* [PATCH v2 1/6] memcg: refactor swap_cgroup_swapon()
@ 2013-01-28 10:54 ` Jeff Liu
[not found] ` <510658E6.9030108-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 1 reply; 26+ messages in thread
From: Jeff Liu @ 2013-01-28 10:54 UTC (permalink / raw)
To: linux-mm; +Cc: Michal Hocko, Glauber Costa, cgroups
Refector swap_cgroup_swapon() to setup the number of pages only, and
move the rest to swap_cgroup_prepare(), so that the later can be used
for allocating buffers when creating the first non-root memcg.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Glauber Costa <glommer@parallels.com>
CC: Michal Hocko <mhocko@suse.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
CC: Johannes Weiner <hannes@cmpxchg.org>
CC: Mel Gorman <mgorman@suse.de>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Sha Zhengju <handai.szj@taobao.com>
---
mm/page_cgroup.c | 17 ++++++-----------
1 file changed, 6 insertions(+), 11 deletions(-)
diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 6d757e3..c945254 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -360,6 +360,9 @@ static int swap_cgroup_prepare(int type)
unsigned long idx, max;
ctrl = &swap_cgroup_ctrl[type];
+ ctrl->map = vzalloc(ctrl->length * sizeof(void *));
+ if (!ctrl->map)
+ goto nomem;
for (idx = 0; idx < ctrl->length; idx++) {
page = alloc_page(GFP_KERNEL | __GFP_ZERO);
@@ -368,11 +371,13 @@ static int swap_cgroup_prepare(int type)
ctrl->map[idx] = page;
}
return 0;
+
not_enough_page:
max = idx;
for (idx = 0; idx < max; idx++)
__free_page(ctrl->map[idx]);
-
+ ctrl->map = NULL;
+nomem:
return -ENOMEM;
}
@@ -460,8 +465,6 @@ unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
int swap_cgroup_swapon(int type, unsigned long max_pages)
{
- void *array;
- unsigned long array_size;
unsigned long length;
struct swap_cgroup_ctrl *ctrl;
@@ -469,23 +472,15 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
return 0;
length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
- array_size = length * sizeof(void *);
-
- array = vzalloc(array_size);
- if (!array)
- goto nomem;
ctrl = &swap_cgroup_ctrl[type];
mutex_lock(&swap_cgroup_mutex);
ctrl->length = length;
- ctrl->map = array;
spin_lock_init(&ctrl->lock);
if (swap_cgroup_prepare(type)) {
/* memory shortage */
- ctrl->map = NULL;
ctrl->length = 0;
mutex_unlock(&swap_cgroup_mutex);
- vfree(array);
goto nomem;
}
mutex_unlock(&swap_cgroup_mutex);
--
1.7.9.5
--
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>
^ permalink raw reply related [flat|nested] 26+ messages in thread* [PATCH v2 3/6] memcg: introduce memsw_accounting_users
@ 2013-01-28 10:54 ` Jeff Liu
[not found] ` <510658F0.9050802-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 1 reply; 26+ messages in thread
From: Jeff Liu @ 2013-01-28 10:54 UTC (permalink / raw)
To: linux-mm; +Cc: Michal Hocko, Glauber Costa, cgroups
As we don't account the swap stat number for the root_mem_cgroup anymore,
here we can just return an invalid CSS ID if there is no non-root memcg
is alive. Also, introduce memsw_accounting_users to track the number of
active non-root memcgs.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Glauber Costa <glommer@parallels.com>
CC: Michal Hocko <mhocko@suse.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
CC: Johannes Weiner <hannes@cmpxchg.org>
CC: Mel Gorman <mgorman@suse.de>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Sha Zhengju <handai.szj@taobao.com>
---
mm/page_cgroup.c | 16 +++++++++++++++-
1 file changed, 15 insertions(+), 1 deletion(-)
diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index c945254..189fbf5 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -336,6 +336,8 @@ struct swap_cgroup {
};
#define SC_PER_PAGE (PAGE_SIZE/sizeof(struct swap_cgroup))
+static atomic_t memsw_accounting_users = ATOMIC_INIT(0);
+
/*
* SwapCgroup implements "lookup" and "exchange" operations.
