xref: /linux/net/mptcp/pm_netlink.c (revision 86287543715ac2a6d92d561cc105d79306511457)
1 // SPDX-License-Identifier: GPL-2.0
2 /* Multipath TCP
3  *
4  * Copyright (c) 2020, Red Hat, Inc.
5  */
6 
7 #include <linux/inet.h>
8 #include <linux/kernel.h>
9 #include <net/tcp.h>
10 #include <net/netns/generic.h>
11 #include <net/mptcp.h>
12 #include <net/genetlink.h>
13 #include <uapi/linux/mptcp.h>
14 
15 #include "protocol.h"
16 
17 /* forward declaration */
18 static struct genl_family mptcp_genl_family;
19 
20 static int pm_nl_pernet_id;
21 
22 struct mptcp_pm_addr_entry {
23 	struct list_head	list;
24 	unsigned int		flags;
25 	int			ifindex;
26 	struct mptcp_addr_info	addr;
27 	struct rcu_head		rcu;
28 };
29 
30 struct pm_nl_pernet {
31 	/* protects pernet updates */
32 	spinlock_t		lock;
33 	struct list_head	local_addr_list;
34 	unsigned int		addrs;
35 	unsigned int		add_addr_signal_max;
36 	unsigned int		add_addr_accept_max;
37 	unsigned int		local_addr_max;
38 	unsigned int		subflows_max;
39 	unsigned int		next_id;
40 };
41 
42 #define MPTCP_PM_ADDR_MAX	8
43 
44 static bool addresses_equal(const struct mptcp_addr_info *a,
45 			    struct mptcp_addr_info *b, bool use_port)
46 {
47 	bool addr_equals = false;
48 
49 	if (a->family != b->family)
50 		return false;
51 
52 	if (a->family == AF_INET)
53 		addr_equals = a->addr.s_addr == b->addr.s_addr;
54 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
55 	else
56 		addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
57 #endif
58 
59 	if (!addr_equals)
60 		return false;
61 	if (!use_port)
62 		return true;
63 
64 	return a->port == b->port;
65 }
66 
67 static void local_address(const struct sock_common *skc,
68 			  struct mptcp_addr_info *addr)
69 {
70 	addr->port = 0;
71 	addr->family = skc->skc_family;
72 	if (addr->family == AF_INET)
73 		addr->addr.s_addr = skc->skc_rcv_saddr;
74 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
75 	else if (addr->family == AF_INET6)
76 		addr->addr6 = skc->skc_v6_rcv_saddr;
77 #endif
78 }
79 
80 static void remote_address(const struct sock_common *skc,
81 			   struct mptcp_addr_info *addr)
82 {
83 	addr->family = skc->skc_family;
84 	addr->port = skc->skc_dport;
85 	if (addr->family == AF_INET)
86 		addr->addr.s_addr = skc->skc_daddr;
87 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
88 	else if (addr->family == AF_INET6)
89 		addr->addr6 = skc->skc_v6_daddr;
90 #endif
91 }
92 
93 static bool lookup_subflow_by_saddr(const struct list_head *list,
94 				    struct mptcp_addr_info *saddr)
95 {
96 	struct mptcp_subflow_context *subflow;
97 	struct mptcp_addr_info cur;
98 	struct sock_common *skc;
99 
100 	list_for_each_entry(subflow, list, node) {
101 		skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
102 
103 		local_address(skc, &cur);
104 		if (addresses_equal(&cur, saddr, false))
105 			return true;
106 	}
107 
108 	return false;
109 }
110 
111 static struct mptcp_pm_addr_entry *
112 select_local_address(const struct pm_nl_pernet *pernet,
113 		     struct mptcp_sock *msk)
114 {
115 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
116 
117 	rcu_read_lock();
118 	spin_lock_bh(&msk->join_list_lock);
119 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
120 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
121 			continue;
122 
123 		/* avoid any address already in use by subflows and
124 		 * pending join
125 		 */
126 		if (entry->addr.family == ((struct sock *)msk)->sk_family &&
127 		    !lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
128 		    !lookup_subflow_by_saddr(&msk->join_list, &entry->addr)) {
129 			ret = entry;
130 			break;
131 		}
132 	}
133 	spin_unlock_bh(&msk->join_list_lock);
134 	rcu_read_unlock();
135 	return ret;
136 }
137 
138 static struct mptcp_pm_addr_entry *
139 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
140 {
141 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
142 	int i = 0;
143 
144 	rcu_read_lock();
145 	/* do not keep any additional per socket state, just signal
146 	 * the address list in order.
