Re: [PATCH V2] mm/thp: Allocate transparent hugepages on local node

["Aneesh Kumar K.V" <aneesh.kumar@linux.vnet.ibm.com>]

Vlastimil Babka <vbabka@suse.cz> writes:

> On 12/01/2014 03:06 PM, Aneesh Kumar K.V wrote:
>> "Kirill A. Shutemov" <kirill@shutemov.name> writes:
>>
>>> On Mon, Dec 01, 2014 at 11:16:43AM +0530, Aneesh Kumar K.V wrote:
>>>> This make sure that we try to allocate hugepages from local node if
>>>> allowed by mempolicy. If we can't, we fallback to small page allocation
>>>> based on mempolicy. This is based on the observation that allocating pages
>>>> on local node is more beneficial that allocating hugepages on remote node.
>>>>
........
......

>>>> index e58725aff7e9..fa96af5b31f7 100644
>>>> --- a/mm/mempolicy.c
>>>> +++ b/mm/mempolicy.c
>>>> @@ -2041,6 +2041,46 @@ retry_cpuset:
>>>>   	return page;
>>>>   }
>>>>
>>>> +struct page *alloc_hugepage_vma(gfp_t gfp, struct vm_area_struct *vma,
>>>> +				unsigned long addr, int order)
>
> It's somewhat confusing that the name talks about hugepages, yet you 
> have to supply the order and gfp. Only the policy handling is tailored 
> for hugepages. But maybe it's better than calling the function 
> "alloc_pages_vma_local_only_unless_interpolate" :/
>

I did try to do an API that does

struct page *alloc_hugepage_vma(struct vm_area_struct *vma, unsigned long addr)

But that will result in further #ifdef in mm/mempolicy, because we will
then introduce transparent_hugepage_defrag(vma) and HPAGE_PMD_ORDER
there. I was not sure whether we really wanted that.

-aneesh

--
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>