[SCSI] megaraid_sas: Add High Availability clustering support using shared Logical Disks

Signed-off-by: Adam Radford <aradford@gmail.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

drivers/scsi/megaraid/megaraid_sas.h

@@ -170,6 +170,7 @@
 
 #define MR_DCMD_CTRL_GET_INFO			0x01010000
 #define MR_DCMD_LD_GET_LIST			0x03010000
+#define MR_DCMD_LD_LIST_QUERY			0x03010100
 
 #define MR_DCMD_CTRL_CACHE_FLUSH		0x01101000
 #define MR_FLUSH_CTRL_CACHE			0x01
@@ -345,6 +346,15 @@ enum MR_PD_QUERY_TYPE {
 	MR_PD_QUERY_TYPE_EXPOSED_TO_HOST    = 5,
 };
 
+enum MR_LD_QUERY_TYPE {
+	MR_LD_QUERY_TYPE_ALL	         = 0,
+	MR_LD_QUERY_TYPE_EXPOSED_TO_HOST = 1,
+	MR_LD_QUERY_TYPE_USED_TGT_IDS    = 2,
+	MR_LD_QUERY_TYPE_CLUSTER_ACCESS  = 3,
+	MR_LD_QUERY_TYPE_CLUSTER_LOCALE  = 4,
+};
+
+
 #define MR_EVT_CFG_CLEARED                              0x0004
 #define MR_EVT_LD_STATE_CHANGE                          0x0051
 #define MR_EVT_PD_INSERTED                              0x005b
@@ -435,6 +445,14 @@ struct MR_LD_LIST {
 	} ldList[MAX_LOGICAL_DRIVES];
 } __packed;
 
+struct MR_LD_TARGETID_LIST {
+	u32	size;
+	u32	count;
+	u8	pad[3];
+	u8	targetId[MAX_LOGICAL_DRIVES];
+};
+
+
 /*
  * SAS controller properties
  */
@@ -863,7 +881,7 @@ struct megasas_ctrl_info {
  * ===============================
  */
 #define MEGASAS_MAX_PD_CHANNELS			2
-#define MEGASAS_MAX_LD_CHANNELS			2
+#define MEGASAS_MAX_LD_CHANNELS			1
 #define MEGASAS_MAX_CHANNELS			(MEGASAS_MAX_PD_CHANNELS + \
 						MEGASAS_MAX_LD_CHANNELS)
 #define MEGASAS_MAX_DEV_PER_CHANNEL		128
@@ -1656,4 +1674,16 @@ struct megasas_mgmt_info {
 	int max_index;
 };
 
+u8
+MR_BuildRaidContext(struct megasas_instance *instance,
+		    struct IO_REQUEST_INFO *io_info,
+		    struct RAID_CONTEXT *pRAID_Context,
+		    struct MR_FW_RAID_MAP_ALL *map, u8 **raidLUN);
+u16 MR_TargetIdToLdGet(u32 ldTgtId, struct MR_FW_RAID_MAP_ALL *map);
+struct MR_LD_RAID *MR_LdRaidGet(u32 ld, struct MR_FW_RAID_MAP_ALL *map);
+u16 MR_ArPdGet(u32 ar, u32 arm, struct MR_FW_RAID_MAP_ALL *map);
+u16 MR_LdSpanArrayGet(u32 ld, u32 span, struct MR_FW_RAID_MAP_ALL *map);
+u16 MR_PdDevHandleGet(u32 pd, struct MR_FW_RAID_MAP_ALL *map);
+u16 MR_GetLDTgtId(u32 ld, struct MR_FW_RAID_MAP_ALL *map);
+
 #endif				/*LSI_MEGARAID_SAS_H */

drivers/scsi/megaraid/megaraid_sas_base.c

@@ -92,6 +92,8 @@ MODULE_DESCRIPTION("LSI MegaRAID SAS Driver");
 
 int megasas_transition_to_ready(struct megasas_instance *instance, int ocr);
 static int megasas_get_pd_list(struct megasas_instance *instance);
+static int megasas_ld_list_query(struct megasas_instance *instance,
+				 u8 query_type);
 static int megasas_issue_init_mfi(struct megasas_instance *instance);
 static int megasas_register_aen(struct megasas_instance *instance,
 				u32 seq_num, u32 class_locale_word);
@@ -3270,6 +3272,84 @@ megasas_get_ld_list(struct megasas_instance *instance)
 	return ret;
 }
 
