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 node_established_contact(struct tipc_node *n_ptr);
76 static void tipc_node_delete(struct tipc_node *node);
77 static void tipc_node_timeout(unsigned long data);
78 static void tipc_node_fsm_evt(struct tipc_node *n, int evt);
80 struct tipc_sock_conn {
84 struct list_head list;
87 static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
88 [TIPC_NLA_NODE_UNSPEC] = { .type = NLA_UNSPEC },
89 [TIPC_NLA_NODE_ADDR] = { .type = NLA_U32 },
90 [TIPC_NLA_NODE_UP] = { .type = NLA_FLAG }
94 * A trivial power-of-two bitmask technique is used for speed, since this
95 * operation is done for every incoming TIPC packet. The number of hash table
96 * entries has been chosen so that no hash chain exceeds 8 nodes and will
97 * usually be much smaller (typically only a single node).
99 static unsigned int tipc_hashfn(u32 addr)
101 return addr & (NODE_HTABLE_SIZE - 1);
104 static void tipc_node_kref_release(struct kref *kref)
106 struct tipc_node *node = container_of(kref, struct tipc_node, kref);
108 tipc_node_delete(node);
111 void tipc_node_put(struct tipc_node *node)
113 kref_put(&node->kref, tipc_node_kref_release);
116 static void tipc_node_get(struct tipc_node *node)
118 kref_get(&node->kref);
122 * tipc_node_find - locate specified node object, if it exists
124 struct tipc_node *tipc_node_find(struct net *net, u32 addr)
126 struct tipc_net *tn = net_generic(net, tipc_net_id);
127 struct tipc_node *node;
129 if (unlikely(!in_own_cluster_exact(net, addr)))
133 hlist_for_each_entry_rcu(node, &tn->node_htable[tipc_hashfn(addr)],
135 if (node->addr == addr) {
145 struct tipc_node *tipc_node_create(struct net *net, u32 addr, u16 capabilities)
147 struct tipc_net *tn = net_generic(net, tipc_net_id);
148 struct tipc_node *n_ptr, *temp_node;
150 spin_lock_bh(&tn->node_list_lock);
151 n_ptr = tipc_node_find(net, addr);
154 n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
156 pr_warn("Node creation failed, no memory\n");
161 n_ptr->capabilities = capabilities;
162 kref_init(&n_ptr->kref);
163 spin_lock_init(&n_ptr->lock);
164 INIT_HLIST_NODE(&n_ptr->hash);
165 INIT_LIST_HEAD(&n_ptr->list);
166 INIT_LIST_HEAD(&n_ptr->publ_list);
167 INIT_LIST_HEAD(&n_ptr->conn_sks);
168 skb_queue_head_init(&n_ptr->bclink.namedq);
169 __skb_queue_head_init(&n_ptr->bclink.deferdq);
170 hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
171 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
172 if (n_ptr->addr < temp_node->addr)
175 list_add_tail_rcu(&n_ptr->list, &temp_node->list);
176 n_ptr->state = SELF_DOWN_PEER_LEAVING;
177 n_ptr->signature = INVALID_NODE_SIG;
178 n_ptr->active_links[0] = INVALID_BEARER_ID;
179 n_ptr->active_links[1] = INVALID_BEARER_ID;
180 tipc_node_get(n_ptr);
181 setup_timer(&n_ptr->timer, tipc_node_timeout, (unsigned long)n_ptr);
182 n_ptr->keepalive_intv = U32_MAX;
184 spin_unlock_bh(&tn->node_list_lock);
188 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
190 unsigned long tol = l->tolerance;
191 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
192 unsigned long keepalive_intv = msecs_to_jiffies(intv);
194 /* Link with lowest tolerance determines timer interval */
195 if (keepalive_intv < n->keepalive_intv)
196 n->keepalive_intv = keepalive_intv;
198 /* Ensure link's abort limit corresponds to current interval */
199 l->abort_limit = l->tolerance / jiffies_to_msecs(n->keepalive_intv);
202 static void tipc_node_delete(struct tipc_node *node)
