xref: /openbmc/linux/net/mptcp/pm_netlink.c (revision bcda5fd3)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Multipath TCP
3  *
4  * Copyright (c) 2020, Red Hat, Inc.
5  */
6 
7 #define pr_fmt(fmt) "MPTCP: " fmt
8 
9 #include <linux/inet.h>
10 #include <linux/kernel.h>
11 #include <net/tcp.h>
12 #include <net/netns/generic.h>
13 #include <net/mptcp.h>
14 #include <net/genetlink.h>
15 #include <uapi/linux/mptcp.h>
16 
17 #include "protocol.h"
18 #include "mib.h"
19 
20 /* forward declaration */
21 static struct genl_family mptcp_genl_family;
22 
23 static int pm_nl_pernet_id;
24 
25 struct mptcp_pm_addr_entry {
26 	struct list_head	list;
27 	struct mptcp_addr_info	addr;
28 	u8			flags;
29 	int			ifindex;
30 	struct socket		*lsk;
31 };
32 
33 struct mptcp_pm_add_entry {
34 	struct list_head	list;
35 	struct mptcp_addr_info	addr;
36 	struct timer_list	add_timer;
37 	struct mptcp_sock	*sock;
38 	u8			retrans_times;
39 };
40 
41 #define MAX_ADDR_ID		255
42 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
43 
44 struct pm_nl_pernet {
45 	/* protects pernet updates */
46 	spinlock_t		lock;
47 	struct list_head	local_addr_list;
48 	unsigned int		addrs;
49 	unsigned int		add_addr_signal_max;
50 	unsigned int		add_addr_accept_max;
51 	unsigned int		local_addr_max;
52 	unsigned int		subflows_max;
53 	unsigned int		next_id;
54 	unsigned long		id_bitmap[BITMAP_SZ];
55 };
56 
57 #define MPTCP_PM_ADDR_MAX	8
58 #define ADD_ADDR_RETRANS_MAX	3
59 
60 static bool addresses_equal(const struct mptcp_addr_info *a,
61 			    struct mptcp_addr_info *b, bool use_port)
62 {
63 	bool addr_equals = false;
64 
65 	if (a->family == b->family) {
66 		if (a->family == AF_INET)
67 			addr_equals = a->addr.s_addr == b->addr.s_addr;
68 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
69 		else
70 			addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
71 	} else if (a->family == AF_INET) {
72 		if (ipv6_addr_v4mapped(&b->addr6))
73 			addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
74 	} else if (b->family == AF_INET) {
75 		if (ipv6_addr_v4mapped(&a->addr6))
76 			addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
77 #endif
78 	}
79 
80 	if (!addr_equals)
81 		return false;
82 	if (!use_port)
83 		return true;
84 
85 	return a->port == b->port;
86 }
87 
88 static bool address_zero(const struct mptcp_addr_info *addr)
89 {
90 	struct mptcp_addr_info zero;
91 
92 	memset(&zero, 0, sizeof(zero));
93 	zero.family = addr->family;
94 
95 	return addresses_equal(addr, &zero, true);
96 }
97 
98 static void local_address(const struct sock_common *skc,
99 			  struct mptcp_addr_info *addr)
100 {
101 	addr->family = skc->skc_family;
102 	addr->port = htons(skc->skc_num);
103 	if (addr->family == AF_INET)
104 		addr->addr.s_addr = skc->skc_rcv_saddr;
105 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
106 	else if (addr->family == AF_INET6)
107 		addr->addr6 = skc->skc_v6_rcv_saddr;
108 #endif
109 }
110 
111 static void remote_address(const struct sock_common *skc,
112 			   struct mptcp_addr_info *addr)
113 {
114 	addr->family = skc->skc_family;
115 	addr->port = skc->skc_dport;
116 	if (addr->family == AF_INET)
117 		addr->addr.s_addr = skc->skc_daddr;
118 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
119 	else if (addr->family == AF_INET6)
120 		addr->addr6 = skc->skc_v6_daddr;
121 #endif
122 }
123 
124 static bool lookup_subflow_by_saddr(const struct list_head *list,
125 				    struct mptcp_addr_info *saddr)
126 {
127 	struct mptcp_subflow_context *subflow;
128 	struct mptcp_addr_info cur;
129 	struct sock_common *skc;
130 
131 	list_for_each_entry(subflow, list, node) {
132 		skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
133 
134 		local_address(skc, &cur);
135 		if (addresses_equal(&cur, saddr, saddr->port))
136 			return true;
137 	}
138 
139 	return false;
140 }
141 
142 static bool lookup_subflow_by_daddr(const struct list_head *list,
143 				    struct mptcp_addr_info *daddr)
144 {
145 	struct mptcp_subflow_context *subflow;
146 	struct mptcp_addr_info cur;
147 	struct sock_common *skc;
148 
149 	list_for_each_entry(subflow, list, node) {
150 		skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
151 
152 		remote_address(skc, &cur);
153 		if (addresses_equal(&cur, daddr, daddr->port))
154 			return true;
155 	}
156 
157 	return false;
158 }
159 
160 static struct mptcp_pm_addr_entry *
161 select_local_address(const struct pm_nl_pernet *pernet,
162 		     struct mptcp_sock *msk)
163 {
164 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
165 	struct sock *sk = (struct sock *)msk;
166 
167 	msk_owned_by_me(msk);
168 
169 	rcu_read_lock();
170 	__mptcp_flush_join_list(msk);
171 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
172 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
173 			continue;
174 
175 		if (entry->addr.family != sk->sk_family) {
176 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
177 			if ((entry->addr.family == AF_INET &&
178 			     !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
179 			    (sk->sk_family == AF_INET &&
180 			     !ipv6_addr_v4mapped(&entry->addr.addr6)))
181 #endif
182 				continue;
183 		}
184 
185 		/* avoid any address already in use by subflows and
186 		 * pending join
187 		 */
188 		if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
189 			ret = entry;
190 			break;
191 		}
192 	}
193 	rcu_read_unlock();
194 	return ret;
195 }
196 
197 static struct mptcp_pm_addr_entry *
198 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
199 {
200 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
201 	int i = 0;
202 
203 	rcu_read_lock();
204 	/* do not keep any additional per socket state, just signal
205 	 * the address list in order.
206 	 * Note: removal from the local address list during the msk life-cycle
207 	 * can lead to additional addresses not being announced.
