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;
32 struct mptcp_pm_add_entry {
33 struct list_head list;
34 struct mptcp_addr_info addr;
35 struct timer_list add_timer;
36 struct mptcp_sock *sock;
40 #define MAX_ADDR_ID 255
41 #define BITMAP_SZ DIV_ROUND_UP(MAX_ADDR_ID + 1, BITS_PER_LONG)
44 /* protects pernet updates */
46 struct list_head local_addr_list;
48 unsigned int add_addr_signal_max;
49 unsigned int add_addr_accept_max;
50 unsigned int local_addr_max;
51 unsigned int subflows_max;
53 unsigned long id_bitmap[BITMAP_SZ];
56 #define MPTCP_PM_ADDR_MAX 8
57 #define ADD_ADDR_RETRANS_MAX 3
59 static bool addresses_equal(const struct mptcp_addr_info *a,
60 struct mptcp_addr_info *b, bool use_port)
62 bool addr_equals = false;
64 if (a->family == b->family) {
65 if (a->family == AF_INET)
66 addr_equals = a->addr.s_addr == b->addr.s_addr;
67 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
69 addr_equals = !ipv6_addr_cmp(&a->addr6, &b->addr6);
70 } else if (a->family == AF_INET) {
71 if (ipv6_addr_v4mapped(&b->addr6))
72 addr_equals = a->addr.s_addr == b->addr6.s6_addr32[3];
73 } else if (b->family == AF_INET) {
74 if (ipv6_addr_v4mapped(&a->addr6))
75 addr_equals = a->addr6.s6_addr32[3] == b->addr.s_addr;
84 return a->port == b->port;
87 static bool address_zero(const struct mptcp_addr_info *addr)
89 struct mptcp_addr_info zero;
91 memset(&zero, 0, sizeof(zero));
92 zero.family = addr->family;
94 return addresses_equal(addr, &zero, true);
97 static void local_address(const struct sock_common *skc,
98 struct mptcp_addr_info *addr)
100 addr->family = skc->skc_family;
101 addr->port = htons(skc->skc_num);
102 if (addr->family == AF_INET)
103 addr->addr.s_addr = skc->skc_rcv_saddr;
104 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
105 else if (addr->family == AF_INET6)
106 addr->addr6 = skc->skc_v6_rcv_saddr;
110 static void remote_address(const struct sock_common *skc,
111 struct mptcp_addr_info *addr)
113 addr->family = skc->skc_family;
114 addr->port = skc->skc_dport;
115 if (addr->family == AF_INET)
116 addr->addr.s_addr = skc->skc_daddr;
117 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
118 else if (addr->family == AF_INET6)
119 addr->addr6 = skc->skc_v6_daddr;
123 static bool lookup_subflow_by_saddr(const struct list_head *list,
124 struct mptcp_addr_info *saddr)
126 struct mptcp_subflow_context *subflow;
127 struct mptcp_addr_info cur;
128 struct sock_common *skc;
130 list_for_each_entry(subflow, list, node) {
131 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
133 local_address(skc, &cur);
134 if (addresses_equal(&cur, saddr, saddr->port))
141 static bool lookup_subflow_by_daddr(const struct list_head *list,
142 struct mptcp_addr_info *daddr)
144 struct mptcp_subflow_context *subflow;
145 struct mptcp_addr_info cur;
146 struct sock_common *skc;
148 list_for_each_entry(subflow, list, node) {
149 skc = (struct sock_common *)mptcp_subflow_tcp_sock(subflow);
151 remote_address(skc, &cur);
152 if (addresses_equal(&cur, daddr, daddr->port))
159 static struct mptcp_pm_addr_entry *
160 select_local_address(const struct pm_nl_pernet *pernet,
161 struct mptcp_sock *msk)
163 struct mptcp_pm_addr_entry *entry, *ret = NULL;
164 struct sock *sk = (struct sock *)msk;
166 msk_owned_by_me(msk);
169 __mptcp_flush_join_list(msk);
170 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
171 if (!(entry->addr.flags & MPTCP_PM_ADDR_FLAG_SUBFLOW))
174 if (entry->addr.family != sk->sk_family) {
175 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
176 if ((entry->addr.family == AF_INET &&
177 !ipv6_addr_v4mapped(&sk->sk_v6_daddr)) ||
178 (sk->sk_family == AF_INET &&
179 !ipv6_addr_v4mapped(&entry->addr.addr6)))
184 /* avoid any address already in use by subflows and
187 if (!lookup_subflow_by_saddr(&msk->conn_list, &entry->addr)) {
196 static struct mptcp_pm_addr_entry *
197 select_signal_address(struct pm_nl_pernet *pernet, unsigned int pos)
199 struct mptcp_pm_addr_entry *entry, *ret = NULL;
203 /* do not keep any additional per socket state, just signal
204 * the address list in order.
205 * Note: removal from the local address list during the msk life-cycle
206 * can lead to additional addresses not being announced.
208 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
209 if (!(entry->addr.flags & MPTCP_PM_ADDR_FLAG_SIGNAL))
220 unsigned int mptcp_pm_get_add_addr_signal_max(struct mptcp_sock *msk)
222 struct pm_nl_pernet *pernet;
224 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
225 return READ_ONCE(pernet->add_addr_signal_max);
227 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_signal_max);
229 unsigned int mptcp_pm_get_add_addr_accept_max(struct mptcp_sock *msk)
231 struct pm_nl_pernet *pernet;
233 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
234 return READ_ONCE(pernet->add_addr_accept_max);
236 EXPORT_SYMBOL_GPL(mptcp_pm_get_add_addr_accept_max);
238 unsigned int mptcp_pm_get_subflows_max(struct mptcp_sock *msk)
240 struct pm_nl_pernet *pernet;
242 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
243 return READ_ONCE(pernet->subflows_max);
245 EXPORT_SYMBOL_GPL(mptcp_pm_get_subflows_max);
247 unsigned int mptcp_pm_get_local_addr_max(struct mptcp_sock *msk)
249 struct pm_nl_pernet *pernet;
251 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
252 return READ_ONCE(pernet->local_addr_max);
254 EXPORT_SYMBOL_GPL(mptcp_pm_get_local_addr_max);
256 static void check_work_pending(struct mptcp_sock *msk)
258 if (msk->pm.add_addr_signaled == mptcp_pm_get_add_addr_signal_max(msk) &&
259 (msk->pm.local_addr_used == mptcp_pm_get_local_addr_max(msk) ||
260 msk->pm.subflows == mptcp_pm_get_subflows_max(msk)))
261 WRITE_ONCE(msk->pm.work_pending, false);
264 struct mptcp_pm_add_entry *
265 mptcp_lookup_anno_list_by_saddr(struct mptcp_sock *msk,
266 struct mptcp_addr_info *addr)
268 struct mptcp_pm_add_entry *entry;
