mlxsw: fix error return code
[linux-2.6/btrfs-unstable.git] / net / tipc / node.c
blob703875fd6cde204ddeaf630b9a6bd11daec6dbfa
1 /*
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
6 * All rights reserved.
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.
37 #include "core.h"
38 #include "link.h"
39 #include "node.h"
40 #include "name_distr.h"
41 #include "socket.h"
42 #include "bcast.h"
43 #include "discover.h"
45 /* Node FSM states and events:
47 enum {
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,
55 NODE_SYNCHING = 0xcc
58 enum {
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,
73 bool delete);
74 static void node_lost_contact(struct tipc_node *n, struct sk_buff_head *inputq);
75 static void node_established_contact(struct tipc_node *n_ptr);
76 static void tipc_node_delete(struct tipc_node *node);
77 static void tipc_node_timeout(unsigned long data);
78 static void tipc_node_fsm_evt(struct tipc_node *n, int evt);
80 struct tipc_sock_conn {
81 u32 port;
82 u32 peer_port;
83 u32 peer_node;
84 struct list_head list;
87 static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
88 [TIPC_NLA_NODE_UNSPEC] = { .type = NLA_UNSPEC },
89 [TIPC_NLA_NODE_ADDR] = { .type = NLA_U32 },
90 [TIPC_NLA_NODE_UP] = { .type = NLA_FLAG }
94 * A trivial power-of-two bitmask technique is used for speed, since this
95 * operation is done for every incoming TIPC packet. The number of hash table
96 * entries has been chosen so that no hash chain exceeds 8 nodes and will
97 * usually be much smaller (typically only a single node).
99 static unsigned int tipc_hashfn(u32 addr)
101 return addr & (NODE_HTABLE_SIZE - 1);
104 static void tipc_node_kref_release(struct kref *kref)
106 struct tipc_node *node = container_of(kref, struct tipc_node, kref);
108 tipc_node_delete(node);
111 void tipc_node_put(struct tipc_node *node)
113 kref_put(&node->kref, tipc_node_kref_release);
116 static void tipc_node_get(struct tipc_node *node)
118 kref_get(&node->kref);
122 * tipc_node_find - locate specified node object, if it exists
124 struct tipc_node *tipc_node_find(struct net *net, u32 addr)
126 struct tipc_net *tn = net_generic(net, tipc_net_id);
127 struct tipc_node *node;
129 if (unlikely(!in_own_cluster_exact(net, addr)))
130 return NULL;
132 rcu_read_lock();
133 hlist_for_each_entry_rcu(node, &tn->node_htable[tipc_hashfn(addr)],
134 hash) {
135 if (node->addr == addr) {
136 tipc_node_get(node);
137 rcu_read_unlock();
138 return node;
141 rcu_read_unlock();
142 return NULL;
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);
152 if (n_ptr)
153 goto exit;
154 n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
155 if (!n_ptr) {
156 pr_warn("Node creation failed, no memory\n");
157 goto exit;
159 n_ptr->addr = addr;
160 n_ptr->net = net;
161 n_ptr->capabilities = capabilities;
162 kref_init(&n_ptr->kref);
163 spin_lock_init(&n_ptr->lock);
164 INIT_HLIST_NODE(&n_ptr->hash);
165 INIT_LIST_HEAD(&n_ptr->list);
166 INIT_LIST_HEAD(&n_ptr->publ_list);
167 INIT_LIST_HEAD(&n_ptr->conn_sks);
168 skb_queue_head_init(&n_ptr->bclink.namedq);
169 __skb_queue_head_init(&n_ptr->bclink.deferdq);
170 hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
171 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
172 if (n_ptr->addr < temp_node->addr)
173 break;
175 list_add_tail_rcu(&n_ptr->list, &temp_node->list);
176 n_ptr->state = SELF_DOWN_PEER_LEAVING;
177 n_ptr->signature = INVALID_NODE_SIG;
178 n_ptr->active_links[0] = INVALID_BEARER_ID;
179 n_ptr->active_links[1] = INVALID_BEARER_ID;
180 tipc_node_get(n_ptr);
181 setup_timer(&n_ptr->timer, tipc_node_timeout, (unsigned long)n_ptr);
182 n_ptr->keepalive_intv = U32_MAX;
183 exit:
184 spin_unlock_bh(&tn->node_list_lock);
185 return n_ptr;
188 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
190 unsigned long tol = l->tolerance;
191 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
192 unsigned long keepalive_intv = msecs_to_jiffies(intv);
194 /* Link with lowest tolerance determines timer interval */
195 if (keepalive_intv < n->keepalive_intv)
196 n->keepalive_intv = keepalive_intv;
198 /* Ensure link's abort limit corresponds to current interval */
199 l->abort_limit = l->tolerance / jiffies_to_msecs(n->keepalive_intv);
202 static void tipc_node_delete(struct tipc_node *node)
204 list_del_rcu(&node->list);
205 hlist_del_rcu(&node->hash);
206 kfree_rcu(node, rcu);
209 void tipc_node_stop(struct net *net)
211 struct tipc_net *tn = net_generic(net, tipc_net_id);
212 struct tipc_node *node, *t_node;
214 spin_lock_bh(&tn->node_list_lock);
215 list_for_each_entry_safe(node, t_node, &tn->node_list, list) {
