,
+	PVRDMA_LINK_LAYER_ETHERNET,
+};
+
+enum pvrdma_mtu {
+	PVRDMA_MTU_256  = 1,
+	PVRDMA_MTU_512  = 2,
+	PVRDMA_MTU_1024 = 3,
+	PVRDMA_MTU_2048 = 4,
+	PVRDMA_MTU_4096 = 5,
+};
+
+static inline int pvrdma_mtu_enum_to_int(enum pvrdma_mtu mtu)
+{
+	switch (mtu) {
+	case PVRDMA_MTU_256:	return  256;
+	case PVRDMA_MTU_512:	return  512;
+	case PVRDMA_MTU_1024:	return 1024;
+	case PVRDMA_MTU_2048:	return 2048;
+	case PVRDMA_MTU_4096:	return 4096;
+	default:		return   -1;
+	}
+}
+
+static inline enum pvrdma_mtu pvrdma_mtu_int_to_enum(int mtu)
+{
+	switch (mtu) {
+	case 256:	return PVRDMA_MTU_256;
+	case 512:	return PVRDMA_MTU_512;
+	case 1024:	return PVRDMA_MTU_1024;
+	case 2048:	return PVRDMA_MTU_2048;
+	case 4096:
+	default:	return PVRDMA_MTU_4096;
+	}
+}
+
+enum pvrdma_port_state {
+	PVRDMA_PORT_NOP			= 0,
+	PVRDMA_PORT_DOWN		= 1,
+	PVRDMA_PORT_INIT		= 2,
+	PVRDMA_PORT_ARMED		= 3,
+	PVRDMA_PORT_ACTIVE		= 4,
+	PVRDMA_PORT_ACTIVE_DEFER	= 5,
+};
+
+enum pvrdma_port_cap_flags {
+	PVRDMA_PORT_SM				= 1 <<  1,
+	PVRDMA_PORT_NOTICE_SUP			= 1 <<  2,
+	PVRDMA_PORT_TRAP_SUP			= 1 <<  3,
+	PVRDMA_PORT_OPT_IPD_SUP			= 1 <<  4,
+	PVRDMA_PORT_AUTO_MIGR_SUP		= 1 <<  5,
+	PVRDMA_PORT_SL_MAP_SUP			= 1 <<  6,
+	PVRDMA_PORT_MKEY_NVRAM			= 1 <<  7,
+	PVRDMA_PORT_PKEY_NVRAM			= 1 <<  8,
+	PVRDMA_PORT_LED_INFO_SUP		= 1 <<  9,
+	PVRDMA_PORT_SM_DISABLED			= 1 << 10,
+	PVRDMA_PORT_SYS_IMAGE_GUID_SUP		= 1 << 11,
+	PVRDMA_PORT_PKEY_SW_EXT_PORT_TRAP_SUP	= 1 << 12,
+	PVRDMA_PORT_EXTENDED_SPEEDS_SUP		= 1 << 14,
+	PVRDMA_PORT_CM_SUP			= 1 << 16,
+	PVRDMA_PORT_SNMP_TUNNEL_SUP		= 1 << 17,
+	PVRDMA_PORT_REINIT_SUP			= 1 << 18,
+	PVRDMA_PORT_DEVICE_MGMT_SUP		= 1 << 19,
+	PVRDMA_PORT_VENDOR_CLASS_SUP		= 1 << 20,
+	PVRDMA_PORT_DR_NOTICE_SUP		= 1 << 21,
+	PVRDMA_PORT_CAP_MASK_NOTICE_SUP		= 1 << 22,
+	PVRDMA_PORT_BOOT_MGMT_SUP		= 1 << 23,
+	PVRDMA_PORT_LINK_LATENCY_SUP		= 1 << 24,
+	PVRDMA_PORT_CLIENT_REG_SUP		= 1 << 25,
+	PVRDMA_PORT_IP_BASED_GIDS		= 1 << 26,
+	PVRDMA_PORT_CAP_FLAGS_MAX		= PVRDMA_PORT_IP_BASED_GIDS,
+};
+
+enum pvrdma_port_width {
+	PVRDMA_WIDTH_1X		= 1,
+	PVRDMA_WIDTH_4X		= 2,
+	PVRDMA_WIDTH_8X		= 4,
+	PVRDMA_WIDTH_12X	= 8,
+};
+
+static inline int pvrdma_width_enum_to_int(enum pvrdma_port_width width)
+{
+	switch (width) {
+	case PVRDMA_WIDTH_1X:	return  1;
+	case PVRDMA_WIDTH_4X:	return  4;
+	case PVRDMA_WIDTH_8X:	return  8;
+	case PVRDMA_WIDTH_12X:	return 12;
+	default:		return -1;
+	}
+}
+
+enum pvrdma_port_speed {
+	PVRDMA_SPEED_SDR	= 1,
+	PVRDMA_SPEED_DDR	= 2,
+	PVRDMA_SPEED_QDR	= 4,
+	PVRDMA_SPEED_FDR10	= 8,
+	PVRDMA_SPEED_FDR	= 16,
+	PVRDMA_SPEED_EDR	= 32,
