scsi: lpfc: Fix poor use of hardware queues if fewer irq vectors

While fixing the resources per socket, realized the driver was not using
hardware queues (up to 1 per cpu) if there were fewer interrupt
vectors. The driver was only using the hardware queue assigned to the cpu
with the vector.

Rework the affinity map check to use the additional hardware queue elements
that had been allocated. If the cpu count exceeds the hardware queue count
- share, but choose what is shared with by: hyperthread peer, core peer,
socket peer, or finally similar cpu in a different socket.

Signed-off-by: Dick Kennedy <[e-mail]>
Signed-off-by: James Smart <[e-mail]>
Signed-off-by: Martin K. Petersen <[e-mail]>

drivers/scsi/lpfc/lpfc_attr.c

···
 
 	/* set the values on the cq's */
 	for (i = 0; i < phba->cfg_irq_chann; i++) {
-		eq = phba->sli4_hba.hdwq[i].hba_eq;
+		/* Get the EQ corresponding to the IRQ vector */
+		eq = phba->sli4_hba.hba_eq_hdl[i].eq;
 		if (!eq)
 			continue;
 

drivers/scsi/lpfc/lpfc_init.c

···
 static void lpfc_sli4_disable_intr(struct lpfc_hba *);
 static uint32_t lpfc_sli4_enable_intr(struct lpfc_hba *, uint32_t);
 static void lpfc_sli4_oas_verify(struct lpfc_hba *phba);
-static uint16_t lpfc_find_eq_handle(struct lpfc_hba *, uint16_t);
 static uint16_t lpfc_find_cpu_handle(struct lpfc_hba *, uint16_t, int);
 
 static struct scsi_transport_template *lpfc_transport_template = NULL;
···
 	if (!eqcnt)
 		goto requeue;
 
+	/* Loop thru all IRQ vectors */
 	for (i = 0; i < phba->cfg_irq_chann; i++) {
-		eq = phba->sli4_hba.hdwq[i].hba_eq;
+		/* Get the EQ corresponding to the IRQ vector */
+		eq = phba->sli4_hba.hba_eq_hdl[i].eq;
 		if (eq && eqcnt[eq->last_cpu] < 2)
 			eqcnt[eq->last_cpu]++;
 		continue;
···
 lpfc_sli4_queue_create(struct lpfc_hba *phba)
 {
 	struct lpfc_queue *qdesc;
-	int idx, eqidx, cpu;
+	int idx, cpu, eqcpu;
 	struct lpfc_sli4_hdw_queue *qp;
+	struct lpfc_vector_map_info *cpup;
+	struct lpfc_vector_map_info *eqcpup;
 	struct lpfc_eq_intr_info *eqi;
 
 	/*
···
 	INIT_LIST_HEAD(&phba->sli4_hba.lpfc_wq_list);
 
 	/* Create HBA Event Queues (EQs) */
-	for (idx = 0; idx < phba->cfg_hdw_queue; idx++) {
-		/* determine EQ affinity */
-		eqidx = lpfc_find_eq_handle(phba, idx);
-		cpu = lpfc_find_cpu_handle(phba, eqidx, LPFC_FIND_BY_EQ);
-		/*
-		 * If there are more Hardware Queues than available
-		 * EQs, multiple Hardware Queues may share a common EQ.
+	for_each_present_cpu(cpu) {
+		/* We only want to create 1 EQ per vector, even though
+		 * multiple CPUs might be using that vector. so only
+		 * selects the CPUs that are LPFC_CPU_FIRST_IRQ.
 		 */
-		if (idx >= phba->cfg_irq_chann) {
-			/* Share an existing EQ */
-			phba->sli4_hba.hdwq[idx].hba_eq =
-				phba->sli4_hba.hdwq[eqidx].hba_eq;
+		cpup = &phba->sli4_hba.cpu_map[cpu];
+		if (!(cpup->flag & LPFC_CPU_FIRST_IRQ))
 			continue;
-		}
-		/* Create an EQ */
+
+		/* Get a ptr to the Hardware Queue associated with this CPU */
+		qp = &phba->sli4_hba.hdwq[cpup->hdwq];
+
+		/* Allocate an EQ */
 		qdesc = lpfc_sli4_queue_alloc(phba, LPFC_DEFAULT_PAGE_SIZE,
 					      phba->sli4_hba.eq_esize,
 					      phba->sli4_hba.eq_ecount, cpu);
 		if (!qdesc) {
 			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
-					"0497 Failed allocate EQ (%d)\n", idx);
+					"0497 Failed allocate EQ (%d)\n",
+					cpup->hdwq);
 			goto out_error;
 		}
 		qdesc->qe_valid = 1;
-		qdesc->hdwq = idx;
-
-		/* Save the CPU this EQ is affinitised to */
-		qdesc->chann = cpu;
-		phba->sli4_hba.hdwq[idx].hba_eq = qdesc;
+		qdesc->hdwq = cpup->hdwq;
+		qdesc->chann = cpu; /* First CPU this EQ is affinitised to */
 		qdesc->last_cpu = qdesc->chann;
+
+		/* Save the allocated EQ in the Hardware Queue */
+		qp->hba_eq = qdesc;
+
 		eqi = per_cpu_ptr(phba->sli4_hba.eq_info, qdesc->last_cpu);
 		list_add(&qdesc->cpu_list, &eqi->list);
 	}
 
+	/* Now we need to populate the other Hardware Queues, that share
+	 * an IRQ vector, with the associated EQ ptr.
+	 */
+	for_each_present_cpu(cpu) {
+		cpup = &phba->sli4_hba.cpu_map[cpu];
+
+		/* Check for EQ already allocated in previous loop */
+		if (cpup->flag & LPFC_CPU_FIRST_IRQ)
+			continue;
+
+		/* Check for multiple CPUs per hdwq */
+		qp = &phba->sli4_hba.hdwq[cpup->hdwq];
+		if (qp->hba_eq)
+			continue;
+
+		/* We need to share an EQ for this hdwq */
+		eqcpu = lpfc_find_cpu_handle(phba, cpup->eq, LPFC_FIND_BY_EQ);
+		eqcpup = &phba->sli4_hba.cpu_map[eqcpu];
+		qp->hba_eq = phba->sli4_hba.hdwq[eqcpup->hdwq].hba_eq;
+	}
 
 	/* Allocate SCSI SLI4 CQ/WQs */
 	for (idx = 0; idx < phba->cfg_hdw_queue; idx++) {
···
 lpfc_sli4_release_hdwq(struct lpfc_hba *phba)
 {
 	struct lpfc_sli4_hdw_queue *hdwq;
+	struct lpfc_queue *eq;
 	uint32_t idx;
 
 	hdwq = phba->sli4_hba.hdwq;
-	for (idx = 0; idx < phba->cfg_hdw_queue; idx++) {
-		if (idx < phba->cfg_irq_chann)
-			lpfc_sli4_queue_free(hdwq[idx].hba_eq);
-		hdwq[idx].hba_eq = NULL;
 
+	/* Loop thru all Hardware Queues */
+	for (idx = 0; idx < phba->cfg_hdw_queue; idx++) {
+		/* Free the CQ/WQ corresponding to the Hardware Queue */
 		lpfc_sli4_queue_free(hdwq[idx].fcp_cq);
 		lpfc_sli4_queue_free(hdwq[idx].nvme_cq);
 		lpfc_sli4_queue_free(hdwq[idx].fcp_wq);
 		lpfc_sli4_queue_free(hdwq[idx].nvme_wq);
+		hdwq[idx].hba_eq = NULL;
 		hdwq[idx].fcp_cq = NULL;
 		hdwq[idx].nvme_cq = NULL;
 		hdwq[idx].fcp_wq = NULL;
 		hdwq[idx].nvme_wq = NULL;
+	}
+	/* Loop thru all IRQ vectors */
+	for (idx = 0; idx < phba->cfg_irq_chann; idx++) {
+		/* Free the EQ corresponding to the IRQ vector */
+		eq = phba->sli4_hba.hba_eq_hdl[idx].eq;
+		lpfc_sli4_queue_free(eq);
+		phba->sli4_hba.hba_eq_hdl[idx].eq = NULL;
 	}
 }
 
···
 	qp = phba->sli4_hba.hdwq;
 	memset(phba->sli4_hba.cq_lookup, 0,
 	       (sizeof(struct lpfc_queue *) * (phba->sli4_hba.cq_max + 1)));
+	/* Loop thru all IRQ vectors */
 	for (qidx = 0; qidx < phba->cfg_irq_chann; qidx++) {
-		eq = qp[qidx].hba_eq;
+		/* Get the EQ corresponding to the IRQ vector */
+		eq = phba->sli4_hba.hba_eq_hdl[qidx].eq;
 		if (!eq)
 			continue;
+		/* Loop through all CQs associated with that EQ */
 		list_for_each_entry(childq, &eq->child_list, list) {
 			if (childq->queue_id > phba->sli4_hba.cq_max)
 				continue;
···
 {
 	uint32_t shdr_status, shdr_add_status;
 	union lpfc_sli4_cfg_shdr *shdr;
+	struct lpfc_vector_map_info *cpup;
 	struct lpfc_sli4_hdw_queue *qp;
 	LPFC_MBOXQ_t *mboxq;
-	int qidx;
+	int qidx, cpu;
 	uint32_t length, usdelay;
 	int rc = -ENOMEM;
 
···
 		rc = -ENOMEM;
 		goto out_error;
 	}
+
+	/* Loop thru all IRQ vectors */
 	for (qidx = 0; qidx < phba->cfg_irq_chann; qidx++) {
-		if (!qp[qidx].hba_eq) {
-			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
-					"0522 Fast-path EQ (%d) not "
-					"allocated\n", qidx);
-			rc = -ENOMEM;
-			goto out_destroy;
+		/* Create HBA Event Queues (EQs) in order */
+		for_each_present_cpu(cpu) {
+			cpup = &phba->sli4_hba.cpu_map[cpu];
+
+			/* Look for the CPU thats using that vector with
+			 * LPFC_CPU_FIRST_IRQ set.
+			 */
+			if (!(cpup->flag & LPFC_CPU_FIRST_IRQ))
+				continue;
+			if (qidx != cpup->eq)
+				continue;
+
+			/* Create an EQ for that vector */
+			rc = lpfc_eq_create(phba, qp[cpup->hdwq].hba_eq,
+					    phba->cfg_fcp_imax);
+			if (rc) {
+				lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+						"0523 Failed setup of fast-path"
+						" EQ (%d), rc = 0x%x\n",
+						cpup->eq, (uint32_t)rc);
+				goto out_destroy;
+			}
+
+			/* Save the EQ for that vector in the hba_eq_hdl */
+			phba->sli4_hba.hba_eq_hdl[cpup->eq].eq =
+				qp[cpup->hdwq].hba_eq;
+
+			lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
+					"2584 HBA EQ setup: queue[%d]-id=%d\n",
+					cpup->eq,
+					qp[cpup->hdwq].hba_eq->queue_id);
 		}
-		rc = lpfc_eq_create(phba, qp[qidx].hba_eq,
-				    phba->cfg_fcp_imax);
-		if (rc) {
-			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
-					"0523 Failed setup of fast-path EQ "
-					"(%d), rc = 0x%x\n", qidx,
-					(uint32_t)rc);
-			goto out_destroy;
-		}
-		lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
-				"2584 HBA EQ setup: queue[%d]-id=%d\n", qidx,
-				qp[qidx].hba_eq->queue_id);
 	}
 
+	/* Loop thru all Hardware Queues */
 	if (phba->cfg_enable_fc4_type & LPFC_ENABLE_NVME) {
 		for (qidx = 0; qidx < phba->cfg_hdw_queue; qidx++) {
+			cpu = lpfc_find_cpu_handle(phba, qidx,
+						   LPFC_FIND_BY_HDWQ);
+			cpup = &phba->sli4_hba.cpu_map[cpu];
+
+			/* Create the CQ/WQ corresponding to the
+			 * Hardware Queue
+			 */
 			rc = lpfc_create_wq_cq(phba,
-					qp[qidx].hba_eq,
+					phba->sli4_hba.hdwq[cpup->hdwq].hba_eq,
 					qp[qidx].nvme_cq,
 					qp[qidx].nvme_wq,
 					&phba->sli4_hba.hdwq[qidx].nvme_cq_map,
···
 	}
 
 	for (qidx = 0; qidx < phba->cfg_hdw_queue; qidx++) {
+		cpu = lpfc_find_cpu_handle(phba, qidx, LPFC_FIND_BY_HDWQ);
+		cpup = &phba->sli4_hba.cpu_map[cpu];
+
+		/* Create the CQ/WQ corresponding to the Hardware Queue */
 		rc = lpfc_create_wq_cq(phba,
-				       qp[qidx].hba_eq,
+				       phba->sli4_hba.hdwq[cpup->hdwq].hba_eq,
 				       qp[qidx].fcp_cq,
 				       qp[qidx].fcp_wq,
 				       &phba->sli4_hba.hdwq[qidx].fcp_cq_map,
···
 lpfc_sli4_queue_unset(struct lpfc_hba *phba)
 {
 	struct lpfc_sli4_hdw_queue *qp;
+	struct lpfc_queue *eq;
 	int qidx;
 