* In typical usage, this swap_cgroup is accessed via memcg's charge/uncharge
@@ -389,6 +391,9 @@ static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
struct page *mappage;
struct swap_cgroup *sc;
+ if (!atomic_read(&memsw_accounting_users))
+ return NULL;
+
ctrl = &swap_cgroup_ctrl[swp_type(ent)];
if (ctrlp)
*ctrlp = ctrl;
@@ -416,6 +421,8 @@ unsigned short swap_cgroup_cmpxchg(swp_entry_t ent,
unsigned short retval;
sc = lookup_swap_cgroup(ent, &ctrl);
+ if (!sc)
+ return 0;
spin_lock_irqsave(&ctrl->lock, flags);
retval = sc->id;
@@ -443,6 +450,8 @@ unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id)
unsigned long flags;
sc = lookup_swap_cgroup(ent, &ctrl);
+ if (!sc)
+ return 0;
spin_lock_irqsave(&ctrl->lock, flags);
old = sc->id;
@@ -460,7 +469,9 @@ unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id)
*/
unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
{
- return lookup_swap_cgroup(ent, NULL)->id;
+ struct swap_cgroup *sc = lookup_swap_cgroup(ent, NULL);
+
+ return sc ? sc->id : 0;
}
int swap_cgroup_swapon(int type, unsigned long max_pages)
@@ -471,6 +482,9 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
if (!do_swap_account)
return 0;
+ if (!atomic_read(&memsw_accounting_users))
+ return 0;
+
length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
ctrl = &swap_cgroup_ctrl[type];
--
1.7.9.5
--
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>
^ permalink raw reply related [flat|nested] 26+ messages in thread* [PATCH v2 4/6] memcg: export nr_swap_files
@ 2013-01-28 10:54 ` Jeff Liu
[not found] ` <510658F6.4010504-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 1 reply; 26+ messages in thread
From: Jeff Liu @ 2013-01-28 10:54 UTC (permalink / raw)
To: linux-mm; +Cc: Michal Hocko, Glauber Costa, cgroups
Export nr_swap_files which would be used for initializing and destorying
swap cgroup structures in the coming patch.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Glauber Costa <glommer@parallels.com>
CC: Michal Hocko <mhocko@suse.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
CC: Johannes Weiner <hannes@cmpxchg.org>
CC: Mel Gorman <mgorman@suse.de>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Sha Zhengju <handai.szj@taobao.com>
---
include/linux/swap.h | 1 +
mm/swapfile.c | 2 +-
2 files changed, 2 insertions(+), 1 deletion(-)
diff --git a/include/linux/swap.h b/include/linux/swap.h
index 68df9c1..6de44c9 100644
--- a/include/linux/swap.h
+++ b/include/linux/swap.h
@@ -348,6 +348,7 @@ extern struct page *swapin_readahead(swp_entry_t, gfp_t,
/* linux/mm/swapfile.c */
extern long nr_swap_pages;
extern long total_swap_pages;
+extern unsigned int nr_swapfiles;
extern void si_swapinfo(struct sysinfo *);
extern swp_entry_t get_swap_page(void);
extern swp_entry_t get_swap_page_of_type(int);
diff --git a/mm/swapfile.c b/mm/swapfile.c
index e97a0e5..89cebcf 100644
--- a/mm/swapfile.c
+++ b/mm/swapfile.c
@@ -46,7 +46,7 @@ static void free_swap_count_continuations(struct swap_info_struct *);
static sector_t map_swap_entry(swp_entry_t, struct block_device**);
DEFINE_SPINLOCK(swap_lock);
-static unsigned int nr_swapfiles;
+unsigned int nr_swapfiles;
long nr_swap_pages;
long total_swap_pages;
static int least_priority;
--
1.7.9.5
--
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>
^ permalink raw reply related [flat|nested] 26+ messages in thread* [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free()
@ 2013-01-28 10:54 ` Jeff Liu
2013-01-29 9:57 ` Lord Glauber Costa of Sealand
[not found] ` <510658F7.6050806-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 2 replies; 26+ messages in thread
From: Jeff Liu @ 2013-01-28 10:54 UTC (permalink / raw)
To: linux-mm; +Cc: Michal Hocko, Glauber Costa, cgroups
Introduce swap_cgroup_init()/swap_cgroup_free() to allocate buffers when creating the first
non-root memcg and deallocate buffers on the last non-root memcg is gone.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Glauber Costa <glommer@parallels.com>
CC: Michal Hocko <mhocko@suse.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
CC: Johannes Weiner <hannes@cmpxchg.org>
CC: Mel Gorman <mgorman@suse.de>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Sha Zhengju <handai.szj@taobao.com>
---
include/linux/page_cgroup.h | 12 +++++
mm/page_cgroup.c | 108 +++++++++++++++++++++++++++++++++++++++----
2 files changed, 110 insertions(+), 10 deletions(-)
diff --git a/include/linux/page_cgroup.h b/include/linux/page_cgroup.h
index 777a524..1255cc9 100644
--- a/include/linux/page_cgroup.h
+++ b/include/linux/page_cgroup.h
@@ -113,6 +113,8 @@ extern unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id);
extern unsigned short lookup_swap_cgroup_id(swp_entry_t ent);
extern int swap_cgroup_swapon(int type, unsigned long max_pages);
extern void swap_cgroup_swapoff(int type);
+extern int swap_cgroup_init(void);
+extern void swap_cgroup_free(void);
#else
static inline
@@ -138,6 +140,16 @@ static inline void swap_cgroup_swapoff(int type)
return;
}
+static inline int swap_cgroup_init(void)
+{
+ return 0;
+}
+
+static inline void swap_cgroup_free(void)
+{
+ return;
+}
+
#endif /* CONFIG_MEMCG_SWAP */
#endif /* !__GENERATING_BOUNDS_H */
diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
index 189fbf5..0ebd127 100644
--- a/mm/page_cgroup.c
+++ b/mm/page_cgroup.c
@@ -362,14 +362,28 @@ static int swap_cgroup_prepare(int type)
unsigned long idx, max;
ctrl = &swap_cgroup_ctrl[type];
+ if (!ctrl->length) {
+ /*
+ * Bypass the buffer allocation if the corresponding swap
+ * partition/file was turned off.
+ */
+ pr_debug("couldn't allocate swap_cgroup on a disabled swap "
+ "partition or file, index: %d\n", type);
+ return 0;
+ }
+
ctrl->map = vzalloc(ctrl->length * sizeof(void *));
- if (!ctrl->map)
+ if (!ctrl->map) {
+ ctrl->length = 0;
goto nomem;
+ }
for (idx = 0; idx < ctrl->length; idx++) {
page = alloc_page(GFP_KERNEL | __GFP_ZERO);
- if (!page)
+ if (!page) {
+ ctrl->length = 0;
goto not_enough_page;
+ }
ctrl->map[idx] = page;
}
return 0;
@@ -383,6 +397,32 @@ nomem:
return -ENOMEM;
}
+/*
+ * free buffer for swap_cgroup.
+ */
+static void swap_cgroup_teardown(int type)
+{
+ struct page **map;
+ unsigned long length;
+ struct swap_cgroup_ctrl *ctrl;
+
+ ctrl = &swap_cgroup_ctrl[type];
+ map = ctrl->map;
+ length = ctrl->length;
+ ctrl->map = NULL;
+ ctrl->length = 0;
+
+ if (map) {
+ unsigned long i;
+ for (i = 0; i < length; i++) {
+ struct page *page = map[i];
+ if (page)
+ __free_page(page);
+ }
+ vfree(map);
+ }
+}
+
static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
struct swap_cgroup_ctrl **ctrlp)
{
@@ -474,6 +514,56 @@ unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
return sc ? sc->id : 0;
}
+/*
+ * Allocate swap cgroup accounting structures when the first non-root
+ * memcg is created.
+ */
+int swap_cgroup_init(void)
+{
+ unsigned int type;
+
+ if (!do_swap_account)
+ return 0;
+
+ if (atomic_add_return(1, &memsw_accounting_users) != 1)
+ return 0;
+
+ mutex_lock(&swap_cgroup_mutex);
+ for (type = 0; type < nr_swapfiles; type++) {
+ if (swap_cgroup_prepare(type) < 0) {
+ mutex_unlock(&swap_cgroup_mutex);
+ goto nomem;
+ }
+ }
+ mutex_unlock(&swap_cgroup_mutex);
+ return 0;
+
+nomem:
+ pr_info("couldn't allocate enough memory for swap_cgroup "
+ "while creating non-root memcg.\n");
+ return -ENOMEM;
+}
+
+/*
+ * Deallocate swap cgroup accounting structures on the last non-root
+ * memcg removal.
+ */
+void swap_cgroup_free(void)
+{
+ unsigned int type;
+
+ if (!do_swap_account)
+ return;
+
+ if (atomic_sub_return(1, &memsw_accounting_users))
+ return;
+
+ mutex_lock(&swap_cgroup_mutex);
+ for (type = 0; type < nr_swapfiles; type++)
+ swap_cgroup_teardown(type);
+ mutex_unlock(&swap_cgroup_mutex);
+}
+
int swap_cgroup_swapon(int type, unsigned long max_pages)
{
unsigned long length;
@@ -482,20 +572,18 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
if (!do_swap_account)
return 0;
- if (!atomic_read(&memsw_accounting_users))
- return 0;
-
length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
ctrl = &swap_cgroup_ctrl[type];
mutex_lock(&swap_cgroup_mutex);
ctrl->length = length;
spin_lock_init(&ctrl->lock);
- if (swap_cgroup_prepare(type)) {
- /* memory shortage */
- ctrl->length = 0;
- mutex_unlock(&swap_cgroup_mutex);
- goto nomem;
+ if (atomic_read(&memsw_accounting_users)) {
+ if (swap_cgroup_prepare(type)) {
+ /* memory shortage */
+ mutex_unlock(&swap_cgroup_mutex);
+ goto nomem;
+ }
}
mutex_unlock(&swap_cgroup_mutex);
--
1.7.9.5
--
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>
^ permalink raw reply related [flat|nested] 26+ messages in thread* Re: [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free()
2013-01-28 10:54 ` [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free() Jeff Liu
@ 2013-01-29 9:57 ` Lord Glauber Costa of Sealand
2013-01-29 10:21 ` Jeff Liu
[not found] ` <510658F7.6050806-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
1 sibling, 1 reply; 26+ messages in thread
From: Lord Glauber Costa of Sealand @ 2013-01-29 9:57 UTC (permalink / raw)
To: Jeff Liu; +Cc: linux-mm, Michal Hocko, cgroups
On 01/28/2013 02:54 PM, Jeff Liu wrote:
> Introduce swap_cgroup_init()/swap_cgroup_free() to allocate buffers when creating the first
> non-root memcg and deallocate buffers on the last non-root memcg is gone.
>
> Signed-off-by: Jie Liu <jeff.liu@oracle.com>
> CC: Glauber Costa <glommer@parallels.com>
> CC: Michal Hocko <mhocko@suse.cz>
> CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
> CC: Johannes Weiner <hannes@cmpxchg.org>
> CC: Mel Gorman <mgorman@suse.de>
> CC: Andrew Morton <akpm@linux-foundation.org>
> CC: Sha Zhengju <handai.szj@taobao.com>
>
Looks sane.
Reviewed-by: Glauber Costa <glommer@parallels.com>
Only:
> #endif /* !__GENERATING_BOUNDS_H */
> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
> index 189fbf5..0ebd127 100644
> --- a/mm/page_cgroup.c
> +++ b/mm/page_cgroup.c
> @@ -362,14 +362,28 @@ static int swap_cgroup_prepare(int type)
> unsigned long idx, max;
>
> ctrl = &swap_cgroup_ctrl[type];
> + if (!ctrl->length) {
> + /*
> + * Bypass the buffer allocation if the corresponding swap
> + * partition/file was turned off.
> + */
> + pr_debug("couldn't allocate swap_cgroup on a disabled swap "
> + "partition or file, index: %d\n", type);
> + return 0;
> + }
> +
> ctrl->map = vzalloc(ctrl->length * sizeof(void *));
> - if (!ctrl->map)
> + if (!ctrl->map) {
> + ctrl->length = 0;
Considering moving this assignment somewhere in the exit path in the
labels region.
