* Tore Anderson

>  I can confirm that this patch solves my problem without any side
> effects (as far as I can tell).

  I'm sorry I have to retract this statement.  When I did some changes
 on my storage array, it generated RSCNs and therefore I/O briefly
 failed, causing timeouts and path failovers.

  I've attached the log of the event from the host running the patched
 kernel, and you'll see the -EEXIST message and call trace near the end.

  I'm sorry that I can't reproduce this much, because the RSCNs affect
 all hosts connected to the SAN, including a number of production
 systems which I can't touch more than necessary - unless if you know of
 any way to make the HBA drivers (qla2xxx/lpfc) to ignore all RSCN
 updates so that I/O is unaffected by them (I'd be forever grateful,
 too).

Regards
-- 
Tore Anderson