204 list_del_rcu(&node->list);
205 hlist_del_rcu(&node->hash);
206 kfree_rcu(node, rcu);
209 void tipc_node_stop(struct net *net)
211 struct tipc_net *tn = net_generic(net, tipc_net_id);
212 struct tipc_node *node, *t_node;
214 spin_lock_bh(&tn->node_list_lock);
215 list_for_each_entry_safe(node, t_node, &tn->node_list, list) {
216 if (del_timer(&node->timer))
220 spin_unlock_bh(&tn->node_list_lock);
223 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
225 struct tipc_node *node;
226 struct tipc_sock_conn *conn;
229 if (in_own_node(net, dnode))
232 node = tipc_node_find(net, dnode);
234 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
235 return -EHOSTUNREACH;
237 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
242 conn->peer_node = dnode;
244 conn->peer_port = peer_port;
246 tipc_node_lock(node);
247 list_add_tail(&conn->list, &node->conn_sks);
248 tipc_node_unlock(node);
254 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
256 struct tipc_node *node;
257 struct tipc_sock_conn *conn, *safe;
259 if (in_own_node(net, dnode))
262 node = tipc_node_find(net, dnode);
266 tipc_node_lock(node);
267 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
268 if (port != conn->port)
270 list_del(&conn->list);
273 tipc_node_unlock(node);
277 /* tipc_node_timeout - handle expiration of node timer
279 static void tipc_node_timeout(unsigned long data)
281 struct tipc_node *n = (struct tipc_node *)data;
282 struct tipc_link_entry *le;
283 struct sk_buff_head xmitq;
287 __skb_queue_head_init(&xmitq);
289 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
291 le = &n->links[bearer_id];
293 /* Link tolerance may change asynchronously: */
294 tipc_node_calculate_timer(n, le->link);
295 rc = tipc_link_timeout(le->link, &xmitq);
298 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
299 if (rc & TIPC_LINK_DOWN_EVT)
300 tipc_node_link_down(n, bearer_id, false);
302 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
308 * __tipc_node_link_up - handle addition of link
309 * Node lock must be held by caller
310 * Link becomes active (alone or shared) or standby, depending on its priority.
312 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
313 struct sk_buff_head *xmitq)
315 int *slot0 = &n->active_links[0];
316 int *slot1 = &n->active_links[1];
317 struct tipc_link *ol = node_active_link(n, 0);
318 struct tipc_link *nl = n->links[bearer_id].link;
323 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
324 if (!tipc_link_is_up(nl))
328 n->action_flags |= TIPC_NOTIFY_LINK_UP;
329 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
331 /* Leave room for tunnel header when returning 'mtu' to users: */
332 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
334 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
336 pr_debug("Established link <%s> on network plane %c\n",
337 nl->name, nl->net_plane);
339 /* First link? => give it both slots */
343 tipc_link_build_bcast_sync_msg(nl, xmitq);
344 node_established_contact(n);
348 /* Second link => redistribute slots */
349 if (nl->priority > ol->priority) {
350 pr_debug("Old link <%s> becomes standby\n", ol->name);
353 } else if (nl->priority == ol->priority) {
356 pr_debug("New link <%s> is standby\n", nl->name);
359 /* Prepare synchronization with first link */
360 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
364 * tipc_node_link_up - handle addition of link
366 * Link becomes active (alone or shared) or standby, depending on its priority.
