net.c 10 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345
  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. #include "core.h"
  37. #include "net.h"
  38. #include "name_distr.h"
  39. #include "subscr.h"
  40. #include "socket.h"
  41. #include "node.h"
  42. #include "bcast.h"
  43. #include "link.h"
  44. #include "netlink.h"
  45. #include "monitor.h"
  46. /*
  47. * The TIPC locking policy is designed to ensure a very fine locking
  48. * granularity, permitting complete parallel access to individual
  49. * port and node/link instances. The code consists of four major
  50. * locking domains, each protected with their own disjunct set of locks.
  51. *
  52. * 1: The bearer level.
  53. * RTNL lock is used to serialize the process of configuring bearer
  54. * on update side, and RCU lock is applied on read side to make
  55. * bearer instance valid on both paths of message transmission and
  56. * reception.
  57. *
  58. * 2: The node and link level.
  59. * All node instances are saved into two tipc_node_list and node_htable
  60. * lists. The two lists are protected by node_list_lock on write side,
  61. * and they are guarded with RCU lock on read side. Especially node
  62. * instance is destroyed only when TIPC module is removed, and we can
  63. * confirm that there has no any user who is accessing the node at the
  64. * moment. Therefore, Except for iterating the two lists within RCU
  65. * protection, it's no needed to hold RCU that we access node instance
  66. * in other places.
  67. *
  68. * In addition, all members in node structure including link instances
  69. * are protected by node spin lock.
  70. *
  71. * 3: The transport level of the protocol.
  72. * This consists of the structures port, (and its user level
  73. * representations, such as user_port and tipc_sock), reference and
  74. * tipc_user (port.c, reg.c, socket.c).
  75. *
  76. * This layer has four different locks:
  77. * - The tipc_port spin_lock. This is protecting each port instance
  78. * from parallel data access and removal. Since we can not place
  79. * this lock in the port itself, it has been placed in the
  80. * corresponding reference table entry, which has the same life
  81. * cycle as the module. This entry is difficult to access from
  82. * outside the TIPC core, however, so a pointer to the lock has
  83. * been added in the port instance, -to be used for unlocking
  84. * only.
  85. * - A read/write lock to protect the reference table itself (teg.c).
  86. * (Nobody is using read-only access to this, so it can just as
  87. * well be changed to a spin_lock)
  88. * - A spin lock to protect the registry of kernel/driver users (reg.c)
  89. * - A global spin_lock (tipc_port_lock), which only task is to ensure
  90. * consistency where more than one port is involved in an operation,
  91. * i.e., when a port is part of a linked list of ports.
  92. * There are two such lists; 'port_list', which is used for management,
  93. * and 'wait_list', which is used to queue ports during congestion.
  94. *
  95. * 4: The name table (name_table.c, name_distr.c, subscription.c)
  96. * - There is one big read/write-lock (tipc_nametbl_lock) protecting the
  97. * overall name table structure. Nothing must be added/removed to
  98. * this structure without holding write access to it.
  99. * - There is one local spin_lock per sub_sequence, which can be seen
  100. * as a sub-domain to the tipc_nametbl_lock domain. It is used only
  101. * for translation operations, and is needed because a translation
  102. * steps the root of the 'publication' linked list between each lookup.
  103. * This is always used within the scope of a tipc_nametbl_lock(read).
  104. * - A local spin_lock protecting the queue of subscriber events.
