[PATCH 1/1] rtas_flash_4gig

[Paul Mackerras]

Jake Moilanen writes:

> Correct me if I'm wrong, but the ibm,flash-block-version will always be
> 1 if we have ibm,update-flash-64-and-reboot (from 270 onwards), and if
> we don't have the ibm,update-flash-64-and-reboot, we will not have
> gotten to this point anyways.

The RPA doesn't say that discontiguous block list support is
required.  To be properly compliant, we should check for the
ibm,flash-block-version property, even if we "know" that every machine
will have it.

> The block list pages do not have the restriction of being under 4G. They
> just have to be page aligned and not cross a LMB.  Since we only
> allocate one page for every memory block using get_zeroed_page, that
> should be fine.  

The block_list has to be below 4GB.  The RPA doesn't actually say
whether the block_list extensions need to be below 4GB also, but I
don't see why there would be the requirement on the first block_list
but not the subsequent ones.  I have sent off a query to the powers
that be to ask.

> Why do you feel we need to check if the pages overlap OF's memory?  

RPA requirement R-7.3.7.2-4.

Regards,
Paul.