LKML Archive on lore.kernel.org
help / color / mirror / Atom feed
* [PATCH] Move memory controller allocations to their own slabs
@ 2008-03-11  4:31 Balbir Singh
  2008-03-11  4:41 ` Andrew Morton
                   ` (2 more replies)
  0 siblings, 3 replies; 7+ messages in thread
From: Balbir Singh @ 2008-03-11  4:31 UTC (permalink / raw)
  To: Paul Menage, Andrew Morton, Pavel Emelianov
  Cc: Hugh Dickins, Sudhir Kumar, YAMAMOTO Takashi, lizf, linux-kernel,
	taka, linux-mm, David Rientjes, Balbir Singh, KAMEZAWA Hiroyuki



Move the memory controller data structures page_cgroup and
mem_cgroup_per_zone to their own slab caches. It saves space on the system,
allocations are not necessarily pushed to order of 2 and should provide
performance benefits. Users who disable the memory controller can also double
check that the memory controller is not allocating page_cgroup's.

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

 linux/memcontrol.h |    0 
 mm/memcontrol.c    |   21 ++++++++++++++-------
 2 files changed, 14 insertions(+), 7 deletions(-)

diff -puN mm/memcontrol.c~memory-controller-move-to-own-slab mm/memcontrol.c
--- linux-2.6.25-rc4/mm/memcontrol.c~memory-controller-move-to-own-slab	2008-03-10 23:22:34.000000000 +0530
+++ linux-2.6.25-rc4-balbir/mm/memcontrol.c	2008-03-10 23:34:42.000000000 +0530
@@ -26,6 +26,7 @@
 #include <linux/backing-dev.h>
 #include <linux/bit_spinlock.h>
 #include <linux/rcupdate.h>
+#include <linux/slab.h>
 #include <linux/swap.h>
 #include <linux/spinlock.h>
 #include <linux/fs.h>
@@ -35,6 +36,8 @@
 
 struct cgroup_subsys mem_cgroup_subsys;
 static const int MEM_CGROUP_RECLAIM_RETRIES = 5;
+static struct kmem_cache *page_cgroup_cache;
+static struct kmem_cache *mem_cgroup_per_zone_cache;
 
 /*
  * Statistics for memory cgroup.
@@ -560,7 +563,7 @@ retry:
 	}
 	unlock_page_cgroup(page);
 
-	pc = kzalloc(sizeof(struct page_cgroup), gfp_mask);
+	pc = kmem_cache_zalloc(page_cgroup_cache, gfp_mask);
 	if (pc == NULL)
 		goto err;
 
@@ -622,7 +625,7 @@ retry:
 		 */
 		res_counter_uncharge(&mem->res, PAGE_SIZE);
 		css_put(&mem->css);
-		kfree(pc);
+		kmem_cache_free(page_cgroup_cache, pc);
 		goto retry;
 	}
 	page_assign_page_cgroup(page, pc);
@@ -637,7 +640,7 @@ done:
 	return 0;
 out:
 	css_put(&mem->css);
-	kfree(pc);
+	kmem_cache_free(page_cgroup_cache, pc);
 err:
 	return -ENOMEM;
 }
@@ -695,7 +698,7 @@ void mem_cgroup_uncharge_page(struct pag
 		res_counter_uncharge(&mem->res, PAGE_SIZE);
 		css_put(&mem->css);
 
-		kfree(pc);
+		kmem_cache_free(page_cgroup_cache, pc);
 		return;
 	}
 
@@ -988,9 +991,10 @@ static int alloc_mem_cgroup_per_zone_inf
 	 *       function.
 	 */
 	if (node_state(node, N_HIGH_MEMORY))
-		pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, node);
+		pn = kmem_cache_alloc_node(mem_cgroup_per_zone_cache,
+						GFP_KERNEL, node);
 	else
-		pn = kmalloc(sizeof(*pn), GFP_KERNEL);
+		pn = kmem_cache_alloc(mem_cgroup_per_zone_cache, GFP_KERNEL);
 	if (!pn)
 		return 1;
 
