diff mbox

[net-next] net: avoid one atomic operation in skb_clone()

Message ID 1412202435.16704.68.camel@edumazet-glaptop2.roam.corp.google.com
State Accepted, archived
Delegated to: David Miller
Headers show

Commit Message

Eric Dumazet Oct. 1, 2014, 10:27 p.m. UTC 
From: Eric Dumazet <edumazet@google.com>

Fast clone cloning can actually avoid an atomic_inc(), if we
guarantee prior clone_ref value is 1.

This requires a change kfree_skbmem(), to perform the 
atomic_dec_and_test() on clone_ref before setting fclone to
SKB_FCLONE_UNAVAILABLE.

Signed-off-by: Eric Dumazet <edumazet@google.com>
---
 net/core/skbuff.c |   23 +++++++++++++++++------
 1 file changed, 17 insertions(+), 6 deletions(-)



--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html

Comments

David Miller Oct. 2, 2014, 1:27 a.m. UTC | #1 
From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 01 Oct 2014 15:27:15 -0700

> From: Eric Dumazet <edumazet@google.com>
> 
> Fast clone cloning can actually avoid an atomic_inc(), if we
> guarantee prior clone_ref value is 1.
> 
> This requires a change kfree_skbmem(), to perform the 
> atomic_dec_and_test() on clone_ref before setting fclone to
> SKB_FCLONE_UNAVAILABLE.
> 
> Signed-off-by: Eric Dumazet <edumazet@google.com>

Applied, thanks Eric.
--
To unsubscribe from this list: send the line "unsubscribe netdev" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
diff mbox

Patch

diff --git a/net/core/skbuff.c b/net/core/skbuff.c
index a8cebb40699c..f77e64873caf 100644
--- a/net/core/skbuff.c
+++ b/net/core/skbuff.c
@@ -541,13 +541,20 @@  static void kfree_skbmem(struct sk_buff *skb)
 	case SKB_FCLONE_CLONE:
 		fclones = container_of(skb, struct sk_buff_fclones, skb2);
 
-		/* The clone portion is available for
-		 * fast-cloning again.
+		/* Warning : We must perform the atomic_dec_and_test() before
+		 * setting skb->fclone back to SKB_FCLONE_UNAVAILABLE, otherwise
+		 * skb_clone() could set clone_ref to 2 before our decrement.
+		 * Anyway, if we are going to free the structure, no need to
+		 * rewrite skb->fclone.
 		 */
-		skb->fclone = SKB_FCLONE_UNAVAILABLE;
-
-		if (atomic_dec_and_test(&fclones->fclone_ref))
+		if (atomic_dec_and_test(&fclones->fclone_ref)) {
 			kmem_cache_free(skbuff_fclone_cache, fclones);
+		} else {
+			/* The clone portion is available for
+			 * fast-cloning again.
+			 */
+			skb->fclone = SKB_FCLONE_UNAVAILABLE;
+		}
 		break;
 	}
 }
@@ -869,7 +876,11 @@  struct sk_buff *skb_clone(struct sk_buff *skb, gfp_t gfp_mask)
 	if (skb->fclone == SKB_FCLONE_ORIG &&
 	    n->fclone == SKB_FCLONE_UNAVAILABLE) {
 		n->fclone = SKB_FCLONE_CLONE;
-		atomic_inc(&fclones->fclone_ref);
+		/* As our fastclone was free, clone_ref must be 1 at this point.
+		 * We could use atomic_inc() here, but it is faster
+		 * to set the final value.
+		 */
+		atomic_set(&fclones->fclone_ref, 2);
 	} else {
 		if (skb_pfmemalloc(skb))
 			gfp_mask |= __GFP_MEMALLOC;