147 	 * Note: removal from the local address list during the msk life-cycle
148 	 * can lead to additional addresses not being announced.
149 	 */
150 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
151 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
152 			continue;
153 		if (i++ == pos) {
154 			ret = entry;
155 			break;
156 		}
157 	}
158 	rcu_read_unlock();
159 	return ret;
160 }
161 
162 static void check_work_pending(struct mptcp_sock *msk)
163 {
164 	if (msk->pm.add_addr_signaled == msk->pm.add_addr_signal_max &&
165 	    (msk->pm.local_addr_used == msk->pm.local_addr_max ||
166 	     msk->pm.subflows == msk->pm.subflows_max))
167 		WRITE_ONCE(msk->pm.work_pending, false);
168 }
169 
170 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
171 {
172 	struct sock *sk = (struct sock *)msk;
173 	struct mptcp_pm_addr_entry *local;
174 	struct mptcp_addr_info remote;
175 	struct pm_nl_pernet *pernet;
176 
177 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
178 
179 	pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
180 		 msk->pm.local_addr_used, msk->pm.local_addr_max,
181 		 msk->pm.add_addr_signaled, msk->pm.add_addr_signal_max,
182 		 msk->pm.subflows, msk->pm.subflows_max);
183 
184 	/* check first for announce */
185 	if (msk->pm.add_addr_signaled < msk->pm.add_addr_signal_max) {
186 		local = select_signal_address(pernet,
187 					      msk->pm.add_addr_signaled);
188 
189 		if (local) {
190 			msk->pm.add_addr_signaled++;
191 			mptcp_pm_announce_addr(msk, &local->addr);
192 		} else {
193 			/* pick failed, avoid fourther attempts later */
194 			msk->pm.local_addr_used = msk->pm.add_addr_signal_max;
195 		}
196 
197 		check_work_pending(msk);
198 	}
199 
200 	/* check if should create a new subflow */
201 	if (msk->pm.local_addr_used < msk->pm.local_addr_max &&
202 	    msk->pm.subflows < msk->pm.subflows_max) {
203 		remote_address((struct sock_common *)sk, &remote);
204 
205 		local = select_local_address(pernet, msk);
206 		if (local) {
207 			msk->pm.local_addr_used++;
208 			msk->pm.subflows++;
209 			check_work_pending(msk);
210 			spin_unlock_bh(&msk->pm.lock);
211 			__mptcp_subflow_connect(sk, local->ifindex,
212 						&local->addr, &remote);
213 			spin_lock_bh(&msk->pm.lock);
214 			return;
215 		}
216 
217 		/* lookup failed, avoid fourther attempts later */
218 		msk->pm.local_addr_used = msk->pm.local_addr_max;
219 		check_work_pending(msk);
220 	}
221 }
222 
223 void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
224 {
225 	mptcp_pm_create_subflow_or_signal_addr(msk);
226 }
227 
228 void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
229 {
230 	mptcp_pm_create_subflow_or_signal_addr(msk);
231 }
232 
233 void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
234 {
235 	struct sock *sk = (struct sock *)msk;
236 	struct mptcp_addr_info remote;
237 	struct mptcp_addr_info local;
238 
239 	pr_debug("accepted %d:%d remote family %d",
240 		 msk->pm.add_addr_accepted, msk->pm.add_addr_accept_max,
241 		 msk->pm.remote.family);
242 	msk->pm.add_addr_accepted++;
243 	msk->pm.subflows++;
244 	if (msk->pm.add_addr_accepted >= msk->pm.add_addr_accept_max ||
245 	    msk->pm.subflows >= msk->pm.subflows_max)
246 		WRITE_ONCE(msk->pm.accept_addr, false);
247 
248 	/* connect to the specified remote address, using whatever
249 	 * local address the routing configuration will pick.
