From: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
To: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Cc: kosaki.motohiro@jp.fujitsu.com,
Wu Fengguang <fengguang.wu@intel.com>,
Minchan Kim <minchan.kim@gmail.com>,
Huang Shijie <shijie8@gmail.com>,
"akpm@linux-foundation.org" <akpm@linux-foundation.org>,
"mel@csn.ul.ie" <mel@csn.ul.ie>,
"linux-mm@kvack.org" <linux-mm@kvack.org>,
Rik van Riel <riel@redhat.com>
Subject: Re: [PATCH 4/4] mm/page_alloc : relieve zone->lock's pressure for memory free
Date: Tue, 12 Jan 2010 14:10:36 +0900 (JST) [thread overview]
Message-ID: <20100112140923.B3A4.A69D9226@jp.fujitsu.com> (raw)
In-Reply-To: <20100112133223.005b81ed.kamezawa.hiroyu@jp.fujitsu.com>
> On Tue, 12 Jan 2010 12:21:16 +0800
> Wu Fengguang <fengguang.wu@intel.com> wrote:
> > > BTW,
> > > Hmm. It's not atomic as Kame pointed out.
> > >
> > > Now, zone->flags have several bit.
> > > * ZONE_ALL_UNRECLAIMALBE
> > > * ZONE_RECLAIM_LOCKED
> > > * ZONE_OOM_LOCKED.
> > >
> > > I think this flags are likely to race when the memory pressure is high.
> > > If we don't prevent race, concurrent reclaim and killing could be happened.
> > > So I think reset zone->flags outside of zone->lock would make our efforts which
> > > prevent current reclaim and killing invalidate.
> >
> > zone_set_flag()/zone_clear_flag() calls set_bit()/clear_bit() which is
> > atomic. Do you mean more high level exclusion?
> >
> Ah, sorry, I missed that.
> In my memory, this wasn't atomic ;) ...maybe recent change.
>
> I don't want to see atomic_ops here...So, how about making this back to be
> zone->all_unreclaimable word ?
>
> Clearing this is not necessary to be atomic because this is cleard at every
> page freeing.
I agree. How about this?
From 751f197ad256c7245151681d7aece591b1dab343 Mon Sep 17 00:00:00 2001
From: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
Date: Tue, 12 Jan 2010 13:53:47 +0900
Subject: [PATCH] mm: Restore zone->all_unreclaimable to independence word
commit e815af95 (change all_unreclaimable zone member to flags) chage
all_unreclaimable member to bit flag. but It have undesireble side
effect.
free_one_page() is one of most hot path in linux kernel and increasing
atomic ops in it can reduce kernel performance a bit.
Thus, this patch revert such commit partially. at least
all_unreclaimable shouldn't share memory word with other zone flags.
Signed-off-by: KOSAKI Motohiro <kosaki.motohiro@jp.fujitsu.com>
---
include/linux/mmzone.h | 7 +------
mm/page_alloc.c | 6 +++---
mm/vmscan.c | 20 ++++++++------------
mm/vmstat.c | 2 +-
4 files changed, 13 insertions(+), 22 deletions(-)
diff --git a/include/linux/mmzone.h b/include/linux/mmzone.h
index 30fe668..4f0c6f1 100644
--- a/include/linux/mmzone.h
+++ b/include/linux/mmzone.h
@@ -341,6 +341,7 @@ struct zone {
unsigned long pages_scanned; /* since last reclaim */
unsigned long flags; /* zone flags, see below */
+ int all_unreclaimable; /* All pages pinned */
/* Zone statistics */
atomic_long_t vm_stat[NR_VM_ZONE_STAT_ITEMS];
@@ -425,7 +426,6 @@ struct zone {
} ____cacheline_internodealigned_in_smp;
