scsi: aacraid: Reload offlined drives after controller reset
authorRaghava Aditya Renukunta <RaghavaAditya.Renukunta@microsemi.com>
Thu, 16 Feb 2017 20:51:18 +0000 (12:51 -0800)
committerMartin K. Petersen <martin.petersen@oracle.com>
Wed, 22 Feb 2017 23:41:42 +0000 (18:41 -0500)
During the IOP reset stress testing, it was found that the drives can be
marked offline when the adapter controller crashes and IO's are running
in parallel. When the controller  does come back from the reset, the drive
that is marked offline is not exposed.

Fixed by removing and adding drives that are marked offline. In addition
invoke a scsi host bus rescan to capture any additional configuration
changes.

Signed-off-by: Raghava Aditya Renukunta <RaghavaAditya.Renukunta@microsemi.com>
Reviewed-by: David Carroll <David.Carroll@microsemi.com>
Reviewed-by: Johannes Thumshirn <jthumshirn@suse.de>
Signed-off-by: Martin K. Petersen <martin.petersen@oracle.com>
drivers/scsi/aacraid/commsup.c

index eb4d8cfc31945cec09f4375a0df92ab5d8756b2d..1f716c01708bc591f156e61e99d85463ddd3c7ea 100644 (file)
@@ -1637,11 +1637,29 @@ static int _aac_reset_adapter(struct aac_dev *aac, int forced, u8 reset_type)
                command->SCp.phase = AAC_OWNER_ERROR_HANDLER;
                command->scsi_done(command);
        }
+       /*
+        * Any Device that was already marked offline needs to be cleaned up
+        */
+       __shost_for_each_device(dev, host) {
+               if (!scsi_device_online(dev)) {
+                       sdev_printk(KERN_INFO, dev, "Removing offline device\n");
+                       scsi_remove_device(dev);
+                       scsi_device_put(dev);
+               }
+       }
        retval = 0;
 
 out:
        aac->in_reset = 0;
        scsi_unblock_requests(host);
+       /*
+        * Issue bus rescan to catch any configuration that might have
+        * occurred
+        */
+       if (!retval) {
+               dev_info(&aac->pdev->dev, "Issuing bus rescan\n");
+               scsi_scan_host(host);
+       }
        if (jafo) {
                spin_lock_irq(host->host_lock);
        }