250 	 */
251 	remote = msk->pm.remote;
252 	if (!remote.port)
253 		remote.port = sk->sk_dport;
254 	memset(&local, 0, sizeof(local));
255 	local.family = remote.family;
256 
257 	spin_unlock_bh(&msk->pm.lock);
258 	__mptcp_subflow_connect((struct sock *)msk, 0, &local, &remote);
259 	spin_lock_bh(&msk->pm.lock);
260 }
261 
262 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
263 {
264 	return (entry->flags &
265 		(MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
266 		MPTCP_PM_ADDR_FLAG_SIGNAL;
267 }
268 
269 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
270 					     struct mptcp_pm_addr_entry *entry)
271 {
272 	struct mptcp_pm_addr_entry *cur;
273 	int ret = -EINVAL;
274 
275 	spin_lock_bh(&pernet->lock);
276 	/* to keep the code simple, don't do IDR-like allocation for address ID,
277 	 * just bail when we exceed limits
278 	 */
279 	if (pernet->next_id > 255)
280 		goto out;
281 	if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
282 		goto out;
283 
284 	/* do not insert duplicate address, differentiate on port only
285 	 * singled addresses
286 	 */
287 	list_for_each_entry(cur, &pernet->local_addr_list, list) {
288 		if (addresses_equal(&cur->addr, &entry->addr,
289 				    address_use_port(entry) &&
290 				    address_use_port(cur)))
291 			goto out;
292 	}
293 
294 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)
295 		pernet->add_addr_signal_max++;
296 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
297 		pernet->local_addr_max++;
298 
299 	entry->addr.id = pernet->next_id++;
300 	pernet->addrs++;
301 	list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
302 	ret = entry->addr.id;
303 
304 out:
305 	spin_unlock_bh(&pernet->lock);
306 	return ret;
307 }
308 
309 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
310 {
311 	struct mptcp_pm_addr_entry *entry;
312 	struct mptcp_addr_info skc_local;
313 	struct mptcp_addr_info msk_local;
314 	struct pm_nl_pernet *pernet;
315 	int ret = -1;
316 
317 	if (WARN_ON_ONCE(!msk))
318 		return -1;
319 
320 	/* The 0 ID mapping is defined by the first subflow, copied into the msk
321 	 * addr
322 	 */
323 	local_address((struct sock_common *)msk, &msk_local);
324 	local_address((struct sock_common *)msk, &skc_local);
325 	if (addresses_equal(&msk_local, &skc_local, false))
326 		return 0;
327 
328 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
329 
330 	rcu_read_lock();
331 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
332 		if (addresses_equal(&entry->addr, &skc_local, false)) {
333 			ret = entry->addr.id;
334 			break;
335 		}
336 	}
337 	rcu_read_unlock();
338 	if (ret >= 0)
339 		return ret;
340 
341 	/* address not found, add to local list */
342 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
343 	if (!entry)
344 		return -ENOMEM;
345 
346 	entry->flags = 0;
347 	entry->addr = skc_local;
348 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
349 	if (ret < 0)
350 		kfree(entry);
351 
352 	return ret;
353 }
354 
355 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
356 {
357 	struct mptcp_pm_data *pm = &msk->pm;
358 	struct pm_nl_pernet *pernet;
359 	bool subflows;
360 
361 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
362 
363 	pm->add_addr_signal_max = READ_ONCE(pernet->add_addr_signal_max);
364 	pm->add_addr_accept_max = READ_ONCE(pernet->add_addr_accept_max);
365 	pm->local_addr_max = READ_ONCE(pernet->local_addr_max);
366 	pm->subflows_max = READ_ONCE(pernet->subflows_max);
367 	subflows = !!pm->subflows_max;
368 	WRITE_ONCE(pm->work_pending, (!!pm->local_addr_max && subflows) ||
369 		   !!pm->add_addr_signal_max);
370 	WRITE_ONCE(pm->accept_addr, !!pm->add_addr_accept_max && subflows);
371 	WRITE_ONCE(pm->accept_subflow, subflows);
372 }
373 
374 #define MPTCP_PM_CMD_GRP_OFFSET	0
375 
376 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
377 	[MPTCP_PM_CMD_GRP_OFFSET]	= { .name = MPTCP_PM_CMD_GRP_NAME, },
378 };
379 
380 static const struct nla_policy
381 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
382 	[MPTCP_PM_ADDR_ATTR_FAMILY]	= { .type	= NLA_U16,	},
383 	[MPTCP_PM_ADDR_ATTR_ID]		= { .type	= NLA_U8,	},
384 	[MPTCP_PM_ADDR_ATTR_ADDR4]	= { .type	= NLA_U32,	},
385 	[MPTCP_PM_ADDR_ATTR_ADDR6]	= { .type	= NLA_EXACT_LEN,
386 					    .len   = sizeof(struct in6_addr), },
387 	[MPTCP_PM_ADDR_ATTR_PORT]	= { .type	= NLA_U16	},
388 	[MPTCP_PM_ADDR_ATTR_FLAGS]	= { .type	= NLA_U32	},
