xref: /openbmc/linux/net/mptcp/pm_netlink.c (revision e0d07278)
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 
19 /* forward declaration */
20 static struct genl_family mptcp_genl_family;
21 
22 static int pm_nl_pernet_id;
23 
24 struct mptcp_pm_addr_entry {
25 	struct list_head	list;
26 	unsigned int		flags;
27 	int			ifindex;
28 	struct mptcp_addr_info	addr;
29 	struct rcu_head		rcu;
30 };
31 
32 struct pm_nl_pernet {
33 	/* protects pernet updates */
34 	spinlock_t		lock;
35 	struct list_head	local_addr_list;
36 	unsigned int		addrs;
37 	unsigned int		add_addr_signal_max;
38 	unsigned int		add_addr_accept_max;
39 	unsigned int		local_addr_max;
40 	unsigned int		subflows_max;
41 	unsigned int		next_id;
42 };
43 
44 #define MPTCP_PM_ADDR_MAX	8
45 
46 static bool addresses_equal(const struct mptcp_addr_info *a,
47 			    struct mptcp_addr_info *b, bool use_port)
48 {
49 	bool addr_equals = false;
50 
51 	if (a->family != b->family)
52 		return false;
53 
54 	if (a->family == AF_INET)
55 		addr_equals = a->addr.s_addr == b->addr.s_addr;
56 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
57 	else
58 		addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
59 #endif
60 
61 	if (!addr_equals)
62 		return false;
63 	if (!use_port)
64 		return true;
65 
66 	return a->port == b->port;
67 }
68 
69 static void local_address(const struct sock_common *skc,
70 			  struct mptcp_addr_info *addr)
71 {
72 	addr->port = 0;
73 	addr->family = skc->skc_family;
74 	if (addr->family == AF_INET)
75 		addr->addr.s_addr = skc->skc_rcv_saddr;
76 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
77 	else if (addr->family == AF_INET6)
78 		addr->addr6 = skc->skc_v6_rcv_saddr;
79 #endif
80 }
81 
82 static void remote_address(const struct sock_common *skc,
83 			   struct mptcp_addr_info *addr)
84 {
85 	addr->family = skc->skc_family;
86 	addr->port = skc->skc_dport;
87 	if (addr->family == AF_INET)
88 		addr->addr.s_addr = skc->skc_daddr;
89 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
90 	else if (addr->family == AF_INET6)
91 		addr->addr6 = skc->skc_v6_daddr;
92 #endif
93 }
94 
95 static bool lookup_subflow_by_saddr(const struct list_head *list,
96 				    struct mptcp_addr_info *saddr)
97 {
98 	struct mptcp_subflow_context *subflow;
99 	struct mptcp_addr_info cur;
100 	struct sock_common *skc;
101 
102 	list_for_each_entry(subflow, list, node) {
103 		skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
104 
105 		local_address(skc, &cur);
106 		if (addresses_equal(&cur, saddr, false))
107 			return true;
108 	}
109 
110 	return false;
111 }
112 
113 static struct mptcp_pm_addr_entry *
114 select_local_address(const struct pm_nl_pernet *pernet,
115 		     struct mptcp_sock *msk)
116 {
117 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
118 
119 	rcu_read_lock();
120 	spin_lock_bh(&msk->join_list_lock);
121 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
122 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
123 			continue;
124 
125 		/* avoid any address already in use by subflows and
126 		 * pending join
127 		 */
128 		if (entry->addr.family == ((struct sock *)msk)->sk_family &&
129 		    !lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
130 		    !lookup_subflow_by_saddr(&msk->join_list, &entry->addr)) {
131 			ret = entry;
132 			break;
133 		}
134 	}
135 	spin_unlock_bh(&msk->join_list_lock);
136 	rcu_read_unlock();
137 	return ret;
138 }
139 
140 static struct mptcp_pm_addr_entry *
141 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
142 {
143 	struct mptcp_pm_addr_entry *entry, *ret = NULL;
144 	int i = 0;
145 
146 	rcu_read_lock();
147 	/* do not keep any additional per socket state, just signal
148 	 * the address list in order.
149 	 * Note: removal from the local address list during the msk life-cycle
150 	 * can lead to additional addresses not being announced.
