LILO dual booting with SATA and PATA drives

raypen · 11-23-2011, 08:55 AM

Recently installed Slack 13.37 on an older 40GB IDE
drive in a system with both SATA and IDE channels.
Win XP is installed on the SATA drive (which is shown
as Disk 1 in Windows Disk management) and Linux is
installed on the IDE drive (Disk 0).

I opted to have LILO installed on the MBR which I
thought would be the MBR of the IDE drive, but it
instead overwrote the MBR of the SATA drive removing
the WinXP boot info. I wanted to keep the boot
sectors separate and only boot one OS at a time by
altering the boot drive sequence in the BIOS.

Dual booting works fine, I just have a LILO screen
appearing before WinXP loads. All of this however,
gives me pause as I now realize that I don't fully
understand how to distinguish boot sectors when
both types of disks are involved.

I used to be able to load LILO on an MBR of a
secondary HD and have a Windows system on the
primary drive boot simply by specifying:

Code:

other = /dev/hda1
  Label = Windows
  table = /dev/hda
  boot-as=0x80

Can anyone provide a detailed explanation on the
various methods of dual booting when SATA and IDE
drives are involved or when multiple SATA drives
are connected?

Martinus2u · 11-23-2011, 01:49 PM

this topic is harder than it appears. There are at least four different numbering schemes at work here, and they can all be different.

(1) the BIOS numbering scheme, and the order in which the BIOS probes devices for executable boot loaders. This depends entitrely on the BIOS itself and can be influenced by your BIOS settings and the BIOS boot menu. In simple cases the BIOS boots the MBR on device 0x80 (the first number). Aim for that if you can.

(2) when executing lilo all your device names are according to the running kernel, except the one after root=. That one is stored with the kernel to be loaded and executed at boot time. There are lilo commands for mapping the BIOS number to the linux device name of the running kernel.

The is one big problem because lilo tries to verify the existence of the root= parameter in terms of the running kernel. If the naming scheme is totally different between kernels (like when moving from /dev/hda to /dev/sda) you need to be creative.

(3) when loading and executing a kernel, the root= parameter is evaluated according to the kernel being booted.

(4) the root file system itself contains references to devices (like in /etc/fstab) which are evaluated according to the kernel just booted.

In your scenario there is no harm in overwriting the MBR of the hdd containing Windows since the actual Windows loader resides in the boot sector of the Windows system partition. If you keep the Windows MBR, you need to change the boot order in the BIOS (if your BIOS allows such a thing) in order to boot from another disk.