> Wahoo!!! We are finally starting in 2.1.x!!! I can tell you about my
> experiences so far very simply... It didn't work at all. :-( I use
> loadlin to boot from an M$-DOS 6.22 boot menu (this makes my wife
> happy). I just glanced through the source for loadlin, and it _DOES_
> use hardware addresses to address kernel addresses, which means it is
> horribly broken using the new kernel.
What 'hardware addresses' do you mean ? 0x90000 ?
Well this is the common bootloader segment for both LILO, Loadlin and others.
The rest (0x1000, 0x100000, also used by LILO) are hardcoded entry points to
the kernel _after_ the loader has done its job.
Loadlin does _not_ fiddle or trick anything in the kernel code, what it
does it does _before_ jumping to the kernel.
Well, I tried pre-2.1.0 and because I just wanted to test the bootstuff
I built a small kernel, only with IDE driver configured in.
It _did_ bomb with Laodlin, however, it also bombs using a DD'ed floppy.
^^^^^^^^^^ ^^^^^^^^^^^^
Now have a look at this strangeness:
el16:/usr/src/linux-pre-2.1.0# ls arch/i386/boot/*zI* -l
-rw-r--r-- 1 root root 5121 Sep 24 11:50 arch/i386/boot/bzImage
-rw-r--r-- 1 root root 18325 Sep 24 12:30 arch/i386/boot/zImage
el16:/usr/src/linux-pre-2.1.0# ls vmlinux -l
-rwxr-xr-x 1 root root 678823 Sep 24 12:30 vmlinux*
Dammned good compression ratio, too good if you ask me.
There is something broken with the boot/compress stuff I think.
The vmlinux looks good, the System.map looks good, but the (b)zImages
are broken.
So long,
Hans
<lermen@fgan.de>