Skip to contentSkip to conversation

net act_vlan: use correct len in skb_pull

RFC PATCH In review Zahari Doychev 2 participants 3 messages Feb 13, 2019 – Feb 14, 2019

Subsystems (4)
lore
Share & tools
lore
#0 lore raw

RFC PATCH 1/1 net act_vlan: use correct len in skb_pull

The bridge and VLAN code expects that skb->data points to the start of the VLAN header instead of the next (network) header. Currently after tcf_vlan_act() on ingress filter skb->data points to the next network header. In this case the Linux bridge does not forward correctly double tagged VLAN packets added using tc vlan action as the outer vlan tag from the skb is inserted at the wrong offset after the vlan tag in the payload. Making skb->data to point to the VLAN header in tcf_vlan_act() by using ETH_HLEN in skb_pull_rcsum() fixes the problem.

The following commands were used for testing:

ip link add name br0 type bridge vlan_filtering 1
ip link set dev br0 up

ip link set dev net0 up
ip link set dev net0 master br0

ip link set dev net1 up
ip link set dev net1 master br0

bridge vlan add dev net0 vid 100 master
bridge vlan add dev br0 vid 100 self
bridge vlan add dev net1 vid 100 master

tc qdisc add dev net0 handle ffff: clsact
tc qdisc add dev net1 handle ffff: clsact

tc filter add dev net0 ingress pref 1 protocol all flower \
action vlan push id 10 pipe action vlan push id 100

tc filter add dev net0 egress pref 1 protocol 802.1q flower \
vlan_id 100 vlan_ethtype 802.1q cvlan_id 10 \
action vlan pop pipe action vlan pop

---
 net/sched/act_vlan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/sched/act_vlan.c +1 -1
index 93fdaf707313..308d7d89f925 100644
@@ -86,7 +86,7 @@ static int tcf_vlan_act(struct sk_buff *skb, const struct tc_action *a,
 
 out:
 	if (skb_at_tc_ingress(skb))
-		skb_pull_rcsum(skb, skb->mac_len);
+		skb_pull_rcsum(skb, ETH_HLEN);
 
 	return action;
 
Signature · 1 line
2.20.1
#1 lore raw
Quoted 44 lines · Zahari Doychev

The bridge and VLAN code expects that skb->data points to the start of the VLAN header instead of the next (network) header. Currently after tcf_vlan_act() on ingress filter skb->data points to the next network header. In this case the Linux bridge does not forward correctly double tagged VLAN packets added using tc vlan action as the outer vlan tag from the skb is inserted at the wrong offset after the vlan tag in the payload. Making skb->data to point to the VLAN header in tcf_vlan_act() by using ETH_HLEN in skb_pull_rcsum() fixes the problem.

The following commands were used for testing:

ip link add name br0 type bridge vlan_filtering 1
ip link set dev br0 up

ip link set dev net0 up
ip link set dev net0 master br0

ip link set dev net1 up
ip link set dev net1 master br0

bridge vlan add dev net0 vid 100 master
bridge vlan add dev br0 vid 100 self
bridge vlan add dev net1 vid 100 master

tc qdisc add dev net0 handle ffff: clsact
tc qdisc add dev net1 handle ffff: clsact

tc filter add dev net0 ingress pref 1 protocol all flower \
action vlan push id 10 pipe action vlan push id 100

tc filter add dev net0 egress pref 1 protocol 802.1q flower \
vlan_id 100 vlan_ethtype 802.1q cvlan_id 10 \
action vlan pop pipe action vlan pop

---
 net/sched/act_vlan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/sched/act_vlan.c +0 -0
index 93fdaf707313..308d7d89f925 100644

net/sched/act_vlan.c static int tcf_vlan_act(struct sk_buff *skb, const struct tc_action *a, · line 89

Quoted 5 lines
net/sched/act_vlan.c +1 -1
@@ -86,7 +86,7 @@ static int tcf_vlan_act(struct sk_buff *skb, const struct tc_action *a,
 
 out:
 	if (skb_at_tc_ingress(skb))
-		skb_pull_rcsum(skb, skb->mac_len);
+		skb_pull_rcsum(skb, ETH_HLEN);

As I said before, it would be safer to remember mac_len and use it later.

	__u16 mac_len = skb->mac_len;
...
	err = skb_vlan_push(...)
...
	if (skb_at_tc_ingress(skb))
		skb_pull_rcsum(skb, mac_len);
Signature · 1 line
Toshiaki Makita
#2 lore raw
Quoted 53 lines · Toshiaki Makita
Quoted 44 lines · Zahari Doychev

The bridge and VLAN code expects that skb->data points to the start of the VLAN header instead of the next (network) header. Currently after tcf_vlan_act() on ingress filter skb->data points to the next network header. In this case the Linux bridge does not forward correctly double tagged VLAN packets added using tc vlan action as the outer vlan tag from the skb is inserted at the wrong offset after the vlan tag in the payload. Making skb->data to point to the VLAN header in tcf_vlan_act() by using ETH_HLEN in skb_pull_rcsum() fixes the problem.

The following commands were used for testing:

ip link add name br0 type bridge vlan_filtering 1
ip link set dev br0 up

ip link set dev net0 up
ip link set dev net0 master br0

ip link set dev net1 up
ip link set dev net1 master br0

bridge vlan add dev net0 vid 100 master
bridge vlan add dev br0 vid 100 self
bridge vlan add dev net1 vid 100 master

tc qdisc add dev net0 handle ffff: clsact
tc qdisc add dev net1 handle ffff: clsact

tc filter add dev net0 ingress pref 1 protocol all flower \
action vlan push id 10 pipe action vlan push id 100

tc filter add dev net0 egress pref 1 protocol 802.1q flower \
vlan_id 100 vlan_ethtype 802.1q cvlan_id 10 \
action vlan pop pipe action vlan pop

---
 net/sched/act_vlan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)
net/sched/act_vlan.c +0 -0
index 93fdaf707313..308d7d89f925 100644

net/sched/act_vlan.c static int tcf_vlan_act(struct sk_buff *skb, const struct tc_action *a, · line 89

Quoted 5 lines
net/sched/act_vlan.c +1 -1
@@ -86,7 +86,7 @@ static int tcf_vlan_act(struct sk_buff *skb, const struct tc_action *a,
 
 out:
 	if (skb_at_tc_ingress(skb))
-		skb_pull_rcsum(skb, skb->mac_len);
+		skb_pull_rcsum(skb, ETH_HLEN);

As I said before, it would be safer to remember mac_len and use it later.

Quoted 7 lines
	__u16 mac_len = skb->mac_len;
...
	err = skb_vlan_push(...)
...
	if (skb_at_tc_ingress(skb))
		skb_pull_rcsum(skb, mac_len);

sorry, I misunderstood it. I will send an update.

Zahari

Quoted 4 lines
Signature · 2 lines
-- 
Toshiaki Makita