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>

16 files changed, 5710 insertions, 0 deletions

diff --git a/drivers/infiniband/Kconfig b/drivers/infiniband/Kconfig
index fb3fb89640e5..670917387eda 100644
--- a/drivers/infiniband/Kconfig
+++ b/drivers/infiniband/Kconfig
@@ -73,6 +73,7 @@ source "drivers/infiniband/hw/mlx4/Kconfig"
 source "drivers/infiniband/hw/mlx5/Kconfig"
 source "drivers/infiniband/hw/nes/Kconfig"
 source "drivers/infiniband/hw/ocrdma/Kconfig"
+source "drivers/infiniband/hw/vmw_pvrdma/Kconfig"
 source "drivers/infiniband/hw/usnic/Kconfig"
 source "drivers/infiniband/hw/hns/Kconfig"
 
diff --git a/drivers/infiniband/hw/Makefile b/drivers/infiniband/hw/Makefile
index e7a5ed9f6f3f..ed553de2ca12 100644
--- a/drivers/infiniband/hw/Makefile
+++ b/drivers/infiniband/hw/Makefile
@@ -7,6 +7,7 @@ obj-$(CONFIG_MLX4_INFINIBAND)		+= mlx4/
 obj-$(CONFIG_MLX5_INFINIBAND)		+= mlx5/
 obj-$(CONFIG_INFINIBAND_NES)		+= nes/
 obj-$(CONFIG_INFINIBAND_OCRDMA)		+= ocrdma/
+obj-$(CONFIG_INFINIBAND_VMWARE_PVRDMA)	+= vmw_pvrdma/
 obj-$(CONFIG_INFINIBAND_USNIC)		+= usnic/
 obj-$(CONFIG_INFINIBAND_HFI1)		+= hfi1/
 obj-$(CONFIG_INFINIBAND_HNS)		+= hns/
diff --git a/drivers/infiniband/hw/vmw_pvrdma/Kconfig b/drivers/infiniband/hw/vmw_pvrdma/Kconfig
new file mode 100644
index 000000000000..5a9790ac0ede
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/Kconfig
@@ -0,0 +1,7 @@
+config INFINIBAND_VMWARE_PVRDMA
+	tristate "VMware Paravirtualized RDMA Driver"
+	depends on NETDEVICES && ETHERNET && PCI && INET && VMXNET3
+	---help---
+	  This driver provides low-level support for VMware Paravirtual
+	  RDMA adapter. It interacts with the VMXNet3 driver to provide
+	  Ethernet capabilities.
diff --git a/drivers/infiniband/hw/vmw_pvrdma/Makefile b/drivers/infiniband/hw/vmw_pvrdma/Makefile
new file mode 100644
index 000000000000..0194ed19f542
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/Makefile
@@ -0,0 +1,3 @@
+obj-$(CONFIG_INFINIBAND_VMWARE_PVRDMA) += vmw_pvrdma.o
+
+vmw_pvrdma-y := pvrdma_cmd.o pvrdma_cq.o pvrdma_doorbell.o pvrdma_main.o pvrdma_misc.o pvrdma_mr.o pvrdma_qp.o pvrdma_verbs.o
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
new file mode 100644
index 000000000000..71e1d55d69d6
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma.h
@@ -0,0 +1,474 @@
+/*
+ * 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_H__
+#define __PVRDMA_H__
+
+#include <linux/compiler.h>
+#include <linux/interrupt.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/pci.h>
+#include <linux/semaphore.h>
+#include <linux/workqueue.h>
+#include <rdma/ib_umem.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/vmw_pvrdma-abi.h>
+
+#include "pvrdma_ring.h"
+#include "pvrdma_dev_api.h"
+#include "pvrdma_verbs.h"
+
+/* NOT the same as BIT_MASK(). */
+#define PVRDMA_MASK(n) ((n << 1) - 1)
+
+/*
+ * VMware PVRDMA PCI device id.