270 lockdep_assert_held(&msk->pm.lock);
272 list_for_each_entry(entry, &msk->pm.anno_list, list) {
273 if (addresses_equal(&entry->addr, addr, true))
280 bool mptcp_pm_sport_in_anno_list(struct mptcp_sock *msk, const struct sock *sk)
282 struct mptcp_pm_add_entry *entry;
283 struct mptcp_addr_info saddr;
286 local_address((struct sock_common *)sk, &saddr);
288 spin_lock_bh(&msk->pm.lock);
289 list_for_each_entry(entry, &msk->pm.anno_list, list) {
290 if (addresses_equal(&entry->addr, &saddr, true)) {
297 spin_unlock_bh(&msk->pm.lock);
301 static void mptcp_pm_add_timer(struct timer_list *timer)
303 struct mptcp_pm_add_entry *entry = from_timer(entry, timer, add_timer);
304 struct mptcp_sock *msk = entry->sock;
305 struct sock *sk = (struct sock *)msk;
307 pr_debug("msk=%p", msk);
312 if (inet_sk_state_load(sk) == TCP_CLOSE)
318 if (mptcp_pm_should_add_signal(msk)) {
319 sk_reset_timer(sk, timer, jiffies + TCP_RTO_MAX / 8);
323 spin_lock_bh(&msk->pm.lock);
325 if (!mptcp_pm_should_add_signal(msk)) {
326 pr_debug("retransmit ADD_ADDR id=%d", entry->addr.id);
327 mptcp_pm_announce_addr(msk, &entry->addr, false);
328 mptcp_pm_add_addr_send_ack(msk);
329 entry->retrans_times++;
332 if (entry->retrans_times < ADD_ADDR_RETRANS_MAX)
333 sk_reset_timer(sk, timer,
334 jiffies + mptcp_get_add_addr_timeout(sock_net(sk)));
336 spin_unlock_bh(&msk->pm.lock);
338 if (entry->retrans_times == ADD_ADDR_RETRANS_MAX)
339 mptcp_pm_subflow_established(msk);
345 struct mptcp_pm_add_entry *
346 mptcp_pm_del_add_timer(struct mptcp_sock *msk,
347 struct mptcp_addr_info *addr)
349 struct mptcp_pm_add_entry *entry;
350 struct sock *sk = (struct sock *)msk;
352 spin_lock_bh(&msk->pm.lock);
353 entry = mptcp_lookup_anno_list_by_saddr(msk, addr);
355 entry->retrans_times = ADD_ADDR_RETRANS_MAX;
356 spin_unlock_bh(&msk->pm.lock);
359 sk_stop_timer_sync(sk, &entry->add_timer);
364 static bool mptcp_pm_alloc_anno_list(struct mptcp_sock *msk,
365 struct mptcp_pm_addr_entry *entry)
367 struct mptcp_pm_add_entry *add_entry = NULL;
368 struct sock *sk = (struct sock *)msk;
369 struct net *net = sock_net(sk);
371 lockdep_assert_held(&msk->pm.lock);
373 if (mptcp_lookup_anno_list_by_saddr(msk, &entry->addr))
376 add_entry = kmalloc(sizeof(*add_entry), GFP_ATOMIC);
380 list_add(&add_entry->list, &msk->pm.anno_list);
382 add_entry->addr = entry->addr;
383 add_entry->sock = msk;
384 add_entry->retrans_times = 0;
386 timer_setup(&add_entry->add_timer, mptcp_pm_add_timer, 0);
387 sk_reset_timer(sk, &add_entry->add_timer,
388 jiffies + mptcp_get_add_addr_timeout(net));
393 void mptcp_pm_free_anno_list(struct mptcp_sock *msk)
395 struct mptcp_pm_add_entry *entry, *tmp;
396 struct sock *sk = (struct sock *)msk;
397 LIST_HEAD(free_list);
399 pr_debug("msk=%p", msk);
401 spin_lock_bh(&msk->pm.lock);
402 list_splice_init(&msk->pm.anno_list, &free_list);
403 spin_unlock_bh(&msk->pm.lock);
405 list_for_each_entry_safe(entry, tmp, &free_list, list) {
406 sk_stop_timer_sync(sk, &entry->add_timer);
411 static void mptcp_pm_create_subflow_or_signal_addr(struct mptcp_sock *msk)
413 struct sock *sk = (struct sock *)msk;
414 struct mptcp_pm_addr_entry *local;
415 unsigned int add_addr_signal_max;
416 unsigned int local_addr_max;
417 struct pm_nl_pernet *pernet;
418 unsigned int subflows_max;
420 pernet = net_generic(sock_net(sk), pm_nl_pernet_id);
422 add_addr_signal_max = mptcp_pm_get_add_addr_signal_max(msk);
423 local_addr_max = mptcp_pm_get_local_addr_max(msk);
424 subflows_max = mptcp_pm_get_subflows_max(msk);
426 pr_debug("local %d:%d signal %d:%d subflows %d:%d\n",
427 msk->pm.local_addr_used, local_addr_max,
428 msk->pm.add_addr_signaled, add_addr_signal_max,
429 msk->pm.subflows, subflows_max);
431 /* check first for announce */
432 if (msk->pm.add_addr_signaled < add_addr_signal_max) {
433 local = select_signal_address(pernet,
434 msk->pm.add_addr_signaled);
437 if (mptcp_pm_alloc_anno_list(msk, local)) {
438 msk->pm.add_addr_signaled++;
439 mptcp_pm_announce_addr(msk, &local->addr, false);
440 mptcp_pm_nl_addr_send_ack(msk);
443 /* pick failed, avoid fourther attempts later */
444 msk->pm.local_addr_used = add_addr_signal_max;
447 check_work_pending(msk);
450 /* check if should create a new subflow */
451 if (msk->pm.local_addr_used < local_addr_max &&
452 msk->pm.subflows < subflows_max) {
453 local = select_local_address(pernet, msk);
455 struct mptcp_addr_info remote = { 0 };
457 msk->pm.local_addr_used++;
459 check_work_pending(msk);
460 remote_address((struct sock_common *)sk, &remote);
461 spin_unlock_bh(&msk->pm.lock);
462 __mptcp_subflow_connect(sk, &local->addr, &remote);
463 spin_lock_bh(&msk->pm.lock);
467 /* lookup failed, avoid fourther attempts later */
468 msk->pm.local_addr_used = local_addr_max;
469 check_work_pending(msk);
473 static void mptcp_pm_nl_fully_established(struct mptcp_sock *msk)
475 mptcp_pm_create_subflow_or_signal_addr(msk);
478 static void mptcp_pm_nl_subflow_established(struct mptcp_sock *msk)
480 mptcp_pm_create_subflow_or_signal_addr(msk);
483 static void mptcp_pm_nl_add_addr_received(struct mptcp_sock *msk)
485 struct sock *sk = (struct sock *)msk;
486 unsigned int add_addr_accept_max;
487 struct mptcp_addr_info remote;
488 struct mptcp_addr_info local;
489 unsigned int subflows_max;
491 add_addr_accept_max = mptcp_pm_get_add_addr_accept_max(msk);
492 subflows_max = mptcp_pm_get_subflows_max(msk);
494 pr_debug("accepted %d:%d remote family %d",
495 msk->pm.add_addr_accepted, add_addr_accept_max,
496 msk->pm.remote.family);
498 if (lookup_subflow_by_daddr(&msk->conn_list, &msk->pm.remote))
501 msk->pm.add_addr_accepted++;
503 if (msk->pm.add_addr_accepted >= add_addr_accept_max ||
504 msk->pm.subflows >= subflows_max)
505 WRITE_ONCE(msk->pm.accept_addr, false);
507 /* connect to the specified remote address, using whatever
508 * local address the routing configuration will pick.
