xref: /linux/net/tipc/name_distr.c (revision 4fd18fc38757217c746aa063ba9e4729814dc737)
1 /*
2  * net/tipc/name_distr.c: TIPC name distribution code
3  *
4  * Copyright (c) 2000-2006, 2014, Ericsson AB
5  * Copyright (c) 2005, 2010-2011, 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 "name_distr.h"
40 
41 int sysctl_tipc_named_timeout __read_mostly = 2000;
42 
43 struct distr_queue_item {
44 	struct distr_item i;
45 	u32 dtype;
46 	u32 node;
47 	unsigned long expires;
48 	struct list_head next;
49 };
50 
51 /**
52  * publ_to_item - add publication info to a publication message
53  * @p: publication info
54  * @i: location of item in the message
55  */
56 static void publ_to_item(struct distr_item *i, struct publication *p)
57 {
58 	i->type = htonl(p->type);
59 	i->lower = htonl(p->lower);
60 	i->upper = htonl(p->upper);
61 	i->port = htonl(p->port);
62 	i->key = htonl(p->key);
63 }
64 
65 /**
66  * named_prepare_buf - allocate & initialize a publication message
67  * @net: the associated network namespace
68  * @type: message type
69  * @size: payload size
70  * @dest: destination node
71  *
72  * The buffer returned is of size INT_H_SIZE + payload size
73  */
74 static struct sk_buff *named_prepare_buf(struct net *net, u32 type, u32 size,
75 					 u32 dest)
76 {
77 	struct sk_buff *buf = tipc_buf_acquire(INT_H_SIZE + size, GFP_ATOMIC);
78 	u32 self = tipc_own_addr(net);
79 	struct tipc_msg *msg;
80 
81 	if (buf != NULL) {
82 		msg = buf_msg(buf);
83 		tipc_msg_init(self, msg, NAME_DISTRIBUTOR,
84 			      type, INT_H_SIZE, dest);
85 		msg_set_size(msg, INT_H_SIZE + size);
86 	}
87 	return buf;
88 }
89 
90 /**
91  * tipc_named_publish - tell other nodes about a new publication by this node
92  * @net: the associated network namespace
93  * @publ: the new publication
94  */
95 struct sk_buff *tipc_named_publish(struct net *net, struct publication *publ)
96 {
97 	struct name_table *nt = tipc_name_table(net);
98 	struct distr_item *item;
99 	struct sk_buff *skb;
100 
101 	if (publ->scope == TIPC_NODE_SCOPE) {
102 		list_add_tail_rcu(&publ->binding_node, &nt->node_scope);
103 		return NULL;
104 	}
105 	write_lock_bh(&nt->cluster_scope_lock);
106 	list_add_tail(&publ->binding_node, &nt->cluster_scope);
107 	write_unlock_bh(&nt->cluster_scope_lock);
108 	skb = named_prepare_buf(net, PUBLICATION, ITEM_SIZE, 0);
109 	if (!skb) {
110 		pr_warn("Publication distribution failure\n");
111 		return NULL;
112 	}
113 	msg_set_named_seqno(buf_msg(skb), nt->snd_nxt++);
114 	msg_set_non_legacy(buf_msg(skb));
115 	item = (struct distr_item *)msg_data(buf_msg(skb));
116 	publ_to_item(item, publ);
117 	return skb;
118 }
119 
120 /**
121  * tipc_named_withdraw - tell other nodes about a withdrawn publication by this node
122  * @net: the associated network namespace
123  * @publ: the withdrawn publication
124  */
125 struct sk_buff *tipc_named_withdraw(struct net *net, struct publication *publ)
126 {
127 	struct name_table *nt = tipc_name_table(net);
128 	struct distr_item *item;
129 	struct sk_buff *skb;
130 
131 	write_lock_bh(&nt->cluster_scope_lock);
132 	list_del(&publ->binding_node);
133 	write_unlock_bh(&nt->cluster_scope_lock);
134 	if (publ->scope == TIPC_NODE_SCOPE)
135 		return NULL;
136 
137 	skb = named_prepare_buf(net, WITHDRAWAL, ITEM_SIZE, 0);
138 	if (!skb) {
139 		pr_warn("Withdrawal distribution failure\n");
140 		return NULL;
141 	}
142 	msg_set_named_seqno(buf_msg(skb), nt->snd_nxt++);
143 	msg_set_non_legacy(buf_msg(skb));
144 	item = (struct distr_item *)msg_data(buf_msg(skb));
145 	publ_to_item(item, publ);
146 	return skb;
147 }
148 
149 /**
150  * named_distribute - prepare name info for bulk distribution to another node
151  * @net: the associated network namespace
152  * @list: list of messages (buffers) to be returned from this function
153  * @dnode: node to be updated
154  * @pls: linked list of publication items to be packed into buffer chain
155  * @seqno: sequence number for this message
156  */
157 static void named_distribute(struct net *net, struct sk_buff_head *list,
158 			     u32 dnode, struct list_head *pls, u16 seqno)
