xref: /linux/net/tipc/node.c (revision e0bf6c5ca2d3281f231c5f0c9bf145e9513644de)
1 /*
2  * net/tipc/node.c: TIPC node management routines
3  *
4  * Copyright (c) 2000-2006, 2012-2014, 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 
43 static void node_lost_contact(struct tipc_node *n_ptr);
44 static void node_established_contact(struct tipc_node *n_ptr);
45 
46 struct tipc_sock_conn {
47 	u32 port;
48 	u32 peer_port;
49 	u32 peer_node;
50 	struct list_head list;
51 };
52 
53 static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
54 	[TIPC_NLA_NODE_UNSPEC]		= { .type = NLA_UNSPEC },
55 	[TIPC_NLA_NODE_ADDR]		= { .type = NLA_U32 },
56 	[TIPC_NLA_NODE_UP]		= { .type = NLA_FLAG }
57 };
58 
59 /*
60  * A trivial power-of-two bitmask technique is used for speed, since this
61  * operation is done for every incoming TIPC packet. The number of hash table
62  * entries has been chosen so that no hash chain exceeds 8 nodes and will
63  * usually be much smaller (typically only a single node).
64  */
65 static unsigned int tipc_hashfn(u32 addr)
66 {
67 	return addr & (NODE_HTABLE_SIZE - 1);
68 }
69 
70 /*
71  * tipc_node_find - locate specified node object, if it exists
72  */
73 struct tipc_node *tipc_node_find(struct net *net, u32 addr)
74 {
75 	struct tipc_net *tn = net_generic(net, tipc_net_id);
76 	struct tipc_node *node;
77 
78 	if (unlikely(!in_own_cluster_exact(net, addr)))
79 		return NULL;
80 
81 	rcu_read_lock();
82 	hlist_for_each_entry_rcu(node, &tn->node_htable[tipc_hashfn(addr)],
83 				 hash) {
84 		if (node->addr == addr) {
85 			rcu_read_unlock();
86 			return node;
87 		}
88 	}
89 	rcu_read_unlock();
90 	return NULL;
91 }
92 
93 struct tipc_node *tipc_node_create(struct net *net, u32 addr)
94 {
95 	struct tipc_net *tn = net_generic(net, tipc_net_id);
96 	struct tipc_node *n_ptr, *temp_node;
97 
98 	spin_lock_bh(&tn->node_list_lock);
99 	n_ptr = tipc_node_find(net, addr);
100 	if (n_ptr)
101 		goto exit;
102 	n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
103 	if (!n_ptr) {
104 		pr_warn("Node creation failed, no memory\n");
105 		goto exit;
106 	}
107 	n_ptr->addr = addr;
108 	n_ptr->net = net;
109 	spin_lock_init(&n_ptr->lock);
110 	INIT_HLIST_NODE(&n_ptr->hash);
111 	INIT_LIST_HEAD(&n_ptr->list);
112 	INIT_LIST_HEAD(&n_ptr->publ_list);
113 	INIT_LIST_HEAD(&n_ptr->conn_sks);
114 	__skb_queue_head_init(&n_ptr->bclink.deferred_queue);
115 	hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
116 	list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
117 		if (n_ptr->addr < temp_node->addr)
118 			break;
119 	}
120 	list_add_tail_rcu(&n_ptr->list, &temp_node->list);
121 	n_ptr->action_flags = TIPC_WAIT_PEER_LINKS_DOWN;
122 	n_ptr->signature = INVALID_NODE_SIG;
123 exit:
124 	spin_unlock_bh(&tn->node_list_lock);
125 	return n_ptr;
126 }
127 
128 static void tipc_node_delete(struct tipc_net *tn, struct tipc_node *n_ptr)
129 {
130 	list_del_rcu(&n_ptr->list);
131 	hlist_del_rcu(&n_ptr->hash);
132 	kfree_rcu(n_ptr, rcu);
133 }
134 
135 void tipc_node_stop(struct net *net)
136 {
137 	struct tipc_net *tn = net_generic(net, tipc_net_id);
138 	struct tipc_node *node, *t_node;
139 
140 	spin_lock_bh(&tn->node_list_lock);
141 	list_for_each_entry_safe(node, t_node, &tn->node_list, list)
142 		tipc_node_delete(tn, node);
143 	spin_unlock_bh(&tn->node_list_lock);
144 }
145 
146 int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
147 {
148 	struct tipc_node *node;
149 	struct tipc_sock_conn *conn;
150 
151 	if (in_own_node(net, dnode))
152 		return 0;
153 
154 	node = tipc_node_find(net, dnode);
155 	if (!node) {
156 		pr_warn("Connecting sock to node 0x%x failed\n", dnode);
157 		return -EHOSTUNREACH;
158 	}
159 	conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
160 	if (!conn)
161 		return -EHOSTUNREACH;
162 	conn->peer_node = dnode;
163 	conn->port = port;
164 	conn->peer_port = peer_port;
165 
166 	tipc_node_lock(node);
167 	list_add_tail(&conn->list, &node->conn_sks);
168 	tipc_node_unlock(node);
169 	return 0;
170 }
171 
172 void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
173 {
174 	struct tipc_node *node;
175 	struct tipc_sock_conn *conn, *safe;
176 
177 	if (in_own_node(net, dnode))
178 		return;
179 
180 	node = tipc_node_find(net, dnode);
181 	if (!node)
182 		return;
183 
184 	tipc_node_lock(node);
185 	list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
186 		if (port != conn->port)
187 			continue;
188 		list_del(&conn->list);
189 		kfree(conn);
190 	}
191 	tipc_node_unlock(node);
192 }
193 
194 /**
195  * tipc_node_link_up - handle addition of link
196  *
197  * Link becomes active (alone or shared) or standby, depending on its priority.
198  */
199 void tipc_node_link_up(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
200 {
201 	struct tipc_link **active = &n_ptr->active_links[0];
202 
203 	n_ptr->working_links++;
204 	n_ptr->action_flags |= TIPC_NOTIFY_LINK_UP;
205 	n_ptr->link_id = l_ptr->peer_bearer_id << 16 | l_ptr->bearer_id;
206 
207 	pr_debug("Established link <%s> on network plane %c\n",
208 		 l_ptr->name, l_ptr->net_plane);
209 
210 	if (!active[0]) {
211 		active[0] = active[1] = l_ptr;
212 		node_established_contact(n_ptr);
213 		goto exit;
214 	}
215 	if (l_ptr->priority < active[0]->priority) {
216 		pr_debug("New link <%s> becomes standby\n", l_ptr->name);
217 		goto exit;
218 	}
219 	tipc_link_dup_queue_xmit(active[0], l_ptr);
220 	if (l_ptr->priority == active[0]->priority) {
221 		active[0] = l_ptr;
222 		goto exit;
223 	}
224 	pr_debug("Old link <%s> becomes standby\n", active[0]->name);
225 	if (active[1] != active[0])
226 		pr_debug("Old link <%s> becomes standby\n", active[1]->name);
227 	active[0] = active[1] = l_ptr;
228 exit:
229 	/* Leave room for changeover header when returning 'mtu' to users: */
230 	n_ptr->act_mtus[0] = active[0]->max_pkt - INT_H_SIZE;
231 	n_ptr->act_mtus[1] = active[1]->max_pkt - INT_H_SIZE;
232 }
233 
234 /**
235  * node_select_active_links - select active link
236  */
237 static void node_select_active_links(struct tipc_node *n_ptr)
238 {
239 	struct tipc_link **active = &n_ptr->active_links[0];
240 	u32 i;
241 	u32 highest_prio = 0;
242 
243 	active[0] = active[1] = NULL;
244 
245 	for (i = 0; i < MAX_BEARERS; i++) {
246 		struct tipc_link *l_ptr = n_ptr->links[i];
247 
248 		if (!l_ptr || !tipc_link_is_up(l_ptr) ||
