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);
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]" : "");
553 spin_lock_bh(&msk->pm.lock);
557 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
558 struct mptcp_addr_info *addr,
561 struct mptcp_subflow_context *subflow;
563 pr_debug("bkup=%d", bkup);
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;
570 local_address((struct sock_common *)ssk, &local);
571 if (!addresses_equal(&local, addr, addr->port))
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);
579 spin_unlock_bh(&msk->pm.lock);
580 pr_debug("send ack for mp_prio");
584 spin_lock_bh(&msk->pm.lock);
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)
596 struct mptcp_subflow_context *subflow, *tmp;
597 struct sock *sk = (struct sock *)msk;
600 pr_debug("%s rm_list_nr %d",
601 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
603 msk_owned_by_me(msk);
608 if (list_empty(&msk->conn_list))
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;
617 if (rm_type == MPTCP_MIB_RMADDR)
618 id = subflow->remote_id;
620 if (rm_list->ids[i] != id)
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);
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--;
638 __MPTCP_INC_STATS(sock_net(sk), rm_type);
643 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
645 mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
648 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
649 const struct mptcp_rm_list *rm_list)
651 mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
654 void mptcp_pm_nl_work(struct mptcp_sock *msk)
656 struct mptcp_pm_data *pm = &msk->pm;
658 msk_owned_by_me(msk);
660 spin_lock_bh(&msk->pm.lock);
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);
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);
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);
675 if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
676 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
677 mptcp_pm_nl_fully_established(msk);
679 if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
680 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
681 mptcp_pm_nl_subflow_established(msk);
684 spin_unlock_bh(&msk->pm.lock);
687 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
689 return (entry->flags &
690 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
691 MPTCP_PM_ADDR_FLAG_SIGNAL;
694 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
695 struct mptcp_pm_addr_entry *entry)
697 struct mptcp_pm_addr_entry *cur;
698 unsigned int addr_max;
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
705 if (pernet->next_id == MAX_ADDR_ID)
707 if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
709 if (test_bit(entry->addr.id, pernet->id_bitmap))
712 /* do not insert duplicate address, differentiate on port only
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)))
722 if (!entry->addr.id) {
724 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
727 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
728 pernet->next_id != 1) {
734 if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
737 __set_bit(entry->addr.id, pernet->id_bitmap);
738 if (entry->addr.id > pernet->next_id)
739 pernet->next_id = entry->addr.id;
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);
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);
751 list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
752 ret = entry->addr.id;
755 spin_unlock_bh(&pernet->lock);
759 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
760 struct mptcp_pm_addr_entry *entry)
762 struct sockaddr_storage addr;
763 struct mptcp_sock *msk;
764 struct socket *ssock;
768 err = sock_create_kern(sock_net(sk), entry->addr.family,
769 SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
773 msk = mptcp_sk(entry->lsk->sk);
779 ssock = __mptcp_nmpc_socket(msk);
785 mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
786 err = kernel_bind(ssock, (struct sockaddr *)&addr,
787 sizeof(struct sockaddr_in));
789 pr_warn("kernel_bind error, err=%d", err);
793 err = kernel_listen(ssock, backlog);
795 pr_warn("kernel_listen error, err=%d", err);
802 sock_release(entry->lsk);
806 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
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;
814 if (WARN_ON_ONCE(!msk))
817 /* The 0 ID mapping is defined by the first subflow, copied into the msk
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))
825 if (address_zero(&skc_local))
828 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
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;
841 /* address not found, add to local list */
842 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
846 entry->addr = skc_local;
848 entry->addr.port = 0;
852 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
859 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
861 struct mptcp_pm_data *pm = &msk->pm;
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);
871 #define MPTCP_PM_CMD_GRP_OFFSET 0
872 #define MPTCP_PM_EV_GRP_OFFSET 1
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,
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 },
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, },
900 static int mptcp_pm_family_to_addr(int family)
