Merge tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma

Pull more rdma updates from Doug Ledford:
 "Round two of 4.6 merge window patches.

  This is a monster pull request.  I held off on the hfi1 driver updates
  (the hfi1 driver is intimately tied to the qib driver and the new
  rdmavt software library that was created to help both of them) in my
  first pull request.  The hfi1/qib/rdmavt update is probably 90% of
  this pull request.  The hfi1 driver is being left in staging so that
  it can be fixed up in regards to the API that Al and yourself didn't
  like.  Intel has agreed to do the work, but in the meantime, this
  clears out 300+ patches in the backlog queue and brings my tree and
  their tree closer to sync.

  This also includes about 10 patches to the core and a few to mlx5 to
  create an infrastructure for configuring SRIOV ports on IB devices.
  That series includes one patch to the net core that we sent to netdev@
  and Dave Miller with each of the three revisions to the series.  We
  didn't get any response to the patch, so we took that as implicit
  approval.

  Finally, this series includes Intel's new iWARP driver for their x722
  cards.  It's not nearly the beast as the hfi1 driver.  It also has a
  linux-next merge issue, but that has been resolved and it now passes
  just fine.

  Summary:

   - A few minor core fixups needed for the next patch series

   - The IB SRIOV series.  This has bounced around for several versions.
     Of note is the fact that the first patch in this series effects the
     net core.  It was directed to netdev and DaveM for each iteration
     of the series (three versions total).  Dave did not object, but did
     not respond either.  I've taken this as permission to move forward
     with the series.

   - The new Intel X722 iWARP driver

   - A huge set of updates to the Intel hfi1 driver.  Of particular
     interest here is that we have left the driver in staging since it
     still has an API that people object to.  Intel is working on a fix,
     but getting these patches in now helps keep me sane as the upstream
     and Intel's trees were over 300 patches apart"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/dledford/rdma: (362 commits)
  IB/ipoib: Allow mcast packets from other VFs
  IB/mlx5: Implement callbacks for manipulating VFs
  net/mlx5_core: Implement modify HCA vport command
  net/mlx5_core: Add VF param when querying vport counter
  IB/ipoib: Add ndo operations for configuring VFs
  IB/core: Add interfaces to control VF attributes
  IB/core: Support accessing SA in virtualized environment
  IB/core: Add subnet prefix to port info
  IB/mlx5: Fix decision on using MAD_IFC
  net/core: Add support for configuring VF GUIDs
  IB/{core, ulp} Support above 32 possible device capability flags
  IB/core: Replace setting the zero values in ib_uverbs_ex_query_device
  net/mlx5_core: Introduce offload arithmetic hardware capabilities
  net/mlx5_core: Refactor device capability function
  net/mlx5_core: Fix caching ATOMIC endian mode capability
  ib_srpt: fix a WARN_ON() message
  i40iw: Replace the obsolete crypto hash interface with shash
  IB/hfi1: Add SDMA cache eviction algorithm
  IB/hfi1: Switch to using the pin query function
  IB/hfi1: Specify mm when releasing pages
  ...

6 files changed, 1194 insertions, 2 deletions

diff --git a/include/rdma/ib_verbs.h b/include/rdma/ib_verbs.h
index 3a03c1d18afa..fb2cef4e9747 100644
--- a/include/rdma/ib_verbs.h
+++ b/include/rdma/ib_verbs.h
@@ -56,6 +56,7 @@
 #include <linux/string.h>
 #include <linux/slab.h>
 
+#include <linux/if_link.h>
 #include <linux/atomic.h>
 #include <linux/mmu_notifier.h>
 #include <asm/uaccess.h>
@@ -97,6 +98,11 @@ enum rdma_node_type {
 	RDMA_NODE_USNIC_UDP,
 };
 
+enum {
+	/* set the local administered indication */
+	IB_SA_WELL_KNOWN_GUID	= BIT_ULL(57) | 2,
+};
+
 enum rdma_transport_type {
 	RDMA_TRANSPORT_IB,
 	RDMA_TRANSPORT_IWARP,
@@ -213,6 +219,7 @@ enum ib_device_cap_flags {
 	IB_DEVICE_SIGNATURE_HANDOVER		= (1 << 30),
 	IB_DEVICE_ON_DEMAND_PAGING		= (1 << 31),
 	IB_DEVICE_SG_GAPS_REG			= (1ULL << 32),
+	IB_DEVICE_VIRTUAL_FUNCTION		= ((u64)1 << 33),
 };
 
 enum ib_signature_prot_cap {
@@ -274,7 +281,7 @@ struct ib_device_attr {
 	u32			hw_ver;
 	int			max_qp;
 	int			max_qp_wr;
-	int			device_cap_flags;
+	u64			device_cap_flags;
 	int			max_sge;
 	int			max_sge_rd;
 	int			max_cq;
@@ -490,6 +497,7 @@ union rdma_protocol_stats {
 					| RDMA_CORE_CAP_OPA_MAD)
 
 struct ib_port_attr {
+	u64			subnet_prefix;
 	enum ib_port_state	state;
 	enum ib_mtu		max_mtu;
 	enum ib_mtu		active_mtu;
@@ -509,6 +517,7 @@ struct ib_port_attr {
 	u8			active_width;
 	u8			active_speed;
 	u8                      phys_state;
+	bool			grh_required;
 };
 
 enum ib_device_modify_flags {
@@ -614,6 +623,7 @@ enum {
 };
 
 #define IB_LID_PERMISSIVE	cpu_to_be16(0xFFFF)
+#define IB_MULTICAST_LID_BASE	cpu_to_be16(0xC000)
 
 enum ib_ah_flags {
 	IB_AH_GRH	= 1
@@ -1860,6 +1870,14 @@ struct ib_device {
 	void			   (*disassociate_ucontext)(struct ib_ucontext *ibcontext);
 	void			   (*drain_rq)(struct ib_qp *qp);
 	void			   (*drain_sq)(struct ib_qp *qp);
+	int			   (*set_vf_link_state)(struct ib_device *device, int vf, u8 port,
+							int state);
+	int			   (*get_vf_config)(struct ib_device *device, int vf, u8 port,
+						   struct ifla_vf_info *ivf);
+	int			   (*get_vf_stats)(struct ib_device *device, int vf, u8 port,
+						   struct ifla_vf_stats *stats);
+	int			   (*set_vf_guid)(struct ib_device *device, int vf, u8 port, u64 guid,
+						  int type);
 
 	struct ib_dma_mapping_ops   *dma_ops;
 