typedef enum {
- ZONE_ALL_UNRECLAIMABLE, /* all pages pinned */
ZONE_RECLAIM_LOCKED, /* prevents concurrent reclaim */
ZONE_OOM_LOCKED, /* zone is in OOM killer zonelist */
} zone_flags_t;
@@ -445,11 +445,6 @@ static inline void zone_clear_flag(struct zone *zone, zone_flags_t flag)
clear_bit(flag, &zone->flags);
}
-static inline int zone_is_all_unreclaimable(const struct zone *zone)
-{
- return test_bit(ZONE_ALL_UNRECLAIMABLE, &zone->flags);
-}
-
static inline int zone_is_reclaim_locked(const struct zone *zone)
{
return test_bit(ZONE_RECLAIM_LOCKED, &zone->flags);
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index 4e9f5cc..19a5b0e 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -530,7 +530,7 @@ static void free_pcppages_bulk(struct zone *zone, int count,
int batch_free = 0;
spin_lock(&zone->lock);
- zone_clear_flag(zone, ZONE_ALL_UNRECLAIMABLE);
+ zone->all_unreclaimable = 0;
zone->pages_scanned = 0;
__mod_zone_page_state(zone, NR_FREE_PAGES, count);
@@ -567,7 +567,7 @@ static void free_one_page(struct zone *zone, struct page *page, int order,
int migratetype)
{
spin_lock(&zone->lock);
- zone_clear_flag(zone, ZONE_ALL_UNRECLAIMABLE);
+ zone->all_unreclaimable = 0;
zone->pages_scanned = 0;
__mod_zone_page_state(zone, NR_FREE_PAGES, 1 << order);
@@ -2270,7 +2270,7 @@ void show_free_areas(void)
K(zone_page_state(zone, NR_BOUNCE)),
K(zone_page_state(zone, NR_WRITEBACK_TEMP)),
zone->pages_scanned,
- (zone_is_all_unreclaimable(zone) ? "yes" : "no")
+ (zone->all_unreclaimable ? "yes" : "no")
);
printk("lowmem_reserve[]:");
for (i = 0; i < MAX_NR_ZONES; i++)
diff --git a/mm/vmscan.c b/mm/vmscan.c
index 885207a..8057d36 100644
--- a/mm/vmscan.c
+++ b/mm/vmscan.c
@@ -1694,8 +1694,7 @@ static void shrink_zones(int priority, struct zonelist *zonelist,
continue;
note_zone_scanning_priority(zone, priority);
- if (zone_is_all_unreclaimable(zone) &&
- priority != DEF_PRIORITY)
+ if (zone->all_unreclaimable && priority != DEF_PRIORITY)
continue; /* Let kswapd poll it */
sc->all_unreclaimable = 0;
} else {
@@ -2009,8 +2008,7 @@ loop_again:
if (!populated_zone(zone))
continue;
- if (zone_is_all_unreclaimable(zone) &&
- priority != DEF_PRIORITY)
+ if (zone->all_unreclaimable && priority != DEF_PRIORITY)
continue;
/*
@@ -2053,8 +2051,7 @@ loop_again:
if (!populated_zone(zone))
continue;
- if (zone_is_all_unreclaimable(zone) &&
- priority != DEF_PRIORITY)
+ if (zone->all_unreclaimable && priority != DEF_PRIORITY)
continue;
if (!zone_watermark_ok(zone, order,
@@ -2084,12 +2081,11 @@ loop_again:
lru_pages);
sc.nr_reclaimed += reclaim_state->reclaimed_slab;
total_scanned += sc.nr_scanned;
- if (zone_is_all_unreclaimable(zone))
+ if (zone->all_unreclaimable)
continue;
- if (nr_slab == 0 && zone->pages_scanned >=
- (zone_reclaimable_pages(zone) * 6))
- zone_set_flag(zone,
- ZONE_ALL_UNRECLAIMABLE);
+ if (nr_slab == 0 &&
+ zone->pages_scanned >= (zone_reclaimable_pages(zone) * 6))
+ zone->all_unreclaimable = 1;
/*
* If we've done a decent amount of scanning and
* the reclaim ratio is low, start doing writepage