510 remote = msk->pm.remote;
512 remote.port = sk->sk_dport;
513 memset(&local, 0, sizeof(local));
514 local.family = remote.family;
516 spin_unlock_bh(&msk->pm.lock);
517 __mptcp_subflow_connect(sk, &local, &remote);
518 spin_lock_bh(&msk->pm.lock);
521 mptcp_pm_announce_addr(msk, &msk->pm.remote, true);
522 mptcp_pm_nl_addr_send_ack(msk);
525 void mptcp_pm_nl_addr_send_ack(struct mptcp_sock *msk)
527 struct mptcp_subflow_context *subflow;
529 msk_owned_by_me(msk);
530 lockdep_assert_held(&msk->pm.lock);
532 if (!mptcp_pm_should_add_signal(msk) &&
533 !mptcp_pm_should_rm_signal(msk))
536 __mptcp_flush_join_list(msk);
537 subflow = list_first_entry_or_null(&msk->conn_list, typeof(*subflow), node);
539 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
541 spin_unlock_bh(&msk->pm.lock);
542 pr_debug("send ack for %s%s%s",
543 mptcp_pm_should_add_signal(msk) ? "add_addr" : "rm_addr",
544 mptcp_pm_should_add_signal_ipv6(msk) ? " [ipv6]" : "",
545 mptcp_pm_should_add_signal_port(msk) ? " [port]" : "");
550 spin_lock_bh(&msk->pm.lock);
554 int mptcp_pm_nl_mp_prio_send_ack(struct mptcp_sock *msk,
555 struct mptcp_addr_info *addr,
558 struct mptcp_subflow_context *subflow;
560 pr_debug("bkup=%d", bkup);
562 mptcp_for_each_subflow(msk, subflow) {
563 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
564 struct sock *sk = (struct sock *)msk;
565 struct mptcp_addr_info local;
567 local_address((struct sock_common *)ssk, &local);
568 if (!addresses_equal(&local, addr, addr->port))
571 subflow->backup = bkup;
572 subflow->send_mp_prio = 1;
573 subflow->request_bkup = bkup;
574 __MPTCP_INC_STATS(sock_net(sk), MPTCP_MIB_MPPRIOTX);
576 spin_unlock_bh(&msk->pm.lock);
577 pr_debug("send ack for mp_prio");
581 spin_lock_bh(&msk->pm.lock);
589 static void mptcp_pm_nl_rm_addr_or_subflow(struct mptcp_sock *msk,
590 const struct mptcp_rm_list *rm_list,
591 enum linux_mptcp_mib_field rm_type)
593 struct mptcp_subflow_context *subflow, *tmp;
594 struct sock *sk = (struct sock *)msk;
597 pr_debug("%s rm_list_nr %d",
598 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow", rm_list->nr);
600 msk_owned_by_me(msk);
605 if (list_empty(&msk->conn_list))
608 for (i = 0; i < rm_list->nr; i++) {
609 list_for_each_entry_safe(subflow, tmp, &msk->conn_list, node) {
610 struct sock *ssk = mptcp_subflow_tcp_sock(subflow);
611 int how = RCV_SHUTDOWN | SEND_SHUTDOWN;
612 u8 id = subflow->local_id;
614 if (rm_type == MPTCP_MIB_RMADDR)
615 id = subflow->remote_id;
617 if (rm_list->ids[i] != id)
620 pr_debug(" -> %s rm_list_ids[%d]=%u local_id=%u remote_id=%u",
621 rm_type == MPTCP_MIB_RMADDR ? "address" : "subflow",
622 i, rm_list->ids[i], subflow->local_id, subflow->remote_id);
623 spin_unlock_bh(&msk->pm.lock);
624 mptcp_subflow_shutdown(sk, ssk, how);
625 mptcp_close_ssk(sk, ssk, subflow);
626 spin_lock_bh(&msk->pm.lock);
628 if (rm_type == MPTCP_MIB_RMADDR) {
629 msk->pm.add_addr_accepted--;
630 WRITE_ONCE(msk->pm.accept_addr, true);
631 } else if (rm_type == MPTCP_MIB_RMSUBFLOW) {
632 msk->pm.local_addr_used--;
635 __MPTCP_INC_STATS(sock_net(sk), rm_type);
640 static void mptcp_pm_nl_rm_addr_received(struct mptcp_sock *msk)
642 mptcp_pm_nl_rm_addr_or_subflow(msk, &msk->pm.rm_list_rx, MPTCP_MIB_RMADDR);
645 void mptcp_pm_nl_rm_subflow_received(struct mptcp_sock *msk,
646 const struct mptcp_rm_list *rm_list)
648 mptcp_pm_nl_rm_addr_or_subflow(msk, rm_list, MPTCP_MIB_RMSUBFLOW);
651 void mptcp_pm_nl_work(struct mptcp_sock *msk)
653 struct mptcp_pm_data *pm = &msk->pm;
655 msk_owned_by_me(msk);
657 spin_lock_bh(&msk->pm.lock);
659 pr_debug("msk=%p status=%x", msk, pm->status);
660 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_RECEIVED)) {
661 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_RECEIVED);
662 mptcp_pm_nl_add_addr_received(msk);
664 if (pm->status & BIT(MPTCP_PM_ADD_ADDR_SEND_ACK)) {
665 pm->status &= ~BIT(MPTCP_PM_ADD_ADDR_SEND_ACK);
666 mptcp_pm_nl_addr_send_ack(msk);
668 if (pm->status & BIT(MPTCP_PM_RM_ADDR_RECEIVED)) {
669 pm->status &= ~BIT(MPTCP_PM_RM_ADDR_RECEIVED);
670 mptcp_pm_nl_rm_addr_received(msk);
672 if (pm->status & BIT(MPTCP_PM_ESTABLISHED)) {
673 pm->status &= ~BIT(MPTCP_PM_ESTABLISHED);
674 mptcp_pm_nl_fully_established(msk);
676 if (pm->status & BIT(MPTCP_PM_SUBFLOW_ESTABLISHED)) {
677 pm->status &= ~BIT(MPTCP_PM_SUBFLOW_ESTABLISHED);
678 mptcp_pm_nl_subflow_established(msk);
681 spin_unlock_bh(&msk->pm.lock);
684 static bool address_use_port(struct mptcp_pm_addr_entry *entry)
686 return (entry->addr.flags &
687 (MPTCP_PM_ADDR_FLAG_SIGNAL | MPTCP_PM_ADDR_FLAG_SUBFLOW)) ==
688 MPTCP_PM_ADDR_FLAG_SIGNAL;
691 static int mptcp_pm_nl_append_new_local_addr(struct pm_nl_pernet *pernet,
692 struct mptcp_pm_addr_entry *entry)
694 struct mptcp_pm_addr_entry *cur;
695 unsigned int addr_max;
698 spin_lock_bh(&pernet->lock);
699 /* to keep the code simple, don't do IDR-like allocation for address ID,
700 * just bail when we exceed limits
702 if (pernet->next_id == MAX_ADDR_ID)
704 if (pernet->addrs >= MPTCP_PM_ADDR_MAX)
706 if (test_bit(entry->addr.id, pernet->id_bitmap))
709 /* do not insert duplicate address, differentiate on port only
712 list_for_each_entry(cur, &pernet->local_addr_list, list) {
713 if (addresses_equal(&cur->addr, &entry->addr,
714 address_use_port(entry) &&
