[SCSI] zfcp: Recover from stalled outbound queue
authorChristof Schmitt <christof.schmitt@de.ibm.com>
Mon, 13 Jul 2009 13:06:10 +0000 (15:06 +0200)
committerJames Bottomley <James.Bottomley@HansenPartnership.com>
Thu, 30 Jul 2009 13:49:57 +0000 (08:49 -0500)
Depending on interruptions on some storage systems, the complete
channel can stall which looks like an outbound queue stall to Linux.
When trying to acquire a free SBAL for a non-SCSI command, zfcp waits
for 5 seconds for a free slot to appear. This is the right place to
detect a queue stall: If the wait times out, we assume a stalled queue
and try to recover this.

The overall strategy should be to trigger the erp from specific
events, and not try an overall escalation from one failed port to a
full-blown queue recovery. If we manage to send a command, the status
codes for this command or a timeout will trigger the right follow-on
actions.

Reviewed-by: Swen Schillig <swen@vnet.ibm.com>
Signed-off-by: Christof Schmitt <christof.schmitt@de.ibm.com>
Signed-off-by: James Bottomley <James.Bottomley@HansenPartnership.com>
drivers/s390/scsi/zfcp_fsf.c

index 0c24695..b7e4884 100644 (file)
@@ -670,8 +670,11 @@ static int zfcp_fsf_req_sbal_get(struct zfcp_adapter *adapter)
                               zfcp_fsf_sbal_check(adapter), 5 * HZ);
        if (ret > 0)
                return 0;
-       if (!ret)
+       if (!ret) {
                atomic_inc(&adapter->qdio_outb_full);
+               /* assume hanging outbound queue, try queue recovery */
+               zfcp_erp_adapter_reopen(adapter, 0, "fsrsg_1", NULL);
+       }
 
        spin_lock_bh(&adapter->req_q_lock);
        return -EIO;