216 if (del_timer(&node->timer))
217 tipc_node_put(node);
218 tipc_node_put(node);
220 spin_unlock_bh(&tn->node_list_lock);
223 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
225 struct tipc_node *node;
226 struct tipc_sock_conn *conn;
227 int err = 0;
229 if (in_own_node(net, dnode))
230 return 0;
232 node = tipc_node_find(net, dnode);
233 if (!node) {
234 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
235 return -EHOSTUNREACH;
237 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
238 if (!conn) {
239 err = -EHOSTUNREACH;
240 goto exit;
242 conn->peer_node = dnode;
243 conn->port = port;
244 conn->peer_port = peer_port;
246 tipc_node_lock(node);
247 list_add_tail(&conn->list, &node->conn_sks);
248 tipc_node_unlock(node);
249 exit:
250 tipc_node_put(node);
251 return err;
254 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
256 struct tipc_node *node;
257 struct tipc_sock_conn *conn, *safe;
259 if (in_own_node(net, dnode))
260 return;
262 node = tipc_node_find(net, dnode);
263 if (!node)
264 return;
266 tipc_node_lock(node);
267 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
268 if (port != conn->port)
269 continue;
270 list_del(&conn->list);
271 kfree(conn);
273 tipc_node_unlock(node);
274 tipc_node_put(node);
277 /* tipc_node_timeout - handle expiration of node timer
279 static void tipc_node_timeout(unsigned long data)
281 struct tipc_node *n = (struct tipc_node *)data;
282 struct tipc_link_entry *le;
283 struct sk_buff_head xmitq;
284 int bearer_id;
285 int rc = 0;
287 __skb_queue_head_init(&xmitq);
289 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
290 tipc_node_lock(n);
291 le = &n->links[bearer_id];
292 if (le->link) {
293 /* Link tolerance may change asynchronously: */
294 tipc_node_calculate_timer(n, le->link);
295 rc = tipc_link_timeout(le->link, &xmitq);
297 tipc_node_unlock(n);
298 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
299 if (rc & TIPC_LINK_DOWN_EVT)
300 tipc_node_link_down(n, bearer_id, false);
302 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
303 tipc_node_get(n);
304 tipc_node_put(n);
308 * __tipc_node_link_up - handle addition of link
309 * Node lock must be held by caller
310 * Link becomes active (alone or shared) or standby, depending on its priority.
312 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
313 struct sk_buff_head *xmitq)
315 int *slot0 = &n->active_links[0];
316 int *slot1 = &n->active_links[1];
317 struct tipc_link *ol = node_active_link(n, 0);
318 struct tipc_link *nl = n->links[bearer_id].link;
320 if (!nl || !tipc_link_is_up(nl))
321 return;
323 n->working_links++;
324 n->action_flags |= TIPC_NOTIFY_LINK_UP;
325 n->link_id = nl->peer_bearer_id << 16 | bearer_id;
327 /* Leave room for tunnel header when returning 'mtu' to users: */
328 n->links[bearer_id].mtu = nl->mtu - INT_H_SIZE;
330 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
332 pr_debug("Established link <%s> on network plane %c\n",
333 nl->name, nl->net_plane);
335 /* First link? => give it both slots */
336 if (!ol) {
337 *slot0 = bearer_id;
338 *slot1 = bearer_id;
339 tipc_link_build_bcast_sync_msg(nl, xmitq);
340 node_established_contact(n);
341 return;
344 /* Second link => redistribute slots */
345 if (nl->priority > ol->priority) {
346 pr_debug("Old link <%s> becomes standby\n", ol->name);
347 *slot0 = bearer_id;
348 *slot1 = bearer_id;
349 } else if (nl->priority == ol->priority) {
350 *slot0 = bearer_id;
351 } else {
352 pr_debug("New link <%s> is standby\n", nl->name);
355 /* Prepare synchronization with first link */
356 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
360 * tipc_node_link_up - handle addition of link
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 tipc_node_lock(n);
368 __tipc_node_link_up(n, bearer_id, xmitq);
369 tipc_node_unlock(n);
373 * __tipc_node_link_down - handle loss of link
375 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
376 struct sk_buff_head *xmitq,
377 struct tipc_media_addr **maddr)
379 struct tipc_link_entry *le = &n->links[*bearer_id];
380 int *slot0 = &n->active_links[0];
381 int *slot1 = &n->active_links[1];
382 int i, highest = 0;
383 struct tipc_link *l, *_l, *tnl;
385 l = n->links[*bearer_id].link;
386 if (!l || tipc_link_is_reset(l))
387 return;
389 n->working_links--;
390 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
391 n->link_id = l->peer_bearer_id << 16 | *bearer_id;
393 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
395 pr_debug("Lost link <%s> on network plane %c\n",
396 l->name, l->net_plane);
398 /* Select new active link if any available */
399 *slot0 = INVALID_BEARER_ID;
400 *slot1 = INVALID_BEARER_ID;