368 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
369 struct sk_buff_head *xmitq)
372 __tipc_node_link_up(n, bearer_id, xmitq);
377 * __tipc_node_link_down - handle loss of link
379 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
380 struct sk_buff_head *xmitq,
381 struct tipc_media_addr **maddr)
383 struct tipc_link_entry *le = &n->links[*bearer_id];
384 int *slot0 = &n->active_links[0];
385 int *slot1 = &n->active_links[1];
387 struct tipc_link *l, *_l, *tnl;
389 l = n->links[*bearer_id].link;
390 if (!l || tipc_link_is_reset(l))
394 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
395 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
397 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
399 pr_debug("Lost link <%s> on network plane %c\n",
400 l->name, l->net_plane);
402 /* Select new active link if any available */
403 *slot0 = INVALID_BEARER_ID;
404 *slot1 = INVALID_BEARER_ID;
405 for (i = 0; i < MAX_BEARERS; i++) {
406 _l = n->links[i].link;
407 if (!_l || !tipc_link_is_up(_l))
411 if (_l->priority < highest)
413 if (_l->priority > highest) {
414 highest = _l->priority;
422 if (!tipc_node_is_up(n)) {
423 if (tipc_link_peer_is_down(l))
424 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
425 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
426 tipc_link_fsm_evt(l, LINK_RESET_EVT);
428 tipc_link_build_reset_msg(l, xmitq);
429 *maddr = &n->links[*bearer_id].maddr;
430 node_lost_contact(n, &le->inputq);
434 /* There is still a working link => initiate failover */
435 tnl = node_active_link(n, 0);
436 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
437 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
438 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
439 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
441 tipc_link_fsm_evt(l, LINK_RESET_EVT);
442 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
443 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
444 *maddr = &n->links[tnl->bearer_id].maddr;
445 *bearer_id = tnl->bearer_id;
448 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
450 struct tipc_link_entry *le = &n->links[bearer_id];
451 struct tipc_link *l = le->link;
452 struct tipc_media_addr *maddr;
453 struct sk_buff_head xmitq;
458 __skb_queue_head_init(&xmitq);
461 if (!tipc_link_is_establishing(l)) {
462 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
469 /* Defuse pending tipc_node_link_up() */
470 tipc_link_fsm_evt(l, LINK_RESET_EVT);
473 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
474 tipc_sk_rcv(n->net, &le->inputq);
477 bool tipc_node_is_up(struct tipc_node *n)
479 return n->active_links[0] != INVALID_BEARER_ID;
482 void tipc_node_check_dest(struct net *net, u32 onode,
483 struct tipc_bearer *b,
484 u16 capabilities, u32 signature,
485 struct tipc_media_addr *maddr,
486 bool *respond, bool *dupl_addr)
490 struct tipc_link_entry *le;
491 bool addr_match = false;
492 bool sign_match = false;
493 bool link_up = false;
494 bool accept_addr = false;
501 n = tipc_node_create(net, onode, capabilities);
507 le = &n->links[b->identity];
509 /* Prepare to validate requesting node's signature and media address */
511 link_up = l && tipc_link_is_up(l);
512 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
513 sign_match = (signature == n->signature);
515 /* These three flags give us eight permutations: */
517 if (sign_match && addr_match && link_up) {
518 /* All is fine. Do nothing. */
520 } else if (sign_match && addr_match && !link_up) {
521 /* Respond. The link will come up in due time */
523 } else if (sign_match && !addr_match && link_up) {
524 /* Peer has changed i/f address without rebooting.
525 * If so, the link will reset soon, and the next
526 * discovery will be accepted. So we can ignore it.
527 * It may also be an cloned or malicious peer having
528 * chosen the same node address and signature as an
530 * Ignore requests until the link goes down, if ever.
533 } else if (sign_match && !addr_match && !link_up) {
534 /* Peer link has changed i/f address without rebooting.
535 * It may also be a cloned or malicious peer; we can't
536 * distinguish between the two.
537 * The signature is correct, so we must accept.
541 } else if (!sign_match && addr_match && link_up) {
542 /* Peer node rebooted. Two possibilities:
543 * - Delayed re-discovery; this link endpoint has already
544 * reset and re-established contact with the peer, before
545 * receiving a discovery message from that node.
546 * (The peer happened to receive one from this node first).
547 * - The peer came back so fast that our side has not
548 * discovered it yet. Probing from this side will soon
549 * reset the link, since there can be no working link
550 * endpoint at the peer end, and the link will re-establish.
551 * Accept the signature, since it comes from a known peer.
553 n->signature = signature;
554 } else if (!sign_match && addr_match && !link_up) {
555 /* The peer node has rebooted.
556 * Accept signature, since it is a known peer.
558 n->signature = signature;
560 } else if (!sign_match && !addr_match && link_up) {
561 /* Peer rebooted with new address, or a new/duplicate peer.
