On Thu, 30 Sep 1999, Harald Koerfgen wrote:
> On 30-Sep-99 Sergey N. Lapin wrote:
> > Thanks, but then I have a question: how a bootloading on DECstations
> > works? so, where do prom finds a bootblock? does it have a direct
> > knowledge of partitions?
> No, it hasn't. The PROM loads sector 0 from the boot device (a 512 byte
> that is) and is able to load a sequence of sectors depending on information in
> the boot sector. Have a look at
> ftp://ftp.linux.sgi.com/pub/linux/mips/test/bootprep.c for the details.
> BTW, I have a newer version of bootprep here which is compatible with MSDOS
> style partitioning (yuck!) and SUN partition tables. I will upload this
> immediately :-).
Will try it...
> 1) Ultrix partition tables, Ultrix bootsector and Ultrix bootloader.
Ultrix disk dead on my machine :(
> 2) NetBSD partition tables, NetBSD/pmax bootsector and second stage
Too painful, IMHO... :)
> 3) MSDOS or SUN partition tables and bootprep. bootprep simply determines the
> sectors of an executable image (objdump -b binary) and puts this in sector 0
> it fits. With "bootprep image /dev/sda" "image" will directly be loaded by the
> PROM. Actually "image" can be a complete kernel but in the long term it would
> be prefereable to have a second stage bootloader like SILO to handle different
> kernels. Any volunteers?
I could, if anyone could explain me, why binutils from decroot-current
(2.8.1-2D) and egcs from ftp.linux.sgi.com are unable to produce working
kernel, while produce working programs...
> Method 3 is obviously somewhat unflexible without a second stage bootloader
> it doesn't make use of information stored outside the bootsector, like methods
> 1 and 2 do. As a consequence you can have any filesystem you like on your
> partition :-)
Simple bootloader could be done in a few hours, just reuse some code
from kernel, libbfd, libe2fs, lilo...