+ */
+#define PCI_DEVICE_ID_VMWARE_PVRDMA	0x0820
+
+struct pvrdma_dev;
+
+struct pvrdma_page_dir {
+	dma_addr_t dir_dma;
+	u64 *dir;
+	int ntables;
+	u64 **tables;
+	u64 npages;
+	void **pages;
+};
+
+struct pvrdma_cq {
+	struct ib_cq ibcq;
+	int offset;
+	spinlock_t cq_lock; /* Poll lock. */
+	struct pvrdma_uar_map *uar;
+	struct ib_umem *umem;
+	struct pvrdma_ring_state *ring_state;
+	struct pvrdma_page_dir pdir;
+	u32 cq_handle;
+	bool is_kernel;
+	atomic_t refcnt;
+	wait_queue_head_t wait;
+};
+
+struct pvrdma_id_table {
+	u32 last;
+	u32 top;
+	u32 max;
+	u32 mask;
+	spinlock_t lock; /* Table lock. */
+	unsigned long *table;
+};
+
+struct pvrdma_uar_map {
+	unsigned long pfn;
+	void __iomem *map;
+	int index;
+};
+
+struct pvrdma_uar_table {
+	struct pvrdma_id_table tbl;
+	int size;
+};
+
+struct pvrdma_ucontext {
+	struct ib_ucontext ibucontext;
+	struct pvrdma_dev *dev;
+	struct pvrdma_uar_map uar;
+	u64 ctx_handle;
+};
+
+struct pvrdma_pd {
+	struct ib_pd ibpd;
+	u32 pdn;
+	u32 pd_handle;
+	int privileged;
+};
+
+struct pvrdma_mr {
+	u32 mr_handle;
+	u64 iova;
+	u64 size;
+};
+
+struct pvrdma_user_mr {
+	struct ib_mr ibmr;
+	struct ib_umem *umem;
+	struct pvrdma_mr mmr;
+	struct pvrdma_page_dir pdir;
+	u64 *pages;
+	u32 npages;
+	u32 max_pages;
+	u32 page_shift;
+};
+
+struct pvrdma_wq {
+	struct pvrdma_ring *ring;
+	spinlock_t lock; /* Work queue lock. */
+	int wqe_cnt;
+	int wqe_size;
+	int max_sg;
+	int offset;
+};
+
+struct pvrdma_ah {
+	struct ib_ah ibah;
+	struct pvrdma_av av;
+};
+
+struct pvrdma_qp {
+	struct ib_qp ibqp;
+	u32 qp_handle;
+	u32 qkey;
+	struct pvrdma_wq sq;
+	struct pvrdma_wq rq;
+	struct ib_umem *rumem;
+	struct ib_umem *sumem;
+	struct pvrdma_page_dir pdir;
+	int npages;
+	int npages_send;
+	int npages_recv;
+	u32 flags;
+	u8 port;
+	u8 state;
+	bool is_kernel;
+	struct mutex mutex; /* QP state mutex. */
+	atomic_t refcnt;
+	wait_queue_head_t wait;
+};
+
+struct pvrdma_dev {
+	/* PCI device-related information. */
+	struct ib_device ib_dev;
+	struct pci_dev *pdev;
+	void __iomem *regs;
+	struct pvrdma_device_shared_region *dsr; /* Shared region pointer */
+	dma_addr_t dsrbase; /* Shared region base address */
+	void *cmd_slot;
+	void *resp_slot;
+	unsigned long flags;
+	struct list_head device_link;
+
+	/* Locking and interrupt information. */
+	spinlock_t cmd_lock; /* Command lock. */
+	struct semaphore cmd_sema;
+	struct completion cmd_done;
+	struct {
+		enum pvrdma_intr_type type; /* Intr type */
+		struct msix_entry msix_entry[PVRDMA_MAX_INTERRUPTS];
+		irq_handler_t handler[PVRDMA_MAX_INTERRUPTS];
+		u8 enabled[PVRDMA_MAX_INTERRUPTS];
+		u8 size;
+	} intr;
+
+	/* RDMA-related device information. */
+	union ib_gid *sgid_tbl;
+	struct pvrdma_ring_state *async_ring_state;
+	struct pvrdma_page_dir async_pdir;
+	struct pvrdma_ring_state *cq_ring_state;
+	struct pvrdma_page_dir cq_pdir;
+	struct pvrdma_cq **cq_tbl;
+	spinlock_t cq_tbl_lock;
+	struct pvrdma_qp **qp_tbl;
+	spinlock_t qp_tbl_lock;
+	struct pvrdma_uar_table uar_table;
+	struct pvrdma_uar_map driver_uar;
+	__be64 sys_image_guid;
+	spinlock_t desc_lock; /* Device modification lock. */
+	u32 port_cap_mask;
