>
>Brian Buhrow writes:
>> One of the things that I've wanted to do but have not yet done is to 
>> make pfdisk, the little program Gordon Ross wrote, work under
>> NetBSD.
>
>We already have an fdisk.
>
>That is not the point. You can't map BIOS partition IDs into NetBSD
>devices reliably, so you can't extract the BIOS geometry information
>and put it into an MBR.

If you asked the controller for the device to give you a reasonable geometry,
this would not be an issue.  In the CAM SCSI layer, this is the mechanism
used to create the in-core label exported by the "sd" device.  If the 
system doesn't find a valid MBR on the disk, it uses the original one
provided by the disk driver, otherwise it uses the one on the disk.  This
makes fdisk automagically default to a bootable geometry.  The bootable
geometry is also what is reported in the probe messages for the device.

Of course, on many architectures the geometry doesn't matter for booting,
but with more and more new machines (alpha, powerpc) picking up this
PC braindamage, it's no longer an x86 problem, so there should be an
MI interface for getting this information.

>Perry

--
Justin