RDMA 5.10 pull request

The typical set of driver updates across the subsystem:
 
  - Driver minor changes and bug fixes for mlx5, efa, rxe, vmw_pvrdma, hns,
    usnic, qib, qedr, cxgb4, hns, bnxt_re
 
  - Various rtrs fixes and updates
 
  - Bug fix for mlx4 CM emulation for virtualization scenarios where MRA
    wasn't working right
 
  - Use tracepoints instead of pr_debug in the CM code
 
  - Scrub the locking in ucma and cma to close more syzkaller bugs
 
  - Use tasklet_setup in the subsystem
 
  - Revert the idea that 'destroy' operations are not allowed to fail at
    the driver level. This proved unworkable from a HW perspective.
 
  - Revise how the umem API works so drivers make fewer mistakes using it
 
  - XRC support for qedr
 
  - Convert uverbs objects RWQ and MW to new the allocation scheme
 
  - Large queue entry sizes for hns
 
  - Use hmm_range_fault() for mlx5 On Demand Paging
 
  - uverbs APIs to inspect the GID table instead of sysfs
 
  - Move some of the RDMA code for building large page SGLs into
    lib/scatterlist
 -----BEGIN PGP SIGNATURE-----
 
 iQIzBAABCgAdFiEEfB7FMLh+8QxL+6i3OG33FX4gmxoFAl+J37MACgkQOG33FX4g
 mxrKfRAAnIecwdE8df0yvVU5k0Eg6qVjMy9MMHq4va9m7g6GpUcNNI0nIlOASxH2
 l+9vnUQS3ebgsPeECaDYzEr0hh/u53+xw2g4WV5ts/hE8KkQ6erruXb9kasCe8yi
 5QWJ9K36T3c03Cd3EeH6JVtytAxuH42ombfo9BkFLPVyfG/R2tsAzvm5pVi73lxk
 46wtU1Bqi4tsLhyCbifn1huNFGbHp08OIBPAIKPUKCA+iBRPaWS+Dpi+93h3g3Bp
 oJwDhL9CBCGcHM+rKWLzek3Dy87FnQn7R1wmTpUFwkK+4AH3U/XazivhX035w1vL
 YJyhakVU0kosHlX9hJTNKDHJGkt0YEV2mS8dxAuqilFBtdnrVszb5/MirvlzC310
 /b5xCPSEusv9UVZV0G4zbySVNA9knZ4YaRiR3VDVMLKl/pJgTOwEiHIIx+vs3ejk
 p8GRWa1SjXw5LfZEQcq39J689ljt6xjCTonyuBSv7vSQq5v8pjBxvHxiAe2FIa2a
 ZyZeSCYoSh0SwJQukO2VO7aprhHP3TcCJ/987+X03LQ8tV2VWPktHqm62YCaDcOl
 fgiQuQdPivRjDDkJgMfDWDGKfZeHoWLKl5XsJhWByt0lablVrsvc+8ylUl1UI7gI
 16hWB/Qtlhfwg10VdApn+aOFpIS+s5P4XIp8ik57MZO+VeJzpmE=
 =LKpl
 -----END PGP SIGNATURE-----

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

Pull rdma updates from Jason Gunthorpe:
 "A usual cycle for RDMA with a typical mix of driver and core subsystem
  updates:

   - Driver minor changes and bug fixes for mlx5, efa, rxe, vmw_pvrdma,
     hns, usnic, qib, qedr, cxgb4, hns, bnxt_re

   - Various rtrs fixes and updates

   - Bug fix for mlx4 CM emulation for virtualization scenarios where
     MRA wasn't working right

   - Use tracepoints instead of pr_debug in the CM code

   - Scrub the locking in ucma and cma to close more syzkaller bugs

   - Use tasklet_setup in the subsystem

   - Revert the idea that 'destroy' operations are not allowed to fail
     at the driver level. This proved unworkable from a HW perspective.

   - Revise how the umem API works so drivers make fewer mistakes using
     it

   - XRC support for qedr

   - Convert uverbs objects RWQ and MW to new the allocation scheme

   - Large queue entry sizes for hns

   - Use hmm_range_fault() for mlx5 On Demand Paging

   - uverbs APIs to inspect the GID table instead of sysfs

   - Move some of the RDMA code for building large page SGLs into
     lib/scatterlist"

* tag 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/rdma/rdma: (191 commits)
  RDMA/ucma: Fix use after free in destroy id flow
  RDMA/rxe: Handle skb_clone() failure in rxe_recv.c
  RDMA/rxe: Move the definitions for rxe_av.network_type to uAPI
  RDMA: Explicitly pass in the dma_device to ib_register_device
  lib/scatterlist: Do not limit max_segment to PAGE_ALIGNED values
  IB/mlx4: Convert rej_tmout radix-tree to XArray
  RDMA/rxe: Fix bug rejecting all multicast packets
  RDMA/rxe: Fix skb lifetime in rxe_rcv_mcast_pkt()
  RDMA/rxe: Remove duplicate entries in struct rxe_mr
  IB/hfi,rdmavt,qib,opa_vnic: Update MAINTAINERS
  IB/rdmavt: Fix sizeof mismatch
  MAINTAINERS: CISCO VIC LOW LATENCY NIC DRIVER
  RDMA/bnxt_re: Fix sizeof mismatch for allocation of pbl_tbl.
  RDMA/bnxt_re: Use rdma_umem_for_each_dma_block()
  RDMA/umem: Move to allocate SG table from pages
  lib/scatterlist: Add support in dynamic allocation of SG table from pages
  tools/testing/scatterlist: Show errors in human readable form
  tools/testing/scatterlist: Rejuvenate bit-rotten test
  RDMA/ipoib: Set rtnl_link_ops for ipoib interfaces
  RDMA/uverbs: Expose the new GID query API to user space
  ...

.clang-format

@@ -429,6 +429,7 @@ ForEachMacros:
   - 'rbtree_postorder_for_each_entry_safe'
   - 'rdma_for_each_block'
   - 'rdma_for_each_port'
+  - 'rdma_umem_for_each_dma_block'
   - 'resource_list_for_each_entry'
   - 'resource_list_for_each_entry_safe'
   - 'rhl_for_each_entry_rcu'

Documentation/ABI/stable/sysfs-class-infiniband

@@ -258,23 +258,6 @@ Description:
 		userspace ABI compatibility of umad & issm devices.
 
 
-What:		/sys/class/infiniband_cm/ucmN/ibdev
-Date:		Oct, 2005
-KernelVersion:	v2.6.14
-Contact:	linux-rdma@vger.kernel.org
-Description:
-		(RO) Display Infiniband (IB) device name
-
-
-What:		/sys/class/infiniband_cm/abi_version
-Date:		Oct, 2005
-KernelVersion:	v2.6.14
-Contact:	linux-rdma@vger.kernel.org
-Description:
-		(RO) Value is incremented if any changes are made that break
-		userspace ABI compatibility of ucm devices.
-
-
 What:		/sys/class/infiniband_verbs/uverbsN/ibdev
 What:		/sys/class/infiniband_verbs/uverbsN/abi_version
 Date:		Sept, 2005

MAINTAINERS

@@ -4256,7 +4256,6 @@ F:	drivers/net/ethernet/cisco/enic/
 CISCO VIC LOW LATENCY NIC DRIVER
 M:	Christian Benvenuti <benve@cisco.com>
 M:	Nelson Escobar <neescoba@cisco.com>
-M:	Parvi Kaustubhi <pkaustub@cisco.com>
 S:	Supported
 F:	drivers/infiniband/hw/usnic/
 
@@ -7793,8 +7792,8 @@ F:	include/linux/cciss*.h
 F:	include/uapi/linux/cciss*.h
 
 HFI1 DRIVER
-M:	Mike Marciniszyn <mike.marciniszyn@intel.com>
+M:	Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>
-M:	Dennis Dalessandro <dennis.dalessandro@intel.com>
+M:	Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
 L:	linux-rdma@vger.kernel.org
 S:	Supported
 F:	drivers/infiniband/hw/hfi1
@@ -12999,8 +12998,8 @@ S:	Maintained
 F:	drivers/char/hw_random/optee-rng.c
 
 OPA-VNIC DRIVER
-M:	Dennis Dalessandro <dennis.dalessandro@intel.com>
+M:	Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
-M:	Niranjana Vishwanathapura <niranjana.vishwanathapura@intel.com>
+M:	Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>
 L:	linux-rdma@vger.kernel.org
 S:	Supported
 F:	drivers/infiniband/ulp/opa_vnic
@@ -14301,8 +14300,8 @@ F:	drivers/firmware/qemu_fw_cfg.c
 F:	include/uapi/linux/qemu_fw_cfg.h
 
 QIB DRIVER
-M:	Dennis Dalessandro <dennis.dalessandro@intel.com>
+M:	Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
-M:	Mike Marciniszyn <mike.marciniszyn@intel.com>
+M:	Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>
 L:	linux-rdma@vger.kernel.org
 S:	Supported
 F:	drivers/infiniband/hw/qib/
@@ -14727,8 +14726,8 @@ S:	Maintained
 F:	drivers/net/ethernet/rdc/r6040.c
 
 RDMAVT - RDMA verbs software
-M:	Dennis Dalessandro <dennis.dalessandro@intel.com>
+M:	Dennis Dalessandro <dennis.dalessandro@cornelisnetworks.com>
-M:	Mike Marciniszyn <mike.marciniszyn@intel.com>
+M:	Mike Marciniszyn <mike.marciniszyn@cornelisnetworks.com>
 L:	linux-rdma@vger.kernel.org
 S:	Supported
 F:	drivers/infiniband/sw/rdmavt

drivers/gpu/drm/drm_prime.c

@@ -806,30 +806,27 @@ static const struct dma_buf_ops drm_gem_prime_dmabuf_ops =  {
 struct sg_table *drm_prime_pages_to_sg(struct drm_device *dev,
 				       struct page **pages, unsigned int nr_pages)
 {
-	struct sg_table *sg = NULL;
+	struct sg_table *sg;
+	struct scatterlist *sge;
 	size_t max_segment = 0;
-	int ret;
 
 	sg = kmalloc(sizeof(struct sg_table), GFP_KERNEL);
-	if (!sg) {
+	if (!sg)
-		ret = -ENOMEM;
+		return ERR_PTR(-ENOMEM);
-		goto out;
-	}
 
 	if (dev)
 		max_segment = dma_max_mapping_size(dev->dev);
 	if (max_segment == 0 || max_segment > SCATTERLIST_MAX_SEGMENT)
 		max_segment = SCATTERLIST_MAX_SEGMENT;
-	ret = __sg_alloc_table_from_pages(sg, pages, nr_pages, 0,
+	sge = __sg_alloc_table_from_pages(sg, pages, nr_pages, 0,
 					  nr_pages << PAGE_SHIFT,
-					  max_segment, GFP_KERNEL);
+					  max_segment,
-	if (ret)
+					  NULL, 0, GFP_KERNEL);
-		goto out;
+	if (IS_ERR(sge)) {
-
+		kfree(sg);
+		sg = ERR_CAST(sge);
+	}
 	return sg;
-out:
-	kfree(sg);
-	return ERR_PTR(ret);
 }
 EXPORT_SYMBOL(drm_prime_pages_to_sg);
 

drivers/gpu/drm/i915/gem/i915_gem_userptr.c

@@ -403,6 +403,7 @@ __i915_gem_userptr_alloc_pages(struct drm_i915_gem_object *obj,
 	unsigned int max_segment = i915_sg_segment_size();
 	struct sg_table *st;
 	unsigned int sg_page_sizes;
+	struct scatterlist *sg;
 	int ret;
 
 	st = kmalloc(sizeof(*st), GFP_KERNEL);
@@ -410,13 +411,12 @@ __i915_gem_userptr_alloc_pages(struct drm_i915_gem_object *obj,
 		return ERR_PTR(-ENOMEM);
 
 alloc_table:
-	ret = __sg_alloc_table_from_pages(st, pvec, num_pages,
+	sg = __sg_alloc_table_from_pages(st, pvec, num_pages, 0,
-					  0, num_pages << PAGE_SHIFT,
+					 num_pages << PAGE_SHIFT, max_segment,
-					  max_segment,
+					 NULL, 0, GFP_KERNEL);
-					  GFP_KERNEL);
+	if (IS_ERR(sg)) {
-	if (ret) {
 		kfree(st);
-		return ERR_PTR(ret);
+		return ERR_CAST(sg);
 	}
 
 	ret = i915_gem_gtt_prepare_pages(obj, st);

drivers/gpu/drm/vmwgfx/vmwgfx_ttm_buffer.c

@@ -432,6 +432,7 @@ static int vmw_ttm_map_dma(struct vmw_ttm_tt *vmw_tt)
 	int ret = 0;
 	static size_t sgl_size;
 	static size_t sgt_size;
+	struct scatterlist *sg;
 
 	if (vmw_tt->mapped)
 		return 0;
@@ -454,13 +455,15 @@ static int vmw_ttm_map_dma(struct vmw_ttm_tt *vmw_tt)
 		if (unlikely(ret != 0))
 			return ret;
 
-		ret = __sg_alloc_table_from_pages
+		sg = __sg_alloc_table_from_pages(&vmw_tt->sgt, vsgt->pages,
-			(&vmw_tt->sgt, vsgt->pages, vsgt->num_pages, 0,
+				vsgt->num_pages, 0,
-			 (unsigned long) vsgt->num_pages << PAGE_SHIFT,
+				(unsigned long) vsgt->num_pages << PAGE_SHIFT,
-			 dma_get_max_seg_size(dev_priv->dev->dev),
+				dma_get_max_seg_size(dev_priv->dev->dev),
-			 GFP_KERNEL);
+				NULL, 0, GFP_KERNEL);
-		if (unlikely(ret != 0))
+		if (IS_ERR(sg)) {
+			ret = PTR_ERR(sg);
 			goto out_sg_alloc_fail;
+		}
 
 		if (vsgt->num_pages > vmw_tt->sgt.orig_nents) {
 			uint64_t over_alloc =

drivers/infiniband/Kconfig

@@ -48,6 +48,7 @@ config INFINIBAND_ON_DEMAND_PAGING
 	depends on INFINIBAND_USER_MEM
 	select MMU_NOTIFIER
 	select INTERVAL_TREE
+	select HMM_MIRROR
 	default y
 	help
 	  On demand paging support for the InfiniBand subsystem.

drivers/infiniband/core/Makefile

@@ -17,7 +17,7 @@ ib_core-y :=			packer.o ud_header.o verbs.o cq.o rw.o sysfs.o \
 ib_core-$(CONFIG_SECURITY_INFINIBAND) += security.o
 ib_core-$(CONFIG_CGROUP_RDMA) += cgroup.o
 
-ib_cm-y :=			cm.o
+ib_cm-y :=			cm.o cm_trace.o
 
 iw_cm-y :=			iwcm.o iwpm_util.o iwpm_msg.o
 

drivers/infiniband/core/addr.c

@@ -647,13 +647,12 @@ static void process_one_req(struct work_struct *_work)
 	req->callback = NULL;
 
 	spin_lock_bh(&lock);
+	/*
+	 * Although the work will normally have been canceled by the workqueue,
+	 * it can still be requeued as long as it is on the req_list.
+	 */
+	cancel_delayed_work(&req->work);
 	if (!list_empty(&req->list)) {
-		/*
-		 * Although the work will normally have been canceled by the
-		 * workqueue, it can still be requeued as long as it is on the
-		 * req_list.
-		 */
-		cancel_delayed_work(&req->work);
 		list_del_init(&req->list);
 		kfree(req);
 	}

drivers/infiniband/core/cache.c

@@ -133,7 +133,11 @@ static void dispatch_gid_change_event(struct ib_device *ib_dev, u8 port)
 }
 
 static const char * const gid_type_str[] = {
+	/* IB/RoCE v1 value is set for IB_GID_TYPE_IB and IB_GID_TYPE_ROCE for
+	 * user space compatibility reasons.
+	 */
 	[IB_GID_TYPE_IB]	= "IB/RoCE v1",
+	[IB_GID_TYPE_ROCE]	= "IB/RoCE v1",
 	[IB_GID_TYPE_ROCE_UDP_ENCAP]	= "RoCE v2",
 };
 
@@ -1220,7 +1224,7 @@ EXPORT_SYMBOL(ib_get_cached_port_state);
 const struct ib_gid_attr *
 rdma_get_gid_attr(struct ib_device *device, u8 port_num, int index)
 {
-	const struct ib_gid_attr *attr = ERR_PTR(-EINVAL);
+	const struct ib_gid_attr *attr = ERR_PTR(-ENODATA);
 	struct ib_gid_table *table;
 	unsigned long flags;
 
@@ -1243,6 +1247,67 @@ done:
 }
 EXPORT_SYMBOL(rdma_get_gid_attr);
 
+/**
+ * rdma_query_gid_table - Reads GID table entries of all the ports of a device up to max_entries.
+ * @device: The device to query.
+ * @entries: Entries where GID entries are returned.
+ * @max_entries: Maximum number of entries that can be returned.
+ * Entries array must be allocated to hold max_entries number of entries.
+ * @num_entries: Updated to the number of entries that were successfully read.
+ *
+ * Returns number of entries on success or appropriate error code.
+ */
+ssize_t rdma_query_gid_table(struct ib_device *device,
+			     struct ib_uverbs_gid_entry *entries,
+			     size_t max_entries)
+{
+	const struct ib_gid_attr *gid_attr;
+	ssize_t num_entries = 0, ret;
+	struct ib_gid_table *table;
+	unsigned int port_num, i;
+	struct net_device *ndev;
+	unsigned long flags;
+
+	rdma_for_each_port(device, port_num) {
+		if (!rdma_ib_or_roce(device, port_num))
+			continue;
+
+		table = rdma_gid_table(device, port_num);
+		read_lock_irqsave(&table->rwlock, flags);
+		for (i = 0; i < table->sz; i++) {
+			if (!is_gid_entry_valid(table->data_vec[i]))
+				continue;
+			if (num_entries >= max_entries) {
+				ret = -EINVAL;
+				goto err;
+			}
+
+			gid_attr = &table->data_vec[i]->attr;
+
+			memcpy(&entries->gid, &gid_attr->gid,
+			       sizeof(gid_attr->gid));
+			entries->gid_index = gid_attr->index;
+			entries->port_num = gid_attr->port_num;
+			entries->gid_type = gid_attr->gid_type;
+			ndev = rcu_dereference_protected(
+				gid_attr->ndev,
+				lockdep_is_held(&table->rwlock));
+			if (ndev)
+				entries->netdev_ifindex = ndev->ifindex;
+
+			num_entries++;
+			entries++;
+		}
+		read_unlock_irqrestore(&table->rwlock, flags);
+	}
+
+	return num_entries;
+err:
+	read_unlock_irqrestore(&table->rwlock, flags);
+	return ret;
+}
+EXPORT_SYMBOL(rdma_query_gid_table);
+
 /**
  * rdma_put_gid_attr - Release reference to the GID attribute
  * @attr:		Pointer to the GID attribute whose reference
@@ -1299,7 +1364,7 @@ struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr)
 	struct ib_gid_table_entry *entry =
 			container_of(attr, struct ib_gid_table_entry, attr);
 	struct ib_device *device = entry->attr.device;
-	struct net_device *ndev = ERR_PTR(-ENODEV);
+	struct net_device *ndev = ERR_PTR(-EINVAL);
 	u8 port_num = entry->attr.port_num;
 	struct ib_gid_table *table;
 	unsigned long flags;
@@ -1311,8 +1376,7 @@ struct net_device *rdma_read_gid_attr_ndev_rcu(const struct ib_gid_attr *attr)
 	valid = is_gid_entry_valid(table->data_vec[attr->index]);
 	if (valid) {
 		ndev = rcu_dereference(attr->ndev);
-		if (!ndev ||
+		if (!ndev)
-		    (ndev && ((READ_ONCE(ndev->flags) & IFF_UP) == 0)))
 			ndev = ERR_PTR(-ENODEV);
 	}
 	read_unlock_irqrestore(&table->rwlock, flags);

drivers/infiniband/core/cm.c

@@ -27,6 +27,7 @@
 #include <rdma/ib_cm.h>
 #include "cm_msgs.h"
 #include "core_priv.h"
+#include "cm_trace.h"
 
 MODULE_AUTHOR("Sean Hefty");
 MODULE_DESCRIPTION("InfiniBand CM");
@@ -201,7 +202,6 @@ static struct attribute *cm_counter_default_attrs[] = {
 struct cm_port {
 	struct cm_device *cm_dev;
 	struct ib_mad_agent *mad_agent;
-	struct kobject port_obj;
 	u8 port_num;
 	struct list_head cm_priv_prim_list;
 	struct list_head cm_priv_altr_list;
@@ -1563,6 +1563,7 @@ int ib_send_cm_req(struct ib_cm_id *cm_id,
 	cm_id_priv->local_qpn = cpu_to_be32(IBA_GET(CM_REQ_LOCAL_QPN, req_msg));
 	cm_id_priv->rq_psn = cpu_to_be32(IBA_GET(CM_REQ_STARTING_PSN, req_msg));
 
+	trace_icm_send_req(&cm_id_priv->id);
 	spin_lock_irqsave(&cm_id_priv->lock, flags);
 	ret = ib_post_send_mad(cm_id_priv->msg, NULL);
 	if (ret) {
@@ -1610,6 +1611,9 @@ static int cm_issue_rej(struct cm_port *port,
 		IBA_SET_MEM(CM_REJ_ARI, rej_msg, ari, ari_length);
 	}
 
+	trace_icm_issue_rej(
+		IBA_GET(CM_REJ_LOCAL_COMM_ID, rcv_msg),
+		IBA_GET(CM_REJ_REMOTE_COMM_ID, rcv_msg));
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		cm_free_msg(msg);
@@ -1961,6 +1965,7 @@ static void cm_dup_req_handler(struct cm_work *work,
 	}
 	spin_unlock_irq(&cm_id_priv->lock);
 
+	trace_icm_send_dup_req(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		goto free;
@@ -2124,8 +2129,7 @@ static int cm_req_handler(struct cm_work *work)
 
 	listen_cm_id_priv = cm_match_req(work, cm_id_priv);
 	if (!listen_cm_id_priv) {
-		pr_debug("%s: local_id %d, no listen_cm_id_priv\n", __func__,
+		trace_icm_no_listener_err(&cm_id_priv->id);
-			 be32_to_cpu(cm_id_priv->id.local_id));
 		cm_id_priv->id.state = IB_CM_IDLE;
 		ret = -EINVAL;
 		goto destroy;
@@ -2274,8 +2278,7 @@ int ib_send_cm_rep(struct ib_cm_id *cm_id,
 	spin_lock_irqsave(&cm_id_priv->lock, flags);
 	if (cm_id->state != IB_CM_REQ_RCVD &&
 	    cm_id->state != IB_CM_MRA_REQ_SENT) {
-		pr_debug("%s: local_comm_id %d, cm_id->state: %d\n", __func__,
+		trace_icm_send_rep_err(cm_id_priv->id.local_id, cm_id->state);
-			 be32_to_cpu(cm_id_priv->id.local_id), cm_id->state);
 		ret = -EINVAL;
 		goto out;
 	}
@@ -2289,6 +2292,7 @@ int ib_send_cm_rep(struct ib_cm_id *cm_id,
 	msg->timeout_ms = cm_id_priv->timeout_ms;
 	msg->context[1] = (void *) (unsigned long) IB_CM_REP_SENT;
 
+	trace_icm_send_rep(cm_id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
@@ -2348,8 +2352,7 @@ int ib_send_cm_rtu(struct ib_cm_id *cm_id,
 	spin_lock_irqsave(&cm_id_priv->lock, flags);
 	if (cm_id->state != IB_CM_REP_RCVD &&
 	    cm_id->state != IB_CM_MRA_REP_SENT) {
-		pr_debug("%s: local_id %d, cm_id->state %d\n", __func__,
+		trace_icm_send_cm_rtu_err(cm_id);
-			 be32_to_cpu(cm_id->local_id), cm_id->state);
 		ret = -EINVAL;
 		goto error;
 	}
@@ -2361,6 +2364,7 @@ int ib_send_cm_rtu(struct ib_cm_id *cm_id,
 	cm_format_rtu((struct cm_rtu_msg *) msg->mad, cm_id_priv,
 		      private_data, private_data_len);
 
+	trace_icm_send_rtu(cm_id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
@@ -2442,6 +2446,7 @@ static void cm_dup_rep_handler(struct cm_work *work)
 		goto unlock;
 	spin_unlock_irq(&cm_id_priv->lock);
 
+	trace_icm_send_dup_rep(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		goto free;
@@ -2465,7 +2470,7 @@ static int cm_rep_handler(struct cm_work *work)
 		cpu_to_be32(IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg)), 0);
 	if (!cm_id_priv) {
 		cm_dup_rep_handler(work);
-		pr_debug("%s: remote_comm_id %d, no cm_id_priv\n", __func__,
+		trace_icm_remote_no_priv_err(
 			 IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg));
 		return -EINVAL;
 	}
@@ -2479,11 +2484,10 @@ static int cm_rep_handler(struct cm_work *work)
 		break;
 	default:
 		ret = -EINVAL;
-		pr_debug(
+		trace_icm_rep_unknown_err(
-			"%s: cm_id_priv->id.state: %d, local_comm_id %d, remote_comm_id %d\n",
-			__func__, cm_id_priv->id.state,
 			IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg),
-			IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg));
+			IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg),
+			cm_id_priv->id.state);
 		spin_unlock_irq(&cm_id_priv->lock);
 		goto error;
 	}
@@ -2500,7 +2504,7 @@ static int cm_rep_handler(struct cm_work *work)
 		spin_unlock(&cm.lock);
 		spin_unlock_irq(&cm_id_priv->lock);
 		ret = -EINVAL;
-		pr_debug("%s: Failed to insert remote id %d\n", __func__,
+		trace_icm_insert_failed_err(
 			 IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg));
 		goto error;
 	}
@@ -2517,9 +2521,8 @@ static int cm_rep_handler(struct cm_work *work)
 			     IB_CM_REJ_STALE_CONN, CM_MSG_RESPONSE_REP,
 			     NULL, 0);
 		ret = -EINVAL;
-		pr_debug(
+		trace_icm_staleconn_err(
-			"%s: Stale connection. local_comm_id %d, remote_comm_id %d\n",
+			IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg),
-			__func__, IBA_GET(CM_REP_LOCAL_COMM_ID, rep_msg),
 			IBA_GET(CM_REP_REMOTE_COMM_ID, rep_msg));
 
 		if (cur_cm_id_priv) {
@@ -2646,9 +2649,7 @@ static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv,
 		return -EINVAL;
 
 	if (cm_id_priv->id.state != IB_CM_ESTABLISHED) {
-		pr_debug("%s: local_id %d, cm_id->state: %d\n", __func__,
+		trace_icm_dreq_skipped(&cm_id_priv->id);
-			 be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		return -EINVAL;
 	}
 
@@ -2667,6 +2668,7 @@ static int cm_send_dreq_locked(struct cm_id_private *cm_id_priv,
 	msg->timeout_ms = cm_id_priv->timeout_ms;
 	msg->context[1] = (void *) (unsigned long) IB_CM_DREQ_SENT;
 
+	trace_icm_send_dreq(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		cm_enter_timewait(cm_id_priv);
@@ -2722,10 +2724,7 @@ static int cm_send_drep_locked(struct cm_id_private *cm_id_priv,
 		return -EINVAL;
 
 	if (cm_id_priv->id.state != IB_CM_DREQ_RCVD) {
-		pr_debug(
+		trace_icm_send_drep_err(&cm_id_priv->id);
-			"%s: local_id %d, cm_idcm_id->state(%d) != IB_CM_DREQ_RCVD\n",
-			__func__, be32_to_cpu(cm_id_priv->id.local_id),
-			cm_id_priv->id.state);
 		kfree(private_data);
 		return -EINVAL;
 	}
@@ -2740,6 +2739,7 @@ static int cm_send_drep_locked(struct cm_id_private *cm_id_priv,
 	cm_format_drep((struct cm_drep_msg *) msg->mad, cm_id_priv,
 		       private_data, private_data_len);
 
+	trace_icm_send_drep(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		cm_free_msg(msg);
@@ -2789,6 +2789,9 @@ static int cm_issue_drep(struct cm_port *port,
 	IBA_SET(CM_DREP_LOCAL_COMM_ID, drep_msg,
 		IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg));
 
+	trace_icm_issue_drep(
+		IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg),
+		IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg));
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret)
 		cm_free_msg(msg);
@@ -2810,9 +2813,8 @@ static int cm_dreq_handler(struct cm_work *work)
 		atomic_long_inc(&work->port->counter_group[CM_RECV_DUPLICATES].
 				counter[CM_DREQ_COUNTER]);
 		cm_issue_drep(work->port, work->mad_recv_wc);
-		pr_debug(
+		trace_icm_no_priv_err(
-			"%s: no cm_id_priv, local_comm_id %d, remote_comm_id %d\n",
+			IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg),
-			__func__, IBA_GET(CM_DREQ_LOCAL_COMM_ID, dreq_msg),
 			IBA_GET(CM_DREQ_REMOTE_COMM_ID, dreq_msg));
 		return -EINVAL;
 	}
@@ -2858,9 +2860,7 @@ static int cm_dreq_handler(struct cm_work *work)
 				counter[CM_DREQ_COUNTER]);
 		goto unlock;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_dreq_unknown_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		goto unlock;
 	}
 	cm_id_priv->id.state = IB_CM_DREQ_RCVD;
@@ -2945,12 +2945,11 @@ static int cm_send_rej_locked(struct cm_id_private *cm_id_priv,
 			      state);
 		break;
 	default:
-		pr_debug("%s: local_id %d, cm_id->state: %d\n", __func__,
+		trace_icm_send_unknown_rej_err(&cm_id_priv->id);
-			 be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		return -EINVAL;
 	}
 
+	trace_icm_send_rej(&cm_id_priv->id, reason);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		cm_free_msg(msg);
@@ -3060,9 +3059,7 @@ static int cm_rej_handler(struct cm_work *work)
 		}
 		fallthrough;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_rej_unknown_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		spin_unlock_irq(&cm_id_priv->lock);
 		goto out;
 	}
@@ -3118,9 +3115,7 @@ int ib_send_cm_mra(struct ib_cm_id *cm_id,
 		}
 		fallthrough;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_send_mra_unknown_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		ret = -EINVAL;
 		goto error1;
 	}
@@ -3133,6 +3128,7 @@ int ib_send_cm_mra(struct ib_cm_id *cm_id,
 		cm_format_mra((struct cm_mra_msg *) msg->mad, cm_id_priv,
 			      msg_response, service_timeout,
 			      private_data, private_data_len);
+		trace_icm_send_mra(cm_id);
 		ret = ib_post_send_mad(msg, NULL);
 		if (ret)
 			goto error2;
@@ -3229,9 +3225,7 @@ static int cm_mra_handler(struct cm_work *work)
 				counter[CM_MRA_COUNTER]);
 		fallthrough;
 	default:
-		pr_debug("%s local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_mra_unknown_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		goto out;
 	}
 
@@ -3505,10 +3499,12 @@ int ib_send_cm_sidr_req(struct ib_cm_id *cm_id,
 	msg->context[1] = (void *) (unsigned long) IB_CM_SIDR_REQ_SENT;
 
 	spin_lock_irqsave(&cm_id_priv->lock, flags);
-	if (cm_id->state == IB_CM_IDLE)
+	if (cm_id->state == IB_CM_IDLE) {
+		trace_icm_send_sidr_req(&cm_id_priv->id);
 		ret = ib_post_send_mad(msg, NULL);
-	else
+	} else {
 		ret = -EINVAL;
+	}
 
 	if (ret) {
 		spin_unlock_irqrestore(&cm_id_priv->lock, flags);
@@ -3670,6 +3666,7 @@ static int cm_send_sidr_rep_locked(struct cm_id_private *cm_id_priv,
 
 	cm_format_sidr_rep((struct cm_sidr_rep_msg *) msg->mad, cm_id_priv,
 			   param);
+	trace_icm_send_sidr_rep(&cm_id_priv->id);
 	ret = ib_post_send_mad(msg, NULL);
 	if (ret) {
 		cm_free_msg(msg);
@@ -3767,8 +3764,7 @@ static void cm_process_send_error(struct ib_mad_send_buf *msg,
 	if (msg != cm_id_priv->msg || state != cm_id_priv->id.state)
 		goto discard;
 
-	pr_debug_ratelimited("CM: failed sending MAD in state %d. (%s)\n",
+	trace_icm_mad_send_err(state, wc_status);
-			     state, ib_wc_status_msg(wc_status));
 	switch (state) {
 	case IB_CM_REQ_SENT:
 	case IB_CM_MRA_REQ_RCVD:
@@ -3891,7 +3887,7 @@ static void cm_work_handler(struct work_struct *_work)
 		ret = cm_timewait_handler(work);
 		break;
 	default:
-		pr_debug("cm_event.event: 0x%x\n", work->cm_event.event);
+		trace_icm_handler_err(work->cm_event.event);
 		ret = -EINVAL;
 		break;
 	}
@@ -3927,8 +3923,7 @@ static int cm_establish(struct ib_cm_id *cm_id)
 		ret = -EISCONN;
 		break;
 	default:
-		pr_debug("%s: local_id %d, cm_id->state: %d\n", __func__,
+		trace_icm_establish_err(cm_id);
-			 be32_to_cpu(cm_id->local_id), cm_id->state);
 		ret = -EINVAL;
 		break;
 	}
@@ -4125,9 +4120,7 @@ static int cm_init_qp_init_attr(struct cm_id_private *cm_id_priv,
 		ret = 0;
 		break;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_qp_init_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		ret = -EINVAL;
 		break;
 	}
@@ -4175,9 +4168,7 @@ static int cm_init_qp_rtr_attr(struct cm_id_private *cm_id_priv,
 		ret = 0;
 		break;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_qp_rtr_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		ret = -EINVAL;
 		break;
 	}
@@ -4237,9 +4228,7 @@ static int cm_init_qp_rts_attr(struct cm_id_private *cm_id_priv,
 		ret = 0;
 		break;
 	default:
-		pr_debug("%s: local_id %d, cm_id_priv->id.state: %d\n",
+		trace_icm_qp_rts_err(&cm_id_priv->id);
-			 __func__, be32_to_cpu(cm_id_priv->id.local_id),
-			 cm_id_priv->id.state);
 		ret = -EINVAL;
 		break;
 	}
@@ -4295,20 +4284,6 @@ static struct kobj_type cm_counter_obj_type = {
 	.default_attrs = cm_counter_default_attrs
 };
 
-static char *cm_devnode(struct device *dev, umode_t *mode)
-{
-	if (mode)
-		*mode = 0666;
-	return kasprintf(GFP_KERNEL, "infiniband/%s", dev_name(dev));
-}
-
-struct class cm_class = {
-	.owner   = THIS_MODULE,
-	.name    = "infiniband_cm",
-	.devnode = cm_devnode,
-};
-EXPORT_SYMBOL(cm_class);
-
 static int cm_create_port_fs(struct cm_port *port)
 {
 	int i, ret;
@@ -4511,12 +4486,6 @@ static int __init ib_cm_init(void)
 	get_random_bytes(&cm.random_id_operand, sizeof cm.random_id_operand);
 	INIT_LIST_HEAD(&cm.timewait_list);
 
-	ret = class_register(&cm_class);
-	if (ret) {
-		ret = -ENOMEM;
-		goto error1;
-	}
-
 	cm.wq = alloc_workqueue("ib_cm", 0, 1);
 	if (!cm.wq) {
 		ret = -ENOMEM;
@@ -4531,8 +4500,6 @@ static int __init ib_cm_init(void)
 error3:
 	destroy_workqueue(cm.wq);
 error2:
-	class_unregister(&cm_class);
-error1:
 	return ret;
 }
 
@@ -4553,7 +4520,6 @@ static void __exit ib_cm_cleanup(void)
 		kfree(timewait_info);
 	}
 
-	class_unregister(&cm_class);
 	WARN_ON(!xa_empty(&cm.local_id_table));
 }
 

drivers/infiniband/core/cm_trace.c

@@ -0,0 +1,15 @@
+// SPDX-License-Identifier: GPL-2.0-only
+/*
+ * Trace points for the IB Connection Manager.
+ *
+ * Author: Chuck Lever <chuck.lever@oracle.com>
+ *
+ * Copyright (c) 2020, Oracle and/or its affiliates.
+ */
+
+#include <rdma/rdma_cm.h>
+#include "cma_priv.h"
+
+#define CREATE_TRACE_POINTS
+
+#include "cm_trace.h"

drivers/infiniband/core/cm_trace.h

@@ -0,0 +1,414 @@
+/* SPDX-License-Identifier: GPL-2.0-only */
+/*
+ * Trace point definitions for the RDMA Connect Manager.
+ *
+ * Author: Chuck Lever <chuck.lever@oracle.com>
+ *
+ * Copyright (c) 2020 Oracle and/or its affiliates.
+ */
+
+#undef TRACE_SYSTEM
+#define TRACE_SYSTEM ib_cma
+
+#if !defined(_TRACE_IB_CMA_H) || defined(TRACE_HEADER_MULTI_READ)
+
+#define _TRACE_IB_CMA_H
+
+#include <linux/tracepoint.h>
+#include <rdma/ib_cm.h>
+#include <trace/events/rdma.h>
+
+/*
+ * enum ib_cm_state, from include/rdma/ib_cm.h
+ */
+#define IB_CM_STATE_LIST					\
+	ib_cm_state(IDLE)					\
+	ib_cm_state(LISTEN)					\
+	ib_cm_state(REQ_SENT)					\
+	ib_cm_state(REQ_RCVD)					\
+	ib_cm_state(MRA_REQ_SENT)				\
+	ib_cm_state(MRA_REQ_RCVD)				\
+	ib_cm_state(REP_SENT)					\
+	ib_cm_state(REP_RCVD)					\
+	ib_cm_state(MRA_REP_SENT)				\
+	ib_cm_state(MRA_REP_RCVD)				\
+	ib_cm_state(ESTABLISHED)				\
+	ib_cm_state(DREQ_SENT)					\
+	ib_cm_state(DREQ_RCVD)					\
+	ib_cm_state(TIMEWAIT)					\
+	ib_cm_state(SIDR_REQ_SENT)				\
+	ib_cm_state_end(SIDR_REQ_RCVD)
+
+#undef  ib_cm_state
+#undef  ib_cm_state_end
+#define ib_cm_state(x)		TRACE_DEFINE_ENUM(IB_CM_##x);
+#define ib_cm_state_end(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
+
+IB_CM_STATE_LIST
+
+#undef  ib_cm_state
+#undef  ib_cm_state_end
+#define ib_cm_state(x)		{ IB_CM_##x, #x },
+#define ib_cm_state_end(x)	{ IB_CM_##x, #x }
+
+#define show_ib_cm_state(x) \
+		__print_symbolic(x, IB_CM_STATE_LIST)
+
+/*
+ * enum ib_cm_lap_state, from include/rdma/ib_cm.h
+ */
+#define IB_CM_LAP_STATE_LIST					\
+	ib_cm_lap_state(LAP_UNINIT)				\
+	ib_cm_lap_state(LAP_IDLE)				\
+	ib_cm_lap_state(LAP_SENT)				\
+	ib_cm_lap_state(LAP_RCVD)				\
+	ib_cm_lap_state(MRA_LAP_SENT)				\
+	ib_cm_lap_state_end(MRA_LAP_RCVD)
+
+#undef  ib_cm_lap_state
+#undef  ib_cm_lap_state_end
+#define ib_cm_lap_state(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
+#define ib_cm_lap_state_end(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
+
+IB_CM_LAP_STATE_LIST
+
+#undef  ib_cm_lap_state
+#undef  ib_cm_lap_state_end
+#define ib_cm_lap_state(x)	{ IB_CM_##x, #x },
+#define ib_cm_lap_state_end(x)	{ IB_CM_##x, #x }
+
+#define show_ib_cm_lap_state(x) \
+		__print_symbolic(x, IB_CM_LAP_STATE_LIST)
+
+/*
+ * enum ib_cm_rej_reason, from include/rdma/ib_cm.h
+ */
+#define IB_CM_REJ_REASON_LIST					\
+	ib_cm_rej_reason(REJ_NO_QP)				\
+	ib_cm_rej_reason(REJ_NO_EEC)				\
+	ib_cm_rej_reason(REJ_NO_RESOURCES)			\
+	ib_cm_rej_reason(REJ_TIMEOUT)				\
+	ib_cm_rej_reason(REJ_UNSUPPORTED)			\
+	ib_cm_rej_reason(REJ_INVALID_COMM_ID)			\
+	ib_cm_rej_reason(REJ_INVALID_COMM_INSTANCE)		\
+	ib_cm_rej_reason(REJ_INVALID_SERVICE_ID)		\
+	ib_cm_rej_reason(REJ_INVALID_TRANSPORT_TYPE)		\
+	ib_cm_rej_reason(REJ_STALE_CONN)			\
+	ib_cm_rej_reason(REJ_RDC_NOT_EXIST)			\
+	ib_cm_rej_reason(REJ_INVALID_GID)			\
+	ib_cm_rej_reason(REJ_INVALID_LID)			\
+	ib_cm_rej_reason(REJ_INVALID_SL)			\
+	ib_cm_rej_reason(REJ_INVALID_TRAFFIC_CLASS)		\
+	ib_cm_rej_reason(REJ_INVALID_HOP_LIMIT)			\
+	ib_cm_rej_reason(REJ_INVALID_PACKET_RATE)		\
+	ib_cm_rej_reason(REJ_INVALID_ALT_GID)			\
+	ib_cm_rej_reason(REJ_INVALID_ALT_LID)			\
+	ib_cm_rej_reason(REJ_INVALID_ALT_SL)			\
+	ib_cm_rej_reason(REJ_INVALID_ALT_TRAFFIC_CLASS)		\
+	ib_cm_rej_reason(REJ_INVALID_ALT_HOP_LIMIT)		\
+	ib_cm_rej_reason(REJ_INVALID_ALT_PACKET_RATE)		\
+	ib_cm_rej_reason(REJ_PORT_CM_REDIRECT)			\
+	ib_cm_rej_reason(REJ_PORT_REDIRECT)			\
+	ib_cm_rej_reason(REJ_INVALID_MTU)			\
+	ib_cm_rej_reason(REJ_INSUFFICIENT_RESP_RESOURCES)	\
+	ib_cm_rej_reason(REJ_CONSUMER_DEFINED)			\
+	ib_cm_rej_reason(REJ_INVALID_RNR_RETRY)			\
+	ib_cm_rej_reason(REJ_DUPLICATE_LOCAL_COMM_ID)		\
+	ib_cm_rej_reason(REJ_INVALID_CLASS_VERSION)		\
+	ib_cm_rej_reason(REJ_INVALID_FLOW_LABEL)		\
+	ib_cm_rej_reason(REJ_INVALID_ALT_FLOW_LABEL)		\
+	ib_cm_rej_reason_end(REJ_VENDOR_OPTION_NOT_SUPPORTED)
+
+#undef  ib_cm_rej_reason
+#undef  ib_cm_rej_reason_end
+#define ib_cm_rej_reason(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
+#define ib_cm_rej_reason_end(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
+
+IB_CM_REJ_REASON_LIST
+
+#undef  ib_cm_rej_reason
+#undef  ib_cm_rej_reason_end
+#define ib_cm_rej_reason(x)	{ IB_CM_##x, #x },
+#define ib_cm_rej_reason_end(x)	{ IB_CM_##x, #x }
+
+#define show_ib_cm_rej_reason(x) \
+		__print_symbolic(x, IB_CM_REJ_REASON_LIST)
+
+DECLARE_EVENT_CLASS(icm_id_class,
+	TP_PROTO(
+		const struct ib_cm_id *cm_id
+	),
+
+	TP_ARGS(cm_id),
+
+	TP_STRUCT__entry(
+		__field(const void *, cm_id)	/* for eBPF scripts */
+		__field(unsigned int, local_id)
+		__field(unsigned int, remote_id)
+		__field(unsigned long, state)
+		__field(unsigned long, lap_state)
+	),
+
+	TP_fast_assign(
+		__entry->cm_id = cm_id;
+		__entry->local_id = be32_to_cpu(cm_id->local_id);
+		__entry->remote_id = be32_to_cpu(cm_id->remote_id);
+		__entry->state = cm_id->state;
+		__entry->lap_state = cm_id->lap_state;
+	),
+
+	TP_printk("local_id=%u remote_id=%u state=%s lap_state=%s",
+		__entry->local_id, __entry->remote_id,
+		show_ib_cm_state(__entry->state),
+		show_ib_cm_lap_state(__entry->lap_state)
+	)
+);
+
+#define DEFINE_CM_SEND_EVENT(name)					\
+		DEFINE_EVENT(icm_id_class,				\
+				icm_send_##name,				\
+				TP_PROTO(				\
+					const struct ib_cm_id *cm_id	\
+				),					\
+				TP_ARGS(cm_id))
+
+DEFINE_CM_SEND_EVENT(req);
+DEFINE_CM_SEND_EVENT(rep);
+DEFINE_CM_SEND_EVENT(dup_req);
+DEFINE_CM_SEND_EVENT(dup_rep);
+DEFINE_CM_SEND_EVENT(rtu);
+DEFINE_CM_SEND_EVENT(mra);
+DEFINE_CM_SEND_EVENT(sidr_req);
+DEFINE_CM_SEND_EVENT(sidr_rep);
+DEFINE_CM_SEND_EVENT(dreq);
+DEFINE_CM_SEND_EVENT(drep);
+
+TRACE_EVENT(icm_send_rej,
+	TP_PROTO(
+		const struct ib_cm_id *cm_id,
+		enum ib_cm_rej_reason reason
+	),
+
+	TP_ARGS(cm_id, reason),
+
+	TP_STRUCT__entry(
+		__field(const void *, cm_id)
+		__field(u32, local_id)
+		__field(u32, remote_id)
+		__field(unsigned long, state)
+		__field(unsigned long, reason)
+	),
+
+	TP_fast_assign(
+		__entry->cm_id = cm_id;
+		__entry->local_id = be32_to_cpu(cm_id->local_id);
+		__entry->remote_id = be32_to_cpu(cm_id->remote_id);
+		__entry->state = cm_id->state;
+		__entry->reason = reason;
+	),
+
+	TP_printk("local_id=%u remote_id=%u state=%s reason=%s",
+		__entry->local_id, __entry->remote_id,
+		show_ib_cm_state(__entry->state),
+		show_ib_cm_rej_reason(__entry->reason)
+	)
+);
+
+#define DEFINE_CM_ERR_EVENT(name)					\
+		DEFINE_EVENT(icm_id_class,				\
+				icm_##name##_err,			\
+				TP_PROTO(				\
+					const struct ib_cm_id *cm_id	\
+				),					\
+				TP_ARGS(cm_id))
+
+DEFINE_CM_ERR_EVENT(send_cm_rtu);
+DEFINE_CM_ERR_EVENT(establish);
+DEFINE_CM_ERR_EVENT(no_listener);
+DEFINE_CM_ERR_EVENT(send_drep);
+DEFINE_CM_ERR_EVENT(dreq_unknown);
+DEFINE_CM_ERR_EVENT(send_unknown_rej);
+DEFINE_CM_ERR_EVENT(rej_unknown);
+DEFINE_CM_ERR_EVENT(send_mra_unknown);
+DEFINE_CM_ERR_EVENT(mra_unknown);
+DEFINE_CM_ERR_EVENT(qp_init);
+DEFINE_CM_ERR_EVENT(qp_rtr);
+DEFINE_CM_ERR_EVENT(qp_rts);
+
+DEFINE_EVENT(icm_id_class,						\
+	icm_dreq_skipped,						\
+	TP_PROTO(							\
+		const struct ib_cm_id *cm_id				\
+	),								\
+	TP_ARGS(cm_id)							\
+);
+
+DECLARE_EVENT_CLASS(icm_local_class,
+	TP_PROTO(
+		unsigned int local_id,
+		unsigned int remote_id
+	),
+
+	TP_ARGS(local_id, remote_id),
+
+	TP_STRUCT__entry(
+		__field(unsigned int, local_id)
+		__field(unsigned int, remote_id)
+	),
+
+	TP_fast_assign(
+		__entry->local_id = local_id;
+		__entry->remote_id = remote_id;
+	),
+
+	TP_printk("local_id=%u remote_id=%u",
+		__entry->local_id, __entry->remote_id
+	)
+);
+
+#define DEFINE_CM_LOCAL_EVENT(name)					\
+		DEFINE_EVENT(icm_local_class,				\
+				icm_##name,				\
+				TP_PROTO(				\
+					unsigned int local_id,			\
+					unsigned int remote_id			\
+				),					\
+				TP_ARGS(local_id, remote_id))
+
+DEFINE_CM_LOCAL_EVENT(issue_rej);
+DEFINE_CM_LOCAL_EVENT(issue_drep);
+DEFINE_CM_LOCAL_EVENT(staleconn_err);
+DEFINE_CM_LOCAL_EVENT(no_priv_err);
+
+DECLARE_EVENT_CLASS(icm_remote_class,
+	TP_PROTO(
+		u32 remote_id
+	),
+
+	TP_ARGS(remote_id),
+
+	TP_STRUCT__entry(
+		__field(u32, remote_id)
+	),
+
+	TP_fast_assign(
+		__entry->remote_id = remote_id;
+	),
+
+	TP_printk("remote_id=%u",
+		__entry->remote_id
+	)
+);
+
+#define DEFINE_CM_REMOTE_EVENT(name)					\
+		DEFINE_EVENT(icm_remote_class,				\
+				icm_##name,				\
+				TP_PROTO(				\
+					u32 remote_id			\
+				),					\
+				TP_ARGS(remote_id))
+
+DEFINE_CM_REMOTE_EVENT(remote_no_priv_err);
+DEFINE_CM_REMOTE_EVENT(insert_failed_err);
+
+TRACE_EVENT(icm_send_rep_err,
+	TP_PROTO(
+		__be32 local_id,
+		enum ib_cm_state state
+	),
+
+	TP_ARGS(local_id, state),
+
+	TP_STRUCT__entry(
+		__field(unsigned int, local_id)
+		__field(unsigned long, state)
+	),
+
+	TP_fast_assign(
+		__entry->local_id = be32_to_cpu(local_id);
+		__entry->state = state;
+	),
+
+	TP_printk("local_id=%u state=%s",
+		__entry->local_id, show_ib_cm_state(__entry->state)
+	)
+);
+
+TRACE_EVENT(icm_rep_unknown_err,
+	TP_PROTO(
+		unsigned int local_id,
+		unsigned int remote_id,
+		enum ib_cm_state state
+	),
+
+	TP_ARGS(local_id, remote_id, state),
+
+	TP_STRUCT__entry(
+		__field(unsigned int, local_id)
+		__field(unsigned int, remote_id)
+		__field(unsigned long, state)
+	),
+
+	TP_fast_assign(
+		__entry->local_id = local_id;
+		__entry->remote_id = remote_id;
+		__entry->state = state;
+	),
+
+	TP_printk("local_id=%u remote_id=%u state=%s",
+		__entry->local_id, __entry->remote_id,
+		show_ib_cm_state(__entry->state)
+	)
+);
+
+TRACE_EVENT(icm_handler_err,
+	TP_PROTO(
+		enum ib_cm_event_type event
+	),
+
+	TP_ARGS(event),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, event)
+	),
+
+	TP_fast_assign(
+		__entry->event = event;
+	),
+
+	TP_printk("unhandled event=%s",
+		rdma_show_ib_cm_event(__entry->event)
+	)
+);
+
+TRACE_EVENT(icm_mad_send_err,
+	TP_PROTO(
+		enum ib_cm_state state,
+		enum ib_wc_status wc_status
+	),
+
+	TP_ARGS(state, wc_status),
+
+	TP_STRUCT__entry(
+		__field(unsigned long, state)
+		__field(unsigned long, wc_status)
+	),
+
+	TP_fast_assign(
+		__entry->state = state;
+		__entry->wc_status = wc_status;
+	),
+
+	TP_printk("state=%s completion status=%s",
+		show_ib_cm_state(__entry->state),
+		rdma_show_wc_status(__entry->wc_status)
+	)
+);
+
+#endif /* _TRACE_IB_CMA_H */
+
+#undef TRACE_INCLUDE_PATH
+#define TRACE_INCLUDE_PATH ../../drivers/infiniband/core
+#define TRACE_INCLUDE_FILE cm_trace
+
+#include <trace/define_trace.h>

drivers/infiniband/core/cma_configfs.c

@@ -123,16 +123,17 @@ static ssize_t default_roce_mode_store(struct config_item *item,
 {
 	struct cma_device *cma_dev;
 	struct cma_dev_port_group *group;
-	int gid_type = ib_cache_gid_parse_type_str(buf);
+	int gid_type;
 	ssize_t ret;
 
-	if (gid_type < 0)
-		return -EINVAL;
-
 	ret = cma_configfs_params_get(item, &cma_dev, &group);
 	if (ret)
 		return ret;
 
+	gid_type = ib_cache_gid_parse_type_str(buf);
+	if (gid_type < 0)
+		return -EINVAL;
+
 	ret = cma_set_default_gid_type(cma_dev, group->port_num, gid_type);
 
 	cma_configfs_params_put(cma_dev);

drivers/infiniband/core/cma_trace.h

@@ -17,46 +17,6 @@
 #include <linux/tracepoint.h>
 #include <trace/events/rdma.h>
 
-/*
- * enum ib_cm_event_type, from include/rdma/ib_cm.h
- */
-#define IB_CM_EVENT_LIST			\
-	ib_cm_event(REQ_ERROR)			\
-	ib_cm_event(REQ_RECEIVED)		\
-	ib_cm_event(REP_ERROR)			\
-	ib_cm_event(REP_RECEIVED)		\
-	ib_cm_event(RTU_RECEIVED)		\
-	ib_cm_event(USER_ESTABLISHED)		\
-	ib_cm_event(DREQ_ERROR)			\
-	ib_cm_event(DREQ_RECEIVED)		\
-	ib_cm_event(DREP_RECEIVED)		\
-	ib_cm_event(TIMEWAIT_EXIT)		\
-	ib_cm_event(MRA_RECEIVED)		\
-	ib_cm_event(REJ_RECEIVED)		\
-	ib_cm_event(LAP_ERROR)			\
-	ib_cm_event(LAP_RECEIVED)		\
-	ib_cm_event(APR_RECEIVED)		\
-	ib_cm_event(SIDR_REQ_ERROR)		\
-	ib_cm_event(SIDR_REQ_RECEIVED)		\
-	ib_cm_event_end(SIDR_REP_RECEIVED)
-
-#undef ib_cm_event
-#undef ib_cm_event_end
-
-#define ib_cm_event(x)		TRACE_DEFINE_ENUM(IB_CM_##x);
-#define ib_cm_event_end(x)	TRACE_DEFINE_ENUM(IB_CM_##x);
-
-IB_CM_EVENT_LIST
-
-#undef ib_cm_event
-#undef ib_cm_event_end
-
-#define ib_cm_event(x)		{ IB_CM_##x, #x },
-#define ib_cm_event_end(x)	{ IB_CM_##x, #x }
-
-#define rdma_show_ib_cm_event(x) \
-		__print_symbolic(x, IB_CM_EVENT_LIST)
-
 
 DECLARE_EVENT_CLASS(cma_fsm_class,
 	TP_PROTO(

drivers/infiniband/core/core_priv.h

@@ -44,6 +44,7 @@
 #include <rdma/ib_mad.h>
 #include <rdma/restrack.h>
 #include "mad_priv.h"
+#include "restrack.h"
 
 /* Total number of ports combined across all struct ib_devices's */
 #define RDMA_MAX_PORTS 8192
@@ -352,6 +353,7 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev,
 	INIT_LIST_HEAD(&qp->rdma_mrs);
 	INIT_LIST_HEAD(&qp->sig_mrs);
 
+	rdma_restrack_new(&qp->res, RDMA_RESTRACK_QP);
 	/*
 	 * We don't track XRC QPs for now, because they don't have PD
 	 * and more importantly they are created internaly by driver,
@@ -359,14 +361,9 @@ static inline struct ib_qp *_ib_create_qp(struct ib_device *dev,
 	 */
 	is_xrc = qp_type == IB_QPT_XRC_INI || qp_type == IB_QPT_XRC_TGT;
 	if ((qp_type < IB_QPT_MAX && !is_xrc) || qp_type == IB_QPT_DRIVER) {
-		qp->res.type = RDMA_RESTRACK_QP;
+		rdma_restrack_parent_name(&qp->res, &pd->res);
-		if (uobj)
+		rdma_restrack_add(&qp->res);
-			rdma_restrack_uadd(&qp->res);
+	}
-		else
-			rdma_restrack_kadd(&qp->res);
-	} else
-		qp->res.valid = false;
-
 	return qp;
 }
 

drivers/infiniband/core/counters.c

@@ -80,8 +80,9 @@ static struct rdma_counter *rdma_counter_alloc(struct ib_device *dev, u8 port,
 
 	counter->device    = dev;
 	counter->port      = port;
-	counter->res.type  = RDMA_RESTRACK_COUNTER;
+
-	counter->stats     = dev->ops.counter_alloc_stats(counter);
+	rdma_restrack_new(&counter->res, RDMA_RESTRACK_COUNTER);
+	counter->stats = dev->ops.counter_alloc_stats(counter);
 	if (!counter->stats)
 		goto err_stats;
 
@@ -107,6 +108,7 @@ err_mode:
 	mutex_unlock(&port_counter->lock);
 	kfree(counter->stats);
 err_stats:
+	rdma_restrack_put(&counter->res);
 	kfree(counter);
 	return NULL;
 }
@@ -248,13 +250,8 @@ next:
 static void rdma_counter_res_add(struct rdma_counter *counter,
 				 struct ib_qp *qp)
 {
-	if (rdma_is_kernel_res(&qp->res)) {
+	rdma_restrack_parent_name(&counter->res, &qp->res);
-		rdma_restrack_set_task(&counter->res, qp->res.kern_name);
+	rdma_restrack_add(&counter->res);
-		rdma_restrack_kadd(&counter->res);
-	} else {
-		rdma_restrack_attach_task(&counter->res, qp->res.task);
-		rdma_restrack_uadd(&counter->res);
-	}
 }
 
 static void counter_release(struct kref *kref)

drivers/infiniband/core/cq.c

@@ -197,24 +197,22 @@ static void ib_cq_completion_workqueue(struct ib_cq *cq, void *private)
 }
 
 /**
- * __ib_alloc_cq_user - allocate a completion queue
+ * __ib_alloc_cq        allocate a completion queue
  * @dev:		device to allocate the CQ for
  * @private:		driver private data, accessible from cq->cq_context
  * @nr_cqe:		number of CQEs to allocate
  * @comp_vector:	HCA completion vectors for this CQ
  * @poll_ctx:		context to poll the CQ from.
  * @caller:		module owner name.
- * @udata:		Valid user data or NULL for kernel object
  *
  * This is the proper interface to allocate a CQ for in-kernel users. A
  * CQ allocated with this interface will automatically be polled from the
  * specified context. The ULP must use wr->wr_cqe instead of wr->wr_id
  * to use this CQ abstraction.
  */
-struct ib_cq *__ib_alloc_cq_user(struct ib_device *dev, void *private,
+struct ib_cq *__ib_alloc_cq(struct ib_device *dev, void *private, int nr_cqe,
-				 int nr_cqe, int comp_vector,
+			    int comp_vector, enum ib_poll_context poll_ctx,
-				 enum ib_poll_context poll_ctx,
+			    const char *caller)
-				 const char *caller, struct ib_udata *udata)
 {
 	struct ib_cq_init_attr cq_attr = {
 		.cqe		= nr_cqe,
@@ -237,15 +235,13 @@ struct ib_cq *__ib_alloc_cq_user(struct ib_device *dev, void *private,
 	if (!cq->wc)
 		goto out_free_cq;
 
-	cq->res.type = RDMA_RESTRACK_CQ;
+	rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ);
-	rdma_restrack_set_task(&cq->res, caller);
+	rdma_restrack_set_name(&cq->res, caller);
 
 	ret = dev->ops.create_cq(cq, &cq_attr, NULL);
 	if (ret)
 		goto out_free_wc;
 
-	rdma_restrack_kadd(&cq->res);
-
 	rdma_dim_init(cq);
 
 	switch (cq->poll_ctx) {
@@ -271,21 +267,22 @@ struct ib_cq *__ib_alloc_cq_user(struct ib_device *dev, void *private,
 		goto out_destroy_cq;
 	}
 
+	rdma_restrack_add(&cq->res);
 	trace_cq_alloc(cq, nr_cqe, comp_vector, poll_ctx);
 	return cq;
 
 out_destroy_cq:
 	rdma_dim_destroy(cq);
-	rdma_restrack_del(&cq->res);
+	cq->device->ops.destroy_cq(cq, NULL);
-	cq->device->ops.destroy_cq(cq, udata);
 out_free_wc:
+	rdma_restrack_put(&cq->res);
 	kfree(cq->wc);
 out_free_cq:
 	kfree(cq);
 	trace_cq_alloc_error(nr_cqe, comp_vector, poll_ctx, ret);
 	return ERR_PTR(ret);
 }
-EXPORT_SYMBOL(__ib_alloc_cq_user);
+EXPORT_SYMBOL(__ib_alloc_cq);
 
 /**
  * __ib_alloc_cq_any - allocate a completion queue
@@ -310,18 +307,19 @@ struct ib_cq *__ib_alloc_cq_any(struct ib_device *dev, void *private,
 			atomic_inc_return(&counter) %
 			min_t(int, dev->num_comp_vectors, num_online_cpus());
 
-	return __ib_alloc_cq_user(dev, private, nr_cqe, comp_vector, poll_ctx,
+	return __ib_alloc_cq(dev, private, nr_cqe, comp_vector, poll_ctx,
-				  caller, NULL);
+			     caller);
 }
 EXPORT_SYMBOL(__ib_alloc_cq_any);
 
 /**
- * ib_free_cq_user - free a completion queue
+ * ib_free_cq - free a completion queue
  * @cq:		completion queue to free.
- * @udata:	User data or NULL for kernel object
  */
-void ib_free_cq_user(struct ib_cq *cq, struct ib_udata *udata)
+void ib_free_cq(struct ib_cq *cq)
 {
+	int ret;
+
 	if (WARN_ON_ONCE(atomic_read(&cq->usecnt)))
 		return;
 	if (WARN_ON_ONCE(cq->cqe_used))
@@ -343,12 +341,13 @@ void ib_free_cq_user(struct ib_cq *cq, struct ib_udata *udata)
 
 	rdma_dim_destroy(cq);
 	trace_cq_free(cq);
+	ret = cq->device->ops.destroy_cq(cq, NULL);
+	WARN_ONCE(ret, "Destroy of kernel CQ shouldn't fail");
 	rdma_restrack_del(&cq->res);
-	cq->device->ops.destroy_cq(cq, udata);
 	kfree(cq->wc);
 	kfree(cq);
 }
-EXPORT_SYMBOL(ib_free_cq_user);
+EXPORT_SYMBOL(ib_free_cq);
 
 void ib_cq_pool_init(struct ib_device *dev)
 {

drivers/infiniband/core/device.c

@@ -1177,58 +1177,23 @@ out:
 	return ret;
 }
 
-static void setup_dma_device(struct ib_device *device)
+static void setup_dma_device(struct ib_device *device,
+			     struct device *dma_device)
 {
-	struct device *parent = device->dev.parent;
+	/*
-
+	 * If the caller does not provide a DMA capable device then the IB
-	WARN_ON_ONCE(device->dma_device);
+	 * device will be used. In this case the caller should fully setup the
-
+	 * ibdev for DMA. This usually means using dma_virt_ops.
-#ifdef CONFIG_DMA_OPS
+	 */
-	if (device->dev.dma_ops) {
+#ifdef CONFIG_DMA_VIRT_OPS
-		/*
+	if (!dma_device) {
-		 * The caller provided custom DMA operations. Copy the
+		device->dev.dma_ops = &dma_virt_ops;
-		 * DMA-related fields that are used by e.g. dma_alloc_coherent()
+		dma_device = &device->dev;
-		 * into device->dev.
-		 */
-		device->dma_device = &device->dev;
-		if (!device->dev.dma_mask) {
-			if (parent)
-				device->dev.dma_mask = parent->dma_mask;
-			else
-				WARN_ON_ONCE(true);
-		}
-		if (!device->dev.coherent_dma_mask) {
-			if (parent)
-				device->dev.coherent_dma_mask =
-					parent->coherent_dma_mask;
-			else
-				WARN_ON_ONCE(true);
-		}
-	} else
-#endif /* CONFIG_DMA_OPS */
-	{
-		/*
-		 * The caller did not provide custom DMA operations. Use the
-		 * DMA mapping operations of the parent device.
-		 */
-		WARN_ON_ONCE(!parent);
-		device->dma_device = parent;
-	}
-
-	if (!device->dev.dma_parms) {
-		if (parent) {
-			/*
-			 * The caller did not provide DMA parameters, so
-			 * 'parent' probably represents a PCI device. The PCI
-			 * core sets the maximum segment size to 64
-			 * KB. Increase this parameter to 2 GB.
-			 */
-			device->dev.dma_parms = parent->dma_parms;
-			dma_set_max_seg_size(device->dma_device, SZ_2G);
-		} else {
-			WARN_ON_ONCE(true);
-		}
 	}
+#endif
+	WARN_ON(!dma_device);
+	device->dma_device = dma_device;
+	WARN_ON(!device->dma_device->dma_parms);
 }
 
 /*
@@ -1241,7 +1206,6 @@ static int setup_device(struct ib_device *device)
 	struct ib_udata uhw = {.outlen = 0, .inlen = 0};
 	int ret;
 
-	setup_dma_device(device);
 	ib_device_check_mandatory(device);
 
 	ret = setup_port_data(device);
@@ -1354,7 +1318,10 @@ static void prevent_dealloc_device(struct ib_device *ib_dev)
  * ib_register_device - Register an IB device with IB core
  * @device: Device to register
  * @name: unique string device name. This may include a '%' which will
- * cause a unique index to be added to the passed device name.
+ * 	  cause a unique index to be added to the passed device name.
+ * @dma_device: pointer to a DMA-capable device. If %NULL, then the IB
+ *	        device will be used. In this case the caller should fully
+ *		setup the ibdev for DMA. This usually means using dma_virt_ops.
  *
  * Low-level drivers use ib_register_device() to register their
  * devices with the IB core.  All registered clients will receive a
@@ -1365,7 +1332,8 @@ static void prevent_dealloc_device(struct ib_device *ib_dev)
  * asynchronously then the device pointer may become freed as soon as this
  * function returns.
  */
-int ib_register_device(struct ib_device *device, const char *name)
+int ib_register_device(struct ib_device *device, const char *name,
+		       struct device *dma_device)
 {
 	int ret;
 
@@ -1373,6 +1341,7 @@ int ib_register_device(struct ib_device *device, const char *name)
 	if (ret)
 		return ret;
 