+/**
+ * megasas_ld_list_query -	Returns FW's ld_list structure
+ * @instance:				Adapter soft state
+ * @ld_list:				ld_list structure
+ *
+ * Issues an internal command (DCMD) to get the FW's controller PD
+ * list structure.  This information is mainly used to find out SYSTEM
+ * supported by the FW.
+ */
+static int
+megasas_ld_list_query(struct megasas_instance *instance, u8 query_type)
+{
+	int ret = 0, ld_index = 0, ids = 0;
+	struct megasas_cmd *cmd;
+	struct megasas_dcmd_frame *dcmd;
+	struct MR_LD_TARGETID_LIST *ci;
+	dma_addr_t ci_h = 0;
+
+	cmd = megasas_get_cmd(instance);
+
+	if (!cmd) {
+		printk(KERN_WARNING
+		       "megasas:(megasas_ld_list_query): Failed to get cmd\n");
+		return -ENOMEM;
+	}
+
+	dcmd = &cmd->frame->dcmd;
+
+	ci = pci_alloc_consistent(instance->pdev,
+				  sizeof(struct MR_LD_TARGETID_LIST), &ci_h);
+
+	if (!ci) {
+		printk(KERN_WARNING
+		       "megasas: Failed to alloc mem for ld_list_query\n");
+		megasas_return_cmd(instance, cmd);
+		return -ENOMEM;
+	}
+
+	memset(ci, 0, sizeof(*ci));
+	memset(dcmd->mbox.b, 0, MFI_MBOX_SIZE);
+
+	dcmd->mbox.b[0] = query_type;
+
+	dcmd->cmd = MFI_CMD_DCMD;
+	dcmd->cmd_status = 0xFF;
+	dcmd->sge_count = 1;
+	dcmd->flags = MFI_FRAME_DIR_READ;
+	dcmd->timeout = 0;
+	dcmd->data_xfer_len = sizeof(struct MR_LD_TARGETID_LIST);
+	dcmd->opcode = MR_DCMD_LD_LIST_QUERY;
+	dcmd->sgl.sge32[0].phys_addr = ci_h;
+	dcmd->sgl.sge32[0].length = sizeof(struct MR_LD_TARGETID_LIST);
+	dcmd->pad_0  = 0;
+
+	if (!megasas_issue_polled(instance, cmd) && !dcmd->cmd_status) {
+		ret = 0;
+	} else {
+		/* On failure, call older LD list DCMD */
+		ret = 1;
+	}
+
+	if ((ret == 0) && (ci->count <= (MAX_LOGICAL_DRIVES))) {
+		memset(instance->ld_ids, 0xff, MEGASAS_MAX_LD_IDS);
+		for (ld_index = 0; ld_index < ci->count; ld_index++) {
+			ids = ci->targetId[ld_index];
+			instance->ld_ids[ids] = ci->targetId[ld_index];
+		}
+
+	}
+
+	pci_free_consistent(instance->pdev, sizeof(struct MR_LD_TARGETID_LIST),
+			    ci, ci_h);
+
+	megasas_return_cmd(instance, cmd);
+
+	return ret;
+}
+
 /**
  * megasas_get_controller_info -	Returns FW's controller structure
  * @instance:				Adapter soft state
@@ -3648,7 +3728,9 @@ static int megasas_init_fw(struct megasas_instance *instance)
 	megasas_get_pd_list(instance);
 
 	memset(instance->ld_ids, 0xff, MEGASAS_MAX_LD_IDS);
-	megasas_get_ld_list(instance);
+	if (megasas_ld_list_query(instance,
+				  MR_LD_QUERY_TYPE_EXPOSED_TO_HOST))
+		megasas_get_ld_list(instance);
 
 	ctrl_info = kmalloc(sizeof(struct megasas_ctrl_info), GFP_KERNEL);
 
@@ -5389,7 +5471,9 @@ megasas_aen_polling(struct work_struct *work)
 		case MR_EVT_LD_OFFLINE:
 		case MR_EVT_CFG_CLEARED:
 		case MR_EVT_LD_DELETED:
-			megasas_get_ld_list(instance);
+			if (megasas_ld_list_query(instance,
+					MR_LD_QUERY_TYPE_EXPOSED_TO_HOST))
+				megasas_get_ld_list(instance);
 			for (i = 0; i < MEGASAS_MAX_LD_CHANNELS; i++) {
 				for (j = 0;
 				j < MEGASAS_MAX_DEV_PER_CHANNEL;
@@ -5399,7 +5483,7 @@ megasas_aen_polling(struct work_struct *work)
 				(i * MEGASAS_MAX_DEV_PER_CHANNEL) + j;
 
 				sdev1 = scsi_device_lookup(host,
-					i + MEGASAS_MAX_LD_CHANNELS,
+					MEGASAS_MAX_PD_CHANNELS + i,
 					j,
 					0);
 
