[PATCH 1/3] memcg: mark init_section_page_cgroup() properly

[PATCH 1/3] memcg: mark init_section_page_cgroup() properly

[Namhyung Kim]

The commit ca371c0d7e23 ("memcg: fix page_cgroup fatal error
in FLATMEM") removes call to alloc_bootmem() in the function
so that it can be marked as __meminit to reduce memory usage
when MEMORY_HOTPLUG=n.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
---
 mm/page_cgroup.c |    3 +--
 1 files changed, 1 insertions(+), 2 deletions(-)

diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 5bffada7cde1..2d1a0fa01d7b 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -105,8 +105,7 @@ struct page_cgroup *lookup_page_cgroup(struct page *page)
 	return section->page_cgroup + pfn;
 }
 
-/* __alloc_bootmem...() is protected by !slab_available() */
-static int __init_refok init_section_page_cgroup(unsigned long pfn)
+static int __meminit init_section_page_cgroup(unsigned long pfn)
 {
 	struct mem_section *section = __pfn_to_section(pfn);
 	struct page_cgroup *base, *pc;
-- 
1.7.4

[PATCH 2/3] memcg: fix off-by-one when calculating swap cgroup map length

[Namhyung Kim]

It allocated one more page than necessary if @max_pages was
a multiple of SC_PER_PAGE.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
---
 mm/page_cgroup.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 2d1a0fa01d7b..29951abc852e 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -446,7 +446,7 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
 	if (!do_swap_account)
 		return 0;
 
-	length = ((max_pages/SC_PER_PAGE) + 1);
+	length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
 	array_size = length * sizeof(void *);
 
 	array = vmalloc(array_size);
-- 
1.7.4

[PATCH 3/3] memcg: move page-freeing code outside of lock

[Namhyung Kim]

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
---
 mm/page_cgroup.c |   22 +++++++++++++---------
 1 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 29951abc852e..17eb5eb95bab 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -463,8 +463,8 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
 		/* memory shortage */
 		ctrl->map = NULL;
 		ctrl->length = 0;
-		vfree(array);
 		mutex_unlock(&swap_cgroup_mutex);
+		vfree(array);
 		goto nomem;
 	}
 	mutex_unlock(&swap_cgroup_mutex);
@@ -479,7 +479,8 @@ nomem:
 
 void swap_cgroup_swapoff(int type)
 {
-	int i;
+	struct page **map;
+	unsigned long i, length;
 	struct swap_cgroup_ctrl *ctrl;
 
 	if (!do_swap_account)
@@ -487,17 +488,20 @@ void swap_cgroup_swapoff(int type)
 
 	mutex_lock(&swap_cgroup_mutex);
 	ctrl = &swap_cgroup_ctrl[type];
-	if (ctrl->map) {
-		for (i = 0; i < ctrl->length; i++) {
-			struct page *page = ctrl->map[i];
+	map = ctrl->map;
+	length = ctrl->length;
+	ctrl->map = NULL;
+	ctrl->length = 0;
+	mutex_unlock(&swap_cgroup_mutex);
+
+	if (map) {
+		for (i = 0; i < length; i++) {
+			struct page *page = map[i];
 			if (page)
 				__free_page(page);
 		}
-		vfree(ctrl->map);
-		ctrl->map = NULL;
-		ctrl->length = 0;
+		vfree(map);
 	}
-	mutex_unlock(&swap_cgroup_mutex);
 }
 
 #endif
-- 
1.7.4

Re: [PATCH 1/3] memcg: mark init_section_page_cgroup() properly

[KAMEZAWA Hiroyuki]

On Fri, 18 Mar 2011 21:54:13 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> The commit ca371c0d7e23 ("memcg: fix page_cgroup fatal error
> in FLATMEM") removes call to alloc_bootmem() in the function
> so that it can be marked as __meminit to reduce memory usage
> when MEMORY_HOTPLUG=n.
> 
> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>


