Netdev List
 help / color / mirror / Atom feed
* [PATCH net] udp: preserve explicit sk_bound_dev_if across __udp_disconnect()
@ 2026-11-19 19:21 Dominik Maier
  0 siblings, 0 replies; only message in thread
From: Dominik Maier @ 2026-11-19 19:21 UTC (permalink / raw)
  To: Willem de Bruijn, Eric Dumazet, Kuniyuki Iwashima, Paolo Abeni,
	David S. Miller, Jakub Kicinski
  Cc: Simon Horman, David Ahern, netdev, dmnk+artist, Dominik Maier

sock_bindtoindex_locked() allows unprivileged users to bind an unbound
socket to a network interface via SO_BINDTODEVICE or SO_BINDTOIFINDEX,
and enforces CAP_NET_RAW before unbinding or changing sk_bound_dev_if
once set. Sockets may also have sk_bound_dev_if set via a scoped bind()
address or via cgroup BPF (such as ip vrf exec).

However, __udp_disconnect() unconditionally zeroes sk->sk_bound_dev_if
when disconnecting via AF_UNSPEC, which is only intended to undo
implicit scoped-address bindings assigned at connect() time by
__ip6_datagram_connect(). When a socket was also explicitly bound to an
address and port (SOCK_BINDADDR_LOCK | SOCK_BINDPORT_LOCK), it remains
hashed at the head of its UDP portaddr hash chain with sk_bound_dev_if
reset to 0. An unprivileged local user can therefore bind to a disjoint
interface index alongside another user's exclusive SO_BINDTODEVICE or
scoped UDP listener, disconnect via AF_UNSPEC, and intercept cross-UID
UDP packets on all interfaces (or clear a BPF-assigned VRF binding).

Add SOCK_CONNECT_IFINDEX to sk->sk_userlocks (matching
SOCK_CONNECT_BIND), set it in __ip6_datagram_connect() only when
sk->sk_bound_dev_if transitions from 0 to non-zero (clearing it again if
ip6_datagram_dst_update() fails), clear it when an explicit interface or
scoped bind overrides sk_bound_dev_if, and only reset
sk->sk_bound_dev_if in __udp_disconnect() when SOCK_CONNECT_IFINDEX is
set.

This issue was discovered with an AI-assisted code review tool (ARTIST)
and reproduced in QEMU/KVM on an unmodified v7.4-rc3 kernel without
kernel modifications or error injection; a standalone reproducer is
available on request. This revised SOCK_CONNECT_IFINDEX patch was
compile-tested with -Werror and checked with checkpatch.pl --strict; it
has not yet been runtime-tested against the reproducer or kernel
selftests.

Fixes: c427bfec18f2 ("net: core: enable SO_BINDTODEVICE for non-root users")
Cc: stable@vger.kernel.org
Signed-off-by: Dominik Maier <dmnk@google.com>
---
 include/net/sock.h  | 4 ++++
 net/core/sock.c     | 1 +
 net/ipv4/udp.c      | 5 ++++-
 net/ipv6/af_inet6.c | 1 +
 net/ipv6/datagram.c | 4 ++++
 net/ipv6/raw.c      | 1 +
 6 files changed, 15 insertions(+), 1 deletion(-)

diff --git a/include/net/sock.h b/include/net/sock.h
index 3144efa..5700807 100644
--- a/include/net/sock.h
+++ b/include/net/sock.h
@@ -1534,6 +1534,10 @@ static inline int __sk_prot_rehash(struct sock *sk)
  * define SOCK_CONNECT_BIND - &sock->sk_userlocks flag for auto-bind at connect() time
  */
 #define SOCK_CONNECT_BIND	16
+/**
+ * define SOCK_CONNECT_IFINDEX - &sock->sk_userlocks flag for scoped connect() ifindex
+ */
+#define SOCK_CONNECT_IFINDEX	32
 
 struct socket_alloc {
 	struct socket socket;
diff --git a/net/core/sock.c b/net/core/sock.c
index ad41487..211b7f1 100644
--- a/net/core/sock.c
+++ b/net/core/sock.c
@@ -650,6 +650,7 @@ static int sock_bindtoindex_locked(struct sock *sk, int ifindex)
 
