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 *n = container_of(kref, struct tipc_node, kref);
107 kfree(n->bc_entry.link);
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->bc_entry.namedq);
169 skb_queue_head_init(&n_ptr->bc_entry.inputq1);
170 __skb_queue_head_init(&n_ptr->bc_entry.arrvq);
171 skb_queue_head_init(&n_ptr->bc_entry.inputq2);
172 n_ptr->state = SELF_DOWN_PEER_LEAVING;
173 n_ptr->signature = INVALID_NODE_SIG;
174 n_ptr->active_links[0] = INVALID_BEARER_ID;
175 n_ptr->active_links[1] = INVALID_BEARER_ID;
176 if (!tipc_link_bc_create(net, tipc_own_addr(net), n_ptr->addr,
177 U16_MAX, tipc_bc_sndlink(net)->window,
179 &n_ptr->bc_entry.inputq1,
180 &n_ptr->bc_entry.namedq,
181 tipc_bc_sndlink(net),
182 &n_ptr->bc_entry.link)) {
183 pr_warn("Broadcast rcv link creation failed, no memory\n");
188 tipc_node_get(n_ptr);
189 setup_timer(&n_ptr->timer, tipc_node_timeout, (unsigned long)n_ptr);
190 n_ptr->keepalive_intv = U32_MAX;
191 hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
192 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
193 if (n_ptr->addr < temp_node->addr)
196 list_add_tail_rcu(&n_ptr->list, &temp_node->list);
198 spin_unlock_bh(&tn->node_list_lock);
202 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
204 unsigned long tol = l->tolerance;
205 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
206 unsigned long keepalive_intv = msecs_to_jiffies(intv);
208 /* Link with lowest tolerance determines timer interval */
209 if (keepalive_intv < n->keepalive_intv)
210 n->keepalive_intv = keepalive_intv;
212 /* Ensure link's abort limit corresponds to current interval */
213 l->abort_limit = l->tolerance / jiffies_to_msecs(n->keepalive_intv);
216 static void tipc_node_delete(struct tipc_node *node)
218 list_del_rcu(&node->list);
219 hlist_del_rcu(&node->hash);
222 del_timer_sync(&node->timer);
226 void tipc_node_stop(struct net *net)
228 struct tipc_net *tn = tipc_net(net);
229 struct tipc_node *node, *t_node;
231 spin_lock_bh(&tn->node_list_lock);
232 list_for_each_entry_safe(node, t_node, &tn->node_list, list)
233 tipc_node_delete(node);
234 spin_unlock_bh(&tn->node_list_lock);
237 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
239 struct tipc_node *node;
240 struct tipc_sock_conn *conn;
243 if (in_own_node(net, dnode))
246 node = tipc_node_find(net, dnode);
248 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
249 return -EHOSTUNREACH;
251 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
256 conn->peer_node = dnode;
258 conn->peer_port = peer_port;
260 tipc_node_lock(node);
261 list_add_tail(&conn->list, &node->conn_sks);
262 tipc_node_unlock(node);
268 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
270 struct tipc_node *node;
271 struct tipc_sock_conn *conn, *safe;
273 if (in_own_node(net, dnode))
276 node = tipc_node_find(net, dnode);
280 tipc_node_lock(node);
281 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
282 if (port != conn->port)
284 list_del(&conn->list);
287 tipc_node_unlock(node);
291 /* tipc_node_timeout - handle expiration of node timer
293 static void tipc_node_timeout(unsigned long data)
295 struct tipc_node *n = (struct tipc_node *)data;
296 struct tipc_link_entry *le;
297 struct sk_buff_head xmitq;
301 __skb_queue_head_init(&xmitq);
303 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
305 le = &n->links[bearer_id];
307 /* Link tolerance may change asynchronously: */
308 tipc_node_calculate_timer(n, le->link);
309 rc = tipc_link_timeout(le->link, &xmitq);
312 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
313 if (rc & TIPC_LINK_DOWN_EVT)
314 tipc_node_link_down(n, bearer_id, false);
316 mod_timer(&n->timer, jiffies + n->keepalive_intv);
320 * __tipc_node_link_up - handle addition of link
321 * Node lock must be held by caller
322 * Link becomes active (alone or shared) or standby, depending on its priority.
