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 rcu_head rcu;
31 struct socket *lsk;
32 };
33
34 struct mptcp_pm_add_entry {
35 struct list_head list;
36 struct mptcp_addr_info addr;
37 struct timer_list add_timer;
38 struct mptcp_sock *sock;
39 u8 retrans_times;
40 };
41
42 #define MAX_ADDR_ID 255
43 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
44
45 struct pm_nl_pernet {
46 /* protects pernet updates */
47 spinlock_t lock;
48 struct list_head local_addr_list;
49 unsigned int addrs;
50 unsigned int add_addr_signal_max;
51 unsigned int add_addr_accept_max;
52 unsigned int local_addr_max;
53 unsigned int subflows_max;
54 unsigned int next_id;
55 unsigned long id_bitmap[BITMAP_SZ];
56 };
57
58 #define MPTCP_PM_ADDR_MAX 8
59 #define ADD_ADDR_RETRANS_MAX 3
60
addresses_equal(const struct mptcp_addr_info * a,struct mptcp_addr_info * b,bool use_port)61 static bool addresses_equal(const struct mptcp_addr_info *a,
62 struct mptcp_addr_info *b, bool use_port)
63 {
64 bool addr_equals = false;
65
66 if (a->family == b->family) {
67 if (a->family == AF_INET)
68 addr_equals = a->addr.s_addr == b->addr.s_addr;
69 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
70 else
71 addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
72 } else if (a->family == AF_INET) {
73 if (ipv6_addr_v4mapped(&b->addr6))
74 addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
75 } else if (b->family == AF_INET) {
76 if (ipv6_addr_v4mapped(&a->addr6))
77 addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
78 #endif
79 }
80
81 if (!addr_equals)
82 return false;
83 if (!use_port)
84 return true;
85
86 return a->port == b->port;
87 }
88
address_zero(const struct mptcp_addr_info * addr)89 static bool address_zero(const struct mptcp_addr_info *addr)
90 {
91 struct mptcp_addr_info zero;
92
93 memset(&zero, 0, sizeof(zero));
94 zero.family = addr->family;
95
96 return addresses_equal(addr, &zero, true);
97 }
98
local_address(const struct sock_common * skc,struct mptcp_addr_info * addr)99 static void local_address(const struct sock_common *skc,
100 struct mptcp_addr_info *addr)
101 {
102 addr->family = skc->skc_family;
103 addr->port = htons(skc->skc_num);
104 if (addr->family == AF_INET)
105 addr->addr.s_addr = skc->skc_rcv_saddr;
106 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
107 else if (addr->family == AF_INET6)
108 addr->addr6 = skc->skc_v6_rcv_saddr;
109 #endif
110 }
111
remote_address(const struct sock_common * skc,struct mptcp_addr_info * addr)112 static void remote_address(const struct sock_common *skc,
113 struct mptcp_addr_info *addr)
114 {
115 addr->family = skc->skc_family;
116 addr->port = skc->skc_dport;
117 if (addr->family == AF_INET)
118 addr->addr.s_addr = skc->skc_daddr;
119 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
120 else if (addr->family == AF_INET6)
121 addr->addr6 = skc->skc_v6_daddr;
122 #endif
123 }
124
lookup_subflow_by_saddr(const struct list_head * list,struct mptcp_addr_info * saddr)125 static bool lookup_subflow_by_saddr(const struct list_head *list,
126 struct mptcp_addr_info *saddr)
127 {
128 struct mptcp_subflow_context *subflow;
129 struct mptcp_addr_info cur;
130 struct sock_common *skc;
131
132 list_for_each_entry(subflow, list, node) {
133 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
134
135 local_address(skc, &cur);
136 if (addresses_equal(&cur, saddr, saddr->port))
137 return true;
138 }
139
140 return false;
141 }
142
lookup_subflow_by_daddr(const struct list_head * list,struct mptcp_addr_info * daddr)143 static bool lookup_subflow_by_daddr(const struct list_head *list,
144 struct mptcp_addr_info *daddr)
145 {
146 struct mptcp_subflow_context *subflow;
147 struct mptcp_addr_info cur;
148 struct sock_common *skc;
149
150 list_for_each_entry(subflow, list, node) {
151 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
152
153 remote_address(skc, &cur);
154 if (addresses_equal(&cur, daddr, daddr->port))
155 return true;
156 }
157
158 return false;
159 }
160
161 static struct mptcp_pm_addr_entry *
select_local_address(const struct pm_nl_pernet * pernet,struct mptcp_sock * msk)162 select_local_address(const struct pm_nl_pernet *pernet,
163 struct mptcp_sock *msk)
164 {
165 struct mptcp_pm_addr_entry *entry, *ret = NULL;
166 struct sock *sk = (struct sock *)msk;
167
168 msk_owned_by_me(msk);
169
170 rcu_read_lock();
171 __mptcp_flush_join_list(msk);
172 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
173 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
174 continue;
175
176 if (entry->addr.family != sk->sk_family) {
177 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
178 if ((entry->addr.family == AF_INET &&
179 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
180 (sk->sk_family == AF_INET &&
181 !ipv6_addr_v4mapped(&entry->addr.addr6)))
182 #endif
183 continue;
184 }
185
186 /* avoid any address already in use by subflows and
187 * pending join
188 */
189 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
190 ret = entry;
191 break;
192 }
193 }
194 rcu_read_unlock();
195 return ret;
196 }
197
198 static struct mptcp_pm_addr_entry *
select_signal_address(struct pm_nl_pernet * pernet,unsigned int pos)199 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
200 {
201 struct mptcp_pm_addr_entry *entry, *ret = NULL;
202 int i = 0;
203
204 rcu_read_lock();
205 /* do not keep any additional per socket state, just signal
206 * the address list in order.
207 * Note: removal from the local address list during the msk life-cycle
208 * can lead to additional addresses not being announced.