401 for (i = 0; i < MAX_BEARERS; i++) {
402 _l = n->links[i].link;
403 if (!_l || !tipc_link_is_up(_l))
404 continue;
405 if (_l == l)
406 continue;
407 if (_l->priority < highest)
408 continue;
409 if (_l->priority > highest) {
410 highest = _l->priority;
411 *slot0 = i;
412 *slot1 = i;
413 continue;
415 *slot1 = i;
418 if (!tipc_node_is_up(n)) {
419 tipc_link_reset(l);
420 node_lost_contact(n, &le->inputq);
421 return;
424 /* There is still a working link => initiate failover */
425 tnl = node_active_link(n, 0);
426 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
427 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
428 n->sync_point = tnl->rcv_nxt + (U16_MAX / 2 - 1);
429 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
430 tipc_link_reset(l);
431 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
432 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
433 *maddr = &n->links[tnl->bearer_id].maddr;
434 *bearer_id = tnl->bearer_id;
437 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
439 struct tipc_link_entry *le = &n->links[bearer_id];
440 struct tipc_media_addr *maddr;
441 struct sk_buff_head xmitq;
443 __skb_queue_head_init(&xmitq);
445 tipc_node_lock(n);
446 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
447 if (delete && le->link) {
448 kfree(le->link);
449 le->link = NULL;
450 n->link_cnt--;
452 tipc_node_unlock(n);
454 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
455 tipc_sk_rcv(n->net, &le->inputq);
458 bool tipc_node_is_up(struct tipc_node *n)
460 return n->active_links[0] != INVALID_BEARER_ID;
463 void tipc_node_check_dest(struct net *net, u32 onode,
464 struct tipc_bearer *b,
465 u16 capabilities, u32 signature,
466 struct tipc_media_addr *maddr,
467 bool *respond, bool *dupl_addr)
469 struct tipc_node *n;
470 struct tipc_link *l;
471 struct tipc_link_entry *le;
472 bool addr_match = false;
473 bool sign_match = false;
474 bool link_up = false;
475 bool accept_addr = false;
476 bool reset = true;
478 *dupl_addr = false;
479 *respond = false;
481 n = tipc_node_create(net, onode, capabilities);
482 if (!n)
483 return;
485 tipc_node_lock(n);
487 le = &n->links[b->identity];
489 /* Prepare to validate requesting node's signature and media address */
490 l = le->link;
491 link_up = l && tipc_link_is_up(l);
492 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
493 sign_match = (signature == n->signature);
495 /* These three flags give us eight permutations: */
497 if (sign_match && addr_match && link_up) {
498 /* All is fine. Do nothing. */
499 reset = false;
500 } else if (sign_match && addr_match && !link_up) {
501 /* Respond. The link will come up in due time */
502 *respond = true;
503 } else if (sign_match && !addr_match && link_up) {
504 /* Peer has changed i/f address without rebooting.
505 * If so, the link will reset soon, and the next
506 * discovery will be accepted. So we can ignore it.
507 * It may also be an cloned or malicious peer having
508 * chosen the same node address and signature as an
509 * existing one.
510 * Ignore requests until the link goes down, if ever.
512 *dupl_addr = true;
513 } else if (sign_match && !addr_match && !link_up) {
514 /* Peer link has changed i/f address without rebooting.
515 * It may also be a cloned or malicious peer; we can't
516 * distinguish between the two.
517 * The signature is correct, so we must accept.
519 accept_addr = true;
520 *respond = true;
521 } else if (!sign_match && addr_match && link_up) {
522 /* Peer node rebooted. Two possibilities:
523 * - Delayed re-discovery; this link endpoint has already
524 * reset and re-established contact with the peer, before
525 * receiving a discovery message from that node.
526 * (The peer happened to receive one from this node first).
527 * - The peer came back so fast that our side has not
528 * discovered it yet. Probing from this side will soon
529 * reset the link, since there can be no working link
530 * endpoint at the peer end, and the link will re-establish.
531 * Accept the signature, since it comes from a known peer.
533 n->signature = signature;
534 } else if (!sign_match && addr_match && !link_up) {
535 /* The peer node has rebooted.
536 * Accept signature, since it is a known peer.
538 n->signature = signature;
539 *respond = true;
540 } else if (!sign_match && !addr_match && link_up) {
541 /* Peer rebooted with new address, or a new/duplicate peer.
542 * Ignore until the link goes down, if ever.
544 *dupl_addr = true;
545 } else if (!sign_match && !addr_match && !link_up) {
546 /* Peer rebooted with new address, or it is a new peer.
547 * Accept signature and address.
