?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
tipc: fix crash during node removal
[linux-2.6/btrfs-unstable.git] / net / tipc / node.c
blobcdb79503d890ed5eef2752d029a38be55fb4ac71
1 /*
2 * net/tipc/node.c: TIPC node management routines
3 *
4 * Copyright (c) 2000-2006, 2012-2015, Ericsson AB
5 * Copyright (c) 2005-2006, 2010-2014, Wind River Systems
6 * All rights reserved.
7 *
8 * Redistribution and use in source and binary forms, with or without
9 * modification, are permitted provided that the following conditions are met:
10 *
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.
19 *
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.
23 *
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.
35 */
36
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"
44
45 #define INVALID_NODE_SIG 0x10000
46
47 /* Flags used to take different actions according to flag type
48 * TIPC_NOTIFY_NODE_DOWN: notify node is down
49 * TIPC_NOTIFY_NODE_UP: notify node is up
50 * TIPC_DISTRIBUTE_NAME: publish or withdraw link state name type
51 */
52 enum {
53 TIPC_NOTIFY_NODE_DOWN = (1 << 3),
54 TIPC_NOTIFY_NODE_UP = (1 << 4),
55 TIPC_NOTIFY_LINK_UP = (1 << 6),
56 TIPC_NOTIFY_LINK_DOWN = (1 << 7)
57 };
58
59 struct tipc_link_entry {
60 struct tipc_link *link;
61 spinlock_t lock; /* per link */
62 u32 mtu;
63 struct sk_buff_head inputq;
64 struct tipc_media_addr maddr;
65 };
66
67 struct tipc_bclink_entry {
68 struct tipc_link *link;
69 struct sk_buff_head inputq1;
70 struct sk_buff_head arrvq;
71 struct sk_buff_head inputq2;
72 struct sk_buff_head namedq;
73 };
74
75 /**
76 * struct tipc_node - TIPC node structure
77 * @addr: network address of node
78 * @ref: reference counter to node object
79 * @lock: rwlock governing access to structure
80 * @net: the applicable net namespace
81 * @hash: links to adjacent nodes in unsorted hash chain
82 * @inputq: pointer to input queue containing messages for msg event
83 * @namedq: pointer to name table input queue with name table messages
84 * @active_links: bearer ids of active links, used as index into links[] array
85 * @links: array containing references to all links to node
86 * @action_flags: bit mask of different types of node actions
87 * @state: connectivity state vs peer node
88 * @sync_point: sequence number where synch/failover is finished
89 * @list: links to adjacent nodes in sorted list of cluster's nodes
90 * @working_links: number of working links to node (both active and standby)
91 * @link_cnt: number of links to node
92 * @capabilities: bitmap, indicating peer node's functional capabilities
93 * @signature: node instance identifier
94 * @link_id: local and remote bearer ids of changing link, if any
95 * @publ_list: list of publications
96 * @rcu: rcu struct for tipc_node
97 */
98 struct tipc_node {
99 u32 addr;
100 struct kref kref;
101 rwlock_t lock;
102 struct net *net;
103 struct hlist_node hash;
104 int active_links[2];
105 struct tipc_link_entry links[MAX_BEARERS];
106 struct tipc_bclink_entry bc_entry;
107 int action_flags;
108 struct list_head list;
109 int state;
110 u16 sync_point;
111 int link_cnt;
112 u16 working_links;
113 u16 capabilities;
114 u32 signature;
115 u32 link_id;
116 struct list_head publ_list;
117 struct list_head conn_sks;
118 unsigned long keepalive_intv;
119 struct timer_list timer;
120 struct rcu_head rcu;
121 };
122
123 /* Node FSM states and events:
124 */
125 enum {
126 SELF_DOWN_PEER_DOWN = 0xdd,
127 SELF_UP_PEER_UP = 0xaa,
128 SELF_DOWN_PEER_LEAVING = 0xd1,
129 SELF_UP_PEER_COMING = 0xac,
130 SELF_COMING_PEER_UP = 0xca,
131 SELF_LEAVING_PEER_DOWN = 0x1d,
132 NODE_FAILINGOVER = 0xf0,
133 NODE_SYNCHING = 0xcc
134 };
135
136 enum {
137 SELF_ESTABL_CONTACT_EVT = 0xece,
138 SELF_LOST_CONTACT_EVT = 0x1ce,
139 PEER_ESTABL_CONTACT_EVT = 0x9ece,
140 PEER_LOST_CONTACT_EVT = 0x91ce,
141 NODE_FAILOVER_BEGIN_EVT = 0xfbe,
142 NODE_FAILOVER_END_EVT = 0xfee,
143 NODE_SYNCH_BEGIN_EVT = 0xcbe,
144 NODE_SYNCH_END_EVT = 0xcee
145 };
146
147 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
148 struct sk_buff_head *xmitq,
149 struct tipc_media_addr **maddr);
150 static void tipc_node_link_down(struct tipc_node *n, int bearer_id,
151 bool delete);
152 static void node_lost_contact(struct tipc_node *n, struct sk_buff_head *inputq);
153 static void tipc_node_delete(struct tipc_node *node);
154 static void tipc_node_timeout(unsigned long data);
155 static void tipc_node_fsm_evt(struct tipc_node *n, int evt);
156 static struct tipc_node *tipc_node_find(struct net *net, u32 addr);
157 static void tipc_node_put(struct tipc_node *node);
158 static bool tipc_node_is_up(struct tipc_node *n);
159
160 struct tipc_sock_conn {
161 u32 port;
162 u32 peer_port;
163 u32 peer_node;
164 struct list_head list;
165 };
166
167 static const struct nla_policy tipc_nl_link_policy[TIPC_NLA_LINK_MAX + 1] = {
168 [TIPC_NLA_LINK_UNSPEC] = { .type = NLA_UNSPEC },
169 [TIPC_NLA_LINK_NAME] = {
170 .type = NLA_STRING,
171 .len = TIPC_MAX_LINK_NAME
172 },
173 [TIPC_NLA_LINK_MTU] = { .type = NLA_U32 },
174 [TIPC_NLA_LINK_BROADCAST] = { .type = NLA_FLAG },
175 [TIPC_NLA_LINK_UP] = { .type = NLA_FLAG },
176 [TIPC_NLA_LINK_ACTIVE] = { .type = NLA_FLAG },
177 [TIPC_NLA_LINK_PROP] = { .type = NLA_NESTED },
178 [TIPC_NLA_LINK_STATS] = { .type = NLA_NESTED },
179 [TIPC_NLA_LINK_RX] = { .type = NLA_U32 },
180 [TIPC_NLA_LINK_TX] = { .type = NLA_U32 }
181 };
182
183 static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
184 [TIPC_NLA_NODE_UNSPEC] = { .type = NLA_UNSPEC },
185 [TIPC_NLA_NODE_ADDR] = { .type = NLA_U32 },
186 [TIPC_NLA_NODE_UP] = { .type = NLA_FLAG }
187 };
188
189 static struct tipc_link *node_active_link(struct tipc_node *n, int sel)
190 {
191 int bearer_id = n->active_links[sel & 1];
192
193 if (unlikely(bearer_id == INVALID_BEARER_ID))
194 return NULL;
195
196 return n->links[bearer_id].link;
197 }
198
199 int tipc_node_get_mtu(struct net *net, u32 addr, u32 sel)
200 {
201 struct tipc_node *n;
202 int bearer_id;
203 unsigned int mtu = MAX_MSG_SIZE;
204
205 n = tipc_node_find(net, addr);
206 if (unlikely(!n))
207 return mtu;
208
209 bearer_id = n->active_links[sel & 1];
210 if (likely(bearer_id != INVALID_BEARER_ID))
211 mtu = n->links[bearer_id].mtu;
212 tipc_node_put(n);
213 return mtu;
214 }
215 /*
216 * A trivial power-of-two bitmask technique is used for speed, since this
217 * operation is done for every incoming TIPC packet. The number of hash table
218 * entries has been chosen so that no hash chain exceeds 8 nodes and will
219 * usually be much smaller (typically only a single node).