+	setup_dma_device(device, dma_device);
 	ret = setup_device(device);
 	if (ret)
 		return ret;
@@ -2697,7 +2666,9 @@ void ib_set_device_ops(struct ib_device *dev, const struct ib_device_ops *ops)
 	SET_OBJ_SIZE(dev_ops, ib_ah);
 	SET_OBJ_SIZE(dev_ops, ib_counters);
 	SET_OBJ_SIZE(dev_ops, ib_cq);
+	SET_OBJ_SIZE(dev_ops, ib_mw);
 	SET_OBJ_SIZE(dev_ops, ib_pd);
+	SET_OBJ_SIZE(dev_ops, ib_rwq_ind_table);
 	SET_OBJ_SIZE(dev_ops, ib_srq);
 	SET_OBJ_SIZE(dev_ops, ib_ucontext);
 	SET_OBJ_SIZE(dev_ops, ib_xrcd);

drivers/infiniband/core/rdma_core.c

@@ -130,17 +130,6 @@ static int uverbs_destroy_uobject(struct ib_uobject *uobj,
 	lockdep_assert_held(&ufile->hw_destroy_rwsem);
 	assert_uverbs_usecnt(uobj, UVERBS_LOOKUP_WRITE);
 
-	if (reason == RDMA_REMOVE_ABORT_HWOBJ) {
-		reason = RDMA_REMOVE_ABORT;
-		ret = uobj->uapi_object->type_class->destroy_hw(uobj, reason,
-								attrs);
-		/*
-		 * Drivers are not permitted to ignore RDMA_REMOVE_ABORT, see
-		 * ib_is_destroy_retryable, cleanup_retryable == false here.
-		 */
-		WARN_ON(ret);
-	}
-
 	if (reason == RDMA_REMOVE_ABORT) {
 		WARN_ON(!list_empty(&uobj->list));
 		WARN_ON(!uobj->context);
@@ -674,11 +663,22 @@ void rdma_alloc_abort_uobject(struct ib_uobject *uobj,
 			      bool hw_obj_valid)
 {
 	struct ib_uverbs_file *ufile = uobj->ufile;
+	int ret;
 
-	uverbs_destroy_uobject(uobj,
+	if (hw_obj_valid) {
-			       hw_obj_valid ? RDMA_REMOVE_ABORT_HWOBJ :
+		ret = uobj->uapi_object->type_class->destroy_hw(
-					      RDMA_REMOVE_ABORT,
+			uobj, RDMA_REMOVE_ABORT, attrs);
-			       attrs);
+		/*
+		 * If the driver couldn't destroy the object then go ahead and
+		 * commit it. Leaking objects that can't be destroyed is only
+		 * done during FD close after the driver has a few more tries to
+		 * destroy it.
+		 */
+		if (WARN_ON(ret))
+			return rdma_alloc_commit_uobject(uobj, attrs);
+	}
+
+	uverbs_destroy_uobject(uobj, RDMA_REMOVE_ABORT, attrs);
 
 	/* Matches the down_read in rdma_alloc_begin_uobject */
 	up_read(&ufile->hw_destroy_rwsem);
@@ -889,14 +889,14 @@ void uverbs_destroy_ufile_hw(struct ib_uverbs_file *ufile,
 	if (!ufile->ucontext)
 		goto done;
 
-	ufile->ucontext->closing = true;
 	ufile->ucontext->cleanup_retryable = true;
 	while (!list_empty(&ufile->uobjects))
 		if (__uverbs_cleanup_ufile(ufile, reason)) {
 			/*
 			 * No entry was cleaned-up successfully during this
-			 * iteration
+			 * iteration. It is a driver bug to fail destruction.
 			 */
+			WARN_ON(!list_empty(&ufile->uobjects));
 			break;
 		}
 

drivers/infiniband/core/restrack.c

@@ -123,32 +123,6 @@ int rdma_restrack_count(struct ib_device *dev, enum rdma_restrack_type type)
 }
 EXPORT_SYMBOL(rdma_restrack_count);
 
-static void set_kern_name(struct rdma_restrack_entry *res)
-{
-	struct ib_pd *pd;
-
-	switch (res->type) {
-	case RDMA_RESTRACK_QP:
-		pd = container_of(res, struct ib_qp, res)->pd;
-		if (!pd) {
-			WARN_ONCE(true, "XRC QPs are not supported\n");
-			/* Survive, despite the programmer's error */
-			res->kern_name = " ";
-		}
-		break;
-	case RDMA_RESTRACK_MR:
-		pd = container_of(res, struct ib_mr, res)->pd;
-		break;
-	default:
-		/* Other types set kern_name directly */
-		pd = NULL;
-		break;
-	}
-
-	if (pd)
-		res->kern_name = pd->res.kern_name;
-}
-
 static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 {
 	switch (res->type) {
@@ -173,36 +147,77 @@ static struct ib_device *res_to_dev(struct rdma_restrack_entry *res)
 	}
 }
 
-void rdma_restrack_set_task(struct rdma_restrack_entry *res,
+/**
-			    const char *caller)
+ * rdma_restrack_attach_task() - attach the task onto this resource,
+ * valid for user space restrack entries.
+ * @res:  resource entry
+ * @task: the task to attach
+ */
+static void rdma_restrack_attach_task(struct rdma_restrack_entry *res,
+				      struct task_struct *task)
+{
+	if (WARN_ON_ONCE(!task))
+		return;
+
+	if (res->task)
+		put_task_struct(res->task);
+	get_task_struct(task);
+	res->task = task;
+	res->user = true;
+}
+
+/**
+ * rdma_restrack_set_name() - set the task for this resource
+ * @res:  resource entry
+ * @caller: kernel name, the current task will be used if the caller is NULL.
+ */
+void rdma_restrack_set_name(struct rdma_restrack_entry *res, const char *caller)
 {
 	if (caller) {
 		res->kern_name = caller;
 		return;
 	}
 
-	if (res->task)
+	rdma_restrack_attach_task(res, current);
-		put_task_struct(res->task);
-	get_task_struct(current);
-	res->task = current;
 }
-EXPORT_SYMBOL(rdma_restrack_set_task);
+EXPORT_SYMBOL(rdma_restrack_set_name);
 
 /**
- * rdma_restrack_attach_task() - attach the task onto this resource
+ * rdma_restrack_parent_name() - set the restrack name properties based
- * @res:  resource entry
+ * on parent restrack
- * @task: the task to attach, the current task will be used if it is NULL.
+ * @dst: destination resource entry
+ * @parent: parent resource entry
  */
-void rdma_restrack_attach_task(struct rdma_restrack_entry *res,
+void rdma_restrack_parent_name(struct rdma_restrack_entry *dst,
-			       struct task_struct *task)
+			       const struct rdma_restrack_entry *parent)
 {
-	if (res->task)
+	if (rdma_is_kernel_res(parent))
-		put_task_struct(res->task);
+		dst->kern_name = parent->kern_name;
-	get_task_struct(task);
+	else
-	res->task = task;
+		rdma_restrack_attach_task(dst, parent->task);
 }
+EXPORT_SYMBOL(rdma_restrack_parent_name);
 
-static void rdma_restrack_add(struct rdma_restrack_entry *res)
+/**
+ * rdma_restrack_new() - Initializes new restrack entry to allow _put() interface
+ * to release memory in fully automatic way.
+ * @res - Entry to initialize
+ * @type - REstrack type
+ */
+void rdma_restrack_new(struct rdma_restrack_entry *res,
+		       enum rdma_restrack_type type)
+{
+	kref_init(&res->kref);
+	init_completion(&res->comp);
+	res->type = type;
+}
+EXPORT_SYMBOL(rdma_restrack_new);
+
+/**
+ * rdma_restrack_add() - add object to the reource tracking database
+ * @res:  resource entry
+ */
+void rdma_restrack_add(struct rdma_restrack_entry *res)
 {
 	struct ib_device *dev = res_to_dev(res);
 	struct rdma_restrack_root *rt;
@@ -213,8 +228,6 @@ static void rdma_restrack_add(struct rdma_restrack_entry *res)
 
 	rt = &dev->res[res->type];
 
-	kref_init(&res->kref);
-	init_completion(&res->comp);
 	if (res->type == RDMA_RESTRACK_QP) {
 		/* Special case to ensure that LQPN points to right QP */
 		struct ib_qp *qp = container_of(res, struct ib_qp, res);
@@ -236,38 +249,7 @@ static void rdma_restrack_add(struct rdma_restrack_entry *res)
 	if (!ret)
 		res->valid = true;
 }
-
+EXPORT_SYMBOL(rdma_restrack_add);
-/**
- * rdma_restrack_kadd() - add kernel object to the reource tracking database
- * @res:  resource entry
- */
-void rdma_restrack_kadd(struct rdma_restrack_entry *res)
-{
-	res->task = NULL;
-	set_kern_name(res);
-	res->user = false;
-	rdma_restrack_add(res);
-}
-EXPORT_SYMBOL(rdma_restrack_kadd);
-
-/**
- * rdma_restrack_uadd() - add user object to the reource tracking database
- * @res:  resource entry
- */
-void rdma_restrack_uadd(struct rdma_restrack_entry *res)
-{
-	if ((res->type != RDMA_RESTRACK_CM_ID) &&
-	    (res->type != RDMA_RESTRACK_COUNTER))
-		res->task = NULL;
-
-	if (!res->task)
-		rdma_restrack_set_task(res, NULL);
-	res->kern_name = NULL;
-
-	res->user = true;
-	rdma_restrack_add(res);
-}
-EXPORT_SYMBOL(rdma_restrack_uadd);
 
 int __must_check rdma_restrack_get(struct rdma_restrack_entry *res)
 {
@@ -305,6 +287,10 @@ static void restrack_release(struct kref *kref)
 	struct rdma_restrack_entry *res;
 
 	res = container_of(kref, struct rdma_restrack_entry, kref);
+	if (res->task) {
+		put_task_struct(res->task);
+		res->task = NULL;
+	}
 	complete(&res->comp);
 }
 
@@ -314,14 +300,23 @@ int rdma_restrack_put(struct rdma_restrack_entry *res)
 }
 EXPORT_SYMBOL(rdma_restrack_put);
 
+/**
+ * rdma_restrack_del() - delete object from the reource tracking database
+ * @res:  resource entry
+ */
 void rdma_restrack_del(struct rdma_restrack_entry *res)
 {
 	struct rdma_restrack_entry *old;
 	struct rdma_restrack_root *rt;
 	struct ib_device *dev;
 
-	if (!res->valid)
+	if (!res->valid) {
-		goto out;
+		if (res->task) {
+			put_task_struct(res->task);
+			res->task = NULL;
+		}
+		return;
+	}
 
 	dev = res_to_dev(res);
 	if (WARN_ON(!dev))
@@ -330,16 +325,12 @@ void rdma_restrack_del(struct rdma_restrack_entry *res)
 	rt = &dev->res[res->type];
 
 	old = xa_erase(&rt->xa, res->id);
+	if (res->type == RDMA_RESTRACK_MR || res->type == RDMA_RESTRACK_QP)
+		return;
 	WARN_ON(old != res);
 	res->valid = false;
 
 	rdma_restrack_put(res);
 	wait_for_completion(&res->comp);
-
-out:
-	if (res->task) {
-		put_task_struct(res->task);
-		res->task = NULL;
-	}
 }
 EXPORT_SYMBOL(rdma_restrack_del);

drivers/infiniband/core/restrack.h

@@ -25,6 +25,12 @@ struct rdma_restrack_root {
 
 int rdma_restrack_init(struct ib_device *dev);
 void rdma_restrack_clean(struct ib_device *dev);
-void rdma_restrack_attach_task(struct rdma_restrack_entry *res,
+void rdma_restrack_add(struct rdma_restrack_entry *res);
-			       struct task_struct *task);
+void rdma_restrack_del(struct rdma_restrack_entry *res);
+void rdma_restrack_new(struct rdma_restrack_entry *res,
+		       enum rdma_restrack_type type);
+void rdma_restrack_set_name(struct rdma_restrack_entry *res,
+			    const char *caller);
+void rdma_restrack_parent_name(struct rdma_restrack_entry *dst,
+			       const struct rdma_restrack_entry *parent);
 #endif /* _RDMA_CORE_RESTRACK_H_ */

drivers/infiniband/core/sysfs.c

@@ -59,7 +59,7 @@ struct ib_port {
 	struct gid_attr_group *gid_attr_group;
 	struct attribute_group gid_group;
 	struct attribute_group *pkey_group;
-	struct attribute_group *pma_table;
+	const struct attribute_group *pma_table;
 	struct attribute_group *hw_stats_ag;
 	struct rdma_hw_stats   *hw_stats;
 	u8                     port_num;
@@ -387,7 +387,8 @@ static ssize_t _show_port_gid_attr(
 
 	gid_attr = rdma_get_gid_attr(p->ibdev, p->port_num, tab_attr->index);
 	if (IS_ERR(gid_attr))
-		return PTR_ERR(gid_attr);
+		/* -EINVAL is returned for user space compatibility reasons. */
+		return -EINVAL;
 
 	ret = print(gid_attr, buf);
 	rdma_put_gid_attr(gid_attr);
@@ -653,17 +654,17 @@ static struct attribute *pma_attrs_noietf[] = {
 	NULL
 };
 
-static struct attribute_group pma_group = {
+static const struct attribute_group pma_group = {
 	.name  = "counters",
 	.attrs  = pma_attrs
 };
 
-static struct attribute_group pma_group_ext = {
+static const struct attribute_group pma_group_ext = {
 	.name  = "counters",
 	.attrs  = pma_attrs_ext
 };
 
-static struct attribute_group pma_group_noietf = {
+static const struct attribute_group pma_group_noietf = {
 	.name  = "counters",
 	.attrs  = pma_attrs_noietf
 };
@@ -778,8 +779,8 @@ err:
  * Figure out which counter table to use depending on
  * the device capabilities.
  */
-static struct attribute_group *get_counter_table(struct ib_device *dev,
+static const struct attribute_group *get_counter_table(struct ib_device *dev,
-						 int port_num)
+						       int port_num)
 {
 	struct ib_class_port_info cpi;
 

drivers/infiniband/core/ucma.c

@@ -80,7 +80,6 @@ struct ucma_file {
 	struct list_head	ctx_list;
 	struct list_head	event_list;
 	wait_queue_head_t	poll_wait;
-	struct workqueue_struct	*close_wq;
 };
 
 struct ucma_context {
@@ -88,7 +87,7 @@ struct ucma_context {
 	struct completion	comp;
 	refcount_t		ref;
 	int			events_reported;
-	int			backlog;
+	atomic_t		backlog;
 
 	struct ucma_file	*file;
 	struct rdma_cm_id	*cm_id;
@@ -96,11 +95,6 @@ struct ucma_context {
 	u64			uid;
 
 	struct list_head	list;
-	struct list_head	mc_list;
-	/* mark that device is in process of destroying the internal HW
-	 * resources, protected by the ctx_table lock
-	 */
-	int			closing;
 	/* sync between removal event and id destroy, protected by file mut */
 	int			destroying;
 	struct work_struct	close_work;
@@ -113,23 +107,22 @@ struct ucma_multicast {
 
 	u64			uid;
 	u8			join_state;
-	struct list_head	list;
 	struct sockaddr_storage	addr;
 };
 
 struct ucma_event {
 	struct ucma_context	*ctx;
+	struct ucma_context	*conn_req_ctx;
 	struct ucma_multicast	*mc;
 	struct list_head	list;
-	struct rdma_cm_id	*cm_id;
 	struct rdma_ucm_event_resp resp;
-	struct work_struct	close_work;
 };
 
 static DEFINE_XARRAY_ALLOC(ctx_table);
 static DEFINE_XARRAY_ALLOC(multicast_table);
 
 static const struct file_operations ucma_fops;
+static int __destroy_id(struct ucma_context *ctx);
 
 static inline struct ucma_context *_ucma_find_context(int id,
 						      struct ucma_file *file)
@@ -139,7 +132,7 @@ static inline struct ucma_context *_ucma_find_context(int id,
 	ctx = xa_load(&ctx_table, id);
 	if (!ctx)
 		ctx = ERR_PTR(-ENOENT);
-	else if (ctx->file != file || !ctx->cm_id)
+	else if (ctx->file != file)
 		ctx = ERR_PTR(-EINVAL);
 	return ctx;
 }
@@ -150,12 +143,9 @@ static struct ucma_context *ucma_get_ctx(struct ucma_file *file, int id)
 
 	xa_lock(&ctx_table);
 	ctx = _ucma_find_context(id, file);
-	if (!IS_ERR(ctx)) {
+	if (!IS_ERR(ctx))
-		if (ctx->closing)
+		if (!refcount_inc_not_zero(&ctx->ref))
-			ctx = ERR_PTR(-EIO);
+			ctx = ERR_PTR(-ENXIO);
-		else
-			refcount_inc(&ctx->ref);
-	}
 	xa_unlock(&ctx_table);
 	return ctx;
 }
@@ -183,14 +173,6 @@ static struct ucma_context *ucma_get_ctx_dev(struct ucma_file *file, int id)
 	return ctx;
 }
 
-static void ucma_close_event_id(struct work_struct *work)
-{
-	struct ucma_event *uevent_close =  container_of(work, struct ucma_event, close_work);
-
-	rdma_destroy_id(uevent_close->cm_id);
-	kfree(uevent_close);
-}
-
 static void ucma_close_id(struct work_struct *work)
 {
 	struct ucma_context *ctx =  container_of(work, struct ucma_context, close_work);
@@ -203,6 +185,14 @@ static void ucma_close_id(struct work_struct *work)
 	wait_for_completion(&ctx->comp);
 	/* No new events will be generated after destroying the id. */
 	rdma_destroy_id(ctx->cm_id);
+
+	/*
+	 * At this point ctx->ref is zero so the only place the ctx can be is in
+	 * a uevent or in __destroy_id(). Since the former doesn't touch
+	 * ctx->cm_id and the latter sync cancels this, there is no races with
+	 * this store.
+	 */
+	ctx->cm_id = NULL;
 }
 
 static struct ucma_context *ucma_alloc_ctx(struct ucma_file *file)
@@ -216,39 +206,23 @@ static struct ucma_context *ucma_alloc_ctx(struct ucma_file *file)
 	INIT_WORK(&ctx->close_work, ucma_close_id);
 	refcount_set(&ctx->ref, 1);
 	init_completion(&ctx->comp);
-	INIT_LIST_HEAD(&ctx->mc_list);
+	/* So list_del() will work if we don't do ucma_finish_ctx() */
+	INIT_LIST_HEAD(&ctx->list);
 	ctx->file = file;
 	mutex_init(&ctx->mutex);
 
-	if (xa_alloc(&ctx_table, &ctx->id, ctx, xa_limit_32b, GFP_KERNEL))
+	if (xa_alloc(&ctx_table, &ctx->id, NULL, xa_limit_32b, GFP_KERNEL)) {
-		goto error;
+		kfree(ctx);
-
+		return NULL;
-	list_add_tail(&ctx->list, &file->ctx_list);
+	}
 	return ctx;
-
-error:
-	kfree(ctx);
-	return NULL;
 }
 
-static struct ucma_multicast* ucma_alloc_multicast(struct ucma_context *ctx)
+static void ucma_finish_ctx(struct ucma_context *ctx)
 {
-	struct ucma_multicast *mc;
+	lockdep_assert_held(&ctx->file->mut);
-
+	list_add_tail(&ctx->list, &ctx->file->ctx_list);
-	mc = kzalloc(sizeof(*mc), GFP_KERNEL);
+	xa_store(&ctx_table, ctx->id, ctx, GFP_KERNEL);
-	if (!mc)
-		return NULL;
-
-	mc->ctx = ctx;
-	if (xa_alloc(&multicast_table, &mc->id, NULL, xa_limit_32b, GFP_KERNEL))
-		goto error;
-
-	list_add_tail(&mc->list, &ctx->mc_list);
-	return mc;
-
-error:
-	kfree(mc);
-	return NULL;
 }
 
 static void ucma_copy_conn_event(struct rdma_ucm_conn_param *dst,
@@ -280,10 +254,15 @@ static void ucma_copy_ud_event(struct ib_device *device,
 	dst->qkey = src->qkey;
 }
 
-static void ucma_set_event_context(struct ucma_context *ctx,
+static struct ucma_event *ucma_create_uevent(struct ucma_context *ctx,
-				   struct rdma_cm_event *event,
+					     struct rdma_cm_event *event)
-				   struct ucma_event *uevent)
 {
+	struct ucma_event *uevent;
+
+	uevent = kzalloc(sizeof(*uevent), GFP_KERNEL);
+	if (!uevent)
+		return NULL;
+
 	uevent->ctx = ctx;
 	switch (event->event) {
 	case RDMA_CM_EVENT_MULTICAST_JOIN:
@@ -298,64 +277,10 @@ static void ucma_set_event_context(struct ucma_context *ctx,
 		uevent->resp.id = ctx->id;
 		break;
 	}
-}
-
-/* Called with file->mut locked for the relevant context. */
-static void ucma_removal_event_handler(struct rdma_cm_id *cm_id)
-{
-	struct ucma_context *ctx = cm_id->context;
-	struct ucma_event *con_req_eve;
-	int event_found = 0;
-
-	if (ctx->destroying)
-		return;
-
-	/* only if context is pointing to cm_id that it owns it and can be
-	 * queued to be closed, otherwise that cm_id is an inflight one that
-	 * is part of that context event list pending to be detached and
-	 * reattached to its new context as part of ucma_get_event,
-	 * handled separately below.
-	 */
-	if (ctx->cm_id == cm_id) {
-		xa_lock(&ctx_table);
-		ctx->closing = 1;
-		xa_unlock(&ctx_table);
-		queue_work(ctx->file->close_wq, &ctx->close_work);
-		return;
-	}
-
-	list_for_each_entry(con_req_eve, &ctx->file->event_list, list) {
-		if (con_req_eve->cm_id == cm_id &&
-		    con_req_eve->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST) {
-			list_del(&con_req_eve->list);
-			INIT_WORK(&con_req_eve->close_work, ucma_close_event_id);
-			queue_work(ctx->file->close_wq, &con_req_eve->close_work);
-			event_found = 1;
-			break;
-		}
-	}
-	if (!event_found)
-		pr_err("ucma_removal_event_handler: warning: connect request event wasn't found\n");
-}
-
-static int ucma_event_handler(struct rdma_cm_id *cm_id,
-			      struct rdma_cm_event *event)
-{
-	struct ucma_event *uevent;
-	struct ucma_context *ctx = cm_id->context;
-	int ret = 0;
-
-	uevent = kzalloc(sizeof(*uevent), GFP_KERNEL);
-	if (!uevent)
-		return event->event == RDMA_CM_EVENT_CONNECT_REQUEST;
-
-	mutex_lock(&ctx->file->mut);
-	uevent->cm_id = cm_id;
-	ucma_set_event_context(ctx, event, uevent);
 	uevent->resp.event = event->event;
 	uevent->resp.status = event->status;
-	if (cm_id->qp_type == IB_QPT_UD)
+	if (ctx->cm_id->qp_type == IB_QPT_UD)
-		ucma_copy_ud_event(cm_id->device, &uevent->resp.param.ud,
+		ucma_copy_ud_event(ctx->cm_id->device, &uevent->resp.param.ud,
 				   &event->param.ud);
 	else
 		ucma_copy_conn_event(&uevent->resp.param.conn,
@@ -363,46 +288,84 @@ static int ucma_event_handler(struct rdma_cm_id *cm_id,
 
 	uevent->resp.ece.vendor_id = event->ece.vendor_id;
 	uevent->resp.ece.attr_mod = event->ece.attr_mod;
+	return uevent;
+}
 
-	if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST) {
+static int ucma_connect_event_handler(struct rdma_cm_id *cm_id,
-		if (!ctx->backlog) {
+				      struct rdma_cm_event *event)
-			ret = -ENOMEM;
+{
-			kfree(uevent);
+	struct ucma_context *listen_ctx = cm_id->context;
-			goto out;
+	struct ucma_context *ctx;
-		}
+	struct ucma_event *uevent;
-		ctx->backlog--;
-	} else if (!ctx->uid || ctx->cm_id != cm_id) {
-		/*
-		 * We ignore events for new connections until userspace has set
-		 * their context.  This can only happen if an error occurs on a
-		 * new connection before the user accepts it.  This is okay,
-		 * since the accept will just fail later. However, we do need
-		 * to release the underlying HW resources in case of a device
-		 * removal event.
-		 */
-		if (event->event == RDMA_CM_EVENT_DEVICE_REMOVAL)
-			ucma_removal_event_handler(cm_id);
 
-		kfree(uevent);
+	if (!atomic_add_unless(&listen_ctx->backlog, -1, 0))
-		goto out;
+		return -ENOMEM;
+	ctx = ucma_alloc_ctx(listen_ctx->file);
+	if (!ctx)
+		goto err_backlog;
+	ctx->cm_id = cm_id;
+
+	uevent = ucma_create_uevent(listen_ctx, event);
+	if (!uevent)
+		goto err_alloc;
+	uevent->conn_req_ctx = ctx;
+	uevent->resp.id = ctx->id;
+
+	ctx->cm_id->context = ctx;
+
+	mutex_lock(&ctx->file->mut);
+	ucma_finish_ctx(ctx);
+	list_add_tail(&uevent->list, &ctx->file->event_list);
+	mutex_unlock(&ctx->file->mut);
+	wake_up_interruptible(&ctx->file->poll_wait);
+	return 0;
+
+err_alloc:
+	xa_erase(&ctx_table, ctx->id);
+	kfree(ctx);
+err_backlog:
+	atomic_inc(&listen_ctx->backlog);
+	/* Returning error causes the new ID to be destroyed */
+	return -ENOMEM;
+}
+
+static int ucma_event_handler(struct rdma_cm_id *cm_id,
+			      struct rdma_cm_event *event)
+{
+	struct ucma_event *uevent;
+	struct ucma_context *ctx = cm_id->context;
+
+	if (event->event == RDMA_CM_EVENT_CONNECT_REQUEST)
+		return ucma_connect_event_handler(cm_id, event);
+
+	/*
+	 * We ignore events for new connections until userspace has set their
+	 * context.  This can only happen if an error occurs on a new connection
+	 * before the user accepts it.  This is okay, since the accept will just
+	 * fail later. However, we do need to release the underlying HW
+	 * resources in case of a device removal event.
+	 */
+	if (ctx->uid) {
+		uevent = ucma_create_uevent(ctx, event);
+		if (!uevent)
+			return 0;
+
+		mutex_lock(&ctx->file->mut);
+		list_add_tail(&uevent->list, &ctx->file->event_list);
+		mutex_unlock(&ctx->file->mut);
+		wake_up_interruptible(&ctx->file->poll_wait);
 	}
 
-	list_add_tail(&uevent->list, &ctx->file->event_list);
+	if (event->event == RDMA_CM_EVENT_DEVICE_REMOVAL && !ctx->destroying)
-	wake_up_interruptible(&ctx->file->poll_wait);
+		queue_work(system_unbound_wq, &ctx->close_work);
-	if (event->event == RDMA_CM_EVENT_DEVICE_REMOVAL)
+	return 0;
-		ucma_removal_event_handler(cm_id);
-out:
-	mutex_unlock(&ctx->file->mut);
-	return ret;
 }
 
 static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
 			      int in_len, int out_len)
 {
-	struct ucma_context *ctx;
 	struct rdma_ucm_get_event cmd;
 	struct ucma_event *uevent;
-	int ret = 0;
 
 	/*
 	 * Old 32 bit user space does not send the 4 byte padding in the
@@ -429,35 +392,25 @@ static ssize_t ucma_get_event(struct ucma_file *file, const char __user *inbuf,
 		mutex_lock(&file->mut);
 	}
 
-	uevent = list_entry(file->event_list.next, struct ucma_event, list);
+	uevent = list_first_entry(&file->event_list, struct ucma_event, list);
-
-	if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST) {
-		ctx = ucma_alloc_ctx(file);
-		if (!ctx) {
-			ret = -ENOMEM;
-			goto done;
-		}
-		uevent->ctx->backlog++;
-		ctx->cm_id = uevent->cm_id;
-		ctx->cm_id->context = ctx;
-		uevent->resp.id = ctx->id;
-	}
 
 	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &uevent->resp,
 			 min_t(size_t, out_len, sizeof(uevent->resp)))) {
-		ret = -EFAULT;
+		mutex_unlock(&file->mut);
-		goto done;
+		return -EFAULT;
 	}
 
 	list_del(&uevent->list);
 	uevent->ctx->events_reported++;
 	if (uevent->mc)
 		uevent->mc->events_reported++;
-	kfree(uevent);
+	if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST)
-done:
+		atomic_inc(&uevent->ctx->backlog);
 	mutex_unlock(&file->mut);
-	return ret;
+
+	kfree(uevent);
+	return 0;
 }
 
 static int ucma_get_qp_type(struct rdma_ucm_create_id *cmd, enum ib_qp_type *qp_type)
@@ -498,58 +451,60 @@ static ssize_t ucma_create_id(struct ucma_file *file, const char __user *inbuf,
 	if (ret)
 		return ret;
 
-	mutex_lock(&file->mut);
 	ctx = ucma_alloc_ctx(file);
-	mutex_unlock(&file->mut);
 	if (!ctx)
 		return -ENOMEM;
 
 	ctx->uid = cmd.uid;
-	cm_id = __rdma_create_id(current->nsproxy->net_ns,
+	cm_id = rdma_create_user_id(ucma_event_handler, ctx, cmd.ps, qp_type);
-				 ucma_event_handler, ctx, cmd.ps, qp_type, NULL);
 	if (IS_ERR(cm_id)) {
 		ret = PTR_ERR(cm_id);
 		goto err1;
 	}
+	ctx->cm_id = cm_id;
 
 	resp.id = ctx->id;
 	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp))) {
-		ret = -EFAULT;
+		xa_erase(&ctx_table, ctx->id);
-		goto err2;
+		__destroy_id(ctx);
+		return -EFAULT;
 	}
 
-	ctx->cm_id = cm_id;
+	mutex_lock(&file->mut);
+	ucma_finish_ctx(ctx);
+	mutex_unlock(&file->mut);
 	return 0;
 
-err2:
-	rdma_destroy_id(cm_id);
 err1:
 	xa_erase(&ctx_table, ctx->id);
-	mutex_lock(&file->mut);
-	list_del(&ctx->list);
-	mutex_unlock(&file->mut);
 	kfree(ctx);
 	return ret;
 }
 
 static void ucma_cleanup_multicast(struct ucma_context *ctx)
 {
-	struct ucma_multicast *mc, *tmp;
+	struct ucma_multicast *mc;
+	unsigned long index;
 
-	mutex_lock(&ctx->file->mut);
+	xa_for_each(&multicast_table, index, mc) {
-	list_for_each_entry_safe(mc, tmp, &ctx->mc_list, list) {
+		if (mc->ctx != ctx)
-		list_del(&mc->list);
+			continue;
-		xa_erase(&multicast_table, mc->id);
+		/*
+		 * At this point mc->ctx->ref is 0 so the mc cannot leave the
+		 * lock on the reader and this is enough serialization
+		 */
+		xa_erase(&multicast_table, index);
 		kfree(mc);
 	}
-	mutex_unlock(&ctx->file->mut);
 }
 
 static void ucma_cleanup_mc_events(struct ucma_multicast *mc)
 {
 	struct ucma_event *uevent, *tmp;
 
+	rdma_lock_handler(mc->ctx->cm_id);
+	mutex_lock(&mc->ctx->file->mut);
 	list_for_each_entry_safe(uevent, tmp, &mc->ctx->file->event_list, list) {
 		if (uevent->mc != mc)
 			continue;
@@ -557,6 +512,8 @@ static void ucma_cleanup_mc_events(struct ucma_multicast *mc)
 		list_del(&uevent->list);
 		kfree(uevent);
 	}
+	mutex_unlock(&mc->ctx->file->mut);
+	rdma_unlock_handler(mc->ctx->cm_id);
 }
 
 /*
@@ -564,10 +521,6 @@ static void ucma_cleanup_mc_events(struct ucma_multicast *mc)
  * this point, no new events will be reported from the hardware. However, we
  * still need to cleanup the UCMA context for this ID. Specifically, there
  * might be events that have not yet been consumed by the user space software.
- * These might include pending connect requests which we have not completed
- * processing.  We cannot call rdma_destroy_id while holding the lock of the
- * context (file->mut), as it might cause a deadlock. We therefore extract all
- * relevant events from the context pending events list while holding the
  * mutex. After that we release them as needed.
  */
 static int ucma_free_ctx(struct ucma_context *ctx)
@@ -576,31 +529,57 @@ static int ucma_free_ctx(struct ucma_context *ctx)
 	struct ucma_event *uevent, *tmp;
 	LIST_HEAD(list);
 
-
 	ucma_cleanup_multicast(ctx);
 
 	/* Cleanup events not yet reported to the user. */
 	mutex_lock(&ctx->file->mut);
 	list_for_each_entry_safe(uevent, tmp, &ctx->file->event_list, list) {
-		if (uevent->ctx == ctx)
+		if (uevent->ctx == ctx || uevent->conn_req_ctx == ctx)
 			list_move_tail(&uevent->list, &list);
 	}
 	list_del(&ctx->list);
+	events_reported = ctx->events_reported;
 	mutex_unlock(&ctx->file->mut);
 
+	/*
+	 * If this was a listening ID then any connections spawned from it
+	 * that have not been delivered to userspace are cleaned up too.
+	 * Must be done outside any locks.
+	 */
 	list_for_each_entry_safe(uevent, tmp, &list, list) {
 		list_del(&uevent->list);
-		if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST)
+		if (uevent->resp.event == RDMA_CM_EVENT_CONNECT_REQUEST &&
-			rdma_destroy_id(uevent->cm_id);
+		    uevent->conn_req_ctx != ctx)
+			__destroy_id(uevent->conn_req_ctx);
 		kfree(uevent);
 	}
 
-	events_reported = ctx->events_reported;
 	mutex_destroy(&ctx->mutex);
 	kfree(ctx);
 	return events_reported;
 }
 
+static int __destroy_id(struct ucma_context *ctx)
+{
+	/*
+	 * If the refcount is already 0 then ucma_close_id() has already
+	 * destroyed the cm_id, otherwise holding the refcount keeps cm_id
+	 * valid. Prevent queue_work() from being called.
+	 */
+	if (refcount_inc_not_zero(&ctx->ref)) {
+		rdma_lock_handler(ctx->cm_id);
+		ctx->destroying = 1;
+		rdma_unlock_handler(ctx->cm_id);
+		ucma_put_ctx(ctx);
+	}
+
+	cancel_work_sync(&ctx->close_work);
+	/* At this point it's guaranteed that there is no inflight closing task */
+	if (ctx->cm_id)
+		ucma_close_id(&ctx->close_work);
+	return ucma_free_ctx(ctx);
+}
+
 static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf,
 			       int in_len, int out_len)
 {
@@ -624,24 +603,7 @@ static ssize_t ucma_destroy_id(struct ucma_file *file, const char __user *inbuf,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
-	mutex_lock(&ctx->file->mut);
+	resp.events_reported = __destroy_id(ctx);
-	ctx->destroying = 1;
-	mutex_unlock(&ctx->file->mut);
-
-	flush_workqueue(ctx->file->close_wq);
-	/* At this point it's guaranteed that there is no inflight
-	 * closing task */
-	xa_lock(&ctx_table);
-	if (!ctx->closing) {
-		xa_unlock(&ctx_table);
-		ucma_put_ctx(ctx);
-		wait_for_completion(&ctx->comp);
-		rdma_destroy_id(ctx->cm_id);
-	} else {
-		xa_unlock(&ctx_table);
-	}
-
-	resp.events_reported = ucma_free_ctx(ctx);
 	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
@@ -1124,10 +1086,12 @@ static ssize_t ucma_listen(struct ucma_file *file, const char __user *inbuf,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
-	ctx->backlog = cmd.backlog > 0 && cmd.backlog < max_backlog ?
+	if (cmd.backlog <= 0 || cmd.backlog > max_backlog)
-		       cmd.backlog : max_backlog;
+		cmd.backlog = max_backlog;
+	atomic_set(&ctx->backlog, cmd.backlog);
+
 	mutex_lock(&ctx->mutex);
-	ret = rdma_listen(ctx->cm_id, ctx->backlog);
+	ret = rdma_listen(ctx->cm_id, cmd.backlog);
 	mutex_unlock(&ctx->mutex);
 	ucma_put_ctx(ctx);
 	return ret;
@@ -1160,16 +1124,20 @@ static ssize_t ucma_accept(struct ucma_file *file, const char __user *inbuf,
 
 	if (cmd.conn_param.valid) {
 		ucma_copy_conn_param(ctx->cm_id, &conn_param, &cmd.conn_param);
-		mutex_lock(&file->mut);
 		mutex_lock(&ctx->mutex);
-		ret = __rdma_accept_ece(ctx->cm_id, &conn_param, NULL, &ece);
+		rdma_lock_handler(ctx->cm_id);
-		mutex_unlock(&ctx->mutex);
+		ret = rdma_accept_ece(ctx->cm_id, &conn_param, &ece);
-		if (!ret)
+		if (!ret) {
+			/* The uid must be set atomically with the handler */
 			ctx->uid = cmd.uid;
-		mutex_unlock(&file->mut);
+		}
+		rdma_unlock_handler(ctx->cm_id);
+		mutex_unlock(&ctx->mutex);
 	} else {
 		mutex_lock(&ctx->mutex);
-		ret = __rdma_accept_ece(ctx->cm_id, NULL, NULL, &ece);
+		rdma_lock_handler(ctx->cm_id);
+		ret = rdma_accept_ece(ctx->cm_id, NULL, &ece);
+		rdma_unlock_handler(ctx->cm_id);
 		mutex_unlock(&ctx->mutex);
 	}
 	ucma_put_ctx(ctx);
@@ -1482,44 +1450,52 @@ static ssize_t ucma_process_join(struct ucma_file *file,
 	if (IS_ERR(ctx))
 		return PTR_ERR(ctx);
 
-	mutex_lock(&file->mut);
+	mc = kzalloc(sizeof(*mc), GFP_KERNEL);
-	mc = ucma_alloc_multicast(ctx);
 	if (!mc) {
 		ret = -ENOMEM;
-		goto err1;
+		goto err_put_ctx;
 	}
+
+	mc->ctx = ctx;
 	mc->join_state = join_state;
 	mc->uid = cmd->uid;
 	memcpy(&mc->addr, addr, cmd->addr_size);
+
+	if (xa_alloc(&multicast_table, &mc->id, NULL, xa_limit_32b,
+		     GFP_KERNEL)) {
+		ret = -ENOMEM;
+		goto err_free_mc;
+	}
+
 	mutex_lock(&ctx->mutex);
 	ret = rdma_join_multicast(ctx->cm_id, (struct sockaddr *)&mc->addr,
 				  join_state, mc);
 	mutex_unlock(&ctx->mutex);
 	if (ret)
-		goto err2;
+		goto err_xa_erase;
 
 	resp.id = mc->id;
 	if (copy_to_user(u64_to_user_ptr(cmd->response),
 			 &resp, sizeof(resp))) {
 		ret = -EFAULT;
-		goto err3;
+		goto err_leave_multicast;
 	}
 
 	xa_store(&multicast_table, mc->id, mc, 0);
 
-	mutex_unlock(&file->mut);
 	ucma_put_ctx(ctx);
 	return 0;
 
-err3:
+err_leave_multicast:
+	mutex_lock(&ctx->mutex);
 	rdma_leave_multicast(ctx->cm_id, (struct sockaddr *) &mc->addr);
+	mutex_unlock(&ctx->mutex);
 	ucma_cleanup_mc_events(mc);
-err2:
+err_xa_erase:
 	xa_erase(&multicast_table, mc->id);
-	list_del(&mc->list);
+err_free_mc:
 	kfree(mc);
-err1:
+err_put_ctx:
-	mutex_unlock(&file->mut);
 	ucma_put_ctx(ctx);
 	return ret;
 }
@@ -1581,7 +1557,7 @@ static ssize_t ucma_leave_multicast(struct ucma_file *file,
 	mc = xa_load(&multicast_table, cmd.id);
 	if (!mc)
 		mc = ERR_PTR(-ENOENT);
-	else if (mc->ctx->file != file)
+	else if (READ_ONCE(mc->ctx->file) != file)
 		mc = ERR_PTR(-EINVAL);
 	else if (!refcount_inc_not_zero(&mc->ctx->ref))
 		mc = ERR_PTR(-ENXIO);
@@ -1598,10 +1574,7 @@ static ssize_t ucma_leave_multicast(struct ucma_file *file,
 	rdma_leave_multicast(mc->ctx->cm_id, (struct sockaddr *) &mc->addr);
 	mutex_unlock(&mc->ctx->mutex);
 
-	mutex_lock(&mc->ctx->file->mut);
 	ucma_cleanup_mc_events(mc);
-	list_del(&mc->list);
-	mutex_unlock(&mc->ctx->file->mut);
 
 	ucma_put_ctx(mc->ctx);
 	resp.events_reported = mc->events_reported;
@@ -1614,45 +1587,15 @@ out:
 	return ret;
 }
 
-static void ucma_lock_files(struct ucma_file *file1, struct ucma_file *file2)
-{
-	/* Acquire mutex's based on pointer comparison to prevent deadlock. */
-	if (file1 < file2) {
-		mutex_lock(&file1->mut);
-		mutex_lock_nested(&file2->mut, SINGLE_DEPTH_NESTING);
-	} else {
-		mutex_lock(&file2->mut);
-		mutex_lock_nested(&file1->mut, SINGLE_DEPTH_NESTING);
-	}
-}
-
-static void ucma_unlock_files(struct ucma_file *file1, struct ucma_file *file2)
-{
-	if (file1 < file2) {
-		mutex_unlock(&file2->mut);
-		mutex_unlock(&file1->mut);
-	} else {
-		mutex_unlock(&file1->mut);
-		mutex_unlock(&file2->mut);
-	}
-}
-
-static void ucma_move_events(struct ucma_context *ctx, struct ucma_file *file)
-{
-	struct ucma_event *uevent, *tmp;
-
-	list_for_each_entry_safe(uevent, tmp, &ctx->file->event_list, list)
-		if (uevent->ctx == ctx)
-			list_move_tail(&uevent->list, &file->event_list);
-}
-
 static ssize_t ucma_migrate_id(struct ucma_file *new_file,
 			       const char __user *inbuf,
 			       int in_len, int out_len)
 {
 	struct rdma_ucm_migrate_id cmd;
 	struct rdma_ucm_migrate_resp resp;
+	struct ucma_event *uevent, *tmp;
 	struct ucma_context *ctx;
+	LIST_HEAD(event_list);
 	struct fd f;
 	struct ucma_file *cur_file;
 	int ret = 0;
@@ -1668,40 +1611,53 @@ static ssize_t ucma_migrate_id(struct ucma_file *new_file,
 		ret = -EINVAL;
 		goto file_put;
 	}
+	cur_file = f.file->private_data;
 
 	/* Validate current fd and prevent destruction of id. */
-	ctx = ucma_get_ctx(f.file->private_data, cmd.id);
+	ctx = ucma_get_ctx(cur_file, cmd.id);
 	if (IS_ERR(ctx)) {
 		ret = PTR_ERR(ctx);
 		goto file_put;
 	}
 
-	cur_file = ctx->file;
+	rdma_lock_handler(ctx->cm_id);
-	if (cur_file == new_file) {
-		resp.events_reported = ctx->events_reported;
-		goto response;
-	}
-
 	/*
-	 * Migrate events between fd's, maintaining order, and avoiding new
+	 * ctx->file can only be changed under the handler & xa_lock. xa_load()
-	 * events being added before existing events.
+	 * must be checked again to ensure the ctx hasn't begun destruction
+	 * since the ucma_get_ctx().
 	 */
-	ucma_lock_files(cur_file, new_file);
 	xa_lock(&ctx_table);
-
+	if (_ucma_find_context(cmd.id, cur_file) != ctx) {
-	list_move_tail(&ctx->list, &new_file->ctx_list);
+		xa_unlock(&ctx_table);
-	ucma_move_events(ctx, new_file);
+		ret = -ENOENT;
+		goto err_unlock;
+	}
 	ctx->file = new_file;
-	resp.events_reported = ctx->events_reported;
-
 	xa_unlock(&ctx_table);
-	ucma_unlock_files(cur_file, new_file);
 
-response:
+	mutex_lock(&cur_file->mut);
+	list_del(&ctx->list);
+	/*
+	 * At this point lock_handler() prevents addition of new uevents for
+	 * this ctx.
+	 */
+	list_for_each_entry_safe(uevent, tmp, &cur_file->event_list, list)
+		if (uevent->ctx == ctx)
+			list_move_tail(&uevent->list, &event_list);
+	resp.events_reported = ctx->events_reported;
+	mutex_unlock(&cur_file->mut);
+
+	mutex_lock(&new_file->mut);
+	list_add_tail(&ctx->list, &new_file->ctx_list);
+	list_splice_tail(&event_list, &new_file->event_list);
+	mutex_unlock(&new_file->mut);
+
 	if (copy_to_user(u64_to_user_ptr(cmd.response),
 			 &resp, sizeof(resp)))
 		ret = -EFAULT;
 
+err_unlock:
+	rdma_unlock_handler(ctx->cm_id);
 	ucma_put_ctx(ctx);
 file_put:
 	fdput(f);
@@ -1801,13 +1757,6 @@ static int ucma_open(struct inode *inode, struct file *filp)
 	if (!file)
 		return -ENOMEM;
 
-	file->close_wq = alloc_ordered_workqueue("ucma_close_id",
-						 WQ_MEM_RECLAIM);
-	if (!file->close_wq) {
-		kfree(file);
-		return -ENOMEM;
-	}
-
 	INIT_LIST_HEAD(&file->event_list);
 	INIT_LIST_HEAD(&file->ctx_list);
 	init_waitqueue_head(&file->poll_wait);
@@ -1822,37 +1771,22 @@ static int ucma_open(struct inode *inode, struct file *filp)
 static int ucma_close(struct inode *inode, struct file *filp)
 {
 	struct ucma_file *file = filp->private_data;
-	struct ucma_context *ctx, *tmp;
 
-	mutex_lock(&file->mut);
+	/*
-	list_for_each_entry_safe(ctx, tmp, &file->ctx_list, list) {
+	 * All paths that touch ctx_list or ctx_list starting from write() are
-		ctx->destroying = 1;
+	 * prevented by this being a FD release function. The list_add_tail() in
-		mutex_unlock(&file->mut);
+	 * ucma_connect_event_handler() can run concurrently, however it only
+	 * adds to the list *after* a listening ID. By only reading the first of
+	 * the list, and relying on __destroy_id() to block
+	 * ucma_connect_event_handler(), no additional locking is needed.
+	 */
+	while (!list_empty(&file->ctx_list)) {
+		struct ucma_context *ctx = list_first_entry(
+			&file->ctx_list, struct ucma_context, list);
 
 		xa_erase(&ctx_table, ctx->id);
-		flush_workqueue(file->close_wq);
+		__destroy_id(ctx);
-		/* At that step once ctx was marked as destroying and workqueue
-		 * was flushed we are safe from any inflights handlers that
-		 * might put other closing task.
-		 */
-		xa_lock(&ctx_table);
-		if (!ctx->closing) {
-			xa_unlock(&ctx_table);
-			ucma_put_ctx(ctx);
-			wait_for_completion(&ctx->comp);
-			/* rdma_destroy_id ensures that no event handlers are
-			 * inflight for that id before releasing it.
-			 */
-			rdma_destroy_id(ctx->cm_id);
-		} else {
-			xa_unlock(&ctx_table);
-		}
-
-		ucma_free_ctx(ctx);
-		mutex_lock(&file->mut);
 	}
-	mutex_unlock(&file->mut);
-	destroy_workqueue(file->close_wq);
 	kfree(file);
 	return 0;
 }

drivers/infiniband/core/umem.c

@@ -39,6 +39,7 @@
 #include <linux/export.h>
 #include <linux/slab.h>
 #include <linux/pagemap.h>
+#include <linux/count_zeros.h>
 #include <rdma/ib_umem_odp.h>
 
 #include "uverbs.h"
@@ -60,73 +61,6 @@ static void __ib_umem_release(struct ib_device *dev, struct ib_umem *umem, int d
 	sg_free_table(&umem->sg_head);
 }
 
-/* ib_umem_add_sg_table - Add N contiguous pages to scatter table
- *
- * sg: current scatterlist entry
- * page_list: array of npage struct page pointers
- * npages: number of pages in page_list
- * max_seg_sz: maximum segment size in bytes
- * nents: [out] number of entries in the scatterlist
- *
- * Return new end of scatterlist
- */
-static struct scatterlist *ib_umem_add_sg_table(struct scatterlist *sg,
-						struct page **page_list,
-						unsigned long npages,
-						unsigned int max_seg_sz,
-						int *nents)
-{
-	unsigned long first_pfn;
-	unsigned long i = 0;
-	bool update_cur_sg = false;
-	bool first = !sg_page(sg);
-
-	/* Check if new page_list is contiguous with end of previous page_list.
-	 * sg->length here is a multiple of PAGE_SIZE and sg->offset is 0.
-	 */
-	if (!first && (page_to_pfn(sg_page(sg)) + (sg->length >> PAGE_SHIFT) ==
-		       page_to_pfn(page_list[0])))
-		update_cur_sg = true;
-
-	while (i != npages) {
-		unsigned long len;
-		struct page *first_page = page_list[i];
-
-		first_pfn = page_to_pfn(first_page);
-
-		/* Compute the number of contiguous pages we have starting
-		 * at i
-		 */
-		for (len = 0; i != npages &&
-			      first_pfn + len == page_to_pfn(page_list[i]) &&
-			      len < (max_seg_sz >> PAGE_SHIFT);
-		     len++)
-			i++;
-
-		/* Squash N contiguous pages from page_list into current sge */
-		if (update_cur_sg) {
-			if ((max_seg_sz - sg->length) >= (len << PAGE_SHIFT)) {
-				sg_set_page(sg, sg_page(sg),
-					    sg->length + (len << PAGE_SHIFT),
-					    0);
-				update_cur_sg = false;
-				continue;
-			}
-			update_cur_sg = false;
-		}
-
-		/* Squash N contiguous pages into next sge or first sge */
-		if (!first)
-			sg = sg_next(sg);
-
-		(*nents)++;
-		sg_set_page(sg, first_page, len << PAGE_SHIFT, 0);
-		first = false;
-	}
-
-	return sg;
-}
-
 /**
  * ib_umem_find_best_pgsz - Find best HW page size to use for this MR
  *
@@ -146,18 +80,28 @@ unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
 				     unsigned long virt)
 {
 	struct scatterlist *sg;
-	unsigned int best_pg_bit;
 	unsigned long va, pgoff;
 	dma_addr_t mask;
 	int i;
 
+	/* rdma_for_each_block() has a bug if the page size is smaller than the
+	 * page size used to build the umem. For now prevent smaller page sizes
+	 * from being returned.
+	 */
+	pgsz_bitmap &= GENMASK(BITS_PER_LONG - 1, PAGE_SHIFT);
+
 	/* At minimum, drivers must support PAGE_SIZE or smaller */
 	if (WARN_ON(!(pgsz_bitmap & GENMASK(PAGE_SHIFT, 0))))
 		return 0;
 
-	va = virt;
+	umem->iova = va = virt;
-	/* max page size not to exceed MR length */
+	/* The best result is the smallest page size that results in the minimum
-	mask = roundup_pow_of_two(umem->length);
+	 * number of required pages. Compute the largest page size that could
+	 * work based on VA address bits that don't change.
+	 */
+	mask = pgsz_bitmap &
+	       GENMASK(BITS_PER_LONG - 1,
+		       bits_per((umem->length - 1 + virt) ^ virt));
 	/* offset into first SGL */
 	pgoff = umem->address & ~PAGE_MASK;
 
@@ -175,9 +119,14 @@ unsigned long ib_umem_find_best_pgsz(struct ib_umem *umem,
 			mask |= va;
 		pgoff = 0;
 	}
-	best_pg_bit = rdma_find_pg_bit(mask, pgsz_bitmap);
 
-	return BIT_ULL(best_pg_bit);
+	/* The mask accumulates 1's in each position where the VA and physical
+	 * address differ, thus the length of trailing 0 is the largest page
+	 * size that can pass the VA through to the physical.
+	 */
+	if (mask)
+		pgsz_bitmap &= GENMASK(count_trailing_zeros(mask), 0);
+	return rounddown_pow_of_two(pgsz_bitmap);
 }
 EXPORT_SYMBOL(ib_umem_find_best_pgsz);
 
@@ -201,7 +150,7 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
 	struct mm_struct *mm;
 	unsigned long npages;
 	int ret;
-	struct scatterlist *sg;
+	struct scatterlist *sg = NULL;
 	unsigned int gup_flags = FOLL_WRITE;
 
 	/*
@@ -224,6 +173,11 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
 	umem->ibdev      = device;
 	umem->length     = size;
 	umem->address    = addr;
+	/*
+	 * Drivers should call ib_umem_find_best_pgsz() to set the iova
+	 * correctly.
+	 */
+	umem->iova = addr;
 	umem->writable   = ib_access_writable(access);
 	umem->owning_mm = mm = current->mm;
 	mmgrab(mm);
@@ -251,15 +205,9 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
 
 	cur_base = addr & PAGE_MASK;
 
-	ret = sg_alloc_table(&umem->sg_head, npages, GFP_KERNEL);
-	if (ret)
-		goto vma;
-
 	if (!umem->writable)
 		gup_flags |= FOLL_FORCE;
 
-	sg = umem->sg_head.sgl;
-
 	while (npages) {
 		cond_resched();
 		ret = pin_user_pages_fast(cur_base,
@@ -271,15 +219,19 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
 			goto umem_release;
 
 		cur_base += ret * PAGE_SIZE;
-		npages   -= ret;
+		npages -= ret;
-
+		sg = __sg_alloc_table_from_pages(
-		sg = ib_umem_add_sg_table(sg, page_list, ret,
+			&umem->sg_head, page_list, ret, 0, ret << PAGE_SHIFT,
-			dma_get_max_seg_size(device->dma_device),
+			dma_get_max_seg_size(device->dma_device), sg, npages,
-			&umem->sg_nents);
+			GFP_KERNEL);
+		umem->sg_nents = umem->sg_head.nents;
+		if (IS_ERR(sg)) {
+			unpin_user_pages_dirty_lock(page_list, ret, 0);
+			ret = PTR_ERR(sg);
+			goto umem_release;
+		}
 	}
 
-	sg_mark_end(sg);
-
 	if (access & IB_ACCESS_RELAXED_ORDERING)
 		dma_attr |= DMA_ATTR_WEAK_ORDERING;
 
@@ -297,7 +249,6 @@ struct ib_umem *ib_umem_get(struct ib_device *device, unsigned long addr,
 
 umem_release:
 	__ib_umem_release(device, umem, 0);
-vma:
 	atomic64_sub(ib_umem_num_pages(umem), &mm->pinned_vm);
 out:
 	free_page((unsigned long) page_list);
@@ -329,18 +280,6 @@ void ib_umem_release(struct ib_umem *umem)
 }
 EXPORT_SYMBOL(ib_umem_release);
 
-int ib_umem_page_count(struct ib_umem *umem)
-{
-	int i, n = 0;
-	struct scatterlist *sg;
-
-	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, i)
-		n += sg_dma_len(sg) >> PAGE_SHIFT;
-
-	return n;
-}
-EXPORT_SYMBOL(ib_umem_page_count);
-
 /*
  * Copy from the given ib_umem's pages to the given buffer.
  *

drivers/infiniband/core/umem_odp.c

@@ -40,6 +40,7 @@
 #include <linux/vmalloc.h>
 #include <linux/hugetlb.h>
 #include <linux/interval_tree.h>
+#include <linux/hmm.h>
 #include <linux/pagemap.h>
 
 #include <rdma/ib_verbs.h>
@@ -60,7 +61,7 @@ static inline int ib_init_umem_odp(struct ib_umem_odp *umem_odp,
 		size_t page_size = 1UL << umem_odp->page_shift;
 		unsigned long start;
 		unsigned long end;
-		size_t pages;
+		size_t ndmas, npfns;
 
 		start = ALIGN_DOWN(umem_odp->umem.address, page_size);
 		if (check_add_overflow(umem_odp->umem.address,
@@ -71,20 +72,21 @@ static inline int ib_init_umem_odp(struct ib_umem_odp *umem_odp,
 		if (unlikely(end < page_size))
 			return -EOVERFLOW;
 
-		pages = (end - start) >> umem_odp->page_shift;
+		ndmas = (end - start) >> umem_odp->page_shift;
-		if (!pages)
+		if (!ndmas)
 			return -EINVAL;
 
-		umem_odp->page_list = kvcalloc(
+		npfns = (end - start) >> PAGE_SHIFT;
-			pages, sizeof(*umem_odp->page_list), GFP_KERNEL);
+		umem_odp->pfn_list = kvcalloc(
-		if (!umem_odp->page_list)
+			npfns, sizeof(*umem_odp->pfn_list), GFP_KERNEL);
+		if (!umem_odp->pfn_list)
 			return -ENOMEM;
 
 		umem_odp->dma_list = kvcalloc(
-			pages, sizeof(*umem_odp->dma_list), GFP_KERNEL);
+			ndmas, sizeof(*umem_odp->dma_list), GFP_KERNEL);
 		if (!umem_odp->dma_list) {
 			ret = -ENOMEM;
-			goto out_page_list;
+			goto out_pfn_list;
 		}
 
 		ret = mmu_interval_notifier_insert(&umem_odp->notifier,
@@ -98,8 +100,8 @@ static inline int ib_init_umem_odp(struct ib_umem_odp *umem_odp,
 
 out_dma_list:
 	kvfree(umem_odp->dma_list);
-out_page_list:
+out_pfn_list:
-	kvfree(umem_odp->page_list);
+	kvfree(umem_odp->pfn_list);
 	return ret;
 }
 
@@ -276,7 +278,7 @@ void ib_umem_odp_release(struct ib_umem_odp *umem_odp)
 		mutex_unlock(&umem_odp->umem_mutex);
 		mmu_interval_notifier_remove(&umem_odp->notifier);
 		kvfree(umem_odp->dma_list);
-		kvfree(umem_odp->page_list);
+		kvfree(umem_odp->pfn_list);
 	}
 	put_pid(umem_odp->tgid);
 	kfree(umem_odp);
@@ -287,87 +289,56 @@ EXPORT_SYMBOL(ib_umem_odp_release);
  * Map for DMA and insert a single page into the on-demand paging page tables.
  *
  * @umem: the umem to insert the page to.
- * @page_index: index in the umem to add the page to.
+ * @dma_index: index in the umem to add the dma to.
  * @page: the page struct to map and add.
  * @access_mask: access permissions needed for this page.
  * @current_seq: sequence number for synchronization with invalidations.
  *               the sequence number is taken from
  *               umem_odp->notifiers_seq.
  *
- * The function returns -EFAULT if the DMA mapping operation fails. It returns
+ * The function returns -EFAULT if the DMA mapping operation fails.
- * -EAGAIN if a concurrent invalidation prevents us from updating the page.
  *
- * The page is released via put_page even if the operation failed. For on-demand
- * pinning, the page is released whenever it isn't stored in the umem.
  */
 static int ib_umem_odp_map_dma_single_page(
 		struct ib_umem_odp *umem_odp,
-		unsigned int page_index,
+		unsigned int dma_index,
 		struct page *page,
-		u64 access_mask,
+		u64 access_mask)
-		unsigned long current_seq)
 {
 	struct ib_device *dev = umem_odp->umem.ibdev;
-	dma_addr_t dma_addr;
+	dma_addr_t *dma_addr = &umem_odp->dma_list[dma_index];
-	int ret = 0;
 
-	if (mmu_interval_check_retry(&umem_odp->notifier, current_seq)) {
+	if (*dma_addr) {
-		ret = -EAGAIN;
-		goto out;
-	}
-	if (!(umem_odp->dma_list[page_index])) {
-		dma_addr =
-			ib_dma_map_page(dev, page, 0, BIT(umem_odp->page_shift),
-					DMA_BIDIRECTIONAL);
-		if (ib_dma_mapping_error(dev, dma_addr)) {
-			ret = -EFAULT;
-			goto out;
-		}
-		umem_odp->dma_list[page_index] = dma_addr | access_mask;
-		umem_odp->page_list[page_index] = page;
-		umem_odp->npages++;
-	} else if (umem_odp->page_list[page_index] == page) {
-		umem_odp->dma_list[page_index] |= access_mask;
-	} else {
 		/*
-		 * This is a race here where we could have done:
+		 * If the page is already dma mapped it means it went through
-		 *
+		 * a non-invalidating trasition, like read-only to writable.
-		 *         CPU0                             CPU1
+		 * Resync the flags.
-		 *   get_user_pages()
-		 *                                       invalidate()
-		 *                                       page_fault()
-		 *   mutex_lock(umem_mutex)
-		 *    page from GUP != page in ODP
-		 *
-		 * It should be prevented by the retry test above as reading
-		 * the seq number should be reliable under the
-		 * umem_mutex. Thus something is really not working right if
-		 * things get here.
 		 */
-		WARN(true,
+		*dma_addr = (*dma_addr & ODP_DMA_ADDR_MASK) | access_mask;
-		     "Got different pages in IB device and from get_user_pages. IB device page: %p, gup page: %p\n",
+		return 0;
-		     umem_odp->page_list[page_index], page);
-		ret = -EAGAIN;
 	}
 
-out:
+	*dma_addr = ib_dma_map_page(dev, page, 0, 1 << umem_odp->page_shift,
-	put_page(page);
+				    DMA_BIDIRECTIONAL);
-	return ret;
+	if (ib_dma_mapping_error(dev, *dma_addr)) {
+		*dma_addr = 0;
+		return -EFAULT;
+	}
+	umem_odp->npages++;
+	*dma_addr |= access_mask;
+	return 0;
 }
 
 /**
- * ib_umem_odp_map_dma_pages - Pin and DMA map userspace memory in an ODP MR.
+ * ib_umem_odp_map_dma_and_lock - DMA map userspace memory in an ODP MR and lock it.
  *
- * Pins the range of pages passed in the argument, and maps them to
+ * Maps the range passed in the argument to DMA addresses.
- * DMA addresses. The DMA addresses of the mapped pages is updated in
+ * The DMA addresses of the mapped pages is updated in umem_odp->dma_list.
- * umem_odp->dma_list.
+ * Upon success the ODP MR will be locked to let caller complete its device
+ * page table update.
  *
  * Returns the number of pages mapped in success, negative error code
  * for failure.
- * An -EAGAIN error code is returned when a concurrent mmu notifier prevents
- * the function from completing its task.
- * An -ENOENT error code indicates that userspace process is being terminated
- * and mm was already destroyed.
  * @umem_odp: the umem to map and pin
  * @user_virt: the address from which we need to map.
  * @bcnt: the minimal number of bytes to pin and map. The mapping might be
@@ -376,21 +347,19 @@ out:
  *        the return value.
  * @access_mask: bit mask of the requested access permissions for the given
  *               range.
- * @current_seq: the MMU notifiers sequance value for synchronization with
+ * @fault: is faulting required for the given range
- *               invalidations. the sequance number is read from
- *               umem_odp->notifiers_seq before calling this function
  */
-int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 user_virt,
+int ib_umem_odp_map_dma_and_lock(struct ib_umem_odp *umem_odp, u64 user_virt,
-			      u64 bcnt, u64 access_mask,
+				 u64 bcnt, u64 access_mask, bool fault)
-			      unsigned long current_seq)
+			__acquires(&umem_odp->umem_mutex)
 {
 	struct task_struct *owning_process  = NULL;
 	struct mm_struct *owning_mm = umem_odp->umem.owning_mm;
-	struct page       **local_page_list = NULL;
+	int pfn_index, dma_index, ret = 0, start_idx;
-	u64 page_mask, off;
+	unsigned int page_shift, hmm_order, pfn_start_idx;
-	int j, k, ret = 0, start_idx, npages = 0;
+	unsigned long num_pfns, current_seq;
-	unsigned int flags = 0, page_shift;
+	struct hmm_range range = {};
-	phys_addr_t p = 0;
+	unsigned long timeout;
 
 	if (access_mask == 0)
 		return -EINVAL;
@@ -399,15 +368,7 @@ int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 user_virt,
 	    user_virt + bcnt > ib_umem_end(umem_odp))
 		return -EFAULT;
 
-	local_page_list = (struct page **)__get_free_page(GFP_KERNEL);
-	if (!local_page_list)
-		return -ENOMEM;
-
 	page_shift = umem_odp->page_shift;
-	page_mask = ~(BIT(page_shift) - 1);
-	off = user_virt & (~page_mask);
-	user_virt = user_virt & page_mask;
-	bcnt += off; /* Charge for the first page offset as well. */
 
 	/*
 	 * owning_process is allowed to be NULL, this means somehow the mm is
@@ -420,99 +381,104 @@ int ib_umem_odp_map_dma_pages(struct ib_umem_odp *umem_odp, u64 user_virt,
 		goto out_put_task;
 	}
 
-	if (access_mask & ODP_WRITE_ALLOWED_BIT)
+	range.notifier = &umem_odp->notifier;
-		flags |= FOLL_WRITE;
+	range.start = ALIGN_DOWN(user_virt, 1UL << page_shift);
+	range.end = ALIGN(user_virt + bcnt, 1UL << page_shift);
+	pfn_start_idx = (range.start - ib_umem_start(umem_odp)) >> PAGE_SHIFT;
+	num_pfns = (range.end - range.start) >> PAGE_SHIFT;
+	if (fault) {
+		range.default_flags = HMM_PFN_REQ_FAULT;
 
-	start_idx = (user_virt - ib_umem_start(umem_odp)) >> page_shift;
+		if (access_mask & ODP_WRITE_ALLOWED_BIT)
-	k = start_idx;
+			range.default_flags |= HMM_PFN_REQ_WRITE;
+	}
 
-	while (bcnt > 0) {
+	range.hmm_pfns = &(umem_odp->pfn_list[pfn_start_idx]);
-		const size_t gup_num_pages = min_t(size_t,
+	timeout = jiffies + msecs_to_jiffies(HMM_RANGE_DEFAULT_TIMEOUT);
-				ALIGN(bcnt, PAGE_SIZE) / PAGE_SIZE,
-				PAGE_SIZE / sizeof(struct page *));
 
-		mmap_read_lock(owning_mm);
+retry:
-		/*
+	current_seq = range.notifier_seq =
-		 * Note: this might result in redundent page getting. We can
+		mmu_interval_read_begin(&umem_odp->notifier);
-		 * avoid this by checking dma_list to be 0 before calling
-		 * get_user_pages. However, this make the code much more
-		 * complex (and doesn't gain us much performance in most use
-		 * cases).
-		 */
-		npages = get_user_pages_remote(owning_mm,
-				user_virt, gup_num_pages,
-				flags, local_page_list, NULL, NULL);
-		mmap_read_unlock(owning_mm);
 
-		if (npages < 0) {
+	mmap_read_lock(owning_mm);
-			if (npages != -EAGAIN)
+	ret = hmm_range_fault(&range);
-				pr_warn("fail to get %zu user pages with error %d\n", gup_num_pages, npages);
+	mmap_read_unlock(owning_mm);
-			else
+	if (unlikely(ret)) {
-				pr_debug("fail to get %zu user pages with error %d\n", gup_num_pages, npages);
+		if (ret == -EBUSY && !time_after(jiffies, timeout))
-			break;
+			goto retry;
-		}
+		goto out_put_mm;
+	}
 