@@ -1008,7 +1012,7 @@ static int alloc_mem_cgroup_per_zone_inf
 
 static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
 {
-	kfree(mem->info.nodeinfo[node]);
+	kmem_cache_free(mem_cgroup_per_zone_cache, mem->info.nodeinfo[node]);
 }
 
 static struct cgroup_subsys_state *
@@ -1020,6 +1024,9 @@ mem_cgroup_create(struct cgroup_subsys *
 	if (unlikely((cont->parent) == NULL)) {
 		mem = &init_mem_cgroup;
 		init_mm.mem_cgroup = mem;
+		page_cgroup_cache = KMEM_CACHE(page_cgroup, SLAB_PANIC);
+		mem_cgroup_per_zone_cache = KMEM_CACHE(mem_cgroup_per_zone,
+							SLAB_PANIC);
 	} else
 		mem = kzalloc(sizeof(struct mem_cgroup), GFP_KERNEL);
 
diff -puN include/linux/memcontrol.h~memory-controller-move-to-own-slab include/linux/memcontrol.h
_

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] Move memory controller allocations to their own slabs
  2008-03-11  4:31 [PATCH] Move memory controller allocations to their own slabs Balbir Singh
@ 2008-03-11  4:41 ` Andrew Morton
  2008-03-11  4:45   ` Balbir Singh
  2008-03-11  4:45 ` KAMEZAWA Hiroyuki
  2008-03-11  5:00 ` + KOSAKI Motohiro
  2 siblings, 1 reply; 7+ messages in thread
From: Andrew Morton @ 2008-03-11  4:41 UTC (permalink / raw)
  To: Balbir Singh
  Cc: Paul Menage, Pavel Emelianov, Hugh Dickins, Sudhir Kumar,
	YAMAMOTO Takashi, lizf, linux-kernel, taka, linux-mm,
	David Rientjes, KAMEZAWA Hiroyuki

On Tue, 11 Mar 2008 10:01:49 +0530 Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> 
> 
> Move the memory controller data structures page_cgroup and
> mem_cgroup_per_zone to their own slab caches. It saves space on the system,
> allocations are not necessarily pushed to order of 2 and should provide
> performance benefits.

eh?  Those structures are tiny.  Which slab allocator has gone and used an
order-2 allocation and for which structure did it (stupidly) do this?



^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] Move memory controller allocations to their own slabs
  2008-03-11  4:41 ` Andrew Morton
@ 2008-03-11  4:45   ` Balbir Singh
  0 siblings, 0 replies; 7+ messages in thread
From: Balbir Singh @ 2008-03-11  4:45 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Paul Menage, Pavel Emelianov, Hugh Dickins, Sudhir Kumar,
	YAMAMOTO Takashi, lizf, linux-kernel, taka, linux-mm,
	David Rientjes, KAMEZAWA Hiroyuki

Andrew Morton wrote:
> On Tue, 11 Mar 2008 10:01:49 +0530 Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
>>
>> Move the memory controller data structures page_cgroup and
>> mem_cgroup_per_zone to their own slab caches. It saves space on the system,
>> allocations are not necessarily pushed to order of 2 and should provide
>> performance benefits.
> 
> eh?  Those structures are tiny.  Which slab allocator has gone and used an
> order-2 allocation and for which structure did it (stupidly) do this?
> 
> 

When I say order of 2, I meant with kmalloc. They are rounded of to 64, 128 byte
boundary. I wanted to pack these objects more closely.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: [PATCH] Move memory controller allocations to their own slabs
  2008-03-11  4:31 [PATCH] Move memory controller allocations to their own slabs Balbir Singh
  2008-03-11  4:41 ` Andrew Morton
@ 2008-03-11  4:45 ` KAMEZAWA Hiroyuki
  2008-03-11  4:46   ` Balbir Singh
  2008-03-11  5:00 ` + KOSAKI Motohiro
  2 siblings, 1 reply; 7+ messages in thread
From: KAMEZAWA Hiroyuki @ 2008-03-11  4:45 UTC (permalink / raw)
  To: Balbir Singh
  Cc: Paul Menage, Andrew Morton, Pavel Emelianov, Hugh Dickins,
	Sudhir Kumar, YAMAMOTO Takashi, lizf, linux-kernel, taka,
	linux-mm, David Rientjes

On Tue, 11 Mar 2008 10:01:49 +0530
Balbir Singh <balbir@linux.vnet.ibm.com> wrote:

> 
> Move the memory controller data structures page_cgroup and
> mem_cgroup_per_zone to their own slab caches. It saves space on the system,
> allocations are not necessarily pushed to order of 2 and should provide
> performance benefits. Users who disable the memory controller can also double
> check that the memory controller is not allocating page_cgroup's.
> 
> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
> ---
I think using its own kmem_cache for mem_cgroup_per_zone is a bit overkill.

