scsi: megaraid_sas: mq_poll support

Implement mq_poll interface support in megaraid_sas. This feature
requires shared host tag support in kernel and driver.

The driver can work in non-IRQ mode which means there will not be any MSI-x
vector associated for poll_queues. The MegaRAID hardware has a single
submission queue and multiple reply queues. However, using the shared host
tagset support will enable the driver to simulate multiple hardware queues.

Change driver to allocate some extra reply queues which will be marked as
poll_queues. These poll_queues will not have associated MSI-x vectors. All
I/O completions on these queues will be done through the IOPOLL interface.

megaraid_sas with 8 poll_queues and using the io_uring hiprio=1 setting can
reach 3.2M IOPS with zero interrupts generated by the hardware.

The IOPOLL feature can be enabled using module parameter poll_queues.

Link: https://lore.kernel.org/r/20210215074048.19424-3-kashyap.desai@broadcom.com
Cc: sumit.saxena@broadcom.com
Cc: chandrakanth.patil@broadcom.com
Cc: linux-block@vger.kernel.org
Reviewed-by: Hannes Reinecke <hare@suse.de>
Signed-off-by: Kashyap Desai <kashyap.desai@broadcom.com>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>

drivers/scsi/megaraid/megaraid_sas.h

@@ -2214,6 +2214,7 @@ struct megasas_irq_context {
 	struct irq_poll irqpoll;
 	bool irq_poll_scheduled;
 	bool irq_line_enable;
+	atomic_t   in_used;
 };
 
 struct MR_DRV_SYSTEM_INFO {
@@ -2448,6 +2449,7 @@ struct megasas_instance {
 	bool support_pci_lane_margining;
 	u8  low_latency_index_start;
 	int perf_mode;
+	int iopoll_q_count;
 };
 
 struct MR_LD_VF_MAP {
@@ -2728,5 +2730,6 @@ void megasas_init_debugfs(void);
 void megasas_exit_debugfs(void);
 void megasas_setup_debugfs(struct megasas_instance *instance);
 void megasas_destroy_debugfs(struct megasas_instance *instance);
+int megasas_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num);
 
 #endif				/*LSI_MEGARAID_SAS_H */

drivers/scsi/megaraid/megaraid_sas_base.c

@@ -114,6 +114,15 @@ unsigned int enable_sdev_max_qd;
 module_param(enable_sdev_max_qd, int, 0444);
 MODULE_PARM_DESC(enable_sdev_max_qd, "Enable sdev max qd as can_queue. Default: 0");
 
+int poll_queues;
+module_param(poll_queues, int, 0444);
+MODULE_PARM_DESC(poll_queues, "Number of queues to be use for io_uring poll mode.\n\t\t"
+		"This parameter is effective only if host_tagset_enable=1 &\n\t\t"
+		"It is not applicable for MFI_SERIES. &\n\t\t"
+		"Driver will work in latency mode. &\n\t\t"
+		"High iops queues are not allocated &\n\t\t"
+		);
+
 int host_tagset_enable = 1;
 module_param(host_tagset_enable, int, 0444);
 MODULE_PARM_DESC(host_tagset_enable, "Shared host tagset enable/disable Default: enable(1)");
@@ -207,6 +216,7 @@ static bool support_pci_lane_margining;
 static spinlock_t poll_aen_lock;
 
 extern struct dentry *megasas_debugfs_root;
+extern int megasas_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num);
 
 void
 megasas_complete_cmd(struct megasas_instance *instance, struct megasas_cmd *cmd,
@@ -3127,14 +3137,37 @@ megasas_bios_param(struct scsi_device *sdev, struct block_device *bdev,
 static int megasas_map_queues(struct Scsi_Host *shost)
 {
 	struct megasas_instance *instance;
+	int qoff = 0, offset;
+	struct blk_mq_queue_map *map;
 
 	instance = (struct megasas_instance *)shost->hostdata;
 
 	if (shost->nr_hw_queues == 1)
 		return 0;
 
-	return blk_mq_pci_map_queues(&shost->tag_set.map[HCTX_TYPE_DEFAULT],
-			instance->pdev, instance->low_latency_index_start);
+	offset = instance->low_latency_index_start;
+
+	/* Setup Default hctx */
+	map = &shost->tag_set.map[HCTX_TYPE_DEFAULT];
+	map->nr_queues = instance->msix_vectors - offset;
+	map->queue_offset = 0;
+	blk_mq_pci_map_queues(map, instance->pdev, offset);
+	qoff += map->nr_queues;
+	offset += map->nr_queues;
+
+	/* Setup Poll hctx */
+	map = &shost->tag_set.map[HCTX_TYPE_POLL];
+	map->nr_queues = instance->iopoll_q_count;
+	if (map->nr_queues) {
+		/*
+		 * The poll queue(s) doesn't have an IRQ (and hence IRQ
+		 * affinity), so use the regular blk-mq cpu mapping
+		 */
+		map->queue_offset = qoff;
+		blk_mq_map_queues(map);
+	}
+
+	return 0;
 }
 
