In message <877klq7lfo.fsf@snark.piermont.com>,
"Perry E. Metzger" writes>:

>As the author of part of our current bootloader, let me note that I
>believe we probably do have enough space for this, but that doing the
>check reliably might be a bit tough.

It'd be great if the bootblocks had enough of a test so that one could
specify an LBA size in the MBR, and leave the c/h/s values unspecified
(or junk).  I recently tried using using non-NetBSD tools to carve out
an MBR slice for NetBSD; the NetBSD bootloader didn't work until I
went back and used NetBSD tools to write C/H/S values for
(if memory serves) the start of the NetBSD MBR partition.

I would file a PR, if only I could reproduce it without toasting
the now-working setup :-)



>> It would actually be better for the kernel to determine what
>> the BIOS can do, and tell the bootloader.
>
>Er, the kernel isn't running when the bootloader runs. Also, the
>kernel isn't capable of making 16 bit BIOS calls (at least not at all
>easily).


I think David meant: at install time, have the install kernel detect
whether the BIOS supports int13 extensions. If it does, then install
the LBA bootblocks; otherwise install the non-LBA bootblocks.