Merge branch 'tcp-rds-fix-use-after-free-around-kernel-tcp-reqsk'

Kuniyuki Iwashima says:

====================
tcp/rds: Fix use-after-free around kernel TCP reqsk.

syzkaller reported an warning of netns ref tracker for RDS TCP listener,
which commit 740ea3c ("tcp: Clean up kernel listener's reqsk in
inet_twsk_purge()") fixed for per-netns ehash.

This series fixes the bug in the partial fix and fixes the reported bug
in the global ehash.

v4: https://lore.kernel.org/netdev/20240307232151.55963-1-kuniyu@amazon.com/
v3: https://lore.kernel.org/netdev/20240307224423.53315-1-kuniyu@amazon.com/
v2: https://lore.kernel.org/netdev/20240227011041.97375-1-kuniyu@amazon.com/
v1: https://lore.kernel.org/netdev/20240223172448.94084-1-kuniyu@amazon.com/
====================

Link: https://lore.kernel.org/r/20240308200122.64357-1-kuniyu@amazon.com
Signed-off-by: Jakub Kicinski <kuba@kernel.org>

Author: kuba-moo

net/ipv4/inet_timewait_sock.c

@@ -263,12 +263,12 @@ void __inet_twsk_schedule(struct inet_timewait_sock *tw, int timeo, bool rearm)
 }
 EXPORT_SYMBOL_GPL(__inet_twsk_schedule);
 
+/* Remove all non full sockets (TIME_WAIT and NEW_SYN_RECV) for dead netns */
 void inet_twsk_purge(struct inet_hashinfo *hashinfo, int family)
 {
-	struct inet_timewait_sock *tw;
-	struct sock *sk;
 	struct hlist_nulls_node *node;
 	unsigned int slot;
+	struct sock *sk;
 
 	for (slot = 0; slot <= hashinfo->ehash_mask; slot++) {
 		struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
@@ -277,38 +277,35 @@ void inet_twsk_purge(struct inet_hashinfo *hashinfo, int family)
 		rcu_read_lock();
 restart:
 		sk_nulls_for_each_rcu(sk, node, &head->chain) {
-			if (sk->sk_state != TCP_TIME_WAIT) {
-				/* A kernel listener socket might not hold refcnt for net,
-				 * so reqsk_timer_handler() could be fired after net is
-				 * freed.  Userspace listener and reqsk never exist here.
-				 */
-				if (unlikely(sk->sk_state == TCP_NEW_SYN_RECV &&
-					     hashinfo->pernet)) {
-					struct request_sock *req = inet_reqsk(sk);
-
-					inet_csk_reqsk_queue_drop_and_put(req->rsk_listener, req);
-				}
+			int state = inet_sk_state_load(sk);
 
+			if ((1 << state) & ~(TCPF_TIME_WAIT |
+					     TCPF_NEW_SYN_RECV))
 				continue;
-			}
 
-			tw = inet_twsk(sk);
-			if ((tw->tw_family != family) ||
-				refcount_read(&twsk_net(tw)->ns.count))
+			if (sk->sk_family != family ||
+			    refcount_read(&sock_net(sk)->ns.count))
 				continue;
 
-			if (unlikely(!refcount_inc_not_zero(&tw->tw_refcnt)))
+			if (unlikely(!refcount_inc_not_zero(&sk->sk_refcnt)))
 				continue;
 
-			if (unlikely((tw->tw_family != family) ||
-				     refcount_read(&twsk_net(tw)->ns.count))) {
-				inet_twsk_put(tw);
+			if (unlikely(sk->sk_family != family ||
+				     refcount_read(&sock_net(sk)->ns.count))) {
+				sock_gen_put(sk);
 				goto restart;
 			}
 
 			rcu_read_unlock();
 			local_bh_disable();
-			inet_twsk_deschedule_put(tw);
+			if (state == TCP_TIME_WAIT) {
+				inet_twsk_deschedule_put(inet_twsk(sk));
+			} else {
+				struct request_sock *req = inet_reqsk(sk);
+
+				inet_csk_reqsk_queue_drop_and_put(req->rsk_listener,
+								  req);
+			}
 			local_bh_enable();
 			goto restart_rcu;
 		}

net/ipv4/tcp_minisocks.c

@@ -398,10 +398,6 @@ void tcp_twsk_purge(struct list_head *net_exit_list, int family)
 			/* Even if tw_refcount == 1, we must clean up kernel reqsk */
 			inet_twsk_purge(net->ipv4.tcp_death_row.hashinfo, family);
 		} else if (!purged_once) {
-			/* The last refcount is decremented in tcp_sk_exit_batch() */
-			if (refcount_read(&net->ipv4.tcp_death_row.tw_refcount) == 1)
-				continue;
-
 			inet_twsk_purge(&tcp_hashinfo, family);
 			purged_once = true;
 		}