[SCSI] qla2xxx: Proper completion to scsi-ml for scsi status task_set…

…_full and busy.

In case of firmmware detected under-run condition and scsi status of
task_set_full or busy_condition, return that to the mid layer for proper error
handling instead of DID_ERROR (which causes error handler activation and a
full retry).

Signed-off-by: Giridhar Malavali <giridhar.malavali@qlogic.com>
Signed-off-by: Chad Dupuis <chad.dupuis@qlogic.com>
Signed-off-by: James Bottomley <JBottomley@Parallels.com>

drivers/scsi/qla2xxx/qla_dbg.c

@@ -15,7 +15,7 @@
  * | Mailbox commands             |       0x113e       | 0x112c-0x112e  |
  * |                              |                    | 0x113a         |
  * | Device Discovery             |       0x2086       | 0x2020-0x2022  |
- * | Queue Command and IO tracing |       0x302f       | 0x3006,0x3008  |
+ * | Queue Command and IO tracing |       0x3030       | 0x3006,0x3008  |
  * |                              |                    | 0x302d-0x302e  |
  * | DPC Thread                   |       0x401c       |		|
  * | Async Events                 |       0x505d       | 0x502b-0x502f  |

drivers/scsi/qla2xxx/qla_isr.c

@@ -1715,13 +1715,24 @@ qla2x00_status_entry(scsi_qla_host_t *vha, struct rsp_que *rsp, void *pkt)
 				res = DID_ERROR << 16;
 				break;
 			}
-		} else {
+		} else if (lscsi_status != SAM_STAT_TASK_SET_FULL &&
+			    lscsi_status != SAM_STAT_BUSY) {
+			/*
+			 * scsi status of task set and busy are considered to be
+			 * task not completed.
+			 */
+
 			ql_dbg(ql_dbg_io, fcport->vha, 0x301f,
 			    "Dropped frame(s) detected (0x%x "
-			    "of 0x%x bytes).\n", resid, scsi_bufflen(cp));
+			    "of 0x%x bytes).\n", resid,
+			    scsi_bufflen(cp));
 
 			res = DID_ERROR << 16 | lscsi_status;
 			goto check_scsi_status;
+		} else {
+			ql_dbg(ql_dbg_io, fcport->vha, 0x3030,
+			    "scsi_status: 0x%x, lscsi_status: 0x%x\n",
+			    scsi_status, lscsi_status);
 		}
 
 		res = DID_OK << 16 | lscsi_status;