* [PATCH nf,v6 2/3] netfilter: flowtable: cache destination hw address for XMIT_NEIGH type
2026-11-20 22:52 [PATCH nf,v6 1/3] netfilter: flowtable: remove redundant ifidx field in tuple output union Pablo Neira Ayuso
@ 2026-11-20 22:52 ` Pablo Neira Ayuso
2026-11-20 22:52 ` [PATCH nf,v6 3/3] netfilter: flowtable: unify direct and neigh xmit path Pablo Neira Ayuso
1 sibling, 0 replies; 3+ messages in thread
From: Pablo Neira Ayuso @ 2026-11-20 22:52 UTC (permalink / raw)
To: netfilter-devel
Since commit b5964aac51e0 ("netfilter: flowtable: consolidate xmit
path"), a flow offloaded with FLOW_OFFLOAD_XMIT_NEIGH keeps being
forwarded with destination MAC 00:00:00:00:00:00 after the next hop's
neighbour entry has been removed. Nothing resolves the next hop again,
so the flow stays black-holed for as long as it keeps sending.
Cache the destination hw address in the flow entry, just like
XMIT_DIRECT does. Use the flowtable GC to validate that there is a
neighbour entry that is still valid. Otherwise, if the neighbour entry
is stale, trigger a probe.
Skip check the stale neighbour check for non-Ethernet devices such as
WireGuard or GRE by checking for all zero destination hw address,
otherwise the flow entries are incorrectly torn down.
Fixes: b5964aac51e0 ("netfilter: flowtable: consolidate xmit path")
Signed-off-by: Pablo Neira Ayuso <pablo@netfilter.org>
---
v6: skip check for stale neighbout entries in case non-ethernet device
is use in the flowtable tx path.
net/ipv4/ipip.c | 18 ++++++++
net/ipv6/ip6_tunnel.c | 41 +++++++++++++-----
net/netfilter/nf_flow_table_core.c | 67 ++++++++++++++++++++++++++++--
net/netfilter/nf_flow_table_ip.c | 41 ++++--------------
net/netfilter/nf_flow_table_path.c | 11 ++++-
5 files changed, 130 insertions(+), 48 deletions(-)
diff --git a/net/ipv4/ipip.c b/net/ipv4/ipip.c
index 15189e1fa068..4079aa385fea 100644
--- a/net/ipv4/ipip.c
+++ b/net/ipv4/ipip.c
@@ -358,6 +358,8 @@ static int ipip_fill_forward_path(struct net_device_path_ctx *ctx,
{
struct ip_tunnel *tunnel = netdev_priv(ctx->dev);
const struct iphdr *tiph = &tunnel->parms.iph;
+ unsigned char ha[ETH_ALEN];
+ struct neighbour *n;
struct rtable *rt;
if (ctx->ether_type != cpu_to_be16(ETH_P_IP))
@@ -375,6 +377,21 @@ static int ipip_fill_forward_path(struct net_device_path_ctx *ctx,
if (IS_ERR(rt))
return PTR_ERR(rt);
+ n = dst_neigh_lookup(&rt->dst, &tiph->daddr);
+ if (!n) {
+ ip_rt_put(rt);
+ return -ENOENT;
+ }
+
+ if (n->dev->addr_len != ETH_ALEN ||
+ !(READ_ONCE(n->nud_state) & NUD_VALID)) {
+ ip_rt_put(rt);
+ neigh_release(n);
+ return -EINVAL;
+ }
+ neigh_ha_snapshot(ha, n, n->dev);
+ neigh_release(n);
+
path->type = DEV_PATH_TUN;
path->tun.src_v4.s_addr = tiph->saddr;
path->tun.dst_v4.s_addr = tiph->daddr;
@@ -383,6 +400,7 @@ static int ipip_fill_forward_path(struct net_device_path_ctx *ctx,
path->tun.dst = &rt->dst;
path->dev = ctx->dev;
+ memcpy(ctx->daddr, ha, ETH_ALEN);
ctx->dev = rt->dst.dev;
return 0;
diff --git a/net/ipv6/ip6_tunnel.c b/net/ipv6/ip6_tunnel.c
index 64a184bca467..84d26d9454a9 100644
--- a/net/ipv6/ip6_tunnel.c
+++ b/net/ipv6/ip6_tunnel.c
@@ -1840,7 +1840,9 @@ static int ip6_tnl_fill_forward_path(struct net_device_path_ctx *ctx,
