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(net, 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 void tipc_node_subscribe(struct net *net, struct list_head *subscr, u32 addr)
241 if (in_own_node(net, addr))
244 n = tipc_node_find(net, addr);
246 pr_warn("Node subscribe rejected, unknown node 0x%x\n", addr);
250 list_add_tail(subscr, &n->publ_list);
255 void tipc_node_unsubscribe(struct net *net, struct list_head *subscr, u32 addr)
259 if (in_own_node(net, addr))
262 n = tipc_node_find(net, addr);
264 pr_warn("Node unsubscribe rejected, unknown node 0x%x\n", addr);
268 list_del_init(subscr);
273 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
275 struct tipc_node *node;
276 struct tipc_sock_conn *conn;
279 if (in_own_node(net, dnode))
282 node = tipc_node_find(net, dnode);
284 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
285 return -EHOSTUNREACH;
287 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
292 conn->peer_node = dnode;
294 conn->peer_port = peer_port;
296 tipc_node_lock(node);
297 list_add_tail(&conn->list, &node->conn_sks);
298 tipc_node_unlock(node);
304 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
306 struct tipc_node *node;
307 struct tipc_sock_conn *conn, *safe;
309 if (in_own_node(net, dnode))
312 node = tipc_node_find(net, dnode);
316 tipc_node_lock(node);
317 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
318 if (port != conn->port)
320 list_del(&conn->list);
323 tipc_node_unlock(node);
327 /* tipc_node_timeout - handle expiration of node timer
329 static void tipc_node_timeout(unsigned long data)
331 struct tipc_node *n = (struct tipc_node *)data;
332 struct tipc_link_entry *le;
333 struct sk_buff_head xmitq;
337 __skb_queue_head_init(&xmitq);
339 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
341 le = &n->links[bearer_id];
342 spin_lock_bh(&le->lock);
344 /* Link tolerance may change asynchronously: */
345 tipc_node_calculate_timer(n, le->link);
346 rc = tipc_link_timeout(le->link, &xmitq);
348 spin_unlock_bh(&le->lock);
350 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
351 if (rc & TIPC_LINK_DOWN_EVT)
352 tipc_node_link_down(n, bearer_id, false);
354 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
360 * __tipc_node_link_up - handle addition of link
361 * Node lock must be held by caller
362 * Link becomes active (alone or shared) or standby, depending on its priority.
364 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
365 struct sk_buff_head *xmitq)
367 int *slot0 = &n->active_links[0];
368 int *slot1 = &n->active_links[1];
369 struct tipc_link *ol = node_active_link(n, 0);
370 struct tipc_link *nl = n->links[bearer_id].link;
375 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
376 if (!tipc_link_is_up(nl))
380 n->action_flags |= TIPC_NOTIFY_LINK_UP;
381 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
383 /* Leave room for tunnel header when returning 'mtu' to users: */
384 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
386 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
387 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
389 pr_debug("Established link <%s> on network plane %c\n",
390 nl->name, nl->net_plane);
392 /* First link? => give it both slots */
396 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
397 n->action_flags |= TIPC_NOTIFY_NODE_UP;
398 tipc_bcast_add_peer(n->net, nl, xmitq);
402 /* Second link => redistribute slots */
403 if (nl->priority > ol->priority) {
404 pr_debug("Old link <%s> becomes standby\n", ol->name);
407 tipc_link_set_active(nl, true);
408 tipc_link_set_active(ol, false);
409 } else if (nl->priority == ol->priority) {
410 tipc_link_set_active(nl, true);
413 pr_debug("New link <%s> is standby\n", nl->name);
416 /* Prepare synchronization with first link */
417 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
421 * tipc_node_link_up - handle addition of link
423 * Link becomes active (alone or shared) or standby, depending on its priority.
