boot2 broken ? (booting from pst fails)

[Toerless Eckert]

> As you pointed out, FreeBSD does try to avoid dirty hacks like this for the
> sake of a cleaner and more flexible design. 

But you can't have it without affecting compatiblity.

> Hardcoding the sectors would make
> things like booting kernels directly from boot2 as well as booting
> /boot/loader.old in the case that /boot/loader breaks impossible.

If booting loader.old is so important, why not boot2.old ? It's no
different. At some point there is something hardcoded. I don't see the
big issue here in hardcoding the loader sectors instead of effetively
the boot2 sectors. Heck, one could even hardcode sectors of both loader and
loader.old for the choice.

Sure, it would be nice to not depend on this. I for once think it's perfectly
valid to spare a full freebsd partition just for the linear encoding
of the bootstrap code, but that's just me.

> Also, apart
> from your hardware, FreeBSD is quite compatible with the large majority of
> PC hardware.  We boot directly off of 3ware ATA RAID controllers where I
> work, so I really think your claims are rather overstated.

Don't forget that neither 3ware nor promise list FreeBSD as a supported
operating system, both are equally listed in FreeBSDs hardware list,
there's no indication that booting for one of them doesn't work but
that it's supposed to work on the other. Neither does the documentation
of boot2 or loader explain the possible issue so as to buy possible buyers.

It would really be nice if this boot-compatiblity/incompatiblity information
would be documented somehwere. Like in pst(4) and boot(8).

Cheers
	Toerless