1// SPDX-License-Identifier: GPL-2.0-only
2/*
3 * INET		An implementation of the TCP/IP protocol suite for the LINUX
4 *		operating system.  INET is implemented using the  BSD Socket
5 *		interface as the means of communication with the user level.
6 *
7 *		Generic TIME_WAIT sockets functions
8 *
9 *		From code orinally in TCP
10 */
11
12#include <linux/kernel.h>
13#include <linux/slab.h>
14#include <linux/module.h>
15#include <net/inet_hashtables.h>
16#include <net/inet_timewait_sock.h>
17#include <net/ip.h>
18
19
20/**
21 *	inet_twsk_bind_unhash - unhash a timewait socket from bind hash
22 *	@tw: timewait socket
23 *	@hashinfo: hashinfo pointer
24 *
25 *	unhash a timewait socket from bind hash, if hashed.
26 *	bind hash lock must be held by caller.
27 *	Returns 1 if caller should call inet_twsk_put() after lock release.
28 */
29void inet_twsk_bind_unhash(struct inet_timewait_sock *tw,
30			  struct inet_hashinfo *hashinfo)
31{
32	struct inet_bind2_bucket *tb2 = tw->tw_tb2;
33	struct inet_bind_bucket *tb = tw->tw_tb;
34
35	if (!tb)
36		return;
37
38	__hlist_del(&tw->tw_bind_node);
39	tw->tw_tb = NULL;
40	inet_bind_bucket_destroy(hashinfo->bind_bucket_cachep, tb);
41
42	__hlist_del(&tw->tw_bind2_node);
43	tw->tw_tb2 = NULL;
44	inet_bind2_bucket_destroy(hashinfo->bind2_bucket_cachep, tb2);
45
46	__sock_put((struct sock *)tw);
47}
48
49/* Must be called with locally disabled BHs. */
50static void inet_twsk_kill(struct inet_timewait_sock *tw)
51{
52	struct inet_hashinfo *hashinfo = tw->tw_dr->hashinfo;
53	spinlock_t *lock = inet_ehash_lockp(hashinfo, tw->tw_hash);
54	struct inet_bind_hashbucket *bhead, *bhead2;
55
56	spin_lock(lock);
57	sk_nulls_del_node_init_rcu((struct sock *)tw);
58	spin_unlock(lock);
59
60	/* Disassociate with bind bucket. */
61	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), tw->tw_num,
62			hashinfo->bhash_size)];
63	bhead2 = inet_bhashfn_portaddr(hashinfo, (struct sock *)tw,
64				       twsk_net(tw), tw->tw_num);
65
66	spin_lock(&bhead->lock);
67	spin_lock(&bhead2->lock);
68	inet_twsk_bind_unhash(tw, hashinfo);
69	spin_unlock(&bhead2->lock);
70	spin_unlock(&bhead->lock);
71
72	refcount_dec(&tw->tw_dr->tw_refcount);
73	inet_twsk_put(tw);
74}
75
76void inet_twsk_free(struct inet_timewait_sock *tw)
77{
78	struct module *owner = tw->tw_prot->owner;
79	twsk_destructor((struct sock *)tw);
80	kmem_cache_free(tw->tw_prot->twsk_prot->twsk_slab, tw);
81	module_put(owner);
82}
83
84void inet_twsk_put(struct inet_timewait_sock *tw)
85{
86	if (refcount_dec_and_test(&tw->tw_refcnt))
87		inet_twsk_free(tw);
88}
89EXPORT_SYMBOL_GPL(inet_twsk_put);
90
91static void inet_twsk_add_node_rcu(struct inet_timewait_sock *tw,
92				   struct hlist_nulls_head *list)
93{
94	hlist_nulls_add_head_rcu(&tw->tw_node, list);
95}
96
97static void inet_twsk_add_bind_node(struct inet_timewait_sock *tw,
98				    struct hlist_head *list)
99{
100	hlist_add_head(&tw->tw_bind_node, list);
101}
102
103static void inet_twsk_add_bind2_node(struct inet_timewait_sock *tw,
104				     struct hlist_head *list)
105{
106	hlist_add_head(&tw->tw_bind2_node, list);
107}
108
109/*
110 * Enter the time wait state. This is called with locally disabled BH.
111 * Essentially we whip up a timewait bucket, copy the relevant info into it
112 * from the SK, and mess with hash chains and list linkage.
113 */
114void inet_twsk_hashdance(struct inet_timewait_sock *tw, struct sock *sk,
115			   struct inet_hashinfo *hashinfo)
116{
117	const struct inet_sock *inet = inet_sk(sk);
118	const struct inet_connection_sock *icsk = inet_csk(sk);
119	struct inet_ehash_bucket *ehead = inet_ehash_bucket(hashinfo, sk->sk_hash);
120	spinlock_t *lock = inet_ehash_lockp(hashinfo, sk->sk_hash);
121	struct inet_bind_hashbucket *bhead, *bhead2;
122
123	/* Step 1: Put TW into bind hash. Original socket stays there too.
