Re: slice weirdness

From: Allan Fields <bsd_at_afields.ca>
Date: Mon, 9 Aug 2004 03:47:32 -0400
On Sun, Aug 08, 2004 at 11:13:37PM +0200, Stijn Hoop wrote:
> Hi,
> 
> while trying to migrate my FreeBSD -CURRENT partition to another disk, I keep
> running into a slice weirdness issue which makes the kernel unable to find
> it's root fs. It seems that something about the partition table is fishy such
> that GEOM doesn't find both slices:

Which kernel: the one from your existing install, the install kernel
(booting off CD), a new kernel?

I've had some partition/slice issues in the past w/ install CD and
missing device entries with non-standard slice (anything except s1
entry), can't recall details, but running sysinstall w/ newfs
flag to N creates appropriate entries.  Else I've found suitable
device entries elsewhere or make them temporarily with mknod.
Maybe this was only an issue in -stable.

> So where's my /dev/ad1s2?
> 
> The disk layout is ad1s1 is my Windows partition, ad1s2 my targeted new
> partition for the FreeBSD installation currently residing at ad0s1.

Out of curiosity, did you disklabel the second slice anew or just
copy/dd existing slice over to new drive?

> I first created ad1s2 by hand using fdisk, but got the exact same result.  The

But, does a newly created slice/disklabel/root filesystem have the same problem?
i.e. if you were to go the sysinstall route and do a fresh install
on ad1, exhibit same behaviour?

> script above shows the values that I obtained when /sbin/sysinstall
> partitioned the drive. After partitioning the device nodes reappear, and I was
> able to install{kernel,world} with DESTDIR pointing to the newly mounted
> ad1s2, but the device nodes disappear after having booted the newly installed
> slice. That boot ends with the kernel unable to find the root file system
> ad1s2a (which is not strange given the above).
> 
> Am I looking at some sort of geometry bug? I've tried setting the BIOS
> geometry settings to LBA (from Auto), that didn't make a difference.  Setting
> them to CHS produced an unbootable Windows so I reverted that. In any case
> I thought that those values were of historical interest only...
> 
> Any clues?
> 
> --Stijn
> 
> PS: I thought that there was a sysctl that showed the GEOM topology in XML;
>     however I was unable to find it in sysctl -a. Is it still around?

I believe so, phk recently (few months back) mentioned it's staying in.

> -- 
> The rain it raineth on the just
> 	And also on the unjust fella,
> But chiefly on the just, because
> 	The unjust steals the just's umbrella.


-- 
 Allan Fields, AFRSL - http://afields.ca
 2D4F 6806 D307 0889 6125  C31D F745 0D72 39B4 5541

Received on Mon Aug 09 2004 - 05:47:33 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:38:05 UTC