208 	 */
209 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
210 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
211 			continue;
212 		if (i++ == pos) {
213 			ret = entry;
214 			break;
215 		}
216 	}
217 	rcu_read_unlock();
218 	return ret;
219 }
220 
221 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
222 {
223 	struct pm_nl_pernet *pernet;
224 
225 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
226 	return READ_ONCE(pernet->add_addr_signal_max);
227 }
228 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
229 
230 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
231 {
232 	struct pm_nl_pernet *pernet;
233 
234 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
235 	return READ_ONCE(pernet->add_addr_accept_max);
236 }
237 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
238 
239 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
240 {
241 	struct pm_nl_pernet *pernet;
242 
243 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
244 	return READ_ONCE(pernet->subflows_max);
245 }
246 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
247 
248 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
249 {
250 	struct pm_nl_pernet *pernet;
251 
252 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
253 	return READ_ONCE(pernet->local_addr_max);
254 }
255 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
256 
257 static void check_work_pending(struct mptcp_sock *msk)
258 {
259 	if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
260 	    (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
261 	     msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
262 		WRITE_ONCE(msk->pm.work_pending, false);
263 }
264 
265 struct mptcp_pm_add_entry *
266 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
267 				struct mptcp_addr_info *addr)
268 {
269 	struct mptcp_pm_add_entry *entry;
270 
271 	lockdep_assert_held(&msk->pm.lock);
272 
273 	list_for_each_entry(entry, &msk->pm.anno_list, list) {
274 		if (addresses_equal(&entry->addr, addr, true))
275 			return entry;
276 	}
277 
278 	return NULL;
279 }
280 
281 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
282 {
283 	struct mptcp_pm_add_entry *entry;
284 	struct mptcp_addr_info saddr;
285 	bool ret = false;
286 
287 	local_address((struct sock_common *)sk, &saddr);
288 
289 	spin_lock_bh(&msk->pm.lock);
290 	list_for_each_entry(entry, &msk->pm.anno_list, list) {
291 		if (addresses_equal(&entry->addr, &saddr, true)) {
292 			ret = true;
293 			goto out;
294 		}
295 	}
296 
297 out:
298 	spin_unlock_bh(&msk->pm.lock);
299 	return ret;
300 }
301 
302 static void mptcp_pm_add_timer(struct timer_list *timer)
303 {
304 	struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
305 	struct mptcp_sock *msk = entry->sock;
306 	struct sock *sk = (struct sock *)msk;
307 
308 	pr_debug("msk=%p", msk);
309 
310 	if (!msk)
311 		return;
312 
313 	if (inet_sk_state_load(sk) == TCP_CLOSE)
314 		return;
315 
316 	if (!entry->addr.id)
317 		return;
318 
319 	if (mptcp_pm_should_add_signal(msk)) {
320 		sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
321 		goto out;
322 	}
323 
324 	spin_lock_bh(&msk->pm.lock);
325 
326 	if (!mptcp_pm_should_add_signal(msk)) {
327 		pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
328 		mptcp_pm_announce_addr(msk, &entry->addr, false);
329 		mptcp_pm_add_addr_send_ack(msk);
330 		entry->retrans_times++;
331 	}
332 
333 	if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
334 		sk_reset_timer(sk, timer,
335 			       jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
336 
337 	spin_unlock_bh(&msk->pm.lock);
338 
339 	if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
340 		mptcp_pm_subflow_established(msk);
341 
342 out:
343 	__sock_put(sk);
344 }
345 
346 struct mptcp_pm_add_entry *
347 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
348 		       struct mptcp_addr_info *addr, bool check_id)
349 {
350 	struct mptcp_pm_add_entry *entry;
351 	struct sock *sk = (struct sock *)msk;
352 
353 	spin_lock_bh(&msk->pm.lock);
354 	entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
355 	if (entry && (!check_id || entry->addr.id == addr->id))
356 		entry->retrans_times = ADD_ADDR_RETRANS_MAX;
357 	spin_unlock_bh(&msk->pm.lock);
358 
359 	if (entry && (!check_id || entry->addr.id == addr->id))
360 		sk_stop_timer_sync(sk, &entry->add_timer);
361 
362 	return entry;
363 }
364 
365 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
366 				     struct mptcp_pm_addr_entry *entry)
367 {
368 	struct mptcp_pm_add_entry *add_entry = NULL;
369 	struct sock *sk = (struct sock *)msk;
370 	struct net *net = sock_net(sk);
371 
372 	lockdep_assert_held(&msk->pm.lock);
373 
374 	if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
375 		return false;
376 
377 	add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
378 	if (!add_entry)
379 		return false;
380 
381 	list_add(&add_entry->list, &msk->pm.anno_list);
382 
383 	add_entry->addr = entry->addr;
384 	add_entry->sock = msk;
385 	add_entry->retrans_times = 0;
386 
387 	timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
388 	sk_reset_timer(sk, &add_entry->add_timer,
389 		       jiffies + mptcp_get_add_addr_timeout(net));
390 
391 	return true;
392 }
393 
394 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
395 {
396 	struct mptcp_pm_add_entry *entry, *tmp;
397 	struct sock *sk = (struct sock *)msk;
398 	LIST_HEAD(free_list);
399 
400 	pr_debug("msk=%p", msk);
401 
402 	spin_lock_bh(&msk->pm.lock);
403 	list_splice_init(&msk->pm.anno_list, &free_list);
404 	spin_unlock_bh(&msk->pm.lock);
405 
406 	list_for_each_entry_safe(entry, tmp, &free_list, list) {
407 		sk_stop_timer_sync(sk, &entry->add_timer);
408 		kfree(entry);
409 	}
410 }
411 
412 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
413 {
414 	struct sock *sk = (struct sock *)msk;
415 	struct mptcp_pm_addr_entry *local;
416 	unsigned int add_addr_signal_max;
417 	unsigned int local_addr_max;
418 	struct pm_nl_pernet *pernet;
419 	unsigned int subflows_max;
420 
421 	pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
422 
423 	add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
424 	local_addr_max = mptcp_pm_get_local_addr_max(msk);
425 	subflows_max = mptcp_pm_get_subflows_max(msk);
426 
427 	pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
428 		 msk->pm.local_addr_used, local_addr_max,
429 		 msk->pm.add_addr_signaled, add_addr_signal_max,
430 		 msk->pm.subflows, subflows_max);
431 
432 	/* check first for announce */
433 	if (msk->pm.add_addr_signaled < add_addr_signal_max) {
434 		local = select_signal_address(pernet,
435 					      msk->pm.add_addr_signaled);
436 
437 		if (local) {
438 			if (mptcp_pm_alloc_anno_list(msk, local)) {
439 				msk->pm.add_addr_signaled++;
440 				mptcp_pm_announce_addr(msk, &local->addr, false);
441 				mptcp_pm_nl_addr_send_ack(msk);
442 			}
443 		} else {
444 			/* pick failed, avoid fourther attempts later */
445 			msk->pm.local_addr_used = add_addr_signal_max;
446 		}
447 
448 		check_work_pending(msk);
449 	}
450 
451 	/* check if should create a new subflow */
452 	if (msk->pm.local_addr_used < local_addr_max &&
453 	    msk->pm.subflows < subflows_max &&
454 	    !READ_ONCE(msk->pm.remote_deny_join_id0)) {
455 		local = select_local_address(pernet, msk);
456 		if (local) {
457 			struct mptcp_addr_info remote = { 0 };
458 
459 			msk->pm.local_addr_used++;
460 			msk->pm.subflows++;
461 			check_work_pending(msk);
462 			remote_address((struct sock_common *)sk, &remote);
463 			spin_unlock_bh(&msk->pm.lock);
464 			__mptcp_subflow_connect(sk, &local->addr, &remote,
465 						local->flags, local->ifindex);
466 			spin_lock_bh(&msk->pm.lock);
467 			return;
468 		}
469 
470 		/* lookup failed, avoid fourther attempts later */
471 		msk->pm.local_addr_used = local_addr_max;
472 		check_work_pending(msk);
473 	}
474 }
475 
476 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
477 {
478 	mptcp_pm_create_subflow_or_signal_addr(msk);
479 }
480 
481 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
482 {
483 	mptcp_pm_create_subflow_or_signal_addr(msk);
484 }
485 
486 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
487 {
488 	struct sock *sk = (struct sock *)msk;
489 	unsigned int add_addr_accept_max;
490 	struct mptcp_addr_info remote;
491 	struct mptcp_addr_info local;
492 	unsigned int subflows_max;
493 
494 	add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
495 	subflows_max = mptcp_pm_get_subflows_max(msk);
496 
497 	pr_debug("accepted %d:%d remote family %d",
498 		 msk->pm.add_addr_accepted, add_addr_accept_max,
499 		 msk->pm.remote.family);
500 
501 	if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
502 		goto add_addr_echo;
503 
504 	msk->pm.add_addr_accepted++;
505 	msk->pm.subflows++;
506 	if (msk->pm.add_addr_accepted >= add_addr_accept_max ||
507 	    msk->pm.subflows >= subflows_max)
508 		WRITE_ONCE(msk->pm.accept_addr, false);
509 
510 	/* connect to the specified remote address, using whatever
511 	 * local address the routing configuration will pick.