> ---
>  mm/page_cgroup.c |    3 +--
>  1 files changed, 1 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
> index 5bffada7cde1..2d1a0fa01d7b 100644
> --- a/mm/page_cgroup.c
> +++ b/mm/page_cgroup.c
> @@ -105,8 +105,7 @@ struct page_cgroup *lookup_page_cgroup(struct page *page)
>  	return section->page_cgroup + pfn;
>  }
>  
> -/* __alloc_bootmem...() is protected by !slab_available() */
> -static int __init_refok init_section_page_cgroup(unsigned long pfn)
> +static int __meminit init_section_page_cgroup(unsigned long pfn)
>  {
>  	struct mem_section *section = __pfn_to_section(pfn);
>  	struct page_cgroup *base, *pc;
> -- 
> 1.7.4
> 
> 

Re: [PATCH 2/3] memcg: fix off-by-one when calculating swap cgroup map length

[KAMEZAWA Hiroyuki]

On Fri, 18 Mar 2011 21:54:14 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> It allocated one more page than necessary if @max_pages was
> a multiple of SC_PER_PAGE.
> 
> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[KAMEZAWA Hiroyuki]

On Fri, 18 Mar 2011 21:54:15 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

What is the benefit of this patch ?

-Kame

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[Namhyung Kim]

2011-03-22 (화), 08:59 +0900, KAMEZAWA Hiroyuki:
> On Fri, 18 Mar 2011 21:54:15 +0900
> Namhyung Kim <namhyung@gmail.com> wrote:
> 
> > Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> > Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> 
> What is the benefit of this patch ?
> 
> -Kame
> 

Oh, I just thought generally it'd better call such a (potentially)
costly function outside of locks and it could reduce few of theoretical
contentions between swapons and/or offs. If it doesn't help any
realistic cases I don't mind discarding it.

Thanks.


-- 
Regards,
Namhyung Kim

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[Balbir Singh]

On Tue, Mar 22, 2011 at 8:34 AM, Namhyung Kim <namhyung@gmail.com> wrote:
>
> 2011-03-22 (화), 08:59 +0900, KAMEZAWA Hiroyuki:
> > On Fri, 18 Mar 2011 21:54:15 +0900
> > Namhyung Kim <namhyung@gmail.com> wrote:
> >
> > > Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> > > Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> >
> > What is the benefit of this patch ?
> >
> > -Kame
> >
>
> Oh, I just thought generally it'd better call such a (potentially)
> costly function outside of locks and it could reduce few of theoretical
> contentions between swapons and/or offs. If it doesn't help any
> realistic cases I don't mind discarding it.

swapoff is a rare path, I would not worry about it too much at all.

Balbir

Re: [PATCH 1/3] memcg: mark init_section_page_cgroup() properly

[Balbir Singh]

On Tue, Mar 22, 2011 at 5:27 AM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Fri, 18 Mar 2011 21:54:13 +0900
> Namhyung Kim <namhyung@gmail.com> wrote:
>
>> The commit ca371c0d7e23 ("memcg: fix page_cgroup fatal error
>> in FLATMEM") removes call to alloc_bootmem() in the function
>> so that it can be marked as __meminit to reduce memory usage
>> when MEMORY_HOTPLUG=n.
>>
>> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
>> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>

Re: [PATCH 2/3] memcg: fix off-by-one when calculating swap cgroup map length

[Balbir Singh]

On Tue, Mar 22, 2011 at 5:28 AM, KAMEZAWA Hiroyuki
<kamezawa.hiroyu@jp.fujitsu.com> wrote:
> On Fri, 18 Mar 2011 21:54:14 +0900
> Namhyung Kim <namhyung@gmail.com> wrote:
>
>> It allocated one more page than necessary if @max_pages was
>> a multiple of SC_PER_PAGE.
>>
>> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
>> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
>
> Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

Acked-by: Balbir Singh <balbir@linux.vnet.ibm.com>

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[KAMEZAWA Hiroyuki]

