Still not booting

Still not booting

[Claude Robitaille]

Hi,

I am still not able to boot my 8xx board via NFS. Actually, I do not
think is it NFS related anymore. I put a lot of printk and was able
to narrow down to the clear_user function.

This is happening while trying to load /sbin/init. init/main.c calls
execve which eventually leads to load_elf_intepr to padzero to
clear_user and __clear_user. (There is other functions inbetween, I
just skipped them end the list here). __clear_user is an assembly
routine which clears a memory block. The arguments are the same as
clear_user, i.e. an address pointer and a length.

I do not get any more printk outputs after the processor enters
__clear_user. I modified clear_user to do the actual clearing of the
memory in C (easier to work with). Right now, I find that the first
access to the memory to clear it ends up in hanging the system. The
address is 3004ddcc which obviously seems to be a logical or virtual
address. If I am not wrong, /sbin/init is the first user land process,
so trying to start it is the first time the kernel is playing with
user land memory space. I do not think this is an hardware unmapped
physical location since I think the bus monitor would eventually kick
in and result into a machine check. Am I right?

Is there s way to easilly convert the virtual address to physical?

Any other idea on how to keep analysing the problem?

Thanks

Claude


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Still not booting

[Dan Malek]

Claude Robitaille wrote:


> ..... Right now, I find that the first
> access to the memory to clear it ends up in hanging the system. The
> address is 3004ddcc which obviously seems to be a logical or virtual
> address.


What are you using for the addresses of the IMMR, control registers
and such?  You can't use the default values that most debuggers
stuff into a FADS board, these have to all be moved into the
kernel virtual address space, which map 1:1 to their
corresponding physical address, and requires properly programming
the memory controller and changing the debugger scripts.



	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Still not booting

[Claude Robitaille]

I works now, thanks Dan

On Mon, 25 Oct 1999, Dan Malek wrote:

> 
> Claude Robitaille wrote:
> 
> 
> > ..... Right now, I find that the first
> > access to the memory to clear it ends up in hanging the system. The
> > address is 3004ddcc which obviously seems to be a logical or virtual
> > address.
> 
> 
> What are you using for the addresses of the IMMR, control registers
> and such?  You can't use the default values that most debuggers
> stuff into a FADS board, these have to all be moved into the
> kernel virtual address space, which map 1:1 to their
> corresponding physical address, and requires properly programming
> the memory controller and changing the debugger scripts.


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Still not booting

[Dan Malek]

Claude Robitaille wrote:
> 
> I works now, thanks Dan

So.....what did you do?


	-- Dan

** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/

Re: Still not booting

[Claude Robitaille]

As you recommended, I moved IMMAP to fa200000 (and the BCSR to fa400000).
I did that in head.S

On a different topic, I want ot compile in the MATH_EMULATION feature
but compiling fails with a error saying that some member does not exist
in a structure (I do not remember the details). I am still using an old
2.3.18 tree from the embedded directory. I am planning to download the
official CVS tree but I am wondering if the soft FP is really stable 
enough or if it is just still a wish. Anybody to comment?

Thanks

Claude

On Thu, 28 Oct 1999, Dan Malek wrote:

> Claude Robitaille wrote:
> > 
> > I works now, thanks Dan
> 
> So.....what did you do?
> 
> 
> 	-- Dan
> 
> 


** Sent via the linuxppc-embedded mail list. See http://lists.linuxppc.org/