[23] in linux-scsi channel archive
Re: Probing luns > 0
daemon@ATHENA.MIT.EDU (Andrew Walker)
Wed Jan 25 05:16:50 1995
From: andy@eng.kvaerner.no (Andrew Walker)
To: linux-scsi@vger.rutgers.edu
Date: Wed, 25 Jan 1995 10:23:42 +0100 (MET)
In-Reply-To: <199501250619.AA01985@rasty.anu.edu.au> from "Paul Gortmaker" at Jan 25, 95 05:19:56 pm
>
> >
> > From: Rik Faith <faith@cs.unc.edu>
> > Subject: Re: Default, no probing of luns > 0
> > Date: Tue, 24 Jan 1995 10:31:09 -0500
> >
> >
> > On Mon 23 Jan 1995 11:17:06 MST,
> > Drew Eckhardt <drew@boulder.openware.com> wrote:
> >
> > > There are still large numbers of broken SCSI devices out there which
> > > are extremely unhappy (ie, lock the SCSI bus up, do an unexpected
> > > disconnect, etc) when you attempt to talk to them at a LUN other
> > > than 0.
> > >
> > > Since very few devices use multiple LUNs (SCSI bridge boards,
> > > fan out adapters, jukeboxes, and CD changers are the only
> > > ones to come to mind), I think that it might be reasonable
> > > to make the default maximum LUN searched 0 instead of 7.
> > >
> > > We could scan for the additional LUNS only when
> > > - The user used the max_scsi_luns= command line
> > > option.
> > > - Certain vendor & model combinations were returned
> > > in result for the IDENTIFY command sent to
> > > LUN 0 of a given target.
> > > Comments?
> >
> > I think this is an excellent idea and will alleviate many of the buf
> > reports and problems that LUN scanning causes.
>
> Another thing that nobody has mentioned (yet) is that we can then throw the
> huge blacklist table in the garbage, which will de-bloat scsi.c
> That is a major plus IMHO. Go for it.
>
> Paul.
Yeah, that's fine until the 'whitelist' of units which *should* be probed
for multiple LUNs gets as long as (or longer than) the current 'blacklist'.
Now, I'm not saying this will happen, and maybe folks will be happy to use
a 'max_scsi_luns=' command line (what do you say Pioneer owners? :-) so we
wouldn't need a new list. In principle I'm very much *for* Drew's suggestion.
The problem with a simple command line option is that you still couldn't
mix, for example, a Pioneer CD changer and one of the duff Maxtor disks.
Of course, the command line syntax could be made a bit more complicated than
just setting the max LUN to probe for. So maybe we're back to a table.
One suggestion is a command line like:
max_scsi_luns=07007070
where each digit is the max. LUN for devices 0-7, but this basically codes
in all the devices on the bus and makes on the fly reconfiguration slightly
trickier.
Alternatively, pair the options for a slightly more dynamic solution:
max_scsi_luns=3757
saying that max. LUN is 7 for SCSI id's 3 and 5.
Now we just have to find out what's most acceptable to most of the people.
Just being a wet blanket as usual :-(
-Andy
--
Andy Walker Kvaerner Engineering a.s.
Andrew.Walker@eng.kvaerner.no P.O. Box 222, N-1324 Lysaker, Norway
......if the answer isn't violence, neither is it silence......
