Skip to contentSkip to conversation

net: clear the dst when changing skb protocol

PATCH Applied Jakub Kicinski 3 participants 5 messages Jun 10, 2025 – Jun 12, 2025

Subsystems (6)
lore
Share & tools
lore

Jakub Kicinski

author maintainer
#0 lore raw

PATCH v3 1/2 net: clear the dst when changing skb protocol

A not-so-careful NAT46 BPF program can crash the kernel if it indiscriminately flips ingress packets from v4 to v6:

  BUG: kernel NULL pointer dereference, address: 0000000000000000
    ip6_rcv_core (net/ipv6/ip6_input.c:190:20)
    ipv6_rcv (net/ipv6/ip6_input.c:306:8)
    process_backlog (net/core/dev.c:6186:4)
    napi_poll (net/core/dev.c:6906:9)
    net_rx_action (net/core/dev.c:7028:13)
    do_softirq (kernel/softirq.c:462:3)
    netif_rx (net/core/dev.c:5326:3)
    dev_loopback_xmit (net/core/dev.c:4015:2)
    ip_mc_finish_output (net/ipv4/ip_output.c:363:8)
    NF_HOOK (./include/linux/netfilter.h:314:9)
    ip_mc_output (net/ipv4/ip_output.c:400:5)
    dst_output (./include/net/dst.h:459:9)
    ip_local_out (net/ipv4/ip_output.c:130:9)
    ip_send_skb (net/ipv4/ip_output.c:1496:8)
    udp_send_skb (net/ipv4/udp.c:1040:8)
    udp_sendmsg (net/ipv4/udp.c:1328:10)

The output interface has a 4->6 program attached at ingress. We try to loop the multicast skb back to the sending socket. Ingress BPF runs as part of netif_rx(), pushes a valid v6 hdr and changes skb->protocol to v6. We enter ip6_rcv_core which tries to use skb_dst(). But the dst is still an IPv4 one left after IPv4 mcast output.

Clear the dst in all BPF helpers which change the protocol. Try to preserve metadata dsts, those may carry non-routing metadata.

Fixes d219df60a70e ("bpf: Add ipip6 and ip6ip decap support for bpf_skb_adjust_room()")
Fixes 1b00e0dfe7d0 ("bpf: update skb->protocol in bpf_skb_net_grow")
Fixes 6578171a7ff0 ("bpf: add bpf_skb_change_proto helper")

---
v3:
- go back to v1, the encap / decap which don't change proto will be added in -next
- split out the test
v2: https://lore.kernel.org/[email protected]
- drop on encap/decap
- fix typo (protcol)
- add the test to the Makefile
v1: https://lore.kernel.org/[email protected]

I wonder if we should not skip ingress (tc_skip_classify?) for looped back packets in the first place. But that doesn't seem robust enough vs multiple redirections to solve the crash.

Ignoring LOOPBACK packets (like the NAT46 prog should) doesn't work either, since BPF can change pkt_type arbitrarily.

---
 net/core/filter.c | 19 +++++++++++++------
 1 file changed, 13 insertions(+), 6 deletions(-)
net/core/filter.c +13 -6
index 327ca73f9cd7..7a72f766aacf 100644
@@ -3233,6 +3233,13 @@ static const struct bpf_func_proto bpf_skb_vlan_pop_proto = {
 	.arg1_type      = ARG_PTR_TO_CTX,
 };
 
+static void bpf_skb_change_protocol(struct sk_buff *skb, u16 proto)
+{
+	skb->protocol = htons(proto);
+	if (skb_valid_dst(skb))
+		skb_dst_drop(skb);
+}
+
 static int bpf_skb_generic_push(struct sk_buff *skb, u32 off, u32 len)
 {
 	/* Caller already did skb_cow() with len as headroom,
@@ -3329,7 +3336,7 @@ static int bpf_skb_proto_4_to_6(struct sk_buff *skb)
 		}
 	}
 
-	skb->protocol = htons(ETH_P_IPV6);
+	bpf_skb_change_protocol(skb, ETH_P_IPV6);
 	skb_clear_hash(skb);
 
 	return 0;
@@ -3359,7 +3366,7 @@ static int bpf_skb_proto_6_to_4(struct sk_buff *skb)
 		}
 	}
 
