[SCSI] don't call ips_eh_reset in ips_queue to avoid deadlock

When the locking was changed in the eh code ips_eh_reset was changed
so that it was a wraper around __ips_eh_reset and all ips_eh_reset
does is grab the host lock and then calls __ips_eh_reset.

In the queuecommand, ips_queue is called with the host_lock held so if
it calls ips_eh_reset we will have a problem. This patch just has
ips_queue call __ips_eh_reset.

Patch is only compile tested. I do not have the HW.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Acked-by: Hammer, Jack <Jack_Hammer@adaptec.com>
Signed-off-by: James Bottomley <James.Bottomley@SteelEye.com>

drivers/scsi/ips.c

@@ -1146,7 +1146,7 @@ ips_queue(Scsi_Cmnd * SC, void (*done) (Scsi_Cmnd *))
 				return (0);
 			}
 			ha->ioctl_reset = 1;	/* This reset request is from an IOCTL */
-			ips_eh_reset(SC);
+			__ips_eh_reset(SC);
 			SC->result = DID_OK << 16;
 			SC->scsi_done(SC);
 			return (0);