[PATCH net 0/4] netfilter: updates for net

[PATCH net 0/4] netfilter: updates for net

[Florian Westphal]

The following patchset contains Netfilter fixes for *net*:

1) Fix crash (call recursion) when nftables synproxy extension is used
   in an object map.  When this feature was added in v5.4 the required
   hook call validation was forgotten.
   Fix from Fernando Fernandez Mancera.
2) bridge br_vlan_fill_forward_path_pvid uses incorrect
   rcu_dereference_protected(); we only have rcu read lock but not
   RTNL.  Fix from Eric Woudstra.

Last two patches address flakes in two existing selftests.

Please, pull these changes from:
The following changes since commit 2c95a756e0cfc19af6d0b32b0c6cf3bada334998:

  net: pse-pd: tps23881: Fix current measurement scaling (2025-10-07 18:30:53 -0700)

are available in the Git repository at:

  https://git.kernel.org/pub/scm/linux/kernel/git/netfilter/nf.git tags/nf-25-10-08

for you to fetch changes up to e84945bdc619ed4243ba4298dbb8ca2062026474:

  selftests: netfilter: query conntrack state to check for port clash resolution (2025-10-08 13:17:31 +0200)

----------------------------------------------------------------
netfilter pull request nf-25-10-08

----------------------------------------------------------------
Eric Woudstra (1):
      bridge: br_vlan_fill_forward_path_pvid: use br_vlan_group_rcu()

Fernando Fernandez Mancera (1):
      netfilter: nft_objref: validate objref and objrefmap expressions

Florian Westphal (2):
      selftests: netfilter: nft_fib.sh: fix spurious test failures
      selftests: netfilter: query conntrack state to check for port clash resolution

 net/bridge/br_vlan.c                               |  2 +-
 net/netfilter/nft_objref.c                         | 39 +++++++++++++++
 .../selftests/net/netfilter/nf_nat_edemux.sh       | 58 +++++++++++++++-------
 tools/testing/selftests/net/netfilter/nft_fib.sh   | 13 +++--
 4 files changed, 89 insertions(+), 23 deletions(-)

[PATCH net 1/4] netfilter: nft_objref: validate objref and objrefmap expressions

[Florian Westphal]

From: Fernando Fernandez Mancera <fmancera@suse.de>

Referencing a synproxy stateful object from OUTPUT hook causes kernel
crash due to infinite recursive calls:

BUG: TASK stack guard page was hit at 000000008bda5b8c (stack is 000000003ab1c4a5..00000000494d8b12)
[...]
Call Trace:
 __find_rr_leaf+0x99/0x230
 fib6_table_lookup+0x13b/0x2d0
 ip6_pol_route+0xa4/0x400
 fib6_rule_lookup+0x156/0x240
 ip6_route_output_flags+0xc6/0x150
 __nf_ip6_route+0x23/0x50
 synproxy_send_tcp_ipv6+0x106/0x200
 synproxy_send_client_synack_ipv6+0x1aa/0x1f0
 nft_synproxy_do_eval+0x263/0x310
 nft_do_chain+0x5a8/0x5f0 [nf_tables
 nft_do_chain_inet+0x98/0x110
 nf_hook_slow+0x43/0xc0
 __ip6_local_out+0xf0/0x170
 ip6_local_out+0x17/0x70
 synproxy_send_tcp_ipv6+0x1a2/0x200
 synproxy_send_client_synack_ipv6+0x1aa/0x1f0
[...]

Implement objref and objrefmap expression validate functions.

Currently, only NFT_OBJECT_SYNPROXY object type requires validation.
This will also handle a jump to a chain using a synproxy object from the
OUTPUT hook.

Now when trying to reference a synproxy object in the OUTPUT hook, nft
will produce the following error:

synproxy_crash.nft: Error: Could not process rule: Operation not supported
  synproxy name mysynproxy
  ^^^^^^^^^^^^^^^^^^^^^^^^

