1/*
2 * net/tipc/name_distr.c: TIPC name distribution code
3 *
4 * Copyright (c) 2000-2006, 2014-2019, Ericsson AB
5 * Copyright (c) 2005, 2010-2011, Wind River Systems
6 * Copyright (c) 2020-2021, Red Hat Inc
7 * All rights reserved.
8 *
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions are met:
11 *
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. Neither the names of the copyright holders nor the names of its
18 * contributors may be used to endorse or promote products derived from
19 * this software without specific prior written permission.
20 *
21 * Alternatively, this software may be distributed under the terms of the
22 * GNU General Public License ("GPL") version 2 as published by the Free
23 * Software Foundation.
24 *
25 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
26 * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
29 * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
30 * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
31 * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
32 * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
33 * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
34 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
35 * POSSIBILITY OF SUCH DAMAGE.
36 */
37
38#include "core.h"
39#include "link.h"
40#include "name_distr.h"
41
42int sysctl_tipc_named_timeout __read_mostly = 2000;
43
44/**
45 * publ_to_item - add publication info to a publication message
46 * @p: publication info
47 * @i: location of item in the message
48 */
49static void publ_to_item(struct distr_item *i, struct publication *p)
50{
51 i->type = htonl(p->sr.type);
52 i->lower = htonl(p->sr.lower);
53 i->upper = htonl(p->sr.upper);
54 i->port = htonl(p->sk.ref);
55 i->key = htonl(p->key);
56}
57
58/**
59 * named_prepare_buf - allocate & initialize a publication message
60 * @net: the associated network namespace
61 * @type: message type
62 * @size: payload size
63 * @dest: destination node
64 *
65 * The buffer returned is of size INT_H_SIZE + payload size
66 */
67static struct sk_buff *named_prepare_buf(struct net *net, u32 type, u32 size,
68 u32 dest)
69{
70 struct sk_buff *buf = tipc_buf_acquire(INT_H_SIZE + size, GFP_ATOMIC);
71 u32 self = tipc_own_addr(net);
72 struct tipc_msg *msg;
73
74 if (buf != NULL) {
75 msg = buf_msg(skb: buf);
76 tipc_msg_init(own_addr: self, m: msg, NAME_DISTRIBUTOR,
77 type, INT_H_SIZE, destnode: dest);
78 msg_set_size(m: msg, INT_H_SIZE + size);
79 }
80 return buf;
81}
82
83/**
84 * tipc_named_publish - tell other nodes about a new publication by this node
85 * @net: the associated network namespace
86 * @p: the new publication
87 */
88struct sk_buff *tipc_named_publish(struct net *net, struct publication *p)
89{
90 struct name_table *nt = tipc_name_table(net);
91 struct distr_item *item;
92 struct sk_buff *skb;
93
94 if (p->scope == TIPC_NODE_SCOPE) {
95 list_add_tail_rcu(new: &p->binding_node, head: &nt->node_scope);
96 return NULL;
97 }
98 write_lock_bh(&nt->cluster_scope_lock);
99 list_add_tail(new: &p->binding_node, head: &nt->cluster_scope);
100 write_unlock_bh(&nt->cluster_scope_lock);
101 skb = named_prepare_buf(net, PUBLICATION, ITEM_SIZE, dest: 0);
102 if (!skb) {
103 pr_warn("Publication distribution failure\n");
104 return NULL;
105 }
106 msg_set_named_seqno(m: buf_msg(skb), n: nt->snd_nxt++);
107 msg_set_non_legacy(m: buf_msg(skb));
108 item = (struct distr_item *)msg_data(m: buf_msg(skb));
109 publ_to_item(i: item, p);
110 return skb;
111}
112
113/**
114 * tipc_named_withdraw - tell other nodes about a withdrawn publication by this node
115 * @net: the associated network namespace
116 * @p: the withdrawn publication
117 */
