Colin Wood <cwood@ichips.intel.com> wrote:

>Bob Nestor wrote:
>> Also on one of my new disks I see that the Installer 1.1e reports 9 
>> partitions, but Installer 1.1g reports only 7. Attempting to access this 
>> disk under NetBSD with either disklabel or newfs causes a system crash 
>> (illegal instruction) at the end of the process. Is it possible there's 
>> an off-by-one error in the fake disklabel code?  I'm about to test this 
>> by further reducing the partiton count on the disk.
>
>i'd say quite possible.  well, at least something seems to be screwy, b/c
>i've noticed that the more you mess with the partition map, the less
>likely netbsd is going to be able to read the disk.  i resized and deleted
>a couple of partitions on one of my disks, and i'm pretty sure i cannot
>access the original netbsd partition on the disk now; it panics the
>machine to do a disklabel on that drive :-(  i haven't looked at it in a
>while now, tho.

This particular disk that's giving me fits was formatted on my PPC 
machine using Drive Setup 1.5.  For some reason it has a tendency of 
leaving a couple of danging "garbage" partitions on the disk much like HD 
Setup does.  I'm also getting warnings from some of the other disk 
formatters that the Partition Map has some "minor" problems.  Having 
looked at this in the past I seem to remember that the Partition Map size 
doesn't agree with the size actually allocated.  Too bad we (the NetBSD 
"we") haven't been monitoring the disk formatter that folks have used 
when they're reported problems with the utilities and NetBSD setup.  I'm 
beginning to suspect that some formatters may be more prone to tripping 
up NetBSD than others are.

Oh, scratch what I wrote about the transient I/O errors and the 
Installer.  Just had the same round of errors on a brand-new disk drive, 
so it's probably something else.

-bob