On Mon 17-10-16 08:06:18, Minchan Kim wrote:
> Hi Michal,
>
> On Sat, Oct 15, 2016 at 09:10:45AM +0200, Michal Hocko wrote:
> > On Sat 15-10-16 00:26:33, Minchan Kim wrote:
> > > On Fri, Oct 14, 2016 at 05:03:55PM +0200, Michal Hocko wrote:
> > [...]
> > > > diff --git a/mm/compaction.c b/mm/compaction.c
> > > > index 0409a4ad6ea1..6584705a46f6 100644
> > > > --- a/mm/compaction.c
> > > > +++ b/mm/compaction.c
> > > > @@ -685,7 +685,8 @@ static bool too_many_isolated(struct zone *zone)
> > > > */
> > > > static unsigned long
> > > > isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
> > > > - unsigned long end_pfn, isolate_mode_t isolate_mode)
> > > > + unsigned long end_pfn, isolate_mode_t isolate_mode,
> > > > + unsigned long *isolated_file, unsigned long *isolated_anon)
> > > > {
> > > > struct zone *zone = cc->zone;
> > > > unsigned long nr_scanned = 0, nr_isolated = 0;
> > > > @@ -866,6 +867,10 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
> > > >
> > > > /* Successfully isolated */
> > > > del_page_from_lru_list(page, lruvec, page_lru(page));
> > > > + if (page_is_file_cache(page))
> > > > + (*isolated_file)++;
> > > > + else
> > > > + (*isolated_anon)++;
> > > >
> > > > isolate_success:
> > > > list_add(&page->lru, &cc->migratepages);
> > > >
> > > > Makes more sense?
> > >
> > > It is doable for isolation part. IOW, maybe we can make acct_isolated
> > > simple with those counters but we need to handle migrate, putback part.
> > > If you want to remove the check of __PageMoable with those counter, it
> > > means we should pass the counter on every functions related migration
> > > where isolate, migrate, putback parts.
> >
> > OK, I see. Can we just get rid of acct_isolated altogether? Why cannot
> > we simply update NR_ISOLATED_* while isolating pages? Just looking at
> > isolate_migratepages_block:
> > acct_isolated(zone, cc);
> > putback_movable_pages(&cc->migratepages);
> >
> > suggests we are doing something suboptimal. I guess we cannot get rid of
> > __PageMoveble checks which is sad because that just adds a lot of
> > confusion because checking for !__PageMovable(page) for LRU pages is
> > just a head scratcher (LRU pages are movable arent' they?). Maybe it
> > would be even good to get rid of this misnomer. PageNonLRUMovable?
>
> Yeah, I hated the naming but didn't have a good idea.
> PageNonLRUMovable, definitely, one I thought as candidate but dropped
> by lenghthy naming. If others don't object, I am happy to change it.
Yes it is long but it is less confusing because it is just utterly
confusing to test for LRU pages with !__PageMovable when in fact they
are movable. Heck even unreclaimable pages are movable unless explicitly
configured to not be.
> > Anyway, I would suggest to do something like this. Batching NR_ISOLATED*
> > just doesn't make all that much sense as these are per-cpu and the
> > resulting code seems to be easier without it.
>
> Agree. Could you resend it as formal patch?
Sure, what do you think about the following? I haven't marked it for
stable because there was no bug report for it AFAIU.
---
>From 3b2bd4486f36ada9f6dc86d3946855281455ba9f Mon Sep 17 00:00:00 2001
From: Ming Ling <ming.ling@xxxxxxxxxxxxxx>
Date: Mon, 17 Oct 2016 10:26:50 +0200
Subject: [PATCH] mm, compaction: fix NR_ISOLATED_* stats for pfn based
migration
Since bda807d44454 ("mm: migrate: support non-lru movable page
migration") isolate_migratepages_block) can isolate !PageLRU pages which
would acct_isolated account as NR_ISOLATED_*. Accounting these non-lru
pages NR_ISOLATED_{ANON,FILE} doesn't make any sense and it can misguide
heuristics based on those counters such as pgdat_reclaimable_pages resp.
too_many_isolated which would lead to unexpected stalls during the
direct reclaim without any good reason. Note that
__alloc_contig_migrate_range can isolate a lot of pages at once.
On mobile devices such as 512M ram android Phone, it may use a big zram
swap. In some cases zram(zsmalloc) uses too many non-lru but migratedable
pages, such as:
MemTotal: 468148 kB
Normal free:5620kB
Free swap:4736kB
Total swap:409596kB
ZRAM: 164616kB(zsmalloc non-lru pages)
active_anon:60700kB
inactive_anon:60744kB
active_file:34420kB
inactive_file:37532kB
Fix this by only accounting lru pages to NR_ISOLATED_* in
isolate_migratepages_block right after they were isolated and we still
know they were on LRU. Drop acct_isolated because it is called after the
fact and we've lost that information. Batching per-cpu counter doesn't
make much improvement anyway. Also make sure that we uncharge only LRU
pages when putting them back on the LRU in putback_movable_pages resp.
when unmap_and_move migrates the page.
