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 struct tipc_node *tipc_node_create(struct net *net, u32 addr, u16 capabilities)
146 struct tipc_net *tn = net_generic(net, tipc_net_id);
147 struct tipc_node *n_ptr, *temp_node;
149 spin_lock_bh(&tn->node_list_lock);
150 n_ptr = tipc_node_find(net, addr);
153 n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
155 pr_warn("Node creation failed, no memory\n");
160 n_ptr->capabilities = capabilities;
161 kref_init(&n_ptr->kref);
162 spin_lock_init(&n_ptr->lock);
163 INIT_HLIST_NODE(&n_ptr->hash);
164 INIT_LIST_HEAD(&n_ptr->list);
165 INIT_LIST_HEAD(&n_ptr->publ_list);
166 INIT_LIST_HEAD(&n_ptr->conn_sks);
167 skb_queue_head_init(&n_ptr->bc_entry.namedq);
168 skb_queue_head_init(&n_ptr->bc_entry.inputq1);
169 __skb_queue_head_init(&n_ptr->bc_entry.arrvq);
170 skb_queue_head_init(&n_ptr->bc_entry.inputq2);
171 hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
172 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
173 if (n_ptr->addr < temp_node->addr)
176 list_add_tail_rcu(&n_ptr->list, &temp_node->list);
177 n_ptr->state = SELF_DOWN_PEER_LEAVING;
178 n_ptr->signature = INVALID_NODE_SIG;
179 n_ptr->active_links[0] = INVALID_BEARER_ID;
180 n_ptr->active_links[1] = INVALID_BEARER_ID;
181 if (!tipc_link_bc_create(n_ptr, tipc_own_addr(net), n_ptr->addr,
182 U16_MAX, tipc_bc_sndlink(net)->window,
184 &n_ptr->bc_entry.inputq1,
185 &n_ptr->bc_entry.namedq,
186 tipc_bc_sndlink(net),
187 &n_ptr->bc_entry.link)) {
188 pr_warn("Broadcast rcv link creation failed, no memory\n");
193 tipc_node_get(n_ptr);
194 setup_timer(&n_ptr->timer, tipc_node_timeout, (unsigned long)n_ptr);
195 n_ptr->keepalive_intv = U32_MAX;
197 spin_unlock_bh(&tn->node_list_lock);
201 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
203 unsigned long tol = l->tolerance;
204 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
205 unsigned long keepalive_intv = msecs_to_jiffies(intv);
207 /* Link with lowest tolerance determines timer interval */
208 if (keepalive_intv < n->keepalive_intv)
209 n->keepalive_intv = keepalive_intv;
211 /* Ensure link's abort limit corresponds to current interval */
212 l->abort_limit = l->tolerance / jiffies_to_msecs(n->keepalive_intv);
215 static void tipc_node_delete(struct tipc_node *node)
217 list_del_rcu(&node->list);
218 hlist_del_rcu(&node->hash);
219 kfree(node->bc_entry.link);
220 kfree_rcu(node, rcu);
223 void tipc_node_stop(struct net *net)
225 struct tipc_net *tn = net_generic(net, tipc_net_id);
226 struct tipc_node *node, *t_node;
228 spin_lock_bh(&tn->node_list_lock);
229 list_for_each_entry_safe(node, t_node, &tn->node_list, list) {
230 if (del_timer(&node->timer))
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 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
322 * __tipc_node_link_up - handle addition of link
323 * Node lock must be held by caller
324 * Link becomes active (alone or shared) or standby, depending on its priority.
326 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
327 struct sk_buff_head *xmitq)
329 int *slot0 = &n->active_links[0];
330 int *slot1 = &n->active_links[1];
331 struct tipc_link *ol = node_active_link(n, 0);
332 struct tipc_link *nl = n->links[bearer_id].link;
337 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
338 if (!tipc_link_is_up(nl))
342 n->action_flags |= TIPC_NOTIFY_LINK_UP;
343 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
345 /* Leave room for tunnel header when returning 'mtu' to users: */
346 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
348 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
349 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
351 pr_debug("Established link <%s> on network plane %c\n",
352 nl->name, nl->net_plane);
354 /* First link? => give it both slots */
358 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
359 n->action_flags |= TIPC_NOTIFY_NODE_UP;
360 tipc_bcast_add_peer(n->net, nl, xmitq);
364 /* Second link => redistribute slots */
365 if (nl->priority > ol->priority) {
366 pr_debug("Old link <%s> becomes standby\n", ol->name);
369 } else if (nl->priority == ol->priority) {
372 pr_debug("New link <%s> is standby\n", nl->name);
375 /* Prepare synchronization with first link */
376 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
380 * tipc_node_link_up - handle addition of link
382 * Link becomes active (alone or shared) or standby, depending on its priority.