-		bcnt -= min_t(size_t, npages << PAGE_SHIFT, bcnt);
+	start_idx = (range.start - ib_umem_start(umem_odp)) >> page_shift;
-		mutex_lock(&umem_odp->umem_mutex);
+	dma_index = start_idx;
-		for (j = 0; j < npages; j++, user_virt += PAGE_SIZE) {
+
-			if (user_virt & ~page_mask) {
+	mutex_lock(&umem_odp->umem_mutex);
-				p += PAGE_SIZE;
+	if (mmu_interval_read_retry(&umem_odp->notifier, current_seq)) {
-				if (page_to_phys(local_page_list[j]) != p) {
+		mutex_unlock(&umem_odp->umem_mutex);
-					ret = -EFAULT;
+		goto retry;
-					break;
+	}
-				}
+
-				put_page(local_page_list[j]);
+	for (pfn_index = 0; pfn_index < num_pfns;
+		pfn_index += 1 << (page_shift - PAGE_SHIFT), dma_index++) {
+
+		if (fault) {
+			/*
+			 * Since we asked for hmm_range_fault() to populate
+			 * pages it shouldn't return an error entry on success.
+			 */
+			WARN_ON(range.hmm_pfns[pfn_index] & HMM_PFN_ERROR);
+			WARN_ON(!(range.hmm_pfns[pfn_index] & HMM_PFN_VALID));
+		} else {
+			if (!(range.hmm_pfns[pfn_index] & HMM_PFN_VALID)) {
+				WARN_ON(umem_odp->dma_list[dma_index]);
 				continue;
 			}
-
+			access_mask = ODP_READ_ALLOWED_BIT;
-			ret = ib_umem_odp_map_dma_single_page(
+			if (range.hmm_pfns[pfn_index] & HMM_PFN_WRITE)
-					umem_odp, k, local_page_list[j],
+				access_mask |= ODP_WRITE_ALLOWED_BIT;
-					access_mask, current_seq);
-			if (ret < 0) {
-				if (ret != -EAGAIN)
-					pr_warn("ib_umem_odp_map_dma_single_page failed with error %d\n", ret);
-				else
-					pr_debug("ib_umem_odp_map_dma_single_page failed with error %d\n", ret);
-				break;
-			}
-
-			p = page_to_phys(local_page_list[j]);
-			k++;
 		}
-		mutex_unlock(&umem_odp->umem_mutex);
 
+		hmm_order = hmm_pfn_to_map_order(range.hmm_pfns[pfn_index]);
+		/* If a hugepage was detected and ODP wasn't set for, the umem
+		 * page_shift will be used, the opposite case is an error.
+		 */
+		if (hmm_order + PAGE_SHIFT < page_shift) {
+			ret = -EINVAL;
+			ibdev_dbg(umem_odp->umem.ibdev,
+				  "%s: un-expected hmm_order %d, page_shift %d\n",
+				  __func__, hmm_order, page_shift);
+			break;
+		}
+
+		ret = ib_umem_odp_map_dma_single_page(
+				umem_odp, dma_index, hmm_pfn_to_page(range.hmm_pfns[pfn_index]),
+				access_mask);
 		if (ret < 0) {
-			/*
+			ibdev_dbg(umem_odp->umem.ibdev,
-			 * Release pages, remembering that the first page
+				  "ib_umem_odp_map_dma_single_page failed with error %d\n", ret);
-			 * to hit an error was already released by
-			 * ib_umem_odp_map_dma_single_page().
-			 */
-			if (npages - (j + 1) > 0)
-				release_pages(&local_page_list[j+1],
-					      npages - (j + 1));
 			break;
 		}
 	}
+	/* upon sucesss lock should stay on hold for the callee */
+	if (!ret)
+		ret = dma_index - start_idx;
+	else
+		mutex_unlock(&umem_odp->umem_mutex);
 
-	if (ret >= 0) {
+out_put_mm:
-		if (npages < 0 && k == start_idx)
-			ret = npages;
-		else
-			ret = k - start_idx;
-	}
-
 	mmput(owning_mm);
 out_put_task:
 	if (owning_process)
 		put_task_struct(owning_process);
-	free_page((unsigned long)local_page_list);
 	return ret;
 }
-EXPORT_SYMBOL(ib_umem_odp_map_dma_pages);
+EXPORT_SYMBOL(ib_umem_odp_map_dma_and_lock);
 
 void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 virt,
 				 u64 bound)
 {
+	dma_addr_t dma_addr;
+	dma_addr_t dma;
 	int idx;
 	u64 addr;
 	struct ib_device *dev = umem_odp->umem.ibdev;
@@ -521,20 +487,16 @@ void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 virt,
 
 	virt = max_t(u64, virt, ib_umem_start(umem_odp));
 	bound = min_t(u64, bound, ib_umem_end(umem_odp));
-	/* Note that during the run of this function, the
-	 * notifiers_count of the MR is > 0, preventing any racing
-	 * faults from completion. We might be racing with other
-	 * invalidations, so we must make sure we free each page only
-	 * once. */
 	for (addr = virt; addr < bound; addr += BIT(umem_odp->page_shift)) {
 		idx = (addr - ib_umem_start(umem_odp)) >> umem_odp->page_shift;
-		if (umem_odp->page_list[idx]) {
+		dma = umem_odp->dma_list[idx];
-			struct page *page = umem_odp->page_list[idx];
-			dma_addr_t dma = umem_odp->dma_list[idx];
-			dma_addr_t dma_addr = dma & ODP_DMA_ADDR_MASK;
 
-			WARN_ON(!dma_addr);
+		/* The access flags guaranteed a valid DMA address in case was NULL */
+		if (dma) {
+			unsigned long pfn_idx = (addr - ib_umem_start(umem_odp)) >> PAGE_SHIFT;
+			struct page *page = hmm_pfn_to_page(umem_odp->pfn_list[pfn_idx]);
 
+			dma_addr = dma & ODP_DMA_ADDR_MASK;
 			ib_dma_unmap_page(dev, dma_addr,
 					  BIT(umem_odp->page_shift),
 					  DMA_BIDIRECTIONAL);
@@ -551,7 +513,6 @@ void ib_umem_odp_unmap_dma_pages(struct ib_umem_odp *umem_odp, u64 virt,
 				 */
 				set_page_dirty(head_page);
 			}
-			umem_odp->page_list[idx] = NULL;
 			umem_odp->dma_list[idx] = 0;
 			umem_odp->npages--;
 		}

drivers/infiniband/core/uverbs_cmd.c

@@ -218,10 +218,12 @@ int ib_alloc_ucontext(struct uverbs_attr_bundle *attrs)
 	if (!ucontext)
 		return -ENOMEM;
 
-	ucontext->res.type = RDMA_RESTRACK_CTX;
 	ucontext->device = ib_dev;
 	ucontext->ufile = ufile;
 	xa_init_flags(&ucontext->mmap_xa, XA_FLAGS_ALLOC);
+
+	rdma_restrack_new(&ucontext->res, RDMA_RESTRACK_CTX);
+	rdma_restrack_set_name(&ucontext->res, NULL);
 	attrs->context = ucontext;
 	return 0;
 }
@@ -250,7 +252,7 @@ int ib_init_ucontext(struct uverbs_attr_bundle *attrs)
 	if (ret)
 		goto err_uncharge;
 
-	rdma_restrack_uadd(&ucontext->res);
+	rdma_restrack_add(&ucontext->res);
 
 	/*
 	 * Make sure that ib_uverbs_get_ucontext() sees the pointer update
@@ -313,6 +315,7 @@ static int ib_uverbs_get_context(struct uverbs_attr_bundle *attrs)
 err_uobj:
 	rdma_alloc_abort_uobject(uobj, attrs, false);
 err_ucontext:
+	rdma_restrack_put(&attrs->context->res);
 	kfree(attrs->context);
 	attrs->context = NULL;
 	return ret;
@@ -439,12 +442,14 @@ static int ib_uverbs_alloc_pd(struct uverbs_attr_bundle *attrs)
 	pd->device  = ib_dev;
 	pd->uobject = uobj;
 	atomic_set(&pd->usecnt, 0);
-	pd->res.type = RDMA_RESTRACK_PD;
+
+	rdma_restrack_new(&pd->res, RDMA_RESTRACK_PD);
+	rdma_restrack_set_name(&pd->res, NULL);
 
 	ret = ib_dev->ops.alloc_pd(pd, &attrs->driver_udata);
 	if (ret)
 		goto err_alloc;
-	rdma_restrack_uadd(&pd->res);
+	rdma_restrack_add(&pd->res);
 
 	uobj->object = pd;
 	uobj_finalize_uobj_create(uobj, attrs);
@@ -453,6 +458,7 @@ static int ib_uverbs_alloc_pd(struct uverbs_attr_bundle *attrs)
 	return uverbs_response(attrs, &resp, sizeof(resp));
 
 err_alloc:
+	rdma_restrack_put(&pd->res);
 	kfree(pd);
 err:
 	uobj_alloc_abort(uobj, attrs);
@@ -742,9 +748,11 @@ static int ib_uverbs_reg_mr(struct uverbs_attr_bundle *attrs)
 	mr->sig_attrs = NULL;
 	mr->uobject = uobj;
 	atomic_inc(&pd->usecnt);
-	mr->res.type = RDMA_RESTRACK_MR;
 	mr->iova = cmd.hca_va;
-	rdma_restrack_uadd(&mr->res);
+
+	rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR);
+	rdma_restrack_set_name(&mr->res, NULL);
+	rdma_restrack_add(&mr->res);
 
 	uobj->object = mr;
 	uobj_put_obj_read(pd);
@@ -858,7 +866,7 @@ static int ib_uverbs_dereg_mr(struct uverbs_attr_bundle *attrs)
 static int ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs)
 {
 	struct ib_uverbs_alloc_mw      cmd;
-	struct ib_uverbs_alloc_mw_resp resp;
+	struct ib_uverbs_alloc_mw_resp resp = {};
 	struct ib_uobject             *uobj;
 	struct ib_pd                  *pd;
 	struct ib_mw                  *mw;
@@ -884,15 +892,21 @@ static int ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs)
 		goto err_put;
 	}
 
-	mw = pd->device->ops.alloc_mw(pd, cmd.mw_type, &attrs->driver_udata);
+	mw = rdma_zalloc_drv_obj(ib_dev, ib_mw);
-	if (IS_ERR(mw)) {
+	if (!mw) {
-		ret = PTR_ERR(mw);
+		ret = -ENOMEM;
 		goto err_put;
 	}
 
-	mw->device  = pd->device;
+	mw->device = ib_dev;
-	mw->pd      = pd;
+	mw->pd = pd;
 	mw->uobject = uobj;
+	mw->type = cmd.mw_type;
+
+	ret = pd->device->ops.alloc_mw(mw, &attrs->driver_udata);
+	if (ret)
+		goto err_alloc;
+
 	atomic_inc(&pd->usecnt);
 
 	uobj->object = mw;
@@ -903,6 +917,8 @@ static int ib_uverbs_alloc_mw(struct uverbs_attr_bundle *attrs)
 	resp.mw_handle = uobj->id;
 	return uverbs_response(attrs, &resp, sizeof(resp));
 
+err_alloc:
+	kfree(mw);
 err_put:
 	uobj_put_obj_read(pd);
 err_free:
@@ -994,12 +1010,14 @@ static int create_cq(struct uverbs_attr_bundle *attrs,
 	cq->event_handler = ib_uverbs_cq_event_handler;
 	cq->cq_context    = ev_file ? &ev_file->ev_queue : NULL;
 	atomic_set(&cq->usecnt, 0);
-	cq->res.type = RDMA_RESTRACK_CQ;
+
+	rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ);
+	rdma_restrack_set_name(&cq->res, NULL);
 
 	ret = ib_dev->ops.create_cq(cq, &attr, &attrs->driver_udata);
 	if (ret)
 		goto err_free;
-	rdma_restrack_uadd(&cq->res);
+	rdma_restrack_add(&cq->res);
 
 	obj->uevent.uobject.object = cq;
 	obj->uevent.event_file = READ_ONCE(attrs->ufile->default_async_file);
@@ -1013,6 +1031,7 @@ static int create_cq(struct uverbs_attr_bundle *attrs,
 	return uverbs_response(attrs, &resp, sizeof(resp));
 
 err_free:
+	rdma_restrack_put(&cq->res);
 	kfree(cq);
 err_file:
 	if (ev_file)
@@ -1237,8 +1256,21 @@ static int create_qp(struct uverbs_attr_bundle *attrs,
 	bool has_sq = true;
 	struct ib_device *ib_dev;
 
-	if (cmd->qp_type == IB_QPT_RAW_PACKET && !capable(CAP_NET_RAW))
+	switch (cmd->qp_type) {
-		return -EPERM;
+	case IB_QPT_RAW_PACKET:
+		if (!capable(CAP_NET_RAW))
+			return -EPERM;
+		break;
+	case IB_QPT_RC:
+	case IB_QPT_UC:
+	case IB_QPT_UD:
+	case IB_QPT_XRC_INI:
+	case IB_QPT_XRC_TGT:
+	case IB_QPT_DRIVER:
+		break;
+	default:
+		return -EINVAL;
+	}
 
 	obj = (struct ib_uqp_object *)uobj_alloc(UVERBS_OBJECT_QP, attrs,
 						 &ib_dev);
@@ -2985,11 +3017,11 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs)
 {
 	struct ib_uverbs_ex_create_rwq_ind_table cmd;
 	struct ib_uverbs_ex_create_rwq_ind_table_resp  resp = {};
-	struct ib_uobject		  *uobj;
+	struct ib_uobject *uobj;
 	int err;
 	struct ib_rwq_ind_table_init_attr init_attr = {};
 	struct ib_rwq_ind_table *rwq_ind_tbl;
-	struct ib_wq	**wqs = NULL;
+	struct ib_wq **wqs = NULL;
 	u32 *wqs_handles = NULL;
 	struct ib_wq	*wq = NULL;
 	int i, num_read_wqs;
@@ -3047,17 +3079,15 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs)
 		goto put_wqs;
 	}
 
-	init_attr.log_ind_tbl_size = cmd.log_ind_tbl_size;
+	rwq_ind_tbl = rdma_zalloc_drv_obj(ib_dev, ib_rwq_ind_table);
-	init_attr.ind_tbl = wqs;
+	if (!rwq_ind_tbl) {
-
+		err = -ENOMEM;
-	rwq_ind_tbl = ib_dev->ops.create_rwq_ind_table(ib_dev, &init_attr,
-						       &attrs->driver_udata);
-
-	if (IS_ERR(rwq_ind_tbl)) {
-		err = PTR_ERR(rwq_ind_tbl);
 		goto err_uobj;
 	}
 
+	init_attr.log_ind_tbl_size = cmd.log_ind_tbl_size;
+	init_attr.ind_tbl = wqs;
+
 	rwq_ind_tbl->ind_tbl = wqs;
 	rwq_ind_tbl->log_ind_tbl_size = init_attr.log_ind_tbl_size;
 	rwq_ind_tbl->uobject = uobj;
@@ -3065,6 +3095,11 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs)
 	rwq_ind_tbl->device = ib_dev;
 	atomic_set(&rwq_ind_tbl->usecnt, 0);
 
+	err = ib_dev->ops.create_rwq_ind_table(rwq_ind_tbl, &init_attr,
+					       &attrs->driver_udata);
+	if (err)
+		goto err_create;
+
 	for (i = 0; i < num_wq_handles; i++)
 		rdma_lookup_put_uobject(&wqs[i]->uobject->uevent.uobject,
 					UVERBS_LOOKUP_READ);
@@ -3076,6 +3111,8 @@ static int ib_uverbs_ex_create_rwq_ind_table(struct uverbs_attr_bundle *attrs)
 	resp.response_length = uverbs_response_length(attrs, sizeof(resp));
 	return uverbs_response(attrs, &resp, sizeof(resp));
 
+err_create:
+	kfree(rwq_ind_tbl);
 err_uobj:
 	uobj_alloc_abort(uobj, attrs);
 put_wqs:
@@ -3232,8 +3269,8 @@ static int ib_uverbs_ex_create_flow(struct uverbs_attr_bundle *attrs)
 		goto err_free;
 	}
 
-	flow_id = qp->device->ops.create_flow(
+	flow_id = qp->device->ops.create_flow(qp, flow_attr,
-		qp, flow_attr, IB_FLOW_DOMAIN_USER, &attrs->driver_udata);
+					      &attrs->driver_udata);
 
 	if (IS_ERR(flow_id)) {
 		err = PTR_ERR(flow_id);

drivers/infiniband/core/uverbs_main.c

@@ -108,8 +108,11 @@ int uverbs_dealloc_mw(struct ib_mw *mw)
 	int ret;
 
 	ret = mw->device->ops.dealloc_mw(mw);
-	if (!ret)
+	if (ret)
-		atomic_dec(&pd->usecnt);
+		return ret;
+
+	atomic_dec(&pd->usecnt);
+	kfree(mw);
 	return ret;
 }
 

drivers/infiniband/core/uverbs_std_types.c

@@ -81,12 +81,20 @@ static int uverbs_free_rwq_ind_tbl(struct ib_uobject *uobject,
 {
 	struct ib_rwq_ind_table *rwq_ind_tbl = uobject->object;
 	struct ib_wq **ind_tbl = rwq_ind_tbl->ind_tbl;
-	int ret;
+	u32 table_size = (1 << rwq_ind_tbl->log_ind_tbl_size);
+	int ret, i;
 
-	ret = ib_destroy_rwq_ind_table(rwq_ind_tbl);
+	if (atomic_read(&rwq_ind_tbl->usecnt))
+		return -EBUSY;
+
+	ret = rwq_ind_tbl->device->ops.destroy_rwq_ind_table(rwq_ind_tbl);
 	if (ib_is_destroy_retryable(ret, why, uobject))
 		return ret;
 
+	for (i = 0; i < table_size; i++)
+		atomic_dec(&ind_tbl[i]->usecnt);
+
+	kfree(rwq_ind_tbl);
 	kfree(ind_tbl);
 	return ret;
 }
@@ -122,8 +130,7 @@ static int uverbs_free_pd(struct ib_uobject *uobject,
 	if (ret)
 		return ret;
 
-	ib_dealloc_pd_user(pd, &attrs->driver_udata);
+	return ib_dealloc_pd_user(pd, &attrs->driver_udata);
-	return 0;
 }
 
 void ib_uverbs_free_event_queue(struct ib_uverbs_event_queue *event_queue)

drivers/infiniband/core/uverbs_std_types_counters.c

@@ -46,7 +46,9 @@ static int uverbs_free_counters(struct ib_uobject *uobject,
 	if (ret)
 		return ret;
 
-	counters->device->ops.destroy_counters(counters);
+	ret = counters->device->ops.destroy_counters(counters);
+	if (ret)
+		return ret;
 	kfree(counters);
 	return 0;
 }

drivers/infiniband/core/uverbs_std_types_cq.c

@@ -33,6 +33,7 @@
 #include <rdma/uverbs_std_types.h>
 #include "rdma_core.h"
 #include "uverbs.h"
+#include "restrack.h"
 
 static int uverbs_free_cq(struct ib_uobject *uobject,
 			  enum rdma_remove_reason why,
@@ -123,7 +124,9 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)(
 	cq->event_handler = ib_uverbs_cq_event_handler;
 	cq->cq_context    = ev_file ? &ev_file->ev_queue : NULL;
 	atomic_set(&cq->usecnt, 0);
-	cq->res.type = RDMA_RESTRACK_CQ;
+
+	rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ);
+	rdma_restrack_set_name(&cq->res, NULL);
 
 	ret = ib_dev->ops.create_cq(cq, &attr, &attrs->driver_udata);
 	if (ret)
@@ -131,7 +134,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)(
 
 	obj->uevent.uobject.object = cq;
 	obj->uevent.uobject.user_handle = user_handle;
-	rdma_restrack_uadd(&cq->res);
+	rdma_restrack_add(&cq->res);
 	uverbs_finalize_uobj_create(attrs, UVERBS_ATTR_CREATE_CQ_HANDLE);
 
 	ret = uverbs_copy_to(attrs, UVERBS_ATTR_CREATE_CQ_RESP_CQE, &cq->cqe,
@@ -139,6 +142,7 @@ static int UVERBS_HANDLER(UVERBS_METHOD_CQ_CREATE)(
 	return ret;
 
 err_free:
+	rdma_restrack_put(&cq->res);
 	kfree(cq);
 err_event_file:
 	if (obj->uevent.event_file)

drivers/infiniband/core/uverbs_std_types_device.c

@@ -3,11 +3,13 @@
  * Copyright (c) 2018, Mellanox Technologies inc.  All rights reserved.
  */
 
+#include <linux/overflow.h>
 #include <rdma/uverbs_std_types.h>
 #include "rdma_core.h"
 #include "uverbs.h"
 #include <rdma/uverbs_ioctl.h>
 #include <rdma/opa_addr.h>
+#include <rdma/ib_cache.h>
 
 /*
  * This ioctl method allows calling any defined write or write_ex
@@ -165,7 +167,8 @@ void copy_port_attr_to_resp(struct ib_port_attr *attr,
 	resp->subnet_timeout = attr->subnet_timeout;
 	resp->init_type_reply = attr->init_type_reply;
 	resp->active_width = attr->active_width;
-	resp->active_speed = attr->active_speed;
+	/* This ABI needs to be extended to provide any speed more than IB_SPEED_NDR */
+	resp->active_speed = min_t(u16, attr->active_speed, IB_SPEED_NDR);
 	resp->phys_state = attr->phys_state;
 	resp->link_layer = rdma_port_get_link_layer(ib_dev, port_num);
 }
@@ -265,6 +268,172 @@ static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_CONTEXT)(
 	return ucontext->device->ops.query_ucontext(ucontext, attrs);
 }
 
+static int copy_gid_entries_to_user(struct uverbs_attr_bundle *attrs,
+				    struct ib_uverbs_gid_entry *entries,
+				    size_t num_entries, size_t user_entry_size)
+{
+	const struct uverbs_attr *attr;
+	void __user *user_entries;
+	size_t copy_len;
+	int ret;
+	int i;
+
+	if (user_entry_size == sizeof(*entries)) {
+		ret = uverbs_copy_to(attrs,
+				     UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES,
+				     entries, sizeof(*entries) * num_entries);
+		return ret;
+	}
+
+	copy_len = min_t(size_t, user_entry_size, sizeof(*entries));
+	attr = uverbs_attr_get(attrs, UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES);
+	if (IS_ERR(attr))
+		return PTR_ERR(attr);
+
+	user_entries = u64_to_user_ptr(attr->ptr_attr.data);
+	for (i = 0; i < num_entries; i++) {
+		if (copy_to_user(user_entries, entries, copy_len))
+			return -EFAULT;
+
+		if (user_entry_size > sizeof(*entries)) {
+			if (clear_user(user_entries + sizeof(*entries),
+				       user_entry_size - sizeof(*entries)))
+				return -EFAULT;
+		}
+
+		entries++;
+		user_entries += user_entry_size;
+	}
+
+	return uverbs_output_written(attrs,
+				     UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES);
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_GID_TABLE)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uverbs_gid_entry *entries;
+	struct ib_ucontext *ucontext;
+	struct ib_device *ib_dev;
+	size_t user_entry_size;
+	ssize_t num_entries;
+	size_t max_entries;
+	size_t num_bytes;
+	u32 flags;
+	int ret;
+
+	ret = uverbs_get_flags32(&flags, attrs,
+				 UVERBS_ATTR_QUERY_GID_TABLE_FLAGS, 0);
+	if (ret)
+		return ret;
+
+	ret = uverbs_get_const(&user_entry_size, attrs,
+			       UVERBS_ATTR_QUERY_GID_TABLE_ENTRY_SIZE);
+	if (ret)
+		return ret;
+
+	max_entries = uverbs_attr_ptr_get_array_size(
+		attrs, UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES,
+		user_entry_size);
+	if (max_entries <= 0)
+		return -EINVAL;
+
+	ucontext = ib_uverbs_get_ucontext(attrs);
+	if (IS_ERR(ucontext))
+		return PTR_ERR(ucontext);
+	ib_dev = ucontext->device;
+
+	if (check_mul_overflow(max_entries, sizeof(*entries), &num_bytes))
+		return -EINVAL;
+
+	entries = uverbs_zalloc(attrs, num_bytes);
+	if (!entries)
+		return -ENOMEM;
+
+	num_entries = rdma_query_gid_table(ib_dev, entries, max_entries);
+	if (num_entries < 0)
+		return -EINVAL;
+
+	ret = copy_gid_entries_to_user(attrs, entries, num_entries,
+				       user_entry_size);
+	if (ret)
+		return ret;
+
+	ret = uverbs_copy_to(attrs,
+			     UVERBS_ATTR_QUERY_GID_TABLE_RESP_NUM_ENTRIES,
+			     &num_entries, sizeof(num_entries));
+	return ret;
+}
+
+static int UVERBS_HANDLER(UVERBS_METHOD_QUERY_GID_ENTRY)(
+	struct uverbs_attr_bundle *attrs)
+{
+	struct ib_uverbs_gid_entry entry = {};
+	const struct ib_gid_attr *gid_attr;
+	struct ib_ucontext *ucontext;
+	struct ib_device *ib_dev;
+	struct net_device *ndev;
+	u32 gid_index;
+	u32 port_num;
+	u32 flags;
+	int ret;
+
+	ret = uverbs_get_flags32(&flags, attrs,
+				 UVERBS_ATTR_QUERY_GID_ENTRY_FLAGS, 0);
+	if (ret)
+		return ret;
+
+	ret = uverbs_get_const(&port_num, attrs,
+			       UVERBS_ATTR_QUERY_GID_ENTRY_PORT);
+	if (ret)
+		return ret;
+
+	ret = uverbs_get_const(&gid_index, attrs,
+			       UVERBS_ATTR_QUERY_GID_ENTRY_GID_INDEX);
+	if (ret)
+		return ret;
+
+	ucontext = ib_uverbs_get_ucontext(attrs);
+	if (IS_ERR(ucontext))
+		return PTR_ERR(ucontext);
+	ib_dev = ucontext->device;
+
+	if (!rdma_is_port_valid(ib_dev, port_num))
+		return -EINVAL;
+
+	if (!rdma_ib_or_roce(ib_dev, port_num))
+		return -EOPNOTSUPP;
+
+	gid_attr = rdma_get_gid_attr(ib_dev, port_num, gid_index);
+	if (IS_ERR(gid_attr))
+		return PTR_ERR(gid_attr);
+
+	memcpy(&entry.gid, &gid_attr->gid, sizeof(gid_attr->gid));
+	entry.gid_index = gid_attr->index;
+	entry.port_num = gid_attr->port_num;
+	entry.gid_type = gid_attr->gid_type;
+
+	rcu_read_lock();
+	ndev = rdma_read_gid_attr_ndev_rcu(gid_attr);
+	if (IS_ERR(ndev)) {
+		if (PTR_ERR(ndev) != -ENODEV) {
+			ret = PTR_ERR(ndev);
+			rcu_read_unlock();
+			goto out;
+		}
+	} else {
+		entry.netdev_ifindex = ndev->ifindex;
+	}
+	rcu_read_unlock();
+
+	ret = uverbs_copy_to_struct_or_zero(
+		attrs, UVERBS_ATTR_QUERY_GID_ENTRY_RESP_ENTRY, &entry,
+		sizeof(entry));
+out:
+	rdma_put_gid_attr(gid_attr);
+	return ret;
+}
+
 DECLARE_UVERBS_NAMED_METHOD(
 	UVERBS_METHOD_GET_CONTEXT,
 	UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_GET_CONTEXT_NUM_COMP_VECTORS,
@@ -299,12 +468,38 @@ DECLARE_UVERBS_NAMED_METHOD(
 				   reserved),
 		UA_MANDATORY));
 
+DECLARE_UVERBS_NAMED_METHOD(
+	UVERBS_METHOD_QUERY_GID_TABLE,
+	UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_TABLE_ENTRY_SIZE, u64,
+			     UA_MANDATORY),
+	UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_QUERY_GID_TABLE_FLAGS, u32,
+			     UA_OPTIONAL),
+	UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_TABLE_RESP_ENTRIES,
+			    UVERBS_ATTR_MIN_SIZE(0), UA_MANDATORY),
+	UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_TABLE_RESP_NUM_ENTRIES,
+			    UVERBS_ATTR_TYPE(u64), UA_MANDATORY));
+
+DECLARE_UVERBS_NAMED_METHOD(
+	UVERBS_METHOD_QUERY_GID_ENTRY,
+	UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_ENTRY_PORT, u32,
+			     UA_MANDATORY),
+	UVERBS_ATTR_CONST_IN(UVERBS_ATTR_QUERY_GID_ENTRY_GID_INDEX, u32,
+			     UA_MANDATORY),
+	UVERBS_ATTR_FLAGS_IN(UVERBS_ATTR_QUERY_GID_ENTRY_FLAGS, u32,
+			     UA_MANDATORY),
+	UVERBS_ATTR_PTR_OUT(UVERBS_ATTR_QUERY_GID_ENTRY_RESP_ENTRY,
+			    UVERBS_ATTR_STRUCT(struct ib_uverbs_gid_entry,
+					       netdev_ifindex),
+			    UA_MANDATORY));
+
 DECLARE_UVERBS_GLOBAL_METHODS(UVERBS_OBJECT_DEVICE,
 			      &UVERBS_METHOD(UVERBS_METHOD_GET_CONTEXT),
 			      &UVERBS_METHOD(UVERBS_METHOD_INVOKE_WRITE),
 			      &UVERBS_METHOD(UVERBS_METHOD_INFO_HANDLES),
 			      &UVERBS_METHOD(UVERBS_METHOD_QUERY_PORT),
-			      &UVERBS_METHOD(UVERBS_METHOD_QUERY_CONTEXT));
+			      &UVERBS_METHOD(UVERBS_METHOD_QUERY_CONTEXT),
+			      &UVERBS_METHOD(UVERBS_METHOD_QUERY_GID_TABLE),
+			      &UVERBS_METHOD(UVERBS_METHOD_QUERY_GID_ENTRY));
 
 const struct uapi_definition uverbs_def_obj_device[] = {
 	UAPI_DEF_CHAIN_OBJ_TREE_NAMED(UVERBS_OBJECT_DEVICE),

drivers/infiniband/core/uverbs_std_types_wq.c

@@ -16,7 +16,7 @@ static int uverbs_free_wq(struct ib_uobject *uobject,
 		container_of(uobject, struct ib_uwq_object, uevent.uobject);
 	int ret;
 
-	ret = ib_destroy_wq(wq, &attrs->driver_udata);
+	ret = ib_destroy_wq_user(wq, &attrs->driver_udata);
 	if (ib_is_destroy_retryable(ret, why, uobject))
 		return ret;
 

drivers/infiniband/core/verbs.c

@@ -272,15 +272,16 @@ struct ib_pd *__ib_alloc_pd(struct ib_device *device, unsigned int flags,
 	atomic_set(&pd->usecnt, 0);
 	pd->flags = flags;
 
-	pd->res.type = RDMA_RESTRACK_PD;
+	rdma_restrack_new(&pd->res, RDMA_RESTRACK_PD);
-	rdma_restrack_set_task(&pd->res, caller);
+	rdma_restrack_set_name(&pd->res, caller);
 
 	ret = device->ops.alloc_pd(pd, NULL);
 	if (ret) {
+		rdma_restrack_put(&pd->res);
 		kfree(pd);
 		return ERR_PTR(ret);
 	}
-	rdma_restrack_kadd(&pd->res);
+	rdma_restrack_add(&pd->res);
 
 	if (device->attrs.device_cap_flags & IB_DEVICE_LOCAL_DMA_LKEY)
 		pd->local_dma_lkey = device->local_dma_lkey;
@@ -329,7 +330,7 @@ EXPORT_SYMBOL(__ib_alloc_pd);
  * exist.  The caller is responsible to synchronously destroy them and
  * guarantee no new allocations will happen.
  */
-void ib_dealloc_pd_user(struct ib_pd *pd, struct ib_udata *udata)
+int ib_dealloc_pd_user(struct ib_pd *pd, struct ib_udata *udata)
 {
 	int ret;
 
@@ -343,9 +344,13 @@ void ib_dealloc_pd_user(struct ib_pd *pd, struct ib_udata *udata)
 	   requires the caller to guarantee we can't race here. */
 	WARN_ON(atomic_read(&pd->usecnt));
 
+	ret = pd->device->ops.dealloc_pd(pd, udata);
+	if (ret)
+		return ret;
+
 	rdma_restrack_del(&pd->res);
-	pd->device->ops.dealloc_pd(pd, udata);
 	kfree(pd);
+	return ret;
 }
 EXPORT_SYMBOL(ib_dealloc_pd_user);
 
@@ -728,7 +733,7 @@ int ib_get_gids_from_rdma_hdr(const union rdma_network_hdr *hdr,
 				       (struct in6_addr *)dgid);
 		return 0;
 	} else if (net_type == RDMA_NETWORK_IPV6 ||
-		   net_type == RDMA_NETWORK_IB) {
+		   net_type == RDMA_NETWORK_IB || RDMA_NETWORK_ROCE_V1) {
 		*dgid = hdr->ibgrh.dgid;
 		*sgid = hdr->ibgrh.sgid;
 		return 0;
@@ -964,18 +969,22 @@ int rdma_destroy_ah_user(struct ib_ah *ah, u32 flags, struct ib_udata *udata)
 {
 	const struct ib_gid_attr *sgid_attr = ah->sgid_attr;
 	struct ib_pd *pd;
+	int ret;
 
 	might_sleep_if(flags & RDMA_DESTROY_AH_SLEEPABLE);
 
 	pd = ah->pd;
 
-	ah->device->ops.destroy_ah(ah, flags);
+	ret = ah->device->ops.destroy_ah(ah, flags);
+	if (ret)
+		return ret;
+
 	atomic_dec(&pd->usecnt);
 	if (sgid_attr)
 		rdma_put_gid_attr(sgid_attr);
 
 	kfree(ah);
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL(rdma_destroy_ah_user);
 
@@ -1060,10 +1069,14 @@ EXPORT_SYMBOL(ib_query_srq);
 
 int ib_destroy_srq_user(struct ib_srq *srq, struct ib_udata *udata)
 {
+	int ret;
+
 	if (atomic_read(&srq->usecnt))
 		return -EBUSY;
 
-	srq->device->ops.destroy_srq(srq, udata);
+	ret = srq->device->ops.destroy_srq(srq, udata);
+	if (ret)
+		return ret;
 
 	atomic_dec(&srq->pd->usecnt);
 	if (srq->srq_type == IB_SRQT_XRC)
@@ -1072,7 +1085,7 @@ int ib_destroy_srq_user(struct ib_srq *srq, struct ib_udata *udata)
 		atomic_dec(&srq->ext.cq->usecnt);
 	kfree(srq);
 
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL(ib_destroy_srq_user);
 
@@ -1781,7 +1794,7 @@ int ib_modify_qp_with_udata(struct ib_qp *ib_qp, struct ib_qp_attr *attr,
 }
 EXPORT_SYMBOL(ib_modify_qp_with_udata);
 
-int ib_get_eth_speed(struct ib_device *dev, u8 port_num, u8 *speed, u8 *width)
+int ib_get_eth_speed(struct ib_device *dev, u8 port_num, u16 *speed, u8 *width)
 {
 	int rc;
 	u32 netdev_speed;
@@ -1984,16 +1997,18 @@ struct ib_cq *__ib_create_cq(struct ib_device *device,
 	cq->event_handler = event_handler;
 	cq->cq_context = cq_context;
 	atomic_set(&cq->usecnt, 0);
-	cq->res.type = RDMA_RESTRACK_CQ;
+
-	rdma_restrack_set_task(&cq->res, caller);
+	rdma_restrack_new(&cq->res, RDMA_RESTRACK_CQ);
+	rdma_restrack_set_name(&cq->res, caller);
 
 	ret = device->ops.create_cq(cq, cq_attr, NULL);
 	if (ret) {
+		rdma_restrack_put(&cq->res);
 		kfree(cq);
 		return ERR_PTR(ret);
 	}
 
-	rdma_restrack_kadd(&cq->res);
+	rdma_restrack_add(&cq->res);
 	return cq;
 }
 EXPORT_SYMBOL(__ib_create_cq);
@@ -2011,16 +2026,21 @@ EXPORT_SYMBOL(rdma_set_cq_moderation);
 
 int ib_destroy_cq_user(struct ib_cq *cq, struct ib_udata *udata)
 {
+	int ret;
+
 	if (WARN_ON_ONCE(cq->shared))
 		return -EOPNOTSUPP;
 
 	if (atomic_read(&cq->usecnt))
 		return -EBUSY;
 
+	ret = cq->device->ops.destroy_cq(cq, udata);
+	if (ret)
+		return ret;
+
 	rdma_restrack_del(&cq->res);
-	cq->device->ops.destroy_cq(cq, udata);
 	kfree(cq);
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL(ib_destroy_cq_user);
 
@@ -2059,8 +2079,10 @@ struct ib_mr *ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	mr->pd = pd;
 	mr->dm = NULL;
 	atomic_inc(&pd->usecnt);
-	mr->res.type = RDMA_RESTRACK_MR;
+
-	rdma_restrack_kadd(&mr->res);
+	rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR);
+	rdma_restrack_parent_name(&mr->res, &pd->res);
+	rdma_restrack_add(&mr->res);
 
 	return mr;
 }
@@ -2139,11 +2161,12 @@ struct ib_mr *ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
 	mr->uobject = NULL;
 	atomic_inc(&pd->usecnt);
 	mr->need_inval = false;
-	mr->res.type = RDMA_RESTRACK_MR;
-	rdma_restrack_kadd(&mr->res);
 	mr->type = mr_type;
 	mr->sig_attrs = NULL;
 
+	rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR);
+	rdma_restrack_parent_name(&mr->res, &pd->res);
+	rdma_restrack_add(&mr->res);
 out:
 	trace_mr_alloc(pd, mr_type, max_num_sg, mr);
 	return mr;
@@ -2199,11 +2222,12 @@ struct ib_mr *ib_alloc_mr_integrity(struct ib_pd *pd,
 	mr->uobject = NULL;
 	atomic_inc(&pd->usecnt);
 	mr->need_inval = false;
-	mr->res.type = RDMA_RESTRACK_MR;
-	rdma_restrack_kadd(&mr->res);
 	mr->type = IB_MR_TYPE_INTEGRITY;
 	mr->sig_attrs = sig_attrs;
 
+	rdma_restrack_new(&mr->res, RDMA_RESTRACK_MR);
+	rdma_restrack_parent_name(&mr->res, &pd->res);
+	rdma_restrack_add(&mr->res);
 out:
 	trace_mr_integ_alloc(pd, max_num_data_sg, max_num_meta_sg, mr);
 	return mr;
@@ -2328,13 +2352,17 @@ EXPORT_SYMBOL(ib_alloc_xrcd_user);
  */
 int ib_dealloc_xrcd_user(struct ib_xrcd *xrcd, struct ib_udata *udata)
 {
+	int ret;
+
 	if (atomic_read(&xrcd->usecnt))
 		return -EBUSY;
 
 	WARN_ON(!xa_empty(&xrcd->tgt_qps));
-	xrcd->device->ops.dealloc_xrcd(xrcd, udata);
+	ret = xrcd->device->ops.dealloc_xrcd(xrcd, udata);
+	if (ret)
+		return ret;
 	kfree(xrcd);
-	return 0;
+	return ret;
 }
 EXPORT_SYMBOL(ib_dealloc_xrcd_user);
 
@@ -2378,25 +2406,28 @@ struct ib_wq *ib_create_wq(struct ib_pd *pd,
 EXPORT_SYMBOL(ib_create_wq);
 
 /**
- * ib_destroy_wq - Destroys the specified user WQ.
+ * ib_destroy_wq_user - Destroys the specified user WQ.
  * @wq: The WQ to destroy.
  * @udata: Valid user data
  */
-int ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata)
+int ib_destroy_wq_user(struct ib_wq *wq, struct ib_udata *udata)
 {
 	struct ib_cq *cq = wq->cq;
 	struct ib_pd *pd = wq->pd;
+	int ret;
 
 	if (atomic_read(&wq->usecnt))
 		return -EBUSY;
 
-	wq->device->ops.destroy_wq(wq, udata);
+	ret = wq->device->ops.destroy_wq(wq, udata);
+	if (ret)
+		return ret;
+
 	atomic_dec(&pd->usecnt);
 	atomic_dec(&cq->usecnt);
-
+	return ret;
-	return 0;
 }
-EXPORT_SYMBOL(ib_destroy_wq);
+EXPORT_SYMBOL(ib_destroy_wq_user);
 
 /**
  * ib_modify_wq - Modifies the specified WQ.
@@ -2419,29 +2450,6 @@ int ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
 }
 EXPORT_SYMBOL(ib_modify_wq);
 
-/*
- * ib_destroy_rwq_ind_table - Destroys the specified Indirection Table.
- * @wq_ind_table: The Indirection Table to destroy.
-*/
-int ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *rwq_ind_table)
-{
-	int err, i;
-	u32 table_size = (1 << rwq_ind_table->log_ind_tbl_size);
-	struct ib_wq **ind_tbl = rwq_ind_table->ind_tbl;
-
-	if (atomic_read(&rwq_ind_table->usecnt))
-		return -EBUSY;
-
-	err = rwq_ind_table->device->ops.destroy_rwq_ind_table(rwq_ind_table);
-	if (!err) {
-		for (i = 0; i < table_size; i++)
-			atomic_dec(&ind_tbl[i]->usecnt);
-	}
-
-	return err;
-}
-EXPORT_SYMBOL(ib_destroy_rwq_ind_table);
-
 int ib_check_mr_status(struct ib_mr *mr, u32 check_mask,
 		       struct ib_mr_status *mr_status)
 {

drivers/infiniband/hw/bnxt_re/bnxt_re.h

@@ -150,7 +150,7 @@ struct bnxt_re_dev {
 
 	struct delayed_work		worker;
 	u8				cur_prio_map;
-	u8				active_speed;
+	u16				active_speed;
 	u8				active_width;
 
 	/* FP Notification Queue (CQ & SRQ) */

drivers/infiniband/hw/bnxt_re/ib_verbs.c

@@ -532,7 +532,7 @@ fail:
 }
 
 /* Protection Domains */
-void bnxt_re_dealloc_pd(struct ib_pd *ib_pd, struct ib_udata *udata)
+int bnxt_re_dealloc_pd(struct ib_pd *ib_pd, struct ib_udata *udata)
 {
 	struct bnxt_re_pd *pd = container_of(ib_pd, struct bnxt_re_pd, ib_pd);
 	struct bnxt_re_dev *rdev = pd->rdev;
@@ -542,6 +542,7 @@ void bnxt_re_dealloc_pd(struct ib_pd *ib_pd, struct ib_udata *udata)
 	if (pd->qplib_pd.id)
 		bnxt_qplib_dealloc_pd(&rdev->qplib_res, &rdev->qplib_res.pd_tbl,
 				      &pd->qplib_pd);
+	return 0;
 }
 
 int bnxt_re_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
@@ -601,13 +602,14 @@ fail:
 }
 
 /* Address Handles */
-void bnxt_re_destroy_ah(struct ib_ah *ib_ah, u32 flags)
+int bnxt_re_destroy_ah(struct ib_ah *ib_ah, u32 flags)
 {
 	struct bnxt_re_ah *ah = container_of(ib_ah, struct bnxt_re_ah, ib_ah);
 	struct bnxt_re_dev *rdev = ah->rdev;
 
 	bnxt_qplib_destroy_ah(&rdev->qplib_res, &ah->qplib_ah,
 			      !(flags & RDMA_DESTROY_AH_SLEEPABLE));
+	return 0;
 }
 
 static u8 bnxt_re_stack_to_dev_nw_type(enum rdma_network_type ntype)
@@ -938,9 +940,7 @@ static int bnxt_re_init_user_qp(struct bnxt_re_dev *rdev, struct bnxt_re_pd *pd,
 		return PTR_ERR(umem);
 
 	qp->sumem = umem;
-	qplib_qp->sq.sg_info.sghead = umem->sg_head.sgl;
+	qplib_qp->sq.sg_info.umem = umem;
-	qplib_qp->sq.sg_info.npages = ib_umem_num_pages(umem);
-	qplib_qp->sq.sg_info.nmap = umem->nmap;
 	qplib_qp->sq.sg_info.pgsize = PAGE_SIZE;
 	qplib_qp->sq.sg_info.pgshft = PAGE_SHIFT;
 	qplib_qp->qp_handle = ureq.qp_handle;
@@ -953,9 +953,7 @@ static int bnxt_re_init_user_qp(struct bnxt_re_dev *rdev, struct bnxt_re_pd *pd,
 		if (IS_ERR(umem))
 			goto rqfail;
 		qp->rumem = umem;
-		qplib_qp->rq.sg_info.sghead = umem->sg_head.sgl;
+		qplib_qp->rq.sg_info.umem = umem;
-		qplib_qp->rq.sg_info.npages = ib_umem_num_pages(umem);
-		qplib_qp->rq.sg_info.nmap = umem->nmap;
 		qplib_qp->rq.sg_info.pgsize = PAGE_SIZE;
 		qplib_qp->rq.sg_info.pgshft = PAGE_SHIFT;
 	}
@@ -1568,7 +1566,7 @@ static enum ib_mtu __to_ib_mtu(u32 mtu)
 }
 
 /* Shared Receive Queues */
-void bnxt_re_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata)
+int bnxt_re_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata)
 {
 	struct bnxt_re_srq *srq = container_of(ib_srq, struct bnxt_re_srq,
 					       ib_srq);
@@ -1583,6 +1581,7 @@ void bnxt_re_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata)
 	atomic_dec(&rdev->srq_count);
 	if (nq)
 		nq->budget--;
+	return 0;
 }
 
 static int bnxt_re_init_user_srq(struct bnxt_re_dev *rdev,
@@ -1608,9 +1607,7 @@ static int bnxt_re_init_user_srq(struct bnxt_re_dev *rdev,
 		return PTR_ERR(umem);
 
 	srq->umem = umem;
-	qplib_srq->sg_info.sghead = umem->sg_head.sgl;
+	qplib_srq->sg_info.umem = umem;
-	qplib_srq->sg_info.npages = ib_umem_num_pages(umem);
-	qplib_srq->sg_info.nmap = umem->nmap;
 	qplib_srq->sg_info.pgsize = PAGE_SIZE;
 	qplib_srq->sg_info.pgshft = PAGE_SHIFT;
 	qplib_srq->srq_handle = ureq.srq_handle;
@@ -2800,7 +2797,7 @@ int bnxt_re_post_recv(struct ib_qp *ib_qp, const struct ib_recv_wr *wr,
 }
 
 /* Completion Queues */
-void bnxt_re_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
+int bnxt_re_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 {
 	struct bnxt_re_cq *cq;
 	struct bnxt_qplib_nq *nq;
@@ -2816,6 +2813,7 @@ void bnxt_re_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 	atomic_dec(&rdev->cq_count);
 	nq->budget--;
 	kfree(cq->cql);
+	return 0;
 }
 
 int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
@@ -2860,9 +2858,7 @@ int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 			rc = PTR_ERR(cq->umem);
 			goto fail;
 		}
-		cq->qplib_cq.sg_info.sghead = cq->umem->sg_head.sgl;
+		cq->qplib_cq.sg_info.umem = cq->umem;
-		cq->qplib_cq.sg_info.npages = ib_umem_num_pages(cq->umem);
-		cq->qplib_cq.sg_info.nmap = cq->umem->nmap;
 		cq->qplib_cq.dpi = &uctx->dpi;
 	} else {
 		cq->max_cql = min_t(u32, entries, MAX_CQL_PER_POLL);
@@ -3774,23 +3770,6 @@ int bnxt_re_dealloc_mw(struct ib_mw *ib_mw)
 	return rc;
 }
 
-static int bnxt_re_page_size_ok(int page_shift)
-{
-	switch (page_shift) {
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4K:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_8K:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_64K:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_2M:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_256K:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1M:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_4M:
-	case CMDQ_REGISTER_MR_LOG2_PBL_PG_SIZE_PG_1G:
-		return 1;
-	default:
-		return 0;
-	}
-}
-
 static int fill_umem_pbl_tbl(struct ib_umem *umem, u64 *pbl_tbl_orig,
 			     int page_shift)
 {
@@ -3798,7 +3777,7 @@ static int fill_umem_pbl_tbl(struct ib_umem *umem, u64 *pbl_tbl_orig,
 	u64 page_size =  BIT_ULL(page_shift);
 	struct ib_block_iter biter;
 
-	rdma_for_each_block(umem->sg_head.sgl, &biter, umem->nmap, page_size)
+	rdma_umem_for_each_dma_block(umem, &biter, page_size)
 		*pbl_tbl++ = rdma_block_iter_dma_address(&biter);
 
 	return pbl_tbl - pbl_tbl_orig;
@@ -3814,7 +3793,8 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
 	struct bnxt_re_mr *mr;
 	struct ib_umem *umem;
 	u64 *pbl_tbl = NULL;
-	int umem_pgs, page_shift, rc;
+	unsigned long page_size;
+	int umem_pgs, rc;
 
 	if (length > BNXT_RE_MAX_MR_SIZE) {
 		ibdev_err(&rdev->ibdev, "MR Size: %lld > Max supported:%lld\n",
@@ -3848,42 +3828,34 @@ struct ib_mr *bnxt_re_reg_user_mr(struct ib_pd *ib_pd, u64 start, u64 length,
 	mr->ib_umem = umem;
 
 	mr->qplib_mr.va = virt_addr;
-	umem_pgs = ib_umem_page_count(umem);
+	page_size = ib_umem_find_best_pgsz(
-	if (!umem_pgs) {
+		umem, BNXT_RE_PAGE_SIZE_4K | BNXT_RE_PAGE_SIZE_2M, virt_addr);
-		ibdev_err(&rdev->ibdev, "umem is invalid!");
+	if (!page_size) {
-		rc = -EINVAL;
+		ibdev_err(&rdev->ibdev, "umem page size unsupported!");
+		rc = -EFAULT;
 		goto free_umem;
 	}
 	mr->qplib_mr.total_size = length;
 
-	pbl_tbl = kcalloc(umem_pgs, sizeof(u64 *), GFP_KERNEL);
+	if (page_size == BNXT_RE_PAGE_SIZE_4K &&
+	    length > BNXT_RE_MAX_MR_SIZE_LOW) {
+		ibdev_err(&rdev->ibdev, "Requested MR Sz:%llu Max sup:%llu",
+			  length, (u64)BNXT_RE_MAX_MR_SIZE_LOW);
+		rc = -EINVAL;
+		goto free_umem;
+	}
+
+	umem_pgs = ib_umem_num_dma_blocks(umem, page_size);
+	pbl_tbl = kcalloc(umem_pgs, sizeof(*pbl_tbl), GFP_KERNEL);
 	if (!pbl_tbl) {
 		rc = -ENOMEM;
 		goto free_umem;
 	}
 
-	page_shift = __ffs(ib_umem_find_best_pgsz(umem,
-				BNXT_RE_PAGE_SIZE_4K | BNXT_RE_PAGE_SIZE_2M,
-				virt_addr));
-
-	if (!bnxt_re_page_size_ok(page_shift)) {
-		ibdev_err(&rdev->ibdev, "umem page size unsupported!");
-		rc = -EFAULT;
-		goto fail;
-	}
-
-	if (page_shift == BNXT_RE_PAGE_SHIFT_4K &&
-	    length > BNXT_RE_MAX_MR_SIZE_LOW) {
-		ibdev_err(&rdev->ibdev, "Requested MR Sz:%llu Max sup:%llu",
-			  length, (u64)BNXT_RE_MAX_MR_SIZE_LOW);
-		rc = -EINVAL;
-		goto fail;
-	}
-
 	/* Map umem buf ptrs to the PBL */
-	umem_pgs = fill_umem_pbl_tbl(umem, pbl_tbl, page_shift);
+	umem_pgs = fill_umem_pbl_tbl(umem, pbl_tbl, order_base_2(page_size));
 	rc = bnxt_qplib_reg_mr(&rdev->qplib_res, &mr->qplib_mr, pbl_tbl,
-			       umem_pgs, false, 1 << page_shift);
+			       umem_pgs, false, page_size);
 	if (rc) {
 		ibdev_err(&rdev->ibdev, "Failed to register user MR");
 		goto fail;

drivers/infiniband/hw/bnxt_re/ib_verbs.h

@@ -163,12 +163,12 @@ int bnxt_re_query_gid(struct ib_device *ibdev, u8 port_num,
 enum rdma_link_layer bnxt_re_get_link_layer(struct ib_device *ibdev,
 					    u8 port_num);
 int bnxt_re_alloc_pd(struct ib_pd *pd, struct ib_udata *udata);
-void bnxt_re_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata);
+int bnxt_re_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata);
 int bnxt_re_create_ah(struct ib_ah *ah, struct rdma_ah_init_attr *init_attr,
 		      struct ib_udata *udata);
 int bnxt_re_modify_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
 int bnxt_re_query_ah(struct ib_ah *ah, struct rdma_ah_attr *ah_attr);
-void bnxt_re_destroy_ah(struct ib_ah *ah, u32 flags);
+int bnxt_re_destroy_ah(struct ib_ah *ah, u32 flags);
 int bnxt_re_create_srq(struct ib_srq *srq,
 		       struct ib_srq_init_attr *srq_init_attr,
 		       struct ib_udata *udata);
@@ -176,7 +176,7 @@ int bnxt_re_modify_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr,
 		       enum ib_srq_attr_mask srq_attr_mask,
 		       struct ib_udata *udata);
 int bnxt_re_query_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr);
-void bnxt_re_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
+int bnxt_re_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
 int bnxt_re_post_srq_recv(struct ib_srq *srq, const struct ib_recv_wr *recv_wr,
 			  const struct ib_recv_wr **bad_recv_wr);
 struct ib_qp *bnxt_re_create_qp(struct ib_pd *pd,
@@ -193,7 +193,7 @@ int bnxt_re_post_recv(struct ib_qp *qp, const struct ib_recv_wr *recv_wr,
 		      const struct ib_recv_wr **bad_recv_wr);
 int bnxt_re_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		      struct ib_udata *udata);
-void bnxt_re_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
+int bnxt_re_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
 int bnxt_re_poll_cq(struct ib_cq *cq, int num_entries, struct ib_wc *wc);
 int bnxt_re_req_notify_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags);
 struct ib_mr *bnxt_re_get_dma_mr(struct ib_pd *pd, int mr_access_flags);

drivers/infiniband/hw/bnxt_re/main.c

@@ -736,7 +736,8 @@ static int bnxt_re_register_ib(struct bnxt_re_dev *rdev)
 	if (ret)
 		return ret;
 
-	return ib_register_device(ibdev, "bnxt_re%d");
+	dma_set_max_seg_size(&rdev->en_dev->pdev->dev, UINT_MAX);
+	return ib_register_device(ibdev, "bnxt_re%d", &rdev->en_dev->pdev->dev);
 }
 
 static void bnxt_re_dev_remove(struct bnxt_re_dev *rdev)

drivers/infiniband/hw/bnxt_re/qplib_fp.c

@@ -295,9 +295,9 @@ static void __wait_for_all_nqes(struct bnxt_qplib_cq *cq, u16 cnq_events)
 	}
 }
 
-static void bnxt_qplib_service_nq(unsigned long data)
+static void bnxt_qplib_service_nq(struct tasklet_struct *t)
 {
-	struct bnxt_qplib_nq *nq = (struct bnxt_qplib_nq *)data;
+	struct bnxt_qplib_nq *nq = from_tasklet(nq, t, nq_tasklet);
 	struct bnxt_qplib_hwq *hwq = &nq->hwq;
 	int num_srqne_processed = 0;
 	int num_cqne_processed = 0;
@@ -448,8 +448,7 @@ int bnxt_qplib_nq_start_irq(struct bnxt_qplib_nq *nq, int nq_indx,
 
 	nq->msix_vec = msix_vector;
 	if (need_init)
-		tasklet_init(&nq->nq_tasklet, bnxt_qplib_service_nq,
+		tasklet_setup(&nq->nq_tasklet, bnxt_qplib_service_nq);
-			     (unsigned long)nq);
 	else
 		tasklet_enable(&nq->nq_tasklet);
 

drivers/infiniband/hw/bnxt_re/qplib_rcfw.c

@@ -50,7 +50,7 @@
 #include "qplib_sp.h"
 #include "qplib_fp.h"
 
-static void bnxt_qplib_service_creq(unsigned long data);
+static void bnxt_qplib_service_creq(struct tasklet_struct *t);
 
 /* Hardware communication channel */
 static int __wait_for_resp(struct bnxt_qplib_rcfw *rcfw, u16 cookie)
@@ -79,7 +79,7 @@ static int __block_for_resp(struct bnxt_qplib_rcfw *rcfw, u16 cookie)
 		goto done;
 	do {
 		mdelay(1); /* 1m sec */
-		bnxt_qplib_service_creq((unsigned long)rcfw);
+		bnxt_qplib_service_creq(&rcfw->creq.creq_tasklet);
 	} while (test_bit(cbit, cmdq->cmdq_bitmap) && --count);
 done:
 	return count ? 0 : -ETIMEDOUT;
@@ -370,9 +370,9 @@ static int bnxt_qplib_process_qp_event(struct bnxt_qplib_rcfw *rcfw,
 }
 
 /* SP - CREQ Completion handlers */
-static void bnxt_qplib_service_creq(unsigned long data)
+static void bnxt_qplib_service_creq(struct tasklet_struct *t)
 {
-	struct bnxt_qplib_rcfw *rcfw = (struct bnxt_qplib_rcfw *)data;
+	struct bnxt_qplib_rcfw *rcfw = from_tasklet(rcfw, t, creq.creq_tasklet);
 	struct bnxt_qplib_creq_ctx *creq = &rcfw->creq;
 	u32 type, budget = CREQ_ENTRY_POLL_BUDGET;
 	struct bnxt_qplib_hwq *hwq = &creq->hwq;
@@ -687,8 +687,7 @@ int bnxt_qplib_rcfw_start_irq(struct bnxt_qplib_rcfw *rcfw, int msix_vector,
 
 	creq->msix_vec = msix_vector;
 	if (need_init)
-		tasklet_init(&creq->creq_tasklet,
+		tasklet_setup(&creq->creq_tasklet, bnxt_qplib_service_creq);
-			     bnxt_qplib_service_creq, (unsigned long)rcfw);
 	else
 		tasklet_enable(&creq->creq_tasklet);
 	rc = request_irq(creq->msix_vec, bnxt_qplib_creq_irq, 0,

drivers/infiniband/hw/bnxt_re/qplib_res.c

@@ -45,6 +45,9 @@
 #include <linux/dma-mapping.h>
 #include <linux/if_vlan.h>
 #include <linux/vmalloc.h>
+#include <rdma/ib_verbs.h>
+#include <rdma/ib_umem.h>
+
 #include "roce_hsi.h"
 #include "qplib_res.h"
 #include "qplib_sp.h"
@@ -87,12 +90,11 @@ static void __free_pbl(struct bnxt_qplib_res *res, struct bnxt_qplib_pbl *pbl,
 static void bnxt_qplib_fill_user_dma_pages(struct bnxt_qplib_pbl *pbl,
 					   struct bnxt_qplib_sg_info *sginfo)
 {
-	struct scatterlist *sghead = sginfo->sghead;
+	struct ib_block_iter biter;
-	struct sg_dma_page_iter sg_iter;
 	int i = 0;
 
-	for_each_sg_dma_page(sghead, &sg_iter, sginfo->nmap, 0) {
+	rdma_umem_for_each_dma_block(sginfo->umem, &biter, sginfo->pgsize) {
-		pbl->pg_map_arr[i] = sg_page_iter_dma_address(&sg_iter);
+		pbl->pg_map_arr[i] = rdma_block_iter_dma_address(&biter);
 		pbl->pg_arr[i] = NULL;
 		pbl->pg_count++;
 		i++;
@@ -104,15 +106,16 @@ static int __alloc_pbl(struct bnxt_qplib_res *res,
 		       struct bnxt_qplib_sg_info *sginfo)
 {
 	struct pci_dev *pdev = res->pdev;
-	struct scatterlist *sghead;
 	bool is_umem = false;
 	u32 pages;
 	int i;
 
 	if (sginfo->nopte)
 		return 0;
-	pages = sginfo->npages;
+	if (sginfo->umem)
-	sghead = sginfo->sghead;
+		pages = ib_umem_num_dma_blocks(sginfo->umem, sginfo->pgsize);
+	else
+		pages = sginfo->npages;
 	/* page ptr arrays */
 	pbl->pg_arr = vmalloc(pages * sizeof(void *));
 	if (!pbl->pg_arr)
@@ -127,7 +130,7 @@ static int __alloc_pbl(struct bnxt_qplib_res *res,
 	pbl->pg_count = 0;
 	pbl->pg_size = sginfo->pgsize;
 
-	if (!sghead) {
+	if (!sginfo->umem) {
 		for (i = 0; i < pages; i++) {
 			pbl->pg_arr[i] = dma_alloc_coherent(&pdev->dev,
 							    pbl->pg_size,
@@ -183,14 +186,12 @@ int bnxt_qplib_alloc_init_hwq(struct bnxt_qplib_hwq *hwq,
 	struct bnxt_qplib_sg_info sginfo = {};
 	u32 depth, stride, npbl, npde;
 	dma_addr_t *src_phys_ptr, **dst_virt_ptr;
-	struct scatterlist *sghead = NULL;
 	struct bnxt_qplib_res *res;
 	struct pci_dev *pdev;
 	int i, rc, lvl;
 
 	res = hwq_attr->res;
 	pdev = res->pdev;
-	sghead = hwq_attr->sginfo->sghead;
 	pg_size = hwq_attr->sginfo->pgsize;
 	hwq->level = PBL_LVL_MAX;
 
@@ -204,7 +205,7 @@ int bnxt_qplib_alloc_init_hwq(struct bnxt_qplib_hwq *hwq,
 			aux_pages++;
 	}
 
-	if (!sghead) {
+	if (!hwq_attr->sginfo->umem) {
 		hwq->is_user = false;
 		npages = (depth * stride) / pg_size + aux_pages;
 		if ((depth * stride) % pg_size)
@@ -213,11 +214,14 @@ int bnxt_qplib_alloc_init_hwq(struct bnxt_qplib_hwq *hwq,
 			return -EINVAL;
 		hwq_attr->sginfo->npages = npages;
 	} else {
+		unsigned long sginfo_num_pages = ib_umem_num_dma_blocks(
+			hwq_attr->sginfo->umem, hwq_attr->sginfo->pgsize);
+
 		hwq->is_user = true;
-		npages = hwq_attr->sginfo->npages;
+		npages = sginfo_num_pages;
 		npages = (npages * PAGE_SIZE) /
 			  BIT_ULL(hwq_attr->sginfo->pgshft);
-		if ((hwq_attr->sginfo->npages * PAGE_SIZE) %
+		if ((sginfo_num_pages * PAGE_SIZE) %
 		     BIT_ULL(hwq_attr->sginfo->pgshft))
 			if (!npages)
 				npages++;

drivers/infiniband/hw/bnxt_re/qplib_res.h

@@ -126,8 +126,7 @@ struct bnxt_qplib_pbl {
 };
 
 struct bnxt_qplib_sg_info {
-	struct scatterlist		*sghead;
+	struct ib_umem			*umem;
-	u32				nmap;
 	u32				npages;
 	u32				pgshft;
 	u32				pgsize;

drivers/infiniband/hw/cxgb4/cm.c

@@ -77,9 +77,9 @@ static int enable_ecn;
 module_param(enable_ecn, int, 0644);
 MODULE_PARM_DESC(enable_ecn, "Enable ECN (default=0/disabled)");
 
-static int dack_mode = 1;
+static int dack_mode;
 module_param(dack_mode, int, 0644);
-MODULE_PARM_DESC(dack_mode, "Delayed ack mode (default=1)");
+MODULE_PARM_DESC(dack_mode, "Delayed ack mode (default=0)");
 
 uint c4iw_max_read_depth = 32;
 module_param(c4iw_max_read_depth, int, 0644);

drivers/infiniband/hw/cxgb4/cq.c

@@ -967,7 +967,7 @@ int c4iw_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc)
 	return !err || err == -ENODATA ? npolled : err;
 }
 
-void c4iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
+int c4iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 {
 	struct c4iw_cq *chp;
 	struct c4iw_ucontext *ucontext;
@@ -985,6 +985,7 @@ void c4iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 		   ucontext ? &ucontext->uctx : &chp->cq.rdev->uctx,
 		   chp->destroy_skb, chp->wr_waitp);
 	c4iw_put_wr_wait(chp->wr_waitp);
+	return 0;
 }
 
 int c4iw_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,

drivers/infiniband/hw/cxgb4/iw_cxgb4.h

@@ -985,21 +985,20 @@ int c4iw_map_mr_sg(struct ib_mr *ibmr, struct scatterlist *sg, int sg_nents,
 		   unsigned int *sg_offset);
 int c4iw_dealloc_mw(struct ib_mw *mw);
 void c4iw_dealloc(struct uld_ctx *ctx);
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int c4iw_alloc_mw(struct ib_mw *mw, struct ib_udata *udata);
-			    struct ib_udata *udata);
 struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start,
 					   u64 length, u64 virt, int acc,
 					   struct ib_udata *udata);
 struct ib_mr *c4iw_get_dma_mr(struct ib_pd *pd, int acc);
 int c4iw_dereg_mr(struct ib_mr *ib_mr, struct ib_udata *udata);
-void c4iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata);
+int c4iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata);
 int c4iw_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		   struct ib_udata *udata);
 int c4iw_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
 int c4iw_modify_srq(struct ib_srq *ib_srq, struct ib_srq_attr *attr,
 		    enum ib_srq_attr_mask srq_attr_mask,
 		    struct ib_udata *udata);
-void c4iw_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata);
+int c4iw_destroy_srq(struct ib_srq *ib_srq, struct ib_udata *udata);
 int c4iw_create_srq(struct ib_srq *srq, struct ib_srq_init_attr *attrs,
 		    struct ib_udata *udata);
 int c4iw_destroy_qp(struct ib_qp *ib_qp, struct ib_udata *udata);

drivers/infiniband/hw/cxgb4/mem.c

@@ -510,7 +510,7 @@ struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	__be64 *pages;
 	int shift, n, i;
 	int err = -ENOMEM;
-	struct sg_dma_page_iter sg_iter;
+	struct ib_block_iter biter;
 	struct c4iw_dev *rhp;
 	struct c4iw_pd *php;
 	struct c4iw_mr *mhp;
@@ -548,7 +548,7 @@ struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 
 	shift = PAGE_SHIFT;
 
-	n = ib_umem_num_pages(mhp->umem);
+	n = ib_umem_num_dma_blocks(mhp->umem, 1 << shift);
 	err = alloc_pbl(mhp, n);
 	if (err)
 		goto err_umem_release;
@@ -561,8 +561,8 @@ struct ib_mr *c4iw_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 
 	i = n = 0;
 
-	for_each_sg_dma_page(mhp->umem->sg_head.sgl, &sg_iter, mhp->umem->nmap, 0) {
+	rdma_umem_for_each_dma_block(mhp->umem, &biter, 1 << shift) {
-		pages[i++] = cpu_to_be64(sg_page_iter_dma_address(&sg_iter));
+		pages[i++] = cpu_to_be64(rdma_block_iter_dma_address(&biter));
 		if (i == PAGE_SIZE / sizeof(*pages)) {
 			err = write_pbl(&mhp->rhp->rdev, pages,
 					mhp->attr.pbl_addr + (n << 3), i,
@@ -611,30 +611,23 @@ err_free_mhp:
 	return ERR_PTR(err);
 }
 
-struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int c4iw_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
-			    struct ib_udata *udata)
 {
+	struct c4iw_mw *mhp = to_c4iw_mw(ibmw);
 	struct c4iw_dev *rhp;
 	struct c4iw_pd *php;
-	struct c4iw_mw *mhp;
 	u32 mmid;
 	u32 stag = 0;
 	int ret;
 
-	if (type != IB_MW_TYPE_1)
+	if (ibmw->type != IB_MW_TYPE_1)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
-	php = to_c4iw_pd(pd);
+	php = to_c4iw_pd(ibmw->pd);
 	rhp = php->rhp;
-	mhp = kzalloc(sizeof(*mhp), GFP_KERNEL);
-	if (!mhp)
-		return ERR_PTR(-ENOMEM);
-
 	mhp->wr_waitp = c4iw_alloc_wr_wait(GFP_KERNEL);
-	if (!mhp->wr_waitp) {
+	if (!mhp->wr_waitp)
-		ret = -ENOMEM;
+		return -ENOMEM;
-		goto free_mhp;
-	}
 
 	mhp->dereg_skb = alloc_skb(SGE_MAX_WR_LEN, GFP_KERNEL);
 	if (!mhp->dereg_skb) {
@@ -645,18 +638,19 @@ struct ib_mw *c4iw_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 	ret = allocate_window(&rhp->rdev, &stag, php->pdid, mhp->wr_waitp);
 	if (ret)
 		goto free_skb;
+
 	mhp->rhp = rhp;
 	mhp->attr.pdid = php->pdid;
 	mhp->attr.type = FW_RI_STAG_MW;
 	mhp->attr.stag = stag;
 	mmid = (stag) >> 8;
-	mhp->ibmw.rkey = stag;
+	ibmw->rkey = stag;
 	if (xa_insert_irq(&rhp->mrs, mmid, mhp, GFP_KERNEL)) {
 		ret = -ENOMEM;
 		goto dealloc_win;
 	}
 	pr_debug("mmid 0x%x mhp %p stag 0x%x\n", mmid, mhp, stag);
-	return &(mhp->ibmw);
+	return 0;
 
 dealloc_win:
 	deallocate_window(&rhp->rdev, mhp->attr.stag, mhp->dereg_skb,
@@ -665,9 +659,7 @@ free_skb:
 	kfree_skb(mhp->dereg_skb);
 free_wr_wait:
 	c4iw_put_wr_wait(mhp->wr_waitp);
-free_mhp:
+	return ret;
-	kfree(mhp);
-	return ERR_PTR(ret);
 }
 
 int c4iw_dealloc_mw(struct ib_mw *mw)
@@ -684,8 +676,6 @@ int c4iw_dealloc_mw(struct ib_mw *mw)
 			  mhp->wr_waitp);
 	kfree_skb(mhp->dereg_skb);
 	c4iw_put_wr_wait(mhp->wr_waitp);
-	pr_debug("ib_mw %p mmid 0x%x ptr %p\n", mw, mmid, mhp);
-	kfree(mhp);
 	return 0;
 }
 

drivers/infiniband/hw/cxgb4/provider.c

@@ -190,7 +190,7 @@ static int c4iw_mmap(struct ib_ucontext *context, struct vm_area_struct *vma)
 	return ret;
 }
 
-static void c4iw_deallocate_pd(struct ib_pd *pd, struct ib_udata *udata)
+static int c4iw_deallocate_pd(struct ib_pd *pd, struct ib_udata *udata)
 {
 	struct c4iw_dev *rhp;
 	struct c4iw_pd *php;
@@ -202,6 +202,7 @@ static void c4iw_deallocate_pd(struct ib_pd *pd, struct ib_udata *udata)
 	mutex_lock(&rhp->rdev.stats.lock);
 	rhp->rdev.stats.pd.cur--;
 	mutex_unlock(&rhp->rdev.stats.lock);
+	return 0;
 }
 
 static int c4iw_allocate_pd(struct ib_pd *pd, struct ib_udata *udata)
@@ -497,8 +498,10 @@ static const struct ib_device_ops c4iw_dev_ops = {
 	.query_qp = c4iw_ib_query_qp,
 	.reg_user_mr = c4iw_reg_user_mr,
 	.req_notify_cq = c4iw_arm_cq,
-	INIT_RDMA_OBJ_SIZE(ib_pd, c4iw_pd, ibpd),
+
 	INIT_RDMA_OBJ_SIZE(ib_cq, c4iw_cq, ibcq),
+	INIT_RDMA_OBJ_SIZE(ib_mw, c4iw_mw, ibmw),
+	INIT_RDMA_OBJ_SIZE(ib_pd, c4iw_pd, ibpd),
 	INIT_RDMA_OBJ_SIZE(ib_srq, c4iw_srq, ibsrq),
 	INIT_RDMA_OBJ_SIZE(ib_ucontext, c4iw_ucontext, ibucontext),
 };
@@ -567,7 +570,9 @@ void c4iw_register_device(struct work_struct *work)
 	ret = set_netdevs(&dev->ibdev, &dev->rdev);
 	if (ret)
 		goto err_dealloc_ctx;
-	ret = ib_register_device(&dev->ibdev, "cxgb4_%d");
+	dma_set_max_seg_size(&dev->rdev.lldi.pdev->dev, UINT_MAX);
+	ret = ib_register_device(&dev->ibdev, "cxgb4_%d",
+				 &dev->rdev.lldi.pdev->dev);
 	if (ret)
 		goto err_dealloc_ctx;
 	return;

drivers/infiniband/hw/cxgb4/qp.c

@@ -2797,7 +2797,7 @@ err_free_wr_wait:
 	return ret;
 }
 
-void c4iw_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
+int c4iw_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
 {
 	struct c4iw_dev *rhp;
 	struct c4iw_srq *srq;
@@ -2813,4 +2813,5 @@ void c4iw_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
 		       srq->wr_waitp);
 	c4iw_free_srq_idx(&rhp->rdev, srq->idx);
 	c4iw_put_wr_wait(srq->wr_waitp);
+	return 0;
 }

drivers/infiniband/hw/efa/efa.h

@@ -33,7 +33,8 @@ struct efa_irq {
 	char name[EFA_IRQNAME_SIZE];
 };
 
-struct efa_sw_stats {
+/* Don't use anything other than atomic64 */
+struct efa_stats {
 	atomic64_t alloc_pd_err;
 	atomic64_t create_qp_err;
 	atomic64_t create_cq_err;
@@ -41,11 +42,6 @@ struct efa_sw_stats {
 	atomic64_t alloc_ucontext_err;
 	atomic64_t create_ah_err;
 	atomic64_t mmap_err;
-};
-
-/* Don't use anything other than atomic64 */
-struct efa_stats {
-	struct efa_sw_stats sw_stats;
 	atomic64_t keep_alive_rcvd;
 };
 
@@ -134,12 +130,12 @@ int efa_query_gid(struct ib_device *ibdev, u8 port, int index,
 int efa_query_pkey(struct ib_device *ibdev, u8 port, u16 index,
 		   u16 *pkey);
 int efa_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata);
-void efa_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata);
+int efa_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata);
 int efa_destroy_qp(struct ib_qp *ibqp, struct ib_udata *udata);
 struct ib_qp *efa_create_qp(struct ib_pd *ibpd,
 			    struct ib_qp_init_attr *init_attr,
 			    struct ib_udata *udata);
-void efa_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata);
+int efa_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata);
 int efa_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		  struct ib_udata *udata);
 struct ib_mr *efa_reg_mr(struct ib_pd *ibpd, u64 start, u64 length,
@@ -156,7 +152,7 @@ void efa_mmap_free(struct rdma_user_mmap_entry *rdma_entry);
 int efa_create_ah(struct ib_ah *ibah,
 		  struct rdma_ah_init_attr *init_attr,
 		  struct ib_udata *udata);
-void efa_destroy_ah(struct ib_ah *ibah, u32 flags);
+int efa_destroy_ah(struct ib_ah *ibah, u32 flags);
 int efa_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 		  int qp_attr_mask, struct ib_udata *udata);
 enum rdma_link_layer efa_port_link_layer(struct ib_device *ibdev,

drivers/infiniband/hw/efa/efa_admin_cmds_defs.h

@@ -61,6 +61,8 @@ enum efa_admin_qp_state {
 
 enum efa_admin_get_stats_type {
 	EFA_ADMIN_GET_STATS_TYPE_BASIC              = 0,
+	EFA_ADMIN_GET_STATS_TYPE_MESSAGES           = 1,
+	EFA_ADMIN_GET_STATS_TYPE_RDMA_READ          = 2,
 };
 
 enum efa_admin_get_stats_scope {
@@ -68,14 +70,6 @@ enum efa_admin_get_stats_scope {
 	EFA_ADMIN_GET_STATS_SCOPE_QUEUE             = 1,
 };
 
-enum efa_admin_modify_qp_mask_bits {
-	EFA_ADMIN_QP_STATE_BIT                      = 0,
-	EFA_ADMIN_CUR_QP_STATE_BIT                  = 1,
-	EFA_ADMIN_QKEY_BIT                          = 2,
-	EFA_ADMIN_SQ_PSN_BIT                        = 3,
-	EFA_ADMIN_SQ_DRAINED_ASYNC_NOTIFY_BIT       = 4,
-};
-
 /*
  * QP allocation sizes, converted by fabric QueuePair (QP) create command
  * from QP capabilities.
@@ -199,8 +193,14 @@ struct efa_admin_modify_qp_cmd {
 	struct efa_admin_aq_common_desc aq_common_desc;
 
 	/*
-	 * Mask indicating which fields should be updated see enum
+	 * Mask indicating which fields should be updated
-	 * efa_admin_modify_qp_mask_bits
+	 * 0 : qp_state
+	 * 1 : cur_qp_state
+	 * 2 : qkey
+	 * 3 : sq_psn
+	 * 4 : sq_drained_async_notify
+	 * 5 : rnr_retry
+	 * 31:6 : reserved
 	 */
 	u32 modify_mask;
 
@@ -222,8 +222,8 @@ struct efa_admin_modify_qp_cmd {
 	/* Enable async notification when SQ is drained */
 	u8 sq_drained_async_notify;
 
-	/* MBZ */
+	/* Number of RNR retries (valid only for SRD QPs) */
-	u8 reserved1;
+	u8 rnr_retry;
 
 	/* MBZ */
 	u16 reserved2;
@@ -258,8 +258,8 @@ struct efa_admin_query_qp_resp {
 	/* Indicates that draining is in progress */
 	u8 sq_draining;
 
-	/* MBZ */
+	/* Number of RNR retries (valid only for SRD QPs) */
-	u8 reserved1;
+	u8 rnr_retry;
 
 	/* MBZ */
 	u16 reserved2;
@@ -530,10 +530,36 @@ struct efa_admin_basic_stats {
 	u64 rx_drops;
 };
 
+struct efa_admin_messages_stats {
+	u64 send_bytes;
+
+	u64 send_wrs;
+
+	u64 recv_bytes;
+
+	u64 recv_wrs;
+};
+
+struct efa_admin_rdma_read_stats {
+	u64 read_wrs;
+
+	u64 read_bytes;
+
+	u64 read_wr_err;
+
+	u64 read_resp_bytes;
+};
+
 struct efa_admin_acq_get_stats_resp {
 	struct efa_admin_acq_common_desc acq_common_desc;
 
-	struct efa_admin_basic_stats basic_stats;
+	union {
+		struct efa_admin_basic_stats basic_stats;
+
+		struct efa_admin_messages_stats messages_stats;
+
+		struct efa_admin_rdma_read_stats rdma_read_stats;
+	} u;
 };
 
 struct efa_admin_get_set_feature_common_desc {
@@ -576,7 +602,9 @@ struct efa_admin_feature_device_attr_desc {
 	/*
 	 * 0 : rdma_read - If set, RDMA Read is supported on
 	 *    TX queues
-	 * 31:1 : reserved - MBZ
+	 * 1 : rnr_retry - If set, RNR retry is supported on
+	 *    modify QP command
+	 * 31:2 : reserved - MBZ
 	 */
 	u32 device_caps;
 
@@ -862,6 +890,14 @@ struct efa_admin_host_info {
 #define EFA_ADMIN_CREATE_QP_CMD_SQ_VIRT_MASK                BIT(0)
 #define EFA_ADMIN_CREATE_QP_CMD_RQ_VIRT_MASK                BIT(1)
 
+/* modify_qp_cmd */
+#define EFA_ADMIN_MODIFY_QP_CMD_QP_STATE_MASK               BIT(0)
+#define EFA_ADMIN_MODIFY_QP_CMD_CUR_QP_STATE_MASK           BIT(1)
+#define EFA_ADMIN_MODIFY_QP_CMD_QKEY_MASK                   BIT(2)
+#define EFA_ADMIN_MODIFY_QP_CMD_SQ_PSN_MASK                 BIT(3)
+#define EFA_ADMIN_MODIFY_QP_CMD_SQ_DRAINED_ASYNC_NOTIFY_MASK BIT(4)
+#define EFA_ADMIN_MODIFY_QP_CMD_RNR_RETRY_MASK              BIT(5)
+
 /* reg_mr_cmd */
 #define EFA_ADMIN_REG_MR_CMD_PHYS_PAGE_SIZE_SHIFT_MASK      GENMASK(4, 0)
 #define EFA_ADMIN_REG_MR_CMD_MEM_ADDR_PHY_MODE_EN_MASK      BIT(7)
@@ -878,6 +914,7 @@ struct efa_admin_host_info {
 
 /* feature_device_attr_desc */
 #define EFA_ADMIN_FEATURE_DEVICE_ATTR_DESC_RDMA_READ_MASK   BIT(0)
+#define EFA_ADMIN_FEATURE_DEVICE_ATTR_DESC_RNR_RETRY_MASK   BIT(1)
 
 /* host_info */
 #define EFA_ADMIN_HOST_INFO_DRIVER_MODULE_TYPE_MASK         GENMASK(7, 0)

drivers/infiniband/hw/efa/efa_com_cmd.c

@@ -76,6 +76,7 @@ int efa_com_modify_qp(struct efa_com_dev *edev,
 	cmd.qkey = params->qkey;
 	cmd.sq_psn = params->sq_psn;
 	cmd.sq_drained_async_notify = params->sq_drained_async_notify;
+	cmd.rnr_retry = params->rnr_retry;
 
 	err = efa_com_cmd_exec(aq,
 			       (struct efa_admin_aq_entry *)&cmd,
@@ -121,6 +122,7 @@ int efa_com_query_qp(struct efa_com_dev *edev,
 	result->qkey = resp.qkey;
 	result->sq_draining = resp.sq_draining;
 	result->sq_psn = resp.sq_psn;
+	result->rnr_retry = resp.rnr_retry;
 
 	return 0;
 }
@@ -750,11 +752,27 @@ int efa_com_get_stats(struct efa_com_dev *edev,
 		return err;
 	}
 
-	result->basic_stats.tx_bytes = resp.basic_stats.tx_bytes;
+	switch (cmd.type) {
-	result->basic_stats.tx_pkts = resp.basic_stats.tx_pkts;
+	case EFA_ADMIN_GET_STATS_TYPE_BASIC:
-	result->basic_stats.rx_bytes = resp.basic_stats.rx_bytes;
+		result->basic_stats.tx_bytes = resp.u.basic_stats.tx_bytes;
-	result->basic_stats.rx_pkts = resp.basic_stats.rx_pkts;
+		result->basic_stats.tx_pkts = resp.u.basic_stats.tx_pkts;
-	result->basic_stats.rx_drops = resp.basic_stats.rx_drops;
+		result->basic_stats.rx_bytes = resp.u.basic_stats.rx_bytes;
+		result->basic_stats.rx_pkts = resp.u.basic_stats.rx_pkts;
+		result->basic_stats.rx_drops = resp.u.basic_stats.rx_drops;
+		break;
+	case EFA_ADMIN_GET_STATS_TYPE_MESSAGES:
+		result->messages_stats.send_bytes = resp.u.messages_stats.send_bytes;
+		result->messages_stats.send_wrs = resp.u.messages_stats.send_wrs;
+		result->messages_stats.recv_bytes = resp.u.messages_stats.recv_bytes;
+		result->messages_stats.recv_wrs = resp.u.messages_stats.recv_wrs;
+		break;
+	case EFA_ADMIN_GET_STATS_TYPE_RDMA_READ:
+		result->rdma_read_stats.read_wrs = resp.u.rdma_read_stats.read_wrs;
+		result->rdma_read_stats.read_bytes = resp.u.rdma_read_stats.read_bytes;
+		result->rdma_read_stats.read_wr_err = resp.u.rdma_read_stats.read_wr_err;
+		result->rdma_read_stats.read_resp_bytes = resp.u.rdma_read_stats.read_resp_bytes;
+		break;
+	}
 
 	return 0;
 }

drivers/infiniband/hw/efa/efa_com_cmd.h

@@ -47,6 +47,7 @@ struct efa_com_modify_qp_params {
 	u32 qkey;
 	u32 sq_psn;
 	u8 sq_drained_async_notify;
+	u8 rnr_retry;
 };
 
 struct efa_com_query_qp_params {
@@ -58,6 +59,7 @@ struct efa_com_query_qp_result {
 	u32 qkey;
 	u32 sq_draining;
 	u32 sq_psn;
+	u8 rnr_retry;
 };
 
 struct efa_com_destroy_qp_params {
@@ -238,8 +240,24 @@ struct efa_com_basic_stats {
 	u64 rx_drops;
 };
 
+struct efa_com_messages_stats {
+	u64 send_bytes;
+	u64 send_wrs;
+	u64 recv_bytes;
+	u64 recv_wrs;
+};
+
+struct efa_com_rdma_read_stats {
+	u64 read_wrs;
+	u64 read_bytes;
+	u64 read_wr_err;
+	u64 read_resp_bytes;
+};
+
 union efa_com_get_stats_result {
 	struct efa_com_basic_stats basic_stats;
+	struct efa_com_messages_stats messages_stats;
+	struct efa_com_rdma_read_stats rdma_read_stats;
 };
 
 void efa_com_set_dma_addr(dma_addr_t addr, u32 *addr_high, u32 *addr_low);

drivers/infiniband/hw/efa/efa_main.c

@@ -331,7 +331,7 @@ static int efa_ib_device_add(struct efa_dev *dev)
 
 	ib_set_device_ops(&dev->ibdev, &efa_dev_ops);
 
-	err = ib_register_device(&dev->ibdev, "efa_%d");
+	err = ib_register_device(&dev->ibdev, "efa_%d", &pdev->dev);
 	if (err)
 		goto err_release_doorbell_bar;
 
@@ -418,7 +418,7 @@ static int efa_device_init(struct efa_com_dev *edev, struct pci_dev *pdev)
 			err);
 		return err;
 	}
-
+	dma_set_max_seg_size(&pdev->dev, UINT_MAX);
 	return 0;
 }
 

drivers/infiniband/hw/efa/efa_verbs.c

@@ -4,6 +4,7 @@
  */
 
 #include <linux/vmalloc.h>
+#include <linux/log2.h>
 
 #include <rdma/ib_addr.h>
 #include <rdma/ib_umem.h>
@@ -35,6 +36,14 @@ struct efa_user_mmap_entry {
 	op(EFA_RX_BYTES, "rx_bytes") \
 	op(EFA_RX_PKTS, "rx_pkts") \
 	op(EFA_RX_DROPS, "rx_drops") \
+	op(EFA_SEND_BYTES, "send_bytes") \
+	op(EFA_SEND_WRS, "send_wrs") \
+	op(EFA_RECV_BYTES, "recv_bytes") \
+	op(EFA_RECV_WRS, "recv_wrs") \
+	op(EFA_RDMA_READ_WRS, "rdma_read_wrs") \
+	op(EFA_RDMA_READ_BYTES, "rdma_read_bytes") \
+	op(EFA_RDMA_READ_WR_ERR, "rdma_read_wr_err") \
+	op(EFA_RDMA_READ_RESP_BYTES, "rdma_read_resp_bytes") \
 	op(EFA_SUBMITTED_CMDS, "submitted_cmds") \
 	op(EFA_COMPLETED_CMDS, "completed_cmds") \
 	op(EFA_CMDS_ERR, "cmds_err") \
@@ -142,10 +151,9 @@ to_emmap(struct rdma_user_mmap_entry *rdma_entry)
 	return container_of(rdma_entry, struct efa_user_mmap_entry, rdma_entry);
 }
 
-static inline bool is_rdma_read_cap(struct efa_dev *dev)
+#define EFA_DEV_CAP(dev, cap) \
-{
+	((dev)->dev_attr.device_caps & \
-	return dev->dev_attr.device_caps & EFA_ADMIN_FEATURE_DEVICE_ATTR_DESC_RDMA_READ_MASK;
+	 EFA_ADMIN_FEATURE_DEVICE_ATTR_DESC_##cap##_MASK)
-}
 
 #define is_reserved_cleared(reserved) \
 	!memchr_inv(reserved, 0, sizeof(reserved))
@@ -221,9 +229,12 @@ int efa_query_device(struct ib_device *ibdev,
 		resp.max_rq_wr = dev_attr->max_rq_depth;
 		resp.max_rdma_size = dev_attr->max_rdma_size;
 
-		if (is_rdma_read_cap(dev))
+		if (EFA_DEV_CAP(dev, RDMA_READ))
 			resp.device_caps |= EFA_QUERY_DEVICE_CAPS_RDMA_READ;
 
+		if (EFA_DEV_CAP(dev, RNR_RETRY))
+			resp.device_caps |= EFA_QUERY_DEVICE_CAPS_RNR_RETRY;
+
 		err = ib_copy_to_udata(udata, &resp,
 				       min(sizeof(resp), udata->outlen));
 		if (err) {
@@ -269,7 +280,7 @@ int efa_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 
 #define EFA_QUERY_QP_SUPP_MASK \
 	(IB_QP_STATE | IB_QP_PKEY_INDEX | IB_QP_PORT | \
-	 IB_QP_QKEY | IB_QP_SQ_PSN | IB_QP_CAP)
+	 IB_QP_QKEY | IB_QP_SQ_PSN | IB_QP_CAP | IB_QP_RNR_RETRY)
 
 	if (qp_attr_mask & ~EFA_QUERY_QP_SUPP_MASK) {
 		ibdev_dbg(&dev->ibdev,
@@ -291,6 +302,7 @@ int efa_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 	qp_attr->sq_psn = result.sq_psn;
 	qp_attr->sq_draining = result.sq_draining;
 	qp_attr->port_num = 1;
+	qp_attr->rnr_retry = result.rnr_retry;
 
 	qp_attr->cap.max_send_wr = qp->max_send_wr;
 	qp_attr->cap.max_recv_wr = qp->max_recv_wr;
@@ -376,17 +388,18 @@ int efa_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 err_dealloc_pd:
 	efa_pd_dealloc(dev, result.pdn);
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.alloc_pd_err);
+	atomic64_inc(&dev->stats.alloc_pd_err);
 	return err;
 }
 
-void efa_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
+int efa_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 {
 	struct efa_dev *dev = to_edev(ibpd->device);
 	struct efa_pd *pd = to_epd(ibpd);
 
 	ibdev_dbg(&dev->ibdev, "Dealloc pd[%d]\n", pd->pdn);
 	efa_pd_dealloc(dev, pd->pdn);
+	return 0;
 }
 
 static int efa_destroy_qp_handle(struct efa_dev *dev, u32 qp_handle)
@@ -737,18 +750,130 @@ err_free_mapped:
 err_free_qp:
 	kfree(qp);
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.create_qp_err);
+	atomic64_inc(&dev->stats.create_qp_err);
 	return ERR_PTR(err);
 }
 
+static const struct {
+	int			valid;
+	enum ib_qp_attr_mask	req_param;
+	enum ib_qp_attr_mask	opt_param;
+} srd_qp_state_table[IB_QPS_ERR + 1][IB_QPS_ERR + 1] = {
+	[IB_QPS_RESET] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_INIT]  = {
+			.valid = 1,
+			.req_param = IB_QP_PKEY_INDEX |
+				     IB_QP_PORT |
+				     IB_QP_QKEY,
+		},
+	},
+	[IB_QPS_INIT] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+		[IB_QPS_INIT]  = {
+			.valid = 1,
+			.opt_param = IB_QP_PKEY_INDEX |
+				     IB_QP_PORT |
+				     IB_QP_QKEY,
+		},
+		[IB_QPS_RTR]   = {
+			.valid = 1,
+			.opt_param = IB_QP_PKEY_INDEX |
+				     IB_QP_QKEY,
+		},
+	},
+	[IB_QPS_RTR] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+		[IB_QPS_RTS]   = {
+			.valid = 1,
+			.req_param = IB_QP_SQ_PSN,
+			.opt_param = IB_QP_CUR_STATE |
+				     IB_QP_QKEY |
+				     IB_QP_RNR_RETRY,
+
+		}
+	},
+	[IB_QPS_RTS] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+		[IB_QPS_RTS]   = {
+			.valid = 1,
+			.opt_param = IB_QP_CUR_STATE |
+				     IB_QP_QKEY,
+		},
+		[IB_QPS_SQD] = {
+			.valid = 1,
+			.opt_param = IB_QP_EN_SQD_ASYNC_NOTIFY,
+		},
+	},
+	[IB_QPS_SQD] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+		[IB_QPS_RTS]   = {
+			.valid = 1,
+			.opt_param = IB_QP_CUR_STATE |
+				     IB_QP_QKEY,
+		},
+		[IB_QPS_SQD] = {
+			.valid = 1,
+			.opt_param = IB_QP_PKEY_INDEX |
+				     IB_QP_QKEY,
+		}
+	},
+	[IB_QPS_SQE] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+		[IB_QPS_RTS]   = {
+			.valid = 1,
+			.opt_param = IB_QP_CUR_STATE |
+				     IB_QP_QKEY,
+		}
+	},
+	[IB_QPS_ERR] = {
+		[IB_QPS_RESET] = { .valid = 1 },
+		[IB_QPS_ERR]   = { .valid = 1 },
+	}
+};
+
+static bool efa_modify_srd_qp_is_ok(enum ib_qp_state cur_state,
+				    enum ib_qp_state next_state,
+				    enum ib_qp_attr_mask mask)
+{
+	enum ib_qp_attr_mask req_param, opt_param;
+
+	if (mask & IB_QP_CUR_STATE  &&
+	    cur_state != IB_QPS_RTR && cur_state != IB_QPS_RTS &&
+	    cur_state != IB_QPS_SQD && cur_state != IB_QPS_SQE)
+		return false;
+
+	if (!srd_qp_state_table[cur_state][next_state].valid)
+		return false;
+
+	req_param = srd_qp_state_table[cur_state][next_state].req_param;
+	opt_param = srd_qp_state_table[cur_state][next_state].opt_param;
+
+	if ((mask & req_param) != req_param)
+		return false;
+
+	if (mask & ~(req_param | opt_param | IB_QP_STATE))
+		return false;
+
+	return true;
+}
+
 static int efa_modify_qp_validate(struct efa_dev *dev, struct efa_qp *qp,
 				  struct ib_qp_attr *qp_attr, int qp_attr_mask,
 				  enum ib_qp_state cur_state,
 				  enum ib_qp_state new_state)
 {
+	int err;
+
 #define EFA_MODIFY_QP_SUPP_MASK \
 	(IB_QP_STATE | IB_QP_CUR_STATE | IB_QP_EN_SQD_ASYNC_NOTIFY | \
-	 IB_QP_PKEY_INDEX | IB_QP_PORT | IB_QP_QKEY | IB_QP_SQ_PSN)
+	 IB_QP_PKEY_INDEX | IB_QP_PORT | IB_QP_QKEY | IB_QP_SQ_PSN | \
+	 IB_QP_RNR_RETRY)
 
 	if (qp_attr_mask & ~EFA_MODIFY_QP_SUPP_MASK) {
 		ibdev_dbg(&dev->ibdev,
@@ -757,8 +882,14 @@ static int efa_modify_qp_validate(struct efa_dev *dev, struct efa_qp *qp,
 		return -EOPNOTSUPP;
 	}
 
-	if (!ib_modify_qp_is_ok(cur_state, new_state, IB_QPT_UD,
+	if (qp->ibqp.qp_type == IB_QPT_DRIVER)
-				qp_attr_mask)) {
+		err = !efa_modify_srd_qp_is_ok(cur_state, new_state,
+					       qp_attr_mask);
+	else
+		err = !ib_modify_qp_is_ok(cur_state, new_state, IB_QPT_UD,
+					  qp_attr_mask);
+
+	if (err) {
 		ibdev_dbg(&dev->ibdev, "Invalid modify QP parameters\n");
 		return -EINVAL;
 	}
@@ -805,28 +936,36 @@ int efa_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 	params.qp_handle = qp->qp_handle;
 
 	if (qp_attr_mask & IB_QP_STATE) {
-		params.modify_mask |= BIT(EFA_ADMIN_QP_STATE_BIT) |
+		EFA_SET(&params.modify_mask, EFA_ADMIN_MODIFY_QP_CMD_QP_STATE,
-				      BIT(EFA_ADMIN_CUR_QP_STATE_BIT);
+			1);
+		EFA_SET(&params.modify_mask,
+			EFA_ADMIN_MODIFY_QP_CMD_CUR_QP_STATE, 1);
 		params.cur_qp_state = qp_attr->cur_qp_state;
 		params.qp_state = qp_attr->qp_state;
 	}
 
 	if (qp_attr_mask & IB_QP_EN_SQD_ASYNC_NOTIFY) {
-		params.modify_mask |=
+		EFA_SET(&params.modify_mask,
-			BIT(EFA_ADMIN_SQ_DRAINED_ASYNC_NOTIFY_BIT);
+			EFA_ADMIN_MODIFY_QP_CMD_SQ_DRAINED_ASYNC_NOTIFY, 1);
 		params.sq_drained_async_notify = qp_attr->en_sqd_async_notify;
 	}
 
 	if (qp_attr_mask & IB_QP_QKEY) {
-		params.modify_mask |= BIT(EFA_ADMIN_QKEY_BIT);
+		EFA_SET(&params.modify_mask, EFA_ADMIN_MODIFY_QP_CMD_QKEY, 1);
 		params.qkey = qp_attr->qkey;
 	}
 
 	if (qp_attr_mask & IB_QP_SQ_PSN) {
-		params.modify_mask |= BIT(EFA_ADMIN_SQ_PSN_BIT);
+		EFA_SET(&params.modify_mask, EFA_ADMIN_MODIFY_QP_CMD_SQ_PSN, 1);
 		params.sq_psn = qp_attr->sq_psn;
 	}
 
+	if (qp_attr_mask & IB_QP_RNR_RETRY) {
+		EFA_SET(&params.modify_mask, EFA_ADMIN_MODIFY_QP_CMD_RNR_RETRY,
+			1);
+		params.rnr_retry = qp_attr->rnr_retry;
+	}
+
 	err = efa_com_modify_qp(&dev->edev, &params);
 	if (err)
 		return err;
@@ -843,7 +982,7 @@ static int efa_destroy_cq_idx(struct efa_dev *dev, int cq_idx)
 	return efa_com_destroy_cq(&dev->edev, &params);
 }
 
-void efa_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
+int efa_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
 {
 	struct efa_dev *dev = to_edev(ibcq->device);
 	struct efa_cq *cq = to_ecq(ibcq);
@@ -856,6 +995,7 @@ void efa_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
 	efa_destroy_cq_idx(dev, cq->cq_idx);
 	efa_free_mapped(dev, cq->cpu_addr, cq->dma_addr, cq->size,
 			DMA_FROM_DEVICE);
+	return 0;
 }
 
 static int cq_mmap_entries_setup(struct efa_dev *dev, struct efa_cq *cq,
@@ -996,7 +1136,7 @@ err_free_mapped:
 			DMA_FROM_DEVICE);
 
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.create_cq_err);
+	atomic64_inc(&dev->stats.create_cq_err);
 	return err;
 }
 
@@ -1013,8 +1153,7 @@ static int umem_to_page_list(struct efa_dev *dev,
 	ibdev_dbg(&dev->ibdev, "hp_cnt[%u], pages_in_hp[%u]\n",
 		  hp_cnt, pages_in_hp);
 
-	rdma_for_each_block(umem->sg_head.sgl, &biter, umem->nmap,
+	rdma_umem_for_each_dma_block(umem, &biter, BIT(hp_shift))
-			    BIT(hp_shift))
 		page_list[hp_idx++] = rdma_block_iter_dma_address(&biter);
 
 	return 0;
@@ -1026,7 +1165,7 @@ static struct scatterlist *efa_vmalloc_buf_to_sg(u64 *buf, int page_cnt)
 	struct page *pg;
 	int i;
 
-	sglist = kcalloc(page_cnt, sizeof(*sglist), GFP_KERNEL);
+	sglist = kmalloc_array(page_cnt, sizeof(*sglist), GFP_KERNEL);
 	if (!sglist)
 		return NULL;
 	sg_init_table(sglist, page_cnt);
@@ -1370,7 +1509,7 @@ struct ib_mr *efa_reg_mr(struct ib_pd *ibpd, u64 start, u64 length,
 
 	supp_access_flags =
 		IB_ACCESS_LOCAL_WRITE |
-		(is_rdma_read_cap(dev) ? IB_ACCESS_REMOTE_READ : 0);
+		(EFA_DEV_CAP(dev, RDMA_READ) ? IB_ACCESS_REMOTE_READ : 0);
 
 	access_flags &= ~IB_ACCESS_OPTIONAL;
 	if (access_flags & ~supp_access_flags) {
@@ -1410,9 +1549,8 @@ struct ib_mr *efa_reg_mr(struct ib_pd *ibpd, u64 start, u64 length,
 		goto err_unmap;
 	}
 
-	params.page_shift = __ffs(pg_sz);
+	params.page_shift = order_base_2(pg_sz);
-	params.page_num = DIV_ROUND_UP(length + (start & (pg_sz - 1)),
+	params.page_num = ib_umem_num_dma_blocks(mr->umem, pg_sz);
-				       pg_sz);
 
 	ibdev_dbg(&dev->ibdev,
 		  "start %#llx length %#llx params.page_shift %u params.page_num %u\n",
@@ -1451,7 +1589,7 @@ err_unmap:
 err_free:
 	kfree(mr);
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.reg_mr_err);
+	atomic64_inc(&dev->stats.reg_mr_err);
 	return ERR_PTR(err);
 }
 
@@ -1569,19 +1707,17 @@ int efa_alloc_ucontext(struct ib_ucontext *ibucontext, struct ib_udata *udata)
 	resp.max_tx_batch = dev->dev_attr.max_tx_batch;
 	resp.min_sq_wr = dev->dev_attr.min_sq_depth;
 
-	if (udata && udata->outlen) {
+	err = ib_copy_to_udata(udata, &resp,
-		err = ib_copy_to_udata(udata, &resp,
+			       min(sizeof(resp), udata->outlen));
-				       min(sizeof(resp), udata->outlen));
+	if (err)
-		if (err)
+		goto err_dealloc_uar;
-			goto err_dealloc_uar;
-	}
 
 	return 0;
 
 err_dealloc_uar:
 	efa_dealloc_uar(dev, result.uarn);
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.alloc_ucontext_err);
+	atomic64_inc(&dev->stats.alloc_ucontext_err);
 	return err;
 }
 
@@ -1614,7 +1750,7 @@ static int __efa_mmap(struct efa_dev *dev, struct efa_ucontext *ucontext,
 		ibdev_dbg(&dev->ibdev,
 			  "pgoff[%#lx] does not have valid entry\n",
 			  vma->vm_pgoff);
-		atomic64_inc(&dev->stats.sw_stats.mmap_err);
+		atomic64_inc(&dev->stats.mmap_err);
 		return -EINVAL;
 	}
 	entry = to_emmap(rdma_entry);
@@ -1656,7 +1792,7 @@ static int __efa_mmap(struct efa_dev *dev, struct efa_ucontext *ucontext,
 			"Couldn't mmap address[%#llx] length[%#zx] mmap_flag[%d] err[%d]\n",
 			entry->address, rdma_entry->npages * PAGE_SIZE,
 			entry->mmap_flag, err);
-		atomic64_inc(&dev->stats.sw_stats.mmap_err);
+		atomic64_inc(&dev->stats.mmap_err);
 	}
 
 	rdma_user_mmap_entry_put(rdma_entry);
@@ -1741,11 +1877,11 @@ int efa_create_ah(struct ib_ah *ibah,
 err_destroy_ah:
 	efa_ah_destroy(dev, ah);
 err_out:
-	atomic64_inc(&dev->stats.sw_stats.create_ah_err);
+	atomic64_inc(&dev->stats.create_ah_err);
 	return err;
 }
 
-void efa_destroy_ah(struct ib_ah *ibah, u32 flags)
+int efa_destroy_ah(struct ib_ah *ibah, u32 flags)
 {
 	struct efa_dev *dev = to_edev(ibah->pd->device);
 	struct efa_ah *ah = to_eah(ibah);
@@ -1755,10 +1891,11 @@ void efa_destroy_ah(struct ib_ah *ibah, u32 flags)
 	if (!(flags & RDMA_DESTROY_AH_SLEEPABLE)) {
 		ibdev_dbg(&dev->ibdev,
 			  "Destroy address handle is not supported in atomic context\n");
-		return;
+		return -EOPNOTSUPP;
 	}
 
 	efa_ah_destroy(dev, ah);
+	return 0;
 }
 
 struct rdma_hw_stats *efa_alloc_hw_stats(struct ib_device *ibdev, u8 port_num)
@@ -1774,13 +1911,15 @@ int efa_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,
 	struct efa_com_get_stats_params params = {};
 	union efa_com_get_stats_result result;
 	struct efa_dev *dev = to_edev(ibdev);
+	struct efa_com_rdma_read_stats *rrs;
+	struct efa_com_messages_stats *ms;
 	struct efa_com_basic_stats *bs;
 	struct efa_com_stats_admin *as;
 	struct efa_stats *s;
 	int err;
 
-	params.type = EFA_ADMIN_GET_STATS_TYPE_BASIC;
 	params.scope = EFA_ADMIN_GET_STATS_SCOPE_ALL;
+	params.type = EFA_ADMIN_GET_STATS_TYPE_BASIC;
 
 	err = efa_com_get_stats(&dev->edev, &params, &result);
 	if (err)
@@ -1793,6 +1932,28 @@ int efa_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,
 	stats->value[EFA_RX_PKTS] = bs->rx_pkts;
 	stats->value[EFA_RX_DROPS] = bs->rx_drops;
 
+	params.type = EFA_ADMIN_GET_STATS_TYPE_MESSAGES;
+	err = efa_com_get_stats(&dev->edev, &params, &result);
+	if (err)
+		return err;
+
+	ms = &result.messages_stats;
+	stats->value[EFA_SEND_BYTES] = ms->send_bytes;
+	stats->value[EFA_SEND_WRS] = ms->send_wrs;
+	stats->value[EFA_RECV_BYTES] = ms->recv_bytes;
+	stats->value[EFA_RECV_WRS] = ms->recv_wrs;
+
+	params.type = EFA_ADMIN_GET_STATS_TYPE_RDMA_READ;
+	err = efa_com_get_stats(&dev->edev, &params, &result);
+	if (err)
+		return err;
+
+	rrs = &result.rdma_read_stats;
+	stats->value[EFA_RDMA_READ_WRS] = rrs->read_wrs;
+	stats->value[EFA_RDMA_READ_BYTES] = rrs->read_bytes;
+	stats->value[EFA_RDMA_READ_WR_ERR] = rrs->read_wr_err;
+	stats->value[EFA_RDMA_READ_RESP_BYTES] = rrs->read_resp_bytes;
+
 	as = &dev->edev.aq.stats;
 	stats->value[EFA_SUBMITTED_CMDS] = atomic64_read(&as->submitted_cmd);
 	stats->value[EFA_COMPLETED_CMDS] = atomic64_read(&as->completed_cmd);
@@ -1801,13 +1962,14 @@ int efa_get_hw_stats(struct ib_device *ibdev, struct rdma_hw_stats *stats,
 
 	s = &dev->stats;
 	stats->value[EFA_KEEP_ALIVE_RCVD] = atomic64_read(&s->keep_alive_rcvd);
-	stats->value[EFA_ALLOC_PD_ERR] = atomic64_read(&s->sw_stats.alloc_pd_err);
+	stats->value[EFA_ALLOC_PD_ERR] = atomic64_read(&s->alloc_pd_err);
-	stats->value[EFA_CREATE_QP_ERR] = atomic64_read(&s->sw_stats.create_qp_err);
+	stats->value[EFA_CREATE_QP_ERR] = atomic64_read(&s->create_qp_err);
-	stats->value[EFA_CREATE_CQ_ERR] = atomic64_read(&s->sw_stats.create_cq_err);
+	stats->value[EFA_CREATE_CQ_ERR] = atomic64_read(&s->create_cq_err);
-	stats->value[EFA_REG_MR_ERR] = atomic64_read(&s->sw_stats.reg_mr_err);
+	stats->value[EFA_REG_MR_ERR] = atomic64_read(&s->reg_mr_err);
-	stats->value[EFA_ALLOC_UCONTEXT_ERR] = atomic64_read(&s->sw_stats.alloc_ucontext_err);
+	stats->value[EFA_ALLOC_UCONTEXT_ERR] =
-	stats->value[EFA_CREATE_AH_ERR] = atomic64_read(&s->sw_stats.create_ah_err);
+		atomic64_read(&s->alloc_ucontext_err);
-	stats->value[EFA_MMAP_ERR] = atomic64_read(&s->sw_stats.mmap_err);
+	stats->value[EFA_CREATE_AH_ERR] = atomic64_read(&s->create_ah_err);
+	stats->value[EFA_MMAP_ERR] = atomic64_read(&s->mmap_err);
 
 	return ARRAY_SIZE(efa_stats_names);
 }

drivers/infiniband/hw/hfi1/sdma.c

@@ -232,11 +232,11 @@ static const struct sdma_set_state_action sdma_action_table[] = {
 static void sdma_complete(struct kref *);
 static void sdma_finalput(struct sdma_state *);
 static void sdma_get(struct sdma_state *);
-static void sdma_hw_clean_up_task(unsigned long);
+static void sdma_hw_clean_up_task(struct tasklet_struct *);
 static void sdma_put(struct sdma_state *);
 static void sdma_set_state(struct sdma_engine *, enum sdma_states);
 static void sdma_start_hw_clean_up(struct sdma_engine *);
-static void sdma_sw_clean_up_task(unsigned long);
+static void sdma_sw_clean_up_task(struct tasklet_struct *);
 static void sdma_sendctrl(struct sdma_engine *, unsigned);
 static void init_sdma_regs(struct sdma_engine *, u32, uint);
 static void sdma_process_event(
@@ -545,9 +545,10 @@ static void sdma_err_progress_check(struct timer_list *t)
 	schedule_work(&sde->err_halt_worker);
 }
 
-static void sdma_hw_clean_up_task(unsigned long opaque)
+static void sdma_hw_clean_up_task(struct tasklet_struct *t)
 {
-	struct sdma_engine *sde = (struct sdma_engine *)opaque;
+	struct sdma_engine *sde = from_tasklet(sde, t,
+					       sdma_hw_clean_up_task);
 	u64 statuscsr;
 
 	while (1) {
@@ -604,9 +605,9 @@ static void sdma_flush_descq(struct sdma_engine *sde)
 		sdma_desc_avail(sde, sdma_descq_freecnt(sde));
 }
 
-static void sdma_sw_clean_up_task(unsigned long opaque)
+static void sdma_sw_clean_up_task(struct tasklet_struct *t)
 {
-	struct sdma_engine *sde = (struct sdma_engine *)opaque;
+	struct sdma_engine *sde = from_tasklet(sde, t, sdma_sw_clean_up_task);
 	unsigned long flags;
 
 	spin_lock_irqsave(&sde->tail_lock, flags);
@@ -1454,11 +1455,10 @@ int sdma_init(struct hfi1_devdata *dd, u8 port)
 		sde->tail_csr =
 			get_kctxt_csr_addr(dd, this_idx, SD(TAIL));
 
-		tasklet_init(&sde->sdma_hw_clean_up_task, sdma_hw_clean_up_task,
+		tasklet_setup(&sde->sdma_hw_clean_up_task,
-			     (unsigned long)sde);
+			      sdma_hw_clean_up_task);
-
+		tasklet_setup(&sde->sdma_sw_clean_up_task,
-		tasklet_init(&sde->sdma_sw_clean_up_task, sdma_sw_clean_up_task,
+			      sdma_sw_clean_up_task);
-			     (unsigned long)sde);
 		INIT_WORK(&sde->err_halt_worker, sdma_err_halt_wait);
 		INIT_WORK(&sde->flush_worker, sdma_field_flush);
 

drivers/infiniband/hw/hfi1/verbs.c

@@ -1424,7 +1424,7 @@ static int query_port(struct rvt_dev_info *rdi, u8 port_num,
 	props->gid_tbl_len = HFI1_GUIDS_PER_PORT;
 	props->active_width = (u8)opa_width_to_ib(ppd->link_width_active);
 	/* see rate_show() in ib core/sysfs.c */
-	props->active_speed = (u8)opa_speed_to_ib(ppd->link_speed_active);
+	props->active_speed = opa_speed_to_ib(ppd->link_speed_active);
 	props->max_vl_num = ppd->vls_supported;
 
 	/* Once we are a "first class" citizen and have added the OPA MTUs to

drivers/infiniband/hw/hns/hns_roce_ah.c

@@ -39,6 +39,22 @@
 #define HNS_ROCE_VLAN_SL_BIT_MASK	7
 #define HNS_ROCE_VLAN_SL_SHIFT		13
 
+static inline u16 get_ah_udp_sport(const struct rdma_ah_attr *ah_attr)
+{
+	u32 fl = ah_attr->grh.flow_label;
+	u16 sport;
+
+	if (!fl)
+		sport = get_random_u32() %
+			(IB_ROCE_UDP_ENCAP_VALID_PORT_MAX + 1 -
+			 IB_ROCE_UDP_ENCAP_VALID_PORT_MIN) +
+			IB_ROCE_UDP_ENCAP_VALID_PORT_MIN;
+	else
+		sport = rdma_flow_label_to_udp_sport(fl);
+
+	return sport;
+}
+
 int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 		       struct ib_udata *udata)
 {
@@ -79,6 +95,8 @@ int hns_roce_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 
 	memcpy(ah->av.dgid, grh->dgid.raw, HNS_ROCE_GID_SIZE);
 	ah->av.sl = rdma_ah_get_sl(ah_attr);
+	ah->av.flowlabel = grh->flow_label;
+	ah->av.udp_sport = get_ah_udp_sport(ah_attr);
 
 	return 0;
 }
@@ -98,8 +116,3 @@ int hns_roce_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr)
 
 	return 0;
 }
-
-void hns_roce_destroy_ah(struct ib_ah *ah, u32 flags)
-{
-	return;
-}

drivers/infiniband/hw/hns/hns_roce_alloc.c

@@ -268,8 +268,7 @@ int hns_roce_get_umem_bufs(struct hns_roce_dev *hr_dev, dma_addr_t *bufs,
 	}
 
 	/* convert system page cnt to hw page cnt */
-	rdma_for_each_block(umem->sg_head.sgl, &biter, umem->nmap,
+	rdma_umem_for_each_dma_block(umem, &biter, 1 << page_shift) {
-			    1 << page_shift) {
 		addr = rdma_block_iter_dma_address(&biter);
 		if (idx >= start) {
 			bufs[total++] = addr;

drivers/infiniband/hw/hns/hns_roce_cq.c

@@ -150,7 +150,7 @@ static int alloc_cq_buf(struct hns_roce_dev *hr_dev, struct hns_roce_cq *hr_cq,
 	int err;
 
 	buf_attr.page_shift = hr_dev->caps.cqe_buf_pg_sz + HNS_HW_PAGE_SHIFT;
-	buf_attr.region[0].size = hr_cq->cq_depth * hr_dev->caps.cq_entry_sz;
+	buf_attr.region[0].size = hr_cq->cq_depth * hr_cq->cqe_size;
 	buf_attr.region[0].hopnum = hr_dev->caps.cqe_hop_num;
 	buf_attr.region_count = 1;
 	buf_attr.fixed_page = true;
@@ -224,6 +224,21 @@ static void free_cq_db(struct hns_roce_dev *hr_dev, struct hns_roce_cq *hr_cq,
 	}
 }
 
+static void set_cqe_size(struct hns_roce_cq *hr_cq, struct ib_udata *udata,
+			 struct hns_roce_ib_create_cq *ucmd)
+{
+	struct hns_roce_dev *hr_dev = to_hr_dev(hr_cq->ib_cq.device);
+
+	if (udata) {
+		if (udata->inlen >= offsetofend(typeof(*ucmd), cqe_size))
+			hr_cq->cqe_size = ucmd->cqe_size;
+		else
+			hr_cq->cqe_size = HNS_ROCE_V2_CQE_SIZE;
+	} else {
+		hr_cq->cqe_size = hr_dev->caps.cqe_sz;
+	}
+}
+
 int hns_roce_create_cq(struct ib_cq *ib_cq, const struct ib_cq_init_attr *attr,
 		       struct ib_udata *udata)
 {
@@ -258,7 +273,8 @@ int hns_roce_create_cq(struct ib_cq *ib_cq, const struct ib_cq_init_attr *attr,
 	INIT_LIST_HEAD(&hr_cq->rq_list);
 
 	if (udata) {
-		ret = ib_copy_from_udata(&ucmd, udata, sizeof(ucmd));
+		ret = ib_copy_from_udata(&ucmd, udata,
+					 min(sizeof(ucmd), udata->inlen));
 		if (ret) {
 			ibdev_err(ibdev, "Failed to copy CQ udata, err %d\n",
 				  ret);
@@ -266,6 +282,8 @@ int hns_roce_create_cq(struct ib_cq *ib_cq, const struct ib_cq_init_attr *attr,
 		}
 	}
 
+	set_cqe_size(hr_cq, udata, &ucmd);
+
 	ret = alloc_cq_buf(hr_dev, hr_cq, udata, ucmd.buf_addr);
 	if (ret) {
 		ibdev_err(ibdev, "Failed to alloc CQ buf, err %d\n", ret);
@@ -287,7 +305,7 @@ int hns_roce_create_cq(struct ib_cq *ib_cq, const struct ib_cq_init_attr *attr,
 	/*
 	 * For the QP created by kernel space, tptr value should be initialized
 	 * to zero; For the QP created by user space, it will cause synchronous
-	 * problems if tptr is set to zero here, so we initialze it in user
+	 * problems if tptr is set to zero here, so we initialize it in user
 	 * space.
 	 */
 	if (!udata && hr_cq->tptr_addr)
@@ -311,7 +329,7 @@ err_cq_buf:
 	return ret;
 }
 
-void hns_roce_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
+int hns_roce_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 {
 	struct hns_roce_dev *hr_dev = to_hr_dev(ib_cq->device);
 	struct hns_roce_cq *hr_cq = to_hr_cq(ib_cq);
@@ -322,6 +340,7 @@ void hns_roce_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 	free_cq_buf(hr_dev, hr_cq);
 	free_cq_db(hr_dev, hr_cq, udata);
 	free_cqc(hr_dev, hr_cq);
+	return 0;
 }
 
 void hns_roce_cq_completion(struct hns_roce_dev *hr_dev, u32 cqn)

drivers/infiniband/hw/hns/hns_roce_device.h

@@ -37,8 +37,8 @@
 
 #define DRV_NAME "hns_roce"
 
-/* hip08 is a pci device */
 #define PCI_REVISION_ID_HIP08			0x21
+#define PCI_REVISION_ID_HIP09			0x30
 
 #define HNS_ROCE_HW_VER1	('h' << 24 | 'i' << 16 | '0' << 8 | '6')
 
@@ -57,7 +57,6 @@
 /* Hardware specification only for v1 engine */
 #define HNS_ROCE_MAX_INNER_MTPT_NUM		0x7
 #define HNS_ROCE_MAX_MTPT_PBL_NUM		0x100000
-#define HNS_ROCE_MAX_SGE_NUM			2
 
 #define HNS_ROCE_EACH_FREE_CQ_WAIT_MSECS	20
 #define HNS_ROCE_MAX_FREE_CQ_WAIT_CNT	\
@@ -76,15 +75,18 @@
 #define HNS_ROCE_CEQ				0
 #define HNS_ROCE_AEQ				1
 
-#define HNS_ROCE_CEQ_ENTRY_SIZE			0x4
+#define HNS_ROCE_CEQE_SIZE 0x4
-#define HNS_ROCE_AEQ_ENTRY_SIZE			0x10
+#define HNS_ROCE_AEQE_SIZE 0x10
 
-#define HNS_ROCE_SL_SHIFT			28
+#define HNS_ROCE_V3_EQE_SIZE 0x40
-#define HNS_ROCE_TCLASS_SHIFT			20
+
-#define HNS_ROCE_FLOW_LABEL_MASK		0xfffff
+#define HNS_ROCE_V2_CQE_SIZE 32
+#define HNS_ROCE_V3_CQE_SIZE 64
+
+#define HNS_ROCE_V2_QPC_SZ 256
+#define HNS_ROCE_V3_QPC_SZ 512
 
 #define HNS_ROCE_MAX_PORTS			6
-#define HNS_ROCE_MAX_GID_NUM			16
 #define HNS_ROCE_GID_SIZE			16
 #define HNS_ROCE_SGE_SIZE			16
 
@@ -112,8 +114,6 @@
 #define PAGES_SHIFT_24				24
 #define PAGES_SHIFT_32				32
 
-#define HNS_ROCE_PCI_BAR_NUM			2
-
 #define HNS_ROCE_IDX_QUE_ENTRY_SZ		4
 #define SRQ_DB_REG				0x230
 
@@ -467,6 +467,7 @@ struct hns_roce_cq {
 	void __iomem			*cq_db_l;
 	u16				*tptr_addr;
 	int				arm_sn;
+	int				cqe_size;
 	unsigned long			cqn;
 	u32				vector;
 	atomic_t			refcount;
@@ -535,17 +536,18 @@ struct hns_roce_raq_table {
 };
 
 struct hns_roce_av {
-	u8          port;
+	u8 port;
-	u8          gid_index;
+	u8 gid_index;
-	u8          stat_rate;
+	u8 stat_rate;
-	u8          hop_limit;
+	u8 hop_limit;
-	u32         flowlabel;
+	u32 flowlabel;
-	u8          sl;
+	u16 udp_sport;
-	u8          tclass;
+	u8 sl;
-	u8          dgid[HNS_ROCE_GID_SIZE];
+	u8 tclass;
-	u8          mac[ETH_ALEN];
+	u8 dgid[HNS_ROCE_GID_SIZE];
-	u16         vlan_id;
+	u8 mac[ETH_ALEN];
-	bool	    vlan_en;
+	u16 vlan_id;
+	bool vlan_en;
 };
 
 struct hns_roce_ah {
@@ -655,6 +657,8 @@ struct hns_roce_qp {
 
 	struct hns_roce_sge	sge;
 	u32			next_sge;
+	enum ib_mtu		path_mtu;
+	u32			max_inline_data;
 
 	/* 0: flush needed, 1: unneeded */
 	unsigned long		flush_flag;
@@ -678,7 +682,8 @@ enum {
 };
 
 struct hns_roce_ceqe {
-	__le32			comp;
+	__le32	comp;
+	__le32	rsv[15];
 };
 
 struct hns_roce_aeqe {
@@ -715,6 +720,7 @@ struct hns_roce_aeqe {
 			u8	rsv0;
 		} __packed cmd;
 	 } event;
+	__le32 rsv[12];
 };
 
 struct hns_roce_eq {
@@ -791,15 +797,15 @@ struct hns_roce_caps {
 	int		num_pds;
 	int		reserved_pds;
 	u32		mtt_entry_sz;
-	u32		cq_entry_sz;
+	u32		cqe_sz;
 	u32		page_size_cap;
 	u32		reserved_lkey;
 	int		mtpt_entry_sz;
-	int		qpc_entry_sz;
+	int		qpc_sz;
 	int		irrl_entry_sz;
 	int		trrl_entry_sz;
 	int		cqc_entry_sz;
-	int		sccc_entry_sz;
+	int		sccc_sz;
 	int		qpc_timer_entry_sz;
 	int		cqc_timer_entry_sz;
 	int		srqc_entry_sz;
@@ -809,6 +815,8 @@ struct hns_roce_caps {
 	u32		pbl_hop_num;
 	int		aeqe_depth;
 	int		ceqe_depth;
+	u32		aeqe_size;
+	u32		ceqe_size;
 	enum ib_mtu	max_mtu;
 	u32		qpc_bt_num;
 	u32		qpc_timer_bt_num;
@@ -930,7 +938,7 @@ struct hns_roce_hw {
 	int (*poll_cq)(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
 	int (*dereg_mr)(struct hns_roce_dev *hr_dev, struct hns_roce_mr *mr,
 			struct ib_udata *udata);
-	void (*destroy_cq)(struct ib_cq *ibcq, struct ib_udata *udata);
+	int (*destroy_cq)(struct ib_cq *ibcq, struct ib_udata *udata);
 	int (*modify_cq)(struct ib_cq *cq, u16 cq_count, u16 cq_period);
 	int (*init_eq)(struct hns_roce_dev *hr_dev);
 	void (*cleanup_eq)(struct hns_roce_dev *hr_dev);
@@ -1178,10 +1186,13 @@ void hns_roce_bitmap_free_range(struct hns_roce_bitmap *bitmap,
 int hns_roce_create_ah(struct ib_ah *ah, struct rdma_ah_init_attr *init_attr,
 		       struct ib_udata *udata);
 int hns_roce_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr);
-void hns_roce_destroy_ah(struct ib_ah *ah, u32 flags);
+static inline int hns_roce_destroy_ah(struct ib_ah *ah, u32 flags)
+{
+	return 0;
+}
 
 int hns_roce_alloc_pd(struct ib_pd *pd, struct ib_udata *udata);
-void hns_roce_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata);
+int hns_roce_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata);
 
 struct ib_mr *hns_roce_get_dma_mr(struct ib_pd *pd, int acc);
 struct ib_mr *hns_roce_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
@@ -1200,8 +1211,7 @@ int hns_roce_hw_destroy_mpt(struct hns_roce_dev *hr_dev,
 			    unsigned long mpt_index);
 unsigned long key_to_hw_index(u32 key);
 
-struct ib_mw *hns_roce_alloc_mw(struct ib_pd *pd, enum ib_mw_type,
+int hns_roce_alloc_mw(struct ib_mw *mw, struct ib_udata *udata);
-				struct ib_udata *udata);
 int hns_roce_dealloc_mw(struct ib_mw *ibmw);
 
 void hns_roce_buf_free(struct hns_roce_dev *hr_dev, struct hns_roce_buf *buf);
@@ -1220,7 +1230,7 @@ int hns_roce_create_srq(struct ib_srq *srq,
 int hns_roce_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr,
 			enum ib_srq_attr_mask srq_attr_mask,
 			struct ib_udata *udata);
-void hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata);
+int hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata);
 
 struct ib_qp *hns_roce_create_qp(struct ib_pd *ib_pd,
 				 struct ib_qp_init_attr *init_attr,
@@ -1247,7 +1257,7 @@ int to_hr_qp_type(int qp_type);
 int hns_roce_create_cq(struct ib_cq *ib_cq, const struct ib_cq_init_attr *attr,
 		       struct ib_udata *udata);
 
-void hns_roce_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata);
+int hns_roce_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata);
 int hns_roce_db_map_user(struct hns_roce_ucontext *context,
 			 struct ib_udata *udata, unsigned long virt,
 			 struct hns_roce_db *db);

drivers/infiniband/hw/hns/hns_roce_hem.c

@@ -338,8 +338,8 @@ static int hns_roce_set_hem(struct hns_roce_dev *hr_dev,
 	void __iomem *bt_cmd;
 	__le32 bt_cmd_val[2];
 	__le32 bt_cmd_h = 0;
-	__le32 bt_cmd_l = 0;
+	__le32 bt_cmd_l;
-	u64 bt_ba = 0;
+	u64 bt_ba;
 	int ret = 0;
 
 	/* Find the HEM(Hardware Entry Memory) entry */
@@ -1027,7 +1027,7 @@ void hns_roce_cleanup_hem(struct hns_roce_dev *hr_dev)
 	if (hr_dev->caps.cqc_timer_entry_sz)
 		hns_roce_cleanup_hem_table(hr_dev,
 					   &hr_dev->cqc_timer_table);
-	if (hr_dev->caps.sccc_entry_sz)
+	if (hr_dev->caps.sccc_sz)
 		hns_roce_cleanup_hem_table(hr_dev,
 					   &hr_dev->qp_table.sccc_table);
 	if (hr_dev->caps.trrl_entry_sz)
@@ -1404,7 +1404,7 @@ int hns_roce_hem_list_request(struct hns_roce_dev *hr_dev,
 {
 	const struct hns_roce_buf_region *r;
 	int ofs, end;
-	int ret = 0;
+	int ret;
 	int unit;
 	int i;
 

drivers/infiniband/hw/hns/hns_roce_hw_v1.c

@@ -70,15 +70,15 @@ static int hns_roce_v1_post_send(struct ib_qp *ibqp,
 	struct hns_roce_qp *qp = to_hr_qp(ibqp);
 	struct device *dev = &hr_dev->pdev->dev;
 	struct hns_roce_sq_db sq_db = {};
-	int ps_opcode = 0, i = 0;
+	int ps_opcode, i;
 	unsigned long flags = 0;
 	void *wqe = NULL;
 	__le32 doorbell[2];
-	u32 wqe_idx = 0;
-	int nreq = 0;
 	int ret = 0;
-	u8 *smac;
 	int loopback;
+	u32 wqe_idx;
+	int nreq;
+	u8 *smac;
 
 	if (unlikely(ibqp->qp_type != IB_QPT_GSI &&
 		ibqp->qp_type != IB_QPT_RC)) {
@@ -271,7 +271,6 @@ static int hns_roce_v1_post_send(struct ib_qp *ibqp,
 				ps_opcode = HNS_ROCE_WQE_OPCODE_SEND;
 				break;
 			case IB_WR_LOCAL_INV:
-				break;
 			case IB_WR_ATOMIC_CMP_AND_SWP:
 			case IB_WR_ATOMIC_FETCH_AND_ADD:
 			case IB_WR_LSO:
@@ -888,7 +887,7 @@ static int hns_roce_db_init(struct hns_roce_dev *hr_dev)
 	u32 odb_ext_mod;
 	u32 sdb_evt_mod;
 	u32 odb_evt_mod;
-	int ret = 0;
+	int ret;
 
 	memset(db, 0, sizeof(*db));
 
@@ -1148,8 +1147,8 @@ static int hns_roce_raq_init(struct hns_roce_dev *hr_dev)
 	struct hns_roce_v1_priv *priv = hr_dev->priv;
 	struct hns_roce_raq_table *raq = &priv->raq_table;
 	struct device *dev = &hr_dev->pdev->dev;
-	int raq_shift = 0;
 	dma_addr_t addr;
+	int raq_shift;
 	__le32 tmp;
 	u32 val;
 	int ret;
@@ -1360,7 +1359,7 @@ static int hns_roce_free_mr_init(struct hns_roce_dev *hr_dev)
 	struct hns_roce_v1_priv *priv = hr_dev->priv;
 	struct hns_roce_free_mr *free_mr = &priv->free_mr;
 	struct device *dev = &hr_dev->pdev->dev;
-	int ret = 0;
+	int ret;
 
 	free_mr->free_mr_wq = create_singlethread_workqueue("hns_roce_free_mr");
 	if (!free_mr->free_mr_wq) {
@@ -1440,8 +1439,8 @@ static int hns_roce_v1_reset(struct hns_roce_dev *hr_dev, bool dereset)
 
 static int hns_roce_v1_profile(struct hns_roce_dev *hr_dev)
 {
-	int i = 0;
 	struct hns_roce_caps *caps = &hr_dev->caps;
+	int i;
 
 	hr_dev->vendor_id = roce_read(hr_dev, ROCEE_VENDOR_ID_REG);
 	hr_dev->vendor_part_id = roce_read(hr_dev, ROCEE_VENDOR_PART_ID_REG);
@@ -1471,12 +1470,12 @@ static int hns_roce_v1_profile(struct hns_roce_dev *hr_dev)
 	caps->max_qp_dest_rdma	= HNS_ROCE_V1_MAX_QP_DEST_RDMA;
 	caps->max_sq_desc_sz	= HNS_ROCE_V1_MAX_SQ_DESC_SZ;
 	caps->max_rq_desc_sz	= HNS_ROCE_V1_MAX_RQ_DESC_SZ;
-	caps->qpc_entry_sz	= HNS_ROCE_V1_QPC_ENTRY_SIZE;
+	caps->qpc_sz		= HNS_ROCE_V1_QPC_SIZE;
 	caps->irrl_entry_sz	= HNS_ROCE_V1_IRRL_ENTRY_SIZE;
 	caps->cqc_entry_sz	= HNS_ROCE_V1_CQC_ENTRY_SIZE;
 	caps->mtpt_entry_sz	= HNS_ROCE_V1_MTPT_ENTRY_SIZE;
 	caps->mtt_entry_sz	= HNS_ROCE_V1_MTT_ENTRY_SIZE;
-	caps->cq_entry_sz	= HNS_ROCE_V1_CQE_ENTRY_SIZE;
+	caps->cqe_sz		= HNS_ROCE_V1_CQE_SIZE;
 	caps->page_size_cap	= HNS_ROCE_V1_PAGE_SIZE_SUPPORT;
 	caps->reserved_lkey	= 0;
 	caps->reserved_pds	= 0;
@@ -1643,7 +1642,7 @@ static int hns_roce_v1_chk_mbox(struct hns_roce_dev *hr_dev,
 				unsigned long timeout)
 {
 	u8 __iomem *hcr = hr_dev->reg_base + ROCEE_MB1_REG;
-	unsigned long end = 0;
+	unsigned long end;
 	u32 status = 0;
 
 	end = msecs_to_jiffies(timeout) + jiffies;
@@ -1671,7 +1670,7 @@ static int hns_roce_v1_set_gid(struct hns_roce_dev *hr_dev, u8 port,
 {
 	unsigned long flags;
 	u32 *p = NULL;
-	u8 gid_idx = 0;
+	u8 gid_idx;
 
 	gid_idx = hns_get_gid_index(hr_dev, port, gid_index);
 
@@ -1897,8 +1896,7 @@ static int hns_roce_v1_write_mtpt(struct hns_roce_dev *hr_dev, void *mb_buf,
 
 static void *get_cqe(struct hns_roce_cq *hr_cq, int n)
 {
-	return hns_roce_buf_offset(hr_cq->mtr.kmem,
+	return hns_roce_buf_offset(hr_cq->mtr.kmem, n * HNS_ROCE_V1_CQE_SIZE);
-				   n * HNS_ROCE_V1_CQE_ENTRY_SIZE);
 }
 
 static void *get_sw_cqe(struct hns_roce_cq *hr_cq, int n)
@@ -2445,7 +2443,7 @@ static int hns_roce_v1_qp_modify(struct hns_roce_dev *hr_dev,
 
 	struct hns_roce_cmd_mailbox *mailbox;
 	struct device *dev = &hr_dev->pdev->dev;
-	int ret = 0;
+	int ret;
 
 	if (cur_state >= HNS_ROCE_QP_NUM_STATE ||
 	    new_state >= HNS_ROCE_QP_NUM_STATE ||
@@ -3394,7 +3392,7 @@ static int hns_roce_v1_q_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 	struct hns_roce_qp *hr_qp = to_hr_qp(ibqp);
 	struct device *dev = &hr_dev->pdev->dev;
 	struct hns_roce_qp_context *context;
-	int tmp_qp_state = 0;
+	int tmp_qp_state;
 	int ret = 0;
 	int state;
 