249 		    (l_ptr->priority < highest_prio))
250 			continue;
251 
252 		if (l_ptr->priority > highest_prio) {
253 			highest_prio = l_ptr->priority;
254 			active[0] = active[1] = l_ptr;
255 		} else {
256 			active[1] = l_ptr;
257 		}
258 	}
259 }
260 
261 /**
262  * tipc_node_link_down - handle loss of link
263  */
264 void tipc_node_link_down(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
265 {
266 	struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
267 	struct tipc_link **active;
268 
269 	n_ptr->working_links--;
270 	n_ptr->action_flags |= TIPC_NOTIFY_LINK_DOWN;
271 	n_ptr->link_id = l_ptr->peer_bearer_id << 16 | l_ptr->bearer_id;
272 
273 	if (!tipc_link_is_active(l_ptr)) {
274 		pr_debug("Lost standby link <%s> on network plane %c\n",
275 			 l_ptr->name, l_ptr->net_plane);
276 		return;
277 	}
278 	pr_debug("Lost link <%s> on network plane %c\n",
279 		 l_ptr->name, l_ptr->net_plane);
280 
281 	active = &n_ptr->active_links[0];
282 	if (active[0] == l_ptr)
283 		active[0] = active[1];
284 	if (active[1] == l_ptr)
285 		active[1] = active[0];
286 	if (active[0] == l_ptr)
287 		node_select_active_links(n_ptr);
288 	if (tipc_node_is_up(n_ptr))
289 		tipc_link_failover_send_queue(l_ptr);
290 	else
291 		node_lost_contact(n_ptr);
292 
293 	/* Leave room for changeover header when returning 'mtu' to users: */
294 	if (active[0]) {
295 		n_ptr->act_mtus[0] = active[0]->max_pkt - INT_H_SIZE;
296 		n_ptr->act_mtus[1] = active[1]->max_pkt - INT_H_SIZE;
297 		return;
298 	}
299 
300 	/* Loopback link went down? No fragmentation needed from now on. */
301 	if (n_ptr->addr == tn->own_addr) {
302 		n_ptr->act_mtus[0] = MAX_MSG_SIZE;
303 		n_ptr->act_mtus[1] = MAX_MSG_SIZE;
304 	}
305 }
306 
307 int tipc_node_active_links(struct tipc_node *n_ptr)
308 {
309 	return n_ptr->active_links[0] != NULL;
310 }
311 
312 int tipc_node_is_up(struct tipc_node *n_ptr)
313 {
314 	return tipc_node_active_links(n_ptr);
315 }
316 
317 void tipc_node_attach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
318 {
319 	n_ptr->links[l_ptr->bearer_id] = l_ptr;
320 	n_ptr->link_cnt++;
321 }
322 
323 void tipc_node_detach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
324 {
325 	int i;
326 
327 	for (i = 0; i < MAX_BEARERS; i++) {
328 		if (l_ptr != n_ptr->links[i])
329 			continue;
330 		n_ptr->links[i] = NULL;
331 		n_ptr->link_cnt--;
332 	}
333 }
334 
335 static void node_established_contact(struct tipc_node *n_ptr)
336 {
337 	n_ptr->action_flags |= TIPC_NOTIFY_NODE_UP;
338 	n_ptr->bclink.oos_state = 0;
339 	n_ptr->bclink.acked = tipc_bclink_get_last_sent(n_ptr->net);
340 	tipc_bclink_add_node(n_ptr->net, n_ptr->addr);
341 }
342 
343 static void node_lost_contact(struct tipc_node *n_ptr)
344 {
345 	char addr_string[16];
346 	struct tipc_sock_conn *conn, *safe;
347 	struct list_head *conns = &n_ptr->conn_sks;
348 	struct sk_buff *skb;
349 	struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
350 	uint i;
351 
352 	pr_debug("Lost contact with %s\n",
353 		 tipc_addr_string_fill(addr_string, n_ptr->addr));
354 
355 	/* Flush broadcast link info associated with lost node */
356 	if (n_ptr->bclink.recv_permitted) {
357 		__skb_queue_purge(&n_ptr->bclink.deferred_queue);
358 
359 		if (n_ptr->bclink.reasm_buf) {
360 			kfree_skb(n_ptr->bclink.reasm_buf);
361 			n_ptr->bclink.reasm_buf = NULL;
362 		}
363 
364 		tipc_bclink_remove_node(n_ptr->net, n_ptr->addr);
365 		tipc_bclink_acknowledge(n_ptr, INVALID_LINK_SEQ);
366 
367 		n_ptr->bclink.recv_permitted = false;
368 	}
369 
370 	/* Abort link changeover */
371 	for (i = 0; i < MAX_BEARERS; i++) {
372 		struct tipc_link *l_ptr = n_ptr->links[i];
373 		if (!l_ptr)
374 			continue;
375 		l_ptr->reset_checkpoint = l_ptr->next_in_no;
376 		l_ptr->exp_msg_count = 0;
377 		tipc_link_reset_fragments(l_ptr);
378 
379 		/* Link marked for deletion after failover? => do it now */
380 		if (l_ptr->flags & LINK_STOPPED)
381 			tipc_link_delete(l_ptr);
382 	}
383 
384 	n_ptr->action_flags &= ~TIPC_WAIT_OWN_LINKS_DOWN;
385 
386 	/* Prevent re-contact with node until cleanup is done */
387 	n_ptr->action_flags |= TIPC_WAIT_PEER_LINKS_DOWN;
388 
389 	/* Notify publications from this node */
390 	n_ptr->action_flags |= TIPC_NOTIFY_NODE_DOWN;
391 
392 	/* Notify sockets connected to node */
393 	list_for_each_entry_safe(conn, safe, conns, list) {
394 		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
395 				      SHORT_H_SIZE, 0, tn->own_addr,
396 				      conn->peer_node, conn->port,
397 				      conn->peer_port, TIPC_ERR_NO_NODE);
398 		if (likely(skb)) {
399 			skb_queue_tail(n_ptr->inputq, skb);
400 			n_ptr->action_flags |= TIPC_MSG_EVT;
401 		}
402 		list_del(&conn->list);
403 		kfree(conn);
404 	}
405 }
406 
407 /**
408  * tipc_node_get_linkname - get the name of a link
409  *
410  * @bearer_id: id of the bearer
411  * @node: peer node address
412  * @linkname: link name output buffer
413  *
414  * Returns 0 on success
415  */
416 int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
417 			   char *linkname, size_t len)
418 {
419 	struct tipc_link *link;
420 	struct tipc_node *node = tipc_node_find(net, addr);
421 
422 	if ((bearer_id >= MAX_BEARERS) || !node)
423 		return -EINVAL;
424 	tipc_node_lock(node);
425 	link = node->links[bearer_id];
426 	if (link) {
427 		strncpy(linkname, link->name, len);
428 		tipc_node_unlock(node);
429 		return 0;
430 	}
431 	tipc_node_unlock(node);
432 	return -EINVAL;
433 }
434 
435 void tipc_node_unlock(struct tipc_node *node)
436 {
437 	struct net *net = node->net;
438 	u32 addr = 0;
439 	u32 flags = node->action_flags;
440 	u32 link_id = 0;
441 	struct list_head *publ_list;
442 	struct sk_buff_head *inputq = node->inputq;
443 	struct sk_buff_head *namedq;
444 
445 	if (likely(!flags || (flags == TIPC_MSG_EVT))) {
446 		node->action_flags = 0;
447 		spin_unlock_bh(&node->lock);
448 		if (flags == TIPC_MSG_EVT)
449 			tipc_sk_rcv(net, inputq);
450 		return;
451 	}
452 
453 	addr = node->addr;
454 	link_id = node->link_id;
455 	namedq = node->namedq;
456 	publ_list = &node->publ_list;
457 
458 	node->action_flags &= ~(TIPC_MSG_EVT |
459 				TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
460 				TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP |
461 				TIPC_WAKEUP_BCAST_USERS | TIPC_BCAST_MSG_EVT |
462 				TIPC_NAMED_MSG_EVT);
463 
