On Wed, 22 Sep 1999 brad@baileylink.net wrote:

> Why can't this guy just remove all the offending partitions with his disk
> partitioning software, quit to make sure the changes are saved, launch again
> and make new partitions?  Then use mkfs to change partitions and format
> these partitions.  Since he has the os on the other hd if he were to want to
> check that everything is good before doing the install he could boot into
> bsd and mount the disks to make sure that they are mounted as ufs instead of
> the ext2.
> 
> From what I understand about the problem the issue is with the partition
> table and dd'ing over the start of the partition is not going to fix that
> problem.
> 
> I am really asking this to understand the ideas involved.

The problem AFAICT is NOT with the partition table. I believe this for two
reasons: 1) we don't encode anything in the disklabel we fake up to say
it's ext2fs or not. 2) I've looked at the ext2fs & ffs mount codes, and
they keep their superblocks in seperate places.

Put another way, we end up with a filesystem which passes both the ext2fs
and the ffs validity tests, when in fact it's really an ffs.

As the reason for the superblock offset is to preserve space for boot
blocks, newfs does NOT overwrite this area (the ext2fs superblock) (for a
system with real boot blocks, machine'd stop booting at that point). 

We can readily test my idea. :-) If it's wrong, it's back to the drawing
board. :-)

Also, mounting the filesystem in the installer won't be a good test as,
AFAIK, the installer doesn't know about ext2fs's, so won't make the wrong
choice.

Take care,

Bill