715 address_use_port(cur)))
719 if (!entry->addr.id) {
721 entry->addr.id = find_next_zero_bit(pernet->id_bitmap,
724 if ((!entry->addr.id || entry->addr.id > MAX_ADDR_ID) &&
725 pernet->next_id != 1) {
731 if (!entry->addr.id || entry->addr.id > MAX_ADDR_ID)
734 __set_bit(entry->addr.id, pernet->id_bitmap);
735 if (entry->addr.id > pernet->next_id)
736 pernet->next_id = entry->addr.id;
738 if (entry->addr.flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
739 addr_max = pernet->add_addr_signal_max;
740 WRITE_ONCE(pernet->add_addr_signal_max, addr_max + 1);
742 if (entry->addr.flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
743 addr_max = pernet->local_addr_max;
744 WRITE_ONCE(pernet->local_addr_max, addr_max + 1);
748 list_add_tail_rcu(&entry->list, &pernet->local_addr_list);
749 ret = entry->addr.id;
752 spin_unlock_bh(&pernet->lock);
756 static int mptcp_pm_nl_create_listen_socket(struct sock *sk,
757 struct mptcp_pm_addr_entry *entry)
759 struct sockaddr_storage addr;
760 struct mptcp_sock *msk;
761 struct socket *ssock;
765 err = sock_create_kern(sock_net(sk), entry->addr.family,
766 SOCK_STREAM, IPPROTO_MPTCP, &entry->lsk);
770 msk = mptcp_sk(entry->lsk->sk);
776 ssock = __mptcp_nmpc_socket(msk);
782 mptcp_info2sockaddr(&entry->addr, &addr, entry->addr.family);
783 err = kernel_bind(ssock, (struct sockaddr *)&addr,
784 sizeof(struct sockaddr_in));
786 pr_warn("kernel_bind error, err=%d", err);
790 err = kernel_listen(ssock, backlog);
792 pr_warn("kernel_listen error, err=%d", err);
799 sock_release(entry->lsk);
803 int mptcp_pm_nl_get_local_id(struct mptcp_sock *msk, struct sock_common *skc)
805 struct mptcp_pm_addr_entry *entry;
806 struct mptcp_addr_info skc_local;
807 struct mptcp_addr_info msk_local;
808 struct pm_nl_pernet *pernet;
811 if (WARN_ON_ONCE(!msk))
814 /* The 0 ID mapping is defined by the first subflow, copied into the msk
817 local_address((struct sock_common *)msk, &msk_local);
818 local_address((struct sock_common *)skc, &skc_local);
819 if (addresses_equal(&msk_local, &skc_local, false))
822 if (address_zero(&skc_local))
825 pernet = net_generic(sock_net((struct sock *)msk), pm_nl_pernet_id);
828 list_for_each_entry_rcu(entry, &pernet->local_addr_list, list) {
829 if (addresses_equal(&entry->addr, &skc_local, entry->addr.port)) {
830 ret = entry->addr.id;
838 /* address not found, add to local list */
839 entry = kmalloc(sizeof(*entry), GFP_ATOMIC);
843 entry->addr = skc_local;
844 entry->addr.ifindex = 0;
845 entry->addr.flags = 0;
847 entry->addr.port = 0;
849 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
856 void mptcp_pm_nl_data_init(struct mptcp_sock *msk)
858 struct mptcp_pm_data *pm = &msk->pm;
861 subflows = !!mptcp_pm_get_subflows_max(msk);
862 WRITE_ONCE(pm->work_pending, (!!mptcp_pm_get_local_addr_max(msk) && subflows) ||
863 !!mptcp_pm_get_add_addr_signal_max(msk));
864 WRITE_ONCE(pm->accept_addr, !!mptcp_pm_get_add_addr_accept_max(msk) && subflows);
865 WRITE_ONCE(pm->accept_subflow, subflows);
868 #define MPTCP_PM_CMD_GRP_OFFSET 0
869 #define MPTCP_PM_EV_GRP_OFFSET 1
871 static const struct genl_multicast_group mptcp_pm_mcgrps[] = {
872 [MPTCP_PM_CMD_GRP_OFFSET] = { .name = MPTCP_PM_CMD_GRP_NAME, },
873 [MPTCP_PM_EV_GRP_OFFSET] = { .name = MPTCP_PM_EV_GRP_NAME,
874 .flags = GENL_UNS_ADMIN_PERM,
878 static const struct nla_policy
879 mptcp_pm_addr_policy[MPTCP_PM_ADDR_ATTR_MAX + 1] = {
880 [MPTCP_PM_ADDR_ATTR_FAMILY] = { .type = NLA_U16, },
881 [MPTCP_PM_ADDR_ATTR_ID] = { .type = NLA_U8, },
882 [MPTCP_PM_ADDR_ATTR_ADDR4] = { .type = NLA_U32, },
883 [MPTCP_PM_ADDR_ATTR_ADDR6] =
884 NLA_POLICY_EXACT_LEN(sizeof(struct in6_addr)),
885 [MPTCP_PM_ADDR_ATTR_PORT] = { .type = NLA_U16 },
886 [MPTCP_PM_ADDR_ATTR_FLAGS] = { .type = NLA_U32 },
887 [MPTCP_PM_ADDR_ATTR_IF_IDX] = { .type = NLA_S32 },
890 static const struct nla_policy mptcp_pm_policy[MPTCP_PM_ATTR_MAX + 1] = {
891 [MPTCP_PM_ATTR_ADDR] =
892 NLA_POLICY_NESTED(mptcp_pm_addr_policy),
893 [MPTCP_PM_ATTR_RCV_ADD_ADDRS] = { .type = NLA_U32, },
894 [MPTCP_PM_ATTR_SUBFLOWS] = { .type = NLA_U32, },
897 static int mptcp_pm_family_to_addr(int family)
899 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
900 if (family == AF_INET6)
901 return MPTCP_PM_ADDR_ATTR_ADDR6;
903 return MPTCP_PM_ADDR_ATTR_ADDR4;
906 static int mptcp_pm_parse_addr(struct nlattr *attr, struct genl_info *info,
908 struct mptcp_pm_addr_entry *entry)
910 struct nlattr *tb[MPTCP_PM_ADDR_ATTR_MAX + 1];
914 GENL_SET_ERR_MSG(info, "missing address info");
918 /* no validation needed - was already done via nested policy */
919 err = nla_parse_nested_deprecated(tb, MPTCP_PM_ADDR_ATTR_MAX, attr,
920 mptcp_pm_addr_policy, info->extack);
924 memset(entry, 0, sizeof(*entry));
925 if (!tb[MPTCP_PM_ADDR_ATTR_FAMILY]) {
929 NL_SET_ERR_MSG_ATTR(info->extack, attr,
934 entry->addr.family = nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_FAMILY]);
935 if (entry->addr.family != AF_INET
936 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
937 && entry->addr.family != AF_INET6
940 NL_SET_ERR_MSG_ATTR(info->extack, attr,
941 "unknown address family");
944 addr_addr = mptcp_pm_family_to_addr(entry->addr.family);
945 if (!tb[addr_addr]) {
946 NL_SET_ERR_MSG_ATTR(info->extack, attr,
947 "missing address data");