--
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>
^ permalink raw reply [flat|nested] 26+ messages in thread* Re: [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free()
2013-01-29 9:57 ` Lord Glauber Costa of Sealand
@ 2013-01-29 10:21 ` Jeff Liu
0 siblings, 0 replies; 26+ messages in thread
From: Jeff Liu @ 2013-01-29 10:21 UTC (permalink / raw)
To: Lord Glauber Costa of Sealand; +Cc: linux-mm, Michal Hocko, cgroups
On 01/29/2013 05:57 PM, Lord Glauber Costa of Sealand wrote:
> On 01/28/2013 02:54 PM, Jeff Liu wrote:
>> Introduce swap_cgroup_init()/swap_cgroup_free() to allocate buffers when creating the first
>> non-root memcg and deallocate buffers on the last non-root memcg is gone.
>>
>> Signed-off-by: Jie Liu <jeff.liu@oracle.com>
>> CC: Glauber Costa <glommer@parallels.com>
>> CC: Michal Hocko <mhocko@suse.cz>
>> CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
>> CC: Johannes Weiner <hannes@cmpxchg.org>
>> CC: Mel Gorman <mgorman@suse.de>
>> CC: Andrew Morton <akpm@linux-foundation.org>
>> CC: Sha Zhengju <handai.szj@taobao.com>
>>
>
> Looks sane.
>
> Reviewed-by: Glauber Costa <glommer@parallels.com>
>
> Only:
>
>> #endif /* !__GENERATING_BOUNDS_H */
>> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
>> index 189fbf5..0ebd127 100644
>> --- a/mm/page_cgroup.c
>> +++ b/mm/page_cgroup.c
>> @@ -362,14 +362,28 @@ static int swap_cgroup_prepare(int type)
>> unsigned long idx, max;
>>
>> ctrl = &swap_cgroup_ctrl[type];
>> + if (!ctrl->length) {
>> + /*
>> + * Bypass the buffer allocation if the corresponding swap
>> + * partition/file was turned off.
>> + */
>> + pr_debug("couldn't allocate swap_cgroup on a disabled swap "
>> + "partition or file, index: %d\n", type);
>> + return 0;
>> + }
>> +
>> ctrl->map = vzalloc(ctrl->length * sizeof(void *));
>> - if (!ctrl->map)
>> + if (!ctrl->map) {
>> + ctrl->length = 0;
>
> Considering moving this assignment somewhere in the exit path in the
> labels region.
Nice point. Both "ctrl->length = 0" statements in this function should be
moved down to the exit path of "nomem:" label.
Thanks,
-Jeff
--
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>
^ permalink raw reply [flat|nested] 26+ messages in thread
[parent not found: <510658F7.6050806-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>]
* Re: [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free()
[not found] ` <510658F7.6050806-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
@ 2013-01-29 14:56 ` Michal Hocko
2013-01-29 15:51 ` Jeff Liu
0 siblings, 1 reply; 26+ messages in thread
From: Michal Hocko @ 2013-01-29 14:56 UTC (permalink / raw)
To: Jeff Liu
Cc: linux-mm-Bw31MaZKKs3YtjvyW6yDsg, Glauber Costa,
cgroups-u79uwXL29TY76Z2rM5mHXA
On Mon 28-01-13 18:54:47, Jeff Liu wrote:
> Introduce swap_cgroup_init()/swap_cgroup_free() to allocate buffers when creating the first
> non-root memcg and deallocate buffers on the last non-root memcg is gone.
I think this deserves more words ;) At least it would be good to
describe contexts from which init and free might be called. What are the
locking rules.
Also swap_cgroup_destroy sounds more in pair with swap_cgroup_init.
Please add the users of those function here as well. It is much easier
to review.
> Signed-off-by: Jie Liu <jeff.liu-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
> CC: Glauber Costa <glommer-bzQdu9zFT3WakBO8gow8eQ@public.gmane.org>
> CC: Michal Hocko <mhocko-AlSwsSmVLrQ@public.gmane.org>
> CC: Kamezawa Hiroyuki <kamezawa.hiroyu-+CUm20s59erQFUHtdCDX3A@public.gmane.org>
> CC: Johannes Weiner <hannes-druUgvl0LCNAfugRpC6u6w@public.gmane.org>
> CC: Mel Gorman <mgorman-l3A5Bk7waGM@public.gmane.org>
> CC: Andrew Morton <akpm-de/tnXTf+JLsfHDXvbKv3WD2FQJk+8+b@public.gmane.org>
> CC: Sha Zhengju <handai.szj-3b8fjiQLQpfQT0dZR+AlfA@public.gmane.org>
>
> ---
> include/linux/page_cgroup.h | 12 +++++
> mm/page_cgroup.c | 108 +++++++++++++++++++++++++++++++++++++++----
> 2 files changed, 110 insertions(+), 10 deletions(-)
>
> diff --git a/include/linux/page_cgroup.h b/include/linux/page_cgroup.h
> index 777a524..1255cc9 100644
> --- a/include/linux/page_cgroup.h
> +++ b/include/linux/page_cgroup.h
> @@ -113,6 +113,8 @@ extern unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id);
> extern unsigned short lookup_swap_cgroup_id(swp_entry_t ent);
> extern int swap_cgroup_swapon(int type, unsigned long max_pages);
> extern void swap_cgroup_swapoff(int type);
> +extern int swap_cgroup_init(void);
> +extern void swap_cgroup_free(void);
> #else
>
> static inline
> @@ -138,6 +140,16 @@ static inline void swap_cgroup_swapoff(int type)
> return;
> }
>
> +static inline int swap_cgroup_init(void)
> +{
> + return 0;
> +}
> +
> +static inline void swap_cgroup_free(void)
> +{
> + return;
> +}
> +
> #endif /* CONFIG_MEMCG_SWAP */
>
> #endif /* !__GENERATING_BOUNDS_H */
> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
> index 189fbf5..0ebd127 100644
> --- a/mm/page_cgroup.c
> +++ b/mm/page_cgroup.c
> @@ -362,14 +362,28 @@ static int swap_cgroup_prepare(int type)
> unsigned long idx, max;
>
> ctrl = &swap_cgroup_ctrl[type];
> + if (!ctrl->length) {
> + /*
> + * Bypass the buffer allocation if the corresponding swap
> + * partition/file was turned off.
> + */
> + pr_debug("couldn't allocate swap_cgroup on a disabled swap "
> + "partition or file, index: %d\n", type);
Do we really need to log this? I guess your scenario is:
swapon part1
swapon part2
swapon part3
swapoff part2
create first non-root cgroup
which is perfectly ok and I do not see any reason to log it.
> + return 0;
> + }
> +
> ctrl->map = vzalloc(ctrl->length * sizeof(void *));
> - if (!ctrl->map)
> + if (!ctrl->map) {
> + ctrl->length = 0;
> goto nomem;
> + }
>
> for (idx = 0; idx < ctrl->length; idx++) {
> page = alloc_page(GFP_KERNEL | __GFP_ZERO);
> - if (!page)
> + if (!page) {
> + ctrl->length = 0;
> goto not_enough_page;
> + }
> ctrl->map[idx] = page;
> }
> return 0;
> @@ -383,6 +397,32 @@ nomem:
ctrl->length = 0 under this label would be probably nicer than keeping
it at two places.
> return -ENOMEM;
> }
>
> +/*
> + * free buffer for swap_cgroup.
> + */
> +static void swap_cgroup_teardown(int type)
> +{
> + struct page **map;
> + unsigned long length;
> + struct swap_cgroup_ctrl *ctrl;
> +
> + ctrl = &swap_cgroup_ctrl[type];
> + map = ctrl->map;
> + length = ctrl->length;
> + ctrl->map = NULL;
> + ctrl->length = 0;
> +
> + if (map) {
allocation path checks for ctrl->length so it would be good to unify
both. They are handling the same case (gone swap).
> + unsigned long i;
> + for (i = 0; i < length; i++) {
> + struct page *page = map[i];
> + if (page)
> + __free_page(page);
> + }
> + vfree(map);
> + }
> +}
> +
> static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
> struct swap_cgroup_ctrl **ctrlp)
> {
> @@ -474,6 +514,56 @@ unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
> return sc ? sc->id : 0;
> }
>
> +/*
> + * Allocate swap cgroup accounting structures when the first non-root
> + * memcg is created.
> + */
> +int swap_cgroup_init(void)
> +{
> + unsigned int type;
> +
> + if (!do_swap_account)
> + return 0;
> +
> + if (atomic_add_return(1, &memsw_accounting_users) != 1)
> + return 0;
> +
> + mutex_lock(&swap_cgroup_mutex);
> + for (type = 0; type < nr_swapfiles; type++) {
> + if (swap_cgroup_prepare(type) < 0) {
> + mutex_unlock(&swap_cgroup_mutex);
> + goto nomem;
> + }
> + }
You should clean up those types that were successful...
> + mutex_unlock(&swap_cgroup_mutex);
> + return 0;
> +
> +nomem:
> + pr_info("couldn't allocate enough memory for swap_cgroup "
> + "while creating non-root memcg.\n");
> + return -ENOMEM;
> +}
> +
> +/*
> + * Deallocate swap cgroup accounting structures on the last non-root
> + * memcg removal.
> + */
> +void swap_cgroup_free(void)
> +{
> + unsigned int type;
> +
> + if (!do_swap_account)
> + return;
> +
> + if (atomic_sub_return(1, &memsw_accounting_users))
> + return;
> +
> + mutex_lock(&swap_cgroup_mutex);
> + for (type = 0; type < nr_swapfiles; type++)
> + swap_cgroup_teardown(type);
> + mutex_unlock(&swap_cgroup_mutex);
> +}
> +
> int swap_cgroup_swapon(int type, unsigned long max_pages)
> {
> unsigned long length;
> @@ -482,20 +572,18 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
> if (!do_swap_account)
> return 0;
>
> - if (!atomic_read(&memsw_accounting_users))
> - return 0;
> -
> length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
>
> ctrl = &swap_cgroup_ctrl[type];
> mutex_lock(&swap_cgroup_mutex);
> ctrl->length = length;
> spin_lock_init(&ctrl->lock);
> - if (swap_cgroup_prepare(type)) {
> - /* memory shortage */
> - ctrl->length = 0;
> - mutex_unlock(&swap_cgroup_mutex);
> - goto nomem;
> + if (atomic_read(&memsw_accounting_users)) {
> + if (swap_cgroup_prepare(type)) {
> + /* memory shortage */
> + mutex_unlock(&swap_cgroup_mutex);
> + goto nomem;
> + }
> }
> mutex_unlock(&swap_cgroup_mutex);
>
> --
> 1.7.9.5
--
Michal Hocko
SUSE Labs
^ permalink raw reply [flat|nested] 26+ messages in thread* Re: [PATCH v2 5/6] memcg: introduce swap_cgroup_init()/swap_cgroup_free()
2013-01-29 14:56 ` Michal Hocko
@ 2013-01-29 15:51 ` Jeff Liu
[not found] ` <5107F00E.7070302-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 1 reply; 26+ messages in thread
From: Jeff Liu @ 2013-01-29 15:51 UTC (permalink / raw)
To: Michal Hocko; +Cc: linux-mm, Glauber Costa, cgroups
On 01/29/2013 10:56 PM, Michal Hocko wrote:
> On Mon 28-01-13 18:54:47, Jeff Liu wrote:
>> Introduce swap_cgroup_init()/swap_cgroup_free() to allocate buffers when creating the first
>> non-root memcg and deallocate buffers on the last non-root memcg is gone.