struct net_device_path *path)
{
struct ip6_tnl *t = netdev_priv(ctx->dev);
+ unsigned char ha[ETH_ALEN];
struct dst_entry *dst;
+ struct neighbour *n;
struct flowi6 fl6;
int err;
@@ -1863,20 +1865,37 @@ static int ip6_tnl_fill_forward_path(struct net_device_path_ctx *ctx,
fl6.flowi6_proto = 0;
dst = ip6_route_output(dev_net(ctx->dev), NULL, &fl6);
- if (!dst->error) {
- path->type = DEV_PATH_TUN;
- path->tun.src_v6 = fl6.saddr;
- path->tun.dst_v6 = fl6.daddr;
- path->tun.inner_proto = IPPROTO_IPV6;
- path->tun.encap_proto = AF_INET6;
- path->tun.dst = dst;
- path->dev = ctx->dev;
- ctx->dev = dst->dev;
+ err = dst->error;
+ if (err) {
+ dst_release(dst);
+ return err;
}
- err = dst->error;
- if (err)
+ n = dst_neigh_lookup(dst, &fl6.daddr);
+ if (!n) {
dst_release(dst);
+ return -ENOENT;
+ }
+
+ if (n->dev->addr_len != ETH_ALEN ||
+ !(READ_ONCE(n->nud_state) & NUD_VALID)) {
+ dst_release(dst);
+ neigh_release(n);
+ return -EINVAL;
+ }
+ neigh_ha_snapshot(ha, n, n->dev);
+ neigh_release(n);
+
+ path->type = DEV_PATH_TUN;
+ path->tun.src_v6 = fl6.saddr;
+ path->tun.dst_v6 = fl6.daddr;
+ path->tun.inner_proto = IPPROTO_IPV6;
+ path->tun.encap_proto = AF_INET6;
+ path->tun.dst = dst;
+ path->dev = ctx->dev;
+
+ memcpy(ctx->daddr, ha, ETH_ALEN);
+ ctx->dev = dst->dev;
return err;
}
diff --git a/net/netfilter/nf_flow_table_core.c b/net/netfilter/nf_flow_table_core.c
index a7b4fc5766a9..80c5784140db 100644
--- a/net/netfilter/nf_flow_table_core.c
+++ b/net/netfilter/nf_flow_table_core.c
@@ -5,6 +5,7 @@
#include <linux/netfilter.h>
#include <linux/rhashtable.h>
#include <linux/netdevice.h>
+#include <linux/etherdevice.h>
#include <net/ip.h>
#include <net/ip6_route.h>
#include <net/netfilter/nf_tables.h>
@@ -140,8 +141,11 @@ static int flow_offload_fill_route(struct flow_offload *flow,
ETH_ALEN);
flow_tuple->out.ifidx = route->tuple[dir].out.ifindex;
break;
- case FLOW_OFFLOAD_XMIT_XFRM:
case FLOW_OFFLOAD_XMIT_NEIGH:
+ memcpy(flow_tuple->out.h_dest, route->tuple[dir].out.h_dest,
+ ETH_ALEN);
+ fallthrough;
+ case FLOW_OFFLOAD_XMIT_XFRM:
flow_tuple->out.ifidx = route->tuple[dir].out.ifindex;
flow_tuple->dst_cache = dst;
flow_tuple->dst_cookie = flow_offload_dst_cookie(flow_tuple);
@@ -573,6 +577,63 @@ static void nf_flow_table_extend_ct_timeout(struct nf_conn *ct)
nf_ct_put(ct);
}
+static inline bool nf_flow_dst_check_gc(struct flow_offload *flow, int dir)
+{
+ struct flow_offload_tuple *tuple = &flow->tuplehash[dir].tuple;
+ struct flow_offload_tuple *other_tuple;
+ unsigned char ha[ETH_ALEN];
+ struct neighbour *n;
+ const void *daddr;
+ u8 state;
+ bool ret;
+
+ if (!tuple->dst_cache)
+ return true;
+
+ if (!dst_check(tuple->dst_cache, tuple->dst_cookie))
+ return false;
+
+ if (tuple->xmit_type != FLOW_OFFLOAD_XMIT_NEIGH)
+ return true;
+
+ /* non-Ethernet device, skip check. */
+ if (is_zero_ether_addr(tuple->out.h_dest))
+ return true;
+
+ other_tuple = &flow->tuplehash[!dir].tuple;
+ if (other_tuple->tun_num)
+ daddr = &other_tuple->tun.src_v6;
+ else
+ daddr = &other_tuple->src_v6;
+
+ n = dst_neigh_lookup(tuple->dst_cache, daddr);
+ if (!n)
+ return false;
+
+ if (n->dev->addr_len != ETH_ALEN) {
+ neigh_release(n);
+ return false;
+ }
+ neigh_ha_snapshot(ha, n, n->dev);
