[SCSI] - fusion - mptfc bug fix's to prevent deadlock situations

mptbase.h

	bump version number to 3.03.09

	remove unneeded flags
	define workq and remove old fc specific locks

mptbase.c

	initialize new lock and don't initialize two removed locks

mptscsih.c

	when firmware reports target is no longer there, return
        DID_REQUEUE for fc hosts so that i/o doesn't get killed until
        the transport has an opportunity to manage the loss via its
        dev loss timer

	when the "eh_abort" routine is called, check to see if the
        driver has the command or not before looking to see if a reset
        is pending.  James Smart and I talked about this and believe
        that the API for this routine is: if driver doesn't have
        command, return SUCCESS.  This change helps prevent a target
        from being taken offline.  SUCCESS is returned because it's
        likely that the command completed after error recovery timed
        it out but before it could be aborted.

	provide a routine to queue work to newly created workq, and
        use it.

	remove "ioc" from mptscsih_abort() it was only used one time.
	the other references were via hd->ioc, so I just moved it....
	net change in references to ioc via hd->ioc is zero

	move hd->resetPending test and hd->timeouts increment to after
	the test for whether the command to be aborted remains known
	to the driver

	Make certain that the workq exists before queuing work to it.

mptfc.c

	no longer need to lock rport data structures as I was able to
	single thread the code!  I fixed up the debug code to
	eliminate compilation messages due to type mismatch in the
	printk.  Got rid of some no longer needed rport flags.
	Initialize and destroy the workq used for the rescan work.

	simplify the logic regarding the increment of
        fc_rescan_work_count.  use post increment and test for zero
        vs. pre increment and test for one; eliminate work_count
        variable: queue_work can be called with the work_lock held as
        it doesn't sleep

Signed-off-by: Michael Reed <mdr@sgi.com>
Signed-off-by: Eric Moore <Eric.Moore@lsil.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

4 files changed, 107 insertions, 77 deletions

diff --git a/drivers/message/fusion/mptbase.c b/drivers/message/fusion/mptbase.c
index ac66a658aa1..5fe6e8df50a 100644
--- a/drivers/message/fusion/mptbase.c
+++ b/drivers/message/fusion/mptbase.c
@@ -1189,7 +1189,6 @@ mpt_attach(struct pci_dev *pdev, const struct pci_device_id *id)
 	ioc->diagPending = 0;
 	spin_lock_init(&ioc->diagLock);
 	spin_lock_init(&ioc->fc_rescan_work_lock);
-	spin_lock_init(&ioc->fc_rport_lock);
 	spin_lock_init(&ioc->initializing_hba_lock);
 
 	/* Initialize the event logging.
diff --git a/drivers/message/fusion/mptbase.h b/drivers/message/fusion/mptbase.h
index be7e8501b53..f673cca507e 100644
--- a/drivers/message/fusion/mptbase.h
+++ b/drivers/message/fusion/mptbase.h
@@ -76,8 +76,8 @@
 #define COPYRIGHT	"Copyright (c) 1999-2005 " MODULEAUTHOR
 #endif
 
-#define MPT_LINUX_VERSION_COMMON	"3.03.08"
-#define MPT_LINUX_PACKAGE_NAME		"@(#)mptlinux-3.03.08"
+#define MPT_LINUX_VERSION_COMMON	"3.03.09"
+#define MPT_LINUX_PACKAGE_NAME		"@(#)mptlinux-3.03.09"
 #define WHAT_MAGIC_STRING		"@" "(" "#" ")"
 
 #define show_mptmod_ver(s,ver)  \
@@ -489,7 +489,6 @@ typedef	struct _RaidCfgData {
 
 #define MPT_RPORT_INFO_FLAGS_REGISTERED	0x01	/* rport registered */
 #define MPT_RPORT_INFO_FLAGS_MISSING	0x02	/* missing from DevPage0 scan */
-#define MPT_RPORT_INFO_FLAGS_MAPPED_VDEV 0x04	/* target mapped in vdev */
 
 /*
  * data allocated for each fc rport device
@@ -501,7 +500,6 @@ struct mptfc_rport_info
 	struct scsi_target *starget;
 	FCDevicePage0_t pg0;
 	u8		flags;
-	u8		remap_needed;
 };
 
 /*
@@ -628,11 +626,11 @@ typedef struct _MPT_ADAPTER
 	struct work_struct	 mptscsih_persistTask;
 
 	struct list_head	 fc_rports;
-	spinlock_t		 fc_rport_lock; /* list and ri flags */
 	spinlock_t		 fc_rescan_work_lock;
 	int			 fc_rescan_work_count;
 	struct work_struct	 fc_rescan_work;
-
+	char			 fc_rescan_work_q_name[KOBJ_NAME_LEN];
+	struct workqueue_struct *fc_rescan_work_q;
 } MPT_ADAPTER;
 
