Loading...
1/*
2 * net/tipc/net.c: TIPC network routing code
3 *
4 * Copyright (c) 1995-2006, 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 "net.h"
39#include "name_distr.h"
40#include "subscr.h"
41#include "socket.h"
42#include "node.h"
43#include "bcast.h"
44#include "netlink.h"
45#include "monitor.h"
46
47/*
48 * The TIPC locking policy is designed to ensure a very fine locking
49 * granularity, permitting complete parallel access to individual
50 * port and node/link instances. The code consists of four major
51 * locking domains, each protected with their own disjunct set of locks.
52 *
53 * 1: The bearer level.
54 * RTNL lock is used to serialize the process of configuring bearer
55 * on update side, and RCU lock is applied on read side to make
56 * bearer instance valid on both paths of message transmission and
57 * reception.
58 *
59 * 2: The node and link level.
60 * All node instances are saved into two tipc_node_list and node_htable
61 * lists. The two lists are protected by node_list_lock on write side,
62 * and they are guarded with RCU lock on read side. Especially node
63 * instance is destroyed only when TIPC module is removed, and we can
64 * confirm that there has no any user who is accessing the node at the
65 * moment. Therefore, Except for iterating the two lists within RCU
66 * protection, it's no needed to hold RCU that we access node instance
67 * in other places.
68 *
69 * In addition, all members in node structure including link instances
70 * are protected by node spin lock.
71 *
72 * 3: The transport level of the protocol.
73 * This consists of the structures port, (and its user level
74 * representations, such as user_port and tipc_sock), reference and
75 * tipc_user (port.c, reg.c, socket.c).
76 *
77 * This layer has four different locks:
78 * - The tipc_port spin_lock. This is protecting each port instance
79 * from parallel data access and removal. Since we can not place
80 * this lock in the port itself, it has been placed in the
81 * corresponding reference table entry, which has the same life
82 * cycle as the module. This entry is difficult to access from
83 * outside the TIPC core, however, so a pointer to the lock has
84 * been added in the port instance, -to be used for unlocking
85 * only.
86 * - A read/write lock to protect the reference table itself (teg.c).
87 * (Nobody is using read-only access to this, so it can just as
88 * well be changed to a spin_lock)
89 * - A spin lock to protect the registry of kernel/driver users (reg.c)
90 * - A global spin_lock (tipc_port_lock), which only task is to ensure
91 * consistency where more than one port is involved in an operation,
92 * i.e., whe a port is part of a linked list of ports.
93 * There are two such lists; 'port_list', which is used for management,
94 * and 'wait_list', which is used to queue ports during congestion.
95 *
96 * 4: The name table (name_table.c, name_distr.c, subscription.c)
97 * - There is one big read/write-lock (tipc_nametbl_lock) protecting the
98 * overall name table structure. Nothing must be added/removed to
99 * this structure without holding write access to it.
100 * - There is one local spin_lock per sub_sequence, which can be seen
101 * as a sub-domain to the tipc_nametbl_lock domain. It is used only
102 * for translation operations, and is needed because a translation
103 * steps the root of the 'publication' linked list between each lookup.
104 * This is always used within the scope of a tipc_nametbl_lock(read).
105 * - A local spin_lock protecting the queue of subscriber events.
106*/
107
108struct tipc_net_work {
109 struct work_struct work;
110 struct net *net;
111 u32 addr;
112};
113
114static void tipc_net_finalize(struct net *net, u32 addr);
115
116int tipc_net_init(struct net *net, u8 *node_id, u32 addr)
117{
118 if (tipc_own_id(net)) {
119 pr_info("Cannot configure node identity twice\n");
120 return -1;
121 }
122 pr_info("Started in network mode\n");
123
124 if (node_id)
125 tipc_set_node_id(net, node_id);
126 if (addr)
127 tipc_net_finalize(net, addr);
128 return 0;
129}
130
131static void tipc_net_finalize(struct net *net, u32 addr)
132{
133 struct tipc_net *tn = tipc_net(net);
134
135 if (cmpxchg(&tn->node_addr, 0, addr))
136 return;
137 tipc_set_node_addr(net, addr);
138 tipc_named_reinit(net);
139 tipc_sk_reinit(net);
140 tipc_mon_reinit_self(net);
141 tipc_nametbl_publish(net, TIPC_CFG_SRV, addr, addr,
142 TIPC_CLUSTER_SCOPE, 0, addr);
143}
144
145static void tipc_net_finalize_work(struct work_struct *work)
146{
147 struct tipc_net_work *fwork;
148
149 fwork = container_of(work, struct tipc_net_work, work);
150 tipc_net_finalize(fwork->net, fwork->addr);
151 kfree(fwork);
152}
153
154void tipc_sched_net_finalize(struct net *net, u32 addr)
155{
156 struct tipc_net_work *fwork = kzalloc(sizeof(*fwork), GFP_ATOMIC);
157
158 if (!fwork)
159 return;
160 INIT_WORK(&fwork->work, tipc_net_finalize_work);
161 fwork->net = net;
162 fwork->addr = addr;
163 schedule_work(&fwork->work);
164}
165
166void tipc_net_stop(struct net *net)
167{
168 if (!tipc_own_id(net))
169 return;
170
171 rtnl_lock();
172 tipc_bearer_stop(net);
173 tipc_node_stop(net);
174 rtnl_unlock();
175
176 pr_info("Left network mode\n");
177}
178
179static int __tipc_nl_add_net(struct net *net, struct tipc_nl_msg *msg)
180{
181 struct tipc_net *tn = net_generic(net, tipc_net_id);
182 u64 *w0 = (u64 *)&tn->node_id[0];
183 u64 *w1 = (u64 *)&tn->node_id[8];
184 struct nlattr *attrs;
185 void *hdr;
186
187 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
188 NLM_F_MULTI, TIPC_NL_NET_GET);
189 if (!hdr)
190 return -EMSGSIZE;
191
192 attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_NET);
193 if (!attrs)
194 goto msg_full;
195
196 if (nla_put_u32(msg->skb, TIPC_NLA_NET_ID, tn->net_id))
197 goto attr_msg_full;
198 if (nla_put_u64_64bit(msg->skb, TIPC_NLA_NET_NODEID, *w0, 0))
199 goto attr_msg_full;
200 if (nla_put_u64_64bit(msg->skb, TIPC_NLA_NET_NODEID_W1, *w1, 0))
201 goto attr_msg_full;
202 nla_nest_end(msg->skb, attrs);
203 genlmsg_end(msg->skb, hdr);
204
205 return 0;
206
207attr_msg_full:
208 nla_nest_cancel(msg->skb, attrs);
209msg_full:
210 genlmsg_cancel(msg->skb, hdr);
211
212 return -EMSGSIZE;
213}
214
215int tipc_nl_net_dump(struct sk_buff *skb, struct netlink_callback *cb)
216{
217 struct net *net = sock_net(skb->sk);
218 int err;
219 int done = cb->args[0];
220 struct tipc_nl_msg msg;
221
222 if (done)
223 return 0;
224
225 msg.skb = skb;
226 msg.portid = NETLINK_CB(cb->skb).portid;
227 msg.seq = cb->nlh->nlmsg_seq;
228
229 err = __tipc_nl_add_net(net, &msg);
230 if (err)
231 goto out;
232
233 done = 1;
234out:
235 cb->args[0] = done;
236
237 return skb->len;
238}
239
240int __tipc_nl_net_set(struct sk_buff *skb, struct genl_info *info)
241{
242 struct nlattr *attrs[TIPC_NLA_NET_MAX + 1];
243 struct net *net = sock_net(skb->sk);
244 struct tipc_net *tn = tipc_net(net);
245 int err;
246
247 if (!info->attrs[TIPC_NLA_NET])
248 return -EINVAL;
249
250 err = nla_parse_nested_deprecated(attrs, TIPC_NLA_NET_MAX,
251 info->attrs[TIPC_NLA_NET],
252 tipc_nl_net_policy, info->extack);
253
254 if (err)
255 return err;
256
257 /* Can't change net id once TIPC has joined a network */
258 if (tipc_own_addr(net))
259 return -EPERM;
260
261 if (attrs[TIPC_NLA_NET_ID]) {
262 u32 val;
263
264 val = nla_get_u32(attrs[TIPC_NLA_NET_ID]);
265 if (val < 1 || val > 9999)
266 return -EINVAL;
267
268 tn->net_id = val;
269 }
270
271 if (attrs[TIPC_NLA_NET_ADDR]) {
272 u32 addr;
273
274 addr = nla_get_u32(attrs[TIPC_NLA_NET_ADDR]);
275 if (!addr)
276 return -EINVAL;
277 tn->legacy_addr_format = true;
278 tipc_net_init(net, NULL, addr);
279 }
280
281 if (attrs[TIPC_NLA_NET_NODEID]) {
282 u8 node_id[NODE_ID_LEN];
283 u64 *w0 = (u64 *)&node_id[0];
284 u64 *w1 = (u64 *)&node_id[8];
285
286 if (!attrs[TIPC_NLA_NET_NODEID_W1])
287 return -EINVAL;
288 *w0 = nla_get_u64(attrs[TIPC_NLA_NET_NODEID]);
289 *w1 = nla_get_u64(attrs[TIPC_NLA_NET_NODEID_W1]);
290 tipc_net_init(net, node_id, 0);
291 }
292 return 0;
293}
294
295int tipc_nl_net_set(struct sk_buff *skb, struct genl_info *info)
296{
297 int err;
298
299 rtnl_lock();
300 err = __tipc_nl_net_set(skb, info);
301 rtnl_unlock();
302
303 return err;
304}
305
306static int __tipc_nl_addr_legacy_get(struct net *net, struct tipc_nl_msg *msg)
307{
308 struct tipc_net *tn = tipc_net(net);
309 struct nlattr *attrs;
