Re: SCSI disk naming

david parsons (o.r.c@p.e.l.l.p.o.r.t.l.a.n.d.o.r.u.s)
14 Jul 1997 12:32:36 -0700

In article <linux.kernel.m0wnNTI-00091BC@seneca>,
Harald Milz <hm@seneca.muc.de> wrote:
>david parsons (o.r.c@p.e.l.l.p.o.r.t.l.a.n.d.o.r.u.s) wrote:
>> that leaves 13 bits for minor devices. Rest assured you will have
>> to do some jumping through hoops to deal with these properly in the
>> kernel, but 13 bits gives you 4 bits for ID, 3 bits for LUN, 3 bits
>
>3 bits for the LUN is definitely not enough for SSA which we will see not
>too far from now.

I'd suggest not mapping SSA onto the parallel SCSI device driver
family; It's different enough from parallel SCSI to cause massive
confusion if you try to shoehorn it into the command set -- for
instance, the SCSI-2 command set is saturated with commands which
have 3 bits for LUN, so a SSA architecture that attempts to map disks
to LUNS will be unpleasantly surprised when you put more than 8 disks
on a chain.

SSA allows, what, 127 devices on a chain? You'll not fit that
into a 3-bit LUN with any degree of success (but you will be
able to fit that into 4 bits of ID and 3 bits of LUN; I suspect
that's how the mapping would go onto a SCSI-2 command set, and
that, at the very least, would be compatable with the addressing
scheme I proposed in my original message.)

____
david parsons \bi/ now if you have more than 8 partitions on a disk,
\/ then you could have trouble in river city.