Re: [PATCH] mm: Check if PTE is already allocated during page fault

[Minchan Kim]

Hi Mel,

On Fri, Apr 15, 2011 at 7:12 PM, Mel Gorman <mgorman@xxxxxxx> wrote:
> With transparent hugepage support, handle_mm_fault() has to be careful
> that a normal PMD has been established before handling a PTE fault. To
> achieve this, it used __pte_alloc() directly instead of pte_alloc_map
> as pte_alloc_map is unsafe to run against a huge PMD. pte_offset_map()
> is called once it is known the PMD is safe.
>
> pte_alloc_map() is smart enough to check if a PTE is already present
> before calling __pte_alloc but this check was lost. As a consequence,
> PTEs may be allocated unnecessarily and the page table lock taken.
> Thi useless PTE does get cleaned up but it's a performance hit which
> is visible in page_test from aim9.
>
> This patch simply re-adds the check normally done by pte_alloc_map to
> check if the PTE needs to be allocated before taking the page table
> lock. The effect is noticable in page_test from aim9.
>
> AIM9
> Â Â Â Â Â Â Â Â2.6.38-vanilla 2.6.38-checkptenone
> creat-clo   Â446.10 ( 0.00%)  424.47 (-5.10%)
> page_test    38.10 ( 0.00%)  Â42.04 ( 9.37%)
> brk_test    Â52.45 ( 0.00%)  Â51.57 (-1.71%)
> exec_test   Â382.00 ( 0.00%)  456.90 (16.39%)
> fork_test    60.11 ( 0.00%)  Â67.79 (11.34%)
> MMTests Statistics: duration
> Total Elapsed Time (seconds) Â Â Â Â Â Â Â Â611.90 Â Â612.22
>
> (While this affects 2.6.38, it is a performance rather than a
> functional bug and normally outside the rules -stable. While the big
> performance differences are to a microbench, the difference in fork
> and exec performance may be significant enough that -stable wants to
> consider the patch)
>
> Reported-by: Raz Ben Yehuda <raziebe@xxxxxxxxx>
> Signed-off-by: Mel Gorman <mgorman@xxxxxxx>
> --
> Âmm/memory.c | Â Â2 +-
> Â1 file changed, 1 insertion(+), 1 deletion(-)
>
> diff --git a/mm/memory.c b/mm/memory.c
> index 5823698..1659574 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -3322,7 +3322,7 @@ int handle_mm_fault(struct mm_struct *mm, struct vm_area_struct *vma,
> Â Â Â Â * run pte_offset_map on the pmd, if an huge pmd could
> Â Â Â Â * materialize from under us from a different thread.
> Â Â Â Â */
> - Â Â Â if (unlikely(__pte_alloc(mm, vma, pmd, address)))
> + Â Â Â if (unlikely(pmd_none(*pmd)) && __pte_alloc(mm, vma, pmd, address))
> Â Â Â Â Â Â Â Âreturn VM_FAULT_OOM;
> Â Â Â Â/* if an huge pmd materialized from under us just retry later */
> Â Â Â Âif (unlikely(pmd_trans_huge(*pmd)))
>

Reviewed-by: Minchan Kim <minchan.kim@xxxxxxxxx>

Sorry for jumping in too late. I have a just nitpick.

We have another place, do_huge_pmd_anonymous_page.
Although it isn't workload of page_test, is it valuable to expand your
patch to cover it?
If there is workload there are many thread and share one shared anon
vma in ALWAYS THP mode, same problem would happen.


-- 
Kind regards,
Minchan Kim
--
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/