IB/{hfi1, rdmavt, qib}: Implement CQ completion vector support

Currently the driver doesn't support completion vectors. These
are used to indicate which sets of CQs should be grouped together
into the same vector. A vector is a CQ processing thread that
runs on a specific CPU.

If an application has several CQs bound to different completion
vectors, and each completion vector runs on different CPUs, then
the completion queue workload is balanced. This helps scale as more
nodes are used.

Implement CQ completion vector support using a global workqueue
where a CQ entry is queued to the CPU corresponding to the CQ's
completion vector. Since the workqueue is global, it's guaranteed
to always be there when queueing CQ entries; Therefore, the RCU
locking for cq->rdi->worker in the hot path is superfluous.

Each completion vector is assigned to a different CPU. The number of
completion vectors available is computed by taking the number of
online, physical CPUs from the local NUMA node and subtracting the
CPUs used for kernel receive queues and the general interrupt.
Special use cases:

  * If there are no CPUs left for completion vectors, the same CPU
    for the general interrupt is used; Therefore, there would only
    be one completion vector available.

  * For multi-HFI systems, the number of completion vectors available
    for each device is the total number of completion vectors in
    the local NUMA node divided by the number of devices in the same
    NUMA node. If there's a division remainder, the first device to
    get initialized gets an extra completion vector.

Upon a CQ creation, an invalid completion vector could be specified.
Handle it as follows:

  * If the completion vector is less than 0, set it to 0.

  * Set the completion vector to the result of the passed completion
    vector moded with the number of device completion vectors
    available.

Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>

9 files changed, 445 insertions, 21 deletions

diff --git a/drivers/infiniband/hw/hfi1/affinity.c b/drivers/infiniband/hw/hfi1/affinity.c
index eca9e6354017..fbe7198a715a 100644
--- a/drivers/infiniband/hw/hfi1/affinity.c
+++ b/drivers/infiniband/hw/hfi1/affinity.c
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015 - 2017 Intel Corporation.
+ * Copyright(c) 2015 - 2018 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -208,7 +208,13 @@ int node_affinity_init(void)
 	return 0;
 }
 
-void node_affinity_destroy(void)
+static void node_affinity_destroy(struct hfi1_affinity_node *entry)
+{
+	free_percpu(entry->comp_vect_affinity);
+	kfree(entry);
+}
+
+void node_affinity_destroy_all(void)
 {
 	struct list_head *pos, *q;
 	struct hfi1_affinity_node *entry;
@@ -218,7 +224,7 @@ void node_affinity_destroy(void)
 		entry = list_entry(pos, struct hfi1_affinity_node,
 				   list);
 		list_del(pos);
-		kfree(entry);
+		node_affinity_destroy(entry);
 	}
 	mutex_unlock(&node_affinity.lock);
 	kfree(hfi1_per_node_cntr);
@@ -232,6 +238,7 @@ static struct hfi1_affinity_node *node_affinity_allocate(int node)
 	if (!entry)
 		return NULL;
 	entry->node = node;
+	entry->comp_vect_affinity = alloc_percpu(u16);
 	INIT_LIST_HEAD(&entry->list);
 
 	return entry;
@@ -261,6 +268,341 @@ static struct hfi1_affinity_node *node_affinity_lookup(int node)
 	return NULL;
 }
 
