[SCSI] qla2xxx: use new host byte transport errors.
authorMike Christie <michaelc@cs.wisc.edu>
Tue, 19 Aug 2008 23:45:29 +0000 (18:45 -0500)
committerJames Bottomley <James.Bottomley@HansenPartnership.com>
Mon, 13 Oct 2008 13:28:51 +0000 (09:28 -0400)
This has qla2xxx use the new transport error values instead of
DID_BUS_BUSY. I am not sure if all the errors
in qla_isr.c I changed are transport related. We end up blocking/deleting
the rport for all of them so it is better to use the new transport error since
the fc classs will decide when to fail the IO.

With this patch if I pull a cable then IO that had reached
the driver, will be failed with DID_TRANSPORT_DISRUPTED (not including
tape). The fc class will then fail the IO when the fast io fail tmo
has fired, and the driver will flush any other commands running.

Signed-off-by: Mike Christie <michaelc@cs.wisc.edu>
Cc: Andrew Vasquez <andrew.vasquez@qlogic.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
drivers/scsi/qla2xxx/qla_isr.c

index fc4bfa7f839c8648e021f643387027d1f67cf1d2..a76efd99d0070fa7d61ee43a5c8d2c539c5db3c6 100644 (file)
@@ -1187,7 +1187,12 @@ qla2x00_status_entry(scsi_qla_host_t *ha, void *pkt)
                    cp->serial_number, comp_status,
                    atomic_read(&fcport->state)));
 
-               cp->result = DID_BUS_BUSY << 16;
+               /*
+                * We are going to have the fc class block the rport
+                * while we try to recover so instruct the mid layer
+                * to requeue until the class decides how to handle this.
+                */
+               cp->result = DID_TRANSPORT_DISRUPTED << 16;
                if (atomic_read(&fcport->state) == FCS_ONLINE)
                        qla2x00_mark_device_lost(fcport->ha, fcport, 1, 1);
                break;
@@ -1214,7 +1219,12 @@ qla2x00_status_entry(scsi_qla_host_t *ha, void *pkt)
                break;
 
        case CS_TIMEOUT:
-               cp->result = DID_BUS_BUSY << 16;
+               /*
+                * We are going to have the fc class block the rport
+                * while we try to recover so instruct the mid layer
+                * to requeue until the class decides how to handle this.
+                */
+               cp->result = DID_TRANSPORT_DISRUPTED << 16;
 
                if (IS_FWI2_CAPABLE(ha)) {
                        DEBUG2(printk(KERN_INFO