220 */
221 static unsigned int tipc_hashfn(u32 addr)
222 {
223 return addr & (NODE_HTABLE_SIZE - 1);
224 }
225
226 static void tipc_node_kref_release(struct kref *kref)
227 {
228 struct tipc_node *n = container_of(kref, struct tipc_node, kref);
229
230 kfree(n->bc_entry.link);
231 kfree_rcu(n, rcu);
232 }
233
234 static void tipc_node_put(struct tipc_node *node)
235 {
236 kref_put(&node->kref, tipc_node_kref_release);
237 }
238
239 static void tipc_node_get(struct tipc_node *node)
240 {
241 kref_get(&node->kref);
242 }
243
244 /*
245 * tipc_node_find - locate specified node object, if it exists
246 */
247 static struct tipc_node *tipc_node_find(struct net *net, u32 addr)
248 {
249 struct tipc_net *tn = tipc_net(net);
250 struct tipc_node *node;
251 unsigned int thash = tipc_hashfn(addr);
252
253 if (unlikely(!in_own_cluster_exact(net, addr)))
254 return NULL;
255
256 rcu_read_lock();
257 hlist_for_each_entry_rcu(node, &tn->node_htable[thash], hash) {
258 if (node->addr != addr)
259 continue;
260 if (!kref_get_unless_zero(&node->kref))
261 node = NULL;
262 break;
263 }
264 rcu_read_unlock();
265 return node;
266 }
267
268 static void tipc_node_read_lock(struct tipc_node *n)
269 {
270 read_lock_bh(&n->lock);
271 }
272
273 static void tipc_node_read_unlock(struct tipc_node *n)
274 {
275 read_unlock_bh(&n->lock);
276 }
277
278 static void tipc_node_write_lock(struct tipc_node *n)
279 {
280 write_lock_bh(&n->lock);
281 }
282
283 static void tipc_node_write_unlock(struct tipc_node *n)
284 {
285 struct net *net = n->net;
286 u32 addr = 0;
287 u32 flags = n->action_flags;
288 u32 link_id = 0;
289 struct list_head *publ_list;
290
291 if (likely(!flags)) {
292 write_unlock_bh(&n->lock);
293 return;
294 }
295
296 addr = n->addr;
297 link_id = n->link_id;
298 publ_list = &n->publ_list;
299
300 n->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
301 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP);
302
303 write_unlock_bh(&n->lock);
304
305 if (flags & TIPC_NOTIFY_NODE_DOWN)
306 tipc_publ_notify(net, publ_list, addr);
307
308 if (flags & TIPC_NOTIFY_NODE_UP)
309 tipc_named_node_up(net, addr);
310
311 if (flags & TIPC_NOTIFY_LINK_UP)
312 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
313 TIPC_NODE_SCOPE, link_id, addr);
314
315 if (flags & TIPC_NOTIFY_LINK_DOWN)
316 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
317 link_id, addr);
318 }
319
320 struct tipc_node *tipc_node_create(struct net *net, u32 addr, u16 capabilities)
321 {
322 struct tipc_net *tn = net_generic(net, tipc_net_id);
323 struct tipc_node *n, *temp_node;
324 int i;
325
326 spin_lock_bh(&tn->node_list_lock);
327 n = tipc_node_find(net, addr);
328 if (n)
329 goto exit;
330 n = kzalloc(sizeof(*n), GFP_ATOMIC);
331 if (!n) {
332 pr_warn("Node creation failed, no memory\n");
333 goto exit;
334 }
335 n->addr = addr;
336 n->net = net;
337 n->capabilities = capabilities;
338 kref_init(&n->kref);
339 rwlock_init(&n->lock);
340 INIT_HLIST_NODE(&n->hash);
341 INIT_LIST_HEAD(&n->list);
342 INIT_LIST_HEAD(&n->publ_list);
343 INIT_LIST_HEAD(&n->conn_sks);
344 skb_queue_head_init(&n->bc_entry.namedq);
345 skb_queue_head_init(&n->bc_entry.inputq1);
346 __skb_queue_head_init(&n->bc_entry.arrvq);
347 skb_queue_head_init(&n->bc_entry.inputq2);
348 for (i = 0; i < MAX_BEARERS; i++)
349 spin_lock_init(&n->links[i].lock);
350 n->state = SELF_DOWN_PEER_LEAVING;
351 n->signature = INVALID_NODE_SIG;
352 n->active_links[0] = INVALID_BEARER_ID;
353 n->active_links[1] = INVALID_BEARER_ID;
354 if (!tipc_link_bc_create(net, tipc_own_addr(net), n->addr,
355 U16_MAX,
356 tipc_link_window(tipc_bc_sndlink(net)),
357 n->capabilities,
358 &n->bc_entry.inputq1,
359 &n->bc_entry.namedq,
360 tipc_bc_sndlink(net),
361 &n->bc_entry.link)) {
362 pr_warn("Broadcast rcv link creation failed, no memory\n");
363 kfree(n);
364 n = NULL;
365 goto exit;
366 }
367 tipc_node_get(n);
368 setup_timer(&n->timer, tipc_node_timeout, (unsigned long)n);
369 n->keepalive_intv = U32_MAX;
370 hlist_add_head_rcu(&n->hash, &tn->node_htable[tipc_hashfn(addr)]);
371 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
372 if (n->addr < temp_node->addr)
373 break;
374 }
375 list_add_tail_rcu(&n->list, &temp_node->list);
376 exit:
377 spin_unlock_bh(&tn->node_list_lock);
378 return n;
379 }
380
381 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
382 {
383 unsigned long tol = tipc_link_tolerance(l);
384 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
385 unsigned long keepalive_intv = msecs_to_jiffies(intv);
386
387 /* Link with lowest tolerance determines timer interval */
388 if (keepalive_intv < n->keepalive_intv)
389 n->keepalive_intv = keepalive_intv;
390
391 /* Ensure link's abort limit corresponds to current interval */
392 tipc_link_set_abort_limit(l, tol / jiffies_to_msecs(n->keepalive_intv));
393 }
394
395 static void tipc_node_delete(struct tipc_node *node)
396 {
397 list_del_rcu(&node->list);
398 hlist_del_rcu(&node->hash);
399 tipc_node_put(node);
400
401 del_timer_sync(&node->timer);
402 tipc_node_put(node);
403 }
404
405 void tipc_node_stop(struct net *net)
406 {
407 struct tipc_net *tn = tipc_net(net);
408 struct tipc_node *node, *t_node;
409
410 spin_lock_bh(&tn->node_list_lock);
411 list_for_each_entry_safe(node, t_node, &tn->node_list, list)
412 tipc_node_delete(node);
413 spin_unlock_bh(&tn->node_list_lock);
414 }
415
416 void tipc_node_subscribe(struct net *net, struct list_head *subscr, u32 addr)
417 {
418 struct tipc_node *n;
419
420 if (in_own_node(net, addr))
421 return;
422
423 n = tipc_node_find(net, addr);
424 if (!n) {
425 pr_warn("Node subscribe rejected, unknown node 0x%x\n", addr);
426 return;
427 }
428 tipc_node_write_lock(n);
429 list_add_tail(subscr, &n->publ_list);
430 tipc_node_write_unlock(n);
431 tipc_node_put(n);
432 }
433
434 void tipc_node_unsubscribe(struct net *net, struct list_head *subscr, u32 addr)
435 {
436 struct tipc_node *n;
437
438 if (in_own_node(net, addr))
439 return;
440
441 n = tipc_node_find(net, addr);
442 if (!n) {
443 pr_warn("Node unsubscribe rejected, unknown node 0x%x\n", addr);
444 return;
445 }
446 tipc_node_write_lock(n);
447 list_del_init(subscr);
448 tipc_node_write_unlock(n);
449 tipc_node_put(n);
450 }
451
452 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
453 {
454 struct tipc_node *node;
455 struct tipc_sock_conn *conn;
456 int err = 0;
457
458 if (in_own_node(net, dnode))
459 return 0;
460
461 node = tipc_node_find(net, dnode);
462 if (!node) {
463 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
464 return -EHOSTUNREACH;
465 }
466 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
467 if (!conn) {
468 err = -EHOSTUNREACH;
469 goto exit;
470 }
471 conn->peer_node = dnode;
472 conn->port = port;
473 conn->peer_port = peer_port;
474
475 tipc_node_write_lock(node);
476 list_add_tail(&conn->list, &node->conn_sks);
477 tipc_node_write_unlock(node);
478 exit:
479 tipc_node_put(node);
480 return err;
481 }
482
483 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
484 {
485 struct tipc_node *node;
486 struct tipc_sock_conn *conn, *safe;
487
488 if (in_own_node(net, dnode))
489 return;
490
491 node = tipc_node_find(net, dnode);
492 if (!node)
493 return;
494
495 tipc_node_write_lock(node);
496 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
497 if (port != conn->port)
498 continue;
499 list_del(&conn->list);
500 kfree(conn);
501 }
502 tipc_node_write_unlock(node);
503 tipc_node_put(node);
504 }
505
506 /* tipc_node_timeout - handle expiration of node timer
507 */
508 static void tipc_node_timeout(unsigned long data)
509 {