+static int per_cpu_affinity_get(cpumask_var_t possible_cpumask,
+				u16 __percpu *comp_vect_affinity)
+{
+	int curr_cpu;
+	u16 cntr;
+	u16 prev_cntr;
+	int ret_cpu;
+
+	if (!possible_cpumask) {
+		ret_cpu = -EINVAL;
+		goto fail;
+	}
+
+	if (!comp_vect_affinity) {
+		ret_cpu = -EINVAL;
+		goto fail;
+	}
+
+	ret_cpu = cpumask_first(possible_cpumask);
+	if (ret_cpu >= nr_cpu_ids) {
+		ret_cpu = -EINVAL;
+		goto fail;
+	}
+
+	prev_cntr = *per_cpu_ptr(comp_vect_affinity, ret_cpu);
+	for_each_cpu(curr_cpu, possible_cpumask) {
+		cntr = *per_cpu_ptr(comp_vect_affinity, curr_cpu);
+
+		if (cntr < prev_cntr) {
+			ret_cpu = curr_cpu;
+			prev_cntr = cntr;
+		}
+	}
+
+	*per_cpu_ptr(comp_vect_affinity, ret_cpu) += 1;
+
+fail:
+	return ret_cpu;
+}
+
+static int per_cpu_affinity_put_max(cpumask_var_t possible_cpumask,
+				    u16 __percpu *comp_vect_affinity)
+{
+	int curr_cpu;
+	int max_cpu;
+	u16 cntr;
+	u16 prev_cntr;
+
+	if (!possible_cpumask)
+		return -EINVAL;
+
+	if (!comp_vect_affinity)
+		return -EINVAL;
+
+	max_cpu = cpumask_first(possible_cpumask);
+	if (max_cpu >= nr_cpu_ids)
+		return -EINVAL;
+
+	prev_cntr = *per_cpu_ptr(comp_vect_affinity, max_cpu);
+	for_each_cpu(curr_cpu, possible_cpumask) {
+		cntr = *per_cpu_ptr(comp_vect_affinity, curr_cpu);
+
+		if (cntr > prev_cntr) {
+			max_cpu = curr_cpu;
+			prev_cntr = cntr;
+		}
+	}
+
+	*per_cpu_ptr(comp_vect_affinity, max_cpu) -= 1;
+
+	return max_cpu;
+}
+
+/*
+ * Non-interrupt CPUs are used first, then interrupt CPUs.
+ * Two already allocated cpu masks must be passed.
+ */
+static int _dev_comp_vect_cpu_get(struct hfi1_devdata *dd,
+				  struct hfi1_affinity_node *entry,
+				  cpumask_var_t non_intr_cpus,
+				  cpumask_var_t available_cpus)
+	__must_hold(&node_affinity.lock)
+{
+	int cpu;
+	struct cpu_mask_set *set = dd->comp_vect;
+
+	lockdep_assert_held(&node_affinity.lock);
+	if (!non_intr_cpus) {
+		cpu = -1;
+		goto fail;
+	}
+
+	if (!available_cpus) {
+		cpu = -1;
+		goto fail;
+	}
+
+	/* Available CPUs for pinning completion vectors */
+	_cpu_mask_set_gen_inc(set);
+	cpumask_andnot(available_cpus, &set->mask, &set->used);
+
+	/* Available CPUs without SDMA engine interrupts */
+	cpumask_andnot(non_intr_cpus, available_cpus,
+		       &entry->def_intr.used);
+
+	/* If there are non-interrupt CPUs available, use them first */
+	if (!cpumask_empty(non_intr_cpus))
+		cpu = cpumask_first(non_intr_cpus);
+	else /* Otherwise, use interrupt CPUs */
+		cpu = cpumask_first(available_cpus);
+
+	if (cpu >= nr_cpu_ids) { /* empty */
+		cpu = -1;
+		goto fail;
+	}
+	cpumask_set_cpu(cpu, &set->used);
+
+fail:
+	return cpu;
+}
+
+static void _dev_comp_vect_cpu_put(struct hfi1_devdata *dd, int cpu)
+{
+	struct cpu_mask_set *set = dd->comp_vect;
+
+	if (cpu < 0)
+		return;
+
+	cpu_mask_set_put(set, cpu);
+}
+
+/* _dev_comp_vect_mappings_destroy() is reentrant */
+static void _dev_comp_vect_mappings_destroy(struct hfi1_devdata *dd)
+{
+	int i, cpu;
+
+	if (!dd->comp_vect_mappings)
+		return;
+
+	for (i = 0; i < dd->comp_vect_possible_cpus; i++) {
+		cpu = dd->comp_vect_mappings[i];
+		_dev_comp_vect_cpu_put(dd, cpu);
+		dd->comp_vect_mappings[i] = -1;
+		hfi1_cdbg(AFFINITY,
+			  "[%s] Release CPU %d from completion vector %d",
+			  rvt_get_ibdev_name(&(dd)->verbs_dev.rdi), cpu, i);
+	}
+
+	kfree(dd->comp_vect_mappings);
+	dd->comp_vect_mappings = NULL;
+}
+
+/*
+ * This function creates the table for looking up CPUs for completion vectors.
+ * num_comp_vectors needs to have been initilized before calling this function.
+ */
+static int _dev_comp_vect_mappings_create(struct hfi1_devdata *dd,
+					  struct hfi1_affinity_node *entry)
+	__must_hold(&node_affinity.lock)
+{
+	int i, cpu, ret;
+	cpumask_var_t non_intr_cpus;
+	cpumask_var_t available_cpus;
+
+	lockdep_assert_held(&node_affinity.lock);
+
+	if (!zalloc_cpumask_var(&non_intr_cpus, GFP_KERNEL))
+		return -ENOMEM;
+
+	if (!zalloc_cpumask_var(&available_cpus, GFP_KERNEL)) {
+		free_cpumask_var(non_intr_cpus);
+		return -ENOMEM;
+	}
+
+	dd->comp_vect_mappings = kcalloc(dd->comp_vect_possible_cpus,
+					 sizeof(*dd->comp_vect_mappings),
+					 GFP_KERNEL);
+	if (!dd->comp_vect_mappings) {
+		ret = -ENOMEM;
+		goto fail;
+	}
+	for (i = 0; i < dd->comp_vect_possible_cpus; i++)
+		dd->comp_vect_mappings[i] = -1;
+
+	for (i = 0; i < dd->comp_vect_possible_cpus; i++) {