389 	[MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type	= NLA_S32	},
390 };
391 
392 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
393 	[MPTCP_PM_ATTR_ADDR]		=
394 					NLA_POLICY_NESTED(mptcp_pm_addr_policy),
395 	[MPTCP_PM_ATTR_RCV_ADD_ADDRS]	= { .type	= NLA_U32,	},
396 	[MPTCP_PM_ATTR_SUBFLOWS]	= { .type	= NLA_U32,	},
397 };
398 
399 static int mptcp_pm_family_to_addr(int family)
400 {
401 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
402 	if (family == AF_INET6)
403 		return MPTCP_PM_ADDR_ATTR_ADDR6;
404 #endif
405 	return MPTCP_PM_ADDR_ATTR_ADDR4;
406 }
407 
408 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
409 			       bool require_family,
410 			       struct mptcp_pm_addr_entry *entry)
411 {
412 	struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
413 	int err, addr_addr;
414 
415 	if (!attr) {
416 		GENL_SET_ERR_MSG(info, "missing address info");
417 		return -EINVAL;
418 	}
419 
420 	/* no validation needed - was already done via nested policy */
421 	err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
422 					  mptcp_pm_addr_policy, info->extack);
423 	if (err)
424 		return err;
425 
426 	memset(entry, 0, sizeof(*entry));
427 	if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
428 		if (!require_family)
429 			goto skip_family;
430 
431 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
432 				    "missing family");
433 		return -EINVAL;
434 	}
435 
436 	entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
437 	if (entry->addr.family != AF_INET
438 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
439 	    && entry->addr.family != AF_INET6
440 #endif
441 	    ) {
442 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
443 				    "unknown address family");
444 		return -EINVAL;
445 	}
446 	addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
447 	if (!tb[addr_addr]) {
448 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
449 				    "missing address data");
450 		return -EINVAL;
451 	}
452 
453 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
454 	if (entry->addr.family == AF_INET6)
455 		entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
456 	else
457 #endif
458 		entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
459 
460 skip_family:
461 	if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX])
462 		entry->ifindex = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
463 
464 	if (tb[MPTCP_PM_ADDR_ATTR_ID])
465 		entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
466 
467 	if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
468 		entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
469 
470 	return 0;
471 }
472 
473 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
474 {
475 	return net_generic(genl_info_net(info), pm_nl_pernet_id);
476 }
477 
478 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
479 {
480 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
481 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
482 	struct mptcp_pm_addr_entry addr, *entry;
483 	int ret;
484 
485 	ret = mptcp_pm_parse_addr(attr, info, true, &addr);
486 	if (ret < 0)
487 		return ret;
488 
489 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
490 	if (!entry) {
491 		GENL_SET_ERR_MSG(info, "can't allocate addr");
492 		return -ENOMEM;
493 	}
494 
495 	*entry = addr;
496 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
497 	if (ret < 0) {
498 		GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
499 		kfree(entry);
500 		return ret;
501 	}
502 
503 	return 0;
504 }
505 
506 static struct mptcp_pm_addr_entry *
507 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
508 {
509 	struct mptcp_pm_addr_entry *entry;
510 
511 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
512 		if (entry->addr.id == id)
513 			return entry;
514 	}
515 	return NULL;
516 }
517 
518 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
519 {
520 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
521 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
522 	struct mptcp_pm_addr_entry addr, *entry;
523 	int ret;
524 
525 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
526 	if (ret < 0)
527 		return ret;
528 
529 	spin_lock_bh(&pernet->lock);
530 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
531 	if (!entry) {
532 		GENL_SET_ERR_MSG(info, "address not found");
533 		ret = -EINVAL;
