Re: [patch] mm, migrate: increment fail count on ENOMEM

[Vlastimil Babka]

On 05/20/2016 03:06 PM, Michal Hocko wrote:
> On Thu 19-05-16 15:11:23, David Rientjes wrote:
> > If page migration fails due to -ENOMEM, nr_failed should still be
> > incremented for proper statistics.
> >
> > This was encountered recently when all page migration vmstats showed 0,
> > and inferred that migrate_pages() was never called, although in reality
> > the first page migration failed because compaction_alloc() failed to find
> > a migration target.
> >
> > This patch increments nr_failed so the vmstat is properly accounted on
> > ENOMEM.
> >
> > Signed-off-by: David Rientjes <rientjes@xxxxxxxxxx>
>
> Acked-by: Michal Hocko <mhocko@xxxxxxxx>

Acked-by: Vlastimil Babka <vbabka@xxxxxxx>

> One question though
>
> > ---
> >   mm/migrate.c | 1 +
> >   1 file changed, 1 insertion(+)
> >
> > diff --git a/mm/migrate.c b/mm/migrate.c
> > --- a/mm/migrate.c
> > +++ b/mm/migrate.c
> > @@ -1171,6 +1171,7 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
> >
> >   			switch(rc) {
> >   			case -ENOMEM:
> > +				nr_failed++;
> >   				goto out;
> >   			case -EAGAIN:
> >   				retry++;
>
> Why don't we need also to count also retries?

We could, but not like you suggest.

> ---
> diff --git a/mm/migrate.c b/mm/migrate.c
> index 53ab6398e7a2..ef9c5211ae3c 100644
> --- a/mm/migrate.c
> +++ b/mm/migrate.c
> @@ -1190,9 +1190,9 @@ int migrate_pages(struct list_head *from, new_page_t get_new_page,
>   			}
>   		}
>   	}
> +out:
>   	nr_failed += retry;
>   	rc = nr_failed;

This overwrites rc == -ENOMEM, which at least compaction needs to 
recognize. But we could duplicate "nr_failed += retry" in the case -ENOMEM.

> -out:
>   	if (nr_succeeded)
>   		count_vm_events(PGMIGRATE_SUCCESS, nr_succeeded);
>   	if (nr_failed)