IB: Add vmw_pvrdma driver

This patch series adds a driver for a paravirtual RDMA device. The
device is developed for VMware's Virtual Machines and allows existing RDMA
applications to continue to use existing Verbs API when deployed in VMs
on ESXi. We recently did a presentation in the OFA Workshop [1] regarding
this device.

Description and RDMA Support
============================
The virtual device is exposed as a dual function PCIe device. One part
is a virtual network device (VMXNet3) which provides networking properties
like MAC, IP addresses to the RDMA part of the device. The networking
properties are used to register GIDs required by RDMA applications to
communicate.

These patches add support and the all required infrastructure for
letting applications use such a device. We support the mandatory Verbs API as
well as the base memory management extensions (Local Inv, Send with Inv and
Fast Register Work Requests). We currently support both Reliable Connected
and Unreliable Datagram QPs but do not support Shared Receive Queues
(SRQs).

Also, we support the following types of Work Requests:
 o Send/Receive (with or without Immediate Data)
 o RDMA Write (with or without Immediate Data)
 o RDMA Read
 o Local Invalidate
 o Send with Invalidate
 o Fast Register Work Requests

This version only adds support for version 1 of RoCE. We will add RoCEv2
support in a future patch. We do support registration of both MAC-based
and IP-based GIDs. I have also created a git tree for our user-level driver
[2].

Testing
=======
We have tested this internally for various types of Guest OS - Red Hat,
Centos, Ubuntu 12.04/14.04/16.04, Oracle Enterprise Linux, SLES 12
using backported versions of this driver. The tests included several
runs of the performance tests (included with OFED), Intel MPI PingPong
benchmark on OpenMPI, krping for FRWRs. Mellanox has been kind enough
to test the backported version of the driver internally on their hardware
using a VMware provided ESX build. I have also applied and tested this
with Doug's k.o/for-4.9 branch (commit 5603910b). Note, that this patch
series should be applied all together. I split out the commits so that
it may be easier to review.

PVRDMA Resources
================
[1] OFA Workshop Presentation -
https://openfabrics.org/images/eventpresos/2016presentations/102parardma.pdf

[2] Libpvrdma User-level library -
http://git.openfabrics.org/?p=~aditr/libpvrdma.git;a=summary

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>
Reviewed-by: Leon Romanovsky <leonro@mellanox.com>
Signed-off-by: Adit Ranadive <aditr@vmware.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

1 files changed, 579 insertions, 0 deletions

diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c
new file mode 100644
index 000000000000..54891370d18a
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_verbs.c
@@ -0,0 +1,579 @@
+/*
+ * 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/vmw_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
+ * @udata: user data blob
+ *
+ * @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 ib_udata *udata)
+{
+	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;
+}