124	   Note, that any socket with inet->num != 0 MUST be bound in
125	   binding cache, even if it is closed.
126	 */
127	bhead = &hashinfo->bhash[inet_bhashfn(twsk_net(tw), inet->inet_num,
128			hashinfo->bhash_size)];
129	bhead2 = inet_bhashfn_portaddr(hashinfo, sk, twsk_net(tw), inet->inet_num);
130
131	spin_lock(&bhead->lock);
132	spin_lock(&bhead2->lock);
133
134	tw->tw_tb = icsk->icsk_bind_hash;
135	WARN_ON(!icsk->icsk_bind_hash);
136	inet_twsk_add_bind_node(tw, &tw->tw_tb->owners);
137
138	tw->tw_tb2 = icsk->icsk_bind2_hash;
139	WARN_ON(!icsk->icsk_bind2_hash);
140	inet_twsk_add_bind2_node(tw, &tw->tw_tb2->deathrow);
141
142	spin_unlock(&bhead2->lock);
143	spin_unlock(&bhead->lock);
144
145	spin_lock(lock);
146
147	inet_twsk_add_node_rcu(tw, &ehead->chain);
148
149	/* Step 3: Remove SK from hash chain */
150	if (__sk_nulls_del_node_init_rcu(sk))
151		sock_prot_inuse_add(sock_net(sk), sk->sk_prot, -1);
152
153	spin_unlock(lock);
154
155	/* tw_refcnt is set to 3 because we have :
156	 * - one reference for bhash chain.
157	 * - one reference for ehash chain.
158	 * - one reference for timer.
159	 * We can use atomic_set() because prior spin_lock()/spin_unlock()
160	 * committed into memory all tw fields.
161	 * Also note that after this point, we lost our implicit reference
162	 * so we are not allowed to use tw anymore.
163	 */
164	refcount_set(&tw->tw_refcnt, 3);
165}
166EXPORT_SYMBOL_GPL(inet_twsk_hashdance);
167
168static void tw_timer_handler(struct timer_list *t)
169{
170	struct inet_timewait_sock *tw = from_timer(tw, t, tw_timer);
171
172	inet_twsk_kill(tw);
173}
174
175struct inet_timewait_sock *inet_twsk_alloc(const struct sock *sk,
176					   struct inet_timewait_death_row *dr,
177					   const int state)
178{
179	struct inet_timewait_sock *tw;
180
181	if (refcount_read(&dr->tw_refcount) - 1 >=
182	    READ_ONCE(dr->sysctl_max_tw_buckets))
183		return NULL;
184
185	tw = kmem_cache_alloc(sk->sk_prot_creator->twsk_prot->twsk_slab,
186			      GFP_ATOMIC);
187	if (tw) {
188		const struct inet_sock *inet = inet_sk(sk);
189
190		tw->tw_dr	    = dr;
191		/* Give us an identity. */
192		tw->tw_daddr	    = inet->inet_daddr;
193		tw->tw_rcv_saddr    = inet->inet_rcv_saddr;
194		tw->tw_bound_dev_if = sk->sk_bound_dev_if;
195		tw->tw_tos	    = inet->tos;
196		tw->tw_num	    = inet->inet_num;
197		tw->tw_state	    = TCP_TIME_WAIT;
198		tw->tw_substate	    = state;
199		tw->tw_sport	    = inet->inet_sport;
200		tw->tw_dport	    = inet->inet_dport;
201		tw->tw_family	    = sk->sk_family;
202		tw->tw_reuse	    = sk->sk_reuse;
203		tw->tw_reuseport    = sk->sk_reuseport;
204		tw->tw_hash	    = sk->sk_hash;
205		tw->tw_ipv6only	    = 0;
206		tw->tw_transparent  = inet_test_bit(TRANSPARENT, sk);
207		tw->tw_prot	    = sk->sk_prot_creator;
208		atomic64_set(&tw->tw_cookie, atomic64_read(&sk->sk_cookie));
209		twsk_net_set(tw, sock_net(sk));
210		timer_setup(&tw->tw_timer, tw_timer_handler, TIMER_PINNED);
211		/*
212		 * Because we use RCU lookups, we should not set tw_refcnt
213		 * to a non null value before everything is setup for this
214		 * timewait socket.
215		 */
216		refcount_set(&tw->tw_refcnt, 0);
217
218		__module_get(tw->tw_prot->owner);
219	}
220
221	return tw;
222}
223EXPORT_SYMBOL_GPL(inet_twsk_alloc);
224
225/* These are always called from BH context.  See callers in
226 * tcp_input.c to verify this.