151 	 */
152 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
153 		if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
154 			continue;
155 		if (i++ == pos) {
156 			ret = entry;
157 			break;
158 		}
159 	}
160 	rcu_read_unlock();
161 	return ret;
162 }
163 
164 static void check_work_pending(struct mptcp_sock *msk)
165 {
166 	if (msk->pm.add_addr_signaled == msk->pm.add_addr_signal_max &&
167 	    (msk->pm.local_addr_used == msk->pm.local_addr_max ||
168 	     msk->pm.subflows == msk->pm.subflows_max))
169 		WRITE_ONCE(msk->pm.work_pending, false);
170 }
171 
172 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
173 {
174 	struct sock *sk = (struct sock *)msk;
175 	struct mptcp_pm_addr_entry *local;
176 	struct mptcp_addr_info remote;
177 	struct pm_nl_pernet *pernet;
178 
179 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
180 
181 	pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
182 		 msk->pm.local_addr_used, msk->pm.local_addr_max,
183 		 msk->pm.add_addr_signaled, msk->pm.add_addr_signal_max,
184 		 msk->pm.subflows, msk->pm.subflows_max);
185 
186 	/* check first for announce */
187 	if (msk->pm.add_addr_signaled < msk->pm.add_addr_signal_max) {
188 		local = select_signal_address(pernet,
189 					      msk->pm.add_addr_signaled);
190 
191 		if (local) {
192 			msk->pm.add_addr_signaled++;
193 			mptcp_pm_announce_addr(msk, &local->addr);
194 		} else {
195 			/* pick failed, avoid fourther attempts later */
196 			msk->pm.local_addr_used = msk->pm.add_addr_signal_max;
197 		}
198 
199 		check_work_pending(msk);
200 	}
201 
202 	/* check if should create a new subflow */
203 	if (msk->pm.local_addr_used < msk->pm.local_addr_max &&
204 	    msk->pm.subflows < msk->pm.subflows_max) {
205 		remote_address((struct sock_common *)sk, &remote);
206 
207 		local = select_local_address(pernet, msk);
208 		if (local) {
209 			msk->pm.local_addr_used++;
210 			msk->pm.subflows++;
211 			check_work_pending(msk);
212 			spin_unlock_bh(&msk->pm.lock);
213 			__mptcp_subflow_connect(sk, local->ifindex,
214 						&local->addr, &remote);
215 			spin_lock_bh(&msk->pm.lock);
216 			return;
217 		}
218 
219 		/* lookup failed, avoid fourther attempts later */
220 		msk->pm.local_addr_used = msk->pm.local_addr_max;
221 		check_work_pending(msk);
222 	}
223 }
224 
225 void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
226 {
227 	mptcp_pm_create_subflow_or_signal_addr(msk);
228 }
229 
230 void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
231 {
232 	mptcp_pm_create_subflow_or_signal_addr(msk);
233 }
234 
235 void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
236 {
237 	struct sock *sk = (struct sock *)msk;
238 	struct mptcp_addr_info remote;
239 	struct mptcp_addr_info local;
240 
241 	pr_debug("accepted %d:%d remote family %d",
242 		 msk->pm.add_addr_accepted, msk->pm.add_addr_accept_max,
243 		 msk->pm.remote.family);
244 	msk->pm.add_addr_accepted++;
245 	msk->pm.subflows++;
246 	if (msk->pm.add_addr_accepted >= msk->pm.add_addr_accept_max ||
247 	    msk->pm.subflows >= msk->pm.subflows_max)
248 		WRITE_ONCE(msk->pm.accept_addr, false);
249 
250 	/* connect to the specified remote address, using whatever
251 	 * local address the routing configuration will pick.
