megaraid_sas : Use Block layer tag support for internal command indexing

megaraid_sas driver will use block layer provided tag for indexing internal
MPT frames to get any unique MPT frame tied with tag.  Each IO request
submitted from SCSI mid layer will get associated MPT frame from MPT framepool
(retrieved and return back using spinlock inside megaraid_sas driver's
submission/completion call back). Getting MPT frame from MPT Frame pool is
very expensive operation because of associated spin lock operation (spinlock
overhead increase on multi NUMA node). This type of locking in driver is very
expensive call considering each IO request need - Acquire and Release of the
same lock.

With this support, in IO path driver will directly provide the unique command
index(which is based on block layer tag) and will get the MPT frame tied to
the tag and this way driver can get rid off lock, which synchronizes the
access to MPT frame pool while fetching and returning MPT frame from the pool.

This support in driver provides siginificant performance improvement(on multi
NUMA node system)on latest upstream with SCSI.MQ as well as on existing linux
distributions.

Here is the data for test executed at Avago-
- IO Tool- FIO
- 4 Socket SMC server. (4 NUMA node server)
- 12 SSDs in JBOD mode .
- 4K Rand READ, QD=32
- SCSI MQ x86_64 (Latest Upstream kernel)
- upto 300% Performance Improvement.

If IOs are running on single Node, perfromance gain is less, but as soon as
increase number of nodes, performance improvement is significant. IOs running
on all 4 NUMA nodes, with this patch applied IOPs observed was 1170K vs 344K
IOPs seen without this patch.

Logically, there are two parts of this patch- 1) Block layer tag support 2)
changes in calling convention of return_cmd.  part 2 will revert the changes
done by patch- 90dc9d98 megaraid_sas : MFI MPT linked list corruption fix
because changes done in part 1 has fixed the problem of MFI MPT linked list
corruption. part 2 is very much dependent on part 1, so we decided to have
single patch for these two logical changes.

[jejb: remove chatty printk pointed out by hch]
Signed-off-by: Kashyap Desai <kashyap.desai@avagotech.com>
Signed-off-by: Sumit Saxena <sumit.saxena@avagotech.com>
Reviewed-by: Christoph Hellwig <hch@lst.de>
Signed-off-by: James Bottomley <JBottomley@Odin.com>

drivers/scsi/megaraid/megaraid_sas.h

@@ -153,6 +153,9 @@
 #define MFI_FRAME_DIR_BOTH			0x0018
 #define MFI_FRAME_IEEE                          0x0020
 
+/* Driver internal */
+#define DRV_DCMD_POLLED_MODE		0x1
+
 /*
  * Definition for cmd_status
  */
@@ -1042,11 +1045,6 @@ struct megasas_ctrl_info {
 
 #define VD_EXT_DEBUG 0
 
-enum MR_MFI_MPT_PTHR_FLAGS {
-	MFI_MPT_DETACHED = 0,
-	MFI_LIST_ADDED = 1,
-	MFI_MPT_ATTACHED = 2,
-};
 
 enum MR_SCSI_CMD_TYPE {
 	READ_WRITE_LDIO = 0,
@@ -1889,10 +1887,7 @@ struct megasas_cmd {
 
 	struct list_head list;
 	struct scsi_cmnd *scmd;
-
-	void *mpt_pthr_cmd_blocked;
-	atomic_t mfi_mpt_pthr;
-	u8 is_wait_event;
+	u8 flags;
 
 	struct megasas_instance *instance;
 	union {

drivers/scsi/megaraid/megaraid_sas_fusion.c

@@ -57,6 +57,7 @@
 #include "megaraid_sas_fusion.h"
 #include "megaraid_sas.h"
 
+
 extern void megasas_free_cmds(struct megasas_instance *instance);
 extern struct megasas_cmd *megasas_get_cmd(struct megasas_instance
 					   *instance);
@@ -156,28 +157,15 @@ megasas_clear_intr_fusion(struct megasas_register_set __iomem *regs)
  * megasas_get_cmd_fusion -	Get a command from the free pool
  * @instance:		Adapter soft state
  *
- * Returns a free command from the pool
+ * Returns a blk_tag indexed mpt frame
  */
-struct megasas_cmd_fusion *megasas_get_cmd_fusion(struct megasas_instance
-						  *instance)
+inline struct megasas_cmd_fusion *megasas_get_cmd_fusion(struct megasas_instance
+						  *instance, u32 blk_tag)
 {
-	unsigned long flags;
-	struct fusion_context *fusion =
-		(struct fusion_context *)instance->ctrl_context;
-	struct megasas_cmd_fusion *cmd = NULL;
-
-	spin_lock_irqsave(&fusion->mpt_pool_lock, flags);
-
-	if (!list_empty(&fusion->cmd_pool)) {
-		cmd = list_entry((&fusion->cmd_pool)->next,
-				 struct megasas_cmd_fusion, list);
-		list_del_init(&cmd->list);
-	} else {
-		printk(KERN_ERR "megasas: Command pool (fusion) empty!\n");
-	}
+	struct fusion_context *fusion;
 
-	spin_unlock_irqrestore(&fusion->mpt_pool_lock, flags);
-	return cmd;
+	fusion = instance->ctrl_context;
+	return fusion->cmd_list[blk_tag];
 }
 
 /**
@@ -188,45 +176,8 @@ struct megasas_cmd_fusion *megasas_get_cmd_fusion(struct megasas_instance
 inline void megasas_return_cmd_fusion(struct megasas_instance *instance,
 	struct megasas_cmd_fusion *cmd)
 {
-	unsigned long flags;
-	struct fusion_context *fusion =
-		(struct fusion_context *)instance->ctrl_context;
-
-	spin_lock_irqsave(&fusion->mpt_pool_lock, flags);
-
 	cmd->scmd = NULL;
-	cmd->sync_cmd_idx = (u32)ULONG_MAX;
 	memset(cmd->io_request, 0, sizeof(struct MPI2_RAID_SCSI_IO_REQUEST));
-	list_add(&cmd->list, (&fusion->cmd_pool)->next);
-
-	spin_unlock_irqrestore(&fusion->mpt_pool_lock, flags);
-}
-
-/**
- * megasas_return_mfi_mpt_pthr - Return a mfi and mpt to free command pool
- * @instance:		Adapter soft state
- * @cmd_mfi:		MFI Command packet to be returned to free command pool
- * @cmd_mpt:		MPT Command packet to be returned to free command pool
- */
-inline void megasas_return_mfi_mpt_pthr(struct megasas_instance *instance,
-		struct megasas_cmd *cmd_mfi,
-		struct megasas_cmd_fusion *cmd_fusion)
-{
-	unsigned long flags;
-
-	/*
-	 * TO DO: optimize this code and use only one lock instead of two
-	 * locks being used currently- mpt_pool_lock is acquired
-	 * inside mfi_pool_lock
-	 */
-	spin_lock_irqsave(&instance->mfi_pool_lock, flags);
-	megasas_return_cmd_fusion(instance, cmd_fusion);
-	if (atomic_read(&cmd_mfi->mfi_mpt_pthr) != MFI_MPT_ATTACHED)
-		dev_err(&instance->pdev->dev, "Possible bug from %s %d\n",
-			__func__, __LINE__);
-	atomic_set(&cmd_mfi->mfi_mpt_pthr, MFI_MPT_DETACHED);
-	__megasas_return_cmd(instance, cmd_mfi);
-	spin_unlock_irqrestore(&instance->mfi_pool_lock, flags);
 }
 
 /**
@@ -326,7 +277,6 @@ megasas_free_cmds_fusion(struct megasas_instance *instance)
 	kfree(fusion->cmd_list);
 	fusion->cmd_list = NULL;
 
-	INIT_LIST_HEAD(&fusion->cmd_pool);
 }
 
 /**
@@ -535,7 +485,9 @@ megasas_alloc_cmds_fusion(struct megasas_instance *instance)
 		memset(cmd, 0, sizeof(struct megasas_cmd_fusion));
 		cmd->index = i + 1;
 		cmd->scmd = NULL;
-		cmd->sync_cmd_idx = (u32)ULONG_MAX; /* Set to Invalid */
+		cmd->sync_cmd_idx = (i >= instance->max_scsi_cmds) ?
+				(i - instance->max_scsi_cmds) :
+				(u32)ULONG_MAX; /* Set to Invalid */
 		cmd->instance = instance;
 		cmd->io_request =
 			(struct MPI2_RAID_SCSI_IO_REQUEST *)
@@ -543,8 +495,6 @@ megasas_alloc_cmds_fusion(struct megasas_instance *instance)
 		memset(cmd->io_request, 0,
 		       sizeof(struct MPI2_RAID_SCSI_IO_REQUEST));
 		cmd->io_request_phys_addr = io_req_base_phys + offset;
-
-		list_add_tail(&cmd->list, &fusion->cmd_pool);
 	}
 
 	/*
@@ -605,12 +555,8 @@ wait_and_poll(struct megasas_instance *instance, struct megasas_cmd *cmd,
 		msleep(20);
 	}
 