Fixes: ee394f96ad75 ("netfilter: nft_synproxy: add synproxy stateful object support")
Reported-by: Georg Pfuetzenreuter <georg.pfuetzenreuter@suse.com>
Closes: https://bugzilla.suse.com/1250237
Signed-off-by: Fernando Fernandez Mancera <fmancera@suse.de>
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 net/netfilter/nft_objref.c | 39 ++++++++++++++++++++++++++++++++++++++
 1 file changed, 39 insertions(+)

diff --git a/net/netfilter/nft_objref.c b/net/netfilter/nft_objref.c
index 8ee66a86c3bc..1a62e384766a 100644
--- a/net/netfilter/nft_objref.c
+++ b/net/netfilter/nft_objref.c
@@ -22,6 +22,35 @@ void nft_objref_eval(const struct nft_expr *expr,
 	obj->ops->eval(obj, regs, pkt);
 }
 
+static int nft_objref_validate_obj_type(const struct nft_ctx *ctx, u32 type)
+{
+	unsigned int hooks;
+
+	switch (type) {
+	case NFT_OBJECT_SYNPROXY:
+		if (ctx->family != NFPROTO_IPV4 &&
+		    ctx->family != NFPROTO_IPV6 &&
+		    ctx->family != NFPROTO_INET)
+			return -EOPNOTSUPP;
+
+		hooks = (1 << NF_INET_LOCAL_IN) | (1 << NF_INET_FORWARD);
+
+		return nft_chain_validate_hooks(ctx->chain, hooks);
+	default:
+		break;
+	}
+
+	return 0;
+}
+
+static int nft_objref_validate(const struct nft_ctx *ctx,
+			       const struct nft_expr *expr)
+{
+	struct nft_object *obj = nft_objref_priv(expr);
+
+	return nft_objref_validate_obj_type(ctx, obj->ops->type->type);
+}
+
 static int nft_objref_init(const struct nft_ctx *ctx,
 			   const struct nft_expr *expr,
 			   const struct nlattr * const tb[])
@@ -93,6 +122,7 @@ static const struct nft_expr_ops nft_objref_ops = {
 	.activate	= nft_objref_activate,
 	.deactivate	= nft_objref_deactivate,
 	.dump		= nft_objref_dump,
+	.validate	= nft_objref_validate,
 	.reduce		= NFT_REDUCE_READONLY,
 };
 
@@ -197,6 +227,14 @@ static void nft_objref_map_destroy(const struct nft_ctx *ctx,
 	nf_tables_destroy_set(ctx, priv->set);
 }
 
+static int nft_objref_map_validate(const struct nft_ctx *ctx,
+				   const struct nft_expr *expr)
+{
+	const struct nft_objref_map *priv = nft_expr_priv(expr);
+
+	return nft_objref_validate_obj_type(ctx, priv->set->objtype);
+}
+
 static const struct nft_expr_ops nft_objref_map_ops = {
 	.type		= &nft_objref_type,
 	.size		= NFT_EXPR_SIZE(sizeof(struct nft_objref_map)),
@@ -206,6 +244,7 @@ static const struct nft_expr_ops nft_objref_map_ops = {
 	.deactivate	= nft_objref_map_deactivate,
 	.destroy	= nft_objref_map_destroy,
 	.dump		= nft_objref_map_dump,
+	.validate	= nft_objref_map_validate,
 	.reduce		= NFT_REDUCE_READONLY,
 };
 
-- 
2.49.1

Re: [PATCH net 1/4] netfilter: nft_objref: validate objref and objrefmap expressions

[patchwork-bot+netdevbpf]

Hello:

This series was applied to netdev/net.git (main)
by Florian Westphal <fw@strlen.de>:

On Wed,  8 Oct 2025 14:59:39 +0200 you wrote:
> From: Fernando Fernandez Mancera <fmancera@suse.de>
> 
> Referencing a synproxy stateful object from OUTPUT hook causes kernel
> crash due to infinite recursive calls:
> 
> BUG: TASK stack guard page was hit at 000000008bda5b8c (stack is 000000003ab1c4a5..00000000494d8b12)
> [...]
> Call Trace:
>  __find_rr_leaf+0x99/0x230
>  fib6_table_lookup+0x13b/0x2d0
>  ip6_pol_route+0xa4/0x400
>  fib6_rule_lookup+0x156/0x240
>  ip6_route_output_flags+0xc6/0x150
>  __nf_ip6_route+0x23/0x50
>  synproxy_send_tcp_ipv6+0x106/0x200
>  synproxy_send_client_synack_ipv6+0x1aa/0x1f0
>  nft_synproxy_do_eval+0x263/0x310
>  nft_do_chain+0x5a8/0x5f0 [nf_tables
>  nft_do_chain_inet+0x98/0x110
>  nf_hook_slow+0x43/0xc0
>  __ip6_local_out+0xf0/0x170
>  ip6_local_out+0x17/0x70
>  synproxy_send_tcp_ipv6+0x1a2/0x200
>  synproxy_send_client_synack_ipv6+0x1aa/0x1f0
> [...]
> 
> [...]