@@ -2612,7 +2608,7 @@ int zone_reclaim(struct zone *zone, gfp_t gfp_mask, unsigned int order)
zone_page_state(zone, NR_SLAB_RECLAIMABLE) <= zone->min_slab_pages)
return ZONE_RECLAIM_FULL;
- if (zone_is_all_unreclaimable(zone))
+ if (zone->all_unreclaimable)
return ZONE_RECLAIM_FULL;
/*
diff --git a/mm/vmstat.c b/mm/vmstat.c
index 6051fba..8175c64 100644
--- a/mm/vmstat.c
+++ b/mm/vmstat.c
@@ -761,7 +761,7 @@ static void zoneinfo_show_print(struct seq_file *m, pg_data_t *pgdat,
"\n prev_priority: %i"
"\n start_pfn: %lu"
"\n inactive_ratio: %u",
- zone_is_all_unreclaimable(zone),
+ zone->all_unreclaimable,
zone->prev_priority,
zone->zone_start_pfn,
zone->inactive_ratio);
--
1.6.5.2
--
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>
next prev parent reply other threads:[~2010-01-12 5:10 UTC|newest]
Thread overview: 41+ messages / expand[flat|nested] mbox.gz Atom feed top
2010-01-11 4:37 [PATCH 1/4] mm/page_alloc : rename rmqueue_bulk to rmqueue_single Huang Shijie
2010-01-11 4:37 ` [PATCH 2/4] mm/page_alloc : relieve the zone->lock's pressure for allocation Huang Shijie
2010-01-11 4:37 ` [PATCH 3/4] mm/page_alloc : modify the return type of __free_one_page Huang Shijie
2010-01-11 4:37 ` [PATCH 4/4] mm/page_alloc : relieve zone->lock's pressure for memory free Huang Shijie
2010-01-11 5:20 ` Minchan Kim
2010-01-11 6:01 ` Huang Shijie
2010-01-11 6:27 ` Huang Shijie
2010-01-11 6:38 ` Minchan Kim
2010-01-11 6:59 ` Huang Shijie
2010-01-12 0:47 ` KAMEZAWA Hiroyuki
2010-01-12 2:02 ` Huang Shijie
2010-01-12 2:07 ` KAMEZAWA Hiroyuki
2010-01-12 2:32 ` Huang Shijie
2010-01-12 2:27 ` Wu Fengguang
2010-01-12 2:56 ` KOSAKI Motohiro
2010-01-12 3:02 ` Huang Shijie
2010-01-12 4:05 ` Minchan Kim
2010-01-12 4:21 ` Wu Fengguang
2010-01-12 4:32 ` KAMEZAWA Hiroyuki
2010-01-12 4:59 ` Wu Fengguang
2010-01-12 5:09 ` Wu Fengguang
2010-01-12 5:10 ` KOSAKI Motohiro [this message]
2010-01-12 7:36 ` David Rientjes
2010-01-12 8:56 ` KOSAKI Motohiro
2010-01-12 21:39 ` David Rientjes
2010-01-13 0:01 ` KOSAKI Motohiro
2010-01-12 4:48 ` Minchan Kim
2010-01-12 2:51 ` Huang Shijie
2010-01-12 3:03 ` Wu Fengguang
2010-01-12 3:05 ` KOSAKI Motohiro
2010-01-11 5:04 ` [PATCH 3/4] mm/page_alloc : modify the return type of __free_one_page Minchan Kim
2010-01-12 2:56 ` KOSAKI Motohiro
2010-01-18 11:25 ` Mel Gorman
2010-01-19 1:49 ` Huang Shijie
2010-01-11 5:02 ` [PATCH 2/4] mm/page_alloc : relieve the zone->lock's pressure for allocation Minchan Kim
2010-01-11 5:13 ` Huang Shijie
2010-01-12 2:54 ` KOSAKI Motohiro
2010-01-18 11:24 ` Mel Gorman
2010-01-11 5:00 ` [PATCH 1/4] mm/page_alloc : rename rmqueue_bulk to rmqueue_single Minchan Kim
2010-01-12 2:52 ` KOSAKI Motohiro
2010-01-18 11:21 ` Mel Gorman
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=20100112140923.B3A4.A69D9226@jp.fujitsu.com \
--to=kosaki.motohiro@jp.fujitsu.com \
--cc=akpm@linux-foundation.org \
--cc=fengguang.wu@intel.com \
--cc=kamezawa.hiroyu@jp.fujitsu.com \
--cc=linux-mm@kvack.org \
--cc=mel@csn.ul.ie \
--cc=minchan.kim@gmail.com \
--cc=riel@redhat.com \
--cc=shijie8@gmail.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox