Re: [PATCH 1/2] Revert "hugetlb: avoid taking i_mmap_mutex in unmap_single_vma() for hugetlb"

[Aneesh Kumar K.V]

Mel Gorman <mgorman@xxxxxxx> writes:

> This reverts the patch "hugetlb: avoid taking i_mmap_mutex in
> unmap_single_vma() for hugetlb" from mmotm.
>
> This patch is possibly a mistake and blocks the merging of a hugetlb fix
> where page tables can get corrupted (https://lkml.org/lkml/2012/7/24/93).
> The motivation of the patch appears to be two-fold.
>
> First, it believes that the i_mmap_mutex is to protect against list
> corruption of the page->lru lock but that is not quite accurate. The
> i_mmap_mutex for shared page tables is meant to protect against races
> when sharing and unsharing the page tables. For example, an important
> use of i_mmap_mutex is to stabilise the page_count of the PMD page
> during huge_pmd_unshare.

I missed that. 

>
> Second, it is protecting against a potential deadlock when
> unmap_unsingle_page is called from unmap_mapping_range(). However, hugetlbfs
> should never be in this path. It has its own setattr and truncate handlers
> where are the paths that use unmap_mapping_range().

I noted this in 

http://article.gmane.org/gmane.linux.kernel.mm/80065


>
> Unless Aneesh has another reason for the patch, it should be reverted
> to preserve hugetlb page sharing locking.
>

I guess we want to take this patch as a revert patch rather than
dropping the one in -mm. That would help in documenting the i_mmap_mutex
locking details in commit message. Or may be we should add necessary
comments around the locking ?

Acked-by: Aneesh Kumar K.V <aneesh.kumar@xxxxxxxxxxxxxxxxxx>

> Signed-off-by: Mel Gorman <mgorman@xxxxxxx>
> ---
>  mm/memory.c |    5 ++++-
>  1 file changed, 4 insertions(+), 1 deletion(-)
>
> diff --git a/mm/memory.c b/mm/memory.c
> index 8a989f1..22bc695 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -1344,8 +1344,11 @@ static void unmap_single_vma(struct mmu_gather *tlb,
>  			 * Since no pte has actually been setup, it is
>  			 * safe to do nothing in this case.
>  			 */
> -			if (vma->vm_file)
> +			if (vma->vm_file) {
> +				mutex_lock(&vma->vm_file->f_mapping->i_mmap_mutex);
>  				__unmap_hugepage_range(tlb, vma, start, end, NULL);
> +				mutex_unlock(&vma->vm_file->f_mapping->i_mmap_mutex);
> +			}
>  		} else
>  			unmap_page_range(tlb, vma, start, end, details);
>  	}
> -- 
> 1.7.9.2

--
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/