[PATCH 1/1] fc class: fix scanning when devs are offline

[PATCH 1/1] fc class: fix scanning when devs are offline

[michaelc]

From: Mike Christie <michaelc@cs.wisc.edu>

When a rport is added back or the role is changed the fc class
will queue a scan and then call scsi_target_unblock. The problem
with this is if the devices are in the SDEV_OFFLINE state and
the scan is run before the scsi_target_unblock, then the scan
will see LUN0 as offline and the scan will fail. This patch moves
the unblock call to before the scan, so we know the device state
will be set correctly when the scan is run.
---
 drivers/scsi/scsi_transport_fc.c |   24 ++++++++++++++----------
 1 files changed, 14 insertions(+), 10 deletions(-)

diff --git a/drivers/scsi/scsi_transport_fc.c b/drivers/scsi/scsi_transport_fc.c
index 80fbe2a..5797604 100644
--- a/drivers/scsi/scsi_transport_fc.c
+++ b/drivers/scsi/scsi_transport_fc.c
@@ -2808,17 +2808,20 @@ fc_remote_port_add(struct Scsi_Host *shost, int channel,
 						  FC_RPORT_DEVLOSS_PENDING |
 						  FC_RPORT_DEVLOSS_CALLBK_DONE);
 
+				spin_unlock_irqrestore(shost->host_lock, flags);
+
 				/* if target, initiate a scan */
 				if (rport->scsi_target_id != -1) {
+					scsi_target_unblock(&rport->dev);
+
+					spin_lock_irqsave(shost->host_lock,
+							  flags);
 					rport->flags |= FC_RPORT_SCAN_PENDING;
 					scsi_queue_work(shost,
 							&rport->scan_work);
 					spin_unlock_irqrestore(shost->host_lock,
 							flags);
-					scsi_target_unblock(&rport->dev);
-				} else
-					spin_unlock_irqrestore(shost->host_lock,
-							flags);
+				}
 
 				fc_bsg_goose_queue(rport);
 
@@ -2876,16 +2879,17 @@ fc_remote_port_add(struct Scsi_Host *shost, int channel,
 			if (fci->f->dd_fcrport_size)
 				memset(rport->dd_data, 0,
 						fci->f->dd_fcrport_size);
+			spin_unlock_irqrestore(shost->host_lock, flags);
+
+			if (ids->roles & FC_PORT_ROLE_FCP_TARGET) {
+				scsi_target_unblock(&rport->dev);
 
-			if (rport->roles & FC_PORT_ROLE_FCP_TARGET) {
 				/* initiate a scan of the target */
+				spin_lock_irqsave(shost->host_lock, flags);
 				rport->flags |= FC_RPORT_SCAN_PENDING;
 				scsi_queue_work(shost, &rport->scan_work);
 				spin_unlock_irqrestore(shost->host_lock, flags);
-				scsi_target_unblock(&rport->dev);
-			} else
-				spin_unlock_irqrestore(shost->host_lock, flags);
-
+			}
 			return rport;
 		}
 	}
@@ -3083,12 +3087,12 @@ fc_remote_port_rolechg(struct fc_rport  *rport, u32 roles)
 		/* ensure any stgt delete functions are done */
 		fc_flush_work(shost);
 
+		scsi_target_unblock(&rport->dev);
 		/* initiate a scan of the target */
 		spin_lock_irqsave(shost->host_lock, flags);
 		rport->flags |= FC_RPORT_SCAN_PENDING;
 		scsi_queue_work(shost, &rport->scan_work);
 		spin_unlock_irqrestore(shost->host_lock, flags);
-		scsi_target_unblock(&rport->dev);
 	}
 }
 EXPORT_SYMBOL(fc_remote_port_rolechg);
-- 
1.7.7.6

Re: [PATCH 1/1] fc class: fix scanning when devs are offline

[Mike Christie]

On 04/15/2012 09:47 PM, michaelc@cs.wisc.edu wrote:
> From: Mike Christie <michaelc@cs.wisc.edu>
> 
> When a rport is added back or the role is changed the fc class
> will queue a scan and then call scsi_target_unblock. The problem
> with this is if the devices are in the SDEV_OFFLINE state and
> the scan is run before the scsi_target_unblock, then the scan
> will see LUN0 as offline and the scan will fail. This patch moves
> the unblock call to before the scan, so we know the device state
> will be set correctly when the scan is run.

I forgot to add my signed off for the patch.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>