+		cpu = _dev_comp_vect_cpu_get(dd, entry, non_intr_cpus,
+					     available_cpus);
+		if (cpu < 0) {
+			ret = -EINVAL;
+			goto fail;
+		}
+
+		dd->comp_vect_mappings[i] = cpu;
+		hfi1_cdbg(AFFINITY,
+			  "[%s] Completion Vector %d -> CPU %d",
+			  rvt_get_ibdev_name(&(dd)->verbs_dev.rdi), i, cpu);
+	}
+
+	return 0;
+
+fail:
+	free_cpumask_var(available_cpus);
+	free_cpumask_var(non_intr_cpus);
+	_dev_comp_vect_mappings_destroy(dd);
+
+	return ret;
+}
+
+int hfi1_comp_vectors_set_up(struct hfi1_devdata *dd)
+{
+	int ret;
+	struct hfi1_affinity_node *entry;
+
+	mutex_lock(&node_affinity.lock);
+	entry = node_affinity_lookup(dd->node);
+	if (!entry) {
+		ret = -EINVAL;
+		goto unlock;
+	}
+	ret = _dev_comp_vect_mappings_create(dd, entry);
+unlock:
+	mutex_unlock(&node_affinity.lock);
+
+	return ret;
+}
+
+void hfi1_comp_vectors_clean_up(struct hfi1_devdata *dd)
+{
+	_dev_comp_vect_mappings_destroy(dd);
+}
+
+int hfi1_comp_vect_mappings_lookup(struct rvt_dev_info *rdi, int comp_vect)
+{
+	struct hfi1_ibdev *verbs_dev = dev_from_rdi(rdi);
+	struct hfi1_devdata *dd = dd_from_dev(verbs_dev);
+
+	if (!dd->comp_vect_mappings)
+		return -EINVAL;
+	if (comp_vect >= dd->comp_vect_possible_cpus)
+		return -EINVAL;
+
+	return dd->comp_vect_mappings[comp_vect];
+}
+
+/*
+ * It assumes dd->comp_vect_possible_cpus is available.
+ */
+static int _dev_comp_vect_cpu_mask_init(struct hfi1_devdata *dd,
+					struct hfi1_affinity_node *entry,
+					bool first_dev_init)
+	__must_hold(&node_affinity.lock)
+{
+	int i, j, curr_cpu;
+	int possible_cpus_comp_vect = 0;
+	struct cpumask *dev_comp_vect_mask = &dd->comp_vect->mask;
+
+	lockdep_assert_held(&node_affinity.lock);
+	/*
+	 * If there's only one CPU available for completion vectors, then
+	 * there will only be one completion vector available. Othewise,
+	 * the number of completion vector available will be the number of
+	 * available CPUs divide it by the number of devices in the
+	 * local NUMA node.
+	 */
+	if (cpumask_weight(&entry->comp_vect_mask) == 1) {
+		possible_cpus_comp_vect = 1;
+		dd_dev_warn(dd,
+			    "Number of kernel receive queues is too large for completion vector affinity to be effective\n");
+	} else {
+		possible_cpus_comp_vect +=
+			cpumask_weight(&entry->comp_vect_mask) /
+				       hfi1_per_node_cntr[dd->node];
+
+		/*
+		 * If the completion vector CPUs available doesn't divide
+		 * evenly among devices, then the first device device to be
+		 * initialized gets an extra CPU.
+		 */
+		if (first_dev_init &&
+		    cpumask_weight(&entry->comp_vect_mask) %
+		    hfi1_per_node_cntr[dd->node] != 0)
+			possible_cpus_comp_vect++;
+	}
+
+	dd->comp_vect_possible_cpus = possible_cpus_comp_vect;
+
+	/* Reserving CPUs for device completion vector */
+	for (i = 0; i < dd->comp_vect_possible_cpus; i++) {
+		curr_cpu = per_cpu_affinity_get(&entry->comp_vect_mask,
+						entry->comp_vect_affinity);
+		if (curr_cpu < 0)
+			goto fail;
+
+		cpumask_set_cpu(curr_cpu, dev_comp_vect_mask);
+	}
+
+	hfi1_cdbg(AFFINITY,
+		  "[%s] Completion vector affinity CPU set(s) %*pbl",
+		  rvt_get_ibdev_name(&(dd)->verbs_dev.rdi),
+		  cpumask_pr_args(dev_comp_vect_mask));
+
+	return 0;
+
+fail:
+	for (j = 0; j < i; j++)
+		per_cpu_affinity_put_max(&entry->comp_vect_mask,
+					 entry->comp_vect_affinity);
+
+	return curr_cpu;
+}
+
+/*
+ * It assumes dd->comp_vect_possible_cpus is available.
+ */
+static void _dev_comp_vect_cpu_mask_clean_up(struct hfi1_devdata *dd,
+					     struct hfi1_affinity_node *entry)
+	__must_hold(&node_affinity.lock)
+{
+	int i, cpu;
+
+	lockdep_assert_held(&node_affinity.lock);
+	if (!dd->comp_vect_possible_cpus)
+		return;
+
+	for (i = 0; i < dd->comp_vect_possible_cpus; i++) {
+		cpu = per_cpu_affinity_put_max(&dd->comp_vect->mask,
+					       entry->comp_vect_affinity);
+		/* Clearing CPU in device completion vector cpu mask */
+		if (cpu >= 0)
+			cpumask_clear_cpu(cpu, &dd->comp_vect->mask);
+	}
+
+	dd->comp_vect_possible_cpus = 0;
+}
+
 /*
  * Interrupt affinity.
  *
@@ -277,7 +619,8 @@ int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 	int node = pcibus_to_node(dd->pcidev->bus);
 	struct hfi1_affinity_node *entry;
 	const struct cpumask *local_mask;
-	int curr_cpu, possible, i;
+	int curr_cpu, possible, i, ret;
+	bool new_entry = false;
 
