]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - net/tipc/node.c
x86/boot: Simplify EBDA-vs-BIOS reservation logic
[mirror_ubuntu-artful-kernel.git] / net / tipc / node.c
1 /*
2 * net/tipc/node.c: TIPC node management routines
3 *
4 * Copyright (c) 2000-2006, 2012-2016, 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 #include "netlink.h"
45
46 #define INVALID_NODE_SIG 0x10000
47
48 /* Flags used to take different actions according to flag type
49 * TIPC_NOTIFY_NODE_DOWN: notify node is down
50 * TIPC_NOTIFY_NODE_UP: notify node is up
51 * TIPC_DISTRIBUTE_NAME: publish or withdraw link state name type
52 */
53 enum {
54 TIPC_NOTIFY_NODE_DOWN = (1 << 3),
55 TIPC_NOTIFY_NODE_UP = (1 << 4),
56 TIPC_NOTIFY_LINK_UP = (1 << 6),
57 TIPC_NOTIFY_LINK_DOWN = (1 << 7)
58 };
59
60 struct tipc_link_entry {
61 struct tipc_link *link;
62 spinlock_t lock; /* per link */
63 u32 mtu;
64 struct sk_buff_head inputq;
65 struct tipc_media_addr maddr;
66 };
67
68 struct tipc_bclink_entry {
69 struct tipc_link *link;
70 struct sk_buff_head inputq1;
71 struct sk_buff_head arrvq;
72 struct sk_buff_head inputq2;
73 struct sk_buff_head namedq;
74 };
75
76 /**
77 * struct tipc_node - TIPC node structure
78 * @addr: network address of node
79 * @ref: reference counter to node object
80 * @lock: rwlock governing access to structure
81 * @net: the applicable net namespace
82 * @hash: links to adjacent nodes in unsorted hash chain
83 * @inputq: pointer to input queue containing messages for msg event
84 * @namedq: pointer to name table input queue with name table messages
85 * @active_links: bearer ids of active links, used as index into links[] array
86 * @links: array containing references to all links to node
87 * @action_flags: bit mask of different types of node actions
88 * @state: connectivity state vs peer node
89 * @sync_point: sequence number where synch/failover is finished
90 * @list: links to adjacent nodes in sorted list of cluster's nodes
91 * @working_links: number of working links to node (both active and standby)
92 * @link_cnt: number of links to node
93 * @capabilities: bitmap, indicating peer node's functional capabilities
94 * @signature: node instance identifier
95 * @link_id: local and remote bearer ids of changing link, if any
96 * @publ_list: list of publications
97 * @rcu: rcu struct for tipc_node
98 */
99 struct tipc_node {
100 u32 addr;
101 struct kref kref;
102 rwlock_t lock;
103 struct net *net;
104 struct hlist_node hash;
105 int active_links[2];
106 struct tipc_link_entry links[MAX_BEARERS];
107 struct tipc_bclink_entry bc_entry;
108 int action_flags;
109 struct list_head list;
110 int state;
111 u16 sync_point;
112 int link_cnt;
113 u16 working_links;
114 u16 capabilities;
115 u32 signature;
116 u32 link_id;
117 struct list_head publ_list;
118 struct list_head conn_sks;
119 unsigned long keepalive_intv;
120 struct timer_list timer;
121 struct rcu_head rcu;
122 };
123
124 /* Node FSM states and events:
125 */
126 enum {
127 SELF_DOWN_PEER_DOWN = 0xdd,
128 SELF_UP_PEER_UP = 0xaa,
129 SELF_DOWN_PEER_LEAVING = 0xd1,
130 SELF_UP_PEER_COMING = 0xac,
131 SELF_COMING_PEER_UP = 0xca,
132 SELF_LEAVING_PEER_DOWN = 0x1d,
133 NODE_FAILINGOVER = 0xf0,
134 NODE_SYNCHING = 0xcc
135 };
136
137 enum {
138 SELF_ESTABL_CONTACT_EVT = 0xece,
139 SELF_LOST_CONTACT_EVT = 0x1ce,
140 PEER_ESTABL_CONTACT_EVT = 0x9ece,
141 PEER_LOST_CONTACT_EVT = 0x91ce,
142 NODE_FAILOVER_BEGIN_EVT = 0xfbe,
143 NODE_FAILOVER_END_EVT = 0xfee,
144 NODE_SYNCH_BEGIN_EVT = 0xcbe,
145 NODE_SYNCH_END_EVT = 0xcee
146 };
147
148 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
149 struct sk_buff_head *xmitq,
150 struct tipc_media_addr **maddr);
151 static void tipc_node_link_down(struct tipc_node *n, int bearer_id,
152 bool delete);
153 static void node_lost_contact(struct tipc_node *n, struct sk_buff_head *inputq);
154 static void tipc_node_delete(struct tipc_node *node);
155 static void tipc_node_timeout(unsigned long data);
156 static void tipc_node_fsm_evt(struct tipc_node *n, int evt);
157 static struct tipc_node *tipc_node_find(struct net *net, u32 addr);
158 static void tipc_node_put(struct tipc_node *node);
159 static bool tipc_node_is_up(struct tipc_node *n);
160
161 struct tipc_sock_conn {
162 u32 port;
163 u32 peer_port;
164 u32 peer_node;
165 struct list_head list;
166 };
167
168 static struct tipc_link *node_active_link(struct tipc_node *n, int sel)
169 {
170 int bearer_id = n->active_links[sel & 1];
171
172 if (unlikely(bearer_id == INVALID_BEARER_ID))
173 return NULL;
174
175 return n->links[bearer_id].link;
176 }
177
178 int tipc_node_get_mtu(struct net *net, u32 addr, u32 sel)
179 {
180 struct tipc_node *n;
181 int bearer_id;
182 unsigned int mtu = MAX_MSG_SIZE;
183
184 n = tipc_node_find(net, addr);
185 if (unlikely(!n))
186 return mtu;
187
188 bearer_id = n->active_links[sel & 1];
189 if (likely(bearer_id != INVALID_BEARER_ID))
190 mtu = n->links[bearer_id].mtu;
191 tipc_node_put(n);
192 return mtu;
193 }
194
195 u16 tipc_node_get_capabilities(struct net *net, u32 addr)
196 {
197 struct tipc_node *n;
198 u16 caps;
199
200 n = tipc_node_find(net, addr);
201 if (unlikely(!n))
202 return TIPC_NODE_CAPABILITIES;
203 caps = n->capabilities;
204 tipc_node_put(n);
205 return caps;
206 }
207
208 /*
209 * A trivial power-of-two bitmask technique is used for speed, since this
210 * operation is done for every incoming TIPC packet. The number of hash table
211 * entries has been chosen so that no hash chain exceeds 8 nodes and will
212 * usually be much smaller (typically only a single node).
