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
43static void node_lost_contact(struct tipc_node *n_ptr);
44static void node_established_contact(struct tipc_node *n_ptr);
45static void tipc_node_delete(struct tipc_node *node);
46
47struct tipc_sock_conn {
48	u32 port;
49	u32 peer_port;
50	u32 peer_node;
51	struct list_head list;
52};
53
54static const struct nla_policy tipc_nl_node_policy[TIPC_NLA_NODE_MAX + 1] = {
55	[TIPC_NLA_NODE_UNSPEC]		= { .type = NLA_UNSPEC },
56	[TIPC_NLA_NODE_ADDR]		= { .type = NLA_U32 },
57	[TIPC_NLA_NODE_UP]		= { .type = NLA_FLAG }
58};
59
60/*
61 * A trivial power-of-two bitmask technique is used for speed, since this
62 * operation is done for every incoming TIPC packet. The number of hash table
63 * entries has been chosen so that no hash chain exceeds 8 nodes and will
64 * usually be much smaller (typically only a single node).
65 */
66static unsigned int tipc_hashfn(u32 addr)
67{
68	return addr & (NODE_HTABLE_SIZE - 1);
69}
70
71static void tipc_node_kref_release(struct kref *kref)
72{
73	struct tipc_node *node = container_of(kref, struct tipc_node, kref);
74
75	tipc_node_delete(node);
76}
77
78void tipc_node_put(struct tipc_node *node)
79{
80	kref_put(&node->kref, tipc_node_kref_release);
81}
82
83static void tipc_node_get(struct tipc_node *node)
84{
85	kref_get(&node->kref);
86}
87
88/*
89 * tipc_node_find - locate specified node object, if it exists
90 */
91struct tipc_node *tipc_node_find(struct net *net, u32 addr)
92{
93	struct tipc_net *tn = net_generic(net, tipc_net_id);
94	struct tipc_node *node;
95
96	if (unlikely(!in_own_cluster_exact(net, addr)))
97		return NULL;
98
99	rcu_read_lock();
100	hlist_for_each_entry_rcu(node, &tn->node_htable[tipc_hashfn(addr)],
101				 hash) {
102		if (node->addr == addr) {
103			tipc_node_get(node);
104			rcu_read_unlock();
105			return node;
106		}
107	}
108	rcu_read_unlock();
109	return NULL;
110}
111
112struct tipc_node *tipc_node_create(struct net *net, u32 addr)
113{
114	struct tipc_net *tn = net_generic(net, tipc_net_id);
115	struct tipc_node *n_ptr, *temp_node;
116
117	spin_lock_bh(&tn->node_list_lock);
118	n_ptr = tipc_node_find(net, addr);
119	if (n_ptr)
120		goto exit;
121	n_ptr = kzalloc(sizeof(*n_ptr), GFP_ATOMIC);
122	if (!n_ptr) {
123		pr_warn("Node creation failed, no memory\n");
124		goto exit;
125	}
126	n_ptr->addr = addr;
127	n_ptr->net = net;
128	kref_init(&n_ptr->kref);
129	spin_lock_init(&n_ptr->lock);
130	INIT_HLIST_NODE(&n_ptr->hash);
131	INIT_LIST_HEAD(&n_ptr->list);
132	INIT_LIST_HEAD(&n_ptr->publ_list);
133	INIT_LIST_HEAD(&n_ptr->conn_sks);
134	__skb_queue_head_init(&n_ptr->bclink.deferdq);
135	hlist_add_head_rcu(&n_ptr->hash, &tn->node_htable[tipc_hashfn(addr)]);
136	list_for_each_entry_rcu(temp_node, &tn->node_list, list) {
137		if (n_ptr->addr < temp_node->addr)
138			break;
139	}
140	list_add_tail_rcu(&n_ptr->list, &temp_node->list);
141	n_ptr->action_flags = TIPC_WAIT_PEER_LINKS_DOWN;
142	n_ptr->signature = INVALID_NODE_SIG;
143	tipc_node_get(n_ptr);
144exit:
145	spin_unlock_bh(&tn->node_list_lock);
146	return n_ptr;
147}
148
149static void tipc_node_delete(struct tipc_node *node)
150{
151	list_del_rcu(&node->list);
