Re: MSDOS vs diskabel regression in -4, -5 and -current

[Stephen Borrill <netbsd%precedence.co.uk@localhost>]

On Fri, 6 Mar 2009, Stephen Borrill wrote:
> > The code is validating the mbr data against the reported media size,
> > detecting that it is invalid so ignoring the mbr.
>
> I've been sent one of the offending cameras now.
>
> I used dd to clear the camera's memory and then fdisk to partition it 
> (accepting defaults in the main). This started partition 0 at 32 instead of 1 
> (as when the camera formatted itself - fdisk didn't like this "Track zero is 
> reserved for the BIOS"):
>
> Disk: /dev/rsd0d
> NetBSD disklabel disk geometry:
> cylinders: 6, heads: 64, sectors/track: 32 (2048 sectors/cylinder)
> total sectors: 14143
>
> BIOS disk geometry:
> cylinders: 7, heads: 64, sectors/track: 32 (2048 sectors/cylinder)
> total sectors: 14143
>
> Partition table:
> 0: Primary DOS with 12 bit FAT (sysid 1)
>    start 32, size 14143 (7 MB, Cyls 0-6/58/31)
> 1: <UNUSED>
> 2: <UNUSED>
> 3: <UNUSED>
> Bootselector disabled.
>
> After newfs_msdos (or format from Windows), disklabel was still unhappy with 
> the offsets set to 0 instead of 32
[snip]

Further info. I tried to recreate this using vnd (dd if=/dev/zero 
of=test.fs count=14175). I was able to successfully fdisk, format and use 
disklabel on it.

Just for kicks, I copied the filesystem image created with vnd back to the 
camera and disklabel didn't work on there.

--
Stephen