[PATCH net v2 0/2] bridge/vxlan: fix reading neigh ha without synchronization

[PATCH net v2 0/2] bridge/vxlan: fix reading neigh ha without synchronization

[Nikolay Aleksandrov]

Hi,
Neigh ha address must be read using the seqlock to get a stable snapshot.
Both the bridge and vxlan read it directly and can see partial updates.
I reproduced both issues with running neigh updates and exercising these
paths in parallel and saw partial addresses, e.g. updating between
neigh A: 02:00:00:00:00:00 neigh B: fe:ff:ff:ff:ff:ff was able to observe
02:00:ff:ff:ff:ff and fe:ff:00:00:00:00 in packets. Noticed this initially
in the bridge, then checked vxlan and its arp/neigh_reduce functions have
the same bug, route_shortcircuit is doing the right thing already.

v1 link: https://lore.kernel.org/netdev/20260817143613.685769-1-razor@blackwall.org/
v2: - use ETH_ALEN instead of MAX_ADDR_LEN, the bridge devices all use
      ETH_ALEN and vxlan allows arp/nd reduce only when not in raw/gpe
      so it also always uses ETH_ALEN
    - align ha to 2 bytes because ether_addr_copy() expects it (Sashiko)

2-byte alignment is not strictly necessary everywhere (e.g. the ARP
suppress side can't reach ether_addr_copy) but it doesn't cost us anything
and is aligned with the rest of the code.

Cheers,
 Nik

Nikolay Aleksandrov (2):
  net: bridge: arp/nd proxy: fix reading neigh ha
  vxlan: fix reading neigh ha

 drivers/net/vxlan/vxlan_core.c | 20 +++++++++++++-------
 net/bridge/br_arp_nd_proxy.c   | 24 ++++++++++++++----------
 2 files changed, 27 insertions(+), 17 deletions(-)

-- 
2.47.3

[PATCH net v2 1/2] net: bridge: arp/nd proxy: fix reading neigh ha

[Nikolay Aleksandrov]

Currently neigh ha address is read directly, but that can result in
torn/partial reads if the neigh is being updated. Use neigh_ha_snapshot
to take a stable snapshot of the address.

Fixes: 057658cb33fb ("bridge: suppress arp pkts on BR_NEIGH_SUPPRESS ports")
Fixes: ed842faeb2bd ("bridge: suppress nd pkts on BR_NEIGH_SUPPRESS ports")
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
---
v2: - use ETH_ALEN instead of MAX_ADDR_LEN, the bridge devices all use
      ETH_ALEN and vxlan allows arp/nd reduce only when not in raw/gpe
      so it also always uses ETH_ALEN
    - align ha to 2 bytes because ether_addr_copy() expects it (Sashiko)

 net/bridge/br_arp_nd_proxy.c | 24 ++++++++++++++----------
 1 file changed, 14 insertions(+), 10 deletions(-)

diff --git a/net/bridge/br_arp_nd_proxy.c b/net/bridge/br_arp_nd_proxy.c
index 23eb6931a2b4..fd65d6af4f35 100644
--- a/net/bridge/br_arp_nd_proxy.c
+++ b/net/bridge/br_arp_nd_proxy.c
@@ -195,13 +195,15 @@ void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
 	n = neigh_lookup(&arp_tbl, &tip, vlandev);
 	if (n) {
 		struct net_bridge_fdb_entry *f;
+		u8 ha[ETH_ALEN] __aligned(2);
 
 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
 			neigh_release(n);
 			return;
 		}
 