+};
+
+struct pvrdma_port_attr {
+	enum pvrdma_port_state	state;
+	enum pvrdma_mtu		max_mtu;
+	enum pvrdma_mtu		active_mtu;
+	u32			gid_tbl_len;
+	u32			port_cap_flags;
+	u32			max_msg_sz;
+	u32			bad_pkey_cntr;
+	u32			qkey_viol_cntr;
+	u16			pkey_tbl_len;
+	u16			lid;
+	u16			sm_lid;
+	u8			lmc;
+	u8			max_vl_num;
+	u8			sm_sl;
+	u8			subnet_timeout;
+	u8			init_type_reply;
+	u8			active_width;
+	u8			active_speed;
+	u8			phys_state;
+	u8			reserved[2];
+};
+
+struct pvrdma_global_route {
+	union pvrdma_gid	dgid;
+	u32			flow_label;
+	u8			sgid_index;
+	u8			hop_limit;
+	u8			traffic_class;
+	u8			reserved;
+};
+
+struct pvrdma_grh {
+	__be32			version_tclass_flow;
+	__be16			paylen;
+	u8			next_hdr;
+	u8			hop_limit;
+	union pvrdma_gid	sgid;
+	union pvrdma_gid	dgid;
+};
+
+enum pvrdma_ah_flags {
+	PVRDMA_AH_GRH = 1,
+};
+
+enum pvrdma_rate {
+	PVRDMA_RATE_PORT_CURRENT	= 0,
+	PVRDMA_RATE_2_5_GBPS		= 2,
+	PVRDMA_RATE_5_GBPS		= 5,
+	PVRDMA_RATE_10_GBPS		= 3,
+	PVRDMA_RATE_20_GBPS		= 6,
+	PVRDMA_RATE_30_GBPS		= 4,
+	PVRDMA_RATE_40_GBPS		= 7,
+	PVRDMA_RATE_60_GBPS		= 8,
+	PVRDMA_RATE_80_GBPS		= 9,
+	PVRDMA_RATE_120_GBPS		= 10,
+	PVRDMA_RATE_14_GBPS		= 11,
+	PVRDMA_RATE_56_GBPS		= 12,
+	PVRDMA_RATE_112_GBPS		= 13,
+	PVRDMA_RATE_168_GBPS		= 14,
+	PVRDMA_RATE_25_GBPS		= 15,
+	PVRDMA_RATE_100_GBPS		= 16,
+	PVRDMA_RATE_200_GBPS		= 17,
+	PVRDMA_RATE_300_GBPS		= 18,
+};
+
+struct pvrdma_ah_attr {
+	struct pvrdma_global_route	grh;
+	u16				dlid;
+	u16				vlan_id;
+	u8				sl;
+	u8				src_path_bits;
+	u8				static_rate;
+	u8				ah_flags;
+	u8				port_num;
+	u8				dmac[6];
+	u8				reserved;
+};
+
+enum pvrdma_cq_notify_flags {
+	PVRDMA_CQ_SOLICITED		= 1 << 0,
+	PVRDMA_CQ_NEXT_COMP		= 1 << 1,
+	PVRDMA_CQ_SOLICITED_MASK	= PVRDMA_CQ_SOLICITED |
+					  PVRDMA_CQ_NEXT_COMP,
+	PVRDMA_CQ_REPORT_MISSED_EVENTS	= 1 << 2,
+};
+
+struct pvrdma_qp_cap {
+	u32	max_send_wr;
+	u32	max_recv_wr;
+	u32	max_send_sge;
+	u32	max_recv_sge;
+	u32	max_inline_data;
+	u32	reserved;
+};
+
+enum pvrdma_sig_type {
+	PVRDMA_SIGNAL_ALL_WR,
+	PVRDMA_SIGNAL_REQ_WR,
+};
+
+enum pvrdma_qp_type {
+	PVRDMA_QPT_SMI,
+	PVRDMA_QPT_GSI,
+	PVRDMA_QPT_RC,
+	PVRDMA_QPT_UC,
+	PVRDMA_QPT_UD,
+	PVRDMA_QPT_RAW_IPV6,
+	PVRDMA_QPT_RAW_ETHERTYPE,
+	PVRDMA_QPT_RAW_PACKET = 8,
+	PVRDMA_QPT_XRC_INI = 9,
+	PVRDMA_QPT_XRC_TGT,
+	PVRDMA_QPT_MAX,
+};
+
+enum pvrdma_qp_create_flags {
+	PVRDMA_QP_CREATE_IPOPVRDMA_UD_LSO		= 1 << 0,
+	PVRDMA_QP_CREATE_BLOCK_MULTICAST_LOOPBACK	= 1 << 1,
+};
+
+enum pvrdma_qp_attr_mask {
+	PVRDMA_QP_STATE			= 1 << 0,
+	PVRDMA_QP_CUR_STATE		= 1 << 1,
+	PVRDMA_QP_EN_SQD_ASYNC_NOTIFY	= 1 << 2,