510 struct tipc_node *n = (struct tipc_node *)data;
511 struct tipc_link_entry *le;
512 struct sk_buff_head xmitq;
513 int bearer_id;
514 int rc = 0;
515
516 __skb_queue_head_init(&xmitq);
517
518 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
519 tipc_node_read_lock(n);
520 le = &n->links[bearer_id];
521 spin_lock_bh(&le->lock);
522 if (le->link) {
523 /* Link tolerance may change asynchronously: */
524 tipc_node_calculate_timer(n, le->link);
525 rc = tipc_link_timeout(le->link, &xmitq);
526 }
527 spin_unlock_bh(&le->lock);
528 tipc_node_read_unlock(n);
529 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
530 if (rc & TIPC_LINK_DOWN_EVT)
531 tipc_node_link_down(n, bearer_id, false);
532 }
533 mod_timer(&n->timer, jiffies + n->keepalive_intv);
534 }
535
536 /**
537 * __tipc_node_link_up - handle addition of link
538 * Node lock must be held by caller
539 * Link becomes active (alone or shared) or standby, depending on its priority.
540 */
541 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
542 struct sk_buff_head *xmitq)
543 {
544 int *slot0 = &n->active_links[0];
545 int *slot1 = &n->active_links[1];
546 struct tipc_link *ol = node_active_link(n, 0);
547 struct tipc_link *nl = n->links[bearer_id].link;
548
549 if (!nl)
550 return;
551
552 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
553 if (!tipc_link_is_up(nl))
554 return;
555
556 n->working_links++;
557 n->action_flags |= TIPC_NOTIFY_LINK_UP;
558 n->link_id = tipc_link_id(nl);
559
560 /* Leave room for tunnel header when returning 'mtu' to users: */
561 n->links[bearer_id].mtu = tipc_link_mtu(nl) - INT_H_SIZE;
562
563 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
564 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
565
566 pr_debug("Established link <%s> on network plane %c\n",
567 tipc_link_name(nl), tipc_link_plane(nl));
568
569 /* First link? => give it both slots */
570 if (!ol) {
571 *slot0 = bearer_id;
572 *slot1 = bearer_id;
573 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
574 n->action_flags |= TIPC_NOTIFY_NODE_UP;
575 tipc_bcast_add_peer(n->net, nl, xmitq);
576 return;
577 }
578
579 /* Second link => redistribute slots */
580 if (tipc_link_prio(nl) > tipc_link_prio(ol)) {
581 pr_debug("Old link <%s> becomes standby\n", tipc_link_name(ol));
582 *slot0 = bearer_id;
583 *slot1 = bearer_id;
584 tipc_link_set_active(nl, true);
585 tipc_link_set_active(ol, false);
586 } else if (tipc_link_prio(nl) == tipc_link_prio(ol)) {
587 tipc_link_set_active(nl, true);
588 *slot1 = bearer_id;
589 } else {
590 pr_debug("New link <%s> is standby\n", tipc_link_name(nl));
591 }
592
593 /* Prepare synchronization with first link */
594 tipc_link_tnl_prepare(ol, nl, SYNCH_MSG, xmitq);
595 }
596
597 /**
598 * tipc_node_link_up - handle addition of link
599 *
600 * Link becomes active (alone or shared) or standby, depending on its priority.
601 */
602 static void tipc_node_link_up(struct tipc_node *n, int bearer_id,
603 struct sk_buff_head *xmitq)
604 {
605 tipc_node_write_lock(n);
606 __tipc_node_link_up(n, bearer_id, xmitq);
607 tipc_node_write_unlock(n);
608 }
609
610 /**
611 * __tipc_node_link_down - handle loss of link
612 */
613 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
614 struct sk_buff_head *xmitq,
615 struct tipc_media_addr **maddr)
616 {
617 struct tipc_link_entry *le = &n->links[*bearer_id];
618 int *slot0 = &n->active_links[0];
619 int *slot1 = &n->active_links[1];
620 int i, highest = 0, prio;
621 struct tipc_link *l, *_l, *tnl;
622
623 l = n->links[*bearer_id].link;
624 if (!l || tipc_link_is_reset(l))
625 return;
626
627 n->working_links--;
628 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
629 n->link_id = tipc_link_id(l);
630
631 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
632
633 pr_debug("Lost link <%s> on network plane %c\n",
634 tipc_link_name(l), tipc_link_plane(l));
635
636 /* Select new active link if any available */
637 *slot0 = INVALID_BEARER_ID;
638 *slot1 = INVALID_BEARER_ID;
639 for (i = 0; i < MAX_BEARERS; i++) {
640 _l = n->links[i].link;
641 if (!_l || !tipc_link_is_up(_l))
642 continue;
643 if (_l == l)
644 continue;
645 prio = tipc_link_prio(_l);
646 if (prio < highest)
647 continue;
648 if (prio > highest) {
649 highest = prio;
650 *slot0 = i;
651 *slot1 = i;
652 continue;
653 }
654 *slot1 = i;
655 }
656
657 if (!tipc_node_is_up(n)) {
658 if (tipc_link_peer_is_down(l))
659 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
660 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
661 tipc_link_fsm_evt(l, LINK_RESET_EVT);
662 tipc_link_reset(l);
663 tipc_link_build_reset_msg(l, xmitq);
664 *maddr = &n->links[*bearer_id].maddr;
665 node_lost_contact(n, &le->inputq);
666 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
667 return;
668 }
669 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
670
671 /* There is still a working link => initiate failover */
672 *bearer_id = n->active_links[0];
673 tnl = n->links[*bearer_id].link;
674 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
675 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
676 n->sync_point = tipc_link_rcv_nxt(tnl) + (U16_MAX / 2 - 1);
677 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
678 tipc_link_reset(l);
679 tipc_link_fsm_evt(l, LINK_RESET_EVT);
680 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
681 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
682 *maddr = &n->links[*bearer_id].maddr;
683 }
684
685 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
686 {
687 struct tipc_link_entry *le = &n->links[bearer_id];
688 struct tipc_link *l = le->link;
689 struct tipc_media_addr *maddr;
690 struct sk_buff_head xmitq;
691
692 if (!l)
693 return;
694
695 __skb_queue_head_init(&xmitq);
696
697 tipc_node_write_lock(n);
698 if (!tipc_link_is_establishing(l)) {
699 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
700 if (delete) {
701 kfree(l);
702 le->link = NULL;
703 n->link_cnt--;
704 }
705 } else {
706 /* Defuse pending tipc_node_link_up() */
707 tipc_link_fsm_evt(l, LINK_RESET_EVT);
708 }
709 tipc_node_write_unlock(n);
710 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
711 tipc_sk_rcv(n->net, &le->inputq);
712 }
713
714 static bool tipc_node_is_up(struct tipc_node *n)
715 {
716 return n->active_links[0] != INVALID_BEARER_ID;
717 }
718
719 void tipc_node_check_dest(struct net *net, u32 onode,
720 struct tipc_bearer *b,
721 u16 capabilities, u32 signature,
722 struct tipc_media_addr *maddr,
723 bool *respond, bool *dupl_addr)
724 {
725 struct tipc_node *n;
726 struct tipc_link *l;
727 struct tipc_link_entry *le;
728 bool addr_match = false;
729 bool sign_match = false;
730 bool link_up = false;
731 bool accept_addr = false;
732 bool reset = true;
733 char *if_name;
734
735 *dupl_addr = false;
736 *respond = false;
737
738 n = tipc_node_create(net, onode, capabilities);
739 if (!n)
740 return;
741
742 tipc_node_write_lock(n);
743
744 le = &n->links[b->identity];
745
746 /* Prepare to validate requesting node's signature and media address */
747 l = le->link;
748 link_up = l && tipc_link_is_up(l);
749 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
750 sign_match = (signature == n->signature);
751
752 /* These three flags give us eight permutations: */
753
754 if (sign_match && addr_match && link_up) {
755 /* All is fine. Do nothing. */
756 reset = false;
757 } else if (sign_match && addr_match && !link_up) {
758 /* Respond. The link will come up in due time */
759 *respond = true;
760 } else if (sign_match && !addr_match && link_up) {
761 /* Peer has changed i/f address without rebooting.
