10.2.4 update hangs MOL
Samuel Rydh
mol-devel@lists.maconlinux.org
Mon, 17 Feb 2003 18:27:25 +0100
On Tue, Feb 18, 2003 at 12:10:13AM +0800, Jamie Clark wrote:
> Samuel Rydh wrote:
>
> >Hmm. Can you run md5sum on /mac_kernel? Just to make
> >sure you have the same 10.2.4 kernel as I have here...
> >
> >
> OK, copied from mac partition back to linux:
>
> root@adele:~/# md5sum mach*
> bcc6fbf5707d1f0be820b5d5659d3497 mach.sym
> 6e2ac40db80ebd2aeebbba97df5a9d1c mach_kernel
Good... it is exactly the same kernel as I have here.
> >If you start osx in verbose mode (startmol --config=v --osx),
> >what is printed out?
> >
> I forgot to mention that the first time. It show nothing, just blanks
> the logo and displays a steady cursor in the top left corner. It does
> _not_ get as far as showing the "standard timeslicing quantum" that you
> would expect.
Great. That means the IOKit drivers have not started any matching
so it is only the kernel that matters.
There are a couple of interesting test you could try:
1) Adding 'kernel: /linux/path/to/mach_kernel' (with a 10.2.4 disk)
2) Ditto, but with the 10.2.3 disk. (Might affect things, if the
the osx boot-loader contains some bug.)
3) Ditto, but with the 10.2.3 kernel and a 10.2.4 disk.
4) Disable the instruction acceleration stuff. Yo can do this by
checking out bk://mol.bkbits.net/bootx-mol and then commenting out
the inst_replacements() call at the end of bootx-mol/main/mol.c.
Build bootx and add 'bootx_image: /path/to/bootx' in the config
file.
5) Turn off AltiVec ('processor: G3')
> I have not yet checked if Apple has released the 6.4
> darwin kernel to anon cvs yet.
Doesn't really matter; the symbol table is available
(nm mach.sym) and can be used with the mol debugger.
The molrc syntax is 'symfile: /file/with/the/output/from/nm'.
> >Another thing you could try is 'startmol -d --osx' which starts
> >the debugger and then 'Meta-G' to start mol. 'Meta-S' will stop
> >mol and show you what OSX is doing (if you know some assembly).
> >
> OK. Meta-G set it running, but Meta-S had no effect (comparing with what
> I see with 10.2.3) the cpu continues to spin. I confess I'm new to MOL
> so forgive me if I have missed something obvious.
Well, Meta-S single-steps the assembly code. If you load the symbols,
then it might be possible to tell what MacOS X thinks it is doing
(e.g. spinning in the panic function, in the DEC calibration loop
or somewhere else).
Cheers,
/Samuel