512 	 */
513 	remote = msk->pm.remote;
514 	if (!remote.port)
515 		remote.port = sk->sk_dport;
516 	memset(&local, 0, sizeof(local));
517 	local.family = remote.family;
518 
519 	spin_unlock_bh(&msk->pm.lock);
520 	__mptcp_subflow_connect(sk, &local, &remote, 0, 0);
521 	spin_lock_bh(&msk->pm.lock);
522 
523 add_addr_echo:
524 	mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
525 	mptcp_pm_nl_addr_send_ack(msk);
526 }
527 
528 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
529 {
530 	struct mptcp_subflow_context *subflow;
531 
532 	msk_owned_by_me(msk);
533 	lockdep_assert_held(&msk->pm.lock);
534 
535 	if (!mptcp_pm_should_add_signal(msk) &&
536 	    !mptcp_pm_should_rm_signal(msk))
537 		return;
538 
539 	__mptcp_flush_join_list(msk);
540 	subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
541 	if (subflow) {
542 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
543 		bool slow;
544 
545 		spin_unlock_bh(&msk->pm.lock);
546 		pr_debug("send ack for %s%s%s",
547 			 mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
548 			 mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
549 			 mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
550 
551 		slow = lock_sock_fast(ssk);
552 		tcp_send_ack(ssk);
553 		unlock_sock_fast(ssk, slow);
554 		spin_lock_bh(&msk->pm.lock);
555 	}
556 }
557 
558 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
559 				 struct mptcp_addr_info *addr,
560 				 u8 bkup)
561 {
562 	struct mptcp_subflow_context *subflow;
563 
564 	pr_debug("bkup=%d", bkup);
565 
566 	mptcp_for_each_subflow(msk, subflow) {
567 		struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
568 		struct sock *sk = (struct sock *)msk;
569 		struct mptcp_addr_info local;
570 		bool slow;
571 
572 		local_address((struct sock_common *)ssk, &local);
573 		if (!addresses_equal(&local, addr, addr->port))
574 			continue;
575 
576 		subflow->backup = bkup;
577 		subflow->send_mp_prio = 1;
578 		subflow->request_bkup = bkup;
579 		__MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
580 
581 		spin_unlock_bh(&msk->pm.lock);
582 		pr_debug("send ack for mp_prio");
583 		slow = lock_sock_fast(ssk);
584 		tcp_send_ack(ssk);
585 		unlock_sock_fast(ssk, slow);
586 		spin_lock_bh(&msk->pm.lock);
587 
588 		return 0;
589 	}
590 
591 	return -EINVAL;
592 }
593 
594 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
595 					   const struct mptcp_rm_list *rm_list,
596 					   enum linux_mptcp_mib_field rm_type)
597 {
598 	struct mptcp_subflow_context *subflow, *tmp;
599 	struct sock *sk = (struct sock *)msk;
600 	u8 i;
601 
602 	pr_debug("%s rm_list_nr %d",
603 		 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
604 
605 	msk_owned_by_me(msk);
606 
607 	if (!rm_list->nr)
608 		return;
609 
610 	if (list_empty(&msk->conn_list))
611 		return;
612 
613 	for (i = 0; i < rm_list->nr; i++) {
614 		list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
615 			struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
616 			int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
617 			u8 id = subflow->local_id;
618 
619 			if (rm_type == MPTCP_MIB_RMADDR)
620 				id = subflow->remote_id;
621 
622 			if (rm_list->ids[i] != id)
623 				continue;
624 
625 			pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
626 				 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
627 				 i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
628 			spin_unlock_bh(&msk->pm.lock);
629 			mptcp_subflow_shutdown(sk, ssk, how);
630 			mptcp_close_ssk(sk, ssk, subflow);
631 			spin_lock_bh(&msk->pm.lock);
632 
633 			if (rm_type == MPTCP_MIB_RMADDR) {
634 				msk->pm.add_addr_accepted--;
635 				WRITE_ONCE(msk->pm.accept_addr, true);
636 			} else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
637 				msk->pm.local_addr_used--;
638 			}
639 			msk->pm.subflows--;
640 			__MPTCP_INC_STATS(sock_net(sk), rm_type);
641 		}
642 	}
643 }
644 
645 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
646 {
647 	mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
648 }
649 
650 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
651 				     const struct mptcp_rm_list *rm_list)
652 {
653 	mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
654 }
655 
656 void mptcp_pm_nl_work(struct mptcp_sock *msk)
657 {
658 	struct mptcp_pm_data *pm = &msk->pm;
659 
660 	msk_owned_by_me(msk);
661 
662 	spin_lock_bh(&msk->pm.lock);
663 
664 	pr_debug("msk=%p status=%x", msk, pm->status);
665 	if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
666 		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
667 		mptcp_pm_nl_add_addr_received(msk);
668 	}
669 	if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
670 		pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
671 		mptcp_pm_nl_addr_send_ack(msk);
672 	}
673 	if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
674 		pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
675 		mptcp_pm_nl_rm_addr_received(msk);
676 	}
677 	if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
678 		pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
679 		mptcp_pm_nl_fully_established(msk);
680 	}
681 	if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
682 		pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
683 		mptcp_pm_nl_subflow_established(msk);
684 	}
685 
686 	spin_unlock_bh(&msk->pm.lock);
687 }
688 
689 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
690 {
691 	return (entry->flags &
692 		(MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
693 		MPTCP_PM_ADDR_FLAG_SIGNAL;
694 }
695 
696 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
697 					     struct mptcp_pm_addr_entry *entry)
698 {
699 	struct mptcp_pm_addr_entry *cur;
700 	unsigned int addr_max;
701 	int ret = -EINVAL;
702 
703 	spin_lock_bh(&pernet->lock);
704 	/* to keep the code simple, don't do IDR-like allocation for address ID,
705 	 * just bail when we exceed limits
706 	 */
707 	if (pernet->next_id == MAX_ADDR_ID)
708 		pernet->next_id = 1;
709 	if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
710 		goto out;
711 	if (test_bit(entry->addr.id, pernet->id_bitmap))
712 		goto out;
713 
714 	/* do not insert duplicate address, differentiate on port only
715 	 * singled addresses
716 	 */
717 	list_for_each_entry(cur, &pernet->local_addr_list, list) {
718 		if (addresses_equal(&cur->addr, &entry->addr,
719 				    address_use_port(entry) &&
720 				    address_use_port(cur)))
721 			goto out;
722 	}
723 
724 	if (!entry->addr.id) {
725 find_next:
726 		entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
727 						    MAX_ADDR_ID + 1,
728 						    pernet->next_id);
729 		if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