534 		goto out;
535 	}
536 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)
537 		pernet->add_addr_signal_max--;
538 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
539 		pernet->local_addr_max--;
540 
541 	pernet->addrs--;
542 	list_del_rcu(&entry->list);
543 	kfree_rcu(entry, rcu);
544 out:
545 	spin_unlock_bh(&pernet->lock);
546 	return ret;
547 }
548 
549 static void __flush_addrs(struct pm_nl_pernet *pernet)
550 {
551 	while (!list_empty(&pernet->local_addr_list)) {
552 		struct mptcp_pm_addr_entry *cur;
553 
554 		cur = list_entry(pernet->local_addr_list.next,
555 				 struct mptcp_pm_addr_entry, list);
556 		list_del_rcu(&cur->list);
557 		kfree_rcu(cur, rcu);
558 	}
559 }
560 
561 static void __reset_counters(struct pm_nl_pernet *pernet)
562 {
563 	pernet->add_addr_signal_max = 0;
564 	pernet->add_addr_accept_max = 0;
565 	pernet->local_addr_max = 0;
566 	pernet->addrs = 0;
567 }
568 
569 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
570 {
571 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
572 
573 	spin_lock_bh(&pernet->lock);
574 	__flush_addrs(pernet);
575 	__reset_counters(pernet);
576 	spin_unlock_bh(&pernet->lock);
577 	return 0;
578 }
579 
580 static int mptcp_nl_fill_addr(struct sk_buff *skb,
581 			      struct mptcp_pm_addr_entry *entry)
582 {
583 	struct mptcp_addr_info *addr = &entry->addr;
584 	struct nlattr *attr;
585 
586 	attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
587 	if (!attr)
588 		return -EMSGSIZE;
589 
590 	if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
591 		goto nla_put_failure;
592 	if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
593 		goto nla_put_failure;
594 	if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
595 		goto nla_put_failure;
596 	if (entry->ifindex &&
597 	    nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
598 		goto nla_put_failure;
599 
600 	if (addr->family == AF_INET)
601 		nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
602 				addr->addr.s_addr);
603 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
604 	else if (addr->family == AF_INET6)
605 		nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6);
606 #endif
607 	nla_nest_end(skb, attr);
608 	return 0;
609 
610 nla_put_failure:
611 	nla_nest_cancel(skb, attr);
612 	return -EMSGSIZE;
613 }
614 
615 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
616 {
617 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
618 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
619 	struct mptcp_pm_addr_entry addr, *entry;
620 	struct sk_buff *msg;
621 	void *reply;
622 	int ret;
623 
624 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
625 	if (ret < 0)
626 		return ret;
627 
628 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
629 	if (!msg)
630 		return -ENOMEM;
631 
632 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
633 				  info->genlhdr->cmd);
634 	if (!reply) {
635 		GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
636 		ret = -EMSGSIZE;
637 		goto fail;
638 	}
639 
640 	spin_lock_bh(&pernet->lock);
641 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
642 	if (!entry) {
643 		GENL_SET_ERR_MSG(info, "address not found");
644 		ret = -EINVAL;
645 		goto unlock_fail;
646 	}
647 
648 	ret = mptcp_nl_fill_addr(msg, entry);
649 	if (ret)
650 		goto unlock_fail;
651 
652 	genlmsg_end(msg, reply);
653 	ret = genlmsg_reply(msg, info);
654 	spin_unlock_bh(&pernet->lock);
655 	return ret;
656 
657 unlock_fail:
658 	spin_unlock_bh(&pernet->lock);
659 
660 fail:
661 	nlmsg_free(msg);
662 	return ret;
663 }
664 
665 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
666 				   struct netlink_callback *cb)
667 {
668 	struct net *net = sock_net(msg->sk);
669 	struct mptcp_pm_addr_entry *entry;
670 	struct pm_nl_pernet *pernet;
671 	int id = cb->args[0];
672 	void *hdr;
673 
674 	pernet = net_generic(net, pm_nl_pernet_id);
675 
676 	spin_lock_bh(&pernet->lock);
677 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
678 		if (entry->addr.id <= id)
679 			continue;
680 
681 		hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
682 				  cb->nlh->nlmsg_seq, &mptcp_genl_family,
683 				  NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
684 		if (!hdr)
685 			break;
686 
687 		if (mptcp_nl_fill_addr(msg, entry) < 0) {
688 			genlmsg_cancel(msg, hdr);
689 			break;
690 		}
691 
692 		id = entry->addr.id;