227 */
228
229/* This is for handling early-kills of TIME_WAIT sockets.
230 * Warning : consume reference.
231 * Caller should not access tw anymore.
232 */
233void inet_twsk_deschedule_put(struct inet_timewait_sock *tw)
234{
235	if (del_timer_sync(&tw->tw_timer))
236		inet_twsk_kill(tw);
237	inet_twsk_put(tw);
238}
239EXPORT_SYMBOL(inet_twsk_deschedule_put);
240
241void __inet_twsk_schedule(struct inet_timewait_sock *tw, int timeo, bool rearm)
242{
243	/* timeout := RTO * 3.5
244	 *
245	 * 3.5 = 1+2+0.5 to wait for two retransmits.
246	 *
247	 * RATIONALE: if FIN arrived and we entered TIME-WAIT state,
248	 * our ACK acking that FIN can be lost. If N subsequent retransmitted
249	 * FINs (or previous seqments) are lost (probability of such event
250	 * is p^(N+1), where p is probability to lose single packet and
251	 * time to detect the loss is about RTO*(2^N - 1) with exponential
252	 * backoff). Normal timewait length is calculated so, that we
253	 * waited at least for one retransmitted FIN (maximal RTO is 120sec).
254	 * [ BTW Linux. following BSD, violates this requirement waiting
255	 *   only for 60sec, we should wait at least for 240 secs.
256	 *   Well, 240 consumes too much of resources 8)
257	 * ]
258	 * This interval is not reduced to catch old duplicate and
259	 * responces to our wandering segments living for two MSLs.
260	 * However, if we use PAWS to detect
261	 * old duplicates, we can reduce the interval to bounds required
262	 * by RTO, rather than MSL. So, if peer understands PAWS, we
263	 * kill tw bucket after 3.5*RTO (it is important that this number
264	 * is greater than TS tick!) and detect old duplicates with help
265	 * of PAWS.
266	 */
267
268	if (!rearm) {
269		bool kill = timeo <= 4*HZ;
270
271		__NET_INC_STATS(twsk_net(tw), kill ? LINUX_MIB_TIMEWAITKILLED :
272						     LINUX_MIB_TIMEWAITED);
273		BUG_ON(mod_timer(&tw->tw_timer, jiffies + timeo));
274		refcount_inc(&tw->tw_dr->tw_refcount);
275	} else {
276		mod_timer_pending(&tw->tw_timer, jiffies + timeo);
277	}
278}
279EXPORT_SYMBOL_GPL(__inet_twsk_schedule);
280
281/* Remove all non full sockets (TIME_WAIT and NEW_SYN_RECV) for dead netns */
282void inet_twsk_purge(struct inet_hashinfo *hashinfo, int family)
283{
284	struct hlist_nulls_node *node;
285	unsigned int slot;
286	struct sock *sk;
287
288	for (slot = 0; slot <= hashinfo->ehash_mask; slot++) {
289		struct inet_ehash_bucket *head = &hashinfo->ehash[slot];
290restart_rcu:
291		cond_resched();
292		rcu_read_lock();
293restart:
294		sk_nulls_for_each_rcu(sk, node, &head->chain) {
295			int state = inet_sk_state_load(sk);
296
297			if ((1 << state) & ~(TCPF_TIME_WAIT |
298					     TCPF_NEW_SYN_RECV))
299				continue;
300
301			if (sk->sk_family != family ||
302			    refcount_read(&sock_net(sk)->ns.count))
303				continue;
304
305			if (unlikely(!refcount_inc_not_zero(&sk->sk_refcnt)))
306				continue;
307
308			if (unlikely(sk->sk_family != family ||
309				     refcount_read(&sock_net(sk)->ns.count))) {
310				sock_gen_put(sk);
311				goto restart;
312			}
313
314			rcu_read_unlock();
315			local_bh_disable();
316			if (state == TCP_TIME_WAIT) {
317				inet_twsk_deschedule_put(inet_twsk(sk));
318			} else {
319				struct request_sock *req = inet_reqsk(sk);
320
321				inet_csk_reqsk_queue_drop_and_put(req->rsk_listener,
322								  req);
323			}
324			local_bh_enable();
325			goto restart_rcu;
326		}
327		/* If the nulls value we got at the end of this lookup is
328		 * not the expected one, we must restart lookup.
329		 * We probably met an item that was moved to another chain.
330		 */
331		if (get_nulls_value(node) != slot)
332			goto restart;
333		rcu_read_unlock();
334	}
335}
336EXPORT_SYMBOL_GPL(inet_twsk_purge);
337