From: Johannes Weiner <hannes@cmpxchg.org>
To: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>,
Andrew Morton <akpm@linux-foundation.org>,
kosaki.motohiro@jp.fujitsu.com,
Maxim Levitsky <maximlevitsky@gmail.com>,
linux-kernel@vger.kernel.org, linux-mm@kvack.org,
Lee Schermerhorn <Lee.Schermerhorn@hp.com>,
Pekka Enberg <penberg@cs.helsinki.fi>,
Jiri Slaby <jirislaby@gmail.com>
Subject: Re: [PATCH] mm: Warn once when a page is freed with PG_mlocked set V2
Date: Thu, 16 Jul 2009 00:04:45 +0200 [thread overview]
Message-ID: <20090715220445.GA1823@cmpxchg.org> (raw)
In-Reply-To: <alpine.DEB.1.10.0907151027410.23643@gentwo.org>
Hello Christoph,
On Wed, Jul 15, 2009 at 10:31:54AM -0400, Christoph Lameter wrote:
> On Wed, 15 Jul 2009, Mel Gorman wrote:
>
> > -static inline int free_pages_check(struct page *page)
> > +static inline int free_pages_check(struct page *page, int wasMlocked)
> > {
> > + WARN_ONCE(wasMlocked, KERN_WARNING
> > + "Page flag mlocked set for process %s at pfn:%05lx\n"
> > + "page:%p flags:0x%lX\n",
> > + current->comm, page_to_pfn(page),
> > + page, page->flags|__PG_MLOCKED);
> > +
> > if (unlikely(page_mapcount(page) |
>
> There is already a free_page_mlocked() that is only called if the mlock
> bit is set. Move it into there to avoid having to run two checks in the
> hot codee path?
>
> Also __free_pages_ok() now has a TestClearMlocked in the hot code path.
> Would it be possible to get rid of the unconditional use of an atomic
> operation? Just check the bit and clear it later in free_page_mlocked()?
That was initially done, but free_pages_check() checks for that flag
and did bad_page() on those mlocked ones. Now, one idea was to not
check the mlocked flag at all in free_pages_check() as we handle it
differently anyway. But I think we might still want to check for it
in tail-pages of higher order blocks.
And if you move that warning after free_pages_check(), the interesting
bits for the warning have been wiped already.
But we can get rid of the locked test-and-clear despite all the other
issues, patch below.
Hannes
>From eee677ddea61b1331a3bd8e402a0d02437fe872a Mon Sep 17 00:00:00 2001
From: Johannes Weiner <hannes@cmpxchg.org>
Date: Wed, 15 Jul 2009 23:40:28 +0200
Subject: [patch] mm: non-atomic test-clear of PG_mlocked on free
By the time PG_mlocked is cleared in the page freeing path, nobody
else is looking at our page->flags anymore.
It is thus safe to make the test-and-clear non-atomic and thereby
removing an unnecessary and expensive operation from a hotpath.
Signed-off-by: Johannes Weiner <hannes@cmpxchg.org>
---
include/linux/page-flags.h | 12 +++++++++---
mm/page_alloc.c | 4 ++--
2 files changed, 11 insertions(+), 5 deletions(-)
diff --git a/include/linux/page-flags.h b/include/linux/page-flags.h
index e2e5ce5..10e6011 100644
--- a/include/linux/page-flags.h
+++ b/include/linux/page-flags.h
@@ -158,6 +158,9 @@ static inline int TestSetPage##uname(struct page *page) \
static inline int TestClearPage##uname(struct page *page) \
{ return test_and_clear_bit(PG_##lname, &page->flags); }
+#define __TESTCLEARFLAG(uname, lname) \
+static inline int __TestClearPage##uname(struct page *page) \
+ { return __test_and_clear_bit(PG_##lname, &page->flags); }
#define PAGEFLAG(uname, lname) TESTPAGEFLAG(uname, lname) \
SETPAGEFLAG(uname, lname) CLEARPAGEFLAG(uname, lname)
@@ -184,6 +187,9 @@ static inline void __ClearPage##uname(struct page *page) { }
#define TESTCLEARFLAG_FALSE(uname) \
static inline int TestClearPage##uname(struct page *page) { return 0; }
+#define __TESTCLEARFLAG_FALSE(uname) \
+static inline int __TestClearPage##uname(struct page *page) { return 0; }
+
struct page; /* forward declaration */
TESTPAGEFLAG(Locked, locked) TESTSETFLAG(Locked, locked)
@@ -250,11 +256,11 @@ PAGEFLAG(Unevictable, unevictable) __CLEARPAGEFLAG(Unevictable, unevictable)
#ifdef CONFIG_HAVE_MLOCKED_PAGE_BIT
#define MLOCK_PAGES 1
PAGEFLAG(Mlocked, mlocked) __CLEARPAGEFLAG(Mlocked, mlocked)
- TESTSCFLAG(Mlocked, mlocked)
+ TESTSCFLAG(Mlocked, mlocked) __TESTCLEARFLAG(Mlocked, mlocked)
#else
#define MLOCK_PAGES 0
-PAGEFLAG_FALSE(Mlocked)
- SETPAGEFLAG_NOOP(Mlocked) TESTCLEARFLAG_FALSE(Mlocked)
+PAGEFLAG_FALSE(Mlocked) SETPAGEFLAG_NOOP(Mlocked)
+ TESTCLEARFLAG_FALSE(Mlocked) __TESTCLEARFLAG_FALSE(Mlocked)
#endif
#ifdef CONFIG_IA64_UNCACHED_ALLOCATOR
diff --git a/mm/page_alloc.c b/mm/page_alloc.c
index caa9268..b0c8758 100644
--- a/mm/page_alloc.c
