bcast.c 23 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623624625626627628629630631632633634635636637638639640641642643644645646647648649650651652653654655656657658659660661662663664665666667668669670671672673674675676677678679680681682683684685686687688689690691692693694695696697698699700701702703704705706707708709710711712713714715716717718719720721722723724725726727728729730731732733734735736737738739740741742743744745746747748749750751752753754755756757758759760761762763764765766767768769770771772773774775776777778779780781782783784785786787788789790791792793794795796797798799800801802803804805806807808809810811812813814815816817818819820821822823824825826827828829830831832833834835836837838839840841842843844845846847848849850851852853854855856857858859860861862863864865866867868869870871872873874875876877878879880881882883884885886887888889890891892893894895896897898899900901902903904905906907908909910911912913914
  1. /*
  2. * net/tipc/bcast.c: TIPC broadcast code
  3. *
  4. * Copyright (c) 2004-2006, Ericsson AB
  5. * Copyright (c) 2004, Intel Corporation.
  6. * Copyright (c) 2005, 2010-2011, Wind River Systems
  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. #include "core.h"
  38. #include "link.h"
  39. #include "port.h"
  40. #include "bcast.h"
  41. #include "name_distr.h"
  42. #define MAX_PKT_DEFAULT_MCAST 1500 /* bcast link max packet size (fixed) */
  43. #define BCLINK_WIN_DEFAULT 20 /* bcast link window size (default) */
  44. /**
  45. * struct tipc_bcbearer_pair - a pair of bearers used by broadcast link
  46. * @primary: pointer to primary bearer
  47. * @secondary: pointer to secondary bearer
  48. *
  49. * Bearers must have same priority and same set of reachable destinations
  50. * to be paired.
  51. */
  52. struct tipc_bcbearer_pair {
  53. struct tipc_bearer *primary;
  54. struct tipc_bearer *secondary;
  55. };
  56. /**
  57. * struct tipc_bcbearer - bearer used by broadcast link
  58. * @bearer: (non-standard) broadcast bearer structure
  59. * @media: (non-standard) broadcast media structure
  60. * @bpairs: array of bearer pairs
  61. * @bpairs_temp: temporary array of bearer pairs used by tipc_bcbearer_sort()
  62. * @remains: temporary node map used by tipc_bcbearer_send()
  63. * @remains_new: temporary node map used tipc_bcbearer_send()
  64. *
  65. * Note: The fields labelled "temporary" are incorporated into the bearer
  66. * to avoid consuming potentially limited stack space through the use of
  67. * large local variables within multicast routines. Concurrent access is
  68. * prevented through use of the spinlock "bc_lock".
  69. */
  70. struct tipc_bcbearer {
  71. struct tipc_bearer bearer;
  72. struct tipc_media media;
  73. struct tipc_bcbearer_pair bpairs[MAX_BEARERS];
  74. struct tipc_bcbearer_pair bpairs_temp[TIPC_MAX_LINK_PRI + 1];
  75. struct tipc_node_map remains;
  76. struct tipc_node_map remains_new;
  77. };
  78. /**
  79. * struct tipc_bclink - link used for broadcast messages
  80. * @link: (non-standard) broadcast link structure
  81. * @node: (non-standard) node structure representing b'cast link's peer node
  82. * @bcast_nodes: map of broadcast-capable nodes
  83. * @retransmit_to: node that most recently requested a retransmit
  84. *
  85. * Handles sequence numbering, fragmentation, bundling, etc.