@@ -2303,6 +2321,15 @@ int ib_query_gid(struct ib_device *device,
 		 u8 port_num, int index, union ib_gid *gid,
 		 struct ib_gid_attr *attr);
 
+int ib_set_vf_link_state(struct ib_device *device, int vf, u8 port,
+			 int state);
+int ib_get_vf_config(struct ib_device *device, int vf, u8 port,
+		     struct ifla_vf_info *info);
+int ib_get_vf_stats(struct ib_device *device, int vf, u8 port,
+		    struct ifla_vf_stats *stats);
+int ib_set_vf_guid(struct ib_device *device, int vf, u8 port, u64 guid,
+		   int type);
+
 int ib_query_pkey(struct ib_device *device,
 		  u8 port_num, u16 index, u16 *pkey);
 
diff --git a/include/rdma/opa_port_info.h b/include/rdma/opa_port_info.h
index a0fa975cd1c1..2b95c2c336eb 100644
--- a/include/rdma/opa_port_info.h
+++ b/include/rdma/opa_port_info.h
@@ -97,7 +97,7 @@
 #define OPA_LINKDOWN_REASON_WIDTH_POLICY			41
 /* 42-48 reserved */
 #define OPA_LINKDOWN_REASON_DISCONNECTED			49
-#define OPA_LINKDOWN_REASONLOCAL_MEDIA_NOT_INSTALLED		50
+#define OPA_LINKDOWN_REASON_LOCAL_MEDIA_NOT_INSTALLED		50
 #define OPA_LINKDOWN_REASON_NOT_INSTALLED			51
 #define OPA_LINKDOWN_REASON_CHASSIS_CONFIG			52
 /* 53 reserved */
diff --git a/include/rdma/rdma_vt.h b/include/rdma/rdma_vt.h
new file mode 100644
index 000000000000..a8696551abb1
--- /dev/null
+++ b/include/rdma/rdma_vt.h
@@ -0,0 +1,481 @@
+#ifndef DEF_RDMA_VT_H
+#define DEF_RDMA_VT_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * 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 for more details.
+ *
+ * BSD 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.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * 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
+ * OWNER 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.
+ *
+ */
+
+/*
+ * Structure that low level drivers will populate in order to register with the
+ * rdmavt layer.
+ */
+
+#include <linux/spinlock.h>
+#include <linux/list.h>
+#include <linux/hash.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/rdmavt_mr.h>
+#include <rdma/rdmavt_qp.h>
+
+#define RVT_MAX_PKEY_VALUES 16
+
+struct rvt_ibport {
+	struct rvt_qp __rcu *qp[2];
+	struct ib_mad_agent *send_agent;	/* agent for SMI (traps) */
+	struct rb_root mcast_tree;
+	spinlock_t lock;		/* protect changes in this struct */
+
+	/* non-zero when timer is set */
+	unsigned long mkey_lease_timeout;
+	unsigned long trap_timeout;
+	__be64 gid_prefix;      /* in network order */
+	__be64 mkey;
+	u64 tid;
+	u32 port_cap_flags;
+	u32 pma_sample_start;
+	u32 pma_sample_interval;
+	__be16 pma_counter_select[5];
+	u16 pma_tag;
+	u16 mkey_lease_period;
+	u16 sm_lid;
+	u8 sm_sl;
+	u8 mkeyprot;
+	u8 subnet_timeout;
+	u8 vl_high_limit;
+
+	/*
+	 * Driver is expected to keep these up to date. These
+	 * counters are informational only and not required to be
+	 * completely accurate.
+	 */
+	u64 n_rc_resends;
+	u64 n_seq_naks;
+	u64 n_rdma_seq;
+	u64 n_rnr_naks;
+	u64 n_other_naks;
+	u64 n_loop_pkts;
+	u64 n_pkt_drops;
+	u64 n_vl15_dropped;
+	u64 n_rc_timeouts;
+	u64 n_dmawait;
+	u64 n_unaligned;
+	u64 n_rc_dupreq;
+	u64 n_rc_seqnak;
+	u16 pkey_violations;
+	u16 qkey_violations;
+	u16 mkey_violations;
+
+	/* Hot-path per CPU counters to avoid cacheline trading to update */
+	u64 z_rc_acks;
+	u64 z_rc_qacks;
+	u64 z_rc_delayed_comp;
+	u64 __percpu *rc_acks;
+	u64 __percpu *rc_qacks;
+	u64 __percpu *rc_delayed_comp;
+
+	void *priv; /* driver private data */
+
+	/*
+	 * The pkey table is allocated and maintained by the driver. Drivers
+	 * need to have access to this before registering with rdmav. However
+	 * rdmavt will need access to it so drivers need to proviee this during
+	 * the attach port API call.
+	 */
+	u16 *pkey_table;
+
+	struct rvt_ah *sm_ah;
+};
+
+#define RVT_CQN_MAX 16 /* maximum length of cq name */
+
+/*
+ * Things that are driver specific, module parameters in hfi1 and qib
+ */
+struct rvt_driver_params {
+	struct ib_device_attr props;
+
+	/*
+	 * Anything driver specific that is not covered by props
+	 * For instance special module parameters. Goes here.
+	 */
+	unsigned int lkey_table_size;
+	unsigned int qp_table_size;
+	int qpn_start;
+	int qpn_inc;
+	int qpn_res_start;
+	int qpn_res_end;
+	int nports;
+	int npkeys;
+	u8 qos_shift;
+	char cq_name[RVT_CQN_MAX];
+	int node;
+	int max_rdma_atomic;
+	int psn_mask;
+	int psn_shift;
+	int psn_modify_mask;
+	u32 core_cap_flags;
+	u32 max_mad_size;
+};
+
+/* Protection domain */
+struct rvt_pd {
+	struct ib_pd ibpd;
+	int user;               /* non-zero if created from user space */
+};
+
+/* Address handle */
+struct rvt_ah {
+	struct ib_ah ibah;
+	struct ib_ah_attr attr;
+	atomic_t refcount;
+	u8 vl;
+	u8 log_pmtu;
+};
+
+struct rvt_dev_info;
+struct rvt_swqe;
+struct rvt_driver_provided {
+	/*
+	 * Which functions are required depends on which verbs rdmavt is
+	 * providing and which verbs the driver is overriding. See