+	struct mutex port_mutex; /* Port modification mutex. */
+	bool ib_active;
+	atomic_t num_qps;
+	atomic_t num_cqs;
+	atomic_t num_pds;
+	atomic_t num_ahs;
+
+	/* Network device information. */
+	struct net_device *netdev;
+	struct notifier_block nb_netdev;
+};
+
+struct pvrdma_netdevice_work {
+	struct work_struct work;
+	struct net_device *event_netdev;
+	unsigned long event;
+};
+
+static inline struct pvrdma_dev *to_vdev(struct ib_device *ibdev)
+{
+	return container_of(ibdev, struct pvrdma_dev, ib_dev);
+}
+
+static inline struct
+pvrdma_ucontext *to_vucontext(struct ib_ucontext *ibucontext)
+{
+	return container_of(ibucontext, struct pvrdma_ucontext, ibucontext);
+}
+
+static inline struct pvrdma_pd *to_vpd(struct ib_pd *ibpd)
+{
+	return container_of(ibpd, struct pvrdma_pd, ibpd);
+}
+
+static inline struct pvrdma_cq *to_vcq(struct ib_cq *ibcq)
+{
+	return container_of(ibcq, struct pvrdma_cq, ibcq);
+}
+
+static inline struct pvrdma_user_mr *to_vmr(struct ib_mr *ibmr)
+{
+	return container_of(ibmr, struct pvrdma_user_mr, ibmr);
+}
+
+static inline struct pvrdma_qp *to_vqp(struct ib_qp *ibqp)
+{
+	return container_of(ibqp, struct pvrdma_qp, ibqp);
+}
+
+static inline struct pvrdma_ah *to_vah(struct ib_ah *ibah)
+{
+	return container_of(ibah, struct pvrdma_ah, ibah);
+}
+
+static inline void pvrdma_write_reg(struct pvrdma_dev *dev, u32 reg, u32 val)
+{
+	writel(cpu_to_le32(val), dev->regs + reg);
+}
+
+static inline u32 pvrdma_read_reg(struct pvrdma_dev *dev, u32 reg)
+{
+	return le32_to_cpu(readl(dev->regs + reg));
+}
+
+static inline void pvrdma_write_uar_cq(struct pvrdma_dev *dev, u32 val)
+{
+	writel(cpu_to_le32(val), dev->driver_uar.map + PVRDMA_UAR_CQ_OFFSET);
+}
+
+static inline void pvrdma_write_uar_qp(struct pvrdma_dev *dev, u32 val)
+{
+	writel(cpu_to_le32(val), dev->driver_uar.map + PVRDMA_UAR_QP_OFFSET);
+}
+
+static inline void *pvrdma_page_dir_get_ptr(struct pvrdma_page_dir *pdir,
+					    u64 offset)
+{
+	return pdir->pages[offset / PAGE_SIZE] + (offset % PAGE_SIZE);
+}
+
+static inline enum pvrdma_mtu ib_mtu_to_pvrdma(enum ib_mtu mtu)
+{
+	return (enum pvrdma_mtu)mtu;
+}
+
+static inline enum ib_mtu pvrdma_mtu_to_ib(enum pvrdma_mtu mtu)
+{
+	return (enum ib_mtu)mtu;
+}
+
+static inline enum pvrdma_port_state ib_port_state_to_pvrdma(
+					enum ib_port_state state)
+{
+	return (enum pvrdma_port_state)state;
+}
+
+static inline enum ib_port_state pvrdma_port_state_to_ib(
+					enum pvrdma_port_state state)
+{
+	return (enum ib_port_state)state;
+}
+
+static inline int ib_port_cap_flags_to_pvrdma(int flags)
+{
+	return flags & PVRDMA_MASK(PVRDMA_PORT_CAP_FLAGS_MAX);
+}
+
+static inline int pvrdma_port_cap_flags_to_ib(int flags)
+{
+	return flags;
+}
+
+static inline enum pvrdma_port_width ib_port_width_to_pvrdma(
+					enum ib_port_width width)
+{
+	return (enum pvrdma_port_width)width;
+}
+
+static inline enum ib_port_width pvrdma_port_width_to_ib(
+					enum pvrdma_port_width width)
+{
+	return (enum ib_port_width)width;
+}
+
+static inline enum pvrdma_port_speed ib_port_speed_to_pvrdma(
+					enum ib_port_speed speed)
+{
+	return (enum pvrdma_port_speed)speed;
+}
+