@@ -5418,7 +5502,9 @@ megasas_aen_polling(struct work_struct *work)
 			doscan = 0;
 			break;
 		case MR_EVT_LD_CREATED:
-			megasas_get_ld_list(instance);
+			if (megasas_ld_list_query(instance,
+					MR_LD_QUERY_TYPE_EXPOSED_TO_HOST))
+				megasas_get_ld_list(instance);
 			for (i = 0; i < MEGASAS_MAX_LD_CHANNELS; i++) {
 				for (j = 0;
 					j < MEGASAS_MAX_DEV_PER_CHANNEL;
@@ -5427,14 +5513,14 @@ megasas_aen_polling(struct work_struct *work)
 					(i * MEGASAS_MAX_DEV_PER_CHANNEL) + j;
 
 					sdev1 = scsi_device_lookup(host,
-						i+MEGASAS_MAX_LD_CHANNELS,
+						MEGASAS_MAX_PD_CHANNELS + i,
 						j, 0);
 
 					if (instance->ld_ids[ld_index] !=
 								0xff) {
 						if (!sdev1) {
 							scsi_add_device(host,
-								i + 2,
+						MEGASAS_MAX_PD_CHANNELS + i,
 								j, 0);
 						}
 					}
@@ -5483,18 +5569,20 @@ megasas_aen_polling(struct work_struct *work)
 			}
 		}
 
-		megasas_get_ld_list(instance);
+		if (megasas_ld_list_query(instance,
+					  MR_LD_QUERY_TYPE_EXPOSED_TO_HOST))
+			megasas_get_ld_list(instance);
 		for (i = 0; i < MEGASAS_MAX_LD_CHANNELS; i++) {
 			for (j = 0; j < MEGASAS_MAX_DEV_PER_CHANNEL; j++) {
 				ld_index =
 				(i * MEGASAS_MAX_DEV_PER_CHANNEL) + j;
 
 				sdev1 = scsi_device_lookup(host,
-					i+MEGASAS_MAX_LD_CHANNELS, j, 0);
+					MEGASAS_MAX_PD_CHANNELS + i, j, 0);
 				if (instance->ld_ids[ld_index] != 0xff) {
 					if (!sdev1) {
 						scsi_add_device(host,
-								i+2,
+						MEGASAS_MAX_PD_CHANNELS + i,
 								j, 0);
 					} else {
 						scsi_device_put(sdev1);

drivers/scsi/megaraid/megaraid_sas_fp.c

@@ -126,17 +126,17 @@ static u8 MR_LdDataArmGet(u32 ld, u32 armIdx, struct MR_FW_RAID_MAP_ALL *map)
 	return map->raidMap.ldSpanMap[ld].dataArmMap[armIdx];
 }
 
-static u16 MR_ArPdGet(u32 ar, u32 arm, struct MR_FW_RAID_MAP_ALL *map)
+u16 MR_ArPdGet(u32 ar, u32 arm, struct MR_FW_RAID_MAP_ALL *map)
 {
 	return map->raidMap.arMapInfo[ar].pd[arm];
 }
 
-static u16 MR_LdSpanArrayGet(u32 ld, u32 span, struct MR_FW_RAID_MAP_ALL *map)
+u16 MR_LdSpanArrayGet(u32 ld, u32 span, struct MR_FW_RAID_MAP_ALL *map)
 {
 	return map->raidMap.ldSpanMap[ld].spanBlock[span].span.arrayRef;
 }
 
