1 // SPDX-License-Identifier: GPL-2.0
4 * Copyright (c) 2020, Red Hat, Inc.
7 #define pr_fmt(fmt) "MPTCP: " fmt
9 #include <linux/inet.h>
10 #include <linux/kernel.h>
12 #include <net/netns/generic.h>
13 #include <net/mptcp.h>
14 #include <net/genetlink.h>
15 #include <uapi/linux/mptcp.h>
20 /* forward declaration */
21 static struct genl_family mptcp_genl_family;
23 static int pm_nl_pernet_id;
25 struct mptcp_pm_addr_entry {
26 struct list_head list;
27 struct mptcp_addr_info addr;
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;
42 #define MAX_ADDR_ID 255
43 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
46 /* protects pernet updates */
48 struct list_head local_addr_list;
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;
55 unsigned long id_bitmap[BITMAP_SZ];
58 #define MPTCP_PM_ADDR_MAX 8
59 #define ADD_ADDR_RETRANS_MAX 3
61 static bool addresses_equal(const struct mptcp_addr_info *a,
62 struct mptcp_addr_info *b, bool use_port)
64 bool addr_equals = false;
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)
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;
86 return a->port == b->port;
89 static bool address_zero(const struct mptcp_addr_info *addr)
91 struct mptcp_addr_info zero;
93 memset(&zero, 0, sizeof(zero));
94 zero.family = addr->family;
96 return addresses_equal(addr, &zero, true);
99 static void local_address(const struct sock_common *skc,
100 struct mptcp_addr_info *addr)
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;
112 static void remote_address(const struct sock_common *skc,
113 struct mptcp_addr_info *addr)
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;
125 static bool lookup_subflow_by_saddr(const struct list_head *list,
126 struct mptcp_addr_info *saddr)
128 struct mptcp_subflow_context *subflow;
129 struct mptcp_addr_info cur;
130 struct sock_common *skc;
132 list_for_each_entry(subflow, list, node) {
133 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
135 local_address(skc, &cur);
136 if (addresses_equal(&cur, saddr, saddr->port))
143 static bool lookup_subflow_by_daddr(const struct list_head *list,
144 struct mptcp_addr_info *daddr)
146 struct mptcp_subflow_context *subflow;
147 struct mptcp_addr_info cur;
148 struct sock_common *skc;
150 list_for_each_entry(subflow, list, node) {
151 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
153 remote_address(skc, &cur);
154 if (addresses_equal(&cur, daddr, daddr->port))
161 static struct mptcp_pm_addr_entry *
162 select_local_address(const struct pm_nl_pernet *pernet,
163 struct mptcp_sock *msk)
165 struct mptcp_pm_addr_entry *entry, *ret = NULL;
166 struct sock *sk = (struct sock *)msk;
168 msk_owned_by_me(msk);
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))
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)))
186 /* avoid any address already in use by subflows and
189 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
198 static struct mptcp_pm_addr_entry *
199 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
201 struct mptcp_pm_addr_entry *entry, *ret = NULL;
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.
210 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
211 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
222 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
224 struct pm_nl_pernet *pernet;
226 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
227 return READ_ONCE(pernet->add_addr_signal_max);
229 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
231 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
233 struct pm_nl_pernet *pernet;
235 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
236 return READ_ONCE(pernet->add_addr_accept_max);
238 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
240 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
242 struct pm_nl_pernet *pernet;
244 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
245 return READ_ONCE(pernet->subflows_max);
247 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
249 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
251 struct pm_nl_pernet *pernet;
253 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
254 return READ_ONCE(pernet->local_addr_max);
256 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
258 static void check_work_pending(struct mptcp_sock *msk)
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);
266 struct mptcp_pm_add_entry *
267 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
268 struct mptcp_addr_info *addr)
270 struct mptcp_pm_add_entry *entry;
272 lockdep_assert_held(&msk->pm.lock);
274 list_for_each_entry(entry, &msk->pm.anno_list, list) {
275 if (addresses_equal(&entry->addr, addr, true))
282 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
284 struct mptcp_pm_add_entry *entry;
285 struct mptcp_addr_info saddr;