549 n->signature = signature;
550 accept_addr = true;
551 *respond = true;
554 if (!accept_addr)
555 goto exit;
557 /* Now create new link if not already existing */
558 if (!l) {
559 if (n->link_cnt == 2) {
560 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
561 goto exit;
563 if (!tipc_link_create(n, b, mod(tipc_net(net)->random),
564 tipc_own_addr(net), onode, &le->maddr,
565 &le->inputq, &n->bclink.namedq, &l)) {
566 *respond = false;
567 goto exit;
569 tipc_link_reset(l);
570 if (n->state == NODE_FAILINGOVER)
571 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
572 le->link = l;
573 n->link_cnt++;
574 tipc_node_calculate_timer(n, l);
575 if (n->link_cnt == 1)
576 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
577 tipc_node_get(n);
579 memcpy(&le->maddr, maddr, sizeof(*maddr));
580 exit:
581 tipc_node_unlock(n);
582 if (reset)
583 tipc_node_link_down(n, b->identity, false);
584 tipc_node_put(n);
587 void tipc_node_delete_links(struct net *net, int bearer_id)
589 struct tipc_net *tn = net_generic(net, tipc_net_id);
590 struct tipc_node *n;
592 rcu_read_lock();
593 list_for_each_entry_rcu(n, &tn->node_list, list) {
594 tipc_node_link_down(n, bearer_id, true);
596 rcu_read_unlock();
599 static void tipc_node_reset_links(struct tipc_node *n)
601 char addr_string[16];
602 int i;
604 pr_warn("Resetting all links to %s\n",
605 tipc_addr_string_fill(addr_string, n->addr));
607 for (i = 0; i < MAX_BEARERS; i++) {
608 tipc_node_link_down(n, i, false);
612 /* tipc_node_fsm_evt - node finite state machine
613 * Determines when contact is allowed with peer node
615 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
617 int state = n->state;
619 switch (state) {
620 case SELF_DOWN_PEER_DOWN:
621 switch (evt) {
622 case SELF_ESTABL_CONTACT_EVT:
623 state = SELF_UP_PEER_COMING;
624 break;
625 case PEER_ESTABL_CONTACT_EVT:
626 state = SELF_COMING_PEER_UP;
627 break;
628 case SELF_LOST_CONTACT_EVT:
629 case PEER_LOST_CONTACT_EVT:
630 break;
631 case NODE_SYNCH_END_EVT:
632 case NODE_SYNCH_BEGIN_EVT:
633 case NODE_FAILOVER_BEGIN_EVT:
634 case NODE_FAILOVER_END_EVT:
635 default:
636 goto illegal_evt;
638 break;
639 case SELF_UP_PEER_UP:
640 switch (evt) {
641 case SELF_LOST_CONTACT_EVT:
642 state = SELF_DOWN_PEER_LEAVING;
643 break;
644 case PEER_LOST_CONTACT_EVT:
645 state = SELF_LEAVING_PEER_DOWN;
646 break;
647 case NODE_SYNCH_BEGIN_EVT:
648 state = NODE_SYNCHING;
649 break;
650 case NODE_FAILOVER_BEGIN_EVT:
651 state = NODE_FAILINGOVER;
652 break;
653 case SELF_ESTABL_CONTACT_EVT:
654 case PEER_ESTABL_CONTACT_EVT:
655 case NODE_SYNCH_END_EVT:
656 case NODE_FAILOVER_END_EVT:
657 break;
658 default:
659 goto illegal_evt;
661 break;
662 case SELF_DOWN_PEER_LEAVING:
663 switch (evt) {
664 case PEER_LOST_CONTACT_EVT:
665 state = SELF_DOWN_PEER_DOWN;
666 break;
667 case SELF_ESTABL_CONTACT_EVT:
668 case PEER_ESTABL_CONTACT_EVT:
669 case SELF_LOST_CONTACT_EVT:
670 break;
671 case NODE_SYNCH_END_EVT:
672 case NODE_SYNCH_BEGIN_EVT:
673 case NODE_FAILOVER_BEGIN_EVT:
674 case NODE_FAILOVER_END_EVT:
675 default:
676 goto illegal_evt;
678 break;
679 case SELF_UP_PEER_COMING:
680 switch (evt) {
681 case PEER_ESTABL_CONTACT_EVT:
682 state = SELF_UP_PEER_UP;
683 break;
684 case SELF_LOST_CONTACT_EVT:
685 state = SELF_DOWN_PEER_LEAVING;
686 break;
687 case SELF_ESTABL_CONTACT_EVT:
688 case PEER_LOST_CONTACT_EVT:
689 break;
690 case NODE_SYNCH_END_EVT:
691 case NODE_SYNCH_BEGIN_EVT:
692 case NODE_FAILOVER_BEGIN_EVT:
693 case NODE_FAILOVER_END_EVT:
694 default:
695 goto illegal_evt;
697 break;
698 case SELF_COMING_PEER_UP:
699 switch (evt) {
700 case SELF_ESTABL_CONTACT_EVT:
701 state = SELF_UP_PEER_UP;
702 break;
703 case PEER_LOST_CONTACT_EVT:
704 state = SELF_LEAVING_PEER_DOWN;
705 break;
706 case SELF_LOST_CONTACT_EVT:
707 case PEER_ESTABL_CONTACT_EVT:
708 break;
709 case NODE_SYNCH_END_EVT:
710 case NODE_SYNCH_BEGIN_EVT:
711 case NODE_FAILOVER_BEGIN_EVT:
712 case NODE_FAILOVER_END_EVT:
713 default:
714 goto illegal_evt;
716 break;
717 case SELF_LEAVING_PEER_DOWN:
718 switch (evt) {
719 case SELF_LOST_CONTACT_EVT:
720 state = SELF_DOWN_PEER_DOWN;
721 break;
722 case SELF_ESTABL_CONTACT_EVT:
723 case PEER_ESTABL_CONTACT_EVT:
724 case PEER_LOST_CONTACT_EVT:
725 break;
726 case NODE_SYNCH_END_EVT:
727 case NODE_SYNCH_BEGIN_EVT:
728 case NODE_FAILOVER_BEGIN_EVT:
729 case NODE_FAILOVER_END_EVT:
730 default:
731 goto illegal_evt;
733 break;
734 case NODE_FAILINGOVER:
735 switch (evt) {
736 case SELF_LOST_CONTACT_EVT:
737 state = SELF_DOWN_PEER_LEAVING;
738 break;
739 case PEER_LOST_CONTACT_EVT:
740 state = SELF_LEAVING_PEER_DOWN;
741 break;
742 case NODE_FAILOVER_END_EVT:
743 state = SELF_UP_PEER_UP;
744 break;
745 case NODE_FAILOVER_BEGIN_EVT:
746 case SELF_ESTABL_CONTACT_EVT:
747 case PEER_ESTABL_CONTACT_EVT:
748 break;
749 case NODE_SYNCH_BEGIN_EVT:
750 case NODE_SYNCH_END_EVT:
751 default:
752 goto illegal_evt;
754 break;
755 case NODE_SYNCHING:
756 switch (evt) {
757 case SELF_LOST_CONTACT_EVT:
758 state = SELF_DOWN_PEER_LEAVING;
759 break;
760 case PEER_LOST_CONTACT_EVT:
761 state = SELF_LEAVING_PEER_DOWN;
762 break;
763 case NODE_SYNCH_END_EVT:
764 state = SELF_UP_PEER_UP;
765 break;
766 case NODE_FAILOVER_BEGIN_EVT:
767 state = NODE_FAILINGOVER;
768 break;
769 case NODE_SYNCH_BEGIN_EVT:
770 case SELF_ESTABL_CONTACT_EVT:
771 case PEER_ESTABL_CONTACT_EVT:
772 break;
773 case NODE_FAILOVER_END_EVT:
774 default:
775 goto illegal_evt;
777 break;
778 default:
779 pr_err("Unknown node fsm state %x\n", state);
780 break;
782 n->state = state;
783 return;
785 illegal_evt:
786 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
789 bool tipc_node_filter_pkt(struct tipc_node *n, struct tipc_msg *hdr)
791 int state = n->state;
793 if (likely(state == SELF_UP_PEER_UP))
794 return true;
796 if (state == SELF_LEAVING_PEER_DOWN)
797 return false;
799 if (state == SELF_DOWN_PEER_LEAVING) {
800 if (msg_peer_node_is_up(hdr))
801 return false;
804 return true;
807 static void node_established_contact(struct tipc_node *n_ptr)
809 tipc_node_fsm_evt(n_ptr, SELF_ESTABL_CONTACT_EVT);
810 n_ptr->action_flags |= TIPC_NOTIFY_NODE_UP;
811 n_ptr->bclink.oos_state = 0;
812 n_ptr->bclink.acked = tipc_bclink_get_last_sent(n_ptr->net);
813 tipc_bclink_add_node(n_ptr->net, n_ptr->addr);
816 static void node_lost_contact(struct tipc_node *n_ptr,
817 struct sk_buff_head *inputq)
819 char addr_string[16];
820 struct tipc_sock_conn *conn, *safe;
821 struct tipc_link *l;
822 struct list_head *conns = &n_ptr->conn_sks;
823 struct sk_buff *skb;
824 struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
825 uint i;
827 pr_debug("Lost contact with %s\n",
828 tipc_addr_string_fill(addr_string, n_ptr->addr));
830 /* Flush broadcast link info associated with lost node */
831 if (n_ptr->bclink.recv_permitted) {
832 __skb_queue_purge(&n_ptr->bclink.deferdq);
834 if (n_ptr->bclink.reasm_buf) {
835 kfree_skb(n_ptr->bclink.reasm_buf);
836 n_ptr->bclink.reasm_buf = NULL;
839 tipc_bclink_remove_node(n_ptr->net, n_ptr->addr);
840 tipc_bclink_acknowledge(n_ptr, INVALID_LINK_SEQ);
842 n_ptr->bclink.recv_permitted = false;
845 /* Abort any ongoing link failover */
846 for (i = 0; i < MAX_BEARERS; i++) {
847 l = n_ptr->links[i].link;
848 if (l)
849 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
852 /* Prevent re-contact with node until cleanup is done */
853 tipc_node_fsm_evt(n_ptr, SELF_LOST_CONTACT_EVT);
855 /* Notify publications from this node */
856 n_ptr->action_flags |= TIPC_NOTIFY_NODE_DOWN;
858 /* Notify sockets connected to node */
859 list_for_each_entry_safe(conn, safe, conns, list) {
860 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
861 SHORT_H_SIZE, 0, tn->own_addr,
862 conn->peer_node, conn->port,
863 conn->peer_port, TIPC_ERR_NO_NODE);
864 if (likely(skb))
865 skb_queue_tail(inputq, skb);
866 list_del(&conn->list);
867 kfree(conn);
872 * tipc_node_get_linkname - get the name of a link
874 * @bearer_id: id of the bearer
875 * @node: peer node address
876 * @linkname: link name output buffer
878 * Returns 0 on success
880 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
881 char *linkname, size_t len)
883 struct tipc_link *link;
884 int err = -EINVAL;
885 struct tipc_node *node = tipc_node_find(net, addr);
887 if (!node)
888 return err;