+static inline enum ib_port_speed pvrdma_port_speed_to_ib(
+					enum pvrdma_port_speed speed)
+{
+	return (enum ib_port_speed)speed;
+}
+
+static inline int pvrdma_qp_attr_mask_to_ib(int attr_mask)
+{
+	return attr_mask;
+}
+
+static inline int ib_qp_attr_mask_to_pvrdma(int attr_mask)
+{
+	return attr_mask & PVRDMA_MASK(PVRDMA_QP_ATTR_MASK_MAX);
+}
+
+static inline enum pvrdma_mig_state ib_mig_state_to_pvrdma(
+					enum ib_mig_state state)
+{
+	return (enum pvrdma_mig_state)state;
+}
+
+static inline enum ib_mig_state pvrdma_mig_state_to_ib(
+					enum pvrdma_mig_state state)
+{
+	return (enum ib_mig_state)state;
+}
+
+static inline int ib_access_flags_to_pvrdma(int flags)
+{
+	return flags;
+}
+
+static inline int pvrdma_access_flags_to_ib(int flags)
+{
+	return flags & PVRDMA_MASK(PVRDMA_ACCESS_FLAGS_MAX);
+}
+
+static inline enum pvrdma_qp_type ib_qp_type_to_pvrdma(enum ib_qp_type type)
+{
+	return (enum pvrdma_qp_type)type;
+}
+
+static inline enum ib_qp_type pvrdma_qp_type_to_ib(enum pvrdma_qp_type type)
+{
+	return (enum ib_qp_type)type;
+}
+
+static inline enum pvrdma_qp_state ib_qp_state_to_pvrdma(enum ib_qp_state state)
+{
+	return (enum pvrdma_qp_state)state;
+}
+
+static inline enum ib_qp_state pvrdma_qp_state_to_ib(enum pvrdma_qp_state state)
+{
+	return (enum ib_qp_state)state;
+}
+
+static inline enum pvrdma_wr_opcode ib_wr_opcode_to_pvrdma(enum ib_wr_opcode op)
+{
+	return (enum pvrdma_wr_opcode)op;
+}
+
+static inline enum ib_wc_status pvrdma_wc_status_to_ib(
+					enum pvrdma_wc_status status)
+{
+	return (enum ib_wc_status)status;
+}
+
+static inline int pvrdma_wc_opcode_to_ib(int opcode)
+{
+	return opcode;
+}
+
+static inline int pvrdma_wc_flags_to_ib(int flags)
+{
+	return flags;
+}
+
+static inline int ib_send_flags_to_pvrdma(int flags)
+{
+	return flags & PVRDMA_MASK(PVRDMA_SEND_FLAGS_MAX);
+}
+
+void pvrdma_qp_cap_to_ib(struct ib_qp_cap *dst,
+			 const struct pvrdma_qp_cap *src);
+void ib_qp_cap_to_pvrdma(struct pvrdma_qp_cap *dst,
+			 const struct ib_qp_cap *src);
+void pvrdma_gid_to_ib(union ib_gid *dst, const union pvrdma_gid *src);
+void ib_gid_to_pvrdma(union pvrdma_gid *dst, const union ib_gid *src);
+void pvrdma_global_route_to_ib(struct ib_global_route *dst,
+			       const struct pvrdma_global_route *src);
+void ib_global_route_to_pvrdma(struct pvrdma_global_route *dst,
+			       const struct ib_global_route *src);
+void pvrdma_ah_attr_to_ib(struct ib_ah_attr *dst,
+			  const struct pvrdma_ah_attr *src);
+void ib_ah_attr_to_pvrdma(struct pvrdma_ah_attr *dst,
+			  const struct ib_ah_attr *src);
+
+int pvrdma_uar_table_init(struct pvrdma_dev *dev);
+void pvrdma_uar_table_cleanup(struct pvrdma_dev *dev);
+
+int pvrdma_uar_alloc(struct pvrdma_dev *dev, struct pvrdma_uar_map *uar);
+void pvrdma_uar_free(struct pvrdma_dev *dev, struct pvrdma_uar_map *uar);
+
+void _pvrdma_flush_cqe(struct pvrdma_qp *qp, struct pvrdma_cq *cq);
+
+int pvrdma_page_dir_init(struct pvrdma_dev *dev, struct pvrdma_page_dir *pdir,
+			 u64 npages, bool alloc_pages);
+void pvrdma_page_dir_cleanup(struct pvrdma_dev *dev,
+			     struct pvrdma_page_dir *pdir);
