diff --git a/qla2x00t-32gbit/qla_def.h b/qla2x00t-32gbit/qla_def.h
index ffd0754e8..e25f8ea2b 100644
--- a/qla2x00t-32gbit/qla_def.h
+++ b/qla2x00t-32gbit/qla_def.h
@@ -3515,6 +3515,7 @@ struct qla_msix_entry {
 	int have_irq;
 	int in_use;
 	uint32_t vector;
+	uint32_t vector_base0;
 	uint16_t entry;
 	char name[30];
 	void *handle;
@@ -4171,6 +4172,7 @@ struct qla_hw_data {
 	struct req_que **req_q_map;
 	struct rsp_que **rsp_q_map;
 	struct qla_qpair **queue_pair_map;
+	struct qla_qpair **qp_cpu_map;
 	unsigned long req_qid_map[(QLA_MAX_QUEUES / 8) / sizeof(unsigned long)];
 	unsigned long rsp_qid_map[(QLA_MAX_QUEUES / 8) / sizeof(unsigned long)];
 	unsigned long qpair_qid_map[(QLA_MAX_QUEUES / 8)
diff --git a/qla2x00t-32gbit/qla_init.c b/qla2x00t-32gbit/qla_init.c
index f56c98b25..46d4d7529 100644
--- a/qla2x00t-32gbit/qla_init.c
+++ b/qla2x00t-32gbit/qla_init.c
@@ -9429,8 +9429,6 @@ struct qla_qpair *qla2xxx_create_qpair(struct scsi_qla_host *vha, int qos,
 		qpair->req = ha->req_q_map[req_id];
 		qpair->rsp->req = qpair->req;
 		qpair->rsp->qpair = qpair;
-		/* init qpair to this cpu. Will adjust at run time. */
-		qla_cpu_update(qpair, raw_smp_processor_id());
 
 		if (IS_T10_PI_CAPABLE(ha) && ql2xenabledif) {
 			if (ha->fw_attributes & BIT_4)
diff --git a/qla2x00t-32gbit/qla_inline.h b/qla2x00t-32gbit/qla_inline.h
index d871dd9eb..7dd8f05c3 100644
--- a/qla2x00t-32gbit/qla_inline.h
+++ b/qla2x00t-32gbit/qla_inline.h
@@ -523,3 +523,58 @@ fcport_is_bigger(fc_port_t *fcport)
 {
 	return !fcport_is_smaller(fcport);
 }
+
+static inline struct qla_qpair *
+qla_mapq_nvme_select_qpair(struct qla_hw_data *ha, struct qla_qpair *qpair)
+{
+	int cpuid = smp_processor_id();
+
+	if (qpair->cpuid != cpuid &&
+	    ha->qp_cpu_map[cpuid]) {
+		qpair = ha->qp_cpu_map[cpuid];
+	}
+	return qpair;
+}
+
+static inline void
+qla_mapq_init_qp_cpu_map(struct qla_hw_data *ha,
+			 struct qla_msix_entry *msix,
+			 struct qla_qpair *qpair)
+{
+	const struct cpumask *mask;
+	unsigned int cpu;
+
+	if (!ha->qp_cpu_map)
+		return;
+	mask = pci_irq_get_affinity(ha->pdev, msix->vector_base0);
+	qpair->cpuid = cpumask_first(mask);
+	for_each_cpu(cpu, mask) {
+		ha->qp_cpu_map[cpu] = qpair;
+	}
+	msix->cpuid = qpair->cpuid;
+}
+
+static inline void
+qla_mapq_free_qp_cpu_map(struct qla_hw_data *ha)
+{
+	if (ha->qp_cpu_map) {
+		kfree(ha->qp_cpu_map);
+		ha->qp_cpu_map = NULL;
+	}
+}
+
+static inline int qla_mapq_alloc_qp_cpu_map(struct qla_hw_data *ha)
+{
+	scsi_qla_host_t *vha = pci_get_drvdata(ha->pdev);
+
+	if (!ha->qp_cpu_map) {
+		ha->qp_cpu_map = kcalloc(NR_CPUS, sizeof(struct qla_qpair *),
+					 GFP_KERNEL);
+		if (!ha->qp_cpu_map) {
+			ql_log(ql_log_fatal, vha, 0x0180,
+			       "Unable to allocate memory for qp_cpu_map ptrs.\n");
+			return -1;
+		}
+	}
+	return 0;
+}
diff --git a/qla2x00t-32gbit/qla_isr.c b/qla2x00t-32gbit/qla_isr.c
index dc5df99a0..e9df2ef3d 100644
--- a/qla2x00t-32gbit/qla_isr.c
+++ b/qla2x00t-32gbit/qla_isr.c
@@ -3793,7 +3793,6 @@ void qla24xx_process_response_queue(struct scsi_qla_host *vha,
 	if (rsp->qpair->cpuid != raw_smp_processor_id() ||
 	    !rsp->qpair->rcv_intr) {
 		rsp->qpair->rcv_intr = 1;
-		qla_cpu_update(rsp->qpair, raw_smp_processor_id());
 	}
 