118struct sk_buff *tipc_named_withdraw(struct net *net, struct publication *p)
119{
120 struct name_table *nt = tipc_name_table(net);
121 struct distr_item *item;
122 struct sk_buff *skb;
123
124 write_lock_bh(&nt->cluster_scope_lock);
125 list_del(entry: &p->binding_node);
126 write_unlock_bh(&nt->cluster_scope_lock);
127 if (p->scope == TIPC_NODE_SCOPE)
128 return NULL;
129
130 skb = named_prepare_buf(net, WITHDRAWAL, ITEM_SIZE, dest: 0);
131 if (!skb) {
132 pr_warn("Withdrawal distribution failure\n");
133 return NULL;
134 }
135 msg_set_named_seqno(m: buf_msg(skb), n: nt->snd_nxt++);
136 msg_set_non_legacy(m: buf_msg(skb));
137 item = (struct distr_item *)msg_data(m: buf_msg(skb));
138 publ_to_item(i: item, p);
139 return skb;
140}
141
142/**
143 * named_distribute - prepare name info for bulk distribution to another node
144 * @net: the associated network namespace
145 * @list: list of messages (buffers) to be returned from this function
146 * @dnode: node to be updated
147 * @pls: linked list of publication items to be packed into buffer chain
148 * @seqno: sequence number for this message
149 */
150static void named_distribute(struct net *net, struct sk_buff_head *list,
151 u32 dnode, struct list_head *pls, u16 seqno)
152{
153 struct publication *publ;
154 struct sk_buff *skb = NULL;
155 struct distr_item *item = NULL;
156 u32 msg_dsz = ((tipc_node_get_mtu(net, addr: dnode, sel: 0, connected: false) - INT_H_SIZE) /
157 ITEM_SIZE) * ITEM_SIZE;
158 u32 msg_rem = msg_dsz;
159 struct tipc_msg *hdr;
160
161 list_for_each_entry(publ, pls, binding_node) {
162 /* Prepare next buffer: */
163 if (!skb) {
164 skb = named_prepare_buf(net, PUBLICATION, size: msg_rem,
165 dest: dnode);
166 if (!skb) {
167 pr_warn("Bulk publication failure\n");
168 return;
169 }
170 hdr = buf_msg(skb);
171 msg_set_bc_ack_invalid(m: hdr, invalid: true);
172 msg_set_bulk(m: hdr);
173 msg_set_non_legacy(m: hdr);
174 item = (struct distr_item *)msg_data(m: hdr);
175 }
176
177 /* Pack publication into message: */
178 publ_to_item(i: item, p: publ);
179 item++;
180 msg_rem -= ITEM_SIZE;
181
182 /* Append full buffer to list: */
183 if (!msg_rem) {
184 __skb_queue_tail(list, newsk: skb);
185 skb = NULL;
186 msg_rem = msg_dsz;
187 }
188 }
189 if (skb) {
190 hdr = buf_msg(skb);
191 msg_set_size(m: hdr, INT_H_SIZE + (msg_dsz - msg_rem));
192 skb_trim(skb, INT_H_SIZE + (msg_dsz - msg_rem));
193 __skb_queue_tail(list, newsk: skb);
194 }
195 hdr = buf_msg(skb: skb_peek_tail(list_: list));
196 msg_set_last_bulk(m: hdr);
197 msg_set_named_seqno(m: hdr, n: seqno);
198}
199
200/**
201 * tipc_named_node_up - tell specified node about all publications by this node
202 * @net: the associated network namespace
203 * @dnode: destination node
204 * @capabilities: peer node's capabilities
205 */
206void tipc_named_node_up(struct net *net, u32 dnode, u16 capabilities)
207{
208 struct name_table *nt = tipc_name_table(net);
209 struct tipc_net *tn = tipc_net(net);
210 struct sk_buff_head head;
211 u16 seqno;
212
213 __skb_queue_head_init(list: &head);
214 spin_lock_bh(lock: &tn->nametbl_lock);
215 if (!(capabilities & TIPC_NAMED_BCAST))
216 nt->rc_dests++;
217 seqno = nt->snd_nxt;
218 spin_unlock_bh(lock: &tn->nametbl_lock);
219
220 read_lock_bh(&nt->cluster_scope_lock);
221 named_distribute(net, list: &head, dnode, pls: &nt->cluster_scope, seqno);
222 tipc_node_xmit(net, list: &head, dnode, selector: 0);
223 read_unlock_bh(&nt->cluster_scope_lock);
224}
225
226/**
227 * tipc_publ_purge - remove publication associated with a failed node
228 * @net: the associated network namespace
229 * @p: the publication to remove
230 * @addr: failed node's address
231 *
232 * Invoked for each publication issued by a newly failed node.
233 * Removes publication structure from name table & deletes it.
234 */
235static void tipc_publ_purge(struct net *net, struct publication *p, u32 addr)
236{
237 struct tipc_net *tn = tipc_net(net);
238 struct publication *_p;
239 struct tipc_uaddr ua;
240
241 tipc_uaddr(ua: &ua, TIPC_SERVICE_RANGE, scope: p->scope, type: p->sr.type,
242 lower: p->sr.lower, upper: p->sr.upper);
243 spin_lock_bh(lock: &tn->nametbl_lock);
244 _p = tipc_nametbl_remove_publ(net, ua: &ua, sk: &p->sk, key: p->key);
245 if (_p)
246 tipc_node_unsubscribe(net, subscr: &_p->binding_node, addr);
247 spin_unlock_bh(lock: &tn->nametbl_lock);
248 if (_p)
249 kfree_rcu(_p, rcu);
250}
251
252void tipc_publ_notify(struct net *net, struct list_head *nsub_list,
253 u32 addr, u16 capabilities)
254{
255 struct name_table *nt = tipc_name_table(net);
256 struct tipc_net *tn = tipc_net(net);
257
258 struct publication *publ, *tmp;
259
260 list_for_each_entry_safe(publ, tmp, nsub_list, binding_node)
261 tipc_publ_purge(net, p: publ, addr);
262 spin_lock_bh(lock: &tn->nametbl_lock);
263 if (!(capabilities & TIPC_NAMED_BCAST))
264 nt->rc_dests--;
265 spin_unlock_bh(lock: &tn->nametbl_lock);
266}
267
268/**
269 * tipc_update_nametbl - try to process a nametable update and notify
270 * subscribers
271 * @net: the associated network namespace
272 * @i: location of item in the message
273 * @node: node address
274 * @dtype: name distributor message type
275 *
276 * tipc_nametbl_lock must be held.
277 * Return: the publication item if successful, otherwise NULL.
278 */
279static bool tipc_update_nametbl(struct net *net, struct distr_item *i,
280 u32 node, u32 dtype)
281{
282 struct publication *p = NULL;
283 struct tipc_socket_addr sk;
284 struct tipc_uaddr ua;
285 u32 key = ntohl(i->key);
286
287 tipc_uaddr(ua: &ua, TIPC_SERVICE_RANGE, scope: TIPC_CLUSTER_SCOPE,
288 ntohl(i->type), ntohl(i->lower), ntohl(i->upper));
289 sk.ref = ntohl(i->port);
290 sk.node = node;
291
292 if (dtype == PUBLICATION) {
293 p = tipc_nametbl_insert_publ(net, ua: &ua, sk: &sk, key);
294 if (p) {
295 tipc_node_subscribe(net, subscr: &p->binding_node, addr: node);
296 return true;
297 }
298 } else if (dtype == WITHDRAWAL) {
299 p = tipc_nametbl_remove_publ(net, ua: &ua, sk: &sk, key);
300 if (p) {
301 tipc_node_unsubscribe(net, subscr: &p->binding_node, addr: node);
302 kfree_rcu(p, rcu);
303 return true;
304 }
305 pr_warn_ratelimited("Failed to remove binding %u,%u from %u\n",
306 ua.sr.type, ua.sr.lower, node);
307 } else {
308 pr_warn_ratelimited("Unknown name table message received\n");
309 }
310 return false;
311}
312
313static struct sk_buff *tipc_named_dequeue(struct sk_buff_head *namedq,
314 u16 *rcv_nxt, bool *open)
315{
316 struct sk_buff *skb, *tmp;
