Skip to contentSkip to conversation

bridge: br_netfilter: give fake rtable its own lifetime

PATCH In review Ren Wei 3 participants 3 messages May 14 – May 26 1 message not in the archive

Subsystems (3)
lore
Share & tools
lore

Ren Wei

author
#1 lore raw

PATCH 1/1 bridge: br_netfilter: give fake rtable its own lifetime

In reply to a message not in the archive

From: Haoze Xie <[email protected]>

The bridge netfilter fake rtable is currently embedded in struct net_bridge even though packets can keep using it after bridge teardown.

Give the fake rtable its own allocated lifetime and make bridge_parent_rtable() return a referenced dst. This way the bridge and any packets that still carry the fake dst each hold their own reference, so bridge teardown no longer leaves a dangling fake dst behind.

We are sending this to the nf tree because the patch touches br_netfilter paths; if the net tree is preferred, please let us know.

Fixes 4adf0af6818f ("bridge: send correct MTU value in PMTU (revised)")
---
 include/net/netfilter/br_netfilter.h | 14 +++++++-
 net/bridge/br_device.c               | 18 ++++++++--
 net/bridge/br_netfilter_hooks.c      |  2 +-
 net/bridge/br_netfilter_ipv6.c       |  2 +-
 net/bridge/br_nf_core.c              | 49 +++++++++++++++++++++++-----
 net/bridge/br_private.h              | 12 +++----
 6 files changed, 76 insertions(+), 21 deletions(-)
include/net/netfilter/br_netfilter.h +13 -1
index 371696ec11b2..ad09dad09da8 100644
@@ -3,6 +3,7 @@
 #define _BR_NETFILTER_H_
 
 #include <linux/netfilter.h>
+#include <net/dst.h>
 
 #include "../../../net/bridge/br_private.h"
 
@@ -44,9 +45,20 @@ static inline struct rtable *bridge_parent_rtable(const struct net_device *dev)
 {
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 	struct net_bridge_port *port;
+	struct rtable *rt;
 
+	rt = NULL;
+	rcu_read_lock();
 	port = br_port_get_rcu(dev);
-	return port ? &port->br->fake_rtable : NULL;
+	if (!port)
+		goto out;
+
+	rt = rcu_dereference(port->br->fake_rtable);
+	if (rt && !dst_hold_safe(&rt->dst))
+		rt = NULL;
+out:
+	rcu_read_unlock();
+	return rt;
 #else
 	return NULL;
 #endif
net/bridge/br_device.c +15 -3
index a35ceae0a6f2..ca2e76929377 100644
@@ -146,6 +146,15 @@ static int br_dev_init(struct net_device *dev)
 		return err;
 	}
 
+	err = br_netfilter_rtable_init(br);
+	if (err) {
+		br_multicast_uninit_stats(br);
+		br_vlan_flush(br);
+		br_mdb_hash_fini(br);
+		br_fdb_hash_fini(br);
+		return err;
+	}
+
 	netdev_lockdep_set_classes(dev);
 	return 0;
 }
@@ -154,6 +163,7 @@ static void br_dev_uninit(struct net_device *dev)
 {
 	struct net_bridge *br = netdev_priv(dev);
 
+	br_netfilter_rtable_fini(br);
 	br_multicast_dev_del(br);
 	br_multicast_uninit_stats(br);
 	br_vlan_flush(br);
@@ -210,8 +220,11 @@ static int br_change_mtu(struct net_device *dev, int new_mtu)
 	/* this flag will be cleared if the MTU was automatically adjusted */
 	br_opt_toggle(br, BROPT_MTU_SET_BY_USER, true);
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
-	/* remember the MTU in the rtable for PMTU */
-	dst_metric_set(&br->fake_rtable.dst, RTAX_MTU, new_mtu);
+	struct rtable *rt;
+
+	rt = rcu_dereference_protected(br->fake_rtable, lockdep_rtnl_is_held());
+	if (rt)
+		dst_metric_set(&rt->dst, RTAX_MTU, new_mtu);
 #endif
 
 	return 0;
@@ -529,7 +542,6 @@ void br_dev_setup(struct net_device *dev)
 	br->bridge_ageing_time = br->ageing_time = BR_DEFAULT_AGEING_TIME;
 	dev->max_mtu = ETH_MAX_MTU;
 
-	br_netfilter_rtable_init(br);
 	br_stp_timer_init(br);
 	br_multicast_init(br);
 	INIT_DELAYED_WORK(&br->gc_work, br_fdb_cleanup);
net/bridge/br_netfilter_hooks.c +1 -1
index 0ab1c94db4b9..8b3b2fb48334 100644
@@ -417,7 +417,7 @@ static int br_nf_pre_routing_finish(struct net *net, struct sock *sk, struct sk_
 			return 0;
 		}
 		skb_dst_drop(skb);
-		skb_dst_set_noref(skb, &rt->dst);
+		skb_dst_set(skb, &rt->dst);
 	}
 