  105. */
  106. static void tipc_net_finalize(struct net *net, u32 addr);
  107. int tipc_net_init(struct net *net, u8 *node_id, u32 addr)
  108. {
  109. if (tipc_own_id(net)) {
  110. pr_info("Cannot configure node identity twice\n");
  111. return -1;
  112. }
  113. pr_info("Started in network mode\n");
  114. if (node_id)
  115. tipc_set_node_id(net, node_id);
  116. if (addr)
  117. tipc_net_finalize(net, addr);
  118. return 0;
  119. }
  120. static void tipc_net_finalize(struct net *net, u32 addr)
  121. {
  122. struct tipc_net *tn = tipc_net(net);
  123. struct tipc_socket_addr sk = {0, addr};
  124. struct tipc_uaddr ua;
  125. tipc_uaddr(&ua, TIPC_SERVICE_RANGE, TIPC_CLUSTER_SCOPE,
  126. TIPC_NODE_STATE, addr, addr);
  127. if (cmpxchg(&tn->node_addr, 0, addr))
  128. return;
  129. tipc_set_node_addr(net, addr);
  130. tipc_named_reinit(net);
  131. tipc_sk_reinit(net);
  132. tipc_mon_reinit_self(net);
  133. tipc_nametbl_publish(net, &ua, &sk, addr);
  134. }
  135. void tipc_net_finalize_work(struct work_struct *work)
  136. {
  137. struct tipc_net *tn = container_of(work, struct tipc_net, work);
  138. tipc_net_finalize(tipc_link_net(tn->bcl), tn->trial_addr);
  139. }
  140. void tipc_net_stop(struct net *net)
  141. {
  142. if (!tipc_own_id(net))
  143. return;
  144. rtnl_lock();
  145. tipc_bearer_stop(net);
  146. tipc_node_stop(net);
  147. rtnl_unlock();
  148. pr_info("Left network mode\n");
  149. }
  150. static int __tipc_nl_add_net(struct net *net, struct tipc_nl_msg *msg)
  151. {
  152. struct tipc_net *tn = net_generic(net, tipc_net_id);
  153. u64 *w0 = (u64 *)&tn->node_id[0];
  154. u64 *w1 = (u64 *)&tn->node_id[8];
  155. struct nlattr *attrs;
  156. void *hdr;
  157. hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
  158. NLM_F_MULTI, TIPC_NL_NET_GET);
  159. if (!hdr)
  160. return -EMSGSIZE;
  161. attrs = nla_nest_start_noflag(msg->skb, TIPC_NLA_NET);
  162. if (!attrs)
  163. goto msg_full;
  164. if (nla_put_u32(msg->skb, TIPC_NLA_NET_ID, tn->net_id))
  165. goto attr_msg_full;
  166. if (nla_put_u64_64bit(msg->skb, TIPC_NLA_NET_NODEID, *w0, 0))
  167. goto attr_msg_full;
  168. if (nla_put_u64_64bit(msg->skb, TIPC_NLA_NET_NODEID_W1, *w1, 0))
  169. goto attr_msg_full;
  170. nla_nest_end(msg->skb, attrs);
  171. genlmsg_end(msg->skb, hdr);
  172. return 0;
  173. attr_msg_full:
  174. nla_nest_cancel(msg->skb, attrs);
  175. msg_full:
  176. genlmsg_cancel(msg->skb, hdr);
  177. return -EMSGSIZE;
  178. }
  179. int tipc_nl_net_dump(struct sk_buff *skb, struct netlink_callback *cb)
  180. {
  181. struct net *net = sock_net(skb->sk);
  182. int err;
  183. int done = cb->args[0];
  184. struct tipc_nl_msg msg;
  185. if (done)
  186. return 0;
  187. msg.skb = skb;
  188. msg.portid = NETLINK_CB(cb->skb).portid;
  189. msg.seq = cb->nlh->nlmsg_seq;
  190. err = __tipc_nl_add_net(net, &msg);
  191. if (err)
  192. goto out;
  193. done = 1;
  194. out:
  195. cb->args[0] = done;
  196. return skb->len;