324 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
325 struct sk_buff_head *xmitq)
327 int *slot0 = &n->active_links[0];
328 int *slot1 = &n->active_links[1];
329 struct tipc_link *ol = node_active_link(n, 0);
330 struct tipc_link *nl = n->links[bearer_id].link;
335 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
336 if (!tipc_link_is_up(nl))
340 n->action_flags |= TIPC_NOTIFY_LINK_UP;
341 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
343 /* Leave room for tunnel header when returning 'mtu' to users: */
344 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
346 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
347 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
349 pr_debug("Established link <%s> on network plane %c\n",
350 nl->name, nl->net_plane);
352 /* First link? => give it both slots */
356 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
357 n->action_flags |= TIPC_NOTIFY_NODE_UP;
358 tipc_bcast_add_peer(n->net, nl, xmitq);
362 /* Second link => redistribute slots */
363 if (nl->priority > ol->priority) {
364 pr_debug("Old link <%s> becomes standby\n", ol->name);
367 tipc_link_set_active(nl, true);
368 tipc_link_set_active(ol, false);
369 } else if (nl->priority == ol->priority) {
370 tipc_link_set_active(nl, true);
373 pr_debug("New link <%s> is standby\n", nl->name);
376 /* Prepare synchronization with first link */
377 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
381 * tipc_node_link_up - handle addition of link
383 * Link becomes active (alone or shared) or standby, depending on its priority.
385 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
386 struct sk_buff_head *xmitq)
389 __tipc_node_link_up(n, bearer_id, xmitq);
394 * __tipc_node_link_down - handle loss of link
396 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
397 struct sk_buff_head *xmitq,
398 struct tipc_media_addr **maddr)
400 struct tipc_link_entry *le = &n->links[*bearer_id];
401 int *slot0 = &n->active_links[0];
402 int *slot1 = &n->active_links[1];
404 struct tipc_link *l, *_l, *tnl;
406 l = n->links[*bearer_id].link;
407 if (!l || tipc_link_is_reset(l))
411 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
412 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
414 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
416 pr_debug("Lost link <%s> on network plane %c\n",
417 l->name, l->net_plane);
419 /* Select new active link if any available */
420 *slot0 = INVALID_BEARER_ID;
421 *slot1 = INVALID_BEARER_ID;
422 for (i = 0; i < MAX_BEARERS; i++) {
423 _l = n->links[i].link;
424 if (!_l || !tipc_link_is_up(_l))
428 if (_l->priority < highest)
430 if (_l->priority > highest) {
431 highest = _l->priority;
439 if (!tipc_node_is_up(n)) {
440 if (tipc_link_peer_is_down(l))
441 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
442 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
443 tipc_link_fsm_evt(l, LINK_RESET_EVT);
445 tipc_link_build_reset_msg(l, xmitq);
446 *maddr = &n->links[*bearer_id].maddr;
447 node_lost_contact(n, &le->inputq);
448 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
451 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
453 /* There is still a working link => initiate failover */
454 tnl = node_active_link(n, 0);
455 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
456 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
457 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
458 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
460 tipc_link_fsm_evt(l, LINK_RESET_EVT);
461 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
462 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
463 *maddr = &n->links[tnl->bearer_id].maddr;
464 *bearer_id = tnl->bearer_id;
467 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
469 struct tipc_link_entry *le = &n->links[bearer_id];
470 struct tipc_link *l = le->link;
471 struct tipc_media_addr *maddr;
472 struct sk_buff_head xmitq;
477 __skb_queue_head_init(&xmitq);
480 if (!tipc_link_is_establishing(l)) {
481 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
488 /* Defuse pending tipc_node_link_up() */
489 tipc_link_fsm_evt(l, LINK_RESET_EVT);
492 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
493 tipc_sk_rcv(n->net, &le->inputq);
496 bool tipc_node_is_up(struct tipc_node *n)
498 return n->active_links[0] != INVALID_BEARER_ID;
501 void tipc_node_check_dest(struct net *net, u32 onode,
502 struct tipc_bearer *b,
503 u16 capabilities, u32 signature,
504 struct tipc_media_addr *maddr,
505 bool *respond, bool *dupl_addr)
509 struct tipc_link_entry *le;
510 bool addr_match = false;
511 bool sign_match = false;
512 bool link_up = false;
513 bool accept_addr = false;
520 n = tipc_node_create(net, onode, capabilities);
526 le = &n->links[b->identity];
528 /* Prepare to validate requesting node's signature and media address */
530 link_up = l && tipc_link_is_up(l);
531 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
532 sign_match = (signature == n->signature);
534 /* These three flags give us eight permutations: */
536 if (sign_match && addr_match && link_up) {
537 /* All is fine. Do nothing. */
539 } else if (sign_match && addr_match && !link_up) {
540 /* Respond. The link will come up in due time */
542 } else if (sign_match && !addr_match && link_up) {
543 /* Peer has changed i/f address without rebooting.