951 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
952 if (entry->addr.family == AF_INET6)
953 entry->addr.addr6 = nla_get_in6_addr(tb[addr_addr]);
956 entry->addr.addr.s_addr = nla_get_in_addr(tb[addr_addr]);
959 if (tb[MPTCP_PM_ADDR_ATTR_IF_IDX]) {
960 u32 val = nla_get_s32(tb[MPTCP_PM_ADDR_ATTR_IF_IDX]);
962 entry->addr.ifindex = val;
965 if (tb[MPTCP_PM_ADDR_ATTR_ID])
966 entry->addr.id = nla_get_u8(tb[MPTCP_PM_ADDR_ATTR_ID]);
968 if (tb[MPTCP_PM_ADDR_ATTR_FLAGS])
969 entry->addr.flags = nla_get_u32(tb[MPTCP_PM_ADDR_ATTR_FLAGS]);
971 if (tb[MPTCP_PM_ADDR_ATTR_PORT])
972 entry->addr.port = htons(nla_get_u16(tb[MPTCP_PM_ADDR_ATTR_PORT]));
977 static struct pm_nl_pernet *genl_info_pm_nl(struct genl_info *info)
979 return net_generic(genl_info_net(info), pm_nl_pernet_id);
982 static int mptcp_nl_add_subflow_or_signal_addr(struct net *net)
984 struct mptcp_sock *msk;
985 long s_slot = 0, s_num = 0;
987 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
988 struct sock *sk = (struct sock *)msk;
990 if (!READ_ONCE(msk->fully_established))
994 spin_lock_bh(&msk->pm.lock);
995 mptcp_pm_create_subflow_or_signal_addr(msk);
996 spin_unlock_bh(&msk->pm.lock);
1007 static int mptcp_nl_cmd_add_addr(struct sk_buff *skb, struct genl_info *info)
1009 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1010 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1011 struct mptcp_pm_addr_entry addr, *entry;
1014 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1018 entry = kmalloc(sizeof(*entry), GFP_KERNEL);
1020 GENL_SET_ERR_MSG(info, "can't allocate addr");
1025 if (entry->addr.port) {
1026 ret = mptcp_pm_nl_create_listen_socket(skb->sk, entry);
1028 GENL_SET_ERR_MSG(info, "create listen socket error");
1033 ret = mptcp_pm_nl_append_new_local_addr(pernet, entry);
1035 GENL_SET_ERR_MSG(info, "too many addresses or duplicate one");
1037 sock_release(entry->lsk);
1042 mptcp_nl_add_subflow_or_signal_addr(sock_net(skb->sk));
1047 static struct mptcp_pm_addr_entry *
1048 __lookup_addr_by_id(struct pm_nl_pernet *pernet, unsigned int id)
1050 struct mptcp_pm_addr_entry *entry;
1052 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1053 if (entry->addr.id == id)
1059 static bool remove_anno_list_by_saddr(struct mptcp_sock *msk,
1060 struct mptcp_addr_info *addr)
1062 struct mptcp_pm_add_entry *entry;
1064 entry = mptcp_pm_del_add_timer(msk, addr);
1066 list_del(&entry->list);
1074 static bool mptcp_pm_remove_anno_addr(struct mptcp_sock *msk,
1075 struct mptcp_addr_info *addr,
1078 struct mptcp_rm_list list = { .nr = 0 };
1081 list.ids[list.nr++] = addr->id;
1083 ret = remove_anno_list_by_saddr(msk, addr);
1085 spin_lock_bh(&msk->pm.lock);
1086 mptcp_pm_remove_addr(msk, &list);
1087 spin_unlock_bh(&msk->pm.lock);
1092 static int mptcp_nl_remove_subflow_and_signal_addr(struct net *net,
1093 struct mptcp_addr_info *addr)
1095 struct mptcp_sock *msk;
1096 long s_slot = 0, s_num = 0;
1097 struct mptcp_rm_list list = { .nr = 0 };
1099 pr_debug("remove_id=%d", addr->id);
1101 list.ids[list.nr++] = addr->id;
1103 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1104 struct sock *sk = (struct sock *)msk;
1105 bool remove_subflow;
1107 if (list_empty(&msk->conn_list)) {
1108 mptcp_pm_remove_anno_addr(msk, addr, false);
1113 remove_subflow = lookup_subflow_by_saddr(&msk->conn_list, addr);
1114 mptcp_pm_remove_anno_addr(msk, addr, remove_subflow);
1116 mptcp_pm_remove_subflow(msk, &list);
1127 struct addr_entry_release_work {
1128 struct rcu_work rwork;
1129 struct mptcp_pm_addr_entry *entry;
1132 static void mptcp_pm_release_addr_entry(struct work_struct *work)
1134 struct addr_entry_release_work *w;
1135 struct mptcp_pm_addr_entry *entry;
1137 w = container_of(to_rcu_work(work), struct addr_entry_release_work, rwork);
1141 sock_release(entry->lsk);
1147 static void mptcp_pm_free_addr_entry(struct mptcp_pm_addr_entry *entry)
1149 struct addr_entry_release_work *w;
1151 w = kmalloc(sizeof(*w), GFP_ATOMIC);
1153 INIT_RCU_WORK(&w->rwork, mptcp_pm_release_addr_entry);
1155 queue_rcu_work(system_wq, &w->rwork);
1159 static int mptcp_nl_remove_id_zero_address(struct net *net,
1160 struct mptcp_addr_info *addr)
1162 struct mptcp_rm_list list = { .nr = 0 };
1163 long s_slot = 0, s_num = 0;
1164 struct mptcp_sock *msk;
1166 list.ids[list.nr++] = 0;
1168 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1169 struct sock *sk = (struct sock *)msk;
1170 struct mptcp_addr_info msk_local;
1172 if (list_empty(&msk->conn_list))
1175 local_address((struct sock_common *)msk, &msk_local);
1176 if (!addresses_equal(&msk_local, addr, addr->port))
1180 spin_lock_bh(&msk->pm.lock);
1181 mptcp_pm_remove_addr(msk, &list);
1182 mptcp_pm_nl_rm_subflow_received(msk, &list);
1183 spin_unlock_bh(&msk->pm.lock);
1194 static int mptcp_nl_cmd_del_addr(struct sk_buff *skb, struct genl_info *info)
1196 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1197 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1198 struct mptcp_pm_addr_entry addr, *entry;
1199 unsigned int addr_max;
1202 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1206 /* the zero id address is special: the first address used by the msk
1207 * always gets such an id, so different subflows can have different zero
1208 * id addresses. Additionally zero id is not accounted for in id_bitmap.