+	PVRDMA_QP_ACCESS_FLAGS		= 1 << 3,
+	PVRDMA_QP_PKEY_INDEX		= 1 << 4,
+	PVRDMA_QP_PORT			= 1 << 5,
+	PVRDMA_QP_QKEY			= 1 << 6,
+	PVRDMA_QP_AV			= 1 << 7,
+	PVRDMA_QP_PATH_MTU		= 1 << 8,
+	PVRDMA_QP_TIMEOUT		= 1 << 9,
+	PVRDMA_QP_RETRY_CNT		= 1 << 10,
+	PVRDMA_QP_RNR_RETRY		= 1 << 11,
+	PVRDMA_QP_RQ_PSN		= 1 << 12,
+	PVRDMA_QP_MAX_QP_RD_ATOMIC	= 1 << 13,
+	PVRDMA_QP_ALT_PATH		= 1 << 14,
+	PVRDMA_QP_MIN_RNR_TIMER		= 1 << 15,
+	PVRDMA_QP_SQ_PSN		= 1 << 16,
+	PVRDMA_QP_MAX_DEST_RD_ATOMIC	= 1 << 17,
+	PVRDMA_QP_PATH_MIG_STATE	= 1 << 18,
+	PVRDMA_QP_CAP			= 1 << 19,
+	PVRDMA_QP_DEST_QPN		= 1 << 20,
+	PVRDMA_QP_ATTR_MASK_MAX		= PVRDMA_QP_DEST_QPN,
+};
+
+enum pvrdma_qp_state {
+	PVRDMA_QPS_RESET,
+	PVRDMA_QPS_INIT,
+	PVRDMA_QPS_RTR,
+	PVRDMA_QPS_RTS,
+	PVRDMA_QPS_SQD,
+	PVRDMA_QPS_SQE,
+	PVRDMA_QPS_ERR,
+};
+
+enum pvrdma_mig_state {
+	PVRDMA_MIG_MIGRATED,
+	PVRDMA_MIG_REARM,
+	PVRDMA_MIG_ARMED,
+};
+
+enum pvrdma_mw_type {
+	PVRDMA_MW_TYPE_1 = 1,
+	PVRDMA_MW_TYPE_2 = 2,
+};
+
+struct pvrdma_qp_attr {
+	enum pvrdma_qp_state	qp_state;
+	enum pvrdma_qp_state	cur_qp_state;
+	enum pvrdma_mtu		path_mtu;
+	enum pvrdma_mig_state	path_mig_state;
+	u32			qkey;
+	u32			rq_psn;
+	u32			sq_psn;
+	u32			dest_qp_num;
+	u32			qp_access_flags;
+	u16			pkey_index;
+	u16			alt_pkey_index;
+	u8			en_sqd_async_notify;
+	u8			sq_draining;
+	u8			max_rd_atomic;
+	u8			max_dest_rd_atomic;
+	u8			min_rnr_timer;
+	u8			port_num;
+	u8			timeout;
+	u8			retry_cnt;
+	u8			rnr_retry;
+	u8			alt_port_num;
+	u8			alt_timeout;
+	u8			reserved[5];
+	struct pvrdma_qp_cap	cap;
+	struct pvrdma_ah_attr	ah_attr;
+	struct pvrdma_ah_attr	alt_ah_attr;
+};
+
+enum pvrdma_send_flags {
+	PVRDMA_SEND_FENCE	= 1 << 0,
+	PVRDMA_SEND_SIGNALED	= 1 << 1,
+	PVRDMA_SEND_SOLICITED	= 1 << 2,
+	PVRDMA_SEND_INLINE	= 1 << 3,
+	PVRDMA_SEND_IP_CSUM	= 1 << 4,
+	PVRDMA_SEND_FLAGS_MAX	= PVRDMA_SEND_IP_CSUM,
+};
+
+enum pvrdma_access_flags {
+	PVRDMA_ACCESS_LOCAL_WRITE	= 1 << 0,
+	PVRDMA_ACCESS_REMOTE_WRITE	= 1 << 1,
+	PVRDMA_ACCESS_REMOTE_READ	= 1 << 2,
+	PVRDMA_ACCESS_REMOTE_ATOMIC	= 1 << 3,
+	PVRDMA_ACCESS_MW_BIND		= 1 << 4,
+	PVRDMA_ZERO_BASED		= 1 << 5,
+	PVRDMA_ACCESS_ON_DEMAND		= 1 << 6,
+	PVRDMA_ACCESS_FLAGS_MAX		= PVRDMA_ACCESS_ON_DEMAND,
+};
+
+int pvrdma_query_device(struct ib_device *ibdev,
+			struct ib_device_attr *props,
+			struct ib_udata *udata);
+int pvrdma_query_port(struct ib_device *ibdev, u8 port,
+		      struct ib_port_attr *props);
+int pvrdma_query_gid(struct ib_device *ibdev, u8 port,
+		     int index, union ib_gid *gid);
+int pvrdma_query_pkey(struct ib_device *ibdev, u8 port,