159 {
160 	struct publication *publ;
161 	struct sk_buff *skb = NULL;
162 	struct distr_item *item = NULL;
163 	u32 msg_dsz = ((tipc_node_get_mtu(net, dnode, 0, false) - INT_H_SIZE) /
164 			ITEM_SIZE) * ITEM_SIZE;
165 	u32 msg_rem = msg_dsz;
166 	struct tipc_msg *hdr;
167 
168 	list_for_each_entry(publ, pls, binding_node) {
169 		/* Prepare next buffer: */
170 		if (!skb) {
171 			skb = named_prepare_buf(net, PUBLICATION, msg_rem,
172 						dnode);
173 			if (!skb) {
174 				pr_warn("Bulk publication failure\n");
175 				return;
176 			}
177 			hdr = buf_msg(skb);
178 			msg_set_bc_ack_invalid(hdr, true);
179 			msg_set_bulk(hdr);
180 			msg_set_non_legacy(hdr);
181 			item = (struct distr_item *)msg_data(hdr);
182 		}
183 
184 		/* Pack publication into message: */
185 		publ_to_item(item, publ);
186 		item++;
187 		msg_rem -= ITEM_SIZE;
188 
189 		/* Append full buffer to list: */
190 		if (!msg_rem) {
191 			__skb_queue_tail(list, skb);
192 			skb = NULL;
193 			msg_rem = msg_dsz;
194 		}
195 	}
196 	if (skb) {
197 		hdr = buf_msg(skb);
198 		msg_set_size(hdr, INT_H_SIZE + (msg_dsz - msg_rem));
199 		skb_trim(skb, INT_H_SIZE + (msg_dsz - msg_rem));
200 		__skb_queue_tail(list, skb);
201 	}
202 	hdr = buf_msg(skb_peek_tail(list));
203 	msg_set_last_bulk(hdr);
204 	msg_set_named_seqno(hdr, seqno);
205 }
206 
207 /**
208  * tipc_named_node_up - tell specified node about all publications by this node
209  * @net: the associated network namespace
210  * @dnode: destination node
211  * @capabilities: peer node's capabilities
212  */
213 void tipc_named_node_up(struct net *net, u32 dnode, u16 capabilities)
214 {
215 	struct name_table *nt = tipc_name_table(net);
216 	struct tipc_net *tn = tipc_net(net);
217 	struct sk_buff_head head;
218 	u16 seqno;
219 
220 	__skb_queue_head_init(&head);
221 	spin_lock_bh(&tn->nametbl_lock);
222 	if (!(capabilities & TIPC_NAMED_BCAST))
223 		nt->rc_dests++;
224 	seqno = nt->snd_nxt;
225 	spin_unlock_bh(&tn->nametbl_lock);
226 
227 	read_lock_bh(&nt->cluster_scope_lock);
228 	named_distribute(net, &head, dnode, &nt->cluster_scope, seqno);
229 	tipc_node_xmit(net, &head, dnode, 0);
230 	read_unlock_bh(&nt->cluster_scope_lock);
231 }
232 
233 /**
234  * tipc_publ_purge - remove publication associated with a failed node
235  * @net: the associated network namespace
236  * @publ: the publication to remove
237  * @addr: failed node's address
238  *
239  * Invoked for each publication issued by a newly failed node.
240  * Removes publication structure from name table & deletes it.
241  */
242 static void tipc_publ_purge(struct net *net, struct publication *publ, u32 addr)
243 {
244 	struct tipc_net *tn = tipc_net(net);
245 	struct publication *p;
246 
247 	spin_lock_bh(&tn->nametbl_lock);
248 	p = tipc_nametbl_remove_publ(net, publ->type, publ->lower, publ->upper,
249 				     publ->node, publ->key);
250 	if (p)
251 		tipc_node_unsubscribe(net, &p->binding_node, addr);
252 	spin_unlock_bh(&tn->nametbl_lock);
253 
254 	if (p != publ) {
255 		pr_err("Unable to remove publication from failed node\n"
256 		       " (type=%u, lower=%u, node=0x%x, port=%u, key=%u)\n",
257 		       publ->type, publ->lower, publ->node, publ->port,
258 		       publ->key);
259 	}
260 
261 	if (p)
262 		kfree_rcu(p, rcu);
263 }
264 
265 void tipc_publ_notify(struct net *net, struct list_head *nsub_list,
266 		      u32 addr, u16 capabilities)
267 {
268 	struct name_table *nt = tipc_name_table(net);
269 	struct tipc_net *tn = tipc_net(net);
270 
271 	struct publication *publ, *tmp;
272 
273 	list_for_each_entry_safe(publ, tmp, nsub_list, binding_node)
274 		tipc_publ_purge(net, publ, addr);
275 	spin_lock_bh(&tn->nametbl_lock);
276 	if (!(capabilities & TIPC_NAMED_BCAST))
277 		nt->rc_dests--;
278 	spin_unlock_bh(&tn->nametbl_lock);
279 }
280 
281 /**
282  * tipc_update_nametbl - try to process a nametable update and notify
283  *			 subscribers
284  * @net: the associated network namespace
285  * @i: location of item in the message
286  * @node: node address
287  * @dtype: name distributor message type
288  *
289  * tipc_nametbl_lock must be held.