On Tue, 22 Mar 2011 12:04:39 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> 2011-03-22 (화), 08:59 +0900, KAMEZAWA Hiroyuki:
> > On Fri, 18 Mar 2011 21:54:15 +0900
> > Namhyung Kim <namhyung@gmail.com> wrote:
> > 
> > > Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> > > Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > 
> > What is the benefit of this patch ?
> > 
> > -Kame
> > 
> 
> Oh, I just thought generally it'd better call such a (potentially)
> costly function outside of locks and it could reduce few of theoretical
> contentions between swapons and/or offs. If it doesn't help any
> realistic cases I don't mind discarding it.
> 

My point is, please write patch description which shows for what this patc is.
All cleanup are okay to me if it reasonable. But without patch description as
"this is just a cleanup, no functional change, and the reason is...."
we cannot maintain patches.

Thanks,
-Kame

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[Namhyung Kim]

2011-03-22 (화), 13:56 +0900, KAMEZAWA Hiroyuki:
> On Tue, 22 Mar 2011 12:04:39 +0900
> Namhyung Kim <namhyung@gmail.com> wrote:
> 
> > 2011-03-22 (화), 08:59 +0900, KAMEZAWA Hiroyuki:
> > > On Fri, 18 Mar 2011 21:54:15 +0900
> > > Namhyung Kim <namhyung@gmail.com> wrote:
> > > 
> > > > Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> > > > Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > > 
> > > What is the benefit of this patch ?
> > > 
> > > -Kame
> > > 
> > 
> > Oh, I just thought generally it'd better call such a (potentially)
> > costly function outside of locks and it could reduce few of theoretical
> > contentions between swapons and/or offs. If it doesn't help any
> > realistic cases I don't mind discarding it.
> > 
> 
> My point is, please write patch description which shows for what this patc is.
> All cleanup are okay to me if it reasonable. But without patch description as
> "this is just a cleanup, no functional change, and the reason is...."
> we cannot maintain patches.
> 
> Thanks,
> -Kame
> 

OK, I will do that in the future. Anyway, do you want me to resend the
patch with new description?

Thanks.


-- 
Regards,
Namhyung Kim

Re: [PATCH 3/3] memcg: move page-freeing code outside of lock

[KAMEZAWA Hiroyuki]

On Tue, 22 Mar 2011 19:06:57 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> 2011-03-22 (화), 13:56 +0900, KAMEZAWA Hiroyuki:
> > On Tue, 22 Mar 2011 12:04:39 +0900
> > Namhyung Kim <namhyung@gmail.com> wrote:
> > 
> > > 2011-03-22 (화), 08:59 +0900, KAMEZAWA Hiroyuki:
> > > > On Fri, 18 Mar 2011 21:54:15 +0900
> > > > Namhyung Kim <namhyung@gmail.com> wrote:
> > > > 
> > > > > Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> > > > > Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> > > > 
> > > > What is the benefit of this patch ?
> > > > 
> > > > -Kame
> > > > 
> > > 
> > > Oh, I just thought generally it'd better call such a (potentially)
> > > costly function outside of locks and it could reduce few of theoretical
> > > contentions between swapons and/or offs. If it doesn't help any
> > > realistic cases I don't mind discarding it.
> > > 
> > 
> > My point is, please write patch description which shows for what this patc is.
> > All cleanup are okay to me if it reasonable. But without patch description as
> > "this is just a cleanup, no functional change, and the reason is...."
> > we cannot maintain patches.
> > 
> > Thanks,
> > -Kame
> > 
> 
> OK, I will do that in the future. Anyway, do you want me to resend the
> patch with new description?
> 

please. I'll never ack a patch without description.

Thanks,
-Kame

[PATCH] memcg: move page-freeing code out of lock

[Namhyung Kim]

Move page-freeing code out of swap_cgroup_mutex in the hope that it
could reduce few of theoretical contentions between swapons and/or
swapoffs.

This is just a cleanup, no functional changes.