288 local_address((struct sock_common *)sk, &saddr);
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)) {
299 spin_unlock_bh(&msk->pm.lock);
303 static void mptcp_pm_add_timer(struct timer_list *timer)
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;
309 pr_debug("msk=%p", msk);
314 if (inet_sk_state_load(sk) == TCP_CLOSE)
320 if (mptcp_pm_should_add_signal(msk)) {
321 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
325 spin_lock_bh(&msk->pm.lock);
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++;
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)));
338 spin_unlock_bh(&msk->pm.lock);
340 if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
341 mptcp_pm_subflow_established(msk);
347 struct mptcp_pm_add_entry *
348 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
349 struct mptcp_addr_info *addr, bool check_id)
351 struct mptcp_pm_add_entry *entry;
352 struct sock *sk = (struct sock *)msk;
354 spin_lock_bh(&msk->pm.lock);
355 entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
356 if (entry && (!check_id || entry->addr.id == addr->id))
357 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
358 spin_unlock_bh(&msk->pm.lock);
360 if (entry && (!check_id || entry->addr.id == addr->id))
361 sk_stop_timer_sync(sk, &entry->add_timer);
366 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
367 struct mptcp_pm_addr_entry *entry)
369 struct mptcp_pm_add_entry *add_entry = NULL;
370 struct sock *sk = (struct sock *)msk;
371 struct net *net = sock_net(sk);
373 lockdep_assert_held(&msk->pm.lock);
375 if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
378 add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
382 list_add(&add_entry->list, &msk->pm.anno_list);
384 add_entry->addr = entry->addr;
385 add_entry->sock = msk;
386 add_entry->retrans_times = 0;
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));
395 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
397 struct mptcp_pm_add_entry *entry, *tmp;
398 struct sock *sk = (struct sock *)msk;
399 LIST_HEAD(free_list);
401 pr_debug("msk=%p", msk);
403 spin_lock_bh(&msk->pm.lock);
404 list_splice_init(&msk->pm.anno_list, &free_list);
405 spin_unlock_bh(&msk->pm.lock);
407 list_for_each_entry_safe(entry, tmp, &free_list, list) {
408 sk_stop_timer_sync(sk, &entry->add_timer);
413 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
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;
422 pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
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);
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);
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);
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);
445 /* pick failed, avoid fourther attempts later */
446 msk->pm.local_addr_used = add_addr_signal_max;
449 check_work_pending(msk);
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);
457 struct mptcp_addr_info remote = { 0 };
459 msk->pm.local_addr_used++;
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);
470 /* lookup failed, avoid fourther attempts later */
471 msk->pm.local_addr_used = local_addr_max;
472 check_work_pending(msk);
476 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
478 mptcp_pm_create_subflow_or_signal_addr(msk);
481 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
483 mptcp_pm_create_subflow_or_signal_addr(msk);
486 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
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;
494 add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
495 subflows_max = mptcp_pm_get_subflows_max(msk);
497 pr_debug("accepted %d:%d remote family %d",
498 msk->pm.add_addr_accepted, add_addr_accept_max,
499 msk->pm.remote.family);
501 if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
504 msk->pm.add_addr_accepted++;
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);
510 /* connect to the specified remote address, using whatever
511 * local address the routing configuration will pick.
513 remote = msk->pm.remote;
515 remote.port = sk->sk_dport;
516 memset(&local, 0, sizeof(local));
517 local.family = remote.family;
519 spin_unlock_bh(&msk->pm.lock);
520 __mptcp_subflow_connect(sk, &local, &remote, 0, 0);
521 spin_lock_bh(&msk->pm.lock);
524 mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
525 mptcp_pm_nl_addr_send_ack(msk);
528 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
530 struct mptcp_subflow_context *subflow;
532 msk_owned_by_me(msk);
533 lockdep_assert_held(&msk->pm.lock);
535 if (!mptcp_pm_should_add_signal(msk) &&
536 !mptcp_pm_should_rm_signal(msk))
539 __mptcp_flush_join_list(msk);
540 subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
542 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
545 spin_unlock_bh(&msk->pm.lock);
546 pr_debug("send ack for %s%s%s",