 	/* Unset mailbox command work queue */
···
 
 	/* Unset fast-path SLI4 queues */
 	if (phba->sli4_hba.hdwq) {
+		/* Loop thru all Hardware Queues */
 		for (qidx = 0; qidx < phba->cfg_hdw_queue; qidx++) {
+			/* Destroy the CQ/WQ corresponding to Hardware Queue */
 			qp = &phba->sli4_hba.hdwq[qidx];
 			lpfc_wq_destroy(phba, qp->fcp_wq);
 			lpfc_wq_destroy(phba, qp->nvme_wq);
 			lpfc_cq_destroy(phba, qp->fcp_cq);
 			lpfc_cq_destroy(phba, qp->nvme_cq);
-			if (qidx < phba->cfg_irq_chann)
-				lpfc_eq_destroy(phba, qp->hba_eq);
+		}
+		/* Loop thru all IRQ vectors */
+		for (qidx = 0; qidx < phba->cfg_irq_chann; qidx++) {
+			/* Destroy the EQ corresponding to the IRQ vector */
+			eq = phba->sli4_hba.hba_eq_hdl[qidx].eq;
+			lpfc_eq_destroy(phba, eq);
 		}
 	}
 
···
 }
 
 /**
- * lpfc_find_cpu_handle - Find the CPU that corresponds to the specified EQ
+ * lpfc_find_cpu_handle - Find the CPU that corresponds to the specified Queue
  * @phba: pointer to lpfc hba data structure.
  * @id: EQ vector index or Hardware Queue index
  * @match: LPFC_FIND_BY_EQ = match by EQ
  *         LPFC_FIND_BY_HDWQ = match by Hardware Queue
+ * Return the CPU that matches the selection criteria
  */
 static uint16_t
 lpfc_find_cpu_handle(struct lpfc_hba *phba, uint16_t id, int match)
···
 	struct lpfc_vector_map_info *cpup;
 	int cpu;
 
-	/* Find the desired phys_id for the specified EQ */
+	/* Loop through all CPUs */
 	for_each_present_cpu(cpu) {
 		cpup = &phba->sli4_hba.cpu_map[cpu];
+
+		/* If we are matching by EQ, there may be multiple CPUs using
+		 * using the same vector, so select the one with
+		 * LPFC_CPU_FIRST_IRQ set.
+		 */
 		if ((match == LPFC_FIND_BY_EQ) &&
+		    (cpup->flag & LPFC_CPU_FIRST_IRQ) &&
 		    (cpup->irq != LPFC_VECTOR_MAP_EMPTY) &&
 		    (cpup->eq == id))
 			return cpu;
+
+		/* If matching by HDWQ, select the first CPU that matches */
 		if ((match == LPFC_FIND_BY_HDWQ) && (cpup->hdwq == id))
 			return cpu;
-	}
-	return 0;
-}
-
-/**
- * lpfc_find_eq_handle - Find the EQ that corresponds to the specified
- *                       Hardware Queue
- * @phba: pointer to lpfc hba data structure.
- * @hdwq: Hardware Queue index
- */
-static uint16_t
-lpfc_find_eq_handle(struct lpfc_hba *phba, uint16_t hdwq)
-{
-	struct lpfc_vector_map_info *cpup;
-	int cpu;
-
-	/* Find the desired phys_id for the specified EQ */
-	for_each_present_cpu(cpu) {
-		cpup = &phba->sli4_hba.cpu_map[cpu];
-		if (cpup->hdwq == hdwq)
-			return cpup->eq;
 	}
 	return 0;
 }
···
 
 	/* This loop sets up all CPUs that are affinitized with a
 	 * irq vector assigned to the driver. All affinitized CPUs
-	 * will get a link to that vectors IRQ and EQ. For now we
-	 * are assuming all CPUs using the same EQ will all share
-	 * the same hardware queue.
+	 * will get a link to that vectors IRQ and EQ.
 	 */
 	for (idx = 0; idx <  phba->cfg_irq_chann; idx++) {
+		/* Get a CPU mask for all CPUs affinitized to this vector */
 		maskp = pci_irq_get_affinity(phba->pcidev, idx);
 		if (!maskp)
 			continue;
 