762 * If so, the link will reset soon, and the next
763 * discovery will be accepted. So we can ignore it.
764 * It may also be an cloned or malicious peer having
765 * chosen the same node address and signature as an
766 * existing one.
767 * Ignore requests until the link goes down, if ever.
768 */
769 *dupl_addr = true;
770 } else if (sign_match && !addr_match && !link_up) {
771 /* Peer link has changed i/f address without rebooting.
772 * It may also be a cloned or malicious peer; we can't
773 * distinguish between the two.
774 * The signature is correct, so we must accept.
775 */
776 accept_addr = true;
777 *respond = true;
778 } else if (!sign_match && addr_match && link_up) {
779 /* Peer node rebooted. Two possibilities:
780 * - Delayed re-discovery; this link endpoint has already
781 * reset and re-established contact with the peer, before
782 * receiving a discovery message from that node.
783 * (The peer happened to receive one from this node first).
784 * - The peer came back so fast that our side has not
785 * discovered it yet. Probing from this side will soon
786 * reset the link, since there can be no working link
787 * endpoint at the peer end, and the link will re-establish.
788 * Accept the signature, since it comes from a known peer.
789 */
790 n->signature = signature;
791 } else if (!sign_match && addr_match && !link_up) {
792 /* The peer node has rebooted.
793 * Accept signature, since it is a known peer.
794 */
795 n->signature = signature;
796 *respond = true;
797 } else if (!sign_match && !addr_match && link_up) {
798 /* Peer rebooted with new address, or a new/duplicate peer.
799 * Ignore until the link goes down, if ever.
800 */
801 *dupl_addr = true;
802 } else if (!sign_match && !addr_match && !link_up) {
803 /* Peer rebooted with new address, or it is a new peer.
804 * Accept signature and address.
805 */
806 n->signature = signature;
807 accept_addr = true;
808 *respond = true;
809 }
810
811 if (!accept_addr)
812 goto exit;
813
814 /* Now create new link if not already existing */
815 if (!l) {
816 if (n->link_cnt == 2) {
817 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
818 goto exit;
819 }
820 if_name = strchr(b->name, ':') + 1;
821 if (!tipc_link_create(net, if_name, b->identity, b->tolerance,
822 b->net_plane, b->mtu, b->priority,
823 b->window, mod(tipc_net(net)->random),
824 tipc_own_addr(net), onode,
825 n->capabilities,
826 tipc_bc_sndlink(n->net), n->bc_entry.link,
827 &le->inputq,
828 &n->bc_entry.namedq, &l)) {
829 *respond = false;
830 goto exit;
831 }
832 tipc_link_reset(l);
833 tipc_link_fsm_evt(l, LINK_RESET_EVT);
834 if (n->state == NODE_FAILINGOVER)
835 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
836 le->link = l;
837 n->link_cnt++;
838 tipc_node_calculate_timer(n, l);
839 if (n->link_cnt == 1)
840 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
841 tipc_node_get(n);
842 }
843 memcpy(&le->maddr, maddr, sizeof(*maddr));
844 exit:
845 tipc_node_write_unlock(n);
846 if (reset && !tipc_link_is_reset(l))
847 tipc_node_link_down(n, b->identity, false);
848 tipc_node_put(n);
849 }
850
851 void tipc_node_delete_links(struct net *net, int bearer_id)
852 {
853 struct tipc_net *tn = net_generic(net, tipc_net_id);
854 struct tipc_node *n;
855
856 rcu_read_lock();
857 list_for_each_entry_rcu(n, &tn->node_list, list) {
858 tipc_node_link_down(n, bearer_id, true);
859 }
860 rcu_read_unlock();
861 }
862
863 static void tipc_node_reset_links(struct tipc_node *n)
864 {
865 char addr_string[16];
866 int i;
867
868 pr_warn("Resetting all links to %s\n",
869 tipc_addr_string_fill(addr_string, n->addr));
870
871 for (i = 0; i < MAX_BEARERS; i++) {
872 tipc_node_link_down(n, i, false);
873 }
874 }
875
876 /* tipc_node_fsm_evt - node finite state machine
877 * Determines when contact is allowed with peer node
878 */
879 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
880 {
881 int state = n->state;
882
883 switch (state) {
884 case SELF_DOWN_PEER_DOWN:
885 switch (evt) {
886 case SELF_ESTABL_CONTACT_EVT:
887 state = SELF_UP_PEER_COMING;
888 break;
889 case PEER_ESTABL_CONTACT_EVT:
890 state = SELF_COMING_PEER_UP;
891 break;
892 case SELF_LOST_CONTACT_EVT:
893 case PEER_LOST_CONTACT_EVT:
894 break;
895 case NODE_SYNCH_END_EVT:
896 case NODE_SYNCH_BEGIN_EVT:
897 case NODE_FAILOVER_BEGIN_EVT:
898 case NODE_FAILOVER_END_EVT:
899 default:
900 goto illegal_evt;
901 }
902 break;
903 case SELF_UP_PEER_UP:
904 switch (evt) {
905 case SELF_LOST_CONTACT_EVT:
906 state = SELF_DOWN_PEER_LEAVING;
907 break;
908 case PEER_LOST_CONTACT_EVT:
909 state = SELF_LEAVING_PEER_DOWN;
910 break;
911 case NODE_SYNCH_BEGIN_EVT:
912 state = NODE_SYNCHING;
913 break;
914 case NODE_FAILOVER_BEGIN_EVT:
915 state = NODE_FAILINGOVER;
916 break;
917 case SELF_ESTABL_CONTACT_EVT:
918 case PEER_ESTABL_CONTACT_EVT:
919 case NODE_SYNCH_END_EVT:
920 case NODE_FAILOVER_END_EVT:
921 break;
922 default:
923 goto illegal_evt;
924 }
925 break;
926 case SELF_DOWN_PEER_LEAVING:
927 switch (evt) {
928 case PEER_LOST_CONTACT_EVT:
929 state = SELF_DOWN_PEER_DOWN;
930 break;
931 case SELF_ESTABL_CONTACT_EVT:
932 case PEER_ESTABL_CONTACT_EVT:
933 case SELF_LOST_CONTACT_EVT:
934 break;
935 case NODE_SYNCH_END_EVT:
936 case NODE_SYNCH_BEGIN_EVT:
937 case NODE_FAILOVER_BEGIN_EVT:
938 case NODE_FAILOVER_END_EVT:
939 default:
940 goto illegal_evt;
941 }
942 break;
943 case SELF_UP_PEER_COMING:
944 switch (evt) {
945 case PEER_ESTABL_CONTACT_EVT:
946 state = SELF_UP_PEER_UP;
947 break;
948 case SELF_LOST_CONTACT_EVT:
949 state = SELF_DOWN_PEER_LEAVING;
950 break;
951 case SELF_ESTABL_CONTACT_EVT:
952 case PEER_LOST_CONTACT_EVT:
953 case NODE_SYNCH_END_EVT:
954 case NODE_FAILOVER_BEGIN_EVT:
955 break;
956 case NODE_SYNCH_BEGIN_EVT:
957 case NODE_FAILOVER_END_EVT:
958 default:
959 goto illegal_evt;
960 }
961 break;
962 case SELF_COMING_PEER_UP:
963 switch (evt) {
964 case SELF_ESTABL_CONTACT_EVT:
965 state = SELF_UP_PEER_UP;
966 break;
967 case PEER_LOST_CONTACT_EVT:
968 state = SELF_LEAVING_PEER_DOWN;
969 break;
970 case SELF_LOST_CONTACT_EVT:
971 case PEER_ESTABL_CONTACT_EVT:
972 break;
973 case NODE_SYNCH_END_EVT:
974 case NODE_SYNCH_BEGIN_EVT:
975 case NODE_FAILOVER_BEGIN_EVT:
976 case NODE_FAILOVER_END_EVT:
977 default:
978 goto illegal_evt;
979 }
980 break;
981 case SELF_LEAVING_PEER_DOWN:
982 switch (evt) {
983 case SELF_LOST_CONTACT_EVT:
984 state = SELF_DOWN_PEER_DOWN;
985 break;
986 case SELF_ESTABL_CONTACT_EVT:
987 case PEER_ESTABL_CONTACT_EVT:
988 case PEER_LOST_CONTACT_EVT:
989 break;
990 case NODE_SYNCH_END_EVT:
991 case NODE_SYNCH_BEGIN_EVT:
992 case NODE_FAILOVER_BEGIN_EVT:
993 case NODE_FAILOVER_END_EVT:
994 default:
995 goto illegal_evt;
996 }
997 break;
998 case NODE_FAILINGOVER:
999 switch (evt) {
1000 case SELF_LOST_CONTACT_EVT:
1001 state = SELF_DOWN_PEER_LEAVING;
1002 break;
1003 case PEER_LOST_CONTACT_EVT:
1004 state = SELF_LEAVING_PEER_DOWN;
1005 break;
1006 case NODE_FAILOVER_END_EVT:
1007 state = SELF_UP_PEER_UP;
1008 break;
1009 case NODE_FAILOVER_BEGIN_EVT:
1010 case SELF_ESTABL_CONTACT_EVT:
1011 case PEER_ESTABL_CONTACT_EVT:
1012 break;
1013 case NODE_SYNCH_BEGIN_EVT:
1014 case NODE_SYNCH_END_EVT:
1015 default:
1016 goto illegal_evt;
1017 }
1018 break;
1019 case NODE_SYNCHING:
1020 switch (evt) {
1021 case SELF_LOST_CONTACT_EVT:
1022 state = SELF_DOWN_PEER_LEAVING;
1023 break;
1024 case PEER_LOST_CONTACT_EVT:
1025 state = SELF_LEAVING_PEER_DOWN;
1026 break;
1027 case NODE_SYNCH_END_EVT:
1028 state = SELF_UP_PEER_UP;
1029 break;
1030 case NODE_FAILOVER_BEGIN_EVT:
1031 state = NODE_FAILINGOVER;
1032 break;
1033 case NODE_SYNCH_BEGIN_EVT:
1034 case SELF_ESTABL_CONTACT_EVT:
1035 case PEER_ESTABL_CONTACT_EVT:
1036 break;
1037 case NODE_FAILOVER_END_EVT:
1038 default:
1039 goto illegal_evt;
1040 }
1041 break;
1042 default:
1043 pr_err("Unknown node fsm state %x\n", state);
1044 break;
1045 }
1046 n->state = state;
1047 return;
1048
1049 illegal_evt:
1050 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
1051 }
1052
1053 static void node_lost_contact(struct tipc_node *n,
1054 struct sk_buff_head *inputq)
1055 {
1056 char addr_string[16];
1057 struct tipc_sock_conn *conn, *safe;
1058 struct tipc_link *l;
1059 struct list_head *conns = &n->conn_sks;
1060 struct sk_buff *skb;
1061 uint i;
1062
1063 pr_debug("Lost contact with %s\n",
1064 tipc_addr_string_fill(addr_string, n->addr));
1065
1066 /* Clean up broadcast state */
1067 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
1068
1069 /* Abort any ongoing link failover */
1070 for (i = 0; i < MAX_BEARERS; i++) {
1071 l = n->links[i].link;
1072 if (l)
1073 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
1074 }
1075
1076 /* Notify publications from this node */
1077 n->action_flags |= TIPC_NOTIFY_NODE_DOWN;
1078
1079 /* Notify sockets connected to node */
1080 list_for_each_entry_safe(conn, safe, conns, list) {
1081 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
1082 SHORT_H_SIZE, 0, tipc_own_addr(n->net),
1083 conn->peer_node, conn->port,
1084 conn->peer_port, TIPC_ERR_NO_NODE);
1085 if (likely(skb))
1086 skb_queue_tail(inputq, skb);
1087 list_del(&conn->list);
1088 kfree(conn);
1089 }
1090 }
1091
1092 /**
1093 * tipc_node_get_linkname - get the name of a link
1094 *
1095 * @bearer_id: id of the bearer
1096 * @node: peer node address
1097 * @linkname: link name output buffer
1098 *
1099 * Returns 0 on success
1100 */
1101 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
1102 char *linkname, size_t len)
1103 {
1104 struct tipc_link *link;
1105 int err = -EINVAL;
1106 struct tipc_node *node = tipc_node_find(net, addr);
1107
1108 if (!node)
1109 return err;
1110
1111 if (bearer_id >= MAX_BEARERS)
1112 goto exit;
1113
1114 tipc_node_read_lock(node);
1115 link = node->links[bearer_id].link;
1116 if (link) {
1117 strncpy(linkname, tipc_link_name(link), len);
1118 err = 0;
1119 }
1120 exit:
1121 tipc_node_read_unlock(node);
1122 tipc_node_put(node);
1123 return err;
1124 }
1125
1126 /* Caller should hold node lock for the passed node */
1127 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
1128 {
1129 void *hdr;
1130 struct nlattr *attrs;
1131
1132 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1133 NLM_F_MULTI, TIPC_NL_NODE_GET);
1134 if (!hdr)
1135 return -EMSGSIZE;
1136
1137 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
1138 if (!attrs)
1139 goto msg_full;
1140
1141 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
1142 goto attr_msg_full;
1143 if (tipc_node_is_up(node))
1144 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
1145 goto attr_msg_full;
1146
1147 nla_nest_end(msg->skb, attrs);
1148 genlmsg_end(msg->skb, hdr);
1149
1150 return 0;
1151
1152 attr_msg_full:
1153 nla_nest_cancel(msg->skb, attrs);
1154 msg_full:
1155 genlmsg_cancel(msg->skb, hdr);
1156
1157 return -EMSGSIZE;
1158 }
1159
1160 /**
1161 * tipc_node_xmit() is the general link level function for message sending
1162 * @net: the applicable net namespace
1163 * @list: chain of buffers containing message
1164 * @dnode: address of destination node
1165 * @selector: a number used for deterministic link selection
1166 * Consumes the buffer chain, except when returning -ELINKCONG
1167 * Returns 0 if success, otherwise: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE,-ENOBUF
1168 */
1169 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1170 u32 dnode, int selector)
1171 {
1172 struct tipc_link_entry *le = NULL;
1173 struct tipc_node *n;
1174 struct sk_buff_head xmitq;
1175 int bearer_id;
1176 int rc;
1177
1178 if (in_own_node(net, dnode)) {
1179 tipc_sk_rcv(net, list);
1180 return 0;
1181 }
1182
1183 n = tipc_node_find(net, dnode);
1184 if (unlikely(!n)) {
1185 skb_queue_purge(list);
1186 return -EHOSTUNREACH;
1187 }
1188
1189 tipc_node_read_lock(n);
1190 bearer_id = n->active_links[selector & 1];
1191 if (unlikely(bearer_id == INVALID_BEARER_ID)) {
1192 tipc_node_read_unlock(n);
1193 tipc_node_put(n);
1194 skb_queue_purge(list);
1195 return -EHOSTUNREACH;
1196 }
1197
1198 __skb_queue_head_init(&xmitq);
1199 le = &n->links[bearer_id];
1200 spin_lock_bh(&le->lock);
1201 rc = tipc_link_xmit(le->link, list, &xmitq);
1202 spin_unlock_bh(&le->lock);
1203 tipc_node_read_unlock(n);
1204
1205 if (likely(rc == 0))
1206 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1207 else if (rc == -ENOBUFS)
1208 tipc_node_link_down(n, bearer_id, false);
1209
1210 tipc_node_put(n);
1211
1212 return rc;
1213 }
1214
1215 /* tipc_node_xmit_skb(): send single buffer to destination
1216 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1217 * messages, which will not be rejected
1218 * The only exception is datagram messages rerouted after secondary
1219 * lookup, which are rare and safe to dispose of anyway.
