[Top] [All Lists]

Re: fdisk/kernel oddity

To: Guido Guenther <>
Subject: Re: fdisk/kernel oddity
From: Keith M Wesolowski <>
Date: Wed, 25 Oct 2000 10:14:53 -0700
In-reply-to: <>; from on Wed, Oct 25, 2000 at 07:01:29PM +0200
References: <>
User-agent: Mutt/1.2i
On Wed, Oct 25, 2000 at 07:01:29PM +0200, Guido Guenther wrote:

> After some writing to it( cp -a /usr /mnt), I see the following:
> Oct 25 16:00:37 bert kernel: 08:11: rw=0, want=1088320776, limit=533919
> Oct 25 16:00:37 bert kernel: attempt to access beyond end of device
> [..snip..]   
> Oct 25 16:00:37 bert kernel: 08:11: rw=0, want=138373542, limit=533919
> Oct 25 16:01:27 bert kernel: [rm:9357] Illegal instruction at 004059a4 
> ra=004028d8

Not to sound defensive, but I'm fairly sure this isn't an
fdisk-related problem. The partitions that fdisk creates follow the
spec as far as I can see.

> 08:11 is /dev/sda11 which is the "SGI volume", according to fdisk -l:
>   Device  Info      Start       End   Sectors  Id  System
>   /dev/sda1  boot         6       937   3872540  83  Linux native
>   /dev/sda2  swap       938      1009    298850  83  Linux native
>   /dev/sda9               0         4     20770   0  SGI volhdr
>   /dev/sda11              0      1008   4191386   6  SGI volume
> So the requsted block is *far* out of bounds.

These partitions look quite sane. I wonder, though, if there's
something about your specific partitioning that is confusing the
kernel's SGI disklabel support.

> What puzzles me even more is that I get illegal instructions for almost 
> all commands I execute afterwards. Any comments on this one?

I rather suspect that this is the same problem that causes the request
for the out-of-bounds block in the first place: kernel memory
corruption. Unfortunately I have few ideas as to what the specific
problem is. I would start bug-hunting in the sgi disklabel kernel
parts. Make sure that it's compatible with what fdisk is doing.

Keith M Wesolowski            
University of Nevada          
Chemistry Department Systems and Network Administrator

<Prev in Thread] Current Thread [Next in Thread>