209 */
210 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
211 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
212 continue;
213 if (i++ == pos) {
214 ret = entry;
215 break;
216 }
217 }
218 rcu_read_unlock();
219 return ret;
220 }
221
mptcp_pm_get_add_addr_signal_max(struct mptcp_sock * msk)222 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
223 {
224 struct pm_nl_pernet *pernet;
225
226 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
227 return READ_ONCE(pernet->add_addr_signal_max);
228 }
229 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
230
mptcp_pm_get_add_addr_accept_max(struct mptcp_sock * msk)231 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
232 {
233 struct pm_nl_pernet *pernet;
234
235 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
236 return READ_ONCE(pernet->add_addr_accept_max);
237 }
238 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
239
mptcp_pm_get_subflows_max(struct mptcp_sock * msk)240 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
241 {
242 struct pm_nl_pernet *pernet;
243
244 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
245 return READ_ONCE(pernet->subflows_max);
246 }
247 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
248
mptcp_pm_get_local_addr_max(struct mptcp_sock * msk)249 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
250 {
251 struct pm_nl_pernet *pernet;
252
253 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
254 return READ_ONCE(pernet->local_addr_max);
255 }
256 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
257
check_work_pending(struct mptcp_sock * msk)258 static void check_work_pending(struct mptcp_sock *msk)
259 {
260 if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
261 (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
262 msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
263 WRITE_ONCE(msk->pm.work_pending, false);
264 }
265
266 struct mptcp_pm_add_entry *
mptcp_lookup_anno_list_by_saddr(struct mptcp_sock * msk,struct mptcp_addr_info * addr)267 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
268 struct mptcp_addr_info *addr)
269 {
270 struct mptcp_pm_add_entry *entry;
271
272 lockdep_assert_held(&msk->pm.lock);
273
274 list_for_each_entry(entry, &msk->pm.anno_list, list) {
275 if (addresses_equal(&entry->addr, addr, true))
276 return entry;
277 }
278
279 return NULL;
280 }
281
mptcp_pm_sport_in_anno_list(struct mptcp_sock * msk,const struct sock * sk)282 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
283 {
284 struct mptcp_pm_add_entry *entry;
285 struct mptcp_addr_info saddr;
286 bool ret = false;
287
288 local_address((struct sock_common *)sk, &saddr);
289
290 spin_lock_bh(&msk->pm.lock);
291 list_for_each_entry(entry, &msk->pm.anno_list, list) {
292 if (addresses_equal(&entry->addr, &saddr, true)) {
293 ret = true;
294 goto out;
295 }
296 }
297
298 out:
299 spin_unlock_bh(&msk->pm.lock);
300 return ret;
301 }
302
mptcp_pm_add_timer(struct timer_list * timer)303 static void mptcp_pm_add_timer(struct timer_list *timer)
304 {
305 struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
306 struct mptcp_sock *msk = entry->sock;
307 struct sock *sk = (struct sock *)msk;
308
309 pr_debug("msk=%p", msk);
310
311 if (!msk)
312 return;
313
314 if (inet_sk_state_load(sk) == TCP_CLOSE)
315 return;
316
317 if (!entry->addr.id)
318 return;
319
320 if (mptcp_pm_should_add_signal(msk)) {
321 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
322 goto out;
323 }
324
325 spin_lock_bh(&msk->pm.lock);
326
327 if (!mptcp_pm_should_add_signal(msk)) {
328 pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
329 mptcp_pm_announce_addr(msk, &entry->addr, false);
330 mptcp_pm_add_addr_send_ack(msk);
331 entry->retrans_times++;
332 }
333
334 if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
335 sk_reset_timer(sk, timer,
336 jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
337
338 spin_unlock_bh(&msk->pm.lock);
339
340 if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
341 mptcp_pm_subflow_established(msk);
342
343 out:
344 __sock_put(sk);
345 }
346
347 struct mptcp_pm_add_entry *
mptcp_pm_del_add_timer(struct mptcp_sock * msk,struct mptcp_addr_info * addr)348 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
349 struct mptcp_addr_info *addr)
350 {
351 struct mptcp_pm_add_entry *entry;
352 struct sock *sk = (struct sock *)msk;
353
354 spin_lock_bh(&msk->pm.lock);
355 entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
356 if (entry)
357 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
358 spin_unlock_bh(&msk->pm.lock);
359
360 if (entry)
361 sk_stop_timer_sync(sk, &entry->add_timer);
362
363 return entry;
364 }
365
mptcp_pm_alloc_anno_list(struct mptcp_sock * msk,struct mptcp_pm_addr_entry * entry)366 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
367 struct mptcp_pm_addr_entry *entry)
368 {
369 struct mptcp_pm_add_entry *add_entry = NULL;
370 struct sock *sk = (struct sock *)msk;
371 struct net *net = sock_net(sk);
372
373 lockdep_assert_held(&msk->pm.lock);
374
375 if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
376 return false;
377
378 add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
379 if (!add_entry)
380 return false;
381
382 list_add(&add_entry->list, &msk->pm.anno_list);
383
384 add_entry->addr = entry->addr;
385 add_entry->sock = msk;
386 add_entry->retrans_times = 0;
387
388 timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
389 sk_reset_timer(sk, &add_entry->add_timer,
390 jiffies + mptcp_get_add_addr_timeout(net));
391
392 return true;
393 }
394
mptcp_pm_free_anno_list(struct mptcp_sock * msk)395 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
396 {
397 struct mptcp_pm_add_entry *entry, *tmp;
398 struct sock *sk = (struct sock *)msk;
399 LIST_HEAD(free_list);
400
401 pr_debug("msk=%p", msk);
402
403 spin_lock_bh(&msk->pm.lock);
404 list_splice_init(&msk->pm.anno_list, &free_list);
405 spin_unlock_bh(&msk->pm.lock);
406
407 list_for_each_entry_safe(entry, tmp, &free_list, list) {
408 sk_stop_timer_sync(sk, &entry->add_timer);
409 kfree(entry);
410 }
411 }
412
mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock * msk)413 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
414 {
415 struct sock *sk = (struct sock *)msk;
416 struct mptcp_pm_addr_entry *local;
417 unsigned int add_addr_signal_max;
418 unsigned int local_addr_max;
419 struct pm_nl_pernet *pernet;
420 unsigned int subflows_max;
421
422 pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
423
424 add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
425 local_addr_max = mptcp_pm_get_local_addr_max(msk);
426 subflows_max = mptcp_pm_get_subflows_max(msk);
427
428 pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
429 msk->pm.local_addr_used, local_addr_max,
430 msk->pm.add_addr_signaled, add_addr_signal_max,
431 msk->pm.subflows, subflows_max);
432
433 /* check first for announce */
434 if (msk->pm.add_addr_signaled < add_addr_signal_max) {
435 local = select_signal_address(pernet,
436 msk->pm.add_addr_signaled);
437
438 if (local) {