902 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
903 if (family == AF_INET6)
904 return MPTCP_PM_ADDR_ATTR_ADDR6;
906 return MPTCP_PM_ADDR_ATTR_ADDR4;
909 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
911 struct mptcp_pm_addr_entry *entry)
913 struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
917 GENL_SET_ERR_MSG(info, "missing address info");
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);
927 memset(entry, 0, sizeof(*entry));
928 if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
932 NL_SET_ERR_MSG_ATTR(info->extack, attr,
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
943 NL_SET_ERR_MSG_ATTR(info->extack, attr,
944 "unknown address family");
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");
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]);
959 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
962 if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
963 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
965 entry->ifindex = val;
968 if (tb[MPTCP_PM_ADDR_ATTR_ID])
969 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
971 if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
972 entry->flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
974 if (tb[MPTCP_PM_ADDR_ATTR_PORT])
975 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
980 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
982 return net_generic(genl_info_net(info), pm_nl_pernet_id);
985 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
987 struct mptcp_sock *msk;
988 long s_slot = 0, s_num = 0;
990 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
991 struct sock *sk = (struct sock *)msk;
993 if (!READ_ONCE(msk->fully_established))
997 spin_lock_bh(&msk->pm.lock);
998 mptcp_pm_create_subflow_or_signal_addr(msk);
999 spin_unlock_bh(&msk->pm.lock);
1010 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
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;
1017 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1021 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1023 GENL_SET_ERR_MSG(info, "can't allocate addr");
1028 if (entry->addr.port) {
1029 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1031 GENL_SET_ERR_MSG(info, "create listen socket error");
1036 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1038 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1040 sock_release(entry->lsk);
1045 mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1050 static struct mptcp_pm_addr_entry *
1051 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1053 struct mptcp_pm_addr_entry *entry;
1055 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1056 if (entry->addr.id == id)
1062 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1063 struct mptcp_addr_info *addr)
1065 struct mptcp_pm_add_entry *entry;
1067 entry = mptcp_pm_del_add_timer(msk, addr, false);
1069 list_del(&entry->list);
1077 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1078 struct mptcp_addr_info *addr,
1081 struct mptcp_rm_list list = { .nr = 0 };
1084 list.ids[list.nr++] = addr->id;
1086 ret = remove_anno_list_by_saddr(msk, addr);
1088 spin_lock_bh(&msk->pm.lock);
1089 mptcp_pm_remove_addr(msk, &list);
1090 spin_unlock_bh(&msk->pm.lock);
1095 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1096 struct mptcp_addr_info *addr)
1098 struct mptcp_sock *msk;
1099 long s_slot = 0, s_num = 0;
1100 struct mptcp_rm_list list = { .nr = 0 };
1102 pr_debug("remove_id=%d", addr->id);
1104 list.ids[list.nr++] = addr->id;
1106 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1107 struct sock *sk = (struct sock *)msk;
1108 bool remove_subflow;
1110 if (list_empty(&msk->conn_list)) {
1111 mptcp_pm_remove_anno_addr(msk, addr, false);
1116 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1117 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1119 mptcp_pm_remove_subflow(msk, &list);
1130 struct addr_entry_release_work {
1131 struct rcu_work rwork;
1132 struct mptcp_pm_addr_entry *entry;
1135 static void mptcp_pm_release_addr_entry(struct work_struct *work)
1137 struct addr_entry_release_work *w;
1138 struct mptcp_pm_addr_entry *entry;
1140 w = container_of(to_rcu_work(work), struct addr_entry_release_work, rwork);
1144 sock_release(entry->lsk);
1150 static void mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry *entry)
1152 struct addr_entry_release_work *w;
1154 w = kmalloc(sizeof(*w), GFP_ATOMIC);
1156 INIT_RCU_WORK(&w->rwork, mptcp_pm_release_addr_entry);
1158 queue_rcu_work(system_wq, &w->rwork);
1162 static int mptcp_nl_remove_id_zero_address(struct net *net,
1163 struct mptcp_addr_info *addr)
1165 struct mptcp_rm_list list = { .nr = 0 };
1166 long s_slot = 0, s_num = 0;
1167 struct mptcp_sock *msk;
1169 list.ids[list.nr++] = 0;
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;
1175 if (list_empty(&msk->conn_list))
1178 local_address((struct sock_common *)msk, &msk_local);
1179 if (!addresses_equal(&msk_local, addr, addr->port))
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);
1197 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
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;
1205 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
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.