562 * Ignore until the link goes down, if ever.
565 } else if (!sign_match && !addr_match && !link_up) {
566 /* Peer rebooted with new address, or it is a new peer.
567 * Accept signature and address.
569 n->signature = signature;
577 /* Now create new link if not already existing */
579 if (n->link_cnt == 2) {
580 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
583 if_name = strchr(b->name, ':') + 1;
584 if (!tipc_link_create(n, if_name, b->identity, b->tolerance,
585 b->net_plane, b->mtu, b->priority,
586 b->window, mod(tipc_net(net)->random),
587 tipc_own_addr(net), onode, &le->maddr,
588 &le->inputq, &n->bclink.namedq, &l)) {
593 tipc_link_fsm_evt(l, LINK_RESET_EVT);
594 if (n->state == NODE_FAILINGOVER)
595 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
598 tipc_node_calculate_timer(n, l);
599 if (n->link_cnt == 1)
600 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
603 memcpy(&le->maddr, maddr, sizeof(*maddr));
606 if (reset && !tipc_link_is_reset(l))
607 tipc_node_link_down(n, b->identity, false);
611 void tipc_node_delete_links(struct net *net, int bearer_id)
613 struct tipc_net *tn = net_generic(net, tipc_net_id);
617 list_for_each_entry_rcu(n, &tn->node_list, list) {
618 tipc_node_link_down(n, bearer_id, true);
623 static void tipc_node_reset_links(struct tipc_node *n)
625 char addr_string[16];
628 pr_warn("Resetting all links to %s\n",
629 tipc_addr_string_fill(addr_string, n->addr));
631 for (i = 0; i < MAX_BEARERS; i++) {
632 tipc_node_link_down(n, i, false);
636 /* tipc_node_fsm_evt - node finite state machine
637 * Determines when contact is allowed with peer node
639 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
641 int state = n->state;
644 case SELF_DOWN_PEER_DOWN:
646 case SELF_ESTABL_CONTACT_EVT:
647 state = SELF_UP_PEER_COMING;
649 case PEER_ESTABL_CONTACT_EVT:
650 state = SELF_COMING_PEER_UP;
652 case SELF_LOST_CONTACT_EVT:
653 case PEER_LOST_CONTACT_EVT:
655 case NODE_SYNCH_END_EVT:
656 case NODE_SYNCH_BEGIN_EVT:
657 case NODE_FAILOVER_BEGIN_EVT:
658 case NODE_FAILOVER_END_EVT:
663 case SELF_UP_PEER_UP:
665 case SELF_LOST_CONTACT_EVT:
666 state = SELF_DOWN_PEER_LEAVING;
668 case PEER_LOST_CONTACT_EVT:
669 state = SELF_LEAVING_PEER_DOWN;
671 case NODE_SYNCH_BEGIN_EVT:
672 state = NODE_SYNCHING;
674 case NODE_FAILOVER_BEGIN_EVT:
675 state = NODE_FAILINGOVER;
677 case SELF_ESTABL_CONTACT_EVT:
678 case PEER_ESTABL_CONTACT_EVT:
679 case NODE_SYNCH_END_EVT:
680 case NODE_FAILOVER_END_EVT:
686 case SELF_DOWN_PEER_LEAVING:
688 case PEER_LOST_CONTACT_EVT:
689 state = SELF_DOWN_PEER_DOWN;
691 case SELF_ESTABL_CONTACT_EVT:
692 case PEER_ESTABL_CONTACT_EVT:
693 case SELF_LOST_CONTACT_EVT:
695 case NODE_SYNCH_END_EVT:
696 case NODE_SYNCH_BEGIN_EVT:
697 case NODE_FAILOVER_BEGIN_EVT:
698 case NODE_FAILOVER_END_EVT:
703 case SELF_UP_PEER_COMING:
705 case PEER_ESTABL_CONTACT_EVT:
706 state = SELF_UP_PEER_UP;
708 case SELF_LOST_CONTACT_EVT:
709 state = SELF_DOWN_PEER_LEAVING;
711 case SELF_ESTABL_CONTACT_EVT:
712 case PEER_LOST_CONTACT_EVT:
713 case NODE_SYNCH_END_EVT:
714 case NODE_FAILOVER_BEGIN_EVT:
716 case NODE_SYNCH_BEGIN_EVT:
717 case NODE_FAILOVER_END_EVT:
722 case SELF_COMING_PEER_UP:
724 case SELF_ESTABL_CONTACT_EVT:
725 state = SELF_UP_PEER_UP;
727 case PEER_LOST_CONTACT_EVT:
728 state = SELF_LEAVING_PEER_DOWN;
730 case SELF_LOST_CONTACT_EVT:
731 case PEER_ESTABL_CONTACT_EVT:
733 case NODE_SYNCH_END_EVT:
734 case NODE_SYNCH_BEGIN_EVT:
735 case NODE_FAILOVER_BEGIN_EVT:
736 case NODE_FAILOVER_END_EVT:
741 case SELF_LEAVING_PEER_DOWN:
743 case SELF_LOST_CONTACT_EVT:
744 state = SELF_DOWN_PEER_DOWN;
746 case SELF_ESTABL_CONTACT_EVT:
747 case PEER_ESTABL_CONTACT_EVT:
748 case PEER_LOST_CONTACT_EVT:
750 case NODE_SYNCH_END_EVT:
751 case NODE_SYNCH_BEGIN_EVT:
752 case NODE_FAILOVER_BEGIN_EVT:
753 case NODE_FAILOVER_END_EVT:
758 case NODE_FAILINGOVER:
760 case SELF_LOST_CONTACT_EVT:
761 state = SELF_DOWN_PEER_LEAVING;
763 case PEER_LOST_CONTACT_EVT:
764 state = SELF_LEAVING_PEER_DOWN;
766 case NODE_FAILOVER_END_EVT:
767 state = SELF_UP_PEER_UP;
769 case NODE_FAILOVER_BEGIN_EVT:
770 case SELF_ESTABL_CONTACT_EVT:
771 case PEER_ESTABL_CONTACT_EVT:
773 case NODE_SYNCH_BEGIN_EVT:
774 case NODE_SYNCH_END_EVT:
781 case SELF_LOST_CONTACT_EVT:
782 state = SELF_DOWN_PEER_LEAVING;
784 case PEER_LOST_CONTACT_EVT:
785 state = SELF_LEAVING_PEER_DOWN;
787 case NODE_SYNCH_END_EVT:
788 state = SELF_UP_PEER_UP;
790 case NODE_FAILOVER_BEGIN_EVT:
791 state = NODE_FAILINGOVER;
793 case NODE_SYNCH_BEGIN_EVT:
794 case SELF_ESTABL_CONTACT_EVT:
795 case PEER_ESTABL_CONTACT_EVT:
797 case NODE_FAILOVER_END_EVT:
803 pr_err("Unknown node fsm state %x\n", state);
810 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
813 bool tipc_node_filter_pkt(struct tipc_node *n, struct tipc_msg *hdr)
815 int state = n->state;
817 if (likely(state == SELF_UP_PEER_UP))
820 if (state == SELF_LEAVING_PEER_DOWN)
823 if (state == SELF_DOWN_PEER_LEAVING) {
824 if (msg_peer_node_is_up(hdr))
831 static void node_established_contact(struct tipc_node *n_ptr)
833 tipc_node_fsm_evt(n_ptr, SELF_ESTABL_CONTACT_EVT);
834 n_ptr->action_flags |= TIPC_NOTIFY_NODE_UP;
835 n_ptr->bclink.oos_state = 0;
836 n_ptr->bclink.acked = tipc_bclink_get_last_sent(n_ptr->net);
837 tipc_bclink_add_node(n_ptr->net, n_ptr->addr);
840 static void node_lost_contact(struct tipc_node *n_ptr,
841 struct sk_buff_head *inputq)
843 char addr_string[16];
844 struct tipc_sock_conn *conn, *safe;
846 struct list_head *conns = &n_ptr->conn_sks;
848 struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
851 pr_debug("Lost contact with %s\n",
852 tipc_addr_string_fill(addr_string, n_ptr->addr));
854 /* Flush broadcast link info associated with lost node */
855 if (n_ptr->bclink.recv_permitted) {
856 __skb_queue_purge(&n_ptr->bclink.deferdq);
858 if (n_ptr->bclink.reasm_buf) {
859 kfree_skb(n_ptr->bclink.reasm_buf);
860 n_ptr->bclink.reasm_buf = NULL;
863 tipc_bclink_remove_node(n_ptr->net, n_ptr->addr);
864 tipc_bclink_acknowledge(n_ptr, INVALID_LINK_SEQ);
866 n_ptr->bclink.recv_permitted = false;
869 /* Abort any ongoing link failover */
870 for (i = 0; i < MAX_BEARERS; i++) {
871 l = n_ptr->links[i].link;
873 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
876 /* Notify publications from this node */
877 n_ptr->action_flags |= TIPC_NOTIFY_NODE_DOWN;
879 /* Notify sockets connected to node */
880 list_for_each_entry_safe(conn, safe, conns, list) {
881 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
882 SHORT_H_SIZE, 0, tn->own_addr,
883 conn->peer_node, conn->port,
884 conn->peer_port, TIPC_ERR_NO_NODE);
886 skb_queue_tail(inputq, skb);
887 list_del(&conn->list);
893 * tipc_node_get_linkname - get the name of a link
895 * @bearer_id: id of the bearer
896 * @node: peer node address
897 * @linkname: link name output buffer
899 * Returns 0 on success
901 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
902 char *linkname, size_t len)
904 struct tipc_link *link;
906 struct tipc_node *node = tipc_node_find(net, addr);
911 if (bearer_id >= MAX_BEARERS)
914 tipc_node_lock(node);
915 link = node->links[bearer_id].link;
917 strncpy(linkname, link->name, len);
921 tipc_node_unlock(node);
