[PATCH] Tight check of pfn_valid on sparsemem - v3

[Minchan Kim]

On Mon, Jul 26, 2010 at 2:03 AM, Balbir Singh <balbir at linux.vnet.ibm.com> wrote:
> On Sun, Jul 25, 2010 at 7:32 PM, Minchan Kim <minchan.kim at gmail.com> wrote:
>> Changelog since v2
>>  o Change some function names
>>  o Remove mark_memmap_hole in memmap bring up
>>  o Change CONFIG_SPARSEMEM with CONFIG_ARCH_HAS_HOLES_MEMORYMODEL
>>
>> I have a plan following as after this patch is acked.
>>
>> TODO:
>> 1) expand pfn_valid to FALTMEM in ARM
>> I think we can enhance pfn_valid of FLATMEM in ARM.
>> Now it is doing binary search and it's expesive.
>> First of all, After we merge this patch, I expand it to FALTMEM of ARM.
>>
>> 2) remove memmap_valid_within
>> We can remove memmap_valid_within by strict pfn_valid's tight check.
>>
>> 3) Optimize hole check in sparsemem
>> In case of spasemem, we can optimize pfn_valid through defining new flag
>> like SECTION_HAS_HOLE of hole mem_section.
>>
>
> Is there an assumption somewhere that assumes that page->private will
> always have MEMMAP_HOLE set when the pfn is invalid, independent of
> the context in which it is invoked? BTW, I'd also recommend moving

zzzzz.... I needed sleep.
Will fix and resend.

> over to using set_page_private() and page_private() wrappers (makes
> the code easier to search)

Okay.
Thanks for pointing me out, Balbir.

>
> Balbir
>



-- 
Kind regards,
Minchan Kim