  86. */
  87. struct tipc_bclink {
  88. struct tipc_link link;
  89. struct tipc_node node;
  90. struct tipc_node_map bcast_nodes;
  91. struct tipc_node *retransmit_to;
  92. };
  93. static struct tipc_bcbearer bcast_bearer;
  94. static struct tipc_bclink bcast_link;
  95. static struct tipc_bcbearer *bcbearer = &bcast_bearer;
  96. static struct tipc_bclink *bclink = &bcast_link;
  97. static struct tipc_link *bcl = &bcast_link.link;
  98. static DEFINE_SPINLOCK(bc_lock);
  99. const char tipc_bclink_name[] = "broadcast-link";
  100. static void tipc_nmap_diff(struct tipc_node_map *nm_a,
  101. struct tipc_node_map *nm_b,
  102. struct tipc_node_map *nm_diff);
  103. static u32 bcbuf_acks(struct sk_buff *buf)
  104. {
  105. return (u32)(unsigned long)TIPC_SKB_CB(buf)->handle;
  106. }
  107. static void bcbuf_set_acks(struct sk_buff *buf, u32 acks)
  108. {
  109. TIPC_SKB_CB(buf)->handle = (void *)(unsigned long)acks;
  110. }
  111. static void bcbuf_decr_acks(struct sk_buff *buf)
  112. {
  113. bcbuf_set_acks(buf, bcbuf_acks(buf) - 1);
  114. }
  115. void tipc_bclink_add_node(u32 addr)
  116. {
  117. spin_lock_bh(&bc_lock);
  118. tipc_nmap_add(&bclink->bcast_nodes, addr);
  119. spin_unlock_bh(&bc_lock);
  120. }
  121. void tipc_bclink_remove_node(u32 addr)
  122. {
  123. spin_lock_bh(&bc_lock);
  124. tipc_nmap_remove(&bclink->bcast_nodes, addr);
  125. spin_unlock_bh(&bc_lock);
  126. }
  127. static void bclink_set_last_sent(void)
  128. {
  129. if (bcl->next_out)
  130. bcl->fsm_msg_cnt = mod(buf_seqno(bcl->next_out) - 1);
  131. else
  132. bcl->fsm_msg_cnt = mod(bcl->next_out_no - 1);
  133. }
  134. u32 tipc_bclink_get_last_sent(void)
  135. {
  136. return bcl->fsm_msg_cnt;
  137. }
  138. static void bclink_update_last_sent(struct tipc_node *node, u32 seqno)
  139. {
  140. node->bclink.last_sent = less_eq(node->bclink.last_sent, seqno) ?
  141. seqno : node->bclink.last_sent;
  142. }
  143. /**
  144. * tipc_bclink_retransmit_to - get most recent node to request retransmission
  145. *
  146. * Called with bc_lock locked
  147. */
  148. struct tipc_node *tipc_bclink_retransmit_to(void)
  149. {
  150. return bclink->retransmit_to;
  151. }
  152. /**
  153. * bclink_retransmit_pkt - retransmit broadcast packets
  154. * @after: sequence number of last packet to *not* retransmit
  155. * @to: sequence number of last packet to retransmit
  156. *
  157. * Called with bc_lock locked
  158. */
  159. static void bclink_retransmit_pkt(u32 after, u32 to)
  160. {
  161. struct sk_buff *buf;
  162. buf = bcl->first_out;
  163. while (buf && less_eq(buf_seqno(buf), after))
  164. buf = buf->next;
  165. tipc_link_retransmit(bcl, buf, mod(to - after));
  166. }
  167. /**
  168. * tipc_bclink_acknowledge - handle acknowledgement of broadcast packets
  169. * @n_ptr: node that sent acknowledgement info
  170. * @acked: broadcast sequence # that has been acknowledged
  171. *
  172. * Node is locked, bc_lock unlocked.
  173. */
  174. void tipc_bclink_acknowledge(struct tipc_node *n_ptr, u32 acked)
  175. {
  176. struct sk_buff *crs;
  177. struct sk_buff *next;
  178. unsigned int released = 0;
  179. spin_lock_bh(&bc_lock);
  180. /* Bail out if tx queue is empty (no clean up is required) */
  181. crs = bcl->first_out;
  182. if (!crs)
  183. goto exit;
  184. /* Determine which messages need to be acknowledged */
  185. if (acked == INVALID_LINK_SEQ) {
  186. /*
  187. * Contact with specified node has been lost, so need to
  188. * acknowledge sent messages only (if other nodes still exist)
  189. * or both sent and unsent messages (otherwise)
  190. */
  191. if (bclink->bcast_nodes.count)
  192. acked = bcl->fsm_msg_cnt;
  193. else
  194. acked = bcl->next_out_no;
  195. } else {
  196. /*
  197. * Bail out if specified sequence number does not correspond
  198. * to a message that has been sent and not yet acknowledged
  199. */
  200. if (less(acked, buf_seqno(crs)) ||
  201. less(bcl->fsm_msg_cnt, acked) ||
  202. less_eq(acked, n_ptr->bclink.acked))
  203. goto exit;
  204. }
  205. /* Skip over packets that node has previously acknowledged */
  206. while (crs && less_eq(buf_seqno(crs), n_ptr->bclink.acked))
  207. crs = crs->next;
  208. /* Update packets that node is now acknowledging */
  209. while (crs && less_eq(buf_seqno(crs), acked)) {
  210. next = crs->next;
  211. if (crs != bcl->next_out)
  212. bcbuf_decr_acks(crs);
  213. else {
  214. bcbuf_set_acks(crs, 0);
  215. bcl->next_out = next;
  216. bclink_set_last_sent();
  217. }
  218. if (bcbuf_acks(crs) == 0) {
  219. bcl->first_out = next;
  220. bcl->out_queue_size--;
  221. kfree_skb(crs);
  222. released = 1;
  223. }
  224. crs = next;
  225. }
  226. n_ptr->bclink.acked = acked;
  227. /* Try resolving broadcast link congestion, if necessary */
  228. if (unlikely(bcl->next_out)) {
  229. tipc_link_push_queue(bcl);
  230. bclink_set_last_sent();
  231. }
  232. if (unlikely(released && !list_empty(&bcl->waiting_ports)))
  233. tipc_link_wakeup_ports(bcl, 0);
  234. exit:
  235. spin_unlock_bh(&bc_lock);
  236. }
  237. /**
  238. * tipc_bclink_update_link_state - update broadcast link state
  239. *
  240. * tipc_net_lock and node lock set
  241. */
  242. void tipc_bclink_update_link_state(struct tipc_node *n_ptr, u32 last_sent)
  243. {
  244. struct sk_buff *buf;
  245. /* Ignore "stale" link state info */
  246. if (less_eq(last_sent, n_ptr->bclink.last_in))
  247. return;
  248. /* Update link synchronization state; quit if in sync */
  249. bclink_update_last_sent(n_ptr, last_sent);
  250. if (n_ptr->bclink.last_sent == n_ptr->bclink.last_in)
  251. return;
  252. /* Update out-of-sync state; quit if loss is still unconfirmed */
  253. if ((++n_ptr->bclink.oos_state) == 1) {
  254. if (n_ptr->bclink.deferred_size < (TIPC_MIN_LINK_WIN / 2))
  255. return;
  256. n_ptr->bclink.oos_state++;
  257. }
  258. /* Don't NACK if one has been recently sent (or seen) */
  259. if (n_ptr->bclink.oos_state & 0x1)
  260. return;
  261. /* Send NACK */
  262. buf = tipc_buf_acquire(INT_H_SIZE);
  263. if (buf) {
  264. struct tipc_msg *msg = buf_msg(buf);
  265. tipc_msg_init(msg, BCAST_PROTOCOL, STATE_MSG,
  266. INT_H_SIZE, n_ptr->addr);
  267. msg_set_non_seq(msg, 1);
  268. msg_set_mc_netid(msg, tipc_net_id);
  269. msg_set_bcast_ack(msg, n_ptr->bclink.last_in);
  270. msg_set_bcgap_after(msg, n_ptr->bclink.last_in);
  271. msg_set_bcgap_to(msg, n_ptr->bclink.deferred_head
  272. ? buf_seqno(n_ptr->bclink.deferred_head) - 1
  273. : n_ptr->bclink.last_sent);
  274. spin_lock_bh(&bc_lock);
  275. tipc_bearer_send(&bcbearer->bearer, buf, NULL);
  276. bcl->stats.sent_nacks++;
  277. spin_unlock_bh(&bc_lock);
  278. kfree_skb(buf);
  279. n_ptr->bclink.oos_state++;
  280. }
  281. }
  282. /**
  283. * bclink_peek_nack - monitor retransmission requests sent by other nodes
  284. *
  285. * Delay any upcoming NACK by this node if another node has already
  286. * requested the first message this node is going to ask for.
  287. *
  288. * Only tipc_net_lock set.
