[PATCH] ethernet: call __skb_pull() in eth_type_trans()

[PATCH] ethernet: call __skb_pull() in eth_type_trans()

[Changli Gao]

call __skb_pull() in eth_type_trans().

Since the callers of eth_type_trans() always feed it long enough packets,
we can use __skb_pull() to save some cycles.

Signed-off-by: Changli Gao <xiaosuo@gmail.com>
----
 net/ethernet/eth.c |    2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
diff --git a/net/ethernet/eth.c b/net/ethernet/eth.c
index 61ec032..aacbaf7 100644
--- a/net/ethernet/eth.c
+++ b/net/ethernet/eth.c
@@ -162,7 +162,7 @@ __be16 eth_type_trans(struct sk_buff *skb, struct net_device *dev)
 
 	skb->dev = dev;
 	skb_reset_mac_header(skb);
-	skb_pull_inline(skb, ETH_HLEN);
+	__skb_pull(skb, ETH_HLEN);
 	eth = eth_hdr(skb);
 
 	if (unlikely(is_multicast_ether_addr(eth->h_dest))) {

Re: [PATCH] ethernet: call __skb_pull() in eth_type_trans()

[David Miller]

From: Changli Gao <xiaosuo@gmail.com>
Date: Mon,  3 May 2010 06:50:17 +0800

> call __skb_pull() in eth_type_trans().
> 
> Since the callers of eth_type_trans() always feed it long enough packets,
> we can use __skb_pull() to save some cycles.
> 
> Signed-off-by: Changli Gao <xiaosuo@gmail.com>

Although they should, assuming that some runts won't show up here and
never check for that condition at all is dangerous.

At least we should have a WARN_ON() check for skb->len < ETH_ZLEN here
or similar.

So many other things get layered into ethernet, which means adding
this length assumption without any checks is bound to lead to
unpleasant surprises for somebody.