425 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
426 struct sk_buff_head *xmitq)
429 __tipc_node_link_up(n, bearer_id, xmitq);
434 * __tipc_node_link_down - handle loss of link
436 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
437 struct sk_buff_head *xmitq,
438 struct tipc_media_addr **maddr)
440 struct tipc_link_entry *le = &n->links[*bearer_id];
441 int *slot0 = &n->active_links[0];
442 int *slot1 = &n->active_links[1];
444 struct tipc_link *l, *_l, *tnl;
446 l = n->links[*bearer_id].link;
447 if (!l || tipc_link_is_reset(l))
451 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
452 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
454 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
456 pr_debug("Lost link <%s> on network plane %c\n",
457 l->name, l->net_plane);
459 /* Select new active link if any available */
460 *slot0 = INVALID_BEARER_ID;
461 *slot1 = INVALID_BEARER_ID;
462 for (i = 0; i < MAX_BEARERS; i++) {
463 _l = n->links[i].link;
464 if (!_l || !tipc_link_is_up(_l))
468 if (_l->priority < highest)
470 if (_l->priority > highest) {
471 highest = _l->priority;
479 if (!tipc_node_is_up(n)) {
480 if (tipc_link_peer_is_down(l))
481 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
482 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
483 tipc_link_fsm_evt(l, LINK_RESET_EVT);
485 tipc_link_build_reset_msg(l, xmitq);
486 *maddr = &n->links[*bearer_id].maddr;
487 node_lost_contact(n, &le->inputq);
488 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
491 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
493 /* There is still a working link => initiate failover */
494 tnl = node_active_link(n, 0);
495 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
496 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
497 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
498 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
500 tipc_link_fsm_evt(l, LINK_RESET_EVT);
501 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
502 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
503 *maddr = &n->links[tnl->bearer_id].maddr;
504 *bearer_id = tnl->bearer_id;
507 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
509 struct tipc_link_entry *le = &n->links[bearer_id];
510 struct tipc_link *l = le->link;
511 struct tipc_media_addr *maddr;
512 struct sk_buff_head xmitq;
517 __skb_queue_head_init(&xmitq);
520 if (!tipc_link_is_establishing(l)) {
521 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
528 /* Defuse pending tipc_node_link_up() */
529 tipc_link_fsm_evt(l, LINK_RESET_EVT);
532 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
533 tipc_sk_rcv(n->net, &le->inputq);
536 bool tipc_node_is_up(struct tipc_node *n)
538 return n->active_links[0] != INVALID_BEARER_ID;
541 void tipc_node_check_dest(struct net *net, u32 onode,
542 struct tipc_bearer *b,
543 u16 capabilities, u32 signature,
544 struct tipc_media_addr *maddr,
545 bool *respond, bool *dupl_addr)
549 struct tipc_link_entry *le;
550 bool addr_match = false;
551 bool sign_match = false;
552 bool link_up = false;
553 bool accept_addr = false;
560 n = tipc_node_create(net, onode, capabilities);
566 le = &n->links[b->identity];
568 /* Prepare to validate requesting node's signature and media address */
570 link_up = l && tipc_link_is_up(l);
571 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
572 sign_match = (signature == n->signature);
574 /* These three flags give us eight permutations: */
576 if (sign_match && addr_match && link_up) {
577 /* All is fine. Do nothing. */
579 } else if (sign_match && addr_match && !link_up) {
580 /* Respond. The link will come up in due time */
582 } else if (sign_match && !addr_match && link_up) {
583 /* Peer has changed i/f address without rebooting.
584 * If so, the link will reset soon, and the next
585 * discovery will be accepted. So we can ignore it.
586 * It may also be an cloned or malicious peer having
587 * chosen the same node address and signature as an
589 * Ignore requests until the link goes down, if ever.
592 } else if (sign_match && !addr_match && !link_up) {
593 /* Peer link has changed i/f address without rebooting.
594 * It may also be a cloned or malicious peer; we can't
595 * distinguish between the two.
596 * The signature is correct, so we must accept.
600 } else if (!sign_match && addr_match && link_up) {
601 /* Peer node rebooted. Two possibilities:
602 * - Delayed re-discovery; this link endpoint has already
603 * reset and re-established contact with the peer, before
604 * receiving a discovery message from that node.
605 * (The peer happened to receive one from this node first).
606 * - The peer came back so fast that our side has not
607 * discovered it yet. Probing from this side will soon
608 * reset the link, since there can be no working link
609 * endpoint at the peer end, and the link will re-establish.
610 * Accept the signature, since it comes from a known peer.
612 n->signature = signature;
613 } else if (!sign_match && addr_match && !link_up) {
614 /* The peer node has rebooted.
615 * Accept signature, since it is a known peer.