890 if (bearer_id >= MAX_BEARERS)
891 goto exit;
893 tipc_node_lock(node);
894 link = node->links[bearer_id].link;
895 if (link) {
896 strncpy(linkname, link->name, len);
897 err = 0;
899 exit:
900 tipc_node_unlock(node);
901 tipc_node_put(node);
902 return err;
905 void tipc_node_unlock(struct tipc_node *node)
907 struct net *net = node->net;
908 u32 addr = 0;
909 u32 flags = node->action_flags;
910 u32 link_id = 0;
911 struct list_head *publ_list;
913 if (likely(!flags)) {
914 spin_unlock_bh(&node->lock);
915 return;
918 addr = node->addr;
919 link_id = node->link_id;
920 publ_list = &node->publ_list;
922 node->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
923 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP |
924 TIPC_WAKEUP_BCAST_USERS | TIPC_BCAST_MSG_EVT |
925 TIPC_BCAST_RESET);
927 spin_unlock_bh(&node->lock);
929 if (flags & TIPC_NOTIFY_NODE_DOWN)
930 tipc_publ_notify(net, publ_list, addr);
932 if (flags & TIPC_WAKEUP_BCAST_USERS)
933 tipc_bclink_wakeup_users(net);
935 if (flags & TIPC_NOTIFY_NODE_UP)
936 tipc_named_node_up(net, addr);
938 if (flags & TIPC_NOTIFY_LINK_UP)
939 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
940 TIPC_NODE_SCOPE, link_id, addr);
942 if (flags & TIPC_NOTIFY_LINK_DOWN)
943 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
944 link_id, addr);
946 if (flags & TIPC_BCAST_MSG_EVT)
947 tipc_bclink_input(net);
949 if (flags & TIPC_BCAST_RESET)
950 tipc_node_reset_links(node);
953 /* Caller should hold node lock for the passed node */
954 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
956 void *hdr;
957 struct nlattr *attrs;
959 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
960 NLM_F_MULTI, TIPC_NL_NODE_GET);
961 if (!hdr)
962 return -EMSGSIZE;
964 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
965 if (!attrs)
966 goto msg_full;
968 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
969 goto attr_msg_full;
970 if (tipc_node_is_up(node))
971 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
972 goto attr_msg_full;
974 nla_nest_end(msg->skb, attrs);
975 genlmsg_end(msg->skb, hdr);
977 return 0;
979 attr_msg_full:
980 nla_nest_cancel(msg->skb, attrs);
981 msg_full:
982 genlmsg_cancel(msg->skb, hdr);
984 return -EMSGSIZE;
987 static struct tipc_link *tipc_node_select_link(struct tipc_node *n, int sel,
988 int *bearer_id,
989 struct tipc_media_addr **maddr)
991 int id = n->active_links[sel & 1];
993 if (unlikely(id < 0))
994 return NULL;
996 *bearer_id = id;
997 *maddr = &n->links[id].maddr;
998 return n->links[id].link;
1002 * tipc_node_xmit() is the general link level function for message sending
1003 * @net: the applicable net namespace
1004 * @list: chain of buffers containing message
1005 * @dnode: address of destination node
1006 * @selector: a number used for deterministic link selection
1007 * Consumes the buffer chain, except when returning -ELINKCONG
1008 * Returns 0 if success, otherwise errno: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE
1010 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1011 u32 dnode, int selector)
1013 struct tipc_link *l = NULL;
1014 struct tipc_node *n;
1015 struct sk_buff_head xmitq;
1016 struct tipc_media_addr *maddr;
1017 int bearer_id;
1018 int rc = -EHOSTUNREACH;
1020 __skb_queue_head_init(&xmitq);
1021 n = tipc_node_find(net, dnode);
1022 if (likely(n)) {
1023 tipc_node_lock(n);
1024 l = tipc_node_select_link(n, selector, &bearer_id, &maddr);
1025 if (likely(l))
1026 rc = tipc_link_xmit(l, list, &xmitq);
1027 tipc_node_unlock(n);
1028 if (unlikely(rc == -ENOBUFS))
1029 tipc_node_link_down(n, bearer_id, false);
1030 tipc_node_put(n);
1032 if (likely(!rc)) {
1033 tipc_bearer_xmit(net, bearer_id, &xmitq, maddr);
1034 return 0;
1036 if (likely(in_own_node(net, dnode))) {
1037 tipc_sk_rcv(net, list);
1038 return 0;
1040 return rc;
1043 /* tipc_node_xmit_skb(): send single buffer to destination
1044 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1045 * messages, which will not be rejected
1046 * The only exception is datagram messages rerouted after secondary
1047 * lookup, which are rare and safe to dispose of anyway.
