Skip to contentSkip to conversation

bridge-netfilter: use the vlan id as part of the connection tracking tuple for bridged traffic

PATCH In review Bart De Schuymer 4 participants 6 messages Mar 30, 2010 – Apr 1, 2010

Subsystems (2)
lore
Share & tools
lore
#0 lore raw

PATCH 5/5 bridge-netfilter: use the vlan id as part of the connection tracking tuple for bridged traffic

bridge-netfilter: use the vlan id as part of the connection tracking tuple for bridged traffic

Currently connection tracking isn't aware of the vlan id that bridged encapsulated IP packets belong to. Since it's possible to enable filtering of bridged IP traffic encapsulated in a vlan header, we have a security risk if an overlapping IP network range is used on different vlan networks. By making the connection tracking code use the vlan id as part of the tuple, we eliminate this security risk. Something similar could be done for IP traffic encapsulated in PPPoE, but I'm not sure it's worth the effort.

net/netfilter/nf_conntrack_core.c +9 -0
@@ -32,6 +32,7 @@
 #include <linux/mm.h>
 #include <linux/nsproxy.h>
 #include <linux/rculist_nulls.h>
+#include <linux/netfilter_bridge.h>
 
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_l3proto.h>
@@ -110,6 +111,11 @@ nf_ct_get_tuple(const struct sk_buff *sk
 
 	tuple->dst.protonum = protonum;
 	tuple->dst.dir = IP_CT_DIR_ORIGINAL;
+#ifdef CONFIG_BRIDGE_NETFILTER
+	/* +1 because 0 identifies no vlan encapsulation */
+	if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_8021Q))
+		tuple->dst.vlan_id = (ntohs(*(u_int16_t *)(skb->data - VLAN_HLEN)) & VLAN_VID_MASK) + 1;
+#endif
 
 	return l4proto->pkt_to_tuple(skb, dataoff, tuple);
 }
@@ -158,6 +164,9 @@ nf_ct_invert_tuple(struct nf_conntrack_t
 	inverse->dst.dir = !orig->dst.dir;
 
 	inverse->dst.protonum = orig->dst.protonum;
+#ifdef CONFIG_BRIDGE_NETFILTER
+	inverse->dst.vlan_id = orig->dst.vlan_id;
+#endif
 	return l4proto->invert_tuple(inverse, orig);
 }
 EXPORT_SYMBOL_GPL(nf_ct_invert_tuple);
include/net/netfilter/nf_conntrack_tuple.h +8 -1
@@ -94,6 +94,9 @@ struct nf_conntrack_tuple {
 
 		/* The direction (for tuplehash) */
 		u_int8_t dir;
+#ifdef CONFIG_BRIDGE_NETFILTER
+		u_int16_t vlan_id;
+#endif
 	} dst;
 };
 
@@ -163,7 +166,11 @@ static inline bool __nf_ct_tuple_dst_equ
 {
 	return (nf_inet_addr_cmp(&t1->dst.u3, &t2->dst.u3) &&
 		t1->dst.u.all == t2->dst.u.all &&
-		t1->dst.protonum == t2->dst.protonum);
+		t1->dst.protonum == t2->dst.protonum
+#ifdef CONFIG_BRIDGE_NETFILTER
+		&& likely(t1->dst.vlan_id == t2->dst.vlan_id)
+#endif
+		);
 }
 
 static inline bool nf_ct_tuple_equal(const struct nf_conntrack_tuple *t1,
Signature · 2 lines
Bart De Schuymer
www.artinalgorithms.be
#1 lore raw
Quoted 61 lines · Bart De Schuymer

bridge-netfilter: use the vlan id as part of the connection tracking tuple for bridged traffic

Currently connection tracking isn't aware of the vlan id that bridged encapsulated IP packets belong to. Since it's possible to enable filtering of bridged IP traffic encapsulated in a vlan header, we have a security risk if an overlapping IP network range is used on different vlan networks. By making the connection tracking code use the vlan id as part of the tuple, we eliminate this security risk. Something similar could be done for IP traffic encapsulated in PPPoE, but I'm not sure it's worth the effort.

net/netfilter/nf_conntrack_core.c +9 -0
@@ -32,6 +32,7 @@
 #include <linux/mm.h>
 #include <linux/nsproxy.h>
 #include <linux/rculist_nulls.h>
+#include <linux/netfilter_bridge.h>
 