+
+ state = READ_ONCE(n->nud_state);
+ if (state & NUD_VALID) {
+ if (ether_addr_equal(ha, tuple->out.h_dest))
+ ret = true;
+ else
+ ret = false;
+
+ if (state & NUD_STALE)
+ neigh_event_send(n, NULL);
+ } else {
+ ret = false;
+ }
+
+ neigh_release(n);
+
+ return ret;
+}
+
static void nf_flow_offload_gc_step(struct nf_flowtable *flow_table,
struct flow_offload *flow, void *data)
{
@@ -585,8 +646,8 @@ static void nf_flow_offload_gc_step(struct nf_flowtable *flow_table,
if (nf_flow_has_expired(flow) ||
nf_ct_is_dying(flow->ct) ||
- !nf_flow_dst_check(&flow->tuplehash[FLOW_OFFLOAD_DIR_ORIGINAL].tuple) ||
- !nf_flow_dst_check(&flow->tuplehash[FLOW_OFFLOAD_DIR_REPLY].tuple) ||
+ !nf_flow_dst_check_gc(flow, FLOW_OFFLOAD_DIR_ORIGINAL) ||
+ !nf_flow_dst_check_gc(flow, FLOW_OFFLOAD_DIR_REPLY) ||
nf_flow_custom_gc(flow_table, flow)) {
flow_offload_teardown(flow);
teardown = true;
diff --git a/net/netfilter/nf_flow_table_ip.c b/net/netfilter/nf_flow_table_ip.c
index 115c54ad4633..5971d2584c5c 100644
--- a/net/netfilter/nf_flow_table_ip.c
+++ b/net/netfilter/nf_flow_table_ip.c
@@ -590,8 +590,7 @@ static int nf_flow_pppoe_push(struct sk_buff *skb, u16 id,
static int nf_flow_tunnel_ip_push(struct net *net, struct sk_buff *skb,
struct flow_offload_tuple *tuple,
- struct dst_entry *dst,
- union nf_inet_addr **ip_addr)
+ struct dst_entry *dst)
{
__be16 frag_off = 0;
struct iphdr *iph;
@@ -651,15 +650,12 @@ static int nf_flow_tunnel_ip_push(struct net *net, struct sk_buff *skb,
__ip_select_ident(net, iph, skb_shinfo(skb)->gso_segs ?: 1);
ip_send_check(iph);
- *ip_addr = (union nf_inet_addr *)&tuple->tun.src_v4;
-
return 0;
}
static int nf_flow_tunnel_ip6_push(struct net *net, struct sk_buff *skb,
struct flow_offload_tuple *tuple,
- struct dst_entry *dst,
- union nf_inet_addr **ip_addr)
+ struct dst_entry *dst)
{
struct flowi6 fl6 = {
.daddr = tuple->tun.src_v6,
@@ -718,21 +714,18 @@ static int nf_flow_tunnel_ip6_push(struct net *net, struct sk_buff *skb,
ipv6_hdr(skb)->payload_len = htons(skb->len - sizeof(*ip6h));
IP6CB(skb)->nhoff = offsetof(struct ipv6hdr, nexthdr);
- *ip_addr = (union nf_inet_addr *)&tuple->tun.src_v6;
-
return 0;
}
static int nf_flow_tunnel_push(struct net *net, struct sk_buff *skb,
struct flow_offload_tuple *tuple,
- struct dst_entry *dst,
- union nf_inet_addr **ip_addr)
+ struct dst_entry *dst)
{
switch (tuple->tun.encap_proto) {
case AF_INET:
- return nf_flow_tunnel_ip_push(net, skb, tuple, dst, ip_addr);
+ return nf_flow_tunnel_ip_push(net, skb, tuple, dst);
case AF_INET6:
- return nf_flow_tunnel_ip6_push(net, skb, tuple, dst, ip_addr);
+ return nf_flow_tunnel_ip6_push(net, skb, tuple, dst);
default:
return 0;
}
@@ -842,9 +835,7 @@ static int nf_flow_queue_xmit4(struct sk_buff *skb,
struct flow_offload_tuple *other_tuple;
enum flow_offload_tuple_dir dir;
struct nf_flow_xmit xmit = {};
- union nf_inet_addr *ip_addr;
struct flow_offload *flow;
- struct neighbour *neigh;
struct rtable *rt;
if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
@@ -858,10 +849,9 @@ static int nf_flow_queue_xmit4(struct sk_buff *skb,
dir = tuplehash->tuple.dir;
flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
other_tuple = &flow->tuplehash[!dir].tuple;