@@ -3572,7 +3570,7 @@ int hns_roce_v1_destroy_qp(struct ib_qp *ibqp, struct ib_udata *udata)
 	return 0;
 }
 
-static void hns_roce_v1_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
+static int hns_roce_v1_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
 {
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibcq->device);
 	struct hns_roce_cq *hr_cq = to_hr_cq(ibcq);
@@ -3603,6 +3601,7 @@ static void hns_roce_v1_destroy_cq(struct ib_cq *ibcq, struct ib_udata *udata)
 		}
 		wait_time++;
 	}
+	return 0;
 }
 
 static void set_eq_cons_index_v1(struct hns_roce_eq *eq, int req_not)
@@ -3775,8 +3774,7 @@ static void hns_roce_v1_db_overflow_handle(struct hns_roce_dev *hr_dev,
 
 static struct hns_roce_aeqe *get_aeqe_v1(struct hns_roce_eq *eq, u32 entry)
 {
-	unsigned long off = (entry & (eq->entries - 1)) *
+	unsigned long off = (entry & (eq->entries - 1)) * HNS_ROCE_AEQE_SIZE;
-			     HNS_ROCE_AEQ_ENTRY_SIZE;
 
 	return (struct hns_roce_aeqe *)((u8 *)
 		(eq->buf_list[off / HNS_ROCE_BA_SIZE].buf) +
@@ -3881,8 +3879,7 @@ static int hns_roce_v1_aeq_int(struct hns_roce_dev *hr_dev,
 
 static struct hns_roce_ceqe *get_ceqe_v1(struct hns_roce_eq *eq, u32 entry)
 {
-	unsigned long off = (entry & (eq->entries - 1)) *
+	unsigned long off = (entry & (eq->entries - 1)) * HNS_ROCE_CEQE_SIZE;
-			     HNS_ROCE_CEQ_ENTRY_SIZE;
 
 	return (struct hns_roce_ceqe *)((u8 *)
 			(eq->buf_list[off / HNS_ROCE_BA_SIZE].buf) +
@@ -3934,7 +3931,7 @@ static irqreturn_t hns_roce_v1_msix_interrupt_eq(int irq, void *eq_ptr)
 {
 	struct hns_roce_eq  *eq  = eq_ptr;
 	struct hns_roce_dev *hr_dev = eq->hr_dev;
-	int int_work = 0;
+	int int_work;
 
 	if (eq->type_flag == HNS_ROCE_CEQ)
 		/* CEQ irq routine, CEQ is pulse irq, not clear */
@@ -4132,9 +4129,9 @@ static int hns_roce_v1_create_eq(struct hns_roce_dev *hr_dev,
 	void __iomem *eqc = hr_dev->eq_table.eqc_base[eq->eqn];
 	struct device *dev = &hr_dev->pdev->dev;
 	dma_addr_t tmp_dma_addr;
-	u32 eqconsindx_val = 0;
 	u32 eqcuridx_val = 0;
-	u32 eqshift_val = 0;
+	u32 eqconsindx_val;
+	u32 eqshift_val;
 	__le32 tmp2 = 0;
 	__le32 tmp1 = 0;
 	__le32 tmp = 0;
@@ -4253,7 +4250,7 @@ static int hns_roce_v1_init_eq_table(struct hns_roce_dev *hr_dev)
 				       CEQ_REG_OFFSET * i;
 			eq->entries = hr_dev->caps.ceqe_depth;
 			eq->log_entries = ilog2(eq->entries);
-			eq->eqe_size = HNS_ROCE_CEQ_ENTRY_SIZE;
+			eq->eqe_size = HNS_ROCE_CEQE_SIZE;
 		} else {
 			/* AEQ */
 			eq_table->eqc_base[i] = hr_dev->reg_base +
@@ -4263,7 +4260,7 @@ static int hns_roce_v1_init_eq_table(struct hns_roce_dev *hr_dev)
 				       ROCEE_CAEP_AEQE_CONS_IDX_REG;
 			eq->entries = hr_dev->caps.aeqe_depth;
 			eq->log_entries = ilog2(eq->entries);
-			eq->eqe_size = HNS_ROCE_AEQ_ENTRY_SIZE;
+			eq->eqe_size = HNS_ROCE_AEQE_SIZE;
 		}
 	}
 

drivers/infiniband/hw/hns/hns_roce_hw_v1.h

@@ -68,13 +68,13 @@
 #define HNS_ROCE_V1_COMP_EQE_NUM			0x8000
 #define HNS_ROCE_V1_ASYNC_EQE_NUM			0x400
 
-#define HNS_ROCE_V1_QPC_ENTRY_SIZE			256
+#define HNS_ROCE_V1_QPC_SIZE				256
 #define HNS_ROCE_V1_IRRL_ENTRY_SIZE			8
 #define HNS_ROCE_V1_CQC_ENTRY_SIZE			64
 #define HNS_ROCE_V1_MTPT_ENTRY_SIZE			64
 #define HNS_ROCE_V1_MTT_ENTRY_SIZE			64
 
-#define HNS_ROCE_V1_CQE_ENTRY_SIZE			32
+#define HNS_ROCE_V1_CQE_SIZE				32
 #define HNS_ROCE_V1_PAGE_SIZE_SUPPORT			0xFFFFF000
 
 #define HNS_ROCE_V1_TABLE_CHUNK_SIZE			(1 << 17)

drivers/infiniband/hw/hns/hns_roce_hw_v2.c

@@ -153,6 +153,67 @@ static void set_atomic_seg(const struct ib_send_wr *wr,
 		       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_S, valid_num_sge);
 }
 
+static int fill_ext_sge_inl_data(struct hns_roce_qp *qp,
+				 const struct ib_send_wr *wr,
+				 unsigned int *sge_idx, u32 msg_len)
+{
+	struct ib_device *ibdev = &(to_hr_dev(qp->ibqp.device))->ib_dev;
+	unsigned int dseg_len = sizeof(struct hns_roce_v2_wqe_data_seg);
+	unsigned int ext_sge_sz = qp->sq.max_gs * dseg_len;
+	unsigned int left_len_in_pg;
+	unsigned int idx = *sge_idx;
+	unsigned int i = 0;
+	unsigned int len;
+	void *addr;
+	void *dseg;
+
+	if (msg_len > ext_sge_sz) {
+		ibdev_err(ibdev,
+			  "no enough extended sge space for inline data.\n");
+		return -EINVAL;
+	}
+
+	dseg = hns_roce_get_extend_sge(qp, idx & (qp->sge.sge_cnt - 1));
+	left_len_in_pg = hr_hw_page_align((uintptr_t)dseg) - (uintptr_t)dseg;
+	len = wr->sg_list[0].length;
+	addr = (void *)(unsigned long)(wr->sg_list[0].addr);
+
+	/* When copying data to extended sge space, the left length in page may
+	 * not long enough for current user's sge. So the data should be
+	 * splited into several parts, one in the first page, and the others in
+	 * the subsequent pages.
+	 */
+	while (1) {
+		if (len <= left_len_in_pg) {
+			memcpy(dseg, addr, len);
+
+			idx += len / dseg_len;
+
+			i++;
+			if (i >= wr->num_sge)
+				break;
+
+			left_len_in_pg -= len;
+			len = wr->sg_list[i].length;
+			addr = (void *)(unsigned long)(wr->sg_list[i].addr);
+			dseg += len;
+		} else {
+			memcpy(dseg, addr, left_len_in_pg);
+
+			len -= left_len_in_pg;
+			addr += left_len_in_pg;
+			idx += left_len_in_pg / dseg_len;
+			dseg = hns_roce_get_extend_sge(qp,
+						idx & (qp->sge.sge_cnt - 1));
+			left_len_in_pg = 1 << HNS_HW_PAGE_SHIFT;
+		}
+	}
+
+	*sge_idx = idx;
+
+	return 0;
+}
+
 static void set_extend_sge(struct hns_roce_qp *qp, const struct ib_send_wr *wr,
 			   unsigned int *sge_ind, unsigned int valid_num_sge)
 {
@@ -177,73 +238,115 @@ static void set_extend_sge(struct hns_roce_qp *qp, const struct ib_send_wr *wr,
 	*sge_ind = idx;
 }
 
+static bool check_inl_data_len(struct hns_roce_qp *qp, unsigned int len)
+{
+	struct hns_roce_dev *hr_dev = to_hr_dev(qp->ibqp.device);
+	int mtu = ib_mtu_enum_to_int(qp->path_mtu);
+
+	if (len > qp->max_inline_data || len > mtu) {
+		ibdev_err(&hr_dev->ib_dev,
+			  "invalid length of data, data len = %u, max inline len = %u, path mtu = %d.\n",
+			  len, qp->max_inline_data, mtu);
+		return false;
+	}
+
+	return true;
+}
+
+static int set_rc_inl(struct hns_roce_qp *qp, const struct ib_send_wr *wr,
+		      struct hns_roce_v2_rc_send_wqe *rc_sq_wqe,
+		      unsigned int *sge_idx)
+{
+	struct hns_roce_dev *hr_dev = to_hr_dev(qp->ibqp.device);
+	u32 msg_len = le32_to_cpu(rc_sq_wqe->msg_len);
+	struct ib_device *ibdev = &hr_dev->ib_dev;
+	unsigned int curr_idx = *sge_idx;
+	void *dseg = rc_sq_wqe;
+	unsigned int i;
+	int ret;
+
+	if (unlikely(wr->opcode == IB_WR_RDMA_READ)) {
+		ibdev_err(ibdev, "invalid inline parameters!\n");
+		return -EINVAL;
+	}
+
+	if (!check_inl_data_len(qp, msg_len))
+		return -EINVAL;
+
+	dseg += sizeof(struct hns_roce_v2_rc_send_wqe);
+
+	roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_INLINE_S, 1);
+
+	if (msg_len <= HNS_ROCE_V2_MAX_RC_INL_INN_SZ) {
+		roce_set_bit(rc_sq_wqe->byte_20,
+			     V2_RC_SEND_WQE_BYTE_20_INL_TYPE_S, 0);
+
+		for (i = 0; i < wr->num_sge; i++) {
+			memcpy(dseg, ((void *)wr->sg_list[i].addr),
+			       wr->sg_list[i].length);
+			dseg += wr->sg_list[i].length;
+		}
+	} else {
+		roce_set_bit(rc_sq_wqe->byte_20,
+			     V2_RC_SEND_WQE_BYTE_20_INL_TYPE_S, 1);
+
+		ret = fill_ext_sge_inl_data(qp, wr, &curr_idx, msg_len);
+		if (ret)
+			return ret;
+
+		roce_set_field(rc_sq_wqe->byte_16,
+			       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_M,
+			       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_S,
+			       curr_idx - *sge_idx);
+	}
+
+	*sge_idx = curr_idx;
+
+	return 0;
+}
+
 static int set_rwqe_data_seg(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 			     struct hns_roce_v2_rc_send_wqe *rc_sq_wqe,
 			     unsigned int *sge_ind,
 			     unsigned int valid_num_sge)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(ibqp->device);
 	struct hns_roce_v2_wqe_data_seg *dseg =
 		(void *)rc_sq_wqe + sizeof(struct hns_roce_v2_rc_send_wqe);
-	struct ib_device *ibdev = &hr_dev->ib_dev;
 	struct hns_roce_qp *qp = to_hr_qp(ibqp);
-	void *wqe = dseg;
 	int j = 0;
 	int i;
 
-	if (wr->send_flags & IB_SEND_INLINE && valid_num_sge) {
+	roce_set_field(rc_sq_wqe->byte_20,
-		if (unlikely(le32_to_cpu(rc_sq_wqe->msg_len) >
+		       V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_M,
-			     hr_dev->caps.max_sq_inline)) {
+		       V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_S,
-			ibdev_err(ibdev, "inline len(1-%d)=%d, illegal",
+		       (*sge_ind) & (qp->sge.sge_cnt - 1));
-				  rc_sq_wqe->msg_len,
-				  hr_dev->caps.max_sq_inline);
-			return -EINVAL;
-		}
 
-		if (unlikely(wr->opcode == IB_WR_RDMA_READ)) {
+	if (wr->send_flags & IB_SEND_INLINE)
-			ibdev_err(ibdev, "Not support inline data!\n");
+		return set_rc_inl(qp, wr, rc_sq_wqe, sge_ind);
-			return -EINVAL;
-		}
 
+	if (valid_num_sge <= HNS_ROCE_SGE_IN_WQE) {
 		for (i = 0; i < wr->num_sge; i++) {
-			memcpy(wqe, ((void *)wr->sg_list[i].addr),
+			if (likely(wr->sg_list[i].length)) {
-			       wr->sg_list[i].length);
+				set_data_seg_v2(dseg, wr->sg_list + i);
-			wqe += wr->sg_list[i].length;
+				dseg++;
+			}
 		}
-
-		roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_INLINE_S,
-			     1);
 	} else {
-		if (valid_num_sge <= HNS_ROCE_SGE_IN_WQE) {
+		for (i = 0; i < wr->num_sge && j < HNS_ROCE_SGE_IN_WQE; i++) {
-			for (i = 0; i < wr->num_sge; i++) {
+			if (likely(wr->sg_list[i].length)) {
-				if (likely(wr->sg_list[i].length)) {
+				set_data_seg_v2(dseg, wr->sg_list + i);
-					set_data_seg_v2(dseg, wr->sg_list + i);
+				dseg++;
-					dseg++;
+				j++;
-				}
 			}
-		} else {
-			roce_set_field(rc_sq_wqe->byte_20,
-				     V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_M,
-				     V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_S,
-				     (*sge_ind) & (qp->sge.sge_cnt - 1));
-
-			for (i = 0; i < wr->num_sge && j < HNS_ROCE_SGE_IN_WQE;
-			     i++) {
-				if (likely(wr->sg_list[i].length)) {
-					set_data_seg_v2(dseg, wr->sg_list + i);
-					dseg++;
-					j++;
-				}
-			}
-
-			set_extend_sge(qp, wr, sge_ind, valid_num_sge);
 		}
 
-		roce_set_field(rc_sq_wqe->byte_16,
+		set_extend_sge(qp, wr, sge_ind, valid_num_sge);
-			       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_M,
-			       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_S, valid_num_sge);
 	}
 
+	roce_set_field(rc_sq_wqe->byte_16,
+		       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_M,
+		       V2_RC_SEND_WQE_BYTE_16_SGE_NUM_S, valid_num_sge);
+
 	return 0;
 }
 
@@ -292,6 +395,33 @@ static unsigned int calc_wr_sge_num(const struct ib_send_wr *wr,
 	return valid_num;
 }
 
+static __le32 get_immtdata(const struct ib_send_wr *wr)
+{
+	switch (wr->opcode) {
+	case IB_WR_SEND_WITH_IMM:
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		return cpu_to_le32(be32_to_cpu(wr->ex.imm_data));
+	default:
+		return 0;
+	}
+}
+
+static int set_ud_opcode(struct hns_roce_v2_ud_send_wqe *ud_sq_wqe,
+			 const struct ib_send_wr *wr)
+{
+	u32 ib_op = wr->opcode;
+
+	if (ib_op != IB_WR_SEND && ib_op != IB_WR_SEND_WITH_IMM)
+		return -EINVAL;
+
+	ud_sq_wqe->immtdata = get_immtdata(wr);
+
+	roce_set_field(ud_sq_wqe->byte_4, V2_UD_SEND_WQE_BYTE_4_OPCODE_M,
+		       V2_UD_SEND_WQE_BYTE_4_OPCODE_S, to_hr_opcode(ib_op));
+
+	return 0;
+}
+
 static inline int set_ud_wqe(struct hns_roce_qp *qp,
 			     const struct ib_send_wr *wr,
 			     void *wqe, unsigned int *sge_idx,
@@ -305,10 +435,15 @@ static inline int set_ud_wqe(struct hns_roce_qp *qp,
 	u32 msg_len = 0;
 	bool loopback;
 	u8 *smac;
+	int ret;
 
 	valid_num_sge = calc_wr_sge_num(wr, &msg_len);
 	memset(ud_sq_wqe, 0, sizeof(*ud_sq_wqe));
 
+	ret = set_ud_opcode(ud_sq_wqe, wr);
+	if (WARN_ON(ret))
+		return ret;
+
 	roce_set_field(ud_sq_wqe->dmac, V2_UD_SEND_WQE_DMAC_0_M,
 		       V2_UD_SEND_WQE_DMAC_0_S, ah->av.mac[0]);
 	roce_set_field(ud_sq_wqe->dmac, V2_UD_SEND_WQE_DMAC_1_M,
@@ -329,23 +464,8 @@ static inline int set_ud_wqe(struct hns_roce_qp *qp,
 	roce_set_bit(ud_sq_wqe->byte_40,
 		     V2_UD_SEND_WQE_BYTE_40_LBI_S, loopback);
 
-	roce_set_field(ud_sq_wqe->byte_4,
-		       V2_UD_SEND_WQE_BYTE_4_OPCODE_M,
-		       V2_UD_SEND_WQE_BYTE_4_OPCODE_S,
-		       HNS_ROCE_V2_WQE_OP_SEND);
-
 	ud_sq_wqe->msg_len = cpu_to_le32(msg_len);
 
-	switch (wr->opcode) {
-	case IB_WR_SEND_WITH_IMM:
-	case IB_WR_RDMA_WRITE_WITH_IMM:
-		ud_sq_wqe->immtdata = cpu_to_le32(be32_to_cpu(wr->ex.imm_data));
-		break;
-	default:
-		ud_sq_wqe->immtdata = 0;
-		break;
-	}
-
 	/* Set sig attr */
 	roce_set_bit(ud_sq_wqe->byte_4, V2_UD_SEND_WQE_BYTE_4_CQE_S,
 		     (wr->send_flags & IB_SEND_SIGNALED) ? 1 : 0);
@@ -369,7 +489,7 @@ static inline int set_ud_wqe(struct hns_roce_qp *qp,
 		       curr_idx & (qp->sge.sge_cnt - 1));
 
 	roce_set_field(ud_sq_wqe->byte_24, V2_UD_SEND_WQE_BYTE_24_UDPSPN_M,
-		       V2_UD_SEND_WQE_BYTE_24_UDPSPN_S, 0);
+		       V2_UD_SEND_WQE_BYTE_24_UDPSPN_S, ah->av.udp_sport);
 	ud_sq_wqe->qkey = cpu_to_le32(ud_wr(wr)->remote_qkey & 0x80000000 ?
 			  qp->qkey : ud_wr(wr)->remote_qkey);
 	roce_set_field(ud_sq_wqe->byte_32, V2_UD_SEND_WQE_BYTE_32_DQPN_M,
@@ -402,6 +522,46 @@ static inline int set_ud_wqe(struct hns_roce_qp *qp,
 	return 0;
 }
 
+static int set_rc_opcode(struct hns_roce_v2_rc_send_wqe *rc_sq_wqe,
+			 const struct ib_send_wr *wr)
+{
+	u32 ib_op = wr->opcode;
+
+	rc_sq_wqe->immtdata = get_immtdata(wr);
+
+	switch (ib_op) {
+	case IB_WR_RDMA_READ:
+	case IB_WR_RDMA_WRITE:
+	case IB_WR_RDMA_WRITE_WITH_IMM:
+		rc_sq_wqe->rkey = cpu_to_le32(rdma_wr(wr)->rkey);
+		rc_sq_wqe->va = cpu_to_le64(rdma_wr(wr)->remote_addr);
+		break;
+	case IB_WR_SEND:
+	case IB_WR_SEND_WITH_IMM:
+		break;
+	case IB_WR_ATOMIC_CMP_AND_SWP:
+	case IB_WR_ATOMIC_FETCH_AND_ADD:
+		rc_sq_wqe->rkey = cpu_to_le32(atomic_wr(wr)->rkey);
+		rc_sq_wqe->va = cpu_to_le64(atomic_wr(wr)->remote_addr);
+		break;
+	case IB_WR_REG_MR:
+		set_frmr_seg(rc_sq_wqe, reg_wr(wr));
+		break;
+	case IB_WR_LOCAL_INV:
+		roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_SO_S, 1);
+		fallthrough;
+	case IB_WR_SEND_WITH_INV:
+		rc_sq_wqe->inv_key = cpu_to_le32(wr->ex.invalidate_rkey);
+		break;
+	default:
+		return -EINVAL;
+	}
+
+	roce_set_field(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_OPCODE_M,
+		       V2_RC_SEND_WQE_BYTE_4_OPCODE_S, to_hr_opcode(ib_op));
+
+	return 0;
+}
 static inline int set_rc_wqe(struct hns_roce_qp *qp,
 			     const struct ib_send_wr *wr,
 			     void *wqe, unsigned int *sge_idx,
@@ -411,25 +571,16 @@ static inline int set_rc_wqe(struct hns_roce_qp *qp,
 	unsigned int curr_idx = *sge_idx;
 	unsigned int valid_num_sge;
 	u32 msg_len = 0;
-	int ret = 0;
+	int ret;
 
 	valid_num_sge = calc_wr_sge_num(wr, &msg_len);
 	memset(rc_sq_wqe, 0, sizeof(*rc_sq_wqe));
 
 	rc_sq_wqe->msg_len = cpu_to_le32(msg_len);
 
-	switch (wr->opcode) {
+	ret = set_rc_opcode(rc_sq_wqe, wr);
-	case IB_WR_SEND_WITH_IMM:
+	if (WARN_ON(ret))
-	case IB_WR_RDMA_WRITE_WITH_IMM:
+		return ret;
-		rc_sq_wqe->immtdata = cpu_to_le32(be32_to_cpu(wr->ex.imm_data));
-		break;
-	case IB_WR_SEND_WITH_INV:
-		rc_sq_wqe->inv_key = cpu_to_le32(wr->ex.invalidate_rkey);
-		break;
-	default:
-		rc_sq_wqe->immtdata = 0;
-		break;
-	}
 
 	roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_FENCE_S,
 		     (wr->send_flags & IB_SEND_FENCE) ? 1 : 0);
@@ -443,33 +594,6 @@ static inline int set_rc_wqe(struct hns_roce_qp *qp,
 	roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_OWNER_S,
 		     owner_bit);
 
-	switch (wr->opcode) {
-	case IB_WR_RDMA_READ:
-	case IB_WR_RDMA_WRITE:
-	case IB_WR_RDMA_WRITE_WITH_IMM:
-		rc_sq_wqe->rkey = cpu_to_le32(rdma_wr(wr)->rkey);
-		rc_sq_wqe->va = cpu_to_le64(rdma_wr(wr)->remote_addr);
-		break;
-	case IB_WR_LOCAL_INV:
-		roce_set_bit(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_SO_S, 1);
-		rc_sq_wqe->inv_key = cpu_to_le32(wr->ex.invalidate_rkey);
-		break;
-	case IB_WR_REG_MR:
-		set_frmr_seg(rc_sq_wqe, reg_wr(wr));
-		break;
-	case IB_WR_ATOMIC_CMP_AND_SWP:
-	case IB_WR_ATOMIC_FETCH_AND_ADD:
-		rc_sq_wqe->rkey = cpu_to_le32(atomic_wr(wr)->rkey);
-		rc_sq_wqe->va = cpu_to_le64(atomic_wr(wr)->remote_addr);
-		break;
-	default:
-		break;
-	}
-
-	roce_set_field(rc_sq_wqe->byte_4, V2_RC_SEND_WQE_BYTE_4_OPCODE_M,
-		       V2_RC_SEND_WQE_BYTE_4_OPCODE_S,
-		       to_hr_opcode(wr->opcode));
-
 	if (wr->opcode == IB_WR_ATOMIC_CMP_AND_SWP ||
 	    wr->opcode == IB_WR_ATOMIC_FETCH_AND_ADD)
 		set_atomic_seg(wr, rc_sq_wqe, valid_num_sge);
@@ -1682,7 +1806,7 @@ static void set_default_caps(struct hns_roce_dev *hr_dev)
 	caps->max_sq_desc_sz	= HNS_ROCE_V2_MAX_SQ_DESC_SZ;
 	caps->max_rq_desc_sz	= HNS_ROCE_V2_MAX_RQ_DESC_SZ;
 	caps->max_srq_desc_sz	= HNS_ROCE_V2_MAX_SRQ_DESC_SZ;
-	caps->qpc_entry_sz	= HNS_ROCE_V2_QPC_ENTRY_SZ;
+	caps->qpc_sz		= HNS_ROCE_V2_QPC_SZ;
 	caps->irrl_entry_sz	= HNS_ROCE_V2_IRRL_ENTRY_SZ;
 	caps->trrl_entry_sz	= HNS_ROCE_V2_EXT_ATOMIC_TRRL_ENTRY_SZ;
 	caps->cqc_entry_sz	= HNS_ROCE_V2_CQC_ENTRY_SZ;
@@ -1690,7 +1814,7 @@ static void set_default_caps(struct hns_roce_dev *hr_dev)
 	caps->mtpt_entry_sz	= HNS_ROCE_V2_MTPT_ENTRY_SZ;
 	caps->mtt_entry_sz	= HNS_ROCE_V2_MTT_ENTRY_SZ;
 	caps->idx_entry_sz	= HNS_ROCE_V2_IDX_ENTRY_SZ;
-	caps->cq_entry_sz	= HNS_ROCE_V2_CQE_ENTRY_SIZE;
+	caps->cqe_sz		= HNS_ROCE_V2_CQE_SIZE;
 	caps->page_size_cap	= HNS_ROCE_V2_PAGE_SIZE_SUPPORTED;
 	caps->reserved_lkey	= 0;
 	caps->reserved_pds	= 0;
@@ -1739,6 +1863,8 @@ static void set_default_caps(struct hns_roce_dev *hr_dev)
 	caps->gid_table_len[0]	= HNS_ROCE_V2_GID_INDEX_NUM;
 	caps->ceqe_depth	= HNS_ROCE_V2_COMP_EQE_NUM;
 	caps->aeqe_depth	= HNS_ROCE_V2_ASYNC_EQE_NUM;
+	caps->aeqe_size		= HNS_ROCE_AEQE_SIZE;
+	caps->ceqe_size		= HNS_ROCE_CEQE_SIZE;
 	caps->local_ca_ack_delay = 0;
 	caps->max_mtu = IB_MTU_4096;
 
@@ -1760,19 +1886,26 @@ static void set_default_caps(struct hns_roce_dev *hr_dev)
 	caps->cqc_timer_buf_pg_sz = 0;
 	caps->cqc_timer_hop_num   = HNS_ROCE_HOP_NUM_0;
 
-	caps->sccc_entry_sz	  = HNS_ROCE_V2_SCCC_ENTRY_SZ;
+	caps->sccc_sz = HNS_ROCE_V2_SCCC_SZ;
 	caps->sccc_ba_pg_sz	  = 0;
 	caps->sccc_buf_pg_sz	  = 0;
 	caps->sccc_hop_num	  = HNS_ROCE_SCCC_HOP_NUM;
+
+	if (hr_dev->pci_dev->revision >= PCI_REVISION_ID_HIP09) {
+		caps->aeqe_size = HNS_ROCE_V3_EQE_SIZE;
+		caps->ceqe_size = HNS_ROCE_V3_EQE_SIZE;
+		caps->cqe_sz = HNS_ROCE_V3_CQE_SIZE;
+		caps->qpc_sz = HNS_ROCE_V3_QPC_SZ;
+	}
 }
 
 static void calc_pg_sz(int obj_num, int obj_size, int hop_num, int ctx_bt_num,
 		       int *buf_page_size, int *bt_page_size, u32 hem_type)
 {
 	u64 obj_per_chunk;
-	int bt_chunk_size = 1 << PAGE_SHIFT;
+	u64 bt_chunk_size = PAGE_SIZE;
-	int buf_chunk_size = 1 << PAGE_SHIFT;
+	u64 buf_chunk_size = PAGE_SIZE;
-	int obj_per_chunk_default = buf_chunk_size / obj_size;
+	u64 obj_per_chunk_default = buf_chunk_size / obj_size;
 
 	*buf_page_size = 0;
 	*bt_page_size = 0;
@@ -1855,7 +1988,7 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 	caps->max_sq_desc_sz	     = resp_a->max_sq_desc_sz;
 	caps->max_rq_desc_sz	     = resp_a->max_rq_desc_sz;
 	caps->max_srq_desc_sz	     = resp_a->max_srq_desc_sz;
-	caps->cq_entry_sz	     = resp_a->cq_entry_sz;
+	caps->cqe_sz		     = HNS_ROCE_V2_CQE_SIZE;
 
 	caps->mtpt_entry_sz	     = resp_b->mtpt_entry_sz;
 	caps->irrl_entry_sz	     = resp_b->irrl_entry_sz;
@@ -1863,9 +1996,9 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 	caps->cqc_entry_sz	     = resp_b->cqc_entry_sz;
 	caps->srqc_entry_sz	     = resp_b->srqc_entry_sz;
 	caps->idx_entry_sz	     = resp_b->idx_entry_sz;
-	caps->sccc_entry_sz	     = resp_b->scc_ctx_entry_sz;
+	caps->sccc_sz		     = resp_b->sccc_sz;
 	caps->max_mtu		     = resp_b->max_mtu;
-	caps->qpc_entry_sz	     = le16_to_cpu(resp_b->qpc_entry_sz);
+	caps->qpc_sz		     = HNS_ROCE_V2_QPC_SZ;
 	caps->min_cqes		     = resp_b->min_cqes;
 	caps->min_wqes		     = resp_b->min_wqes;
 	caps->page_size_cap	     = le32_to_cpu(resp_b->page_size_cap);
@@ -1958,6 +2091,8 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 	caps->cqc_timer_entry_sz = HNS_ROCE_V2_CQC_TIMER_ENTRY_SZ;
 	caps->mtt_entry_sz = HNS_ROCE_V2_MTT_ENTRY_SZ;
 	caps->num_mtt_segs = HNS_ROCE_V2_MAX_MTT_SEGS;
+	caps->ceqe_size = HNS_ROCE_CEQE_SIZE;
+	caps->aeqe_size = HNS_ROCE_AEQE_SIZE;
 	caps->mtt_ba_pg_sz = 0;
 	caps->num_cqe_segs = HNS_ROCE_V2_MAX_CQE_SEGS;
 	caps->num_srqwqe_segs = HNS_ROCE_V2_MAX_SRQWQE_SEGS;
@@ -1981,7 +2116,15 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 					  V2_QUERY_PF_CAPS_D_RQWQE_HOP_NUM_M,
 					  V2_QUERY_PF_CAPS_D_RQWQE_HOP_NUM_S);
 
-	calc_pg_sz(caps->num_qps, caps->qpc_entry_sz, caps->qpc_hop_num,
+	if (hr_dev->pci_dev->revision >= PCI_REVISION_ID_HIP09) {
+		caps->ceqe_size = HNS_ROCE_V3_EQE_SIZE;
+		caps->aeqe_size = HNS_ROCE_V3_EQE_SIZE;
+		caps->cqe_sz = HNS_ROCE_V3_CQE_SIZE;
+		caps->qpc_sz = HNS_ROCE_V3_QPC_SZ;
+		caps->sccc_sz = HNS_ROCE_V3_SCCC_SZ;
+	}
+
+	calc_pg_sz(caps->num_qps, caps->qpc_sz, caps->qpc_hop_num,
 		   caps->qpc_bt_num, &caps->qpc_buf_pg_sz, &caps->qpc_ba_pg_sz,
 		   HEM_TYPE_QPC);
 	calc_pg_sz(caps->num_mtpts, caps->mtpt_entry_sz, caps->mpt_hop_num,
@@ -1998,7 +2141,7 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 	caps->qpc_timer_hop_num = HNS_ROCE_HOP_NUM_0;
 	caps->cqc_timer_hop_num = HNS_ROCE_HOP_NUM_0;
 
-	calc_pg_sz(caps->num_qps, caps->sccc_entry_sz,
+	calc_pg_sz(caps->num_qps, caps->sccc_sz,
 		   caps->sccc_hop_num, caps->sccc_bt_num,
 		   &caps->sccc_buf_pg_sz, &caps->sccc_ba_pg_sz,
 		   HEM_TYPE_SCCC);
@@ -2018,6 +2161,56 @@ static int hns_roce_query_pf_caps(struct hns_roce_dev *hr_dev)
 	return 0;
 }
 
+static int hns_roce_config_qpc_size(struct hns_roce_dev *hr_dev)
+{
+	struct hns_roce_cmq_desc desc;
+	struct hns_roce_cfg_entry_size *cfg_size =
+				  (struct hns_roce_cfg_entry_size *)desc.data;
+
+	hns_roce_cmq_setup_basic_desc(&desc, HNS_ROCE_OPC_CFG_ENTRY_SIZE,
+				      false);
+
+	cfg_size->type = cpu_to_le32(HNS_ROCE_CFG_QPC_SIZE);
+	cfg_size->size = cpu_to_le32(hr_dev->caps.qpc_sz);
+
+	return hns_roce_cmq_send(hr_dev, &desc, 1);
+}
+
+static int hns_roce_config_sccc_size(struct hns_roce_dev *hr_dev)
+{
+	struct hns_roce_cmq_desc desc;
+	struct hns_roce_cfg_entry_size *cfg_size =
+				  (struct hns_roce_cfg_entry_size *)desc.data;
+
+	hns_roce_cmq_setup_basic_desc(&desc, HNS_ROCE_OPC_CFG_ENTRY_SIZE,
+				      false);
+
+	cfg_size->type = cpu_to_le32(HNS_ROCE_CFG_SCCC_SIZE);
+	cfg_size->size = cpu_to_le32(hr_dev->caps.sccc_sz);
+
+	return hns_roce_cmq_send(hr_dev, &desc, 1);
+}
+
+static int hns_roce_config_entry_size(struct hns_roce_dev *hr_dev)
+{
+	int ret;
+
+	if (hr_dev->pci_dev->revision < PCI_REVISION_ID_HIP09)
+		return 0;
+
+	ret = hns_roce_config_qpc_size(hr_dev);
+	if (ret) {
+		dev_err(hr_dev->dev, "failed to cfg qpc sz, ret = %d.\n", ret);
+		return ret;
+	}
+
+	ret = hns_roce_config_sccc_size(hr_dev);
+	if (ret)
+		dev_err(hr_dev->dev, "failed to cfg sccc sz, ret = %d.\n", ret);
+
+	return ret;
+}
+
 static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 {
 	struct hns_roce_caps *caps = &hr_dev->caps;
@@ -2090,9 +2283,14 @@ static int hns_roce_v2_profile(struct hns_roce_dev *hr_dev)
 	}
 
 	ret = hns_roce_v2_set_bt(hr_dev);
-	if (ret)
+	if (ret) {
-		dev_err(hr_dev->dev, "Configure bt attribute fail, ret = %d.\n",
+		dev_err(hr_dev->dev,
-			ret);
+			"Configure bt attribute fail, ret = %d.\n", ret);
+		return ret;
+	}
+
+	/* Configure the size of QPC, SCCC, etc. */
+	ret = hns_roce_config_entry_size(hr_dev);
 
 	return ret;
 }
@@ -2757,8 +2955,7 @@ static int hns_roce_v2_mw_write_mtpt(void *mb_buf, struct hns_roce_mw *mw)
 
 static void *get_cqe_v2(struct hns_roce_cq *hr_cq, int n)
 {
-	return hns_roce_buf_offset(hr_cq->mtr.kmem,
+	return hns_roce_buf_offset(hr_cq->mtr.kmem, n * hr_cq->cqe_size);
-				   n * HNS_ROCE_V2_CQE_ENTRY_SIZE);
 }
 
 static void *get_sw_cqe_v2(struct hns_roce_cq *hr_cq, int n)
@@ -2858,6 +3055,10 @@ static void hns_roce_v2_write_cqc(struct hns_roce_dev *hr_dev,
 	roce_set_field(cq_context->byte_8_cqn, V2_CQC_BYTE_8_CQN_M,
 		       V2_CQC_BYTE_8_CQN_S, hr_cq->cqn);
 
+	roce_set_field(cq_context->byte_8_cqn, V2_CQC_BYTE_8_CQE_SIZE_M,
+		       V2_CQC_BYTE_8_CQE_SIZE_S, hr_cq->cqe_size ==
+		       HNS_ROCE_V3_CQE_SIZE ? 1 : 0);
+
 	cq_context->cqe_cur_blk_addr = cpu_to_le32(to_hr_hw_page_addr(mtts[0]));
 
 	roce_set_field(cq_context->byte_16_hop_addr,
@@ -3025,7 +3226,8 @@ out:
 }
 
 static void get_cqe_status(struct hns_roce_dev *hr_dev, struct hns_roce_qp *qp,
-			   struct hns_roce_v2_cqe *cqe, struct ib_wc *wc)
+			   struct hns_roce_cq *cq, struct hns_roce_v2_cqe *cqe,
+			   struct ib_wc *wc)
 {
 	static const struct {
 		u32 cqe_status;
@@ -3066,7 +3268,7 @@ static void get_cqe_status(struct hns_roce_dev *hr_dev, struct hns_roce_qp *qp,
 
 	ibdev_err(&hr_dev->ib_dev, "error cqe status 0x%x:\n", cqe_status);
 	print_hex_dump(KERN_ERR, "", DUMP_PREFIX_NONE, 16, 4, cqe,
-		       sizeof(*cqe), false);
+		       cq->cqe_size, false);
 
 	/*
 	 * For hns ROCEE, GENERAL_ERR is an error type that is not defined in
@@ -3163,7 +3365,7 @@ static int hns_roce_v2_poll_one(struct hns_roce_cq *hr_cq,
 		++wq->tail;
 	}
 
-	get_cqe_status(hr_dev, *cur_qp, cqe, wc);
+	get_cqe_status(hr_dev, *cur_qp, hr_cq, cqe, wc);
 	if (unlikely(wc->status != IB_WC_SUCCESS))
 		return 0;
 
@@ -3514,16 +3716,21 @@ static int hns_roce_v2_clear_hem(struct hns_roce_dev *hr_dev,
 
 static int hns_roce_v2_qp_modify(struct hns_roce_dev *hr_dev,
 				 struct hns_roce_v2_qp_context *context,
+				 struct hns_roce_v2_qp_context *qpc_mask,
 				 struct hns_roce_qp *hr_qp)
 {
 	struct hns_roce_cmd_mailbox *mailbox;
+	int qpc_size;
 	int ret;
 
 	mailbox = hns_roce_alloc_cmd_mailbox(hr_dev);
 	if (IS_ERR(mailbox))
 		return PTR_ERR(mailbox);
 
-	memcpy(mailbox->buf, context, sizeof(*context) * 2);
+	/* The qpc size of HIP08 is only 256B, which is half of HIP09 */
+	qpc_size = hr_dev->caps.qpc_sz;
+	memcpy(mailbox->buf, context, qpc_size);
+	memcpy(mailbox->buf + qpc_size, qpc_mask, qpc_size);
 
 	ret = hns_roce_cmd_mbox(hr_dev, mailbox->dma, 0, hr_qp->qpn, 0,
 				HNS_ROCE_CMD_MODIFY_QPC,
@@ -3641,9 +3848,6 @@ static void modify_qp_reset_to_init(struct ib_qp *ibqp,
 			     V2_QPC_BYTE_76_SRQ_EN_S, 1);
 	}
 
-	roce_set_field(context->byte_172_sq_psn, V2_QPC_BYTE_172_ACK_REQ_FREQ_M,
-		       V2_QPC_BYTE_172_ACK_REQ_FREQ_S, 4);
-
 	roce_set_bit(context->byte_172_sq_psn, V2_QPC_BYTE_172_FRE_S, 1);
 
 	hr_qp->access_flags = attr->qp_access_flags;
@@ -3954,6 +4158,7 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 	dma_addr_t trrl_ba;
 	dma_addr_t irrl_ba;
 	enum ib_mtu mtu;
+	u8 lp_pktn_ini;
 	u8 port_num;
 	u64 *mtts;
 	u8 *dmac;
@@ -4052,6 +4257,7 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 		       V2_QPC_BYTE_52_DMAC_S, 0);
 
 	mtu = get_mtu(ibqp, attr);
+	hr_qp->path_mtu = mtu;
 
 	if (attr_mask & IB_QP_PATH_MTU) {
 		roce_set_field(context->byte_24_mtu_tc, V2_QPC_BYTE_24_MTU_M,
@@ -4061,13 +4267,21 @@ static int modify_qp_init_to_rtr(struct ib_qp *ibqp,
 	}
 
 #define MAX_LP_MSG_LEN 65536
-	/* MTU*(2^LP_PKTN_INI) shouldn't be bigger than 64kb */
+	/* MTU * (2 ^ LP_PKTN_INI) shouldn't be bigger than 64KB */
+	lp_pktn_ini = ilog2(MAX_LP_MSG_LEN / ib_mtu_enum_to_int(mtu));
+
 	roce_set_field(context->byte_56_dqpn_err, V2_QPC_BYTE_56_LP_PKTN_INI_M,
-		       V2_QPC_BYTE_56_LP_PKTN_INI_S,
+		       V2_QPC_BYTE_56_LP_PKTN_INI_S, lp_pktn_ini);
-		       ilog2(MAX_LP_MSG_LEN / ib_mtu_enum_to_int(mtu)));
 	roce_set_field(qpc_mask->byte_56_dqpn_err, V2_QPC_BYTE_56_LP_PKTN_INI_M,
 		       V2_QPC_BYTE_56_LP_PKTN_INI_S, 0);
 
+	/* ACK_REQ_FREQ should be larger than or equal to LP_PKTN_INI */
+	roce_set_field(context->byte_172_sq_psn, V2_QPC_BYTE_172_ACK_REQ_FREQ_M,
+		       V2_QPC_BYTE_172_ACK_REQ_FREQ_S, lp_pktn_ini);
+	roce_set_field(qpc_mask->byte_172_sq_psn,
+		       V2_QPC_BYTE_172_ACK_REQ_FREQ_M,
+		       V2_QPC_BYTE_172_ACK_REQ_FREQ_S, 0);
+
 	roce_set_bit(qpc_mask->byte_108_rx_reqepsn,
 		     V2_QPC_BYTE_108_RX_REQ_PSN_ERR_S, 0);
 	roce_set_field(qpc_mask->byte_96_rx_reqmsn, V2_QPC_BYTE_96_RX_REQ_MSN_M,
@@ -4164,6 +4378,14 @@ static int modify_qp_rtr_to_rts(struct ib_qp *ibqp,
 	return 0;
 }
 
+static inline u16 get_udp_sport(u32 fl, u32 lqpn, u32 rqpn)
+{
+	if (!fl)
+		fl = rdma_calc_flow_label(lqpn, rqpn);
+
+	return rdma_flow_label_to_udp_sport(fl);
+}
+
 static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 				const struct ib_qp_attr *attr,
 				int attr_mask,
@@ -4227,7 +4449,8 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 
 	roce_set_field(context->byte_52_udpspn_dmac, V2_QPC_BYTE_52_UDPSPN_M,
 		       V2_QPC_BYTE_52_UDPSPN_S,
-		       is_udp ? 0x12b7 : 0);
+		       is_udp ? get_udp_sport(grh->flow_label, ibqp->qp_num,
+					      attr->dest_qp_num) : 0);
 
 	roce_set_field(qpc_mask->byte_52_udpspn_dmac, V2_QPC_BYTE_52_UDPSPN_M,
 		       V2_QPC_BYTE_52_UDPSPN_S, 0);
@@ -4259,11 +4482,19 @@ static int hns_roce_v2_set_path(struct ib_qp *ibqp,
 		       V2_QPC_BYTE_28_FL_S, 0);
 	memcpy(context->dgid, grh->dgid.raw, sizeof(grh->dgid.raw));
 	memset(qpc_mask->dgid, 0, sizeof(grh->dgid.raw));
+
+	hr_qp->sl = rdma_ah_get_sl(&attr->ah_attr);
+	if (unlikely(hr_qp->sl > MAX_SERVICE_LEVEL)) {
+		ibdev_err(ibdev,
+			  "failed to fill QPC, sl (%d) shouldn't be larger than %d.\n",
+			  hr_qp->sl, MAX_SERVICE_LEVEL);
+		return -EINVAL;
+	}
+
 	roce_set_field(context->byte_28_at_fl, V2_QPC_BYTE_28_SL_M,
-		       V2_QPC_BYTE_28_SL_S, rdma_ah_get_sl(&attr->ah_attr));
+		       V2_QPC_BYTE_28_SL_S, hr_qp->sl);
 	roce_set_field(qpc_mask->byte_28_at_fl, V2_QPC_BYTE_28_SL_M,
 		       V2_QPC_BYTE_28_SL_S, 0);
-	hr_qp->sl = rdma_ah_get_sl(&attr->ah_attr);
 
 	return 0;
 }
@@ -4309,7 +4540,7 @@ static int hns_roce_v2_set_abs_fields(struct ib_qp *ibqp,
 	}
 
 	if (cur_state == IB_QPS_RESET && new_state == IB_QPS_INIT) {
-		memset(qpc_mask, 0, sizeof(*qpc_mask));
+		memset(qpc_mask, 0, hr_dev->caps.qpc_sz);
 		modify_qp_reset_to_init(ibqp, attr, attr_mask, context,
 					qpc_mask);
 	} else if (cur_state == IB_QPS_INIT && new_state == IB_QPS_INIT) {
@@ -4532,8 +4763,9 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp,
 	 * we should set all bits of the relevant fields in context mask to
 	 * 0 at the same time, else set them to 0x1.
 	 */
-	memset(context, 0, sizeof(*context));
+	memset(context, 0, hr_dev->caps.qpc_sz);
-	memset(qpc_mask, 0xff, sizeof(*qpc_mask));
+	memset(qpc_mask, 0xff, hr_dev->caps.qpc_sz);
+
 	ret = hns_roce_v2_set_abs_fields(ibqp, attr, attr_mask, cur_state,
 					 new_state, context, qpc_mask);
 	if (ret)
@@ -4583,7 +4815,7 @@ static int hns_roce_v2_modify_qp(struct ib_qp *ibqp,
 		       V2_QPC_BYTE_60_QP_ST_S, 0);
 
 	/* SW pass context to HW */
-	ret = hns_roce_v2_qp_modify(hr_dev, ctx, hr_qp);
+	ret = hns_roce_v2_qp_modify(hr_dev, context, qpc_mask, hr_qp);
 	if (ret) {
 		ibdev_err(ibdev, "failed to modify QP, ret = %d\n", ret);
 		goto out;
@@ -4646,7 +4878,7 @@ static int hns_roce_v2_query_qpc(struct hns_roce_dev *hr_dev,
 	if (ret)
 		goto out;
 
-	memcpy(hr_context, mailbox->buf, sizeof(*hr_context));
+	memcpy(hr_context, mailbox->buf, hr_dev->caps.qpc_sz);
 
 out:
 	hns_roce_free_cmd_mailbox(hr_dev, mailbox);
@@ -4759,7 +4991,9 @@ static int hns_roce_v2_query_qp(struct ib_qp *ibqp, struct ib_qp_attr *qp_attr,
 	qp_attr->retry_cnt = roce_get_field(context.byte_212_lsn,
 					    V2_QPC_BYTE_212_RETRY_CNT_M,
 					    V2_QPC_BYTE_212_RETRY_CNT_S);
-	qp_attr->rnr_retry = le32_to_cpu(context.rq_rnr_timer);
+	qp_attr->rnr_retry = roce_get_field(context.byte_244_rnr_rxack,
+					    V2_QPC_BYTE_244_RNR_CNT_M,
+					    V2_QPC_BYTE_244_RNR_CNT_S);
 
 done:
 	qp_attr->cur_qp_state = qp_attr->qp_state;
@@ -4775,6 +5009,7 @@ done:
 	}
 
 	qp_init_attr->cap = qp_attr->cap;
+	qp_init_attr->sq_sig_type = hr_qp->sq_signal_bits;
 
 out:
 	mutex_unlock(&hr_qp->mutex);
@@ -5004,6 +5239,10 @@ static int hns_roce_v2_modify_srq(struct ib_srq *ibsrq,
 	struct hns_roce_cmd_mailbox *mailbox;
 	int ret;
 
+	/* Resizing SRQs is not supported yet */
+	if (srq_attr_mask & IB_SRQ_MAX_WR)
+		return -EINVAL;
+
 	if (srq_attr_mask & IB_SRQ_LIMIT) {
 		if (srq_attr->srq_limit >= srq->wqe_cnt)
 			return -EINVAL;
@@ -5233,7 +5472,7 @@ static struct hns_roce_aeqe *next_aeqe_sw_v2(struct hns_roce_eq *eq)
 
 	aeqe = hns_roce_buf_offset(eq->mtr.kmem,
 				   (eq->cons_index & (eq->entries - 1)) *
-				   HNS_ROCE_AEQ_ENTRY_SIZE);
+				   eq->eqe_size);
 
 	return (roce_get_bit(aeqe->asyn, HNS_ROCE_V2_AEQ_AEQE_OWNER_S) ^
 		!!(eq->cons_index & eq->entries)) ? aeqe : NULL;
@@ -5333,7 +5572,8 @@ static struct hns_roce_ceqe *next_ceqe_sw_v2(struct hns_roce_eq *eq)
 
 	ceqe = hns_roce_buf_offset(eq->mtr.kmem,
 				   (eq->cons_index & (eq->entries - 1)) *
-				   HNS_ROCE_CEQ_ENTRY_SIZE);
+				   eq->eqe_size);
+
 	return (!!(roce_get_bit(ceqe->comp, HNS_ROCE_V2_CEQ_CEQE_OWNER_S))) ^
 		(!!(eq->cons_index & eq->entries)) ? ceqe : NULL;
 }
@@ -5374,7 +5614,7 @@ static irqreturn_t hns_roce_v2_msix_interrupt_eq(int irq, void *eq_ptr)
 {
 	struct hns_roce_eq *eq = eq_ptr;
 	struct hns_roce_dev *hr_dev = eq->hr_dev;
-	int int_work = 0;
+	int int_work;
 
 	if (eq->type_flag == HNS_ROCE_CEQ)
 		/* Completion event interrupt */
@@ -5609,14 +5849,16 @@ static int config_eqc(struct hns_roce_dev *hr_dev, struct hns_roce_eq *eq,
 	roce_set_field(eqc->byte_36, HNS_ROCE_EQC_CONS_INDX_M,
 		       HNS_ROCE_EQC_CONS_INDX_S, HNS_ROCE_EQ_INIT_CONS_IDX);
 
-	/* set nex_eqe_ba[43:12] */
+	roce_set_field(eqc->byte_40, HNS_ROCE_EQC_NXT_EQE_BA_L_M,
-	roce_set_field(eqc->nxt_eqe_ba0, HNS_ROCE_EQC_NXT_EQE_BA_L_M,
 		       HNS_ROCE_EQC_NXT_EQE_BA_L_S, eqe_ba[1] >> 12);
 
-	/* set nex_eqe_ba[63:44] */
+	roce_set_field(eqc->byte_44, HNS_ROCE_EQC_NXT_EQE_BA_H_M,
-	roce_set_field(eqc->nxt_eqe_ba1, HNS_ROCE_EQC_NXT_EQE_BA_H_M,
 		       HNS_ROCE_EQC_NXT_EQE_BA_H_S, eqe_ba[1] >> 44);
 
+	roce_set_field(eqc->byte_44, HNS_ROCE_EQC_EQE_SIZE_M,
+		       HNS_ROCE_EQC_EQE_SIZE_S,
+		       eq->eqe_size == HNS_ROCE_V3_EQE_SIZE ? 1 : 0);
+
 	return 0;
 }
 
@@ -5807,7 +6049,7 @@ static int hns_roce_v2_init_eq_table(struct hns_roce_dev *hr_dev)
 			eq_cmd = HNS_ROCE_CMD_CREATE_CEQC;
 			eq->type_flag = HNS_ROCE_CEQ;
 			eq->entries = hr_dev->caps.ceqe_depth;
-			eq->eqe_size = HNS_ROCE_CEQ_ENTRY_SIZE;
+			eq->eqe_size = hr_dev->caps.ceqe_size;
 			eq->irq = hr_dev->irq[i + other_num + aeq_num];
 			eq->eq_max_cnt = HNS_ROCE_CEQ_DEFAULT_BURST_NUM;
 			eq->eq_period = HNS_ROCE_CEQ_DEFAULT_INTERVAL;
@@ -5816,7 +6058,7 @@ static int hns_roce_v2_init_eq_table(struct hns_roce_dev *hr_dev)
 			eq_cmd = HNS_ROCE_CMD_CREATE_AEQC;
 			eq->type_flag = HNS_ROCE_AEQ;
 			eq->entries = hr_dev->caps.aeqe_depth;
-			eq->eqe_size = HNS_ROCE_AEQ_ENTRY_SIZE;
+			eq->eqe_size = hr_dev->caps.aeqe_size;
 			eq->irq = hr_dev->irq[i - comp_num + other_num];
 			eq->eq_max_cnt = HNS_ROCE_AEQ_DEFAULT_BURST_NUM;
 			eq->eq_period = HNS_ROCE_AEQ_DEFAULT_INTERVAL;

drivers/infiniband/hw/hns/hns_roce_hw_v2.h

@@ -60,6 +60,7 @@
 #define HNS_ROCE_V2_MAX_SQ_SGE_NUM		64
 #define HNS_ROCE_V2_MAX_EXTEND_SGE_NUM		0x200000
 #define HNS_ROCE_V2_MAX_SQ_INLINE		0x20
+#define HNS_ROCE_V2_MAX_RC_INL_INN_SZ		32
 #define HNS_ROCE_V2_UAR_NUM			256
 #define HNS_ROCE_V2_PHY_UAR_NUM			1
 #define HNS_ROCE_V2_MAX_IRQ_NUM			65
@@ -77,7 +78,6 @@
 #define HNS_ROCE_V2_MAX_SQ_DESC_SZ		64
 #define HNS_ROCE_V2_MAX_RQ_DESC_SZ		16
 #define HNS_ROCE_V2_MAX_SRQ_DESC_SZ		64
-#define HNS_ROCE_V2_QPC_ENTRY_SZ		256
 #define HNS_ROCE_V2_IRRL_ENTRY_SZ		64
 #define HNS_ROCE_V2_TRRL_ENTRY_SZ		48
 #define HNS_ROCE_V2_EXT_ATOMIC_TRRL_ENTRY_SZ	100
@@ -86,8 +86,10 @@
 #define HNS_ROCE_V2_MTPT_ENTRY_SZ		64
 #define HNS_ROCE_V2_MTT_ENTRY_SZ		64
 #define HNS_ROCE_V2_IDX_ENTRY_SZ		4
-#define HNS_ROCE_V2_CQE_ENTRY_SIZE		32
+
-#define HNS_ROCE_V2_SCCC_ENTRY_SZ		32
+#define HNS_ROCE_V2_SCCC_SZ			32
+#define HNS_ROCE_V3_SCCC_SZ			64
+
 #define HNS_ROCE_V2_QPC_TIMER_ENTRY_SZ		PAGE_SIZE
 #define HNS_ROCE_V2_CQC_TIMER_ENTRY_SZ		PAGE_SIZE
 #define HNS_ROCE_V2_PAGE_SIZE_SUPPORTED		0xFFFFF000
@@ -229,6 +231,7 @@ enum hns_roce_opcode_type {
 	HNS_ROCE_OPC_CFG_TMOUT_LLM			= 0x8404,
 	HNS_ROCE_OPC_QUERY_PF_TIMER_RES			= 0x8406,
 	HNS_ROCE_OPC_QUERY_PF_CAPS_NUM                  = 0x8408,
+	HNS_ROCE_OPC_CFG_ENTRY_SIZE			= 0x8409,
 	HNS_ROCE_OPC_CFG_SGID_TB			= 0x8500,
 	HNS_ROCE_OPC_CFG_SMAC_TB			= 0x8501,
 	HNS_ROCE_OPC_POST_MB				= 0x8504,
@@ -309,6 +312,9 @@ struct hns_roce_v2_cq_context {
 #define	V2_CQC_BYTE_8_CQN_S 0
 #define V2_CQC_BYTE_8_CQN_M GENMASK(23, 0)
 
+#define V2_CQC_BYTE_8_CQE_SIZE_S 27
+#define V2_CQC_BYTE_8_CQE_SIZE_M GENMASK(28, 27)
+
 #define	V2_CQC_BYTE_16_CQE_CUR_BLK_ADDR_S 0
 #define V2_CQC_BYTE_16_CQE_CUR_BLK_ADDR_M GENMASK(19, 0)
 
@@ -512,6 +518,7 @@ struct hns_roce_v2_qp_context {
 	__le32	byte_248_ack_psn;
 	__le32	byte_252_err_txcqn;
 	__le32	byte_256_sqflush_rqcqe;
+	__le32	ext[64];
 };
 
 #define	V2_QPC_BYTE_4_TST_S 0
@@ -896,6 +903,7 @@ struct hns_roce_v2_cqe {
 	u8	smac[4];
 	__le32	byte_28;
 	__le32	byte_32;
+	__le32	rsv[8];
 };
 
 #define	V2_CQE_BYTE_4_OPCODE_S 0
@@ -1187,6 +1195,8 @@ struct hns_roce_v2_rc_send_wqe {
 #define V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_S 0
 #define V2_RC_SEND_WQE_BYTE_20_MSG_START_SGE_IDX_M GENMASK(23, 0)
 
+#define V2_RC_SEND_WQE_BYTE_20_INL_TYPE_S 31
+
 struct hns_roce_wqe_frmr_seg {
 	__le32	pbl_size;
 	__le32	mode_buf_pg_sz;
@@ -1537,6 +1547,18 @@ struct hns_roce_cfg_sgid_tb {
 	__le32	vf_sgid_h;
 	__le32	vf_sgid_type_rsv;
 };
+
+enum {
+	HNS_ROCE_CFG_QPC_SIZE = BIT(0),
+	HNS_ROCE_CFG_SCCC_SIZE = BIT(1),
+};
+
+struct hns_roce_cfg_entry_size {
+	__le32	type;
+	__le32	rsv[4];
+	__le32	size;
+};
+
 #define CFG_SGID_TB_TABLE_IDX_S 0
 #define CFG_SGID_TB_TABLE_IDX_M GENMASK(7, 0)
 
@@ -1571,7 +1593,7 @@ struct hns_roce_query_pf_caps_a {
 	u8 max_sq_desc_sz;
 	u8 max_rq_desc_sz;
 	u8 max_srq_desc_sz;
-	u8 cq_entry_sz;
+	u8 cqe_sz;
 };
 
 struct hns_roce_query_pf_caps_b {
@@ -1581,9 +1603,9 @@ struct hns_roce_query_pf_caps_b {
 	u8 cqc_entry_sz;
 	u8 srqc_entry_sz;
 	u8 idx_entry_sz;
-	u8 scc_ctx_entry_sz;
+	u8 sccc_sz;
 	u8 max_mtu;
-	__le16 qpc_entry_sz;
+	__le16 qpc_sz;
 	__le16 qpc_timer_entry_sz;
 	__le16 cqc_timer_entry_sz;
 	u8 min_cqes;
@@ -1777,8 +1799,8 @@ struct hns_roce_eq_context {
 	__le32	byte_28;
 	__le32	byte_32;
 	__le32	byte_36;
-	__le32	nxt_eqe_ba0;
+	__le32	byte_40;
-	__le32	nxt_eqe_ba1;
+	__le32	byte_44;
 	__le32	rsv[5];
 };
 
@@ -1920,6 +1942,9 @@ struct hns_roce_eq_context {
 #define HNS_ROCE_EQC_NXT_EQE_BA_H_S 0
 #define HNS_ROCE_EQC_NXT_EQE_BA_H_M GENMASK(19, 0)
 
+#define HNS_ROCE_EQC_EQE_SIZE_S 20
+#define HNS_ROCE_EQC_EQE_SIZE_M GENMASK(21, 20)
+
 #define HNS_ROCE_V2_CEQE_COMP_CQN_S 0
 #define HNS_ROCE_V2_CEQE_COMP_CQN_M GENMASK(23, 0)
 
@@ -1941,6 +1966,8 @@ struct hns_roce_eq_context {
 #define HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_S 0
 #define HNS_ROCE_V2_AEQE_EVENT_QUEUE_NUM_M GENMASK(23, 0)
 
+#define MAX_SERVICE_LEVEL 0x7
+
 struct hns_roce_wqe_atomic_seg {
 	__le64          fetchadd_swap_data;
 	__le64          cmp_data;

drivers/infiniband/hw/hns/hns_roce_main.c

@@ -141,8 +141,8 @@ static int hns_roce_netdev_event(struct notifier_block *self,
 	struct net_device *dev = netdev_notifier_info_to_dev(ptr);
 	struct hns_roce_ib_iboe *iboe = NULL;
 	struct hns_roce_dev *hr_dev = NULL;
-	u8 port = 0;
+	int ret;
-	int ret = 0;
+	u8 port;
 
 	hr_dev = container_of(self, struct hns_roce_dev, iboe.nb);
 	iboe = &hr_dev->iboe;
@@ -323,6 +323,8 @@ static int hns_roce_alloc_ucontext(struct ib_ucontext *uctx,
 		mutex_init(&context->page_mutex);
 	}
 
+	resp.cqe_size = hr_dev->caps.cqe_sz;
+
 	ret = ib_copy_to_udata(udata, &resp, sizeof(resp));
 	if (ret)
 		goto error_fail_copy_to_udata;
@@ -454,6 +456,8 @@ static const struct ib_device_ops hns_roce_dev_mr_ops = {
 static const struct ib_device_ops hns_roce_dev_mw_ops = {
 	.alloc_mw = hns_roce_alloc_mw,
 	.dealloc_mw = hns_roce_dealloc_mw,
+
+	INIT_RDMA_OBJ_SIZE(ib_mw, hns_roce_mw, ibmw),
 };
 
 static const struct ib_device_ops hns_roce_dev_frmr_ops = {
@@ -545,7 +549,8 @@ static int hns_roce_register_device(struct hns_roce_dev *hr_dev)
 		if (ret)
 			return ret;
 	}
-	ret = ib_register_device(ib_dev, "hns_%d");
+	dma_set_max_seg_size(dev, UINT_MAX);
+	ret = ib_register_device(ib_dev, "hns_%d", dev);
 	if (ret) {
 		dev_err(dev, "ib_register_device failed!\n");
 		return ret;
@@ -587,7 +592,7 @@ static int hns_roce_init_hem(struct hns_roce_dev *hr_dev)
 	}
 
 	ret = hns_roce_init_hem_table(hr_dev, &hr_dev->qp_table.qp_table,
-				      HEM_TYPE_QPC, hr_dev->caps.qpc_entry_sz,
+				      HEM_TYPE_QPC, hr_dev->caps.qpc_sz,
 				      hr_dev->caps.num_qps, 1);
 	if (ret) {
 		dev_err(dev, "Failed to init QP context memory, aborting.\n");
@@ -638,11 +643,11 @@ static int hns_roce_init_hem(struct hns_roce_dev *hr_dev)
 		}
 	}
 
-	if (hr_dev->caps.sccc_entry_sz) {
+	if (hr_dev->caps.sccc_sz) {
 		ret = hns_roce_init_hem_table(hr_dev,
 					      &hr_dev->qp_table.sccc_table,
 					      HEM_TYPE_SCCC,
-					      hr_dev->caps.sccc_entry_sz,
+					      hr_dev->caps.sccc_sz,
 					      hr_dev->caps.num_qps, 1);
 		if (ret) {
 			dev_err(dev,
@@ -682,7 +687,7 @@ err_unmap_qpc_timer:
 		hns_roce_cleanup_hem_table(hr_dev, &hr_dev->qpc_timer_table);
 
 err_unmap_ctx:
-	if (hr_dev->caps.sccc_entry_sz)
+	if (hr_dev->caps.sccc_sz)
 		hns_roce_cleanup_hem_table(hr_dev,
 					   &hr_dev->qp_table.sccc_table);
 err_unmap_srq:

drivers/infiniband/hw/hns/hns_roce_mr.c

@@ -589,28 +589,22 @@ err_table:
 	return ret;
 }
 
-struct ib_mw *hns_roce_alloc_mw(struct ib_pd *ib_pd, enum ib_mw_type type,
+int hns_roce_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
-				struct ib_udata *udata)
 {
-	struct hns_roce_dev *hr_dev = to_hr_dev(ib_pd->device);
+	struct hns_roce_dev *hr_dev = to_hr_dev(ibmw->device);
-	struct hns_roce_mw *mw;
+	struct hns_roce_mw *mw = to_hr_mw(ibmw);
 	unsigned long index = 0;
 	int ret;
 
-	mw = kmalloc(sizeof(*mw), GFP_KERNEL);
-	if (!mw)
-		return ERR_PTR(-ENOMEM);
-
 	/* Allocate a key for mw from bitmap */
 	ret = hns_roce_bitmap_alloc(&hr_dev->mr_table.mtpt_bitmap, &index);
 	if (ret)
-		goto err_bitmap;
+		return ret;
 
 	mw->rkey = hw_index_to_key(index);
 
-	mw->ibmw.rkey = mw->rkey;
+	ibmw->rkey = mw->rkey;
-	mw->ibmw.type = type;
+	mw->pdn = to_hr_pd(ibmw->pd)->pdn;
-	mw->pdn = to_hr_pd(ib_pd)->pdn;
 	mw->pbl_hop_num = hr_dev->caps.pbl_hop_num;
 	mw->pbl_ba_pg_sz = hr_dev->caps.pbl_ba_pg_sz;
 	mw->pbl_buf_pg_sz = hr_dev->caps.pbl_buf_pg_sz;
@@ -619,15 +613,11 @@ struct ib_mw *hns_roce_alloc_mw(struct ib_pd *ib_pd, enum ib_mw_type type,
 	if (ret)
 		goto err_mw;
 
-	return &mw->ibmw;
+	return 0;
 
 err_mw:
 	hns_roce_mw_free(hr_dev, mw);
-
+	return ret;
-err_bitmap:
-	kfree(mw);
-
-	return ERR_PTR(ret);
 }
 
 int hns_roce_dealloc_mw(struct ib_mw *ibmw)
@@ -636,8 +626,6 @@ int hns_roce_dealloc_mw(struct ib_mw *ibmw)
 	struct hns_roce_mw *mw = to_hr_mw(ibmw);
 
 	hns_roce_mw_free(hr_dev, mw);
-	kfree(mw);
-
 	return 0;
 }
 
@@ -707,19 +695,6 @@ static inline size_t mtr_bufs_size(struct hns_roce_buf_attr *attr)
 	return size;
 }
 