1209 * Let's use an 'mptcp_rm_list' instead of the common remove code.
1211 if (addr.addr.id == 0)
1212 return mptcp_nl_remove_id_zero_address(sock_net(skb->sk), &addr.addr);
1214 spin_lock_bh(&pernet->lock);
1215 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1217 GENL_SET_ERR_MSG(info, "address not found");
1218 spin_unlock_bh(&pernet->lock);
1221 if (entry->addr.flags & MPTCP_PM_ADDR_FLAG_SIGNAL) {
1222 addr_max = pernet->add_addr_signal_max;
1223 WRITE_ONCE(pernet->add_addr_signal_max, addr_max - 1);
1225 if (entry->addr.flags & MPTCP_PM_ADDR_FLAG_SUBFLOW) {
1226 addr_max = pernet->local_addr_max;
1227 WRITE_ONCE(pernet->local_addr_max, addr_max - 1);
1231 list_del_rcu(&entry->list);
1232 __clear_bit(entry->addr.id, pernet->id_bitmap);
1233 spin_unlock_bh(&pernet->lock);
1235 mptcp_nl_remove_subflow_and_signal_addr(sock_net(skb->sk), &entry->addr);
1236 mptcp_pm_free_addr_entry(entry);
1241 static void mptcp_pm_remove_addrs_and_subflows(struct mptcp_sock *msk,
1242 struct list_head *rm_list)
1244 struct mptcp_rm_list alist = { .nr = 0 }, slist = { .nr = 0 };
1245 struct mptcp_pm_addr_entry *entry;
1247 list_for_each_entry(entry, rm_list, list) {
1248 if (lookup_subflow_by_saddr(&msk->conn_list, &entry->addr) &&
1249 alist.nr < MPTCP_RM_IDS_MAX &&
1250 slist.nr < MPTCP_RM_IDS_MAX) {
1251 alist.ids[alist.nr++] = entry->addr.id;
1252 slist.ids[slist.nr++] = entry->addr.id;
1253 } else if (remove_anno_list_by_saddr(msk, &entry->addr) &&
1254 alist.nr < MPTCP_RM_IDS_MAX) {
1255 alist.ids[alist.nr++] = entry->addr.id;
1260 spin_lock_bh(&msk->pm.lock);
1261 mptcp_pm_remove_addr(msk, &alist);
1262 spin_unlock_bh(&msk->pm.lock);
1265 mptcp_pm_remove_subflow(msk, &slist);
1268 static void mptcp_nl_remove_addrs_list(struct net *net,
1269 struct list_head *rm_list)
1271 long s_slot = 0, s_num = 0;
1272 struct mptcp_sock *msk;
1274 if (list_empty(rm_list))
1277 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1278 struct sock *sk = (struct sock *)msk;
1281 mptcp_pm_remove_addrs_and_subflows(msk, rm_list);
1289 static void __flush_addrs(struct list_head *list)
1291 while (!list_empty(list)) {
1292 struct mptcp_pm_addr_entry *cur;
1294 cur = list_entry(list->next,
1295 struct mptcp_pm_addr_entry, list);
1296 list_del_rcu(&cur->list);
1297 mptcp_pm_free_addr_entry(cur);
1301 static void __reset_counters(struct pm_nl_pernet *pernet)
1303 WRITE_ONCE(pernet->add_addr_signal_max, 0);
1304 WRITE_ONCE(pernet->add_addr_accept_max, 0);
1305 WRITE_ONCE(pernet->local_addr_max, 0);
1309 static int mptcp_nl_cmd_flush_addrs(struct sk_buff *skb, struct genl_info *info)
1311 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1312 LIST_HEAD(free_list);
1314 spin_lock_bh(&pernet->lock);
1315 list_splice_init(&pernet->local_addr_list, &free_list);
1316 __reset_counters(pernet);
1317 pernet->next_id = 1;
1318 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1319 spin_unlock_bh(&pernet->lock);
1320 mptcp_nl_remove_addrs_list(sock_net(skb->sk), &free_list);
1321 __flush_addrs(&free_list);
1325 static int mptcp_nl_fill_addr(struct sk_buff *skb,
1326 struct mptcp_pm_addr_entry *entry)
1328 struct mptcp_addr_info *addr = &entry->addr;
1329 struct nlattr *attr;
1331 attr = nla_nest_start(skb, MPTCP_PM_ATTR_ADDR);
1335 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_FAMILY, addr->family))
1336 goto nla_put_failure;
1337 if (nla_put_u16(skb, MPTCP_PM_ADDR_ATTR_PORT, ntohs(addr->port)))
1338 goto nla_put_failure;
1339 if (nla_put_u8(skb, MPTCP_PM_ADDR_ATTR_ID, addr->id))
1340 goto nla_put_failure;
1341 if (nla_put_u32(skb, MPTCP_PM_ADDR_ATTR_FLAGS, entry->addr.flags))
1342 goto nla_put_failure;
1343 if (entry->addr.ifindex &&
1344 nla_put_s32(skb, MPTCP_PM_ADDR_ATTR_IF_IDX, entry->addr.ifindex))
1345 goto nla_put_failure;
1347 if (addr->family == AF_INET &&
1348 nla_put_in_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR4,
1350 goto nla_put_failure;
1351 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1352 else if (addr->family == AF_INET6 &&
1353 nla_put_in6_addr(skb, MPTCP_PM_ADDR_ATTR_ADDR6, &addr->addr6))
1354 goto nla_put_failure;
1356 nla_nest_end(skb, attr);
1360 nla_nest_cancel(skb, attr);
1364 static int mptcp_nl_cmd_get_addr(struct sk_buff *skb, struct genl_info *info)
1366 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1367 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1368 struct mptcp_pm_addr_entry addr, *entry;
1369 struct sk_buff *msg;
1373 ret = mptcp_pm_parse_addr(attr, info, false, &addr);
1377 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1381 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1382 info->genlhdr->cmd);
1384 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1389 spin_lock_bh(&pernet->lock);
1390 entry = __lookup_addr_by_id(pernet, addr.addr.id);
1392 GENL_SET_ERR_MSG(info, "address not found");
1397 ret = mptcp_nl_fill_addr(msg, entry);
1401 genlmsg_end(msg, reply);
1402 ret = genlmsg_reply(msg, info);
1403 spin_unlock_bh(&pernet->lock);
1407 spin_unlock_bh(&pernet->lock);
1414 static int mptcp_nl_cmd_dump_addrs(struct sk_buff *msg,
1415 struct netlink_callback *cb)
1417 struct net *net = sock_net(msg->sk);
1418 struct mptcp_pm_addr_entry *entry;
1419 struct pm_nl_pernet *pernet;
1420 int id = cb->args[0];
1424 pernet = net_generic(net, pm_nl_pernet_id);
1426 spin_lock_bh(&pernet->lock);
1427 for (i = id; i < MAX_ADDR_ID + 1; i++) {
