Re: Booter for 9.3 kernel not working

On Wed, Jan 24, 2024 at 7:54 AM Martin Husemann <[email protected]> wrote:
>
> On Wed, Jan 24, 2024 at 06:45:21AM -0500, Matthew Stock wrote:
> > This doesn't appear to be a filesystem issue, as I can get prebuilt
> > GENERICSBC to load from the root filesystem, and my GENERICSBC will
> > not load from the root filesystem nor the native partition.
>
> Can you compare the output of
>         file netbsd

Identical for all kernels:

lciii$ file /netbsd /netbsd.crosssbc
/netbsd:          ELF 32-bit MSB executable, Motorola m68k, 68020,
version 1 (SYSV), statically linked, for NetBSD 9.3, not stripped
/netbsd.crosssbc: ELF 32-bit MSB executable, Motorola m68k, 68020,
version 1 (SYSV), statically linked, for NetBSD 9.3, not stripped


>         readelf -hlS netbsd
> for the working and the non working kernels?

Comparing the prebuild GENERICSBC and my cross compiled version, the
only difference here is that the .rodata segment is 62 bytes longer.
The other segments are the same size, but have a different offset
obviously.  Some of the changes are likely the build string changing,
but it doesn't explain all of it.  I'll write a script later tonight
to compare the sizes of the individual symbols in rodata and try to
narrow it down.  For sure "config" is different, but a simple diff
won't show the answer since beyond that point the offsets make the
diff useless:

lciii$ grep config rodata.*sym.txt
rodata.crosssbc.sym.txt:002bfe34 l     O .rodata        000058d4 config
rodata.crosssbc.sym.txt:002d59ec l     O .rodata        00000078
configuration_table
rodata.sym.txt:002bfe34 l     O .rodata 00005898 config
rodata.sym.txt:002d59b4 l     O .rodata 00000078 configuration_table

I'll report back when I have more info, but without understanding what
Booter is actually validating it's hard to know where the problem
might be.  Does anyone have the Booter 2.0.0 source?  I am happy to
put some legwork into trying to resolve this.  I'm honestly suspicious
that it's user error since I'd expect more bug reports if it wasn't.

Thanks,

Matt