+	 * check_support() for details.
+	 */
+
+	/* Passed to ib core registration. Callback to create syfs files */
+	int (*port_callback)(struct ib_device *, u8, struct kobject *);
+
+	/*
+	 * Returns a string to represent the device for which is being
+	 * registered. This is primarily used for error and debug messages on
+	 * the console.
+	 */
+	const char * (*get_card_name)(struct rvt_dev_info *rdi);
+
+	/*
+	 * Returns a pointer to the undelying hardware's PCI device. This is
+	 * used to display information as to what hardware is being referenced
+	 * in an output message
+	 */
+	struct pci_dev * (*get_pci_dev)(struct rvt_dev_info *rdi);
+
+	/*
+	 * Allocate a private queue pair data structure for driver specific
+	 * information which is opaque to rdmavt.
+	 */
+	void * (*qp_priv_alloc)(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+				gfp_t gfp);
+
+	/*
+	 * Free the driver's private qp structure.
+	 */
+	void (*qp_priv_free)(struct rvt_dev_info *rdi, struct rvt_qp *qp);
+
+	/*
+	 * Inform the driver the particular qp in quesiton has been reset so
+	 * that it can clean up anything it needs to.
+	 */
+	void (*notify_qp_reset)(struct rvt_qp *qp);
+
+	/*
+	 * Give the driver a notice that there is send work to do. It is up to
+	 * the driver to generally push the packets out, this just queues the
+	 * work with the driver. There are two variants here. The no_lock
+	 * version requires the s_lock not to be held. The other assumes the
+	 * s_lock is held.
+	 */
+	void (*schedule_send)(struct rvt_qp *qp);
+	void (*schedule_send_no_lock)(struct rvt_qp *qp);
+
+	/*
+	 * Sometimes rdmavt needs to kick the driver's send progress. That is
+	 * done by this call back.
+	 */
+	void (*do_send)(struct rvt_qp *qp);
+
+	/*
+	 * Get a path mtu from the driver based on qp attributes.
+	 */
+	int (*get_pmtu_from_attr)(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+				  struct ib_qp_attr *attr);
+
+	/*
+	 * Notify driver that it needs to flush any outstanding IO requests that
+	 * are waiting on a qp.
+	 */
+	void (*flush_qp_waiters)(struct rvt_qp *qp);
+
+	/*
+	 * Notify driver to stop its queue of sending packets. Nothing else
+	 * should be posted to the queue pair after this has been called.
+	 */
+	void (*stop_send_queue)(struct rvt_qp *qp);
+
+	/*
+	 * Have the drivr drain any in progress operations
+	 */
+	void (*quiesce_qp)(struct rvt_qp *qp);
+
+	/*
+	 * Inform the driver a qp has went to error state.
+	 */
+	void (*notify_error_qp)(struct rvt_qp *qp);
+
+	/*
+	 * Get an MTU for a qp.
+	 */
+	u32 (*mtu_from_qp)(struct rvt_dev_info *rdi, struct rvt_qp *qp,
+			   u32 pmtu);
+	/*
+	 * Convert an mtu to a path mtu
+	 */
+	int (*mtu_to_path_mtu)(u32 mtu);
+
+	/*
+	 * Get the guid of a port in big endian byte order
+	 */
+	int (*get_guid_be)(struct rvt_dev_info *rdi, struct rvt_ibport *rvp,
+			   int guid_index, __be64 *guid);
+
+	/*
+	 * Query driver for the state of the port.
+	 */
+	int (*query_port_state)(struct rvt_dev_info *rdi, u8 port_num,
+				struct ib_port_attr *props);
+
+	/*
+	 * Tell driver to shutdown a port
+	 */
+	int (*shut_down_port)(struct rvt_dev_info *rdi, u8 port_num);
+
+	/* Tell driver to send a trap for changed  port capabilities */
+	void (*cap_mask_chg)(struct rvt_dev_info *rdi, u8 port_num);
+
+	/*
+	 * The following functions can be safely ignored completely. Any use of
+	 * these is checked for NULL before blindly calling. Rdmavt should also
+	 * be functional if drivers omit these.
+	 */
+
+	/* Called to inform the driver that all qps should now be freed. */
+	unsigned (*free_all_qps)(struct rvt_dev_info *rdi);
+
+	/* Driver specific AH validation */
+	int (*check_ah)(struct ib_device *, struct ib_ah_attr *);
+
+	/* Inform the driver a new AH has been created */
+	void (*notify_new_ah)(struct ib_device *, struct ib_ah_attr *,
+			      struct rvt_ah *);
+
+	/* Let the driver pick the next queue pair number*/
+	int (*alloc_qpn)(struct rvt_dev_info *rdi, struct rvt_qpn_table *qpt,
+			 enum ib_qp_type type, u8 port_num, gfp_t gfp);
+
+	/* Determine if its safe or allowed to modify the qp */
+	int (*check_modify_qp)(struct rvt_qp *qp, struct ib_qp_attr *attr,
+			       int attr_mask, struct ib_udata *udata);
+
+	/* Driver specific QP modification/notification-of */
+	void (*modify_qp)(struct rvt_qp *qp, struct ib_qp_attr *attr,
+			  int attr_mask, struct ib_udata *udata);
+
+	/* Driver specific work request checking */
+	int (*check_send_wqe)(struct rvt_qp *qp, struct rvt_swqe *wqe);
+
+	/* Notify driver a mad agent has been created */
+	void (*notify_create_mad_agent)(struct rvt_dev_info *rdi, int port_idx);
+
+	/* Notify driver a mad agent has been removed */
+	void (*notify_free_mad_agent)(struct rvt_dev_info *rdi, int port_idx);
+
+};
+
+struct rvt_dev_info {