 	if (node < 0)
 		node = numa_node_id();
@@ -299,11 +642,14 @@ int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 		if (!entry) {
 			dd_dev_err(dd,
 				   "Unable to allocate global affinity node\n");
-			mutex_unlock(&node_affinity.lock);
-			return -ENOMEM;
+			ret = -ENOMEM;
+			goto fail;
 		}
+		new_entry = true;
+
 		init_cpu_mask_set(&entry->def_intr);
 		init_cpu_mask_set(&entry->rcv_intr);
+		cpumask_clear(&entry->comp_vect_mask);
 		cpumask_clear(&entry->general_intr_mask);
 		/* Use the "real" cpu mask of this node as the default */
 		cpumask_and(&entry->def_intr.mask, &node_affinity.real_cpu_mask,
@@ -356,10 +702,64 @@ int hfi1_dev_affinity_init(struct hfi1_devdata *dd)
 					     &entry->general_intr_mask);
 		}
 
-		node_affinity_add_tail(entry);
+		/* Determine completion vector CPUs for the entire node */
+		cpumask_and(&entry->comp_vect_mask,
+			    &node_affinity.real_cpu_mask, local_mask);
+		cpumask_andnot(&entry->comp_vect_mask,
+			       &entry->comp_vect_mask,
+			       &entry->rcv_intr.mask);
+		cpumask_andnot(&entry->comp_vect_mask,
+			       &entry->comp_vect_mask,
+			       &entry->general_intr_mask);
+
+		/*
+		 * If there ends up being 0 CPU cores leftover for completion
+		 * vectors, use the same CPU core as the general/control
+		 * context.
+		 */
+		if (cpumask_weight(&entry->comp_vect_mask) == 0)
+			cpumask_copy(&entry->comp_vect_mask,
+				     &entry->general_intr_mask);
 	}
+
+	ret = _dev_comp_vect_cpu_mask_init(dd, entry, new_entry);
+	if (ret < 0)
+		goto fail;
+
+	if (new_entry)
+		node_affinity_add_tail(entry);
+
 	mutex_unlock(&node_affinity.lock);
+
 	return 0;
+
+fail:
+	if (new_entry)
+		node_affinity_destroy(entry);
+	mutex_unlock(&node_affinity.lock);
+	return ret;
+}
+
+void hfi1_dev_affinity_clean_up(struct hfi1_devdata *dd)
+{
+	struct hfi1_affinity_node *entry;
+
+	if (dd->node < 0)
+		return;
+
+	mutex_lock(&node_affinity.lock);
+	entry = node_affinity_lookup(dd->node);
+	if (!entry)
+		goto unlock;
+
+	/*
+	 * Free device completion vector CPUs to be used by future
+	 * completion vectors
+	 */
+	_dev_comp_vect_cpu_mask_clean_up(dd, entry);
+unlock:
+	mutex_unlock(&node_affinity.lock);
+	dd->node = -1;
 }
 