-static inline int mtr_umem_page_count(struct ib_umem *umem,
-				      unsigned int page_shift)
-{
-	int count = ib_umem_page_count(umem);
-
-	if (page_shift >= PAGE_SHIFT)
-		count >>= page_shift - PAGE_SHIFT;
-	else
-		count <<= PAGE_SHIFT - page_shift;
-
-	return count;
-}
-
 static inline size_t mtr_kmem_direct_size(bool is_direct, size_t alloc_size,
 					  unsigned int page_shift)
 {
@@ -767,13 +742,11 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr,
 			  struct ib_udata *udata, unsigned long user_addr)
 {
 	struct ib_device *ibdev = &hr_dev->ib_dev;
-	unsigned int max_pg_shift = buf_attr->page_shift;
+	unsigned int best_pg_shift;
-	unsigned int best_pg_shift = 0;
 	int all_pg_count = 0;
 	size_t direct_size;
 	size_t total_size;
-	unsigned long tmp;
+	int ret;
-	int ret = 0;
 
 	total_size = mtr_bufs_size(buf_attr);
 	if (total_size < 1) {
@@ -782,6 +755,9 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr,
 	}
 
 	if (udata) {
+		unsigned long pgsz_bitmap;
+		unsigned long page_size;
+
 		mtr->kmem = NULL;
 		mtr->umem = ib_umem_get(ibdev, user_addr, total_size,
 					buf_attr->user_access);
@@ -790,15 +766,17 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr,
 				  PTR_ERR(mtr->umem));
 			return -ENOMEM;
 		}
-		if (buf_attr->fixed_page) {
+		if (buf_attr->fixed_page)
-			best_pg_shift = max_pg_shift;
+			pgsz_bitmap = 1 << buf_attr->page_shift;
-		} else {
+		else
-			tmp = GENMASK(max_pg_shift, 0);
+			pgsz_bitmap = GENMASK(buf_attr->page_shift, PAGE_SHIFT);
-			ret = ib_umem_find_best_pgsz(mtr->umem, tmp, user_addr);
+
-			best_pg_shift = (ret <= PAGE_SIZE) ?
+		page_size = ib_umem_find_best_pgsz(mtr->umem, pgsz_bitmap,
-					PAGE_SHIFT : ilog2(ret);
+						   user_addr);
-		}
+		if (!page_size)
-		all_pg_count = mtr_umem_page_count(mtr->umem, best_pg_shift);
+			return -EINVAL;
+		best_pg_shift = order_base_2(page_size);
+		all_pg_count = ib_umem_num_dma_blocks(mtr->umem, page_size);
 		ret = 0;
 	} else {
 		mtr->umem = NULL;
@@ -808,16 +786,15 @@ static int mtr_alloc_bufs(struct hns_roce_dev *hr_dev, struct hns_roce_mtr *mtr,
 			return -ENOMEM;
 		}
 		direct_size = mtr_kmem_direct_size(is_direct, total_size,
-						   max_pg_shift);
+						   buf_attr->page_shift);
 		ret = hns_roce_buf_alloc(hr_dev, total_size, direct_size,
-					 mtr->kmem, max_pg_shift);
+					 mtr->kmem, buf_attr->page_shift);
 		if (ret) {
 			ibdev_err(ibdev, "Failed to alloc kmem, ret %d\n", ret);
 			goto err_alloc_mem;
-		} else {
-			best_pg_shift = max_pg_shift;
-			all_pg_count = mtr->kmem->npages;
 		}
+		best_pg_shift = buf_attr->page_shift;
+		all_pg_count = mtr->kmem->npages;
 	}
 
 	/* must bigger than minimum hardware page shift */
@@ -967,7 +944,7 @@ static int mtr_init_buf_cfg(struct hns_roce_dev *hr_dev,
 			    unsigned int *buf_page_shift)
 {
 	struct hns_roce_buf_region *r;
-	unsigned int page_shift = 0;
+	unsigned int page_shift;
 	int page_cnt = 0;
 	size_t buf_size;
 	int region_cnt;

drivers/infiniband/hw/hns/hns_roce_pd.c

@@ -82,9 +82,10 @@ int hns_roce_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 	return 0;
 }
 
-void hns_roce_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
+int hns_roce_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
 {
 	hns_roce_pd_free(to_hr_dev(pd->device), to_hr_pd(pd)->pdn);
+	return 0;
 }
 
 int hns_roce_uar_alloc(struct hns_roce_dev *hr_dev, struct hns_roce_uar *uar)

drivers/infiniband/hw/hns/hns_roce_qp.c

@@ -41,8 +41,6 @@
 #include "hns_roce_hem.h"
 #include <rdma/hns-abi.h>
 
-#define SQP_NUM				(2 * HNS_ROCE_MAX_PORTS)
-
 static void flush_work_handle(struct work_struct *work)
 {
 	struct hns_roce_work *flush_work = container_of(work,
@@ -288,7 +286,7 @@ static int alloc_qpc(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp)
 		}
 	}
 
-	if (hr_dev->caps.sccc_entry_sz) {
+	if (hr_dev->caps.sccc_sz) {
 		/* Alloc memory for SCC CTX */
 		ret = hns_roce_table_get(hr_dev, &qp_table->sccc_table,
 					 hr_qp->qpn);
@@ -551,10 +549,9 @@ static int set_kernel_sq_size(struct hns_roce_dev *hr_dev,
 	int ret;
 
 	if (!cap->max_send_wr || cap->max_send_wr > hr_dev->caps.max_wqes ||
-	    cap->max_send_sge > hr_dev->caps.max_sq_sg ||
+	    cap->max_send_sge > hr_dev->caps.max_sq_sg) {
-	    cap->max_inline_data > hr_dev->caps.max_sq_inline) {
 		ibdev_err(ibdev,
-			  "failed to check SQ WR, SGE or inline num, ret = %d.\n",
+			  "failed to check SQ WR or SGE num, ret = %d.\n",
 			  -EINVAL);
 		return -EINVAL;
 	}
@@ -577,9 +574,6 @@ static int set_kernel_sq_size(struct hns_roce_dev *hr_dev,
 	cap->max_send_wr = cnt;
 	cap->max_send_sge = hr_qp->sq.max_gs;
 
-	/* We don't support inline sends for kernel QPs (yet) */
-	cap->max_inline_data = 0;
-
 	return 0;
 }
 
@@ -847,6 +841,11 @@ static int set_qp_param(struct hns_roce_dev *hr_dev, struct hns_roce_qp *hr_qp,
 
 	hr_qp->ibqp.qp_type = init_attr->qp_type;
 
+	if (init_attr->cap.max_inline_data > hr_dev->caps.max_sq_inline)
+		init_attr->cap.max_inline_data = hr_dev->caps.max_sq_inline;
+
+	hr_qp->max_inline_data = init_attr->cap.max_inline_data;
+
 	if (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR)
 		hr_qp->sq_signal_bits = IB_SIGNAL_ALL_WR;
 	else
@@ -1014,53 +1013,32 @@ struct ib_qp *hns_roce_create_qp(struct ib_pd *pd,
 	int ret;
 
 	switch (init_attr->qp_type) {
-	case IB_QPT_RC: {
+	case IB_QPT_RC:
-		hr_qp = kzalloc(sizeof(*hr_qp), GFP_KERNEL);
+	case IB_QPT_GSI:
-		if (!hr_qp)
-			return ERR_PTR(-ENOMEM);
-
-		ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata,
-						hr_qp);
-		if (ret) {
-			ibdev_err(ibdev, "Create QP 0x%06lx failed(%d)\n",
-				  hr_qp->qpn, ret);
-			kfree(hr_qp);
-			return ERR_PTR(ret);
-		}
-
 		break;
-	}
+	default:
-	case IB_QPT_GSI: {
-		/* Userspace is not allowed to create special QPs: */
-		if (udata) {
-			ibdev_err(ibdev, "not support usr space GSI\n");
-			return ERR_PTR(-EINVAL);
-		}
-
-		hr_qp = kzalloc(sizeof(*hr_qp), GFP_KERNEL);
-		if (!hr_qp)
-			return ERR_PTR(-ENOMEM);
-
-		hr_qp->port = init_attr->port_num - 1;
-		hr_qp->phy_port = hr_dev->iboe.phy_port[hr_qp->port];
-
-		ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata,
-						hr_qp);
-		if (ret) {
-			ibdev_err(ibdev, "Create GSI QP failed!\n");
-			kfree(hr_qp);
-			return ERR_PTR(ret);
-		}
-
-		break;
-	}
-	default:{
 		ibdev_err(ibdev, "not support QP type %d\n",
 			  init_attr->qp_type);
 		return ERR_PTR(-EOPNOTSUPP);
 	}
+
+	hr_qp = kzalloc(sizeof(*hr_qp), GFP_KERNEL);
+	if (!hr_qp)
+		return ERR_PTR(-ENOMEM);
+
+	if (init_attr->qp_type == IB_QPT_GSI) {
+		hr_qp->port = init_attr->port_num - 1;
+		hr_qp->phy_port = hr_dev->iboe.phy_port[hr_qp->port];
 	}
 
+	ret = hns_roce_create_qp_common(hr_dev, pd, init_attr, udata, hr_qp);
+	if (ret) {
+		ibdev_err(ibdev, "Create QP type 0x%x failed(%d)\n",
+			  init_attr->qp_type, ret);
+		ibdev_err(ibdev, "Create GSI QP failed!\n");
+		kfree(hr_qp);
+		return ERR_PTR(ret);
+	}
 	return &hr_qp->ibqp;
 }
 
@@ -1161,8 +1139,10 @@ int hns_roce_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 
 	mutex_lock(&hr_qp->mutex);
 
-	cur_state = attr_mask & IB_QP_CUR_STATE ?
+	if (attr_mask & IB_QP_CUR_STATE && attr->cur_qp_state != hr_qp->state)
-		    attr->cur_qp_state : (enum ib_qp_state)hr_qp->state;
+		goto out;
+
+	cur_state = hr_qp->state;
 	new_state = attr_mask & IB_QP_STATE ? attr->qp_state : cur_state;
 
 	if (ibqp->uobject &&

drivers/infiniband/hw/hns/hns_roce_srq.c

@@ -285,7 +285,7 @@ int hns_roce_create_srq(struct ib_srq *ib_srq,
 	struct hns_roce_srq *srq = to_hr_srq(ib_srq);
 	struct ib_device *ibdev = &hr_dev->ib_dev;
 	struct hns_roce_ib_create_srq ucmd = {};
-	int ret = 0;
+	int ret;
 	u32 cqn;
 
 	/* Check the actual SRQ wqe and SRQ sge num */
@@ -363,7 +363,7 @@ err_buf_alloc:
 	return ret;
 }
 
-void hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
+int hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
 {
 	struct hns_roce_dev *hr_dev = to_hr_dev(ibsrq->device);
 	struct hns_roce_srq *srq = to_hr_srq(ibsrq);
@@ -372,6 +372,7 @@ void hns_roce_destroy_srq(struct ib_srq *ibsrq, struct ib_udata *udata)
 	free_srq_idx(hr_dev, srq);
 	free_srq_wrid(srq);
 	free_srq_buf(hr_dev, srq);
+	return 0;
 }
 
 int hns_roce_init_srq_table(struct hns_roce_dev *hr_dev)

drivers/infiniband/hw/i40iw/i40iw.h

@@ -409,8 +409,8 @@ static inline struct i40iw_qp *to_iwqp(struct ib_qp *ibqp)
 }
 
 /* i40iw.c */
-void i40iw_add_ref(struct ib_qp *);
+void i40iw_qp_add_ref(struct ib_qp *ibqp);
-void i40iw_rem_ref(struct ib_qp *);
+void i40iw_qp_rem_ref(struct ib_qp *ibqp);
 struct ib_qp *i40iw_get_qp(struct ib_device *, int);
 
 void i40iw_flush_wqes(struct i40iw_device *iwdev,
@@ -554,9 +554,8 @@ enum i40iw_status_code i40iw_manage_qhash(struct i40iw_device *iwdev,
 					  bool wait);
 void i40iw_receive_ilq(struct i40iw_sc_vsi *vsi, struct i40iw_puda_buf *rbuf);
 void i40iw_free_sqbuf(struct i40iw_sc_vsi *vsi, void *bufp);
-void i40iw_free_qp_resources(struct i40iw_device *iwdev,
+void i40iw_free_qp_resources(struct i40iw_qp *iwqp);
-			     struct i40iw_qp *iwqp,
+
-			     u32 qp_num);
 enum i40iw_status_code i40iw_obj_aligned_mem(struct i40iw_device *iwdev,
 					     struct i40iw_dma_mem *memptr,
 					     u32 size, u32 mask);

drivers/infiniband/hw/i40iw/i40iw_cm.c

@@ -2322,7 +2322,7 @@ static void i40iw_rem_ref_cm_node(struct i40iw_cm_node *cm_node)
 	iwqp = cm_node->iwqp;
 	if (iwqp) {
 		iwqp->cm_node = NULL;
-		i40iw_rem_ref(&iwqp->ibqp);
+		i40iw_qp_rem_ref(&iwqp->ibqp);
 		cm_node->iwqp = NULL;
 	} else if (cm_node->qhash_set) {
 		i40iw_get_addr_info(cm_node, &nfo);
@@ -3452,7 +3452,7 @@ void i40iw_cm_disconn(struct i40iw_qp *iwqp)
 		kfree(work);
 		return;
 	}
-	i40iw_add_ref(&iwqp->ibqp);
+	i40iw_qp_add_ref(&iwqp->ibqp);
 	spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 
 	work->iwqp = iwqp;
@@ -3623,7 +3623,7 @@ static void i40iw_disconnect_worker(struct work_struct *work)
 
 	kfree(dwork);
 	i40iw_cm_disconn_true(iwqp);
-	i40iw_rem_ref(&iwqp->ibqp);
+	i40iw_qp_rem_ref(&iwqp->ibqp);
 }
 
 /**
@@ -3745,7 +3745,7 @@ int i40iw_accept(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	cm_node->lsmm_size = accept.size + conn_param->private_data_len;
 	i40iw_cm_init_tsa_conn(iwqp, cm_node);
 	cm_id->add_ref(cm_id);
-	i40iw_add_ref(&iwqp->ibqp);
+	i40iw_qp_add_ref(&iwqp->ibqp);
 
 	attr.qp_state = IB_QPS_RTS;
 	cm_node->qhash_set = false;
@@ -3908,7 +3908,7 @@ int i40iw_connect(struct iw_cm_id *cm_id, struct iw_cm_conn_param *conn_param)
 	iwqp->cm_node = cm_node;
 	cm_node->iwqp = iwqp;
 	iwqp->cm_id = cm_id;
-	i40iw_add_ref(&iwqp->ibqp);
+	i40iw_qp_add_ref(&iwqp->ibqp);
 
 	if (cm_node->state != I40IW_CM_STATE_OFFLOADED) {
 		cm_node->state = I40IW_CM_STATE_SYN_SENT;

drivers/infiniband/hw/i40iw/i40iw_hw.c

@@ -313,7 +313,7 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 					    __func__, info->qp_cq_id);
 				continue;
 			}
-			i40iw_add_ref(&iwqp->ibqp);
+			i40iw_qp_add_ref(&iwqp->ibqp);
 			spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 			qp = &iwqp->sc_qp;
 			spin_lock_irqsave(&iwqp->lock, flags);
@@ -426,7 +426,7 @@ void i40iw_process_aeq(struct i40iw_device *iwdev)
 			break;
 		}
 		if (info->qp)
-			i40iw_rem_ref(&iwqp->ibqp);
+			i40iw_qp_rem_ref(&iwqp->ibqp);
 	} while (1);
 
 	if (aeqcnt)

drivers/infiniband/hw/i40iw/i40iw_main.c

@@ -192,9 +192,9 @@ static void i40iw_enable_intr(struct i40iw_sc_dev *dev, u32 msix_id)
  * i40iw_dpc - tasklet for aeq and ceq 0
  * @data: iwarp device
  */
-static void i40iw_dpc(unsigned long data)
+static void i40iw_dpc(struct tasklet_struct *t)
 {
-	struct i40iw_device *iwdev = (struct i40iw_device *)data;
+	struct i40iw_device *iwdev = from_tasklet(iwdev, t, dpc_tasklet);
 
 	if (iwdev->msix_shared)
 		i40iw_process_ceq(iwdev, iwdev->ceqlist);
@@ -206,9 +206,9 @@ static void i40iw_dpc(unsigned long data)
  * i40iw_ceq_dpc - dpc handler for CEQ
  * @data: data points to CEQ
  */
-static void i40iw_ceq_dpc(unsigned long data)
+static void i40iw_ceq_dpc(struct tasklet_struct *t)
 {
-	struct i40iw_ceq *iwceq = (struct i40iw_ceq *)data;
+	struct i40iw_ceq *iwceq = from_tasklet(iwceq, t, dpc_tasklet);
 	struct i40iw_device *iwdev = iwceq->iwdev;
 
 	i40iw_process_ceq(iwdev, iwceq);
@@ -689,10 +689,10 @@ static enum i40iw_status_code i40iw_configure_ceq_vector(struct i40iw_device *iw
 	enum i40iw_status_code status;
 
 	if (iwdev->msix_shared && !ceq_id) {
-		tasklet_init(&iwdev->dpc_tasklet, i40iw_dpc, (unsigned long)iwdev);
+		tasklet_setup(&iwdev->dpc_tasklet, i40iw_dpc);
 		status = request_irq(msix_vec->irq, i40iw_irq_handler, 0, "AEQCEQ", iwdev);
 	} else {
-		tasklet_init(&iwceq->dpc_tasklet, i40iw_ceq_dpc, (unsigned long)iwceq);
+		tasklet_setup(&iwceq->dpc_tasklet, i40iw_ceq_dpc);
 		status = request_irq(msix_vec->irq, i40iw_ceq_handler, 0, "CEQ", iwceq);
 	}
 
@@ -841,7 +841,7 @@ static enum i40iw_status_code i40iw_configure_aeq_vector(struct i40iw_device *iw
 	u32 ret = 0;
 
 	if (!iwdev->msix_shared) {
-		tasklet_init(&iwdev->dpc_tasklet, i40iw_dpc, (unsigned long)iwdev);
+		tasklet_setup(&iwdev->dpc_tasklet, i40iw_dpc);
 		ret = request_irq(msix_vec->irq, i40iw_irq_handler, 0, "i40iw", iwdev);
 	}
 	if (ret) {
@@ -1573,7 +1573,7 @@ static enum i40iw_status_code i40iw_setup_init_state(struct i40iw_handler *hdl,
 	status = i40iw_save_msix_info(iwdev, ldev);
 	if (status)
 		return status;
-	iwdev->hw.dev_context = (void *)ldev->pcidev;
+	iwdev->hw.pcidev = ldev->pcidev;
 	iwdev->hw.hw_addr = ldev->hw_addr;
 	status = i40iw_allocate_dma_mem(&iwdev->hw,
 					&iwdev->obj_mem, 8192, 4096);

drivers/infiniband/hw/i40iw/i40iw_pble.c

@@ -167,7 +167,7 @@ static enum i40iw_status_code add_sd_direct(struct i40iw_sc_dev *dev,
  */
 static void i40iw_free_vmalloc_mem(struct i40iw_hw *hw, struct i40iw_chunk *chunk)
 {
-	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	struct pci_dev *pcidev = hw->pcidev;
 	int i;
 
 	if (!chunk->pg_cnt)
@@ -193,7 +193,7 @@ static enum i40iw_status_code i40iw_get_vmalloc_mem(struct i40iw_hw *hw,
 						    struct i40iw_chunk *chunk,
 						    int pg_cnt)
 {
-	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	struct pci_dev *pcidev = hw->pcidev;
 	struct page *page;
 	u8 *addr;
 	u32 size;

drivers/infiniband/hw/i40iw/i40iw_type.h

@@ -73,6 +73,7 @@ struct i40iw_pd_ops;
 struct i40iw_priv_qp_ops;
 struct i40iw_priv_cq_ops;
 struct i40iw_hmc_ops;
+struct pci_dev;
 
 enum i40iw_page_size {
 	I40IW_PAGE_SIZE_4K,
@@ -261,7 +262,7 @@ struct i40iw_vsi_pestat {
 
 struct i40iw_hw {
 	u8 __iomem *hw_addr;
-	void *dev_context;
+	struct pci_dev *pcidev;
 	struct i40iw_hmc_info hmc;
 };
 

drivers/infiniband/hw/i40iw/i40iw_utils.c

@@ -477,25 +477,6 @@ void i40iw_cleanup_pending_cqp_op(struct i40iw_device *iwdev)
 	}
 }
 
-/**
- * i40iw_free_qp - callback after destroy cqp completes
- * @cqp_request: cqp request for destroy qp
- * @num: not used
- */
-static void i40iw_free_qp(struct i40iw_cqp_request *cqp_request, u32 num)
-{
-	struct i40iw_sc_qp *qp = (struct i40iw_sc_qp *)cqp_request->param;
-	struct i40iw_qp *iwqp = (struct i40iw_qp *)qp->back_qp;
-	struct i40iw_device *iwdev;
-	u32 qp_num = iwqp->ibqp.qp_num;
-
-	iwdev = iwqp->iwdev;
-
-	i40iw_rem_pdusecount(iwqp->iwpd, iwdev);
-	i40iw_free_qp_resources(iwdev, iwqp, qp_num);
-	i40iw_rem_devusecount(iwdev);
-}
-
 /**
  * i40iw_wait_event - wait for completion
  * @iwdev: iwarp device
@@ -616,26 +597,23 @@ void i40iw_rem_pdusecount(struct i40iw_pd *iwpd, struct i40iw_device *iwdev)
 }
 
 /**
- * i40iw_add_ref - add refcount for qp
+ * i40iw_qp_add_ref - add refcount for qp
  * @ibqp: iqarp qp
  */
-void i40iw_add_ref(struct ib_qp *ibqp)
+void i40iw_qp_add_ref(struct ib_qp *ibqp)
 {
 	struct i40iw_qp *iwqp = (struct i40iw_qp *)ibqp;
 
-	atomic_inc(&iwqp->refcount);
+	refcount_inc(&iwqp->refcount);
 }
 
 /**
- * i40iw_rem_ref - rem refcount for qp and free if 0
+ * i40iw_qp_rem_ref - rem refcount for qp and free if 0
  * @ibqp: iqarp qp
  */
-void i40iw_rem_ref(struct ib_qp *ibqp)
+void i40iw_qp_rem_ref(struct ib_qp *ibqp)
 {
 	struct i40iw_qp *iwqp;
-	enum i40iw_status_code status;
-	struct i40iw_cqp_request *cqp_request;
-	struct cqp_commands_info *cqp_info;
 	struct i40iw_device *iwdev;
 	u32 qp_num;
 	unsigned long flags;
@@ -643,7 +621,7 @@ void i40iw_rem_ref(struct ib_qp *ibqp)
 	iwqp = to_iwqp(ibqp);
 	iwdev = iwqp->iwdev;
 	spin_lock_irqsave(&iwdev->qptable_lock, flags);
-	if (!atomic_dec_and_test(&iwqp->refcount)) {
+	if (!refcount_dec_and_test(&iwqp->refcount)) {
 		spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
 		return;
 	}
@@ -651,25 +629,8 @@ void i40iw_rem_ref(struct ib_qp *ibqp)
 	qp_num = iwqp->ibqp.qp_num;
 	iwdev->qp_table[qp_num] = NULL;
 	spin_unlock_irqrestore(&iwdev->qptable_lock, flags);
-	cqp_request = i40iw_get_cqp_request(&iwdev->cqp, false);
+	complete(&iwqp->free_qp);
-	if (!cqp_request)
-		return;
 
-	cqp_request->callback_fcn = i40iw_free_qp;
-	cqp_request->param = (void *)&iwqp->sc_qp;
-	cqp_info = &cqp_request->info;
-	cqp_info->cqp_cmd = OP_QP_DESTROY;
-	cqp_info->post_sq = 1;
-	cqp_info->in.u.qp_destroy.qp = &iwqp->sc_qp;
-	cqp_info->in.u.qp_destroy.scratch = (uintptr_t)cqp_request;
-	cqp_info->in.u.qp_destroy.remove_hash_idx = true;
-	status = i40iw_handle_cqp_op(iwdev, cqp_request);
-	if (!status)
-		return;
-
-	i40iw_rem_pdusecount(iwqp->iwpd, iwdev);
-	i40iw_free_qp_resources(iwdev, iwqp, qp_num);
-	i40iw_rem_devusecount(iwdev);
 }
 
 /**
@@ -751,7 +712,7 @@ enum i40iw_status_code i40iw_allocate_dma_mem(struct i40iw_hw *hw,
 					      u64 size,
 					      u32 alignment)
 {
-	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	struct pci_dev *pcidev = hw->pcidev;
 
 	if (!mem)
 		return I40IW_ERR_PARAM;
@@ -770,7 +731,7 @@ enum i40iw_status_code i40iw_allocate_dma_mem(struct i40iw_hw *hw,
  */
 void i40iw_free_dma_mem(struct i40iw_hw *hw, struct i40iw_dma_mem *mem)
 {
-	struct pci_dev *pcidev = (struct pci_dev *)hw->dev_context;
+	struct pci_dev *pcidev = hw->pcidev;
 
 	if (!mem || !mem->va)
 		return;
@@ -936,7 +897,7 @@ static void i40iw_terminate_timeout(struct timer_list *t)
 	struct i40iw_sc_qp *qp = (struct i40iw_sc_qp *)&iwqp->sc_qp;
 
 	i40iw_terminate_done(qp, 1);
-	i40iw_rem_ref(&iwqp->ibqp);
+	i40iw_qp_rem_ref(&iwqp->ibqp);
 }
 
 /**
@@ -948,7 +909,7 @@ void i40iw_terminate_start_timer(struct i40iw_sc_qp *qp)
 	struct i40iw_qp *iwqp;
 
 	iwqp = (struct i40iw_qp *)qp->back_qp;
-	i40iw_add_ref(&iwqp->ibqp);
+	i40iw_qp_add_ref(&iwqp->ibqp);
 	timer_setup(&iwqp->terminate_timer, i40iw_terminate_timeout, 0);
 	iwqp->terminate_timer.expires = jiffies + HZ;
 	add_timer(&iwqp->terminate_timer);
@@ -964,7 +925,7 @@ void i40iw_terminate_del_timer(struct i40iw_sc_qp *qp)
 
 	iwqp = (struct i40iw_qp *)qp->back_qp;
 	if (del_timer(&iwqp->terminate_timer))
-		i40iw_rem_ref(&iwqp->ibqp);
+		i40iw_qp_rem_ref(&iwqp->ibqp);
 }
 
 /**

drivers/infiniband/hw/i40iw/i40iw_verbs.c

@@ -328,12 +328,13 @@ error:
  * @ibpd: ptr of pd to be deallocated
  * @udata: user data or null for kernel object
  */
-static void i40iw_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
+static int i40iw_dealloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 {
 	struct i40iw_pd *iwpd = to_iwpd(ibpd);
 	struct i40iw_device *iwdev = to_iwdev(ibpd->device);
 
 	i40iw_rem_pdusecount(iwpd, iwdev);
+	return 0;
 }
 
 /**
@@ -363,11 +364,11 @@ static struct i40iw_pbl *i40iw_get_pbl(unsigned long va,
  * @iwqp: qp ptr (user or kernel)
  * @qp_num: qp number assigned
  */
-void i40iw_free_qp_resources(struct i40iw_device *iwdev,
+void i40iw_free_qp_resources(struct i40iw_qp *iwqp)
-			     struct i40iw_qp *iwqp,
-			     u32 qp_num)
 {
 	struct i40iw_pbl *iwpbl = &iwqp->iwpbl;
+	struct i40iw_device *iwdev = iwqp->iwdev;
+	u32 qp_num = iwqp->ibqp.qp_num;
 
 	i40iw_ieq_cleanup_qp(iwdev->vsi.ieq, &iwqp->sc_qp);
 	i40iw_dealloc_push_page(iwdev, &iwqp->sc_qp);
@@ -379,7 +380,7 @@ void i40iw_free_qp_resources(struct i40iw_device *iwdev,
 	i40iw_free_dma_mem(iwdev->sc_dev.hw, &iwqp->kqp.dma_mem);
 	kfree(iwqp->kqp.wrid_mem);
 	iwqp->kqp.wrid_mem = NULL;
-	kfree(iwqp->allocated_buffer);
+	kfree(iwqp);
 }
 
 /**
@@ -401,6 +402,10 @@ static void i40iw_clean_cqes(struct i40iw_qp *iwqp, struct i40iw_cq *iwcq)
 static int i40iw_destroy_qp(struct ib_qp *ibqp, struct ib_udata *udata)
 {
 	struct i40iw_qp *iwqp = to_iwqp(ibqp);
+	struct ib_qp_attr attr;
+	struct i40iw_device *iwdev = iwqp->iwdev;
+
+	memset(&attr, 0, sizeof(attr));
 
 	iwqp->destroyed = 1;
 
@@ -415,7 +420,15 @@ static int i40iw_destroy_qp(struct ib_qp *ibqp, struct ib_udata *udata)
 		}
 	}
 
-	i40iw_rem_ref(&iwqp->ibqp);
+	attr.qp_state = IB_QPS_ERR;
+	i40iw_modify_qp(&iwqp->ibqp, &attr, IB_QP_STATE, NULL);
+	i40iw_qp_rem_ref(&iwqp->ibqp);
+	wait_for_completion(&iwqp->free_qp);
+	i40iw_cqp_qp_destroy_cmd(&iwdev->sc_dev, &iwqp->sc_qp);
+	i40iw_rem_pdusecount(iwqp->iwpd, iwdev);
+	i40iw_free_qp_resources(iwqp);
+	i40iw_rem_devusecount(iwdev);
+
 	return 0;
 }
 
@@ -524,7 +537,6 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	struct i40iw_create_qp_req req;
 	struct i40iw_create_qp_resp uresp;
 	u32 qp_num = 0;
-	void *mem;
 	enum i40iw_status_code ret;
 	int err_code;
 	int sq_size;
@@ -566,16 +578,15 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	init_info.qp_uk_init_info.max_rq_frag_cnt = init_attr->cap.max_recv_sge;
 	init_info.qp_uk_init_info.max_inline_data = init_attr->cap.max_inline_data;
 
-	mem = kzalloc(sizeof(*iwqp), GFP_KERNEL);
+	iwqp = kzalloc(sizeof(*iwqp), GFP_KERNEL);
-	if (!mem)
+	if (!iwqp)
 		return ERR_PTR(-ENOMEM);
 
-	iwqp = (struct i40iw_qp *)mem;
-	iwqp->allocated_buffer = mem;
 	qp = &iwqp->sc_qp;
 	qp->back_qp = (void *)iwqp;
 	qp->push_idx = I40IW_INVALID_PUSH_PAGE_INDEX;
 
+	iwqp->iwdev = iwdev;
 	iwqp->ctx_info.iwarp_info = &iwqp->iwarp_info;
 
 	if (i40iw_allocate_dma_mem(dev->hw,
@@ -600,7 +611,6 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 		goto error;
 	}
 
-	iwqp->iwdev = iwdev;
 	iwqp->iwpd = iwpd;
 	iwqp->ibqp.qp_num = qp_num;
 	qp = &iwqp->sc_qp;
@@ -714,7 +724,7 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 		goto error;
 	}
 
-	i40iw_add_ref(&iwqp->ibqp);
+	refcount_set(&iwqp->refcount, 1);
 	spin_lock_init(&iwqp->lock);
 	iwqp->sig_all = (init_attr->sq_sig_type == IB_SIGNAL_ALL_WR) ? 1 : 0;
 	iwdev->qp_table[qp_num] = iwqp;
@@ -736,10 +746,11 @@ static struct ib_qp *i40iw_create_qp(struct ib_pd *ibpd,
 	}
 	init_completion(&iwqp->sq_drained);
 	init_completion(&iwqp->rq_drained);
+	init_completion(&iwqp->free_qp);
 
 	return &iwqp->ibqp;
 error:
-	i40iw_free_qp_resources(iwdev, iwqp, qp_num);
+	i40iw_free_qp_resources(iwqp);
 	return ERR_PTR(err_code);
 }
 
@@ -1052,7 +1063,7 @@ void i40iw_cq_wq_destroy(struct i40iw_device *iwdev, struct i40iw_sc_cq *cq)
  * @ib_cq: cq pointer
  * @udata: user data or NULL for kernel object
  */
-static void i40iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
+static int i40iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 {
 	struct i40iw_cq *iwcq;
 	struct i40iw_device *iwdev;
@@ -1064,6 +1075,7 @@ static void i40iw_destroy_cq(struct ib_cq *ib_cq, struct ib_udata *udata)
 	i40iw_cq_wq_destroy(iwdev, cq);
 	cq_free_resources(iwdev, iwcq);
 	i40iw_rem_devusecount(iwdev);
+	return 0;
 }
 
 /**
@@ -1320,8 +1332,7 @@ static void i40iw_copy_user_pgaddrs(struct i40iw_mr *iwmr,
 	if (iwmr->type == IW_MEMREG_TYPE_QP)
 		iwpbl->qp_mr.sq_page = sg_page(region->sg_head.sgl);
 
-	rdma_for_each_block(region->sg_head.sgl, &biter, region->nmap,
+	rdma_umem_for_each_dma_block(region, &biter, iwmr->page_size) {
-			    iwmr->page_size) {
 		*pbl = rdma_block_iter_dma_address(&biter);
 		pbl = i40iw_next_pbl_addr(pbl, &pinfo, &idx);
 	}
@@ -1744,15 +1755,12 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
 	struct i40iw_mr *iwmr;
 	struct ib_umem *region;
 	struct i40iw_mem_reg_req req;
-	u64 pbl_depth = 0;
 	u32 stag = 0;
 	u16 access;
-	u64 region_length;
 	bool use_pbles = false;
 	unsigned long flags;
 	int err = -ENOSYS;
 	int ret;
-	int pg_shift;
 
 	if (!udata)
 		return ERR_PTR(-EOPNOTSUPP);
@@ -1787,18 +1795,13 @@ static struct ib_mr *i40iw_reg_user_mr(struct ib_pd *pd,
 	if (req.reg_type == IW_MEMREG_TYPE_MEM)
 		iwmr->page_size = ib_umem_find_best_pgsz(region, SZ_4K | SZ_2M,
 							 virt);
-
-	region_length = region->length + (start & (iwmr->page_size - 1));
-	pg_shift = ffs(iwmr->page_size) - 1;
-	pbl_depth = region_length >> pg_shift;
-	pbl_depth += (region_length & (iwmr->page_size - 1)) ? 1 : 0;
 	iwmr->length = region->length;
 
 	iwpbl->user_base = virt;
 	palloc = &iwpbl->pble_alloc;
 
 	iwmr->type = req.reg_type;
-	iwmr->page_cnt = (u32)pbl_depth;
+	iwmr->page_cnt = ib_umem_num_dma_blocks(region, iwmr->page_size);
 
 	switch (req.reg_type) {
 	case IW_MEMREG_TYPE_QP:
@@ -2636,13 +2639,13 @@ static const struct ib_device_ops i40iw_dev_ops = {
 	.get_hw_stats = i40iw_get_hw_stats,
 	.get_port_immutable = i40iw_port_immutable,
 	.iw_accept = i40iw_accept,
-	.iw_add_ref = i40iw_add_ref,
+	.iw_add_ref = i40iw_qp_add_ref,
 	.iw_connect = i40iw_connect,
 	.iw_create_listen = i40iw_create_listen,
 	.iw_destroy_listen = i40iw_destroy_listen,
 	.iw_get_qp = i40iw_get_qp,
 	.iw_reject = i40iw_reject,
-	.iw_rem_ref = i40iw_rem_ref,
+	.iw_rem_ref = i40iw_qp_rem_ref,
 	.map_mr_sg = i40iw_map_mr_sg,
 	.mmap = i40iw_mmap,
 	.modify_qp = i40iw_modify_qp,
@@ -2668,7 +2671,7 @@ static struct i40iw_ib_device *i40iw_init_rdma_device(struct i40iw_device *iwdev
 {
 	struct i40iw_ib_device *iwibdev;
 	struct net_device *netdev = iwdev->netdev;
-	struct pci_dev *pcidev = (struct pci_dev *)iwdev->hw.dev_context;
+	struct pci_dev *pcidev = iwdev->hw.pcidev;
 
 	iwibdev = ib_alloc_device(i40iw_ib_device, ibdev);
 	if (!iwibdev) {
@@ -2758,7 +2761,8 @@ int i40iw_register_rdma_device(struct i40iw_device *iwdev)
 	if (ret)
 		goto error;
 
-	ret = ib_register_device(&iwibdev->ibdev, "i40iw%d");
+	dma_set_max_seg_size(&iwdev->hw.pcidev->dev, UINT_MAX);
+	ret = ib_register_device(&iwibdev->ibdev, "i40iw%d", &iwdev->hw.pcidev->dev);
 	if (ret)
 		goto error;
 

drivers/infiniband/hw/i40iw/i40iw_verbs.h

@@ -139,7 +139,7 @@ struct i40iw_qp {
 	struct i40iw_qp_host_ctx_info ctx_info;
 	struct i40iwarp_offload_info iwarp_info;
 	void *allocated_buffer;
-	atomic_t refcount;
+	refcount_t refcount;
 	struct iw_cm_id *cm_id;
 	void *cm_node;
 	struct ib_mr *lsmm_mr;
@@ -174,5 +174,6 @@ struct i40iw_qp {
 	struct i40iw_dma_mem ietf_mem;
 	struct completion sq_drained;
 	struct completion rq_drained;
+	struct completion free_qp;
 };
 #endif

drivers/infiniband/hw/mlx4/ah.c

@@ -232,8 +232,3 @@ int mlx4_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr)
 
 	return 0;
 }
-
-void mlx4_ib_destroy_ah(struct ib_ah *ah, u32 flags)
-{
-	return;
-}

drivers/infiniband/hw/mlx4/cm.c

@@ -54,11 +54,20 @@ struct id_map_entry {
 	struct delayed_work timeout;
 };
 
+struct rej_tmout_entry {
+	int slave;
+	u32 rem_pv_cm_id;
+	struct delayed_work timeout;
+	struct xarray *xa_rej_tmout;
+};
+
 struct cm_generic_msg {
 	struct ib_mad_hdr hdr;
 
 	__be32 local_comm_id;
 	__be32 remote_comm_id;
+	unsigned char unused[2];
+	__be16 rej_reason;
 };
 
 struct cm_sidr_generic_msg {
@@ -280,11 +289,15 @@ static void schedule_delayed(struct ib_device *ibdev, struct id_map_entry *id)
 	if (!sriov->is_going_down && !id->scheduled_delete) {
 		id->scheduled_delete = 1;
 		schedule_delayed_work(&id->timeout, CM_CLEANUP_CACHE_TIMEOUT);
+	} else if (id->scheduled_delete) {
+		/* Adjust timeout if already scheduled */
+		mod_delayed_work(system_wq, &id->timeout, CM_CLEANUP_CACHE_TIMEOUT);
 	}
 	spin_unlock_irqrestore(&sriov->going_down_lock, flags);
 	spin_unlock(&sriov->id_map_lock);
 }
 
+#define REJ_REASON(m) be16_to_cpu(((struct cm_generic_msg *)(m))->rej_reason)
 int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id,
 		struct ib_mad *mad)
 {
@@ -293,8 +306,10 @@ int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id
 	int pv_cm_id = -1;
 
 	if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID ||
-			mad->mad_hdr.attr_id == CM_REP_ATTR_ID ||
+	    mad->mad_hdr.attr_id == CM_REP_ATTR_ID ||
-			mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
+	    mad->mad_hdr.attr_id == CM_MRA_ATTR_ID ||
+	    mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID ||
+	    (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID && REJ_REASON(mad) == IB_CM_REJ_TIMEOUT)) {
 		sl_cm_id = get_local_comm_id(mad);
 		id = id_map_get(ibdev, &pv_cm_id, slave_id, sl_cm_id);
 		if (id)
@@ -314,8 +329,8 @@ int mlx4_ib_multiplex_cm_handler(struct ib_device *ibdev, int port, int slave_id
 	}
 
 	if (!id) {
-		pr_debug("id{slave: %d, sl_cm_id: 0x%x} is NULL!\n",
+		pr_debug("id{slave: %d, sl_cm_id: 0x%x} is NULL! attr_id: 0x%x\n",
-			 slave_id, sl_cm_id);
+			 slave_id, sl_cm_id, be16_to_cpu(mad->mad_hdr.attr_id));
 		return -EINVAL;
 	}
 
@@ -327,11 +342,94 @@ cont:
 	return 0;
 }
 
+static void rej_tmout_timeout(struct work_struct *work)
+{
+	struct delayed_work *delay = to_delayed_work(work);
+	struct rej_tmout_entry *item = container_of(delay, struct rej_tmout_entry, timeout);
+	struct rej_tmout_entry *deleted;
+
+	deleted = xa_cmpxchg(item->xa_rej_tmout, item->rem_pv_cm_id, item, NULL, 0);
+
+	if (deleted != item)
+		pr_debug("deleted(%p) != item(%p)\n", deleted, item);
+
+	kfree(item);
+}
+
+static int alloc_rej_tmout(struct mlx4_ib_sriov *sriov, u32 rem_pv_cm_id, int slave)
+{
+	struct rej_tmout_entry *item;
+	struct rej_tmout_entry *old;
+	int ret = 0;
+
+	xa_lock(&sriov->xa_rej_tmout);
+	item = xa_load(&sriov->xa_rej_tmout, (unsigned long)rem_pv_cm_id);
+
+	if (item) {
+		if (xa_err(item))
+			ret =  xa_err(item);
+		else
+			/* If a retry, adjust delayed work */
+			mod_delayed_work(system_wq, &item->timeout, CM_CLEANUP_CACHE_TIMEOUT);
+		goto err_or_exists;
+	}
+	xa_unlock(&sriov->xa_rej_tmout);
+
+	item = kmalloc(sizeof(*item), GFP_KERNEL);
+	if (!item)
+		return -ENOMEM;
+
+	INIT_DELAYED_WORK(&item->timeout, rej_tmout_timeout);
+	item->slave = slave;
+	item->rem_pv_cm_id = rem_pv_cm_id;
+	item->xa_rej_tmout = &sriov->xa_rej_tmout;
+
+	old = xa_cmpxchg(&sriov->xa_rej_tmout, (unsigned long)rem_pv_cm_id, NULL, item, GFP_KERNEL);
+	if (old) {
+		pr_debug(
+			"Non-null old entry (%p) or error (%d) when inserting\n",
+			old, xa_err(old));
+		kfree(item);
+		return xa_err(old);
+	}
+
+	schedule_delayed_work(&item->timeout, CM_CLEANUP_CACHE_TIMEOUT);
+
+	return 0;
+
+err_or_exists:
+	xa_unlock(&sriov->xa_rej_tmout);
+	return ret;
+}
+
+static int lookup_rej_tmout_slave(struct mlx4_ib_sriov *sriov, u32 rem_pv_cm_id)
+{
+	struct rej_tmout_entry *item;
+	int slave;
+
+	xa_lock(&sriov->xa_rej_tmout);
+	item = xa_load(&sriov->xa_rej_tmout, (unsigned long)rem_pv_cm_id);
+
+	if (!item || xa_err(item)) {
+		pr_debug("Could not find slave. rem_pv_cm_id 0x%x error: %d\n",
+			 rem_pv_cm_id, xa_err(item));
+		slave = !item ? -ENOENT : xa_err(item);
+	} else {
+		slave = item->slave;
+	}
+	xa_unlock(&sriov->xa_rej_tmout);
+
+	return slave;
+}
+
 int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
 			     struct ib_mad *mad)
 {
+	struct mlx4_ib_sriov *sriov = &to_mdev(ibdev)->sriov;
+	u32 rem_pv_cm_id = get_local_comm_id(mad);
 	u32 pv_cm_id;
 	struct id_map_entry *id;
+	int sts;
 
 	if (mad->mad_hdr.attr_id == CM_REQ_ATTR_ID ||
 	    mad->mad_hdr.attr_id == CM_SIDR_REQ_ATTR_ID) {
@@ -347,6 +445,13 @@ int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
 				     be64_to_cpu(gid.global.interface_id));
 			return -ENOENT;
 		}
+
+		sts = alloc_rej_tmout(sriov, rem_pv_cm_id, *slave);
+		if (sts)
+			/* Even if this fails, we pass on the REQ to the slave */
+			pr_debug("Could not allocate rej_tmout entry. rem_pv_cm_id 0x%x slave %d status %d\n",
+				 rem_pv_cm_id, *slave, sts);
+
 		return 0;
 	}
 
@@ -354,7 +459,14 @@ int mlx4_ib_demux_cm_handler(struct ib_device *ibdev, int port, int *slave,
 	id = id_map_get(ibdev, (int *)&pv_cm_id, -1, -1);
 
 	if (!id) {
-		pr_debug("Couldn't find an entry for pv_cm_id 0x%x\n", pv_cm_id);
+		if (mad->mad_hdr.attr_id == CM_REJ_ATTR_ID &&
+		    REJ_REASON(mad) == IB_CM_REJ_TIMEOUT && slave) {
+			*slave = lookup_rej_tmout_slave(sriov, rem_pv_cm_id);
+
+			return (*slave < 0) ? *slave : 0;
+		}
+		pr_debug("Couldn't find an entry for pv_cm_id 0x%x, attr_id 0x%x\n",
+			 pv_cm_id, be16_to_cpu(mad->mad_hdr.attr_id));
 		return -ENOENT;
 	}
 
@@ -375,6 +487,34 @@ void mlx4_ib_cm_paravirt_init(struct mlx4_ib_dev *dev)
 	INIT_LIST_HEAD(&dev->sriov.cm_list);
 	dev->sriov.sl_id_map = RB_ROOT;
 	xa_init_flags(&dev->sriov.pv_id_table, XA_FLAGS_ALLOC);
+	xa_init(&dev->sriov.xa_rej_tmout);
+}
+
+static void rej_tmout_xa_cleanup(struct mlx4_ib_sriov *sriov, int slave)
+{
+	struct rej_tmout_entry *item;
+	bool flush_needed = false;
+	unsigned long id;
+	int cnt = 0;
+
+	xa_lock(&sriov->xa_rej_tmout);
+	xa_for_each(&sriov->xa_rej_tmout, id, item) {
+		if (slave < 0 || slave == item->slave) {
+			mod_delayed_work(system_wq, &item->timeout, 0);
+			flush_needed = true;
+			++cnt;
+		}
+	}
+	xa_unlock(&sriov->xa_rej_tmout);
+
+	if (flush_needed) {
+		flush_scheduled_work();
+		pr_debug("Deleted %d entries in xarray for slave %d during cleanup\n",
+			 cnt, slave);
+	}
+
+	if (slave < 0)
+		WARN_ON(!xa_empty(&sriov->xa_rej_tmout));
 }
 
 /* slave = -1 ==> all slaves */
@@ -444,4 +584,6 @@ void mlx4_ib_cm_paravirt_clean(struct mlx4_ib_dev *dev, int slave)
 		list_del(&map->list);
 		kfree(map);
 	}
+
+	rej_tmout_xa_cleanup(sriov, slave);
 }

drivers/infiniband/hw/mlx4/cq.c

@@ -149,7 +149,6 @@ static int mlx4_ib_get_cq_umem(struct mlx4_ib_dev *dev, struct ib_udata *udata,
 	if (IS_ERR(*umem))
 		return PTR_ERR(*umem);
 
-	n = ib_umem_page_count(*umem);
 	shift = mlx4_ib_umem_calc_optimal_mtt_size(*umem, 0, &n);
 	err = mlx4_mtt_init(dev->dev, n, shift, &buf->mtt);
 
@@ -475,7 +474,7 @@ out:
 	return err;
 }
 
-void mlx4_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
+int mlx4_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
 {
 	struct mlx4_ib_dev *dev = to_mdev(cq->device);
 	struct mlx4_ib_cq *mcq = to_mcq(cq);
@@ -495,6 +494,7 @@ void mlx4_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
 		mlx4_db_free(dev->dev, &mcq->db);
 	}
 	ib_umem_release(mcq->umem);
+	return 0;
 }
 
 static void dump_cqe(void *cqe)

drivers/infiniband/hw/mlx4/mad.c

@@ -500,6 +500,13 @@ static int get_gids_from_l3_hdr(struct ib_grh *grh, union ib_gid *sgid,
 					 sgid, dgid);
 }
 
+static int is_proxy_qp0(struct mlx4_ib_dev *dev, int qpn, int slave)
+{
+	int proxy_start = dev->dev->phys_caps.base_proxy_sqpn + 8 * slave;
+
+	return (qpn >= proxy_start && qpn <= proxy_start + 1);
+}
+
 int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
 			  enum ib_qp_type dest_qpt, struct ib_wc *wc,
 			  struct ib_grh *grh, struct ib_mad *mad)
@@ -520,8 +527,10 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
 	u16 cached_pkey;
 	u8 is_eth = dev->dev->caps.port_type[port] == MLX4_PORT_TYPE_ETH;
 
-	if (dest_qpt > IB_QPT_GSI)
+	if (dest_qpt > IB_QPT_GSI) {
+		pr_debug("dest_qpt (%d) > IB_QPT_GSI\n", dest_qpt);
 		return -EINVAL;
+	}
 
 	tun_ctx = dev->sriov.demux[port-1].tun[slave];
 
@@ -538,12 +547,20 @@ int mlx4_ib_send_to_slave(struct mlx4_ib_dev *dev, int slave, u8 port,
 	if (dest_qpt) {
 		u16 pkey_ix;
 		ret = ib_get_cached_pkey(&dev->ib_dev, port, wc->pkey_index, &cached_pkey);
-		if (ret)
+		if (ret) {
+			pr_debug("unable to get %s cached pkey for index %d, ret %d\n",
+				 is_proxy_qp0(dev, wc->src_qp, slave) ? "SMI" : "GSI",
+				 wc->pkey_index, ret);
 			return -EINVAL;
+		}
 
 		ret = find_slave_port_pkey_ix(dev, slave, port, cached_pkey, &pkey_ix);
-		if (ret)
+		if (ret) {
+			pr_debug("unable to get %s pkey ix for pkey 0x%x, ret %d\n",
+				 is_proxy_qp0(dev, wc->src_qp, slave) ? "SMI" : "GSI",
+				 cached_pkey, ret);
 			return -EINVAL;
+		}
 		tun_pkey_ix = pkey_ix;
 	} else
 		tun_pkey_ix = dev->pkeys.virt2phys_pkey[slave][port - 1][0];
@@ -715,7 +732,8 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port,
 
 		err = mlx4_ib_send_to_slave(dev, slave, port, wc->qp->qp_type, wc, grh, mad);
 		if (err)
-			pr_debug("failed sending to slave %d via tunnel qp (%d)\n",
+			pr_debug("failed sending %s to slave %d via tunnel qp (%d)\n",
+				 is_proxy_qp0(dev, wc->src_qp, slave) ? "SMI" : "GSI",
 				 slave, err);
 		return 0;
 	}
@@ -794,7 +812,8 @@ static int mlx4_ib_demux_mad(struct ib_device *ibdev, u8 port,
 
 	err = mlx4_ib_send_to_slave(dev, slave, port, wc->qp->qp_type, wc, grh, mad);
 	if (err)
-		pr_debug("failed sending to slave %d via tunnel qp (%d)\n",
+		pr_debug("failed sending %s to slave %d via tunnel qp (%d)\n",
+			 is_proxy_qp0(dev, wc->src_qp, slave) ? "SMI" : "GSI",
 			 slave, err);
 	return 0;
 }
@@ -807,27 +826,6 @@ static int ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
 	int err;
 	struct ib_port_attr pattr;
 
-	if (in_wc && in_wc->qp) {
-		pr_debug("received MAD: port:%d slid:%d sqpn:%d "
-			 "dlid_bits:%d dqpn:%d wc_flags:0x%x tid:%016llx cls:%x mtd:%x atr:%x\n",
-			 port_num,
-			 in_wc->slid, in_wc->src_qp,
-			 in_wc->dlid_path_bits,
-			 in_wc->qp->qp_num,
-			 in_wc->wc_flags,
-			 be64_to_cpu(in_mad->mad_hdr.tid),
-			 in_mad->mad_hdr.mgmt_class, in_mad->mad_hdr.method,
-			 be16_to_cpu(in_mad->mad_hdr.attr_id));
-		if (in_wc->wc_flags & IB_WC_GRH) {
-			pr_debug("sgid_hi:0x%016llx sgid_lo:0x%016llx\n",
-				 be64_to_cpu(in_grh->sgid.global.subnet_prefix),
-				 be64_to_cpu(in_grh->sgid.global.interface_id));
-			pr_debug("dgid_hi:0x%016llx dgid_lo:0x%016llx\n",
-				 be64_to_cpu(in_grh->dgid.global.subnet_prefix),
-				 be64_to_cpu(in_grh->dgid.global.interface_id));
-		}
-	}
-
 	slid = in_wc ? ib_lid_cpu16(in_wc->slid) : be16_to_cpu(IB_LID_PERMISSIVE);
 
 	if (in_mad->mad_hdr.method == IB_MGMT_METHOD_TRAP && slid == 0) {
@@ -1299,6 +1297,18 @@ static void mlx4_ib_tunnel_comp_handler(struct ib_cq *cq, void *arg)
 	spin_unlock_irqrestore(&dev->sriov.going_down_lock, flags);
 }
 
+static void mlx4_ib_wire_comp_handler(struct ib_cq *cq, void *arg)
+{
+	unsigned long flags;
+	struct mlx4_ib_demux_pv_ctx *ctx = cq->cq_context;
+	struct mlx4_ib_dev *dev = to_mdev(ctx->ib_dev);
+
+	spin_lock_irqsave(&dev->sriov.going_down_lock, flags);
+	if (!dev->sriov.is_going_down && ctx->state == DEMUX_PV_STATE_ACTIVE)
+		queue_work(ctx->wi_wq, &ctx->work);
+	spin_unlock_irqrestore(&dev->sriov.going_down_lock, flags);
+}
+
 static int mlx4_ib_post_pv_qp_buf(struct mlx4_ib_demux_pv_ctx *ctx,
 				  struct mlx4_ib_demux_pv_qp *tun_qp,
 				  int index)
@@ -1341,14 +1351,6 @@ static int mlx4_ib_multiplex_sa_handler(struct ib_device *ibdev, int port,
 	return ret;
 }
 
-static int is_proxy_qp0(struct mlx4_ib_dev *dev, int qpn, int slave)
-{
-	int proxy_start = dev->dev->phys_caps.base_proxy_sqpn + 8 * slave;
-
-	return (qpn >= proxy_start && qpn <= proxy_start + 1);
-}
-
-
 int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
 			 enum ib_qp_type dest_qpt, u16 pkey_index,
 			 u32 remote_qpn, u32 qkey, struct rdma_ah_attr *attr,
@@ -1401,10 +1403,10 @@ int mlx4_ib_send_to_wire(struct mlx4_ib_dev *dev, int slave, u8 port,
 
 	spin_lock(&sqp->tx_lock);
 	if (sqp->tx_ix_head - sqp->tx_ix_tail >=
-	    (MLX4_NUM_TUNNEL_BUFS - 1))
+	    (MLX4_NUM_WIRE_BUFS - 1))
 		ret = -EAGAIN;
 	else
-		wire_tx_ix = (++sqp->tx_ix_head) & (MLX4_NUM_TUNNEL_BUFS - 1);
+		wire_tx_ix = (++sqp->tx_ix_head) & (MLX4_NUM_WIRE_BUFS - 1);
 	spin_unlock(&sqp->tx_lock);
 	if (ret)
 		goto out;
@@ -1484,6 +1486,7 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
 	u16 vlan_id;
 	u8 qos;
 	u8 *dmac;
+	int sts;
 
 	/* Get slave that sent this packet */
 	if (wc->src_qp < dev->dev->phys_caps.base_proxy_sqpn ||
@@ -1580,13 +1583,17 @@ static void mlx4_ib_multiplex_mad(struct mlx4_ib_demux_pv_ctx *ctx, struct ib_wc
 					&vlan_id, &qos))
 		rdma_ah_set_sl(&ah_attr, qos);
 
-	mlx4_ib_send_to_wire(dev, slave, ctx->port,
+	sts = mlx4_ib_send_to_wire(dev, slave, ctx->port,
-			     is_proxy_qp0(dev, wc->src_qp, slave) ?
+				   is_proxy_qp0(dev, wc->src_qp, slave) ?
-			     IB_QPT_SMI : IB_QPT_GSI,
+				   IB_QPT_SMI : IB_QPT_GSI,
-			     be16_to_cpu(tunnel->hdr.pkey_index),
+				   be16_to_cpu(tunnel->hdr.pkey_index),
-			     be32_to_cpu(tunnel->hdr.remote_qpn),
+				   be32_to_cpu(tunnel->hdr.remote_qpn),
-			     be32_to_cpu(tunnel->hdr.qkey),
+				   be32_to_cpu(tunnel->hdr.qkey),
-			     &ah_attr, wc->smac, vlan_id, &tunnel->mad);
+				   &ah_attr, wc->smac, vlan_id, &tunnel->mad);
+	if (sts)
+		pr_debug("failed sending %s to wire on behalf of slave %d (%d)\n",
+			 is_proxy_qp0(dev, wc->src_qp, slave) ? "SMI" : "GSI",
+			 slave, sts);
 }
 
 static int mlx4_ib_alloc_pv_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
@@ -1595,19 +1602,20 @@ static int mlx4_ib_alloc_pv_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
 	int i;
 	struct mlx4_ib_demux_pv_qp *tun_qp;
 	int rx_buf_size, tx_buf_size;
+	const int nmbr_bufs = is_tun ? MLX4_NUM_TUNNEL_BUFS : MLX4_NUM_WIRE_BUFS;
 
 	if (qp_type > IB_QPT_GSI)
 		return -EINVAL;
 
 	tun_qp = &ctx->qp[qp_type];
 
-	tun_qp->ring = kcalloc(MLX4_NUM_TUNNEL_BUFS,
+	tun_qp->ring = kcalloc(nmbr_bufs,
 			       sizeof(struct mlx4_ib_buf),
 			       GFP_KERNEL);
 	if (!tun_qp->ring)
 		return -ENOMEM;
 
-	tun_qp->tx_ring = kcalloc(MLX4_NUM_TUNNEL_BUFS,
+	tun_qp->tx_ring = kcalloc(nmbr_bufs,
 				  sizeof (struct mlx4_ib_tun_tx_buf),
 				  GFP_KERNEL);
 	if (!tun_qp->tx_ring) {
@@ -1624,7 +1632,7 @@ static int mlx4_ib_alloc_pv_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
 		tx_buf_size = sizeof (struct mlx4_mad_snd_buf);
 	}
 
-	for (i = 0; i < MLX4_NUM_TUNNEL_BUFS; i++) {
+	for (i = 0; i < nmbr_bufs; i++) {
 		tun_qp->ring[i].addr = kmalloc(rx_buf_size, GFP_KERNEL);
 		if (!tun_qp->ring[i].addr)
 			goto err;
@@ -1638,7 +1646,7 @@ static int mlx4_ib_alloc_pv_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
 		}
 	}
 
-	for (i = 0; i < MLX4_NUM_TUNNEL_BUFS; i++) {
+	for (i = 0; i < nmbr_bufs; i++) {
 		tun_qp->tx_ring[i].buf.addr =
 			kmalloc(tx_buf_size, GFP_KERNEL);
 		if (!tun_qp->tx_ring[i].buf.addr)
@@ -1669,7 +1677,7 @@ tx_err:
 				    tx_buf_size, DMA_TO_DEVICE);
 		kfree(tun_qp->tx_ring[i].buf.addr);
 	}
-	i = MLX4_NUM_TUNNEL_BUFS;
+	i = nmbr_bufs;
 err:
 	while (i > 0) {
 		--i;
@@ -1690,6 +1698,7 @@ static void mlx4_ib_free_pv_qp_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
 	int i;
 	struct mlx4_ib_demux_pv_qp *tun_qp;
 	int rx_buf_size, tx_buf_size;
+	const int nmbr_bufs = is_tun ? MLX4_NUM_TUNNEL_BUFS : MLX4_NUM_WIRE_BUFS;
 
 	if (qp_type > IB_QPT_GSI)
 		return;
@@ -1704,13 +1713,13 @@ static void mlx4_ib_free_pv_qp_bufs(struct mlx4_ib_demux_pv_ctx *ctx,
 	}
 
 
-	for (i = 0; i < MLX4_NUM_TUNNEL_BUFS; i++) {
+	for (i = 0; i < nmbr_bufs; i++) {
 		ib_dma_unmap_single(ctx->ib_dev, tun_qp->ring[i].map,
 				    rx_buf_size, DMA_FROM_DEVICE);
 		kfree(tun_qp->ring[i].addr);
 	}
 
-	for (i = 0; i < MLX4_NUM_TUNNEL_BUFS; i++) {
+	for (i = 0; i < nmbr_bufs; i++) {
 		ib_dma_unmap_single(ctx->ib_dev, tun_qp->tx_ring[i].buf.map,
 				    tx_buf_size, DMA_TO_DEVICE);
 		kfree(tun_qp->tx_ring[i].buf.addr);
@@ -1744,9 +1753,6 @@ static void mlx4_ib_tunnel_comp_worker(struct work_struct *work)
 					       "buf:%lld\n", wc.wr_id);
 				break;
 			case IB_WC_SEND:
-				pr_debug("received tunnel send completion:"
-					 "wrid=0x%llx, status=0x%x\n",
-					 wc.wr_id, wc.status);
 				rdma_destroy_ah(tun_qp->tx_ring[wc.wr_id &
 					      (MLX4_NUM_TUNNEL_BUFS - 1)].ah, 0);
 				tun_qp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah
@@ -1793,6 +1799,7 @@ static int create_pv_sqp(struct mlx4_ib_demux_pv_ctx *ctx,
 	struct mlx4_ib_qp_tunnel_init_attr qp_init_attr;
 	struct ib_qp_attr attr;
 	int qp_attr_mask_INIT;
+	const int nmbr_bufs = create_tun ? MLX4_NUM_TUNNEL_BUFS : MLX4_NUM_WIRE_BUFS;
 
 	if (qp_type > IB_QPT_GSI)
 		return -EINVAL;
@@ -1803,8 +1810,8 @@ static int create_pv_sqp(struct mlx4_ib_demux_pv_ctx *ctx,
 	qp_init_attr.init_attr.send_cq = ctx->cq;
 	qp_init_attr.init_attr.recv_cq = ctx->cq;
 	qp_init_attr.init_attr.sq_sig_type = IB_SIGNAL_ALL_WR;
-	qp_init_attr.init_attr.cap.max_send_wr = MLX4_NUM_TUNNEL_BUFS;
+	qp_init_attr.init_attr.cap.max_send_wr = nmbr_bufs;
-	qp_init_attr.init_attr.cap.max_recv_wr = MLX4_NUM_TUNNEL_BUFS;
+	qp_init_attr.init_attr.cap.max_recv_wr = nmbr_bufs;
 	qp_init_attr.init_attr.cap.max_send_sge = 1;
 	qp_init_attr.init_attr.cap.max_recv_sge = 1;
 	if (create_tun) {
@@ -1866,7 +1873,7 @@ static int create_pv_sqp(struct mlx4_ib_demux_pv_ctx *ctx,
 		goto err_qp;
 	}
 
-	for (i = 0; i < MLX4_NUM_TUNNEL_BUFS; i++) {
+	for (i = 0; i < nmbr_bufs; i++) {
 		ret = mlx4_ib_post_pv_qp_buf(ctx, tun_qp, i);
 		if (ret) {
 			pr_err(" mlx4_ib_post_pv_buf error"
@@ -1902,8 +1909,8 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work)
 			switch (wc.opcode) {
 			case IB_WC_SEND:
 				kfree(sqp->tx_ring[wc.wr_id &
-				      (MLX4_NUM_TUNNEL_BUFS - 1)].ah);
+				      (MLX4_NUM_WIRE_BUFS - 1)].ah);
-				sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah
+				sqp->tx_ring[wc.wr_id & (MLX4_NUM_WIRE_BUFS - 1)].ah
 					= NULL;
 				spin_lock(&sqp->tx_lock);
 				sqp->tx_ix_tail++;
@@ -1912,13 +1919,13 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work)
 			case IB_WC_RECV:
 				mad = (struct ib_mad *) &(((struct mlx4_mad_rcv_buf *)
 						(sqp->ring[wc.wr_id &
-						(MLX4_NUM_TUNNEL_BUFS - 1)].addr))->payload);
+						(MLX4_NUM_WIRE_BUFS - 1)].addr))->payload);
 				grh = &(((struct mlx4_mad_rcv_buf *)
 						(sqp->ring[wc.wr_id &
-						(MLX4_NUM_TUNNEL_BUFS - 1)].addr))->grh);
+						(MLX4_NUM_WIRE_BUFS - 1)].addr))->grh);
 				mlx4_ib_demux_mad(ctx->ib_dev, ctx->port, &wc, grh, mad);
 				if (mlx4_ib_post_pv_qp_buf(ctx, sqp, wc.wr_id &
-							   (MLX4_NUM_TUNNEL_BUFS - 1)))
+							   (MLX4_NUM_WIRE_BUFS - 1)))
 					pr_err("Failed reposting SQP "
 					       "buf:%lld\n", wc.wr_id);
 				break;
@@ -1931,8 +1938,8 @@ static void mlx4_ib_sqp_comp_worker(struct work_struct *work)
 				 ctx->slave, wc.status, wc.wr_id);
 			if (!MLX4_TUN_IS_RECV(wc.wr_id)) {
 				kfree(sqp->tx_ring[wc.wr_id &
-				      (MLX4_NUM_TUNNEL_BUFS - 1)].ah);
+				      (MLX4_NUM_WIRE_BUFS - 1)].ah);
-				sqp->tx_ring[wc.wr_id & (MLX4_NUM_TUNNEL_BUFS - 1)].ah
+				sqp->tx_ring[wc.wr_id & (MLX4_NUM_WIRE_BUFS - 1)].ah
 					= NULL;
 				spin_lock(&sqp->tx_lock);
 				sqp->tx_ix_tail++;
@@ -1972,6 +1979,7 @@ static int create_pv_resources(struct ib_device *ibdev, int slave, int port,
 {
 	int ret, cq_size;
 	struct ib_cq_init_attr cq_attr = {};
+	const int nmbr_bufs = create_tun ? MLX4_NUM_TUNNEL_BUFS : MLX4_NUM_WIRE_BUFS;
 
 	if (ctx->state != DEMUX_PV_STATE_DOWN)
 		return -EEXIST;
@@ -1996,12 +2004,13 @@ static int create_pv_resources(struct ib_device *ibdev, int slave, int port,
 		goto err_out_qp0;
 	}
 
-	cq_size = 2 * MLX4_NUM_TUNNEL_BUFS;
+	cq_size = 2 * nmbr_bufs;
 	if (ctx->has_smi)
 		cq_size *= 2;
 
 	cq_attr.cqe = cq_size;
-	ctx->cq = ib_create_cq(ctx->ib_dev, mlx4_ib_tunnel_comp_handler,
+	ctx->cq = ib_create_cq(ctx->ib_dev,
+			       create_tun ? mlx4_ib_tunnel_comp_handler : mlx4_ib_wire_comp_handler,
 			       NULL, ctx, &cq_attr);
 	if (IS_ERR(ctx->cq)) {
 		ret = PTR_ERR(ctx->cq);
@@ -2038,6 +2047,7 @@ static int create_pv_resources(struct ib_device *ibdev, int slave, int port,
 		INIT_WORK(&ctx->work, mlx4_ib_sqp_comp_worker);
 
 	ctx->wq = to_mdev(ibdev)->sriov.demux[port - 1].wq;
+	ctx->wi_wq = to_mdev(ibdev)->sriov.demux[port - 1].wi_wq;
 
 	ret = ib_req_notify_cq(ctx->cq, IB_CQ_NEXT_COMP);
 	if (ret) {
@@ -2181,7 +2191,7 @@ static int mlx4_ib_alloc_demux_ctx(struct mlx4_ib_dev *dev,
 		goto err_mcg;
 	}
 