617 n->signature = signature;
619 } else if (!sign_match && !addr_match && link_up) {
620 /* Peer rebooted with new address, or a new/duplicate peer.
621 * Ignore until the link goes down, if ever.
624 } else if (!sign_match && !addr_match && !link_up) {
625 /* Peer rebooted with new address, or it is a new peer.
626 * Accept signature and address.
628 n->signature = signature;
636 /* Now create new link if not already existing */
638 if (n->link_cnt == 2) {
639 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
642 if_name = strchr(b->name, ':') + 1;
643 if (!tipc_link_create(net, if_name, b->identity, b->tolerance,
644 b->net_plane, b->mtu, b->priority,
645 b->window, mod(tipc_net(net)->random),
646 tipc_own_addr(net), onode,
648 tipc_bc_sndlink(n->net), n->bc_entry.link,
650 &n->bc_entry.namedq, &l)) {
655 tipc_link_fsm_evt(l, LINK_RESET_EVT);
656 if (n->state == NODE_FAILINGOVER)
657 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
659 spin_lock_init(&le->lock);
661 tipc_node_calculate_timer(n, l);
662 if (n->link_cnt == 1)
663 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
666 memcpy(&le->maddr, maddr, sizeof(*maddr));
669 if (reset && !tipc_link_is_reset(l))
670 tipc_node_link_down(n, b->identity, false);
674 void tipc_node_delete_links(struct net *net, int bearer_id)
676 struct tipc_net *tn = net_generic(net, tipc_net_id);
680 list_for_each_entry_rcu(n, &tn->node_list, list) {
681 tipc_node_link_down(n, bearer_id, true);
686 static void tipc_node_reset_links(struct tipc_node *n)
688 char addr_string[16];
691 pr_warn("Resetting all links to %s\n",
692 tipc_addr_string_fill(addr_string, n->addr));
694 for (i = 0; i < MAX_BEARERS; i++) {
695 tipc_node_link_down(n, i, false);
699 /* tipc_node_fsm_evt - node finite state machine
700 * Determines when contact is allowed with peer node
702 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
704 int state = n->state;
707 case SELF_DOWN_PEER_DOWN:
709 case SELF_ESTABL_CONTACT_EVT:
710 state = SELF_UP_PEER_COMING;
712 case PEER_ESTABL_CONTACT_EVT:
713 state = SELF_COMING_PEER_UP;
715 case SELF_LOST_CONTACT_EVT:
716 case PEER_LOST_CONTACT_EVT:
718 case NODE_SYNCH_END_EVT:
719 case NODE_SYNCH_BEGIN_EVT:
720 case NODE_FAILOVER_BEGIN_EVT:
721 case NODE_FAILOVER_END_EVT:
726 case SELF_UP_PEER_UP:
728 case SELF_LOST_CONTACT_EVT:
729 state = SELF_DOWN_PEER_LEAVING;
731 case PEER_LOST_CONTACT_EVT:
732 state = SELF_LEAVING_PEER_DOWN;
734 case NODE_SYNCH_BEGIN_EVT:
735 state = NODE_SYNCHING;
737 case NODE_FAILOVER_BEGIN_EVT:
738 state = NODE_FAILINGOVER;
740 case SELF_ESTABL_CONTACT_EVT:
741 case PEER_ESTABL_CONTACT_EVT:
742 case NODE_SYNCH_END_EVT:
743 case NODE_FAILOVER_END_EVT:
749 case SELF_DOWN_PEER_LEAVING:
751 case PEER_LOST_CONTACT_EVT:
752 state = SELF_DOWN_PEER_DOWN;
754 case SELF_ESTABL_CONTACT_EVT:
755 case PEER_ESTABL_CONTACT_EVT:
756 case SELF_LOST_CONTACT_EVT:
758 case NODE_SYNCH_END_EVT:
759 case NODE_SYNCH_BEGIN_EVT:
760 case NODE_FAILOVER_BEGIN_EVT:
761 case NODE_FAILOVER_END_EVT:
766 case SELF_UP_PEER_COMING:
768 case PEER_ESTABL_CONTACT_EVT:
769 state = SELF_UP_PEER_UP;
771 case SELF_LOST_CONTACT_EVT:
772 state = SELF_DOWN_PEER_LEAVING;
774 case SELF_ESTABL_CONTACT_EVT:
775 case PEER_LOST_CONTACT_EVT:
776 case NODE_SYNCH_END_EVT:
777 case NODE_FAILOVER_BEGIN_EVT:
779 case NODE_SYNCH_BEGIN_EVT:
780 case NODE_FAILOVER_END_EVT:
785 case SELF_COMING_PEER_UP:
787 case SELF_ESTABL_CONTACT_EVT:
788 state = SELF_UP_PEER_UP;
790 case PEER_LOST_CONTACT_EVT:
791 state = SELF_LEAVING_PEER_DOWN;