730 		    pernet->next_id != 1) {
731 			pernet->next_id = 1;
732 			goto find_next;
733 		}
734 	}
735 
736 	if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
737 		goto out;
738 
739 	__set_bit(entry->addr.id, pernet->id_bitmap);
740 	if (entry->addr.id > pernet->next_id)
741 		pernet->next_id = entry->addr.id;
742 
743 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
744 		addr_max = pernet->add_addr_signal_max;
745 		WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
746 	}
747 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
748 		addr_max = pernet->local_addr_max;
749 		WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
750 	}
751 
752 	pernet->addrs++;
753 	list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
754 	ret = entry->addr.id;
755 
756 out:
757 	spin_unlock_bh(&pernet->lock);
758 	return ret;
759 }
760 
761 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
762 					    struct mptcp_pm_addr_entry *entry)
763 {
764 	struct sockaddr_storage addr;
765 	struct mptcp_sock *msk;
766 	struct socket *ssock;
767 	int backlog = 1024;
768 	int err;
769 
770 	err = sock_create_kern(sock_net(sk), entry->addr.family,
771 			       SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
772 	if (err)
773 		return err;
774 
775 	msk = mptcp_sk(entry->lsk->sk);
776 	if (!msk) {
777 		err = -EINVAL;
778 		goto out;
779 	}
780 
781 	ssock = __mptcp_nmpc_socket(msk);
782 	if (!ssock) {
783 		err = -EINVAL;
784 		goto out;
785 	}
786 
787 	mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
788 	err = kernel_bind(ssock, (struct sockaddr *)&addr,
789 			  sizeof(struct sockaddr_in));
790 	if (err) {
791 		pr_warn("kernel_bind error, err=%d", err);
792 		goto out;
793 	}
794 
795 	err = kernel_listen(ssock, backlog);
796 	if (err) {
797 		pr_warn("kernel_listen error, err=%d", err);
798 		goto out;
799 	}
800 
801 	return 0;
802 
803 out:
804 	sock_release(entry->lsk);
805 	return err;
806 }
807 
808 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
809 {
810 	struct mptcp_pm_addr_entry *entry;
811 	struct mptcp_addr_info skc_local;
812 	struct mptcp_addr_info msk_local;
813 	struct pm_nl_pernet *pernet;
814 	int ret = -1;
815 
816 	if (WARN_ON_ONCE(!msk))
817 		return -1;
818 
819 	/* The 0 ID mapping is defined by the first subflow, copied into the msk
820 	 * addr
821 	 */
822 	local_address((struct sock_common *)msk, &msk_local);
823 	local_address((struct sock_common *)skc, &skc_local);
824 	if (addresses_equal(&msk_local, &skc_local, false))
825 		return 0;
826 
827 	if (address_zero(&skc_local))
828 		return 0;
829 
830 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
831 
832 	rcu_read_lock();
833 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
834 		if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
835 			ret = entry->addr.id;
836 			break;
837 		}
838 	}
839 	rcu_read_unlock();
840 	if (ret >= 0)
841 		return ret;
842 
843 	/* address not found, add to local list */
844 	entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
845 	if (!entry)
846 		return -ENOMEM;
847 
848 	entry->addr = skc_local;
849 	entry->addr.id = 0;
850 	entry->addr.port = 0;
851 	entry->ifindex = 0;
852 	entry->flags = 0;
853 	entry->lsk = NULL;
854 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
855 	if (ret < 0)
856 		kfree(entry);
857 
858 	return ret;
859 }
860 
861 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
862 {
863 	struct mptcp_pm_data *pm = &msk->pm;
864 	bool subflows;
865 
866 	subflows = !!mptcp_pm_get_subflows_max(msk);
867 	WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
868 		   !!mptcp_pm_get_add_addr_signal_max(msk));
869 	WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
870 	WRITE_ONCE(pm->accept_subflow, subflows);
871 }
872 
873 #define MPTCP_PM_CMD_GRP_OFFSET       0
874 #define MPTCP_PM_EV_GRP_OFFSET        1
875 
876 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
877 	[MPTCP_PM_CMD_GRP_OFFSET]	= { .name = MPTCP_PM_CMD_GRP_NAME, },
878 	[MPTCP_PM_EV_GRP_OFFSET]        = { .name = MPTCP_PM_EV_GRP_NAME,
879 					    .flags = GENL_UNS_ADMIN_PERM,
880 					  },
881 };
882 
883 static const struct nla_policy
884 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
885 	[MPTCP_PM_ADDR_ATTR_FAMILY]	= { .type	= NLA_U16,	},
886 	[MPTCP_PM_ADDR_ATTR_ID]		= { .type	= NLA_U8,	},
887 	[MPTCP_PM_ADDR_ATTR_ADDR4]	= { .type	= NLA_U32,	},
888 	[MPTCP_PM_ADDR_ATTR_ADDR6]	=
889 		NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
890 	[MPTCP_PM_ADDR_ATTR_PORT]	= { .type	= NLA_U16	},
891 	[MPTCP_PM_ADDR_ATTR_FLAGS]	= { .type	= NLA_U32	},
892 	[MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type	= NLA_S32	},
893 };
894 
895 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
896 	[MPTCP_PM_ATTR_ADDR]		=
897 					NLA_POLICY_NESTED(mptcp_pm_addr_policy),
898 	[MPTCP_PM_ATTR_RCV_ADD_ADDRS]	= { .type	= NLA_U32,	},
899 	[MPTCP_PM_ATTR_SUBFLOWS]	= { .type	= NLA_U32,	},
900 };
901 
902 static int mptcp_pm_family_to_addr(int family)
903 {
904 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
905 	if (family == AF_INET6)
906 		return MPTCP_PM_ADDR_ATTR_ADDR6;
907 #endif
908 	return MPTCP_PM_ADDR_ATTR_ADDR4;
909 }
910 
911 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
912 			       bool require_family,
913 			       struct mptcp_pm_addr_entry *entry)
914 {
915 	struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
916 	int err, addr_addr;
917 
918 	if (!attr) {
919 		GENL_SET_ERR_MSG(info, "missing address info");
920 		return -EINVAL;
921 	}
922 
923 	/* no validation needed - was already done via nested policy */
924 	err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
925 					  mptcp_pm_addr_policy, info->extack);
926 	if (err)
927 		return err;
928 
929 	memset(entry, 0, sizeof(*entry));
930 	if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
931 		if (!require_family)
932 			goto skip_family;
933 
934 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
935 				    "missing family");
936 		return -EINVAL;
937 	}
938 
939 	entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
940 	if (entry->addr.family != AF_INET
941 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
942 	    && entry->addr.family != AF_INET6
943 #endif
944 	    ) {
945 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
946 				    "unknown address family");
947 		return -EINVAL;
948 	}
949 	addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
950 	if (!tb[addr_addr]) {
951 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
952 				    "missing address data");
953 		return -EINVAL;
954 	}
955 
956 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
957 	if (entry->addr.family == AF_INET6)