384 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
385 struct sk_buff_head *xmitq)
388 __tipc_node_link_up(n, bearer_id, xmitq);
393 * __tipc_node_link_down - handle loss of link
395 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
396 struct sk_buff_head *xmitq,
397 struct tipc_media_addr **maddr)
399 struct tipc_link_entry *le = &n->links[*bearer_id];
400 int *slot0 = &n->active_links[0];
401 int *slot1 = &n->active_links[1];
403 struct tipc_link *l, *_l, *tnl;
405 l = n->links[*bearer_id].link;
406 if (!l || tipc_link_is_reset(l))
410 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
411 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
413 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
415 pr_debug("Lost link <%s> on network plane %c\n",
416 l->name, l->net_plane);
418 /* Select new active link if any available */
419 *slot0 = INVALID_BEARER_ID;
420 *slot1 = INVALID_BEARER_ID;
421 for (i = 0; i < MAX_BEARERS; i++) {
422 _l = n->links[i].link;
423 if (!_l || !tipc_link_is_up(_l))
427 if (_l->priority < highest)
429 if (_l->priority > highest) {
430 highest = _l->priority;
438 if (!tipc_node_is_up(n)) {
439 if (tipc_link_peer_is_down(l))
440 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
441 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
442 tipc_link_fsm_evt(l, LINK_RESET_EVT);
444 tipc_link_build_reset_msg(l, xmitq);
445 *maddr = &n->links[*bearer_id].maddr;
446 node_lost_contact(n, &le->inputq);
447 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
450 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
452 /* There is still a working link => initiate failover */
453 tnl = node_active_link(n, 0);
454 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
455 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
456 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
457 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
459 tipc_link_fsm_evt(l, LINK_RESET_EVT);
460 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
461 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
462 *maddr = &n->links[tnl->bearer_id].maddr;
463 *bearer_id = tnl->bearer_id;
466 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
468 struct tipc_link_entry *le = &n->links[bearer_id];
469 struct tipc_link *l = le->link;
470 struct tipc_media_addr *maddr;
471 struct sk_buff_head xmitq;
476 __skb_queue_head_init(&xmitq);
479 if (!tipc_link_is_establishing(l)) {
480 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
487 /* Defuse pending tipc_node_link_up() */
488 tipc_link_fsm_evt(l, LINK_RESET_EVT);
491 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
492 tipc_sk_rcv(n->net, &le->inputq);
495 bool tipc_node_is_up(struct tipc_node *n)
497 return n->active_links[0] != INVALID_BEARER_ID;
500 void tipc_node_check_dest(struct net *net, u32 onode,
501 struct tipc_bearer *b,
502 u16 capabilities, u32 signature,
503 struct tipc_media_addr *maddr,
504 bool *respond, bool *dupl_addr)
508 struct tipc_link_entry *le;
509 bool addr_match = false;
510 bool sign_match = false;
511 bool link_up = false;
512 bool accept_addr = false;
519 n = tipc_node_create(net, onode, capabilities);
525 le = &n->links[b->identity];
527 /* Prepare to validate requesting node's signature and media address */
529 link_up = l && tipc_link_is_up(l);
530 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
531 sign_match = (signature == n->signature);
533 /* These three flags give us eight permutations: */
535 if (sign_match && addr_match && link_up) {
536 /* All is fine. Do nothing. */
538 } else if (sign_match && addr_match && !link_up) {
539 /* Respond. The link will come up in due time */
541 } else if (sign_match && !addr_match && link_up) {
542 /* Peer has changed i/f address without rebooting.
543 * If so, the link will reset soon, and the next
544 * discovery will be accepted. So we can ignore it.
545 * It may also be an cloned or malicious peer having
546 * chosen the same node address and signature as an
548 * Ignore requests until the link goes down, if ever.
551 } else if (sign_match && !addr_match && !link_up) {
552 /* Peer link has changed i/f address without rebooting.
553 * It may also be a cloned or malicious peer; we can't
554 * distinguish between the two.