Here is the summary with links:
  - [net,1/4] netfilter: nft_objref: validate objref and objrefmap expressions
    https://git.kernel.org/netdev/net/c/f359b809d54c
  - [net,2/4] bridge: br_vlan_fill_forward_path_pvid: use br_vlan_group_rcu()
    https://git.kernel.org/netdev/net/c/bbf0c98b3ad9
  - [net,3/4] selftests: netfilter: nft_fib.sh: fix spurious test failures
    https://git.kernel.org/netdev/net/c/a126ab6b26f1
  - [net,4/4] selftests: netfilter: query conntrack state to check for port clash resolution
    https://git.kernel.org/netdev/net/c/e84945bdc619

You are awesome, thank you!
-- 
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html

[PATCH net 2/4] bridge: br_vlan_fill_forward_path_pvid: use br_vlan_group_rcu()

[Florian Westphal]

From: Eric Woudstra <ericwouds@gmail.com>

net/bridge/br_private.h:1627 suspicious rcu_dereference_protected() usage!
other info that might help us debug this:

rcu_scheduler_active = 2, debug_locks = 1
7 locks held by socat/410:
 #0: ffff88800d7a9c90 (sk_lock-AF_INET){+.+.}-{0:0}, at: inet_stream_connect+0x43/0xa0
 #1: ffffffff9a779900 (rcu_read_lock){....}-{1:3}, at: __ip_queue_xmit+0x62/0x1830
 [..]
 #6: ffffffff9a779900 (rcu_read_lock){....}-{1:3}, at: nf_hook.constprop.0+0x8a/0x440

Call Trace:
 lockdep_rcu_suspicious.cold+0x4f/0xb1
 br_vlan_fill_forward_path_pvid+0x32c/0x410 [bridge]
 br_fill_forward_path+0x7a/0x4d0 [bridge]

Use to correct helper, non _rcu variant requires RTNL mutex.

Fixes: bcf2766b1377 ("net: bridge: resolve forwarding path for VLAN tag actions in bridge devices")
Signed-off-by: Eric Woudstra <ericwouds@gmail.com>
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 net/bridge/br_vlan.c | 2 +-
 1 file changed, 1 insertion(+), 1 deletion(-)

diff --git a/net/bridge/br_vlan.c b/net/bridge/br_vlan.c
index ae911220cb3c..ce72b837ff8e 100644
--- a/net/bridge/br_vlan.c
+++ b/net/bridge/br_vlan.c
@@ -1457,7 +1457,7 @@ void br_vlan_fill_forward_path_pvid(struct net_bridge *br,
 	if (!br_opt_get(br, BROPT_VLAN_ENABLED))
 		return;
 
-	vg = br_vlan_group(br);
+	vg = br_vlan_group_rcu(br);
 
 	if (idx >= 0 &&
 	    ctx->vlan[idx].proto == br->vlan_proto) {
-- 
2.49.1

[PATCH net 3/4] selftests: netfilter: nft_fib.sh: fix spurious test failures

[Florian Westphal]

Jakub reports spurious failure of nft_fib.sh test.
This is caused by a subtle bug inherited when i moved faulty ping
from one test case to another.

nft_fib.sh not only checks that the fib expression matched, it also
records the number of matches and then validates we have the expected
count.  When I did this it was under the assumption that we would
have 0 to n matching packets.  In case of the failure, the entry has
n+1 matching packets.

This happens because ping_unreachable helper uses "ping -c 1 -w 1",
instead of the intended "-W".  -w alters the meaning of -c (count),
namely, its then treated as number of wanted *replies* instead of
"number of packets to send".

So, in some cases, ping -c 1 -w 1 ends up sending two packets which then
makes the test fail due to the higher-than-expected packet count.

Fix the actual bug (s/-w/-W) and also change the error handling:
1. Show the number of expected packets in the error message
2. Always try to delete the key from the set.
   Else, later test that makes sure we don't have unexpected keys
   in there will always fail as well.