  197. }
  198. int __tipc_nl_net_set(struct sk_buff *skb, struct genl_info *info)
  199. {
  200. struct nlattr *attrs[TIPC_NLA_NET_MAX + 1];
  201. struct net *net = sock_net(skb->sk);
  202. struct tipc_net *tn = tipc_net(net);
  203. int err;
  204. if (!info->attrs[TIPC_NLA_NET])
  205. return -EINVAL;
  206. err = nla_parse_nested_deprecated(attrs, TIPC_NLA_NET_MAX,
  207. info->attrs[TIPC_NLA_NET],
  208. tipc_nl_net_policy, info->extack);
  209. if (err)
  210. return err;
  211. /* Can't change net id once TIPC has joined a network */
  212. if (tipc_own_addr(net))
  213. return -EPERM;
  214. if (attrs[TIPC_NLA_NET_ID]) {
  215. u32 val;
  216. val = nla_get_u32(attrs[TIPC_NLA_NET_ID]);
  217. if (val < 1 || val > 9999)
  218. return -EINVAL;
  219. tn->net_id = val;
  220. }
  221. if (attrs[TIPC_NLA_NET_ADDR]) {
  222. u32 addr;
  223. addr = nla_get_u32(attrs[TIPC_NLA_NET_ADDR]);
  224. if (!addr)
  225. return -EINVAL;
  226. tn->legacy_addr_format = true;
  227. tipc_net_init(net, NULL, addr);
  228. }
  229. if (attrs[TIPC_NLA_NET_NODEID]) {
  230. u8 node_id[NODE_ID_LEN];
  231. u64 *w0 = (u64 *)&node_id[0];
  232. u64 *w1 = (u64 *)&node_id[8];
  233. if (!attrs[TIPC_NLA_NET_NODEID_W1])
  234. return -EINVAL;
  235. *w0 = nla_get_u64(attrs[TIPC_NLA_NET_NODEID]);
  236. *w1 = nla_get_u64(attrs[TIPC_NLA_NET_NODEID_W1]);
  237. tipc_net_init(net, node_id, 0);
  238. }
  239. return 0;
  240. }
  241. int tipc_nl_net_set(struct sk_buff *skb, struct genl_info *info)
  242. {
  243. int err;
  244. rtnl_lock();
  245. err = __tipc_nl_net_set(skb, info);
  246. rtnl_unlock();
  247. return err;
  248. }
  249. static int __tipc_nl_addr_legacy_get(struct net *net, struct tipc_nl_msg *msg)
  250. {
  251. struct tipc_net *tn = tipc_net(net);
  252. struct nlattr *attrs;
  253. void *hdr;
  254. hdr = genlmsg_put(msg->skb, msg->portid, msg->seq, &tipc_genl_family,
  255. 0, TIPC_NL_ADDR_LEGACY_GET);
  256. if (!hdr)
  257. return -EMSGSIZE;
  258. attrs = nla_nest_start(msg->skb, TIPC_NLA_NET);
  259. if (!attrs)
  260. goto msg_full;
  261. if (tn->legacy_addr_format)
  262. if (nla_put_flag(msg->skb, TIPC_NLA_NET_ADDR_LEGACY))
  263. goto attr_msg_full;
  264. nla_nest_end(msg->skb, attrs);
  265. genlmsg_end(msg->skb, hdr);
  266. return 0;
  267. attr_msg_full:
  268. nla_nest_cancel(msg->skb, attrs);
  269. msg_full:
  270. genlmsg_cancel(msg->skb, hdr);
  271. return -EMSGSIZE;
  272. }
  273. int tipc_nl_net_addr_legacy_get(struct sk_buff *skb, struct genl_info *info)
  274. {
  275. struct net *net = sock_net(skb->sk);
  276. struct tipc_nl_msg msg;
  277. struct sk_buff *rep;
  278. int err;
  279. rep = nlmsg_new(NLMSG_GOODSIZE, GFP_KERNEL);
  280. if (!rep)
  281. return -ENOMEM;
  282. msg.skb = rep;
  283. msg.portid = info->snd_portid;
  284. msg.seq = info->snd_seq;
  285. err = __tipc_nl_addr_legacy_get(net, &msg);
  286. if (err) {
  287. nlmsg_free(msg.skb);
  288. return err;
  289. }
  290. return genlmsg_reply(msg.skb, info);
  291. }