On Sun, Oct 14, 2018 at 06:38:53PM +0200, Mark Kettenis wrote:
> > Date: Sun, 14 Oct 2018 17:43:47 +0200
> > From: "Peter J. Philipp" <[email protected]>
> >
> > On Tue, Sep 18, 2018 at 10:04:01AM +0200, Peter J. Philipp wrote:
> > > Hi OpenBSD PPC Community,
> >
> > Hi again! I've managed to get a bit further. I'm a step away from linking
> > the kernel for the first time, but I have error messages and it seems to
> > come
> > from the linker. I'm a bit dumb about this right now so I thought I'd ask
> > if
> > you guys can make out what the error wants me to do.
> >
> > http://enhancedrisc.de/#20181014
>
> The instructions you use in locore0.S to refer to fwargsave, esym and
> proc0paddr only have limited range and when the final link is done
> these symbols are put at an offset that is to large. You probably
> have to rewrite the bits of code in locore0.S that deal with the
> global variables to use instruction sequences that can reach larger
> offsets.
Awesome explanation thanks! And I noticed I strayed too far from the
example given to me by the 32 bit macppc code. I corrected that and I got
this:
...
ld -N -Ttext 100114 -e start --warn-common -nopie -o bsd ${SYSTEM_HEAD} vers.o
${OBJS}
text data bss dec hex
6367624 2074912 671936 9114472 8b1368
mv bsd bsd.gdb
ctfstrip -S -o bsd bsd.gdb
ct4#
So the very first cross-compiled kernel. I doubt it works but its so tempting
to put it on. I need a 64 bit loader anyhow so I'll work on that for a bit,
I hope I can reuse the macppc loader. Finally I'm able to test code with the
compiler! This is often a great help to me.
I've committed everything that wasn't committed in my aim64 and powerpc64 tree
and with this change which I'll commit in a sec there should be at least the
compilable sources (buggy or not).
Best Regards,
-peter