547 mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
548 mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
549 mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
551 slow = lock_sock_fast(ssk);
553 unlock_sock_fast(ssk, slow);
554 spin_lock_bh(&msk->pm.lock);
558 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
559 struct mptcp_addr_info *addr,
562 struct mptcp_subflow_context *subflow;
564 pr_debug("bkup=%d", bkup);
566 mptcp_for_each_subflow(msk, subflow) {
567 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
568 struct sock *sk = (struct sock *)msk;
569 struct mptcp_addr_info local;
572 local_address((struct sock_common *)ssk, &local);
573 if (!addresses_equal(&local, addr, addr->port))
576 subflow->backup = bkup;
577 subflow->send_mp_prio = 1;
578 subflow->request_bkup = bkup;
579 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
581 spin_unlock_bh(&msk->pm.lock);
582 pr_debug("send ack for mp_prio");
583 slow = lock_sock_fast(ssk);
585 unlock_sock_fast(ssk, slow);
586 spin_lock_bh(&msk->pm.lock);
594 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
595 const struct mptcp_rm_list *rm_list,
596 enum linux_mptcp_mib_field rm_type)
598 struct mptcp_subflow_context *subflow, *tmp;
599 struct sock *sk = (struct sock *)msk;
602 pr_debug("%s rm_list_nr %d",
603 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
605 msk_owned_by_me(msk);
610 if (list_empty(&msk->conn_list))
613 for (i = 0; i < rm_list->nr; i++) {
614 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
615 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
616 int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
617 u8 id = subflow->local_id;
619 if (rm_type == MPTCP_MIB_RMADDR)
620 id = subflow->remote_id;
622 if (rm_list->ids[i] != id)
625 pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
626 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
627 i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
628 spin_unlock_bh(&msk->pm.lock);
629 mptcp_subflow_shutdown(sk, ssk, how);
630 mptcp_close_ssk(sk, ssk, subflow);
631 spin_lock_bh(&msk->pm.lock);
633 if (rm_type == MPTCP_MIB_RMADDR) {
634 msk->pm.add_addr_accepted--;
635 WRITE_ONCE(msk->pm.accept_addr, true);
636 } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
637 msk->pm.local_addr_used--;
640 __MPTCP_INC_STATS(sock_net(sk), rm_type);
645 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
647 mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
650 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
651 const struct mptcp_rm_list *rm_list)
653 mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
656 void mptcp_pm_nl_work(struct mptcp_sock *msk)
658 struct mptcp_pm_data *pm = &msk->pm;
660 msk_owned_by_me(msk);
662 spin_lock_bh(&msk->pm.lock);
664 pr_debug("msk=%p status=%x", msk, pm->status);
665 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
666 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
667 mptcp_pm_nl_add_addr_received(msk);
669 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
670 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
671 mptcp_pm_nl_addr_send_ack(msk);
673 if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
674 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
675 mptcp_pm_nl_rm_addr_received(msk);
677 if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
678 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
679 mptcp_pm_nl_fully_established(msk);
681 if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
682 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
683 mptcp_pm_nl_subflow_established(msk);
686 spin_unlock_bh(&msk->pm.lock);
689 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
691 return (entry->flags &
692 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
693 MPTCP_PM_ADDR_FLAG_SIGNAL;
696 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
697 struct mptcp_pm_addr_entry *entry)
699 struct mptcp_pm_addr_entry *cur;
700 unsigned int addr_max;
703 spin_lock_bh(&pernet->lock);
704 /* to keep the code simple, don't do IDR-like allocation for address ID,
705 * just bail when we exceed limits
707 if (pernet->next_id == MAX_ADDR_ID)
709 if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
711 if (test_bit(entry->addr.id, pernet->id_bitmap))
714 /* do not insert duplicate address, differentiate on port only
717 list_for_each_entry(cur, &pernet->local_addr_list, list) {
718 if (addresses_equal(&cur->addr, &entry->addr,
719 address_use_port(entry) &&
720 address_use_port(cur)))
724 if (!entry->addr.id) {
726 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
729 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
730 pernet->next_id != 1) {
736 if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
739 __set_bit(entry->addr.id, pernet->id_bitmap);
740 if (entry->addr.id > pernet->next_id)