>
> I think this deserves more words ;) At least it would be good to
> describe contexts from which init and free might be called. What are the
> locking rules.
> Also swap_cgroup_destroy sounds more in pair with swap_cgroup_init.
Will improve the comments log as well as fix the naming. Btw, I named
it as swap_cgroup_free() because we have mem_cgroup_free() corresponding
to mem_cgroup_init(). :)
>
> Please add the users of those function here as well. It is much easier
> to review.
Sure.
>
>> Signed-off-by: Jie Liu <jeff.liu@oracle.com>
>> CC: Glauber Costa <glommer@parallels.com>
>> CC: Michal Hocko <mhocko@suse.cz>
>> CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
>> CC: Johannes Weiner <hannes@cmpxchg.org>
>> CC: Mel Gorman <mgorman@suse.de>
>> CC: Andrew Morton <akpm@linux-foundation.org>
>> CC: Sha Zhengju <handai.szj@taobao.com>
>>
>> ---
>> include/linux/page_cgroup.h | 12 +++++
>> mm/page_cgroup.c | 108 +++++++++++++++++++++++++++++++++++++++----
>> 2 files changed, 110 insertions(+), 10 deletions(-)
>>
>> diff --git a/include/linux/page_cgroup.h b/include/linux/page_cgroup.h
>> index 777a524..1255cc9 100644
>> --- a/include/linux/page_cgroup.h
>> +++ b/include/linux/page_cgroup.h
>> @@ -113,6 +113,8 @@ extern unsigned short swap_cgroup_record(swp_entry_t ent, unsigned short id);
>> extern unsigned short lookup_swap_cgroup_id(swp_entry_t ent);
>> extern int swap_cgroup_swapon(int type, unsigned long max_pages);
>> extern void swap_cgroup_swapoff(int type);
>> +extern int swap_cgroup_init(void);
>> +extern void swap_cgroup_free(void);
>> #else
>>
>> static inline
>> @@ -138,6 +140,16 @@ static inline void swap_cgroup_swapoff(int type)
>> return;
>> }
>>
>> +static inline int swap_cgroup_init(void)
>> +{
>> + return 0;
>> +}
>> +
>> +static inline void swap_cgroup_free(void)
>> +{
>> + return;
>> +}
>> +
>> #endif /* CONFIG_MEMCG_SWAP */
>>
>> #endif /* !__GENERATING_BOUNDS_H */
>> diff --git a/mm/page_cgroup.c b/mm/page_cgroup.c
>> index 189fbf5..0ebd127 100644
>> --- a/mm/page_cgroup.c
>> +++ b/mm/page_cgroup.c
>> @@ -362,14 +362,28 @@ static int swap_cgroup_prepare(int type)
>> unsigned long idx, max;
>>
>> ctrl = &swap_cgroup_ctrl[type];
>> + if (!ctrl->length) {
>> + /*
>> + * Bypass the buffer allocation if the corresponding swap
>> + * partition/file was turned off.
>> + */
>> + pr_debug("couldn't allocate swap_cgroup on a disabled swap "
>> + "partition or file, index: %d\n", type);
>
> Do we really need to log this? I guess your scenario is:
> swapon part1
> swapon part2
> swapon part3
> swapoff part2
> create first non-root cgroup
Yesss!!
>
> which is perfectly ok and I do not see any reason to log it.
okay, I'll remove this redundant debug info.
>
>> + return 0;
>> + }
>> +
>> ctrl->map = vzalloc(ctrl->length * sizeof(void *));
>> - if (!ctrl->map)
>> + if (!ctrl->map) {
>> + ctrl->length = 0;
>> goto nomem;
>> + }
>>
>> for (idx = 0; idx < ctrl->length; idx++) {
>> page = alloc_page(GFP_KERNEL | __GFP_ZERO);
>> - if (!page)
>> + if (!page) {
>> + ctrl->length = 0;
>> goto not_enough_page;
>> + }
>> ctrl->map[idx] = page;
>> }
>> return 0;
>> @@ -383,6 +397,32 @@ nomem:
>
> ctrl->length = 0 under this label would be probably nicer than keeping
> it at two places.
Will take care of it.
>
>> return -ENOMEM;
>> }
>>
>> +/*
>> + * free buffer for swap_cgroup.
>> + */
>> +static void swap_cgroup_teardown(int type)
>> +{
>> + struct page **map;
>> + unsigned long length;
>> + struct swap_cgroup_ctrl *ctrl;
>> +
>> + ctrl = &swap_cgroup_ctrl[type];
>> + map = ctrl->map;
>> + length = ctrl->length;
>> + ctrl->map = NULL;
>> + ctrl->length = 0;
>> +
>> + if (map) {
>
> allocation path checks for ctrl->length so it would be good to unify
> both. They are handling the same case (gone swap).
yes, will fix it accordingly.