555 * The signature is correct, so we must accept.
559 } else if (!sign_match && addr_match && link_up) {
560 /* Peer node rebooted. Two possibilities:
561 * - Delayed re-discovery; this link endpoint has already
562 * reset and re-established contact with the peer, before
563 * receiving a discovery message from that node.
564 * (The peer happened to receive one from this node first).
565 * - The peer came back so fast that our side has not
566 * discovered it yet. Probing from this side will soon
567 * reset the link, since there can be no working link
568 * endpoint at the peer end, and the link will re-establish.
569 * Accept the signature, since it comes from a known peer.
571 n->signature = signature;
572 } else if (!sign_match && addr_match && !link_up) {
573 /* The peer node has rebooted.
574 * Accept signature, since it is a known peer.
576 n->signature = signature;
578 } else if (!sign_match && !addr_match && link_up) {
579 /* Peer rebooted with new address, or a new/duplicate peer.
580 * Ignore until the link goes down, if ever.
583 } else if (!sign_match && !addr_match && !link_up) {
584 /* Peer rebooted with new address, or it is a new peer.
585 * Accept signature and address.
587 n->signature = signature;
595 /* Now create new link if not already existing */
597 if (n->link_cnt == 2) {
598 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
601 if_name = strchr(b->name, ':') + 1;
602 if (!tipc_link_create(n, if_name, b->identity, b->tolerance,
603 b->net_plane, b->mtu, b->priority,
604 b->window, mod(tipc_net(net)->random),
605 tipc_own_addr(net), onode,
606 n->capabilities, &le->maddr,
607 tipc_bc_sndlink(n->net), n->bc_entry.link,
609 &n->bc_entry.namedq, &l)) {
614 tipc_link_fsm_evt(l, LINK_RESET_EVT);
615 if (n->state == NODE_FAILINGOVER)
616 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
619 tipc_node_calculate_timer(n, l);
620 if (n->link_cnt == 1)
621 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
624 memcpy(&le->maddr, maddr, sizeof(*maddr));
627 if (reset && !tipc_link_is_reset(l))
628 tipc_node_link_down(n, b->identity, false);
632 void tipc_node_delete_links(struct net *net, int bearer_id)
634 struct tipc_net *tn = net_generic(net, tipc_net_id);
638 list_for_each_entry_rcu(n, &tn->node_list, list) {
639 tipc_node_link_down(n, bearer_id, true);
644 static void tipc_node_reset_links(struct tipc_node *n)
646 char addr_string[16];
649 pr_warn("Resetting all links to %s\n",
650 tipc_addr_string_fill(addr_string, n->addr));
652 for (i = 0; i < MAX_BEARERS; i++) {
653 tipc_node_link_down(n, i, false);
657 /* tipc_node_fsm_evt - node finite state machine
658 * Determines when contact is allowed with peer node
660 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
662 int state = n->state;
665 case SELF_DOWN_PEER_DOWN:
667 case SELF_ESTABL_CONTACT_EVT:
668 state = SELF_UP_PEER_COMING;
670 case PEER_ESTABL_CONTACT_EVT:
671 state = SELF_COMING_PEER_UP;
673 case SELF_LOST_CONTACT_EVT:
674 case PEER_LOST_CONTACT_EVT:
676 case NODE_SYNCH_END_EVT:
677 case NODE_SYNCH_BEGIN_EVT:
678 case NODE_FAILOVER_BEGIN_EVT:
679 case NODE_FAILOVER_END_EVT:
684 case SELF_UP_PEER_UP:
686 case SELF_LOST_CONTACT_EVT:
687 state = SELF_DOWN_PEER_LEAVING;
689 case PEER_LOST_CONTACT_EVT:
690 state = SELF_LEAVING_PEER_DOWN;
692 case NODE_SYNCH_BEGIN_EVT:
693 state = NODE_SYNCHING;
695 case NODE_FAILOVER_BEGIN_EVT:
696 state = NODE_FAILINGOVER;
698 case SELF_ESTABL_CONTACT_EVT:
699 case PEER_ESTABL_CONTACT_EVT:
700 case NODE_SYNCH_END_EVT:
701 case NODE_FAILOVER_END_EVT:
707 case SELF_DOWN_PEER_LEAVING:
