George Michaelson wrote:
> for 64k, I really do suggest that we could set aside some space. people doing
> this kind of thing for umass could stipulate the <n> instances they need, if
> they need to do umass0 to umass1 copies for instance.
> 
> but surely a modern UNIX has to be able to do this at runtime? and, if this is
> a problem for umass, why isn't it a problem for ieee394, or any other detachable
> device interface? i2c? any of them? (or any other u* USB device?)
None of the other u* devices need large contiguous buffers (ulpt did,
but I changed it).  In actual fact, the umass driver doesn't need
contiguous memory either, but it's much simpler that way.  And it used
to work fine.

Reserving memory would be possible, but it's a nasty hack that I would
rather avoid.

	-- Lennart