 /*
diff --git a/drivers/message/fusion/mptfc.c b/drivers/message/fusion/mptfc.c
index b343f2a68b1..e1fb5a16636 100644
--- a/drivers/message/fusion/mptfc.c
+++ b/drivers/message/fusion/mptfc.c
@@ -355,15 +355,13 @@ mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
 	struct fc_rport		*rport;
 	struct mptfc_rport_info	*ri;
 	int			new_ri = 1;
-	u64			pn;
-	unsigned long		flags;
+	u64			pn, nn;
 	VirtTarget		*vtarget;
 
 	if (mptfc_generate_rport_ids(pg0, &rport_ids) < 0)
 		return;
 
 	/* scan list looking for a match */
-	spin_lock_irqsave(&ioc->fc_rport_lock, flags);
 	list_for_each_entry(ri, &ioc->fc_rports, list) {
 		pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
 		if (pn == rport_ids.port_name) {	/* match */
@@ -373,11 +371,9 @@ mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
 		}
 	}
 	if (new_ri) {	/* allocate one */
-		spin_unlock_irqrestore(&ioc->fc_rport_lock, flags);
 		ri = kzalloc(sizeof(struct mptfc_rport_info), GFP_KERNEL);
 		if (!ri)
 			return;
-		spin_lock_irqsave(&ioc->fc_rport_lock, flags);
 		list_add_tail(&ri->list, &ioc->fc_rports);
 	}
 
@@ -387,14 +383,11 @@ mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
 	/* MPT_RPORT_INFO_FLAGS_REGISTERED - rport not previously deleted */
 	if (!(ri->flags & MPT_RPORT_INFO_FLAGS_REGISTERED)) {
 		ri->flags |= MPT_RPORT_INFO_FLAGS_REGISTERED;
-		spin_unlock_irqrestore(&ioc->fc_rport_lock, flags);
 		rport = fc_remote_port_add(ioc->sh, channel, &rport_ids);
-		spin_lock_irqsave(&ioc->fc_rport_lock, flags);
 		if (rport) {
 			ri->rport = rport;
 			if (new_ri) /* may have been reset by user */
 				rport->dev_loss_tmo = mptfc_dev_loss_tmo;
-			*((struct mptfc_rport_info **)rport->dd_data) = ri;
 			/*
 			 * if already mapped, remap here.  If not mapped,
 			 * target_alloc will allocate vtarget and map,
@@ -406,16 +399,20 @@ mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
 					vtarget->target_id = pg0->CurrentTargetID;
 					vtarget->bus_id = pg0->CurrentBus;
 				}
-				ri->remap_needed = 0;
 			}
+			/* once dd_data is filled in, commands will issue to hardware */
+			*((struct mptfc_rport_info **)rport->dd_data) = ri;
+
+			pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
+			nn = (u64)ri->pg0.WWNN.High << 32 | (u64)ri->pg0.WWNN.Low;
 			dfcprintk ((MYIOC_s_INFO_FMT
 				"mptfc_reg_dev.%d: %x, %llx / %llx, tid %d, "
 				"rport tid %d, tmo %d\n",
 					ioc->name,
 					ioc->sh->host_no,
 					pg0->PortIdentifier,
-					pg0->WWNN,
-					pg0->WWPN,
+					(unsigned long long)nn,
+					(unsigned long long)pn,
 					pg0->CurrentTargetID,
 					ri->rport->scsi_target_id,
 					ri->rport->dev_loss_tmo));
@@ -425,8 +422,6 @@ mptfc_register_dev(MPT_ADAPTER *ioc, int channel, FCDevicePage0_t *pg0)
 			ri = NULL;
 		}
 	}
-	spin_unlock_irqrestore(&ioc->fc_rport_lock,flags);
-
 }
 
 /*
@@ -476,7 +471,6 @@ mptfc_target_alloc(struct scsi_target *starget)
 			vtarget->target_id = ri->pg0.CurrentTargetID;
 			vtarget->bus_id = ri->pg0.CurrentBus;
 			ri->starget = starget;
-			ri->remap_needed = 0;
 			rc = 0;
 		}
 	}
@@ -502,10 +496,10 @@ mptfc_slave_alloc(struct scsi_device *sdev)
 	VirtDevice		*vdev;
 	struct scsi_target	*starget;
 	struct fc_rport		*rport;
-	unsigned long		flags;
 
