Re: [PATCH] mm: cma: fix accounting of CMA pages placed in high memory

[Marek Szyprowski <m.szyprowski@samsung.com>]

Hello,

On 2/5/2013 1:40 AM, Minchan Kim wrote:

...

> > Previous time, it's not fully tested and now we checked it with
> > highmem support patches.
>
> I get it. Sigh. then [1] inline attached below wan't good.
> We have to code like this?
>
> [1] 6a6dccba, mm: cma: don't replace lowmem pages with highmem
>
> diff --git a/mm/page_alloc.c b/mm/page_alloc.c
> index b97cf12..0707e0a 100644
> --- a/mm/page_alloc.c
> +++ b/mm/page_alloc.c
> @@ -5671,11 +5671,10 @@ static struct page *
>   __alloc_contig_migrate_alloc(struct page *page, unsigned long private,
>                               int **resultp)
>   {
> -       gfp_t gfp_mask = GFP_USER | __GFP_MOVABLE;
> -
> -       if (PageHighMem(page))
> -               gfp_mask |= __GFP_HIGHMEM;
> -
> +       gfp_t gfp_mask = GFP_HIGHUSER_MOVABLE;
> +       struct address_space *mapping = page_mapping(page);
> +       if (mapping)
> +               gfp_mask = mapping_gfp_mask(mapping);
>          return alloc_page(gfp_mask);
>   }

Am I right that this code will allocate more pages from himem? Old approach
never migrate lowmem page to himem, what is now possible as gfp mask is 
always
taken from mapping_gfp flags. I only wonder if forcing GFP_HIGHUSER_MOVABLE
for pages without the mapping is a correct. Shouldn't we use avoid himem in
such case?

Best regards
-- 
Marek Szyprowski
Samsung Poland R&D Center


--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>