741 pernet->next_id = entry->addr.id;
743 if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
744 addr_max = pernet->add_addr_signal_max;
745 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
747 if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
748 addr_max = pernet->local_addr_max;
749 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
753 list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
754 ret = entry->addr.id;
757 spin_unlock_bh(&pernet->lock);
761 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
762 struct mptcp_pm_addr_entry *entry)
764 struct sockaddr_storage addr;
765 struct mptcp_sock *msk;
766 struct socket *ssock;
770 err = sock_create_kern(sock_net(sk), entry->addr.family,
771 SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
775 msk = mptcp_sk(entry->lsk->sk);
781 ssock = __mptcp_nmpc_socket(msk);
787 mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
788 err = kernel_bind(ssock, (struct sockaddr *)&addr,
789 sizeof(struct sockaddr_in));
791 pr_warn("kernel_bind error, err=%d", err);
795 err = kernel_listen(ssock, backlog);
797 pr_warn("kernel_listen error, err=%d", err);
804 sock_release(entry->lsk);
808 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
810 struct mptcp_pm_addr_entry *entry;
811 struct mptcp_addr_info skc_local;
812 struct mptcp_addr_info msk_local;
813 struct pm_nl_pernet *pernet;
816 if (WARN_ON_ONCE(!msk))
819 /* The 0 ID mapping is defined by the first subflow, copied into the msk
822 local_address((struct sock_common *)msk, &msk_local);
823 local_address((struct sock_common *)skc, &skc_local);
824 if (addresses_equal(&msk_local, &skc_local, false))
827 if (address_zero(&skc_local))
830 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
833 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
834 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
835 ret = entry->addr.id;
843 /* address not found, add to local list */
844 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
848 entry->addr = skc_local;
850 entry->addr.port = 0;
854 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
861 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
863 struct mptcp_pm_data *pm = &msk->pm;
866 subflows = !!mptcp_pm_get_subflows_max(msk);
867 WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
868 !!mptcp_pm_get_add_addr_signal_max(msk));
869 WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
870 WRITE_ONCE(pm->accept_subflow, subflows);
873 #define MPTCP_PM_CMD_GRP_OFFSET 0
874 #define MPTCP_PM_EV_GRP_OFFSET 1
876 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
877 [MPTCP_PM_CMD_GRP_OFFSET] = { .name = MPTCP_PM_CMD_GRP_NAME, },
878 [MPTCP_PM_EV_GRP_OFFSET] = { .name = MPTCP_PM_EV_GRP_NAME,
879 .flags = GENL_UNS_ADMIN_PERM,
883 static const struct nla_policy
884 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
885 [MPTCP_PM_ADDR_ATTR_FAMILY] = { .type = NLA_U16, },
886 [MPTCP_PM_ADDR_ATTR_ID] = { .type = NLA_U8, },
887 [MPTCP_PM_ADDR_ATTR_ADDR4] = { .type = NLA_U32, },
888 [MPTCP_PM_ADDR_ATTR_ADDR6] =
889 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
890 [MPTCP_PM_ADDR_ATTR_PORT] = { .type = NLA_U16 },
891 [MPTCP_PM_ADDR_ATTR_FLAGS] = { .type = NLA_U32 },
892 [MPTCP_PM_ADDR_ATTR_IF_IDX] = { .type = NLA_S32 },
895 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
896 [MPTCP_PM_ATTR_ADDR] =
897 NLA_POLICY_NESTED(mptcp_pm_addr_policy),
898 [MPTCP_PM_ATTR_RCV_ADD_ADDRS] = { .type = NLA_U32, },
899 [MPTCP_PM_ATTR_SUBFLOWS] = { .type = NLA_U32, },
902 static int mptcp_pm_family_to_addr(int family)
904 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
905 if (family == AF_INET6)
906 return MPTCP_PM_ADDR_ATTR_ADDR6;
908 return MPTCP_PM_ADDR_ATTR_ADDR4;
911 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
913 struct mptcp_pm_addr_entry *entry)
915 struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
919 GENL_SET_ERR_MSG(info, "missing address info");
923 /* no validation needed - was already done via nested policy */
924 err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
925 mptcp_pm_addr_policy, info->extack);
929 memset(entry, 0, sizeof(*entry));
930 if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
934 NL_SET_ERR_MSG_ATTR(info->extack, attr,
939 entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
940 if (entry->addr.family != AF_INET
941 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
942 && entry->addr.family != AF_INET6
945 NL_SET_ERR_MSG_ATTR(info->extack, attr,
946 "unknown address family");
949 addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
950 if (!tb[addr_addr]) {
951 NL_SET_ERR_MSG_ATTR(info->extack, attr,
952 "missing address data");
956 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
957 if (entry->addr.family == AF_INET6)
958 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
