Re: Bug#24168: Hamm: lilo installed to partition, not MBR
Hi,
On 21 May 1999, Adam Di Carlo wrote:
> Joey, AFAIK, the stock lilo.conf puts in place the MBR than puts the
> OS loader (2nd stage boot-loader, I think is the right term) on the
> partition.
AFAIK it's /usr/sbin/liloconfig which installs /boot/mbr.b by default.
-----
grep mbr.b $(
for a in /{,usr/}{,s}bin/*; do
file $a | sed -e '/script/s/\([^:]*\):.*/\1/p' -e'd' ;
done )
/usr/sbin/liloconfig: system("dd if=/boot/mbr.b of=$disk bs=444 count=1");
-----
The file mbr.b is BTW in its own package, "mbr", but is only(*) installed
by a script in the "lilo" package.
(*) All of my homebrew kernel-image.debs have a postinst script which does
the same as liloconfig. I've noticed that the /vmlinuz shipped with the
base system is not "owned" by a package. It would be nice for several
reasons if this were the case, one being that it would be possible to
uninstall the generic kernel image and modules with dpkg instead of rm,
another one that the included liloconf could be used by dinstall, IIRC
kernel-package already supports many types of bootloader. OTOH I do
realise that this would make the bootfloppies slightly bigger :-P
As I understand it, if you have a mbr.b in the bootsector of the harddisk,
that file can chain the bootprocess on to the lilo bootblock in any
partition bootsector. The alternative to mbr.b in the hardisk's
bootsector is writing lilo's own bootblock there.
/sbin/lilo only writes mbr.b to the bootsector if you have both of
these lines in /etc/lilo.conf:
boot = /dev/hda
install = /boot/mbr.b
Unless specified differently, lilo.conf defaults to a "boot" value equal
to the current root device and an "install" value of /boot/boot.b.
> The real bug is that lilo sucks, and silo sucks, and god knows what
> else, the the free software world needs a better boot-loader.
Grub? Or does that work on x86 only?
Cheers,
Joost
Reply to: