3. Preparing Linux

3.1 Building a new kernel

Some people don´t like building a custom kernel. For those of you: you must build your own kernel if you want to share data between Solaris and Linux on one hard disk. If, however, you do not want to share data, you are safe with your old kernel.

Here are your kernel options for ``Solaris compatibility'':

• Under code maturity:

  Prompt for development drivers

  Unfortunately UFS writing is still experimental.

• In file systems:

  UFS file system support

  UFS is the Solaris file system

  UFS file system write support

  The only way to pass data to Solaris

  partition types / Solaris (x86) partition table support

  Is needed for the slices within your Solaris partition to be found.

3.2 Preparing your boot loader

Before messing with your boot loader, please make sure that you have an emergency disk ready which enables you to boot your old system.

Unfortunately the Solaris install overwrites the master boot record (MBR) and overwrites your old boot manager.

Fortunately, the Solaris boot manager is able to chain-boot. For this to work you have to put your old boot loader (e.g. lilo) into the boot sector of one primary Linux partition.

For lilo this means: look at your /etc/lilo.conf. Look for the boot= line. If it is something like boot=/dev/hda1 everything is fine. But if it´s something like boot=/dev/hda it points to the MBR. Please change it to point to a primary Linux partition (e.g. boot=/dev/hda1)

A way to check if your boot-loader is chain-loadable is installing lilo in to the MBR and trying to chain load your other boot-manager. I used the following file called lilo.conf.mbr for this:

#lilo.conf.mbr 
#Lilo in Master Boot Record doing nothing but chain-loading another lilo 
boot=/dev/hda
root=/dev/hda5 
install=/boot/boot.b 
map=/boot/map 
vga=ask
delay=50 
other=/dev/hda1 
label=lilochain

If you reboot now, you should have a lilo-prompt. When you select other this chain loads the old lilo, which in turn loads Linux.