 static void megasas_aen_polling(struct work_struct *work);
@@ -3446,6 +3479,7 @@ static struct scsi_host_template megasas_template = {
 	.shost_attrs = megaraid_host_attrs,
 	.bios_param = megasas_bios_param,
 	.map_queues = megasas_map_queues,
+	.mq_poll = megasas_blk_mq_poll,
 	.change_queue_depth = scsi_change_queue_depth,
 	.max_segment_size = 0xffffffff,
 };
@@ -5834,13 +5868,16 @@ __megasas_alloc_irq_vectors(struct megasas_instance *instance)
 	irq_flags = PCI_IRQ_MSIX;
 
 	if (instance->smp_affinity_enable)
-		irq_flags |= PCI_IRQ_AFFINITY;
+		irq_flags |= PCI_IRQ_AFFINITY | PCI_IRQ_ALL_TYPES;
 	else
 		descp = NULL;
 
+	/* Do not allocate msix vectors for poll_queues.
+	 * msix_vectors is always within a range of FW supported reply queue.
+	 */
 	i = pci_alloc_irq_vectors_affinity(instance->pdev,
 		instance->low_latency_index_start,
-		instance->msix_vectors, irq_flags, descp);
+		instance->msix_vectors - instance->iopoll_q_count, irq_flags, descp);
 
 	return i;
 }
@@ -5856,10 +5893,30 @@ megasas_alloc_irq_vectors(struct megasas_instance *instance)
 	int i;
 	unsigned int num_msix_req;
 
+	instance->iopoll_q_count = 0;
+	if ((instance->adapter_type != MFI_SERIES) &&
+		poll_queues) {
+
+		instance->perf_mode = MR_LATENCY_PERF_MODE;
+		instance->low_latency_index_start = 1;
+
+		/* reserve for default and non-mananged pre-vector. */
+		if (instance->msix_vectors > (poll_queues + 2))
+			instance->iopoll_q_count = poll_queues;
+		else
+			instance->iopoll_q_count = 0;
+
+		num_msix_req = num_online_cpus() + instance->low_latency_index_start;
+		instance->msix_vectors = min(num_msix_req,
+				instance->msix_vectors);
+
+	}
+
 	i = __megasas_alloc_irq_vectors(instance);
 
-	if ((instance->perf_mode == MR_BALANCED_PERF_MODE) &&
-	    (i != instance->msix_vectors)) {
+	if (((instance->perf_mode == MR_BALANCED_PERF_MODE)
+		|| instance->iopoll_q_count) &&
+	    (i != (instance->msix_vectors - instance->iopoll_q_count))) {
 		if (instance->msix_vectors)
 			pci_free_irq_vectors(instance->pdev);
 		/* Disable Balanced IOPS mode and try realloc vectors */
@@ -5870,12 +5927,15 @@ megasas_alloc_irq_vectors(struct megasas_instance *instance)
 		instance->msix_vectors = min(num_msix_req,
 				instance->msix_vectors);
 
+		instance->iopoll_q_count = 0;
 		i = __megasas_alloc_irq_vectors(instance);
 
 	}
 
 	dev_info(&instance->pdev->dev,
-		"requested/available msix %d/%d\n", instance->msix_vectors, i);
+		"requested/available msix %d/%d poll_queue %d\n",
+			instance->msix_vectors - instance->iopoll_q_count,
+			i, instance->iopoll_q_count);
 
 	if (i > 0)
 		instance->msix_vectors = i;
@@ -6841,12 +6901,18 @@ static int megasas_io_attach(struct megasas_instance *instance)
 		instance->smp_affinity_enable) {
 		host->host_tagset = 1;
 		host->nr_hw_queues = instance->msix_vectors -
-			instance->low_latency_index_start;
+			instance->low_latency_index_start + instance->iopoll_q_count;
+		if (instance->iopoll_q_count)
+			host->nr_maps = 3;
+	} else {
+		instance->iopoll_q_count = 0;
 	}
 
 	dev_info(&instance->pdev->dev,
-		"Max firmware commands: %d shared with nr_hw_queues = %d\n",
-		instance->max_fw_cmds, host->nr_hw_queues);
+		"Max firmware commands: %d shared with default "
+		"hw_queues = %d poll_queues %d\n", instance->max_fw_cmds,
+		host->nr_hw_queues - instance->iopoll_q_count,
+		instance->iopoll_q_count);
 	/*
 	 * Notify the mid-layer about the new controller
 	 */
@@ -8859,6 +8925,7 @@ static int __init megasas_init(void)
 		msix_vectors = 1;
 		rdpq_enable = 0;
 		dual_qdepth_disable = 1;
+		poll_queues = 0;
 	}
 