-		f = br_fdb_find_rcu(br, n->ha, vid);
+		neigh_ha_snapshot(ha, n, n->dev);
+		f = br_fdb_find_rcu(br, ha, vid);
 		if (f) {
 			const struct net_bridge_port *dst = READ_ONCE(f->dst);
 			bool replied = false;
@@ -211,10 +213,10 @@ void br_do_proxy_suppress_arp(struct sk_buff *skb, struct net_bridge *br,
 			    br_is_neigh_suppress_enabled(dst, vid)) {
 				if (!vid)
 					br_arp_send(br, p, skb->dev, sip, tip,
-						    sha, n->ha, sha, 0, 0);
+						    sha, ha, sha, 0, 0);
 				else
 					br_arp_send(br, p, skb->dev, sip, tip,
-						    sha, n->ha, sha,
+						    sha, ha, sha,
 						    skb->vlan_proto,
 						    skb_vlan_tag_get(skb));
 				replied = true;
@@ -252,7 +254,7 @@ struct nd_msg *br_is_nd_neigh_msg(const struct sk_buff *skb, struct nd_msg *msg)
 }
 
 static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
-		       struct sk_buff *request, struct neighbour *n,
+		       struct sk_buff *request, struct neighbour *n, u8 *ha,
 		       __be16 vlan_proto, u16 vlan_tci)
 {
 	struct net_device *dev = request->dev;
@@ -310,7 +312,7 @@ static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
 		ipv6_eth_mc_map(&in6addr_linklocal_allnodes, eth_hdr(reply)->h_dest);
 	else
 		ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
-	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
+	ether_addr_copy(eth_hdr(reply)->h_source, ha);
 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
 	reply->protocol = htons(ETH_P_IPV6);
 
@@ -340,7 +342,7 @@ static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
 	na->icmph.icmp6_override = 1;
 	na->icmph.icmp6_solicited = dad ? 0 : 1;
 	na->target = ns->target;
-	ether_addr_copy(&na->opt[2], n->ha);
+	ether_addr_copy(&na->opt[2], ha);
 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
 	na->opt[1] = na_olen >> 3;
 
@@ -369,7 +371,7 @@ static void br_nd_send(struct net_bridge *br, struct net_bridge_port *p,
 		__vlan_hwaccel_put_tag(reply, vlan_proto, vlan_tci);
 
 	netdev_dbg(dev, "nd send dev %s dst %pI6 dst_hw %pM src %pI6 src_hw %pM\n",
-		   dev->name, &pip6->daddr, daddr, &pip6->saddr, n->ha);
+		   dev->name, &pip6->daddr, daddr, &pip6->saddr, ha);
 
 	if (p) {
 		dev_queue_xmit(reply);
@@ -472,24 +474,26 @@ void br_do_suppress_nd(struct sk_buff *skb, struct net_bridge *br,
 	n = neigh_lookup(&nd_tbl, &msg->target, vlandev);
 	if (n) {
 		struct net_bridge_fdb_entry *f;
+		u8 ha[ETH_ALEN] __aligned(2);
 
 		if (!(READ_ONCE(n->nud_state) & NUD_VALID)) {
 			neigh_release(n);
 			return;
 		}
 
-		f = br_fdb_find_rcu(br, n->ha, vid);
+		neigh_ha_snapshot(ha, n, n->dev);
+		f = br_fdb_find_rcu(br, ha, vid);
 		if (f) {
 			const struct net_bridge_port *dst = READ_ONCE(f->dst);
 			bool replied = false;
 
 			if (br_is_neigh_suppress_enabled(dst, vid)) {
 				if (vid != 0)
-					br_nd_send(br, p, skb, n,
+					br_nd_send(br, p, skb, n, ha,
 						   skb->vlan_proto,
 						   skb_vlan_tag_get(skb));
 				else
-					br_nd_send(br, p, skb, n, 0, 0);
+					br_nd_send(br, p, skb, n, ha, 0, 0);
 				replied = true;
 			}
 
-- 
2.47.3

[PATCH net v2 2/2] vxlan: fix reading neigh ha

[Nikolay Aleksandrov]

Currently arp/neigh_reduce read neigh ha directly which can lead to
partial reads while the neigh is being updated. Use neigh_ha_snapshot to
take a stable snapshot of the address similar to route_shortcircuit which
already does the right thing.

