__inet6_bind() already rejects native IPv6 multicast addresses on SOCK_STREAM sockets with -EINVAL. However, __inet_bind() and the IPV6_ADDR_MAPPED branch of __inet6_bind() allow binding SOCK_STREAM sockets to IPv4 multicast and broadcast addresses. When binding a non-INADDR_ANY RTN_MULTICAST or RTN_BROADCAST address, __inet_bind() stores the non-unicast address in inet->inet_rcv_saddr while clearing inet->inet_saddr to 0. Because bind conflict checks compare receive addresses, binding 224.0.0.1:PORT does not conflict with another user's exclusive unicast 127.0.0.1:PORT socket. On a subsequent active tcp_v4_connect(), seeing !inet->inet_saddr causes the kernel to call inet_bhash2_update_saddr(), relocating the active-open socket into the unicast (port, saddr) bhash2 bucket alongside the exclusive socket without re-evaluating inet_csk_bind_conflict() (note that this affects active-open bhash2 rebinding and port occupancy rather than passive lhash2 SYN delivery). Reject non-INADDR_ANY RTN_MULTICAST and RTN_BROADCAST addresses on SOCK_STREAM sockets with -EINVAL in __inet_bind() and in the IPV6_ADDR_MAPPED path of __inet6_bind(), matching native IPv6 stream socket validation while preserving INADDR_ANY (0.0.0.0) wildcard binds. 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 upon maintainers' request. The fix was compile-tested with -Werror and checked with checkpatch.pl --strict; an equivalent fix was verified against the reproducer in QEMU/KVM. Fixes: 28044fc1d495 ("net: Add a bhash2 table hashed by port and address") Cc: stable@vger.kernel.org Signed-off-by: Dominik Maier --- net/ipv4/af_inet.c | 6 ++++++ net/ipv6/af_inet6.c | 8 ++++++++ 2 files changed, 14 insertions(+) diff --git a/net/ipv4/af_inet.c b/net/ipv4/af_inet.c index 8822bae..b1ea554 100644 --- a/net/ipv4/af_inet.c +++ b/net/ipv4/af_inet.c @@ -500,6 +500,12 @@ int __inet_bind(struct sock *sk, struct sockaddr_unsized *uaddr, int addr_len, tb_id = l3mdev_fib_table_by_index(net, sk->sk_bound_dev_if) ? : tb_id; chk_addr_ret = inet_addr_type_table(net, addr->sin_addr.s_addr, tb_id); + err = -EINVAL; + if (sk->sk_type == SOCK_STREAM && + addr->sin_addr.s_addr != htonl(INADDR_ANY) && + (chk_addr_ret == RTN_MULTICAST || chk_addr_ret == RTN_BROADCAST)) + goto out; + /* Not specified by any standard per-se, however it breaks too * many applications when removed. It is unfortunate since * allowing applications to make a non-local bind solves diff --git a/net/ipv6/af_inet6.c b/net/ipv6/af_inet6.c index a9b8b4b..ab7da0c 100644 --- a/net/ipv6/af_inet6.c +++ b/net/ipv6/af_inet6.c @@ -324,6 +324,14 @@ int __inet6_bind(struct sock *sk, struct sockaddr_unsized *uaddr, int addr_len, chk_addr_ret = inet_addr_type_dev_table(net, dev, v4addr); rcu_read_unlock(); + if (sk->sk_type == SOCK_STREAM && + v4addr != htonl(INADDR_ANY) && + (chk_addr_ret == RTN_MULTICAST || + chk_addr_ret == RTN_BROADCAST)) { + err = -EINVAL; + goto out; + } + if (!inet_addr_valid_or_nonlocal(net, inet, v4addr, chk_addr_ret)) { err = -EADDRNOTAVAIL; -- 2.56.1.750.gf70dc682cf-goog