+		i = 0;
+		/* Loop through all CPUs associated with vector idx */
 		for_each_cpu_and(cpu, maskp, cpu_present_mask) {
+			/* Set the EQ index and IRQ for that vector */
 			cpup = &phba->sli4_hba.cpu_map[cpu];
 			cpup->eq = idx;
-			cpup->hdwq = idx;
 			cpup->irq = pci_irq_vector(phba->pcidev, idx);
 
-			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+			lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
 					"3336 Set Affinity: CPU %d "
-					"hdwq %d irq %d\n",
-					cpu, cpup->hdwq, cpup->irq);
+					"irq %d eq %d\n",
+					cpu, cpup->irq, cpup->eq);
+
+			/* If this is the first CPU thats assigned to this
+			 * vector, set LPFC_CPU_FIRST_IRQ.
+			 */
+			if (!i)
+				cpup->flag |= LPFC_CPU_FIRST_IRQ;
+			i++;
 		}
 	}
 
 	/* After looking at each irq vector assigned to this pcidev, its
 	 * possible to see that not ALL CPUs have been accounted for.
-	 * Next we will set any unassigned cpu map entries to a IRQ
-	 * on the same phys_id
+	 * Next we will set any unassigned (unaffinitized) cpu map
+	 * entries to a IRQ on the same phys_id.
 	 */
 	first_cpu = cpumask_first(cpu_present_mask);
 	start_cpu = first_cpu;
···
 			/* Mark CPU as IRQ not assigned by the kernel */
 			cpup->flag |= LPFC_CPU_MAP_UNASSIGN;
 
-			/* If so, find a new_cpup thats on the the same
+			/* If so, find a new_cpup thats on the the SAME
 			 * phys_id as cpup. start_cpu will start where we
 			 * left off so all unassigned entries don't get assgined
 			 * the IRQ of the first entry.
···
 found_same:
 			/* We found a matching phys_id, so copy the IRQ info */
 			cpup->eq = new_cpup->eq;
-			cpup->hdwq = new_cpup->hdwq;
 			cpup->irq = new_cpup->irq;
 
 			/* Bump start_cpu to the next slot to minmize the
···
 			if (start_cpu == nr_cpumask_bits)
 				start_cpu = first_cpu;
 
-			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+			lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
 					"3337 Set Affinity: CPU %d "
-					"hdwq %d irq %d from id %d same "
+					"irq %d from id %d same "
 					"phys_id (%d)\n",
-					cpu, cpup->hdwq, cpup->irq,
-					new_cpu, cpup->phys_id);
+					cpu, cpup->irq, new_cpu, cpup->phys_id);
 		}
 	}
 
···
 			/* Mark it as IRQ not assigned by the kernel */
 			cpup->flag |= LPFC_CPU_MAP_UNASSIGN;
 
-			/* If so, find a new_cpup thats on any phys_id
+			/* If so, find a new_cpup thats on ANY phys_id
 			 * as the cpup. start_cpu will start where we
 			 * left off so all unassigned entries don't get
 			 * assigned the IRQ of the first entry.
···
 			/* We should never leave an entry unassigned */
 			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
 					"3339 Set Affinity: CPU %d "
-					"hdwq %d irq %d UNASSIGNED\n",
-					cpu, cpup->hdwq, cpup->irq);
+					"irq %d UNASSIGNED\n",
+					cpup->hdwq, cpup->irq);
 			continue;
 found_any:
 			/* We found an available entry, copy the IRQ info */
 			cpup->eq = new_cpup->eq;
-			cpup->hdwq = new_cpup->hdwq;
 			cpup->irq = new_cpup->irq;
 
 			/* Bump start_cpu to the next slot to minmize the
···
 			if (start_cpu == nr_cpumask_bits)
 				start_cpu = first_cpu;
 