1220 * TODO: Return real return value, and let callers use
1221 * tipc_wait_for_sendpkt() where applicable
1222 */
1223 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1224 u32 selector)
1225 {
1226 struct sk_buff_head head;
1227 int rc;
1228
1229 skb_queue_head_init(&head);
1230 __skb_queue_tail(&head, skb);
1231 rc = tipc_node_xmit(net, &head, dnode, selector);
1232 if (rc == -ELINKCONG)
1233 kfree_skb(skb);
1234 return 0;
1235 }
1236
1237 void tipc_node_broadcast(struct net *net, struct sk_buff *skb)
1238 {
1239 struct sk_buff *txskb;
1240 struct tipc_node *n;
1241 u32 dst;
1242
1243 rcu_read_lock();
1244 list_for_each_entry_rcu(n, tipc_nodes(net), list) {
1245 dst = n->addr;
1246 if (in_own_node(net, dst))
1247 continue;
1248 if (!tipc_node_is_up(n))
1249 continue;
1250 txskb = pskb_copy(skb, GFP_ATOMIC);
1251 if (!txskb)
1252 break;
1253 msg_set_destnode(buf_msg(txskb), dst);
1254 tipc_node_xmit_skb(net, txskb, dst, 0);
1255 }
1256 rcu_read_unlock();
1257
1258 kfree_skb(skb);
1259 }
1260
1261 /**
1262 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1263 * @net: the applicable net namespace
1264 * @skb: TIPC packet
1265 * @bearer_id: id of bearer message arrived on
1266 *
1267 * Invoked with no locks held.
1268 */
1269 static void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1270 {
1271 int rc;
1272 struct sk_buff_head xmitq;
1273 struct tipc_bclink_entry *be;
1274 struct tipc_link_entry *le;
1275 struct tipc_msg *hdr = buf_msg(skb);
1276 int usr = msg_user(hdr);
1277 u32 dnode = msg_destnode(hdr);
1278 struct tipc_node *n;
1279
1280 __skb_queue_head_init(&xmitq);
1281
1282 /* If NACK for other node, let rcv link for that node peek into it */
1283 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1284 n = tipc_node_find(net, dnode);
1285 else
1286 n = tipc_node_find(net, msg_prevnode(hdr));
1287 if (!n) {
1288 kfree_skb(skb);
1289 return;
1290 }
1291 be = &n->bc_entry;
1292 le = &n->links[bearer_id];
1293
1294 rc = tipc_bcast_rcv(net, be->link, skb);
1295
1296 /* Broadcast link reset may happen at reassembly failure */
1297 if (rc & TIPC_LINK_DOWN_EVT)
1298 tipc_node_reset_links(n);
1299
1300 /* Broadcast ACKs are sent on a unicast link */
1301 if (rc & TIPC_LINK_SND_BC_ACK) {
1302 tipc_node_read_lock(n);
1303 tipc_link_build_ack_msg(le->link, &xmitq);
1304 tipc_node_read_unlock(n);
1305 }
1306
1307 if (!skb_queue_empty(&xmitq))
1308 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1309
1310 /* Deliver. 'arrvq' is under inputq2's lock protection */
1311 if (!skb_queue_empty(&be->inputq1)) {
1312 spin_lock_bh(&be->inputq2.lock);
1313 spin_lock_bh(&be->inputq1.lock);
1314 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1315 spin_unlock_bh(&be->inputq1.lock);
1316 spin_unlock_bh(&be->inputq2.lock);
1317 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1318 }
1319 tipc_node_put(n);
1320 }
1321
1322 /**
1323 * tipc_node_check_state - check and if necessary update node state
1324 * @skb: TIPC packet
1325 * @bearer_id: identity of bearer delivering the packet
1326 * Returns true if state is ok, otherwise consumes buffer and returns false
1327 */
1328 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1329 int bearer_id, struct sk_buff_head *xmitq)
1330 {
1331 struct tipc_msg *hdr = buf_msg(skb);
1332 int usr = msg_user(hdr);
1333 int mtyp = msg_type(hdr);
1334 u16 oseqno = msg_seqno(hdr);
1335 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1336 u16 exp_pkts = msg_msgcnt(hdr);
1337 u16 rcv_nxt, syncpt, dlv_nxt, inputq_len;
1338 int state = n->state;
1339 struct tipc_link *l, *tnl, *pl = NULL;
1340 struct tipc_media_addr *maddr;
1341 int pb_id;
1342
1343 l = n->links[bearer_id].link;
1344 if (!l)
1345 return false;
1346 rcv_nxt = tipc_link_rcv_nxt(l);
1347
1348
1349 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1350 return true;
1351
1352 /* Find parallel link, if any */
1353 for (pb_id = 0; pb_id < MAX_BEARERS; pb_id++) {
1354 if ((pb_id != bearer_id) && n->links[pb_id].link) {
1355 pl = n->links[pb_id].link;
1356 break;
1357 }
1358 }
1359
1360 /* Check and update node accesibility if applicable */
1361 if (state == SELF_UP_PEER_COMING) {
1362 if (!tipc_link_is_up(l))
1363 return true;
1364 if (!msg_peer_link_is_up(hdr))
1365 return true;
1366 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1367 }
1368
1369 if (state == SELF_DOWN_PEER_LEAVING) {
1370 if (msg_peer_node_is_up(hdr))
1371 return false;
1372 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1373 return true;
1374 }
1375
1376 if (state == SELF_LEAVING_PEER_DOWN)
1377 return false;
1378
1379 /* Ignore duplicate packets */
1380 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1381 return true;
1382
1383 /* Initiate or update failover mode if applicable */
1384 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1385 syncpt = oseqno + exp_pkts - 1;
1386 if (pl && tipc_link_is_up(pl)) {
1387 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1388 tipc_skb_queue_splice_tail_init(tipc_link_inputq(pl),
1389 tipc_link_inputq(l));
1390 }
1391 /* If pkts arrive out of order, use lowest calculated syncpt */
1392 if (less(syncpt, n->sync_point))
1393 n->sync_point = syncpt;
1394 }
1395
1396 /* Open parallel link when tunnel link reaches synch point */
1397 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1398 if (!more(rcv_nxt, n->sync_point))
1399 return true;
1400 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1401 if (pl)
1402 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1403 return true;
1404 }
1405
1406 /* No synching needed if only one link */
1407 if (!pl || !tipc_link_is_up(pl))
1408 return true;
1409
1410 /* Initiate synch mode if applicable */
1411 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1412 syncpt = iseqno + exp_pkts - 1;
1413 if (!tipc_link_is_up(l)) {
1414 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1415 __tipc_node_link_up(n, bearer_id, xmitq);
1416 }
1417 if (n->state == SELF_UP_PEER_UP) {
1418 n->sync_point = syncpt;
1419 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1420 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1421 }
1422 }
1423
1424 /* Open tunnel link when parallel link reaches synch point */
1425 if (n->state == NODE_SYNCHING) {
1426 if (tipc_link_is_synching(l)) {
1427 tnl = l;
1428 } else {
1429 tnl = pl;
1430 pl = l;
1431 }
1432 inputq_len = skb_queue_len(tipc_link_inputq(pl));
1433 dlv_nxt = tipc_link_rcv_nxt(pl) - inputq_len;
1434 if (more(dlv_nxt, n->sync_point)) {
1435 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1436 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1437 return true;
1438 }
1439 if (l == pl)
1440 return true;
1441 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1442 return true;
1443 if (usr == LINK_PROTOCOL)
1444 return true;
1445 return false;
1446 }
1447 return true;
1448 }
1449
1450 /**
1451 * tipc_rcv - process TIPC packets/messages arriving from off-node
1452 * @net: the applicable net namespace
1453 * @skb: TIPC packet
1454 * @bearer: pointer to bearer message arrived on
1455 *
1456 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1457 * structure (i.e. cannot be NULL), but bearer can be inactive.