464 	spin_unlock_bh(&node->lock);
465 
466 	if (flags & TIPC_NOTIFY_NODE_DOWN)
467 		tipc_publ_notify(net, publ_list, addr);
468 
469 	if (flags & TIPC_WAKEUP_BCAST_USERS)
470 		tipc_bclink_wakeup_users(net);
471 
472 	if (flags & TIPC_NOTIFY_NODE_UP)
473 		tipc_named_node_up(net, addr);
474 
475 	if (flags & TIPC_NOTIFY_LINK_UP)
476 		tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
477 				     TIPC_NODE_SCOPE, link_id, addr);
478 
479 	if (flags & TIPC_NOTIFY_LINK_DOWN)
480 		tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
481 				      link_id, addr);
482 
483 	if (flags & TIPC_MSG_EVT)
484 		tipc_sk_rcv(net, inputq);
485 
486 	if (flags & TIPC_NAMED_MSG_EVT)
487 		tipc_named_rcv(net, namedq);
488 
489 	if (flags & TIPC_BCAST_MSG_EVT)
490 		tipc_bclink_input(net);
491 }
492 
493 /* Caller should hold node lock for the passed node */
494 static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
495 {
496 	void *hdr;
497 	struct nlattr *attrs;
498 
499 	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
500 			  NLM_F_MULTI, TIPC_NL_NODE_GET);
501 	if (!hdr)
502 		return -EMSGSIZE;
503 
504 	attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
505 	if (!attrs)
506 		goto msg_full;
507 
508 	if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
509 		goto attr_msg_full;
510 	if (tipc_node_is_up(node))
511 		if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
512 			goto attr_msg_full;
513 
514 	nla_nest_end(msg->skb, attrs);
515 	genlmsg_end(msg->skb, hdr);
516 
517 	return 0;
518 
519 attr_msg_full:
520 	nla_nest_cancel(msg->skb, attrs);
521 msg_full:
522 	genlmsg_cancel(msg->skb, hdr);
523 
524 	return -EMSGSIZE;
525 }
526 
527 int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
528 {
529 	int err;
530 	struct net *net = sock_net(skb->sk);
531 	struct tipc_net *tn = net_generic(net, tipc_net_id);
532 	int done = cb->args[0];
533 	int last_addr = cb->args[1];
534 	struct tipc_node *node;
535 	struct tipc_nl_msg msg;
536 
537 	if (done)
538 		return 0;
539 
540 	msg.skb = skb;
541 	msg.portid = NETLINK_CB(cb->skb).portid;
542 	msg.seq = cb->nlh->nlmsg_seq;
543 
544 	rcu_read_lock();
545 
546 	if (last_addr && !tipc_node_find(net, last_addr)) {
547 		rcu_read_unlock();
548 		/* We never set seq or call nl_dump_check_consistent() this
549 		 * means that setting prev_seq here will cause the consistence
550 		 * check to fail in the netlink callback handler. Resulting in
551 		 * the NLMSG_DONE message having the NLM_F_DUMP_INTR flag set if
552 		 * the node state changed while we released the lock.
553 		 */
554 		cb->prev_seq = 1;
555 		return -EPIPE;
556 	}
557 
558 	list_for_each_entry_rcu(node, &tn->node_list, list) {
559 		if (last_addr) {
560 			if (node->addr == last_addr)
561 				last_addr = 0;
562 			else
563 				continue;
564 		}
565 
566 		tipc_node_lock(node);
567 		err = __tipc_nl_add_node(&msg, node);
568 		if (err) {
569 			last_addr = node->addr;
570 			tipc_node_unlock(node);
571 			goto out;
572 		}
573 
574 		tipc_node_unlock(node);
575 	}
576 	done = 1;
577 out:
578 	cb->args[0] = done;
579 	cb->args[1] = last_addr;
580 	rcu_read_unlock();
581 
582 	return skb->len;
583 }
584