+	struct ib_device ibdev; /* Keep this first. Nothing above here */
+
+	/*
+	 * Prior to calling for registration the driver will be responsible for
+	 * allocating space for this structure.
+	 *
+	 * The driver will also be responsible for filling in certain members of
+	 * dparms.props. The driver needs to fill in dparms exactly as it would
+	 * want values reported to a ULP. This will be returned to the caller
+	 * in rdmavt's device. The driver should also therefore refrain from
+	 * modifying this directly after registration with rdmavt.
+	 */
+
+	/* Driver specific properties */
+	struct rvt_driver_params dparms;
+
+	struct rvt_mregion __rcu *dma_mr;
+	struct rvt_lkey_table lkey_table;
+
+	/* Driver specific helper functions */
+	struct rvt_driver_provided driver_f;
+
+	/* Internal use */
+	int n_pds_allocated;
+	spinlock_t n_pds_lock; /* Protect pd allocated count */
+
+	int n_ahs_allocated;
+	spinlock_t n_ahs_lock; /* Protect ah allocated count */
+
+	u32 n_srqs_allocated;
+	spinlock_t n_srqs_lock; /* Protect srqs allocated count */
+
+	int flags;
+	struct rvt_ibport **ports;
+
+	/* QP */
+	struct rvt_qp_ibdev *qp_dev;
+	u32 n_qps_allocated;    /* number of QPs allocated for device */
+	u32 n_rc_qps;		/* number of RC QPs allocated for device */
+	u32 busy_jiffies;	/* timeout scaling based on RC QP count */
+	spinlock_t n_qps_lock;	/* protect qps, rc qps and busy jiffy counts */
+
+	/* memory maps */
+	struct list_head pending_mmaps;
+	spinlock_t mmap_offset_lock; /* protect mmap_offset */
+	u32 mmap_offset;
+	spinlock_t pending_lock; /* protect pending mmap list */
+
+	/* CQ */
+	struct kthread_worker *worker; /* per device cq worker */
+	u32 n_cqs_allocated;    /* number of CQs allocated for device */
+	spinlock_t n_cqs_lock; /* protect count of in use cqs */
+
+	/* Multicast */
+	u32 n_mcast_grps_allocated; /* number of mcast groups allocated */
+	spinlock_t n_mcast_grps_lock;
+
+};
+
+static inline struct rvt_pd *ibpd_to_rvtpd(struct ib_pd *ibpd)
+{
+	return container_of(ibpd, struct rvt_pd, ibpd);
+}
+
+static inline struct rvt_ah *ibah_to_rvtah(struct ib_ah *ibah)
+{
+	return container_of(ibah, struct rvt_ah, ibah);
+}
+
+static inline struct rvt_dev_info *ib_to_rvt(struct ib_device *ibdev)
+{
+	return  container_of(ibdev, struct rvt_dev_info, ibdev);
+}
+
+static inline struct rvt_srq *ibsrq_to_rvtsrq(struct ib_srq *ibsrq)
+{
+	return container_of(ibsrq, struct rvt_srq, ibsrq);
+}
+
+static inline struct rvt_qp *ibqp_to_rvtqp(struct ib_qp *ibqp)
+{
+	return container_of(ibqp, struct rvt_qp, ibqp);
+}
+
+static inline unsigned rvt_get_npkeys(struct rvt_dev_info *rdi)
+{
+	/*
+	 * All ports have same number of pkeys.
+	 */
+	return rdi->dparms.npkeys;
+}
+
+/*
+ * Return the indexed PKEY from the port PKEY table.
+ */
+static inline u16 rvt_get_pkey(struct rvt_dev_info *rdi,
+			       int port_index,
+			       unsigned index)
+{
+	if (index >= rvt_get_npkeys(rdi))
+		return 0;
+	else
+		return rdi->ports[port_index]->pkey_table[index];
+}
+
+/**
+ * rvt_lookup_qpn - return the QP with the given QPN
+ * @ibp: the ibport
+ * @qpn: the QP number to look up
+ *
+ * The caller must hold the rcu_read_lock(), and keep the lock until
+ * the returned qp is no longer in use.
+ */
+/* TODO: Remove this and put in rdmavt/qp.h when no longer needed by drivers */
+static inline struct rvt_qp *rvt_lookup_qpn(struct rvt_dev_info *rdi,
+					    struct rvt_ibport *rvp,
+					    u32 qpn) __must_hold(RCU)
+{
+	struct rvt_qp *qp = NULL;
+
+	if (unlikely(qpn <= 1)) {
+		qp = rcu_dereference(rvp->qp[qpn]);
+	} else {
+		u32 n = hash_32(qpn, rdi->qp_dev->qp_table_bits);
+
+		for (qp = rcu_dereference(rdi->qp_dev->qp_table[n]); qp;
+			qp = rcu_dereference(qp->next))
+			if (qp->ibqp.qp_num == qpn)
+				break;
+	}
+	return qp;
+}
+
+struct rvt_dev_info *rvt_alloc_device(size_t size, int nports);
+int rvt_register_device(struct rvt_dev_info *rvd);
+void rvt_unregister_device(struct rvt_dev_info *rvd);
+int rvt_check_ah(struct ib_device *ibdev, struct ib_ah_attr *ah_attr);
+int rvt_init_port(struct rvt_dev_info *rdi, struct rvt_ibport *port,
+		  int port_index, u16 *pkey_table);
+int rvt_rkey_ok(struct rvt_qp *qp, struct rvt_sge *sge,
+		u32 len, u64 vaddr, u32 rkey, int acc);
+int rvt_lkey_ok(struct rvt_lkey_table *rkt, struct rvt_pd *pd,
+		struct rvt_sge *isge, struct ib_sge *sge, int acc);
+struct rvt_mcast *rvt_mcast_find(struct rvt_ibport *ibp, union ib_gid *mgid);
+
+#endif          /* DEF_RDMA_VT_H */
diff --git a/include/rdma/rdmavt_cq.h b/include/rdma/rdmavt_cq.h
new file mode 100644
index 000000000000..51fd00b243d0
--- /dev/null
+++ b/include/rdma/rdmavt_cq.h
@@ -0,0 +1,99 @@
+#ifndef DEF_RDMAVT_INCCQ_H
+#define DEF_RDMAVT_INCCQ_H
+
+/*
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * 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 for more details.
+ *
+ * BSD LICENSE
+ *