926 void tipc_node_unlock(struct tipc_node *node)
928 struct net *net = node->net;
930 u32 flags = node->action_flags;
932 struct list_head *publ_list;
934 if (likely(!flags)) {
935 spin_unlock_bh(&node->lock);
940 link_id = node->link_id;
941 publ_list = &node->publ_list;
943 node->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
944 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP |
945 TIPC_WAKEUP_BCAST_USERS | TIPC_BCAST_MSG_EVT |
948 spin_unlock_bh(&node->lock);
950 if (flags & TIPC_NOTIFY_NODE_DOWN)
951 tipc_publ_notify(net, publ_list, addr);
953 if (flags & TIPC_WAKEUP_BCAST_USERS)
954 tipc_bclink_wakeup_users(net);
956 if (flags & TIPC_NOTIFY_NODE_UP)
957 tipc_named_node_up(net, addr);
959 if (flags & TIPC_NOTIFY_LINK_UP)
960 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
961 TIPC_NODE_SCOPE, link_id, addr);
963 if (flags & TIPC_NOTIFY_LINK_DOWN)
964 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
967 if (flags & TIPC_BCAST_MSG_EVT)
968 tipc_bclink_input(net);
970 if (flags & TIPC_BCAST_RESET)
971 tipc_node_reset_links(node);
974 /* Caller should hold node lock for the passed node */
975 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
978 struct nlattr *attrs;
980 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
981 NLM_F_MULTI, TIPC_NL_NODE_GET);
985 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
989 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
991 if (tipc_node_is_up(node))
992 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
995 nla_nest_end(msg->skb, attrs);
996 genlmsg_end(msg->skb, hdr);
1001 nla_nest_cancel(msg->skb, attrs);
1003 genlmsg_cancel(msg->skb, hdr);
1008 static struct tipc_link *tipc_node_select_link(struct tipc_node *n, int sel,
1010 struct tipc_media_addr **maddr)
1012 int id = n->active_links[sel & 1];
1014 if (unlikely(id < 0))
1018 *maddr = &n->links[id].maddr;
1019 return n->links[id].link;
1023 * tipc_node_xmit() is the general link level function for message sending
1024 * @net: the applicable net namespace
1025 * @list: chain of buffers containing message
1026 * @dnode: address of destination node
1027 * @selector: a number used for deterministic link selection
1028 * Consumes the buffer chain, except when returning -ELINKCONG
1029 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1031 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1032 u32 dnode, int selector)
1034 struct tipc_link *l = NULL;
1035 struct tipc_node *n;
1036 struct sk_buff_head xmitq;
1037 struct tipc_media_addr *maddr;
1039 int rc = -EHOSTUNREACH;
1041 __skb_queue_head_init(&xmitq);
1042 n = tipc_node_find(net, dnode);
1045 l = tipc_node_select_link(n, selector, &bearer_id, &maddr);
1047 rc = tipc_link_xmit(l, list, &xmitq);
1048 tipc_node_unlock(n);
1049 if (unlikely(rc == -ENOBUFS))
1050 tipc_node_link_down(n, bearer_id, false);
1054 tipc_bearer_xmit(net, bearer_id, &xmitq, maddr);
1057 if (likely(in_own_node(net, dnode))) {
1058 tipc_sk_rcv(net, list);
1064 /* tipc_node_xmit_skb(): send single buffer to destination
1065 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1066 * messages, which will not be rejected
1067 * The only exception is datagram messages rerouted after secondary
1068 * lookup, which are rare and safe to dispose of anyway.
1069 * TODO: Return real return value, and let callers use
1070 * tipc_wait_for_sendpkt() where applicable
1072 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1075 struct sk_buff_head head;
1078 skb_queue_head_init(&head);
1079 __skb_queue_tail(&head, skb);
1080 rc = tipc_node_xmit(net, &head, dnode, selector);
