2 * net/tipc/node.c: TIPC node management routines
4 * Copyright (c) 2000-2006, 2012-2015, Ericsson AB
5 * Copyright (c) 2005-2006, 2010-2014, Wind River Systems
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
11 * 1. Redistributions of source code must retain the above copyright
12 * notice, this list of conditions and the following disclaimer.
13 * 2. Redistributions in binary form must reproduce the above copyright
14 * notice, this list of conditions and the following disclaimer in the
15 * documentation and/or other materials provided with the distribution.
16 * 3. Neither the names of the copyright holders nor the names of its
17 * contributors may be used to endorse or promote products derived from
18 * this software without specific prior written permission.
20 * Alternatively, this software may be distributed under the terms of the
21 * GNU General Public License ("GPL") version 2 as published by the Free
22 * Software Foundation.
24 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
25 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
26 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
27 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
28 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
29 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
30 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
31 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
32 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
33 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
34 * POSSIBILITY OF SUCH DAMAGE.
40 #include "name_distr.h"
45 /* Node FSM states and events:
48 SELF_DOWN_PEER_DOWN = 0xdd,
49 SELF_UP_PEER_UP = 0xaa,
50 SELF_DOWN_PEER_LEAVING = 0xd1,
51 SELF_UP_PEER_COMING = 0xac,
52 SELF_COMING_PEER_UP = 0xca,
53 SELF_LEAVING_PEER_DOWN = 0x1d,
54 NODE_FAILINGOVER = 0xf0,
59 SELF_ESTABL_CONTACT_EVT = 0xece,
60 SELF_LOST_CONTACT_EVT = 0x1ce,
61 PEER_ESTABL_CONTACT_EVT = 0x9ece,
62 PEER_LOST_CONTACT_EVT = 0x91ce,
63 NODE_FAILOVER_BEGIN_EVT = 0xfbe,
64 NODE_FAILOVER_END_EVT = 0xfee,
65 NODE_SYNCH_BEGIN_EVT = 0xcbe,
66 NODE_SYNCH_END_EVT = 0xcee
69 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
70 struct sk_buff_head *xmitq,
71 struct tipc_media_addr **maddr);
72 static void tipc_node_link_down(struct tipc_node *n, int bearer_id,
74 static void node_lost_contact(struct tipc_node *n, struct sk_buff_head *inputq);
75 static void tipc_node_delete(struct tipc_node *node);
76 static void tipc_node_timeout(unsigned long data);
77 static void tipc_node_fsm_evt(struct tipc_node *n, int evt);
79 struct tipc_sock_conn {
83 struct list_head list;
86 static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
87 [TIPC_NLA_NODE_UNSPEC] = { .type = NLA_UNSPEC },
88 [TIPC_NLA_NODE_ADDR] = { .type = NLA_U32 },
89 [TIPC_NLA_NODE_UP] = { .type = NLA_FLAG }
93 * A trivial power-of-two bitmask technique is used for speed, since this
94 * operation is done for every incoming TIPC packet. The number of hash table
95 * entries has been chosen so that no hash chain exceeds 8 nodes and will
96 * usually be much smaller (typically only a single node).
98 static unsigned int tipc_hashfn(u32 addr)
100 return addr & (NODE_HTABLE_SIZE - 1);
103 static void tipc_node_kref_release(struct kref *kref)
105 struct tipc_node *node = container_of(kref, struct tipc_node, kref);
107 tipc_node_delete(node);
110 void tipc_node_put(struct tipc_node *node)
112 kref_put(&node->kref, tipc_node_kref_release);
115 static void tipc_node_get(struct tipc_node *node)
117 kref_get(&node->kref);
121 * tipc_node_find - locate specified node object, if it exists
123 struct tipc_node *tipc_node_find(struct net *net, u32 addr)
125 struct tipc_net *tn = net_generic(net, tipc_net_id);
126 struct tipc_node *node;
128 if (unlikely(!in_own_cluster_exact(net, addr)))
132 hlist_for_each_entry_rcu(node, &tn->node_htable[tipc_hashfn(addr)],
134 if (node->addr == addr) {
144 void tipc_node_read_lock(struct tipc_node *n)
146 read_lock_bh(&n->lock);
149 void tipc_node_read_unlock(struct tipc_node *n)
151 read_unlock_bh(&n->lock);
154 static void tipc_node_write_lock(struct tipc_node *n)
156 write_lock_bh(&n->lock);
159 static void tipc_node_write_unlock(struct tipc_node *n)
161 struct net *net = n->net;
163 u32 flags = n->action_flags;
165 struct list_head *publ_list;
167 if (likely(!flags)) {