Reported-by: Jakub Kicinski <kuba@kernel.org>
Closes: https://lore.kernel.org/netfilter-devel/20250927090709.0b3cd783@kernel.org/
Fixes: 98287045c979 ("selftests: netfilter: move fib vrf test to nft_fib.sh")
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 tools/testing/selftests/net/netfilter/nft_fib.sh | 13 ++++++++-----
 1 file changed, 8 insertions(+), 5 deletions(-)

diff --git a/tools/testing/selftests/net/netfilter/nft_fib.sh b/tools/testing/selftests/net/netfilter/nft_fib.sh
index 9929a9ffef65..04544905c216 100755
--- a/tools/testing/selftests/net/netfilter/nft_fib.sh
+++ b/tools/testing/selftests/net/netfilter/nft_fib.sh
@@ -256,12 +256,12 @@ test_ping_unreachable() {
   local daddr4=$1
   local daddr6=$2
 
-  if ip netns exec "$ns1" ping -c 1 -w 1 -q "$daddr4" > /dev/null; then
+  if ip netns exec "$ns1" ping -c 1 -W 0.1 -q "$daddr4" > /dev/null; then
 	echo "FAIL: ${ns1} could reach $daddr4" 1>&2
 	return 1
   fi
 
-  if ip netns exec "$ns1" ping -c 1 -w 1 -q "$daddr6" > /dev/null; then
+  if ip netns exec "$ns1" ping -c 1 -W 0.1 -q "$daddr6" > /dev/null; then
 	echo "FAIL: ${ns1} could reach $daddr6" 1>&2
 	return 1
   fi
@@ -437,14 +437,17 @@ check_type()
 	local addr="$3"
 	local type="$4"
 	local count="$5"
+	local lret=0
 
 	[ -z "$count" ] && count=1
 
 	if ! ip netns exec "$nsrouter" nft get element inet t "$setname" { "$iifname" . "$addr" . "$type" } |grep -q "counter packets $count";then
-		echo "FAIL: did not find $iifname . $addr . $type in $setname"
+		echo "FAIL: did not find $iifname . $addr . $type in $setname with $count packets"
 		ip netns exec "$nsrouter" nft list set inet t "$setname"
 		ret=1
-		return 1
+		# do not fail right away, delete entry if it exists so later test that
+		# checks for unwanted keys don't get confused by this *expected* key.
+		lret=1
 	fi
 
 	# delete the entry, this allows to check if anything unexpected appeared
@@ -456,7 +459,7 @@ check_type()
 		return 1
 	fi
 
-	return 0
+	return $lret
 }
 
 check_local()
-- 
2.49.1

[PATCH net 4/4] selftests: netfilter: query conntrack state to check for port clash resolution

[Florian Westphal]

Jakub reported this self test flaking occasionally (fails, but passes on
re-run) on debug kernels.

This is because the test checks for elapsed time to determine if both
connections were established in parallel.

Rework this to no longer depend on timing.
Use busywait helper to check that both sockets have moved to established
state and then query the conntrack engine for the two entries.

Reported-by: Jakub Kicinski <kuba@kernel.org>
Closes: https://lore.kernel.org/netfilter-devel/20250926163318.40d1a502@kernel.org/
Fixes: 117e149e26d1 ("selftests: netfilter: test nat source port clash resolution interaction with tcp early demux")
Signed-off-by: Florian Westphal <fw@strlen.de>
---
 .../selftests/net/netfilter/nf_nat_edemux.sh  | 58 +++++++++++++------
 1 file changed, 41 insertions(+), 17 deletions(-)

diff --git a/tools/testing/selftests/net/netfilter/nf_nat_edemux.sh b/tools/testing/selftests/net/netfilter/nf_nat_edemux.sh
index 1014551dd769..6731fe1eaf2e 100755
--- a/tools/testing/selftests/net/netfilter/nf_nat_edemux.sh
+++ b/tools/testing/selftests/net/netfilter/nf_nat_edemux.sh
@@ -17,9 +17,31 @@ cleanup()
 
 checktool "socat -h" "run test without socat"
 checktool "iptables --version" "run test without iptables"
+checktool "conntrack --version" "run test without conntrack"
 
 trap cleanup EXIT
 
+connect_done()
+{
+	local ns="$1"
+	local port="$2"
+
+	ip netns exec "$ns" ss -nt -o state established "dport = :$port" | grep -q "$port"
+}
+
+check_ctstate()
+{
+	local ns="$1"
+	local dp="$2"
+
+	if ! ip netns exec "$ns" conntrack --get -s 192.168.1.2 -d 192.168.1.1 -p tcp \
+	     --sport 10000 --dport "$dp" --state ESTABLISHED > /dev/null 2>&1;then
+		echo "FAIL: Did not find expected state for dport $2"
+		ip netns exec "$ns" bash -c 'conntrack -L; conntrack -S; ss -nt'
+		ret=1
+	fi
+}
+
 setup_ns ns1 ns2
 