 	/*

drivers/scsi/megaraid/megaraid_sas_fusion.c

@@ -713,6 +713,8 @@ megasas_alloc_reply_fusion(struct megasas_instance *instance)
 	fusion = instance->ctrl_context;
 
 	count = instance->msix_vectors > 0 ? instance->msix_vectors : 1;
+	count += instance->iopoll_q_count;
+
 	fusion->reply_frames_desc_pool =
 			dma_pool_create("mr_reply", &instance->pdev->dev,
 				fusion->reply_alloc_sz * count, 16, 0);
@@ -807,6 +809,7 @@ megasas_alloc_rdpq_fusion(struct megasas_instance *instance)
 	}
 
 	msix_count = instance->msix_vectors > 0 ? instance->msix_vectors : 1;
+	msix_count += instance->iopoll_q_count;
 
 	fusion->reply_frames_desc_pool = dma_pool_create("mr_rdpq",
 							 &instance->pdev->dev,
@@ -1157,7 +1160,7 @@ megasas_ioc_init_fusion(struct megasas_instance *instance)
 			MPI2_IOCINIT_MSGFLAG_RDPQ_ARRAY_MODE : 0;
 	IOCInitMessage->SystemRequestFrameBaseAddress = cpu_to_le64(fusion->io_request_frames_phys);
 	IOCInitMessage->SenseBufferAddressHigh = cpu_to_le32(upper_32_bits(fusion->sense_phys_addr));
-	IOCInitMessage->HostMSIxVectors = instance->msix_vectors;
+	IOCInitMessage->HostMSIxVectors = instance->msix_vectors + instance->iopoll_q_count;
 	IOCInitMessage->HostPageSize = MR_DEFAULT_NVME_PAGE_SHIFT;
 
 	time = ktime_get_real();
@@ -1851,6 +1854,8 @@ megasas_init_adapter_fusion(struct megasas_instance *instance)
 		 sizeof(union MPI2_SGE_IO_UNION))/16;
 
 	count = instance->msix_vectors > 0 ? instance->msix_vectors : 1;
+	count += instance->iopoll_q_count;
+
 	for (i = 0 ; i < count; i++)
 		fusion->last_reply_idx[i] = 0;
 
@@ -1863,6 +1868,9 @@ megasas_init_adapter_fusion(struct megasas_instance *instance)
 				MEGASAS_FUSION_IOCTL_CMDS);
 	sema_init(&instance->ioctl_sem, MEGASAS_FUSION_IOCTL_CMDS);
 
+	for (i = 0; i < MAX_MSIX_QUEUES_FUSION; i++)
+		atomic_set(&fusion->busy_mq_poll[i], 0);
+
 	if (megasas_alloc_ioc_init_frame(instance))
 		return 1;
 
@@ -3530,6 +3538,9 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex,
 	if (reply_descript_type == MPI2_RPY_DESCRIPT_FLAGS_UNUSED)
 		return IRQ_NONE;
 
+	if (irq_context && !atomic_add_unless(&irq_context->in_used, 1, 1))
+		return 0;
+
 	num_completed = 0;
 
 	while (d_val.u.low != cpu_to_le32(UINT_MAX) &&
@@ -3644,6 +3655,7 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex,
 					irq_context->irq_line_enable = true;
 					irq_poll_sched(&irq_context->irqpoll);
 				}
+				atomic_dec(&irq_context->in_used);
 				return num_completed;
 			}
 		}
@@ -3661,9 +3673,35 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex,
 				instance->reply_post_host_index_addr[0]);
 		megasas_check_and_restore_queue_depth(instance);
 	}
+
+	if (irq_context)
+		atomic_dec(&irq_context->in_used);
+
 	return num_completed;
 }
 
+int megasas_blk_mq_poll(struct Scsi_Host *shost, unsigned int queue_num)
+{
+
+	struct megasas_instance *instance;
+	int num_entries = 0;
+	struct fusion_context *fusion;
+
+	instance = (struct megasas_instance *)shost->hostdata;
+
+	fusion = instance->ctrl_context;
+
+	queue_num = queue_num + instance->low_latency_index_start;
+
+	if (!atomic_add_unless(&fusion->busy_mq_poll[queue_num], 1, 1))
+		return 0;
+
+	num_entries = complete_cmd_fusion(instance, queue_num, NULL);
+	atomic_dec(&fusion->busy_mq_poll[queue_num]);
+
+	return num_entries;
+}
+
 /**
  * megasas_enable_irq_poll() - enable irqpoll
  * @instance:			Adapter soft state
@@ -4194,6 +4232,8 @@ void  megasas_reset_reply_desc(struct megasas_instance *instance)
 
 	fusion = instance->ctrl_context;
 	count = instance->msix_vectors > 0 ? instance->msix_vectors : 1;
+	count += instance->iopoll_q_count;
+
 	for (i = 0 ; i < count ; i++) {
 		fusion->last_reply_idx[i] = 0;
 		reply_desc = fusion->reply_frames_desc[i];

drivers/scsi/megaraid/megaraid_sas_fusion.h

@@ -1303,6 +1303,8 @@ struct fusion_context {
 	u8 *sense;
 	dma_addr_t sense_phys_addr;
 
+	atomic_t   busy_mq_poll[MAX_MSIX_QUEUES_FUSION];
+
 	dma_addr_t reply_frames_desc_phys[MAX_MSIX_QUEUES_FUSION];
 	union MPI2_REPLY_DESCRIPTORS_UNION *reply_frames_desc[MAX_MSIX_QUEUES_FUSION];
 	struct rdpq_alloc_detail rdpq_tracker[RDPQ_MAX_CHUNK_COUNT];