252 	 */
253 	remote = msk->pm.remote;
254 	if (!remote.port)
255 		remote.port = sk->sk_dport;
256 	memset(&local, 0, sizeof(local));
257 	local.family = remote.family;
258 
259 	spin_unlock_bh(&msk->pm.lock);
260 	__mptcp_subflow_connect((struct sock *)msk, 0, &local, &remote);
261 	spin_lock_bh(&msk->pm.lock);
262 }
263 
264 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
265 {
266 	return (entry->flags &
267 		(MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
268 		MPTCP_PM_ADDR_FLAG_SIGNAL;
269 }
270 
271 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
272 					     struct mptcp_pm_addr_entry *entry)
273 {
274 	struct mptcp_pm_addr_entry *cur;
275 	int ret = -EINVAL;
276 
277 	spin_lock_bh(&pernet->lock);
278 	/* to keep the code simple, don't do IDR-like allocation for address ID,
279 	 * just bail when we exceed limits
280 	 */
281 	if (pernet->next_id > 255)
282 		goto out;
283 	if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
284 		goto out;
285 
286 	/* do not insert duplicate address, differentiate on port only
287 	 * singled addresses
288 	 */
289 	list_for_each_entry(cur, &pernet->local_addr_list, list) {
290 		if (addresses_equal(&cur->addr, &entry->addr,
291 				    address_use_port(entry) &&
292 				    address_use_port(cur)))
293 			goto out;
294 	}
295 
296 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)
297 		pernet->add_addr_signal_max++;
298 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
299 		pernet->local_addr_max++;
300 
301 	entry->addr.id = pernet->next_id++;
302 	pernet->addrs++;
303 	list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
304 	ret = entry->addr.id;
305 
306 out:
307 	spin_unlock_bh(&pernet->lock);
308 	return ret;
309 }
310 
311 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
312 {
313 	struct mptcp_pm_addr_entry *entry;
314 	struct mptcp_addr_info skc_local;
315 	struct mptcp_addr_info msk_local;
316 	struct pm_nl_pernet *pernet;
317 	int ret = -1;
318 
319 	if (WARN_ON_ONCE(!msk))
320 		return -1;
321 
322 	/* The 0 ID mapping is defined by the first subflow, copied into the msk
323 	 * addr
324 	 */
325 	local_address((struct sock_common *)msk, &msk_local);
326 	local_address((struct sock_common *)msk, &skc_local);
327 	if (addresses_equal(&msk_local, &skc_local, false))
328 		return 0;
329 
330 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
331 
332 	rcu_read_lock();
333 	list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
334 		if (addresses_equal(&entry->addr, &skc_local, false)) {
335 			ret = entry->addr.id;
336 			break;
337 		}
338 	}
339 	rcu_read_unlock();
340 	if (ret >= 0)
341 		return ret;
342 
343 	/* address not found, add to local list */
344 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
345 	if (!entry)
346 		return -ENOMEM;
347 
348 	entry->flags = 0;
349 	entry->addr = skc_local;
350 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
351 	if (ret < 0)
352 		kfree(entry);
353 
354 	return ret;
355 }
356 
357 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
358 {
359 	struct mptcp_pm_data *pm = &msk->pm;
360 	struct pm_nl_pernet *pernet;
361 	bool subflows;
362 
363 	pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
364 
365 	pm->add_addr_signal_max = READ_ONCE(pernet->add_addr_signal_max);
366 	pm->add_addr_accept_max = READ_ONCE(pernet->add_addr_accept_max);
367 	pm->local_addr_max = READ_ONCE(pernet->local_addr_max);
368 	pm->subflows_max = READ_ONCE(pernet->subflows_max);
369 	subflows = !!pm->subflows_max;
370 	WRITE_ONCE(pm->work_pending, (!!pm->local_addr_max && subflows) ||
371 		   !!pm->add_addr_signal_max);
372 	WRITE_ONCE(pm->accept_addr, !!pm->add_addr_accept_max && subflows);
373 	WRITE_ONCE(pm->accept_subflow, subflows);
374 }
375 
376 #define MPTCP_PM_CMD_GRP_OFFSET	0
377 
378 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
379 	[MPTCP_PM_CMD_GRP_OFFSET]	= { .name = MPTCP_PM_CMD_GRP_NAME, },
380 };
381 
382 static const struct nla_policy
383 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
384 	[MPTCP_PM_ADDR_ATTR_FAMILY]	= { .type	= NLA_U16,	},
385 	[MPTCP_PM_ADDR_ATTR_ID]		= { .type	= NLA_U8,	},
386 	[MPTCP_PM_ADDR_ATTR_ADDR4]	= { .type	= NLA_U32,	},
387 	[MPTCP_PM_ADDR_ATTR_ADDR6]	= { .type	= NLA_EXACT_LEN,
388 					    .len   = sizeof(struct in6_addr), },
389 	[MPTCP_PM_ADDR_ATTR_PORT]	= { .type	= NLA_U16	},
390 	[MPTCP_PM_ADDR_ATTR_FLAGS]	= { .type	= NLA_U32	},