>
>> + unsigned long i;
>> + for (i = 0; i < length; i++) {
>> + struct page *page = map[i];
>> + if (page)
>> + __free_page(page);
>> + }
>> + vfree(map);
>> + }
>> +}
>> +
>> static struct swap_cgroup *lookup_swap_cgroup(swp_entry_t ent,
>> struct swap_cgroup_ctrl **ctrlp)
>> {
>> @@ -474,6 +514,56 @@ unsigned short lookup_swap_cgroup_id(swp_entry_t ent)
>> return sc ? sc->id : 0;
>> }
>>
>> +/*
>> + * Allocate swap cgroup accounting structures when the first non-root
>> + * memcg is created.
>> + */
>> +int swap_cgroup_init(void)
>> +{
>> + unsigned int type;
>> +
>> + if (!do_swap_account)
>> + return 0;
>> +
>> + if (atomic_add_return(1, &memsw_accounting_users) != 1)
>> + return 0;
>> +
>> + mutex_lock(&swap_cgroup_mutex);
>> + for (type = 0; type < nr_swapfiles; type++) {
>> + if (swap_cgroup_prepare(type) < 0) {
>> + mutex_unlock(&swap_cgroup_mutex);
>> + goto nomem;
>> + }
>> + }
>
> You should clean up those types that were successful...
Oops, those types should be deallocated...
Thanks,
-Jeff
>
>> + mutex_unlock(&swap_cgroup_mutex);
>> + return 0;
>> +
>> +nomem:
>> + pr_info("couldn't allocate enough memory for swap_cgroup "
>> + "while creating non-root memcg.\n");
>> + return -ENOMEM;
>> +}
>> +
>> +/*
>> + * Deallocate swap cgroup accounting structures on the last non-root
>> + * memcg removal.
>> + */
>> +void swap_cgroup_free(void)
>> +{
>> + unsigned int type;
>> +
>> + if (!do_swap_account)
>> + return;
>> +
>> + if (atomic_sub_return(1, &memsw_accounting_users))
>> + return;
>> +
>> + mutex_lock(&swap_cgroup_mutex);
>> + for (type = 0; type < nr_swapfiles; type++)
>> + swap_cgroup_teardown(type);
>> + mutex_unlock(&swap_cgroup_mutex);
>> +}
>> +
>> int swap_cgroup_swapon(int type, unsigned long max_pages)
>> {
>> unsigned long length;
>> @@ -482,20 +572,18 @@ int swap_cgroup_swapon(int type, unsigned long max_pages)
>> if (!do_swap_account)
>> return 0;
>>
>> - if (!atomic_read(&memsw_accounting_users))
>> - return 0;
>> -
>> length = DIV_ROUND_UP(max_pages, SC_PER_PAGE);
>>
>> ctrl = &swap_cgroup_ctrl[type];
>> mutex_lock(&swap_cgroup_mutex);
>> ctrl->length = length;
>> spin_lock_init(&ctrl->lock);
>> - if (swap_cgroup_prepare(type)) {
>> - /* memory shortage */
>> - ctrl->length = 0;
>> - mutex_unlock(&swap_cgroup_mutex);
>> - goto nomem;
>> + if (atomic_read(&memsw_accounting_users)) {
>> + if (swap_cgroup_prepare(type)) {
>> + /* memory shortage */
>> + mutex_unlock(&swap_cgroup_mutex);
>> + goto nomem;
>> + }
>> }
>> mutex_unlock(&swap_cgroup_mutex);
>>
>> --
>> 1.7.9.5
>
--
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>
^ permalink raw reply [flat|nested] 26+ messages in thread
* [PATCH v2 6/6] memcg: init/free swap cgroup strucutres upon create/free child memcg
@ 2013-01-28 10:54 ` Jeff Liu
[not found] ` <510658FC.50009-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
0 siblings, 1 reply; 26+ messages in thread
From: Jeff Liu @ 2013-01-28 10:54 UTC (permalink / raw)
To: linux-mm; +Cc: Michal Hocko, Glauber Costa, cgroups
Initialize swap_cgroup strucutres when creating a non-root memcg,
swap_cgroup_init() will be called for multiple times but only does
buffer allocation per the first non-root memcg.
Free swap_cgroup structures correspondingly on the last non-root memcg
removal.
Signed-off-by: Jie Liu <jeff.liu@oracle.com>
CC: Glauber Costa <glommer@parallels.com>
CC: Michal Hocko <mhocko@suse.cz>
CC: Kamezawa Hiroyuki <kamezawa.hiroyu@jp.fujitsu.com>
CC: Johannes Weiner <hannes@cmpxchg.org>
CC: Mel Gorman <mgorman@suse.de>
CC: Andrew Morton <akpm@linux-foundation.org>
CC: Sha Zhengju <handai.szj@taobao.com>
---
mm/memcontrol.c | 3 +++
1 file changed, 3 insertions(+)
diff --git a/mm/memcontrol.c b/mm/memcontrol.c
index afe5e86..031d242 100644
--- a/mm/memcontrol.c
+++ b/mm/memcontrol.c
@@ -5998,6 +5998,7 @@ static void free_work(struct work_struct *work)
memcg = container_of(work, struct mem_cgroup, work_freeing);
__mem_cgroup_free(memcg);
+ swap_cgroup_free();
}
static void free_rcu(struct rcu_head *rcu_head)
@@ -6116,6 +6117,8 @@ mem_cgroup_css_alloc(struct cgroup *cont)
INIT_WORK(&stock->work, drain_local_stock);
}
} else {
+ if (swap_cgroup_init())
+ goto free_out;
parent = mem_cgroup_from_cont(cont->parent);
memcg->use_hierarchy = parent->use_hierarchy;
memcg->oom_kill_disable = parent->oom_kill_disable;
--
1.7.9.5
--
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>
^ permalink raw reply related [flat|nested] 26+ messages in thread[parent not found: <510658E3.1020306-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>]
* Re: [PATCH v2 0/6] memcg: disable swap cgroup allocation at swapon
[not found] ` <510658E3.1020306-QHcLZuEGTsvQT0dZR+AlfA@public.gmane.org>
@ 2013-01-29 15:15 ` Michal Hocko
2013-01-29 16:50 ` Jeff Liu
0 siblings, 1 reply; 26+ messages in thread
From: Michal Hocko @ 2013-01-29 15:15 UTC (permalink / raw)
To: Jeff Liu
Cc: linux-mm-Bw31MaZKKs3YtjvyW6yDsg, Glauber Costa,
cgroups-u79uwXL29TY76Z2rM5mHXA
On Mon 28-01-13 18:54:27, Jeff Liu wrote:
> Hello,
>
> Here is the v2 patch set for disabling swap_cgroup structures allocation
> per swapon.