1458 */
1459 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1460 {
1461 struct sk_buff_head xmitq;
1462 struct tipc_node *n;
1463 struct tipc_msg *hdr = buf_msg(skb);
1464 int usr = msg_user(hdr);
1465 int bearer_id = b->identity;
1466 struct tipc_link_entry *le;
1467 u16 bc_ack = msg_bcast_ack(hdr);
1468 int rc = 0;
1469
1470 __skb_queue_head_init(&xmitq);
1471
1472 /* Ensure message is well-formed */
1473 if (unlikely(!tipc_msg_validate(skb)))
1474 goto discard;
1475
1476 /* Handle arrival of discovery or broadcast packet */
1477 if (unlikely(msg_non_seq(hdr))) {
1478 if (unlikely(usr == LINK_CONFIG))
1479 return tipc_disc_rcv(net, skb, b);
1480 else
1481 return tipc_node_bc_rcv(net, skb, bearer_id);
1482 }
1483
1484 /* Locate neighboring node that sent packet */
1485 n = tipc_node_find(net, msg_prevnode(hdr));
1486 if (unlikely(!n))
1487 goto discard;
1488 le = &n->links[bearer_id];
1489
1490 /* Ensure broadcast reception is in synch with peer's send state */
1491 if (unlikely(usr == LINK_PROTOCOL))
1492 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1493 else if (unlikely(tipc_link_acked(n->bc_entry.link) != bc_ack))
1494 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1495
1496 /* Receive packet directly if conditions permit */
1497 tipc_node_read_lock(n);
1498 if (likely((n->state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL))) {
1499 spin_lock_bh(&le->lock);
1500 if (le->link) {
1501 rc = tipc_link_rcv(le->link, skb, &xmitq);
1502 skb = NULL;
1503 }
1504 spin_unlock_bh(&le->lock);
1505 }
1506 tipc_node_read_unlock(n);
1507
1508 /* Check/update node state before receiving */
1509 if (unlikely(skb)) {
1510 tipc_node_write_lock(n);
1511 if (tipc_node_check_state(n, skb, bearer_id, &xmitq)) {
1512 if (le->link) {
1513 rc = tipc_link_rcv(le->link, skb, &xmitq);
1514 skb = NULL;
1515 }
1516 }
1517 tipc_node_write_unlock(n);
1518 }
1519
1520 if (unlikely(rc & TIPC_LINK_UP_EVT))
1521 tipc_node_link_up(n, bearer_id, &xmitq);
1522
1523 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1524 tipc_node_link_down(n, bearer_id, false);
1525
1526 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1527 tipc_named_rcv(net, &n->bc_entry.namedq);
1528
1529 if (!skb_queue_empty(&le->inputq))
1530 tipc_sk_rcv(net, &le->inputq);
1531
1532 if (!skb_queue_empty(&xmitq))
1533 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1534
1535 tipc_node_put(n);
1536 discard:
1537 kfree_skb(skb);
1538 }
1539
1540 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1541 {
1542 int err;
1543 struct net *net = sock_net(skb->sk);
1544 struct tipc_net *tn = net_generic(net, tipc_net_id);
1545 int done = cb->args[0];
1546 int last_addr = cb->args[1];
1547 struct tipc_node *node;
1548 struct tipc_nl_msg msg;
1549
1550 if (done)
1551 return 0;
1552
1553 msg.skb = skb;
1554 msg.portid = NETLINK_CB(cb->skb).portid;
1555 msg.seq = cb->nlh->nlmsg_seq;
1556
1557 rcu_read_lock();
1558 if (last_addr) {
1559 node = tipc_node_find(net, last_addr);
1560 if (!node) {
1561 rcu_read_unlock();
1562 /* We never set seq or call nl_dump_check_consistent()
1563 * this means that setting prev_seq here will cause the
1564 * consistence check to fail in the netlink callback
1565 * handler. Resulting in the NLMSG_DONE message having
1566 * the NLM_F_DUMP_INTR flag set if the node state
1567 * changed while we released the lock.
1568 */
1569 cb->prev_seq = 1;
1570 return -EPIPE;
1571 }
1572 tipc_node_put(node);
1573 }
1574
1575 list_for_each_entry_rcu(node, &tn->node_list, list) {
1576 if (last_addr) {
1577 if (node->addr == last_addr)
1578 last_addr = 0;
1579 else
1580 continue;
1581 }
1582
1583 tipc_node_read_lock(node);
1584 err = __tipc_nl_add_node(&msg, node);
1585 if (err) {
1586 last_addr = node->addr;
1587 tipc_node_read_unlock(node);
1588 goto out;
1589 }
1590
1591 tipc_node_read_unlock(node);
1592 }
1593 done = 1;
1594 out:
1595 cb->args[0] = done;
1596 cb->args[1] = last_addr;
1597 rcu_read_unlock();
1598
1599 return skb->len;
1600 }
1601
1602 /* tipc_node_find_by_name - locate owner node of link by link's name
1603 * @net: the applicable net namespace
1604 * @name: pointer to link name string
1605 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1606 *
1607 * Returns pointer to node owning the link, or 0 if no matching link is found.