 
-	rport = starget_to_rport(scsi_target(sdev));
+	starget = scsi_target(sdev);
+	rport = starget_to_rport(starget);
 
 	if (!rport || fc_remote_port_chkready(rport))
 		return -ENXIO;
@@ -519,10 +513,8 @@ mptfc_slave_alloc(struct scsi_device *sdev)
 		return -ENOMEM;
 	}
 
-	spin_lock_irqsave(&hd->ioc->fc_rport_lock,flags);
 
 	sdev->hostdata = vdev;
-	starget = scsi_target(sdev);
 	vtarget = starget->hostdata;
 
 	if (vtarget->num_luns == 0) {
@@ -535,14 +527,16 @@ mptfc_slave_alloc(struct scsi_device *sdev)
 	vdev->vtarget = vtarget;
 	vdev->lun = sdev->lun;
 
-	spin_unlock_irqrestore(&hd->ioc->fc_rport_lock,flags);
-
 	vtarget->num_luns++;
 
+
 #ifdef DMPT_DEBUG_FC
-	 {
+	{
+	u64 nn, pn;
 	struct mptfc_rport_info *ri;
 	ri = *((struct mptfc_rport_info **)rport->dd_data);
+	pn = (u64)ri->pg0.WWPN.High << 32 | (u64)ri->pg0.WWPN.Low;
+	nn = (u64)ri->pg0.WWNN.High << 32 | (u64)ri->pg0.WWNN.Low;
 	dfcprintk ((MYIOC_s_INFO_FMT
 		"mptfc_slv_alloc.%d: num_luns %d, sdev.id %d, "
 	        "CurrentTargetID %d, %x %llx %llx\n",
@@ -550,7 +544,9 @@ mptfc_slave_alloc(struct scsi_device *sdev)
 		sdev->host->host_no,
 		vtarget->num_luns,
 		sdev->id, ri->pg0.CurrentTargetID,
-		ri->pg0.PortIdentifier, ri->pg0.WWPN, ri->pg0.WWNN));
+		ri->pg0.PortIdentifier,
+		(unsigned long long)pn,
+		(unsigned long long)nn));
 	}
 #endif
 
@@ -570,11 +566,31 @@ mptfc_qcmd(struct scsi_cmnd *SCpnt, void (*done)(struct scsi_cmnd *))
 		done(SCpnt);
 		return 0;
 	}
+
+	/* dd_data is null until finished adding target */
 	ri = *((struct mptfc_rport_info **)rport->dd_data);
-	if (unlikely(ri->remap_needed))
-		return SCSI_MLQUEUE_HOST_BUSY;
+	if (unlikely(!ri)) {
+		dfcprintk ((MYIOC_s_INFO_FMT
+			"mptfc_qcmd.%d: %d:%d, dd_data is null.\n",
+			((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->name,
+			((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->sh->host_no,
+			SCpnt->device->id,SCpnt->device->lun));
+		SCpnt->result = DID_IMM_RETRY << 16;
+		done(SCpnt);
+		return 0;
+	}
 
-	return mptscsih_qcmd(SCpnt,done);
+	err = mptscsih_qcmd(SCpnt,done);
+#ifdef DMPT_DEBUG_FC
+	if (unlikely(err)) {
+		dfcprintk ((MYIOC_s_INFO_FMT
+			"mptfc_qcmd.%d: %d:%d, mptscsih_qcmd returns non-zero.\n",
+			((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->name,
+			((MPT_SCSI_HOST *) SCpnt->device->host->hostdata)->ioc->sh->host_no,
+			SCpnt->device->id,SCpnt->device->lun));
+	}
+#endif
+	return err;
 }
 
 static void
@@ -615,18 +631,17 @@ mptfc_rescan_devices(void *arg)
 	MPT_ADAPTER		*ioc = (MPT_ADAPTER *)arg;
 	int			ii;
 	int			work_to_do;
+	u64			pn;
 	unsigned long		flags;
 	struct mptfc_rport_info *ri;
 
 	do {
 		/* start by tagging all ports as missing */
-		spin_lock_irqsave(&ioc->fc_rport_lock,flags);
 		list_for_each_entry(ri, &ioc->fc_rports, list) {
 			if (ri->flags & MPT_RPORT_INFO_FLAGS_REGISTERED) {
 				ri->flags |= MPT_RPORT_INFO_FLAGS_MISSING;
 			}
 		}
-		spin_unlock_irqrestore(&ioc->fc_rport_lock,flags);
 
 		/*
 		 * now rescan devices known to adapter,
@@ -639,33 +654,24 @@ mptfc_rescan_devices(void *arg)
 		}
 
