Re: [PATCH 10/11] Direct compact when a high-order allocation fails

[Mel Gorman <mel@csn.ul.ie>]

On Wed, Mar 24, 2010 at 08:59:45PM +0900, Minchan Kim wrote:
> On Wed, Mar 24, 2010 at 8:11 PM, Mel Gorman <mel@csn.ul.ie> wrote:
> > On Wed, Mar 24, 2010 at 08:10:40AM +0900, Minchan Kim wrote:
> >> Hi, Mel.
> >>
> >> On Tue, Mar 23, 2010 at 9:25 PM, Mel Gorman <mel@csn.ul.ie> wrote:
> >> > Ordinarily when a high-order allocation fails, direct reclaim is entered to
> >> > free pages to satisfy the allocation.  With this patch, it is determined if
> >> > an allocation failed due to external fragmentation instead of low memory
> >> > and if so, the calling process will compact until a suitable page is
> >> > freed. Compaction by moving pages in memory is considerably cheaper than
> >> > paging out to disk and works where there are locked pages or no swap. If
> >> > compaction fails to free a page of a suitable size, then reclaim will
> >> > still occur.
> >> >
> >> > Direct compaction returns as soon as possible. As each block is compacted,
> >> > it is checked if a suitable page has been freed and if so, it returns.
> >> >
> >> > Signed-off-by: Mel Gorman <mel@csn.ul.ie>
> >> > Acked-by: Rik van Riel <riel@redhat.com>
> >> > ---
> >> >  include/linux/compaction.h |   16 +++++-
> >> >  include/linux/vmstat.h     |    1 +
> >> >  mm/compaction.c            |  118 ++++++++++++++++++++++++++++++++++++++++++++
> >> >  mm/page_alloc.c            |   26 ++++++++++
> >> >  mm/vmstat.c                |   15 +++++-
> >> >  5 files changed, 172 insertions(+), 4 deletions(-)
> >> >
> >> > diff --git a/include/linux/compaction.h b/include/linux/compaction.h
> >> > index c94890b..b851428 100644
> >> > --- a/include/linux/compaction.h
> >> > +++ b/include/linux/compaction.h
> >> > @@ -1,14 +1,26 @@
> >> >  #ifndef _LINUX_COMPACTION_H
> >> >  #define _LINUX_COMPACTION_H
> >> >
> >> > -/* Return values for compact_zone() */
> >> > +/* Return values for compact_zone() and try_to_compact_pages() */
> >> >  #define COMPACT_INCOMPLETE     0
> >> > -#define COMPACT_COMPLETE       1
> >> > +#define COMPACT_PARTIAL                1
> >> > +#define COMPACT_COMPLETE       2
> >> >
> >> >  #ifdef CONFIG_COMPACTION
> >> >  extern int sysctl_compact_memory;
> >> >  extern int sysctl_compaction_handler(struct ctl_table *table, int write,
> >> >                        void __user *buffer, size_t *length, loff_t *ppos);
> >> > +
> >> > +extern int fragmentation_index(struct zone *zone, unsigned int order);
> >> > +extern unsigned long try_to_compact_pages(struct zonelist *zonelist,
> >> > +                       int order, gfp_t gfp_mask, nodemask_t *mask);
> >> > +#else
> >> > +static inline unsigned long try_to_compact_pages(struct zonelist *zonelist,
> >> > +                       int order, gfp_t gfp_mask, nodemask_t *nodemask)
> >> > +{
> >> > +       return COMPACT_INCOMPLETE;
> >> > +}
> >> > +
> >> >  #endif /* CONFIG_COMPACTION */
> >> >
> >> >  #if defined(CONFIG_COMPACTION) && defined(CONFIG_SYSFS) && defined(CONFIG_NUMA)
> >> > diff --git a/include/linux/vmstat.h b/include/linux/vmstat.h
> >> > index 56e4b44..b4b4d34 100644
> >> > --- a/include/linux/vmstat.h
> >> > +++ b/include/linux/vmstat.h
> >> > @@ -44,6 +44,7 @@ enum vm_event_item { PGPGIN, PGPGOUT, PSWPIN, PSWPOUT,
> >> >                KSWAPD_SKIP_CONGESTION_WAIT,
> >> >                PAGEOUTRUN, ALLOCSTALL, PGROTATED,
> >> >                COMPACTBLOCKS, COMPACTPAGES, COMPACTPAGEFAILED,
> >> > +               COMPACTSTALL, COMPACTFAIL, COMPACTSUCCESS,
> >> >  #ifdef CONFIG_HUGETLB_PAGE
> >> >                HTLB_BUDDY_PGALLOC, HTLB_BUDDY_PGALLOC_FAIL,
> >> >  #endif
> >> > diff --git a/mm/compaction.c b/mm/compaction.c
> >> > index 8df6e3d..6688700 100644
> >> > --- a/mm/compaction.c
> >> > +++ b/mm/compaction.c
> >> > @@ -34,6 +34,8 @@ struct compact_control {
> >> >        unsigned long nr_anon;
> >> >        unsigned long nr_file;
> >> >
> >> > +       unsigned int order;             /* order a direct compactor needs */
> >> > +       int migratetype;                /* MOVABLE, RECLAIMABLE etc */
> >> >        struct zone *zone;
> >> >  };
> >> >
> >> > @@ -301,10 +303,31 @@ static void update_nr_listpages(struct compact_control *cc)
> >> >  static inline int compact_finished(struct zone *zone,
> >> >                                                struct compact_control *cc)
> >> >  {
> >> > +       unsigned int order;
> >> > +       unsigned long watermark = low_wmark_pages(zone) + (1 << cc->order);
> >> > +
> >> >        /* Compaction run completes if the migrate and free scanner meet */
> >> >        if (cc->free_pfn <= cc->migrate_pfn)
> >> >                return COMPACT_COMPLETE;
> >> >
> >> > +       /* Compaction run is not finished if the watermark is not met */
> >> > +       if (!zone_watermark_ok(zone, cc->order, watermark, 0, 0))
> >> > +               return COMPACT_INCOMPLETE;
> >> > +