1608 */
1609 static struct tipc_node *tipc_node_find_by_name(struct net *net,
1610 const char *link_name,
1611 unsigned int *bearer_id)
1612 {
1613 struct tipc_net *tn = net_generic(net, tipc_net_id);
1614 struct tipc_link *l;
1615 struct tipc_node *n;
1616 struct tipc_node *found_node = NULL;
1617 int i;
1618
1619 *bearer_id = 0;
1620 rcu_read_lock();
1621 list_for_each_entry_rcu(n, &tn->node_list, list) {
1622 tipc_node_read_lock(n);
1623 for (i = 0; i < MAX_BEARERS; i++) {
1624 l = n->links[i].link;
1625 if (l && !strcmp(tipc_link_name(l), link_name)) {
1626 *bearer_id = i;
1627 found_node = n;
1628 break;
1629 }
1630 }
1631 tipc_node_read_unlock(n);
1632 if (found_node)
1633 break;
1634 }
1635 rcu_read_unlock();
1636
1637 return found_node;
1638 }
1639
1640 int tipc_nl_node_set_link(struct sk_buff *skb, struct genl_info *info)
1641 {
1642 int err;
1643 int res = 0;
1644 int bearer_id;
1645 char *name;
1646 struct tipc_link *link;
1647 struct tipc_node *node;
1648 struct sk_buff_head xmitq;
1649 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1650 struct net *net = sock_net(skb->sk);
1651
1652 __skb_queue_head_init(&xmitq);
1653
1654 if (!info->attrs[TIPC_NLA_LINK])
1655 return -EINVAL;
1656
1657 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1658 info->attrs[TIPC_NLA_LINK],
1659 tipc_nl_link_policy);
1660 if (err)
1661 return err;
1662
1663 if (!attrs[TIPC_NLA_LINK_NAME])
1664 return -EINVAL;
1665
1666 name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1667
1668 if (strcmp(name, tipc_bclink_name) == 0)
1669 return tipc_nl_bc_link_set(net, attrs);
1670
1671 node = tipc_node_find_by_name(net, name, &bearer_id);
1672 if (!node)
1673 return -EINVAL;
1674
1675 tipc_node_read_lock(node);
1676
1677 link = node->links[bearer_id].link;
1678 if (!link) {
1679 res = -EINVAL;
1680 goto out;
1681 }
1682
1683 if (attrs[TIPC_NLA_LINK_PROP]) {
1684 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1685
1686 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1687 props);
1688 if (err) {
1689 res = err;
1690 goto out;
1691 }
1692
1693 if (props[TIPC_NLA_PROP_TOL]) {
1694 u32 tol;
1695
1696 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1697 tipc_link_set_tolerance(link, tol, &xmitq);
1698 }
1699 if (props[TIPC_NLA_PROP_PRIO]) {
1700 u32 prio;
1701
1702 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1703 tipc_link_set_prio(link, prio, &xmitq);
1704 }
1705 if (props[TIPC_NLA_PROP_WIN]) {
1706 u32 win;
1707
1708 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1709 tipc_link_set_queue_limits(link, win);
1710 }
1711 }
1712
1713 out:
1714 tipc_node_read_unlock(node);
1715 tipc_bearer_xmit(net, bearer_id, &xmitq, &node->links[bearer_id].maddr);
1716 return res;
1717 }
1718
1719 int tipc_nl_node_get_link(struct sk_buff *skb, struct genl_info *info)
1720 {
1721 struct net *net = genl_info_net(info);
1722 struct tipc_nl_msg msg;
1723 char *name;
1724 int err;
1725
1726 msg.portid = info->snd_portid;
1727 msg.seq = info->snd_seq;
1728
1729 if (!info->attrs[TIPC_NLA_LINK_NAME])
1730 return -EINVAL;
1731 name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
1732
1733 msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1734 if (!msg.skb)
1735 return -ENOMEM;
1736
1737 if (strcmp(name, tipc_bclink_name) == 0) {
1738 err = tipc_nl_add_bc_link(net, &msg);
1739 if (err) {
1740 nlmsg_free(msg.skb);
1741 return err;
1742 }
1743 } else {
1744 int bearer_id;
1745 struct tipc_node *node;
1746 struct tipc_link *link;
1747
1748 node = tipc_node_find_by_name(net, name, &bearer_id);
1749 if (!node)
1750 return -EINVAL;
1751
1752 tipc_node_read_lock(node);
1753 link = node->links[bearer_id].link;
1754 if (!link) {
1755 tipc_node_read_unlock(node);
1756 nlmsg_free(msg.skb);
1757 return -EINVAL;
1758 }
1759
1760 err = __tipc_nl_add_link(net, &msg, link, 0);
1761 tipc_node_read_unlock(node);
1762 if (err) {
1763 nlmsg_free(msg.skb);
1764 return err;
1765 }
1766 }
1767
1768 return genlmsg_reply(msg.skb, info);
1769 }
1770
1771 int tipc_nl_node_reset_link_stats(struct sk_buff *skb, struct genl_info *info)
1772 {
1773 int err;
1774 char *link_name;
1775 unsigned int bearer_id;
1776 struct tipc_link *link;
1777 struct tipc_node *node;
1778 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1779 struct net *net = sock_net(skb->sk);
1780 struct tipc_link_entry *le;
1781
1782 if (!info->attrs[TIPC_NLA_LINK])
1783 return -EINVAL;
1784
1785 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1786 info->attrs[TIPC_NLA_LINK],
1787 tipc_nl_link_policy);
1788 if (err)
1789 return err;
1790
1791 if (!attrs[TIPC_NLA_LINK_NAME])
1792 return -EINVAL;
1793
1794 link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1795
1796 if (strcmp(link_name, tipc_bclink_name) == 0) {
1797 err = tipc_bclink_reset_stats(net);
1798 if (err)
1799 return err;
1800 return 0;
1801 }
1802
1803 node = tipc_node_find_by_name(net, link_name, &bearer_id);
1804 if (!node)
1805 return -EINVAL;
1806
1807 le = &node->links[bearer_id];
1808 tipc_node_read_lock(node);
1809 spin_lock_bh(&le->lock);
1810 link = node->links[bearer_id].link;
1811 if (!link) {
1812 spin_unlock_bh(&le->lock);
1813 tipc_node_read_unlock(node);
1814 return -EINVAL;
1815 }
1816 tipc_link_reset_stats(link);
1817 spin_unlock_bh(&le->lock);
1818 tipc_node_read_unlock(node);
1819 return 0;
1820 }
1821
1822 /* Caller should hold node lock */
1823 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
1824 struct tipc_node *node, u32 *prev_link)
1825 {
1826 u32 i;
1827 int err;
1828
1829 for (i = *prev_link; i < MAX_BEARERS; i++) {
1830 *prev_link = i;
1831
1832 if (!node->links[i].link)
1833 continue;
1834
1835 err = __tipc_nl_add_link(net, msg,
1836 node->links[i].link, NLM_F_MULTI);
1837 if (err)
1838 return err;
1839 }
1840 *prev_link = 0;
1841
1842 return 0;
1843 }
1844
1845 int tipc_nl_node_dump_link(struct sk_buff *skb, struct netlink_callback *cb)
1846 {
1847 struct net *net = sock_net(skb->sk);
1848 struct tipc_net *tn = net_generic(net, tipc_net_id);
1849 struct tipc_node *node;
1850 struct tipc_nl_msg msg;
1851 u32 prev_node = cb->args[0];
1852 u32 prev_link = cb->args[1];
1853 int done = cb->args[2];
1854 int err;
1855
1856 if (done)
1857 return 0;
1858
1859 msg.skb = skb;
1860 msg.portid = NETLINK_CB(cb->skb).portid;
1861 msg.seq = cb->nlh->nlmsg_seq;
1862
1863 rcu_read_lock();
1864 if (prev_node) {
1865 node = tipc_node_find(net, prev_node);
1866 if (!node) {
1867 /* We never set seq or call nl_dump_check_consistent()
1868 * this means that setting prev_seq here will cause the
1869 * consistence check to fail in the netlink callback
1870 * handler. Resulting in the last NLMSG_DONE message
1871 * having the NLM_F_DUMP_INTR flag set.
1872 */
1873 cb->prev_seq = 1;
1874 goto out;
1875 }
1876 tipc_node_put(node);
1877
1878 list_for_each_entry_continue_rcu(node, &tn->node_list,
1879 list) {
1880 tipc_node_read_lock(node);
1881 err = __tipc_nl_add_node_links(net, &msg, node,
1882 &prev_link);
1883 tipc_node_read_unlock(node);
1884 if (err)
1885 goto out;
1886
1887 prev_node = node->addr;
1888 }
1889 } else {
1890 err = tipc_nl_add_bc_link(net, &msg);
1891 if (err)
1892 goto out;
1893
1894 list_for_each_entry_rcu(node, &tn->node_list, list) {
1895 tipc_node_read_lock(node);
1896 err = __tipc_nl_add_node_links(net, &msg, node,
1897 &prev_link);
1898 tipc_node_read_unlock(node);
1899 if (err)
1900 goto out;
1901
1902 prev_node = node->addr;
1903 }
1904 }
1905 done = 1;
1906 out:
1907 rcu_read_unlock();
1908
1909 cb->args[0] = prev_node;
1910 cb->args[1] = prev_link;
1911 cb->args[2] = done;
1912
1913 return skb->len;
1914 }
(deprecated) Btrfs devel tree