-	skb->protocol = htons(ETH_P_IP);
+	bpf_skb_change_protocol(skb, ETH_P_IP);
 	skb_clear_hash(skb);
 
 	return 0;
@@ -3550,10 +3557,10 @@ static int bpf_skb_net_grow(struct sk_buff *skb, u32 off, u32 len_diff,
 		/* Match skb->protocol to new outer l3 protocol */
 		if (skb->protocol == htons(ETH_P_IP) &&
 		    flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV6)
-			skb->protocol = htons(ETH_P_IPV6);
+			bpf_skb_change_protocol(skb, ETH_P_IPV6);
 		else if (skb->protocol == htons(ETH_P_IPV6) &&
 			 flags & BPF_F_ADJ_ROOM_ENCAP_L3_IPV4)
-			skb->protocol = htons(ETH_P_IP);
+			bpf_skb_change_protocol(skb, ETH_P_IP);
 	}
 
 	if (skb_is_gso(skb)) {
@@ -3606,10 +3613,10 @@ static int bpf_skb_net_shrink(struct sk_buff *skb, u32 off, u32 len_diff,
 	/* Match skb->protocol to new outer l3 protocol */
 	if (skb->protocol == htons(ETH_P_IP) &&
 	    flags & BPF_F_ADJ_ROOM_DECAP_L3_IPV6)
-		skb->protocol = htons(ETH_P_IPV6);
+		bpf_skb_change_protocol(skb, ETH_P_IPV6);
 	else if (skb->protocol == htons(ETH_P_IPV6) &&
 		 flags & BPF_F_ADJ_ROOM_DECAP_L3_IPV4)
-		skb->protocol = htons(ETH_P_IP);
+		bpf_skb_change_protocol(skb, ETH_P_IP);
 
 	if (skb_is_gso(skb)) {
 		struct skb_shared_info *shinfo = skb_shinfo(skb);
Signature · 1 line
2.49.0
author maintainer
#1 lore raw

PATCH v3 2/2 selftests: net: add test case for NAT46 looping back dst

Simple test for crash involving multicast loopback and stale dst. Reuse exising NAT46 program.

---
 tools/testing/selftests/net/Makefile   |  1 +
 tools/testing/selftests/net/nat6to4.sh | 15 +++++++++++++++
 2 files changed, 16 insertions(+)
 create mode 100755 tools/testing/selftests/net/nat6to4.sh
tools/testing/selftests/net/Makefile +1 -0
index ea84b88bcb30..ab996bd22a5f 100644
@@ -27,6 +27,7 @@ TEST_PROGS += amt.sh
 TEST_PROGS += unicast_extensions.sh
 TEST_PROGS += udpgro_fwd.sh
 TEST_PROGS += udpgro_frglist.sh
+TEST_PROGS += nat6to4.sh
 TEST_PROGS += veth.sh
 TEST_PROGS += ioam6.sh
 TEST_PROGS += gro.sh
tools/testing/selftests/net/nat6to4.sh new file +15 -0
new file mode 100755
index 000000000000..0ee859b622a4
@@ -0,0 +1,15 @@
+#!/bin/bash
+# SPDX-License-Identifier: GPL-2.0
+
+NS="ns-peer-$(mktemp -u XXXXXX)"
+
+ip netns add "${NS}"
+ip -netns "${NS}" link set lo up
+ip -netns "${NS}" route add default via 127.0.0.2 dev lo
+
+tc -n "${NS}" qdisc add dev lo ingress
+tc -n "${NS}" filter add dev lo ingress prio 4 protocol ip \
+   bpf object-file nat6to4.bpf.o section schedcls/egress4/snat4 direct-action
+
+ip netns exec "${NS}" \
+   bash -c 'echo 012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789012345678901234567890123456789abc | socat - UDP4-DATAGRAM:224.1.0.1:6666,ip-multicast-loop=1'
Signature · 1 line
2.49.0
Reviewed-by on v3 1/2
#3 lore raw
In reply to Jakub Kicinski
Quoted 32 lines · Jakub Kicinski