439 if (mptcp_pm_alloc_anno_list(msk, local)) {
440 msk->pm.add_addr_signaled++;
441 mptcp_pm_announce_addr(msk, &local->addr, false);
442 mptcp_pm_nl_addr_send_ack(msk);
443 }
444 } else {
445 /* pick failed, avoid fourther attempts later */
446 msk->pm.local_addr_used = add_addr_signal_max;
447 }
448
449 check_work_pending(msk);
450 }
451
452 /* check if should create a new subflow */
453 if (msk->pm.local_addr_used < local_addr_max &&
454 msk->pm.subflows < subflows_max) {
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
mptcp_pm_nl_fully_established(struct mptcp_sock * msk)476 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
477 {
478 mptcp_pm_create_subflow_or_signal_addr(msk);
479 }
480
mptcp_pm_nl_subflow_established(struct mptcp_sock * msk)481 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
482 {
483 mptcp_pm_create_subflow_or_signal_addr(msk);
484 }
485
mptcp_pm_nl_add_addr_received(struct mptcp_sock * msk)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
mptcp_pm_nl_addr_send_ack(struct mptcp_sock * msk)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
544 spin_unlock_bh(&msk->pm.lock);
545 pr_debug("send ack for %s%s%s",
546 mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
547 mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
548 mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
549
550 lock_sock(ssk);
551 tcp_send_ack(ssk);
552 release_sock(ssk);
553 spin_lock_bh(&msk->pm.lock);
554 }
555 }
556
mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock * msk,struct mptcp_addr_info * addr,u8 bkup)557 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
558 struct mptcp_addr_info *addr,
559 u8 bkup)
560 {
561 struct mptcp_subflow_context *subflow;
562
563 pr_debug("bkup=%d", bkup);
564
565 mptcp_for_each_subflow(msk, subflow) {
566 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
567 struct sock *sk = (struct sock *)msk;
568 struct mptcp_addr_info local;
569
570 local_address((struct sock_common *)ssk, &local);
571 if (!addresses_equal(&local, addr, addr->port))
572 continue;
573
574 subflow->backup = bkup;
575 subflow->send_mp_prio = 1;
576 subflow->request_bkup = bkup;
577 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
578
579 spin_unlock_bh(&msk->pm.lock);
580 pr_debug("send ack for mp_prio");
581 lock_sock(ssk);
582 tcp_send_ack(ssk);
583 release_sock(ssk);
584 spin_lock_bh(&msk->pm.lock);
585
586 return 0;
587 }
588
589 return -EINVAL;
590 }
591
mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock * msk,const struct mptcp_rm_list * rm_list,enum linux_mptcp_mib_field rm_type)592 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
593 const struct mptcp_rm_list *rm_list,
594 enum linux_mptcp_mib_field rm_type)
595 {
596 struct mptcp_subflow_context *subflow, *tmp;
597 struct sock *sk = (struct sock *)msk;
598 u8 i;
599
600 pr_debug("%s rm_list_nr %d",
601 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
602
603 msk_owned_by_me(msk);
604
605 if (!rm_list->nr)
606 return;
607
608 if (list_empty(&msk->conn_list))
609 return;
610
611 for (i = 0; i < rm_list->nr; i++) {
612 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
613 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
614 int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
615 u8 id = subflow->local_id;
616
617 if (rm_type == MPTCP_MIB_RMADDR)
618 id = subflow->remote_id;
619
620 if (rm_list->ids[i] != id)
621 continue;
622
623 pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
624 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
625 i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
626 spin_unlock_bh(&msk->pm.lock);
627 mptcp_subflow_shutdown(sk, ssk, how);
628 mptcp_close_ssk(sk, ssk, subflow);
629 spin_lock_bh(&msk->pm.lock);
630
631 if (rm_type == MPTCP_MIB_RMADDR) {
632 msk->pm.add_addr_accepted--;
633 WRITE_ONCE(msk->pm.accept_addr, true);
634 } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
635 msk->pm.local_addr_used--;
636 }
637 msk->pm.subflows--;
638 __MPTCP_INC_STATS(sock_net(sk), rm_type);
639 }
640 }
641 }
642
mptcp_pm_nl_rm_addr_received(struct mptcp_sock * msk)643 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
644 {
645 mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
646 }
647
mptcp_pm_nl_rm_subflow_received(struct mptcp_sock * msk,const struct mptcp_rm_list * rm_list)648 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
649 const struct mptcp_rm_list *rm_list)
650 {
651 mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
652 }
653
mptcp_pm_nl_work(struct mptcp_sock * msk)654 void mptcp_pm_nl_work(struct mptcp_sock *msk)
655 {
656 struct mptcp_pm_data *pm = &msk->pm;
657
658 msk_owned_by_me(msk);
659
660 spin_lock_bh(&msk->pm.lock);
661
662 pr_debug("msk=%p status=%x", msk, pm->status);
663 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
664 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
665 mptcp_pm_nl_add_addr_received(msk);
666 }
667 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
668 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
669 mptcp_pm_nl_addr_send_ack(msk);
670 }
671 if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
672 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
673 mptcp_pm_nl_rm_addr_received(msk);
674 }
675 if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
676 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
677 mptcp_pm_nl_fully_established(msk);
678 }
679 if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
680 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
681 mptcp_pm_nl_subflow_established(msk);
682 }
683
684 spin_unlock_bh(&msk->pm.lock);
685 }
686
address_use_port(struct mptcp_pm_addr_entry * entry)687 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
688 {
689 return (entry->flags &
690 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
691 MPTCP_PM_ADDR_FLAG_SIGNAL;
692 }
693
mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet * pernet,struct mptcp_pm_addr_entry * entry)694 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
695 struct mptcp_pm_addr_entry *entry)
696 {
697 struct mptcp_pm_addr_entry *cur;
698 unsigned int addr_max;
699 int ret = -EINVAL;
700
701 spin_lock_bh(&pernet->lock);
702 /* to keep the code simple, don't do IDR-like allocation for address ID,
703 * just bail when we exceed limits
704 */
705 if (pernet->next_id == MAX_ADDR_ID)
706 pernet->next_id = 1;
707 if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
708 goto out;
709 if (test_bit(entry->addr.id, pernet->id_bitmap))
710 goto out;
711
712 /* do not insert duplicate address, differentiate on port only
713 * singled addresses
714 */
715 list_for_each_entry(cur, &pernet->local_addr_list, list) {
716 if (addresses_equal(&cur->addr, &entry->addr,
717 address_use_port(entry) &&
718 address_use_port(cur)))
719 goto out;
720 }
721
722 if (!entry->addr.id) {
723 find_next:
724 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
725 MAX_ADDR_ID + 1,
726 pernet->next_id);
727 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
728 pernet->next_id != 1) {