1048 * TODO: Return real return value, and let callers use
1049 * tipc_wait_for_sendpkt() where applicable
1051 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1052 u32 selector)
1054 struct sk_buff_head head;
1055 int rc;
1057 skb_queue_head_init(&head);
1058 __skb_queue_tail(&head, skb);
1059 rc = tipc_node_xmit(net, &head, dnode, selector);
1060 if (rc == -ELINKCONG)
1061 kfree_skb(skb);
1062 return 0;
1066 * tipc_node_check_state - check and if necessary update node state
1067 * @skb: TIPC packet
1068 * @bearer_id: identity of bearer delivering the packet
1069 * Returns true if state is ok, otherwise consumes buffer and returns false
1071 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1072 int bearer_id, struct sk_buff_head *xmitq)
1074 struct tipc_msg *hdr = buf_msg(skb);
1075 int usr = msg_user(hdr);
1076 int mtyp = msg_type(hdr);
1077 u16 oseqno = msg_seqno(hdr);
1078 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1079 u16 exp_pkts = msg_msgcnt(hdr);
1080 u16 rcv_nxt, syncpt, dlv_nxt;
1081 int state = n->state;
1082 struct tipc_link *l, *tnl, *pl = NULL;
1083 struct tipc_media_addr *maddr;
1084 int i, pb_id;
1086 l = n->links[bearer_id].link;
1087 if (!l)
1088 return false;
1089 rcv_nxt = l->rcv_nxt;
1092 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1093 return true;
1095 /* Find parallel link, if any */
1096 for (i = 0; i < MAX_BEARERS; i++) {
1097 if ((i != bearer_id) && n->links[i].link) {
1098 pl = n->links[i].link;
1099 break;
1103 /* Update node accesibility if applicable */
1104 if (state == SELF_UP_PEER_COMING) {
1105 if (!tipc_link_is_up(l))
1106 return true;
1107 if (!msg_peer_link_is_up(hdr))
1108 return true;
1109 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1112 if (state == SELF_DOWN_PEER_LEAVING) {
1113 if (msg_peer_node_is_up(hdr))
1114 return false;
1115 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1118 /* Ignore duplicate packets */
1119 if (less(oseqno, rcv_nxt))
1120 return true;
1122 /* Initiate or update failover mode if applicable */
1123 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1124 syncpt = oseqno + exp_pkts - 1;
1125 if (pl && tipc_link_is_up(pl)) {
1126 pb_id = pl->bearer_id;
1127 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1128 tipc_skb_queue_splice_tail_init(pl->inputq, l->inputq);
1130 /* If pkts arrive out of order, use lowest calculated syncpt */
1131 if (less(syncpt, n->sync_point))
1132 n->sync_point = syncpt;
1135 /* Open parallel link when tunnel link reaches synch point */
1136 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1137 if (!more(rcv_nxt, n->sync_point))
1138 return true;
1139 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1140 if (pl)
1141 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1142 return true;
1145 /* No synching needed if only one link */
1146 if (!pl || !tipc_link_is_up(pl))
1147 return true;
1149 /* Initiate or update synch mode if applicable */
1150 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG)) {
1151 syncpt = iseqno + exp_pkts - 1;
1152 if (!tipc_link_is_up(l)) {
1153 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1154 __tipc_node_link_up(n, bearer_id, xmitq);
1156 if (n->state == SELF_UP_PEER_UP) {
1157 n->sync_point = syncpt;
1158 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1159 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1161 if (less(syncpt, n->sync_point))
1162 n->sync_point = syncpt;
1165 /* Open tunnel link when parallel link reaches synch point */
1166 if ((n->state == NODE_SYNCHING) && tipc_link_is_synching(l)) {
1167 if (tipc_link_is_synching(l)) {
1168 tnl = l;
1169 } else {
1170 tnl = pl;
1171 pl = l;
1173 dlv_nxt = pl->rcv_nxt - mod(skb_queue_len(pl->inputq));
1174 if (more(dlv_nxt, n->sync_point)) {
1175 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1176 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1177 return true;
1179 if (l == pl)
1180 return true;
1181 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1182 return true;
1183 if (usr == LINK_PROTOCOL)
1184 return true;
1185 return false;
1187 return true;
1191 * tipc_rcv - process TIPC packets/messages arriving from off-node
1192 * @net: the applicable net namespace
1193 * @skb: TIPC packet
1194 * @bearer: pointer to bearer message arrived on
1196 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1197 * structure (i.e. cannot be NULL), but bearer can be inactive.