317 struct tipc_msg *hdr;
318 u16 seqno;
319
320 spin_lock_bh(lock: &namedq->lock);
321 skb_queue_walk_safe(namedq, skb, tmp) {
322 if (unlikely(skb_linearize(skb))) {
323 __skb_unlink(skb, list: namedq);
324 kfree_skb(skb);
325 continue;
326 }
327 hdr = buf_msg(skb);
328 seqno = msg_named_seqno(m: hdr);
329 if (msg_is_last_bulk(m: hdr)) {
330 *rcv_nxt = seqno;
331 *open = true;
332 }
333
334 if (msg_is_bulk(m: hdr) || msg_is_legacy(m: hdr)) {
335 __skb_unlink(skb, list: namedq);
336 spin_unlock_bh(lock: &namedq->lock);
337 return skb;
338 }
339
340 if (*open && (*rcv_nxt == seqno)) {
341 (*rcv_nxt)++;
342 __skb_unlink(skb, list: namedq);
343 spin_unlock_bh(lock: &namedq->lock);
344 return skb;
345 }
346
347 if (less(left: seqno, right: *rcv_nxt)) {
348 __skb_unlink(skb, list: namedq);
349 kfree_skb(skb);
350 continue;
351 }
352 }
353 spin_unlock_bh(lock: &namedq->lock);
354 return NULL;
355}
356
357/**
358 * tipc_named_rcv - process name table update messages sent by another node
359 * @net: the associated network namespace
360 * @namedq: queue to receive from
361 * @rcv_nxt: store last received seqno here
362 * @open: last bulk msg was received (FIXME)
363 */
364void tipc_named_rcv(struct net *net, struct sk_buff_head *namedq,
365 u16 *rcv_nxt, bool *open)
366{
367 struct tipc_net *tn = tipc_net(net);
368 struct distr_item *item;
369 struct tipc_msg *hdr;
370 struct sk_buff *skb;
371 u32 count, node;
372
373 spin_lock_bh(lock: &tn->nametbl_lock);
374 while ((skb = tipc_named_dequeue(namedq, rcv_nxt, open))) {
375 hdr = buf_msg(skb);
376 node = msg_orignode(m: hdr);
377 item = (struct distr_item *)msg_data(m: hdr);
378 count = msg_data_sz(m: hdr) / ITEM_SIZE;
379 while (count--) {
380 tipc_update_nametbl(net, i: item, node, dtype: msg_type(m: hdr));
381 item++;
382 }
383 kfree_skb(skb);
384 }
385 spin_unlock_bh(lock: &tn->nametbl_lock);
386}
387
388/**
389 * tipc_named_reinit - re-initialize local publications
390 * @net: the associated network namespace
391 *
392 * This routine is called whenever TIPC networking is enabled.
393 * All name table entries published by this node are updated to reflect
394 * the node's new network address.
395 */
396void tipc_named_reinit(struct net *net)
397{
398 struct name_table *nt = tipc_name_table(net);
399 struct tipc_net *tn = tipc_net(net);
400 struct publication *p;
401 u32 self = tipc_own_addr(net);
402
403 spin_lock_bh(lock: &tn->nametbl_lock);
404
405 list_for_each_entry_rcu(p, &nt->node_scope, binding_node)
406 p->sk.node = self;
407 list_for_each_entry_rcu(p, &nt->cluster_scope, binding_node)
408 p->sk.node = self;
409 nt->rc_dests = 0;
410 spin_unlock_bh(lock: &tn->nametbl_lock);
411}
412

source code of linux/net/tipc/name_distr.c