1214 if (addr.addr.id == 0)
1215 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1217 spin_lock_bh(&pernet->lock);
1218 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1220 GENL_SET_ERR_MSG(info, "address not found");
1221 spin_unlock_bh(&pernet->lock);
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);
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);
1234 list_del_rcu(&entry->list);
1235 __clear_bit(entry->addr.id, pernet->id_bitmap);
1236 spin_unlock_bh(&pernet->lock);
1238 mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1239 mptcp_pm_free_addr_entry(entry);
1244 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1245 struct list_head *rm_list)
1247 struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1248 struct mptcp_pm_addr_entry *entry;
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;
1263 spin_lock_bh(&msk->pm.lock);
1264 mptcp_pm_remove_addr(msk, &alist);
1265 spin_unlock_bh(&msk->pm.lock);
1268 mptcp_pm_remove_subflow(msk, &slist);
1271 static void mptcp_nl_remove_addrs_list(struct net *net,
1272 struct list_head *rm_list)
1274 long s_slot = 0, s_num = 0;
1275 struct mptcp_sock *msk;
1277 if (list_empty(rm_list))
1280 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1281 struct sock *sk = (struct sock *)msk;
1284 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1292 static void __flush_addrs(struct list_head *list)
1294 while (!list_empty(list)) {
1295 struct mptcp_pm_addr_entry *cur;
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);
1304 static void __reset_counters(struct pm_nl_pernet *pernet)
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);
1312 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1314 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1315 LIST_HEAD(free_list);
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);
1328 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1329 struct mptcp_pm_addr_entry *entry)
1331 struct mptcp_addr_info *addr = &entry->addr;
1332 struct nlattr *attr;
1334 attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
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;
1350 if (addr->family == AF_INET &&
1351 nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
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;
1359 nla_nest_end(skb, attr);
1363 nla_nest_cancel(skb, attr);
1367 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
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;
1376 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1380 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1384 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1385 info->genlhdr->cmd);
1387 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1392 spin_lock_bh(&pernet->lock);
1393 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1395 GENL_SET_ERR_MSG(info, "address not found");
1400 ret = mptcp_nl_fill_addr(msg, entry);
1404 genlmsg_end(msg, reply);
1405 ret = genlmsg_reply(msg, info);
1406 spin_unlock_bh(&pernet->lock);
1410 spin_unlock_bh(&pernet->lock);
1417 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1418 struct netlink_callback *cb)
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];
1427 pernet = net_generic(net, pm_nl_pernet_id);
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);
1436 if (entry->addr.id <= id)
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);
1445 if (mptcp_nl_fill_addr(msg, entry) < 0) {
1446 genlmsg_cancel(msg, hdr);
1450 id = entry->addr.id;
1451 genlmsg_end(msg, hdr);
1454 spin_unlock_bh(&pernet->lock);
1460 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1462 struct nlattr *attr = info->attrs[id];
1467 *limit = nla_get_u32(attr);
1468 if (*limit > MPTCP_PM_ADDR_MAX) {
1469 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1476 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1478 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1479 unsigned int rcv_addrs, subflows;
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);
1488 subflows = pernet->subflows_max;
1489 ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1493 WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1494 WRITE_ONCE(pernet->subflows_max, subflows);
1497 spin_unlock_bh(&pernet->lock);
1502 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1504 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1505 struct sk_buff *msg;
1508 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1512 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1513 MPTCP_PM_CMD_GET_LIMITS);
1517 if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1518 READ_ONCE(pernet->add_addr_accept_max)))
1521 if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1522 READ_ONCE(pernet->subflows_max)))
1525 genlmsg_end(msg, reply);
1526 return genlmsg_reply(msg, info);
1529 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1534 static int mptcp_nl_addr_backup(struct net *net,
1535 struct mptcp_addr_info *addr,
1538 long s_slot = 0, s_num = 0;
1539 struct mptcp_sock *msk;
1542 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1543 struct sock *sk = (struct sock *)msk;
1545 if (list_empty(&msk->conn_list))
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);
1562 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
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);
1571 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1575 if (addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
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);
1585 entry->flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1587 entry->flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1594 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1596 genlmsg_multicast_netns(&mptcp_genl_family, net,
1597 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1600 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1602 const struct inet_sock *issk = inet_sk(ssk);
1603 const struct mptcp_subflow_context *sf;
1605 if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1608 switch (ssk->sk_family) {
1610 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1612 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1615 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1617 const struct ipv6_pinfo *np = inet6_sk(ssk);