+int pvrdma_page_dir_insert_dma(struct pvrdma_page_dir *pdir, u64 idx,
+			       dma_addr_t daddr);
+int pvrdma_page_dir_insert_umem(struct pvrdma_page_dir *pdir,
+				struct ib_umem *umem, u64 offset);
+dma_addr_t pvrdma_page_dir_get_dma(struct pvrdma_page_dir *pdir, u64 idx);
+int pvrdma_page_dir_insert_page_list(struct pvrdma_page_dir *pdir,
+				     u64 *page_list, int num_pages);
+
+int pvrdma_cmd_post(struct pvrdma_dev *dev, union pvrdma_cmd_req *req,
+		    union pvrdma_cmd_resp *rsp, unsigned resp_code);
+
+#endif /* __PVRDMA_H__ */
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cmd.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cmd.c
new file mode 100644
index 000000000000..4a78c537d8a1
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cmd.c
@@ -0,0 +1,119 @@
+/*
+ * 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 <linux/list.h>
+
+#include "pvrdma.h"
+
+#define PVRDMA_CMD_TIMEOUT	10000 /* ms */
+
+static inline int pvrdma_cmd_recv(struct pvrdma_dev *dev,
+				  union pvrdma_cmd_resp *resp,
+				  unsigned resp_code)
+{
+	int err;
+
+	dev_dbg(&dev->pdev->dev, "receive response from device\n");
+
+	err = wait_for_completion_interruptible_timeout(&dev->cmd_done,
+			msecs_to_jiffies(PVRDMA_CMD_TIMEOUT));
+	if (err == 0 || err == -ERESTARTSYS) {
+		dev_warn(&dev->pdev->dev,
+			 "completion timeout or interrupted\n");
+		return -ETIMEDOUT;
+	}
+
+	spin_lock(&dev->cmd_lock);
+	memcpy(resp, dev->resp_slot, sizeof(*resp));
+	spin_unlock(&dev->cmd_lock);
+
+	if (resp->hdr.ack != resp_code) {
+		dev_warn(&dev->pdev->dev,
+			 "unknown response %#x expected %#x\n",
+			 resp->hdr.ack, resp_code);
+		return -EFAULT;
+	}
+
+	return 0;
+}
+
+int
+pvrdma_cmd_post(struct pvrdma_dev *dev, union pvrdma_cmd_req *req,
+		union pvrdma_cmd_resp *resp, unsigned resp_code)
+{
+	int err;
+
+	dev_dbg(&dev->pdev->dev, "post request to device\n");
+
+	/* Serializiation */
+	down(&dev->cmd_sema);
+
+	BUILD_BUG_ON(sizeof(union pvrdma_cmd_req) !=
+		     sizeof(struct pvrdma_cmd_modify_qp));
+
+	spin_lock(&dev->cmd_lock);
+	memcpy(dev->cmd_slot, req, sizeof(*req));
+	spin_unlock(&dev->cmd_lock);
+
+	init_completion(&dev->cmd_done);
+	pvrdma_write_reg(dev, PVRDMA_REG_REQUEST, 0);
+
+	/* Make sure the request is written before reading status. */
+	mb();
+
+	err = pvrdma_read_reg(dev, PVRDMA_REG_ERR);
+	if (err == 0) {
+		if (resp != NULL)
+			err = pvrdma_cmd_recv(dev, resp, resp_code);
+	} else {
+		dev_warn(&dev->pdev->dev,
+			 "failed to write request error reg: %d\n", err);
+		err = -EFAULT;
+	}
+
+	up(&dev->cmd_sema);
+
+	return err;
+}
diff --git a/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c
new file mode 100644
index 000000000000..e429ca5b16aa
--- /dev/null
+++ b/drivers/infiniband/hw/vmw_pvrdma/pvrdma_cq.c
@@ -0,0 +1,425 @@
+/*
+ * 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/io.h>
+#include <linux/wait.h>
+#include <rdma/ib_addr.h>
+#include <rdma/ib_smi.h>
+#include <rdma/ib_user_verbs.h>
+
+#include "pvrdma.h"
+
+/**
+ * pvrdma_req_notify_cq - request notification for a completion queue
+ * @ibcq: the completion queue
+ * @notify_flags: notification flags
+ *
+ * @return: 0 for success.