-	if (frame_hdr->cmd_status == 0xff) {
-		if (fusion)
-			megasas_return_mfi_mpt_pthr(instance, cmd,
-				cmd->mpt_pthr_cmd_blocked);
+	if (frame_hdr->cmd_status == 0xff)
 		return -ETIME;
-	}
 
 	return 0;
 }
@@ -820,11 +766,7 @@ megasas_get_ld_map_info(struct megasas_instance *instance)
 	else
 		ret = megasas_issue_polled(instance, cmd);
 
-	if (instance->ctrl_context && cmd->mpt_pthr_cmd_blocked)
-		megasas_return_mfi_mpt_pthr(instance, cmd,
-			cmd->mpt_pthr_cmd_blocked);
-	else
-		megasas_return_cmd(instance, cmd);
+	megasas_return_cmd(instance, cmd);
 
 	return ret;
 }
@@ -1953,9 +1895,7 @@ megasas_build_and_issue_cmd_fusion(struct megasas_instance *instance,
 
 	fusion = instance->ctrl_context;
 
-	cmd = megasas_get_cmd_fusion(instance);
-	if (!cmd)
-		return SCSI_MLQUEUE_HOST_BUSY;
+	cmd = megasas_get_cmd_fusion(instance, scmd->request->tag);
 
 	index = cmd->index;
 
@@ -2013,6 +1953,7 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex)
 	union desc_value d_val;
 	struct LD_LOAD_BALANCE_INFO *lbinfo;
 	int threshold_reply_count = 0;
+	struct scsi_cmnd *scmd_local = NULL;
 
 	fusion = instance->ctrl_context;
 
@@ -2048,13 +1989,14 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex)
 		if (cmd_fusion->scmd)
 			cmd_fusion->scmd->SCp.ptr = NULL;
 
+		scmd_local = cmd_fusion->scmd;
 		status = scsi_io_req->RaidContext.status;
 		extStatus = scsi_io_req->RaidContext.exStatus;
 
 		switch (scsi_io_req->Function) {
 		case MPI2_FUNCTION_SCSI_IO_REQUEST:  /*Fast Path IO.*/
 			/* Update load balancing info */
-			device_id = MEGASAS_DEV_INDEX(cmd_fusion->scmd);
+			device_id = MEGASAS_DEV_INDEX(scmd_local);
 			lbinfo = &fusion->load_balance_info[device_id];
 			if (cmd_fusion->scmd->SCp.Status &
 			    MEGASAS_LOAD_BALANCE_FLAG) {
@@ -2072,29 +2014,25 @@ complete_cmd_fusion(struct megasas_instance *instance, u32 MSIxIndex)
 		case MEGASAS_MPI2_FUNCTION_LD_IO_REQUEST: /* LD-IO Path */
 			/* Map the FW Cmd Status */
 			map_cmd_status(cmd_fusion, status, extStatus);
-			scsi_dma_unmap(cmd_fusion->scmd);
-			cmd_fusion->scmd->scsi_done(cmd_fusion->scmd);
 			scsi_io_req->RaidContext.status = 0;
 			scsi_io_req->RaidContext.exStatus = 0;
 			megasas_return_cmd_fusion(instance, cmd_fusion);
+			scsi_dma_unmap(scmd_local);
+			scmd_local->scsi_done(scmd_local);
 			atomic_dec(&instance->fw_outstanding);
 
 			break;
 		case MEGASAS_MPI2_FUNCTION_PASSTHRU_IO_REQUEST: /*MFI command */
 			cmd_mfi = instance->cmd_list[cmd_fusion->sync_cmd_idx];
 
-			if (!cmd_mfi->mpt_pthr_cmd_blocked) {
-				if (megasas_dbg_lvl == 5)
-					dev_info(&instance->pdev->dev,
-						"freeing mfi/mpt pass-through "
-						"from %s %d\n",
-						 __func__, __LINE__);
-				megasas_return_mfi_mpt_pthr(instance, cmd_mfi,
-					cmd_fusion);
-			}
-
-			megasas_complete_cmd(instance, cmd_mfi, DID_OK);
-			cmd_fusion->flags = 0;
+			/* Poll mode. Dummy free.
+			 * In case of Interrupt mode, caller has reverse check.
+			 */
+			if (cmd_mfi->flags & DRV_DCMD_POLLED_MODE) {
+				cmd_mfi->flags &= ~DRV_DCMD_POLLED_MODE;
+				megasas_return_cmd(instance, cmd_mfi);
+			} else
+				megasas_complete_cmd(instance, cmd_mfi, DID_OK);
 			break;
 		}
 