709 case PEER_LOST_CONTACT_EVT:
710 state = SELF_DOWN_PEER_DOWN;
712 case SELF_ESTABL_CONTACT_EVT:
713 case PEER_ESTABL_CONTACT_EVT:
714 case SELF_LOST_CONTACT_EVT:
716 case NODE_SYNCH_END_EVT:
717 case NODE_SYNCH_BEGIN_EVT:
718 case NODE_FAILOVER_BEGIN_EVT:
719 case NODE_FAILOVER_END_EVT:
724 case SELF_UP_PEER_COMING:
726 case PEER_ESTABL_CONTACT_EVT:
727 state = SELF_UP_PEER_UP;
729 case SELF_LOST_CONTACT_EVT:
730 state = SELF_DOWN_PEER_LEAVING;
732 case SELF_ESTABL_CONTACT_EVT:
733 case PEER_LOST_CONTACT_EVT:
734 case NODE_SYNCH_END_EVT:
735 case NODE_FAILOVER_BEGIN_EVT:
737 case NODE_SYNCH_BEGIN_EVT:
738 case NODE_FAILOVER_END_EVT:
743 case SELF_COMING_PEER_UP:
745 case SELF_ESTABL_CONTACT_EVT:
746 state = SELF_UP_PEER_UP;
748 case PEER_LOST_CONTACT_EVT:
749 state = SELF_LEAVING_PEER_DOWN;
751 case SELF_LOST_CONTACT_EVT:
752 case PEER_ESTABL_CONTACT_EVT:
754 case NODE_SYNCH_END_EVT:
755 case NODE_SYNCH_BEGIN_EVT:
756 case NODE_FAILOVER_BEGIN_EVT:
757 case NODE_FAILOVER_END_EVT:
762 case SELF_LEAVING_PEER_DOWN:
764 case SELF_LOST_CONTACT_EVT:
765 state = SELF_DOWN_PEER_DOWN;
767 case SELF_ESTABL_CONTACT_EVT:
768 case PEER_ESTABL_CONTACT_EVT:
769 case PEER_LOST_CONTACT_EVT:
771 case NODE_SYNCH_END_EVT:
772 case NODE_SYNCH_BEGIN_EVT:
773 case NODE_FAILOVER_BEGIN_EVT:
774 case NODE_FAILOVER_END_EVT:
779 case NODE_FAILINGOVER:
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_FAILOVER_END_EVT:
788 state = SELF_UP_PEER_UP;
790 case NODE_FAILOVER_BEGIN_EVT:
791 case SELF_ESTABL_CONTACT_EVT:
792 case PEER_ESTABL_CONTACT_EVT:
794 case NODE_SYNCH_BEGIN_EVT:
795 case NODE_SYNCH_END_EVT:
802 case SELF_LOST_CONTACT_EVT:
803 state = SELF_DOWN_PEER_LEAVING;
805 case PEER_LOST_CONTACT_EVT:
806 state = SELF_LEAVING_PEER_DOWN;
808 case NODE_SYNCH_END_EVT:
809 state = SELF_UP_PEER_UP;
811 case NODE_FAILOVER_BEGIN_EVT:
812 state = NODE_FAILINGOVER;
814 case NODE_SYNCH_BEGIN_EVT:
815 case SELF_ESTABL_CONTACT_EVT:
816 case PEER_ESTABL_CONTACT_EVT:
818 case NODE_FAILOVER_END_EVT:
824 pr_err("Unknown node fsm state %x\n", state);
831 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
834 bool tipc_node_filter_pkt(struct tipc_node *n, struct tipc_msg *hdr)
836 int state = n->state;
838 if (likely(state == SELF_UP_PEER_UP))
841 if (state == SELF_LEAVING_PEER_DOWN)
844 if (state == SELF_DOWN_PEER_LEAVING) {
845 if (msg_peer_node_is_up(hdr))
852 static void node_lost_contact(struct tipc_node *n,
853 struct sk_buff_head *inputq)
855 char addr_string[16];
856 struct tipc_sock_conn *conn, *safe;
858 struct list_head *conns = &n->conn_sks;
862 pr_debug("Lost contact with %s\n",
863 tipc_addr_string_fill(addr_string, n->addr));
865 /* Clean up broadcast state */
866 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
868 /* Abort any ongoing link failover */
869 for (i = 0; i < MAX_BEARERS; i++) {
870 l = n->links[i].link;
872 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
875 /* Notify publications from this node */
876 n->action_flags |= TIPC_NOTIFY_NODE_DOWN;
878 /* Notify sockets connected to node */
879 list_for_each_entry_safe(conn, safe, conns, list) {
880 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
881 SHORT_H_SIZE, 0, tipc_own_addr(n->net),
882 conn->peer_node, conn->port,
883 conn->peer_port, TIPC_ERR_NO_NODE);
885 skb_queue_tail(inputq, skb);
886 list_del(&conn->list);
892 * tipc_node_get_linkname - get the name of a link
894 * @bearer_id: id of the bearer
895 * @node: peer node address
896 * @linkname: link name output buffer
898 * Returns 0 on success
900 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
