udp_bpf_recvmsg() calls sk_msg_recvmsg() without holding lock_sock(), unlike tcp_bpf_recvmsg() which properly acquires lock_sock() before calling __sk_msg_recvmsg(). This allows concurrent tasks to race inside sk_msg_recvmsg() on the same psock ingress queue, where one task can free msg_rx via kfree_sk_msg() while another task is still reading it via sk_msg_elem(), causing a slab-use-after-free. Fix this by adding lock_sock()/release_sock() around the sk_msg_recvmsg() path in udp_bpf_recvmsg(), consistent with tcp_bpf_recvmsg(). Also make udp_msg_wait_data() release lock_sock() before sleeping and reacquire it after waking, so it can be called with the socket lock held, consistent with how tcp_msg_wait_data() uses sk_wait_event() which does the same internally. Note: syzbot testing shows a separate pre-existing warning: sk->sk_forward_alloc WARNING: net/ipv4/af_inet.c:162 inet_sock_destruct This warning triggers from the idle CPU path (pv_native_safe_halt) and is unrelated to this patch. It appears to be a pre-existing memory accounting issue in the UDP sockmap path that requires separate investigation. Reported-by: syzbot+431f9a9e3f5227fbb904@syzkaller.appspotmail.com Closes: https://syzkaller.appspot.com/bug?extid=431f9a9e3f5227fbb904 Fixes: 1f5be6b3b063 ("udp: Implement udp_bpf_recvmsg() for sockmap") Signed-off-by: Deepanshu Kartikey --- net/ipv4/udp_bpf.c | 5 +++++ 1 file changed, 5 insertions(+) diff --git a/net/ipv4/udp_bpf.c b/net/ipv4/udp_bpf.c index 9f33b07b1481..f924b255cee6 100644 --- a/net/ipv4/udp_bpf.c +++ b/net/ipv4/udp_bpf.c @@ -50,7 +50,9 @@ static int udp_msg_wait_data(struct sock *sk, struct sk_psock *psock, sk_set_bit(SOCKWQ_ASYNC_WAITDATA, sk); ret = udp_msg_has_data(sk, psock); if (!ret) { + release_sock(sk); wait_woken(&wait, TASK_INTERRUPTIBLE, timeo); + lock_sock(sk); ret = udp_msg_has_data(sk, psock); } sk_clear_bit(SOCKWQ_ASYNC_WAITDATA, sk); @@ -79,6 +81,7 @@ static int udp_bpf_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, goto out; } + lock_sock(sk); msg_bytes_ready: copied = sk_msg_recvmsg(sk, psock, msg, len, flags); if (!copied) { @@ -90,12 +93,14 @@ static int udp_bpf_recvmsg(struct sock *sk, struct msghdr *msg, size_t len, if (data) { if (psock_has_data(psock)) goto msg_bytes_ready; + release_sock(sk); ret = sk_udp_recvmsg(sk, msg, len, flags); goto out; } copied = -EAGAIN; } ret = copied; + release_sock(sk); out: sk_psock_put(sk, psock); return ret; -- 2.43.0