On Mon, 24 Jan 2005, Steven M. Bellovin wrote:

>> mbrlabel did it. Thanks. However now I have something strange
>> (disklabel):
>>
>>  a:   4915953        63     4.2BSD   1024  8192    86   # (Cyl.    0*-
>> 4876)
>>  b:     88704   4919040       swap                      # (Cyl. 4880 -
>> 4967)
>>  c:     16384        63     4.2BSD   1024  8192    16   # (Cyl.    0*-
>> 16*)
>>
>> Note that c: is overlapping a:, yet I get no warning. Huh? I will
>> disable c: to prevent accidents. Any comments ?
>
> An overlap is normal, but your is weird -- normally, the c partition is
> the entire NetBSD portion of the drive, but yours is too small.  You
> should also have a d partition that covers the entire drive, starting
> at offset 0.
>
> I'm not sure what you mean by 'disable c:' (and it's c, not c:) -- you
> certainly shouldn't mount

Disable c: means setting size and offset to 0, and type to unknown, for 
the c: entry in disklabel -e. This makes the relevant slice 'disappear' 
from the normal listing and avoids mounting accidents. I hae done this.

Why is an overlap normal ? Overlaps are not supposed to happen and they 
are supposed to be flagged as errors (I think I saw such an error before 
I did mbrlabel).

The table above is incomplete (I clipped it). The d: slice represents 
the entire disk acc. disklabel manual page, at least for i386 machines. 
It exists and it is correct in my disklabel.

There are also two other slices, e: and f: which contain a Linux Ext2fs 
partition each. netbsd (1.6.1) refuses to mount them with IO errors and 
coredumps on fsck_ext2fs. The partitions are correct (I mounted the disk 
under another machine).

Are interoperability problems known with 1.6.1 ? I already know that 
writing to a ffs slice mounted under linux wreaks havoc (linux machine 
locks up hard). Now it seems that the other direction is also blocked, 
but, strangely, I was able to read and write to/from a ext2fs partition 
before, on *another* disk (wd1 vs wd0 which I am working on now).

Peter