901 char *linkname, size_t len)
903 struct tipc_link *link;
905 struct tipc_node *node = tipc_node_find(net, addr);
910 if (bearer_id >= MAX_BEARERS)
913 tipc_node_lock(node);
914 link = node->links[bearer_id].link;
916 strncpy(linkname, link->name, len);
920 tipc_node_unlock(node);
925 void tipc_node_unlock(struct tipc_node *node)
927 struct net *net = node->net;
929 u32 flags = node->action_flags;
931 struct list_head *publ_list;
933 if (likely(!flags)) {
934 spin_unlock_bh(&node->lock);
939 link_id = node->link_id;
940 publ_list = &node->publ_list;
942 node->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
943 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP |
944 TIPC_WAKEUP_BCAST_USERS | TIPC_BCAST_MSG_EVT |
947 spin_unlock_bh(&node->lock);
949 if (flags & TIPC_NOTIFY_NODE_DOWN)
950 tipc_publ_notify(net, publ_list, addr);
952 if (flags & TIPC_WAKEUP_BCAST_USERS)
953 tipc_bclink_wakeup_users(net);
955 if (flags & TIPC_NOTIFY_NODE_UP)
956 tipc_named_node_up(net, addr);
958 if (flags & TIPC_NOTIFY_LINK_UP)
959 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
960 TIPC_NODE_SCOPE, link_id, addr);
962 if (flags & TIPC_NOTIFY_LINK_DOWN)
963 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
966 if (flags & TIPC_BCAST_MSG_EVT)
967 tipc_bclink_input(net);
969 if (flags & TIPC_BCAST_RESET)
970 tipc_node_reset_links(node);
973 /* Caller should hold node lock for the passed node */
974 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
977 struct nlattr *attrs;
979 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
980 NLM_F_MULTI, TIPC_NL_NODE_GET);
984 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
988 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
990 if (tipc_node_is_up(node))
991 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
994 nla_nest_end(msg->skb, attrs);
995 genlmsg_end(msg->skb, hdr);
1000 nla_nest_cancel(msg->skb, attrs);
1002 genlmsg_cancel(msg->skb, hdr);
1007 static struct tipc_link *tipc_node_select_link(struct tipc_node *n, int sel,
1009 struct tipc_media_addr **maddr)
1011 int id = n->active_links[sel & 1];
1013 if (unlikely(id < 0))
1017 *maddr = &n->links[id].maddr;
1018 return n->links[id].link;
1022 * tipc_node_xmit() is the general link level function for message sending
1023 * @net: the applicable net namespace
1024 * @list: chain of buffers containing message
1025 * @dnode: address of destination node
1026 * @selector: a number used for deterministic link selection
1027 * Consumes the buffer chain, except when returning -ELINKCONG
1028 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1030 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1031 u32 dnode, int selector)
1033 struct tipc_link *l = NULL;
1034 struct tipc_node *n;
1035 struct sk_buff_head xmitq;
1036 struct tipc_media_addr *maddr;
1038 int rc = -EHOSTUNREACH;
1040 __skb_queue_head_init(&xmitq);
1041 n = tipc_node_find(net, dnode);
1044 l = tipc_node_select_link(n, selector, &bearer_id, &maddr);
1046 rc = tipc_link_xmit(l, list, &xmitq);
1047 tipc_node_unlock(n);
1048 if (unlikely(rc == -ENOBUFS))
1049 tipc_node_link_down(n, bearer_id, false);
1053 tipc_bearer_xmit(net, bearer_id, &xmitq, maddr);
1056 if (likely(in_own_node(net, dnode))) {
1057 tipc_sk_rcv(net, list);
1063 /* tipc_node_xmit_skb(): send single buffer to destination
1064 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1065 * messages, which will not be rejected
1066 * The only exception is datagram messages rerouted after secondary
1067 * lookup, which are rare and safe to dispose of anyway.
1068 * TODO: Return real return value, and let callers use
1069 * tipc_wait_for_sendpkt() where applicable
1071 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1074 struct sk_buff_head head;
1077 skb_queue_head_init(&head);
1078 __skb_queue_tail(&head, skb);
1079 rc = tipc_node_xmit(net, &head, dnode, selector);
1080 if (rc == -ELINKCONG)
1086 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1087 * @net: the applicable net namespace
1089 * @bearer_id: id of bearer message arrived on
1091 * Invoked with no locks held.
1093 void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1096 struct sk_buff_head xmitq;
1097 struct tipc_bclink_entry *be;
1098 struct tipc_link_entry *le;
1099 struct tipc_msg *hdr = buf_msg(skb);
1100 int usr = msg_user(hdr);
1101 u32 dnode = msg_destnode(hdr);
1102 struct tipc_node *n;
1104 __skb_queue_head_init(&xmitq);
1106 /* If NACK for other node, let rcv link for that node peek into it */
1107 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1108 n = tipc_node_find(net, dnode);
1110 n = tipc_node_find(net, msg_prevnode(hdr));
1116 le = &n->links[bearer_id];
1118 rc = tipc_bcast_rcv(net, be->link, skb);
1120 /* Broadcast link reset may happen at reassembly failure */
1121 if (rc & TIPC_LINK_DOWN_EVT)
1122 tipc_node_reset_links(n);
1124 /* Broadcast ACKs are sent on a unicast link */
1125 if (rc & TIPC_LINK_SND_BC_ACK) {
1127 tipc_link_build_ack_msg(le->link, &xmitq);
1128 tipc_node_unlock(n);
1131 if (!skb_queue_empty(&xmitq))
1132 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1134 /* Deliver. 'arrvq' is under inputq2's lock protection */
1135 if (!skb_queue_empty(&be->inputq1)) {
1136 spin_lock_bh(&be->inputq2.lock);
1137 spin_lock_bh(&be->inputq1.lock);
1138 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1139 spin_unlock_bh(&be->inputq1.lock);
1140 spin_unlock_bh(&be->inputq2.lock);
1141 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1147 * tipc_node_check_state - check and if necessary update node state
1149 * @bearer_id: identity of bearer delivering the packet
1150 * Returns true if state is ok, otherwise consumes buffer and returns false
1152 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1153 int bearer_id, struct sk_buff_head *xmitq)
1155 struct tipc_msg *hdr = buf_msg(skb);
1156 int usr = msg_user(hdr);
1157 int mtyp = msg_type(hdr);
1158 u16 oseqno = msg_seqno(hdr);
1159 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1160 u16 exp_pkts = msg_msgcnt(hdr);
1161 u16 rcv_nxt, syncpt, dlv_nxt;
1162 int state = n->state;
1163 struct tipc_link *l, *tnl, *pl = NULL;
1164 struct tipc_media_addr *maddr;
1167 l = n->links[bearer_id].link;
1170 rcv_nxt = l->rcv_nxt;
1173 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1176 /* Find parallel link, if any */
1177 for (i = 0; i < MAX_BEARERS; i++) {
1178 if ((i != bearer_id) && n->links[i].link) {
1179 pl = n->links[i].link;
1184 /* Update node accesibility if applicable */
1185 if (state == SELF_UP_PEER_COMING) {
1186 if (!tipc_link_is_up(l))
1188 if (!msg_peer_link_is_up(hdr))
1190 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1193 if (state == SELF_DOWN_PEER_LEAVING) {
1194 if (msg_peer_node_is_up(hdr))
1196 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1199 /* Ignore duplicate packets */
1200 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1203 /* Initiate or update failover mode if applicable */
1204 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1205 syncpt = oseqno + exp_pkts - 1;
1206 if (pl && tipc_link_is_up(pl)) {
1207 pb_id = pl->bearer_id;
1208 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1209 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1211 /* If pkts arrive out of order, use lowest calculated syncpt */
1212 if (less(syncpt, n->sync_point))
1213 n->sync_point = syncpt;
1216 /* Open parallel link when tunnel link reaches synch point */
1217 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1218 if (!more(rcv_nxt, n->sync_point))
1220 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1222 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1226 /* No synching needed if only one link */
1227 if (!pl || !tipc_link_is_up(pl))
1230 /* Initiate synch mode if applicable */
1231 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1232 syncpt = iseqno + exp_pkts - 1;
1233 if (!tipc_link_is_up(l)) {
1234 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1235 __tipc_node_link_up(n, bearer_id, xmitq);
1237 if (n->state == SELF_UP_PEER_UP) {
1238 n->sync_point = syncpt;
1239 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1240 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1242 if (less(syncpt, n->sync_point))
1243 n->sync_point = syncpt;
1246 /* Open tunnel link when parallel link reaches synch point */
1247 if ((n->state == NODE_SYNCHING) && tipc_link_is_synching(l)) {
1248 if (tipc_link_is_synching(l)) {
1254 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1255 if (more(dlv_nxt, n->sync_point)) {
1256 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1257 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1262 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1264 if (usr == LINK_PROTOCOL)
1272 * tipc_rcv - process TIPC packets/messages arriving from off-node
1273 * @net: the applicable net namespace
1275 * @bearer: pointer to bearer message arrived on
1277 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1278 * structure (i.e. cannot be NULL), but bearer can be inactive.