- ip_addr = (union nf_inet_addr *)&other_tuple->src_v4;
if (nf_flow_tunnel_push(state->net, skb, other_tuple,
- tuplehash->tuple.dst_cache, &ip_addr) < 0)
+ tuplehash->tuple.dst_cache) < 0)
return NF_DROP;
switch (tuplehash->tuple.xmit_type) {
@@ -872,12 +862,7 @@ static int nf_flow_queue_xmit4(struct sk_buff *skb,
flow_offload_teardown(flow);
return NF_DROP;
}
- neigh = ip_neigh_gw4(rt->dst.dev, rt_nexthop(rt, ip_addr->ip));
- if (IS_ERR(neigh)) {
- flow_offload_teardown(flow);
- return NF_DROP;
- }
- xmit.dest = neigh->ha;
+ xmit.dest = tuplehash->tuple.out.h_dest;
skb_dst_drop(skb);
skb_dst_set_noref(skb, &rt->dst);
break;
@@ -1171,9 +1156,7 @@ static int nf_flow_queue_xmit6(struct sk_buff *skb,
struct flow_offload_tuple *other_tuple;
enum flow_offload_tuple_dir dir;
struct nf_flow_xmit xmit = {};
- union nf_inet_addr *ip_addr;
struct flow_offload *flow;
- struct neighbour *neigh;
struct rt6_info *rt;
if (unlikely(tuplehash->tuple.xmit_type == FLOW_OFFLOAD_XMIT_XFRM)) {
@@ -1187,10 +1170,9 @@ static int nf_flow_queue_xmit6(struct sk_buff *skb,
dir = tuplehash->tuple.dir;
flow = container_of(tuplehash, struct flow_offload, tuplehash[dir]);
other_tuple = &flow->tuplehash[!dir].tuple;
- ip_addr = (union nf_inet_addr *)&other_tuple->src_v6;
if (nf_flow_tunnel_push(state->net, skb, other_tuple,
- tuplehash->tuple.dst_cache, &ip_addr) < 0)
+ tuplehash->tuple.dst_cache) < 0)
return NF_DROP;
switch (tuplehash->tuple.xmit_type) {
@@ -1201,12 +1183,7 @@ static int nf_flow_queue_xmit6(struct sk_buff *skb,
flow_offload_teardown(flow);
return NF_DROP;
}
- neigh = ip_neigh_gw6(rt->dst.dev, rt6_nexthop(rt, &ip_addr->in6));
- if (IS_ERR(neigh)) {
- flow_offload_teardown(flow);
- return NF_DROP;
- }
- xmit.dest = neigh->ha;
+ xmit.dest = tuplehash->tuple.out.h_dest;
skb_dst_drop(skb);
skb_dst_set_noref(skb, &rt->dst);
break;
diff --git a/net/netfilter/nf_flow_table_path.c b/net/netfilter/nf_flow_table_path.c
index 841f099434d9..5923ef9cdaef 100644
--- a/net/netfilter/nf_flow_table_path.c
+++ b/net/netfilter/nf_flow_table_path.c
@@ -57,6 +57,7 @@ static int nft_dev_fill_forward_path(const struct dst_entry *dst_cache,
};
struct neighbour *n;
u8 nud_state;
+ int ret;
if (!nft_is_valid_ether_device(dev)) {
eth_zero_addr(ha);
@@ -79,7 +80,12 @@ static int nft_dev_fill_forward_path(const struct dst_entry *dst_cache,
out:
ether_addr_copy(ctx.daddr, ha);
- return dev_fill_forward_path(&ctx, stack);
+ ret = dev_fill_forward_path(&ctx, stack);
+
+ if (memcmp(ctx.daddr, ha, ETH_ALEN))
+ ether_addr_copy(ha, ctx.daddr);
+
+ return ret;
}
struct nft_forward_info {
@@ -259,9 +265,10 @@ static int nft_dev_forward_path(const struct nft_pktinfo *pkt,
route->tuple[!dir].in.num_encaps = info.num_encaps;
route->tuple[!dir].in.ingress_vlans = info.ingress_vlans;
+ memcpy(route->tuple[dir].out.h_dest, info.h_dest, ETH_ALEN);
+
if (info.xmit_type == FLOW_OFFLOAD_XMIT_DIRECT) {
memcpy(route->tuple[dir].out.h_source, info.h_source, ETH_ALEN);
- memcpy(route->tuple[dir].out.h_dest, info.h_dest, ETH_ALEN);
route->tuple[dir].xmit_type = info.xmit_type;
}
route->tuple[dir].out.needs_gso_segment = info.needs_gso_segment;
--
2.47.3
^ permalink raw reply related [flat|nested] 3+ messages in thread