1081 if (rc == -ELINKCONG)
1087 * tipc_node_check_state - check and if necessary update node state
1089 * @bearer_id: identity of bearer delivering the packet
1090 * Returns true if state is ok, otherwise consumes buffer and returns false
1092 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1093 int bearer_id, struct sk_buff_head *xmitq)
1095 struct tipc_msg *hdr = buf_msg(skb);
1096 int usr = msg_user(hdr);
1097 int mtyp = msg_type(hdr);
1098 u16 oseqno = msg_seqno(hdr);
1099 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1100 u16 exp_pkts = msg_msgcnt(hdr);
1101 u16 rcv_nxt, syncpt, dlv_nxt;
1102 int state = n->state;
1103 struct tipc_link *l, *tnl, *pl = NULL;
1104 struct tipc_media_addr *maddr;
1107 l = n->links[bearer_id].link;
1110 rcv_nxt = l->rcv_nxt;
1113 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1116 /* Find parallel link, if any */
1117 for (i = 0; i < MAX_BEARERS; i++) {
1118 if ((i != bearer_id) && n->links[i].link) {
1119 pl = n->links[i].link;
1124 /* Update node accesibility if applicable */
1125 if (state == SELF_UP_PEER_COMING) {
1126 if (!tipc_link_is_up(l))
1128 if (!msg_peer_link_is_up(hdr))
1130 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1133 if (state == SELF_DOWN_PEER_LEAVING) {
1134 if (msg_peer_node_is_up(hdr))
1136 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1139 /* Ignore duplicate packets */
1140 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1143 /* Initiate or update failover mode if applicable */
1144 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1145 syncpt = oseqno + exp_pkts - 1;
1146 if (pl && tipc_link_is_up(pl)) {
1147 pb_id = pl->bearer_id;
1148 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1149 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1151 /* If pkts arrive out of order, use lowest calculated syncpt */
1152 if (less(syncpt, n->sync_point))
1153 n->sync_point = syncpt;
1156 /* Open parallel link when tunnel link reaches synch point */
1157 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1158 if (!more(rcv_nxt, n->sync_point))
1160 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1162 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1166 /* No synching needed if only one link */
1167 if (!pl || !tipc_link_is_up(pl))
1170 /* Initiate synch mode if applicable */
1171 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1172 syncpt = iseqno + exp_pkts - 1;
1173 if (!tipc_link_is_up(l)) {
1174 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1175 __tipc_node_link_up(n, bearer_id, xmitq);
1177 if (n->state == SELF_UP_PEER_UP) {
1178 n->sync_point = syncpt;
1179 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1180 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1182 if (less(syncpt, n->sync_point))
1183 n->sync_point = syncpt;
1186 /* Open tunnel link when parallel link reaches synch point */
1187 if ((n->state == NODE_SYNCHING) && tipc_link_is_synching(l)) {
1188 if (tipc_link_is_synching(l)) {
1194 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1195 if (more(dlv_nxt, n->sync_point)) {
1196 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1197 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1202 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1204 if (usr == LINK_PROTOCOL)
1212 * tipc_rcv - process TIPC packets/messages arriving from off-node
1213 * @net: the applicable net namespace
1215 * @bearer: pointer to bearer message arrived on
1217 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1218 * structure (i.e. cannot be NULL), but bearer can be inactive.