168 write_unlock_bh(&n->lock);
173 link_id = n->link_id;
174 publ_list = &n->publ_list;
176 n->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
177 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP);
179 write_unlock_bh(&n->lock);
181 if (flags & TIPC_NOTIFY_NODE_DOWN)
182 tipc_publ_notify(net, publ_list, addr);
184 if (flags & TIPC_NOTIFY_NODE_UP)
185 tipc_named_node_up(net, addr);
187 if (flags & TIPC_NOTIFY_LINK_UP)
188 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
189 TIPC_NODE_SCOPE, link_id, addr);
191 if (flags & TIPC_NOTIFY_LINK_DOWN)
192 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
196 struct tipc_node *tipc_node_create(struct net *net, u32 addr, u16 capabilities)
198 struct tipc_net *tn = net_generic(net, tipc_net_id);
199 struct tipc_node *n_ptr, *temp_node;
202 spin_lock_bh(&tn->node_list_lock);
203 n_ptr = tipc_node_find(net, addr);
206 n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
208 pr_warn("Node creation failed, no memory\n");
213 n_ptr->capabilities = capabilities;
214 kref_init(&n_ptr->kref);
215 rwlock_init(&n_ptr->lock);
216 INIT_HLIST_NODE(&n_ptr->hash);
217 INIT_LIST_HEAD(&n_ptr->list);
218 INIT_LIST_HEAD(&n_ptr->publ_list);
219 INIT_LIST_HEAD(&n_ptr->conn_sks);
220 skb_queue_head_init(&n_ptr->bc_entry.namedq);
221 skb_queue_head_init(&n_ptr->bc_entry.inputq1);
222 __skb_queue_head_init(&n_ptr->bc_entry.arrvq);
223 skb_queue_head_init(&n_ptr->bc_entry.inputq2);
224 for (i = 0; i < MAX_BEARERS; i++)
225 spin_lock_init(&n_ptr->links[i].lock);
226 hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
227 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
228 if (n_ptr->addr < temp_node->addr)
231 list_add_tail_rcu(&n_ptr->list, &temp_node->list);
232 n_ptr->state = SELF_DOWN_PEER_LEAVING;
233 n_ptr->signature = INVALID_NODE_SIG;
234 n_ptr->active_links[0] = INVALID_BEARER_ID;
235 n_ptr->active_links[1] = INVALID_BEARER_ID;
236 if (!tipc_link_bc_create(net, tipc_own_addr(net), n_ptr->addr,
237 U16_MAX, tipc_bc_sndlink(net)->window,
239 &n_ptr->bc_entry.inputq1,
240 &n_ptr->bc_entry.namedq,
241 tipc_bc_sndlink(net),
242 &n_ptr->bc_entry.link)) {
243 pr_warn("Broadcast rcv link creation failed, no memory\n");
248 tipc_node_get(n_ptr);
249 setup_timer(&n_ptr->timer, tipc_node_timeout, (unsigned long)n_ptr);
250 n_ptr->keepalive_intv = U32_MAX;
252 spin_unlock_bh(&tn->node_list_lock);
256 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
258 unsigned long tol = l->tolerance;
259 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
260 unsigned long keepalive_intv = msecs_to_jiffies(intv);
262 /* Link with lowest tolerance determines timer interval */
263 if (keepalive_intv < n->keepalive_intv)
264 n->keepalive_intv = keepalive_intv;
266 /* Ensure link's abort limit corresponds to current interval */
267 l->abort_limit = l->tolerance / jiffies_to_msecs(n->keepalive_intv);
270 static void tipc_node_delete(struct tipc_node *node)
272 list_del_rcu(&node->list);
273 hlist_del_rcu(&node->hash);
274 kfree(node->bc_entry.link);
275 kfree_rcu(node, rcu);
278 void tipc_node_stop(struct net *net)
280 struct tipc_net *tn = net_generic(net, tipc_net_id);
281 struct tipc_node *node, *t_node;
283 spin_lock_bh(&tn->node_list_lock);
284 list_for_each_entry_safe(node, t_node, &tn->node_list, list) {
285 if (del_timer(&node->timer))
289 spin_unlock_bh(&tn->node_list_lock);
292 void tipc_node_subscribe(struct net *net, struct list_head *subscr, u32 addr)
296 if (in_own_node(net, addr))
299 n = tipc_node_find(net, addr);
301 pr_warn("Node subscribe rejected, unknown node 0x%x\n", addr);
304 tipc_node_write_lock(n);
305 list_add_tail(subscr, &n->publ_list);
306 tipc_node_write_unlock(n);
310 void tipc_node_unsubscribe(struct net *net, struct list_head *subscr, u32 addr)
314 if (in_own_node(net, addr))
317 n = tipc_node_find(net, addr);
319 pr_warn("Node unsubscribe rejected, unknown node 0x%x\n", addr);
322 tipc_node_write_lock(n);
323 list_del_init(subscr);
324 tipc_node_write_unlock(n);
328 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
330 struct tipc_node *node;
331 struct tipc_sock_conn *conn;
334 if (in_own_node(net, dnode))
337 node = tipc_node_find(net, dnode);
339 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
340 return -EHOSTUNREACH;
342 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
347 conn->peer_node = dnode;