213 */
214 static unsigned int tipc_hashfn(u32 addr)
215 {
216 return addr & (NODE_HTABLE_SIZE - 1);
217 }
218
219 static void tipc_node_kref_release(struct kref *kref)
220 {
221 struct tipc_node *n = container_of(kref, struct tipc_node, kref);
222
223 kfree(n->bc_entry.link);
224 kfree_rcu(n, rcu);
225 }
226
227 static void tipc_node_put(struct tipc_node *node)
228 {
229 kref_put(&node->kref, tipc_node_kref_release);
230 }
231
232 static void tipc_node_get(struct tipc_node *node)
233 {
234 kref_get(&node->kref);
235 }
236
237 /*
238 * tipc_node_find - locate specified node object, if it exists
239 */
240 static struct tipc_node *tipc_node_find(struct net *net, u32 addr)
241 {
242 struct tipc_net *tn = tipc_net(net);
243 struct tipc_node *node;
244 unsigned int thash = tipc_hashfn(addr);
245
246 if (unlikely(!in_own_cluster_exact(net, addr)))
247 return NULL;
248
249 rcu_read_lock();
250 hlist_for_each_entry_rcu(node, &tn->node_htable[thash], hash) {
251 if (node->addr != addr)
252 continue;
253 if (!kref_get_unless_zero(&node->kref))
254 node = NULL;
255 break;
256 }
257 rcu_read_unlock();
258 return node;
259 }
260
261 static void tipc_node_read_lock(struct tipc_node *n)
262 {
263 read_lock_bh(&n->lock);
264 }
265
266 static void tipc_node_read_unlock(struct tipc_node *n)
267 {
268 read_unlock_bh(&n->lock);
269 }
270
271 static void tipc_node_write_lock(struct tipc_node *n)
272 {
273 write_lock_bh(&n->lock);
274 }
275
276 static void tipc_node_write_unlock(struct tipc_node *n)
277 {
278 struct net *net = n->net;
279 u32 addr = 0;
280 u32 flags = n->action_flags;
281 u32 link_id = 0;
282 struct list_head *publ_list;
283
284 if (likely(!flags)) {
285 write_unlock_bh(&n->lock);
286 return;
287 }
288
289 addr = n->addr;
290 link_id = n->link_id;
291 publ_list = &n->publ_list;
292
293 n->action_flags &= ~(TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
294 TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP);
295
296 write_unlock_bh(&n->lock);
297
298 if (flags & TIPC_NOTIFY_NODE_DOWN)
299 tipc_publ_notify(net, publ_list, addr);
300
301 if (flags & TIPC_NOTIFY_NODE_UP)
302 tipc_named_node_up(net, addr);
303
304 if (flags & TIPC_NOTIFY_LINK_UP)
305 tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
306 TIPC_NODE_SCOPE, link_id, addr);
307
308 if (flags & TIPC_NOTIFY_LINK_DOWN)
309 tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
310 link_id, addr);
311 }
312
313 struct tipc_node *tipc_node_create(struct net *net, u32 addr, u16 capabilities)
314 {
315 struct tipc_net *tn = net_generic(net, tipc_net_id);
316 struct tipc_node *n, *temp_node;
317 int i;
318
319 spin_lock_bh(&tn->node_list_lock);
320 n = tipc_node_find(net, addr);
321 if (n) {
322 /* Same node may come back with new capabilities */
323 n->capabilities = capabilities;
324 goto exit;
325 }
326 n = kzalloc(sizeof(*n), GFP_ATOMIC);
327 if (!n) {
328 pr_warn("Node creation failed, no memory\n");
329 goto exit;
330 }
331 n->addr = addr;
332 n->net = net;
333 n->capabilities = capabilities;
334 kref_init(&n->kref);
335 rwlock_init(&n->lock);
336 INIT_HLIST_NODE(&n->hash);
337 INIT_LIST_HEAD(&n->list);
338 INIT_LIST_HEAD(&n->publ_list);
339 INIT_LIST_HEAD(&n->conn_sks);
340 skb_queue_head_init(&n->bc_entry.namedq);
341 skb_queue_head_init(&n->bc_entry.inputq1);
342 __skb_queue_head_init(&n->bc_entry.arrvq);
343 skb_queue_head_init(&n->bc_entry.inputq2);
344 for (i = 0; i < MAX_BEARERS; i++)
345 spin_lock_init(&n->links[i].lock);
346 n->state = SELF_DOWN_PEER_LEAVING;
347 n->signature = INVALID_NODE_SIG;
348 n->active_links[0] = INVALID_BEARER_ID;
349 n->active_links[1] = INVALID_BEARER_ID;
350 if (!tipc_link_bc_create(net, tipc_own_addr(net), n->addr,
351 U16_MAX,
352 tipc_link_window(tipc_bc_sndlink(net)),
353 n->capabilities,
354 &n->bc_entry.inputq1,
355 &n->bc_entry.namedq,
356 tipc_bc_sndlink(net),
357 &n->bc_entry.link)) {
358 pr_warn("Broadcast rcv link creation failed, no memory\n");
359 kfree(n);
360 n = NULL;
361 goto exit;
362 }
363 tipc_node_get(n);
364 setup_timer(&n->timer, tipc_node_timeout, (unsigned long)n);
365 n->keepalive_intv = U32_MAX;
366 hlist_add_head_rcu(&n->hash, &tn->node_htable[tipc_hashfn(addr)]);
367 list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
368 if (n->addr < temp_node->addr)
369 break;
370 }
371 list_add_tail_rcu(&n->list, &temp_node->list);
372 exit:
373 spin_unlock_bh(&tn->node_list_lock);
374 return n;
375 }
376
377 static void tipc_node_calculate_timer(struct tipc_node *n, struct tipc_link *l)
378 {
379 unsigned long tol = tipc_link_tolerance(l);
380 unsigned long intv = ((tol / 4) > 500) ? 500 : tol / 4;
381 unsigned long keepalive_intv = msecs_to_jiffies(intv);
382
383 /* Link with lowest tolerance determines timer interval */
384 if (keepalive_intv < n->keepalive_intv)
385 n->keepalive_intv = keepalive_intv;
386
387 /* Ensure link's abort limit corresponds to current interval */
388 tipc_link_set_abort_limit(l, tol / jiffies_to_msecs(n->keepalive_intv));
389 }
390
391 static void tipc_node_delete(struct tipc_node *node)
392 {
393 list_del_rcu(&node->list);
394 hlist_del_rcu(&node->hash);
395 tipc_node_put(node);
396
397 del_timer_sync(&node->timer);
398 tipc_node_put(node);
399 }
400
401 void tipc_node_stop(struct net *net)
402 {
403 struct tipc_net *tn = tipc_net(net);
404 struct tipc_node *node, *t_node;
405
406 spin_lock_bh(&tn->node_list_lock);
407 list_for_each_entry_safe(node, t_node, &tn->node_list, list)
408 tipc_node_delete(node);
409 spin_unlock_bh(&tn->node_list_lock);
410 }
411
412 void tipc_node_subscribe(struct net *net, struct list_head *subscr, u32 addr)
413 {
414 struct tipc_node *n;
415
416 if (in_own_node(net, addr))
417 return;
418
419 n = tipc_node_find(net, addr);
420 if (!n) {
421 pr_warn("Node subscribe rejected, unknown node 0x%x\n", addr);
422 return;
423 }
424 tipc_node_write_lock(n);
425 list_add_tail(subscr, &n->publ_list);
426 tipc_node_write_unlock(n);
427 tipc_node_put(n);
428 }
429
430 void tipc_node_unsubscribe(struct net *net, struct list_head *subscr, u32 addr)
431 {
432 struct tipc_node *n;
433
434 if (in_own_node(net, addr))
435 return;
436
437 n = tipc_node_find(net, addr);
438 if (!n) {
439 pr_warn("Node unsubscribe rejected, unknown node 0x%x\n", addr);
440 return;
441 }
442 tipc_node_write_lock(n);
443 list_del_init(subscr);
444 tipc_node_write_unlock(n);
445 tipc_node_put(n);
446 }
447
448 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
449 {
450 struct tipc_node *node;
451 struct tipc_sock_conn *conn;
452 int err = 0;
453
454 if (in_own_node(net, dnode))
455 return 0;
456
457 node = tipc_node_find(net, dnode);
458 if (!node) {
459 pr_warn("Connecting sock to node 0x%x failed\n", dnode);
460 return -EHOSTUNREACH;
461 }
462 conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
463 if (!conn) {
464 err = -EHOSTUNREACH;
465 goto exit;
466 }
467 conn->peer_node = dnode;
468 conn->port = port;
469 conn->peer_port = peer_port;
470
471 tipc_node_write_lock(node);
472 list_add_tail(&conn->list, &node->conn_sks);
473 tipc_node_write_unlock(node);
474 exit:
475 tipc_node_put(node);
476 return err;
477 }
478
479 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
480 {
481 struct tipc_node *node;
482 struct tipc_sock_conn *conn, *safe;
483
484 if (in_own_node(net, dnode))
485 return;
486
487 node = tipc_node_find(net, dnode);
488 if (!node)
489 return;
490
491 tipc_node_write_lock(node);
492 list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
493 if (port != conn->port)
494 continue;
495 list_del(&conn->list);
496 kfree(conn);
497 }
498 tipc_node_write_unlock(node);
499 tipc_node_put(node);
500 }
501
502 /* tipc_node_timeout - handle expiration of node timer
503 */
504 static void tipc_node_timeout(unsigned long data)
505 {
506 struct tipc_node *n = (struct tipc_node *)data;
507 struct tipc_link_entry *le;
508 struct sk_buff_head xmitq;
509 int bearer_id;
510 int rc = 0;
511
512 __skb_queue_head_init(&xmitq);
513
514 for (bearer_id = 0; bearer_id < MAX_BEARERS; bearer_id++) {
515 tipc_node_read_lock(n);
516 le = &n->links[bearer_id];
517 spin_lock_bh(&le->lock);
518 if (le->link) {
519 /* Link tolerance may change asynchronously: */
520 tipc_node_calculate_timer(n, le->link);
521 rc = tipc_link_timeout(le->link, &xmitq);
522 }
523 spin_unlock_bh(&le->lock);
524 tipc_node_read_unlock(n);
525 tipc_bearer_xmit(n->net, bearer_id, &xmitq, &le->maddr);
526 if (rc & TIPC_LINK_DOWN_EVT)
527 tipc_node_link_down(n, bearer_id, false);
528 }
529 mod_timer(&n->timer, jiffies + n->keepalive_intv);
530 }
531
532 /**
533 * __tipc_node_link_up - handle addition of link
534 * Node lock must be held by caller
535 * Link becomes active (alone or shared) or standby, depending on its priority.
536 */
537 static void __tipc_node_link_up(struct tipc_node *n, int bearer_id,
538 struct sk_buff_head *xmitq)
539 {
540 int *slot0 = &n->active_links[0];
541 int *slot1 = &n->active_links[1];
542 struct tipc_link *ol = node_active_link(n, 0);
543 struct tipc_link *nl = n->links[bearer_id].link;
544
545 if (!nl || tipc_link_is_up(nl))
546 return;
547
548 tipc_link_fsm_evt(nl, LINK_ESTABLISH_EVT);
549 if (!tipc_link_is_up(nl))
550 return;
551
552 n->working_links++;
553 n->action_flags |= TIPC_NOTIFY_LINK_UP;
554 n->link_id = tipc_link_id(nl);
555
556 /* Leave room for tunnel header when returning 'mtu' to users: */
557 n->links[bearer_id].mtu = tipc_link_mtu(nl) - INT_H_SIZE;
558
559 tipc_bearer_add_dest(n->net, bearer_id, n->addr);
560 tipc_bcast_inc_bearer_dst_cnt(n->net, bearer_id);
561
562 pr_debug("Established link <%s> on network plane %c\n",
563 tipc_link_name(nl), tipc_link_plane(nl));
564
565 /* Ensure that a STATE message goes first */
566 tipc_link_build_state_msg(nl, xmitq);
567
568 /* First link? => give it both slots */
569 if (!ol) {
570 *slot0 = bearer_id;
571 *slot1 = bearer_id;
572 tipc_node_fsm_evt(n, SELF_ESTABL_CONTACT_EVT);
573 n->action_flags |= TIPC_NOTIFY_NODE_UP;
574 tipc_link_set_active(nl, true);
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 struct tipc_media_addr *maddr;
606
607 tipc_node_write_lock(n);
608 __tipc_node_link_up(n, bearer_id, xmitq);
609 maddr = &n->links[bearer_id].maddr;
610 tipc_bearer_xmit(n->net, bearer_id, xmitq, maddr);
611 tipc_node_write_unlock(n);
612 }
613
614 /**
615 * __tipc_node_link_down - handle loss of link
616 */
617 static void __tipc_node_link_down(struct tipc_node *n, int *bearer_id,
618 struct sk_buff_head *xmitq,
619 struct tipc_media_addr **maddr)
620 {
621 struct tipc_link_entry *le = &n->links[*bearer_id];
622 int *slot0 = &n->active_links[0];
623 int *slot1 = &n->active_links[1];
624 int i, highest = 0, prio;
625 struct tipc_link *l, *_l, *tnl;
626
627 l = n->links[*bearer_id].link;
628 if (!l || tipc_link_is_reset(l))
629 return;
630
631 n->working_links--;
632 n->action_flags |= TIPC_NOTIFY_LINK_DOWN;
633 n->link_id = tipc_link_id(l);
634
635 tipc_bearer_remove_dest(n->net, *bearer_id, n->addr);
636
637 pr_debug("Lost link <%s> on network plane %c\n",
638 tipc_link_name(l), tipc_link_plane(l));
639
640 /* Select new active link if any available */
641 *slot0 = INVALID_BEARER_ID;
642 *slot1 = INVALID_BEARER_ID;
643 for (i = 0; i < MAX_BEARERS; i++) {
644 _l = n->links[i].link;
645 if (!_l || !tipc_link_is_up(_l))
646 continue;
647 if (_l == l)
648 continue;
649 prio = tipc_link_prio(_l);
650 if (prio < highest)
651 continue;
652 if (prio > highest) {
653 highest = prio;
654 *slot0 = i;
655 *slot1 = i;
656 continue;
657 }
658 *slot1 = i;
659 }
660
661 if (!tipc_node_is_up(n)) {
662 if (tipc_link_peer_is_down(l))
663 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
664 tipc_node_fsm_evt(n, SELF_LOST_CONTACT_EVT);
665 tipc_link_fsm_evt(l, LINK_RESET_EVT);
666 tipc_link_reset(l);
667 tipc_link_build_reset_msg(l, xmitq);
668 *maddr = &n->links[*bearer_id].maddr;
669 node_lost_contact(n, &le->inputq);
670 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
671 return;
672 }
673 tipc_bcast_dec_bearer_dst_cnt(n->net, *bearer_id);
674
675 /* There is still a working link => initiate failover */
676 *bearer_id = n->active_links[0];
677 tnl = n->links[*bearer_id].link;
678 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
679 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
680 n->sync_point = tipc_link_rcv_nxt(tnl) + (U16_MAX / 2 - 1);
681 tipc_link_tnl_prepare(l, tnl, FAILOVER_MSG, xmitq);
682 tipc_link_reset(l);
683 tipc_link_fsm_evt(l, LINK_RESET_EVT);
684 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
685 tipc_node_fsm_evt(n, NODE_FAILOVER_BEGIN_EVT);
686 *maddr = &n->links[*bearer_id].maddr;
687 }
688
689 static void tipc_node_link_down(struct tipc_node *n, int bearer_id, bool delete)
690 {
691 struct tipc_link_entry *le = &n->links[bearer_id];
692 struct tipc_link *l = le->link;
693 struct tipc_media_addr *maddr;
694 struct sk_buff_head xmitq;
695
696 if (!l)
697 return;
698
699 __skb_queue_head_init(&xmitq);
700
701 tipc_node_write_lock(n);
702 if (!tipc_link_is_establishing(l)) {
703 __tipc_node_link_down(n, &bearer_id, &xmitq, &maddr);
704 if (delete) {
705 kfree(l);
706 le->link = NULL;
707 n->link_cnt--;
708 }
709 } else {
710 /* Defuse pending tipc_node_link_up() */
711 tipc_link_fsm_evt(l, LINK_RESET_EVT);
712 }
713 tipc_node_write_unlock(n);
714 tipc_bearer_xmit(n->net, bearer_id, &xmitq, maddr);
715 tipc_sk_rcv(n->net, &le->inputq);
716 }
717
718 static bool tipc_node_is_up(struct tipc_node *n)
719 {
720 return n->active_links[0] != INVALID_BEARER_ID;
721 }
722
723 void tipc_node_check_dest(struct net *net, u32 onode,
724 struct tipc_bearer *b,
725 u16 capabilities, u32 signature,
726 struct tipc_media_addr *maddr,
727 bool *respond, bool *dupl_addr)
728 {
729 struct tipc_node *n;
730 struct tipc_link *l;
731 struct tipc_link_entry *le;
732 bool addr_match = false;
733 bool sign_match = false;
734 bool link_up = false;
735 bool accept_addr = false;
736 bool reset = true;
737 char *if_name;
738
739 *dupl_addr = false;
740 *respond = false;
741
742 n = tipc_node_create(net, onode, capabilities);
743 if (!n)
744 return;
745
746 tipc_node_write_lock(n);
747
748 le = &n->links[b->identity];
749
750 /* Prepare to validate requesting node's signature and media address */
751 l = le->link;
752 link_up = l && tipc_link_is_up(l);
753 addr_match = l && !memcmp(&le->maddr, maddr, sizeof(*maddr));
754 sign_match = (signature == n->signature);
755
756 /* These three flags give us eight permutations: */
757
758 if (sign_match && addr_match && link_up) {
759 /* All is fine. Do nothing. */
760 reset = false;
761 } else if (sign_match && addr_match && !link_up) {
762 /* Respond. The link will come up in due time */
763 *respond = true;
764 } else if (sign_match && !addr_match && link_up) {
765 /* Peer has changed i/f address without rebooting.
766 * If so, the link will reset soon, and the next
767 * discovery will be accepted. So we can ignore it.
768 * It may also be an cloned or malicious peer having
769 * chosen the same node address and signature as an
770 * existing one.
771 * Ignore requests until the link goes down, if ever.
772 */
773 *dupl_addr = true;
774 } else if (sign_match && !addr_match && !link_up) {
775 /* Peer link has changed i/f address without rebooting.
776 * It may also be a cloned or malicious peer; we can't
777 * distinguish between the two.
778 * The signature is correct, so we must accept.
779 */
780 accept_addr = true;
781 *respond = true;
782 } else if (!sign_match && addr_match && link_up) {
783 /* Peer node rebooted. Two possibilities:
784 * - Delayed re-discovery; this link endpoint has already
785 * reset and re-established contact with the peer, before
786 * receiving a discovery message from that node.
787 * (The peer happened to receive one from this node first).
788 * - The peer came back so fast that our side has not
789 * discovered it yet. Probing from this side will soon
790 * reset the link, since there can be no working link
791 * endpoint at the peer end, and the link will re-establish.
792 * Accept the signature, since it comes from a known peer.
793 */
794 n->signature = signature;
795 } else if (!sign_match && addr_match && !link_up) {
796 /* The peer node has rebooted.
797 * Accept signature, since it is a known peer.
798 */
799 n->signature = signature;
800 *respond = true;
801 } else if (!sign_match && !addr_match && link_up) {
802 /* Peer rebooted with new address, or a new/duplicate peer.
803 * Ignore until the link goes down, if ever.
804 */
805 *dupl_addr = true;
806 } else if (!sign_match && !addr_match && !link_up) {
807 /* Peer rebooted with new address, or it is a new peer.
808 * Accept signature and address.
809 */
810 n->signature = signature;
811 accept_addr = true;
812 *respond = true;
813 }
814
815 if (!accept_addr)
816 goto exit;
817
818 /* Now create new link if not already existing */
819 if (!l) {
820 if (n->link_cnt == 2) {
821 pr_warn("Cannot establish 3rd link to %x\n", n->addr);
822 goto exit;
823 }
824 if_name = strchr(b->name, ':') + 1;
825 if (!tipc_link_create(net, if_name, b->identity, b->tolerance,
826 b->net_plane, b->mtu, b->priority,
827 b->window, mod(tipc_net(net)->random),
828 tipc_own_addr(net), onode,
829 n->capabilities,
830 tipc_bc_sndlink(n->net), n->bc_entry.link,
831 &le->inputq,
832 &n->bc_entry.namedq, &l)) {
833 *respond = false;
834 goto exit;
835 }
836 tipc_link_reset(l);
837 tipc_link_fsm_evt(l, LINK_RESET_EVT);
838 if (n->state == NODE_FAILINGOVER)
839 tipc_link_fsm_evt(l, LINK_FAILOVER_BEGIN_EVT);
840 le->link = l;
841 n->link_cnt++;
842 tipc_node_calculate_timer(n, l);
843 if (n->link_cnt == 1)
844 if (!mod_timer(&n->timer, jiffies + n->keepalive_intv))
845 tipc_node_get(n);
846 }
847 memcpy(&le->maddr, maddr, sizeof(*maddr));
848 exit:
849 tipc_node_write_unlock(n);
850 if (reset && l && !tipc_link_is_reset(l))
851 tipc_node_link_down(n, b->identity, false);
852 tipc_node_put(n);
853 }
854
855 void tipc_node_delete_links(struct net *net, int bearer_id)
856 {
857 struct tipc_net *tn = net_generic(net, tipc_net_id);
858 struct tipc_node *n;
859
860 rcu_read_lock();
861 list_for_each_entry_rcu(n, &tn->node_list, list) {
862 tipc_node_link_down(n, bearer_id, true);
863 }
864 rcu_read_unlock();
865 }
866
867 static void tipc_node_reset_links(struct tipc_node *n)
868 {
869 char addr_string[16];
870 int i;
871
872 pr_warn("Resetting all links to %s\n",
873 tipc_addr_string_fill(addr_string, n->addr));
874
875 for (i = 0; i < MAX_BEARERS; i++) {
876 tipc_node_link_down(n, i, false);
877 }
878 }
879
880 /* tipc_node_fsm_evt - node finite state machine
881 * Determines when contact is allowed with peer node
882 */
883 static void tipc_node_fsm_evt(struct tipc_node *n, int evt)
884 {
885 int state = n->state;
886
887 switch (state) {
888 case SELF_DOWN_PEER_DOWN:
889 switch (evt) {
890 case SELF_ESTABL_CONTACT_EVT:
891 state = SELF_UP_PEER_COMING;
892 break;
893 case PEER_ESTABL_CONTACT_EVT:
894 state = SELF_COMING_PEER_UP;
895 break;
896 case SELF_LOST_CONTACT_EVT:
897 case PEER_LOST_CONTACT_EVT:
898 break;
899 case NODE_SYNCH_END_EVT:
900 case NODE_SYNCH_BEGIN_EVT:
901 case NODE_FAILOVER_BEGIN_EVT:
902 case NODE_FAILOVER_END_EVT:
903 default:
904 goto illegal_evt;
905 }
906 break;
907 case SELF_UP_PEER_UP:
908 switch (evt) {
909 case SELF_LOST_CONTACT_EVT:
910 state = SELF_DOWN_PEER_LEAVING;
911 break;
912 case PEER_LOST_CONTACT_EVT:
913 state = SELF_LEAVING_PEER_DOWN;
914 break;
915 case NODE_SYNCH_BEGIN_EVT:
916 state = NODE_SYNCHING;
917 break;
918 case NODE_FAILOVER_BEGIN_EVT:
919 state = NODE_FAILINGOVER;
920 break;
921 case SELF_ESTABL_CONTACT_EVT:
922 case PEER_ESTABL_CONTACT_EVT:
923 case NODE_SYNCH_END_EVT:
924 case NODE_FAILOVER_END_EVT:
925 break;
926 default:
927 goto illegal_evt;
928 }
929 break;
930 case SELF_DOWN_PEER_LEAVING:
931 switch (evt) {
932 case PEER_LOST_CONTACT_EVT:
933 state = SELF_DOWN_PEER_DOWN;
934 break;
935 case SELF_ESTABL_CONTACT_EVT:
936 case PEER_ESTABL_CONTACT_EVT:
937 case SELF_LOST_CONTACT_EVT:
938 break;
939 case NODE_SYNCH_END_EVT:
940 case NODE_SYNCH_BEGIN_EVT:
941 case NODE_FAILOVER_BEGIN_EVT:
942 case NODE_FAILOVER_END_EVT:
943 default:
944 goto illegal_evt;
945 }
946 break;
947 case SELF_UP_PEER_COMING:
948 switch (evt) {
949 case PEER_ESTABL_CONTACT_EVT:
950 state = SELF_UP_PEER_UP;
951 break;
952 case SELF_LOST_CONTACT_EVT:
953 state = SELF_DOWN_PEER_LEAVING;
954 break;
955 case SELF_ESTABL_CONTACT_EVT:
956 case PEER_LOST_CONTACT_EVT:
957 case NODE_SYNCH_END_EVT:
958 case NODE_FAILOVER_BEGIN_EVT:
959 break;
960 case NODE_SYNCH_BEGIN_EVT:
961 case NODE_FAILOVER_END_EVT:
962 default:
963 goto illegal_evt;
964 }
965 break;
966 case SELF_COMING_PEER_UP:
967 switch (evt) {
968 case SELF_ESTABL_CONTACT_EVT:
969 state = SELF_UP_PEER_UP;
970 break;
971 case PEER_LOST_CONTACT_EVT:
972 state = SELF_LEAVING_PEER_DOWN;
973 break;
974 case SELF_LOST_CONTACT_EVT:
975 case PEER_ESTABL_CONTACT_EVT:
976 break;
977 case NODE_SYNCH_END_EVT:
978 case NODE_SYNCH_BEGIN_EVT:
979 case NODE_FAILOVER_BEGIN_EVT:
980 case NODE_FAILOVER_END_EVT:
981 default:
982 goto illegal_evt;
983 }
984 break;
985 case SELF_LEAVING_PEER_DOWN:
986 switch (evt) {
987 case SELF_LOST_CONTACT_EVT:
988 state = SELF_DOWN_PEER_DOWN;
989 break;
990 case SELF_ESTABL_CONTACT_EVT:
991 case PEER_ESTABL_CONTACT_EVT:
992 case PEER_LOST_CONTACT_EVT:
993 break;
994 case NODE_SYNCH_END_EVT:
995 case NODE_SYNCH_BEGIN_EVT:
996 case NODE_FAILOVER_BEGIN_EVT:
997 case NODE_FAILOVER_END_EVT:
998 default:
999 goto illegal_evt;
1000 }
1001 break;
1002 case NODE_FAILINGOVER:
1003 switch (evt) {
1004 case SELF_LOST_CONTACT_EVT:
1005 state = SELF_DOWN_PEER_LEAVING;
1006 break;
1007 case PEER_LOST_CONTACT_EVT:
1008 state = SELF_LEAVING_PEER_DOWN;
1009 break;
1010 case NODE_FAILOVER_END_EVT:
1011 state = SELF_UP_PEER_UP;
1012 break;
1013 case NODE_FAILOVER_BEGIN_EVT:
1014 case SELF_ESTABL_CONTACT_EVT:
1015 case PEER_ESTABL_CONTACT_EVT:
1016 break;
1017 case NODE_SYNCH_BEGIN_EVT:
1018 case NODE_SYNCH_END_EVT:
1019 default:
1020 goto illegal_evt;
1021 }
1022 break;
1023 case NODE_SYNCHING:
1024 switch (evt) {
1025 case SELF_LOST_CONTACT_EVT:
1026 state = SELF_DOWN_PEER_LEAVING;
1027 break;
1028 case PEER_LOST_CONTACT_EVT:
1029 state = SELF_LEAVING_PEER_DOWN;
1030 break;
1031 case NODE_SYNCH_END_EVT:
1032 state = SELF_UP_PEER_UP;
1033 break;
1034 case NODE_FAILOVER_BEGIN_EVT:
1035 state = NODE_FAILINGOVER;
1036 break;
1037 case NODE_SYNCH_BEGIN_EVT:
1038 case SELF_ESTABL_CONTACT_EVT:
1039 case PEER_ESTABL_CONTACT_EVT:
1040 break;
1041 case NODE_FAILOVER_END_EVT:
1042 default:
1043 goto illegal_evt;
1044 }
1045 break;
1046 default:
1047 pr_err("Unknown node fsm state %x\n", state);
1048 break;
1049 }
1050 n->state = state;
1051 return;
1052
1053 illegal_evt:
1054 pr_err("Illegal node fsm evt %x in state %x\n", evt, state);
1055 }
1056
1057 static void node_lost_contact(struct tipc_node *n,
1058 struct sk_buff_head *inputq)
1059 {
1060 char addr_string[16];
1061 struct tipc_sock_conn *conn, *safe;
1062 struct tipc_link *l;
1063 struct list_head *conns = &n->conn_sks;
1064 struct sk_buff *skb;
1065 uint i;
1066
1067 pr_debug("Lost contact with %s\n",
1068 tipc_addr_string_fill(addr_string, n->addr));
1069
1070 /* Clean up broadcast state */
1071 tipc_bcast_remove_peer(n->net, n->bc_entry.link);
1072
1073 /* Abort any ongoing link failover */
1074 for (i = 0; i < MAX_BEARERS; i++) {
1075 l = n->links[i].link;
1076 if (l)
1077 tipc_link_fsm_evt(l, LINK_FAILOVER_END_EVT);
1078 }
1079
1080 /* Notify publications from this node */
1081 n->action_flags |= TIPC_NOTIFY_NODE_DOWN;
1082
1083 /* Notify sockets connected to node */
1084 list_for_each_entry_safe(conn, safe, conns, list) {
1085 skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
1086 SHORT_H_SIZE, 0, tipc_own_addr(n->net),
1087 conn->peer_node, conn->port,
1088 conn->peer_port, TIPC_ERR_NO_NODE);
1089 if (likely(skb))
1090 skb_queue_tail(inputq, skb);
1091 list_del(&conn->list);
1092 kfree(conn);
1093 }
1094 }
1095
1096 /**
1097 * tipc_node_get_linkname - get the name of a link
1098 *
1099 * @bearer_id: id of the bearer
1100 * @node: peer node address
1101 * @linkname: link name output buffer
1102 *
1103 * Returns 0 on success
1104 */
1105 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
1106 char *linkname, size_t len)
1107 {
1108 struct tipc_link *link;
1109 int err = -EINVAL;
1110 struct tipc_node *node = tipc_node_find(net, addr);
1111
1112 if (!node)
1113 return err;
1114
1115 if (bearer_id >= MAX_BEARERS)
1116 goto exit;
1117
1118 tipc_node_read_lock(node);
1119 link = node->links[bearer_id].link;
1120 if (link) {
1121 strncpy(linkname, tipc_link_name(link), len);
1122 err = 0;
1123 }
1124 exit:
1125 tipc_node_read_unlock(node);
1126 tipc_node_put(node);
1127 return err;
1128 }
1129
1130 /* Caller should hold node lock for the passed node */
1131 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
1132 {
1133 void *hdr;
1134 struct nlattr *attrs;
1135
1136 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
1137 NLM_F_MULTI, TIPC_NL_NODE_GET);
1138 if (!hdr)
1139 return -EMSGSIZE;
1140
1141 attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
1142 if (!attrs)
1143 goto msg_full;
1144
1145 if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
1146 goto attr_msg_full;
1147 if (tipc_node_is_up(node))
1148 if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
1149 goto attr_msg_full;
1150
1151 nla_nest_end(msg->skb, attrs);
1152 genlmsg_end(msg->skb, hdr);
1153
1154 return 0;
1155
1156 attr_msg_full:
1157 nla_nest_cancel(msg->skb, attrs);
1158 msg_full:
1159 genlmsg_cancel(msg->skb, hdr);
1160
1161 return -EMSGSIZE;
1162 }
1163
1164 /**
1165 * tipc_node_xmit() is the general link level function for message sending
1166 * @net: the applicable net namespace
1167 * @list: chain of buffers containing message
1168 * @dnode: address of destination node
1169 * @selector: a number used for deterministic link selection
1170 * Consumes the buffer chain, except when returning -ELINKCONG
1171 * Returns 0 if success, otherwise: -ELINKCONG,-EHOSTUNREACH,-EMSGSIZE,-ENOBUF
1172 */
1173 int tipc_node_xmit(struct net *net, struct sk_buff_head *list,
1174 u32 dnode, int selector)
1175 {
1176 struct tipc_link_entry *le = NULL;
1177 struct tipc_node *n;
1178 struct sk_buff_head xmitq;
1179 int bearer_id;
1180 int rc;
1181
1182 if (in_own_node(net, dnode)) {
1183 tipc_sk_rcv(net, list);
1184 return 0;
1185 }
1186
1187 n = tipc_node_find(net, dnode);
1188 if (unlikely(!n)) {
1189 skb_queue_purge(list);
1190 return -EHOSTUNREACH;
1191 }
1192
1193 tipc_node_read_lock(n);
1194 bearer_id = n->active_links[selector & 1];
1195 if (unlikely(bearer_id == INVALID_BEARER_ID)) {
1196 tipc_node_read_unlock(n);
1197 tipc_node_put(n);
1198 skb_queue_purge(list);
1199 return -EHOSTUNREACH;
1200 }
1201
1202 __skb_queue_head_init(&xmitq);
1203 le = &n->links[bearer_id];
1204 spin_lock_bh(&le->lock);
1205 rc = tipc_link_xmit(le->link, list, &xmitq);
1206 spin_unlock_bh(&le->lock);
1207 tipc_node_read_unlock(n);
1208
1209 if (likely(rc == 0))
1210 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1211 else if (rc == -ENOBUFS)
1212 tipc_node_link_down(n, bearer_id, false);
1213
1214 tipc_node_put(n);
1215
1216 return rc;
1217 }
1218
1219 /* tipc_node_xmit_skb(): send single buffer to destination
1220 * Buffers sent via this functon are generally TIPC_SYSTEM_IMPORTANCE
1221 * messages, which will not be rejected
1222 * The only exception is datagram messages rerouted after secondary
1223 * lookup, which are rare and safe to dispose of anyway.
1224 * TODO: Return real return value, and let callers use
1225 * tipc_wait_for_sendpkt() where applicable
1226 */
1227 int tipc_node_xmit_skb(struct net *net, struct sk_buff *skb, u32 dnode,
1228 u32 selector)
1229 {
1230 struct sk_buff_head head;
1231 int rc;
1232
1233 skb_queue_head_init(&head);
1234 __skb_queue_tail(&head, skb);
1235 rc = tipc_node_xmit(net, &head, dnode, selector);
1236 if (rc == -ELINKCONG)
1237 kfree_skb(skb);
1238 return 0;
1239 }
1240
1241 void tipc_node_broadcast(struct net *net, struct sk_buff *skb)
1242 {
1243 struct sk_buff *txskb;
1244 struct tipc_node *n;
1245 u32 dst;
1246
1247 rcu_read_lock();
1248 list_for_each_entry_rcu(n, tipc_nodes(net), list) {
1249 dst = n->addr;
1250 if (in_own_node(net, dst))
1251 continue;
1252 if (!tipc_node_is_up(n))
1253 continue;
1254 txskb = pskb_copy(skb, GFP_ATOMIC);
1255 if (!txskb)
1256 break;
1257 msg_set_destnode(buf_msg(txskb), dst);
1258 tipc_node_xmit_skb(net, txskb, dst, 0);
1259 }
1260 rcu_read_unlock();
1261
1262 kfree_skb(skb);
1263 }
1264
1265 /**
1266 * tipc_node_bc_rcv - process TIPC broadcast packet arriving from off-node
1267 * @net: the applicable net namespace
1268 * @skb: TIPC packet
1269 * @bearer_id: id of bearer message arrived on
1270 *
1271 * Invoked with no locks held.
1272 */
1273 static void tipc_node_bc_rcv(struct net *net, struct sk_buff *skb, int bearer_id)
1274 {
1275 int rc;
1276 struct sk_buff_head xmitq;
1277 struct tipc_bclink_entry *be;
1278 struct tipc_link_entry *le;
1279 struct tipc_msg *hdr = buf_msg(skb);
1280 int usr = msg_user(hdr);
1281 u32 dnode = msg_destnode(hdr);
1282 struct tipc_node *n;
1283
1284 __skb_queue_head_init(&xmitq);
1285
1286 /* If NACK for other node, let rcv link for that node peek into it */
1287 if ((usr == BCAST_PROTOCOL) && (dnode != tipc_own_addr(net)))
1288 n = tipc_node_find(net, dnode);
1289 else
1290 n = tipc_node_find(net, msg_prevnode(hdr));
1291 if (!n) {
1292 kfree_skb(skb);
1293 return;
1294 }
1295 be = &n->bc_entry;
1296 le = &n->links[bearer_id];
1297
1298 rc = tipc_bcast_rcv(net, be->link, skb);
1299
1300 /* Broadcast link reset may happen at reassembly failure */
1301 if (rc & TIPC_LINK_DOWN_EVT)
1302 tipc_node_reset_links(n);
1303
1304 /* Broadcast ACKs are sent on a unicast link */
1305 if (rc & TIPC_LINK_SND_BC_ACK) {
1306 tipc_node_read_lock(n);
1307 tipc_link_build_state_msg(le->link, &xmitq);
1308 tipc_node_read_unlock(n);
1309 }
1310
1311 if (!skb_queue_empty(&xmitq))
1312 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1313
1314 /* Deliver. 'arrvq' is under inputq2's lock protection */
1315 if (!skb_queue_empty(&be->inputq1)) {
1316 spin_lock_bh(&be->inputq2.lock);
1317 spin_lock_bh(&be->inputq1.lock);
1318 skb_queue_splice_tail_init(&be->inputq1, &be->arrvq);
1319 spin_unlock_bh(&be->inputq1.lock);
1320 spin_unlock_bh(&be->inputq2.lock);
1321 tipc_sk_mcast_rcv(net, &be->arrvq, &be->inputq2);
1322 }
1323 tipc_node_put(n);
1324 }
1325
1326 /**
1327 * tipc_node_check_state - check and if necessary update node state
1328 * @skb: TIPC packet
1329 * @bearer_id: identity of bearer delivering the packet
1330 * Returns true if state is ok, otherwise consumes buffer and returns false
1331 */
1332 static bool tipc_node_check_state(struct tipc_node *n, struct sk_buff *skb,
1333 int bearer_id, struct sk_buff_head *xmitq)
1334 {
1335 struct tipc_msg *hdr = buf_msg(skb);
1336 int usr = msg_user(hdr);
1337 int mtyp = msg_type(hdr);
1338 u16 oseqno = msg_seqno(hdr);
1339 u16 iseqno = msg_seqno(msg_get_wrapped(hdr));
1340 u16 exp_pkts = msg_msgcnt(hdr);
1341 u16 rcv_nxt, syncpt, dlv_nxt, inputq_len;
1342 int state = n->state;
1343 struct tipc_link *l, *tnl, *pl = NULL;
1344 struct tipc_media_addr *maddr;
1345 int pb_id;
1346
1347 l = n->links[bearer_id].link;
1348 if (!l)
1349 return false;
1350 rcv_nxt = tipc_link_rcv_nxt(l);
1351
1352
1353 if (likely((state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL)))
1354 return true;
1355
1356 /* Find parallel link, if any */
1357 for (pb_id = 0; pb_id < MAX_BEARERS; pb_id++) {
1358 if ((pb_id != bearer_id) && n->links[pb_id].link) {
1359 pl = n->links[pb_id].link;
1360 break;
1361 }
1362 }
1363
1364 /* Check and update node accesibility if applicable */
1365 if (state == SELF_UP_PEER_COMING) {
1366 if (!tipc_link_is_up(l))
1367 return true;
1368 if (!msg_peer_link_is_up(hdr))
1369 return true;
1370 tipc_node_fsm_evt(n, PEER_ESTABL_CONTACT_EVT);
1371 }
1372
1373 if (state == SELF_DOWN_PEER_LEAVING) {
1374 if (msg_peer_node_is_up(hdr))
1375 return false;
1376 tipc_node_fsm_evt(n, PEER_LOST_CONTACT_EVT);
1377 return true;
1378 }
1379
1380 if (state == SELF_LEAVING_PEER_DOWN)
1381 return false;
1382
1383 /* Ignore duplicate packets */
1384 if ((usr != LINK_PROTOCOL) && less(oseqno, rcv_nxt))
1385 return true;
1386
1387 /* Initiate or update failover mode if applicable */
1388 if ((usr == TUNNEL_PROTOCOL) && (mtyp == FAILOVER_MSG)) {
1389 syncpt = oseqno + exp_pkts - 1;
1390 if (pl && tipc_link_is_up(pl)) {
1391 __tipc_node_link_down(n, &pb_id, xmitq, &maddr);
1392 tipc_skb_queue_splice_tail_init(tipc_link_inputq(pl),
1393 tipc_link_inputq(l));
1394 }
1395 /* If pkts arrive out of order, use lowest calculated syncpt */
1396 if (less(syncpt, n->sync_point))
1397 n->sync_point = syncpt;
1398 }
1399
1400 /* Open parallel link when tunnel link reaches synch point */
1401 if ((n->state == NODE_FAILINGOVER) && tipc_link_is_up(l)) {
1402 if (!more(rcv_nxt, n->sync_point))
1403 return true;
1404 tipc_node_fsm_evt(n, NODE_FAILOVER_END_EVT);
1405 if (pl)
1406 tipc_link_fsm_evt(pl, LINK_FAILOVER_END_EVT);
1407 return true;
1408 }
1409
1410 /* No synching needed if only one link */
1411 if (!pl || !tipc_link_is_up(pl))
1412 return true;
1413
1414 /* Initiate synch mode if applicable */
1415 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG) && (oseqno == 1)) {
1416 syncpt = iseqno + exp_pkts - 1;
1417 if (!tipc_link_is_up(l)) {
1418 tipc_link_fsm_evt(l, LINK_ESTABLISH_EVT);
1419 __tipc_node_link_up(n, bearer_id, xmitq);
1420 }
1421 if (n->state == SELF_UP_PEER_UP) {
1422 n->sync_point = syncpt;
1423 tipc_link_fsm_evt(l, LINK_SYNCH_BEGIN_EVT);
1424 tipc_node_fsm_evt(n, NODE_SYNCH_BEGIN_EVT);
1425 }
1426 }
1427
1428 /* Open tunnel link when parallel link reaches synch point */
1429 if (n->state == NODE_SYNCHING) {
1430 if (tipc_link_is_synching(l)) {
1431 tnl = l;
1432 } else {
1433 tnl = pl;
1434 pl = l;
1435 }
1436 inputq_len = skb_queue_len(tipc_link_inputq(pl));
1437 dlv_nxt = tipc_link_rcv_nxt(pl) - inputq_len;
1438 if (more(dlv_nxt, n->sync_point)) {
1439 tipc_link_fsm_evt(tnl, LINK_SYNCH_END_EVT);
1440 tipc_node_fsm_evt(n, NODE_SYNCH_END_EVT);
1441 return true;
1442 }
1443 if (l == pl)
1444 return true;
1445 if ((usr == TUNNEL_PROTOCOL) && (mtyp == SYNCH_MSG))
1446 return true;
1447 if (usr == LINK_PROTOCOL)
1448 return true;
1449 return false;
1450 }
1451 return true;
1452 }
1453
1454 /**
1455 * tipc_rcv - process TIPC packets/messages arriving from off-node
1456 * @net: the applicable net namespace
1457 * @skb: TIPC packet
1458 * @bearer: pointer to bearer message arrived on
1459 *
1460 * Invoked with no locks held. Bearer pointer must point to a valid bearer
1461 * structure (i.e. cannot be NULL), but bearer can be inactive.