290  * Return: the publication item if successful, otherwise NULL.
291  */
292 static bool tipc_update_nametbl(struct net *net, struct distr_item *i,
293 				u32 node, u32 dtype)
294 {
295 	struct publication *p = NULL;
296 	u32 lower = ntohl(i->lower);
297 	u32 upper = ntohl(i->upper);
298 	u32 type = ntohl(i->type);
299 	u32 port = ntohl(i->port);
300 	u32 key = ntohl(i->key);
301 
302 	if (dtype == PUBLICATION) {
303 		p = tipc_nametbl_insert_publ(net, type, lower, upper,
304 					     TIPC_CLUSTER_SCOPE, node,
305 					     port, key);
306 		if (p) {
307 			tipc_node_subscribe(net, &p->binding_node, node);
308 			return true;
309 		}
310 	} else if (dtype == WITHDRAWAL) {
311 		p = tipc_nametbl_remove_publ(net, type, lower,
312 					     upper, node, key);
313 		if (p) {
314 			tipc_node_unsubscribe(net, &p->binding_node, node);
315 			kfree_rcu(p, rcu);
316 			return true;
317 		}
318 		pr_warn_ratelimited("Failed to remove binding %u,%u from %x\n",
319 				    type, lower, node);
320 	} else {
321 		pr_warn("Unrecognized name table message received\n");
322 	}
323 	return false;
324 }
325 
326 static struct sk_buff *tipc_named_dequeue(struct sk_buff_head *namedq,
327 					  u16 *rcv_nxt, bool *open)
328 {
329 	struct sk_buff *skb, *tmp;
330 	struct tipc_msg *hdr;
331 	u16 seqno;
332 
333 	spin_lock_bh(&namedq->lock);
334 	skb_queue_walk_safe(namedq, skb, tmp) {
335 		if (unlikely(skb_linearize(skb))) {
336 			__skb_unlink(skb, namedq);
337 			kfree_skb(skb);
338 			continue;
339 		}
340 		hdr = buf_msg(skb);
341 		seqno = msg_named_seqno(hdr);
342 		if (msg_is_last_bulk(hdr)) {
343 			*rcv_nxt = seqno;
344 			*open = true;
345 		}
346 
347 		if (msg_is_bulk(hdr) || msg_is_legacy(hdr)) {
348 			__skb_unlink(skb, namedq);
349 			spin_unlock_bh(&namedq->lock);
350 			return skb;
351 		}
352 
353 		if (*open && (*rcv_nxt == seqno)) {
354 			(*rcv_nxt)++;
355 			__skb_unlink(skb, namedq);
356 			spin_unlock_bh(&namedq->lock);
357 			return skb;
358 		}
359 
360 		if (less(seqno, *rcv_nxt)) {
361 			__skb_unlink(skb, namedq);
362 			kfree_skb(skb);
363 			continue;
364 		}
365 	}
366 	spin_unlock_bh(&namedq->lock);
367 	return NULL;
368 }
369 
370 /**
371  * tipc_named_rcv - process name table update messages sent by another node
372  * @net: the associated network namespace
373  * @namedq: queue to receive from
374  * @rcv_nxt: store last received seqno here
375  * @open: last bulk msg was received (FIXME)
376  */
377 void tipc_named_rcv(struct net *net, struct sk_buff_head *namedq,
378 		    u16 *rcv_nxt, bool *open)
379 {
380 	struct tipc_net *tn = tipc_net(net);
381 	struct distr_item *item;
382 	struct tipc_msg *hdr;
383 	struct sk_buff *skb;
384 	u32 count, node;
385 
386 	spin_lock_bh(&tn->nametbl_lock);
387 	while ((skb = tipc_named_dequeue(namedq, rcv_nxt, open))) {
388 		hdr = buf_msg(skb);
389 		node = msg_orignode(hdr);
390 		item = (struct distr_item *)msg_data(hdr);
391 		count = msg_data_sz(hdr) / ITEM_SIZE;
392 		while (count--) {
393 			tipc_update_nametbl(net, item, node, msg_type(hdr));
394 			item++;
395 		}
396 		kfree_skb(skb);
397 	}
398 	spin_unlock_bh(&tn->nametbl_lock);
399 }
400 
401 /**
402  * tipc_named_reinit - re-initialize local publications
403  * @net: the associated network namespace
404  *
405  * This routine is called whenever TIPC networking is enabled.
406  * All name table entries published by this node are updated to reflect
407  * the node's new network address.
408  */
409 void tipc_named_reinit(struct net *net)
410 {
411 	struct name_table *nt = tipc_name_table(net);
412 	struct tipc_net *tn = tipc_net(net);
413 	struct publication *publ;
414 	u32 self = tipc_own_addr(net);
415 
416 	spin_lock_bh(&tn->nametbl_lock);
417 
418 	list_for_each_entry_rcu(publ, &nt->node_scope, binding_node)
419 		publ->node = self;
420 	list_for_each_entry_rcu(publ, &nt->cluster_scope, binding_node)
421 		publ->node = self;
422 	nt->rc_dests = 0;
423 	spin_unlock_bh(&tn->nametbl_lock);
424 }
425