Re: [PATCH] net: inet: prevent lhash2 lookups from escaping into ehash

From: David Lee <[email protected]>
Date: Fri, 31 Jul 2026 14:05:12 +0000
> The lhash2 and ehash tables share sk_nulls_node. When a listening
> socket is unhashed and rehashed as a connected socket

This is only fuzzing scenario, and there is another series trying
to address the same class of issue by touching the fast path as well.

https://lore.kernel.org/netdev/[email protected]/


> while an RCU
> listener lookup is walking it, the reader can follow the node's new
> next pointer into ehash. inet_lhash2_lookup() and
> inet6_lhash2_lookup() do not validate the terminal nulls marker, so
> they can return an ehash entry under the listener lookup's
> unreferenced ownership contract.
> 
> In particular, returning a TIME_WAIT socket makes the TCP receive
> path consume a reference that the lookup never acquired. Repeated
> races can free the object while it remains linked in the hash tables.
> 
> Skip non-listening sockets before scoring them. Also restart the walk
> when its terminal nulls marker does not match the expected lhash2 slot,
> as the established lookup already does.
> 
> Fixes: cae3873c5b3a ("net: inet: Retire port only listening_hash")
> Bug found and triaged by OpenAI Security Research and 
> validated by Trail of Bits.
> 
> Assisted-by: Codex:gpt-5.6-sol gpt-5.5-cyber
> Signed-off-by: Kyle Zeng <[email protected]>
> ---
> Trail of Bits has a reproducer for this bug that triggers a
> KASAN use-after-free and can share if needed.

Can you test this diff ?

---8<---
diff --git a/include/net/inet_connection_sock.h b/include/net/inet_connection_sock.h
index 433c2df23076..903499581db7 100644
--- a/include/net/inet_connection_sock.h
+++ b/include/net/inet_connection_sock.h
@@ -94,6 +94,7 @@ struct inet_connection_sock {
 	u32			  icsk_rto_max;
 	__u32                     icsk_delack_max;
 	__u32			  icsk_pmtu_cookie;
+	unsigned char		  unhashed_state;
 	const struct tcp_congestion_ops *icsk_ca_ops;
 	const struct inet_connection_sock_af_ops *icsk_af_ops;
 	const struct tcp_ulp_ops  *icsk_ulp_ops;
diff --git a/net/ipv4/inet_hashtables.c b/net/ipv4/inet_hashtables.c
index ba0faa9ae2bb..3ed8d5833c3b 100644
--- a/net/ipv4/inet_hashtables.c
+++ b/net/ipv4/inet_hashtables.c
@@ -803,6 +803,14 @@ int inet_hash(struct sock *sk)
 	inet_init_ehash_secret();
 
 	WARN_ON(!sk_unhashed(sk));
+
+	if (unlikely(inet_csk(sk)->unhashed_state)) {
+		if (inet_csk(sk)->unhashed_state != TCP_LISTEN)
+			synchronize_rcu();
+
+		inet_csk(sk)->unhashed_state = 0;
+	}
+
 	ilb2 = inet_lhash2_bucket_sk(hashinfo, sk);
 
 	spin_lock(&ilb2->lock);
@@ -832,6 +840,9 @@ void inet_unhash(struct sock *sk)
 		return;
 
 	sock_rps_delete_flow(sk);
+
+	inet_csk(sk)->unhashed_state = sk->sk_state;
+
 	if (sk->sk_state == TCP_LISTEN) {
 		struct inet_listen_hashbucket *ilb2;
 
@@ -1058,6 +1069,13 @@ int __inet_hash_connect(struct inet_timewait_death_row *death_row,
 	int ret, i, low, high;
 	bool local_ports;
 
+	if (unlikely(inet_csk(sk)->unhashed_state)) {
+		if (inet_csk(sk)->unhashed_state == TCP_LISTEN)
+			synchronize_rcu();
+
+		inet_csk(sk)->unhashed_state = 0;
+	}
+
 	if (port) {
 		local_bh_disable();
 		ret = check_established(death_row, sk, port, NULL, false,
---8<---