391 	[MPTCP_PM_ADDR_ATTR_IF_IDX]     = { .type	= NLA_S32	},
392 };
393 
394 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
395 	[MPTCP_PM_ATTR_ADDR]		=
396 					NLA_POLICY_NESTED(mptcp_pm_addr_policy),
397 	[MPTCP_PM_ATTR_RCV_ADD_ADDRS]	= { .type	= NLA_U32,	},
398 	[MPTCP_PM_ATTR_SUBFLOWS]	= { .type	= NLA_U32,	},
399 };
400 
401 static int mptcp_pm_family_to_addr(int family)
402 {
403 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
404 	if (family == AF_INET6)
405 		return MPTCP_PM_ADDR_ATTR_ADDR6;
406 #endif
407 	return MPTCP_PM_ADDR_ATTR_ADDR4;
408 }
409 
410 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
411 			       bool require_family,
412 			       struct mptcp_pm_addr_entry *entry)
413 {
414 	struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
415 	int err, addr_addr;
416 
417 	if (!attr) {
418 		GENL_SET_ERR_MSG(info, "missing address info");
419 		return -EINVAL;
420 	}
421 
422 	/* no validation needed - was already done via nested policy */
423 	err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
424 					  mptcp_pm_addr_policy, info->extack);
425 	if (err)
426 		return err;
427 
428 	memset(entry, 0, sizeof(*entry));
429 	if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
430 		if (!require_family)
431 			goto skip_family;
432 
433 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
434 				    "missing family");
435 		return -EINVAL;
436 	}
437 
438 	entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
439 	if (entry->addr.family != AF_INET
440 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
441 	    && entry->addr.family != AF_INET6
442 #endif
443 	    ) {
444 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
445 				    "unknown address family");
446 		return -EINVAL;
447 	}
448 	addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
449 	if (!tb[addr_addr]) {
450 		NL_SET_ERR_MSG_ATTR(info->extack, attr,
451 				    "missing address data");
452 		return -EINVAL;
453 	}
454 
455 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
456 	if (entry->addr.family == AF_INET6)
457 		entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
458 	else
459 #endif
460 		entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
461 
462 skip_family:
463 	if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX])
464 		entry->ifindex = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
465 
466 	if (tb[MPTCP_PM_ADDR_ATTR_ID])
467 		entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
468 
469 	if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
470 		entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
471 
472 	return 0;
473 }
474 
475 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
476 {
477 	return net_generic(genl_info_net(info), pm_nl_pernet_id);
478 }
479 
480 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
481 {
482 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
483 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
484 	struct mptcp_pm_addr_entry addr, *entry;
485 	int ret;
486 
487 	ret = mptcp_pm_parse_addr(attr, info, true, &addr);
488 	if (ret < 0)
489 		return ret;
490 
491 	entry = kmalloc(sizeof(*entry), GFP_KERNEL);
492 	if (!entry) {
493 		GENL_SET_ERR_MSG(info, "can't allocate addr");
494 		return -ENOMEM;
495 	}
496 
497 	*entry = addr;
498 	ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
499 	if (ret < 0) {
500 		GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
501 		kfree(entry);
502 		return ret;
503 	}
504 
505 	return 0;
506 }
507 
508 static struct mptcp_pm_addr_entry *
509 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
510 {
511 	struct mptcp_pm_addr_entry *entry;
512 
513 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
514 		if (entry->addr.id == id)
515 			return entry;
516 	}
517 	return NULL;
518 }
519 
520 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
521 {
522 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
523 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
524 	struct mptcp_pm_addr_entry addr, *entry;
525 	int ret;
526 
527 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
528 	if (ret < 0)
529 		return ret;
530 
531 	spin_lock_bh(&pernet->lock);
532 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
533 	if (!entry) {
534 		GENL_SET_ERR_MSG(info, "address not found");
535 		ret = -EINVAL;
536 		goto out;
537 	}