961 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
964 if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
965 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
967 entry->ifindex = val;
970 if (tb[MPTCP_PM_ADDR_ATTR_ID])
971 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
973 if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
974 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
976 if (tb[MPTCP_PM_ADDR_ATTR_PORT]) {
977 if (!(entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL)) {
978 NL_SET_ERR_MSG_ATTR(info->extack, attr,
979 "flags must have signal when using port");
982 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
988 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
990 return net_generic(genl_info_net(info), pm_nl_pernet_id);
993 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
995 struct mptcp_sock *msk;
996 long s_slot = 0, s_num = 0;
998 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
999 struct sock *sk = (struct sock *)msk;
1001 if (!READ_ONCE(msk->fully_established))
1005 spin_lock_bh(&msk->pm.lock);
1006 mptcp_pm_create_subflow_or_signal_addr(msk);
1007 spin_unlock_bh(&msk->pm.lock);
1018 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1020 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1021 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1022 struct mptcp_pm_addr_entry addr, *entry;
1025 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1029 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1031 GENL_SET_ERR_MSG(info, "can't allocate addr");
1036 if (entry->addr.port) {
1037 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1039 GENL_SET_ERR_MSG(info, "create listen socket error");
1044 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1046 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1048 sock_release(entry->lsk);
1053 mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1058 static struct mptcp_pm_addr_entry *
1059 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1061 struct mptcp_pm_addr_entry *entry;
1063 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1064 if (entry->addr.id == id)
1070 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1071 struct mptcp_addr_info *addr)
1073 struct mptcp_pm_add_entry *entry;
1075 entry = mptcp_pm_del_add_timer(msk, addr, false);
1077 list_del(&entry->list);
1085 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1086 struct mptcp_addr_info *addr,
1089 struct mptcp_rm_list list = { .nr = 0 };
1092 list.ids[list.nr++] = addr->id;
1094 ret = remove_anno_list_by_saddr(msk, addr);
1096 spin_lock_bh(&msk->pm.lock);
1097 mptcp_pm_remove_addr(msk, &list);
1098 spin_unlock_bh(&msk->pm.lock);
1103 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1104 struct mptcp_addr_info *addr)
1106 struct mptcp_sock *msk;
1107 long s_slot = 0, s_num = 0;
1108 struct mptcp_rm_list list = { .nr = 0 };
1110 pr_debug("remove_id=%d", addr->id);
1112 list.ids[list.nr++] = addr->id;
1114 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1115 struct sock *sk = (struct sock *)msk;
1116 bool remove_subflow;
1118 if (list_empty(&msk->conn_list)) {
1119 mptcp_pm_remove_anno_addr(msk, addr, false);
1124 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1125 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1127 mptcp_pm_remove_subflow(msk, &list);
1138 /* caller must ensure the RCU grace period is already elapsed */
1139 static void __mptcp_pm_release_addr_entry(struct mptcp_pm_addr_entry *entry)
1142 sock_release(entry->lsk);
1146 static int mptcp_nl_remove_id_zero_address(struct net *net,
1147 struct mptcp_addr_info *addr)
1149 struct mptcp_rm_list list = { .nr = 0 };
1150 long s_slot = 0, s_num = 0;
1151 struct mptcp_sock *msk;
1153 list.ids[list.nr++] = 0;
1155 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1156 struct sock *sk = (struct sock *)msk;
1157 struct mptcp_addr_info msk_local;
1159 if (list_empty(&msk->conn_list))
1162 local_address((struct sock_common *)msk, &msk_local);
1163 if (!addresses_equal(&msk_local, addr, addr->port))
1167 spin_lock_bh(&msk->pm.lock);
1168 mptcp_pm_remove_addr(msk, &list);
1169 mptcp_pm_nl_rm_subflow_received(msk, &list);
1170 spin_unlock_bh(&msk->pm.lock);
1181 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1183 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1184 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1185 struct mptcp_pm_addr_entry addr, *entry;
1186 unsigned int addr_max;
1189 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1193 /* the zero id address is special: the first address used by the msk
1194 * always gets such an id, so different subflows can have different zero
1195 * id addresses. Additionally zero id is not accounted for in id_bitmap.
1196 * Let's use an 'mptcp_rm_list' instead of the common remove code.