729 pernet->next_id = 1;
730 goto find_next;
731 }
732 }
733
734 if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
735 goto out;
736
737 __set_bit(entry->addr.id, pernet->id_bitmap);
738 if (entry->addr.id > pernet->next_id)
739 pernet->next_id = entry->addr.id;
740
741 if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
742 addr_max = pernet->add_addr_signal_max;
743 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
744 }
745 if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
746 addr_max = pernet->local_addr_max;
747 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
748 }
749
750 pernet->addrs++;
751 list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
752 ret = entry->addr.id;
753
754 out:
755 spin_unlock_bh(&pernet->lock);
756 return ret;
757 }
758
mptcp_pm_nl_create_listen_socket(struct sock * sk,struct mptcp_pm_addr_entry * entry)759 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
760 struct mptcp_pm_addr_entry *entry)
761 {
762 struct sockaddr_storage addr;
763 struct mptcp_sock *msk;
764 struct socket *ssock;
765 int backlog = 1024;
766 int err;
767
768 err = sock_create_kern(sock_net(sk), entry->addr.family,
769 SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
770 if (err)
771 return err;
772
773 msk = mptcp_sk(entry->lsk->sk);
774 if (!msk) {
775 err = -EINVAL;
776 goto out;
777 }
778
779 ssock = __mptcp_nmpc_socket(msk);
780 if (!ssock) {
781 err = -EINVAL;
782 goto out;
783 }
784
785 mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
786 err = kernel_bind(ssock, (struct sockaddr *)&addr,
787 sizeof(struct sockaddr_in));
788 if (err) {
789 pr_warn("kernel_bind error, err=%d", err);
790 goto out;
791 }
792
793 err = kernel_listen(ssock, backlog);
794 if (err) {
795 pr_warn("kernel_listen error, err=%d", err);
796 goto out;
797 }
798
799 return 0;
800
801 out:
802 sock_release(entry->lsk);
803 return err;
804 }
805
mptcp_pm_nl_get_local_id(struct mptcp_sock * msk,struct sock_common * skc)806 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
807 {
808 struct mptcp_pm_addr_entry *entry;
809 struct mptcp_addr_info skc_local;
810 struct mptcp_addr_info msk_local;
811 struct pm_nl_pernet *pernet;
812 int ret = -1;
813
814 if (WARN_ON_ONCE(!msk))
815 return -1;
816
817 /* The 0 ID mapping is defined by the first subflow, copied into the msk
818 * addr
819 */
820 local_address((struct sock_common *)msk, &msk_local);
821 local_address((struct sock_common *)skc, &skc_local);
822 if (addresses_equal(&msk_local, &skc_local, false))
823 return 0;
824
825 if (address_zero(&skc_local))
826 return 0;
827
828 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
829
830 rcu_read_lock();
831 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
832 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
833 ret = entry->addr.id;
834 break;
835 }
836 }
837 rcu_read_unlock();
838 if (ret >= 0)
839 return ret;
840
841 /* address not found, add to local list */
842 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
843 if (!entry)
844 return -ENOMEM;
845
846 entry->addr = skc_local;
847 entry->addr.id = 0;
848 entry->addr.port = 0;
849 entry->ifindex = 0;
850 entry->flags = 0;
851 entry->lsk = NULL;
852 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
853 if (ret < 0)
854 kfree(entry);
855
856 return ret;
857 }
858
mptcp_pm_nl_data_init(struct mptcp_sock * msk)859 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
860 {
861 struct mptcp_pm_data *pm = &msk->pm;
862 bool subflows;
863
864 subflows = !!mptcp_pm_get_subflows_max(msk);
865 WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
866 !!mptcp_pm_get_add_addr_signal_max(msk));
867 WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
868 WRITE_ONCE(pm->accept_subflow, subflows);
869 }
870
871 #define MPTCP_PM_CMD_GRP_OFFSET 0
872 #define MPTCP_PM_EV_GRP_OFFSET 1
873
874 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
875 [MPTCP_PM_CMD_GRP_OFFSET] = { .name = MPTCP_PM_CMD_GRP_NAME, },
876 [MPTCP_PM_EV_GRP_OFFSET] = { .name = MPTCP_PM_EV_GRP_NAME,
877 .flags = GENL_UNS_ADMIN_PERM,
878 },
879 };
880
881 static const struct nla_policy
882 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
883 [MPTCP_PM_ADDR_ATTR_FAMILY] = { .type = NLA_U16, },
884 [MPTCP_PM_ADDR_ATTR_ID] = { .type = NLA_U8, },
885 [MPTCP_PM_ADDR_ATTR_ADDR4] = { .type = NLA_U32, },
886 [MPTCP_PM_ADDR_ATTR_ADDR6] =
887 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
888 [MPTCP_PM_ADDR_ATTR_PORT] = { .type = NLA_U16 },
889 [MPTCP_PM_ADDR_ATTR_FLAGS] = { .type = NLA_U32 },
890 [MPTCP_PM_ADDR_ATTR_IF_IDX] = { .type = NLA_S32 },
891 };
892
893 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
894 [MPTCP_PM_ATTR_ADDR] =
895 NLA_POLICY_NESTED(mptcp_pm_addr_policy),
896 [MPTCP_PM_ATTR_RCV_ADD_ADDRS] = { .type = NLA_U32, },
897 [MPTCP_PM_ATTR_SUBFLOWS] = { .type = NLA_U32, },
898 };
899
mptcp_pm_family_to_addr(int family)900 static int mptcp_pm_family_to_addr(int family)
901 {
902 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
903 if (family == AF_INET6)
904 return MPTCP_PM_ADDR_ATTR_ADDR6;
905 #endif
906 return MPTCP_PM_ADDR_ATTR_ADDR4;
907 }
908
mptcp_pm_parse_addr(struct nlattr * attr,struct genl_info * info,bool require_family,struct mptcp_pm_addr_entry * entry)909 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
910 bool require_family,
911 struct mptcp_pm_addr_entry *entry)
912 {
913 struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
914 int err, addr_addr;
915
916 if (!attr) {
917 GENL_SET_ERR_MSG(info, "missing address info");
918 return -EINVAL;
919 }
920
921 /* no validation needed - was already done via nested policy */
922 err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
923 mptcp_pm_addr_policy, info->extack);
924 if (err)
925 return err;
926
927 memset(entry, 0, sizeof(*entry));
928 if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
929 if (!require_family)
930 goto skip_family;
931
932 NL_SET_ERR_MSG_ATTR(info->extack, attr,
933 "missing family");
934 return -EINVAL;
935 }
936
937 entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
938 if (entry->addr.family != AF_INET
939 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
940 && entry->addr.family != AF_INET6
941 #endif
942 ) {
943 NL_SET_ERR_MSG_ATTR(info->extack, attr,
944 "unknown address family");
945 return -EINVAL;
946 }
947 addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
948 if (!tb[addr_addr]) {
949 NL_SET_ERR_MSG_ATTR(info->extack, attr,
950 "missing address data");
951 return -EINVAL;
952 }
953
954 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
955 if (entry->addr.family == AF_INET6)
956 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
957 else
958 #endif
959 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
960
961 skip_family:
962 if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
963 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
964
965 entry->ifindex = val;
966 }
967
968 if (tb[MPTCP_PM_ADDR_ATTR_ID])
969 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
970
