Re: [PATCH v5 1/4] mm/page_alloc: fix incorrect isolation behavior by rechecking migratetype

[Mel Gorman <mgorman@suse.de>]

> > diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
> > index 4593567..3d090af 100644
> > --- a/include/linux/mmzone.h
> > +++ b/include/linux/mmzone.h
> > @@ -431,6 +431,15 @@ struct zone {
> >          */
> >         int                     nr_migrate_reserve_block;
> >
> > +#ifdef CONFIG_MEMORY_ISOLATION
> > +       /*
> > +        * Number of isolated pageblock. It is used to solve incorrect
> > +        * freepage counting problem due to racy retrieving migratetype
> > +        * of pageblock. Protected by zone->lock.
> > +        */
> > +       unsigned long           nr_isolate_pageblock;
> > +#endif
> > +
> 
> First sorry for this deferred reply, I see these patches have been merged
> into the mainline.
> However, I still have a tiny question:
> Why use ZONE_PADDING(_pad1_)  seperate it and zone->lock?
> How about move it to the same cacheline with zone->lock, because it is
> accessed under zone->lock?
> 

zone->lock is currently sharing lines with the data that is frequently
updated under zone lock and some of the dirty data cache line bouncing has
completed when the lock is acquired. nr_isolate_pageblock is a read-mostly
field and in some cases will never be used. It's fine where it is beside
other read-mostly fields.

-- 
Mel Gorman
SUSE Labs

--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@kvack.org.  For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@kvack.org"> email@kvack.org </a>