1198 if (addr.addr.id == 0)
1199 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1201 spin_lock_bh(&pernet->lock);
1202 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1204 GENL_SET_ERR_MSG(info, "address not found");
1205 spin_unlock_bh(&pernet->lock);
1208 if (entry->flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1209 addr_max = pernet->add_addr_signal_max;
1210 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1212 if (entry->flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1213 addr_max = pernet->local_addr_max;
1214 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1218 list_del_rcu(&entry->list);
1219 __clear_bit(entry->addr.id, pernet->id_bitmap);
1220 spin_unlock_bh(&pernet->lock);
1222 mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1224 __mptcp_pm_release_addr_entry(entry);
1229 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1230 struct list_head *rm_list)
1232 struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1233 struct mptcp_pm_addr_entry *entry;
1235 list_for_each_entry(entry, rm_list, list) {
1236 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1237 alist.nr < MPTCP_RM_IDS_MAX &&
1238 slist.nr < MPTCP_RM_IDS_MAX) {
1239 alist.ids[alist.nr++] = entry->addr.id;
1240 slist.ids[slist.nr++] = entry->addr.id;
1241 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1242 alist.nr < MPTCP_RM_IDS_MAX) {
1243 alist.ids[alist.nr++] = entry->addr.id;
1248 spin_lock_bh(&msk->pm.lock);
1249 mptcp_pm_remove_addr(msk, &alist);
1250 spin_unlock_bh(&msk->pm.lock);
1253 mptcp_pm_remove_subflow(msk, &slist);
1256 static void mptcp_nl_remove_addrs_list(struct net *net,
1257 struct list_head *rm_list)
1259 long s_slot = 0, s_num = 0;
1260 struct mptcp_sock *msk;
1262 if (list_empty(rm_list))
1265 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1266 struct sock *sk = (struct sock *)msk;
1269 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1277 /* caller must ensure the RCU grace period is already elapsed */
1278 static void __flush_addrs(struct list_head *list)
1280 while (!list_empty(list)) {
1281 struct mptcp_pm_addr_entry *cur;
1283 cur = list_entry(list->next,
1284 struct mptcp_pm_addr_entry, list);
1285 list_del_rcu(&cur->list);
1286 __mptcp_pm_release_addr_entry(cur);
1290 static void __reset_counters(struct pm_nl_pernet *pernet)
1292 WRITE_ONCE(pernet->add_addr_signal_max, 0);
1293 WRITE_ONCE(pernet->add_addr_accept_max, 0);
1294 WRITE_ONCE(pernet->local_addr_max, 0);
1298 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1300 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1301 LIST_HEAD(free_list);
1303 spin_lock_bh(&pernet->lock);
1304 list_splice_init(&pernet->local_addr_list, &free_list);
1305 __reset_counters(pernet);
1306 pernet->next_id = 1;
1307 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1308 spin_unlock_bh(&pernet->lock);
1309 mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1311 __flush_addrs(&free_list);
1315 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1316 struct mptcp_pm_addr_entry *entry)
1318 struct mptcp_addr_info *addr = &entry->addr;
1319 struct nlattr *attr;
1321 attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1325 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1326 goto nla_put_failure;
1327 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1328 goto nla_put_failure;
1329 if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1330 goto nla_put_failure;
1331 if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->flags))
1332 goto nla_put_failure;
1333 if (entry->ifindex &&
1334 nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->ifindex))
1335 goto nla_put_failure;
1337 if (addr->family == AF_INET &&
1338 nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1340 goto nla_put_failure;
1341 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1342 else if (addr->family == AF_INET6 &&
1343 nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1344 goto nla_put_failure;
1346 nla_nest_end(skb, attr);
1350 nla_nest_cancel(skb, attr);
1354 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1356 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1357 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1358 struct mptcp_pm_addr_entry addr, *entry;
1359 struct sk_buff *msg;
1363 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1367 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1371 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1372 info->genlhdr->cmd);
1374 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1379 spin_lock_bh(&pernet->lock);
1380 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1382 GENL_SET_ERR_MSG(info, "address not found");
1387 ret = mptcp_nl_fill_addr(msg, entry);
1391 genlmsg_end(msg, reply);
1392 ret = genlmsg_reply(msg, info);
1393 spin_unlock_bh(&pernet->lock);
1397 spin_unlock_bh(&pernet->lock);
1404 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1405 struct netlink_callback *cb)
1407 struct net *net = sock_net(msg->sk);
1408 struct mptcp_pm_addr_entry *entry;
1409 struct pm_nl_pernet *pernet;
1410 int id = cb->args[0];
1414 pernet = net_generic(net, pm_nl_pernet_id);
1416 spin_lock_bh(&pernet->lock);
1417 for (i = id; i < MAX_ADDR_ID + 1; i++) {
1418 if (test_bit(i, pernet->id_bitmap)) {