-	snprintf(name, sizeof name, "mlx4_ibt%d", port);
+	snprintf(name, sizeof(name), "mlx4_ibt%d", port);
 	ctx->wq = alloc_ordered_workqueue(name, WQ_MEM_RECLAIM);
 	if (!ctx->wq) {
 		pr_err("Failed to create tunnelling WQ for port %d\n", port);
@@ -2189,7 +2199,15 @@ static int mlx4_ib_alloc_demux_ctx(struct mlx4_ib_dev *dev,
 		goto err_wq;
 	}
 
-	snprintf(name, sizeof name, "mlx4_ibud%d", port);
+	snprintf(name, sizeof(name), "mlx4_ibwi%d", port);
+	ctx->wi_wq = alloc_ordered_workqueue(name, WQ_MEM_RECLAIM);
+	if (!ctx->wi_wq) {
+		pr_err("Failed to create wire WQ for port %d\n", port);
+		ret = -ENOMEM;
+		goto err_wiwq;
+	}
+
+	snprintf(name, sizeof(name), "mlx4_ibud%d", port);
 	ctx->ud_wq = alloc_ordered_workqueue(name, WQ_MEM_RECLAIM);
 	if (!ctx->ud_wq) {
 		pr_err("Failed to create up/down WQ for port %d\n", port);
@@ -2200,6 +2218,10 @@ static int mlx4_ib_alloc_demux_ctx(struct mlx4_ib_dev *dev,
 	return 0;
 
 err_udwq:
+	destroy_workqueue(ctx->wi_wq);
+	ctx->wi_wq = NULL;
+
+err_wiwq:
 	destroy_workqueue(ctx->wq);
 	ctx->wq = NULL;
 
@@ -2247,12 +2269,14 @@ static void mlx4_ib_free_demux_ctx(struct mlx4_ib_demux_ctx *ctx)
 				ctx->tun[i]->state = DEMUX_PV_STATE_DOWNING;
 		}
 		flush_workqueue(ctx->wq);
+		flush_workqueue(ctx->wi_wq);
 		for (i = 0; i < dev->dev->caps.sqp_demux; i++) {
 			destroy_pv_resources(dev, i, ctx->port, ctx->tun[i], 0);
 			free_pv_object(dev, i, ctx->port);
 		}
 		kfree(ctx->tun);
 		destroy_workqueue(ctx->ud_wq);
+		destroy_workqueue(ctx->wi_wq);
 		destroy_workqueue(ctx->wq);
 	}
 }

drivers/infiniband/hw/mlx4/main.c

@@ -1215,9 +1215,10 @@ static int mlx4_ib_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 	return 0;
 }
 
-static void mlx4_ib_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
+static int mlx4_ib_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
 {
 	mlx4_pd_free(to_mdev(pd->device)->dev, to_mpd(pd)->pdn);
+	return 0;
 }
 
 static int mlx4_ib_alloc_xrcd(struct ib_xrcd *ibxrcd, struct ib_udata *udata)
@@ -1256,11 +1257,12 @@ err2:
 	return err;
 }
 
-static void mlx4_ib_dealloc_xrcd(struct ib_xrcd *xrcd, struct ib_udata *udata)
+static int mlx4_ib_dealloc_xrcd(struct ib_xrcd *xrcd, struct ib_udata *udata)
 {
 	ib_destroy_cq(to_mxrcd(xrcd)->cq);
 	ib_dealloc_pd(to_mxrcd(xrcd)->pd);
 	mlx4_xrcd_free(to_mdev(xrcd->device)->dev, to_mxrcd(xrcd)->xrcdn);
+	return 0;
 }
 
 static int add_gid_entry(struct ib_qp *ibqp, union ib_gid *gid)
@@ -1533,23 +1535,11 @@ static int __mlx4_ib_create_flow(struct ib_qp *qp, struct ib_flow_attr *flow_att
 	struct mlx4_net_trans_rule_hw_ctrl *ctrl;
 	int default_flow;
 
-	static const u16 __mlx4_domain[] = {
-		[IB_FLOW_DOMAIN_USER] = MLX4_DOMAIN_UVERBS,
-		[IB_FLOW_DOMAIN_ETHTOOL] = MLX4_DOMAIN_ETHTOOL,
-		[IB_FLOW_DOMAIN_RFS] = MLX4_DOMAIN_RFS,
-		[IB_FLOW_DOMAIN_NIC] = MLX4_DOMAIN_NIC,
-	};
-
 	if (flow_attr->priority > MLX4_IB_FLOW_MAX_PRIO) {
 		pr_err("Invalid priority value %d\n", flow_attr->priority);
 		return -EINVAL;
 	}
 
-	if (domain >= IB_FLOW_DOMAIN_NUM) {
-		pr_err("Invalid domain value %d\n", domain);
-		return -EINVAL;
-	}
-
 	if (mlx4_map_sw_to_hw_steering_mode(mdev->dev, flow_type) < 0)
 		return -EINVAL;
 
@@ -1558,8 +1548,7 @@ static int __mlx4_ib_create_flow(struct ib_qp *qp, struct ib_flow_attr *flow_att
 		return PTR_ERR(mailbox);
 	ctrl = mailbox->buf;
 
-	ctrl->prio = cpu_to_be16(__mlx4_domain[domain] |
+	ctrl->prio = cpu_to_be16(domain | flow_attr->priority);
-				 flow_attr->priority);
 	ctrl->type = mlx4_map_sw_to_hw_steering_mode(mdev->dev, flow_type);
 	ctrl->port = flow_attr->port;
 	ctrl->qpn = cpu_to_be32(qp->qp_num);
@@ -1701,8 +1690,8 @@ static int mlx4_ib_add_dont_trap_rule(struct mlx4_dev *dev,
 }
 
 static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp,
-				    struct ib_flow_attr *flow_attr,
+					   struct ib_flow_attr *flow_attr,
-				    int domain, struct ib_udata *udata)
+					   struct ib_udata *udata)
 {
 	int err = 0, i = 0, j = 0;
 	struct mlx4_ib_flow *mflow;
@@ -1768,8 +1757,8 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp,
 	}
 
 	while (i < ARRAY_SIZE(type) && type[i]) {
-		err = __mlx4_ib_create_flow(qp, flow_attr, domain, type[i],
+		err = __mlx4_ib_create_flow(qp, flow_attr, MLX4_DOMAIN_UVERBS,
-					    &mflow->reg_id[i].id);
+					    type[i], &mflow->reg_id[i].id);
 		if (err)
 			goto err_create_flow;
 		if (is_bonded) {
@@ -1778,7 +1767,7 @@ static struct ib_flow *mlx4_ib_create_flow(struct ib_qp *qp,
 			 */
 			flow_attr->port = 2;
 			err = __mlx4_ib_create_flow(qp, flow_attr,
-						    domain, type[j],
+						    MLX4_DOMAIN_UVERBS, type[j],
 						    &mflow->reg_id[j].mirror);
 			flow_attr->port = 1;
 			if (err)
@@ -2589,11 +2578,16 @@ static const struct ib_device_ops mlx4_ib_dev_wq_ops = {
 	.destroy_rwq_ind_table = mlx4_ib_destroy_rwq_ind_table,
 	.destroy_wq = mlx4_ib_destroy_wq,
 	.modify_wq = mlx4_ib_modify_wq,
+
+	INIT_RDMA_OBJ_SIZE(ib_rwq_ind_table, mlx4_ib_rwq_ind_table,
+			   ib_rwq_ind_tbl),
 };
 
 static const struct ib_device_ops mlx4_ib_dev_mw_ops = {
 	.alloc_mw = mlx4_ib_alloc_mw,
 	.dealloc_mw = mlx4_ib_dealloc_mw,
+
+	INIT_RDMA_OBJ_SIZE(ib_mw, mlx4_ib_mw, ibmw),
 };
 
 static const struct ib_device_ops mlx4_ib_dev_xrc_ops = {
@@ -2847,7 +2841,8 @@ static void *mlx4_ib_add(struct mlx4_dev *dev)
 		goto err_steer_free_bitmap;
 
 	rdma_set_device_sysfs_group(&ibdev->ib_dev, &mlx4_attr_group);
-	if (ib_register_device(&ibdev->ib_dev, "mlx4_%d"))
+	if (ib_register_device(&ibdev->ib_dev, "mlx4_%d",
+			       &dev->persist->pdev->dev))
 		goto err_diag_counters;
 
 	if (mlx4_ib_mad_init(ibdev))
@@ -2989,10 +2984,8 @@ int mlx4_ib_steer_qp_reg(struct mlx4_ib_dev *mdev, struct mlx4_ib_qp *mqp,
 		/* Add an empty rule for IB L2 */
 		memset(&ib_spec->mask, 0, sizeof(ib_spec->mask));
 
-		err = __mlx4_ib_create_flow(&mqp->ibqp, flow,
+		err = __mlx4_ib_create_flow(&mqp->ibqp, flow, MLX4_DOMAIN_NIC,
-					    IB_FLOW_DOMAIN_NIC,
+					    MLX4_FS_REGULAR, &mqp->reg_id);
-					    MLX4_FS_REGULAR,
-					    &mqp->reg_id);
 	} else {
 		err = __mlx4_ib_destroy_flow(mdev->dev, mqp->reg_id);
 	}

drivers/infiniband/hw/mlx4/mlx4_ib.h

@@ -233,7 +233,8 @@ enum mlx4_ib_mad_ifc_flags {
 };
 
 enum {
-	MLX4_NUM_TUNNEL_BUFS		= 256,
+	MLX4_NUM_TUNNEL_BUFS		= 512,
+	MLX4_NUM_WIRE_BUFS		= 2048,
 };
 
 struct mlx4_ib_tunnel_header {
@@ -298,6 +299,26 @@ struct mlx4_ib_rss {
 	u8			rss_key[MLX4_EN_RSS_KEY_SIZE];
 };
 
+enum {
+	/*
+	 * Largest possible UD header: send with GRH and immediate
+	 * data plus 18 bytes for an Ethernet header with VLAN/802.1Q
+	 * tag.  (LRH would only use 8 bytes, so Ethernet is the
+	 * biggest case)
+	 */
+	MLX4_IB_UD_HEADER_SIZE		= 82,
+	MLX4_IB_LSO_HEADER_SPARE	= 128,
+};
+
+struct mlx4_ib_sqp {
+	int pkey_index;
+	u32 qkey;
+	u32 send_psn;
+	struct ib_ud_header ud_header;
+	u8 header_buf[MLX4_IB_UD_HEADER_SIZE];
+	struct ib_qp *roce_v2_gsi;
+};
+
 struct mlx4_ib_qp {
 	union {
 		struct ib_qp	ibqp;
@@ -343,7 +364,10 @@ struct mlx4_ib_qp {
 	struct mlx4_wqn_range	*wqn_range;
 	/* Number of RSS QP parents that uses this WQ */
 	u32			rss_usecnt;
-	struct mlx4_ib_rss	*rss_ctx;
+	union {
+		struct mlx4_ib_rss *rss_ctx;
+		struct mlx4_ib_sqp *sqp;
+	};
 };
 
 struct mlx4_ib_srq {
@@ -366,6 +390,10 @@ struct mlx4_ib_ah {
 	union mlx4_ext_av       av;
 };
 
+struct mlx4_ib_rwq_ind_table {
+	struct ib_rwq_ind_table ib_rwq_ind_tbl;
+};
+
 /****************************************/
 /* alias guid support */
 /****************************************/
@@ -454,6 +482,7 @@ struct mlx4_ib_demux_pv_ctx {
 	struct ib_pd *pd;
 	struct work_struct work;
 	struct workqueue_struct *wq;
+	struct workqueue_struct *wi_wq;
 	struct mlx4_ib_demux_pv_qp qp[2];
 };
 
@@ -461,6 +490,7 @@ struct mlx4_ib_demux_ctx {
 	struct ib_device *ib_dev;
 	int port;
 	struct workqueue_struct *wq;
+	struct workqueue_struct *wi_wq;
 	struct workqueue_struct *ud_wq;
 	spinlock_t ud_lock;
 	atomic64_t subnet_prefix;
@@ -492,6 +522,7 @@ struct mlx4_ib_sriov {
 	spinlock_t id_map_lock;
 	struct rb_root sl_id_map;
 	struct list_head cm_list;
+	struct xarray xa_rej_tmout;
 };
 
 struct gid_cache_context {
@@ -725,8 +756,7 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 				  u64 virt_addr, int access_flags,
 				  struct ib_udata *udata);
 int mlx4_ib_dereg_mr(struct ib_mr *mr, struct ib_udata *udata);
-struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int mlx4_ib_alloc_mw(struct ib_mw *mw, struct ib_udata *udata);
-			       struct ib_udata *udata);
 int mlx4_ib_dealloc_mw(struct ib_mw *mw);
 struct ib_mr *mlx4_ib_alloc_mr(struct ib_pd *pd, enum ib_mr_type mr_type,
 			       u32 max_num_sg);
@@ -736,7 +766,7 @@ int mlx4_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
 int mlx4_ib_resize_cq(struct ib_cq *ibcq, int entries, struct ib_udata *udata);
 int mlx4_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		      struct ib_udata *udata);
-void mlx4_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
+int mlx4_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
 int mlx4_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
 int mlx4_ib_arm_cq(struct ib_cq *cq, enum ib_cq_notify_flags flags);
 void __mlx4_ib_cq_clean(struct mlx4_ib_cq *cq, u32 qpn, struct mlx4_ib_srq *srq);
@@ -747,14 +777,17 @@ int mlx4_ib_create_ah(struct ib_ah *ah, struct rdma_ah_init_attr *init_attr,
 int mlx4_ib_create_ah_slave(struct ib_ah *ah, struct rdma_ah_attr *ah_attr,
 			    int slave_sgid_index, u8 *s_mac, u16 vlan_tag);
 int mlx4_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr);
-void mlx4_ib_destroy_ah(struct ib_ah *ah, u32 flags);
+static inline int mlx4_ib_destroy_ah(struct ib_ah *ah, u32 flags)
+{
+	return 0;
+}
 
 int mlx4_ib_create_srq(struct ib_srq *srq, struct ib_srq_init_attr *init_attr,
 		       struct ib_udata *udata);
 int mlx4_ib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		       enum ib_srq_attr_mask attr_mask, struct ib_udata *udata);
 int mlx4_ib_query_srq(struct ib_srq *srq, struct ib_srq_attr *srq_attr);
-void mlx4_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
+int mlx4_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
 void mlx4_ib_free_srq_wqe(struct mlx4_ib_srq *srq, int wqe_index);
 int mlx4_ib_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr,
 			  const struct ib_recv_wr **bad_wr);
@@ -890,15 +923,18 @@ void mlx4_ib_sl2vl_update(struct mlx4_ib_dev *mdev, int port);
 struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd,
 				struct ib_wq_init_attr *init_attr,
 				struct ib_udata *udata);
-void mlx4_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
+int mlx4_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
 int mlx4_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
 		      u32 wq_attr_mask, struct ib_udata *udata);
 
-struct ib_rwq_ind_table
+int mlx4_ib_create_rwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_tbl,
-*mlx4_ib_create_rwq_ind_table(struct ib_device *device,
+				 struct ib_rwq_ind_table_init_attr *init_attr,
-			      struct ib_rwq_ind_table_init_attr *init_attr,
+				 struct ib_udata *udata);
-			      struct ib_udata *udata);
+static inline int
-int mlx4_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
+mlx4_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table)
+{
+	return 0;
+}
 int mlx4_ib_umem_calc_optimal_mtt_size(struct ib_umem *umem, u64 start_va,
 				       int *num_of_mtts);
 

drivers/infiniband/hw/mlx4/mr.c

@@ -271,6 +271,8 @@ int mlx4_ib_umem_calc_optimal_mtt_size(struct ib_umem *umem, u64 start_va,
 	u64 total_len = 0;
 	int i;
 
+	*num_of_mtts = ib_umem_num_dma_blocks(umem, PAGE_SIZE);
+
 	for_each_sg(umem->sg_head.sgl, sg, umem->nmap, i) {
 		/*
 		 * Initialization - save the first chunk start as the
@@ -421,7 +423,6 @@ struct ib_mr *mlx4_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 		goto err_free;
 	}
 
-	n = ib_umem_page_count(mr->umem);
 	shift = mlx4_ib_umem_calc_optimal_mtt_size(mr->umem, start, &n);
 
 	err = mlx4_mr_alloc(dev->dev, to_mpd(pd)->pdn, virt_addr, length,
@@ -511,7 +512,7 @@ int mlx4_ib_rereg_user_mr(struct ib_mr *mr, int flags,
 			mmr->umem = NULL;
 			goto release_mpt_entry;
 		}
-		n = ib_umem_page_count(mmr->umem);
+		n = ib_umem_num_dma_blocks(mmr->umem, PAGE_SIZE);
 		shift = PAGE_SHIFT;
 
 		err = mlx4_mr_rereg_mem_write(dev->dev, &mmr->mmr,
@@ -610,37 +611,27 @@ int mlx4_ib_dereg_mr(struct ib_mr *ibmr, struct ib_udata *udata)
 	return 0;
 }
 
-struct ib_mw *mlx4_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int mlx4_ib_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
-			       struct ib_udata *udata)
 {
-	struct mlx4_ib_dev *dev = to_mdev(pd->device);
+	struct mlx4_ib_dev *dev = to_mdev(ibmw->device);
-	struct mlx4_ib_mw *mw;
+	struct mlx4_ib_mw *mw = to_mmw(ibmw);
 	int err;
 
-	mw = kmalloc(sizeof(*mw), GFP_KERNEL);
+	err = mlx4_mw_alloc(dev->dev, to_mpd(ibmw->pd)->pdn,
-	if (!mw)
+			    to_mlx4_type(ibmw->type), &mw->mmw);
-		return ERR_PTR(-ENOMEM);
-
-	err = mlx4_mw_alloc(dev->dev, to_mpd(pd)->pdn,
-			    to_mlx4_type(type), &mw->mmw);
 	if (err)
-		goto err_free;
+		return err;
 
 	err = mlx4_mw_enable(dev->dev, &mw->mmw);
 	if (err)
 		goto err_mw;
 
-	mw->ibmw.rkey = mw->mmw.key;
+	ibmw->rkey = mw->mmw.key;
-
+	return 0;
-	return &mw->ibmw;
 
 err_mw:
 	mlx4_mw_free(dev->dev, &mw->mmw);
-
+	return err;
-err_free:
-	kfree(mw);
-
-	return ERR_PTR(err);
 }
 
 int mlx4_ib_dealloc_mw(struct ib_mw *ibmw)
@@ -648,8 +639,6 @@ int mlx4_ib_dealloc_mw(struct ib_mw *ibmw)
 	struct mlx4_ib_mw *mw = to_mmw(ibmw);
 
 	mlx4_mw_free(to_mdev(ibmw->device)->dev, &mw->mmw);
-	kfree(mw);
-
 	return 0;
 }
 

drivers/infiniband/hw/mlx4/qp.c

@@ -67,27 +67,6 @@ enum {
 	MLX4_IB_LINK_TYPE_ETH		= 1
 };
 
-enum {
-	/*
-	 * Largest possible UD header: send with GRH and immediate
-	 * data plus 18 bytes for an Ethernet header with VLAN/802.1Q
-	 * tag.  (LRH would only use 8 bytes, so Ethernet is the
-	 * biggest case)
-	 */
-	MLX4_IB_UD_HEADER_SIZE		= 82,
-	MLX4_IB_LSO_HEADER_SPARE	= 128,
-};
-
-struct mlx4_ib_sqp {
-	struct mlx4_ib_qp	qp;
-	int			pkey_index;
-	u32			qkey;
-	u32			send_psn;
-	struct ib_ud_header	ud_header;
-	u8			header_buf[MLX4_IB_UD_HEADER_SIZE];
-	struct ib_qp		*roce_v2_gsi;
-};
-
 enum {
 	MLX4_IB_MIN_SQ_STRIDE	= 6,
 	MLX4_IB_CACHE_LINE_SIZE	= 64,
@@ -123,11 +102,6 @@ enum mlx4_ib_source_type {
 	MLX4_IB_RWQ_SRC	= 1,
 };
 
-static struct mlx4_ib_sqp *to_msqp(struct mlx4_ib_qp *mqp)
-{
-	return container_of(mqp, struct mlx4_ib_sqp, qp);
-}
-
 static int is_tunnel_qp(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
 {
 	if (!mlx4_is_master(dev->dev))
@@ -656,8 +630,6 @@ static int create_qp_rss(struct mlx4_ib_dev *dev,
 	if (err)
 		goto err_qpn;
 
-	mutex_init(&qp->mutex);
-
 	INIT_LIST_HEAD(&qp->gid_list);
 	INIT_LIST_HEAD(&qp->steering_rules);
 
@@ -696,80 +668,72 @@ err_qpn:
 	return err;
 }
 
-static struct ib_qp *_mlx4_ib_create_qp_rss(struct ib_pd *pd,
+static int _mlx4_ib_create_qp_rss(struct ib_pd *pd, struct mlx4_ib_qp *qp,
-					    struct ib_qp_init_attr *init_attr,
+				  struct ib_qp_init_attr *init_attr,
-					    struct ib_udata *udata)
+				  struct ib_udata *udata)
 {
-	struct mlx4_ib_qp *qp;
 	struct mlx4_ib_create_qp_rss ucmd = {};
 	size_t required_cmd_sz;
 	int err;
 
 	if (!udata) {
 		pr_debug("RSS QP with NULL udata\n");
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 	}
 
 	if (udata->outlen)
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	required_cmd_sz = offsetof(typeof(ucmd), reserved1) +
 					sizeof(ucmd.reserved1);
 	if (udata->inlen < required_cmd_sz) {
 		pr_debug("invalid inlen\n");
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 	}
 
 	if (ib_copy_from_udata(&ucmd, udata, min(sizeof(ucmd), udata->inlen))) {
 		pr_debug("copy failed\n");
-		return ERR_PTR(-EFAULT);
+		return -EFAULT;
 	}
 
 	if (memchr_inv(ucmd.reserved, 0, sizeof(ucmd.reserved)))
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	if (ucmd.comp_mask || ucmd.reserved1)
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	if (udata->inlen > sizeof(ucmd) &&
 	    !ib_is_udata_cleared(udata, sizeof(ucmd),
 				 udata->inlen - sizeof(ucmd))) {
 		pr_debug("inlen is not supported\n");
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 	}
 
 	if (init_attr->qp_type != IB_QPT_RAW_PACKET) {
 		pr_debug("RSS QP with unsupported QP type %d\n",
 			 init_attr->qp_type);
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 	}
 
 	if (init_attr->create_flags) {
 		pr_debug("RSS QP doesn't support create flags\n");
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 	}
 
 	if (init_attr->send_cq || init_attr->cap.max_send_wr) {
 		pr_debug("RSS QP with unsupported send attributes\n");
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 	}
 
-	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
-	if (!qp)
-		return ERR_PTR(-ENOMEM);
-
 	qp->pri.vid = 0xFFFF;
 	qp->alt.vid = 0xFFFF;
 
 	err = create_qp_rss(to_mdev(pd->device), init_attr, &ucmd, qp);
-	if (err) {
+	if (err)
-		kfree(qp);
+		return err;
-		return ERR_PTR(err);
-	}
 
 	qp->ibqp.qp_num = qp->mqp.qpn;
-
+	return 0;
-	return &qp->ibqp;
 }
 
 /*
@@ -873,7 +837,6 @@ static int create_rq(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 
 	qp->mlx4_ib_qp_type = MLX4_IB_QPT_RAW_PACKET;
 
-	mutex_init(&qp->mutex);
 	spin_lock_init(&qp->sq.lock);
 	spin_lock_init(&qp->rq.lock);
 	INIT_LIST_HEAD(&qp->gid_list);
@@ -922,7 +885,6 @@ static int create_rq(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 		goto err;
 	}
 
-	n = ib_umem_page_count(qp->umem);
 	shift = mlx4_ib_umem_calc_optimal_mtt_size(qp->umem, 0, &n);
 	err = mlx4_mtt_init(dev->dev, n, shift, &qp->mtt);
 
@@ -989,13 +951,11 @@ err:
 
 static int create_qp_common(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 			    struct ib_udata *udata, int sqpn,
-			    struct mlx4_ib_qp **caller_qp)
+			    struct mlx4_ib_qp *qp)
 {
 	struct mlx4_ib_dev *dev = to_mdev(pd->device);
 	int qpn;
 	int err;
-	struct mlx4_ib_sqp *sqp = NULL;
-	struct mlx4_ib_qp *qp;
 	struct mlx4_ib_ucontext *context = rdma_udata_to_drv_context(
 		udata, struct mlx4_ib_ucontext, ibucontext);
 	enum mlx4_ib_qp_type qp_type = (enum mlx4_ib_qp_type) init_attr->qp_type;
@@ -1043,27 +1003,18 @@ static int create_qp_common(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 		sqpn = qpn;
 	}
 
-	if (!*caller_qp) {
+	if (init_attr->qp_type == IB_QPT_SMI ||
-		if (qp_type == MLX4_IB_QPT_SMI || qp_type == MLX4_IB_QPT_GSI ||
+	    init_attr->qp_type == IB_QPT_GSI || qp_type == MLX4_IB_QPT_SMI ||
-		    (qp_type & (MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_SMI_OWNER |
+	    qp_type == MLX4_IB_QPT_GSI ||
-				MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER))) {
+	    (qp_type & (MLX4_IB_QPT_PROXY_SMI | MLX4_IB_QPT_PROXY_SMI_OWNER |
-			sqp = kzalloc(sizeof(struct mlx4_ib_sqp), GFP_KERNEL);
+			MLX4_IB_QPT_PROXY_GSI | MLX4_IB_QPT_TUN_SMI_OWNER))) {
-			if (!sqp)
+		qp->sqp = kzalloc(sizeof(struct mlx4_ib_sqp), GFP_KERNEL);
-				return -ENOMEM;
+		if (!qp->sqp)
-			qp = &sqp->qp;
+			return -ENOMEM;
-		} else {
+	}
-			qp = kzalloc(sizeof(struct mlx4_ib_qp), GFP_KERNEL);
-			if (!qp)
-				return -ENOMEM;
-		}
-		qp->pri.vid = 0xFFFF;
-		qp->alt.vid = 0xFFFF;
-	} else
-		qp = *caller_qp;
 
 	qp->mlx4_ib_qp_type = qp_type;
 
-	mutex_init(&qp->mutex);
 	spin_lock_init(&qp->sq.lock);
 	spin_lock_init(&qp->rq.lock);
 	INIT_LIST_HEAD(&qp->gid_list);
@@ -1117,7 +1068,6 @@ static int create_qp_common(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 			goto err;
 		}
 
-		n = ib_umem_page_count(qp->umem);
 		shift = mlx4_ib_umem_calc_optimal_mtt_size(qp->umem, 0, &n);
 		err = mlx4_mtt_init(dev->dev, n, shift, &qp->mtt);
 
@@ -1239,9 +1189,6 @@ static int create_qp_common(struct ib_pd *pd, struct ib_qp_init_attr *init_attr,
 
 	qp->mqp.event = mlx4_ib_qp_event;
 
-	if (!*caller_qp)
-		*caller_qp = qp;
-
 	spin_lock_irqsave(&dev->reset_flow_resource_lock, flags);
 	mlx4_ib_lock_cqs(to_mcq(init_attr->send_cq),
 			 to_mcq(init_attr->recv_cq));
@@ -1293,10 +1240,7 @@ err_db:
 		mlx4_db_free(dev->dev, &qp->db);
 
 err:
-	if (!sqp && !*caller_qp)
+	kfree(qp->sqp);
-		kfree(qp);
-	kfree(sqp);
-
 	return err;
 }
 
@@ -1410,7 +1354,6 @@ static void destroy_qp_rss(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp)
 	mlx4_qp_free(dev->dev, &qp->mqp);
 	mlx4_qp_release_range(dev->dev, qp->mqp.qpn, 1);
 	del_gid_entries(qp);
-	kfree(qp->rss_ctx);
 }
 
 static void destroy_qp_common(struct mlx4_ib_dev *dev, struct mlx4_ib_qp *qp,
@@ -1529,17 +1472,16 @@ static u32 get_sqp_num(struct mlx4_ib_dev *dev, struct ib_qp_init_attr *attr)
 		return dev->dev->caps.spec_qps[attr->port_num - 1].qp1_proxy;
 }
 
-static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd,
+static int _mlx4_ib_create_qp(struct ib_pd *pd, struct mlx4_ib_qp *qp,
-					struct ib_qp_init_attr *init_attr,
+			      struct ib_qp_init_attr *init_attr,
-					struct ib_udata *udata)
+			      struct ib_udata *udata)
 {
-	struct mlx4_ib_qp *qp = NULL;
 	int err;
 	int sup_u_create_flags = MLX4_IB_QP_BLOCK_MULTICAST_LOOPBACK;
 	u16 xrcdn = 0;
 
 	if (init_attr->rwq_ind_tbl)
-		return _mlx4_ib_create_qp_rss(pd, init_attr, udata);
+		return _mlx4_ib_create_qp_rss(pd, qp, init_attr, udata);
 
 	/*
 	 * We only support LSO, vendor flag1, and multicast loopback blocking,
@@ -1551,16 +1493,16 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd,
 					MLX4_IB_SRIOV_SQP |
 					MLX4_IB_QP_NETIF |
 					MLX4_IB_QP_CREATE_ROCE_V2_GSI))
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	if (init_attr->create_flags & IB_QP_CREATE_NETIF_QP) {
 		if (init_attr->qp_type != IB_QPT_UD)
-			return ERR_PTR(-EINVAL);
+			return -EINVAL;
 	}
 
 	if (init_attr->create_flags) {
 		if (udata && init_attr->create_flags & ~(sup_u_create_flags))
-			return ERR_PTR(-EINVAL);
+			return -EINVAL;
 
 		if ((init_attr->create_flags & ~(MLX4_IB_SRIOV_SQP |
 						 MLX4_IB_QP_CREATE_ROCE_V2_GSI  |
@@ -1570,7 +1512,7 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd,
 		     init_attr->qp_type > IB_QPT_GSI) ||
 		    (init_attr->create_flags & MLX4_IB_QP_CREATE_ROCE_V2_GSI &&
 		     init_attr->qp_type != IB_QPT_GSI))
-			return ERR_PTR(-EINVAL);
+			return -EINVAL;
 	}
 
 	switch (init_attr->qp_type) {
@@ -1581,53 +1523,43 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd,
 		fallthrough;
 	case IB_QPT_XRC_INI:
 		if (!(to_mdev(pd->device)->dev->caps.flags & MLX4_DEV_CAP_FLAG_XRC))
-			return ERR_PTR(-ENOSYS);
+			return -ENOSYS;
 		init_attr->recv_cq = init_attr->send_cq;
 		fallthrough;
 	case IB_QPT_RC:
 	case IB_QPT_UC:
 	case IB_QPT_RAW_PACKET:
-		qp = kzalloc(sizeof(*qp), GFP_KERNEL);
+	case IB_QPT_UD:
-		if (!qp)
-			return ERR_PTR(-ENOMEM);
 		qp->pri.vid = 0xFFFF;
 		qp->alt.vid = 0xFFFF;
-		fallthrough;
+		err = create_qp_common(pd, init_attr, udata, 0, qp);
-	case IB_QPT_UD:
+		if (err)
-	{
+			return err;
-		err = create_qp_common(pd, init_attr, udata, 0, &qp);
-		if (err) {
-			kfree(qp);
-			return ERR_PTR(err);
-		}
 
 		qp->ibqp.qp_num = qp->mqp.qpn;
 		qp->xrcdn = xrcdn;
-
 		break;
-	}
 	case IB_QPT_SMI:
 	case IB_QPT_GSI:
 	{
 		int sqpn;
 
-		/* Userspace is not allowed to create special QPs: */
-		if (udata)
-			return ERR_PTR(-EINVAL);
 		if (init_attr->create_flags & MLX4_IB_QP_CREATE_ROCE_V2_GSI) {
 			int res = mlx4_qp_reserve_range(to_mdev(pd->device)->dev,
 							1, 1, &sqpn, 0,
 							MLX4_RES_USAGE_DRIVER);
 
 			if (res)
-				return ERR_PTR(res);
+				return res;
 		} else {
 			sqpn = get_sqp_num(to_mdev(pd->device), init_attr);
 		}
 
-		err = create_qp_common(pd, init_attr, udata, sqpn, &qp);
+		qp->pri.vid = 0xFFFF;
+		qp->alt.vid = 0xFFFF;
+		err = create_qp_common(pd, init_attr, udata, sqpn, qp);
 		if (err)
-			return ERR_PTR(err);
+			return err;
 
 		qp->port	= init_attr->port_num;
 		qp->ibqp.qp_num = init_attr->qp_type == IB_QPT_SMI ? 0 :
@@ -1636,25 +1568,33 @@ static struct ib_qp *_mlx4_ib_create_qp(struct ib_pd *pd,
 	}
 	default:
 		/* Don't support raw QPs */
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 	}
-
+	return 0;
-	return &qp->ibqp;
 }
 
 struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 				struct ib_qp_init_attr *init_attr,
 				struct ib_udata *udata) {
 	struct ib_device *device = pd ? pd->device : init_attr->xrcd->device;
-	struct ib_qp *ibqp;
 	struct mlx4_ib_dev *dev = to_mdev(device);
+	struct mlx4_ib_qp *qp;
+	int ret;
 
-	ibqp = _mlx4_ib_create_qp(pd, init_attr, udata);
+	qp = kzalloc(sizeof(*qp), GFP_KERNEL);
+	if (!qp)
+		return ERR_PTR(-ENOMEM);
 
-	if (!IS_ERR(ibqp) &&
+	mutex_init(&qp->mutex);
-	    (init_attr->qp_type == IB_QPT_GSI) &&
+	ret = _mlx4_ib_create_qp(pd, qp, init_attr, udata);
+	if (ret) {
+		kfree(qp);
+		return ERR_PTR(ret);
+	}
+
+	if (init_attr->qp_type == IB_QPT_GSI &&
 	    !(init_attr->create_flags & MLX4_IB_QP_CREATE_ROCE_V2_GSI)) {
-		struct mlx4_ib_sqp *sqp = to_msqp((to_mqp(ibqp)));
+		struct mlx4_ib_sqp *sqp = qp->sqp;
 		int is_eth = rdma_cap_eth_ah(&dev->ib_dev, init_attr->port_num);
 
 		if (is_eth &&
@@ -1666,14 +1606,14 @@ struct ib_qp *mlx4_ib_create_qp(struct ib_pd *pd,
 				pr_err("Failed to create GSI QP for RoCEv2 (%ld)\n", PTR_ERR(sqp->roce_v2_gsi));
 				sqp->roce_v2_gsi = NULL;
 			} else {
-				sqp = to_msqp(to_mqp(sqp->roce_v2_gsi));
+				to_mqp(sqp->roce_v2_gsi)->flags |=
-				sqp->qp.flags |= MLX4_IB_ROCE_V2_GSI_QP;
+					MLX4_IB_ROCE_V2_GSI_QP;
 			}
 
 			init_attr->create_flags &= ~MLX4_IB_QP_CREATE_ROCE_V2_GSI;
 		}
 	}
-	return ibqp;
+	return &qp->ibqp;
 }
 
 static int _mlx4_ib_destroy_qp(struct ib_qp *qp, struct ib_udata *udata)
@@ -1700,10 +1640,8 @@ static int _mlx4_ib_destroy_qp(struct ib_qp *qp, struct ib_udata *udata)
 		destroy_qp_common(dev, mqp, MLX4_IB_QP_SRC, udata);
 	}
 
-	if (is_sqp(dev, mqp))
+	kfree(mqp->sqp);
-		kfree(to_msqp(mqp));
+	kfree(mqp);
-	else
-		kfree(mqp);
 
 	return 0;
 }
@@ -1713,7 +1651,7 @@ int mlx4_ib_destroy_qp(struct ib_qp *qp, struct ib_udata *udata)
 	struct mlx4_ib_qp *mqp = to_mqp(qp);
 
 	if (mqp->mlx4_ib_qp_type == MLX4_IB_QPT_GSI) {
-		struct mlx4_ib_sqp *sqp = to_msqp(mqp);
+		struct mlx4_ib_sqp *sqp = mqp->sqp;
 
 		if (sqp->roce_v2_gsi)
 			ib_destroy_qp(sqp->roce_v2_gsi);
@@ -2575,7 +2513,7 @@ static int __mlx4_ib_modify_qp(void *src, enum mlx4_ib_source_type src_type,
 		qp->alt_port = attr->alt_port_num;
 
 	if (is_sqp(dev, qp))
-		store_sqp_attrs(to_msqp(qp), attr, attr_mask);
+		store_sqp_attrs(qp->sqp, attr, attr_mask);
 
 	/*
 	 * If we moved QP0 to RTR, bring the IB link up; if we moved
@@ -2852,7 +2790,7 @@ int mlx4_ib_modify_qp(struct ib_qp *ibqp, struct ib_qp_attr *attr,
 	ret = _mlx4_ib_modify_qp(ibqp, attr, attr_mask, udata);
 
 	if (mqp->mlx4_ib_qp_type == MLX4_IB_QPT_GSI) {
-		struct mlx4_ib_sqp *sqp = to_msqp(mqp);
+		struct mlx4_ib_sqp *sqp = mqp->sqp;
 		int err = 0;
 
 		if (sqp->roce_v2_gsi)
@@ -2877,12 +2815,13 @@ static int vf_get_qp0_qkey(struct mlx4_dev *dev, int qpn, u32 *qkey)
 	return -EINVAL;
 }
 
-static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
+static int build_sriov_qp0_header(struct mlx4_ib_qp *qp,
 				  const struct ib_ud_wr *wr,
 				  void *wqe, unsigned *mlx_seg_len)
 {
-	struct mlx4_ib_dev *mdev = to_mdev(sqp->qp.ibqp.device);
+	struct mlx4_ib_dev *mdev = to_mdev(qp->ibqp.device);
-	struct ib_device *ib_dev = &mdev->ib_dev;
+	struct mlx4_ib_sqp *sqp = qp->sqp;
+	struct ib_device *ib_dev = qp->ibqp.device;
 	struct mlx4_wqe_mlx_seg *mlx = wqe;
 	struct mlx4_wqe_inline_seg *inl = wqe + sizeof *mlx;
 	struct mlx4_ib_ah *ah = to_mah(wr->ah);
@@ -2904,12 +2843,12 @@ static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
 
 	/* for proxy-qp0 sends, need to add in size of tunnel header */
 	/* for tunnel-qp0 sends, tunnel header is already in s/g list */
-	if (sqp->qp.mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_SMI_OWNER)
+	if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_SMI_OWNER)
 		send_size += sizeof (struct mlx4_ib_tunnel_header);
 
 	ib_ud_header_init(send_size, 1, 0, 0, 0, 0, 0, 0, &sqp->ud_header);
 
-	if (sqp->qp.mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_SMI_OWNER) {
+	if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_PROXY_SMI_OWNER) {
 		sqp->ud_header.lrh.service_level =
 			be32_to_cpu(ah->av.ib.sl_tclass_flowlabel) >> 28;
 		sqp->ud_header.lrh.destination_lid =
@@ -2926,26 +2865,26 @@ static int build_sriov_qp0_header(struct mlx4_ib_sqp *sqp,
 
 	sqp->ud_header.lrh.virtual_lane    = 0;
 	sqp->ud_header.bth.solicited_event = !!(wr->wr.send_flags & IB_SEND_SOLICITED);
-	err = ib_get_cached_pkey(ib_dev, sqp->qp.port, 0, &pkey);
+	err = ib_get_cached_pkey(ib_dev, qp->port, 0, &pkey);
 	if (err)
 		return err;
 	sqp->ud_header.bth.pkey = cpu_to_be16(pkey);
-	if (sqp->qp.mlx4_ib_qp_type == MLX4_IB_QPT_TUN_SMI_OWNER)
+	if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_TUN_SMI_OWNER)
 		sqp->ud_header.bth.destination_qpn = cpu_to_be32(wr->remote_qpn);
 	else
 		sqp->ud_header.bth.destination_qpn =
-			cpu_to_be32(mdev->dev->caps.spec_qps[sqp->qp.port - 1].qp0_tunnel);
+			cpu_to_be32(mdev->dev->caps.spec_qps[qp->port - 1].qp0_tunnel);
 
 	sqp->ud_header.bth.psn = cpu_to_be32((sqp->send_psn++) & ((1 << 24) - 1));
 	if (mlx4_is_master(mdev->dev)) {
-		if (mlx4_get_parav_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+		if (mlx4_get_parav_qkey(mdev->dev, qp->mqp.qpn, &qkey))
 			return -EINVAL;
 	} else {
-		if (vf_get_qp0_qkey(mdev->dev, sqp->qp.mqp.qpn, &qkey))
+		if (vf_get_qp0_qkey(mdev->dev, qp->mqp.qpn, &qkey))
 			return -EINVAL;
 	}
 	sqp->ud_header.deth.qkey = cpu_to_be32(qkey);
-	sqp->ud_header.deth.source_qpn = cpu_to_be32(sqp->qp.mqp.qpn);
+	sqp->ud_header.deth.source_qpn = cpu_to_be32(qp->mqp.qpn);
 
 	sqp->ud_header.bth.opcode        = IB_OPCODE_UD_SEND_ONLY;
 	sqp->ud_header.immediate_present = 0;
@@ -3029,10 +2968,11 @@ static int fill_gid_by_hw_index(struct mlx4_ib_dev *ibdev, u8 port_num,
 }
 
 #define MLX4_ROCEV2_QP1_SPORT 0xC000
-static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
+static int build_mlx_header(struct mlx4_ib_qp *qp, const struct ib_ud_wr *wr,
 			    void *wqe, unsigned *mlx_seg_len)
 {
-	struct ib_device *ib_dev = sqp->qp.ibqp.device;
+	struct mlx4_ib_sqp *sqp = qp->sqp;
+	struct ib_device *ib_dev = qp->ibqp.device;
 	struct mlx4_ib_dev *ibdev = to_mdev(ib_dev);
 	struct mlx4_wqe_mlx_seg *mlx = wqe;
 	struct mlx4_wqe_ctrl_seg *ctrl = wqe;
@@ -3056,7 +2996,7 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 	for (i = 0; i < wr->wr.num_sge; ++i)
 		send_size += wr->wr.sg_list[i].length;
 
-	is_eth = rdma_port_get_link_layer(sqp->qp.ibqp.device, sqp->qp.port) == IB_LINK_LAYER_ETHERNET;
+	is_eth = rdma_port_get_link_layer(qp->ibqp.device, qp->port) == IB_LINK_LAYER_ETHERNET;
 	is_grh = mlx4_ib_ah_grh_present(ah);
 	if (is_eth) {
 		enum ib_gid_type gid_type;
@@ -3070,9 +3010,9 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 			if (err)
 				return err;
 		} else  {
-			err = fill_gid_by_hw_index(ibdev, sqp->qp.port,
+			err = fill_gid_by_hw_index(ibdev, qp->port,
-					    ah->av.ib.gid_index,
+						   ah->av.ib.gid_index, &sgid,
-					    &sgid, &gid_type);
+						   &gid_type);
 			if (!err) {
 				is_udp = gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP;
 				if (is_udp) {
@@ -3117,13 +3057,18 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 				 * indexes don't necessarily match the hw ones, so
 				 * we must use our own cache
 				 */
-				sqp->ud_header.grh.source_gid.global.subnet_prefix =
+				sqp->ud_header.grh.source_gid.global
-					cpu_to_be64(atomic64_read(&(to_mdev(ib_dev)->sriov.
+					.subnet_prefix =
-								    demux[sqp->qp.port - 1].
+					cpu_to_be64(atomic64_read(
-								    subnet_prefix)));
+						&(to_mdev(ib_dev)
-				sqp->ud_header.grh.source_gid.global.interface_id =
+							  ->sriov
-					to_mdev(ib_dev)->sriov.demux[sqp->qp.port - 1].
+							  .demux[qp->port - 1]
-						       guid_cache[ah->av.ib.gid_index];
+							  .subnet_prefix)));
+				sqp->ud_header.grh.source_gid.global
+					.interface_id =
+					to_mdev(ib_dev)
+						->sriov.demux[qp->port - 1]
+						.guid_cache[ah->av.ib.gid_index];
 			} else {
 				sqp->ud_header.grh.source_gid =
 					ah->ibah.sgid_attr->gid;
@@ -3155,10 +3100,13 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 	mlx->flags &= cpu_to_be32(MLX4_WQE_CTRL_CQ_UPDATE);
 
 	if (!is_eth) {
-		mlx->flags |= cpu_to_be32((!sqp->qp.ibqp.qp_num ? MLX4_WQE_MLX_VL15 : 0) |
+		mlx->flags |=
-					  (sqp->ud_header.lrh.destination_lid ==
+			cpu_to_be32((!qp->ibqp.qp_num ? MLX4_WQE_MLX_VL15 : 0) |
-					   IB_LID_PERMISSIVE ? MLX4_WQE_MLX_SLR : 0) |
+				    (sqp->ud_header.lrh.destination_lid ==
-					  (sqp->ud_header.lrh.service_level << 8));
+						     IB_LID_PERMISSIVE ?
+					     MLX4_WQE_MLX_SLR :
+					     0) |
+				    (sqp->ud_header.lrh.service_level << 8));
 		if (ah->av.ib.port_pd & cpu_to_be32(0x80000000))
 			mlx->flags |= cpu_to_be32(0x1); /* force loopback */
 		mlx->rlid = sqp->ud_header.lrh.destination_lid;
@@ -3204,21 +3152,23 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 			sqp->ud_header.vlan.tag = cpu_to_be16(vlan | pcp);
 		}
 	} else {
-		sqp->ud_header.lrh.virtual_lane    = !sqp->qp.ibqp.qp_num ? 15 :
+		sqp->ud_header.lrh.virtual_lane =
-							sl_to_vl(to_mdev(ib_dev),
+			!qp->ibqp.qp_num ?
-								 sqp->ud_header.lrh.service_level,
+				15 :
-								 sqp->qp.port);
+				sl_to_vl(to_mdev(ib_dev),
-		if (sqp->qp.ibqp.qp_num && sqp->ud_header.lrh.virtual_lane == 15)
+					 sqp->ud_header.lrh.service_level,
+					 qp->port);
+		if (qp->ibqp.qp_num && sqp->ud_header.lrh.virtual_lane == 15)
 			return -EINVAL;
 		if (sqp->ud_header.lrh.destination_lid == IB_LID_PERMISSIVE)
 			sqp->ud_header.lrh.source_lid = IB_LID_PERMISSIVE;
 	}
 	sqp->ud_header.bth.solicited_event = !!(wr->wr.send_flags & IB_SEND_SOLICITED);
-	if (!sqp->qp.ibqp.qp_num)
+	if (!qp->ibqp.qp_num)
-		err = ib_get_cached_pkey(ib_dev, sqp->qp.port, sqp->pkey_index,
+		err = ib_get_cached_pkey(ib_dev, qp->port, sqp->pkey_index,
 					 &pkey);
 	else
-		err = ib_get_cached_pkey(ib_dev, sqp->qp.port, wr->pkey_index,
+		err = ib_get_cached_pkey(ib_dev, qp->port, wr->pkey_index,
 					 &pkey);
 	if (err)
 		return err;
@@ -3228,7 +3178,7 @@ static int build_mlx_header(struct mlx4_ib_sqp *sqp, const struct ib_ud_wr *wr,
 	sqp->ud_header.bth.psn = cpu_to_be32((sqp->send_psn++) & ((1 << 24) - 1));
 	sqp->ud_header.deth.qkey = cpu_to_be32(wr->remote_qkey & 0x80000000 ?
 					       sqp->qkey : wr->remote_qkey);
-	sqp->ud_header.deth.source_qpn = cpu_to_be32(sqp->qp.ibqp.qp_num);
+	sqp->ud_header.deth.source_qpn = cpu_to_be32(qp->ibqp.qp_num);
 
 	header_size = ib_ud_header_pack(&sqp->ud_header, sqp->header_buf);
 
@@ -3551,14 +3501,14 @@ static int _mlx4_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 	struct mlx4_ib_dev *mdev = to_mdev(ibqp->device);
 
 	if (qp->mlx4_ib_qp_type == MLX4_IB_QPT_GSI) {
-		struct mlx4_ib_sqp *sqp = to_msqp(qp);
+		struct mlx4_ib_sqp *sqp = qp->sqp;
 
 		if (sqp->roce_v2_gsi) {
 			struct mlx4_ib_ah *ah = to_mah(ud_wr(wr)->ah);
 			enum ib_gid_type gid_type;
 			union ib_gid gid;
 
-			if (!fill_gid_by_hw_index(mdev, sqp->qp.port,
+			if (!fill_gid_by_hw_index(mdev, qp->port,
 					   ah->av.ib.gid_index,
 					   &gid, &gid_type))
 				qp = (gid_type == IB_GID_TYPE_ROCE_UDP_ENCAP) ?
@@ -3678,8 +3628,8 @@ static int _mlx4_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 			break;
 
 		case MLX4_IB_QPT_TUN_SMI_OWNER:
-			err =  build_sriov_qp0_header(to_msqp(qp), ud_wr(wr),
+			err = build_sriov_qp0_header(qp, ud_wr(wr), ctrl,
-					ctrl, &seglen);
+						     &seglen);
 			if (unlikely(err)) {
 				*bad_wr = wr;
 				goto out;
@@ -3715,8 +3665,8 @@ static int _mlx4_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 			break;
 
 		case MLX4_IB_QPT_PROXY_SMI_OWNER:
-			err = build_sriov_qp0_header(to_msqp(qp), ud_wr(wr),
+			err = build_sriov_qp0_header(qp, ud_wr(wr), ctrl,
-					ctrl, &seglen);
+						     &seglen);
 			if (unlikely(err)) {
 				*bad_wr = wr;
 				goto out;
@@ -3749,8 +3699,7 @@ static int _mlx4_ib_post_send(struct ib_qp *ibqp, const struct ib_send_wr *wr,
 
 		case MLX4_IB_QPT_SMI:
 		case MLX4_IB_QPT_GSI:
-			err = build_mlx_header(to_msqp(qp), ud_wr(wr), ctrl,
+			err = build_mlx_header(qp, ud_wr(wr), ctrl, &seglen);
-					&seglen);
 			if (unlikely(err)) {
 				*bad_wr = wr;
 				goto out;
@@ -4172,6 +4121,7 @@ struct ib_wq *mlx4_ib_create_wq(struct ib_pd *pd,
 	if (!qp)
 		return ERR_PTR(-ENOMEM);
 
+	mutex_init(&qp->mutex);
 	qp->pri.vid = 0xFFFF;
 	qp->alt.vid = 0xFFFF;
 
@@ -4327,7 +4277,7 @@ int mlx4_ib_modify_wq(struct ib_wq *ibwq, struct ib_wq_attr *wq_attr,
 	return err;
 }
 
-void mlx4_ib_destroy_wq(struct ib_wq *ibwq, struct ib_udata *udata)
+int mlx4_ib_destroy_wq(struct ib_wq *ibwq, struct ib_udata *udata)
 {
 	struct mlx4_ib_dev *dev = to_mdev(ibwq->device);
 	struct mlx4_ib_qp *qp = to_mqp((struct ib_qp *)ibwq);
@@ -4338,36 +4288,35 @@ void mlx4_ib_destroy_wq(struct ib_wq *ibwq, struct ib_udata *udata)
 	destroy_qp_common(dev, qp, MLX4_IB_RWQ_SRC, udata);
 
 	kfree(qp);
+	return 0;
 }
 
-struct ib_rwq_ind_table
+int mlx4_ib_create_rwq_ind_table(struct ib_rwq_ind_table *rwq_ind_table,
-*mlx4_ib_create_rwq_ind_table(struct ib_device *device,
+				 struct ib_rwq_ind_table_init_attr *init_attr,
-			      struct ib_rwq_ind_table_init_attr *init_attr,
+				 struct ib_udata *udata)
-			      struct ib_udata *udata)
 {
-	struct ib_rwq_ind_table *rwq_ind_table;
 	struct mlx4_ib_create_rwq_ind_tbl_resp resp = {};
 	unsigned int ind_tbl_size = 1 << init_attr->log_ind_tbl_size;
+	struct ib_device *device = rwq_ind_table->device;
 	unsigned int base_wqn;
 	size_t min_resp_len;
-	int i;
+	int i, err = 0;
-	int err;
 
 	if (udata->inlen > 0 &&
 	    !ib_is_udata_cleared(udata, 0,
 				 udata->inlen))
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	min_resp_len = offsetof(typeof(resp), reserved) + sizeof(resp.reserved);
 	if (udata->outlen && udata->outlen < min_resp_len)
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 
 	if (ind_tbl_size >
 	    device->attrs.rss_caps.max_rwq_indirection_table_size) {
 		pr_debug("log_ind_tbl_size = %d is bigger than supported = %d\n",
 			 ind_tbl_size,
 			 device->attrs.rss_caps.max_rwq_indirection_table_size);
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 	}
 
 	base_wqn = init_attr->ind_tbl[0]->wq_num;
@@ -4375,39 +4324,23 @@ struct ib_rwq_ind_table
 	if (base_wqn % ind_tbl_size) {
 		pr_debug("WQN=0x%x isn't aligned with indirection table size\n",
 			 base_wqn);
-		return ERR_PTR(-EINVAL);
+		return -EINVAL;
 	}
 
 	for (i = 1; i < ind_tbl_size; i++) {
 		if (++base_wqn != init_attr->ind_tbl[i]->wq_num) {
 			pr_debug("indirection table's WQNs aren't consecutive\n");
-			return ERR_PTR(-EINVAL);
+			return -EINVAL;
 		}
 	}
 
-	rwq_ind_table = kzalloc(sizeof(*rwq_ind_table), GFP_KERNEL);
-	if (!rwq_ind_table)
-		return ERR_PTR(-ENOMEM);
-
 	if (udata->outlen) {
 		resp.response_length = offsetof(typeof(resp), response_length) +
 					sizeof(resp.response_length);
 		err = ib_copy_to_udata(udata, &resp, resp.response_length);
-		if (err)
-			goto err;
 	}
 
-	return rwq_ind_table;
+	return err;
-
-err:
-	kfree(rwq_ind_table);
-	return ERR_PTR(err);
-}
-
-int mlx4_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_tbl)
-{
-	kfree(ib_rwq_ind_tbl);
-	return 0;
 }
 
 struct mlx4_ib_drain_cqe {

drivers/infiniband/hw/mlx4/srq.c

@@ -115,8 +115,9 @@ int mlx4_ib_create_srq(struct ib_srq *ib_srq,
 		if (IS_ERR(srq->umem))
 			return PTR_ERR(srq->umem);
 
-		err = mlx4_mtt_init(dev->dev, ib_umem_page_count(srq->umem),
+		err = mlx4_mtt_init(
-				    PAGE_SHIFT, &srq->mtt);
+			dev->dev, ib_umem_num_dma_blocks(srq->umem, PAGE_SIZE),
+			PAGE_SHIFT, &srq->mtt);
 		if (err)
 			goto err_buf;
 
@@ -260,7 +261,7 @@ int mlx4_ib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr)
 	return 0;
 }
 
-void mlx4_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata)
+int mlx4_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata)
 {
 	struct mlx4_ib_dev *dev = to_mdev(srq->device);
 	struct mlx4_ib_srq *msrq = to_msrq(srq);
@@ -282,6 +283,7 @@ void mlx4_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata)
 		mlx4_db_free(dev->dev, &msrq->db);
 	}
 	ib_umem_release(msrq->umem);
+	return 0;
 }
 
 void mlx4_ib_free_srq_wqe(struct mlx4_ib_srq *srq, int wqe_index)

drivers/infiniband/hw/mlx5/ah.c

@@ -106,8 +106,8 @@ int mlx5_ib_create_ah(struct ib_ah *ibah, struct rdma_ah_init_attr *init_attr,
 	if (ah_type == RDMA_AH_ATTR_TYPE_ROCE && udata) {
 		int err;
 		struct mlx5_ib_create_ah_resp resp = {};
-		u32 min_resp_len = offsetof(typeof(resp), dmac) +
+		u32 min_resp_len =
-				   sizeof(resp.dmac);
+			offsetofend(struct mlx5_ib_create_ah_resp, dmac);
 
 		if (udata->outlen < min_resp_len)
 			return -EINVAL;
@@ -147,8 +147,3 @@ int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr)
 
 	return 0;
 }
-
-void mlx5_ib_destroy_ah(struct ib_ah *ah, u32 flags)
-{
-	return;
-}

drivers/infiniband/hw/mlx5/cmd.c

@@ -168,14 +168,14 @@ void mlx5_cmd_destroy_tis(struct mlx5_core_dev *dev, u32 tisn, u16 uid)
 	mlx5_cmd_exec_in(dev, destroy_tis, in);
 }
 
-void mlx5_cmd_destroy_rqt(struct mlx5_core_dev *dev, u32 rqtn, u16 uid)
+int mlx5_cmd_destroy_rqt(struct mlx5_core_dev *dev, u32 rqtn, u16 uid)
 {
 	u32 in[MLX5_ST_SZ_DW(destroy_rqt_in)] = {};
 
 	MLX5_SET(destroy_rqt_in, in, opcode, MLX5_CMD_OP_DESTROY_RQT);
 	MLX5_SET(destroy_rqt_in, in, rqtn, rqtn);
 	MLX5_SET(destroy_rqt_in, in, uid, uid);
-	mlx5_cmd_exec_in(dev, destroy_rqt, in);
+	return mlx5_cmd_exec_in(dev, destroy_rqt, in);
 }
 
 int mlx5_cmd_alloc_transport_domain(struct mlx5_core_dev *dev, u32 *tdn,
@@ -209,14 +209,14 @@ void mlx5_cmd_dealloc_transport_domain(struct mlx5_core_dev *dev, u32 tdn,
 	mlx5_cmd_exec_in(dev, dealloc_transport_domain, in);
 }
 
-void mlx5_cmd_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn, u16 uid)
+int mlx5_cmd_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn, u16 uid)
 {
 	u32 in[MLX5_ST_SZ_DW(dealloc_pd_in)] = {};
 
 	MLX5_SET(dealloc_pd_in, in, opcode, MLX5_CMD_OP_DEALLOC_PD);
 	MLX5_SET(dealloc_pd_in, in, pd, pdn);
 	MLX5_SET(dealloc_pd_in, in, uid, uid);
-	mlx5_cmd_exec_in(dev, dealloc_pd, in);
+	return mlx5_cmd_exec_in(dev, dealloc_pd, in);
 }
 
 int mlx5_cmd_attach_mcg(struct mlx5_core_dev *dev, union ib_gid *mgid,

drivers/infiniband/hw/mlx5/cmd.h

@@ -44,10 +44,10 @@ int mlx5_cmd_query_cong_params(struct mlx5_core_dev *dev, int cong_point,
 int mlx5_cmd_alloc_memic(struct mlx5_dm *dm, phys_addr_t *addr,
 			 u64 length, u32 alignment);
 void mlx5_cmd_dealloc_memic(struct mlx5_dm *dm, phys_addr_t addr, u64 length);
-void mlx5_cmd_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn, u16 uid);
+int mlx5_cmd_dealloc_pd(struct mlx5_core_dev *dev, u32 pdn, u16 uid);
 void mlx5_cmd_destroy_tir(struct mlx5_core_dev *dev, u32 tirn, u16 uid);
 void mlx5_cmd_destroy_tis(struct mlx5_core_dev *dev, u32 tisn, u16 uid);
-void mlx5_cmd_destroy_rqt(struct mlx5_core_dev *dev, u32 rqtn, u16 uid);
+int mlx5_cmd_destroy_rqt(struct mlx5_core_dev *dev, u32 rqtn, u16 uid);
 int mlx5_cmd_alloc_transport_domain(struct mlx5_core_dev *dev, u32 *tdn,
 				    u16 uid);
 void mlx5_cmd_dealloc_transport_domain(struct mlx5_core_dev *dev, u32 tdn,

drivers/infiniband/hw/mlx5/counters.c

@@ -117,7 +117,7 @@ err_bound:
 	return ret;
 }
 
-static void mlx5_ib_destroy_counters(struct ib_counters *counters)
+static int mlx5_ib_destroy_counters(struct ib_counters *counters)
 {
 	struct mlx5_ib_mcounters *mcounters = to_mcounters(counters);
 
@@ -125,6 +125,7 @@ static void mlx5_ib_destroy_counters(struct ib_counters *counters)
 	if (mcounters->hw_cntrs_hndl)
 		mlx5_fc_destroy(to_mdev(counters->device)->mdev,
 				mcounters->hw_cntrs_hndl);
+	return 0;
 }
 
 static int mlx5_ib_create_counters(struct ib_counters *counters,
@@ -456,12 +457,12 @@ static int __mlx5_ib_alloc_counters(struct mlx5_ib_dev *dev,
 		cnts->num_ext_ppcnt_counters = ARRAY_SIZE(ext_ppcnt_cnts);
 		num_counters += ARRAY_SIZE(ext_ppcnt_cnts);
 	}
-	cnts->names = kcalloc(num_counters, sizeof(cnts->names), GFP_KERNEL);
+	cnts->names = kcalloc(num_counters, sizeof(*cnts->names), GFP_KERNEL);
 	if (!cnts->names)
 		return -ENOMEM;
 
 	cnts->offsets = kcalloc(num_counters,
-				sizeof(cnts->offsets), GFP_KERNEL);
+				sizeof(*cnts->offsets), GFP_KERNEL);
 	if (!cnts->offsets)
 		goto err_names;
 

drivers/infiniband/hw/mlx5/cq.c

@@ -168,7 +168,7 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
 {
 	enum rdma_link_layer ll = rdma_port_get_link_layer(qp->ibqp.device, 1);
 	struct mlx5_ib_dev *dev = to_mdev(qp->ibqp.device);
-	struct mlx5_ib_srq *srq;
+	struct mlx5_ib_srq *srq = NULL;
 	struct mlx5_ib_wq *wq;
 	u16 wqe_ctr;
 	u8  roce_packet_type;
@@ -180,7 +180,8 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
 
 		if (qp->ibqp.xrcd) {
 			msrq = mlx5_cmd_get_srq(dev, be32_to_cpu(cqe->srqn));
-			srq = to_mibsrq(msrq);
+			if (msrq)
+				srq = to_mibsrq(msrq);
 		} else {
 			srq = to_msrq(qp->ibqp.srq);
 		}
@@ -254,7 +255,7 @@ static void handle_responder(struct ib_wc *wc, struct mlx5_cqe64 *cqe,
 
 	switch (roce_packet_type) {
 	case MLX5_CQE_ROCE_L3_HEADER_TYPE_GRH:
-		wc->network_hdr_type = RDMA_NETWORK_IB;
+		wc->network_hdr_type = RDMA_NETWORK_ROCE_V1;
 		break;
 	case MLX5_CQE_ROCE_L3_HEADER_TYPE_IPV6:
 		wc->network_hdr_type = RDMA_NETWORK_IPV6;
@@ -1023,16 +1024,21 @@ err_cqb:
 	return err;
 }
 
-void mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
+int mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *dev = to_mdev(cq->device);
 	struct mlx5_ib_cq *mcq = to_mcq(cq);
+	int ret;
+
+	ret = mlx5_core_destroy_cq(dev->mdev, &mcq->mcq);
+	if (ret)
+		return ret;
 
-	mlx5_core_destroy_cq(dev->mdev, &mcq->mcq);
 	if (udata)
 		destroy_cq_user(mcq, udata);
 	else
 		destroy_cq_kernel(dev, mcq);
+	return 0;
 }
 
 static int is_equal_rsn(struct mlx5_cqe64 *cqe64, u32 rsn)

drivers/infiniband/hw/mlx5/fs.c

@@ -136,12 +136,9 @@ static int check_mpls_supp_fields(u32 field_support, const __be32 *set_mask)
 #define LAST_COUNTERS_FIELD counters
 
 /* Field is the last supported field */
-#define FIELDS_NOT_SUPPORTED(filter, field)\
+#define FIELDS_NOT_SUPPORTED(filter, field)                                    \
-	memchr_inv((void *)&filter.field  +\
+	memchr_inv((void *)&filter.field + sizeof(filter.field), 0,            \
-		   sizeof(filter.field), 0,\
+		   sizeof(filter) - offsetofend(typeof(filter), field))
-		   sizeof(filter) -\
-		   offsetof(typeof(filter), field) -\
-		   sizeof(filter.field))
 
 int parse_flow_flow_action(struct mlx5_ib_flow_action *maction,
 			   bool is_egress,
@@ -767,6 +764,7 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 {
 	bool dont_trap = flow_attr->flags & IB_FLOW_ATTR_FLAGS_DONT_TRAP;
 	struct mlx5_flow_namespace *ns = NULL;
+	enum mlx5_flow_namespace_type fn_type;
 	struct mlx5_ib_flow_prio *prio;
 	struct mlx5_flow_table *ft;
 	int max_table_size;
@@ -780,11 +778,9 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 						       log_max_ft_size));
 	esw_encap = mlx5_eswitch_get_encap_mode(dev->mdev) !=
 		DEVLINK_ESWITCH_ENCAP_MODE_NONE;
-	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
+	switch (flow_attr->type) {
-		enum mlx5_flow_namespace_type fn_type;
+	case IB_FLOW_ATTR_NORMAL:
-
+		if (flow_is_multicast_only(flow_attr) && !dont_trap)
-		if (flow_is_multicast_only(flow_attr) &&
-		    !dont_trap)
 			priority = MLX5_IB_FLOW_MCAST_PRIO;
 		else
 			priority = ib_prio_to_core_prio(flow_attr->priority,
@@ -797,12 +793,11 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 				flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
 			if (!dev->is_rep && !esw_encap &&
 			    MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
-					reformat_l3_tunnel_to_l2))
+						      reformat_l3_tunnel_to_l2))
 				flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
 		} else {
-			max_table_size =
+			max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_TX(
-				BIT(MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev,
+				dev->mdev, log_max_ft_size));
-							      log_max_ft_size));
 			fn_type = MLX5_FLOW_NAMESPACE_EGRESS;
 			prio = &dev->flow_db->egress_prios[priority];
 			if (!dev->is_rep && !esw_encap &&
@@ -812,27 +807,31 @@ static struct mlx5_ib_flow_prio *get_flow_table(struct mlx5_ib_dev *dev,
 		ns = mlx5_get_flow_namespace(dev->mdev, fn_type);
 		num_entries = MLX5_FS_MAX_ENTRIES;
 		num_groups = MLX5_FS_MAX_TYPES;
-	} else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+		break;
-		   flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) {
+	case IB_FLOW_ATTR_ALL_DEFAULT:
+	case IB_FLOW_ATTR_MC_DEFAULT:
 		ns = mlx5_get_flow_namespace(dev->mdev,
 					     MLX5_FLOW_NAMESPACE_LEFTOVERS);
-		build_leftovers_ft_param(&priority,
+		build_leftovers_ft_param(&priority, &num_entries, &num_groups);
-					 &num_entries,
-					 &num_groups);
 		prio = &dev->flow_db->prios[MLX5_IB_FLOW_LEFTOVERS_PRIO];
-	} else if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) {
+		break;
+	case IB_FLOW_ATTR_SNIFFER:
 		if (!MLX5_CAP_FLOWTABLE(dev->mdev,
 					allow_sniffer_and_nic_rx_shared_tir))
 			return ERR_PTR(-EOPNOTSUPP);
 