 	/* Paired with all READ_ONCE() done locklessly. */
 	WRITE_ONCE(sk->sk_bound_dev_if, ifindex);
+	sk->sk_userlocks &= ~SOCK_CONNECT_IFINDEX;
 
 	if (sk->sk_prot->rehash)
 		sk->sk_prot->rehash(sk);
diff --git a/net/ipv4/udp.c b/net/ipv4/udp.c
index 2e5562f..4943afe 100644
--- a/net/ipv4/udp.c
+++ b/net/ipv4/udp.c
@@ -2190,7 +2190,10 @@ int __udp_disconnect(struct sock *sk, int flags)
 	WRITE_ONCE(inet->inet_daddr, 0);
 	inet->inet_dport = 0;
 	sock_rps_reset_rxhash(sk);
-	WRITE_ONCE(sk->sk_bound_dev_if, 0);
+	if (sk->sk_userlocks & SOCK_CONNECT_IFINDEX) {
+		WRITE_ONCE(sk->sk_bound_dev_if, 0);
+		sk->sk_userlocks &= ~SOCK_CONNECT_IFINDEX;
+	}
 	if (!(sk->sk_userlocks & SOCK_BINDADDR_LOCK)) {
 		inet_reset_saddr(sk);
 		if (sk->sk_prot->rehash &&
diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c
index a9b8b4b..b402f0e 100644
--- a/net/ipv6/af_inet6.c
+++ b/net/ipv6/af_inet6.c
@@ -341,6 +341,7 @@ int __inet6_bind(struct sock *sk, struct sockaddr_unsized *uaddr, int addr_len,
 					 * is supplied by user.
 					 */
 					sk->sk_bound_dev_if = addr->sin6_scope_id;
+					sk->sk_userlocks &= ~SOCK_CONNECT_IFINDEX;
 				}
 
 				/* Binding to link-local address requires an interface */
diff --git a/net/ipv6/datagram.c b/net/ipv6/datagram.c
index ff7a7ff..36581c6 100644
--- a/net/ipv6/datagram.c
+++ b/net/ipv6/datagram.c
@@ -240,6 +240,8 @@ int __ip6_datagram_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
 			err = -EINVAL;
 			goto out;
 		}
+		if (!old_bound_dev_if)
+			sk->sk_userlocks |= SOCK_CONNECT_IFINDEX;
 	}
 
 	/* save the current peer information before updating it */
@@ -264,6 +266,8 @@ int __ip6_datagram_connect(struct sock *sk, struct sockaddr_unsized *uaddr,
 		sk->sk_v6_daddr = old_daddr;
 		np->flow_label = old_fl6_flowlabel;
 		inet->inet_dport = old_dport;
+		if (!old_bound_dev_if)
+			sk->sk_userlocks &= ~SOCK_CONNECT_IFINDEX;
 		WRITE_ONCE(sk->sk_bound_dev_if, old_bound_dev_if);
 		goto out;
 	}
diff --git a/net/ipv6/raw.c b/net/ipv6/raw.c
index 00d832c..643f732 100644
--- a/net/ipv6/raw.c
+++ b/net/ipv6/raw.c
@@ -255,6 +255,7 @@ static int rawv6_bind(struct sock *sk, struct sockaddr_unsized *uaddr,
 				 * one is supplied by user.
 				 */
 				sk->sk_bound_dev_if = addr->sin6_scope_id;
+				sk->sk_userlocks &= ~SOCK_CONNECT_IFINDEX;
 			}
 
 			/* Binding to link-local address requires an interface */
-- 
2.56.1.750.gf70dc682cf-goog


^ permalink raw reply related	[flat|nested] only message in thread

only message in thread, other threads:[~2026-11-19 19:21 UTC | newest]

Thread overview: (only message) (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2026-11-19 19:21 [PATCH net] udp: preserve explicit sk_bound_dev_if across __udp_disconnect() Dominik Maier

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox