linux-kernel.vger.kernel.org archive mirror
 help / color / mirror / Atom feed
* [PATCH -mm 4/6] slqb: record slab's start address in slqb_page
@ 2009-03-03 13:44 Lai Jiangshan
  2009-03-23  7:54 ` Pekka Enberg
  0 siblings, 1 reply; 2+ messages in thread
From: Lai Jiangshan @ 2009-03-03 13:44 UTC (permalink / raw)
  To: Andrew Morton
  Cc: Pekka Enberg, Christoph Lameter, Nick Piggin, Paul E. McKenney,
	Manfred Spraul, Ingo Molnar, Peter Zijlstra, linux-kernel


slqb's slab is coloring, it's will be useful that record the
color value or start address in struct slqb_page.

This fix adds a field in struct slqb_page to record slab's
start address, It will be used by kfree_rcu().


struct slqb_page has a field:
	struct kmem_cache_list *list;

the corresponding field in struct page is:

	union {
	    struct {
		unsigned long private;
		struct address_space *mapping;
	    };
	    struct kmem_cache *slab;
	    struct page *first_page;
	};

It's large than 'struct kmem_cache_list *list;'
*So we are safe to add a field in struct page.*


(This field will not become a burden in future:)
struct page is very small, this fix uses the last room of
struct page. But if we have to add a more important
field in struct slqb_page in future, We can record slab's
color value in slqb_page instead of start address.
like this:
	struct {
		unsigned long	flags;
		atomic_t	_count;
		u16		inuse;		/* use u16 instead */
		u16		color;  	/* of unsigned int */
		void 		*a_more_important_field;
		struct kmem_cache_list *list;
		void		 **freelist;
		union {
			struct list_head lru;
			struct rcu_head rcu_head;
		};
	};

Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
---
diff --git a/mm/slqb.c b/mm/slqb.c
index 48bdf3b..c258ddc 100644
--- a/mm/slqb.c
+++ b/mm/slqb.c
@@ -41,6 +41,7 @@ struct slqb_page {
 			unsigned long	flags;		/* mandatory */
 			atomic_t	_count;		/* mandatory */
 			unsigned int	inuse;		/* Nr of objects */
+			void 		*s_mem;		/* start address */
 			struct kmem_cache_list *list;	/* Pointer to list */
 			void		 **freelist;	/* LIFO freelist */
 			union {
@@ -1010,6 +1011,7 @@ static struct slqb_page *new_slab_page(struct kmem_cache *s,
 	}
 	set_freepointer(s, last, NULL);
 
+	page->s_mem = start;
 	page->freelist = start;
 	page->inuse = 0;
 out:









^ permalink raw reply related	[flat|nested] 2+ messages in thread

* Re: [PATCH -mm 4/6] slqb: record slab's start address in slqb_page
  2009-03-03 13:44 [PATCH -mm 4/6] slqb: record slab's start address in slqb_page Lai Jiangshan
@ 2009-03-23  7:54 ` Pekka Enberg
  0 siblings, 0 replies; 2+ messages in thread
From: Pekka Enberg @ 2009-03-23  7:54 UTC (permalink / raw)
  To: Lai Jiangshan
  Cc: Andrew Morton, Christoph Lameter, Nick Piggin, Paul E. McKenney,
	Manfred Spraul, Ingo Molnar, Peter Zijlstra, linux-kernel

On Tue, 2009-03-03 at 21:44 +0800, Lai Jiangshan wrote:
> slqb's slab is coloring, it's will be useful that record the
> color value or start address in struct slqb_page.
> 
> This fix adds a field in struct slqb_page to record slab's
> start address, It will be used by kfree_rcu().
> 
> 
> struct slqb_page has a field:
> 	struct kmem_cache_list *list;
> 
> the corresponding field in struct page is:
> 
> 	union {
> 	    struct {
> 		unsigned long private;
> 		struct address_space *mapping;
> 	    };
> 	    struct kmem_cache *slab;
> 	    struct page *first_page;
> 	};
> 
> It's large than 'struct kmem_cache_list *list;'
> *So we are safe to add a field in struct page.*
> 
> 
> (This field will not become a burden in future:)
> struct page is very small, this fix uses the last room of
> struct page. But if we have to add a more important
> field in struct slqb_page in future, We can record slab's
> color value in slqb_page instead of start address.
> like this:
> 	struct {
> 		unsigned long	flags;
> 		atomic_t	_count;
> 		u16		inuse;		/* use u16 instead */
> 		u16		color;  	/* of unsigned int */
> 		void 		*a_more_important_field;
> 		struct kmem_cache_list *list;
> 		void		 **freelist;
> 		union {
> 			struct list_head lru;
> 			struct rcu_head rcu_head;
> 		};
> 	};
> 
> Signed-off-by: Lai Jiangshan <laijs@cn.fujitsu.com>
> ---
> diff --git a/mm/slqb.c b/mm/slqb.c
> index 48bdf3b..c258ddc 100644
> --- a/mm/slqb.c
> +++ b/mm/slqb.c
> @@ -41,6 +41,7 @@ struct slqb_page {
>  			unsigned long	flags;		/* mandatory */
>  			atomic_t	_count;		/* mandatory */
>  			unsigned int	inuse;		/* Nr of objects */
> +			void 		*s_mem;		/* start address */
>  			struct kmem_cache_list *list;	/* Pointer to list */
>  			void		 **freelist;	/* LIFO freelist */
>  			union {
> @@ -1010,6 +1011,7 @@ static struct slqb_page *new_slab_page(struct kmem_cache *s,
>  	}
>  	set_freepointer(s, last, NULL);
>  
> +	page->s_mem = start;
>  	page->freelist = start;
>  	page->inuse = 0;
>  out:

I am okay with this but I think Nick was against it. Nick?

			Pekka


^ permalink raw reply	[flat|nested] 2+ messages in thread

end of thread, other threads:[~2009-03-23  7:54 UTC | newest]

Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2009-03-03 13:44 [PATCH -mm 4/6] slqb: record slab's start address in slqb_page Lai Jiangshan
2009-03-23  7:54 ` Pekka Enberg

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).