+ * Copyright(c) 2015 Intel Corporation.
+ *
+ * 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.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * 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
+ * OWNER 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/kthread.h>
+#include <rdma/ib_user_verbs.h>
+
+/*
+ * Define an ib_cq_notify value that is not valid so we know when CQ
+ * notifications are armed.
+ */
+#define RVT_CQ_NONE      (IB_CQ_NEXT_COMP + 1)
+
+/*
+ * This structure is used to contain the head pointer, tail pointer,
+ * and completion queue entries as a single memory allocation so
+ * it can be mmap'ed into user space.
+ */
+struct rvt_cq_wc {
+	u32 head;               /* index of next entry to fill */
+	u32 tail;               /* index of next ib_poll_cq() entry */
+	union {
+		/* these are actually size ibcq.cqe + 1 */
+		struct ib_uverbs_wc uqueue[0];
+		struct ib_wc kqueue[0];
+	};
+};
+
+/*
+ * The completion queue structure.
+ */
+struct rvt_cq {
+	struct ib_cq ibcq;
+	struct kthread_work comptask;
+	spinlock_t lock; /* protect changes in this struct */
+	u8 notify;
+	u8 triggered;
+	struct rvt_dev_info *rdi;
+	struct rvt_cq_wc *queue;
+	struct rvt_mmap_info *ip;
+};
+
+static inline struct rvt_cq *ibcq_to_rvtcq(struct ib_cq *ibcq)
+{
+	return container_of(ibcq, struct rvt_cq, ibcq);
+}
+
+void rvt_cq_enter(struct rvt_cq *cq, struct ib_wc *entry, bool solicited);
+
+#endif          /* DEF_RDMAVT_INCCQH */
diff --git a/include/rdma/rdmavt_mr.h b/include/rdma/rdmavt_mr.h
new file mode 100644
index 000000000000..5edffdca8c53
--- /dev/null
+++ b/include/rdma/rdmavt_mr.h
@@ -0,0 +1,139 @@
+#ifndef DEF_RDMAVT_INCMR_H
+#define DEF_RDMAVT_INCMR_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * 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 for more details.
+ *
+ * BSD 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.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * 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
+ * OWNER 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.
+ *
+ */
+
+/*
+ * For Memory Regions. This stuff should probably be moved into rdmavt/mr.h once
+ * drivers no longer need access to the MR directly.
+ */
+
+/*
+ * A segment is a linear region of low physical memory.
+ * Used by the verbs layer.
+ */
+struct rvt_seg {
+	void *vaddr;
+	size_t length;
+};
+
+/* The number of rvt_segs that fit in a page. */
+#define RVT_SEGSZ     (PAGE_SIZE / sizeof(struct rvt_seg))
+
+struct rvt_segarray {
+	struct rvt_seg segs[RVT_SEGSZ];
+};
+
+struct rvt_mregion {
+	struct ib_pd *pd;       /* shares refcnt of ibmr.pd */
+	u64 user_base;          /* User's address for this region */
+	u64 iova;               /* IB start address of this region */
+	size_t length;
+	u32 lkey;
+	u32 offset;             /* offset (bytes) to start of region */
+	int access_flags;
+	u32 max_segs;           /* number of rvt_segs in all the arrays */
+	u32 mapsz;              /* size of the map array */
+	u8  page_shift;         /* 0 - non unform/non powerof2 sizes */
+	u8  lkey_published;     /* in global table */
+	struct completion comp; /* complete when refcount goes to zero */
+	atomic_t refcount;
+	struct rvt_segarray *map[0];    /* the segments */
+};
+
+#define RVT_MAX_LKEY_TABLE_BITS 23
+
+struct rvt_lkey_table {
+	spinlock_t lock; /* protect changes in this struct */
+	u32 next;               /* next unused index (speeds search) */
+	u32 gen;                /* generation count */
+	u32 max;                /* size of the table */
+	struct rvt_mregion __rcu **table;
+};
+
+/*
+ * These keep track of the copy progress within a memory region.
+ * Used by the verbs layer.
+ */
+struct rvt_sge {
+	struct rvt_mregion *mr;
+	void *vaddr;            /* kernel virtual address of segment */
+	u32 sge_length;         /* length of the SGE */
+	u32 length;             /* remaining length of the segment */
+	u16 m;                  /* current index: mr->map[m] */
+	u16 n;                  /* current index: mr->map[m]->segs[n] */
+};
+
+struct rvt_sge_state {
+	struct rvt_sge *sg_list;      /* next SGE to be used if any */
+	struct rvt_sge sge;   /* progress state for the current SGE */
+	u32 total_len;
+	u8 num_sge;
+};
+
+static inline void rvt_put_mr(struct rvt_mregion *mr)
+{
+	if (unlikely(atomic_dec_and_test(&mr->refcount)))
+		complete(&mr->comp);
+}
+
+static inline void rvt_get_mr(struct rvt_mregion *mr)
+{
+	atomic_inc(&mr->refcount);
+}
+
+static inline void rvt_put_ss(struct rvt_sge_state *ss)
+{
+	while (ss->num_sge) {
+		rvt_put_mr(ss->sge.mr);
+		if (--ss->num_sge)
+			ss->sge = *ss->sg_list++;
+	}
+}
+
+#endif          /* DEF_RDMAVT_INCMRH */
diff --git a/include/rdma/rdmavt_qp.h b/include/rdma/rdmavt_qp.h
new file mode 100644
index 000000000000..497e59065c2c
--- /dev/null
+++ b/include/rdma/rdmavt_qp.h