1428 if (test_bit(i, pernet->id_bitmap)) {
1429 entry = __lookup_addr_by_id(pernet, i);
1433 if (entry->addr.id <= id)
1436 hdr = genlmsg_put(msg, NETLINK_CB(cb->skb).portid,
1437 cb->nlh->nlmsg_seq, &mptcp_genl_family,
1438 NLM_F_MULTI, MPTCP_PM_CMD_GET_ADDR);
1442 if (mptcp_nl_fill_addr(msg, entry) < 0) {
1443 genlmsg_cancel(msg, hdr);
1447 id = entry->addr.id;
1448 genlmsg_end(msg, hdr);
1451 spin_unlock_bh(&pernet->lock);
1457 static int parse_limit(struct genl_info *info, int id, unsigned int *limit)
1459 struct nlattr *attr = info->attrs[id];
1464 *limit = nla_get_u32(attr);
1465 if (*limit > MPTCP_PM_ADDR_MAX) {
1466 GENL_SET_ERR_MSG(info, "limit greater than maximum");
1473 mptcp_nl_cmd_set_limits(struct sk_buff *skb, struct genl_info *info)
1475 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1476 unsigned int rcv_addrs, subflows;
1479 spin_lock_bh(&pernet->lock);
1480 rcv_addrs = pernet->add_addr_accept_max;
1481 ret = parse_limit(info, MPTCP_PM_ATTR_RCV_ADD_ADDRS, &rcv_addrs);
1485 subflows = pernet->subflows_max;
1486 ret = parse_limit(info, MPTCP_PM_ATTR_SUBFLOWS, &subflows);
1490 WRITE_ONCE(pernet->add_addr_accept_max, rcv_addrs);
1491 WRITE_ONCE(pernet->subflows_max, subflows);
1494 spin_unlock_bh(&pernet->lock);
1499 mptcp_nl_cmd_get_limits(struct sk_buff *skb, struct genl_info *info)
1501 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1502 struct sk_buff *msg;
1505 msg = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_KERNEL);
1509 reply = genlmsg_put_reply(msg, info, &mptcp_genl_family, 0,
1510 MPTCP_PM_CMD_GET_LIMITS);
1514 if (nla_put_u32(msg, MPTCP_PM_ATTR_RCV_ADD_ADDRS,
1515 READ_ONCE(pernet->add_addr_accept_max)))
1518 if (nla_put_u32(msg, MPTCP_PM_ATTR_SUBFLOWS,
1519 READ_ONCE(pernet->subflows_max)))
1522 genlmsg_end(msg, reply);
1523 return genlmsg_reply(msg, info);
1526 GENL_SET_ERR_MSG(info, "not enough space in Netlink message");
1531 static int mptcp_nl_addr_backup(struct net *net,
1532 struct mptcp_addr_info *addr,
1535 long s_slot = 0, s_num = 0;
1536 struct mptcp_sock *msk;
1539 while ((msk = mptcp_token_iter_next(net, &s_slot, &s_num)) != NULL) {
1540 struct sock *sk = (struct sock *)msk;
1542 if (list_empty(&msk->conn_list))
1546 spin_lock_bh(&msk->pm.lock);
1547 ret = mptcp_pm_nl_mp_prio_send_ack(msk, addr, bkup);
1548 spin_unlock_bh(&msk->pm.lock);
1559 static int mptcp_nl_cmd_set_flags(struct sk_buff *skb, struct genl_info *info)
1561 struct nlattr *attr = info->attrs[MPTCP_PM_ATTR_ADDR];
1562 struct pm_nl_pernet *pernet = genl_info_pm_nl(info);
1563 struct mptcp_pm_addr_entry addr, *entry;
1564 struct net *net = sock_net(skb->sk);
1568 ret = mptcp_pm_parse_addr(attr, info, true, &addr);
1572 if (addr.addr.flags & MPTCP_PM_ADDR_FLAG_BACKUP)
1575 list_for_each_entry(entry, &pernet->local_addr_list, list) {
1576 if (addresses_equal(&entry->addr, &addr.addr, true)) {
1577 ret = mptcp_nl_addr_backup(net, &entry->addr, bkup);
1582 entry->addr.flags |= MPTCP_PM_ADDR_FLAG_BACKUP;
1584 entry->addr.flags &= ~MPTCP_PM_ADDR_FLAG_BACKUP;
1591 static void mptcp_nl_mcast_send(struct net *net, struct sk_buff *nlskb, gfp_t gfp)
1593 genlmsg_multicast_netns(&mptcp_genl_family, net,
1594 nlskb, 0, MPTCP_PM_EV_GRP_OFFSET, gfp);
1597 static int mptcp_event_add_subflow(struct sk_buff *skb, const struct sock *ssk)
1599 const struct inet_sock *issk = inet_sk(ssk);
1600 const struct mptcp_subflow_context *sf;
1602 if (nla_put_u16(skb, MPTCP_ATTR_FAMILY, ssk->sk_family))
1605 switch (ssk->sk_family) {
1607 if (nla_put_in_addr(skb, MPTCP_ATTR_SADDR4, issk->inet_saddr))
1609 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, issk->inet_daddr))
1612 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1614 const struct ipv6_pinfo *np = inet6_sk(ssk);
1616 if (nla_put_in6_addr(skb, MPTCP_ATTR_SADDR6, &np->saddr))
1618 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &ssk->sk_v6_daddr))
1628 if (nla_put_be16(skb, MPTCP_ATTR_SPORT, issk->inet_sport))
1630 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, issk->inet_dport))
1633 sf = mptcp_subflow_ctx(ssk);
1634 if (WARN_ON_ONCE(!sf))
1637 if (nla_put_u8(skb, MPTCP_ATTR_LOC_ID, sf->local_id))
1640 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, sf->remote_id))
1646 static int mptcp_event_put_token_and_ssk(struct sk_buff *skb,
1647 const struct mptcp_sock *msk,
1648 const struct sock *ssk)
1650 const struct sock *sk = (const struct sock *)msk;
1651 const struct mptcp_subflow_context *sf;
1654 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1657 if (mptcp_event_add_subflow(skb, ssk))
1660 sf = mptcp_subflow_ctx(ssk);
1661 if (WARN_ON_ONCE(!sf))
1664 if (nla_put_u8(skb, MPTCP_ATTR_BACKUP, sf->backup))
1667 if (ssk->sk_bound_dev_if &&
1668 nla_put_s32(skb, MPTCP_ATTR_IF_IDX, ssk->sk_bound_dev_if))
1671 sk_err = ssk->sk_err;
1672 if (sk_err && sk->sk_state == TCP_ESTABLISHED &&
1673 nla_put_u8(skb, MPTCP_ATTR_ERROR, sk_err))
1679 static int mptcp_event_sub_established(struct sk_buff *skb,
1680 const struct mptcp_sock *msk,
1681 const struct sock *ssk)
1683 return mptcp_event_put_token_and_ssk(skb, msk, ssk);
1686 static int mptcp_event_sub_closed(struct sk_buff *skb,
1687 const struct mptcp_sock *msk,
1688 const struct sock *ssk)
1690 if (mptcp_event_put_token_and_ssk(skb, msk, ssk))
1696 static int mptcp_event_created(struct sk_buff *skb,
1697 const struct mptcp_sock *msk,