544 * If so, the link will reset soon, and the next
545 * discovery will be accepted. So we can ignore it.
546 * It may also be an cloned or malicious peer having
547 * chosen the same node address and signature as an
549 * Ignore requests until the link goes down, if ever.
552 } else if (sign_match && !addr_match && !link_up) {
553 /* Peer link has changed i/f address without rebooting.
554 * It may also be a cloned or malicious peer; we can't
555 * distinguish between the two.
556 * The signature is correct, so we must accept.
560 } else if (!sign_match && addr_match && link_up) {
561 /* Peer node rebooted. Two possibilities:
562 * - Delayed re-discovery; this link endpoint has already
563 * reset and re-established contact with the peer, before
564 * receiving a discovery message from that node.
565 * (The peer happened to receive one from this node first).
566 * - The peer came back so fast that our side has not
567 * discovered it yet. Probing from this side will soon
568 * reset the link, since there can be no working link
569 * endpoint at the peer end, and the link will re-establish.
570 * Accept the signature, since it comes from a known peer.
572 n->signature = signature;
573 } else if (!sign_match && addr_match && !link_up) {
574 /* The peer node has rebooted.
575 * Accept signature, since it is a known peer.
577 n->signature = signature;
579 } else if (!sign_match && !addr_match && link_up) {
580 /* Peer rebooted with new address, or a new/duplicate peer.
581 * Ignore until the link goes down, if ever.
584 } else if (!sign_match && !addr_match && !link_up) {
585 /* Peer rebooted with new address, or it is a new peer.
586 * Accept signature and address.
588 n->signature = signature;
596 /* Now create new link if not already existing */
598 if (n->link_cnt == 2) {
599 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
602 if_name = strchr(b->name, ':') + 1;
603 if (!tipc_link_create(net, if_name, b->identity, b->tolerance,
604 b->net_plane, b->mtu, b->priority,
605 b->window, mod(tipc_net(net)->random),
606 tipc_own_addr(net), onode,
608 tipc_bc_sndlink(n->net), n->bc_entry.link,
610 &n->bc_entry.namedq, &l)) {
615 tipc_link_fsm_evt(l, LINK_RESET_EVT);
616 if (n->state == NODE_FAILINGOVER)
617 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
620 tipc_node_calculate_timer(n, l);
621 if (n->link_cnt == 1)
622 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
625 memcpy(&le->maddr, maddr, sizeof(*maddr));
628 if (reset && !tipc_link_is_reset(l))
629 tipc_node_link_down(n, b->identity, false);
633 void tipc_node_delete_links(struct net *net, int bearer_id)
635 struct tipc_net *tn = net_generic(net, tipc_net_id);
639 list_for_each_entry_rcu(n, &tn->node_list, list) {
640 tipc_node_link_down(n, bearer_id, true);
645 static void tipc_node_reset_links(struct tipc_node *n)
647 char addr_string[16];
650 pr_warn("Resetting all links to %s\n",
651 tipc_addr_string_fill(addr_string, n->addr));
653 for (i = 0; i < MAX_BEARERS; i++) {
654 tipc_node_link_down(n, i, false);
658 /* tipc_node_fsm_evt - node finite state machine
659 * Determines when contact is allowed with peer node
661 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
663 int state = n->state;
666 case SELF_DOWN_PEER_DOWN:
668 case SELF_ESTABL_CONTACT_EVT:
669 state = SELF_UP_PEER_COMING;
671 case PEER_ESTABL_CONTACT_EVT:
672 state = SELF_COMING_PEER_UP;