A not-so-careful NAT46 BPF program can crash the kernel if it indiscriminately flips ingress packets from v4 to v6:

  BUG: kernel NULL pointer dereference, address: 0000000000000000
    ip6_rcv_core (net/ipv6/ip6_input.c:190:20)
    ipv6_rcv (net/ipv6/ip6_input.c:306:8)
    process_backlog (net/core/dev.c:6186:4)
    napi_poll (net/core/dev.c:6906:9)
    net_rx_action (net/core/dev.c:7028:13)
    do_softirq (kernel/softirq.c:462:3)
    netif_rx (net/core/dev.c:5326:3)
    dev_loopback_xmit (net/core/dev.c:4015:2)
    ip_mc_finish_output (net/ipv4/ip_output.c:363:8)
    NF_HOOK (./include/linux/netfilter.h:314:9)
    ip_mc_output (net/ipv4/ip_output.c:400:5)
    dst_output (./include/net/dst.h:459:9)
    ip_local_out (net/ipv4/ip_output.c:130:9)
    ip_send_skb (net/ipv4/ip_output.c:1496:8)
    udp_send_skb (net/ipv4/udp.c:1040:8)
    udp_sendmsg (net/ipv4/udp.c:1328:10)

The output interface has a 4->6 program attached at ingress. We try to loop the multicast skb back to the sending socket. Ingress BPF runs as part of netif_rx(), pushes a valid v6 hdr and changes skb->protocol to v6. We enter ip6_rcv_core which tries to use skb_dst(). But the dst is still an IPv4 one left after IPv4 mcast output.

Clear the dst in all BPF helpers which change the protocol. Try to preserve metadata dsts, those may carry non-routing metadata.

Quoted 7 lines
Fixes d219df60a70e ("bpf: Add ipip6 and ip6ip decap support for bpf_skb_adjust_room()")
Fixes 1b00e0dfe7d0 ("bpf: update skb->protocol in bpf_skb_net_grow")
Fixes 6578171a7ff0 ("bpf: add bpf_skb_change_proto helper")
#4 lore raw
In reply to Jakub Kicinski
This series was applied to netdev/net.git (main) by Jakub Kicinski <[email protected]>:

Hello:

This series was applied to netdev/net.git (main)
by Jakub Kicinski <[email protected]>:

Quoted 21 lines · On Mon, 9 Jun 2025 17:12:44 -0700 you wrote:

A not-so-careful NAT46 BPF program can crash the kernel if it indiscriminately flips ingress packets from v4 to v6:

  BUG: kernel NULL pointer dereference, address: 0000000000000000
    ip6_rcv_core (net/ipv6/ip6_input.c:190:20)
    ipv6_rcv (net/ipv6/ip6_input.c:306:8)
    process_backlog (net/core/dev.c:6186:4)
    napi_poll (net/core/dev.c:6906:9)
    net_rx_action (net/core/dev.c:7028:13)
    do_softirq (kernel/softirq.c:462:3)
    netif_rx (net/core/dev.c:5326:3)
    dev_loopback_xmit (net/core/dev.c:4015:2)
    ip_mc_finish_output (net/ipv4/ip_output.c:363:8)
    NF_HOOK (./include/linux/netfilter.h:314:9)
    ip_mc_output (net/ipv4/ip_output.c:400:5)
    dst_output (./include/net/dst.h:459:9)
    ip_local_out (net/ipv4/ip_output.c:130:9)
    ip_send_skb (net/ipv4/ip_output.c:1496:8)
    udp_send_skb (net/ipv4/udp.c:1040:8)
    udp_sendmsg (net/ipv4/udp.c:1328:10)
Quoted 1 line

[...]

Here is the summary with links:
- [net,v3,1/2] net: clear the dst when changing skb protocol

    https://git.kernel.org/netdev/net/c/ba9db6f907ac
  - [net,v3,2/2] selftests: net: add test case for NAT46 looping back dst
    https://git.kernel.org/netdev/net/c/567766954b2d

You are awesome, thank you!

Signature · 2 lines
Deet-doot-dot, I am a bot.
https://korg.docs.kernel.org/patchwork/pwbot.html