 		/* delete devices still missing */
-		spin_lock_irqsave(&ioc->fc_rport_lock, flags);
 		list_for_each_entry(ri, &ioc->fc_rports, list) {
 			/* if newly missing, delete it */
-			if ((ri->flags & (MPT_RPORT_INFO_FLAGS_REGISTERED |
-					  MPT_RPORT_INFO_FLAGS_MISSING))
-			  == (MPT_RPORT_INFO_FLAGS_REGISTERED |
-			      MPT_RPORT_INFO_FLAGS_MISSING)) {
+			if (ri->flags & MPT_RPORT_INFO_FLAGS_MISSING) {
 
 				ri->flags &= ~(MPT_RPORT_INFO_FLAGS_REGISTERED|
 					       MPT_RPORT_INFO_FLAGS_MISSING);
-				ri->remap_needed = 1;
-				fc_remote_port_delete(ri->rport);
-				/*
-				 * remote port not really deleted 'cause
-				 * binding is by WWPN and driver only
-				 * registers FCP_TARGETs but cannot trust
-				 * data structures.
-				 */
+				fc_remote_port_delete(ri->rport);	/* won't sleep */
 				ri->rport = NULL;
+
+				pn = (u64)ri->pg0.WWPN.High << 32 |
+				     (u64)ri->pg0.WWPN.Low;
 				dfcprintk ((MYIOC_s_INFO_FMT
 					"mptfc_rescan.%d: %llx deleted\n",
 					ioc->name,
 					ioc->sh->host_no,
-					ri->pg0.WWPN));
+					(unsigned long long)pn));
 			}
 		}
-		spin_unlock_irqrestore(&ioc->fc_rport_lock,flags);
 
 		/*
 		 * allow multiple passes as target state
@@ -870,10 +876,23 @@ mptfc_probe(struct pci_dev *pdev, const struct pci_device_id *id)
 		goto out_mptfc_probe;
 	}
 
-	for (ii=0; ii < ioc->facts.NumberOfPorts; ii++) {
-		mptfc_init_host_attr(ioc,ii);
-		mptfc_GetFcDevPage0(ioc,ii,mptfc_register_dev);
-	}
+	/* initialize workqueue */
+
+	snprintf(ioc->fc_rescan_work_q_name, KOBJ_NAME_LEN, "mptfc_wq_%d",
+		sh->host_no);
+	ioc->fc_rescan_work_q =
+		create_singlethread_workqueue(ioc->fc_rescan_work_q_name);
+	if (!ioc->fc_rescan_work_q)
+		goto out_mptfc_probe;
+
+	/*
+	 * scan for rports -
+	 *	by doing it via the workqueue, some locking is eliminated
+	 */
+
+	ioc->fc_rescan_work_count = 1;
+	queue_work(ioc->fc_rescan_work_q, &ioc->fc_rescan_work);
+	flush_workqueue(ioc->fc_rescan_work_q);
 
 	return 0;
 
@@ -949,8 +968,18 @@ mptfc_init(void)
 static void __devexit
 mptfc_remove(struct pci_dev *pdev)
 {
-	MPT_ADAPTER *ioc = pci_get_drvdata(pdev);
-	struct mptfc_rport_info *p, *n;
+	MPT_ADAPTER		*ioc = pci_get_drvdata(pdev);
+	struct mptfc_rport_info	*p, *n;
+	struct workqueue_struct *work_q;
+	unsigned long		flags;
+
+	/* destroy workqueue */
+	if ((work_q=ioc->fc_rescan_work_q)) {
+		spin_lock_irqsave(&ioc->fc_rescan_work_lock, flags);
+		ioc->fc_rescan_work_q = NULL;
+		spin_unlock_irqrestore(&ioc->fc_rescan_work_lock, flags);
+		destroy_workqueue(work_q);
+	}
 
 	fc_remove_host(ioc->sh);
 
diff --git a/drivers/message/fusion/mptscsih.c b/drivers/message/fusion/mptscsih.c
index 6a71b066bd2..84fa271eb8f 100644
--- a/drivers/message/fusion/mptscsih.c
+++ b/drivers/message/fusion/mptscsih.c
@@ -632,7 +632,11 @@ mptscsih_io_done(MPT_ADAPTER *ioc, MPT_FRAME_HDR *mf, MPT_FRAME_HDR *mr)
 
 		case MPI_IOCSTATUS_SCSI_DEVICE_NOT_THERE:	/* 0x0043 */
 			/* Spoof to SCSI Selection Timeout! */
-			sc->result = DID_NO_CONNECT << 16;
+			if (ioc->bus_type != FC)
+				sc->result = DID_NO_CONNECT << 16;
+			/* else fibre, just stall until rescan event */
+			else
+				sc->result = DID_REQUEUE << 16;
 
