On Fri, Aug 26, 2005 at 08:59:11PM +0100, David Laight wrote:
> 
> Only the four main mbr partitions (and excluding the base of the extended
> partition chain) can be set 'active'.  This is enforced because some system
> BIOS won't boot the system at all unless one of the main partitions is
> marked 'active' (and even looks as though it has code in it).

Ah, so this is kind of a forced safety net? To prevent the system from
becoming unbootable?

> 
> Remember that the 'active' flag is used by simple mbr code to determine
> which pbr to load (the pbr code has to re-read the mbr table but will
> normally look at the partition type - which is one thing that makes it
> difficult to have 2 copies of a similar OS on the same disk).
> 
> If you need to boot from one of the extended partitions you typically
> have multiple OS installed, and some form of boot-select program in use.
> Whether a partition is marked 'active' is then usually totally irrelevant.

So, can I go ahead and install netbsd into the extended partition hdb11 and
use LILO to boot it? I currently have Win2K and Slackware installed on
hda, and LILO installed in the mbr of hda.

> It is also worth remembering that both lilo and grub are much, much larger
> than space actually reserver for the first level disk bootstrap, and
> probably also much larger that allocated by most filesystems for boot code.
> (The mbr code has less that 446 bytes, UFS typically allows 8k.)
> This means that they will be reading code out of (one of) the filesystems,
> if you re-einstall (or newfs) that filesystem you disk is immediately
> unbootable.

Understood.

> 
> The netbsd bootselect code resides entirely within its allocated space,
> so can (ok should) be able to boot (the pbr code) for any partition
> on the disk, even if you reinstall the netbsd area.
> (this is the main reason its UI is so terse)

So, should I install the netbsd boot code into the mbr of hdb?