310 void *hdr;
311
312 hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
313 0, TIPC_NL_ADDR_LEGACY_GET);
314 if (!hdr)
315 return -EMSGSIZE;
316
317 attrs = nla_nest_start(msg->skb, TIPC_NLA_NET);
318 if (!attrs)
319 goto msg_full;
320
321 if (tn->legacy_addr_format)
322 if (nla_put_flag(msg->skb, TIPC_NLA_NET_ADDR_LEGACY))
323 goto attr_msg_full;
324
325 nla_nest_end(msg->skb, attrs);
326 genlmsg_end(msg->skb, hdr);
327
328 return 0;
329
330attr_msg_full:
331 nla_nest_cancel(msg->skb, attrs);
332msg_full:
333 genlmsg_cancel(msg->skb, hdr);
334
335 return -EMSGSIZE;
336}
337
338int tipc_nl_net_addr_legacy_get(struct sk_buff *skb, struct genl_info *info)
339{
340 struct net *net = sock_net(skb->sk);
341 struct tipc_nl_msg msg;
342 struct sk_buff *rep;
343 int err;
344
345 rep = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
346 if (!rep)
347 return -ENOMEM;
348
349 msg.skb = rep;
350 msg.portid = info->snd_portid;
351 msg.seq = info->snd_seq;
352
353 err = __tipc_nl_addr_legacy_get(net, &msg);
354 if (err) {
355 nlmsg_free(msg.skb);
356 return err;
357 }
358
359 return genlmsg_reply(msg.skb, info);
360}
1/*
2 * net/tipc/net.c: TIPC network routing code
3 *
4 * Copyright (c) 1995-2006, 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 "net.h"
39#include "name_distr.h"
40#include "subscr.h"
41#include "port.h"
42#include "node.h"
43#include "config.h"
44
45/*
46 * The TIPC locking policy is designed to ensure a very fine locking
47 * granularity, permitting complete parallel access to individual
48 * port and node/link instances. The code consists of three major
49 * locking domains, each protected with their own disjunct set of locks.
50 *
51 * 1: The routing hierarchy.
52 * Comprises the structures 'zone', 'cluster', 'node', 'link'
53 * and 'bearer'. The whole hierarchy is protected by a big
54 * read/write lock, tipc_net_lock, to enssure that nothing is added
55 * or removed while code is accessing any of these structures.
56 * This layer must not be called from the two others while they
57 * hold any of their own locks.
58 * Neither must it itself do any upcalls to the other two before
59 * it has released tipc_net_lock and other protective locks.
60 *
61 * Within the tipc_net_lock domain there are two sub-domains;'node' and
62 * 'bearer', where local write operations are permitted,
63 * provided that those are protected by individual spin_locks
64 * per instance. Code holding tipc_net_lock(read) and a node spin_lock
65 * is permitted to poke around in both the node itself and its
66 * subordinate links. I.e, it can update link counters and queues,
67 * change link state, send protocol messages, and alter the
68 * "active_links" array in the node; but it can _not_ remove a link
69 * or a node from the overall structure.
70 * Correspondingly, individual bearers may change status within a
71 * tipc_net_lock(read), protected by an individual spin_lock ber bearer
72 * instance, but it needs tipc_net_lock(write) to remove/add any bearers.
73 *
74 *
75 * 2: The transport level of the protocol.
76 * This consists of the structures port, (and its user level
77 * representations, such as user_port and tipc_sock), reference and
78 * tipc_user (port.c, reg.c, socket.c).
79 *
80 * This layer has four different locks:
81 * - The tipc_port spin_lock. This is protecting each port instance
82 * from parallel data access and removal. Since we can not place
83 * this lock in the port itself, it has been placed in the
84 * corresponding reference table entry, which has the same life
85 * cycle as the module. This entry is difficult to access from
86 * outside the TIPC core, however, so a pointer to the lock has
87 * been added in the port instance, -to be used for unlocking
88 * only.
89 * - A read/write lock to protect the reference table itself (teg.c).
90 * (Nobody is using read-only access to this, so it can just as
91 * well be changed to a spin_lock)
92 * - A spin lock to protect the registry of kernel/driver users (reg.c)
93 * - A global spin_lock (tipc_port_lock), which only task is to ensure
94 * consistency where more than one port is involved in an operation,
95 * i.e., whe a port is part of a linked list of ports.