958 		entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
959 	else
960 #endif
961 		entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
962 
963 skip_family:
964 	if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
965 		u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
966 
967 		entry->ifindex = val;
968 	}
969 
970 	if (tb[MPTCP_PM_ADDR_ATTR_ID])
971 		entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
972 
973 	if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
974 		entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
975 
976 	if (tb[MPTCP_PM_ADDR_ATTR_PORT]) {
977 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) {
978 			NL_SET_ERR_MSG_ATTR(info->extack, attr,
979 					    "flags must have signal when using port");
980 			return -EINVAL;
981 		}
982 		entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
983 	}
984 
985 	return 0;
986 }
987 
988 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
989 {
990 	return net_generic(genl_info_net(info), pm_nl_pernet_id);
991 }
992 
993 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
994 {
995 	struct mptcp_sock *msk;
996 	long s_slot = 0, s_num = 0;
997 
998 	while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
999 		struct sock *sk = (struct sock *)msk;
1000 
1001 		if (!READ_ONCE(msk->fully_established))
1002 			goto next;
1003 
1004 		lock_sock(sk);
1005 		spin_lock_bh(&msk->pm.lock);
1006 		mptcp_pm_create_subflow_or_signal_addr(msk);
1007 		spin_unlock_bh(&msk->pm.lock);
1008 		release_sock(sk);
1009 
1010 next:
1011 		sock_put(sk);
1012 		cond_resched();
1013 	}
1014 
1015 	return 0;
1016 }
1017 
1018 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1019 {
1020 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1021 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1022 	struct mptcp_pm_addr_entry addr, *entry;
1023 	int ret;
1024 
1025 	ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1026 	if (ret < 0)
1027 		return ret;
1028 
1029 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1030 	if (!entry) {
1031 		GENL_SET_ERR_MSG(info, "can't allocate addr");
1032 		return -ENOMEM;
1033 	}
1034 
1035 	*entry = addr;
1036 	if (entry->addr.port) {
1037 		ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1038 		if (ret) {
1039 			GENL_SET_ERR_MSG(info, "create listen socket error");
1040 			kfree(entry);
1041 			return ret;
1042 		}
1043 	}
1044 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1045 	if (ret < 0) {
1046 		GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1047 		if (entry->lsk)
1048 			sock_release(entry->lsk);
1049 		kfree(entry);
1050 		return ret;
1051 	}
1052 
1053 	mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1054 
1055 	return 0;
1056 }
1057 
1058 static struct mptcp_pm_addr_entry *
1059 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1060 {
1061 	struct mptcp_pm_addr_entry *entry;
1062 
1063 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
1064 		if (entry->addr.id == id)
1065 			return entry;
1066 	}
1067 	return NULL;
1068 }
1069 
1070 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1071 				      struct mptcp_addr_info *addr)
1072 {
1073 	struct mptcp_pm_add_entry *entry;
1074 
1075 	entry = mptcp_pm_del_add_timer(msk, addr, false);
1076 	if (entry) {
1077 		list_del(&entry->list);
1078 		kfree(entry);
1079 		return true;
1080 	}
1081 
1082 	return false;
1083 }
1084 
1085 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1086 				      struct mptcp_addr_info *addr,
1087 				      bool force)
1088 {
1089 	struct mptcp_rm_list list = { .nr = 0 };
1090 	bool ret;
1091 
1092 	list.ids[list.nr++] = addr->id;
1093 
1094 	ret = remove_anno_list_by_saddr(msk, addr);
1095 	if (ret || force) {
1096 		spin_lock_bh(&msk->pm.lock);
1097 		mptcp_pm_remove_addr(msk, &list);
1098 		spin_unlock_bh(&msk->pm.lock);
1099 	}
1100 	return ret;
1101 }
1102 
1103 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1104 						   struct mptcp_addr_info *addr)
1105 {
1106 	struct mptcp_sock *msk;
1107 	long s_slot = 0, s_num = 0;
1108 	struct mptcp_rm_list list = { .nr = 0 };
1109 
1110 	pr_debug("remove_id=%d", addr->id);
1111 
1112 	list.ids[list.nr++] = addr->id;
1113 
1114 	while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1115 		struct sock *sk = (struct sock *)msk;
1116 		bool remove_subflow;
1117 
1118 		if (list_empty(&msk->conn_list)) {
1119 			mptcp_pm_remove_anno_addr(msk, addr, false);
1120 			goto next;
1121 		}
1122 
1123 		lock_sock(sk);
1124 		remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1125 		mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1126 		if (remove_subflow)
1127 			mptcp_pm_remove_subflow(msk, &list);
1128 		release_sock(sk);
1129 
1130 next:
1131 		sock_put(sk);
1132 		cond_resched();
1133 	}
1134 
1135 	return 0;
1136 }
1137 
1138 /* caller must ensure the RCU grace period is already elapsed */
1139 static void __mptcp_pm_release_addr_entry(struct mptcp_pm_addr_entry *entry)
1140 {
1141 	if (entry->lsk)
1142 		sock_release(entry->lsk);
1143 	kfree(entry);
1144 }
1145 
1146 static int mptcp_nl_remove_id_zero_address(struct net *net,
1147 					   struct mptcp_addr_info *addr)
1148 {
1149 	struct mptcp_rm_list list = { .nr = 0 };
1150 	long s_slot = 0, s_num = 0;
1151 	struct mptcp_sock *msk;
1152 
1153 	list.ids[list.nr++] = 0;
1154 
1155 	while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1156 		struct sock *sk = (struct sock *)msk;
1157 		struct mptcp_addr_info msk_local;
1158 
1159 		if (list_empty(&msk->conn_list))
1160 			goto next;
1161 
1162 		local_address((struct sock_common *)msk, &msk_local);
1163 		if (!addresses_equal(&msk_local, addr, addr->port))
1164 			goto next;
1165 
1166 		lock_sock(sk);
1167 		spin_lock_bh(&msk->pm.lock);
1168 		mptcp_pm_remove_addr(msk, &list);
1169 		mptcp_pm_nl_rm_subflow_received(msk, &list);
1170 		spin_unlock_bh(&msk->pm.lock);
1171 		release_sock(sk);
1172 
1173 next:
1174 		sock_put(sk);
1175 		cond_resched();
1176 	}
1177 
1178 	return 0;
1179 }
1180 
1181 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1182 {
1183 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1184 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1185 	struct mptcp_pm_addr_entry addr, *entry;
1186 	unsigned int addr_max;
1187 	int ret;
1188 
1189 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1190 	if (ret < 0)
1191 		return ret;
1192 
1193 	/* the zero id address is special: the first address used by the msk
1194 	 * always gets such an id, so different subflows can have different zero
1195 	 * id addresses. Additionally zero id is not accounted for in id_bitmap.
1196 	 * Let's use an 'mptcp_rm_list' instead of the common remove code.