-static u16 MR_PdDevHandleGet(u32 pd, struct MR_FW_RAID_MAP_ALL *map)
+u16 MR_PdDevHandleGet(u32 pd, struct MR_FW_RAID_MAP_ALL *map)
 {
 	return map->raidMap.devHndlInfo[pd].curDevHdl;
 }
@@ -784,7 +784,7 @@ u8
 MR_BuildRaidContext(struct megasas_instance *instance,
 		    struct IO_REQUEST_INFO *io_info,
 		    struct RAID_CONTEXT *pRAID_Context,
-		    struct MR_FW_RAID_MAP_ALL *map)
+		    struct MR_FW_RAID_MAP_ALL *map, u8 **raidLUN)
 {
 	struct MR_LD_RAID  *raid;
 	u32         ld, stripSize, stripe_mask;
@@ -977,6 +977,9 @@ MR_BuildRaidContext(struct megasas_instance *instance,
 	pRAID_Context->regLockRowLBA    = regStart;
 	pRAID_Context->regLockLength    = regSize;
 	pRAID_Context->configSeqNum	= raid->seqNum;
+	/* save pointer to raid->LUN array */
+	*raidLUN = raid->LUN;
+
 
 	/*Get Phy Params only if FP capable, or else leave it to MR firmware
 	  to do the calculation.*/

drivers/scsi/megaraid/megaraid_sas_fusion.c

@@ -72,17 +72,6 @@ megasas_clear_intr_fusion(struct megasas_register_set __iomem *regs);
 int
 megasas_issue_polled(struct megasas_instance *instance,
 		     struct megasas_cmd *cmd);
-
-u8
-MR_BuildRaidContext(struct megasas_instance *instance,
-		    struct IO_REQUEST_INFO *io_info,
-		    struct RAID_CONTEXT *pRAID_Context,
-		    struct MR_FW_RAID_MAP_ALL *map);
-u16 MR_TargetIdToLdGet(u32 ldTgtId, struct MR_FW_RAID_MAP_ALL *map);
-struct MR_LD_RAID *MR_LdRaidGet(u32 ld, struct MR_FW_RAID_MAP_ALL *map);
-
-u16 MR_GetLDTgtId(u32 ld, struct MR_FW_RAID_MAP_ALL *map);
-
 void
 megasas_check_and_restore_queue_depth(struct megasas_instance *instance);
 
@@ -652,6 +641,10 @@ megasas_ioc_init_fusion(struct megasas_instance *instance)
 		(instance->pdev->device == PCI_DEVICE_ID_LSI_FURY))
 		init_frame->driver_operations.
 			mfi_capabilities.support_additional_msix = 1;
+	/* driver supports HA / Remote LUN over Fast Path interface */
+	init_frame->driver_operations.mfi_capabilities.support_fp_remote_lun
+		= 1;
+
 
 	init_frame->queue_info_new_phys_addr_lo = ioc_init_handle;
 	init_frame->data_xfer_len = sizeof(struct MPI2_IOC_INIT_REQUEST);
@@ -1410,6 +1403,7 @@ megasas_build_ldio_fusion(struct megasas_instance *instance,
 	struct IO_REQUEST_INFO io_info;
 	struct fusion_context *fusion;
 	struct MR_FW_RAID_MAP_ALL *local_map_ptr;
+	u8 *raidLUN;
 
 	device_id = MEGASAS_DEV_INDEX(instance, scp);
 
@@ -1494,7 +1488,7 @@ megasas_build_ldio_fusion(struct megasas_instance *instance,
 	} else {
 		if (MR_BuildRaidContext(instance, &io_info,
 					&io_request->RaidContext,
-					local_map_ptr))
+					local_map_ptr, &raidLUN))
 			fp_possible = io_info.fpOkForIo;
 	}
 
@@ -1537,6 +1531,8 @@ megasas_build_ldio_fusion(struct megasas_instance *instance,
 			scp->SCp.Status &= ~MEGASAS_LOAD_BALANCE_FLAG;
 		cmd->request_desc->SCSIIO.DevHandle = io_info.devHandle;
 		io_request->DevHandle = io_info.devHandle;
+		/* populate the LUN field */
+		memcpy(io_request->LUN, raidLUN, 8);
 	} else {
 		io_request->RaidContext.timeoutValue =
 			local_map_ptr->raidMap.fpPdIoTimeoutSec;
@@ -1579,6 +1575,11 @@ megasas_build_dcdb_fusion(struct megasas_instance *instance,
 	u16 pd_index = 0;
 	struct MR_FW_RAID_MAP_ALL *local_map_ptr;
 	struct fusion_context *fusion = instance->ctrl_context;
+	u8                          span, physArm;
+	u16                         devHandle;
+	u32                         ld, arRef, pd;
+	struct MR_LD_RAID                  *raid;
+	struct RAID_CONTEXT                *pRAID_Context;
 
 	io_request = cmd->io_request;
 	device_id = MEGASAS_DEV_INDEX(instance, scmd);
@@ -1586,6 +1587,9 @@ megasas_build_dcdb_fusion(struct megasas_instance *instance,
 		+scmd->device->id;
 	local_map_ptr = fusion->ld_map[(instance->map_id & 1)];
 