349 conn->peer_port = peer_port;
351 tipc_node_write_lock(node);
352 list_add_tail(&conn->list, &node->conn_sks);
353 tipc_node_write_unlock(node);
359 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
361 struct tipc_node *node;
362 struct tipc_sock_conn *conn, *safe;
364 if (in_own_node(net, dnode))
367 node = tipc_node_find(net, dnode);
371 tipc_node_write_lock(node);
372 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
373 if (port != conn->port)
375 list_del(&conn->list);
378 tipc_node_write_unlock(node);
382 /* tipc_node_timeout - handle expiration of node timer
384 static void tipc_node_timeout(unsigned long data)
386 struct tipc_node *n = (struct tipc_node *)data;
387 struct tipc_link_entry *le;
388 struct sk_buff_head xmitq;
392 __skb_queue_head_init(&xmitq);
394 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
395 tipc_node_read_lock(n);
396 le = &n->links[bearer_id];
397 spin_lock_bh(&le->lock);
399 /* Link tolerance may change asynchronously: */
400 tipc_node_calculate_timer(n, le->link);
401 rc = tipc_link_timeout(le->link, &xmitq);
403 spin_unlock_bh(&le->lock);
404 tipc_node_read_unlock(n);
405 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
406 if (rc & TIPC_LINK_DOWN_EVT)
407 tipc_node_link_down(n, bearer_id, false);
409 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
415 * __tipc_node_link_up - handle addition of link
416 * Node lock must be held by caller
417 * Link becomes active (alone or shared) or standby, depending on its priority.
419 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
420 struct sk_buff_head *xmitq)
422 int *slot0 = &n->active_links[0];
423 int *slot1 = &n->active_links[1];
424 struct tipc_link *ol = node_active_link(n, 0);
425 struct tipc_link *nl = n->links[bearer_id].link;
430 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
431 if (!tipc_link_is_up(nl))
435 n->action_flags |= TIPC_NOTIFY_LINK_UP;
436 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
438 /* Leave room for tunnel header when returning 'mtu' to users: */
439 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
441 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
442 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
444 pr_debug("Established link <%s> on network plane %c\n",
445 nl->name, nl->net_plane);
447 /* First link? => give it both slots */
451 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
452 n->action_flags |= TIPC_NOTIFY_NODE_UP;
453 tipc_bcast_add_peer(n->net, nl, xmitq);
457 /* Second link => redistribute slots */
458 if (nl->priority > ol->priority) {
459 pr_debug("Old link <%s> becomes standby\n", ol->name);
462 tipc_link_set_active(nl, true);
463 tipc_link_set_active(ol, false);
464 } else if (nl->priority == ol->priority) {
465 tipc_link_set_active(nl, true);
468 pr_debug("New link <%s> is standby\n", nl->name);
471 /* Prepare synchronization with first link */
472 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
476 * tipc_node_link_up - handle addition of link
478 * Link becomes active (alone or shared) or standby, depending on its priority.
480 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
481 struct sk_buff_head *xmitq)
483 tipc_node_write_lock(n);
484 __tipc_node_link_up(n, bearer_id, xmitq);
485 tipc_node_write_unlock(n);
489 * __tipc_node_link_down - handle loss of link
491 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
492 struct sk_buff_head *xmitq,
493 struct tipc_media_addr **maddr)
495 struct tipc_link_entry *le = &n->links[*bearer_id];
496 int *slot0 = &n->active_links[0];
497 int *slot1 = &n->active_links[1];
499 struct tipc_link *l, *_l, *tnl;
501 l = n->links[*bearer_id].link;
502 if (!l || tipc_link_is_reset(l))
506 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
507 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
509 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
511 pr_debug("Lost link <%s> on network plane %c\n",
512 l->name, l->net_plane);
514 /* Select new active link if any available */
515 *slot0 = INVALID_BEARER_ID;
516 *slot1 = INVALID_BEARER_ID;
517 for (i = 0; i < MAX_BEARERS; i++) {
518 _l = n->links[i].link;
519 if (!_l || !tipc_link_is_up(_l))
523 if (_l->priority < highest)
525 if (_l->priority > highest) {
526 highest = _l->priority;
534 if (!tipc_node_is_up(n)) {
535 if (tipc_link_peer_is_down(l))