 #include <net/netfilter/nf_conntrack.h>
 #include <net/netfilter/nf_conntrack_l3proto.h>
@@ -110,6 +111,11 @@ nf_ct_get_tuple(const struct sk_buff *sk
 
 	tuple->dst.protonum = protonum;
 	tuple->dst.dir = IP_CT_DIR_ORIGINAL;
+#ifdef CONFIG_BRIDGE_NETFILTER
+	/* +1 because 0 identifies no vlan encapsulation */
+	if (skb->nf_bridge && (skb->nf_bridge->mask & BRNF_8021Q))
+		tuple->dst.vlan_id = (ntohs(*(u_int16_t *)(skb->data - VLAN_HLEN)) & VLAN_VID_MASK) + 1;
+#endif
 
 	return l4proto->pkt_to_tuple(skb, dataoff, tuple);
 }
@@ -158,6 +164,9 @@ nf_ct_invert_tuple(struct nf_conntrack_t
 	inverse->dst.dir = !orig->dst.dir;
 
 	inverse->dst.protonum = orig->dst.protonum;
+#ifdef CONFIG_BRIDGE_NETFILTER
+	inverse->dst.vlan_id = orig->dst.vlan_id;
+#endif
 	return l4proto->invert_tuple(inverse, orig);
 }
 EXPORT_SYMBOL_GPL(nf_ct_invert_tuple);
include/net/netfilter/nf_conntrack_tuple.h +3 -0
@@ -94,6 +94,9 @@ struct nf_conntrack_tuple {
 
 		/* The direction (for tuplehash) */
 		u_int8_t dir;
+#ifdef CONFIG_BRIDGE_NETFILTER
+		u_int16_t vlan_id;
+#endif
 	} dst;
 };
 
@@ -163,7 +166,11 @@ static inline bool __nf_ct_tuple_dst_equ
 {
 	return (nf_inet_addr_cmp(&t1->dst.u3, &t2->dst.u3) &&

include/net/netfilter/nf_conntrack_tuple.h static inline bool __nf_ct_tuple_dst_equ · line 176

Quoted 10 lines
include/net/netfilter/nf_conntrack_tuple.h +5 -1
 		t1->dst.u.all == t2->dst.u.all &&
-		t1->dst.protonum == t2->dst.protonum);
+		t1->dst.protonum == t2->dst.protonum
+#ifdef CONFIG_BRIDGE_NETFILTER
+		&& likely(t1->dst.vlan_id == t2->dst.vlan_id)
+#endif
+		);
 }
 
 static inline bool nf_ct_tuple_equal(const struct nf_conntrack_tuple *t1,

This really sounds very strange, layering violation or something.

You mix conntracking, bridge and vlan here.

Why setups without bridge should not care of vlan + conntracking side effects ?

This whole idea was discussed last November :

http://www.spinics.net/lists/netfilter-devel/msg10692.html

Patrick spoke of 'conntrack zone', and we added this concept.

commit 5d0aa2ccd4699a01cfdf14886191c249d7b45a01

Author: Patrick McHardy <[email protected]>
Date:   Mon Feb 15 18:13:33 2010 +0100

netfilter: nf_conntrack: add support for "conntrack zones"

Normally, each connection needs a unique identity. Conntrack zones allow to specify a numerical zone using the CT target, connections in different zones can use the same identity.

Example:

iptables -t raw -A PREROUTING -i veth0 -j CT --zone 1 iptables -t raw -A OUTPUT -o veth1 -j CT --zone 1

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#2 lore raw
Quoted 13 lines · Eric Dumazet
Quoted 12 lines · Bart De Schuymer
include/net/netfilter/nf_conntrack_tuple.h +5 -1
@@ -163,7 +166,11 @@ static inline bool __nf_ct_tuple_dst_equ
 {
 	return (nf_inet_addr_cmp(&t1->dst.u3, &t2->dst.u3) &&
 		t1->dst.u.all == t2->dst.u.all &&
-		t1->dst.protonum == t2->dst.protonum);
+		t1->dst.protonum == t2->dst.protonum
+#ifdef CONFIG_BRIDGE_NETFILTER
+		&& likely(t1->dst.vlan_id == t2->dst.vlan_id)
+#endif
+		);
 }
 
Quoted 6 lines

include/net/netfilter/nf_conntrack_tuple.h static inline bool __nf_ct_tuple_dst_equ · line 176

Quoted 2 lines
include/net/netfilter/nf_conntrack_tuple.h +0 -0
 static inline bool nf_ct_tuple_equal(const struct nf_conntrack_tuple *t1,

This really sounds very strange, layering violation or something.

You mix conntracking, bridge and vlan here.

I agree, this is really wrong.

Quoted 8 lines

Why setups without bridge should not care of vlan + conntracking side effects ?

This whole idea was discussed last November :

http://www.spinics.net/lists/netfilter-devel/msg10692.html

Patrick spoke of 'conntrack zone', and we added this concept.