Fixes: bda807d44454 ("mm: migrate: support non-lru movable page migration")
Signed-off-by: Ming Ling <ming.ling@xxxxxxxxxxxxxx>
Signed-off-by: Michal Hocko <mhocko@xxxxxxxx>
---
mm/compaction.c | 25 +++----------------------
mm/migrate.c | 15 +++++++++++----
2 files changed, 14 insertions(+), 26 deletions(-)
diff --git a/mm/compaction.c b/mm/compaction.c
index 0409a4ad6ea1..df1fd0c20e5c 100644
--- a/mm/compaction.c
+++ b/mm/compaction.c
@@ -634,22 +634,6 @@ isolate_freepages_range(struct compact_control *cc,
return pfn;
}
-/* Update the number of anon and file isolated pages in the zone */
-static void acct_isolated(struct zone *zone, struct compact_control *cc)
-{
- struct page *page;
- unsigned int count[2] = { 0, };
-
- if (list_empty(&cc->migratepages))
- return;
-
- list_for_each_entry(page, &cc->migratepages, lru)
- count[!!page_is_file_cache(page)]++;
-
- mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_ANON, count[0]);
- mod_node_page_state(zone->zone_pgdat, NR_ISOLATED_FILE, count[1]);
-}
-
/* Similar to reclaim, but different enough that they don't share logic */
static bool too_many_isolated(struct zone *zone)
{
@@ -866,6 +850,8 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
/* Successfully isolated */
del_page_from_lru_list(page, lruvec, page_lru(page));
+ inc_node_page_state(zone->zone_pgdat,
+ NR_ISOLATED_ANON + page_is_file_cache(page));
isolate_success:
list_add(&page->lru, &cc->migratepages);
@@ -902,7 +888,6 @@ isolate_migratepages_block(struct compact_control *cc, unsigned long low_pfn,
spin_unlock_irqrestore(zone_lru_lock(zone), flags);
locked = false;
}
- acct_isolated(zone, cc);
putback_movable_pages(&cc->migratepages);
cc->nr_migratepages = 0;
cc->last_migrated_pfn = 0;
@@ -988,7 +973,6 @@ isolate_migratepages_range(struct compact_control *cc, unsigned long start_pfn,
if (cc->nr_migratepages == COMPACT_CLUSTER_MAX)
break;
}
- acct_isolated(cc->zone, cc);
return pfn;
}
@@ -1258,10 +1242,8 @@ static isolate_migrate_t isolate_migratepages(struct zone *zone,
low_pfn = isolate_migratepages_block(cc, low_pfn,
block_end_pfn, isolate_mode);
- if (!low_pfn || cc->contended) {
- acct_isolated(zone, cc);
+ if (!low_pfn || cc->contended)
return ISOLATE_ABORT;
- }
/*
* Either we isolated something and proceed with migration. Or
@@ -1271,7 +1253,6 @@ static isolate_migrate_t isolate_migratepages(struct zone *zone,
break;
}
- acct_isolated(zone, cc);
/* Record where migration scanner will be restarted. */
cc->migrate_pfn = low_pfn;
diff --git a/mm/migrate.c b/mm/migrate.c
index 99250aee1ac1..66ce6b490b13 100644
--- a/mm/migrate.c
+++ b/mm/migrate.c
@@ -168,8 +168,6 @@ void putback_movable_pages(struct list_head *l)
continue;
}
list_del(&page->lru);
- dec_node_page_state(page, NR_ISOLATED_ANON +
- page_is_file_cache(page));
/*
* We isolated non-lru movable page so here we can use
* __PageMovable because LRU page's mapping cannot have
@@ -186,6 +184,8 @@ void putback_movable_pages(struct list_head *l)
put_page(page);
} else {
putback_lru_page(page);
+ dec_node_page_state(page, NR_ISOLATED_ANON +
+ page_is_file_cache(page));
}
}
}
@@ -1121,8 +1121,15 @@ static ICE_noinline int unmap_and_move(new_page_t get_new_page,
* restored.
*/
list_del(&page->lru);
- dec_node_page_state(page, NR_ISOLATED_ANON +
- page_is_file_cache(page));
+
+ /*
+ * Compaction can migrate also non-LRU pages which are
+ * not accounted to NR_ISOLATED_*. They can be recognized
+ * as __PageMovable
+ */
+ if (likely(!__PageMovable(page)))
+ dec_node_page_state(page, NR_ISOLATED_ANON +
+ page_is_file_cache(page));
}
/*
--
2.9.3
--
Michal Hocko
SUSE Labs
--
To unsubscribe, send a message with 'unsubscribe linux-mm' in
the body to majordomo@xxxxxxxxx. For more info on Linux MM,
see: http://www.linux-mm.org/ .
Don't email: <a href=mailto:"dont@xxxxxxxxx";> email@xxxxxxxxx </a>