152	hlist_del_rcu(&node->hash);
153	kfree_rcu(node, rcu);
154}
155
156void tipc_node_stop(struct net *net)
157{
158	struct tipc_net *tn = net_generic(net, tipc_net_id);
159	struct tipc_node *node, *t_node;
160
161	spin_lock_bh(&tn->node_list_lock);
162	list_for_each_entry_safe(node, t_node, &tn->node_list, list)
163		tipc_node_put(node);
164	spin_unlock_bh(&tn->node_list_lock);
165}
166
167int tipc_node_add_conn(struct net *net, u32 dnode, u32 port, u32 peer_port)
168{
169	struct tipc_node *node;
170	struct tipc_sock_conn *conn;
171	int err = 0;
172
173	if (in_own_node(net, dnode))
174		return 0;
175
176	node = tipc_node_find(net, dnode);
177	if (!node) {
178		pr_warn("Connecting sock to node 0x%x failed\n", dnode);
179		return -EHOSTUNREACH;
180	}
181	conn = kmalloc(sizeof(*conn), GFP_ATOMIC);
182	if (!conn) {
183		err = -EHOSTUNREACH;
184		goto exit;
185	}
186	conn->peer_node = dnode;
187	conn->port = port;
188	conn->peer_port = peer_port;
189
190	tipc_node_lock(node);
191	list_add_tail(&conn->list, &node->conn_sks);
192	tipc_node_unlock(node);
193exit:
194	tipc_node_put(node);
195	return err;
196}
197
198void tipc_node_remove_conn(struct net *net, u32 dnode, u32 port)
199{
200	struct tipc_node *node;
201	struct tipc_sock_conn *conn, *safe;
202
203	if (in_own_node(net, dnode))
204		return;
205
206	node = tipc_node_find(net, dnode);
207	if (!node)
208		return;
209
210	tipc_node_lock(node);
211	list_for_each_entry_safe(conn, safe, &node->conn_sks, list) {
212		if (port != conn->port)
213			continue;
214		list_del(&conn->list);
215		kfree(conn);
216	}
217	tipc_node_unlock(node);
218	tipc_node_put(node);
219}
220
221/**
222 * tipc_node_link_up - handle addition of link
223 *
224 * Link becomes active (alone or shared) or standby, depending on its priority.
225 */
226void tipc_node_link_up(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
227{
228	struct tipc_link **active = &n_ptr->active_links[0];
229
230	n_ptr->working_links++;
231	n_ptr->action_flags |= TIPC_NOTIFY_LINK_UP;
232	n_ptr->link_id = l_ptr->peer_bearer_id << 16 | l_ptr->bearer_id;
233
234	pr_debug("Established link <%s> on network plane %c\n",
235		 l_ptr->name, l_ptr->net_plane);
236
237	if (!active[0]) {
238		active[0] = active[1] = l_ptr;
239		node_established_contact(n_ptr);
240		goto exit;
241	}
242	if (l_ptr->priority < active[0]->priority) {
243		pr_debug("New link <%s> becomes standby\n", l_ptr->name);
244		goto exit;
245	}
246	tipc_link_dup_queue_xmit(active[0], l_ptr);
247	if (l_ptr->priority == active[0]->priority) {
248		active[0] = l_ptr;
249		goto exit;
250	}
251	pr_debug("Old link <%s> becomes standby\n", active[0]->name);
252	if (active[1] != active[0])
253		pr_debug("Old link <%s> becomes standby\n", active[1]->name);
254	active[0] = active[1] = l_ptr;
255exit:
256	/* Leave room for changeover header when returning 'mtu' to users: */
257	n_ptr->act_mtus[0] = active[0]->mtu - INT_H_SIZE;
258	n_ptr->act_mtus[1] = active[1]->mtu - INT_H_SIZE;
259}
260
261/**
262 * node_select_active_links - select active link
263 */
264static void node_select_active_links(struct tipc_node *n_ptr)
265{
266	struct tipc_link **active = &n_ptr->active_links[0];
267	u32 i;