@@ -0,0 +1,446 @@
+#ifndef DEF_RDMAVT_INCQP_H
+#define DEF_RDMAVT_INCQP_H
+
+/*
+ * Copyright(c) 2016 Intel Corporation.
+ *
+ * This file is provided under a dual BSD/GPLv2 license.  When using or
+ * redistributing this file, you may do so under either license.
+ *
+ * GPL LICENSE SUMMARY
+ *
+ * This program is free software; you can redistribute it and/or modify
+ * it under the terms of version 2 of the GNU General Public License as
+ * published by the Free Software Foundation.
+ *
+ * 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 for more details.
+ *
+ * BSD 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.
+ *  - Neither the name of Intel Corporation nor the names of its
+ *    contributors may be used to endorse or promote products derived
+ *    from this software without specific prior written permission.
+ *
+ * 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
+ * OWNER 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 <rdma/rdma_vt.h>
+#include <rdma/ib_pack.h>
+#include <rdma/ib_verbs.h>
+/*
+ * Atomic bit definitions for r_aflags.
+ */
+#define RVT_R_WRID_VALID        0
+#define RVT_R_REWIND_SGE        1
+
+/*
+ * Bit definitions for r_flags.
+ */
+#define RVT_R_REUSE_SGE 0x01
+#define RVT_R_RDMAR_SEQ 0x02
+#define RVT_R_RSP_NAK   0x04
+#define RVT_R_RSP_SEND  0x08
+#define RVT_R_COMM_EST  0x10
+
+/*
+ * Bit definitions for s_flags.
+ *
+ * RVT_S_SIGNAL_REQ_WR - set if QP send WRs contain completion signaled
+ * RVT_S_BUSY - send tasklet is processing the QP
+ * RVT_S_TIMER - the RC retry timer is active
+ * RVT_S_ACK_PENDING - an ACK is waiting to be sent after RDMA read/atomics
+ * RVT_S_WAIT_FENCE - waiting for all prior RDMA read or atomic SWQEs
+ *                         before processing the next SWQE
+ * RVT_S_WAIT_RDMAR - waiting for a RDMA read or atomic SWQE to complete
+ *                         before processing the next SWQE
+ * RVT_S_WAIT_RNR - waiting for RNR timeout
+ * RVT_S_WAIT_SSN_CREDIT - waiting for RC credits to process next SWQE
+ * RVT_S_WAIT_DMA - waiting for send DMA queue to drain before generating
+ *                  next send completion entry not via send DMA
+ * RVT_S_WAIT_PIO - waiting for a send buffer to be available
+ * RVT_S_WAIT_PIO_DRAIN - waiting for a qp to drain pio packets
+ * RVT_S_WAIT_TX - waiting for a struct verbs_txreq to be available
+ * RVT_S_WAIT_DMA_DESC - waiting for DMA descriptors to be available
+ * RVT_S_WAIT_KMEM - waiting for kernel memory to be available
+ * RVT_S_WAIT_PSN - waiting for a packet to exit the send DMA queue
+ * RVT_S_WAIT_ACK - waiting for an ACK packet before sending more requests
+ * RVT_S_SEND_ONE - send one packet, request ACK, then wait for ACK
+ * RVT_S_ECN - a BECN was queued to the send engine
+ */
+#define RVT_S_SIGNAL_REQ_WR	0x0001
+#define RVT_S_BUSY		0x0002
+#define RVT_S_TIMER		0x0004
+#define RVT_S_RESP_PENDING	0x0008
+#define RVT_S_ACK_PENDING	0x0010
+#define RVT_S_WAIT_FENCE	0x0020
+#define RVT_S_WAIT_RDMAR	0x0040
+#define RVT_S_WAIT_RNR		0x0080
+#define RVT_S_WAIT_SSN_CREDIT	0x0100
+#define RVT_S_WAIT_DMA		0x0200
+#define RVT_S_WAIT_PIO		0x0400
+#define RVT_S_WAIT_PIO_DRAIN    0x0800
+#define RVT_S_WAIT_TX		0x1000
+#define RVT_S_WAIT_DMA_DESC	0x2000
+#define RVT_S_WAIT_KMEM		0x4000
+#define RVT_S_WAIT_PSN		0x8000
+#define RVT_S_WAIT_ACK		0x10000
+#define RVT_S_SEND_ONE		0x20000
+#define RVT_S_UNLIMITED_CREDIT	0x40000
+#define RVT_S_AHG_VALID		0x80000
+#define RVT_S_AHG_CLEAR		0x100000
+#define RVT_S_ECN		0x200000
+
+/*
+ * Wait flags that would prevent any packet type from being sent.
+ */
+#define RVT_S_ANY_WAIT_IO (RVT_S_WAIT_PIO | RVT_S_WAIT_TX | \
+	RVT_S_WAIT_DMA_DESC | RVT_S_WAIT_KMEM)
+
+/*
+ * Wait flags that would prevent send work requests from making progress.
+ */
+#define RVT_S_ANY_WAIT_SEND (RVT_S_WAIT_FENCE | RVT_S_WAIT_RDMAR | \
+	RVT_S_WAIT_RNR | RVT_S_WAIT_SSN_CREDIT | RVT_S_WAIT_DMA | \
+	RVT_S_WAIT_PSN | RVT_S_WAIT_ACK)
+
+#define RVT_S_ANY_WAIT (RVT_S_ANY_WAIT_IO | RVT_S_ANY_WAIT_SEND)
+
+/* Number of bits to pay attention to in the opcode for checking qp type */
+#define RVT_OPCODE_QP_MASK 0xE0
+
+/* Flags for checking QP state (see ib_rvt_state_ops[]) */
+#define RVT_POST_SEND_OK                0x01
+#define RVT_POST_RECV_OK                0x02
+#define RVT_PROCESS_RECV_OK             0x04
+#define RVT_PROCESS_SEND_OK             0x08
+#define RVT_PROCESS_NEXT_SEND_OK        0x10
+#define RVT_FLUSH_SEND			0x20
+#define RVT_FLUSH_RECV			0x40
+#define RVT_PROCESS_OR_FLUSH_SEND \
+	(RVT_PROCESS_SEND_OK | RVT_FLUSH_SEND)
+
+/*
+ * Send work request queue entry.
+ * The size of the sg_list is determined when the QP is created and stored
+ * in qp->s_max_sge.
+ */
+struct rvt_swqe {
+	union {
+		struct ib_send_wr wr;   /* don't use wr.sg_list */
+		struct ib_ud_wr ud_wr;
+		struct ib_reg_wr reg_wr;
+		struct ib_rdma_wr rdma_wr;