 #define __update_rsp_in(_is_shadow_hba, _rsp, _rsp_in)			\
@@ -4480,6 +4479,7 @@ msix_failed:
 		qentry->vector = entries[i].vector;
 		qentry->entry = entries[i].entry;
 #endif
+		qentry->vector_base0 = i;
 		qentry->have_irq = 0;
 		qentry->in_use = 0;
 		qentry->handle = NULL;
@@ -4737,5 +4737,6 @@ int qla25xx_request_irq(struct qla_hw_data *ha, struct qla_qpair *qpair,
 	}
 	msix->have_irq = 1;
 	msix->handle = qpair;
+	qla_mapq_init_qp_cpu_map(ha, msix, qpair);
 	return ret;
 }
diff --git a/qla2x00t-32gbit/qla_nvme.c b/qla2x00t-32gbit/qla_nvme.c
index a04594c5f..de6948680 100644
--- a/qla2x00t-32gbit/qla_nvme.c
+++ b/qla2x00t-32gbit/qla_nvme.c
@@ -624,6 +624,7 @@ static int qla_nvme_post_cmd(struct nvme_fc_local_port *lport,
 	fc_port_t *fcport;
 	struct srb_iocb *nvme;
 	struct scsi_qla_host *vha;
+	struct qla_hw_data *ha;
 	int rval;
 	srb_t *sp;
 	struct qla_qpair *qpair = hw_queue_handle;
@@ -644,6 +645,7 @@ static int qla_nvme_post_cmd(struct nvme_fc_local_port *lport,
 		return -ENODEV;
 
 	vha = fcport->vha;
+	ha = vha->hw;
 
 	if (test_bit(ABORT_ISP_ACTIVE, &vha->dpc_flags))
 		return -EBUSY;
@@ -658,6 +660,8 @@ static int qla_nvme_post_cmd(struct nvme_fc_local_port *lport,
 	if (fcport->nvme_flag & NVME_FLAG_RESETTING)
 		return -EBUSY;
 
+	qpair = qla_mapq_nvme_select_qpair(ha, qpair);
+
 	/* Alloc SRB structure */
 	sp = qla2xxx_get_qpair_sp(vha, qpair, fcport, GFP_ATOMIC);
 	if (!sp)
diff --git a/qla2x00t-32gbit/qla_os.c b/qla2x00t-32gbit/qla_os.c
index 26324f5d2..ff722e505 100644
--- a/qla2x00t-32gbit/qla_os.c
+++ b/qla2x00t-32gbit/qla_os.c
@@ -509,6 +509,11 @@ static int qla2x00_alloc_queues(struct qla_hw_data *ha, struct req_que *req,
 			    "Unable to allocate memory for queue pair ptrs.\n");
 			goto fail_qpair_map;
 		}
+		if (qla_mapq_alloc_qp_cpu_map(ha) != 0) {
+			kfree(ha->queue_pair_map);
+			ha->queue_pair_map = NULL;
+			goto fail_qpair_map;
+		}
 	}
 
 	/*
@@ -583,6 +588,7 @@ static void qla2x00_free_queues(struct qla_hw_data *ha)
 		ha->base_qpair = NULL;
 	}
 
+	qla_mapq_free_qp_cpu_map(ha);
 	spin_lock_irqsave(&ha->hardware_lock, flags);
 	for (cnt = 0; cnt < ha->max_req_queues; cnt++) {
 		if (!test_bit(cnt, ha->req_qid_map))