268	u32 highest_prio = 0;
269
270	active[0] = active[1] = NULL;
271
272	for (i = 0; i < MAX_BEARERS; i++) {
273		struct tipc_link *l_ptr = n_ptr->links[i];
274
275		if (!l_ptr || !tipc_link_is_up(l_ptr) ||
276		    (l_ptr->priority < highest_prio))
277			continue;
278
279		if (l_ptr->priority > highest_prio) {
280			highest_prio = l_ptr->priority;
281			active[0] = active[1] = l_ptr;
282		} else {
283			active[1] = l_ptr;
284		}
285	}
286}
287
288/**
289 * tipc_node_link_down - handle loss of link
290 */
291void tipc_node_link_down(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
292{
293	struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
294	struct tipc_link **active;
295
296	n_ptr->working_links--;
297	n_ptr->action_flags |= TIPC_NOTIFY_LINK_DOWN;
298	n_ptr->link_id = l_ptr->peer_bearer_id << 16 | l_ptr->bearer_id;
299
300	if (!tipc_link_is_active(l_ptr)) {
301		pr_debug("Lost standby link <%s> on network plane %c\n",
302			 l_ptr->name, l_ptr->net_plane);
303		return;
304	}
305	pr_debug("Lost link <%s> on network plane %c\n",
306		 l_ptr->name, l_ptr->net_plane);
307
308	active = &n_ptr->active_links[0];
309	if (active[0] == l_ptr)
310		active[0] = active[1];
311	if (active[1] == l_ptr)
312		active[1] = active[0];
313	if (active[0] == l_ptr)
314		node_select_active_links(n_ptr);
315	if (tipc_node_is_up(n_ptr))
316		tipc_link_failover_send_queue(l_ptr);
317	else
318		node_lost_contact(n_ptr);
319
320	/* Leave room for changeover header when returning 'mtu' to users: */
321	if (active[0]) {
322		n_ptr->act_mtus[0] = active[0]->mtu - INT_H_SIZE;
323		n_ptr->act_mtus[1] = active[1]->mtu - INT_H_SIZE;
324		return;
325	}
326	/* Loopback link went down? No fragmentation needed from now on. */
327	if (n_ptr->addr == tn->own_addr) {
328		n_ptr->act_mtus[0] = MAX_MSG_SIZE;
329		n_ptr->act_mtus[1] = MAX_MSG_SIZE;
330	}
331}
332
333int tipc_node_active_links(struct tipc_node *n_ptr)
334{
335	return n_ptr->active_links[0] != NULL;
336}
337
338int tipc_node_is_up(struct tipc_node *n_ptr)
339{
340	return tipc_node_active_links(n_ptr);
341}
342
343void tipc_node_attach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
344{
345	n_ptr->links[l_ptr->bearer_id] = l_ptr;
346	n_ptr->link_cnt++;
347}
348
349void tipc_node_detach_link(struct tipc_node *n_ptr, struct tipc_link *l_ptr)
350{
351	int i;
352
353	for (i = 0; i < MAX_BEARERS; i++) {
354		if (l_ptr != n_ptr->links[i])
355			continue;
356		n_ptr->links[i] = NULL;
357		n_ptr->link_cnt--;
358	}
359}
360
361static void node_established_contact(struct tipc_node *n_ptr)
362{
363	n_ptr->action_flags |= TIPC_NOTIFY_NODE_UP;
364	n_ptr->bclink.oos_state = 0;
365	n_ptr->bclink.acked = tipc_bclink_get_last_sent(n_ptr->net);
366	tipc_bclink_add_node(n_ptr->net, n_ptr->addr);
367}
368
369static void node_lost_contact(struct tipc_node *n_ptr)
370{
371	char addr_string[16];
372	struct tipc_sock_conn *conn, *safe;
373	struct list_head *conns = &n_ptr->conn_sks;
374	struct sk_buff *skb;
375	struct tipc_net *tn = net_generic(n_ptr->net, tipc_net_id);
376	uint i;
377
378	pr_debug("Lost contact with %s\n",
379		 tipc_addr_string_fill(addr_string, n_ptr->addr));
380
381	/* Flush broadcast link info associated with lost node */