> >> > +       if (cc->order == -1)
> >> > +               return COMPACT_INCOMPLETE;
> >> > +
> >> > +       /* Direct compactor: Is a suitable page free? */
> >> > +       for (order = cc->order; order < MAX_ORDER; order++) {
> >> > +               /* Job done if page is free of the right migratetype */
> >> > +               if (!list_empty(&zone->free_area[order].free_list[cc->migratetype]))
> >> > +                       return COMPACT_PARTIAL;
> >> > +
> >> > +               /* Job done if allocation would set block type */
> >> > +               if (order >= pageblock_order && zone->free_area[order].nr_free)
> >> > +                       return COMPACT_PARTIAL;
> >> > +       }
> >> > +
> >> >        return COMPACT_INCOMPLETE;
> >> >  }
> >> >
> >> > @@ -348,6 +371,101 @@ static int compact_zone(struct zone *zone, struct compact_control *cc)
> >> >        return ret;
> >> >  }
> >> >
> >> > +static inline unsigned long compact_zone_order(struct zone *zone,
> >> > +                                               int order, gfp_t gfp_mask)
> >> > +{
> >> > +       struct compact_control cc = {
> >> > +               .nr_freepages = 0,
> >> > +               .nr_migratepages = 0,
> >> > +               .order = order,
> >> > +               .migratetype = allocflags_to_migratetype(gfp_mask),
> >> > +               .zone = zone,
> >> > +       };
> >> > +       INIT_LIST_HEAD(&cc.freepages);
> >> > +       INIT_LIST_HEAD(&cc.migratepages);
> >> > +
> >> > +       return compact_zone(zone, &cc);
> >> > +}
> >> > +
> >> > +/**
> >> > + * try_to_compact_pages - Direct compact to satisfy a high-order allocation
> >> > + * @zonelist: The zonelist used for the current allocation
> >> > + * @order: The order of the current allocation
> >> > + * @gfp_mask: The GFP mask of the current allocation
> >> > + * @nodemask: The allowed nodes to allocate from
> >> > + *
> >> > + * This is the main entry point for direct page compaction.
> >> > + */
> >> > +unsigned long try_to_compact_pages(struct zonelist *zonelist,
> >> > +                       int order, gfp_t gfp_mask, nodemask_t *nodemask)
> >> > +{
> >> > +       enum zone_type high_zoneidx = gfp_zone(gfp_mask);
> >> > +       int may_enter_fs = gfp_mask & __GFP_FS;
> >> > +       int may_perform_io = gfp_mask & __GFP_IO;
> >> > +       unsigned long watermark;
> >> > +       struct zoneref *z;
> >> > +       struct zone *zone;
> >> > +       int rc = COMPACT_INCOMPLETE;
> >> > +
> >> > +       /* Check whether it is worth even starting compaction */
> >> > +       if (order == 0 || !may_enter_fs || !may_perform_io)
> >> > +               return rc;
> >> > +
> >> > +       /*
> >> > +        * We will not stall if the necessary conditions are not met for
> >> > +        * migration but direct reclaim seems to account stalls similarly
> >> > +        */
> >>
> >> I can't understand this comment.
> >> In case of direct reclaim, shrink_zones's long time is just stall
> >> by view point of allocation customer.
> >> So "Allocation is stalled" makes sense to me.
> >>
> >> But "Compaction is stalled" doesn't make sense to me.
> >
> > I considered a "stall" to be when the allocator is doing work that is not
> > allocation-related such as page reclaim or in this case - memory compaction.
> 
> I agree.
> 
> >
> >> How about "COMPACTION_DIRECT" like "PGSCAN_DIRECT"?
> >
> > PGSCAN_DIRECT is page-based counter on the number of pages scanned. The
> > similar naming but very different meaning could be confusing to someone not
> > familar with the counters. The event being counted here is the number of
> > times compaction happened just like ALLOCSTALL counts the number of times
> > direct reclaim happened.
> 
> You're right. I just wanted to change the name as one which imply
> direct compaction.

I think I'd fully agree with your point if there was more than one way to
stall a process due to compaction. As it is, direct compaction is the only
way to meaningfully stall a process and I can't think of alternative stalls
in the future. Technically, a process using the sysfs or proc triggers for
compaction also stalls but it's not interesting to count those events.

> That's because I believe we will implement it by backgroud, too.

This is a possibility but in that case it would be a separate process
like kcompactd and I wouldn't count it as a stall as such.

> Then It's more straightforward, I think. :-)
> 
> > How about COMPACTSTALL like ALLOCSTALL? :/
> 
> I wouldn't have a strong objection any more if you insist on it.
> 

I'm not insisting as such, I just don't think renaming it to
PGSCAN_COMPACT_X would be easier to understand.

> >> I think It's straightforward.
> >> Naming is important since it makes ABI.
> >>
> >> > +       count_vm_event(COMPACTSTALL);
> >> > +
> >>

-- 
Mel Gorman
Part-time Phd Student                          Linux Technology Center
University of Limerick                         IBM Dublin Software Lab