1220 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1222 struct sk_buff_head xmitq;
1223 struct tipc_node *n;
1224 struct tipc_msg *hdr = buf_msg(skb);
1225 int usr = msg_user(hdr);
1226 int bearer_id = b->identity;
1227 struct tipc_link_entry *le;
1230 __skb_queue_head_init(&xmitq);
1232 /* Ensure message is well-formed */
1233 if (unlikely(!tipc_msg_validate(skb)))
1236 /* Handle arrival of a non-unicast link packet */
1237 if (unlikely(msg_non_seq(hdr))) {
1238 if (usr == LINK_CONFIG)
1239 tipc_disc_rcv(net, skb, b);
1241 tipc_bclink_rcv(net, skb);
1245 /* Locate neighboring node that sent packet */
1246 n = tipc_node_find(net, msg_prevnode(hdr));
1249 le = &n->links[bearer_id];
1253 /* Is reception permitted at the moment ? */
1254 if (!tipc_node_filter_pkt(n, hdr))
1257 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1258 tipc_bclink_sync_state(n, hdr);
1260 /* Release acked broadcast packets */
1261 if (unlikely(n->bclink.acked != msg_bcast_ack(hdr)))
1262 tipc_bclink_acknowledge(n, msg_bcast_ack(hdr));
1264 /* Check and if necessary update node state */
1265 if (likely(tipc_node_check_state(n, skb, bearer_id, &xmitq))) {
1266 rc = tipc_link_rcv(le->link, skb, &xmitq);
1270 tipc_node_unlock(n);
1272 if (unlikely(rc & TIPC_LINK_UP_EVT))
1273 tipc_node_link_up(n, bearer_id, &xmitq);
1275 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1276 tipc_node_link_down(n, bearer_id, false);
1278 if (unlikely(!skb_queue_empty(&n->bclink.namedq)))
1279 tipc_named_rcv(net, &n->bclink.namedq);
1281 if (!skb_queue_empty(&le->inputq))
1282 tipc_sk_rcv(net, &le->inputq);
1284 if (!skb_queue_empty(&xmitq))
1285 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1292 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1295 struct net *net = sock_net(skb->sk);
1296 struct tipc_net *tn = net_generic(net, tipc_net_id);
1297 int done = cb->args[0];
1298 int last_addr = cb->args[1];
1299 struct tipc_node *node;
1300 struct tipc_nl_msg msg;
1306 msg.portid = NETLINK_CB(cb->skb).portid;
1307 msg.seq = cb->nlh->nlmsg_seq;
1311 node = tipc_node_find(net, last_addr);
1314 /* We never set seq or call nl_dump_check_consistent()
1315 * this means that setting prev_seq here will cause the
1316 * consistence check to fail in the netlink callback
1317 * handler. Resulting in the NLMSG_DONE message having
1318 * the NLM_F_DUMP_INTR flag set if the node state
1319 * changed while we released the lock.
1324 tipc_node_put(node);
1327 list_for_each_entry_rcu(node, &tn->node_list, list) {
1329 if (node->addr == last_addr)
1335 tipc_node_lock(node);
1336 err = __tipc_nl_add_node(&msg, node);
1338 last_addr = node->addr;
1339 tipc_node_unlock(node);
1343 tipc_node_unlock(node);
1348 cb->args[1] = last_addr;