1197 	 */
1198 	if (addr.addr.id == 0)
1199 		return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1200 
1201 	spin_lock_bh(&pernet->lock);
1202 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
1203 	if (!entry) {
1204 		GENL_SET_ERR_MSG(info, "address not found");
1205 		spin_unlock_bh(&pernet->lock);
1206 		return -EINVAL;
1207 	}
1208 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1209 		addr_max = pernet->add_addr_signal_max;
1210 		WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1211 	}
1212 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1213 		addr_max = pernet->local_addr_max;
1214 		WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1215 	}
1216 
1217 	pernet->addrs--;
1218 	list_del_rcu(&entry->list);
1219 	__clear_bit(entry->addr.id, pernet->id_bitmap);
1220 	spin_unlock_bh(&pernet->lock);
1221 
1222 	mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1223 	synchronize_rcu();
1224 	__mptcp_pm_release_addr_entry(entry);
1225 
1226 	return ret;
1227 }
1228 
1229 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1230 					       struct list_head *rm_list)
1231 {
1232 	struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1233 	struct mptcp_pm_addr_entry *entry;
1234 
1235 	list_for_each_entry(entry, rm_list, list) {
1236 		if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1237 		    alist.nr < MPTCP_RM_IDS_MAX &&
1238 		    slist.nr < MPTCP_RM_IDS_MAX) {
1239 			alist.ids[alist.nr++] = entry->addr.id;
1240 			slist.ids[slist.nr++] = entry->addr.id;
1241 		} else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1242 			 alist.nr < MPTCP_RM_IDS_MAX) {
1243 			alist.ids[alist.nr++] = entry->addr.id;
1244 		}
1245 	}
1246 
1247 	if (alist.nr) {
1248 		spin_lock_bh(&msk->pm.lock);
1249 		mptcp_pm_remove_addr(msk, &alist);
1250 		spin_unlock_bh(&msk->pm.lock);
1251 	}
1252 	if (slist.nr)
1253 		mptcp_pm_remove_subflow(msk, &slist);
1254 }
1255 
1256 static void mptcp_nl_remove_addrs_list(struct net *net,
1257 				       struct list_head *rm_list)
1258 {
1259 	long s_slot = 0, s_num = 0;
1260 	struct mptcp_sock *msk;
1261 
1262 	if (list_empty(rm_list))
1263 		return;
1264 
1265 	while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1266 		struct sock *sk = (struct sock *)msk;
1267 
1268 		lock_sock(sk);
1269 		mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1270 		release_sock(sk);
1271 
1272 		sock_put(sk);
1273 		cond_resched();
1274 	}
1275 }
1276 
1277 /* caller must ensure the RCU grace period is already elapsed */
1278 static void __flush_addrs(struct list_head *list)
1279 {
1280 	while (!list_empty(list)) {
1281 		struct mptcp_pm_addr_entry *cur;
1282 
1283 		cur = list_entry(list->next,
1284 				 struct mptcp_pm_addr_entry, list);
1285 		list_del_rcu(&cur->list);
1286 		__mptcp_pm_release_addr_entry(cur);
1287 	}
1288 }
1289 
1290 static void __reset_counters(struct pm_nl_pernet *pernet)
1291 {
1292 	WRITE_ONCE(pernet->add_addr_signal_max, 0);
1293 	WRITE_ONCE(pernet->add_addr_accept_max, 0);
1294 	WRITE_ONCE(pernet->local_addr_max, 0);
1295 	pernet->addrs = 0;
1296 }
1297 
1298 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1299 {
1300 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1301 	LIST_HEAD(free_list);
1302 
1303 	spin_lock_bh(&pernet->lock);
1304 	list_splice_init(&pernet->local_addr_list, &free_list);
1305 	__reset_counters(pernet);
1306 	pernet->next_id = 1;
1307 	bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1308 	spin_unlock_bh(&pernet->lock);
1309 	mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1310 	synchronize_rcu();
1311 	__flush_addrs(&free_list);
1312 	return 0;
1313 }
1314 
1315 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1316 			      struct mptcp_pm_addr_entry *entry)
1317 {
1318 	struct mptcp_addr_info *addr = &entry->addr;
1319 	struct nlattr *attr;
1320 
1321 	attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1322 	if (!attr)
1323 		return -EMSGSIZE;
1324 
1325 	if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1326 		goto nla_put_failure;
1327 	if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1328 		goto nla_put_failure;
1329 	if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1330 		goto nla_put_failure;
1331 	if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1332 		goto nla_put_failure;
1333 	if (entry->ifindex &&
1334 	    nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1335 		goto nla_put_failure;
1336 
1337 	if (addr->family == AF_INET &&
1338 	    nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1339 			    addr->addr.s_addr))
1340 		goto nla_put_failure;
1341 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1342 	else if (addr->family == AF_INET6 &&
1343 		 nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1344 		goto nla_put_failure;
1345 #endif
1346 	nla_nest_end(skb, attr);
1347 	return 0;
1348 
1349 nla_put_failure:
1350 	nla_nest_cancel(skb, attr);
1351 	return -EMSGSIZE;
1352 }
1353 
1354 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1355 {
1356 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1357 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1358 	struct mptcp_pm_addr_entry addr, *entry;
1359 	struct sk_buff *msg;
1360 	void *reply;
1361 	int ret;
1362 
1363 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1364 	if (ret < 0)
1365 		return ret;
1366 
1367 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1368 	if (!msg)
1369 		return -ENOMEM;
1370 
1371 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1372 				  info->genlhdr->cmd);
1373 	if (!reply) {
1374 		GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1375 		ret = -EMSGSIZE;
1376 		goto fail;
1377 	}
1378 
1379 	spin_lock_bh(&pernet->lock);
1380 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
1381 	if (!entry) {
1382 		GENL_SET_ERR_MSG(info, "address not found");
1383 		ret = -EINVAL;
1384 		goto unlock_fail;
1385 	}
1386 
1387 	ret = mptcp_nl_fill_addr(msg, entry);
1388 	if (ret)
1389 		goto unlock_fail;
1390 
1391 	genlmsg_end(msg, reply);
1392 	ret = genlmsg_reply(msg, info);
1393 	spin_unlock_bh(&pernet->lock);
1394 	return ret;
1395 
1396 unlock_fail:
1397 	spin_unlock_bh(&pernet->lock);
1398 
1399 fail:
1400 	nlmsg_free(msg);
1401 	return ret;
1402 }
1403 
1404 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1405 				   struct netlink_callback *cb)
1406 {
1407 	struct net *net = sock_net(msg->sk);
1408 	struct mptcp_pm_addr_entry *entry;
1409 	struct pm_nl_pernet *pernet;
1410 	int id = cb->args[0];
1411 	void *hdr;
1412 	int i;
1413 
1414 	pernet = net_generic(net, pm_nl_pernet_id);
1415 
1416 	spin_lock_bh(&pernet->lock);
1417 	for (i = id; i < MAX_ADDR_ID + 1; i++) {
1418 		if (test_bit(i, pernet->id_bitmap)) {
1419 			entry = __lookup_addr_by_id(pernet, i);
1420 			if (!entry)
1421 				break;
1422 
1423 			if (entry->addr.id <= id)
1424 				continue;
1425 
1426 			hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1427 					  cb->nlh->nlmsg_seq, &mptcp_genl_family,
1428 					  NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1429 			if (!hdr)
1430 				break;
1431 