Indeed, this seems like a better way.
--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html

#3 lore raw
Quoted 21 lines · Patrick McHardy
Quoted 19 lines · Eric Dumazet
Quoted 5 lines · Bart De Schuymer
include/net/netfilter/nf_conntrack_tuple.h +0 -0
@@ -163,7 +166,11 @@ static inline bool __nf_ct_tuple_dst_equ
 {
 	return (nf_inet_addr_cmp(&t1->dst.u3, &t2->dst.u3) &&

include/net/netfilter/nf_conntrack_tuple.h static inline bool __nf_ct_tuple_dst_equ · line 176

Quoted 10 lines
include/net/netfilter/nf_conntrack_tuple.h +5 -1
 		t1->dst.u.all == t2->dst.u.all &&
-		t1->dst.protonum == t2->dst.protonum);
+		t1->dst.protonum == t2->dst.protonum
+#ifdef CONFIG_BRIDGE_NETFILTER
+		&& likely(t1->dst.vlan_id == t2->dst.vlan_id)
+#endif
+		);
 }
 
 static inline bool nf_ct_tuple_equal(const struct nf_conntrack_tuple *t1,

This really sounds very strange, layering violation or something.

Quoted 5 lines
Quoted 2 lines

You mix conntracking, bridge and vlan here.

I agree, this is really wrong.

Using the conntrack zone workaround to achieve what my patch does is basically doing the same thing, IMHO.
But it's indeed much cleaner to use the generic CT target scheme, which also solves the "multiple bridges" scenario mentioned by Pascal. I wasn't yet aware of this target.

I've tested the following setup with 2.6.34-rc3 and it successfully separates the networks (without my vlan patch).

# set up the connection tracking zones
iptables -t raw -A PREROUTING -m mark --mark 1 -j CT --zone 1
iptables -t raw -A PREROUTING -m mark --mark 2 -j CT --zone 2
# mark packets according to the vlan id
ebtables -t nat -A PREROUTING -p 802_1Q --vlan-id 1 -j mark --mark-set 1
ebtables -t nat -A PREROUTING -p 802_1Q --vlan-id 5 -j mark --mark-set 2

So this is good news. The security risk is solvable starting from v2.6.34. I'll need to mention this clearly in the documentation.

What about the other patches? I'm aware they don't all cleanly patch versus the most recent kernel, but do you have any objections apart from that?

cheers,
Bart

Signature · 7 lines
Bart De Schuymer
www.artinalgorithms.be

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in
the body of a message to [email protected]
More majordomo info at  http://vger.kernel.org/majordomo-info.html
#4 lore raw
Quoted 25 lines · Bart De Schuymer
Quoted 8 lines · Patrick McHardy
Quoted 5 lines · Eric Dumazet
Quoted 1 line

This really sounds very strange, layering violation or something.

You mix conntracking, bridge and vlan here.

I agree, this is really wrong.

Using the conntrack zone workaround to achieve what my patch does is basically doing the same thing, IMHO.
But it's indeed much cleaner to use the generic CT target scheme, which also solves the "multiple bridges" scenario mentioned by Pascal. I wasn't yet aware of this target.

I've tested the following setup with 2.6.34-rc3 and it successfully separates the networks (without my vlan patch).

# set up the connection tracking zones
iptables -t raw -A PREROUTING -m mark --mark 1 -j CT --zone 1
iptables -t raw -A PREROUTING -m mark --mark 2 -j CT --zone 2
# mark packets according to the vlan id
ebtables -t nat -A PREROUTING -p 802_1Q --vlan-id 1 -j mark --mark-set 1
ebtables -t nat -A PREROUTING -p 802_1Q --vlan-id 5 -j mark --mark-set 2
Quoted 2 lines

So this is good news. The security risk is solvable starting from v2.6.34. I'll need to mention this clearly in the documentation.

Great.

Quoted 3 lines

What about the other patches? I'm aware they don't all cleanly patch versus the most recent kernel, but do you have any objections apart from that?

I haven't fully reviewed them yet, but that should happen soon.

#5 lore raw
In reply to Eric Dumazet
Quoted 2 lines · Eric Dumazet

This really sounds very strange, layering violation or something.

Isn't the whole bridge-netfilter concept already a layering violation by design ?

Quoted 4 lines

You mix conntracking, bridge and vlan here.

Why setups without bridge should not care of vlan + conntracking side effects ?

Because without bridge, the host is attached at the IP layer level to the VLANs, so their IP ranges are not supposed to overlap.

Anyway your objection applies to hosts with multiple bridges without VLAN so the bridges may see overlapping IP ranges. Conntrack zones with a dedicated target seems a more generic approach.

--
To unsubscribe from this list: send the line "unsubscribe netfilter-devel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html