382	if (n_ptr->bclink.recv_permitted) {
383		__skb_queue_purge(&n_ptr->bclink.deferdq);
384
385		if (n_ptr->bclink.reasm_buf) {
386			kfree_skb(n_ptr->bclink.reasm_buf);
387			n_ptr->bclink.reasm_buf = NULL;
388		}
389
390		tipc_bclink_remove_node(n_ptr->net, n_ptr->addr);
391		tipc_bclink_acknowledge(n_ptr, INVALID_LINK_SEQ);
392
393		n_ptr->bclink.recv_permitted = false;
394	}
395
396	/* Abort any ongoing link failover */
397	for (i = 0; i < MAX_BEARERS; i++) {
398		struct tipc_link *l_ptr = n_ptr->links[i];
399		if (!l_ptr)
400			continue;
401		l_ptr->flags &= ~LINK_FAILINGOVER;
402		l_ptr->failover_checkpt = 0;
403		l_ptr->failover_pkts = 0;
404		kfree_skb(l_ptr->failover_skb);
405		l_ptr->failover_skb = NULL;
406		tipc_link_reset_fragments(l_ptr);
407	}
408
409	n_ptr->action_flags &= ~TIPC_WAIT_OWN_LINKS_DOWN;
410
411	/* Prevent re-contact with node until cleanup is done */
412	n_ptr->action_flags |= TIPC_WAIT_PEER_LINKS_DOWN;
413
414	/* Notify publications from this node */
415	n_ptr->action_flags |= TIPC_NOTIFY_NODE_DOWN;
416
417	/* Notify sockets connected to node */
418	list_for_each_entry_safe(conn, safe, conns, list) {
419		skb = tipc_msg_create(TIPC_CRITICAL_IMPORTANCE, TIPC_CONN_MSG,
420				      SHORT_H_SIZE, 0, tn->own_addr,
421				      conn->peer_node, conn->port,
422				      conn->peer_port, TIPC_ERR_NO_NODE);
423		if (likely(skb)) {
424			skb_queue_tail(n_ptr->inputq, skb);
425			n_ptr->action_flags |= TIPC_MSG_EVT;
426		}
427		list_del(&conn->list);
428		kfree(conn);
429	}
430}
431
432/**
433 * tipc_node_get_linkname - get the name of a link
434 *
435 * @bearer_id: id of the bearer
436 * @node: peer node address
437 * @linkname: link name output buffer
438 *
439 * Returns 0 on success
440 */
441int tipc_node_get_linkname(struct net *net, u32 bearer_id, u32 addr,
442			   char *linkname, size_t len)
443{
444	struct tipc_link *link;
445	int err = -EINVAL;
446	struct tipc_node *node = tipc_node_find(net, addr);
447
448	if (!node)
449		return err;
450
451	if (bearer_id >= MAX_BEARERS)
452		goto exit;
453
454	tipc_node_lock(node);
455	link = node->links[bearer_id];
456	if (link) {
457		strncpy(linkname, link->name, len);
458		err = 0;
459	}
460exit:
461	tipc_node_unlock(node);
462	tipc_node_put(node);
463	return err;
464}
465
466void tipc_node_unlock(struct tipc_node *node)
467{
468	struct net *net = node->net;
469	u32 addr = 0;
470	u32 flags = node->action_flags;
471	u32 link_id = 0;
472	struct list_head *publ_list;
473	struct sk_buff_head *inputq = node->inputq;
474	struct sk_buff_head *namedq;
475
476	if (likely(!flags || (flags == TIPC_MSG_EVT))) {
477		node->action_flags = 0;
478		spin_unlock_bh(&node->lock);
479		if (flags == TIPC_MSG_EVT)
480			tipc_sk_rcv(net, inputq);
481		return;
482	}
483
484	addr = node->addr;
485	link_id = node->link_id;
486	namedq = node->namedq;
487	publ_list = &node->publ_list;
488
489	node->action_flags &= ~(TIPC_MSG_EVT |
490				TIPC_NOTIFY_NODE_DOWN | TIPC_NOTIFY_NODE_UP |
491				TIPC_NOTIFY_LINK_DOWN | TIPC_NOTIFY_LINK_UP |
492				TIPC_WAKEUP_BCAST_USERS | TIPC_BCAST_MSG_EVT |
493				TIPC_NAMED_MSG_EVT | TIPC_BCAST_RESET);
494
495	spin_unlock_bh(&node->lock);