-			lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+			lpfc_printf_log(phba, KERN_INFO, LOG_INIT,
 					"3338 Set Affinity: CPU %d "
-					"hdwq %d irq %d from id %d (%d/%d)\n",
-					cpu, cpup->hdwq, cpup->irq, new_cpu,
+					"irq %d from id %d (%d/%d)\n",
+					cpu, cpup->irq, new_cpu,
 					new_cpup->phys_id, new_cpup->core_id);
 		}
 	}
+
+	/* Finally we need to associate a hdwq with each cpu_map entry
+	 * This will be 1 to 1 - hdwq to cpu, unless there are less
+	 * hardware queues then CPUs. For that case we will just round-robin
+	 * the available hardware queues as they get assigned to CPUs.
+	 */
+	idx = 0;
+	start_cpu = 0;
+	for_each_present_cpu(cpu) {
+		cpup = &phba->sli4_hba.cpu_map[cpu];
+		if (idx >=  phba->cfg_hdw_queue) {
+			/* We need to reuse a Hardware Queue for another CPU,
+			 * so be smart about it and pick one that has its
+			 * IRQ/EQ mapped to the same phys_id (CPU package).
+			 * and core_id.
+			 */
+			new_cpu = start_cpu;
+			for (i = 0; i < phba->sli4_hba.num_present_cpu; i++) {
+				new_cpup = &phba->sli4_hba.cpu_map[new_cpu];
+				if ((new_cpup->hdwq != LPFC_VECTOR_MAP_EMPTY) &&
+				    (new_cpup->phys_id == cpup->phys_id) &&
+				    (new_cpup->core_id == cpup->core_id))
+					goto found_hdwq;
+				new_cpu = cpumask_next(
+					new_cpu, cpu_present_mask);
+				if (new_cpu == nr_cpumask_bits)
+					new_cpu = first_cpu;
+			}
+
+			/* If we can't match both phys_id and core_id,
+			 * settle for just a phys_id match.
+			 */
+			new_cpu = start_cpu;
+			for (i = 0; i < phba->sli4_hba.num_present_cpu; i++) {
+				new_cpup = &phba->sli4_hba.cpu_map[new_cpu];
+				if ((new_cpup->hdwq != LPFC_VECTOR_MAP_EMPTY) &&
+				    (new_cpup->phys_id == cpup->phys_id))
+					goto found_hdwq;
+				new_cpu = cpumask_next(
+					new_cpu, cpu_present_mask);
+				if (new_cpu == nr_cpumask_bits)
+					new_cpu = first_cpu;
+			}
+
+			/* Otherwise just round robin on cfg_hdw_queue */
+			cpup->hdwq = idx % phba->cfg_hdw_queue;
+			goto logit;
+found_hdwq:
+			/* We found an available entry, copy the IRQ info */
+			start_cpu = cpumask_next(new_cpu, cpu_present_mask);
+			if (start_cpu == nr_cpumask_bits)
+				start_cpu = first_cpu;
+			cpup->hdwq = new_cpup->hdwq;
+		} else {
+			/* 1 to 1, CPU to hdwq */
+			cpup->hdwq = idx;
+		}
+logit:
+		lpfc_printf_log(phba, KERN_ERR, LOG_INIT,
+				"3335 Set Affinity: CPU %d (phys %d core %d): "
+				"hdwq %d eq %d irq %d flg x%x\n",
+				cpu, cpup->phys_id, cpup->core_id,
+				cpup->hdwq, cpup->eq, cpup->irq, cpup->flag);
+		idx++;
+	}
+
+	/* The cpu_map array will be used later during initialization
+	 * when EQ / CQ / WQs are allocated and configured.
+	 */
 	return;
 }
 

drivers/scsi/lpfc/lpfc_sli.c

···
 	int qidx;
 	struct lpfc_sli4_hba *sli4_hba = &phba->sli4_hba;
 	struct lpfc_sli4_hdw_queue *qp;
+	struct lpfc_queue *eq;
 
 	sli4_hba->sli4_write_cq_db(phba, sli4_hba->mbx_cq, 0, LPFC_QUEUE_REARM);
 	sli4_hba->sli4_write_cq_db(phba, sli4_hba->els_cq, 0, LPFC_QUEUE_REARM);
···
 		sli4_hba->sli4_write_cq_db(phba, sli4_hba->nvmels_cq, 0,
 					   LPFC_QUEUE_REARM);
 
