Subject: Re: Another changer, another changer problem
To: NetBSD-current Discussion List <current-users@netbsd.org>
From: Greg A. Woods <woods@most.weird.com>
List: current-users
Date: 10/03/1998 00:23:06
[ On Fri, October 2, 1998 at 20:27:21 (-0700), Curt Sampson wrote: ]
> Subject: Re: Another changer, another changer problem 
>
> Well, it's not likely you're going to. Given an Adaptec and a
> Buslogic controller, each on the PCI bus can you tell me which one
> will be attached first? Will the order of the attaching change if
> you reverse the slots? If the two controllers are the same, can
> you tell from the motherboard which slot is the lowest numbered?
> If your controllers are on different buses, it gets even worse
> because you have to know the bus attach order, which may vary
> depending on machine architecture and BIOS/Boot PROM. And Lord help
> you if you're on a bus such as P&P ISA, where the probe order is
> indeterminate; each card in that case has its own driver-dependent
> order to attach in.

Yeah, well I guess that's why they call it "plug & pray" and why I've
avoided it in general to date...

I'd not yet thought of the problem of multiple SCSI controllers on PCI,
or the potential problem of multiple PCI busses, etc.

I'd kinda hoped there would be a predictable order, just as there is
supposedly for NICs (i.e. the ethernet address sort order) (which of
course currently only works for NICs of all the same type, but is
probably easier to fix generically than it would be to fix for multiple
SCSI host adapters of different types).

I really do like jumpers and switches (or hard-coded slot order or some
other predictable physical identifier that's out of the meddling hands
of the firmware and/or driver authors).

> Well, you Just Can't; see above.

Perhaps not on machines with ISA or PCI or other simplistic busses.

What does Digital Unix do?  That's one platform I know where the big
monster servers have *many* busses and the potential for dozens, or
hundreds, of similar controllers?

> > LUNs are simply logical volumes -- equivalent to partitions but managed
> > by the external device, not the device driver in the kernel.
> 
> Well, if you're going to take the view that the two disks, one to
> a LUN, on the Emulex controller in my Sun shoebox are `logical
> volumes,' why not say that both disks on the SCSI controller in my
> PC are `logical volumes'?

In the context of RAID, which is where my sentence above belongs, LUNs
are "logical", but in the context of SCSI-ESDI controllers or similer
then they're obviously representing "physical" devices.

-- 
							Greg A. Woods

+1 416 218-0098      VE3TCP      <gwoods@acm.org>      <robohack!woods>
Planix, Inc. <woods@planix.com>; Secrets of the Weird <woods@weird.com>