693 		genlmsg_end(msg, hdr);
694 	}
695 	spin_unlock_bh(&pernet->lock);
696 
697 	cb->args[0] = id;
698 	return msg->len;
699 }
700 
701 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
702 {
703 	struct nlattr *attr = info->attrs[id];
704 
705 	if (!attr)
706 		return 0;
707 
708 	*limit = nla_get_u32(attr);
709 	if (*limit > MPTCP_PM_ADDR_MAX) {
710 		GENL_SET_ERR_MSG(info, "limit greater than maximum");
711 		return -EINVAL;
712 	}
713 	return 0;
714 }
715 
716 static int
717 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
718 {
719 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
720 	unsigned int rcv_addrs, subflows;
721 	int ret;
722 
723 	spin_lock_bh(&pernet->lock);
724 	rcv_addrs = pernet->add_addr_accept_max;
725 	ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
726 	if (ret)
727 		goto unlock;
728 
729 	subflows = pernet->subflows_max;
730 	ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
731 	if (ret)
732 		goto unlock;
733 
734 	WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
735 	WRITE_ONCE(pernet->subflows_max, subflows);
736 
737 unlock:
738 	spin_unlock_bh(&pernet->lock);
739 	return ret;
740 }
741 
742 static int
743 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
744 {
745 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
746 	struct sk_buff *msg;
747 	void *reply;
748 
749 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
750 	if (!msg)
751 		return -ENOMEM;
752 
753 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
754 				  MPTCP_PM_CMD_GET_LIMITS);
755 	if (!reply)
756 		goto fail;
757 
758 	if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
759 			READ_ONCE(pernet->add_addr_accept_max)))
760 		goto fail;
761 
762 	if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
763 			READ_ONCE(pernet->subflows_max)))
764 		goto fail;
765 
766 	genlmsg_end(msg, reply);
767 	return genlmsg_reply(msg, info);
768 
769 fail:
770 	GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
771 	nlmsg_free(msg);
772 	return -EMSGSIZE;
773 }
774 
775 static struct genl_ops mptcp_pm_ops[] = {
776 	{
777 		.cmd    = MPTCP_PM_CMD_ADD_ADDR,
778 		.doit   = mptcp_nl_cmd_add_addr,
779 		.flags  = GENL_ADMIN_PERM,
780 	},
781 	{
782 		.cmd    = MPTCP_PM_CMD_DEL_ADDR,
783 		.doit   = mptcp_nl_cmd_del_addr,
784 		.flags  = GENL_ADMIN_PERM,
785 	},
786 	{
787 		.cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
788 		.doit   = mptcp_nl_cmd_flush_addrs,
789 		.flags  = GENL_ADMIN_PERM,
790 	},
791 	{
792 		.cmd    = MPTCP_PM_CMD_GET_ADDR,
793 		.doit   = mptcp_nl_cmd_get_addr,
794 		.dumpit   = mptcp_nl_cmd_dump_addrs,
795 	},
796 	{
797 		.cmd    = MPTCP_PM_CMD_SET_LIMITS,
798 		.doit   = mptcp_nl_cmd_set_limits,
799 		.flags  = GENL_ADMIN_PERM,
800 	},
801 	{
802 		.cmd    = MPTCP_PM_CMD_GET_LIMITS,
803 		.doit   = mptcp_nl_cmd_get_limits,
804 	},
805 };
806 
807 static struct genl_family mptcp_genl_family __ro_after_init = {
808 	.name		= MPTCP_PM_NAME,
809 	.version	= MPTCP_PM_VER,
810 	.maxattr	= MPTCP_PM_ATTR_MAX,
811 	.policy		= mptcp_pm_policy,
812 	.netnsok	= true,
813 	.module		= THIS_MODULE,
814 	.ops		= mptcp_pm_ops,
815 	.n_ops		= ARRAY_SIZE(mptcp_pm_ops),
816 	.mcgrps		= mptcp_pm_mcgrps,
817 	.n_mcgrps	= ARRAY_SIZE(mptcp_pm_mcgrps),
818 };
819 
820 static int __net_init pm_nl_init_net(struct net *net)
821 {
822 	struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
823 
824 	INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
825 	__reset_counters(pernet);
826 	pernet->next_id = 1;
827 	spin_lock_init(&pernet->lock);
828 	return 0;
829 }
830 
831 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
832 {
833 	struct net *net;
834 
835 	list_for_each_entry(net, net_list, exit_list) {
836 		/* net is removed from namespace list, can't race with
837 		 * other modifiers
838 		 */
839 		__flush_addrs(net_generic(net, pm_nl_pernet_id));
840 	}
841 }
842 
843 static struct pernet_operations mptcp_pm_pernet_ops = {
844 	.init = pm_nl_init_net,
845 	.exit_batch = pm_nl_exit_net,
846 	.id = &pm_nl_pernet_id,
847 	.size = sizeof(struct pm_nl_pernet),
848 };
849 
850 void mptcp_pm_nl_init(void)
851 {
852 	if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
853 		panic("Failed to register MPTCP PM pernet subsystem.\n");
854 
855 	if (genl_register_family(&mptcp_genl_family))
856 		panic("Failed to register MPTCP PM netlink family\n");
857 }
858