971 if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
972 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
973
974 if (tb[MPTCP_PM_ADDR_ATTR_PORT])
975 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
976
977 return 0;
978 }
979
genl_info_pm_nl(struct genl_info * info)980 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
981 {
982 return net_generic(genl_info_net(info), pm_nl_pernet_id);
983 }
984
mptcp_nl_add_subflow_or_signal_addr(struct net * net)985 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
986 {
987 struct mptcp_sock *msk;
988 long s_slot = 0, s_num = 0;
989
990 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
991 struct sock *sk = (struct sock *)msk;
992
993 if (!READ_ONCE(msk->fully_established))
994 goto next;
995
996 lock_sock(sk);
997 spin_lock_bh(&msk->pm.lock);
998 mptcp_pm_create_subflow_or_signal_addr(msk);
999 spin_unlock_bh(&msk->pm.lock);
1000 release_sock(sk);
1001
1002 next:
1003 sock_put(sk);
1004 cond_resched();
1005 }
1006
1007 return 0;
1008 }
1009
mptcp_nl_cmd_add_addr(struct sk_buff * skb,struct genl_info * info)1010 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1011 {
1012 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1013 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1014 struct mptcp_pm_addr_entry addr, *entry;
1015 int ret;
1016
1017 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1018 if (ret < 0)
1019 return ret;
1020
1021 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1022 if (!entry) {
1023 GENL_SET_ERR_MSG(info, "can't allocate addr");
1024 return -ENOMEM;
1025 }
1026
1027 *entry = addr;
1028 if (entry->addr.port) {
1029 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1030 if (ret) {
1031 GENL_SET_ERR_MSG(info, "create listen socket error");
1032 kfree(entry);
1033 return ret;
1034 }
1035 }
1036 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1037 if (ret < 0) {
1038 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1039 if (entry->lsk)
1040 sock_release(entry->lsk);
1041 kfree(entry);
1042 return ret;
1043 }
1044
1045 mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1046
1047 return 0;
1048 }
1049
1050 static struct mptcp_pm_addr_entry *
__lookup_addr_by_id(struct pm_nl_pernet * pernet,unsigned int id)1051 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1052 {
1053 struct mptcp_pm_addr_entry *entry;
1054
1055 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1056 if (entry->addr.id == id)
1057 return entry;
1058 }
1059 return NULL;
1060 }
1061
remove_anno_list_by_saddr(struct mptcp_sock * msk,struct mptcp_addr_info * addr)1062 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1063 struct mptcp_addr_info *addr)
1064 {
1065 struct mptcp_pm_add_entry *entry;
1066
1067 entry = mptcp_pm_del_add_timer(msk, addr);
1068 if (entry) {
1069 list_del(&entry->list);
1070 kfree(entry);
1071 return true;
1072 }
1073
1074 return false;
1075 }
1076
mptcp_pm_remove_anno_addr(struct mptcp_sock * msk,struct mptcp_addr_info * addr,bool force)1077 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1078 struct mptcp_addr_info *addr,
1079 bool force)
1080 {
1081 struct mptcp_rm_list list = { .nr = 0 };
1082 bool ret;
1083
1084 list.ids[list.nr++] = addr->id;
1085
1086 ret = remove_anno_list_by_saddr(msk, addr);
1087 if (ret || force) {
1088 spin_lock_bh(&msk->pm.lock);
1089 mptcp_pm_remove_addr(msk, &list);
1090 spin_unlock_bh(&msk->pm.lock);
1091 }
1092 return ret;
1093 }
1094
mptcp_nl_remove_subflow_and_signal_addr(struct net * net,struct mptcp_addr_info * addr)1095 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1096 struct mptcp_addr_info *addr)
1097 {
1098 struct mptcp_sock *msk;
1099 long s_slot = 0, s_num = 0;
1100 struct mptcp_rm_list list = { .nr = 0 };
1101
1102 pr_debug("remove_id=%d", addr->id);
1103
1104 list.ids[list.nr++] = addr->id;
1105
1106 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1107 struct sock *sk = (struct sock *)msk;
1108 bool remove_subflow;
1109
1110 if (list_empty(&msk->conn_list)) {
1111 mptcp_pm_remove_anno_addr(msk, addr, false);
1112 goto next;
1113 }
1114
1115 lock_sock(sk);
1116 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1117 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1118 if (remove_subflow)
1119 mptcp_pm_remove_subflow(msk, &list);
1120 release_sock(sk);
1121
1122 next:
1123 sock_put(sk);
1124 cond_resched();
1125 }
1126
1127 return 0;
1128 }
1129
1130 struct addr_entry_release_work {
1131 struct rcu_work rwork;
1132 struct mptcp_pm_addr_entry *entry;
1133 };
1134
mptcp_pm_release_addr_entry(struct work_struct * work)1135 static void mptcp_pm_release_addr_entry(struct work_struct *work)
1136 {
1137 struct addr_entry_release_work *w;
1138 struct mptcp_pm_addr_entry *entry;
1139
1140 w = container_of(to_rcu_work(work), struct addr_entry_release_work, rwork);
1141 entry = w->entry;
1142 if (entry) {
1143 if (entry->lsk)
1144 sock_release(entry->lsk);
1145 kfree(entry);
1146 }
1147 kfree(w);
1148 }
1149
mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry * entry)1150 static void mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry *entry)
1151 {
1152 struct addr_entry_release_work *w;
1153
1154 w = kmalloc(sizeof(*w), GFP_ATOMIC);
1155 if (w) {
1156 INIT_RCU_WORK(&w->rwork, mptcp_pm_release_addr_entry);
1157 w->entry = entry;
1158 queue_rcu_work(system_wq, &w->rwork);
1159 }
1160 }
1161
mptcp_nl_remove_id_zero_address(struct net * net,struct mptcp_addr_info * addr)1162 static int mptcp_nl_remove_id_zero_address(struct net *net,
1163 struct mptcp_addr_info *addr)
1164 {
1165 struct mptcp_rm_list list = { .nr = 0 };
1166 long s_slot = 0, s_num = 0;
1167 struct mptcp_sock *msk;
1168
1169 list.ids[list.nr++] = 0;
1170
1171 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1172 struct sock *sk = (struct sock *)msk;
1173 struct mptcp_addr_info msk_local;
1174
1175 if (list_empty(&msk->conn_list))
1176 goto next;
1177
1178 local_address((struct sock_common *)msk, &msk_local);
1179 if (!addresses_equal(&msk_local, addr, addr->port))
1180 goto next;
1181
1182 lock_sock(sk);
1183 spin_lock_bh(&msk->pm.lock);
1184 mptcp_pm_remove_addr(msk, &list);
1185 mptcp_pm_nl_rm_subflow_received(msk, &list);
1186 spin_unlock_bh(&msk->pm.lock);
1187 release_sock(sk);
1188
1189 next:
1190 sock_put(sk);
1191 cond_resched();
1192 }
1193
1194 return 0;
1195 }
1196
mptcp_nl_cmd_del_addr(struct sk_buff * skb,struct genl_info * info)1197 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1198 {
1199 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1200 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1201 struct mptcp_pm_addr_entry addr, *entry;
1202 unsigned int addr_max;
1203 int ret;
1204
1205 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1206 if (ret < 0)
1207 return ret;
1208
1209 /* the zero id address is special: the first address used by the msk
1210 * always gets such an id, so different subflows can have different zero
1211 * id addresses. Additionally zero id is not accounted for in id_bitmap.
1212 * Let's use an 'mptcp_rm_list' instead of the common remove code.