 /*
diff --git a/drivers/infiniband/hw/hfi1/affinity.h b/drivers/infiniband/hw/hfi1/affinity.h
index 2a1e374169c0..6a7e6ea4e426 100644
--- a/drivers/infiniband/hw/hfi1/affinity.h
+++ b/drivers/infiniband/hw/hfi1/affinity.h
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015 - 2017 Intel Corporation.
+ * Copyright(c) 2015 - 2018 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -98,9 +98,11 @@ void hfi1_put_proc_affinity(int cpu);
 
 struct hfi1_affinity_node {
 	int node;
+	u16 __percpu *comp_vect_affinity;
 	struct cpu_mask_set def_intr;
 	struct cpu_mask_set rcv_intr;
 	struct cpumask general_intr_mask;
+	struct cpumask comp_vect_mask;
 	struct list_head list;
 };
 
@@ -116,7 +118,11 @@ struct hfi1_affinity_node_list {
 };
 
 int node_affinity_init(void);
-void node_affinity_destroy(void);
+void node_affinity_destroy_all(void);
 extern struct hfi1_affinity_node_list node_affinity;
+void hfi1_dev_affinity_clean_up(struct hfi1_devdata *dd);
+int hfi1_comp_vect_mappings_lookup(struct rvt_dev_info *rdi, int comp_vect);
+int hfi1_comp_vectors_set_up(struct hfi1_devdata *dd);
+void hfi1_comp_vectors_clean_up(struct hfi1_devdata *dd);
 
 #endif /* _HFI1_AFFINITY_H */
diff --git a/drivers/infiniband/hw/hfi1/chip.c b/drivers/infiniband/hw/hfi1/chip.c
index 0fab6df0a345..46e9e4ffcba4 100644
--- a/drivers/infiniband/hw/hfi1/chip.c
+++ b/drivers/infiniband/hw/hfi1/chip.c
@@ -15233,6 +15233,10 @@ struct hfi1_devdata *hfi1_init_dd(struct pci_dev *pdev,
 	if (ret)
 		goto bail_cleanup;
 
+	ret = hfi1_comp_vectors_set_up(dd);
+	if (ret)
+		goto bail_clear_intr;
+
 	/* set up LCB access - must be after set_up_interrupts() */
 	init_lcb_access(dd);
 
