Re: [PATCH 1/11] Add __GFP_MOVABLE flag and update callers

[Mel Gorman <mel@csn.ul.ie>]

On Tue, 21 Nov 2006, Christoph Lameter wrote:

> On Tue, 21 Nov 2006, Mel Gorman wrote:
>
> > On Tue, 21 Nov 2006, Christoph Lameter wrote:
> >
> > > Are GFP_HIGHUSER allocations always movable? It would reduce the size of
> > > the patch if this would be added to GFP_HIGHUSER.
> > No, they aren't. Page tables allocated with HIGHPTE are currently not movable
> > for example. A number of drivers (infiniband for example) also use
> > __GFP_HIGHMEM that are not movable.
>
> HIGHPTE with __GFP_USER set? This is a page table page right?
> pte_alloc_one does currently not set GFP_USER:

What is __GFP_USER? The difference between GFP_USER and GFP_KERNEL is only 
in the use of __GFP_HARDWALL. But HARDWALL on it's own is not enough to 
distinguish movable and non-movable.

> struct page *pte_alloc_one(struct mm_struct *mm, unsigned long address)
> {
>        struct page *pte;
>
> #ifdef CONFIG_HIGHPTE
>        pte =
> alloc_pages(GFP_KERNEL|__GFP_HIGHMEM|__GFP_REPEAT|__GFP_ZERO, 0);
> #else
>        pte = alloc_pages(GFP_KERNEL|__GFP_REPEAT|__GFP_ZERO, 0);
> #endif
>        return pte;
> }
>
> How does infiniband insure that page migration does not move those pages?

I have not looked closely at infiniband and how it uses it's pages.

--
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/