674 case SELF_LOST_CONTACT_EVT:
675 case PEER_LOST_CONTACT_EVT:
677 case NODE_SYNCH_END_EVT:
678 case NODE_SYNCH_BEGIN_EVT:
679 case NODE_FAILOVER_BEGIN_EVT:
680 case NODE_FAILOVER_END_EVT:
685 case SELF_UP_PEER_UP:
687 case SELF_LOST_CONTACT_EVT:
688 state = SELF_DOWN_PEER_LEAVING;
690 case PEER_LOST_CONTACT_EVT:
691 state = SELF_LEAVING_PEER_DOWN;
693 case NODE_SYNCH_BEGIN_EVT:
694 state = NODE_SYNCHING;
696 case NODE_FAILOVER_BEGIN_EVT:
697 state = NODE_FAILINGOVER;
699 case SELF_ESTABL_CONTACT_EVT:
700 case PEER_ESTABL_CONTACT_EVT:
701 case NODE_SYNCH_END_EVT:
702 case NODE_FAILOVER_END_EVT:
708 case SELF_DOWN_PEER_LEAVING:
710 case PEER_LOST_CONTACT_EVT:
711 state = SELF_DOWN_PEER_DOWN;
713 case SELF_ESTABL_CONTACT_EVT:
714 case PEER_ESTABL_CONTACT_EVT:
715 case SELF_LOST_CONTACT_EVT:
717 case NODE_SYNCH_END_EVT:
718 case NODE_SYNCH_BEGIN_EVT:
719 case NODE_FAILOVER_BEGIN_EVT:
720 case NODE_FAILOVER_END_EVT:
725 case SELF_UP_PEER_COMING:
727 case PEER_ESTABL_CONTACT_EVT:
728 state = SELF_UP_PEER_UP;
730 case SELF_LOST_CONTACT_EVT:
731 state = SELF_DOWN_PEER_DOWN;
733 case SELF_ESTABL_CONTACT_EVT:
734 case PEER_LOST_CONTACT_EVT:
735 case NODE_SYNCH_END_EVT:
736 case NODE_FAILOVER_BEGIN_EVT:
738 case NODE_SYNCH_BEGIN_EVT:
739 case NODE_FAILOVER_END_EVT:
744 case SELF_COMING_PEER_UP:
746 case SELF_ESTABL_CONTACT_EVT:
747 state = SELF_UP_PEER_UP;
749 case PEER_LOST_CONTACT_EVT:
750 state = SELF_DOWN_PEER_DOWN;
752 case SELF_LOST_CONTACT_EVT:
753 case PEER_ESTABL_CONTACT_EVT:
755 case NODE_SYNCH_END_EVT:
756 case NODE_SYNCH_BEGIN_EVT:
757 case NODE_FAILOVER_BEGIN_EVT:
758 case NODE_FAILOVER_END_EVT:
763 case SELF_LEAVING_PEER_DOWN:
765 case SELF_LOST_CONTACT_EVT:
766 state = SELF_DOWN_PEER_DOWN;
768 case SELF_ESTABL_CONTACT_EVT:
769 case PEER_ESTABL_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 NODE_FAILINGOVER:
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_FAILOVER_END_EVT:
789 state = SELF_UP_PEER_UP;
791 case NODE_FAILOVER_BEGIN_EVT:
792 case SELF_ESTABL_CONTACT_EVT:
793 case PEER_ESTABL_CONTACT_EVT:
795 case NODE_SYNCH_BEGIN_EVT:
796 case NODE_SYNCH_END_EVT:
803 case SELF_LOST_CONTACT_EVT:
804 state = SELF_DOWN_PEER_LEAVING;
806 case PEER_LOST_CONTACT_EVT:
807 state = SELF_LEAVING_PEER_DOWN;
809 case NODE_SYNCH_END_EVT:
810 state = SELF_UP_PEER_UP;
812 case NODE_FAILOVER_BEGIN_EVT:
813 state = NODE_FAILINGOVER;
815 case NODE_SYNCH_BEGIN_EVT:
816 case SELF_ESTABL_CONTACT_EVT:
817 case PEER_ESTABL_CONTACT_EVT:
819 case NODE_FAILOVER_END_EVT:
825 pr_err("Unknown node fsm state %x\n", state);
832 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
835 bool tipc_node_filter_pkt(struct tipc_node *n, struct tipc_msg *hdr)
837 int state = n->state;
839 if (likely(state == SELF_UP_PEER_UP))
842 if (state == SELF_LEAVING_PEER_DOWN)
845 if (state == SELF_DOWN_PEER_LEAVING) {
846 if (msg_peer_node_is_up(hdr))
853 static void node_lost_contact(struct tipc_node *n,
854 struct sk_buff_head *inputq)
856 char addr_string[16];
857 struct tipc_sock_conn *conn, *safe;
859 struct list_head *conns = &n->conn_sks;
863 pr_debug("Lost contact with %s\n",
864 tipc_addr_string_fill(addr_string, n->addr));
866 /* Clean up broadcast state */
867 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
869 /* Abort any ongoing link failover */
870 for (i = 0; i < MAX_BEARERS; i++) {
871 l = n->links[i].link;
