Re: [PATCH][net-next] net: slightly optimize eth_type_trans

[Eric Dumazet]

On 11/15/2018 03:11 PM, David Miller wrote:

> 
> Applied.
> 

While reviewing this stuff, I found we have a bug.

If napi_reuse_skb() is called, we might inherit from prior skb->pkt_type value.

It seems that GRO could aggregate packets with pkt_type != PACKET_HOST, right ?

David, any objection if I submit the following fix ?

diff --git a/net/core/dev.c b/net/core/dev.c
index 
5927f6a7c301ed90af21a4b82b443f30f00bb483..f2bfd2eda7b2734d29d30f0e82c1a48c1b5b166a
 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -5664,6 +5664,10 @@ static void napi_reuse_skb(struct napi_struct *napi, 
struct sk_buff *skb)
        __vlan_hwaccel_clear_tag(skb);
        skb->dev = napi->dev;
        skb->skb_iif = 0;
+
+       /* eth_type_trans() assumes initial pkt_type is PACKET_HOST */
+       skb->pkt_type = PACKET_HOST;
+
        skb->encapsulation = 0;
        skb_shinfo(skb)->gso_type = 0;
        skb->truesize = SKB_TRUESIZE(skb_end_offset(skb));