 			if (hd->sel_timeout[pScsiReq->TargetID] < 0xFFFF)
 				hd->sel_timeout[pScsiReq->TargetID]++;
@@ -1645,7 +1649,6 @@ int
 mptscsih_abort(struct scsi_cmnd * SCpnt)
 {
 	MPT_SCSI_HOST	*hd;
-	MPT_ADAPTER	*ioc;
 	MPT_FRAME_HDR	*mf;
 	u32		 ctx2abort;
 	int		 scpnt_idx;
@@ -1663,14 +1666,6 @@ mptscsih_abort(struct scsi_cmnd * SCpnt)
 		return FAILED;
 	}
 
-	ioc = hd->ioc;
-	if (hd->resetPending) {
-		return FAILED;
-	}
-
-	if (hd->timeouts < -1)
-		hd->timeouts++;
-
 	/* Find this command
 	 */
 	if ((scpnt_idx = SCPNT_TO_LOOKUP_IDX(SCpnt)) < 0) {
@@ -1684,6 +1679,13 @@ mptscsih_abort(struct scsi_cmnd * SCpnt)
 		return SUCCESS;
 	}
 
+	if (hd->resetPending) {
+		return FAILED;
+	}
+
+	if (hd->timeouts < -1)
+		hd->timeouts++;
+
 	printk(KERN_WARNING MYNAM ": %s: attempting task abort! (sc=%p)\n",
 	       hd->ioc->name, SCpnt);
 	scsi_print_command(SCpnt);
@@ -1703,7 +1705,7 @@ mptscsih_abort(struct scsi_cmnd * SCpnt)
 	vdev = SCpnt->device->hostdata;
 	retval = mptscsih_TMHandler(hd, MPI_SCSITASKMGMT_TASKTYPE_ABORT_TASK,
 		vdev->vtarget->bus_id, vdev->vtarget->target_id, vdev->lun,
-		ctx2abort, mptscsih_get_tm_timeout(ioc));
+		ctx2abort, mptscsih_get_tm_timeout(hd->ioc));
 
 	printk (KERN_WARNING MYNAM ": %s: task abort: %s (sc=%p)\n",
 		hd->ioc->name,
@@ -2521,15 +2523,15 @@ mptscsih_ioc_reset(MPT_ADAPTER *ioc, int reset_phase)
 
 		/* 7. FC: Rescan for blocked rports which might have returned.
 		 */
-		else if (ioc->bus_type == FC) {
-			int work_count;
-			unsigned long flags;
-
+		if (ioc->bus_type == FC) {
 			spin_lock_irqsave(&ioc->fc_rescan_work_lock, flags);
-			work_count = ++ioc->fc_rescan_work_count;
+			if (ioc->fc_rescan_work_q) {
+				if (ioc->fc_rescan_work_count++ == 0) {
+					queue_work(ioc->fc_rescan_work_q,
+						   &ioc->fc_rescan_work);
+				}
+			}
 			spin_unlock_irqrestore(&ioc->fc_rescan_work_lock, flags);
-			if (work_count == 1)
-				schedule_work(&ioc->fc_rescan_work);
 		}
 		dtmprintk((MYIOC_s_WARN_FMT "Post-Reset complete.\n", ioc->name));
 
@@ -2544,7 +2546,6 @@ mptscsih_event_process(MPT_ADAPTER *ioc, EventNotificationReply_t *pEvReply)
 {
 	MPT_SCSI_HOST *hd;
 	u8 event = le32_to_cpu(pEvReply->Event) & 0xFF;
-	int work_count;
 	unsigned long flags;
 
 	devtverboseprintk((MYIOC_s_INFO_FMT "MPT event (=%02Xh) routed to SCSI host driver!\n",
@@ -2569,10 +2570,13 @@ mptscsih_event_process(MPT_ADAPTER *ioc, EventNotificationReply_t *pEvReply)
 
 	case MPI_EVENT_RESCAN:				/* 06 */
 		spin_lock_irqsave(&ioc->fc_rescan_work_lock, flags);
-		work_count = ++ioc->fc_rescan_work_count;
+		if (ioc->fc_rescan_work_q) {
+			if (ioc->fc_rescan_work_count++ == 0) {
+				queue_work(ioc->fc_rescan_work_q,
+					   &ioc->fc_rescan_work);
+			}
+		}
 		spin_unlock_irqrestore(&ioc->fc_rescan_work_lock, flags);
-		if (work_count == 1)
-			schedule_work(&ioc->fc_rescan_work);
 		break;
 
 		/*