Subject: Re: Latest arm32 ksrc does not find the motherboard HD i/f
To: Dave Millen <firstname.lastname@example.org>
From: Richard Earnshaw <email@example.com>
Date: 10/27/1998 09:30:11
> I am presently running a 1.3H arm32(riscpc) kernel compiled on Oct. 12th from
> sources supped the previous night.
> This kernel correctly identifies the hard drive controllers and drives on both
> the motherboard(wdc0: wd0a......) and the ICS IDE interface(wdc1:
> wd1a...., wd2a.....). The system is set up to boot from wd0a and works fine.
> However ......
> I have compiled a new kernel from sources supped on Oct 24th and find that wdc0
> now appears to be the controller on the ICS IDE interface and that wd0 now
> refers to the drive on channel 0 of the same controller. The kernel
> obviously fails to boot, since the 'new' wd0a has nothing to boot, normally only
> being used as extra storage mounted on /mnt.
> Both *should* or *used to* refer to the controller and drive on the motherboard.
> The symptoms persist whether booted as a 'native kernel' or via unixfs:netbsd.
> The ICS interface is found at the point in the boot sequence where the
> motherboard controller normally comes up. This is much earlier than
> previously! and it seems as though the motherboard controller is not being
> found by the kernel.
> Has the order in which the devices are probed changed, resulting in a change of
> device numbers, or is this a serious fault?
> BTW, going back to my earlier kernel works fine.
My PR (6357) Contains a patch which should fix the motherboard driver
issues -- at least it does for me. Don't have an ICS card, so I can't
help there, though the fix may be similar -- there was a similar problem
on the Sharks as well.