Subject: Re: bad144, take two
To: Takahiro Kambe <firstname.lastname@example.org>
From: Ken Hornstein <email@example.com>
Date: 07/17/2001 13:19:46
>> - Fix bad144(8) on the i386 so it agrees with the disklabel method of
>> calculating the location of the error table. I think at the end of
>> the disk (rather than the end of the c partition) is a more
>> reasonable place for it.
>On i386 (really PC/AT clone), NetBSD might be installed in FDISK
>partition ("slice" called by FreeBSD people).
> partition "d": whole disk drive
> partition "c": NetBSD using area
I've always felt this was a HUGE mistake on the i386 port; it's not
consistant with anything else and always causes a ton of confusion to
new users. I'm not sure what the right thing to do _is_, though.
>In this case, NetBSD shouldn't modify or rely on other than the
>partition "c" (exception is fdisk(8) and MBR). Since bad144 remapping
>is valid only for NetBSD, it should be complete in NetBSD using area.
Havard said the same thing to me privately, but let me explain my thinking.
The whole concept of the "NetBSD using area" in disklabel is in my
thinking rather silly; disklabel knows about the "whole disk", and you
can even use partitions outside of "c" with NetBSD just fine. Tieing
the bad144 area to the magical "c" partition not only makes partitioning
the disk harder, but also prevents easy repartitioning of the disk
while keeping bad144 information intact. If we ever migrate away from
c/d on i386, that's just something else that's going to lose.
I know that there's concern that other operating systems don't know about
bad144; that has a simple solution: Simply leave off the last track (or
two) when you're doing fdisk/disklabel partitioning. That will make
it even easier, since you will want both fdisk and disklabel partitions to
end on the same place.
Since this has apparantly been broken since the beginning of the NetBSD
epoch, I'm not really worried about maintaining compatibility with
previous installations. But I am willing to hear more opinions about