+++ b/mm/page_alloc.c
@@ -557,7 +557,7 @@ static void __free_pages_ok(struct page *page, unsigned int order)
unsigned long flags;
int i;
int bad = 0;
- int wasMlocked = TestClearPageMlocked(page);
+ int wasMlocked = __TestClearPageMlocked(page);
kmemcheck_free_shadow(page, order);
@@ -1021,7 +1021,7 @@ static void free_hot_cold_page(struct page *page, int cold)
struct zone *zone = page_zone(page);
struct per_cpu_pages *pcp;
unsigned long flags;
- int wasMlocked = TestClearPageMlocked(page);
+ int wasMlocked = __TestClearPageMlocked(page);
kmemcheck_free_shadow(page, 0);
--
1.6.3
WARNING: multiple messages have this Message-ID (diff)
From: Johannes Weiner <hannes@cmpxchg.org>
To: Christoph Lameter <cl@linux-foundation.org>
Cc: Mel Gorman <mel@csn.ul.ie>,
Andrew Morton <akpm@linux-foundation.org>,
kosaki.motohiro@jp.fujitsu.com,
Maxim Levitsky <maximlevitsky@gmail.com>,
linux-kernel@vger.kernel.org, linux-mm@kvack.org,
Lee Schermerhorn <Lee.Schermerhorn@hp.com>,
Pekka Enberg <penberg@cs.helsinki.fi>,
Jiri Slaby <jirislaby@gmail.com>
Subject: Re: [PATCH] mm: Warn once when a page is freed with PG_mlocked set V2
Date: Thu, 16 Jul 2009 00:04:45 +0200 [thread overview]
Message-ID: <20090715220445.GA1823@cmpxchg.org> (raw)
In-Reply-To: <alpine.DEB.1.10.0907151027410.23643@gentwo.org>
Hello Christoph,
On Wed, Jul 15, 2009 at 10:31:54AM -0400, Christoph Lameter wrote:
> On Wed, 15 Jul 2009, Mel Gorman wrote:
>
> > -static inline int free_pages_check(struct page *page)
> > +static inline int free_pages_check(struct page *page, int wasMlocked)
> > {
> > + WARN_ONCE(wasMlocked, KERN_WARNING
> > + "Page flag mlocked set for process %s at pfn:%05lx\n"
> > + "page:%p flags:0x%lX\n",
> > + current->comm, page_to_pfn(page),
> > + page, page->flags|__PG_MLOCKED);
> > +
> > if (unlikely(page_mapcount(page) |
>
> There is already a free_page_mlocked() that is only called if the mlock
> bit is set. Move it into there to avoid having to run two checks in the
> hot codee path?
>
> Also __free_pages_ok() now has a TestClearMlocked in the hot code path.
> Would it be possible to get rid of the unconditional use of an atomic
> operation? Just check the bit and clear it later in free_page_mlocked()?
That was initially done, but free_pages_check() checks for that flag
and did bad_page() on those mlocked ones. Now, one idea was to not
check the mlocked flag at all in free_pages_check() as we handle it
differently anyway. But I think we might still want to check for it
in tail-pages of higher order blocks.
And if you move that warning after free_pages_check(), the interesting
bits for the warning have been wiped already.
But we can get rid of the locked test-and-clear despite all the other
issues, patch below.
Hannes
next prev parent reply other threads:[~2009-07-15 22:05 UTC|newest]
Thread overview: 28+ messages / expand[flat|nested] mbox.gz Atom feed top
2009-07-15 12:58 [PATCH] mm: Warn once when a page is freed with PG_mlocked set V2 Mel Gorman
2009-07-15 12:58 ` Mel Gorman
2009-07-15 14:31 ` Christoph Lameter
2009-07-15 14:31 ` Christoph Lameter
2009-07-15 22:04 ` Johannes Weiner [this message]
2009-07-15 22:04 ` Johannes Weiner
2009-07-16 7:37 ` KOSAKI Motohiro
2009-07-16 7:37 ` KOSAKI Motohiro
2009-07-16 13:54 ` Christoph Lameter
2009-07-16 13:54 ` Christoph Lameter
2009-07-16 16:01 ` Johannes Weiner
2009-07-16 16:01 ` Johannes Weiner
2009-07-17 0:17 ` KOSAKI Motohiro
2009-07-17 0:17 ` KOSAKI Motohiro
2009-07-22 23:06 ` Andrew Morton
2009-07-22 23:06 ` Andrew Morton
2009-07-23 10:29 ` Mel Gorman
2009-07-23 10:29 ` Mel Gorman
2009-07-23 17:23 ` Andrew Morton
2009-07-23 17:23 ` Andrew Morton
2009-07-24 10:36 ` Mel Gorman
2009-07-24 10:36 ` Mel Gorman
2009-07-24 11:31 ` Christoph Lameter
2009-07-24 11:31 ` Christoph Lameter
2009-07-24 12:00 ` Johannes Weiner
2009-07-24 12:00 ` Johannes Weiner
2009-07-24 12:59 ` Mel Gorman
2009-07-24 12:59 ` 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=20090715220445.GA1823@cmpxchg.org \
--to=hannes@cmpxchg.org \
--cc=Lee.Schermerhorn@hp.com \
--cc=akpm@linux-foundation.org \
--cc=cl@linux-foundation.org \
--cc=jirislaby@gmail.com \
--cc=kosaki.motohiro@jp.fujitsu.com \
--cc=linux-kernel@vger.kernel.org \
--cc=linux-mm@kvack.org \
--cc=maximlevitsky@gmail.com \
--cc=mel@csn.ul.ie \
--cc=penberg@cs.helsinki.fi \
/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 an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.