Re: [RFC] Reserve huge pages for reliable MAP_PRIVATE hugetlbfs mappings

[Andi Kleen]

Mel Gorman <mel@xxxxxxxxx> writes:

> MAP_SHARED mappings on hugetlbfs reserve huge pages at mmap() time. This is
> so that all future faults will be guaranteed to succeed. Applications are not
> expected to use mlock() as this can result in poor NUMA placement.
>
> MAP_PRIVATE mappings do not reserve pages. This can result in an application
> being SIGKILLed later if a large page is not available at fault time. This
> makes huge pages usage very ill-advised in some cases as the unexpected
> application failure is intolerable. Forcing potential poor placement with
> mlock() is not a great solution either.
>
> This patch reserves huge pages at mmap() time for MAP_PRIVATE mappings similar
> to what happens for MAP_SHARED mappings. 

This will break all applications that mmap more hugetlbpages than they
actually use. How do you know these don't exist?

> Opinions?

Seems like a risky interface change to me.

-Andi
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/