-	qp = sli4_hba->hdwq;
 	if (sli4_hba->hdwq) {
+		/* Loop thru all Hardware Queues */
 		for (qidx = 0; qidx < phba->cfg_hdw_queue; qidx++) {
-			sli4_hba->sli4_write_cq_db(phba, qp[qidx].fcp_cq, 0,
+			qp = &sli4_hba->hdwq[qidx];
+			/* ARM the corresponding CQ */
+			sli4_hba->sli4_write_cq_db(phba, qp->fcp_cq, 0,
 						   LPFC_QUEUE_REARM);
-			sli4_hba->sli4_write_cq_db(phba, qp[qidx].nvme_cq, 0,
+			sli4_hba->sli4_write_cq_db(phba, qp->nvme_cq, 0,
 						   LPFC_QUEUE_REARM);
 		}
 
-		for (qidx = 0; qidx < phba->cfg_irq_chann; qidx++)
-			sli4_hba->sli4_write_eq_db(phba, qp[qidx].hba_eq,
-						0, LPFC_QUEUE_REARM);
+		/* Loop thru all IRQ vectors */
+		for (qidx = 0; qidx < phba->cfg_irq_chann; qidx++) {
+			eq = sli4_hba->hba_eq_hdl[qidx].eq;
+			/* ARM the corresponding EQ */
+			sli4_hba->sli4_write_eq_db(phba, eq,
+						   0, LPFC_QUEUE_REARM);
+		}
 	}
 
 	if (phba->nvmet_support) {
···
 	struct lpfc_sli4_hba *sli4_hba = &phba->sli4_hba;
 	uint32_t eqidx;
 	struct lpfc_queue *fpeq = NULL;
+	struct lpfc_queue *eq;
 	bool mbox_pending;
 
 	if (unlikely(!phba) || (phba->sli_rev != LPFC_SLI_REV4))
 		return false;
 
-	/* Find the eq associated with the mcq */
-
-	if (sli4_hba->hdwq)
-		for (eqidx = 0; eqidx < phba->cfg_irq_chann; eqidx++)
-			if (sli4_hba->hdwq[eqidx].hba_eq->queue_id ==
-			    sli4_hba->mbx_cq->assoc_qid) {
-				fpeq = sli4_hba->hdwq[eqidx].hba_eq;
+	/* Find the EQ associated with the mbox CQ */
+	if (sli4_hba->hdwq) {
+		for (eqidx = 0; eqidx < phba->cfg_irq_chann; eqidx++) {
+			eq = phba->sli4_hba.hba_eq_hdl[eqidx].eq;
+			if (eq->queue_id == sli4_hba->mbx_cq->assoc_qid) {
+				fpeq = eq;
 				break;
 			}
+		}
+	}
 	if (!fpeq)
 		return false;
 
···
 		return IRQ_NONE;
 
 	/* Get to the EQ struct associated with this vector */
-	fpeq = phba->sli4_hba.hdwq[hba_eqidx].hba_eq;
+	fpeq = phba->sli4_hba.hba_eq_hdl[hba_eqidx].eq;
 	if (unlikely(!fpeq))
 		return IRQ_NONE;
 
···
 	/* set values by EQ_DELAY register if supported */
 	if (phba->sli.sli_flag & LPFC_SLI_USE_EQDR) {
 		for (qidx = startq; qidx < phba->cfg_irq_chann; qidx++) {
-			eq = phba->sli4_hba.hdwq[qidx].hba_eq;
+			eq = phba->sli4_hba.hba_eq_hdl[qidx].eq;
 			if (!eq)
 				continue;
 
···
 			if (++cnt >= numq)
 				break;
 		}
-
 		return;
 	}
 
···
 		dmult = LPFC_DMULT_MAX;
 
 	for (qidx = startq; qidx < phba->cfg_irq_chann; qidx++) {
-		eq = phba->sli4_hba.hdwq[qidx].hba_eq;
+		eq = phba->sli4_hba.hba_eq_hdl[qidx].eq;
 		if (!eq)
 			continue;
 		eq->q_mode = usdelay;

drivers/scsi/lpfc/lpfc_sli4.h

···
 	uint32_t idx;
 	char handler_name[LPFC_SLI4_HANDLER_NAME_SZ];
 	struct lpfc_hba *phba;
+	struct lpfc_queue *eq;
 };
 
 /*BB Credit recovery value*/
···
 	uint16_t	flag;
 #define LPFC_CPU_MAP_HYPER	0x1
 #define LPFC_CPU_MAP_UNASSIGN	0x2
+#define LPFC_CPU_FIRST_IRQ	0x4
 };
 #define LPFC_VECTOR_MAP_EMPTY	0xffff