+		struct ib_atomic_wr atomic_wr;
+	};
+	u32 psn;                /* first packet sequence number */
+	u32 lpsn;               /* last packet sequence number */
+	u32 ssn;                /* send sequence number */
+	u32 length;             /* total length of data in sg_list */
+	struct rvt_sge sg_list[0];
+};
+
+/*
+ * Receive work request queue entry.
+ * The size of the sg_list is determined when the QP (or SRQ) is created
+ * and stored in qp->r_rq.max_sge (or srq->rq.max_sge).
+ */
+struct rvt_rwqe {
+	u64 wr_id;
+	u8 num_sge;
+	struct ib_sge sg_list[0];
+};
+
+/*
+ * This structure is used to contain the head pointer, tail pointer,
+ * and receive work queue entries as a single memory allocation so
+ * it can be mmap'ed into user space.
+ * Note that the wq array elements are variable size so you can't
+ * just index into the array to get the N'th element;
+ * use get_rwqe_ptr() instead.
+ */
+struct rvt_rwq {
+	u32 head;               /* new work requests posted to the head */
+	u32 tail;               /* receives pull requests from here. */
+	struct rvt_rwqe wq[0];
+};
+
+struct rvt_rq {
+	struct rvt_rwq *wq;
+	u32 size;               /* size of RWQE array */
+	u8 max_sge;
+	/* protect changes in this struct */
+	spinlock_t lock ____cacheline_aligned_in_smp;
+};
+
+/*
+ * This structure is used by rvt_mmap() to validate an offset
+ * when an mmap() request is made.  The vm_area_struct then uses
+ * this as its vm_private_data.
+ */
+struct rvt_mmap_info {
+	struct list_head pending_mmaps;
+	struct ib_ucontext *context;
+	void *obj;
+	__u64 offset;
+	struct kref ref;
+	unsigned size;
+};
+
+#define RVT_MAX_RDMA_ATOMIC	16
+
+/*
+ * This structure holds the information that the send tasklet needs
+ * to send a RDMA read response or atomic operation.
+ */
+struct rvt_ack_entry {
+	u8 opcode;
+	u8 sent;
+	u32 psn;
+	u32 lpsn;
+	union {
+		struct rvt_sge rdma_sge;
+		u64 atomic_data;
+	};
+};
+
+#define	RC_QP_SCALING_INTERVAL	5
+
+/*
+ * Variables prefixed with s_ are for the requester (sender).
+ * Variables prefixed with r_ are for the responder (receiver).
+ * Variables prefixed with ack_ are for responder replies.
+ *
+ * Common variables are protected by both r_rq.lock and s_lock in that order
+ * which only happens in modify_qp() or changing the QP 'state'.
+ */
+struct rvt_qp {
+	struct ib_qp ibqp;
+	void *priv; /* Driver private data */
+	/* read mostly fields above and below */
+	struct ib_ah_attr remote_ah_attr;
+	struct ib_ah_attr alt_ah_attr;
+	struct rvt_qp __rcu *next;           /* link list for QPN hash table */
+	struct rvt_swqe *s_wq;  /* send work queue */
+	struct rvt_mmap_info *ip;
+
+	unsigned long timeout_jiffies;  /* computed from timeout */
+
+	enum ib_mtu path_mtu;
+	int srate_mbps;		/* s_srate (below) converted to Mbit/s */
+	pid_t pid;		/* pid for user mode QPs */
+	u32 remote_qpn;
+	u32 qkey;               /* QKEY for this QP (for UD or RD) */
+	u32 s_size;             /* send work queue size */
+	u32 s_ahgpsn;           /* set to the psn in the copy of the header */
+
+	u16 pmtu;		/* decoded from path_mtu */
+	u8 log_pmtu;		/* shift for pmtu */
+	u8 state;               /* QP state */
+	u8 allowed_ops;		/* high order bits of allowed opcodes */
+	u8 qp_access_flags;
+	u8 alt_timeout;         /* Alternate path timeout for this QP */
+	u8 timeout;             /* Timeout for this QP */
+	u8 s_srate;
+	u8 s_mig_state;
+	u8 port_num;
+	u8 s_pkey_index;        /* PKEY index to use */
+	u8 s_alt_pkey_index;    /* Alternate path PKEY index to use */
+	u8 r_max_rd_atomic;     /* max number of RDMA read/atomic to receive */
+	u8 s_max_rd_atomic;     /* max number of RDMA read/atomic to send */
+	u8 s_retry_cnt;         /* number of times to retry */
+	u8 s_rnr_retry_cnt;
+	u8 r_min_rnr_timer;     /* retry timeout value for RNR NAKs */
+	u8 s_max_sge;           /* size of s_wq->sg_list */
+	u8 s_draining;
+
+	/* start of read/write fields */
+	atomic_t refcount ____cacheline_aligned_in_smp;
+	wait_queue_head_t wait;
+
+	struct rvt_ack_entry s_ack_queue[RVT_MAX_RDMA_ATOMIC + 1]
+		____cacheline_aligned_in_smp;
+	struct rvt_sge_state s_rdma_read_sge;
+
+	spinlock_t r_lock ____cacheline_aligned_in_smp;      /* used for APM */
+	u32 r_psn;              /* expected rcv packet sequence number */
+	unsigned long r_aflags;
+	u64 r_wr_id;            /* ID for current receive WQE */
+	u32 r_ack_psn;          /* PSN for next ACK or atomic ACK */
+	u32 r_len;              /* total length of r_sge */
+	u32 r_rcv_len;          /* receive data len processed */
+	u32 r_msn;              /* message sequence number */
+
+	u8 r_state;             /* opcode of last packet received */
+	u8 r_flags;
+	u8 r_head_ack_queue;    /* index into s_ack_queue[] */
+
+	struct list_head rspwait;       /* link for waiting to respond */
+
