[PATCH net v2] net: vxlan: prevent NULL deref in vxlan_xmit_one

[PATCH net v2] net: vxlan: prevent NULL deref in vxlan_xmit_one

[Antoine Tenart]

Neither sock4 nor sock6 pointers are guaranteed to be non-NULL in
vxlan_xmit_one, e.g. if the iface is brought down. This can lead to the
following NULL dereference:

  BUG: kernel NULL pointer dereference, address: 0000000000000010
  Oops: Oops: 0000 [#1] SMP NOPTI
  RIP: 0010:vxlan_xmit_one+0xbb3/0x1580
  Call Trace:
   vxlan_xmit+0x429/0x610
   dev_hard_start_xmit+0x55/0xa0
   __dev_queue_xmit+0x6d0/0x7f0
   ip_finish_output2+0x24b/0x590
   ip_output+0x63/0x110

Mentioned commits changed the code path in vxlan_xmit_one and as a side
effect the sock4/6 pointer validity checks in vxlan(6)_get_route were
lost. Fix this by adding back checks.

Since both commits being fixed were released in the same version (v6.7)
and are strongly related, bundle the fixes in a single commit.

Reported-by: Liang Li <liali@redhat.com>
Fixes: 6f19b2c136d9 ("vxlan: use generic function for tunnel IPv4 route lookup")
Fixes: 2aceb896ee18 ("vxlan: use generic function for tunnel IPv6 route lookup")
Cc: Beniamino Galvani <b.galvani@gmail.com>
Signed-off-by: Antoine Tenart <atenart@kernel.org>
---

Since v1:
- Fixed uninitialized err variable reported by clang [Jakub]

 drivers/net/vxlan/vxlan_core.c | 18 +++++++++++++++---
 1 file changed, 15 insertions(+), 3 deletions(-)

diff --git a/drivers/net/vxlan/vxlan_core.c b/drivers/net/vxlan/vxlan_core.c
index a5c55e7e4d79..e957aa12a8a4 100644
--- a/drivers/net/vxlan/vxlan_core.c
+++ b/drivers/net/vxlan/vxlan_core.c
@@ -2349,7 +2349,7 @@ void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
 	int addr_family;
 	__u8 tos, ttl;
 	int ifindex;
-	int err;
+	int err = 0;
 	u32 flags = vxlan->cfg.flags;
 	bool use_cache;
 	bool udp_sum = false;
@@ -2454,12 +2454,18 @@ void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
 
 	rcu_read_lock();
 	if (addr_family == AF_INET) {
-		struct vxlan_sock *sock4 = rcu_dereference(vxlan->vn4_sock);
+		struct vxlan_sock *sock4;
 		u16 ipcb_flags = 0;
 		struct rtable *rt;
 		__be16 df = 0;
 		__be32 saddr;
 
+		sock4 = rcu_dereference(vxlan->vn4_sock);
+		if (unlikely(!sock4)) {
+			reason = SKB_DROP_REASON_DEV_READY;
+			goto tx_error;
+		}
+
 		if (!ifindex)
 			ifindex = sock4->sock->sk->sk_bound_dev_if;
 
@@ -2534,10 +2540,16 @@ void vxlan_xmit_one(struct sk_buff *skb, struct net_device *dev,
 				    ipcb_flags);
 #if IS_ENABLED(CONFIG_IPV6)
 	} else {
-		struct vxlan_sock *sock6 = rcu_dereference(vxlan->vn6_sock);
+		struct vxlan_sock *sock6;
 		struct in6_addr saddr;
 		u16 ip6cb_flags = 0;
 
+		sock6 = rcu_dereference(vxlan->vn6_sock);
+		if (unlikely(!sock6)) {
+			reason = SKB_DROP_REASON_DEV_READY;
+			goto tx_error;
+		}
+
 		if (!ifindex)
 			ifindex = sock6->sock->sk->sk_bound_dev_if;
 
-- 
2.52.0

Re: [PATCH net v2] net: vxlan: prevent NULL deref in vxlan_xmit_one

[Ido Schimmel]

On Wed, Nov 26, 2025 at 11:26:25AM +0100, Antoine Tenart wrote:
> Neither sock4 nor sock6 pointers are guaranteed to be non-NULL in
> vxlan_xmit_one, e.g. if the iface is brought down. This can lead to the
> following NULL dereference:
> 
>   BUG: kernel NULL pointer dereference, address: 0000000000000010
>   Oops: Oops: 0000 [#1] SMP NOPTI
>   RIP: 0010:vxlan_xmit_one+0xbb3/0x1580
>   Call Trace:
>    vxlan_xmit+0x429/0x610
>    dev_hard_start_xmit+0x55/0xa0
>    __dev_queue_xmit+0x6d0/0x7f0
>    ip_finish_output2+0x24b/0x590
>    ip_output+0x63/0x110
> 
> Mentioned commits changed the code path in vxlan_xmit_one and as a side
> effect the sock4/6 pointer validity checks in vxlan(6)_get_route were
> lost. Fix this by adding back checks.
> 
> Since both commits being fixed were released in the same version (v6.7)
> and are strongly related, bundle the fixes in a single commit.
> 
> Reported-by: Liang Li <liali@redhat.com>
> Fixes: 6f19b2c136d9 ("vxlan: use generic function for tunnel IPv4 route lookup")
> Fixes: 2aceb896ee18 ("vxlan: use generic function for tunnel IPv6 route lookup")
> Cc: Beniamino Galvani <b.galvani@gmail.com>
> Signed-off-by: Antoine Tenart <atenart@kernel.org>

Reviewed-by: Ido Schimmel <idosch@nvidia.com>
Tested-by: Ido Schimmel <idosch@nvidia.com>

Tested using:

ip link add name vx0 up type vxlan id 10010 local 192.0.2.1 remote 198.51.100.1 dstport 4789
for i in $(seq 0 $(($(nproc) - 2))); do
	taskset -c $i mausezahn vx0 -a own -b 00:11:22:33:44:55 -c 0 -q &
done
taskset -c $(($(nproc) - 1)) bash -c "while true; do ip link set dev vx0 down && ip link set dev vx0 up; done"

Re: [PATCH net v2] net: vxlan: prevent NULL deref in vxlan_xmit_one

[patchwork-bot+netdevbpf]

Hello:

This patch was applied to netdev/net.git (main)
by Jakub Kicinski <kuba@kernel.org>:

On Wed, 26 Nov 2025 11:26:25 +0100 you wrote:
> Neither sock4 nor sock6 pointers are guaranteed to be non-NULL in
> vxlan_xmit_one, e.g. if the iface is brought down. This can lead to the
> following NULL dereference:
> 
>   BUG: kernel NULL pointer dereference, address: 0000000000000010
>   Oops: Oops: 0000 [#1] SMP NOPTI
>   RIP: 0010:vxlan_xmit_one+0xbb3/0x1580
>   Call Trace:
>    vxlan_xmit+0x429/0x610
>    dev_hard_start_xmit+0x55/0xa0
>    __dev_queue_xmit+0x6d0/0x7f0
>    ip_finish_output2+0x24b/0x590
>    ip_output+0x63/0x110
> 
> [...]

Here is the summary with links:
  - [net,v2] net: vxlan: prevent NULL deref in vxlan_xmit_one
    https://git.kernel.org/netdev/net/c/1f73a56f9860

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