1213 */
1214 if (addr.addr.id == 0)
1215 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1216
1217 spin_lock_bh(&pernet->lock);
1218 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1219 if (!entry) {
1220 GENL_SET_ERR_MSG(info, "address not found");
1221 spin_unlock_bh(&pernet->lock);
1222 return -EINVAL;
1223 }
1224 if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1225 addr_max = pernet->add_addr_signal_max;
1226 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1227 }
1228 if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1229 addr_max = pernet->local_addr_max;
1230 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1231 }
1232
1233 pernet->addrs--;
1234 list_del_rcu(&entry->list);
1235 __clear_bit(entry->addr.id, pernet->id_bitmap);
1236 spin_unlock_bh(&pernet->lock);
1237
1238 mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1239 mptcp_pm_free_addr_entry(entry);
1240
1241 return ret;
1242 }
1243
mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock * msk,struct list_head * rm_list)1244 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1245 struct list_head *rm_list)
1246 {
1247 struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1248 struct mptcp_pm_addr_entry *entry;
1249
1250 list_for_each_entry(entry, rm_list, list) {
1251 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1252 alist.nr < MPTCP_RM_IDS_MAX &&
1253 slist.nr < MPTCP_RM_IDS_MAX) {
1254 alist.ids[alist.nr++] = entry->addr.id;
1255 slist.ids[slist.nr++] = entry->addr.id;
1256 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1257 alist.nr < MPTCP_RM_IDS_MAX) {
1258 alist.ids[alist.nr++] = entry->addr.id;
1259 }
1260 }
1261
1262 if (alist.nr) {
1263 spin_lock_bh(&msk->pm.lock);
1264 mptcp_pm_remove_addr(msk, &alist);
1265 spin_unlock_bh(&msk->pm.lock);
1266 }
1267 if (slist.nr)
1268 mptcp_pm_remove_subflow(msk, &slist);
1269 }
1270
mptcp_nl_remove_addrs_list(struct net * net,struct list_head * rm_list)1271 static void mptcp_nl_remove_addrs_list(struct net *net,
1272 struct list_head *rm_list)
1273 {
1274 long s_slot = 0, s_num = 0;
1275 struct mptcp_sock *msk;
1276
1277 if (list_empty(rm_list))
1278 return;
1279
1280 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1281 struct sock *sk = (struct sock *)msk;
1282
1283 lock_sock(sk);
1284 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1285 release_sock(sk);
1286
1287 sock_put(sk);
1288 cond_resched();
1289 }
1290 }
1291
__flush_addrs(struct list_head * list)1292 static void __flush_addrs(struct list_head *list)
1293 {
1294 while (!list_empty(list)) {
1295 struct mptcp_pm_addr_entry *cur;
1296
1297 cur = list_entry(list->next,
1298 struct mptcp_pm_addr_entry, list);
1299 list_del_rcu(&cur->list);
1300 mptcp_pm_free_addr_entry(cur);
1301 }
1302 }
1303
__reset_counters(struct pm_nl_pernet * pernet)1304 static void __reset_counters(struct pm_nl_pernet *pernet)
1305 {
1306 WRITE_ONCE(pernet->add_addr_signal_max, 0);
1307 WRITE_ONCE(pernet->add_addr_accept_max, 0);
1308 WRITE_ONCE(pernet->local_addr_max, 0);
1309 pernet->addrs = 0;
1310 }
1311
mptcp_nl_cmd_flush_addrs(struct sk_buff * skb,struct genl_info * info)1312 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1313 {
1314 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1315 LIST_HEAD(free_list);
1316
1317 spin_lock_bh(&pernet->lock);
1318 list_splice_init(&pernet->local_addr_list, &free_list);
1319 __reset_counters(pernet);
1320 pernet->next_id = 1;
1321 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1322 spin_unlock_bh(&pernet->lock);
1323 mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1324 __flush_addrs(&free_list);
1325 return 0;
1326 }
1327
mptcp_nl_fill_addr(struct sk_buff * skb,struct mptcp_pm_addr_entry * entry)1328 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1329 struct mptcp_pm_addr_entry *entry)
1330 {
1331 struct mptcp_addr_info *addr = &entry->addr;
1332 struct nlattr *attr;
1333
1334 attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1335 if (!attr)
1336 return -EMSGSIZE;
1337
1338 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1339 goto nla_put_failure;
1340 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1341 goto nla_put_failure;
1342 if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1343 goto nla_put_failure;
1344 if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1345 goto nla_put_failure;
1346 if (entry->ifindex &&
1347 nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1348 goto nla_put_failure;
1349
1350 if (addr->family == AF_INET &&
1351 nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1352 addr->addr.s_addr))
1353 goto nla_put_failure;
1354 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1355 else if (addr->family == AF_INET6 &&
1356 nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1357 goto nla_put_failure;
1358 #endif
1359 nla_nest_end(skb, attr);
1360 return 0;
1361
1362 nla_put_failure:
1363 nla_nest_cancel(skb, attr);
1364 return -EMSGSIZE;
1365 }
1366
mptcp_nl_cmd_get_addr(struct sk_buff * skb,struct genl_info * info)1367 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1368 {
1369 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1370 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1371 struct mptcp_pm_addr_entry addr, *entry;
1372 struct sk_buff *msg;
1373 void *reply;
1374 int ret;
1375
1376 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1377 if (ret < 0)
1378 return ret;
1379
1380 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1381 if (!msg)
1382 return -ENOMEM;
1383
1384 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1385 info->genlhdr->cmd);
1386 if (!reply) {
1387 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1388 ret = -EMSGSIZE;
1389 goto fail;
1390 }
1391
1392 spin_lock_bh(&pernet->lock);
1393 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1394 if (!entry) {
1395 GENL_SET_ERR_MSG(info, "address not found");
1396 ret = -EINVAL;
1397 goto unlock_fail;
1398 }
1399
1400 ret = mptcp_nl_fill_addr(msg, entry);
1401 if (ret)
1402 goto unlock_fail;
1403
1404 genlmsg_end(msg, reply);
1405 ret = genlmsg_reply(msg, info);
1406 spin_unlock_bh(&pernet->lock);
1407 return ret;
1408
1409 unlock_fail:
1410 spin_unlock_bh(&pernet->lock);
1411
1412 fail:
1413 nlmsg_free(msg);
1414 return ret;
1415 }
1416
mptcp_nl_cmd_dump_addrs(struct sk_buff * msg,struct netlink_callback * cb)1417 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1418 struct netlink_callback *cb)
1419 {
1420 struct net *net = sock_net(msg->sk);
1421 struct mptcp_pm_addr_entry *entry;
1422 struct pm_nl_pernet *pernet;
1423 int id = cb->args[0];
1424 void *hdr;
1425 int i;
1426
1427 pernet = net_generic(net, pm_nl_pernet_id);
1428
1429 spin_lock_bh(&pernet->lock);
1430 for (i = id; i < MAX_ADDR_ID + 1; i++) {
1431 if (test_bit(i, pernet->id_bitmap)) {
1432 entry = __lookup_addr_by_id(pernet, i);
1433 if (!entry)
1434 break;
1435
1436 if (entry->addr.id <= id)
1437 continue;
1438
1439 hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1440 cb->nlh->nlmsg_seq, &mptcp_genl_family,
1441 NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1442 if (!hdr)
1443 break;
1444
1445 if (mptcp_nl_fill_addr(msg, entry) < 0) {
1446 genlmsg_cancel(msg, hdr);
1447 break;
1448 }
1449