Fixes: e4f67addf158 ("add DOVE extensions for VXLAN")
Fixes: f564f45c4518 ("vxlan: add ipv6 proxy support")
Signed-off-by: Nikolay Aleksandrov <razor@blackwall.org>
---
v2: - use ETH_ALEN instead of MAX_ADDR_LEN, the bridge devices all use
      ETH_ALEN and vxlan allows arp/nd reduce only when not in raw/gpe
      so it also always uses ETH_ALEN
    - align ha to 2 bytes because ether_addr_copy() expects it (Sashiko)

 drivers/net/vxlan/vxlan_core.c | 20 +++++++++++++-------
 1 file changed, 13 insertions(+), 7 deletions(-)

diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c
index 824144bb7774..a94168f7a18e 100644
--- a/drivers/net/vxlan/vxlan_core.c
+++ b/drivers/net/vxlan/vxlan_core.c
@@ -1881,6 +1881,7 @@ static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 
 	if (n) {
 		struct vxlan_rdst *rdst = NULL;
+		u8 ha[ETH_ALEN] __aligned(2);
 		struct vxlan_fdb *f;
 		struct sk_buff	*reply;
 
@@ -1889,8 +1890,10 @@ static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 			goto out;
 		}
 
+		neigh_ha_snapshot(ha, n, n->dev);
+
 		rcu_read_lock();
-		f = vxlan_find_mac_tx(vxlan, n->ha, vni);
+		f = vxlan_find_mac_tx(vxlan, ha, vni);
 		if (f)
 			rdst = first_remote_rcu(f);
 		if (rdst && vxlan_addr_any(&rdst->remote_ip)) {
@@ -1902,7 +1905,7 @@ static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 		rcu_read_unlock();
 
 		reply = arp_create(ARPOP_REPLY, ETH_P_ARP, sip, dev, tip, sha,
-				n->ha, sha);
+				   ha, sha);
 
 		neigh_release(n);
 
@@ -1935,7 +1938,8 @@ static int arp_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 
 #if IS_ENABLED(CONFIG_IPV6)
 static struct sk_buff *vxlan_na_create(struct sk_buff *request,
-	struct neighbour *n, bool isrouter)
+				       struct neighbour *n, u8 *ha,
+				       bool isrouter)
 {
 	struct net_device *dev = request->dev;
 	struct sk_buff *reply;
@@ -1981,7 +1985,7 @@ static struct sk_buff *vxlan_na_create(struct sk_buff *request,
 
 	/* Ethernet header */
 	ether_addr_copy(eth_hdr(reply)->h_dest, daddr);
-	ether_addr_copy(eth_hdr(reply)->h_source, n->ha);
+	ether_addr_copy(eth_hdr(reply)->h_source, ha);
 	eth_hdr(reply)->h_proto = htons(ETH_P_IPV6);
 	reply->protocol = htons(ETH_P_IPV6);
 
@@ -2010,7 +2014,7 @@ static struct sk_buff *vxlan_na_create(struct sk_buff *request,
 	na->icmph.icmp6_override = 1;
 	na->icmph.icmp6_solicited = 1;
 	na->target = ns->target;
-	ether_addr_copy(&na->opt[2], n->ha);
+	ether_addr_copy(&na->opt[2], ha);
 	na->opt[0] = ND_OPT_TARGET_LL_ADDR;
 	na->opt[1] = na_olen >> 3;
 
@@ -2051,6 +2055,7 @@ static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 
 	if (n) {
 		struct vxlan_rdst *rdst = NULL;
+		u8 ha[ETH_ALEN] __aligned(2);
 		struct vxlan_fdb *f;
 		struct sk_buff *reply;
 
@@ -2059,7 +2064,8 @@ static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 			goto out;
 		}
 
-		f = vxlan_find_mac_tx(vxlan, n->ha, vni);
+		neigh_ha_snapshot(ha, n, n->dev);
+		f = vxlan_find_mac_tx(vxlan, ha, vni);
 		if (f)
 			rdst = first_remote_rcu(f);
 		if (rdst && vxlan_addr_any(&rdst->remote_ip)) {
@@ -2068,7 +2074,7 @@ static int neigh_reduce(struct net_device *dev, struct sk_buff *skb, __be32 vni)
 			goto out;
 		}
 
-		reply = vxlan_na_create(skb, n,
+		reply = vxlan_na_create(skb, n, ha,
 					!!(f ? f->flags & NTF_ROUTER : 0));
 
 		neigh_release(n);
-- 
2.47.3