Quirk for this?

[Scott Long]

Warner Losh wrote:
>> It may be a property specific to USB devices, but the code affected is
>> a property of the end target at the end of a transport, not the
>> transport itself.
> 
> The transport already does so much futzing with the scsi
> request/replies that go down to the device that one more wouldn't be
> so horrible and would confine this brain-damage to the devices that
> have it...
> 
> Warner

Strongly agreed.  If the fudging only needed to happen at the 
READ_CAPACITY that is done for device creation in GEOM then having
it be in the da driver would be appropriate.  But since it needs to
happen for all READ_CAPACITY requests, it's much better to have it
isolated in the umass driver.

Scott