Global ID sources:

inquiry vital product info

mode sense unit serial # page (80h)

mode sense device identification page (only one with ww unique defs) (83h)

All these are optional, and inquiry data may depend on device spin-up.

On lookup:

send inquiry
  if unit exists, send start
  loop on test unit ready till ready or timeout
  inquiry again, read data

Alternatively one could send mode sense, but the same spinup need seems
to exist.



The problem here is that world unique IDs are optional so that both
techniques will need to be tried, and that moreover the unique ID
will require drives to be spun up before the information will be
available. 

It is possible that some of this extra delay time can be avoided,
and autoconfigure use inquiry only, where either:


* Sufficient information exists to unambiguously identify a device
	without a world unique ID, or
* A site unique ID can be obtained, e.g. from FC protocols, which can
	be used instead during configuration.

The use of a site ID is complicated by the fact that fibre channel devices
are expected to be on two loops. This reportedly means that the addresses
returned will be identical on the two loops, and we must keep track
"by hand" that they are in fact the same.
   The failover switching driver techniques can be used to do this,
choosing a primary connection at first and switching as needed. A more
dynamic connection can be handled with similar techniques once some
boundary conditions of i/o rundown are dealt with.