Re: [PATCH] mm: fix special swap entry handling on copy mm

[Naoya Horiguchi]

Hi Jerome,

On Mon, Aug 12, 2013 at 11:43:24AM -0400, j.glisse@xxxxxxxxx wrote:
> From: Jerome Glisse <jglisse@xxxxxxxxxx>
> 
> Prior to this copy_one_pte will never reach the special swap file
> handling code because swap_duplicate will return invalid value.
> 
> Note this is not fatal so nothing bad ever happen because of that.
> Reason is that copy_pte_range would break of its loop and call
> add_swap_count_continuation which would see its a special swap
> file and return 0 triggering copy_pte_range to try again. Because
> we try again there is a huge chance that the temporarily special
> migration pte is now again valid and pointing to a new valid page.
> 
> This patch just split handling of special swap entry from regular
> one inside copy_one_pte.
> 
> (Note i spotted that while reading code i haven't tested my theory.)
> 
> Signed-off-by: Jerome Glisse <jglisse@xxxxxxxxxx>

non_swap_entry() means not only migration entry, but also hwpoison entry,
so it seems to me that simply moving the swap_duplicate() into the
if(!non_swap_entry) block can change the behavior for hwpoison entry.
Would it be nice to add check for such a case?

Thanks,
Naoya Horiguchi

> ---
>  mm/memory.c | 26 +++++++++++++-------------
>  1 file changed, 13 insertions(+), 13 deletions(-)
> 
> diff --git a/mm/memory.c b/mm/memory.c
> index 1ce2e2a..9f907dd 100644
> --- a/mm/memory.c
> +++ b/mm/memory.c
> @@ -833,20 +833,20 @@ copy_one_pte(struct mm_struct *dst_mm, struct mm_struct *src_mm,
>  		if (!pte_file(pte)) {
>  			swp_entry_t entry = pte_to_swp_entry(pte);
>  
> -			if (swap_duplicate(entry) < 0)
> -				return entry.val;
> -
> -			/* make sure dst_mm is on swapoff's mmlist. */
> -			if (unlikely(list_empty(&dst_mm->mmlist))) {
> -				spin_lock(&mmlist_lock);
> -				if (list_empty(&dst_mm->mmlist))
> -					list_add(&dst_mm->mmlist,
> -						 &src_mm->mmlist);
> -				spin_unlock(&mmlist_lock);
> -			}
> -			if (likely(!non_swap_entry(entry)))
> +			if (likely(!non_swap_entry(entry))) {
> +				if (swap_duplicate(entry) < 0)
> +					return entry.val;
> +
> +				/* make sure dst_mm is on swapoff's mmlist. */
> +				if (unlikely(list_empty(&dst_mm->mmlist))) {
> +					spin_lock(&mmlist_lock);
> +					if (list_empty(&dst_mm->mmlist))
> +						list_add(&dst_mm->mmlist,
> +							 &src_mm->mmlist);
> +					spin_unlock(&mmlist_lock);
> +				}
>  				rss[MM_SWAPENTS]++;
> -			else if (is_migration_entry(entry)) {
> +			} else if (is_migration_entry(entry)) {
>  				page = migration_entry_to_page(entry);
>  
>  				if (PageAnon(page))
> -- 
> 1.8.3.1
> 
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo@xxxxxxxxxx  For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>
>
--
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/