+ */
+int pvrdma_req_notify_cq(struct ib_cq *ibcq,
+			 enum ib_cq_notify_flags notify_flags)
+{
+	struct pvrdma_dev *dev = to_vdev(ibcq->device);
+	struct pvrdma_cq *cq = to_vcq(ibcq);
+	u32 val = cq->cq_handle;
+
+	val |= (notify_flags & IB_CQ_SOLICITED_MASK) == IB_CQ_SOLICITED ?
+		PVRDMA_UAR_CQ_ARM_SOL : PVRDMA_UAR_CQ_ARM;
+
+	pvrdma_write_uar_cq(dev, val);
+
+	return 0;
+}
+
+/**
+ * pvrdma_create_cq - create completion queue
+ * @ibdev: the device
+ * @attr: completion queue attributes
+ * @context: user context
+ * @udata: user data
+ *
+ * @return: ib_cq completion queue pointer on success,
+ *          otherwise returns negative errno.
+ */
+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 entries = attr->cqe;
+	struct pvrdma_dev *dev = to_vdev(ibdev);
+	struct pvrdma_cq *cq;
+	int ret;
+	int npages;
+	unsigned long flags;
+	union pvrdma_cmd_req req;
+	union pvrdma_cmd_resp rsp;
+	struct pvrdma_cmd_create_cq *cmd = &req.create_cq;
+	struct pvrdma_cmd_create_cq_resp *resp = &rsp.create_cq_resp;
+	struct pvrdma_create_cq ucmd;
+
+	BUILD_BUG_ON(sizeof(struct pvrdma_cqe) != 64);
+
+	entries = roundup_pow_of_two(entries);
+	if (entries < 1 || entries > dev->dsr->caps.max_cqe)
+		return ERR_PTR(-EINVAL);
+
+	if (!atomic_add_unless(&dev->num_cqs, 1, dev->dsr->caps.max_cq))
+		return ERR_PTR(-ENOMEM);
+
+	cq = kzalloc(sizeof(*cq), GFP_KERNEL);
+	if (!cq) {
+		atomic_dec(&dev->num_cqs);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	cq->ibcq.cqe = entries;
+
+	if (context) {
+		if (ib_copy_from_udata(&ucmd, udata, sizeof(ucmd))) {
+			ret = -EFAULT;
+			goto err_cq;
+		}
+
+		cq->umem = ib_umem_get(context, ucmd.buf_addr, ucmd.buf_size,
+				       IB_ACCESS_LOCAL_WRITE, 1);
+		if (IS_ERR(cq->umem)) {
+			ret = PTR_ERR(cq->umem);
+			goto err_cq;
+		}
+
+		npages = ib_umem_page_count(cq->umem);
+	} else {
+		cq->is_kernel = true;
+
+		/* One extra page for shared ring state */
+		npages = 1 + (entries * sizeof(struct pvrdma_cqe) +
+			      PAGE_SIZE - 1) / PAGE_SIZE;
+
+		/* Skip header page. */
+		cq->offset = PAGE_SIZE;
+	}
+
+	if (npages < 0 || npages > PVRDMA_PAGE_DIR_MAX_PAGES) {
+		dev_warn(&dev->pdev->dev,
+			 "overflow pages in completion queue\n");
+		ret = -EINVAL;
+		goto err_umem;
+	}
+
+	ret = pvrdma_page_dir_init(dev, &cq->pdir, npages, cq->is_kernel);
+	if (ret) {
+		dev_warn(&dev->pdev->dev,
+			 "could not allocate page directory\n");
+		goto err_umem;
+	}
+
+	/* Ring state is always the first page. Set in library for user cq. */
+	if (cq->is_kernel)
+		cq->ring_state = cq->pdir.pages[0];
+	else
+		pvrdma_page_dir_insert_umem(&cq->pdir, cq->umem, 0);
+
+	atomic_set(&cq->refcnt, 1);
+	init_waitqueue_head(&cq->wait);
+	spin_lock_init(&cq->cq_lock);
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_CREATE_CQ;
+	cmd->nchunks = npages;
+	cmd->ctx_handle = (context) ?