1462 */
1463 void tipc_rcv(struct net *net, struct sk_buff *skb, struct tipc_bearer *b)
1464 {
1465 struct sk_buff_head xmitq;
1466 struct tipc_node *n;
1467 struct tipc_msg *hdr = buf_msg(skb);
1468 int usr = msg_user(hdr);
1469 int bearer_id = b->identity;
1470 struct tipc_link_entry *le;
1471 u16 bc_ack = msg_bcast_ack(hdr);
1472 u32 self = tipc_own_addr(net);
1473 int rc = 0;
1474
1475 __skb_queue_head_init(&xmitq);
1476
1477 /* Ensure message is well-formed */
1478 if (unlikely(!tipc_msg_validate(skb)))
1479 goto discard;
1480
1481 /* Handle arrival of discovery or broadcast packet */
1482 if (unlikely(msg_non_seq(hdr))) {
1483 if (unlikely(usr == LINK_CONFIG))
1484 return tipc_disc_rcv(net, skb, b);
1485 else
1486 return tipc_node_bc_rcv(net, skb, bearer_id);
1487 }
1488
1489 /* Discard unicast link messages destined for another node */
1490 if (unlikely(!msg_short(hdr) && (msg_destnode(hdr) != self)))
1491 goto discard;
1492
1493 /* Locate neighboring node that sent packet */
1494 n = tipc_node_find(net, msg_prevnode(hdr));
1495 if (unlikely(!n))
1496 goto discard;
1497 le = &n->links[bearer_id];
1498
1499 /* Ensure broadcast reception is in synch with peer's send state */
1500 if (unlikely(usr == LINK_PROTOCOL))
1501 tipc_bcast_sync_rcv(net, n->bc_entry.link, hdr);
1502 else if (unlikely(tipc_link_acked(n->bc_entry.link) != bc_ack))
1503 tipc_bcast_ack_rcv(net, n->bc_entry.link, bc_ack);
1504
1505 /* Receive packet directly if conditions permit */
1506 tipc_node_read_lock(n);
1507 if (likely((n->state == SELF_UP_PEER_UP) && (usr != TUNNEL_PROTOCOL))) {
1508 spin_lock_bh(&le->lock);
1509 if (le->link) {
1510 rc = tipc_link_rcv(le->link, skb, &xmitq);
1511 skb = NULL;
1512 }
1513 spin_unlock_bh(&le->lock);
1514 }
1515 tipc_node_read_unlock(n);
1516
1517 /* Check/update node state before receiving */
1518 if (unlikely(skb)) {
1519 tipc_node_write_lock(n);
1520 if (tipc_node_check_state(n, skb, bearer_id, &xmitq)) {
1521 if (le->link) {
1522 rc = tipc_link_rcv(le->link, skb, &xmitq);
1523 skb = NULL;
1524 }
1525 }
1526 tipc_node_write_unlock(n);
1527 }
1528
1529 if (unlikely(rc & TIPC_LINK_UP_EVT))
1530 tipc_node_link_up(n, bearer_id, &xmitq);
1531
1532 if (unlikely(rc & TIPC_LINK_DOWN_EVT))
1533 tipc_node_link_down(n, bearer_id, false);
1534
1535 if (unlikely(!skb_queue_empty(&n->bc_entry.namedq)))
1536 tipc_named_rcv(net, &n->bc_entry.namedq);
1537
1538 if (!skb_queue_empty(&le->inputq))
1539 tipc_sk_rcv(net, &le->inputq);
1540
1541 if (!skb_queue_empty(&xmitq))
1542 tipc_bearer_xmit(net, bearer_id, &xmitq, &le->maddr);
1543
1544 tipc_node_put(n);
1545 discard:
1546 kfree_skb(skb);
1547 }
1548
1549 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
1550 {
1551 int err;
1552 struct net *net = sock_net(skb->sk);
1553 struct tipc_net *tn = net_generic(net, tipc_net_id);
1554 int done = cb->args[0];
1555 int last_addr = cb->args[1];
1556 struct tipc_node *node;
1557 struct tipc_nl_msg msg;
1558
1559 if (done)
1560 return 0;
1561
1562 msg.skb = skb;
1563 msg.portid = NETLINK_CB(cb->skb).portid;
1564 msg.seq = cb->nlh->nlmsg_seq;
1565
1566 rcu_read_lock();
1567 if (last_addr) {
1568 node = tipc_node_find(net, last_addr);
1569 if (!node) {
1570 rcu_read_unlock();
1571 /* We never set seq or call nl_dump_check_consistent()
1572 * this means that setting prev_seq here will cause the
1573 * consistence check to fail in the netlink callback
1574 * handler. Resulting in the NLMSG_DONE message having
1575 * the NLM_F_DUMP_INTR flag set if the node state
1576 * changed while we released the lock.
1577 */
1578 cb->prev_seq = 1;
1579 return -EPIPE;
1580 }
1581 tipc_node_put(node);
1582 }
1583
1584 list_for_each_entry_rcu(node, &tn->node_list, list) {
1585 if (last_addr) {
1586 if (node->addr == last_addr)
1587 last_addr = 0;
1588 else
1589 continue;
1590 }
1591
1592 tipc_node_read_lock(node);
1593 err = __tipc_nl_add_node(&msg, node);
1594 if (err) {
1595 last_addr = node->addr;
1596 tipc_node_read_unlock(node);
1597 goto out;
1598 }
1599
1600 tipc_node_read_unlock(node);
1601 }
1602 done = 1;
1603 out:
1604 cb->args[0] = done;
1605 cb->args[1] = last_addr;
1606 rcu_read_unlock();
1607
1608 return skb->len;
1609 }
1610
1611 /* tipc_node_find_by_name - locate owner node of link by link's name
1612 * @net: the applicable net namespace
1613 * @name: pointer to link name string
1614 * @bearer_id: pointer to index in 'node->links' array where the link was found.
1615 *
1616 * Returns pointer to node owning the link, or 0 if no matching link is found.
1617 */
1618 static struct tipc_node *tipc_node_find_by_name(struct net *net,
1619 const char *link_name,
1620 unsigned int *bearer_id)
1621 {
1622 struct tipc_net *tn = net_generic(net, tipc_net_id);
1623 struct tipc_link *l;
1624 struct tipc_node *n;
1625 struct tipc_node *found_node = NULL;
1626 int i;
1627
1628 *bearer_id = 0;
1629 rcu_read_lock();
1630 list_for_each_entry_rcu(n, &tn->node_list, list) {
1631 tipc_node_read_lock(n);
1632 for (i = 0; i < MAX_BEARERS; i++) {
1633 l = n->links[i].link;
1634 if (l && !strcmp(tipc_link_name(l), link_name)) {
1635 *bearer_id = i;
1636 found_node = n;
1637 break;
1638 }
1639 }
1640 tipc_node_read_unlock(n);
1641 if (found_node)
1642 break;
1643 }
1644 rcu_read_unlock();
1645
1646 return found_node;
1647 }
1648
1649 int tipc_nl_node_set_link(struct sk_buff *skb, struct genl_info *info)
1650 {
1651 int err;
1652 int res = 0;
1653 int bearer_id;
1654 char *name;
1655 struct tipc_link *link;
1656 struct tipc_node *node;
1657 struct sk_buff_head xmitq;
1658 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1659 struct net *net = sock_net(skb->sk);
1660
1661 __skb_queue_head_init(&xmitq);
1662
1663 if (!info->attrs[TIPC_NLA_LINK])
1664 return -EINVAL;
1665
1666 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1667 info->attrs[TIPC_NLA_LINK],
1668 tipc_nl_link_policy);
1669 if (err)
1670 return err;
1671
1672 if (!attrs[TIPC_NLA_LINK_NAME])
1673 return -EINVAL;
1674
1675 name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1676
1677 if (strcmp(name, tipc_bclink_name) == 0)
1678 return tipc_nl_bc_link_set(net, attrs);
1679
1680 node = tipc_node_find_by_name(net, name, &bearer_id);
1681 if (!node)
1682 return -EINVAL;
1683
1684 tipc_node_read_lock(node);
1685
1686 link = node->links[bearer_id].link;
1687 if (!link) {
1688 res = -EINVAL;
1689 goto out;
1690 }
1691
1692 if (attrs[TIPC_NLA_LINK_PROP]) {
1693 struct nlattr *props[TIPC_NLA_PROP_MAX + 1];
1694
1695 err = tipc_nl_parse_link_prop(attrs[TIPC_NLA_LINK_PROP],
1696 props);
1697 if (err) {
1698 res = err;
1699 goto out;
1700 }
1701
1702 if (props[TIPC_NLA_PROP_TOL]) {
1703 u32 tol;
1704
1705 tol = nla_get_u32(props[TIPC_NLA_PROP_TOL]);
1706 tipc_link_set_tolerance(link, tol, &xmitq);
1707 }
1708 if (props[TIPC_NLA_PROP_PRIO]) {
1709 u32 prio;
1710
1711 prio = nla_get_u32(props[TIPC_NLA_PROP_PRIO]);
1712 tipc_link_set_prio(link, prio, &xmitq);
1713 }
1714 if (props[TIPC_NLA_PROP_WIN]) {
1715 u32 win;
1716
1717 win = nla_get_u32(props[TIPC_NLA_PROP_WIN]);
1718 tipc_link_set_queue_limits(link, win);
1719 }
1720 }
1721
1722 out:
1723 tipc_node_read_unlock(node);
1724 tipc_bearer_xmit(net, bearer_id, &xmitq, &node->links[bearer_id].maddr);
1725 return res;
1726 }
1727
1728 int tipc_nl_node_get_link(struct sk_buff *skb, struct genl_info *info)
1729 {
1730 struct net *net = genl_info_net(info);
1731 struct tipc_nl_msg msg;
1732 char *name;
1733 int err;
1734
1735 msg.portid = info->snd_portid;
1736 msg.seq = info->snd_seq;
1737
1738 if (!info->attrs[TIPC_NLA_LINK_NAME])
1739 return -EINVAL;
1740 name = nla_data(info->attrs[TIPC_NLA_LINK_NAME]);
1741
1742 msg.skb = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
1743 if (!msg.skb)
1744 return -ENOMEM;
1745
1746 if (strcmp(name, tipc_bclink_name) == 0) {
1747 err = tipc_nl_add_bc_link(net, &msg);
1748 if (err) {
1749 nlmsg_free(msg.skb);
1750 return err;
1751 }
1752 } else {
1753 int bearer_id;
1754 struct tipc_node *node;
1755 struct tipc_link *link;
1756
1757 node = tipc_node_find_by_name(net, name, &bearer_id);
1758 if (!node)
1759 return -EINVAL;
1760
1761 tipc_node_read_lock(node);
1762 link = node->links[bearer_id].link;
1763 if (!link) {
1764 tipc_node_read_unlock(node);
1765 nlmsg_free(msg.skb);
1766 return -EINVAL;
1767 }
1768
1769 err = __tipc_nl_add_link(net, &msg, link, 0);
1770 tipc_node_read_unlock(node);
1771 if (err) {
1772 nlmsg_free(msg.skb);
1773 return err;
1774 }
1775 }
1776
1777 return genlmsg_reply(msg.skb, info);
1778 }
1779
1780 int tipc_nl_node_reset_link_stats(struct sk_buff *skb, struct genl_info *info)
1781 {
1782 int err;
1783 char *link_name;
1784 unsigned int bearer_id;
1785 struct tipc_link *link;
1786 struct tipc_node *node;
1787 struct nlattr *attrs[TIPC_NLA_LINK_MAX + 1];
1788 struct net *net = sock_net(skb->sk);
1789 struct tipc_link_entry *le;
1790
1791 if (!info->attrs[TIPC_NLA_LINK])
1792 return -EINVAL;
1793
1794 err = nla_parse_nested(attrs, TIPC_NLA_LINK_MAX,
1795 info->attrs[TIPC_NLA_LINK],
1796 tipc_nl_link_policy);
1797 if (err)
1798 return err;
1799
1800 if (!attrs[TIPC_NLA_LINK_NAME])
1801 return -EINVAL;
1802
1803 link_name = nla_data(attrs[TIPC_NLA_LINK_NAME]);
1804
1805 if (strcmp(link_name, tipc_bclink_name) == 0) {
1806 err = tipc_bclink_reset_stats(net);
1807 if (err)
1808 return err;
1809 return 0;
1810 }
1811
1812 node = tipc_node_find_by_name(net, link_name, &bearer_id);
1813 if (!node)
1814 return -EINVAL;
1815
1816 le = &node->links[bearer_id];
1817 tipc_node_read_lock(node);
1818 spin_lock_bh(&le->lock);
1819 link = node->links[bearer_id].link;
1820 if (!link) {
1821 spin_unlock_bh(&le->lock);
1822 tipc_node_read_unlock(node);
1823 return -EINVAL;
1824 }
1825 tipc_link_reset_stats(link);
1826 spin_unlock_bh(&le->lock);
1827 tipc_node_read_unlock(node);
1828 return 0;
1829 }
1830
1831 /* Caller should hold node lock */
1832 static int __tipc_nl_add_node_links(struct net *net, struct tipc_nl_msg *msg,
1833 struct tipc_node *node, u32 *prev_link)
1834 {
1835 u32 i;
1836 int err;
1837
1838 for (i = *prev_link; i < MAX_BEARERS; i++) {
1839 *prev_link = i;
1840
1841 if (!node->links[i].link)
1842 continue;
1843
1844 err = __tipc_nl_add_link(net, msg,
1845 node->links[i].link, NLM_F_MULTI);
1846 if (err)
1847 return err;
1848 }
1849 *prev_link = 0;
1850
1851 return 0;
1852 }
1853
1854 int tipc_nl_node_dump_link(struct sk_buff *skb, struct netlink_callback *cb)
1855 {
1856 struct net *net = sock_net(skb->sk);
1857 struct tipc_net *tn = net_generic(net, tipc_net_id);
1858 struct tipc_node *node;
1859 struct tipc_nl_msg msg;
1860 u32 prev_node = cb->args[0];
1861 u32 prev_link = cb->args[1];
1862 int done = cb->args[2];
1863 int err;
1864
1865 if (done)
1866 return 0;
1867
1868 msg.skb = skb;
1869 msg.portid = NETLINK_CB(cb->skb).portid;
1870 msg.seq = cb->nlh->nlmsg_seq;
1871
1872 rcu_read_lock();
1873 if (prev_node) {
1874 node = tipc_node_find(net, prev_node);
1875 if (!node) {
1876 /* We never set seq or call nl_dump_check_consistent()
1877 * this means that setting prev_seq here will cause the
1878 * consistence check to fail in the netlink callback
1879 * handler. Resulting in the last NLMSG_DONE message
1880 * having the NLM_F_DUMP_INTR flag set.
1881 */
1882 cb->prev_seq = 1;
1883 goto out;
1884 }
1885 tipc_node_put(node);
1886
1887 list_for_each_entry_continue_rcu(node, &tn->node_list,
1888 list) {
1889 tipc_node_read_lock(node);
1890 err = __tipc_nl_add_node_links(net, &msg, node,
1891 &prev_link);
1892 tipc_node_read_unlock(node);
1893 if (err)
1894 goto out;
1895
1896 prev_node = node->addr;
1897 }
1898 } else {
1899 err = tipc_nl_add_bc_link(net, &msg);
1900 if (err)
1901 goto out;
1902
1903 list_for_each_entry_rcu(node, &tn->node_list, list) {
1904 tipc_node_read_lock(node);
1905 err = __tipc_nl_add_node_links(net, &msg, node,
1906 &prev_link);
1907 tipc_node_read_unlock(node);
1908 if (err)
1909 goto out;
1910
1911 prev_node = node->addr;
1912 }
1913 }
1914 done = 1;
1915 out:
1916 rcu_read_unlock();
1917
1918 cb->args[0] = prev_node;
1919 cb->args[1] = prev_link;
1920 cb->args[2] = done;
1921
1922 return skb->len;
1923 }