1199 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1201 struct sk_buff_head xmitq;
1202 struct tipc_node *n;
1203 struct tipc_msg *hdr = buf_msg(skb);
1204 int usr = msg_user(hdr);
1205 int bearer_id = b->identity;
1206 struct tipc_link_entry *le;
1207 int rc = 0;
1209 __skb_queue_head_init(&xmitq);
1211 /* Ensure message is well-formed */
1212 if (unlikely(!tipc_msg_validate(skb)))
1213 goto discard;
1215 /* Handle arrival of a non-unicast link packet */
1216 if (unlikely(msg_non_seq(hdr))) {
1217 if (usr == LINK_CONFIG)
1218 tipc_disc_rcv(net, skb, b);
1219 else
1220 tipc_bclink_rcv(net, skb);
1221 return;
1224 /* Locate neighboring node that sent packet */
1225 n = tipc_node_find(net, msg_prevnode(hdr));
1226 if (unlikely(!n))
1227 goto discard;
1228 le = &n->links[bearer_id];
1230 tipc_node_lock(n);
1232 /* Is reception permitted at the moment ? */
1233 if (!tipc_node_filter_pkt(n, hdr))
1234 goto unlock;
1236 if (unlikely(msg_user(hdr) == LINK_PROTOCOL))
1237 tipc_bclink_sync_state(n, hdr);
1239 /* Release acked broadcast packets */
1240 if (unlikely(n->bclink.acked != msg_bcast_ack(hdr)))
1241 tipc_bclink_acknowledge(n, msg_bcast_ack(hdr));
1243 /* Check and if necessary update node state */
1244 if (likely(tipc_node_check_state(n, skb, bearer_id, &xmitq))) {
1245 rc = tipc_link_rcv(le->link, skb, &xmitq);
1246 skb = NULL;
1248 unlock:
1249 tipc_node_unlock(n);
1251 if (unlikely(rc & TIPC_LINK_UP_EVT))
1252 tipc_node_link_up(n, bearer_id, &xmitq);
1254 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1255 tipc_node_link_down(n, bearer_id, false);
1257 if (unlikely(!skb_queue_empty(&n->bclink.namedq)))
1258 tipc_named_rcv(net, &n->bclink.namedq);
1260 if (!skb_queue_empty(&le->inputq))
1261 tipc_sk_rcv(net, &le->inputq);
1263 if (!skb_queue_empty(&xmitq))
1264 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1266 tipc_node_put(n);
1267 discard:
1268 kfree_skb(skb);
1271 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1273 int err;
1274 struct net *net = sock_net(skb->sk);
1275 struct tipc_net *tn = net_generic(net, tipc_net_id);
1276 int done = cb->args[0];
1277 int last_addr = cb->args[1];
1278 struct tipc_node *node;
1279 struct tipc_nl_msg msg;
1281 if (done)
1282 return 0;
1284 msg.skb = skb;
1285 msg.portid = NETLINK_CB(cb->skb).portid;
1286 msg.seq = cb->nlh->nlmsg_seq;
1288 rcu_read_lock();
1289 if (last_addr) {
1290 node = tipc_node_find(net, last_addr);
1291 if (!node) {
1292 rcu_read_unlock();
1293 /* We never set seq or call nl_dump_check_consistent()
1294 * this means that setting prev_seq here will cause the
1295 * consistence check to fail in the netlink callback
1296 * handler. Resulting in the NLMSG_DONE message having
1297 * the NLM_F_DUMP_INTR flag set if the node state
1298 * changed while we released the lock.
1300 cb->prev_seq = 1;
1301 return -EPIPE;
1303 tipc_node_put(node);
1306 list_for_each_entry_rcu(node, &tn->node_list, list) {
1307 if (last_addr) {
1308 if (node->addr == last_addr)
1309 last_addr = 0;
1310 else
1311 continue;
1314 tipc_node_lock(node);
1315 err = __tipc_nl_add_node(&msg, node);
1316 if (err) {
1317 last_addr = node->addr;
1318 tipc_node_unlock(node);
1319 goto out;
1322 tipc_node_unlock(node);
1324 done = 1;
1325 out:
1326 cb->args[0] = done;
1327 cb->args[1] = last_addr;
1328 rcu_read_unlock();
1330 return skb->len;