1450 id = entry->addr.id;
1451 genlmsg_end(msg, hdr);
1452 }
1453 }
1454 spin_unlock_bh(&pernet->lock);
1455
1456 cb->args[0] = id;
1457 return msg->len;
1458 }
1459
parse_limit(struct genl_info * info,int id,unsigned int * limit)1460 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1461 {
1462 struct nlattr *attr = info->attrs[id];
1463
1464 if (!attr)
1465 return 0;
1466
1467 *limit = nla_get_u32(attr);
1468 if (*limit > MPTCP_PM_ADDR_MAX) {
1469 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1470 return -EINVAL;
1471 }
1472 return 0;
1473 }
1474
1475 static int
mptcp_nl_cmd_set_limits(struct sk_buff * skb,struct genl_info * info)1476 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1477 {
1478 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1479 unsigned int rcv_addrs, subflows;
1480 int ret;
1481
1482 spin_lock_bh(&pernet->lock);
1483 rcv_addrs = pernet->add_addr_accept_max;
1484 ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1485 if (ret)
1486 goto unlock;
1487
1488 subflows = pernet->subflows_max;
1489 ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1490 if (ret)
1491 goto unlock;
1492
1493 WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1494 WRITE_ONCE(pernet->subflows_max, subflows);
1495
1496 unlock:
1497 spin_unlock_bh(&pernet->lock);
1498 return ret;
1499 }
1500
1501 static int
mptcp_nl_cmd_get_limits(struct sk_buff * skb,struct genl_info * info)1502 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1503 {
1504 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1505 struct sk_buff *msg;
1506 void *reply;
1507
1508 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1509 if (!msg)
1510 return -ENOMEM;
1511
1512 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1513 MPTCP_PM_CMD_GET_LIMITS);
1514 if (!reply)
1515 goto fail;
1516
1517 if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1518 READ_ONCE(pernet->add_addr_accept_max)))
1519 goto fail;
1520
1521 if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1522 READ_ONCE(pernet->subflows_max)))
1523 goto fail;
1524
1525 genlmsg_end(msg, reply);
1526 return genlmsg_reply(msg, info);
1527
1528 fail:
1529 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1530 nlmsg_free(msg);
1531 return -EMSGSIZE;
1532 }
1533
mptcp_nl_addr_backup(struct net * net,struct mptcp_addr_info * addr,u8 bkup)1534 static int mptcp_nl_addr_backup(struct net *net,
1535 struct mptcp_addr_info *addr,
1536 u8 bkup)
1537 {
1538 long s_slot = 0, s_num = 0;
1539 struct mptcp_sock *msk;
1540 int ret = -EINVAL;
1541
1542 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1543 struct sock *sk = (struct sock *)msk;
1544
1545 if (list_empty(&msk->conn_list))
1546 goto next;
1547
1548 lock_sock(sk);
1549 spin_lock_bh(&msk->pm.lock);
1550 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1551 spin_unlock_bh(&msk->pm.lock);
1552 release_sock(sk);
1553
1554 next:
1555 sock_put(sk);
1556 cond_resched();
1557 }
1558
1559 return ret;
1560 }
1561
mptcp_nl_cmd_set_flags(struct sk_buff * skb,struct genl_info * info)1562 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1563 {
1564 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1565 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1566 struct mptcp_pm_addr_entry addr, *entry;
1567 struct net *net = sock_net(skb->sk);
1568 u8 bkup = 0;
1569 int ret;
1570
1571 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1572 if (ret < 0)
1573 return ret;
1574
1575 if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1576 bkup = 1;
1577
1578 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1579 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1580 ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1581 if (ret)
1582 return ret;
1583
1584 if (bkup)
1585 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1586 else
1587 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1588 }
1589 }
1590
1591 return 0;
1592 }
1593
mptcp_nl_mcast_send(struct net * net,struct sk_buff * nlskb,gfp_t gfp)1594 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1595 {
1596 genlmsg_multicast_netns(&mptcp_genl_family, net,
1597 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1598 }
1599
mptcp_event_add_subflow(struct sk_buff * skb,const struct sock * ssk)1600 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1601 {
1602 const struct inet_sock *issk = inet_sk(ssk);
1603 const struct mptcp_subflow_context *sf;
1604
1605 if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1606 return -EMSGSIZE;
1607
1608 switch (ssk->sk_family) {
1609 case AF_INET:
1610 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1611 return -EMSGSIZE;
1612 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1613 return -EMSGSIZE;
1614 break;
1615 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1616 case AF_INET6: {
1617 const struct ipv6_pinfo *np = inet6_sk(ssk);
1618
1619 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1620 return -EMSGSIZE;
1621 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1622 return -EMSGSIZE;
1623 break;
1624 }
1625 #endif
1626 default:
1627 WARN_ON_ONCE(1);
1628 return -EMSGSIZE;
1629 }
1630
1631 if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1632 return -EMSGSIZE;
1633 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1634 return -EMSGSIZE;
1635
1636 sf = mptcp_subflow_ctx(ssk);
1637 if (WARN_ON_ONCE(!sf))
1638 return -EINVAL;
1639
1640 if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1641 return -EMSGSIZE;
1642
1643 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1644 return -EMSGSIZE;
1645
1646 return 0;
1647 }
1648
mptcp_event_put_token_and_ssk(struct sk_buff * skb,const struct mptcp_sock * msk,const struct sock * ssk)1649 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1650 const struct mptcp_sock *msk,
1651 const struct sock *ssk)
1652 {
1653 const struct sock *sk = (const struct sock *)msk;
1654 const struct mptcp_subflow_context *sf;
1655 u8 sk_err;
1656
1657 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1658 return -EMSGSIZE;
1659
1660 if (mptcp_event_add_subflow(skb, ssk))
1661 return -EMSGSIZE;
1662
1663 sf = mptcp_subflow_ctx(ssk);
1664 if (WARN_ON_ONCE(!sf))
1665 return -EINVAL;
1666
1667 if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1668 return -EMSGSIZE;
1669
1670 if (ssk->sk_bound_dev_if &&
1671 nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1672 return -EMSGSIZE;
1673
1674 sk_err = ssk->sk_err;
1675 if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1676 nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1677 return -EMSGSIZE;
1678
1679 return 0;
1680 }
1681
mptcp_event_sub_established(struct sk_buff * skb,const struct mptcp_sock * msk,const struct sock * ssk)1682 static int mptcp_event_sub_established(struct sk_buff *skb,
1683 const struct mptcp_sock *msk,
1684 const struct sock *ssk)
1685 {
1686 return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1687 }
1688
mptcp_event_sub_closed(struct sk_buff * skb,const struct mptcp_sock * msk,const struct sock * ssk)1689 static int mptcp_event_sub_closed(struct sk_buff *skb,
1690 const struct mptcp_sock *msk,
1691 const struct sock *ssk)
1692 {
1693 const struct mptcp_subflow_context *sf;
1694
1695 if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1696 return -EMSGSIZE;