536 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
537 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
538 tipc_link_fsm_evt(l, LINK_RESET_EVT);
540 tipc_link_build_reset_msg(l, xmitq);
541 *maddr = &n->links[*bearer_id].maddr;
542 node_lost_contact(n, &le->inputq);
543 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
546 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
548 /* There is still a working link => initiate failover */
549 tnl = node_active_link(n, 0);
550 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
551 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
552 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
553 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
555 tipc_link_fsm_evt(l, LINK_RESET_EVT);
556 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
557 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
558 *maddr = &n->links[tnl->bearer_id].maddr;
559 *bearer_id = tnl->bearer_id;
562 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
564 struct tipc_link_entry *le = &n->links[bearer_id];
565 struct tipc_link *l = le->link;
566 struct tipc_media_addr *maddr;
567 struct sk_buff_head xmitq;
572 __skb_queue_head_init(&xmitq);
574 tipc_node_write_lock(n);
575 if (!tipc_link_is_establishing(l)) {
576 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
583 /* Defuse pending tipc_node_link_up() */
584 tipc_link_fsm_evt(l, LINK_RESET_EVT);
586 tipc_node_write_unlock(n);
587 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
588 tipc_sk_rcv(n->net, &le->inputq);
591 bool tipc_node_is_up(struct tipc_node *n)
593 return n->active_links[0] != INVALID_BEARER_ID;
596 void tipc_node_check_dest(struct net *net, u32 onode,
597 struct tipc_bearer *b,
598 u16 capabilities, u32 signature,
599 struct tipc_media_addr *maddr,
600 bool *respond, bool *dupl_addr)
604 struct tipc_link_entry *le;
605 bool addr_match = false;
606 bool sign_match = false;
607 bool link_up = false;
608 bool accept_addr = false;
615 n = tipc_node_create(net, onode, capabilities);
619 tipc_node_write_lock(n);
621 le = &n->links[b->identity];
623 /* Prepare to validate requesting node's signature and media address */
625 link_up = l && tipc_link_is_up(l);
626 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
627 sign_match = (signature == n->signature);
629 /* These three flags give us eight permutations: */
631 if (sign_match && addr_match && link_up) {
632 /* All is fine. Do nothing. */
634 } else if (sign_match && addr_match && !link_up) {
635 /* Respond. The link will come up in due time */
637 } else if (sign_match && !addr_match && link_up) {
638 /* Peer has changed i/f address without rebooting.
639 * If so, the link will reset soon, and the next
640 * discovery will be accepted. So we can ignore it.
641 * It may also be an cloned or malicious peer having
642 * chosen the same node address and signature as an
644 * Ignore requests until the link goes down, if ever.
647 } else if (sign_match && !addr_match && !link_up) {
648 /* Peer link has changed i/f address without rebooting.
649 * It may also be a cloned or malicious peer; we can't
650 * distinguish between the two.
651 * The signature is correct, so we must accept.
655 } else if (!sign_match && addr_match && link_up) {
656 /* Peer node rebooted. Two possibilities:
657 * - Delayed re-discovery; this link endpoint has already
658 * reset and re-established contact with the peer, before
659 * receiving a discovery message from that node.
660 * (The peer happened to receive one from this node first).
661 * - The peer came back so fast that our side has not
662 * discovered it yet. Probing from this side will soon
663 * reset the link, since there can be no working link
664 * endpoint at the peer end, and the link will re-establish.
665 * Accept the signature, since it comes from a known peer.
667 n->signature = signature;
668 } else if (!sign_match && addr_match && !link_up) {
669 /* The peer node has rebooted.
670 * Accept signature, since it is a known peer.
672 n->signature = signature;
674 } else if (!sign_match && !addr_match && link_up) {
675 /* Peer rebooted with new address, or a new/duplicate peer.
676 * Ignore until the link goes down, if ever.
679 } else if (!sign_match && !addr_match && !link_up) {
680 /* Peer rebooted with new address, or it is a new peer.
681 * Accept signature and address.