1280 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1282 struct sk_buff_head xmitq;
1283 struct tipc_node *n;
1284 struct tipc_msg *hdr = buf_msg(skb);
1285 int usr = msg_user(hdr);
1286 int bearer_id = b->identity;
1287 struct tipc_link_entry *le;
1288 u16 bc_ack = msg_bcast_ack(hdr);
1291 __skb_queue_head_init(&xmitq);
1293 /* Ensure message is well-formed */
1294 if (unlikely(!tipc_msg_validate(skb)))
1297 /* Handle arrival of discovery or broadcast packet */
1298 if (unlikely(msg_non_seq(hdr))) {
1299 if (unlikely(usr == LINK_CONFIG))
1300 return tipc_disc_rcv(net, skb, b);
1302 return tipc_node_bc_rcv(net, skb, bearer_id);
1305 /* Locate neighboring node that sent packet */
1306 n = tipc_node_find(net, msg_prevnode(hdr));
1309 le = &n->links[bearer_id];
1311 /* Ensure broadcast reception is in synch with peer's send state */
1312 if (unlikely(usr == LINK_PROTOCOL))
1313 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1314 else if (unlikely(n->bc_entry.link->acked != bc_ack))
1315 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1319 /* Is reception permitted at the moment ? */
1320 if (!tipc_node_filter_pkt(n, hdr))
1323 /* Check and if necessary update node state */
1324 if (likely(tipc_node_check_state(n, skb, bearer_id, &xmitq))) {
1325 rc = tipc_link_rcv(le->link, skb, &xmitq);
1329 tipc_node_unlock(n);
1331 if (unlikely(rc & TIPC_LINK_UP_EVT))
1332 tipc_node_link_up(n, bearer_id, &xmitq);
1334 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1335 tipc_node_link_down(n, bearer_id, false);
1337 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1338 tipc_named_rcv(net, &n->bc_entry.namedq);
1340 if (!skb_queue_empty(&le->inputq))
1341 tipc_sk_rcv(net, &le->inputq);
1343 if (!skb_queue_empty(&xmitq))
1344 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1351 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1354 struct net *net = sock_net(skb->sk);
1355 struct tipc_net *tn = net_generic(net, tipc_net_id);
1356 int done = cb->args[0];
1357 int last_addr = cb->args[1];
1358 struct tipc_node *node;
1359 struct tipc_nl_msg msg;
1365 msg.portid = NETLINK_CB(cb->skb).portid;
1366 msg.seq = cb->nlh->nlmsg_seq;
1370 node = tipc_node_find(net, last_addr);
1373 /* We never set seq or call nl_dump_check_consistent()
1374 * this means that setting prev_seq here will cause the
1375 * consistence check to fail in the netlink callback
1376 * handler. Resulting in the NLMSG_DONE message having
1377 * the NLM_F_DUMP_INTR flag set if the node state
1378 * changed while we released the lock.
1383 tipc_node_put(node);
1386 list_for_each_entry_rcu(node, &tn->node_list, list) {
1388 if (node->addr == last_addr)
1394 tipc_node_lock(node);
1395 err = __tipc_nl_add_node(&msg, node);
1397 last_addr = node->addr;
1398 tipc_node_unlock(node);
1402 tipc_node_unlock(node);
1407 cb->args[1] = last_addr;