+		      u16 index, u16 *pkey);
+enum rdma_link_layer pvrdma_port_link_layer(struct ib_device *ibdev,
+					    u8 port);
+int pvrdma_modify_device(struct ib_device *ibdev, int mask,
+			 struct ib_device_modify *props);
+int pvrdma_modify_port(struct ib_device *ibdev, u8 port,
+		       int mask, struct ib_port_modify *props);
+int pvrdma_mmap(struct ib_ucontext *context, struct vm_area_struct *vma);
+struct ib_ucontext *pvrdma_alloc_ucontext(struct ib_device *ibdev,
+					  struct ib_udata *udata);
+int pvrdma_dealloc_ucontext(struct ib_ucontext *context);
+struct ib_pd *pvrdma_alloc_pd(struct ib_device *ibdev,
+			      struct ib_ucontext *context,
+			      struct ib_udata *udata);
+int pvrdma_dealloc_pd(struct ib_pd *ibpd);
+struct ib_mr *pvrdma_get_dma_mr(struct ib_pd *pd, int acc);
+struct ib_mr *pvrdma_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
+				 u64 virt_addr, int access_flags,
+				 struct ib_udata *udata);
+int pvrdma_dereg_mr(struct ib_mr *mr);
+struct ib_mr *pvrdma_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
+			      u32 max_num_sg);
+int pvrdma_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg,
+		     int sg_nents, unsigned int *sg_offset);
+int pvrdma_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
+int pvrdma_resize_cq(struct ib_cq *ibcq, int entries,
+		     struct ib_udata *udata);
+struct ib_cq *pvrdma_create_cq(struct ib_device *ibdev,
+			       const struct ib_cq_init_attr *attr,
+			       struct ib_ucontext *context,
+			       struct ib_udata *udata);
+int pvrdma_resize_cq(struct ib_cq *ibcq, int entries,
+		     struct ib_udata *udata);
+int pvrdma_destroy_cq(struct ib_cq *cq);
+int pvrdma_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
+int pvrdma_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags);
+struct ib_ah *pvrdma_create_ah(struct ib_pd *pd, struct ib_ah_attr *ah_attr);
+int pvrdma_destroy_ah(struct ib_ah *ah);
+struct ib_qp *pvrdma_create_qp(struct ib_pd *pd,
+			       struct ib_qp_init_attr *init_attr,
+			       struct ib_udata *udata);
+int pvrdma_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
+		     int attr_mask, struct ib_udata *udata);
+int pvrdma_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
+		    int qp_attr_mask, struct ib_qp_init_attr *qp_init_attr);
+int pvrdma_destroy_qp(struct ib_qp *qp);
+int pvrdma_post_send(struct ib_qp *ibqp, struct ib_send_wr *wr,
+		     struct ib_send_wr **bad_wr);
+int pvrdma_post_recv(struct ib_qp *ibqp, struct ib_recv_wr *wr,
+		     struct ib_recv_wr **bad_wr);
+
+#endif /* __PVRDMA_VERBS_H__ */