@@ -15275,6 +15279,7 @@ bail_free_rcverr:
 bail_free_cntrs:
 	free_cntrs(dd);
 bail_clear_intr:
+	hfi1_comp_vectors_clean_up(dd);
 	hfi1_clean_up_interrupts(dd);
 bail_cleanup:
 	hfi1_pcie_ddcleanup(dd);
diff --git a/drivers/infiniband/hw/hfi1/hfi.h b/drivers/infiniband/hw/hfi1/hfi.h
index 9cd758ce7764..dd84238c1aac 100644
--- a/drivers/infiniband/hw/hfi1/hfi.h
+++ b/drivers/infiniband/hw/hfi1/hfi.h
@@ -1263,6 +1263,9 @@ struct hfi1_devdata {
 
 	/* Save the enabled LCB error bits */
 	u64 lcb_err_en;
+	struct cpu_mask_set *comp_vect;
+	int *comp_vect_mappings;
+	u32 comp_vect_possible_cpus;
 
 	/*
 	 * Capability to have different send engines simply by changing a
diff --git a/drivers/infiniband/hw/hfi1/init.c b/drivers/infiniband/hw/hfi1/init.c
index 790542ce89a5..5d1adfc450d3 100644
--- a/drivers/infiniband/hw/hfi1/init.c
+++ b/drivers/infiniband/hw/hfi1/init.c
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015-2017 Intel Corporation.
+ * Copyright(c) 2015 - 2018 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -1244,6 +1244,8 @@ static void hfi1_clean_devdata(struct hfi1_devdata *dd)
 	dd->rcv_limit     = NULL;
 	dd->send_schedule = NULL;
 	dd->tx_opstats    = NULL;
+	kfree(dd->comp_vect);
+	dd->comp_vect = NULL;
 	sdma_clean(dd, dd->num_sdma);
 	rvt_dealloc_device(&dd->verbs_dev.rdi);
 }
@@ -1300,6 +1302,7 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 		dd->unit = ret;
 		list_add(&dd->list, &hfi1_dev_list);
 	}
+	dd->node = -1;
 
 	spin_unlock_irqrestore(&hfi1_devs_lock, flags);
 	idr_preload_end();
@@ -1352,6 +1355,12 @@ struct hfi1_devdata *hfi1_alloc_devdata(struct pci_dev *pdev, size_t extra)
 		goto bail;
 	}
 
+	dd->comp_vect = kzalloc(sizeof(*dd->comp_vect), GFP_KERNEL);
+	if (!dd->comp_vect) {
+		ret = -ENOMEM;
+		goto bail;
+	}
+
 	kobject_init(&dd->kobj, &hfi1_devdata_type);
 	return dd;
 
@@ -1521,7 +1530,7 @@ module_init(hfi1_mod_init);
 static void __exit hfi1_mod_cleanup(void)
 {
 	pci_unregister_driver(&hfi1_pci_driver);
-	node_affinity_destroy();
+	node_affinity_destroy_all();
 	hfi1_wss_exit();
 	hfi1_dbg_exit();
 
@@ -1605,6 +1614,8 @@ static void cleanup_device_data(struct hfi1_devdata *dd)
 static void postinit_cleanup(struct hfi1_devdata *dd)
 {
 	hfi1_start_cleanup(dd);
+	hfi1_comp_vectors_clean_up(dd);
+	hfi1_dev_affinity_clean_up(dd);
 
 	hfi1_pcie_ddcleanup(dd);
 	hfi1_pcie_cleanup(dd->pcidev);
diff --git a/drivers/infiniband/hw/hfi1/trace.c b/drivers/infiniband/hw/hfi1/trace.c
index 89bd9851065b..332b9b7c554a 100644
--- a/drivers/infiniband/hw/hfi1/trace.c
+++ b/drivers/infiniband/hw/hfi1/trace.c
@@ -1,5 +1,5 @@
 /*
- * Copyright(c) 2015 - 2017 Intel Corporation.
+ * Copyright(c) 2015 - 2018 Intel Corporation.
  *
  * This file is provided under a dual BSD/GPLv2 license.  When using or
  * redistributing this file, you may do so under either license.
@@ -374,6 +374,7 @@ const char *print_u32_array(
 	return ret;
 }
 