1432 			if (mptcp_nl_fill_addr(msg, entry) < 0) {
1433 				genlmsg_cancel(msg, hdr);
1434 				break;
1435 			}
1436 
1437 			id = entry->addr.id;
1438 			genlmsg_end(msg, hdr);
1439 		}
1440 	}
1441 	spin_unlock_bh(&pernet->lock);
1442 
1443 	cb->args[0] = id;
1444 	return msg->len;
1445 }
1446 
1447 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1448 {
1449 	struct nlattr *attr = info->attrs[id];
1450 
1451 	if (!attr)
1452 		return 0;
1453 
1454 	*limit = nla_get_u32(attr);
1455 	if (*limit > MPTCP_PM_ADDR_MAX) {
1456 		GENL_SET_ERR_MSG(info, "limit greater than maximum");
1457 		return -EINVAL;
1458 	}
1459 	return 0;
1460 }
1461 
1462 static int
1463 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1464 {
1465 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1466 	unsigned int rcv_addrs, subflows;
1467 	int ret;
1468 
1469 	spin_lock_bh(&pernet->lock);
1470 	rcv_addrs = pernet->add_addr_accept_max;
1471 	ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1472 	if (ret)
1473 		goto unlock;
1474 
1475 	subflows = pernet->subflows_max;
1476 	ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1477 	if (ret)
1478 		goto unlock;
1479 
1480 	WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1481 	WRITE_ONCE(pernet->subflows_max, subflows);
1482 
1483 unlock:
1484 	spin_unlock_bh(&pernet->lock);
1485 	return ret;
1486 }
1487 
1488 static int
1489 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1490 {
1491 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1492 	struct sk_buff *msg;
1493 	void *reply;
1494 
1495 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1496 	if (!msg)
1497 		return -ENOMEM;
1498 
1499 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1500 				  MPTCP_PM_CMD_GET_LIMITS);
1501 	if (!reply)
1502 		goto fail;
1503 
1504 	if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1505 			READ_ONCE(pernet->add_addr_accept_max)))
1506 		goto fail;
1507 
1508 	if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1509 			READ_ONCE(pernet->subflows_max)))
1510 		goto fail;
1511 
1512 	genlmsg_end(msg, reply);
1513 	return genlmsg_reply(msg, info);
1514 
1515 fail:
1516 	GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1517 	nlmsg_free(msg);
1518 	return -EMSGSIZE;
1519 }
1520 
1521 static int mptcp_nl_addr_backup(struct net *net,
1522 				struct mptcp_addr_info *addr,
1523 				u8 bkup)
1524 {
1525 	long s_slot = 0, s_num = 0;
1526 	struct mptcp_sock *msk;
1527 	int ret = -EINVAL;
1528 
1529 	while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1530 		struct sock *sk = (struct sock *)msk;
1531 
1532 		if (list_empty(&msk->conn_list))
1533 			goto next;
1534 
1535 		lock_sock(sk);
1536 		spin_lock_bh(&msk->pm.lock);
1537 		ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1538 		spin_unlock_bh(&msk->pm.lock);
1539 		release_sock(sk);
1540 
1541 next:
1542 		sock_put(sk);
1543 		cond_resched();
1544 	}
1545 
1546 	return ret;
1547 }
1548 
1549 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1550 {
1551 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1552 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1553 	struct mptcp_pm_addr_entry addr, *entry;
1554 	struct net *net = sock_net(skb->sk);
1555 	u8 bkup = 0;
1556 	int ret;
1557 
1558 	ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1559 	if (ret < 0)
1560 		return ret;
1561 
1562 	if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1563 		bkup = 1;
1564 
1565 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
1566 		if (addresses_equal(&entry->addr, &addr.addr, true)) {
1567 			ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1568 			if (ret)
1569 				return ret;
1570 
1571 			if (bkup)
1572 				entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1573 			else
1574 				entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1575 		}
1576 	}
1577 
1578 	return 0;
1579 }
1580 
1581 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1582 {
1583 	genlmsg_multicast_netns(&mptcp_genl_family, net,
1584 				nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1585 }
1586 
1587 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1588 {
1589 	const struct inet_sock *issk = inet_sk(ssk);
1590 	const struct mptcp_subflow_context *sf;
1591 
1592 	if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1593 		return -EMSGSIZE;
1594 
1595 	switch (ssk->sk_family) {
1596 	case AF_INET:
1597 		if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1598 			return -EMSGSIZE;
1599 		if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1600 			return -EMSGSIZE;
1601 		break;
1602 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1603 	case AF_INET6: {
1604 		const struct ipv6_pinfo *np = inet6_sk(ssk);
1605 
1606 		if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1607 			return -EMSGSIZE;
1608 		if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1609 			return -EMSGSIZE;
1610 		break;
1611 	}
1612 #endif
1613 	default:
1614 		WARN_ON_ONCE(1);
1615 		return -EMSGSIZE;
1616 	}
1617 
1618 	if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1619 		return -EMSGSIZE;
1620 	if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1621 		return -EMSGSIZE;
1622 
1623 	sf = mptcp_subflow_ctx(ssk);
1624 	if (WARN_ON_ONCE(!sf))
1625 		return -EINVAL;
1626 
1627 	if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1628 		return -EMSGSIZE;
1629 
1630 	if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1631 		return -EMSGSIZE;
1632 
1633 	return 0;
1634 }
1635 
1636 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1637 					 const struct mptcp_sock *msk,
1638 					 const struct sock *ssk)
1639 {
1640 	const struct sock *sk = (const struct sock *)msk;
1641 	const struct mptcp_subflow_context *sf;
1642 	u8 sk_err;
1643 
1644 	if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1645 		return -EMSGSIZE;
1646 
1647 	if (mptcp_event_add_subflow(skb, ssk))
1648 		return -EMSGSIZE;
1649 
1650 	sf = mptcp_subflow_ctx(ssk);
1651 	if (WARN_ON_ONCE(!sf))
1652 		return -EINVAL;
1653 
1654 	if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1655 		return -EMSGSIZE;
1656 
1657 	if (ssk->sk_bound_dev_if &&
1658 	    nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1659 		return -EMSGSIZE;
1660 
1661 	sk_err = ssk->sk_err;
1662 	if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1663 	    nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1664 		return -EMSGSIZE;
1665 
1666 	return 0;
1667 }
1668 
1669 static int mptcp_event_sub_established(struct sk_buff *skb,
1670 				       const struct mptcp_sock *msk,
1671 				       const struct sock *ssk)
1672 {
1673 	return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1674 }
1675 
1676 static int mptcp_event_sub_closed(struct sk_buff *skb,
1677 				  const struct mptcp_sock *msk,
1678 				  const struct sock *ssk)
1679 {
1680 	const struct mptcp_subflow_context *sf;
1681 
1682 	if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1683 		return -EMSGSIZE;
1684 
1685 	sf = mptcp_subflow_ctx(ssk);
1686 	if (!sf->reset_seen)
1687 		return 0;
1688 
1689 	if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1690 		return -EMSGSIZE;
