Re: fdisk, lilo problems
On Wed, 10 Jan 1996, Bruce Perens wrote:
> You might check what fdisk version 3 says about it. It's on Sunsite. I'm
> going to package it with my new front-end, but said front-end is not done
I just tried that, and came across what looks like a source
package problem. fdisk-3.03 won't build with the source-1.2.13-5
package installed. It fails because linux/autoconf.h, included
from /usr/linclude/linux/config.h doesn't exist.
Also, since lilo has problems with this partition table too, the
problem seems to go beyond fdisk. I note that the bios autoconf
offered me configurations for the disk with both 16 and 32 heads.
I tried the 16 head configuration, and it didn't work, so I went
on to a 32 head configuration and that worked until lilo and
fdisk had problems under linux.
I note from the fdisk messages that it appears to be using 16 heads
to compute the physical mappings and 32 heads to compute the 32
head mappings. I'm not sure what the lilo messages are trying
to tell me, but I see that some of the figures it gives for
3D address are half of those it gives for the Linear address.
My guess is that the DOS partition size I chose (using the 32
head geometry I chose when defining the disk to the BIOS) ended
up with a partition halfway through what would have been a cylinder
if I had chosen the 16 head geometry. It's a mystery to me why fdisk
and lilo would be trying to use the 16 head geometry, but I haven't
looked at the sources for either of those programs to try to figure
out what they think they're doing.
Oh well, I know you're busy on other things, as am I. I haven't
decided yet whether to dig into this or to just start over with
a DOS partition size one 32-head-geometry cylinder larger in
hopes that the linux fdisk and lilo programs won't barf on that.