793 case SELF_LOST_CONTACT_EVT:
794 case PEER_ESTABL_CONTACT_EVT:
796 case NODE_SYNCH_END_EVT:
797 case NODE_SYNCH_BEGIN_EVT:
798 case NODE_FAILOVER_BEGIN_EVT:
799 case NODE_FAILOVER_END_EVT:
804 case SELF_LEAVING_PEER_DOWN:
806 case SELF_LOST_CONTACT_EVT:
807 state = SELF_DOWN_PEER_DOWN;
809 case SELF_ESTABL_CONTACT_EVT:
810 case PEER_ESTABL_CONTACT_EVT:
811 case PEER_LOST_CONTACT_EVT:
813 case NODE_SYNCH_END_EVT:
814 case NODE_SYNCH_BEGIN_EVT:
815 case NODE_FAILOVER_BEGIN_EVT:
816 case NODE_FAILOVER_END_EVT:
821 case NODE_FAILINGOVER:
823 case SELF_LOST_CONTACT_EVT:
824 state = SELF_DOWN_PEER_LEAVING;
826 case PEER_LOST_CONTACT_EVT:
827 state = SELF_LEAVING_PEER_DOWN;
829 case NODE_FAILOVER_END_EVT:
830 state = SELF_UP_PEER_UP;
832 case NODE_FAILOVER_BEGIN_EVT:
833 case SELF_ESTABL_CONTACT_EVT:
834 case PEER_ESTABL_CONTACT_EVT:
836 case NODE_SYNCH_BEGIN_EVT:
837 case NODE_SYNCH_END_EVT:
844 case SELF_LOST_CONTACT_EVT:
845 state = SELF_DOWN_PEER_LEAVING;
847 case PEER_LOST_CONTACT_EVT:
848 state = SELF_LEAVING_PEER_DOWN;
850 case NODE_SYNCH_END_EVT:
851 state = SELF_UP_PEER_UP;
853 case NODE_FAILOVER_BEGIN_EVT:
854 state = NODE_FAILINGOVER;
856 case NODE_SYNCH_BEGIN_EVT:
857 case SELF_ESTABL_CONTACT_EVT:
858 case PEER_ESTABL_CONTACT_EVT:
860 case NODE_FAILOVER_END_EVT:
866 pr_err("Unknown node fsm state %x\n", state);
873 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
876 bool tipc_node_filter_pkt(struct tipc_node *n, struct tipc_msg *hdr)
878 int state = n->state;
880 if (likely(state == SELF_UP_PEER_UP))
883 if (state == SELF_LEAVING_PEER_DOWN)
886 if (state == SELF_DOWN_PEER_LEAVING) {
887 if (msg_peer_node_is_up(hdr))
894 static void node_lost_contact(struct tipc_node *n,
895 struct sk_buff_head *inputq)
897 char addr_string[16];
898 struct tipc_sock_conn *conn, *safe;
900 struct list_head *conns = &n->conn_sks;
904 pr_debug("Lost contact with %s\n",
905 tipc_addr_string_fill(addr_string, n->addr));
907 /* Clean up broadcast state */
908 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
910 /* Abort any ongoing link failover */
911 for (i = 0; i < MAX_BEARERS; i++) {
912 l = n->links[i].link;
914 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
917 /* Notify publications from this node */
918 n->action_flags |= TIPC_NOTIFY_NODE_DOWN;
920 /* Notify sockets connected to node */
921 list_for_each_entry_safe(conn, safe, conns, list) {
922 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
923 SHORT_H_SIZE, 0, tipc_own_addr(n->net),
924 conn->peer_node, conn->port,
925 conn->peer_port, TIPC_ERR_NO_NODE);
927 skb_queue_tail(inputq, skb);
928 list_del(&conn->list);
934 * tipc_node_get_linkname - get the name of a link
936 * @bearer_id: id of the bearer
937 * @node: peer node address
938 * @linkname: link name output buffer
940 * Returns 0 on success
942 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
943 char *linkname, size_t len)
945 struct tipc_link *link;
947 struct tipc_node *node = tipc_node_find(net, addr);
952 if (bearer_id >= MAX_BEARERS)
955 tipc_node_lock(node);
956 link = node->links[bearer_id].link;
958 strncpy(linkname, link->name, len);
962 tipc_node_unlock(node);
967 void tipc_node_unlock(struct tipc_node *node)
969 struct net *net = node->net;
971 u32 flags = node->action_flags;
973 struct list_head *publ_list;
975 if (likely(!flags)) {
976 spin_unlock_bh(&node->lock);
981 link_id = node->link_id;
982 publ_list = &node->publ_list;
984 node->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
985 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP);
987 spin_unlock_bh(&node->lock);
989 if (flags & TIPC_NOTIFY_NODE_DOWN)
990 tipc_publ_notify(net, publ_list, addr);
992 if (flags & TIPC_NOTIFY_NODE_UP)
993 tipc_named_node_up(net, addr);
995 if (flags & TIPC_NOTIFY_LINK_UP)
996 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
997 TIPC_NODE_SCOPE, link_id, addr);
999 if (flags & TIPC_NOTIFY_LINK_DOWN)
1000 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
1005 /* Caller should hold node lock for the passed node */
1006 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
1009 struct nlattr *attrs;
1011 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1012 NLM_F_MULTI, TIPC_NL_NODE_GET);
1016 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
1020 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
1022 if (tipc_node_is_up(node))
1023 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
1026 nla_nest_end(msg->skb, attrs);
1027 genlmsg_end(msg->skb, hdr);
1032 nla_nest_cancel(msg->skb, attrs);
1034 genlmsg_cancel(msg->skb, hdr);
1040 * tipc_node_xmit() is the general link level function for message sending
1041 * @net: the applicable net namespace
1042 * @list: chain of buffers containing message
1043 * @dnode: address of destination node
1044 * @selector: a number used for deterministic link selection
1045 * Consumes the buffer chain, except when returning -ELINKCONG
1046 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1048 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1049 u32 dnode, int selector)
1051 struct tipc_link_entry *le;
1052 struct tipc_node *n;
1053 struct sk_buff_head xmitq;
1054 struct tipc_media_addr *maddr = NULL;
1056 int rc = -EHOSTUNREACH;
1058 __skb_queue_head_init(&xmitq);
1059 n = tipc_node_find(net, dnode);
1062 bearer_id = n->active_links[selector & 1];
1063 if (bearer_id >= 0) {
1064 le = &n->links[bearer_id];
1066 spin_lock_bh(&le->lock);
1067 if (likely(le->link))
1068 rc = tipc_link_xmit(le->link, list, &xmitq);
1069 spin_unlock_bh(&le->lock);
1071 tipc_node_unlock(n);
1072 if (unlikely(rc == -ENOBUFS))
1073 tipc_node_link_down(n, bearer_id, false);
1076 if (likely(!skb_queue_empty(&xmitq))) {
1077 tipc_bearer_xmit(net, bearer_id, &xmitq, maddr);
1080 if (likely(in_own_node(net, dnode))) {
1081 tipc_sk_rcv(net, list);
1087 /* tipc_node_xmit_skb(): send single buffer to destination
1088 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1089 * messages, which will not be rejected
1090 * The only exception is datagram messages rerouted after secondary
1091 * lookup, which are rare and safe to dispose of anyway.
1092 * TODO: Return real return value, and let callers use
1093 * tipc_wait_for_sendpkt() where applicable
1095 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1098 struct sk_buff_head head;
1101 skb_queue_head_init(&head);
1102 __skb_queue_tail(&head, skb);
1103 rc = tipc_node_xmit(net, &head, dnode, selector);
1104 if (rc == -ELINKCONG)
1109 void tipc_node_broadcast(struct net *net, struct sk_buff *skb)
1111 struct sk_buff *txskb;
1112 struct tipc_node *n;
1116 list_for_each_entry_rcu(n, tipc_nodes(net), list) {
1118 if (in_own_node(net, dst))
1120 if (!tipc_node_is_up(n))
1122 txskb = pskb_copy(skb, GFP_ATOMIC);
1125 msg_set_destnode(buf_msg(txskb), dst);
1126 tipc_node_xmit_skb(net, txskb, dst, 0);
1134 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1135 * @net: the applicable net namespace
1137 * @bearer_id: id of bearer message arrived on
1139 * Invoked with no locks held.
