Re: [Patch v2 4/4] mm/migrate.c: handle same node and add failure in the same way

[David Hildenbrand]

On 22.01.20 02:16, Wei Yang wrote:
> When page is not queued for migration, there are two possible cases:
> 
>   * page already on the target node
>   * failed to add to migration queue
> 
> Current code handle them differently, this leads to a behavior
> inconsistency.
> 
> Usually for each page's status, we just do store for once. While for the
> page already on the target node, we might store the node information for
> twice:
> 
>   * once when we found the page is on the target node
>   * second when moving the pages to target node successfully after above
>     action
> 
> The reason is even we don't add the page to pagelist, but store_status()
> does store in a range which still contains the page.
> 
> This patch handles these two cases in the same way to reduce this
> inconsistency and also make the code a little easier to read.
> 

I'd rephrase to

"mm/migrate.c: unify "not queued for migration" handling in do_pages_move()

It can currently happen that we store the status of a page twice:
* Once we detect that it is already on the target node
* Once we moved a bunch of pages, and a page that's already on the
  target node is contained in the current interval.

Let's simplify the code and always call do_move_pages_to_node() in
case we did not queue a page for migration. Note that pages that are
already on the target node are not added to the pagelist and are,
therefore, ignored by do_move_pages_to_node() - there is no functional
change.

The status of such a page is now only stored once.
"

> Signed-off-by: Wei Yang <richardw.yang@xxxxxxxxxxxxxxx>
> Acked-by: Michal Hocko <mhocko@xxxxxxxx>
> ---
>  mm/migrate.c | 16 ++++++++--------
>  1 file changed, 8 insertions(+), 8 deletions(-)
> 
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 80d2bba57265..591f2e5caed6 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1654,18 +1654,18 @@ static int do_pages_move(struct mm_struct *mm, nodemask_t task_nodes,
>  		err = add_page_for_migration(mm, addr, current_node,
>  				&pagelist, flags & MPOL_MF_MOVE_ALL);
>  
> -		if (!err) {
> -			/* The page is already on the target node */
> -			err = store_status(status, i, current_node, 1);
> -			if (err)
> -				goto out_flush;
> -			continue;
> -		} else if (err > 0) {
> +		if (err > 0) {
>  			/* The page is successfully queued for migration */
>  			continue;
>  		}
>  
> -		err = store_status(status, i, err, 1);
> +		/*
> +		 * Two possible cases for err here:
> +		 * == 0: page is already on the target node, then store
> +		 *       current_node to status
> +		 * <  0: failed to add page to list, then store err to status
> +		 */

I'd shorten that to

/*
 * If the page is already on the target node (!err), store the node,
 * otherwise, store the err.
*/

> +		err = store_status(status, i, err ? : current_node, 1);
>  		if (err)
>  			goto out_flush;
>  
> 

Thanks!

Reviewed-by: David Hildenbrand <david@xxxxxxxxxx>

-- 
Thanks,

David / dhildenb