Re: porting ARM: remap and MMU questions

Hello, Rainer
You wrote on Thu, 15 Jan 2009 10:05:07 +0100:

Thanks, low-level debugging facility helped me to verify the code flow.

As I mentioned in my original post, kernel dies after printing
"Uncompressing Lin...................... done, booting t" -- yes, exactly
like this, with "ux" and "he kernel" stripped off. Seems a buffer limit to
me, but why on an early stage and how -- do you have any clues?

RW> Getting across the MMU init itself is more difficult: You
RW> must ascertain that the serial port registers are actually mapped
RW> somewhere and that the output code can dynamically use either the
RW> virtual or physical address.

I suspect this might be my case. With the early stage debug macros I found
that problem resides in $(linux)/arch/arm/kernel/head.S, particularly
arouund the code preparing and enabling MMU.

But I _guess_ this is only consequence, and the real problem happens earlier
at page tables setup, because that is the only place the code uses platform
specific settings -- PHYS_OFFSET, which is physical RAM offset. Hence, this
value may depend on RAM being remapped or not.

-- Mark


.