  289. */
  290. static void bclink_peek_nack(struct tipc_msg *msg)
  291. {
  292. struct tipc_node *n_ptr = tipc_node_find(msg_destnode(msg));
  293. if (unlikely(!n_ptr))
  294. return;
  295. tipc_node_lock(n_ptr);
  296. if (n_ptr->bclink.recv_permitted &&
  297. (n_ptr->bclink.last_in != n_ptr->bclink.last_sent) &&
  298. (n_ptr->bclink.last_in == msg_bcgap_after(msg)))
  299. n_ptr->bclink.oos_state = 2;
  300. tipc_node_unlock(n_ptr);
  301. }
  302. /*
  303. * tipc_bclink_send_msg - broadcast a packet to all nodes in cluster
  304. */
  305. int tipc_bclink_send_msg(struct sk_buff *buf)
  306. {
  307. int res;
  308. spin_lock_bh(&bc_lock);
  309. if (!bclink->bcast_nodes.count) {
  310. res = msg_data_sz(buf_msg(buf));
  311. kfree_skb(buf);
  312. goto exit;
  313. }
  314. res = tipc_link_send_buf(bcl, buf);
  315. if (likely(res >= 0)) {
  316. bclink_set_last_sent();
  317. bcl->stats.queue_sz_counts++;
  318. bcl->stats.accu_queue_sz += bcl->out_queue_size;
  319. }
  320. exit:
  321. spin_unlock_bh(&bc_lock);
  322. return res;
  323. }
  324. /**
  325. * bclink_accept_pkt - accept an incoming, in-sequence broadcast packet
  326. *
  327. * Called with both sending node's lock and bc_lock taken.
  328. */
  329. static void bclink_accept_pkt(struct tipc_node *node, u32 seqno)
  330. {
  331. bclink_update_last_sent(node, seqno);
  332. node->bclink.last_in = seqno;
  333. node->bclink.oos_state = 0;
  334. bcl->stats.recv_info++;
  335. /*
  336. * Unicast an ACK periodically, ensuring that
  337. * all nodes in the cluster don't ACK at the same time
  338. */
  339. if (((seqno - tipc_own_addr) % TIPC_MIN_LINK_WIN) == 0) {
  340. tipc_link_send_proto_msg(
  341. node->active_links[node->addr & 1],
  342. STATE_MSG, 0, 0, 0, 0, 0);
  343. bcl->stats.sent_acks++;
  344. }
  345. }
  346. /**
  347. * tipc_bclink_recv_pkt - receive a broadcast packet, and deliver upwards
  348. *
  349. * tipc_net_lock is read_locked, no other locks set
  350. */
  351. void tipc_bclink_recv_pkt(struct sk_buff *buf)
  352. {
  353. struct tipc_msg *msg = buf_msg(buf);
  354. struct tipc_node *node;
  355. u32 next_in;
  356. u32 seqno;
  357. int deferred;
  358. /* Screen out unwanted broadcast messages */
  359. if (msg_mc_netid(msg) != tipc_net_id)
  360. goto exit;
  361. node = tipc_node_find(msg_prevnode(msg));
  362. if (unlikely(!node))
  363. goto exit;
  364. tipc_node_lock(node);
  365. if (unlikely(!node->bclink.recv_permitted))
  366. goto unlock;
  367. /* Handle broadcast protocol message */
  368. if (unlikely(msg_user(msg) == BCAST_PROTOCOL)) {
  369. if (msg_type(msg) != STATE_MSG)
  370. goto unlock;
  371. if (msg_destnode(msg) == tipc_own_addr) {
  372. tipc_bclink_acknowledge(node, msg_bcast_ack(msg));
  373. tipc_node_unlock(node);
  374. spin_lock_bh(&bc_lock);
  375. bcl->stats.recv_nacks++;
  376. bclink->retransmit_to = node;
  377. bclink_retransmit_pkt(msg_bcgap_after(msg),
  378. msg_bcgap_to(msg));
  379. spin_unlock_bh(&bc_lock);
  380. } else {
  381. tipc_node_unlock(node);
  382. bclink_peek_nack(msg);
  383. }
  384. goto exit;
  385. }
  386. /* Handle in-sequence broadcast message */
  387. seqno = msg_seqno(msg);
  388. next_in = mod(node->bclink.last_in + 1);
  389. if (likely(seqno == next_in)) {
  390. receive:
  391. /* Deliver message to destination */
  392. if (likely(msg_isdata(msg))) {
  393. spin_lock_bh(&bc_lock);
  394. bclink_accept_pkt(node, seqno);
  395. spin_unlock_bh(&bc_lock);
  396. tipc_node_unlock(node);
  397. if (likely(msg_mcast(msg)))
  398. tipc_port_recv_mcast(buf, NULL);
  399. else
  400. kfree_skb(buf);
  401. } else if (msg_user(msg) == MSG_BUNDLER) {
  402. spin_lock_bh(&bc_lock);
  403. bclink_accept_pkt(node, seqno);
  404. bcl->stats.recv_bundles++;
  405. bcl->stats.recv_bundled += msg_msgcnt(msg);
  406. spin_unlock_bh(&bc_lock);
  407. tipc_node_unlock(node);
  408. tipc_link_recv_bundle(buf);
  409. } else if (msg_user(msg) == MSG_FRAGMENTER) {
  410. int ret;
  411. ret = tipc_link_recv_fragment(&node->bclink.reasm_head,
  412. &node->bclink.reasm_tail,
  413. &buf);
  414. if (ret == LINK_REASM_ERROR)
  415. goto unlock;
  416. spin_lock_bh(&bc_lock);
  417. bclink_accept_pkt(node, seqno);
  418. bcl->stats.recv_fragments++;
  419. if (ret == LINK_REASM_COMPLETE) {
  420. bcl->stats.recv_fragmented++;
  421. /* Point msg to inner header */
  422. msg = buf_msg(buf);
  423. spin_unlock_bh(&bc_lock);
  424. goto receive;
  425. }
  426. spin_unlock_bh(&bc_lock);
  427. tipc_node_unlock(node);
  428. } else if (msg_user(msg) == NAME_DISTRIBUTOR) {
  429. spin_lock_bh(&bc_lock);
  430. bclink_accept_pkt(node, seqno);
  431. spin_unlock_bh(&bc_lock);
  432. tipc_node_unlock(node);
  433. tipc_named_recv(buf);
  434. } else {
  435. spin_lock_bh(&bc_lock);
  436. bclink_accept_pkt(node, seqno);
  437. spin_unlock_bh(&bc_lock);
  438. tipc_node_unlock(node);
  439. kfree_skb(buf);
  440. }
  441. buf = NULL;
  442. /* Determine new synchronization state */
  443. tipc_node_lock(node);
  444. if (unlikely(!tipc_node_is_up(node)))
  445. goto unlock;
  446. if (node->bclink.last_in == node->bclink.last_sent)
  447. goto unlock;
  448. if (!node->bclink.deferred_head) {
  449. node->bclink.oos_state = 1;
  450. goto unlock;
  451. }
  452. msg = buf_msg(node->bclink.deferred_head);
  453. seqno = msg_seqno(msg);
  454. next_in = mod(next_in + 1);
  455. if (seqno != next_in)
  456. goto unlock;
  457. /* Take in-sequence message from deferred queue & deliver it */
  458. buf = node->bclink.deferred_head;
  459. node->bclink.deferred_head = buf->next;
  460. node->bclink.deferred_size--;
  461. goto receive;
  462. }
  463. /* Handle out-of-sequence broadcast message */
  464. if (less(next_in, seqno)) {
  465. deferred = tipc_link_defer_pkt(&node->bclink.deferred_head,
  466. &node->bclink.deferred_tail,
  467. buf);
  468. node->bclink.deferred_size += deferred;
  469. bclink_update_last_sent(node, seqno);
  470. buf = NULL;
  471. } else
  472. deferred = 0;
  473. spin_lock_bh(&bc_lock);
  474. if (deferred)
  475. bcl->stats.deferred_recv++;
  476. else
  477. bcl->stats.duplicates++;
  478. spin_unlock_bh(&bc_lock);
  479. unlock:
  480. tipc_node_unlock(node);
  481. exit:
  482. kfree_skb(buf);
  483. }
  484. u32 tipc_bclink_acks_missing(struct tipc_node *n_ptr)
  485. {
  486. return (n_ptr->bclink.recv_permitted &&
  487. (tipc_bclink_get_last_sent() != n_ptr->bclink.acked));
  488. }
  489. /**
  490. * tipc_bcbearer_send - send a packet through the broadcast pseudo-bearer
  491. *
  492. * Send packet over as many bearers as necessary to reach all nodes
  493. * that have joined the broadcast link.
