[RFC] vlan: mask vlan prio bits

[RFC] vlan: mask vlan prio bits

[Eric Dumazet]

In commit 48cc32d38a52d0b68f91a171a8d00531edc6a46e
("vlan: don't deliver frames for unknown vlans to protocols")
Florian made sure we set PACKET_OTHERHOST if the vlan id is set
and we could find a vlan device for this particular id.

But we also have a problem if prio bits are set.

Steinar has an issue on a router receiving IPv6 frames with a vlan tag
of 4000 (prio = 2), and tunneled into a sit device, because
skb->vlan_tci is set :

We insert a VLAN tag in forwarded frame after sit tunnel.

Frame is completely corrupted : We can see (8100:4000) being inserted in
the middle of IPv6 source address :

16:48:00.780413 IP6 2001:16d8:8100:4000:ee1c:0:9d9:bc87 >
9f94:4d95:2001:67c:29f4::: ICMP6, unknown icmp6 type (0), length 64
       0x0000:  0000 0029 8000 c7c3 7103 0001 a0ae e651
       0x0010:  0000 0000 ccce 0b00 0000 0000 1011 1213
       0x0020:  1415 1617 1819 1a1b 1c1d 1e1f 2021 2223
       0x0030:  2425 2627 2829 2a2b 2c2d 2e2f 3031 3233

It seems we are not really ready to properly cope with this right now.

Should we clear skb->vlan_tci in iptunnel_xmit(), or should we clear it
in __netif_receive_skb_core() like in this patch ?

diff --git a/include/linux/if_vlan.h b/include/linux/if_vlan.h
index cdcbafa..715c343 100644
--- a/include/linux/if_vlan.h
+++ b/include/linux/if_vlan.h
@@ -79,9 +79,8 @@ static inline int is_vlan_dev(struct net_device *dev)
 }
 
 #define vlan_tx_tag_present(__skb)	((__skb)->vlan_tci & VLAN_TAG_PRESENT)
-#define vlan_tx_nonzero_tag_present(__skb) \
-	(vlan_tx_tag_present(__skb) && ((__skb)->vlan_tci & VLAN_VID_MASK))
 #define vlan_tx_tag_get(__skb)		((__skb)->vlan_tci & ~VLAN_TAG_PRESENT)
+#define vlan_tx_tag_get_id(__skb)	((__skb)->vlan_tci & VLAN_VID_MASK)
 
 #if defined(CONFIG_VLAN_8021Q) || defined(CONFIG_VLAN_8021Q_MODULE)
 
diff --git a/net/8021q/vlan_core.c b/net/8021q/vlan_core.c
index 8a15eaa..4a78c4d 100644
--- a/net/8021q/vlan_core.c
+++ b/net/8021q/vlan_core.c
@@ -9,7 +9,7 @@ bool vlan_do_receive(struct sk_buff **skbp)
 {
 	struct sk_buff *skb = *skbp;
 	__be16 vlan_proto = skb->vlan_proto;
-	u16 vlan_id = skb->vlan_tci & VLAN_VID_MASK;
+	u16 vlan_id = vlan_tx_tag_get_id(skb);
 	struct net_device *vlan_dev;
 	struct vlan_pcpu_stats *rx_stats;
 
diff --git a/net/core/dev.c b/net/core/dev.c
index a3d8d44..133b75a 100644
--- a/net/core/dev.c
+++ b/net/core/dev.c
@@ -3580,8 +3580,11 @@ ncls:
 		}
 	}
 
-	if (vlan_tx_nonzero_tag_present(skb))
-		skb->pkt_type = PACKET_OTHERHOST;
+	if (vlan_tx_tag_present(skb)) {
+		if (vlan_tx_tag_get_id(skb))
+			skb->pkt_type = PACKET_OTHERHOST;
+		skb->vlan_tci = 0;
+	}
 
 	/* deliver only exact match when indicated */
 	null_or_dev = deliver_exact ? skb->dev : NULL;

Re: [RFC] vlan: mask vlan prio bits

[David Miller]

From: Eric Dumazet <eric.dumazet@gmail.com>
Date: Wed, 17 Jul 2013 10:44:34 -0700

> Should we clear skb->vlan_tci in iptunnel_xmit(), or should we clear it
> in __netif_receive_skb_core() like in this patch ?

The safest thing to do is as early as possible, so yes your patch here
doing it in __netif_receive_skb_core() looks good to me.

Re: [RFC] vlan: mask vlan prio bits

[Eric Dumazet]

On Wed, 2013-07-17 at 17:58 -0700, David Miller wrote:
> From: Eric Dumazet <eric.dumazet@gmail.com>
> Date: Wed, 17 Jul 2013 10:44:34 -0700
> 
> > Should we clear skb->vlan_tci in iptunnel_xmit(), or should we clear it
> > in __netif_receive_skb_core() like in this patch ?
> 
> The safest thing to do is as early as possible, so yes your patch here
> doing it in __netif_receive_skb_core() looks good to me.

OK then, I'll send an official patch ;)

Thanks !