+		(u64)to_vucontext(context)->ctx_handle : 0;
+	cmd->cqe = entries;
+	cmd->pdir_dma = cq->pdir.dir_dma;
+	ret = pvrdma_cmd_post(dev, &req, &rsp, PVRDMA_CMD_CREATE_CQ_RESP);
+	if (ret < 0) {
+		dev_warn(&dev->pdev->dev,
+			 "could not create completion queue, error: %d\n", ret);
+		goto err_page_dir;
+	}
+
+	cq->ibcq.cqe = resp->cqe;
+	cq->cq_handle = resp->cq_handle;
+	spin_lock_irqsave(&dev->cq_tbl_lock, flags);
+	dev->cq_tbl[cq->cq_handle % dev->dsr->caps.max_cq] = cq;
+	spin_unlock_irqrestore(&dev->cq_tbl_lock, flags);
+
+	if (context) {
+		cq->uar = &(to_vucontext(context)->uar);
+
+		/* Copy udata back. */
+		if (ib_copy_to_udata(udata, &cq->cq_handle, sizeof(__u32))) {
+			dev_warn(&dev->pdev->dev,
+				 "failed to copy back udata\n");
+			pvrdma_destroy_cq(&cq->ibcq);
+			return ERR_PTR(-EINVAL);
+		}
+	}
+
+	return &cq->ibcq;
+
+err_page_dir:
+	pvrdma_page_dir_cleanup(dev, &cq->pdir);
+err_umem:
+	if (context)
+		ib_umem_release(cq->umem);
+err_cq:
+	atomic_dec(&dev->num_cqs);
+	kfree(cq);
+
+	return ERR_PTR(ret);
+}
+
+static void pvrdma_free_cq(struct pvrdma_dev *dev, struct pvrdma_cq *cq)
+{
+	atomic_dec(&cq->refcnt);
+	wait_event(cq->wait, !atomic_read(&cq->refcnt));
+
+	if (!cq->is_kernel)
+		ib_umem_release(cq->umem);
+
+	pvrdma_page_dir_cleanup(dev, &cq->pdir);
+	kfree(cq);
+}
+
+/**
+ * pvrdma_destroy_cq - destroy completion queue
+ * @cq: the completion queue to destroy.
+ *
+ * @return: 0 for success.
+ */
+int pvrdma_destroy_cq(struct ib_cq *cq)
+{
+	struct pvrdma_cq *vcq = to_vcq(cq);
+	union pvrdma_cmd_req req;
+	struct pvrdma_cmd_destroy_cq *cmd = &req.destroy_cq;
+	struct pvrdma_dev *dev = to_vdev(cq->device);
+	unsigned long flags;
+	int ret;
+
+	memset(cmd, 0, sizeof(*cmd));
+	cmd->hdr.cmd = PVRDMA_CMD_DESTROY_CQ;
+	cmd->cq_handle = vcq->cq_handle;
+
+	ret = pvrdma_cmd_post(dev, &req, NULL, 0);
+	if (ret < 0)
+		dev_warn(&dev->pdev->dev,
+			 "could not destroy completion queue, error: %d\n",
+			 ret);
+
+	/* free cq's resources */
+	spin_lock_irqsave(&dev->cq_tbl_lock, flags);
+	dev->cq_tbl[vcq->cq_handle] = NULL;
+	spin_unlock_irqrestore(&dev->cq_tbl_lock, flags);
+
+	pvrdma_free_cq(dev, vcq);
+	atomic_dec(&dev->num_cqs);
+
+	return ret;
+}
+
+/**
+ * pvrdma_modify_cq - modify the CQ moderation parameters
+ * @ibcq: the CQ to modify
+ * @cq_count: number of CQEs that will trigger an event
+ * @cq_period: max period of time in usec before triggering an event
+ *
+ * @return: -EOPNOTSUPP as CQ resize is not supported.
+ */
+int pvrdma_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period)
+{
+	return -EOPNOTSUPP;
+}
+
+static inline struct pvrdma_cqe *get_cqe(struct pvrdma_cq *cq, int i)
+{
+	return (struct pvrdma_cqe *)pvrdma_page_dir_get_ptr(
+					&cq->pdir,
+					cq->offset +
+					sizeof(struct pvrdma_cqe) * i);
+}
+
+void _pvrdma_flush_cqe(struct pvrdma_qp *qp, struct pvrdma_cq *cq)
+{
+	int head;