>
> In the initial version, one big issue is that I have missed the swap tracking
> for the root memcg, thanks Michal pointing it out. :)
>
> In order to solve it, the easiest approach I can think out is to bypass the root
> memcg swap accounting during the business and figure it out with some global stats,
> which means that we always return 0 per root memcg swap charge/uncharge stage, and
> this is inspired by another proposal from Zhengju:
> "memcg: Don't account root memcg page statistics -- https://lkml.org/lkml/2013/1/2/71"
>
> Besides that, another major fix is deallocate swap accounting structures on the last
> non-root memcg remove after all references to it are gone rather than doing it on
> mem_cgroup_destroy().
>
> Any comment are welcome!
Could you also post your testing methodology and results, please? It
would be also really great if you could sum up memory savings.
Anyway thanks this second version looks really promising.
> v1->v2:
> - Refactor swap_cgroup_swapon()/swap_cgroup_prepare(), to make the later can be
> used for allocating buffers per the first non-root memcg creation.
> - Bypass root memcg swap statistics, using the global stats to figure it out instead.
> - Export nr_swap_files which would be used when creating/freeing swap_cgroup
> - Deallocate swap accounting structures on the last non-root memcg removal
>
> Old patch set:
> v1:
> http://marc.info/?l=linux-mm&m=135461016823964&w=2
>
>
> Thanks,
> -Jeff
>
> --
> To unsubscribe, send a message with 'unsubscribe linux-mm' in
> the body to majordomo-Bw31MaZKKs0EbZ0PF+XxCw@public.gmane.org For more info on Linux MM,
> see: http://www.linux-mm.org/ .
> Don't email: <a href=mailto:"dont-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org"> email-Bw31MaZKKs3YtjvyW6yDsg@public.gmane.org </a>
--
Michal Hocko
SUSE Labs
^ permalink raw reply [flat|nested] 26+ messages in thread* Re: [PATCH v2 0/6] memcg: disable swap cgroup allocation at swapon
2013-01-29 15:15 ` [PATCH v2 0/6] memcg: disable swap cgroup allocation at swapon Michal Hocko
@ 2013-01-29 16:50 ` Jeff Liu
0 siblings, 0 replies; 26+ messages in thread
From: Jeff Liu @ 2013-01-29 16:50 UTC (permalink / raw)
To: Michal Hocko; +Cc: linux-mm, Glauber Costa, cgroups
On 01/29/2013 11:15 PM, Michal Hocko wrote:
> On Mon 28-01-13 18:54:27, Jeff Liu wrote:
>> Hello,
>>
>> Here is the v2 patch set for disabling swap_cgroup structures allocation
>> per swapon.
>>
>> In the initial version, one big issue is that I have missed the swap tracking
>> for the root memcg, thanks Michal pointing it out. :)
>>
>> In order to solve it, the easiest approach I can think out is to bypass the root
>> memcg swap accounting during the business and figure it out with some global stats,
>> which means that we always return 0 per root memcg swap charge/uncharge stage, and
>> this is inspired by another proposal from Zhengju:
>> "memcg: Don't account root memcg page statistics -- https://lkml.org/lkml/2013/1/2/71"
>>
>> Besides that, another major fix is deallocate swap accounting structures on the last
>> non-root memcg remove after all references to it are gone rather than doing it on
>> mem_cgroup_destroy().
>>
>> Any comment are welcome!
>
> Could you also post your testing methodology and results, please? It
> would be also really great if you could sum up memory savings.
Sure, I'll post those info in the next try, and will revisit [PATCH v2
2/6] and answer the comments from you and Glauber since my head is not
very clear now. :)
>
> Anyway thanks this second version looks really promising.
Thanks for your kind review, have a nice day!
-Jeff
>
>> v1->v2:
>> - Refactor swap_cgroup_swapon()/swap_cgroup_prepare(), to make the later can be
>> used for allocating buffers per the first non-root memcg creation.
>> - Bypass root memcg swap statistics, using the global stats to figure it out instead.
>> - Export nr_swap_files which would be used when creating/freeing swap_cgroup
>> - Deallocate swap accounting structures on the last non-root memcg removal
>>
>> Old patch set:
>> v1:
>> http://marc.info/?l=linux-mm&m=135461016823964&w=2
>>
>>
>> Thanks,
>> -Jeff
>>
>> --
>> 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>
>
--
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>
^ permalink raw reply [flat|nested] 26+ messages in thread