+__hfi1_trace_fn(AFFINITY);
 __hfi1_trace_fn(PKT);
 __hfi1_trace_fn(PROC);
 __hfi1_trace_fn(SDMA);
diff --git a/drivers/infiniband/hw/hfi1/trace_dbg.h b/drivers/infiniband/hw/hfi1/trace_dbg.h
index 0e7d929530c5..e62171fb7379 100644
--- a/drivers/infiniband/hw/hfi1/trace_dbg.h
+++ b/drivers/infiniband/hw/hfi1/trace_dbg.h
@@ -1,5 +1,5 @@
 /*
-* Copyright(c) 2015, 2016 Intel Corporation.
+* Copyright(c) 2015 - 2018 Intel Corporation.
 *
 * This file is provided under a dual BSD/GPLv2 license.  When using or
 * redistributing this file, you may do so under either license.
@@ -113,6 +113,7 @@ void __hfi1_trace_##lvl(const char *func, char *fmt, ...)		\
  * hfi1_cdbg(LVL, fmt, ...); as well as take care of all
  * the debugfs stuff.
  */
+__hfi1_trace_def(AFFINITY);
 __hfi1_trace_def(PKT);
 __hfi1_trace_def(PROC);
 __hfi1_trace_def(SDMA);
diff --git a/drivers/infiniband/hw/hfi1/verbs.c b/drivers/infiniband/hw/hfi1/verbs.c
index 9554e912af98..fc2e44cde161 100644
--- a/drivers/infiniband/hw/hfi1/verbs.c
+++ b/drivers/infiniband/hw/hfi1/verbs.c
@@ -64,6 +64,7 @@
 #include "debugfs.h"
 #include "vnic.h"
 #include "fault.h"
+#include "affinity.h"
 
 static unsigned int hfi1_lkey_table_size = 16;
 module_param_named(lkey_table_size, hfi1_lkey_table_size, uint,
@@ -1934,11 +1935,11 @@ int hfi1_register_ib_device(struct hfi1_devdata *dd)
 	dd->verbs_dev.rdi.driver_f.modify_qp = hfi1_modify_qp;
 	dd->verbs_dev.rdi.driver_f.notify_restart_rc = hfi1_restart_rc;
 	dd->verbs_dev.rdi.driver_f.check_send_wqe = hfi1_check_send_wqe;
+	dd->verbs_dev.rdi.driver_f.comp_vect_cpu_lookup =
+						hfi1_comp_vect_mappings_lookup;
 
 	/* completeion queue */
-	snprintf(dd->verbs_dev.rdi.dparms.cq_name,
-		 sizeof(dd->verbs_dev.rdi.dparms.cq_name),
-		 "hfi1_cq%d", dd->unit);
+	dd->verbs_dev.rdi.ibdev.num_comp_vectors = dd->comp_vect_possible_cpus;
 	dd->verbs_dev.rdi.dparms.node = dd->node;
 
 	/* misc settings */
diff --git a/drivers/infiniband/hw/qib/qib_verbs.c b/drivers/infiniband/hw/qib/qib_verbs.c
index 3977abbc83ad..14b4057a2b8f 100644
--- a/drivers/infiniband/hw/qib/qib_verbs.c
+++ b/drivers/infiniband/hw/qib/qib_verbs.c
@@ -1,5 +1,5 @@
 /*
- * Copyright (c) 2012, 2013 Intel Corporation.  All rights reserved.
+ * Copyright (c) 2012 - 2018 Intel Corporation.  All rights reserved.
  * Copyright (c) 2006 - 2012 QLogic Corporation. All rights reserved.
  * Copyright (c) 2005, 2006 PathScale, Inc. All rights reserved.
  *
@@ -1631,10 +1631,6 @@ int qib_register_ib_device(struct qib_devdata *dd)
 	dd->verbs_dev.rdi.dparms.core_cap_flags = RDMA_CORE_PORT_IBA_IB;
 	dd->verbs_dev.rdi.dparms.max_mad_size = IB_MGMT_MAD_SIZE;
 
-	snprintf(dd->verbs_dev.rdi.dparms.cq_name,
-		 sizeof(dd->verbs_dev.rdi.dparms.cq_name),
-		 "qib_cq%d", dd->unit);
-
 	qib_fill_device_attr(dd);
 
 	ppd = dd->pport;