873 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
876 /* Notify publications from this node */
877 n->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, tipc_own_addr(n->net),
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);
946 spin_unlock_bh(&node->lock);
948 if (flags & TIPC_NOTIFY_NODE_DOWN)
949 tipc_publ_notify(net, publ_list, addr);
951 if (flags & TIPC_NOTIFY_NODE_UP)
952 tipc_named_node_up(net, addr);
954 if (flags & TIPC_NOTIFY_LINK_UP)
955 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
956 TIPC_NODE_SCOPE, link_id, addr);
958 if (flags & TIPC_NOTIFY_LINK_DOWN)
959 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
964 /* Caller should hold node lock for the passed node */
965 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
968 struct nlattr *attrs;
970 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
971 NLM_F_MULTI, TIPC_NL_NODE_GET);
975 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
979 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
981 if (tipc_node_is_up(node))
982 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
985 nla_nest_end(msg->skb, attrs);
986 genlmsg_end(msg->skb, hdr);
991 nla_nest_cancel(msg->skb, attrs);
993 genlmsg_cancel(msg->skb, hdr);
998 static struct tipc_link *tipc_node_select_link(struct tipc_node *n, int sel,
1000 struct tipc_media_addr **maddr)
1002 int id = n->active_links[sel & 1];
1004 if (unlikely(id < 0))
1008 *maddr = &n->links[id].maddr;
1009 return n->links[id].link;
1013 * tipc_node_xmit() is the general link level function for message sending
1014 * @net: the applicable net namespace
1015 * @list: chain of buffers containing message
1016 * @dnode: address of destination node
1017 * @selector: a number used for deterministic link selection
1018 * Consumes the buffer chain, except when returning -ELINKCONG
1019 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1021 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1022 u32 dnode, int selector)
1024 struct tipc_link *l = NULL;
1025 struct tipc_node *n;
1026 struct sk_buff_head xmitq;
1027 struct tipc_media_addr *maddr;
1029 int rc = -EHOSTUNREACH;
1031 __skb_queue_head_init(&xmitq);
1032 n = tipc_node_find(net, dnode);
1035 l = tipc_node_select_link(n, selector, &bearer_id, &maddr);
1037 rc = tipc_link_xmit(l, list, &xmitq);
1038 tipc_node_unlock(n);
1039 if (unlikely(rc == -ENOBUFS))
1040 tipc_node_link_down(n, bearer_id, false);
1044 tipc_bearer_xmit(net, bearer_id, &xmitq, maddr);
1047 if (likely(in_own_node(net, dnode))) {
1048 tipc_sk_rcv(net, list);
1054 /* tipc_node_xmit_skb(): send single buffer to destination
1055 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1056 * messages, which will not be rejected
1057 * The only exception is datagram messages rerouted after secondary
1058 * lookup, which are rare and safe to dispose of anyway.
1059 * TODO: Return real return value, and let callers use
1060 * tipc_wait_for_sendpkt() where applicable
1062 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1065 struct sk_buff_head head;
1068 skb_queue_head_init(&head);
1069 __skb_queue_tail(&head, skb);
1070 rc = tipc_node_xmit(net, &head, dnode, selector);
1071 if (rc == -ELINKCONG)
1077 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1078 * @net: the applicable net namespace
1080 * @bearer_id: id of bearer message arrived on
1082 * Invoked with no locks held.