1419 entry = __lookup_addr_by_id(pernet, i);
1423 if (entry->addr.id <= id)
1426 hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1427 cb->nlh->nlmsg_seq, &mptcp_genl_family,
1428 NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1432 if (mptcp_nl_fill_addr(msg, entry) < 0) {
1433 genlmsg_cancel(msg, hdr);
1437 id = entry->addr.id;
1438 genlmsg_end(msg, hdr);
1441 spin_unlock_bh(&pernet->lock);
1447 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1449 struct nlattr *attr = info->attrs[id];
1454 *limit = nla_get_u32(attr);
1455 if (*limit > MPTCP_PM_ADDR_MAX) {
1456 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1463 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1465 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1466 unsigned int rcv_addrs, subflows;
1469 spin_lock_bh(&pernet->lock);
1470 rcv_addrs = pernet->add_addr_accept_max;
1471 ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1475 subflows = pernet->subflows_max;
1476 ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1480 WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1481 WRITE_ONCE(pernet->subflows_max, subflows);
1484 spin_unlock_bh(&pernet->lock);
1489 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1491 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1492 struct sk_buff *msg;
1495 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1499 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1500 MPTCP_PM_CMD_GET_LIMITS);
1504 if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1505 READ_ONCE(pernet->add_addr_accept_max)))
1508 if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1509 READ_ONCE(pernet->subflows_max)))
1512 genlmsg_end(msg, reply);
1513 return genlmsg_reply(msg, info);
1516 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1521 static int mptcp_nl_addr_backup(struct net *net,
1522 struct mptcp_addr_info *addr,
1525 long s_slot = 0, s_num = 0;
1526 struct mptcp_sock *msk;
1529 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1530 struct sock *sk = (struct sock *)msk;
1532 if (list_empty(&msk->conn_list))
1536 spin_lock_bh(&msk->pm.lock);
1537 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1538 spin_unlock_bh(&msk->pm.lock);
1549 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1551 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1552 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1553 struct mptcp_pm_addr_entry addr, *entry;
1554 struct net *net = sock_net(skb->sk);
1558 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1562 if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1565 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1566 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1567 ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1572 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1574 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1581 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1583 genlmsg_multicast_netns(&mptcp_genl_family, net,
1584 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1587 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1589 const struct inet_sock *issk = inet_sk(ssk);
1590 const struct mptcp_subflow_context *sf;
1592 if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1595 switch (ssk->sk_family) {
1597 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1599 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1602 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1604 const struct ipv6_pinfo *np = inet6_sk(ssk);
1606 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1608 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1618 if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1620 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1623 sf = mptcp_subflow_ctx(ssk);
1624 if (WARN_ON_ONCE(!sf))
1627 if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1630 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1636 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1637 const struct mptcp_sock *msk,
1638 const struct sock *ssk)
1640 const struct sock *sk = (const struct sock *)msk;
1641 const struct mptcp_subflow_context *sf;
1644 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1647 if (mptcp_event_add_subflow(skb, ssk))
1650 sf = mptcp_subflow_ctx(ssk);
1651 if (WARN_ON_ONCE(!sf))
1654 if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1657 if (ssk->sk_bound_dev_if &&
1658 nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1661 sk_err = ssk->sk_err;
1662 if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1663 nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1669 static int mptcp_event_sub_established(struct sk_buff *skb,
1670 const struct mptcp_sock *msk,
1671 const struct sock *ssk)
1673 return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1676 static int mptcp_event_sub_closed(struct sk_buff *skb,
1677 const struct mptcp_sock *msk,
1678 const struct sock *ssk)
1680 const struct mptcp_subflow_context *sf;
1682 if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1685 sf = mptcp_subflow_ctx(ssk);
1686 if (!sf->reset_seen)
1689 if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1692 if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1698 static int mptcp_event_created(struct sk_buff *skb,
1699 const struct mptcp_sock *msk,