Signed-off-by: Namhyung Kim <namhyung@gmail.com>
Cc: Paul Menage <menage@google.com>
Cc: Li Zefan <lizf@cn.fujitsu.com>
Cc: containers@lists.linux-foundation.org
---
 mm/page_cgroup.c |   22 +++++++++++++---------
 1 files changed, 13 insertions(+), 9 deletions(-)

diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 29951abc852e..17eb5eb95bab 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -463,8 +463,8 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
 		/* memory shortage */
 		ctrl->map = NULL;
 		ctrl->length = 0;
-		vfree(array);
 		mutex_unlock(&swap_cgroup_mutex);
+		vfree(array);
 		goto nomem;
 	}
 	mutex_unlock(&swap_cgroup_mutex);
@@ -479,7 +479,8 @@ nomem:
 
 void swap_cgroup_swapoff(int type)
 {
-	int i;
+	struct page **map;
+	unsigned long i, length;
 	struct swap_cgroup_ctrl *ctrl;
 
 	if (!do_swap_account)
@@ -487,17 +488,20 @@ void swap_cgroup_swapoff(int type)
 
 	mutex_lock(&swap_cgroup_mutex);
 	ctrl = &swap_cgroup_ctrl[type];
-	if (ctrl->map) {
-		for (i = 0; i < ctrl->length; i++) {
-			struct page *page = ctrl->map[i];
+	map = ctrl->map;
+	length = ctrl->length;
+	ctrl->map = NULL;
+	ctrl->length = 0;
+	mutex_unlock(&swap_cgroup_mutex);
+
+	if (map) {
+		for (i = 0; i < length; i++) {
+			struct page *page = map[i];
 			if (page)
 				__free_page(page);
 		}
-		vfree(ctrl->map);
-		ctrl->map = NULL;
-		ctrl->length = 0;
+		vfree(map);
 	}
-	mutex_unlock(&swap_cgroup_mutex);
 }
 
 #endif
-- 
1.7.4

Re: [PATCH] memcg: move page-freeing code out of lock

[KAMEZAWA Hiroyuki]

On Wed, 23 Mar 2011 20:59:18 +0900
Namhyung Kim <namhyung@gmail.com> wrote:

> Move page-freeing code out of swap_cgroup_mutex in the hope that it
> could reduce few of theoretical contentions between swapons and/or
> swapoffs.
> 
> This is just a cleanup, no functional changes.
> 
> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> Cc: Paul Menage <menage@google.com>
> Cc: Li Zefan <lizf@cn.fujitsu.com>
> Cc: containers@lists.linux-foundation.org

Acked-by: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>

Re: [PATCH 1/3] memcg: mark init_section_page_cgroup() properly

[Namhyung Kim]

2011-03-18 (금), 21:54 +0900, Namhyung Kim:
> The commit ca371c0d7e23 ("memcg: fix page_cgroup fatal error
> in FLATMEM") removes call to alloc_bootmem() in the function
> so that it can be marked as __meminit to reduce memory usage
> when MEMORY_HOTPLUG=n.
> 
> Signed-off-by: Namhyung Kim <namhyung@gmail.com>
> Cc: KAMEZAWA Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> ---
>  mm/page_cgroup.c |    3 +--
>  1 files changed, 1 insertions(+), 2 deletions(-)
> 
> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
> index 5bffada7cde1..2d1a0fa01d7b 100644
> --- a/mm/page_cgroup.c
> +++ b/mm/page_cgroup.c
> @@ -105,8 +105,7 @@ struct page_cgroup *lookup_page_cgroup(struct page *page)
>  	return section->page_cgroup + pfn;
>  }
>  
> -/* __alloc_bootmem...() is protected by !slab_available() */
> -static int __init_refok init_section_page_cgroup(unsigned long pfn)
> +static int __meminit init_section_page_cgroup(unsigned long pfn)
>  {
>  	struct mem_section *section = __pfn_to_section(pfn);
>  	struct page_cgroup *base, *pc;

Andrew, could you please have a look these patches too and consider
applying them in your tree? I can resend them (with given Acked-by
lines) if you want.

Thanks.


-- 
Regards,
Namhyung Kim