1141 static void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1144 struct sk_buff_head xmitq;
1145 struct tipc_bclink_entry *be;
1146 struct tipc_link_entry *le;
1147 struct tipc_msg *hdr = buf_msg(skb);
1148 int usr = msg_user(hdr);
1149 u32 dnode = msg_destnode(hdr);
1150 struct tipc_node *n;
1152 __skb_queue_head_init(&xmitq);
1154 /* If NACK for other node, let rcv link for that node peek into it */
1155 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1156 n = tipc_node_find(net, dnode);
1158 n = tipc_node_find(net, msg_prevnode(hdr));
1164 le = &n->links[bearer_id];
1166 rc = tipc_bcast_rcv(net, be->link, skb);
1168 /* Broadcast link reset may happen at reassembly failure */
1169 if (rc & TIPC_LINK_DOWN_EVT)
1170 tipc_node_reset_links(n);
1172 /* Broadcast ACKs are sent on a unicast link */
1173 if (rc & TIPC_LINK_SND_BC_ACK) {
1175 tipc_link_build_ack_msg(le->link, &xmitq);
1176 tipc_node_unlock(n);
1179 if (!skb_queue_empty(&xmitq))
1180 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1182 /* Deliver. 'arrvq' is under inputq2's lock protection */
1183 if (!skb_queue_empty(&be->inputq1)) {
1184 spin_lock_bh(&be->inputq2.lock);
1185 spin_lock_bh(&be->inputq1.lock);
1186 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1187 spin_unlock_bh(&be->inputq1.lock);
1188 spin_unlock_bh(&be->inputq2.lock);
1189 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1195 * tipc_node_check_state - check and if necessary update node state
1197 * @bearer_id: identity of bearer delivering the packet
1198 * Returns true if state is ok, otherwise consumes buffer and returns false
1200 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1201 int bearer_id, struct sk_buff_head *xmitq)
1203 struct tipc_msg *hdr = buf_msg(skb);
1204 int usr = msg_user(hdr);
1205 int mtyp = msg_type(hdr);
1206 u16 oseqno = msg_seqno(hdr);
1207 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1208 u16 exp_pkts = msg_msgcnt(hdr);
1209 u16 rcv_nxt, syncpt, dlv_nxt;
1210 int state = n->state;
1211 struct tipc_link *l, *tnl, *pl = NULL;
1212 struct tipc_media_addr *maddr;
1215 l = n->links[bearer_id].link;
1218 rcv_nxt = l->rcv_nxt;
1221 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1224 /* Find parallel link, if any */
1225 for (i = 0; i < MAX_BEARERS; i++) {
1226 if ((i != bearer_id) && n->links[i].link) {
1227 pl = n->links[i].link;
1232 /* Update node accesibility if applicable */
1233 if (state == SELF_UP_PEER_COMING) {
1234 if (!tipc_link_is_up(l))
1236 if (!msg_peer_link_is_up(hdr))
1238 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1241 if (state == SELF_DOWN_PEER_LEAVING) {
1242 if (msg_peer_node_is_up(hdr))
1244 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1248 /* Ignore duplicate packets */
1249 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1252 /* Initiate or update failover mode if applicable */
1253 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1254 syncpt = oseqno + exp_pkts - 1;
1255 if (pl && tipc_link_is_up(pl)) {
1256 pb_id = pl->bearer_id;
1257 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1258 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1260 /* If pkts arrive out of order, use lowest calculated syncpt */
1261 if (less(syncpt, n->sync_point))
1262 n->sync_point = syncpt;
1265 /* Open parallel link when tunnel link reaches synch point */
1266 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1267 if (!more(rcv_nxt, n->sync_point))
1269 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1271 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1275 /* No synching needed if only one link */
1276 if (!pl || !tipc_link_is_up(pl))
1279 /* Initiate synch mode if applicable */
1280 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1281 syncpt = iseqno + exp_pkts - 1;
1282 if (!tipc_link_is_up(l)) {
1283 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1284 __tipc_node_link_up(n, bearer_id, xmitq);
1286 if (n->state == SELF_UP_PEER_UP) {
1287 n->sync_point = syncpt;
1288 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1289 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1293 /* Open tunnel link when parallel link reaches synch point */
1294 if (n->state == NODE_SYNCHING) {
1295 if (tipc_link_is_synching(l)) {
1301 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1302 if (more(dlv_nxt, n->sync_point)) {
1303 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1304 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1309 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1311 if (usr == LINK_PROTOCOL)
1319 * tipc_rcv - process TIPC packets/messages arriving from off-node
1320 * @net: the applicable net namespace
1322 * @bearer: pointer to bearer message arrived on
1324 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1325 * structure (i.e. cannot be NULL), but bearer can be inactive.
1327 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1329 struct sk_buff_head xmitq;
1330 struct tipc_node *n;
1331 struct tipc_msg *hdr = buf_msg(skb);
1332 int usr = msg_user(hdr);
1333 int bearer_id = b->identity;
1334 struct tipc_link_entry *le;
1335 u16 bc_ack = msg_bcast_ack(hdr);
1338 __skb_queue_head_init(&xmitq);
1340 /* Ensure message is well-formed */
1341 if (unlikely(!tipc_msg_validate(skb)))
1344 /* Handle arrival of discovery or broadcast packet */
1345 if (unlikely(msg_non_seq(hdr))) {
1346 if (unlikely(usr == LINK_CONFIG))
1347 return tipc_disc_rcv(net, skb, b);
1349 return tipc_node_bc_rcv(net, skb, bearer_id);
1352 /* Locate neighboring node that sent packet */
1353 n = tipc_node_find(net, msg_prevnode(hdr));
1356 le = &n->links[bearer_id];
1358 /* Ensure broadcast reception is in synch with peer's send state */
1359 if (unlikely(usr == LINK_PROTOCOL))
1360 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1361 else if (unlikely(n->bc_entry.link->acked != bc_ack))
1362 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1366 /* Is reception permitted at the moment ? */
1367 if (!tipc_node_filter_pkt(n, hdr))
1370 /* Check and if necessary update node state */
1371 if (likely(tipc_node_check_state(n, skb, bearer_id, &xmitq))) {
1372 spin_lock_bh(&le->lock);
1373 rc = tipc_link_rcv(le->link, skb, &xmitq);
1374 spin_unlock_bh(&le->lock);
1378 tipc_node_unlock(n);
1380 if (unlikely(rc & TIPC_LINK_UP_EVT))
1381 tipc_node_link_up(n, bearer_id, &xmitq);
1383 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1384 tipc_node_link_down(n, bearer_id, false);
1386 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1387 tipc_named_rcv(net, &n->bc_entry.namedq);
1389 if (!skb_queue_empty(&le->inputq))
1390 tipc_sk_rcv(net, &le->inputq);
1392 if (!skb_queue_empty(&xmitq))
1393 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1400 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1403 struct net *net = sock_net(skb->sk);
1404 struct tipc_net *tn = net_generic(net, tipc_net_id);
1405 int done = cb->args[0];
1406 int last_addr = cb->args[1];
1407 struct tipc_node *node;
1408 struct tipc_nl_msg msg;
1414 msg.portid = NETLINK_CB(cb->skb).portid;
1415 msg.seq = cb->nlh->nlmsg_seq;
1419 node = tipc_node_find(net, last_addr);
1422 /* We never set seq or call nl_dump_check_consistent()
1423 * this means that setting prev_seq here will cause the
1424 * consistence check to fail in the netlink callback
1425 * handler. Resulting in the NLMSG_DONE message having
1426 * the NLM_F_DUMP_INTR flag set if the node state
1427 * changed while we released the lock.
1432 tipc_node_put(node);
1435 list_for_each_entry_rcu(node, &tn->node_list, list) {
1437 if (node->addr == last_addr)
1443 tipc_node_lock(node);
1444 err = __tipc_nl_add_node(&msg, node);
1446 last_addr = node->addr;
1447 tipc_node_unlock(node);
1451 tipc_node_unlock(node);
1456 cb->args[1] = last_addr;