1619 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1621 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1631 if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1633 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1636 sf = mptcp_subflow_ctx(ssk);
1637 if (WARN_ON_ONCE(!sf))
1640 if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1643 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1649 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1650 const struct mptcp_sock *msk,
1651 const struct sock *ssk)
1653 const struct sock *sk = (const struct sock *)msk;
1654 const struct mptcp_subflow_context *sf;
1657 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1660 if (mptcp_event_add_subflow(skb, ssk))
1663 sf = mptcp_subflow_ctx(ssk);
1664 if (WARN_ON_ONCE(!sf))
1667 if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1670 if (ssk->sk_bound_dev_if &&
1671 nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
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))
1682 static int mptcp_event_sub_established(struct sk_buff *skb,
1683 const struct mptcp_sock *msk,
1684 const struct sock *ssk)
1686 return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1689 static int mptcp_event_sub_closed(struct sk_buff *skb,
1690 const struct mptcp_sock *msk,
1691 const struct sock *ssk)
1693 const struct mptcp_subflow_context *sf;
1695 if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1698 sf = mptcp_subflow_ctx(ssk);
1699 if (!sf->reset_seen)
1702 if (nla_put_u32(skb, MPTCP_ATTR_RESET_REASON, sf->reset_reason))
1705 if (nla_put_u32(skb, MPTCP_ATTR_RESET_FLAGS, sf->reset_transient))
1711 static int mptcp_event_created(struct sk_buff *skb,
1712 const struct mptcp_sock *msk,
1713 const struct sock *ssk)
1715 int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1720 return mptcp_event_add_subflow(skb, ssk);
1723 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1725 struct net *net = sock_net((const struct sock *)msk);
1726 struct nlmsghdr *nlh;
1727 struct sk_buff *skb;
1729 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1732 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1736 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1738 goto nla_put_failure;
1740 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1741 goto nla_put_failure;
1743 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1744 goto nla_put_failure;
1746 genlmsg_end(skb, nlh);
1747 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1754 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1755 const struct mptcp_addr_info *info)
1757 struct net *net = sock_net((const struct sock *)msk);
1758 struct nlmsghdr *nlh;
1759 struct sk_buff *skb;
1761 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1764 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1768 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1769 MPTCP_EVENT_ANNOUNCED);
1771 goto nla_put_failure;
1773 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1774 goto nla_put_failure;
1776 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1777 goto nla_put_failure;
1779 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1780 goto nla_put_failure;
1782 switch (info->family) {
1784 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1785 goto nla_put_failure;
1787 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1789 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1790 goto nla_put_failure;
1795 goto nla_put_failure;
1798 genlmsg_end(skb, nlh);
1799 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1806 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1807 const struct sock *ssk, gfp_t gfp)
1809 struct net *net = sock_net((const struct sock *)msk);
1810 struct nlmsghdr *nlh;
1811 struct sk_buff *skb;
1813 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1816 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1820 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1822 goto nla_put_failure;
1825 case MPTCP_EVENT_UNSPEC:
1828 case MPTCP_EVENT_CREATED:
1829 case MPTCP_EVENT_ESTABLISHED:
1830 if (mptcp_event_created(skb, msk, ssk) < 0)
1831 goto nla_put_failure;
1833 case MPTCP_EVENT_CLOSED:
1834 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1835 goto nla_put_failure;
1837 case MPTCP_EVENT_ANNOUNCED:
1838 case MPTCP_EVENT_REMOVED:
1839 /* call mptcp_event_addr_announced()/removed instead */
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;
1847 case MPTCP_EVENT_SUB_CLOSED:
1848 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1849 goto nla_put_failure;
1853 genlmsg_end(skb, nlh);
1854 mptcp_nl_mcast_send(net, skb, gfp);
1861 static const struct genl_small_ops mptcp_pm_ops[] = {
1863 .cmd = MPTCP_PM_CMD_ADD_ADDR,
1864 .doit = mptcp_nl_cmd_add_addr,
1865 .flags = GENL_ADMIN_PERM,
1868 .cmd = MPTCP_PM_CMD_DEL_ADDR,
1869 .doit = mptcp_nl_cmd_del_addr,
1870 .flags = GENL_ADMIN_PERM,
1873 .cmd = MPTCP_PM_CMD_FLUSH_ADDRS,
1874 .doit = mptcp_nl_cmd_flush_addrs,
1875 .flags = GENL_ADMIN_PERM,
1878 .cmd = MPTCP_PM_CMD_GET_ADDR,
1879 .doit = mptcp_nl_cmd_get_addr,
1880 .dumpit = mptcp_nl_cmd_dump_addrs,
1883 .cmd = MPTCP_PM_CMD_SET_LIMITS,
1884 .doit = mptcp_nl_cmd_set_limits,
1885 .flags = GENL_ADMIN_PERM,
1888 .cmd = MPTCP_PM_CMD_GET_LIMITS,
1889 .doit = mptcp_nl_cmd_get_limits,
1892 .cmd = MPTCP_PM_CMD_SET_FLAGS,
1893 .doit = mptcp_nl_cmd_set_flags,
1894 .flags = GENL_ADMIN_PERM,
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,
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),
1911 static int __net_init pm_nl_init_net(struct net *net)
1913 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
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);
1923 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1927 list_for_each_entry(net, net_list, exit_list) {
1928 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1930 /* net is removed from namespace list, can't race with
1933 __flush_addrs(&pernet->local_addr_list);
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),
1944 void __init mptcp_pm_nl_init(void)
1946 if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1947 panic("Failed to register MPTCP PM pernet subsystem.\n");
1949 if (genl_register_family(&mptcp_genl_family))
1950 panic("Failed to register MPTCP PM netlink family\n");