Subject: Re: kern/18481: Someone broke RAIDframe autoconfig
To: Manuel Bouyer <bouyer@antioche.eu.org>
From: Greg Oster <oster@cs.usask.ca>
List: netbsd-bugs
Date: 10/01/2002 12:21:20
Manuel Bouyer writes:
> On Mon, Sep 30, 2002 at 04:19:52PM -0600, oster@cs.usask.ca wrote:
> > NetBSD rizzo 1.6I NetBSD 1.6I (GENERIC) #0: Mon Sep 30 16:05:41 CST 2002   
>   oster@merlin:/cheetah/current2/src/sys/arch/i386/compile/GENERIC i386
> > Architecture: i386
> > Machine: i386
> > >Description:
> > 	Someone has Changed Things so that RAIDframe's autoconfig now
> > occurs *before* most of the disk devices are present.  This means that
> > when raidattach() is called, and the autoconfig code is run, there
> > arn't any disks around, and the autoconfig bits won't find anything to
> > configure!!!! 
> 
> Do you have SCSI disks ?

Yes.

> I guess this is related to recent changes to have SCSI disks probed from the
> kernel threads instead of proc0.

Yup.

> We probably need to have a global flag (or better: counter) which tells when
> the disks autoconfig is done. We need this because I'd also like to defer IDE
> devices probing.
> What, actually, make sure that raidframe autoconfig has run before
> cpu_rootconf() is called ?

See Jason's latest changes.. it takes care of that :)

Later...

Greg Oster