 # Connect the namespaces using a veth pair
@@ -44,15 +66,18 @@ socatpid=$!
 ip netns exec "$ns2" sysctl -q net.ipv4.ip_local_port_range="10000 10000"
 
 # add a virtual IP using DNAT
-ip netns exec "$ns2" iptables -t nat -A OUTPUT -d 10.96.0.1/32 -p tcp --dport 443 -j DNAT --to-destination 192.168.1.1:5201
+ip netns exec "$ns2" iptables -t nat -A OUTPUT -d 10.96.0.1/32 -p tcp --dport 443 -j DNAT --to-destination 192.168.1.1:5201 || exit 1
 
 # ... and route it to the other namespace
 ip netns exec "$ns2" ip route add 10.96.0.1 via 192.168.1.1
 
-# add a persistent connection from the other namespace
-ip netns exec "$ns2" socat -t 10 - TCP:192.168.1.1:5201 > /dev/null &
+# listener should be up by now, wait if it isn't yet.
+wait_local_port_listen "$ns1" 5201 tcp
 
-sleep 1
+# add a persistent connection from the other namespace
+sleep 10 | ip netns exec "$ns2" socat -t 10 - TCP:192.168.1.1:5201 > /dev/null &
+cpid0=$!
+busywait "$BUSYWAIT_TIMEOUT" connect_done "$ns2" "5201"
 
 # ip daddr:dport will be rewritten to 192.168.1.1 5201
 # NAT must reallocate source port 10000 because
@@ -71,26 +96,25 @@ fi
 ip netns exec "$ns1" iptables -t nat -A PREROUTING -p tcp --dport 5202 -j REDIRECT --to-ports 5201
 ip netns exec "$ns1" iptables -t nat -A PREROUTING -p tcp --dport 5203 -j REDIRECT --to-ports 5201
 
-sleep 5 | ip netns exec "$ns2" socat -t 5 -u STDIN TCP:192.168.1.1:5202,connect-timeout=5 >/dev/null &
+sleep 5 | ip netns exec "$ns2" socat -T 5 -u STDIN TCP:192.168.1.1:5202,connect-timeout=5 >/dev/null &
+cpid1=$!
 
-# if connect succeeds, client closes instantly due to EOF on stdin.
-# if connect hangs, it will time out after 5s.
-echo | ip netns exec "$ns2" socat -t 3 -u STDIN TCP:192.168.1.1:5203,connect-timeout=5 >/dev/null &
+sleep 5 | ip netns exec "$ns2" socat -T 5 -u STDIN TCP:192.168.1.1:5203,connect-timeout=5 >/dev/null &
 cpid2=$!
 
-time_then=$(date +%s)
-wait $cpid2
-rv=$?
-time_now=$(date +%s)
+busywait "$BUSYWAIT_TIMEOUT" connect_done "$ns2" 5202
+busywait "$BUSYWAIT_TIMEOUT" connect_done "$ns2" 5203
 
-# Check how much time has elapsed, expectation is for
-# 'cpid2' to connect and then exit (and no connect delay).
-delta=$((time_now - time_then))
+check_ctstate "$ns1" 5202
+check_ctstate "$ns1" 5203
 
-if [ $delta -lt 2 ] && [ $rv -eq 0 ]; then
+kill $socatpid $cpid0 $cpid1 $cpid2
+socatpid=0
+
+if [ $ret -eq 0 ]; then
 	echo "PASS: could connect to service via redirected ports"
 else
-	echo "FAIL: socat cannot connect to service via redirect ($delta seconds elapsed, returned $rv)"
+	echo "FAIL: socat cannot connect to service via redirect"
 	ret=1
 fi
 
-- 
2.49.1