96 * There are two such lists; 'port_list', which is used for management,
97 * and 'wait_list', which is used to queue ports during congestion.
98 *
99 * 3: The name table (name_table.c, name_distr.c, subscription.c)
100 * - There is one big read/write-lock (tipc_nametbl_lock) protecting the
101 * overall name table structure. Nothing must be added/removed to
102 * this structure without holding write access to it.
103 * - There is one local spin_lock per sub_sequence, which can be seen
104 * as a sub-domain to the tipc_nametbl_lock domain. It is used only
105 * for translation operations, and is needed because a translation
106 * steps the root of the 'publication' linked list between each lookup.
107 * This is always used within the scope of a tipc_nametbl_lock(read).
108 * - A local spin_lock protecting the queue of subscriber events.
109*/
110
111DEFINE_RWLOCK(tipc_net_lock);
112
113static void net_route_named_msg(struct sk_buff *buf)
114{
115 struct tipc_msg *msg = buf_msg(buf);
116 u32 dnode;
117 u32 dport;
118
119 if (!msg_named(msg)) {
120 buf_discard(buf);
121 return;
122 }
123
124 dnode = addr_domain(msg_lookup_scope(msg));
125 dport = tipc_nametbl_translate(msg_nametype(msg), msg_nameinst(msg), &dnode);
126 if (dport) {
127 msg_set_destnode(msg, dnode);
128 msg_set_destport(msg, dport);
129 tipc_net_route_msg(buf);
130 return;
131 }
132 tipc_reject_msg(buf, TIPC_ERR_NO_NAME);
133}
134
135void tipc_net_route_msg(struct sk_buff *buf)
136{
137 struct tipc_msg *msg;
138 u32 dnode;
139
140 if (!buf)
141 return;
142 msg = buf_msg(buf);
143
144 msg_incr_reroute_cnt(msg);
145 if (msg_reroute_cnt(msg) > 6) {
146 if (msg_errcode(msg)) {
147 buf_discard(buf);
148 } else {
149 tipc_reject_msg(buf, msg_destport(msg) ?
150 TIPC_ERR_NO_PORT : TIPC_ERR_NO_NAME);
151 }
152 return;
153 }
154
155 /* Handle message for this node */
156 dnode = msg_short(msg) ? tipc_own_addr : msg_destnode(msg);
157 if (tipc_in_scope(dnode, tipc_own_addr)) {
158 if (msg_isdata(msg)) {
159 if (msg_mcast(msg))
160 tipc_port_recv_mcast(buf, NULL);
161 else if (msg_destport(msg))
162 tipc_port_recv_msg(buf);
163 else
164 net_route_named_msg(buf);
165 return;
166 }
167 switch (msg_user(msg)) {
168 case NAME_DISTRIBUTOR:
169 tipc_named_recv(buf);
170 break;
171 case CONN_MANAGER:
172 tipc_port_recv_proto_msg(buf);
173 break;
174 default:
175 buf_discard(buf);
176 }
177 return;
178 }
179
180 /* Handle message for another node */
181 skb_trim(buf, msg_size(msg));
182 tipc_link_send(buf, dnode, msg_link_selector(msg));
183}
184
185int tipc_net_start(u32 addr)
186{
187 char addr_string[16];
188 int res;
189
190 if (tipc_mode != TIPC_NODE_MODE)
191 return -ENOPROTOOPT;
192
193 tipc_subscr_stop();
194 tipc_cfg_stop();
195
196 tipc_own_addr = addr;
197 tipc_mode = TIPC_NET_MODE;
198 tipc_named_reinit();
199 tipc_port_reinit();
200
201 res = tipc_bclink_init();
202 if (res)
203 return res;
204
205 tipc_k_signal((Handler)tipc_subscr_start, 0);
206 tipc_k_signal((Handler)tipc_cfg_init, 0);
207
208 info("Started in network mode\n");
209 info("Own node address %s, network identity %u\n",
210 tipc_addr_string_fill(addr_string, tipc_own_addr), tipc_net_id);
211 return 0;
212}
213
214void tipc_net_stop(void)
215{
216 struct tipc_node *node, *t_node;
217
218 if (tipc_mode != TIPC_NET_MODE)
219 return;
220 write_lock_bh(&tipc_net_lock);
221 tipc_bearer_stop();
222 tipc_mode = TIPC_NODE_MODE;
223 tipc_bclink_stop();
224 list_for_each_entry_safe(node, t_node, &tipc_node_list, list)
225 tipc_node_delete(node);
226 write_unlock_bh(&tipc_net_lock);
227 info("Left network mode\n");
228}