+	struct rvt_sge_state r_sge;     /* current receive data */
+	struct rvt_rq r_rq;             /* receive work queue */
+
+	/* post send line */
+	spinlock_t s_hlock ____cacheline_aligned_in_smp;
+	u32 s_head;             /* new entries added here */
+	u32 s_next_psn;         /* PSN for next request */
+	u32 s_avail;            /* number of entries avail */
+	u32 s_ssn;              /* SSN of tail entry */
+
+	spinlock_t s_lock ____cacheline_aligned_in_smp;
+	u32 s_flags;
+	struct rvt_sge_state *s_cur_sge;
+	struct rvt_swqe *s_wqe;
+	struct rvt_sge_state s_sge;     /* current send request data */
+	struct rvt_mregion *s_rdma_mr;
+	u32 s_cur_size;         /* size of send packet in bytes */
+	u32 s_len;              /* total length of s_sge */
+	u32 s_rdma_read_len;    /* total length of s_rdma_read_sge */
+	u32 s_last_psn;         /* last response PSN processed */
+	u32 s_sending_psn;      /* lowest PSN that is being sent */
+	u32 s_sending_hpsn;     /* highest PSN that is being sent */
+	u32 s_psn;              /* current packet sequence number */
+	u32 s_ack_rdma_psn;     /* PSN for sending RDMA read responses */
+	u32 s_ack_psn;          /* PSN for acking sends and RDMA writes */
+	u32 s_tail;             /* next entry to process */
+	u32 s_cur;              /* current work queue entry */
+	u32 s_acked;            /* last un-ACK'ed entry */
+	u32 s_last;             /* last completed entry */
+	u32 s_lsn;              /* limit sequence number (credit) */
+	u16 s_hdrwords;         /* size of s_hdr in 32 bit words */
+	u16 s_rdma_ack_cnt;
+	s8 s_ahgidx;
+	u8 s_state;             /* opcode of last packet sent */
+	u8 s_ack_state;         /* opcode of packet to ACK */
+	u8 s_nak_state;         /* non-zero if NAK is pending */
+	u8 r_nak_state;         /* non-zero if NAK is pending */
+	u8 s_retry;             /* requester retry counter */
+	u8 s_rnr_retry;         /* requester RNR retry counter */
+	u8 s_num_rd_atomic;     /* number of RDMA read/atomic pending */
+	u8 s_tail_ack_queue;    /* index into s_ack_queue[] */
+
+	struct rvt_sge_state s_ack_rdma_sge;
+	struct timer_list s_timer;
+
+	/*
+	 * This sge list MUST be last. Do not add anything below here.
+	 */
+	struct rvt_sge r_sg_list[0] /* verified SGEs */
+		____cacheline_aligned_in_smp;
+};
+
+struct rvt_srq {
+	struct ib_srq ibsrq;
+	struct rvt_rq rq;
+	struct rvt_mmap_info *ip;
+	/* send signal when number of RWQEs < limit */
+	u32 limit;
+};
+
+#define RVT_QPN_MAX                 BIT(24)
+#define RVT_QPNMAP_ENTRIES          (RVT_QPN_MAX / PAGE_SIZE / BITS_PER_BYTE)
+#define RVT_BITS_PER_PAGE           (PAGE_SIZE * BITS_PER_BYTE)
+#define RVT_BITS_PER_PAGE_MASK      (RVT_BITS_PER_PAGE - 1)
+#define RVT_QPN_MASK		    0xFFFFFF
+
+/*
+ * QPN-map pages start out as NULL, they get allocated upon
+ * first use and are never deallocated. This way,
+ * large bitmaps are not allocated unless large numbers of QPs are used.
+ */
+struct rvt_qpn_map {
+	void *page;
+};
+
+struct rvt_qpn_table {
+	spinlock_t lock; /* protect changes to the qp table */
+	unsigned flags;         /* flags for QP0/1 allocated for each port */
+	u32 last;               /* last QP number allocated */
+	u32 nmaps;              /* size of the map table */
+	u16 limit;
+	u8  incr;
+	/* bit map of free QP numbers other than 0/1 */
+	struct rvt_qpn_map map[RVT_QPNMAP_ENTRIES];
+};
+
+struct rvt_qp_ibdev {
+	u32 qp_table_size;
+	u32 qp_table_bits;
+	struct rvt_qp __rcu **qp_table;
+	spinlock_t qpt_lock; /* qptable lock */
+	struct rvt_qpn_table qpn_table;
+};
+
+/*
+ * There is one struct rvt_mcast for each multicast GID.
+ * All attached QPs are then stored as a list of
+ * struct rvt_mcast_qp.
+ */
+struct rvt_mcast_qp {
+	struct list_head list;
+	struct rvt_qp *qp;
+};
+
+struct rvt_mcast {
+	struct rb_node rb_node;
+	union ib_gid mgid;
+	struct list_head qp_list;
+	wait_queue_head_t wait;
+	atomic_t refcount;
+	int n_attached;
+};
+
+/*
+ * Since struct rvt_swqe is not a fixed size, we can't simply index into
+ * struct rvt_qp.s_wq.  This function does the array index computation.
+ */
+static inline struct rvt_swqe *rvt_get_swqe_ptr(struct rvt_qp *qp,
+						unsigned n)
+{
+	return (struct rvt_swqe *)((char *)qp->s_wq +
+				     (sizeof(struct rvt_swqe) +
+				      qp->s_max_sge *
+				      sizeof(struct rvt_sge)) * n);
+}
+
+/*
+ * Since struct rvt_rwqe is not a fixed size, we can't simply index into
+ * struct rvt_rwq.wq.  This function does the array index computation.
+ */
+static inline struct rvt_rwqe *rvt_get_rwqe_ptr(struct rvt_rq *rq, unsigned n)
+{
+	return (struct rvt_rwqe *)
+		((char *)rq->wq->wq +
+		 (sizeof(struct rvt_rwqe) +
+		  rq->max_sge * sizeof(struct ib_sge)) * n);
+}
+
+extern const int  ib_rvt_state_ops[];
+
+struct rvt_dev_info;
+int rvt_error_qp(struct rvt_qp *qp, enum ib_wc_status err);
+
+#endif          /* DEF_RDMAVT_INCQP_H */