+	io_request->DataLength = scsi_bufflen(scmd);
+
+
 	/* Check if this is a system PD I/O */
 	if (scmd->device->channel < MEGASAS_MAX_PD_CHANNELS &&
 	    instance->pd_list[pd_index].driveState == MR_PD_STATE_SYSTEM) {
@@ -1623,6 +1627,54 @@ megasas_build_dcdb_fusion(struct megasas_instance *instance,
 					scmd->request->timeout / HZ;
 		}
 	} else {
+		if (scmd->device->channel < MEGASAS_MAX_PD_CHANNELS)
+			goto NonFastPath;
+
+		ld = MR_TargetIdToLdGet(device_id, local_map_ptr);
+		if ((ld >= MAX_LOGICAL_DRIVES) || (!fusion->fast_path_io))
+			goto NonFastPath;
+
+		raid = MR_LdRaidGet(ld, local_map_ptr);
+
+		/* check if this LD is FP capable */
+		if (!(raid->capability.fpNonRWCapable))
+			/* not FP capable, send as non-FP */
+			goto NonFastPath;
+
+		/* get RAID_Context pointer */
+		pRAID_Context = &io_request->RaidContext;
+
+		/* set RAID context values */
+		pRAID_Context->regLockFlags     = REGION_TYPE_SHARED_READ;
+		pRAID_Context->timeoutValue     = raid->fpIoTimeoutForLd;
+		pRAID_Context->VirtualDiskTgtId = device_id;
+		pRAID_Context->regLockRowLBA    = 0;
+		pRAID_Context->regLockLength    = 0;
+		pRAID_Context->configSeqNum     = raid->seqNum;
+
+		/* get the DevHandle for the PD (since this is
+		   fpNonRWCapable, this is a single disk RAID0) */
+		span = physArm = 0;
+		arRef = MR_LdSpanArrayGet(ld, span, local_map_ptr);
+		pd = MR_ArPdGet(arRef, physArm, local_map_ptr);
+		devHandle = MR_PdDevHandleGet(pd, local_map_ptr);
+
+		/* build request descriptor */
+		cmd->request_desc->SCSIIO.RequestFlags =
+			(MPI2_REQ_DESCRIPT_FLAGS_HIGH_PRIORITY <<
+			 MEGASAS_REQ_DESCRIPT_FLAGS_TYPE_SHIFT);
+		cmd->request_desc->SCSIIO.DevHandle = devHandle;
+
+		/* populate the LUN field */
+		memcpy(io_request->LUN, raid->LUN, 8);
+
+		/* build the raidScsiIO structure */
+		io_request->Function = MPI2_FUNCTION_SCSI_IO_REQUEST;
+		io_request->DevHandle = devHandle;
+
+		return;
+
+NonFastPath:
 		io_request->Function  = MEGASAS_MPI2_FUNCTION_LD_IO_REQUEST;
 		io_request->DevHandle = device_id;
 		cmd->request_desc->SCSIIO.RequestFlags =
@@ -1631,7 +1683,6 @@ megasas_build_dcdb_fusion(struct megasas_instance *instance,
 	}
 	io_request->RaidContext.VirtualDiskTgtId = device_id;
 	io_request->LUN[1] = scmd->device->lun;
-	io_request->DataLength = scsi_bufflen(scmd);
 }
 
 /**

drivers/scsi/megaraid/megaraid_sas_fusion.h

@@ -527,7 +527,8 @@ struct MR_LD_RAID {
 		u32     fpReadCapable:1;
 		u32     fpWriteAcrossStripe:1;
 		u32     fpReadAcrossStripe:1;
-		u32     reserved4:8;
+		u32	fpNonRWCapable:1;
+		u32     reserved4:7;
 	} capability;
 	u32     reserved6;
 	u64     size;
@@ -551,7 +552,9 @@ struct MR_LD_RAID {
 		u32 reserved:31;
 	} flags;
 
-	u8      reserved3[0x5C];
+	u8	LUN[8]; /* 0x24 8 byte LUN field used for SCSI IO's */
+	u8	fpIoTimeoutForLd;/*0x2C timeout value used by driver in FP IO*/
+	u8      reserved3[0x80-0x2D]; /* 0x2D */
 };
 
 struct MR_LD_SPAN_MAP {