538 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)
539 		pernet->add_addr_signal_max--;
540 	if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW)
541 		pernet->local_addr_max--;
542 
543 	pernet->addrs--;
544 	list_del_rcu(&entry->list);
545 	kfree_rcu(entry, rcu);
546 out:
547 	spin_unlock_bh(&pernet->lock);
548 	return ret;
549 }
550 
551 static void __flush_addrs(struct pm_nl_pernet *pernet)
552 {
553 	while (!list_empty(&pernet->local_addr_list)) {
554 		struct mptcp_pm_addr_entry *cur;
555 
556 		cur = list_entry(pernet->local_addr_list.next,
557 				 struct mptcp_pm_addr_entry, list);
558 		list_del_rcu(&cur->list);
559 		kfree_rcu(cur, rcu);
560 	}
561 }
562 
563 static void __reset_counters(struct pm_nl_pernet *pernet)
564 {
565 	pernet->add_addr_signal_max = 0;
566 	pernet->add_addr_accept_max = 0;
567 	pernet->local_addr_max = 0;
568 	pernet->addrs = 0;
569 }
570 
571 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
572 {
573 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
574 
575 	spin_lock_bh(&pernet->lock);
576 	__flush_addrs(pernet);
577 	__reset_counters(pernet);
578 	spin_unlock_bh(&pernet->lock);
579 	return 0;
580 }
581 
582 static int mptcp_nl_fill_addr(struct sk_buff *skb,
583 			      struct mptcp_pm_addr_entry *entry)
584 {
585 	struct mptcp_addr_info *addr = &entry->addr;
586 	struct nlattr *attr;
587 
588 	attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
589 	if (!attr)
590 		return -EMSGSIZE;
591 
592 	if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
593 		goto nla_put_failure;
594 	if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
595 		goto nla_put_failure;
596 	if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
597 		goto nla_put_failure;
598 	if (entry->ifindex &&
599 	    nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
600 		goto nla_put_failure;
601 
602 	if (addr->family == AF_INET &&
603 	    nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
604 			    addr->addr.s_addr))
605 		goto nla_put_failure;
606 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
607 	else if (addr->family == AF_INET6 &&
608 		 nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
609 		goto nla_put_failure;
610 #endif
611 	nla_nest_end(skb, attr);
612 	return 0;
613 
614 nla_put_failure:
615 	nla_nest_cancel(skb, attr);
616 	return -EMSGSIZE;
617 }
618 
619 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
620 {
621 	struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
622 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
623 	struct mptcp_pm_addr_entry addr, *entry;
624 	struct sk_buff *msg;
625 	void *reply;
626 	int ret;
627 
628 	ret = mptcp_pm_parse_addr(attr, info, false, &addr);
629 	if (ret < 0)
630 		return ret;
631 
632 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
633 	if (!msg)
634 		return -ENOMEM;
635 
636 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
637 				  info->genlhdr->cmd);
638 	if (!reply) {
639 		GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
640 		ret = -EMSGSIZE;
641 		goto fail;
642 	}
643 
644 	spin_lock_bh(&pernet->lock);
645 	entry = __lookup_addr_by_id(pernet, addr.addr.id);
646 	if (!entry) {
647 		GENL_SET_ERR_MSG(info, "address not found");
648 		ret = -EINVAL;
649 		goto unlock_fail;
650 	}
651 
652 	ret = mptcp_nl_fill_addr(msg, entry);
653 	if (ret)
654 		goto unlock_fail;
655 
656 	genlmsg_end(msg, reply);
657 	ret = genlmsg_reply(msg, info);
658 	spin_unlock_bh(&pernet->lock);
659 	return ret;
660 
661 unlock_fail:
662 	spin_unlock_bh(&pernet->lock);
663 
664 fail:
665 	nlmsg_free(msg);
666 	return ret;
667 }
668 
669 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
670 				   struct netlink_callback *cb)
671 {
672 	struct net *net = sock_net(msg->sk);
673 	struct mptcp_pm_addr_entry *entry;
674 	struct pm_nl_pernet *pernet;
675 	int id = cb->args[0];
676 	void *hdr;
677 
678 	pernet = net_generic(net, pm_nl_pernet_id);
679 
680 	spin_lock_bh(&pernet->lock);
681 	list_for_each_entry(entry, &pernet->local_addr_list, list) {
682 		if (entry->addr.id <= id)
683 			continue;
684 
685 		hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
686 				  cb->nlh->nlmsg_seq, &mptcp_genl_family,
687 				  NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
688 		if (!hdr)
689 			break;
690 
691 		if (mptcp_nl_fill_addr(msg, entry) < 0) {
692 			genlmsg_cancel(msg, hdr);
693 			break;
694 		}
695 