683 n->signature = signature;
691 /* Now create new link if not already existing */
693 if (n->link_cnt == 2) {
694 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
697 if_name = strchr(b->name, ':') + 1;
698 if (!tipc_link_create(net, if_name, b->identity, b->tolerance,
699 b->net_plane, b->mtu, b->priority,
700 b->window, mod(tipc_net(net)->random),
701 tipc_own_addr(net), onode,
703 tipc_bc_sndlink(n->net), n->bc_entry.link,
705 &n->bc_entry.namedq, &l)) {
710 tipc_link_fsm_evt(l, LINK_RESET_EVT);
711 if (n->state == NODE_FAILINGOVER)
712 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
715 tipc_node_calculate_timer(n, l);
716 if (n->link_cnt == 1)
717 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
720 memcpy(&le->maddr, maddr, sizeof(*maddr));
722 tipc_node_write_unlock(n);
723 if (reset && !tipc_link_is_reset(l))
724 tipc_node_link_down(n, b->identity, false);
728 void tipc_node_delete_links(struct net *net, int bearer_id)
730 struct tipc_net *tn = net_generic(net, tipc_net_id);
734 list_for_each_entry_rcu(n, &tn->node_list, list) {
735 tipc_node_link_down(n, bearer_id, true);
740 static void tipc_node_reset_links(struct tipc_node *n)
742 char addr_string[16];
745 pr_warn("Resetting all links to %s\n",
746 tipc_addr_string_fill(addr_string, n->addr));
748 for (i = 0; i < MAX_BEARERS; i++) {
749 tipc_node_link_down(n, i, false);
753 /* tipc_node_fsm_evt - node finite state machine
754 * Determines when contact is allowed with peer node
756 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
758 int state = n->state;
761 case SELF_DOWN_PEER_DOWN:
763 case SELF_ESTABL_CONTACT_EVT:
764 state = SELF_UP_PEER_COMING;
766 case PEER_ESTABL_CONTACT_EVT:
767 state = SELF_COMING_PEER_UP;
769 case SELF_LOST_CONTACT_EVT:
770 case PEER_LOST_CONTACT_EVT:
772 case NODE_SYNCH_END_EVT:
773 case NODE_SYNCH_BEGIN_EVT:
774 case NODE_FAILOVER_BEGIN_EVT:
775 case NODE_FAILOVER_END_EVT:
780 case SELF_UP_PEER_UP:
782 case SELF_LOST_CONTACT_EVT:
783 state = SELF_DOWN_PEER_LEAVING;
785 case PEER_LOST_CONTACT_EVT:
786 state = SELF_LEAVING_PEER_DOWN;
788 case NODE_SYNCH_BEGIN_EVT:
789 state = NODE_SYNCHING;
791 case NODE_FAILOVER_BEGIN_EVT:
792 state = NODE_FAILINGOVER;
794 case SELF_ESTABL_CONTACT_EVT:
795 case PEER_ESTABL_CONTACT_EVT:
796 case NODE_SYNCH_END_EVT:
797 case NODE_FAILOVER_END_EVT:
803 case SELF_DOWN_PEER_LEAVING:
805 case PEER_LOST_CONTACT_EVT:
806 state = SELF_DOWN_PEER_DOWN;
808 case SELF_ESTABL_CONTACT_EVT:
809 case PEER_ESTABL_CONTACT_EVT:
810 case SELF_LOST_CONTACT_EVT:
812 case NODE_SYNCH_END_EVT:
813 case NODE_SYNCH_BEGIN_EVT:
814 case NODE_FAILOVER_BEGIN_EVT:
815 case NODE_FAILOVER_END_EVT:
820 case SELF_UP_PEER_COMING:
822 case PEER_ESTABL_CONTACT_EVT:
823 state = SELF_UP_PEER_UP;
825 case SELF_LOST_CONTACT_EVT:
826 state = SELF_DOWN_PEER_LEAVING;
828 case SELF_ESTABL_CONTACT_EVT:
829 case PEER_LOST_CONTACT_EVT:
830 case NODE_SYNCH_END_EVT:
831 case NODE_FAILOVER_BEGIN_EVT:
833 case NODE_SYNCH_BEGIN_EVT:
834 case NODE_FAILOVER_END_EVT:
839 case SELF_COMING_PEER_UP:
841 case SELF_ESTABL_CONTACT_EVT:
842 state = SELF_UP_PEER_UP;
844 case PEER_LOST_CONTACT_EVT:
845 state = SELF_LEAVING_PEER_DOWN;
847 case SELF_LOST_CONTACT_EVT:
848 case PEER_ESTABL_CONTACT_EVT:
850 case NODE_SYNCH_END_EVT:
851 case NODE_SYNCH_BEGIN_EVT:
852 case NODE_FAILOVER_BEGIN_EVT:
853 case NODE_FAILOVER_END_EVT:
858 case SELF_LEAVING_PEER_DOWN:
860 case SELF_LOST_CONTACT_EVT:
861 state = SELF_DOWN_PEER_DOWN;
863 case SELF_ESTABL_CONTACT_EVT:
864 case PEER_ESTABL_CONTACT_EVT:
865 case PEER_LOST_CONTACT_EVT:
867 case NODE_SYNCH_END_EVT:
868 case NODE_SYNCH_BEGIN_EVT:
869 case NODE_FAILOVER_BEGIN_EVT:
870 case NODE_FAILOVER_END_EVT:
875 case NODE_FAILINGOVER:
877 case SELF_LOST_CONTACT_EVT:
878 state = SELF_DOWN_PEER_LEAVING;
880 case PEER_LOST_CONTACT_EVT:
881 state = SELF_LEAVING_PEER_DOWN;
883 case NODE_FAILOVER_END_EVT:
884 state = SELF_UP_PEER_UP;
886 case NODE_FAILOVER_BEGIN_EVT:
887 case SELF_ESTABL_CONTACT_EVT:
888 case PEER_ESTABL_CONTACT_EVT:
890 case NODE_SYNCH_BEGIN_EVT:
891 case NODE_SYNCH_END_EVT:
898 case SELF_LOST_CONTACT_EVT:
899 state = SELF_DOWN_PEER_LEAVING;
901 case PEER_LOST_CONTACT_EVT:
902 state = SELF_LEAVING_PEER_DOWN;
904 case NODE_SYNCH_END_EVT:
905 state = SELF_UP_PEER_UP;
907 case NODE_FAILOVER_BEGIN_EVT:
908 state = NODE_FAILINGOVER;
910 case NODE_SYNCH_BEGIN_EVT:
911 case SELF_ESTABL_CONTACT_EVT:
912 case PEER_ESTABL_CONTACT_EVT:
914 case NODE_FAILOVER_END_EVT:
920 pr_err("Unknown node fsm state %x\n", state);
927 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
930 static void node_lost_contact(struct tipc_node *n,
931 struct sk_buff_head *inputq)
933 char addr_string[16];
934 struct tipc_sock_conn *conn, *safe;
936 struct list_head *conns = &n->conn_sks;
940 pr_debug("Lost contact with %s\n",
941 tipc_addr_string_fill(addr_string, n->addr));
943 /* Clean up broadcast state */
944 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
946 /* Abort any ongoing link failover */
947 for (i = 0; i < MAX_BEARERS; i++) {
948 l = n->links[i].link;
950 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
953 /* Notify publications from this node */
954 n->action_flags |= TIPC_NOTIFY_NODE_DOWN;
956 /* Notify sockets connected to node */
957 list_for_each_entry_safe(conn, safe, conns, list) {
958 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
959 SHORT_H_SIZE, 0, tipc_own_addr(n->net),
960 conn->peer_node, conn->port,
961 conn->peer_port, TIPC_ERR_NO_NODE);
963 skb_queue_tail(inputq, skb);
964 list_del(&conn->list);
970 * tipc_node_get_linkname - get the name of a link
972 * @bearer_id: id of the bearer
973 * @node: peer node address
974 * @linkname: link name output buffer
976 * Returns 0 on success
978 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
979 char *linkname, size_t len)
981 struct tipc_link *link;
983 struct tipc_node *node = tipc_node_find(net, addr);
988 if (bearer_id >= MAX_BEARERS)
991 tipc_node_read_lock(node);
992 link = node->links[bearer_id].link;
994 strncpy(linkname, link->name, len);
998 tipc_node_read_unlock(node);
1003 /* Caller should hold node lock for the passed node */
1004 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
1007 struct nlattr *attrs;
1009 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1010 NLM_F_MULTI, TIPC_NL_NODE_GET);
1014 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
1018 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
1020 if (tipc_node_is_up(node))
1021 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
1024 nla_nest_end(msg->skb, attrs);
1025 genlmsg_end(msg->skb, hdr);
1030 nla_nest_cancel(msg->skb, attrs);
1032 genlmsg_cancel(msg->skb, hdr);
1038 * tipc_node_xmit() is the general link level function for message sending
1039 * @net: the applicable net namespace
1040 * @list: chain of buffers containing message
1041 * @dnode: address of destination node
1042 * @selector: a number used for deterministic link selection
1043 * Consumes the buffer chain, except when returning -ELINKCONG
1044 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1046 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1047 u32 dnode, int selector)
1049 struct tipc_link_entry *le = NULL;
1050 struct tipc_node *n;
1051 struct sk_buff_head xmitq;
1053 int rc = -EHOSTUNREACH;
1055 __skb_queue_head_init(&xmitq);
1056 n = tipc_node_find(net, dnode);
1058 tipc_node_read_lock(n);
1059 bearer_id = n->active_links[selector & 1];
1060 if (bearer_id >= 0) {
1061 le = &n->links[bearer_id];
1062 spin_lock_bh(&le->lock);
1063 rc = tipc_link_xmit(le->link, list, &xmitq);
1064 spin_unlock_bh(&le->lock);
1066 tipc_node_read_unlock(n);
1067 if (likely(!skb_queue_empty(&xmitq))) {
1068 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1071 if (unlikely(rc == -ENOBUFS))
1072 tipc_node_link_down(n, bearer_id, false);
1077 if (unlikely(!in_own_node(net, dnode)))
1079 tipc_sk_rcv(net, list);
1083 /* tipc_node_xmit_skb(): send single buffer to destination
1084 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1085 * messages, which will not be rejected
1086 * The only exception is datagram messages rerouted after secondary
1087 * lookup, which are rare and safe to dispose of anyway.