496
497	if (flags & TIPC_NOTIFY_NODE_DOWN)
498		tipc_publ_notify(net, publ_list, addr);
499
500	if (flags & TIPC_WAKEUP_BCAST_USERS)
501		tipc_bclink_wakeup_users(net);
502
503	if (flags & TIPC_NOTIFY_NODE_UP)
504		tipc_named_node_up(net, addr);
505
506	if (flags & TIPC_NOTIFY_LINK_UP)
507		tipc_nametbl_publish(net, TIPC_LINK_STATE, addr, addr,
508				     TIPC_NODE_SCOPE, link_id, addr);
509
510	if (flags & TIPC_NOTIFY_LINK_DOWN)
511		tipc_nametbl_withdraw(net, TIPC_LINK_STATE, addr,
512				      link_id, addr);
513
514	if (flags & TIPC_MSG_EVT)
515		tipc_sk_rcv(net, inputq);
516
517	if (flags & TIPC_NAMED_MSG_EVT)
518		tipc_named_rcv(net, namedq);
519
520	if (flags & TIPC_BCAST_MSG_EVT)
521		tipc_bclink_input(net);
522
523	if (flags & TIPC_BCAST_RESET)
524		tipc_link_reset_all(node);
525}
526
527/* Caller should hold node lock for the passed node */
528static int __tipc_nl_add_node(struct tipc_nl_msg *msg, struct tipc_node *node)
529{
530	void *hdr;
531	struct nlattr *attrs;
532
533	hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
534			  NLM_F_MULTI, TIPC_NL_NODE_GET);
535	if (!hdr)
536		return -EMSGSIZE;
537
538	attrs = nla_nest_start(msg->skb, TIPC_NLA_NODE);
539	if (!attrs)
540		goto msg_full;
541
542	if (nla_put_u32(msg->skb, TIPC_NLA_NODE_ADDR, node->addr))
543		goto attr_msg_full;
544	if (tipc_node_is_up(node))
545		if (nla_put_flag(msg->skb, TIPC_NLA_NODE_UP))
546			goto attr_msg_full;
547
548	nla_nest_end(msg->skb, attrs);
549	genlmsg_end(msg->skb, hdr);
550
551	return 0;
552
553attr_msg_full:
554	nla_nest_cancel(msg->skb, attrs);
555msg_full:
556	genlmsg_cancel(msg->skb, hdr);
557
558	return -EMSGSIZE;
559}
560
561int tipc_nl_node_dump(struct sk_buff *skb, struct netlink_callback *cb)
562{
563	int err;
564	struct net *net = sock_net(skb->sk);
565	struct tipc_net *tn = net_generic(net, tipc_net_id);
566	int done = cb->args[0];
567	int last_addr = cb->args[1];
568	struct tipc_node *node;
569	struct tipc_nl_msg msg;
570
571	if (done)
572		return 0;
573
574	msg.skb = skb;
575	msg.portid = NETLINK_CB(cb->skb).portid;
576	msg.seq = cb->nlh->nlmsg_seq;
577
578	rcu_read_lock();
579	if (last_addr) {
580		node = tipc_node_find(net, last_addr);
581		if (!node) {
582			rcu_read_unlock();
583			/* We never set seq or call nl_dump_check_consistent()
584			 * this means that setting prev_seq here will cause the
585			 * consistence check to fail in the netlink callback
586			 * handler. Resulting in the NLMSG_DONE message having
587			 * the NLM_F_DUMP_INTR flag set if the node state
588			 * changed while we released the lock.
589			 */
590			cb->prev_seq = 1;
591			return -EPIPE;
592		}
593		tipc_node_put(node);
594	}
595
596	list_for_each_entry_rcu(node, &tn->node_list, list) {
597		if (last_addr) {
598			if (node->addr == last_addr)
599				last_addr = 0;
600			else
601				continue;
602		}
603
604		tipc_node_lock(node);
605		err = __tipc_nl_add_node(&msg, node);
606		if (err) {
607			last_addr = node->addr;
608			tipc_node_unlock(node);
609			goto out;
610		}
611
612		tipc_node_unlock(node);
613	}
614	done = 1;
615out:
616	cb->args[0] = done;
617	cb->args[1] = last_addr;
618	rcu_read_unlock();
619
620	return skb->len;
621}
622