scsi: megaraid_sas: Load balance completions across all MSI-X

Driver will use "reply descriptor post queues" in round robin fashion when
the combined MSI-X mode is not enabled. With this IO completions are
distributed and load balanced across all the available reply descriptor
post queues equally.

This is enabled only if combined MSI-X mode is not enabled in firmware.
This improves performance and also fixes soft lockups.

When load balancing is enabled, IRQ affinity from driver needs to be
disabled.

Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Shivasharan S <shivasharan.srikanteshwara@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

3 files changed, 35 insertions, 8 deletions

diff --git a/drivers/scsi/megaraid/megaraid_sas.h b/drivers/scsi/megaraid/megaraid_sas.h
index 8d6a9c511455..6be748f302cf 100644
--- a/drivers/scsi/megaraid/megaraid_sas.h
+++ b/drivers/scsi/megaraid/megaraid_sas.h
@@ -2262,6 +2262,7 @@ struct megasas_instance {
 	u32 secure_jbod_support;
 	u32 support_morethan256jbod; /* FW support for more than 256 PD/JBOD */
 	bool use_seqnum_jbod_fp;   /* Added for PD sequence */
+	bool smp_affinity_enable;
 	spinlock_t crashdump_lock;
 
 	struct megasas_register_set __iomem *reg_set;
@@ -2279,6 +2280,7 @@ struct megasas_instance {
 	u16 ldio_threshold;
 	u16 cur_can_queue;
 	u32 max_sectors_per_req;
+	bool msix_load_balance;
 	struct megasas_aen_event *ev;
 
 	struct megasas_cmd **cmd_list;
@@ -2316,6 +2318,7 @@ struct megasas_instance {
 	atomic_t sge_holes_type1;
 	atomic_t sge_holes_type2;
 	atomic_t sge_holes_type3;
+	atomic64_t total_io_count;
 
 	struct megasas_instance_template *instancet;
 	struct tasklet_struct isr_tasklet;
diff --git a/drivers/scsi/megaraid/megaraid_sas_base.c b/drivers/scsi/megaraid/megaraid_sas_base.c
index ec1259f69683..6d2390bf7271 100644
--- a/drivers/scsi/megaraid/megaraid_sas_base.c
+++ b/drivers/scsi/megaraid/megaraid_sas_base.c
@@ -5349,6 +5349,7 @@ megasas_setup_irqs_msix(struct megasas_instance *instance, u8 is_probe)
 					 &instance->irq_context[j]);
 			/* Retry irq register for IO_APIC*/
 			instance->msix_vectors = 0;
+			instance->msix_load_balance = false;
 			if (is_probe) {
 				pci_free_irq_vectors(instance->pdev);
 				return megasas_setup_irqs_ioapic(instance);
@@ -5357,6 +5358,7 @@ megasas_setup_irqs_msix(struct megasas_instance *instance, u8 is_probe)
 			}
 		}
 	}
+
 	return 0;
 }
 
@@ -5660,6 +5662,12 @@ static int megasas_init_fw(struct megasas_instance *instance)
 				if (rdpq_enable)
 					instance->is_rdpq = (scratch_pad_1 & MR_RDPQ_MODE_OFFSET) ?
 								1 : 0;
+
+				if (!instance->msix_combined) {
+					instance->msix_load_balance = true;
+					instance->smp_affinity_enable = false;
+				}
+
 				fw_msix_count = instance->msix_vectors;
 				/* Save 1-15 reply post index address to local memory
 				 * Index 0 is already saved from reg offset
@@ -5678,17 +5686,20 @@ static int megasas_init_fw(struct megasas_instance *instance)
 					instance->msix_vectors);
 		} else /* MFI adapters */
 			instance->msix_vectors = 1;
+
 		/* Don't bother allocating more MSI-X vectors than cpus */
 		instance->msix_vectors = min(instance->msix_vectors,
 					     (unsigned int)num_online_cpus());
-		if (smp_affinity_enable)
+		if (instance->smp_affinity_enable)
 			irq_flags |= PCI_IRQ_AFFINITY;
 		i = pci_alloc_irq_vectors(instance->pdev, 1,
 					  instance->msix_vectors, irq_flags);
-		if (i > 0)
+		if (i > 0) {
 			instance->msix_vectors = i;
-		else
+		} else {
 			instance->msix_vectors = 0;
+			instance->msix_load_balance = false;
+		}
 	}
 	/*
 	 * MSI-X host index 0 is common for all adapter.
@@ -6795,6 +6806,7 @@ static inline void megasas_init_ctrl_params(struct megasas_instance *instance)
 	INIT_LIST_HEAD(&instance->internal_reset_pending_q);
 
 	atomic_set(&instance->fw_outstanding, 0);
+	atomic64_set(&instance->total_io_count, 0);
 
 	init_waitqueue_head(&instance->int_cmd_wait_q);
 	init_waitqueue_head(&instance->abort_cmd_wait_q);
@@ -6817,6 +6829,8 @@ static inline void megasas_init_ctrl_params(struct megasas_instance *instance)
 	instance->last_time = 0;
 	instance->disableOnlineCtrlReset = 1;
 	instance->UnevenSpanSupport = 0;
+	instance->smp_affinity_enable = smp_affinity_enable ? true : false;
+	instance->msix_load_balance = false;
 
 	if (instance->adapter_type != MFI_SERIES)
 		INIT_WORK(&instance->work_init, megasas_fusion_ocr_wq);
@@ -7180,7 +7194,7 @@ megasas_resume(struct pci_dev *pdev)
 	/* Now re-enable MSI-X */
 	if (instance->msix_vectors) {
 		irq_flags = PCI_IRQ_MSIX;
-		if (smp_affinity_enable)
+		if (instance->smp_affinity_enable)
 			irq_flags |= PCI_IRQ_AFFINITY;
 	}
 	rval = pci_alloc_irq_vectors(instance->pdev, 1,
diff --git a/drivers/scsi/megaraid/megaraid_sas_fusion.c b/drivers/scsi/megaraid/megaraid_sas_fusion.c
index 5fdba1556039..66306ffb34a5 100644
--- a/drivers/scsi/megaraid/megaraid_sas_fusion.c
+++ b/drivers/scsi/megaraid/megaraid_sas_fusion.c
@@ -2768,8 +2768,13 @@ megasas_build_ldio_fusion(struct megasas_instance *instance,
 			fp_possible = (io_info.fpOkForIo > 0) ? true : false;
 	}
 
-	cmd->request_desc->SCSIIO.MSIxIndex =
-		instance->reply_map[raw_smp_processor_id()];
+	if (instance->msix_load_balance)
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			(mega_mod64(atomic64_add_return(1, &instance->total_io_count),
+				    instance->msix_vectors));
+	else
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			instance->reply_map[raw_smp_processor_id()];
 
 	if (instance->adapter_type >= VENTURA_SERIES) {
 		/* FP for Optimal raid level 1.
@@ -3082,8 +3087,13 @@ megasas_build_syspd_fusion(struct megasas_instance *instance,
 
 	cmd->request_desc->SCSIIO.DevHandle = io_request->DevHandle;
 
-	cmd->request_desc->SCSIIO.MSIxIndex =
-		instance->reply_map[raw_smp_processor_id()];
+	if (instance->msix_load_balance)
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			(mega_mod64(atomic64_add_return(1, &instance->total_io_count),
+				    instance->msix_vectors));
+	else
+		cmd->request_desc->SCSIIO.MSIxIndex =
+			instance->reply_map[raw_smp_processor_id()];
 
 	if (!fp_possible) {
 		/* system pd firmware path */