1088 * TODO: Return real return value, and let callers use
1089 * tipc_wait_for_sendpkt() where applicable
1091 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1094 struct sk_buff_head head;
1097 skb_queue_head_init(&head);
1098 __skb_queue_tail(&head, skb);
1099 rc = tipc_node_xmit(net, &head, dnode, selector);
1100 if (rc == -ELINKCONG)
1105 void tipc_node_broadcast(struct net *net, struct sk_buff *skb)
1107 struct sk_buff *txskb;
1108 struct tipc_node *n;
1112 list_for_each_entry_rcu(n, tipc_nodes(net), list) {
1114 if (in_own_node(net, dst))
1116 if (!tipc_node_is_up(n))
1118 txskb = pskb_copy(skb, GFP_ATOMIC);
1121 msg_set_destnode(buf_msg(txskb), dst);
1122 tipc_node_xmit_skb(net, txskb, dst, 0);
1130 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1131 * @net: the applicable net namespace
1133 * @bearer_id: id of bearer message arrived on
1135 * Invoked with no locks held.
1137 static void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1140 struct sk_buff_head xmitq;
1141 struct tipc_bclink_entry *be;
1142 struct tipc_link_entry *le;
1143 struct tipc_msg *hdr = buf_msg(skb);
1144 int usr = msg_user(hdr);
1145 u32 dnode = msg_destnode(hdr);
1146 struct tipc_node *n;
1148 __skb_queue_head_init(&xmitq);
1150 /* If NACK for other node, let rcv link for that node peek into it */
1151 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1152 n = tipc_node_find(net, dnode);
1154 n = tipc_node_find(net, msg_prevnode(hdr));
1160 le = &n->links[bearer_id];
1162 rc = tipc_bcast_rcv(net, be->link, skb);
1164 /* Broadcast link reset may happen at reassembly failure */
1165 if (rc & TIPC_LINK_DOWN_EVT)
1166 tipc_node_reset_links(n);
1168 /* Broadcast ACKs are sent on a unicast link */
1169 if (rc & TIPC_LINK_SND_BC_ACK) {
1170 tipc_node_read_lock(n);
1171 tipc_link_build_ack_msg(le->link, &xmitq);
1172 tipc_node_read_unlock(n);
1175 if (!skb_queue_empty(&xmitq))
1176 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1178 /* Deliver. 'arrvq' is under inputq2's lock protection */
1179 if (!skb_queue_empty(&be->inputq1)) {
1180 spin_lock_bh(&be->inputq2.lock);
1181 spin_lock_bh(&be->inputq1.lock);
1182 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1183 spin_unlock_bh(&be->inputq1.lock);
1184 spin_unlock_bh(&be->inputq2.lock);
1185 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1191 * tipc_node_check_state - check and if necessary update node state
1193 * @bearer_id: identity of bearer delivering the packet
1194 * Returns true if state is ok, otherwise consumes buffer and returns false
1196 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1197 int bearer_id, struct sk_buff_head *xmitq)
1199 struct tipc_msg *hdr = buf_msg(skb);
1200 int usr = msg_user(hdr);
1201 int mtyp = msg_type(hdr);
1202 u16 oseqno = msg_seqno(hdr);
1203 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1204 u16 exp_pkts = msg_msgcnt(hdr);
1205 u16 rcv_nxt, syncpt, dlv_nxt;
1206 int state = n->state;
1207 struct tipc_link *l, *tnl, *pl = NULL;
1208 struct tipc_media_addr *maddr;
1211 l = n->links[bearer_id].link;
1214 rcv_nxt = l->rcv_nxt;
1217 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1220 /* Find parallel link, if any */
1221 for (i = 0; i < MAX_BEARERS; i++) {
1222 if ((i != bearer_id) && n->links[i].link) {
1223 pl = n->links[i].link;
1228 /* Check and update node accesibility if applicable */
1229 if (state == SELF_UP_PEER_COMING) {
1230 if (!tipc_link_is_up(l))
1232 if (!msg_peer_link_is_up(hdr))
1234 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1237 if (state == SELF_DOWN_PEER_LEAVING) {
1238 if (msg_peer_node_is_up(hdr))
1240 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1244 if (state == SELF_LEAVING_PEER_DOWN)
1247 /* Ignore duplicate packets */
1248 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1251 /* Initiate or update failover mode if applicable */
1252 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1253 syncpt = oseqno + exp_pkts - 1;
1254 if (pl && tipc_link_is_up(pl)) {
1255 pb_id = pl->bearer_id;
1256 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1257 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1259 /* If pkts arrive out of order, use lowest calculated syncpt */
1260 if (less(syncpt, n->sync_point))
1261 n->sync_point = syncpt;
1264 /* Open parallel link when tunnel link reaches synch point */
1265 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1266 if (!more(rcv_nxt, n->sync_point))
1268 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1270 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1274 /* No synching needed if only one link */
1275 if (!pl || !tipc_link_is_up(pl))
1278 /* Initiate synch mode if applicable */
1279 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1280 syncpt = iseqno + exp_pkts - 1;
1281 if (!tipc_link_is_up(l)) {
1282 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1283 __tipc_node_link_up(n, bearer_id, xmitq);
1285 if (n->state == SELF_UP_PEER_UP) {
1286 n->sync_point = syncpt;
1287 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1288 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1292 /* Open tunnel link when parallel link reaches synch point */
1293 if (n->state == NODE_SYNCHING) {
1294 if (tipc_link_is_synching(l)) {
1300 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1301 if (more(dlv_nxt, n->sync_point)) {
1302 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1303 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1308 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1310 if (usr == LINK_PROTOCOL)
1318 * tipc_rcv - process TIPC packets/messages arriving from off-node
1319 * @net: the applicable net namespace
1321 * @bearer: pointer to bearer message arrived on
1323 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1324 * structure (i.e. cannot be NULL), but bearer can be inactive.