696 		id = entry->addr.id;
697 		genlmsg_end(msg, hdr);
698 	}
699 	spin_unlock_bh(&pernet->lock);
700 
701 	cb->args[0] = id;
702 	return msg->len;
703 }
704 
705 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
706 {
707 	struct nlattr *attr = info->attrs[id];
708 
709 	if (!attr)
710 		return 0;
711 
712 	*limit = nla_get_u32(attr);
713 	if (*limit > MPTCP_PM_ADDR_MAX) {
714 		GENL_SET_ERR_MSG(info, "limit greater than maximum");
715 		return -EINVAL;
716 	}
717 	return 0;
718 }
719 
720 static int
721 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
722 {
723 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
724 	unsigned int rcv_addrs, subflows;
725 	int ret;
726 
727 	spin_lock_bh(&pernet->lock);
728 	rcv_addrs = pernet->add_addr_accept_max;
729 	ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
730 	if (ret)
731 		goto unlock;
732 
733 	subflows = pernet->subflows_max;
734 	ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
735 	if (ret)
736 		goto unlock;
737 
738 	WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
739 	WRITE_ONCE(pernet->subflows_max, subflows);
740 
741 unlock:
742 	spin_unlock_bh(&pernet->lock);
743 	return ret;
744 }
745 
746 static int
747 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
748 {
749 	struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
750 	struct sk_buff *msg;
751 	void *reply;
752 
753 	msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
754 	if (!msg)
755 		return -ENOMEM;
756 
757 	reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
758 				  MPTCP_PM_CMD_GET_LIMITS);
759 	if (!reply)
760 		goto fail;
761 
762 	if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
763 			READ_ONCE(pernet->add_addr_accept_max)))
764 		goto fail;
765 
766 	if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
767 			READ_ONCE(pernet->subflows_max)))
768 		goto fail;
769 
770 	genlmsg_end(msg, reply);
771 	return genlmsg_reply(msg, info);
772 
773 fail:
774 	GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
775 	nlmsg_free(msg);
776 	return -EMSGSIZE;
777 }
778 
779 static struct genl_ops mptcp_pm_ops[] = {
780 	{
781 		.cmd    = MPTCP_PM_CMD_ADD_ADDR,
782 		.doit   = mptcp_nl_cmd_add_addr,
783 		.flags  = GENL_ADMIN_PERM,
784 	},
785 	{
786 		.cmd    = MPTCP_PM_CMD_DEL_ADDR,
787 		.doit   = mptcp_nl_cmd_del_addr,
788 		.flags  = GENL_ADMIN_PERM,
789 	},
790 	{
791 		.cmd    = MPTCP_PM_CMD_FLUSH_ADDRS,
792 		.doit   = mptcp_nl_cmd_flush_addrs,
793 		.flags  = GENL_ADMIN_PERM,
794 	},
795 	{
796 		.cmd    = MPTCP_PM_CMD_GET_ADDR,
797 		.doit   = mptcp_nl_cmd_get_addr,
798 		.dumpit   = mptcp_nl_cmd_dump_addrs,
799 	},
800 	{
801 		.cmd    = MPTCP_PM_CMD_SET_LIMITS,
802 		.doit   = mptcp_nl_cmd_set_limits,
803 		.flags  = GENL_ADMIN_PERM,
804 	},
805 	{
806 		.cmd    = MPTCP_PM_CMD_GET_LIMITS,
807 		.doit   = mptcp_nl_cmd_get_limits,
808 	},
809 };
810 
811 static struct genl_family mptcp_genl_family __ro_after_init = {
812 	.name		= MPTCP_PM_NAME,
813 	.version	= MPTCP_PM_VER,
814 	.maxattr	= MPTCP_PM_ATTR_MAX,
815 	.policy		= mptcp_pm_policy,
816 	.netnsok	= true,
817 	.module		= THIS_MODULE,
818 	.ops		= mptcp_pm_ops,
819 	.n_ops		= ARRAY_SIZE(mptcp_pm_ops),
820 	.mcgrps		= mptcp_pm_mcgrps,
821 	.n_mcgrps	= ARRAY_SIZE(mptcp_pm_mcgrps),
822 };
823 
824 static int __net_init pm_nl_init_net(struct net *net)
825 {
826 	struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
827 
828 	INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
829 	__reset_counters(pernet);
830 	pernet->next_id = 1;
831 	spin_lock_init(&pernet->lock);
832 	return 0;
833 }
834 
835 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
836 {
837 	struct net *net;
838 
839 	list_for_each_entry(net, net_list, exit_list) {
840 		/* net is removed from namespace list, can't race with
841 		 * other modifiers
842 		 */
843 		__flush_addrs(net_generic(net, pm_nl_pernet_id));
844 	}
845 }
846 
847 static struct pernet_operations mptcp_pm_pernet_ops = {
848 	.init = pm_nl_init_net,
849 	.exit_batch = pm_nl_exit_net,
850 	.id = &pm_nl_pernet_id,
851 	.size = sizeof(struct pm_nl_pernet),
852 };
853 
854 void __init mptcp_pm_nl_init(void)
855 {
856 	if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
857 		panic("Failed to register MPTCP PM pernet subsystem.\n");
858 
859 	if (genl_register_family(&mptcp_genl_family))
860 		panic("Failed to register MPTCP PM netlink family\n");
861 }
862