Thanks,
-Kame



> 
>  linux/memcontrol.h |    0 
>  mm/memcontrol.c    |   21 ++++++++++++++-------
>  2 files changed, 14 insertions(+), 7 deletions(-)
> 
> diff -puN mm/memcontrol.c~memory-controller-move-to-own-slab mm/memcontrol.c
> --- linux-2.6.25-rc4/mm/memcontrol.c~memory-controller-move-to-own-slab	2008-03-10 23:22:34.000000000 +0530
> +++ linux-2.6.25-rc4-balbir/mm/memcontrol.c	2008-03-10 23:34:42.000000000 +0530
> @@ -26,6 +26,7 @@
>  #include <linux/backing-dev.h>
>  #include <linux/bit_spinlock.h>
>  #include <linux/rcupdate.h>
> +#include <linux/slab.h>
>  #include <linux/swap.h>
>  #include <linux/spinlock.h>
>  #include <linux/fs.h>
> @@ -35,6 +36,8 @@
>  
>  struct cgroup_subsys mem_cgroup_subsys;
>  static const int MEM_CGROUP_RECLAIM_RETRIES = 5;
> +static struct kmem_cache *page_cgroup_cache;
> +static struct kmem_cache *mem_cgroup_per_zone_cache;
>  
>  /*
>   * Statistics for memory cgroup.
> @@ -560,7 +563,7 @@ retry:
>  	}
>  	unlock_page_cgroup(page);
>  
> -	pc = kzalloc(sizeof(struct page_cgroup), gfp_mask);
> +	pc = kmem_cache_zalloc(page_cgroup_cache, gfp_mask);
>  	if (pc == NULL)
>  		goto err;
>  
> @@ -622,7 +625,7 @@ retry:
>  		 */
>  		res_counter_uncharge(&mem->res, PAGE_SIZE);
>  		css_put(&mem->css);
> -		kfree(pc);
> +		kmem_cache_free(page_cgroup_cache, pc);
>  		goto retry;
>  	}
>  	page_assign_page_cgroup(page, pc);
> @@ -637,7 +640,7 @@ done:
>  	return 0;
>  out:
>  	css_put(&mem->css);
> -	kfree(pc);
> +	kmem_cache_free(page_cgroup_cache, pc);
>  err:
>  	return -ENOMEM;
>  }
> @@ -695,7 +698,7 @@ void mem_cgroup_uncharge_page(struct pag
>  		res_counter_uncharge(&mem->res, PAGE_SIZE);
>  		css_put(&mem->css);
>  
> -		kfree(pc);
> +		kmem_cache_free(page_cgroup_cache, pc);
>  		return;
>  	}
>  
> @@ -988,9 +991,10 @@ static int alloc_mem_cgroup_per_zone_inf
>  	 *       function.
>  	 */
>  	if (node_state(node, N_HIGH_MEMORY))
> -		pn = kmalloc_node(sizeof(*pn), GFP_KERNEL, node);
> +		pn = kmem_cache_alloc_node(mem_cgroup_per_zone_cache,
> +						GFP_KERNEL, node);
>  	else
> -		pn = kmalloc(sizeof(*pn), GFP_KERNEL);
> +		pn = kmem_cache_alloc(mem_cgroup_per_zone_cache, GFP_KERNEL);
>  	if (!pn)
>  		return 1;
>  
> @@ -1008,7 +1012,7 @@ static int alloc_mem_cgroup_per_zone_inf
>  
>  static void free_mem_cgroup_per_zone_info(struct mem_cgroup *mem, int node)
>  {
> -	kfree(mem->info.nodeinfo[node]);
> +	kmem_cache_free(mem_cgroup_per_zone_cache, mem->info.nodeinfo[node]);
>  }
>  
>  static struct cgroup_subsys_state *
> @@ -1020,6 +1024,9 @@ mem_cgroup_create(struct cgroup_subsys *
>  	if (unlikely((cont->parent) == NULL)) {
>  		mem = &init_mem_cgroup;
>  		init_mm.mem_cgroup = mem;
> +		page_cgroup_cache = KMEM_CACHE(page_cgroup, SLAB_PANIC);
> +		mem_cgroup_per_zone_cache = KMEM_CACHE(mem_cgroup_per_zone,
> +							SLAB_PANIC);
>  	} else
>  		mem = kzalloc(sizeof(struct mem_cgroup), GFP_KERNEL);
>  
> diff -puN include/linux/memcontrol.h~memory-controller-move-to-own-slab include/linux/memcontrol.h
> _
> 
> -- 
> 	Warm Regards,
> 	Balbir Singh
> 	Linux Technology Center
> 	IBM, ISTL
> 
> --
> 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] 7+ messages in thread