1698 const struct sock *ssk)
1700 int err = nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token);
1705 return mptcp_event_add_subflow(skb, ssk);
1708 void mptcp_event_addr_removed(const struct mptcp_sock *msk, uint8_t id)
1710 struct net *net = sock_net((const struct sock *)msk);
1711 struct nlmsghdr *nlh;
1712 struct sk_buff *skb;
1714 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1717 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1721 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, MPTCP_EVENT_REMOVED);
1723 goto nla_put_failure;
1725 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1726 goto nla_put_failure;
1728 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, id))
1729 goto nla_put_failure;
1731 genlmsg_end(skb, nlh);
1732 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1739 void mptcp_event_addr_announced(const struct mptcp_sock *msk,
1740 const struct mptcp_addr_info *info)
1742 struct net *net = sock_net((const struct sock *)msk);
1743 struct nlmsghdr *nlh;
1744 struct sk_buff *skb;
1746 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1749 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, GFP_ATOMIC);
1753 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0,
1754 MPTCP_EVENT_ANNOUNCED);
1756 goto nla_put_failure;
1758 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token))
1759 goto nla_put_failure;
1761 if (nla_put_u8(skb, MPTCP_ATTR_REM_ID, info->id))
1762 goto nla_put_failure;
1764 if (nla_put_be16(skb, MPTCP_ATTR_DPORT, info->port))
1765 goto nla_put_failure;
1767 switch (info->family) {
1769 if (nla_put_in_addr(skb, MPTCP_ATTR_DADDR4, info->addr.s_addr))
1770 goto nla_put_failure;
1772 #if IS_ENABLED(CONFIG_MPTCP_IPV6)
1774 if (nla_put_in6_addr(skb, MPTCP_ATTR_DADDR6, &info->addr6))
1775 goto nla_put_failure;
1780 goto nla_put_failure;
1783 genlmsg_end(skb, nlh);
1784 mptcp_nl_mcast_send(net, skb, GFP_ATOMIC);
1791 void mptcp_event(enum mptcp_event_type type, const struct mptcp_sock *msk,
1792 const struct sock *ssk, gfp_t gfp)
1794 struct net *net = sock_net((const struct sock *)msk);
1795 struct nlmsghdr *nlh;
1796 struct sk_buff *skb;
1798 if (!genl_has_listeners(&mptcp_genl_family, net, MPTCP_PM_EV_GRP_OFFSET))
1801 skb = nlmsg_new(NLMSG_DEFAULT_SIZE, gfp);
1805 nlh = genlmsg_put(skb, 0, 0, &mptcp_genl_family, 0, type);
1807 goto nla_put_failure;
1810 case MPTCP_EVENT_UNSPEC:
1813 case MPTCP_EVENT_CREATED:
1814 case MPTCP_EVENT_ESTABLISHED:
1815 if (mptcp_event_created(skb, msk, ssk) < 0)
1816 goto nla_put_failure;
1818 case MPTCP_EVENT_CLOSED:
1819 if (nla_put_u32(skb, MPTCP_ATTR_TOKEN, msk->token) < 0)
1820 goto nla_put_failure;
1822 case MPTCP_EVENT_ANNOUNCED:
1823 case MPTCP_EVENT_REMOVED:
1824 /* call mptcp_event_addr_announced()/removed instead */
1827 case MPTCP_EVENT_SUB_ESTABLISHED:
1828 case MPTCP_EVENT_SUB_PRIORITY:
1829 if (mptcp_event_sub_established(skb, msk, ssk) < 0)
1830 goto nla_put_failure;
1832 case MPTCP_EVENT_SUB_CLOSED:
1833 if (mptcp_event_sub_closed(skb, msk, ssk) < 0)
1834 goto nla_put_failure;
1838 genlmsg_end(skb, nlh);
1839 mptcp_nl_mcast_send(net, skb, gfp);
1846 static const struct genl_small_ops mptcp_pm_ops[] = {
1848 .cmd = MPTCP_PM_CMD_ADD_ADDR,
1849 .doit = mptcp_nl_cmd_add_addr,
1850 .flags = GENL_ADMIN_PERM,
1853 .cmd = MPTCP_PM_CMD_DEL_ADDR,
1854 .doit = mptcp_nl_cmd_del_addr,
1855 .flags = GENL_ADMIN_PERM,
1858 .cmd = MPTCP_PM_CMD_FLUSH_ADDRS,
1859 .doit = mptcp_nl_cmd_flush_addrs,
1860 .flags = GENL_ADMIN_PERM,
1863 .cmd = MPTCP_PM_CMD_GET_ADDR,
1864 .doit = mptcp_nl_cmd_get_addr,
1865 .dumpit = mptcp_nl_cmd_dump_addrs,
1868 .cmd = MPTCP_PM_CMD_SET_LIMITS,
1869 .doit = mptcp_nl_cmd_set_limits,
1870 .flags = GENL_ADMIN_PERM,
1873 .cmd = MPTCP_PM_CMD_GET_LIMITS,
1874 .doit = mptcp_nl_cmd_get_limits,
1877 .cmd = MPTCP_PM_CMD_SET_FLAGS,
1878 .doit = mptcp_nl_cmd_set_flags,
1879 .flags = GENL_ADMIN_PERM,
1883 static struct genl_family mptcp_genl_family __ro_after_init = {
1884 .name = MPTCP_PM_NAME,
1885 .version = MPTCP_PM_VER,
1886 .maxattr = MPTCP_PM_ATTR_MAX,
1887 .policy = mptcp_pm_policy,
1889 .module = THIS_MODULE,
1890 .small_ops = mptcp_pm_ops,
1891 .n_small_ops = ARRAY_SIZE(mptcp_pm_ops),
1892 .mcgrps = mptcp_pm_mcgrps,
1893 .n_mcgrps = ARRAY_SIZE(mptcp_pm_mcgrps),
1896 static int __net_init pm_nl_init_net(struct net *net)
1898 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1900 INIT_LIST_HEAD_RCU(&pernet->local_addr_list);
1901 __reset_counters(pernet);
1902 pernet->next_id = 1;
1903 bitmap_zero(pernet->id_bitmap, MAX_ADDR_ID + 1);
1904 spin_lock_init(&pernet->lock);
1908 static void __net_exit pm_nl_exit_net(struct list_head *net_list)
1912 list_for_each_entry(net, net_list, exit_list) {
1913 struct pm_nl_pernet *pernet = net_generic(net, pm_nl_pernet_id);
1915 /* net is removed from namespace list, can't race with
1918 __flush_addrs(&pernet->local_addr_list);
1922 static struct pernet_operations mptcp_pm_pernet_ops = {
1923 .init = pm_nl_init_net,
1924 .exit_batch = pm_nl_exit_net,
1925 .id = &pm_nl_pernet_id,
1926 .size = sizeof(struct pm_nl_pernet),
1929 void __init mptcp_pm_nl_init(void)
1931 if (register_pernet_subsys(&mptcp_pm_pernet_ops) < 0)
1932 panic("Failed to register MPTCP PM pernet subsystem.\n");
1934 if (genl_register_family(&mptcp_genl_family))
1935 panic("Failed to register MPTCP PM netlink family\n");