 	skb->dev = br_indev;
net/bridge/br_netfilter_ipv6.c +1 -1
index d8548428929e..4e245645f7e6 100644
@@ -144,7 +144,7 @@ static int br_nf_pre_routing_finish_ipv6(struct net *net, struct sock *sk, struc
 			return 0;
 		}
 		skb_dst_drop(skb);
-		skb_dst_set_noref(skb, &rt->dst);
+		skb_dst_set(skb, &rt->dst);
 	}
 
 	skb->dev = br_indev;
net/bridge/br_nf_core.c +41 -8

77 lines not on this page · raw message

net/bridge/br_private.h +5 -7
index bed1b1d9b282..bb4aa408f232 100644
@@ -508,11 +508,7 @@ struct net_bridge {
 	struct rhashtable		fdb_hash_tbl;
 	struct list_head		port_list;
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
-	union {
-		struct rtable		fake_rtable;
-		struct rt6_info		fake_rt6_info;
-	};
-	u32				metrics[RTAX_MAX];
+	struct rtable __rcu		*fake_rtable;
 #endif
 	u16				group_fwd_mask;
 	u16				group_fwd_mask_required;
@@ -2018,11 +2014,13 @@ extern const struct nf_br_ops __rcu *nf_br_ops;
 #if IS_ENABLED(CONFIG_BRIDGE_NETFILTER)
 int br_nf_core_init(void);
 void br_nf_core_fini(void);
-void br_netfilter_rtable_init(struct net_bridge *);
+int br_netfilter_rtable_init(struct net_bridge *br);
+void br_netfilter_rtable_fini(struct net_bridge *br);
 #else
 static inline int br_nf_core_init(void) { return 0; }
 static inline void br_nf_core_fini(void) {}
-#define br_netfilter_rtable_init(x)
+static inline int br_netfilter_rtable_init(struct net_bridge *br) { return 0; }
+static inline void br_netfilter_rtable_fini(struct net_bridge *br) {}
 #endif
 
 /* br_stp.c */
Signature · 1 line
2.47.3
maintainer
#2 lore raw
Quoted 4 lines · Ren Wei

From: Haoze Xie <[email protected]>

The bridge netfilter fake rtable is currently embedded in struct net_bridge even though packets can keep using it after bridge teardown.

How? Please elaborate a bit, it is unexpected.

Quoted 4 lines

Give the fake rtable its own allocated lifetime and make bridge_parent_rtable() return a referenced dst. This way the bridge and any packets that still carry the fake dst each hold their own reference, so bridge teardown no longer leaves a dangling fake dst behind.

If we have to do this it would be better to move this kludge into br_netfilter.c completely and get rid of the fake rtable hack in bridge for good.

Please also see various AI comments at
https://sashiko.dev/#/patchset/783d76ac83917b7302c1ec647794bd773bb1875a.1778687139.git.royenheart%40gmail.com

[ I would like to zap bridge_netfilter but it seems its too popular ... ]

#3 lore raw
Quoted 8 lines · Florian Westphal
Quoted 4 lines · Ren Wei

From: Haoze Xie <[email protected]>

The bridge netfilter fake rtable is currently embedded in struct net_bridge even though packets can keep using it after bridge teardown.

How? Please elaborate a bit, it is unexpected.

The fake rtable is attached to bridged skbs in the br_netfilter prerouting restore path via bridge_parent_rtable() + skb_dst_set_noref().

If such a packet is queued to NFQUEUE, __nf_queue() upgrades that fake dst with skb_dst_force(). From that point on, the queued skb can hold a real dst reference even after bridge teardown starts freeing the backing struct net_bridge storage. When the verdict path later drops or reinjects the skb, dst_release() can still touch that freed fake dst.

Quoted 9 lines
Quoted 4 lines

Give the fake rtable its own allocated lifetime and make bridge_parent_rtable() return a referenced dst. This way the bridge and any packets that still carry the fake dst each hold their own reference, so bridge teardown no longer leaves a dangling fake dst behind.

If we have to do this it would be better to move this kludge into br_netfilter.c completely and get rid of the fake rtable hack in bridge for good.

I reworked the patch for v2 to keep the lifetime handling in br_netfilter and to address the issues from the previous version.

Instead of adding a separate br_netfilter-private dst lifetime scheme, v2 moves the fake rtable out of struct net_bridge, makes bridge_parent_rtable() return a held dst reference, and switches the callers to skb_dst_set().

To avoid the extra lifetime issues in the previous version, the new fake dst is allocated with rt_dst_alloc(), so it reuses the core IPv4 rtable lifecycle instead of custom br_netfilter dst_ops state. During teardown, br_netfilter first detaches the fake rtable from the bridge, then calls dst_dev_put() before dropping the bridge-owned dst reference.

Quoted 3 lines

Please also see various AI comments at
https://sashiko.dev/#/patchset/783d76ac83917b7302c1ec647794bd773bb1875a.1778687139.git.royenheart%40gmail.com

I also went through the Sashiko comments you pointed out. The v2 change is specifically meant to avoid both the bridge-private storage UAF and the follow-up lifetime problems from the previous approach.

I will send v2.

Thanks,
Haoze

Quoted 1 line

[ I would like to zap bridge_netfilter but it seems its too popular ... ]