1700 const struct sock *ssk)
1702 int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1707 return mptcp_event_add_subflow(skb, ssk);
1710 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1712 struct net *net = sock_net((const struct sock *)msk);
1713 struct nlmsghdr *nlh;
1714 struct sk_buff *skb;
1716 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1719 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1723 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1725 goto nla_put_failure;
1727 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1728 goto nla_put_failure;
1730 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1731 goto nla_put_failure;
1733 genlmsg_end(skb, nlh);
1734 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1741 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1742 const struct mptcp_addr_info *info)
1744 struct net *net = sock_net((const struct sock *)msk);
1745 struct nlmsghdr *nlh;
1746 struct sk_buff *skb;
1748 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1751 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1755 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1756 MPTCP_EVENT_ANNOUNCED);
1758 goto nla_put_failure;
1760 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1761 goto nla_put_failure;
1763 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1764 goto nla_put_failure;
1766 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1767 goto nla_put_failure;
1769 switch (info->family) {
1771 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1772 goto nla_put_failure;
1774 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1776 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1777 goto nla_put_failure;
1782 goto nla_put_failure;
1785 genlmsg_end(skb, nlh);
1786 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1793 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1794 const struct sock *ssk, gfp_t gfp)
1796 struct net *net = sock_net((const struct sock *)msk);
1797 struct nlmsghdr *nlh;
1798 struct sk_buff *skb;
1800 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1803 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1807 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1809 goto nla_put_failure;
1812 case MPTCP_EVENT_UNSPEC:
1815 case MPTCP_EVENT_CREATED:
1816 case MPTCP_EVENT_ESTABLISHED:
1817 if (mptcp_event_created(skb, msk, ssk) < 0)
1818 goto nla_put_failure;
1820 case MPTCP_EVENT_CLOSED:
1821 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1822 goto nla_put_failure;
1824 case MPTCP_EVENT_ANNOUNCED:
1825 case MPTCP_EVENT_REMOVED:
1826 /* call mptcp_event_addr_announced()/removed instead */
1829 case MPTCP_EVENT_SUB_ESTABLISHED:
1830 case MPTCP_EVENT_SUB_PRIORITY:
1831 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1832 goto nla_put_failure;
1834 case MPTCP_EVENT_SUB_CLOSED:
1835 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1836 goto nla_put_failure;
1840 genlmsg_end(skb, nlh);
1841 mptcp_nl_mcast_send(net, skb, gfp);
1848 static const struct genl_small_ops mptcp_pm_ops[] = {
1850 .cmd = MPTCP_PM_CMD_ADD_ADDR,
1851 .doit = mptcp_nl_cmd_add_addr,
1852 .flags = GENL_ADMIN_PERM,
1855 .cmd = MPTCP_PM_CMD_DEL_ADDR,
1856 .doit = mptcp_nl_cmd_del_addr,
1857 .flags = GENL_ADMIN_PERM,
1860 .cmd = MPTCP_PM_CMD_FLUSH_ADDRS,
1861 .doit = mptcp_nl_cmd_flush_addrs,
1862 .flags = GENL_ADMIN_PERM,
1865 .cmd = MPTCP_PM_CMD_GET_ADDR,
1866 .doit = mptcp_nl_cmd_get_addr,
1867 .dumpit = mptcp_nl_cmd_dump_addrs,
1870 .cmd = MPTCP_PM_CMD_SET_LIMITS,
1871 .doit = mptcp_nl_cmd_set_limits,
1872 .flags = GENL_ADMIN_PERM,
1875 .cmd = MPTCP_PM_CMD_GET_LIMITS,
1876 .doit = mptcp_nl_cmd_get_limits,
1879 .cmd = MPTCP_PM_CMD_SET_FLAGS,
1880 .doit = mptcp_nl_cmd_set_flags,
1881 .flags = GENL_ADMIN_PERM,
1885 static struct genl_family mptcp_genl_family __ro_after_init = {
1886 .name = MPTCP_PM_NAME,
1887 .version = MPTCP_PM_VER,
1888 .maxattr = MPTCP_PM_ATTR_MAX,
1889 .policy = mptcp_pm_policy,
1891 .module = THIS_MODULE,
1892 .small_ops = mptcp_pm_ops,
1893 .n_small_ops = ARRAY_SIZE(mptcp_pm_ops),
1894 .mcgrps = mptcp_pm_mcgrps,
1895 .n_mcgrps = ARRAY_SIZE(mptcp_pm_mcgrps),
1898 static int __net_init pm_nl_init_net(struct net *net)
1900 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1902 INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1903 __reset_counters(pernet);
1904 pernet->next_id = 1;
1905 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1906 spin_lock_init(&pernet->lock);
1910 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1914 list_for_each_entry(net, net_list, exit_list) {
1915 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1917 /* net is removed from namespace list, can't race with
1918 * other modifiers, also netns core already waited for a
1921 __flush_addrs(&pernet->local_addr_list);
1925 static struct pernet_operations mptcp_pm_pernet_ops = {
1926 .init = pm_nl_init_net,
1927 .exit_batch = pm_nl_exit_net,
1928 .id = &pm_nl_pernet_id,
1929 .size = sizeof(struct pm_nl_pernet),
1932 void __init mptcp_pm_nl_init(void)
1934 if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1935 panic("Failed to register MPTCP PM pernet subsystem.\n");
1937 if (genl_register_family(&mptcp_genl_family))
1938 panic("Failed to register MPTCP PM netlink family\n");