@@ -2254,27 +2192,14 @@ build_mpt_mfi_pass_thru(struct megasas_instance *instance,
 	struct megasas_cmd_fusion *cmd;
 	struct fusion_context *fusion;
 	struct megasas_header *frame_hdr = &mfi_cmd->frame->hdr;
-	u32 opcode;
 
-	cmd = megasas_get_cmd_fusion(instance);
-	if (!cmd)
-		return 1;
+	fusion = instance->ctrl_context;
+
+	cmd = megasas_get_cmd_fusion(instance,
+			instance->max_scsi_cmds + mfi_cmd->index);
 
 	/*  Save the smid. To be used for returning the cmd */
 	mfi_cmd->context.smid = cmd->index;
-	cmd->sync_cmd_idx = mfi_cmd->index;
-
-	/* Set this only for Blocked commands */
-	opcode = le32_to_cpu(mfi_cmd->frame->dcmd.opcode);
-	if ((opcode == MR_DCMD_LD_MAP_GET_INFO)
-		&& (mfi_cmd->frame->dcmd.mbox.b[1] == 1))
-		mfi_cmd->is_wait_event = 1;
-
-	if (opcode == MR_DCMD_CTRL_EVENT_WAIT)
-		mfi_cmd->is_wait_event = 1;
-
-	if (mfi_cmd->is_wait_event)
-		mfi_cmd->mpt_pthr_cmd_blocked = cmd;
 
 	/*
 	 * For cmds where the flag is set, store the flag and check
@@ -2283,9 +2208,8 @@ build_mpt_mfi_pass_thru(struct megasas_instance *instance,
 	 */
 
 	if (frame_hdr->flags & cpu_to_le16(MFI_FRAME_DONT_POST_IN_REPLY_QUEUE))
-		cmd->flags = MFI_FRAME_DONT_POST_IN_REPLY_QUEUE;
+		mfi_cmd->flags |= DRV_DCMD_POLLED_MODE;
 
-	fusion = instance->ctrl_context;
 	io_req = cmd->io_request;
 
 	if ((instance->pdev->device == PCI_DEVICE_ID_LSI_INVADER) ||
@@ -2364,7 +2288,6 @@ megasas_issue_dcmd_fusion(struct megasas_instance *instance,
 		printk(KERN_ERR "Couldn't issue MFI pass thru cmd\n");
 		return;
 	}
-	atomic_set(&cmd->mfi_mpt_pthr, MFI_MPT_ATTACHED);
 	instance->instancet->fire_cmd(instance, req_desc->u.low,
 				      req_desc->u.high, instance->reg_set);
 }
@@ -2618,6 +2541,7 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int iotimeout)
 	struct fusion_context *fusion;
 	u32 host_diag, abs_state, status_reg, reset_adapter;
 	u32 io_timeout_in_crash_mode = 0;
+	struct scsi_cmnd *scmd_local = NULL;
 
 	instance = (struct megasas_instance *)shost->hostdata;
 	fusion = instance->ctrl_context;
@@ -2685,15 +2609,16 @@ int megasas_reset_fusion(struct Scsi_Host *shost, int iotimeout)
 			iotimeout = 0;
 
 		/* Now return commands back to the OS */
-		for (i = 0 ; i < instance->max_fw_cmds; i++) {
+		for (i = 0 ; i < instance->max_scsi_cmds; i++) {
 			cmd_fusion = fusion->cmd_list[i];
+			scmd_local = cmd_fusion->scmd;
 			if (cmd_fusion->scmd) {
-				scsi_dma_unmap(cmd_fusion->scmd);
-				cmd_fusion->scmd->result =
+				scmd_local->result =
 					megasas_check_mpio_paths(instance,
-								 cmd_fusion->scmd);
-				cmd_fusion->scmd->scsi_done(cmd_fusion->scmd);
+							scmd_local);
 				megasas_return_cmd_fusion(instance, cmd_fusion);
+				scsi_dma_unmap(scmd_local);
+				scmd_local->scsi_done(scmd_local);
 				atomic_dec(&instance->fw_outstanding);
 			}
 		}

drivers/scsi/megaraid/megaraid_sas_fusion.h

@@ -679,7 +679,6 @@ struct megasas_cmd_fusion {
 	 */
 	u32 sync_cmd_idx;
 	u32 index;
-	u8 flags;
 	u8 pd_r1_lb;
 };
 
@@ -792,10 +791,6 @@ struct MR_FW_RAID_MAP_EXT {
 
 struct fusion_context {
 	struct megasas_cmd_fusion **cmd_list;
-	struct list_head cmd_pool;
-
-	spinlock_t mpt_pool_lock;
-
 	dma_addr_t req_frames_desc_phys;
 	u8 *req_frames_desc;