1326 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1328 struct sk_buff_head xmitq;
1329 struct tipc_node *n;
1330 struct tipc_msg *hdr = buf_msg(skb);
1331 int usr = msg_user(hdr);
1332 int bearer_id = b->identity;
1333 struct tipc_link_entry *le;
1334 u16 bc_ack = msg_bcast_ack(hdr);
1337 __skb_queue_head_init(&xmitq);
1339 /* Ensure message is well-formed */
1340 if (unlikely(!tipc_msg_validate(skb)))
1343 /* Handle arrival of discovery or broadcast packet */
1344 if (unlikely(msg_non_seq(hdr))) {
1345 if (unlikely(usr == LINK_CONFIG))
1346 return tipc_disc_rcv(net, skb, b);
1348 return tipc_node_bc_rcv(net, skb, bearer_id);
1351 /* Locate neighboring node that sent packet */
1352 n = tipc_node_find(net, msg_prevnode(hdr));
1355 le = &n->links[bearer_id];
1357 /* Ensure broadcast reception is in synch with peer's send state */
1358 if (unlikely(usr == LINK_PROTOCOL))
1359 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1360 else if (unlikely(n->bc_entry.link->acked != bc_ack))
1361 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1363 /* Receive packet directly if conditions permit */
1364 tipc_node_read_lock(n);
1365 if (likely((n->state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL))) {
1366 spin_lock_bh(&le->lock);
1368 rc = tipc_link_rcv(le->link, skb, &xmitq);
1371 spin_unlock_bh(&le->lock);
1373 tipc_node_read_unlock(n);
1375 /* Check/update node state before receiving */
1376 if (unlikely(skb)) {
1377 tipc_node_write_lock(n);
1378 if (tipc_node_check_state(n, skb, bearer_id, &xmitq)) {
1380 rc = tipc_link_rcv(le->link, skb, &xmitq);
1384 tipc_node_write_unlock(n);
1387 if (unlikely(rc & TIPC_LINK_UP_EVT))
1388 tipc_node_link_up(n, bearer_id, &xmitq);
1390 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1391 tipc_node_link_down(n, bearer_id, false);
1393 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1394 tipc_named_rcv(net, &n->bc_entry.namedq);
1396 if (!skb_queue_empty(&le->inputq))
1397 tipc_sk_rcv(net, &le->inputq);
1399 if (!skb_queue_empty(&xmitq))
1400 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1407 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1410 struct net *net = sock_net(skb->sk);
1411 struct tipc_net *tn = net_generic(net, tipc_net_id);
1412 int done = cb->args[0];
1413 int last_addr = cb->args[1];
1414 struct tipc_node *node;
1415 struct tipc_nl_msg msg;
1421 msg.portid = NETLINK_CB(cb->skb).portid;
1422 msg.seq = cb->nlh->nlmsg_seq;
1426 node = tipc_node_find(net, last_addr);
1429 /* We never set seq or call nl_dump_check_consistent()
1430 * this means that setting prev_seq here will cause the
1431 * consistence check to fail in the netlink callback
1432 * handler. Resulting in the NLMSG_DONE message having
1433 * the NLM_F_DUMP_INTR flag set if the node state
1434 * changed while we released the lock.
1439 tipc_node_put(node);
1442 list_for_each_entry_rcu(node, &tn->node_list, list) {
1444 if (node->addr == last_addr)
1450 tipc_node_read_lock(node);
1451 err = __tipc_nl_add_node(&msg, node);
1453 last_addr = node->addr;
1454 tipc_node_read_unlock(node);
1458 tipc_node_read_unlock(node);
1463 cb->args[1] = last_addr;