[okl4-developer] elf_entry (was: More NSLU2 debugging)

[Geoffrey Lee]

On Wed, Jun 25, 2008 at 09:39:07PM +1000, Joshua Root wrote:
> Geoffrey Lee wrote:
>> On Wed, Jun 25, 2008 at 05:56:21PM +1000, Joshua Root wrote:
>>> I didn't think to check this because I specified "elf_entry =   
>>> 0x01d00000L" in the nslu2 class in machines.py. Does that not work?   
>>> Changing the bootloader default is possible, but it's a bit of a 
>>> pain.
>>
>> I think we did away with that flag because most bootloaders would allow
>> you to hand control over to an arbitary location.
>
> So would it be hard to add it back in? (Read: harder than rebuilding  
> RedBoot and flashing a whole bunch of slugs?)

Josh - 


The physical memory that backs the image by default is allocated
out of the physical pool as defined in the machines description which
leads to the image being backed at the start of the address at the
physical pool.

There are several ways that you can solve this.  One is to define two 
different physical pools, and modify the build system such that the 
image is always backed from the pool starting at the elf entry.

Another method is to have some stub code at bootstrap that copies
the code to the correct location, then jump to it.  You can do this 
because image load location and boot location are well-defined.

I would have thought that the sequence of commands to boot an
image would be stored in non-volatile memory somewhere, but I
am not too familiar with the slug or RedBoot.


>
> - Josh

	-gl
>