1691 
1692 	if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1693 		return -EMSGSIZE;
1694 
1695 	return 0;
1696 }
1697 
1698 static int mptcp_event_created(struct sk_buff *skb,
1699 			       const struct mptcp_sock *msk,
1700 			       const struct sock *ssk)
1701 {
1702 	int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1703 
1704 	if (err)
1705 		return err;
1706 
1707 	return mptcp_event_add_subflow(skb, ssk);
1708 }
1709 
1710 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1711 {
1712 	struct net *net = sock_net((const struct sock *)msk);
1713 	struct nlmsghdr *nlh;
1714 	struct sk_buff *skb;
1715 
1716 	if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1717 		return;
1718 
1719 	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1720 	if (!skb)
1721 		return;
1722 
1723 	nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1724 	if (!nlh)
1725 		goto nla_put_failure;
1726 
1727 	if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1728 		goto nla_put_failure;
1729 
1730 	if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1731 		goto nla_put_failure;
1732 
1733 	genlmsg_end(skb, nlh);
1734 	mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1735 	return;
1736 
1737 nla_put_failure:
1738 	kfree_skb(skb);
1739 }
1740 
1741 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1742 				const struct mptcp_addr_info *info)
1743 {
1744 	struct net *net = sock_net((const struct sock *)msk);
1745 	struct nlmsghdr *nlh;
1746 	struct sk_buff *skb;
1747 
1748 	if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1749 		return;
1750 
1751 	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1752 	if (!skb)
1753 		return;
1754 
1755 	nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1756 			  MPTCP_EVENT_ANNOUNCED);
1757 	if (!nlh)
1758 		goto nla_put_failure;
1759 
1760 	if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1761 		goto nla_put_failure;
1762 
1763 	if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1764 		goto nla_put_failure;
1765 
1766 	if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1767 		goto nla_put_failure;
1768 
1769 	switch (info->family) {
1770 	case AF_INET:
1771 		if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1772 			goto nla_put_failure;
1773 		break;
1774 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1775 	case AF_INET6:
1776 		if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1777 			goto nla_put_failure;
1778 		break;
1779 #endif
1780 	default:
1781 		WARN_ON_ONCE(1);
1782 		goto nla_put_failure;
1783 	}
1784 
1785 	genlmsg_end(skb, nlh);
1786 	mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1787 	return;
1788 
1789 nla_put_failure:
1790 	kfree_skb(skb);
1791 }
1792 
1793 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1794 		 const struct sock *ssk, gfp_t gfp)
1795 {
1796 	struct net *net = sock_net((const struct sock *)msk);
1797 	struct nlmsghdr *nlh;
1798 	struct sk_buff *skb;
1799 
1800 	if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1801 		return;
1802 
1803 	skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1804 	if (!skb)
1805 		return;
1806 
1807 	nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1808 	if (!nlh)
1809 		goto nla_put_failure;
1810 
1811 	switch (type) {
1812 	case MPTCP_EVENT_UNSPEC:
1813 		WARN_ON_ONCE(1);
1814 		break;
1815 	case MPTCP_EVENT_CREATED:
1816 	case MPTCP_EVENT_ESTABLISHED:
1817 		if (mptcp_event_created(skb, msk, ssk) < 0)
1818 			goto nla_put_failure;
1819 		break;
1820 	case MPTCP_EVENT_CLOSED:
1821 		if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1822 			goto nla_put_failure;
1823 		break;
1824 	case MPTCP_EVENT_ANNOUNCED:
1825 	case MPTCP_EVENT_REMOVED:
1826 		/* call mptcp_event_addr_announced()/removed instead */
1827 		WARN_ON_ONCE(1);
1828 		break;
1829 	case MPTCP_EVENT_SUB_ESTABLISHED:
1830 	case MPTCP_EVENT_SUB_PRIORITY:
1831 		if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1832 			goto nla_put_failure;
1833 		break;
1834 	case MPTCP_EVENT_SUB_CLOSED:
1835 		if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1836 			goto nla_put_failure;
1837 		break;
1838 	}
1839 
1840 	genlmsg_end(skb, nlh);
1841 	mptcp_nl_mcast_send(net, skb, gfp);
1842 	return;
1843 
1844 nla_put_failure:
1845 	kfree_skb(skb);
1846 }
1847 
1848 static const struct genl_small_ops mptcp_pm_ops[] = {
1849 	{
1850 		.cmd    = MPTCP_PM_CMD_ADD_ADDR,
1851 		.doit   = mptcp_nl_cmd_add_addr,
1852 		.flags  = GENL_ADMIN_PERM,
1853 	},
1854 	{
1855 		.cmd    = MPTCP_PM_CMD_DEL_ADDR,
1856 		.doit   = mptcp_nl_cmd_del_addr,
1857 		.flags  = GENL_ADMIN_PERM,
1858 	},
1859 	{
1860 		.cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
1861 		.doit   = mptcp_nl_cmd_flush_addrs,
1862 		.flags  = GENL_ADMIN_PERM,
1863 	},
1864 	{
1865 		.cmd    = MPTCP_PM_CMD_GET_ADDR,
1866 		.doit   = mptcp_nl_cmd_get_addr,
1867 		.dumpit   = mptcp_nl_cmd_dump_addrs,
1868 	},
1869 	{
1870 		.cmd    = MPTCP_PM_CMD_SET_LIMITS,
1871 		.doit   = mptcp_nl_cmd_set_limits,
1872 		.flags  = GENL_ADMIN_PERM,
1873 	},
1874 	{
1875 		.cmd    = MPTCP_PM_CMD_GET_LIMITS,
1876 		.doit   = mptcp_nl_cmd_get_limits,
1877 	},
1878 	{
1879 		.cmd    = MPTCP_PM_CMD_SET_FLAGS,
1880 		.doit   = mptcp_nl_cmd_set_flags,
1881 		.flags  = GENL_ADMIN_PERM,
1882 	},
1883 };
1884 
1885 static struct genl_family mptcp_genl_family __ro_after_init = {
1886 	.name		= MPTCP_PM_NAME,
1887 	.version	= MPTCP_PM_VER,
1888 	.maxattr	= MPTCP_PM_ATTR_MAX,
1889 	.policy		= mptcp_pm_policy,
1890 	.netnsok	= true,
1891 	.module		= THIS_MODULE,
1892 	.small_ops	= mptcp_pm_ops,
1893 	.n_small_ops	= ARRAY_SIZE(mptcp_pm_ops),
1894 	.mcgrps		= mptcp_pm_mcgrps,
1895 	.n_mcgrps	= ARRAY_SIZE(mptcp_pm_mcgrps),
1896 };
1897 
1898 static int __net_init pm_nl_init_net(struct net *net)
1899 {
1900 	struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1901 
1902 	INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1903 	pernet->next_id = 1;
1904 	spin_lock_init(&pernet->lock);
1905 
1906 	/* No need to initialize other pernet fields, the struct is zeroed at
1907 	 * allocation time.
1908 	 */
1909 
1910 	return 0;
1911 }
1912 
1913 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1914 {
1915 	struct net *net;
1916 
1917 	list_for_each_entry(net, net_list, exit_list) {
1918 		struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1919 
1920 		/* net is removed from namespace list, can't race with
1921 		 * other modifiers, also netns core already waited for a
1922 		 * RCU grace period.
1923 		 */
1924 		__flush_addrs(&pernet->local_addr_list);
1925 	}
1926 }
1927 
1928 static struct pernet_operations mptcp_pm_pernet_ops = {
1929 	.init = pm_nl_init_net,
1930 	.exit_batch = pm_nl_exit_net,
1931 	.id = &pm_nl_pernet_id,
1932 	.size = sizeof(struct pm_nl_pernet),
1933 };
1934 
1935 void __init mptcp_pm_nl_init(void)
1936 {
1937 	if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1938 		panic("Failed to register MPTCP PM pernet subsystem.\n");
1939 
1940 	if (genl_register_family(&mptcp_genl_family))
1941 		panic("Failed to register MPTCP PM netlink family\n");
1942 }
1943