netdev - [PATCH v2 net 3/4] af_unix: Don't set -ECONNRESET for consumed OOB skb.

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Message-ID: <20250619041457.1132791-4-kuni1840@gmail.com>
Date: Wed, 18 Jun 2025 21:13:57 -0700
From: Kuniyuki Iwashima <kuni1840@...il.com>
To: "David S. Miller" <davem@...emloft.net>,
	Eric Dumazet <edumazet@...gle.com>,
	Jakub Kicinski <kuba@...nel.org>,
	Paolo Abeni <pabeni@...hat.com>
Cc: Simon Horman <horms@...nel.org>,
	Rao Shoaib <rao.shoaib@...cle.com>,
	Kuniyuki Iwashima <kuniyu@...gle.com>,
	Kuniyuki Iwashima <kuni1840@...il.com>,
	netdev@...r.kernel.org,
	Christian Brauner <brauner@...nel.org>
Subject: [PATCH v2 net 3/4] af_unix: Don't set -ECONNRESET for consumed OOB skb.

From: Kuniyuki Iwashima <kuniyu@...gle.com>

Christian Brauner reported that even after MSG_OOB data is consumed,
calling close() on the receiver socket causes the peer's recv() to
return -ECONNRESET:

  1. send() and recv() an OOB data.

    >>> from socket import *
    >>> s1, s2 = socketpair(AF_UNIX, SOCK_STREAM)
    >>> s1.send(b'x', MSG_OOB)
    1
    >>> s2.recv(1, MSG_OOB)
    b'x'

  2. close() for s2 sets ECONNRESET to s1->sk_err even though
     s2 consumed the OOB data

    >>> s2.close()
    >>> s1.recv(10, MSG_DONTWAIT)
    ...
    ConnectionResetError: [Errno 104] Connection reset by peer

Even after being consumed, the skb holding the OOB 1-byte data stays in
the recv queue to mark the OOB boundary and break recv() at that point.

This must be considered while close()ing a socket.

Let's skip the leading consumed OOB skb while checking the -ECONNRESET
condition in unix_release_sock().

Fixes: 314001f0bf92 ("af_unix: Add OOB support")
Reported-by: Christian Brauner <brauner@...nel.org>
Closes: https://lore.kernel.org/netdev/20250529-sinkt-abfeuern-e7b08200c6b0@brauner/
Signed-off-by: Kuniyuki Iwashima <kuniyu@...gle.com>
---
v2: Reuse the existing path for non-listener and skip the leading
    consumed OOB skb instead of freeing it

v1: https://lore.kernel.org/netdev/20250618043453.281247-1-kuni1840@gmail.com/
---
 net/unix/af_unix.c | 18 ++++++++++++------
 1 file changed, 12 insertions(+), 6 deletions(-)

diff --git a/net/unix/af_unix.c b/net/unix/af_unix.c
index 5392aa53cbc8..52b155123985 100644
--- a/net/unix/af_unix.c
+++ b/net/unix/af_unix.c
@@ -660,6 +660,11 @@ static void unix_sock_destructor(struct sock *sk)
 #endif
 }
 
+static unsigned int unix_skb_len(const struct sk_buff *skb)
+{
+	return skb->len - UNIXCB(skb).consumed;
+}
+
 static void unix_release_sock(struct sock *sk, int embrion)
 {
 	struct unix_sock *u = unix_sk(sk);
@@ -694,10 +699,16 @@ static void unix_release_sock(struct sock *sk, int embrion)
 
 	if (skpair != NULL) {
 		if (sk->sk_type == SOCK_STREAM || sk->sk_type == SOCK_SEQPACKET) {
+			struct sk_buff *skb = skb_peek(&sk->sk_receive_queue);
+
+#if IS_ENABLED(CONFIG_AF_UNIX_OOB)
+			if (skb && !unix_skb_len(skb))
+				skb = skb_peek_next(skb, &sk->sk_receive_queue);
+#endif
 			unix_state_lock(skpair);
 			/* No more writes */
 			WRITE_ONCE(skpair->sk_shutdown, SHUTDOWN_MASK);
-			if (!skb_queue_empty_lockless(&sk->sk_receive_queue) || embrion)
+			if (skb || embrion)
 				WRITE_ONCE(skpair->sk_err, ECONNRESET);
 			unix_state_unlock(skpair);
 			skpair->sk_state_change(skpair);
@@ -2661,11 +2672,6 @@ static long unix_stream_data_wait(struct sock *sk, long timeo,
 	return timeo;
 }
 
-static unsigned int unix_skb_len(const struct sk_buff *skb)
-{
-	return skb->len - UNIXCB(skb).consumed;
-}
-
 struct unix_stream_read_state {
 	int (*recv_actor)(struct sk_buff *, int, int,
 			  struct unix_stream_read_state *);
-- 
2.49.0