1084 static void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1087 struct sk_buff_head xmitq;
1088 struct tipc_bclink_entry *be;
1089 struct tipc_link_entry *le;
1090 struct tipc_msg *hdr = buf_msg(skb);
1091 int usr = msg_user(hdr);
1092 u32 dnode = msg_destnode(hdr);
1093 struct tipc_node *n;
1095 __skb_queue_head_init(&xmitq);
1097 /* If NACK for other node, let rcv link for that node peek into it */
1098 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1099 n = tipc_node_find(net, dnode);
1101 n = tipc_node_find(net, msg_prevnode(hdr));
1107 le = &n->links[bearer_id];
1109 rc = tipc_bcast_rcv(net, be->link, skb);
1111 /* Broadcast link reset may happen at reassembly failure */
1112 if (rc & TIPC_LINK_DOWN_EVT)
1113 tipc_node_reset_links(n);
1115 /* Broadcast ACKs are sent on a unicast link */
1116 if (rc & TIPC_LINK_SND_BC_ACK) {
1118 tipc_link_build_ack_msg(le->link, &xmitq);
1119 tipc_node_unlock(n);
1122 if (!skb_queue_empty(&xmitq))
1123 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1125 /* Deliver. 'arrvq' is under inputq2's lock protection */
1126 if (!skb_queue_empty(&be->inputq1)) {
1127 spin_lock_bh(&be->inputq2.lock);
1128 spin_lock_bh(&be->inputq1.lock);
1129 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1130 spin_unlock_bh(&be->inputq1.lock);
1131 spin_unlock_bh(&be->inputq2.lock);
1132 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1138 * tipc_node_check_state - check and if necessary update node state
1140 * @bearer_id: identity of bearer delivering the packet
1141 * Returns true if state is ok, otherwise consumes buffer and returns false
1143 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1144 int bearer_id, struct sk_buff_head *xmitq)
1146 struct tipc_msg *hdr = buf_msg(skb);
1147 int usr = msg_user(hdr);
1148 int mtyp = msg_type(hdr);
1149 u16 oseqno = msg_seqno(hdr);
1150 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1151 u16 exp_pkts = msg_msgcnt(hdr);
1152 u16 rcv_nxt, syncpt, dlv_nxt;
1153 int state = n->state;
1154 struct tipc_link *l, *tnl, *pl = NULL;
1155 struct tipc_media_addr *maddr;
1158 l = n->links[bearer_id].link;
1161 rcv_nxt = l->rcv_nxt;
1164 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1167 /* Find parallel link, if any */
1168 for (i = 0; i < MAX_BEARERS; i++) {
1169 if ((i != bearer_id) && n->links[i].link) {
1170 pl = n->links[i].link;
1175 /* Update node accesibility if applicable */
1176 if (state == SELF_UP_PEER_COMING) {
1177 if (!tipc_link_is_up(l))
1179 if (!msg_peer_link_is_up(hdr))
1181 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1184 if (state == SELF_DOWN_PEER_LEAVING) {
1185 if (msg_peer_node_is_up(hdr))
1187 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1190 /* Ignore duplicate packets */
1191 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1194 /* Initiate or update failover mode if applicable */
1195 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1196 syncpt = oseqno + exp_pkts - 1;
1197 if (pl && tipc_link_is_up(pl)) {
1198 pb_id = pl->bearer_id;
1199 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1200 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1202 /* If pkts arrive out of order, use lowest calculated syncpt */
1203 if (less(syncpt, n->sync_point))
1204 n->sync_point = syncpt;
1207 /* Open parallel link when tunnel link reaches synch point */
1208 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1209 if (!more(rcv_nxt, n->sync_point))
1211 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1213 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1217 /* No synching needed if only one link */
1218 if (!pl || !tipc_link_is_up(pl))
1221 /* Initiate synch mode if applicable */
1222 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1223 syncpt = iseqno + exp_pkts - 1;
1224 if (!tipc_link_is_up(l)) {
1225 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1226 __tipc_node_link_up(n, bearer_id, xmitq);
1228 if (n->state == SELF_UP_PEER_UP) {
1229 n->sync_point = syncpt;
1230 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1231 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1233 if (less(syncpt, n->sync_point))
1234 n->sync_point = syncpt;
1237 /* Open tunnel link when parallel link reaches synch point */
1238 if ((n->state == NODE_SYNCHING) && tipc_link_is_synching(l)) {
1239 if (tipc_link_is_synching(l)) {
1245 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1246 if (more(dlv_nxt, n->sync_point)) {
1247 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1248 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1253 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1255 if (usr == LINK_PROTOCOL)
1263 * tipc_rcv - process TIPC packets/messages arriving from off-node
1264 * @net: the applicable net namespace
1266 * @bearer: pointer to bearer message arrived on
1268 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1269 * structure (i.e. cannot be NULL), but bearer can be inactive.
1271 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1273 struct sk_buff_head xmitq;
1274 struct tipc_node *n;
1275 struct tipc_msg *hdr = buf_msg(skb);
1276 int usr = msg_user(hdr);
1277 int bearer_id = b->identity;
1278 struct tipc_link_entry *le;
1279 u16 bc_ack = msg_bcast_ack(hdr);
1282 __skb_queue_head_init(&xmitq);
1284 /* Ensure message is well-formed */
1285 if (unlikely(!tipc_msg_validate(skb)))
1288 /* Handle arrival of discovery or broadcast packet */
1289 if (unlikely(msg_non_seq(hdr))) {
1290 if (unlikely(usr == LINK_CONFIG))
1291 return tipc_disc_rcv(net, skb, b);
1293 return tipc_node_bc_rcv(net, skb, bearer_id);
1296 /* Locate neighboring node that sent packet */
1297 n = tipc_node_find(net, msg_prevnode(hdr));
1300 le = &n->links[bearer_id];
1302 /* Ensure broadcast reception is in synch with peer's send state */
1303 if (unlikely(usr == LINK_PROTOCOL))
1304 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1305 else if (unlikely(n->bc_entry.link->acked != bc_ack))
1306 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1310 /* Is reception permitted at the moment ? */
1311 if (!tipc_node_filter_pkt(n, hdr))
1314 /* Check and if necessary update node state */
1315 if (likely(tipc_node_check_state(n, skb, bearer_id, &xmitq))) {
1316 rc = tipc_link_rcv(le->link, skb, &xmitq);
1320 tipc_node_unlock(n);
1322 if (unlikely(rc & TIPC_LINK_UP_EVT))
1323 tipc_node_link_up(n, bearer_id, &xmitq);
1325 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1326 tipc_node_link_down(n, bearer_id, false);
1328 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1329 tipc_named_rcv(net, &n->bc_entry.namedq);
1331 if (!skb_queue_empty(&le->inputq))
1332 tipc_sk_rcv(net, &le->inputq);
1334 if (!skb_queue_empty(&xmitq))
1335 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1342 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1345 struct net *net = sock_net(skb->sk);
1346 struct tipc_net *tn = net_generic(net, tipc_net_id);
1347 int done = cb->args[0];
1348 int last_addr = cb->args[1];
1349 struct tipc_node *node;
1350 struct tipc_nl_msg msg;
1356 msg.portid = NETLINK_CB(cb->skb).portid;
1357 msg.seq = cb->nlh->nlmsg_seq;
1361 node = tipc_node_find(net, last_addr);
1364 /* We never set seq or call nl_dump_check_consistent()
1365 * this means that setting prev_seq here will cause the
1366 * consistence check to fail in the netlink callback
1367 * handler. Resulting in the NLMSG_DONE message having
1368 * the NLM_F_DUMP_INTR flag set if the node state
1369 * changed while we released the lock.
1374 tipc_node_put(node);
1377 list_for_each_entry_rcu(node, &tn->node_list, list) {
1379 if (node->addr == last_addr)
1385 tipc_node_lock(node);
1386 err = __tipc_nl_add_node(&msg, node);
1388 last_addr = node->addr;
1389 tipc_node_unlock(node);
1393 tipc_node_unlock(node);
1398 cb->args[1] = last_addr;