mm/slab.c: use 'ac' from the caller
diff mbox series

Message ID 20210421022518.67451-1-yejune.deng@gmail.com
State New
Headers show
Series
  • mm/slab.c: use 'ac' from the caller
Related show

Commit Message

Yejune Deng April 21, 2021, 2:25 a.m. UTC
It can use 'ac' from ____cache_alloc() in cache_alloc_refill().
This saves call cpu_cache_get() twice.

Signed-off-by: Yejune Deng <yejune.deng@gmail.com>
---
 mm/slab.c | 12 +++---------
 1 file changed, 3 insertions(+), 9 deletions(-)

Comments

Matthew Wilcox April 21, 2021, 3:10 a.m. UTC | #1
On Wed, Apr 21, 2021 at 10:25:17AM +0800, Yejune Deng wrote:
> @@ -3045,12 +3044,7 @@ static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
>  	}
>  
>  	STATS_INC_ALLOCMISS(cachep);
> -	objp = cache_alloc_refill(cachep, flags);
> -	/*
> -	 * the 'ac' may be updated by cache_alloc_refill(),
> -	 * and kmemleak_erase() requires its correct value.
> -	 */
> -	ac = cpu_cache_get(cachep);
> +	objp = cache_alloc_refill(cachep, ac, flags);

I think passing 'ac' in is fine (probably?  I don't know this code
deeply), but deleting this call to 'ac' is clearly wrong.  The comment
even tells you that!  I just verified the code, and the comment is
correct.
Christoph Lameter April 21, 2021, 7:20 a.m. UTC | #2
On Wed, 21 Apr 2021, Matthew Wilcox wrote:

> On Wed, Apr 21, 2021 at 10:25:17AM +0800, Yejune Deng wrote:
> > @@ -3045,12 +3044,7 @@ static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
> >  	}
> >
> >  	STATS_INC_ALLOCMISS(cachep);
> > -	objp = cache_alloc_refill(cachep, flags);
> > -	/*
> > -	 * the 'ac' may be updated by cache_alloc_refill(),
> > -	 * and kmemleak_erase() requires its correct value.
> > -	 */
> > -	ac = cpu_cache_get(cachep);
> > +	objp = cache_alloc_refill(cachep, ac, flags);
>
> I think passing 'ac' in is fine (probably?  I don't know this code
> deeply), but deleting this call to 'ac' is clearly wrong.  The comment
> even tells you that!  I just verified the code, and the comment is
> correct.


Yep the delete of the ac assignment is wrong.

But even without that issue: There is no point to passing ac to
cache_alloc_refill since cpu_cache_get is rather trivial and does not
even require memory access since "cachep" is usually in some register.

Patch
diff mbox series

diff --git a/mm/slab.c b/mm/slab.c
index d0f725637663..4b2dc8f8cc37 100644
--- a/mm/slab.c
+++ b/mm/slab.c
@@ -2896,11 +2896,11 @@  static __always_inline int alloc_block(struct kmem_cache *cachep,
 	return batchcount;
 }
 
-static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
+static void *cache_alloc_refill(struct kmem_cache *cachep, struct array_cache *ac, gfp_t flags)
 {
 	int batchcount;
 	struct kmem_cache_node *n;
-	struct array_cache *ac, *shared;
+	struct array_cache *shared;
 	int node;
 	void *list = NULL;
 	struct page *page;
@@ -2908,7 +2908,6 @@  static void *cache_alloc_refill(struct kmem_cache *cachep, gfp_t flags)
 	check_irq_off();
 	node = numa_mem_id();
 
-	ac = cpu_cache_get(cachep);
 	batchcount = ac->batchcount;
 	if (!ac->touched && batchcount > BATCHREFILL_LIMIT) {
 		/*
@@ -3045,12 +3044,7 @@  static inline void *____cache_alloc(struct kmem_cache *cachep, gfp_t flags)
 	}
 
 	STATS_INC_ALLOCMISS(cachep);
-	objp = cache_alloc_refill(cachep, flags);
-	/*
-	 * the 'ac' may be updated by cache_alloc_refill(),
-	 * and kmemleak_erase() requires its correct value.
-	 */
-	ac = cpu_cache_get(cachep);
+	objp = cache_alloc_refill(cachep, ac, flags);
 
 out:
 	/*