On Wed, Feb 21, 2018 at 04:02:25PM +0100, Christophe LEROY wrote:
[...]

quoted

quoted

My question might me stupid, as I don't know PCC64 in deep, but when looking
at page_is_ram() in arch/powerpc/mm/mem.c, I have the feeling the PPC64
implements ram by blocks. Isn't it what you are trying to achieve ? Wouldn't
it be feasible to map to what's done in PPC64 for PPC32 ?

Using page_is_ram in __ioremap_caller and the same memblock-based
approach that's used on PPC64 on PPC32 *should* work, but I think due to
the following line in initmem_init, it won't:

	memblock_set_node(0, (phys_addr_t)ULLONG_MAX, &memblock.memory, 0);

Can't we just fix that ?

Turns out I was completely wrong about this. memblock_set_node as called
above only assigns all memory to node 0 and merges *adjacent* memblocks.
It doesn't merge the memblocks on the Wii, which are far apart.

So now I actually have a working patchset (coming soon), that's a good
deal shorter than this patchset, and hopefully won't break
CONFIG_HIGHMEM in the same way.

Thanks for your input! :)


Jonathan Neuschäfer