* Re: [PATCH] Move memory controller allocations to their own slabs
  2008-03-11  4:45 ` KAMEZAWA Hiroyuki
@ 2008-03-11  4:46   ` Balbir Singh
  0 siblings, 0 replies; 7+ messages in thread
From: Balbir Singh @ 2008-03-11  4:46 UTC (permalink / raw)
  To: KAMEZAWA Hiroyuki
  Cc: Paul Menage, Andrew Morton, Pavel Emelianov, Hugh Dickins,
	Sudhir Kumar, YAMAMOTO Takashi, lizf, linux-kernel, taka,
	linux-mm, David Rientjes

KAMEZAWA Hiroyuki wrote:
> On Tue, 11 Mar 2008 10:01:49 +0530
> Balbir Singh <balbir@linux.vnet.ibm.com> wrote:
> 
>> Move the memory controller data structures page_cgroup and
>> mem_cgroup_per_zone to their own slab caches. It saves space on the system,
>> allocations are not necessarily pushed to order of 2 and should provide
>> performance benefits. Users who disable the memory controller can also double
>> check that the memory controller is not allocating page_cgroup's.
>>
>> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
>> ---
> I think using its own kmem_cache for mem_cgroup_per_zone is a bit overkill.

OK, sure, we can move back to the kmalloc.

-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

^ permalink raw reply	[flat|nested] 7+ messages in thread

* +
  2008-03-11  4:31 [PATCH] Move memory controller allocations to their own slabs Balbir Singh
  2008-03-11  4:41 ` Andrew Morton
  2008-03-11  4:45 ` KAMEZAWA Hiroyuki
@ 2008-03-11  5:00 ` KOSAKI Motohiro
  2008-03-11  5:07   ` + Balbir Singh
  2 siblings, 1 reply; 7+ messages in thread
From: KOSAKI Motohiro @ 2008-03-11  5:00 UTC (permalink / raw)
  To: Balbir Singh
  Cc: kosaki.motohiro, Paul Menage, Andrew Morton, Pavel Emelianov,
	Hugh Dickins, Sudhir Kumar, YAMAMOTO Takashi, lizf, linux-kernel,
	taka, linux-mm, David Rientjes, KAMEZAWA Hiroyuki

Hi

> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
> ---
> 
>  linux/memcontrol.h |    0 

???
unnecessary hunk?
or diff comannd bug?

>  mm/memcontrol.c    |   21 ++++++++++++++-------
>  2 files changed, 14 insertions(+), 7 deletions(-)



^ permalink raw reply	[flat|nested] 7+ messages in thread

* Re: +
  2008-03-11  5:00 ` + KOSAKI Motohiro
@ 2008-03-11  5:07   ` Balbir Singh
  0 siblings, 0 replies; 7+ messages in thread
From: Balbir Singh @ 2008-03-11  5:07 UTC (permalink / raw)
  To: KOSAKI Motohiro
  Cc: Paul Menage, Andrew Morton, Pavel Emelianov, Hugh Dickins,
	Sudhir Kumar, YAMAMOTO Takashi, lizf, linux-kernel, taka,
	linux-mm, David Rientjes, KAMEZAWA Hiroyuki

KOSAKI Motohiro wrote:
> Hi
> 
>> Signed-off-by: Balbir Singh <balbir@linux.vnet.ibm.com>
>> ---
>>
>>  linux/memcontrol.h |    0 
> 
> ???
> unnecessary hunk?
> or diff comannd bug?

I use Andrew Morton's patchutils and specified both include/linux/memcontrol.h
and mm/memcontrol.c

$ cat pc/memory-controller-move-to-own-slab.pc
mm/memcontrol.c
include/linux/memcontrol.h

But refpatch generates linux/memcontrol.h. I am using diffstat v1.41. I'll
reboot to FC8 and see if that makes any difference.


-- 
	Warm Regards,
	Balbir Singh
	Linux Technology Center
	IBM, ISTL

^ permalink raw reply	[flat|nested] 7+ messages in thread

end of thread, other threads:[~2008-03-11  5:07 UTC | newest]

Thread overview: 7+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2008-03-11  4:31 [PATCH] Move memory controller allocations to their own slabs Balbir Singh
2008-03-11  4:41 ` Andrew Morton
2008-03-11  4:45   ` Balbir Singh
2008-03-11  4:45 ` KAMEZAWA Hiroyuki
2008-03-11  4:46   ` Balbir Singh
2008-03-11  5:00 ` + KOSAKI Motohiro
2008-03-11  5:07   ` + Balbir Singh

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).