1697
1698 sf = mptcp_subflow_ctx(ssk);
1699 if (!sf->reset_seen)
1700 return 0;
1701
1702 if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1703 return -EMSGSIZE;
1704
1705 if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1706 return -EMSGSIZE;
1707
1708 return 0;
1709 }
1710
mptcp_event_created(struct sk_buff * skb,const struct mptcp_sock * msk,const struct sock * ssk)1711 static int mptcp_event_created(struct sk_buff *skb,
1712 const struct mptcp_sock *msk,
1713 const struct sock *ssk)
1714 {
1715 int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1716
1717 if (err)
1718 return err;
1719
1720 return mptcp_event_add_subflow(skb, ssk);
1721 }
1722
mptcp_event_addr_removed(const struct mptcp_sock * msk,uint8_t id)1723 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1724 {
1725 struct net *net = sock_net((const struct sock *)msk);
1726 struct nlmsghdr *nlh;
1727 struct sk_buff *skb;
1728
1729 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1730 return;
1731
1732 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1733 if (!skb)
1734 return;
1735
1736 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1737 if (!nlh)
1738 goto nla_put_failure;
1739
1740 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1741 goto nla_put_failure;
1742
1743 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1744 goto nla_put_failure;
1745
1746 genlmsg_end(skb, nlh);
1747 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1748 return;
1749
1750 nla_put_failure:
1751 kfree_skb(skb);
1752 }
1753
mptcp_event_addr_announced(const struct mptcp_sock * msk,const struct mptcp_addr_info * info)1754 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1755 const struct mptcp_addr_info *info)
1756 {
1757 struct net *net = sock_net((const struct sock *)msk);
1758 struct nlmsghdr *nlh;
1759 struct sk_buff *skb;
1760
1761 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1762 return;
1763
1764 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1765 if (!skb)
1766 return;
1767
1768 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1769 MPTCP_EVENT_ANNOUNCED);
1770 if (!nlh)
1771 goto nla_put_failure;
1772
1773 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1774 goto nla_put_failure;
1775
1776 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1777 goto nla_put_failure;
1778
1779 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1780 goto nla_put_failure;
1781
1782 switch (info->family) {
1783 case AF_INET:
1784 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1785 goto nla_put_failure;
1786 break;
1787 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1788 case AF_INET6:
1789 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1790 goto nla_put_failure;
1791 break;
1792 #endif
1793 default:
1794 WARN_ON_ONCE(1);
1795 goto nla_put_failure;
1796 }
1797
1798 genlmsg_end(skb, nlh);
1799 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1800 return;
1801
1802 nla_put_failure:
1803 kfree_skb(skb);
1804 }
1805
mptcp_event(enum mptcp_event_type type,const struct mptcp_sock * msk,const struct sock * ssk,gfp_t gfp)1806 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1807 const struct sock *ssk, gfp_t gfp)
1808 {
1809 struct net *net = sock_net((const struct sock *)msk);
1810 struct nlmsghdr *nlh;
1811 struct sk_buff *skb;
1812
1813 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1814 return;
1815
1816 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1817 if (!skb)
1818 return;
1819
1820 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1821 if (!nlh)
1822 goto nla_put_failure;
1823
1824 switch (type) {
1825 case MPTCP_EVENT_UNSPEC:
1826 WARN_ON_ONCE(1);
1827 break;
1828 case MPTCP_EVENT_CREATED:
1829 case MPTCP_EVENT_ESTABLISHED:
1830 if (mptcp_event_created(skb, msk, ssk) < 0)
1831 goto nla_put_failure;
1832 break;
1833 case MPTCP_EVENT_CLOSED:
1834 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1835 goto nla_put_failure;
1836 break;
1837 case MPTCP_EVENT_ANNOUNCED:
1838 case MPTCP_EVENT_REMOVED:
1839 /* call mptcp_event_addr_announced()/removed instead */
1840 WARN_ON_ONCE(1);
1841 break;
1842 case MPTCP_EVENT_SUB_ESTABLISHED:
1843 case MPTCP_EVENT_SUB_PRIORITY:
1844 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1845 goto nla_put_failure;
1846 break;
1847 case MPTCP_EVENT_SUB_CLOSED:
1848 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1849 goto nla_put_failure;
1850 break;
1851 }
1852
1853 genlmsg_end(skb, nlh);
1854 mptcp_nl_mcast_send(net, skb, gfp);
1855 return;
1856
1857 nla_put_failure:
1858 kfree_skb(skb);
1859 }
1860
1861 static const struct genl_small_ops mptcp_pm_ops[] = {
1862 {
1863 .cmd = MPTCP_PM_CMD_ADD_ADDR,
1864 .doit = mptcp_nl_cmd_add_addr,
1865 .flags = GENL_ADMIN_PERM,
1866 },
1867 {
1868 .cmd = MPTCP_PM_CMD_DEL_ADDR,
1869 .doit = mptcp_nl_cmd_del_addr,
1870 .flags = GENL_ADMIN_PERM,
1871 },
1872 {
1873 .cmd = MPTCP_PM_CMD_FLUSH_ADDRS,
1874 .doit = mptcp_nl_cmd_flush_addrs,
1875 .flags = GENL_ADMIN_PERM,
1876 },
1877 {
1878 .cmd = MPTCP_PM_CMD_GET_ADDR,
1879 .doit = mptcp_nl_cmd_get_addr,
1880 .dumpit = mptcp_nl_cmd_dump_addrs,
1881 },
1882 {
1883 .cmd = MPTCP_PM_CMD_SET_LIMITS,
1884 .doit = mptcp_nl_cmd_set_limits,
1885 .flags = GENL_ADMIN_PERM,
1886 },
1887 {
1888 .cmd = MPTCP_PM_CMD_GET_LIMITS,
1889 .doit = mptcp_nl_cmd_get_limits,
1890 },
1891 {
1892 .cmd = MPTCP_PM_CMD_SET_FLAGS,
1893 .doit = mptcp_nl_cmd_set_flags,
1894 .flags = GENL_ADMIN_PERM,
1895 },
1896 };
1897
1898 static struct genl_family mptcp_genl_family __ro_after_init = {
1899 .name = MPTCP_PM_NAME,
1900 .version = MPTCP_PM_VER,
1901 .maxattr = MPTCP_PM_ATTR_MAX,
1902 .policy = mptcp_pm_policy,
1903 .netnsok = true,
1904 .module = THIS_MODULE,
1905 .small_ops = mptcp_pm_ops,
1906 .n_small_ops = ARRAY_SIZE(mptcp_pm_ops),
1907 .mcgrps = mptcp_pm_mcgrps,
1908 .n_mcgrps = ARRAY_SIZE(mptcp_pm_mcgrps),
1909 };
1910
pm_nl_init_net(struct net * net)1911 static int __net_init pm_nl_init_net(struct net *net)
1912 {
1913 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1914
1915 INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1916 __reset_counters(pernet);
1917 pernet->next_id = 1;
1918 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1919 spin_lock_init(&pernet->lock);
1920 return 0;
1921 }
1922
pm_nl_exit_net(struct list_head * net_list)1923 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1924 {
1925 struct net *net;
1926
1927 list_for_each_entry(net, net_list, exit_list) {
1928 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1929
1930 /* net is removed from namespace list, can't race with
1931 * other modifiers
1932 */
1933 __flush_addrs(&pernet->local_addr_list);
1934 }
1935 }
1936
1937 static struct pernet_operations mptcp_pm_pernet_ops = {
1938 .init = pm_nl_init_net,
1939 .exit_batch = pm_nl_exit_net,
1940 .id = &pm_nl_pernet_id,
1941 .size = sizeof(struct pm_nl_pernet),
1942 };
1943
mptcp_pm_nl_init(void)1944 void __init mptcp_pm_nl_init(void)
1945 {
1946 if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1947 panic("Failed to register MPTCP PM pernet subsystem.\n");
1948
1949 if (genl_register_family(&mptcp_genl_family))
1950 panic("Failed to register MPTCP PM netlink family\n");
1951 }
1952