-		ns = mlx5_get_flow_namespace(dev->mdev, ft_type == MLX5_IB_FT_RX ?
+		ns = mlx5_get_flow_namespace(
-					     MLX5_FLOW_NAMESPACE_SNIFFER_RX :
+			dev->mdev, ft_type == MLX5_IB_FT_RX ?
-					     MLX5_FLOW_NAMESPACE_SNIFFER_TX);
+					   MLX5_FLOW_NAMESPACE_SNIFFER_RX :
+					   MLX5_FLOW_NAMESPACE_SNIFFER_TX);
 
 		prio = &dev->flow_db->sniffer[ft_type];
 		priority = 0;
 		num_entries = 1;
 		num_groups = 1;
+		break;
+	default:
+		break;
 	}
 
 	if (!ns)
@@ -954,7 +953,7 @@ static struct mlx5_ib_flow_handler *_create_flow_rule(struct mlx5_ib_dev *dev,
 	if (!flow_is_multicast_only(flow_attr))
 		set_underlay_qp(dev, spec, underlay_qpn);
 
-	if (dev->is_rep) {
+	if (dev->is_rep && flow_attr->type != IB_FLOW_ATTR_SNIFFER) {
 		struct mlx5_eswitch_rep *rep;
 
 		rep = dev->port[flow_attr->port - 1].rep;
@@ -1116,6 +1115,7 @@ static struct mlx5_ib_flow_handler *create_sniffer_rule(struct mlx5_ib_dev *dev,
 	int err;
 	static const struct ib_flow_attr flow_attr  = {
 		.num_of_specs = 0,
+		.type = IB_FLOW_ATTR_SNIFFER,
 		.size = sizeof(flow_attr)
 	};
 
@@ -1143,10 +1143,8 @@ err:
 	return ERR_PTR(err);
 }
 
-
 static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 					   struct ib_flow_attr *flow_attr,
-					   int domain,
 					   struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *dev = to_mdev(qp->device);
@@ -1162,8 +1160,7 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 	int underlay_qpn;
 
 	if (udata && udata->inlen) {
-		min_ucmd_sz = offsetof(typeof(ucmd_hdr), reserved) +
+		min_ucmd_sz = offsetofend(struct mlx5_ib_create_flow, reserved);
-				sizeof(ucmd_hdr.reserved);
 		if (udata->inlen < min_ucmd_sz)
 			return ERR_PTR(-EOPNOTSUPP);
 
@@ -1197,10 +1194,9 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 		goto free_ucmd;
 	}
 
-	if (domain != IB_FLOW_DOMAIN_USER ||
+	if (flow_attr->port > dev->num_ports ||
-	    flow_attr->port > dev->num_ports ||
+	    (flow_attr->flags &
-	    (flow_attr->flags & ~(IB_FLOW_ATTR_FLAGS_DONT_TRAP |
+	     ~(IB_FLOW_ATTR_FLAGS_DONT_TRAP | IB_FLOW_ATTR_FLAGS_EGRESS))) {
-				  IB_FLOW_ATTR_FLAGS_EGRESS))) {
 		err = -EINVAL;
 		goto free_ucmd;
 	}
@@ -1245,19 +1241,22 @@ static struct ib_flow *mlx5_ib_create_flow(struct ib_qp *qp,
 			dst->tir_num = mqp->raw_packet_qp.rq.tirn;
 	}
 
-	if (flow_attr->type == IB_FLOW_ATTR_NORMAL) {
+	switch (flow_attr->type) {
+	case IB_FLOW_ATTR_NORMAL:
 		underlay_qpn = (mqp->flags & IB_QP_CREATE_SOURCE_QPN) ?
 				       mqp->underlay_qpn :
 				       0;
 		handler = _create_flow_rule(dev, ft_prio, flow_attr, dst,
 					    underlay_qpn, ucmd);
-	} else if (flow_attr->type == IB_FLOW_ATTR_ALL_DEFAULT ||
+		break;
-		   flow_attr->type == IB_FLOW_ATTR_MC_DEFAULT) {
+	case IB_FLOW_ATTR_ALL_DEFAULT:
-		handler = create_leftovers_rule(dev, ft_prio, flow_attr,
+	case IB_FLOW_ATTR_MC_DEFAULT:
-						dst);
+		handler = create_leftovers_rule(dev, ft_prio, flow_attr, dst);
-	} else if (flow_attr->type == IB_FLOW_ATTR_SNIFFER) {
+		break;
+	case IB_FLOW_ATTR_SNIFFER:
 		handler = create_sniffer_rule(dev, ft_prio, ft_prio_tx, dst);
-	} else {
+		break;
+	default:
 		err = -EINVAL;
 		goto destroy_ft;
 	}
@@ -1305,39 +1304,47 @@ _get_flow_table(struct mlx5_ib_dev *dev,
 
 	esw_encap = mlx5_eswitch_get_encap_mode(dev->mdev) !=
 		DEVLINK_ESWITCH_ENCAP_MODE_NONE;
-	if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_BYPASS) {
+	switch (fs_matcher->ns_type) {
-		max_table_size = BIT(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
+	case MLX5_FLOW_NAMESPACE_BYPASS:
-					log_max_ft_size));
+		max_table_size = BIT(
+			MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, log_max_ft_size));
 		if (MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, decap) && !esw_encap)
 			flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
 		if (MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev,
 					      reformat_l3_tunnel_to_l2) &&
 		    !esw_encap)
 			flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
-	} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS) {
+		break;
+	case MLX5_FLOW_NAMESPACE_EGRESS:
 		max_table_size = BIT(
 			MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, log_max_ft_size));
-		if (MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, reformat) && !esw_encap)
+		if (MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, reformat) &&
+		    !esw_encap)
 			flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
-	} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_FDB) {
+		break;
+	case MLX5_FLOW_NAMESPACE_FDB:
 		max_table_size = BIT(
 			MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, log_max_ft_size));
 		if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, decap) && esw_encap)
 			flags |= MLX5_FLOW_TABLE_TUNNEL_EN_DECAP;
-		if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev, reformat_l3_tunnel_to_l2) &&
+		if (MLX5_CAP_ESW_FLOWTABLE_FDB(dev->mdev,
+					       reformat_l3_tunnel_to_l2) &&
 		    esw_encap)
 			flags |= MLX5_FLOW_TABLE_TUNNEL_EN_REFORMAT;
 		priority = FDB_BYPASS_PATH;
-	} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_RDMA_RX) {
+		break;
-		max_table_size =
+	case MLX5_FLOW_NAMESPACE_RDMA_RX:
-			BIT(MLX5_CAP_FLOWTABLE_RDMA_RX(dev->mdev,
+		max_table_size = BIT(
-						       log_max_ft_size));
+			MLX5_CAP_FLOWTABLE_RDMA_RX(dev->mdev, log_max_ft_size));
 		priority = fs_matcher->priority;
-	} else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_RDMA_TX) {
+		break;
-		max_table_size =
+	case MLX5_FLOW_NAMESPACE_RDMA_TX:
-			BIT(MLX5_CAP_FLOWTABLE_RDMA_TX(dev->mdev,
+		max_table_size = BIT(
-						       log_max_ft_size));
+			MLX5_CAP_FLOWTABLE_RDMA_TX(dev->mdev, log_max_ft_size));
 		priority = fs_matcher->priority;
+		break;
+	default:
+		break;
 	}
 
 	max_table_size = min_t(int, max_table_size, MLX5_FS_MAX_ENTRIES);
@@ -1346,16 +1353,24 @@ _get_flow_table(struct mlx5_ib_dev *dev,
 	if (!ns)
 		return ERR_PTR(-EOPNOTSUPP);
 
-	if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_BYPASS)
+	switch (fs_matcher->ns_type) {
+	case MLX5_FLOW_NAMESPACE_BYPASS:
 		prio = &dev->flow_db->prios[priority];
-	else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_EGRESS)
+		break;
+	case MLX5_FLOW_NAMESPACE_EGRESS:
 		prio = &dev->flow_db->egress_prios[priority];
-	else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_FDB)
+		break;
+	case MLX5_FLOW_NAMESPACE_FDB:
 		prio = &dev->flow_db->fdb;
-	else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_RDMA_RX)
+		break;
+	case MLX5_FLOW_NAMESPACE_RDMA_RX:
 		prio = &dev->flow_db->rdma_rx[priority];
-	else if (fs_matcher->ns_type == MLX5_FLOW_NAMESPACE_RDMA_TX)
+		break;
+	case MLX5_FLOW_NAMESPACE_RDMA_TX:
 		prio = &dev->flow_db->rdma_tx[priority];
+		break;
+	default: return ERR_PTR(-EINVAL);
+	}
 
 	if (!prio)
 		return ERR_PTR(-EINVAL);
@@ -1488,20 +1503,25 @@ static struct mlx5_ib_flow_handler *raw_fs_rule_add(
 		goto unlock;
 	}
 
-	if (dest_type == MLX5_FLOW_DESTINATION_TYPE_TIR) {
+	switch (dest_type) {
+	case MLX5_FLOW_DESTINATION_TYPE_TIR:
 		dst[dst_num].type = dest_type;
 		dst[dst_num++].tir_num = dest_id;
 		flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
-	} else if (dest_type == MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE) {
+		break;
+	case MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE:
 		dst[dst_num].type = MLX5_FLOW_DESTINATION_TYPE_FLOW_TABLE_NUM;
 		dst[dst_num++].ft_num = dest_id;
 		flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_FWD_DEST;
-	} else  if (dest_type == MLX5_FLOW_DESTINATION_TYPE_PORT) {
+		break;
+	case MLX5_FLOW_DESTINATION_TYPE_PORT:
 		dst[dst_num++].type = MLX5_FLOW_DESTINATION_TYPE_PORT;
 		flow_act->action |= MLX5_FLOW_CONTEXT_ACTION_ALLOW;
+		break;
+	default:
+		break;
 	}
 
-
 	if (flow_act->action & MLX5_FLOW_CONTEXT_ACTION_COUNT) {
 		dst[dst_num].type = MLX5_FLOW_DESTINATION_TYPE_COUNTER;
 		dst[dst_num].counter_id = counter_id;

drivers/infiniband/hw/mlx5/gsi.c

@@ -35,44 +35,19 @@
 struct mlx5_ib_gsi_wr {
 	struct ib_cqe cqe;
 	struct ib_wc wc;
-	int send_flags;
 	bool completed:1;
 };
 
-struct mlx5_ib_gsi_qp {
-	struct ib_qp ibqp;
-	struct ib_qp *rx_qp;
-	u8 port_num;
-	struct ib_qp_cap cap;
-	enum ib_sig_type sq_sig_type;
-	/* Serialize qp state modifications */
-	struct mutex mutex;
-	struct ib_cq *cq;
-	struct mlx5_ib_gsi_wr *outstanding_wrs;
-	u32 outstanding_pi, outstanding_ci;
-	int num_qps;
-	/* Protects access to the tx_qps. Post send operations synchronize
-	 * with tx_qp creation in setup_qp(). Also protects the
-	 * outstanding_wrs array and indices.
-	 */
-	spinlock_t lock;
-	struct ib_qp **tx_qps;
-};
-
-static struct mlx5_ib_gsi_qp *gsi_qp(struct ib_qp *qp)
-{
-	return container_of(qp, struct mlx5_ib_gsi_qp, ibqp);
-}
-
 static bool mlx5_ib_deth_sqpn_cap(struct mlx5_ib_dev *dev)
 {
 	return MLX5_CAP_GEN(dev->mdev, set_deth_sqpn);
 }
 
 /* Call with gsi->lock locked */
-static void generate_completions(struct mlx5_ib_gsi_qp *gsi)
+static void generate_completions(struct mlx5_ib_qp *mqp)
 {
-	struct ib_cq *gsi_cq = gsi->ibqp.send_cq;
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
+	struct ib_cq *gsi_cq = mqp->ibqp.send_cq;
 	struct mlx5_ib_gsi_wr *wr;
 	u32 index;
 
@@ -83,10 +58,7 @@ static void generate_completions(struct mlx5_ib_gsi_qp *gsi)
 		if (!wr->completed)
 			break;
 
-		if (gsi->sq_sig_type == IB_SIGNAL_ALL_WR ||
+		WARN_ON_ONCE(mlx5_ib_generate_wc(gsi_cq, &wr->wc));
-		    wr->send_flags & IB_SEND_SIGNALED)
-			WARN_ON_ONCE(mlx5_ib_generate_wc(gsi_cq, &wr->wc));
-
 		wr->completed = false;
 	}
 
@@ -98,6 +70,7 @@ static void handle_single_completion(struct ib_cq *cq, struct ib_wc *wc)
 	struct mlx5_ib_gsi_qp *gsi = cq->cq_context;
 	struct mlx5_ib_gsi_wr *wr =
 		container_of(wc->wr_cqe, struct mlx5_ib_gsi_wr, cqe);
+	struct mlx5_ib_qp *mqp = container_of(gsi, struct mlx5_ib_qp, gsi);
 	u64 wr_id;
 	unsigned long flags;
 
@@ -106,19 +79,19 @@ static void handle_single_completion(struct ib_cq *cq, struct ib_wc *wc)
 	wr_id = wr->wc.wr_id;
 	wr->wc = *wc;
 	wr->wc.wr_id = wr_id;
-	wr->wc.qp = &gsi->ibqp;
+	wr->wc.qp = &mqp->ibqp;
 
-	generate_completions(gsi);
+	generate_completions(mqp);
 	spin_unlock_irqrestore(&gsi->lock, flags);
 }
 
-struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
+int mlx5_ib_create_gsi(struct ib_pd *pd, struct mlx5_ib_qp *mqp,
-				    struct ib_qp_init_attr *init_attr)
+		       struct ib_qp_init_attr *attr)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct mlx5_ib_gsi_qp *gsi;
-	struct ib_qp_init_attr hw_init_attr = *init_attr;
+	struct ib_qp_init_attr hw_init_attr = *attr;
-	const u8 port_num = init_attr->port_num;
+	const u8 port_num = attr->port_num;
 	int num_qps = 0;
 	int ret;
 
@@ -130,26 +103,19 @@ struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
 			num_qps = MLX5_MAX_PORTS;
 	}
 
-	gsi = kzalloc(sizeof(*gsi), GFP_KERNEL);
+	gsi = &mqp->gsi;
-	if (!gsi)
-		return ERR_PTR(-ENOMEM);
-
 	gsi->tx_qps = kcalloc(num_qps, sizeof(*gsi->tx_qps), GFP_KERNEL);
-	if (!gsi->tx_qps) {
+	if (!gsi->tx_qps)
-		ret = -ENOMEM;
+		return -ENOMEM;
-		goto err_free;
-	}
 
-	gsi->outstanding_wrs = kcalloc(init_attr->cap.max_send_wr,
+	gsi->outstanding_wrs =
-				       sizeof(*gsi->outstanding_wrs),
+		kcalloc(attr->cap.max_send_wr, sizeof(*gsi->outstanding_wrs),
-				       GFP_KERNEL);
+			GFP_KERNEL);
 	if (!gsi->outstanding_wrs) {
 		ret = -ENOMEM;
 		goto err_free_tx;
 	}
 
-	mutex_init(&gsi->mutex);
-
 	mutex_lock(&dev->devr.mutex);
 
 	if (dev->devr.ports[port_num - 1].gsi) {
@@ -161,12 +127,10 @@ struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
 	gsi->num_qps = num_qps;
 	spin_lock_init(&gsi->lock);
 
-	gsi->cap = init_attr->cap;
+	gsi->cap = attr->cap;
-	gsi->sq_sig_type = init_attr->sq_sig_type;
-	gsi->ibqp.qp_num = 1;
 	gsi->port_num = port_num;
 
-	gsi->cq = ib_alloc_cq(pd->device, gsi, init_attr->cap.max_send_wr, 0,
+	gsi->cq = ib_alloc_cq(pd->device, gsi, attr->cap.max_send_wr, 0,
 			      IB_POLL_SOFTIRQ);
 	if (IS_ERR(gsi->cq)) {
 		mlx5_ib_warn(dev, "unable to create send CQ for GSI QP. error %ld\n",
@@ -182,19 +146,31 @@ struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
 		hw_init_attr.cap.max_send_sge = 0;
 		hw_init_attr.cap.max_inline_data = 0;
 	}
-	gsi->rx_qp = ib_create_qp(pd, &hw_init_attr);
+
+	gsi->rx_qp = mlx5_ib_create_qp(pd, &hw_init_attr, NULL);
 	if (IS_ERR(gsi->rx_qp)) {
 		mlx5_ib_warn(dev, "unable to create hardware GSI QP. error %ld\n",
 			     PTR_ERR(gsi->rx_qp));
 		ret = PTR_ERR(gsi->rx_qp);
 		goto err_destroy_cq;
 	}
+	gsi->rx_qp->device = pd->device;
+	gsi->rx_qp->pd = pd;
+	gsi->rx_qp->real_qp = gsi->rx_qp;
 
-	dev->devr.ports[init_attr->port_num - 1].gsi = gsi;
+	gsi->rx_qp->qp_type = hw_init_attr.qp_type;
+	gsi->rx_qp->send_cq = hw_init_attr.send_cq;
+	gsi->rx_qp->recv_cq = hw_init_attr.recv_cq;
+	gsi->rx_qp->event_handler = hw_init_attr.event_handler;
+	spin_lock_init(&gsi->rx_qp->mr_lock);
+	INIT_LIST_HEAD(&gsi->rx_qp->rdma_mrs);
+	INIT_LIST_HEAD(&gsi->rx_qp->sig_mrs);
+
+	dev->devr.ports[attr->port_num - 1].gsi = gsi;
 
 	mutex_unlock(&dev->devr.mutex);
 
-	return &gsi->ibqp;
+	return 0;
 
 err_destroy_cq:
 	ib_free_cq(gsi->cq);
@@ -203,23 +179,19 @@ err_free_wrs:
 	kfree(gsi->outstanding_wrs);
 err_free_tx:
 	kfree(gsi->tx_qps);
-err_free:
+	return ret;
-	kfree(gsi);
-	return ERR_PTR(ret);
 }
 
-int mlx5_ib_gsi_destroy_qp(struct ib_qp *qp)
+int mlx5_ib_destroy_gsi(struct mlx5_ib_qp *mqp)
 {
-	struct mlx5_ib_dev *dev = to_mdev(qp->device);
+	struct mlx5_ib_dev *dev = to_mdev(mqp->ibqp.device);
-	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 	const int port_num = gsi->port_num;
 	int qp_index;
 	int ret;
 
-	mlx5_ib_dbg(dev, "destroying GSI QP\n");
-
 	mutex_lock(&dev->devr.mutex);
-	ret = ib_destroy_qp(gsi->rx_qp);
+	ret = mlx5_ib_destroy_qp(gsi->rx_qp, NULL);
 	if (ret) {
 		mlx5_ib_warn(dev, "unable to destroy hardware GSI QP. error %d\n",
 			     ret);
@@ -241,7 +213,7 @@ int mlx5_ib_gsi_destroy_qp(struct ib_qp *qp)
 
 	kfree(gsi->outstanding_wrs);
 	kfree(gsi->tx_qps);
-	kfree(gsi);
+	kfree(mqp);
 
 	return 0;
 }
@@ -259,7 +231,6 @@ static struct ib_qp *create_gsi_ud_qp(struct mlx5_ib_gsi_qp *gsi)
 			.max_send_sge = gsi->cap.max_send_sge,
 			.max_inline_data = gsi->cap.max_inline_data,
 		},
-		.sq_sig_type = gsi->sq_sig_type,
 		.qp_type = IB_QPT_UD,
 		.create_flags = MLX5_IB_QP_CREATE_SQPN_QP1,
 	};
@@ -370,56 +341,54 @@ err_destroy_qp:
 
 static void setup_qps(struct mlx5_ib_gsi_qp *gsi)
 {
+	struct mlx5_ib_dev *dev = to_mdev(gsi->rx_qp->device);
 	u16 qp_index;
 
+	mutex_lock(&dev->devr.mutex);
 	for (qp_index = 0; qp_index < gsi->num_qps; ++qp_index)
 		setup_qp(gsi, qp_index);
+	mutex_unlock(&dev->devr.mutex);
 }
 
 int mlx5_ib_gsi_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr,
 			  int attr_mask)
 {
 	struct mlx5_ib_dev *dev = to_mdev(qp->device);
-	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 	int ret;
 
 	mlx5_ib_dbg(dev, "modifying GSI QP to state %d\n", attr->qp_state);
 
-	mutex_lock(&gsi->mutex);
 	ret = ib_modify_qp(gsi->rx_qp, attr, attr_mask);
 	if (ret) {
 		mlx5_ib_warn(dev, "unable to modify GSI rx QP: %d\n", ret);
-		goto unlock;
+		return ret;
 	}
 
 	if (to_mqp(gsi->rx_qp)->state == IB_QPS_RTS)
 		setup_qps(gsi);
-
+	return 0;
-unlock:
-	mutex_unlock(&gsi->mutex);
-
-	return ret;
 }
 
 int mlx5_ib_gsi_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
 			 int qp_attr_mask,
 			 struct ib_qp_init_attr *qp_init_attr)
 {
-	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 	int ret;
 
-	mutex_lock(&gsi->mutex);
 	ret = ib_query_qp(gsi->rx_qp, qp_attr, qp_attr_mask, qp_init_attr);
 	qp_init_attr->cap = gsi->cap;
-	mutex_unlock(&gsi->mutex);
-
 	return ret;
 }
 
 /* Call with gsi->lock locked */
-static int mlx5_ib_add_outstanding_wr(struct mlx5_ib_gsi_qp *gsi,
+static int mlx5_ib_add_outstanding_wr(struct mlx5_ib_qp *mqp,
 				      struct ib_ud_wr *wr, struct ib_wc *wc)
 {
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 	struct mlx5_ib_dev *dev = to_mdev(gsi->rx_qp->device);
 	struct mlx5_ib_gsi_wr *gsi_wr;
 
@@ -448,22 +417,21 @@ static int mlx5_ib_add_outstanding_wr(struct mlx5_ib_gsi_qp *gsi,
 }
 
 /* Call with gsi->lock locked */
-static int mlx5_ib_gsi_silent_drop(struct mlx5_ib_gsi_qp *gsi,
+static int mlx5_ib_gsi_silent_drop(struct mlx5_ib_qp *mqp, struct ib_ud_wr *wr)
-				    struct ib_ud_wr *wr)
 {
 	struct ib_wc wc = {
 		{ .wr_id = wr->wr.wr_id },
 		.status = IB_WC_SUCCESS,
 		.opcode = IB_WC_SEND,
-		.qp = &gsi->ibqp,
+		.qp = &mqp->ibqp,
 	};
 	int ret;
 
-	ret = mlx5_ib_add_outstanding_wr(gsi, wr, &wc);
+	ret = mlx5_ib_add_outstanding_wr(mqp, wr, &wc);
 	if (ret)
 		return ret;
 
-	generate_completions(gsi);
+	generate_completions(mqp);
 
 	return 0;
 }
@@ -490,7 +458,8 @@ static struct ib_qp *get_tx_qp(struct mlx5_ib_gsi_qp *gsi, struct ib_ud_wr *wr)
 int mlx5_ib_gsi_post_send(struct ib_qp *qp, const struct ib_send_wr *wr,
 			  const struct ib_send_wr **bad_wr)
 {
-	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 	struct ib_qp *tx_qp;
 	unsigned long flags;
 	int ret;
@@ -503,14 +472,14 @@ int mlx5_ib_gsi_post_send(struct ib_qp *qp, const struct ib_send_wr *wr,
 		spin_lock_irqsave(&gsi->lock, flags);
 		tx_qp = get_tx_qp(gsi, &cur_wr);
 		if (!tx_qp) {
-			ret = mlx5_ib_gsi_silent_drop(gsi, &cur_wr);
+			ret = mlx5_ib_gsi_silent_drop(mqp, &cur_wr);
 			if (ret)
 				goto err;
 			spin_unlock_irqrestore(&gsi->lock, flags);
 			continue;
 		}
 
-		ret = mlx5_ib_add_outstanding_wr(gsi, &cur_wr, NULL);
+		ret = mlx5_ib_add_outstanding_wr(mqp, &cur_wr, NULL);
 		if (ret)
 			goto err;
 
@@ -534,7 +503,8 @@ err:
 int mlx5_ib_gsi_post_recv(struct ib_qp *qp, const struct ib_recv_wr *wr,
 			  const struct ib_recv_wr **bad_wr)
 {
-	struct mlx5_ib_gsi_qp *gsi = gsi_qp(qp);
+	struct mlx5_ib_qp *mqp = to_mqp(qp);
+	struct mlx5_ib_gsi_qp *gsi = &mqp->gsi;
 
 	return ib_post_recv(gsi->rx_qp, wr, bad_wr);
 }
@@ -544,7 +514,5 @@ void mlx5_ib_gsi_pkey_change(struct mlx5_ib_gsi_qp *gsi)
 	if (!gsi)
 		return;
 
-	mutex_lock(&gsi->mutex);
 	setup_qps(gsi);
-	mutex_unlock(&gsi->mutex);
 }

drivers/infiniband/hw/mlx5/main.c

@@ -326,8 +326,8 @@ out:
 	spin_unlock(&port->mp.mpi_lock);
 }
 
-static int translate_eth_legacy_proto_oper(u32 eth_proto_oper, u8 *active_speed,
+static int translate_eth_legacy_proto_oper(u32 eth_proto_oper,
-					   u8 *active_width)
+					   u16 *active_speed, u8 *active_width)
 {
 	switch (eth_proto_oper) {
 	case MLX5E_PROT_MASK(MLX5E_1000BASE_CX_SGMII):
@@ -384,7 +384,7 @@ static int translate_eth_legacy_proto_oper(u32 eth_proto_oper, u8 *active_speed,
 	return 0;
 }
 
-static int translate_eth_ext_proto_oper(u32 eth_proto_oper, u8 *active_speed,
+static int translate_eth_ext_proto_oper(u32 eth_proto_oper, u16 *active_speed,
 					u8 *active_width)
 {
 	switch (eth_proto_oper) {
@@ -436,7 +436,7 @@ static int translate_eth_ext_proto_oper(u32 eth_proto_oper, u8 *active_speed,
 	return 0;
 }
 
-static int translate_eth_proto_oper(u32 eth_proto_oper, u8 *active_speed,
+static int translate_eth_proto_oper(u32 eth_proto_oper, u16 *active_speed,
 				    u8 *active_width, bool ext)
 {
 	return ext ?
@@ -546,7 +546,7 @@ static int set_roce_addr(struct mlx5_ib_dev *dev, u8 port_num,
 			 unsigned int index, const union ib_gid *gid,
 			 const struct ib_gid_attr *attr)
 {
-	enum ib_gid_type gid_type = IB_GID_TYPE_IB;
+	enum ib_gid_type gid_type = IB_GID_TYPE_ROCE;
 	u16 vlan_id = 0xffff;
 	u8 roce_version = 0;
 	u8 roce_l3_type = 0;
@@ -561,7 +561,7 @@ static int set_roce_addr(struct mlx5_ib_dev *dev, u8 port_num,
 	}
 
 	switch (gid_type) {
-	case IB_GID_TYPE_IB:
+	case IB_GID_TYPE_ROCE:
 		roce_version = MLX5_ROCE_VERSION_1;
 		break;
 	case IB_GID_TYPE_ROCE_UDP_ENCAP:
@@ -840,7 +840,9 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 		/* We support 'Gappy' memory registration too */
 		props->device_cap_flags |= IB_DEVICE_SG_GAPS_REG;
 	}
-	props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
+	/* IB_WR_REG_MR always requires changing the entity size with UMR */
+	if (!MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled))
+		props->device_cap_flags |= IB_DEVICE_MEM_MGT_EXTENSIONS;
 	if (MLX5_CAP_GEN(mdev, sho)) {
 		props->device_cap_flags |= IB_DEVICE_INTEGRITY_HANDOVER;
 		/* At this stage no support for signature handover */
@@ -1175,32 +1177,24 @@ static int mlx5_ib_query_device(struct ib_device *ibdev,
 	return 0;
 }
 
-enum mlx5_ib_width {
+static void translate_active_width(struct ib_device *ibdev, u16 active_width,
-	MLX5_IB_WIDTH_1X	= 1 << 0,
+				   u8 *ib_width)
-	MLX5_IB_WIDTH_2X	= 1 << 1,
-	MLX5_IB_WIDTH_4X	= 1 << 2,
-	MLX5_IB_WIDTH_8X	= 1 << 3,
-	MLX5_IB_WIDTH_12X	= 1 << 4
-};
-
-static void translate_active_width(struct ib_device *ibdev, u8 active_width,
-				  u8 *ib_width)
 {
 	struct mlx5_ib_dev *dev = to_mdev(ibdev);
 
-	if (active_width & MLX5_IB_WIDTH_1X)
+	if (active_width & MLX5_PTYS_WIDTH_1X)
 		*ib_width = IB_WIDTH_1X;
-	else if (active_width & MLX5_IB_WIDTH_2X)
+	else if (active_width & MLX5_PTYS_WIDTH_2X)
 		*ib_width = IB_WIDTH_2X;
-	else if (active_width & MLX5_IB_WIDTH_4X)
+	else if (active_width & MLX5_PTYS_WIDTH_4X)
 		*ib_width = IB_WIDTH_4X;
-	else if (active_width & MLX5_IB_WIDTH_8X)
+	else if (active_width & MLX5_PTYS_WIDTH_8X)
 		*ib_width = IB_WIDTH_8X;
-	else if (active_width & MLX5_IB_WIDTH_12X)
+	else if (active_width & MLX5_PTYS_WIDTH_12X)
 		*ib_width = IB_WIDTH_12X;
 	else {
 		mlx5_ib_dbg(dev, "Invalid active_width %d, setting width to default value: 4x\n",
-			    (int)active_width);
+			    active_width);
 		*ib_width = IB_WIDTH_4X;
 	}
 
@@ -1277,7 +1271,7 @@ static int mlx5_query_hca_port(struct ib_device *ibdev, u8 port,
 	u16 max_mtu;
 	u16 oper_mtu;
 	int err;
-	u8 ib_link_width_oper;
+	u16 ib_link_width_oper;
 	u8 vl_hw_cap;
 
 	rep = kzalloc(sizeof(*rep), GFP_KERNEL);
@@ -1310,16 +1304,13 @@ static int mlx5_query_hca_port(struct ib_device *ibdev, u8 port,
 	if (props->port_cap_flags & IB_PORT_CAP_MASK2_SUP)
 		props->port_cap_flags2 = rep->cap_mask2;
 
-	err = mlx5_query_port_link_width_oper(mdev, &ib_link_width_oper, port);
+	err = mlx5_query_ib_port_oper(mdev, &ib_link_width_oper,
+				      &props->active_speed, port);
 	if (err)
 		goto out;
 
 	translate_active_width(ibdev, ib_link_width_oper, &props->active_width);
 
-	err = mlx5_query_port_ib_proto_oper(mdev, &props->active_speed, port);
-	if (err)
-		goto out;
-
 	mlx5_query_port_max_mtu(mdev, &max_mtu, port);
 
 	props->max_mtu = mlx5_mtu_to_ib_mtu(max_mtu);
@@ -2354,7 +2345,9 @@ static inline int check_dm_type_support(struct mlx5_ib_dev *dev,
 			return -EPERM;
 
 		if (!(MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, sw_owner) ||
-		      MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, sw_owner)))
+		      MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, sw_owner) ||
+		      MLX5_CAP_FLOWTABLE_NIC_RX(dev->mdev, sw_owner_v2) ||
+		      MLX5_CAP_FLOWTABLE_NIC_TX(dev->mdev, sw_owner_v2)))
 			return -EOPNOTSUPP;
 		break;
 	}
@@ -2569,12 +2562,12 @@ static int mlx5_ib_alloc_pd(struct ib_pd *ibpd, struct ib_udata *udata)
 	return 0;
 }
 
-static void mlx5_ib_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
+static int mlx5_ib_dealloc_pd(struct ib_pd *pd, struct ib_udata *udata)
 {
 	struct mlx5_ib_dev *mdev = to_mdev(pd->device);
 	struct mlx5_ib_pd *mpd = to_mpd(pd);
 
-	mlx5_cmd_dealloc_pd(mdev->mdev, mpd->pdn, mpd->uid);
+	return mlx5_cmd_dealloc_pd(mdev->mdev, mpd->pdn, mpd->uid);
 }
 
 static int mlx5_ib_mcg_attach(struct ib_qp *ibqp, union ib_gid *gid, u16 lid)
@@ -2699,9 +2692,7 @@ static void pkey_change_handler(struct work_struct *work)
 		container_of(work, struct mlx5_ib_port_resources,
 			     pkey_change_work);
 
-	mutex_lock(&ports->devr->mutex);
 	mlx5_ib_gsi_pkey_change(ports->gsi);
-	mutex_unlock(&ports->devr->mutex);
 }
 
 static void mlx5_ib_handle_internal_error(struct mlx5_ib_dev *ibdev)
@@ -3127,11 +3118,9 @@ static int mlx5_ib_dev_res_init(struct mlx5_ib_dev *dev)
 	atomic_inc(&devr->p0->usecnt);
 	atomic_set(&devr->s1->usecnt, 0);
 
-	for (port = 0; port < ARRAY_SIZE(devr->ports); ++port) {
+	for (port = 0; port < ARRAY_SIZE(devr->ports); ++port)
 		INIT_WORK(&devr->ports[port].pkey_change_work,
 			  pkey_change_handler);
-		devr->ports[port].devr = devr;
-	}
 
 	return 0;
 
@@ -4098,6 +4087,8 @@ static const struct ib_device_ops mlx5_ib_dev_sriov_ops = {
 static const struct ib_device_ops mlx5_ib_dev_mw_ops = {
 	.alloc_mw = mlx5_ib_alloc_mw,
 	.dealloc_mw = mlx5_ib_dealloc_mw,
+
+	INIT_RDMA_OBJ_SIZE(ib_mw, mlx5_ib_mw, ibmw),
 };
 
 static const struct ib_device_ops mlx5_ib_dev_xrc_ops = {
@@ -4268,6 +4259,9 @@ static const struct ib_device_ops mlx5_ib_dev_common_roce_ops = {
 	.destroy_wq = mlx5_ib_destroy_wq,
 	.get_netdev = mlx5_ib_get_netdev,
 	.modify_wq = mlx5_ib_modify_wq,
+
+	INIT_RDMA_OBJ_SIZE(ib_rwq_ind_table, mlx5_ib_rwq_ind_table,
+			   ib_rwq_ind_tbl),
 };
 
 static int mlx5_ib_roce_init(struct mlx5_ib_dev *dev)
@@ -4386,7 +4380,7 @@ static int mlx5_ib_stage_ib_reg_init(struct mlx5_ib_dev *dev)
 		name = "mlx5_%d";
 	else
 		name = "mlx5_bond_%d";
-	return ib_register_device(&dev->ib_dev, name);
+	return ib_register_device(&dev->ib_dev, name, &dev->mdev->pdev->dev);
 }
 
 static void mlx5_ib_stage_pre_ib_reg_umr_cleanup(struct mlx5_ib_dev *dev)

drivers/infiniband/hw/mlx5/mem.c

@@ -169,8 +169,8 @@ void mlx5_ib_populate_pas(struct mlx5_ib_dev *dev, struct ib_umem *umem,
 			  int page_shift, __be64 *pas, int access_flags)
 {
 	return __mlx5_ib_populate_pas(dev, umem, page_shift, 0,
-				      ib_umem_num_pages(umem), pas,
+				      ib_umem_num_dma_blocks(umem, PAGE_SIZE),
-				      access_flags);
+				      pas, access_flags);
 }
 int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset)
 {

drivers/infiniband/hw/mlx5/mlx5_ib.h

@@ -384,6 +384,22 @@ struct mlx5_ib_dct {
 	u32                     *in;
 };
 
+struct mlx5_ib_gsi_qp {
+	struct ib_qp *rx_qp;
+	u8 port_num;
+	struct ib_qp_cap cap;
+	struct ib_cq *cq;
+	struct mlx5_ib_gsi_wr *outstanding_wrs;
+	u32 outstanding_pi, outstanding_ci;
+	int num_qps;
+	/* Protects access to the tx_qps. Post send operations synchronize
+	 * with tx_qp creation in setup_qp(). Also protects the
+	 * outstanding_wrs array and indices.
+	 */
+	spinlock_t lock;
+	struct ib_qp **tx_qps;
+};
+
 struct mlx5_ib_qp {
 	struct ib_qp		ibqp;
 	union {
@@ -391,6 +407,7 @@ struct mlx5_ib_qp {
 		struct mlx5_ib_raw_packet_qp raw_packet_qp;
 		struct mlx5_ib_rss_qp rss_qp;
 		struct mlx5_ib_dct dct;
+		struct mlx5_ib_gsi_qp gsi;
 	};
 	struct mlx5_frag_buf	buf;
 
@@ -693,10 +710,7 @@ struct mlx5_mr_cache {
 	unsigned long		last_add;
 };
 
-struct mlx5_ib_gsi_qp;
-
 struct mlx5_ib_port_resources {
-	struct mlx5_ib_resources *devr;
 	struct mlx5_ib_gsi_qp *gsi;
 	struct work_struct pkey_change_work;
 };
@@ -1119,13 +1133,16 @@ void mlx5_ib_free_srq_wqe(struct mlx5_ib_srq *srq, int wqe_index);
 int mlx5_ib_create_ah(struct ib_ah *ah, struct rdma_ah_init_attr *init_attr,
 		      struct ib_udata *udata);
 int mlx5_ib_query_ah(struct ib_ah *ibah, struct rdma_ah_attr *ah_attr);
-void mlx5_ib_destroy_ah(struct ib_ah *ah, u32 flags);
+static inline int mlx5_ib_destroy_ah(struct ib_ah *ah, u32 flags)
+{
+	return 0;
+}
 int mlx5_ib_create_srq(struct ib_srq *srq, struct ib_srq_init_attr *init_attr,
 		       struct ib_udata *udata);
 int mlx5_ib_modify_srq(struct ib_srq *ibsrq, struct ib_srq_attr *attr,
 		       enum ib_srq_attr_mask attr_mask, struct ib_udata *udata);
 int mlx5_ib_query_srq(struct ib_srq *ibsrq, struct ib_srq_attr *srq_attr);
-void mlx5_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
+int mlx5_ib_destroy_srq(struct ib_srq *srq, struct ib_udata *udata);
 int mlx5_ib_post_srq_recv(struct ib_srq *ibsrq, const struct ib_recv_wr *wr,
 			  const struct ib_recv_wr **bad_wr);
 int mlx5_ib_enable_lb(struct mlx5_ib_dev *dev, bool td, bool qp);
@@ -1148,7 +1165,7 @@ int mlx5_ib_read_wqe_srq(struct mlx5_ib_srq *srq, int wqe_index, void *buffer,
 			 size_t buflen, size_t *bc);
 int mlx5_ib_create_cq(struct ib_cq *ibcq, const struct ib_cq_init_attr *attr,
 		      struct ib_udata *udata);
-void mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
+int mlx5_ib_destroy_cq(struct ib_cq *cq, struct ib_udata *udata);
 int mlx5_ib_poll_cq(struct ib_cq *ibcq, int num_entries, struct ib_wc *wc);
 int mlx5_ib_arm_cq(struct ib_cq *ibcq, enum ib_cq_notify_flags flags);
 int mlx5_ib_modify_cq(struct ib_cq *cq, u16 cq_count, u16 cq_period);
@@ -1163,8 +1180,7 @@ int mlx5_ib_advise_mr(struct ib_pd *pd,
 		      struct ib_sge *sg_list,
 		      u32 num_sge,
 		      struct uverbs_attr_bundle *attrs);
-struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int mlx5_ib_alloc_mw(struct ib_mw *mw, struct ib_udata *udata);
-			       struct ib_udata *udata);
 int mlx5_ib_dealloc_mw(struct ib_mw *mw);
 int mlx5_ib_update_xlt(struct mlx5_ib_mr *mr, u64 idx, int npages,
 		       int page_shift, int flags);
@@ -1193,7 +1209,7 @@ int mlx5_ib_process_mad(struct ib_device *ibdev, int mad_flags, u8 port_num,
 			const struct ib_mad *in, struct ib_mad *out,
 			size_t *out_mad_size, u16 *out_mad_pkey_index);
 int mlx5_ib_alloc_xrcd(struct ib_xrcd *xrcd, struct ib_udata *udata);
-void mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd, struct ib_udata *udata);
+int mlx5_ib_dealloc_xrcd(struct ib_xrcd *xrcd, struct ib_udata *udata);
 int mlx5_ib_get_buf_offset(u64 addr, int page_shift, u32 *offset);
 int mlx5_query_ext_port_caps(struct mlx5_ib_dev *dev, u8 port);
 int mlx5_query_mad_ifc_smp_attr_node_info(struct ib_device *ibdev,
@@ -1229,7 +1245,7 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev);
 int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev);
 
 struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev,
-				       unsigned int entry);
+				       unsigned int entry, int access_flags);
 void mlx5_mr_cache_free(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr);
 int mlx5_mr_cache_invalidate(struct mlx5_ib_mr *mr);
 
@@ -1238,12 +1254,12 @@ int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,
 struct ib_wq *mlx5_ib_create_wq(struct ib_pd *pd,
 				struct ib_wq_init_attr *init_attr,
 				struct ib_udata *udata);
-void mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
+int mlx5_ib_destroy_wq(struct ib_wq *wq, struct ib_udata *udata);
 int mlx5_ib_modify_wq(struct ib_wq *wq, struct ib_wq_attr *wq_attr,
 		      u32 wq_attr_mask, struct ib_udata *udata);
-struct ib_rwq_ind_table *mlx5_ib_create_rwq_ind_table(struct ib_device *device,
+int mlx5_ib_create_rwq_ind_table(struct ib_rwq_ind_table *ib_rwq_ind_table,
-						      struct ib_rwq_ind_table_init_attr *init_attr,
+				 struct ib_rwq_ind_table_init_attr *init_attr,
-						      struct ib_udata *udata);
+				 struct ib_udata *udata);
 int mlx5_ib_destroy_rwq_ind_table(struct ib_rwq_ind_table *wq_ind_table);
 struct ib_dm *mlx5_ib_alloc_dm(struct ib_device *ibdev,
 			       struct ib_ucontext *context,
@@ -1267,6 +1283,7 @@ void mlx5_odp_populate_xlt(void *xlt, size_t idx, size_t nentries,
 int mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
 			       enum ib_uverbs_advise_mr_advice advice,
 			       u32 flags, struct ib_sge *sg_list, u32 num_sge);
+int mlx5_ib_init_odp_mr(struct mlx5_ib_mr *mr, bool enable);
 #else /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
 static inline void mlx5_ib_internal_fill_odp_caps(struct mlx5_ib_dev *dev)
 {
@@ -1288,6 +1305,10 @@ mlx5_ib_advise_mr_prefetch(struct ib_pd *pd,
 {
 	return -EOPNOTSUPP;
 }
+static inline int mlx5_ib_init_odp_mr(struct mlx5_ib_mr *mr, bool enable)
+{
+	return -EOPNOTSUPP;
+}
 #endif /* CONFIG_INFINIBAND_ON_DEMAND_PAGING */
 
 extern const struct mmu_interval_notifier_ops mlx5_mn_ops;
@@ -1318,9 +1339,9 @@ void mlx5_ib_cleanup_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num);
 void mlx5_ib_init_cong_debugfs(struct mlx5_ib_dev *dev, u8 port_num);
 
 /* GSI QP helper functions */
-struct ib_qp *mlx5_ib_gsi_create_qp(struct ib_pd *pd,
+int mlx5_ib_create_gsi(struct ib_pd *pd, struct mlx5_ib_qp *mqp,
-				    struct ib_qp_init_attr *init_attr);
+		       struct ib_qp_init_attr *attr);
-int mlx5_ib_gsi_destroy_qp(struct ib_qp *qp);
+int mlx5_ib_destroy_gsi(struct mlx5_ib_qp *mqp);
 int mlx5_ib_gsi_modify_qp(struct ib_qp *qp, struct ib_qp_attr *attr,
 			  int attr_mask);
 int mlx5_ib_gsi_query_qp(struct ib_qp *qp, struct ib_qp_attr *qp_attr,
@@ -1358,7 +1379,7 @@ static inline void init_query_mad(struct ib_smp *mad)
 
 static inline int is_qp1(enum ib_qp_type qp_type)
 {
-	return qp_type == MLX5_IB_QPT_HW_GSI;
+	return qp_type == MLX5_IB_QPT_HW_GSI || qp_type == IB_QPT_GSI;
 }
 
 #define MLX5_MAX_UMR_SHIFT 16
@@ -1442,25 +1463,54 @@ int bfregn_to_uar_index(struct mlx5_ib_dev *dev,
 			struct mlx5_bfreg_info *bfregi, u32 bfregn,
 			bool dyn_bfreg);
 
-static inline bool mlx5_ib_can_use_umr(struct mlx5_ib_dev *dev,
+static inline bool mlx5_ib_can_load_pas_with_umr(struct mlx5_ib_dev *dev,
-				       bool do_modify_atomic, int access_flags)
+						 size_t length)
 {
+	/*
+	 * umr_check_mkey_mask() rejects MLX5_MKEY_MASK_PAGE_SIZE which is
+	 * always set if MLX5_IB_SEND_UMR_UPDATE_TRANSLATION (aka
+	 * MLX5_IB_UPD_XLT_ADDR and MLX5_IB_UPD_XLT_ENABLE) is set. Thus, a mkey
+	 * can never be enabled without this capability. Simplify this weird
+	 * quirky hardware by just saying it can't use PAS lists with UMR at
+	 * all.
+	 */
 	if (MLX5_CAP_GEN(dev->mdev, umr_modify_entity_size_disabled))
 		return false;
 
-	if (do_modify_atomic &&
+	/*
+	 * length is the size of the MR in bytes when mlx5_ib_update_xlt() is
+	 * used.
+	 */
+	if (!MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset) &&
+	    length >= MLX5_MAX_UMR_PAGES * PAGE_SIZE)
+		return false;
+	return true;
+}
+
+/*
+ * true if an existing MR can be reconfigured to new access_flags using UMR.
+ * Older HW cannot use UMR to update certain elements of the MKC. See
+ * umr_check_mkey_mask(), get_umr_update_access_mask() and umr_check_mkey_mask()
+ */
+static inline bool mlx5_ib_can_reconfig_with_umr(struct mlx5_ib_dev *dev,
+						 unsigned int current_access_flags,
+						 unsigned int target_access_flags)
+{
+	unsigned int diffs = current_access_flags ^ target_access_flags;
+
+	if ((diffs & IB_ACCESS_REMOTE_ATOMIC) &&
 	    MLX5_CAP_GEN(dev->mdev, atomic) &&
 	    MLX5_CAP_GEN(dev->mdev, umr_modify_atomic_disabled))
 		return false;
 
-	if (access_flags & IB_ACCESS_RELAXED_ORDERING &&
+	if ((diffs & IB_ACCESS_RELAXED_ORDERING) &&
 	    MLX5_CAP_GEN(dev->mdev, relaxed_ordering_write) &&
 	    !MLX5_CAP_GEN(dev->mdev, relaxed_ordering_write_umr))
 		return false;
 
-	if (access_flags & IB_ACCESS_RELAXED_ORDERING &&
+	if ((diffs & IB_ACCESS_RELAXED_ORDERING) &&
-	     MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read) &&
+	    MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read) &&
-	     !MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read_umr))
+	    !MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read_umr))
 		return false;
 
 	return true;

drivers/infiniband/hw/mlx5/mr.c

@@ -50,6 +50,29 @@ enum {
 static void
 create_mkey_callback(int status, struct mlx5_async_work *context);
 
+static void set_mkc_access_pd_addr_fields(void *mkc, int acc, u64 start_addr,
+					  struct ib_pd *pd)
+{
+	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+
+	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
+	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
+	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
+	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
+	MLX5_SET(mkc, mkc, lr, 1);
+
+	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_write))
+		MLX5_SET(mkc, mkc, relaxed_ordering_write,
+			 !!(acc & IB_ACCESS_RELAXED_ORDERING));
+	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read))
+		MLX5_SET(mkc, mkc, relaxed_ordering_read,
+			 !!(acc & IB_ACCESS_RELAXED_ORDERING));
+
+	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, qpn, 0xffffff);
+	MLX5_SET64(mkc, mkc, start_addr, start_addr);
+}
+
 static void
 assign_mkey_variant(struct mlx5_ib_dev *dev, struct mlx5_core_mkey *mkey,
 		    u32 *in)
@@ -100,7 +123,8 @@ static int destroy_mkey(struct mlx5_ib_dev *dev, struct mlx5_ib_mr *mr)
 	return mlx5_core_destroy_mkey(dev->mdev, &mr->mmkey);
 }
 
-static bool use_umr_mtt_update(struct mlx5_ib_mr *mr, u64 start, u64 length)
+static inline bool mlx5_ib_pas_fits_in_mr(struct mlx5_ib_mr *mr, u64 start,
+					  u64 length)
 {
 	return ((u64)1 << mr->order) * MLX5_ADAPTER_PAGE_SIZE >=
 		length + (start & (MLX5_ADAPTER_PAGE_SIZE - 1));
@@ -152,12 +176,12 @@ static struct mlx5_ib_mr *alloc_cache_mr(struct mlx5_cache_ent *ent, void *mkc)
 	mr->cache_ent = ent;
 	mr->dev = ent->dev;
 
+	set_mkc_access_pd_addr_fields(mkc, 0, 0, ent->dev->umrc.pd);
 	MLX5_SET(mkc, mkc, free, 1);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 	MLX5_SET(mkc, mkc, access_mode_1_0, ent->access_mode & 0x3);
 	MLX5_SET(mkc, mkc, access_mode_4_2, (ent->access_mode >> 2) & 0x7);
 
-	MLX5_SET(mkc, mkc, qpn, 0xffffff);
 	MLX5_SET(mkc, mkc, translations_octword_size, ent->xlt);
 	MLX5_SET(mkc, mkc, log_page_size, ent->page);
 	return mr;
@@ -534,7 +558,7 @@ static void cache_work_func(struct work_struct *work)
 
 /* Allocate a special entry from the cache */
 struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev,
-				       unsigned int entry)
+				       unsigned int entry, int access_flags)
 {
 	struct mlx5_mr_cache *cache = &dev->cache;
 	struct mlx5_cache_ent *ent;
@@ -544,6 +568,10 @@ struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev,
 		    entry >= ARRAY_SIZE(cache->ent)))
 		return ERR_PTR(-EINVAL);
 
+	/* Matches access in alloc_cache_mr() */
+	if (!mlx5_ib_can_reconfig_with_umr(dev, 0, access_flags))
+		return ERR_PTR(-EOPNOTSUPP);
+
 	ent = &cache->ent[entry];
 	spin_lock_irq(&ent->lock);
 	if (list_empty(&ent->head)) {
@@ -558,6 +586,7 @@ struct mlx5_ib_mr *mlx5_mr_cache_alloc(struct mlx5_ib_dev *dev,
 		queue_adjust_cache_locked(ent);
 		spin_unlock_irq(&ent->lock);
 	}
+	mr->access_flags = access_flags;
 	return mr;
 }
 
@@ -730,8 +759,8 @@ int mlx5_mr_cache_init(struct mlx5_ib_dev *dev)
 			   MLX5_IB_UMR_OCTOWORD;
 		ent->access_mode = MLX5_MKC_ACCESS_MODE_MTT;
 		if ((dev->mdev->profile->mask & MLX5_PROF_MASK_MR_CACHE) &&
-		    !dev->is_rep &&
+		    !dev->is_rep && mlx5_core_is_pf(dev->mdev) &&
-		    mlx5_core_is_pf(dev->mdev))
+		    mlx5_ib_can_load_pas_with_umr(dev, 0))
 			ent->limit = dev->mdev->profile->mr_cache[i].limit;
 		else
 			ent->limit = 0;
@@ -774,29 +803,6 @@ int mlx5_mr_cache_cleanup(struct mlx5_ib_dev *dev)
 	return 0;
 }
 
-static void set_mkc_access_pd_addr_fields(void *mkc, int acc, u64 start_addr,
-					  struct ib_pd *pd)
-{
-	struct mlx5_ib_dev *dev = to_mdev(pd->device);
-
-	MLX5_SET(mkc, mkc, a, !!(acc & IB_ACCESS_REMOTE_ATOMIC));
-	MLX5_SET(mkc, mkc, rw, !!(acc & IB_ACCESS_REMOTE_WRITE));
-	MLX5_SET(mkc, mkc, rr, !!(acc & IB_ACCESS_REMOTE_READ));
-	MLX5_SET(mkc, mkc, lw, !!(acc & IB_ACCESS_LOCAL_WRITE));
-	MLX5_SET(mkc, mkc, lr, 1);
-
-	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_write))
-		MLX5_SET(mkc, mkc, relaxed_ordering_write,
-			 !!(acc & IB_ACCESS_RELAXED_ORDERING));
-	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read))
-		MLX5_SET(mkc, mkc, relaxed_ordering_read,
-			 !!(acc & IB_ACCESS_RELAXED_ORDERING));
-
-	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
-	MLX5_SET(mkc, mkc, qpn, 0xffffff);
-	MLX5_SET64(mkc, mkc, start_addr, start_addr);
-}
-
 struct ib_mr *mlx5_ib_get_dma_mr(struct ib_pd *pd, int acc)
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
@@ -979,6 +985,11 @@ alloc_mr_from_cache(struct ib_pd *pd, struct ib_umem *umem, u64 virt_addr,
 
 	if (!ent)
 		return ERR_PTR(-E2BIG);
+
+	/* Matches access in alloc_cache_mr() */
+	if (!mlx5_ib_can_reconfig_with_umr(dev, 0, access_flags))
+		return ERR_PTR(-EOPNOTSUPP);
+
 	mr = get_cache_mr(ent);
 	if (!mr) {
 		mr = create_cache_mr(ent);
@@ -1181,38 +1192,31 @@ static struct mlx5_ib_mr *reg_create(struct ib_mr *ibmr, struct ib_pd *pd,
 		goto err_1;
 	}
 	pas = (__be64 *)MLX5_ADDR_OF(create_mkey_in, in, klm_pas_mtt);
-	if (populate && !(access_flags & IB_ACCESS_ON_DEMAND))
+	if (populate) {
+		if (WARN_ON(access_flags & IB_ACCESS_ON_DEMAND)) {
+			err = -EINVAL;
+			goto err_2;
+		}
 		mlx5_ib_populate_pas(dev, umem, page_shift, pas,
 				     pg_cap ? MLX5_IB_MTT_PRESENT : 0);
+	}
 
 	/* The pg_access bit allows setting the access flags
 	 * in the page list submitted with the command. */
 	MLX5_SET(create_mkey_in, in, pg_access, !!(pg_cap));
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
+	set_mkc_access_pd_addr_fields(mkc, access_flags, virt_addr,
+				      populate ? pd : dev->umrc.pd);
 	MLX5_SET(mkc, mkc, free, !populate);
 	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_MTT);
-	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_write))
-		MLX5_SET(mkc, mkc, relaxed_ordering_write,
-			 !!(access_flags & IB_ACCESS_RELAXED_ORDERING));
-	if (MLX5_CAP_GEN(dev->mdev, relaxed_ordering_read))
-		MLX5_SET(mkc, mkc, relaxed_ordering_read,
-			 !!(access_flags & IB_ACCESS_RELAXED_ORDERING));
-	MLX5_SET(mkc, mkc, a, !!(access_flags & IB_ACCESS_REMOTE_ATOMIC));
-	MLX5_SET(mkc, mkc, rw, !!(access_flags & IB_ACCESS_REMOTE_WRITE));
-	MLX5_SET(mkc, mkc, rr, !!(access_flags & IB_ACCESS_REMOTE_READ));
-	MLX5_SET(mkc, mkc, lw, !!(access_flags & IB_ACCESS_LOCAL_WRITE));
-	MLX5_SET(mkc, mkc, lr, 1);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 
-	MLX5_SET64(mkc, mkc, start_addr, virt_addr);
 	MLX5_SET64(mkc, mkc, len, length);
-	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
 	MLX5_SET(mkc, mkc, bsf_octword_size, 0);
 	MLX5_SET(mkc, mkc, translations_octword_size,
 		 get_octo_len(virt_addr, length, page_shift));
 	MLX5_SET(mkc, mkc, log_page_size, page_shift);
-	MLX5_SET(mkc, mkc, qpn, 0xffffff);
 	if (populate) {
 		MLX5_SET(create_mkey_in, in, translations_octword_actual_size,
 			 get_octo_len(virt_addr, length, page_shift));
@@ -1308,7 +1312,8 @@ int mlx5_ib_advise_mr(struct ib_pd *pd,
 		      struct uverbs_attr_bundle *attrs)
 {
 	if (advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH &&
-	    advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE)
+	    advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_WRITE &&
+	    advice != IB_UVERBS_ADVISE_MR_ADVICE_PREFETCH_NO_FAULT)
 		return -EOPNOTSUPP;
 
 	return mlx5_ib_advise_mr_prefetch(pd, advice, flags,
@@ -1353,7 +1358,7 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 {
 	struct mlx5_ib_dev *dev = to_mdev(pd->device);
 	struct mlx5_ib_mr *mr = NULL;
-	bool use_umr;
+	bool xlt_with_umr;
 	struct ib_umem *umem;
 	int page_shift;
 	int npages;
@@ -1367,6 +1372,11 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	mlx5_ib_dbg(dev, "start 0x%llx, virt_addr 0x%llx, length 0x%llx, access_flags 0x%x\n",
 		    start, virt_addr, length, access_flags);
 
+	xlt_with_umr = mlx5_ib_can_load_pas_with_umr(dev, length);
+	/* ODP requires xlt update via umr to work. */
+	if (!xlt_with_umr && (access_flags & IB_ACCESS_ON_DEMAND))
+		return ERR_PTR(-EINVAL);
+
 	if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING) && !start &&
 	    length == U64_MAX) {
 		if (virt_addr != start)
@@ -1387,28 +1397,17 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	if (err < 0)
 		return ERR_PTR(err);
 
-	use_umr = mlx5_ib_can_use_umr(dev, true, access_flags);
+	if (xlt_with_umr) {
-
-	if (order <= mr_cache_max_order(dev) && use_umr) {
 		mr = alloc_mr_from_cache(pd, umem, virt_addr, length, ncont,
 					 page_shift, order, access_flags);
-		if (PTR_ERR(mr) == -EAGAIN) {
+		if (IS_ERR(mr))
-			mlx5_ib_dbg(dev, "cache empty for order %d\n", order);
 			mr = NULL;
-		}
-	} else if (!MLX5_CAP_GEN(dev->mdev, umr_extended_translation_offset)) {
-		if (access_flags & IB_ACCESS_ON_DEMAND) {
-			err = -EINVAL;
-			pr_err("Got MR registration for ODP MR > 512MB, not supported for Connect-IB\n");
-			goto error;
-		}
-		use_umr = false;
 	}
 
 	if (!mr) {
 		mutex_lock(&dev->slow_path_mutex);
 		mr = reg_create(NULL, pd, virt_addr, length, umem, ncont,
-				page_shift, access_flags, !use_umr);
+				page_shift, access_flags, !xlt_with_umr);
 		mutex_unlock(&dev->slow_path_mutex);
 	}
 
@@ -1422,15 +1421,16 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 	mr->umem = umem;
 	set_mr_fields(dev, mr, npages, length, access_flags);
 
-	if (use_umr) {
+	if (xlt_with_umr && !(access_flags & IB_ACCESS_ON_DEMAND)) {
+		/*
+		 * If the MR was created with reg_create then it will be
+		 * configured properly but left disabled. It is safe to go ahead
+		 * and configure it again via UMR while enabling it.
+		 */
 		int update_xlt_flags = MLX5_IB_UPD_XLT_ENABLE;
 
-		if (access_flags & IB_ACCESS_ON_DEMAND)
-			update_xlt_flags |= MLX5_IB_UPD_XLT_ZAP;
-
 		err = mlx5_ib_update_xlt(mr, 0, ncont, page_shift,
 					 update_xlt_flags);
-
 		if (err) {
 			dereg_mr(dev, mr);
 			return ERR_PTR(err);
@@ -1448,6 +1448,12 @@ struct ib_mr *mlx5_ib_reg_user_mr(struct ib_pd *pd, u64 start, u64 length,
 			dereg_mr(dev, mr);
 			return ERR_PTR(err);
 		}
+
+		err = mlx5_ib_init_odp_mr(mr, xlt_with_umr);
+		if (err) {
+			dereg_mr(dev, mr);
+			return ERR_PTR(err);
+		}
 	}
 
 	return &mr->ibmr;
@@ -1555,8 +1561,11 @@ int mlx5_ib_rereg_user_mr(struct ib_mr *ib_mr, int flags, u64 start,
 			goto err;
 	}
 
-	if (!mlx5_ib_can_use_umr(dev, true, access_flags) ||
+	if (!mlx5_ib_can_reconfig_with_umr(dev, mr->access_flags,
-	    (flags & IB_MR_REREG_TRANS && !use_umr_mtt_update(mr, addr, len))) {
+					   access_flags) ||
+	    !mlx5_ib_can_load_pas_with_umr(dev, len) ||
+	    (flags & IB_MR_REREG_TRANS &&
+	     !mlx5_ib_pas_fits_in_mr(mr, addr, len))) {
 		/*
 		 * UMR can't be used - MKey needs to be replaced.
 		 */
@@ -1727,9 +1736,9 @@ static void mlx5_set_umr_free_mkey(struct ib_pd *pd, u32 *in, int ndescs,
 
 	mkc = MLX5_ADDR_OF(create_mkey_in, in, memory_key_mkey_entry);
 
+	/* This is only used from the kernel, so setting the PD is OK. */
+	set_mkc_access_pd_addr_fields(mkc, 0, 0, pd);
 	MLX5_SET(mkc, mkc, free, 1);
-	MLX5_SET(mkc, mkc, qpn, 0xffffff);
-	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
 	MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
 	MLX5_SET(mkc, mkc, access_mode_1_0, access_mode & 0x3);
 	MLX5_SET(mkc, mkc, access_mode_4_2, (access_mode >> 2) & 0x7);
@@ -1973,12 +1982,11 @@ struct ib_mr *mlx5_ib_alloc_mr_integrity(struct ib_pd *pd,
 				  max_num_meta_sg);
 }
 
-struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
+int mlx5_ib_alloc_mw(struct ib_mw *ibmw, struct ib_udata *udata)
-			       struct ib_udata *udata)
 {
-	struct mlx5_ib_dev *dev = to_mdev(pd->device);
+	struct mlx5_ib_dev *dev = to_mdev(ibmw->device);
 	int inlen = MLX5_ST_SZ_BYTES(create_mkey_in);
-	struct mlx5_ib_mw *mw = NULL;
+	struct mlx5_ib_mw *mw = to_mmw(ibmw);
 	u32 *in = NULL;
 	void *mkc;
 	int ndescs;
@@ -1991,21 +1999,20 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 
 	err = ib_copy_from_udata(&req, udata, min(udata->inlen, sizeof(req)));
 	if (err)
-		return ERR_PTR(err);
+		return err;
 
 	if (req.comp_mask || req.reserved1 || req.reserved2)
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	if (udata->inlen > sizeof(req) &&
 	    !ib_is_udata_cleared(udata, sizeof(req),
 				 udata->inlen - sizeof(req)))
-		return ERR_PTR(-EOPNOTSUPP);
+		return -EOPNOTSUPP;
 
 	ndescs = req.num_klms ? roundup(req.num_klms, 4) : roundup(1, 4);
 
-	mw = kzalloc(sizeof(*mw), GFP_KERNEL);
 	in = kzalloc(inlen, GFP_KERNEL);
-	if (!mw || !in) {
+	if (!in) {
 		err = -ENOMEM;
 		goto free;
 	}
@@ -2014,11 +2021,11 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 
 	MLX5_SET(mkc, mkc, free, 1);
 	MLX5_SET(mkc, mkc, translations_octword_size, ndescs);
-	MLX5_SET(mkc, mkc, pd, to_mpd(pd)->pdn);
+	MLX5_SET(mkc, mkc, pd, to_mpd(ibmw->pd)->pdn);
 	MLX5_SET(mkc, mkc, umr_en, 1);
 	MLX5_SET(mkc, mkc, lr, 1);
 	MLX5_SET(mkc, mkc, access_mode_1_0, MLX5_MKC_ACCESS_MODE_KLMS);
-	MLX5_SET(mkc, mkc, en_rinval, !!((type == IB_MW_TYPE_2)));
+	MLX5_SET(mkc, mkc, en_rinval, !!((ibmw->type == IB_MW_TYPE_2)));
 	MLX5_SET(mkc, mkc, qpn, 0xffffff);
 
 	err = mlx5_ib_create_mkey(dev, &mw->mmkey, in, inlen);
@@ -2026,17 +2033,15 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 		goto free;
 
 	mw->mmkey.type = MLX5_MKEY_MW;
-	mw->ibmw.rkey = mw->mmkey.key;
+	ibmw->rkey = mw->mmkey.key;
 	mw->ndescs = ndescs;
 
-	resp.response_length = min(offsetof(typeof(resp), response_length) +
+	resp.response_length =
-				   sizeof(resp.response_length), udata->outlen);
+		min(offsetofend(typeof(resp), response_length), udata->outlen);
 	if (resp.response_length) {
 		err = ib_copy_to_udata(udata, &resp, resp.response_length);
-		if (err) {
+		if (err)
-			mlx5_core_destroy_mkey(dev->mdev, &mw->mmkey);
+			goto free_mkey;
-			goto free;
-		}
 	}
 
 	if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) {
@@ -2048,21 +2053,19 @@ struct ib_mw *mlx5_ib_alloc_mw(struct ib_pd *pd, enum ib_mw_type type,
 	}
 
 	kfree(in);
-	return &mw->ibmw;
+	return 0;
 
 free_mkey:
 	mlx5_core_destroy_mkey(dev->mdev, &mw->mmkey);
 free:
-	kfree(mw);
 	kfree(in);
-	return ERR_PTR(err);
+	return err;
 }
 
 int mlx5_ib_dealloc_mw(struct ib_mw *mw)
 {
 	struct mlx5_ib_dev *dev = to_mdev(mw->device);
 	struct mlx5_ib_mw *mmw = to_mmw(mw);
-	int err;
 
 	if (IS_ENABLED(CONFIG_INFINIBAND_ON_DEMAND_PAGING)) {
 		xa_erase(&dev->odp_mkeys, mlx5_base_mkey(mmw->mmkey.key));
@@ -2073,11 +2076,7 @@ int mlx5_ib_dealloc_mw(struct ib_mw *mw)
 		synchronize_srcu(&dev->odp_srcu);
 	}
 
-	err = mlx5_core_destroy_mkey(dev->mdev, &mmw->mmkey);
+	return mlx5_core_destroy_mkey(dev->mdev, &mmw->mmkey);
-	if (err)
-		return err;
-	kfree(mmw);
-	return 0;
 }
 
 int mlx5_ib_check_mr_status(struct ib_mr *ibmr, u32 check_mask,