  494. *
  495. * Returns 0 (packet sent successfully) under all circumstances,
  496. * since the broadcast link's pseudo-bearer never blocks
  497. */
  498. static int tipc_bcbearer_send(struct sk_buff *buf, struct tipc_bearer *unused1,
  499. struct tipc_media_addr *unused2)
  500. {
  501. int bp_index;
  502. /* Prepare broadcast link message for reliable transmission,
  503. * if first time trying to send it;
  504. * preparation is skipped for broadcast link protocol messages
  505. * since they are sent in an unreliable manner and don't need it
  506. */
  507. if (likely(!msg_non_seq(buf_msg(buf)))) {
  508. struct tipc_msg *msg;
  509. bcbuf_set_acks(buf, bclink->bcast_nodes.count);
  510. msg = buf_msg(buf);
  511. msg_set_non_seq(msg, 1);
  512. msg_set_mc_netid(msg, tipc_net_id);
  513. bcl->stats.sent_info++;
  514. if (WARN_ON(!bclink->bcast_nodes.count)) {
  515. dump_stack();
  516. return 0;
  517. }
  518. }
  519. /* Send buffer over bearers until all targets reached */
  520. bcbearer->remains = bclink->bcast_nodes;
  521. for (bp_index = 0; bp_index < MAX_BEARERS; bp_index++) {
  522. struct tipc_bearer *p = bcbearer->bpairs[bp_index].primary;
  523. struct tipc_bearer *s = bcbearer->bpairs[bp_index].secondary;
  524. struct tipc_bearer *b = p;
  525. struct sk_buff *tbuf;
  526. if (!p)
  527. break; /* No more bearers to try */
  528. if (tipc_bearer_blocked(p)) {
  529. if (!s || tipc_bearer_blocked(s))
  530. continue; /* Can't use either bearer */
  531. b = s;
  532. }
  533. tipc_nmap_diff(&bcbearer->remains, &b->nodes,
  534. &bcbearer->remains_new);
  535. if (bcbearer->remains_new.count == bcbearer->remains.count)
  536. continue; /* Nothing added by bearer pair */
  537. if (bp_index == 0) {
  538. /* Use original buffer for first bearer */
  539. tipc_bearer_send(b, buf, &b->bcast_addr);
  540. } else {
  541. /* Avoid concurrent buffer access */
  542. tbuf = pskb_copy(buf, GFP_ATOMIC);
  543. if (!tbuf)
  544. break;
  545. tipc_bearer_send(b, tbuf, &b->bcast_addr);
  546. kfree_skb(tbuf); /* Bearer keeps a clone */
  547. }
  548. /* Swap bearers for next packet */
  549. if (s) {
  550. bcbearer->bpairs[bp_index].primary = s;
  551. bcbearer->bpairs[bp_index].secondary = p;
  552. }
  553. if (bcbearer->remains_new.count == 0)
  554. break; /* All targets reached */
  555. bcbearer->remains = bcbearer->remains_new;
  556. }
  557. return 0;
  558. }
  559. /**
  560. * tipc_bcbearer_sort - create sets of bearer pairs used by broadcast bearer
  561. */
  562. void tipc_bcbearer_sort(void)
  563. {
  564. struct tipc_bcbearer_pair *bp_temp = bcbearer->bpairs_temp;
  565. struct tipc_bcbearer_pair *bp_curr;
  566. int b_index;
  567. int pri;
  568. spin_lock_bh(&bc_lock);
  569. /* Group bearers by priority (can assume max of two per priority) */
  570. memset(bp_temp, 0, sizeof(bcbearer->bpairs_temp));
  571. for (b_index = 0; b_index < MAX_BEARERS; b_index++) {
  572. struct tipc_bearer *b = &tipc_bearers[b_index];
  573. if (!b->active || !b->nodes.count)
  574. continue;
  575. if (!bp_temp[b->priority].primary)
  576. bp_temp[b->priority].primary = b;
  577. else
  578. bp_temp[b->priority].secondary = b;
  579. }
  580. /* Create array of bearer pairs for broadcasting */
  581. bp_curr = bcbearer->bpairs;
  582. memset(bcbearer->bpairs, 0, sizeof(bcbearer->bpairs));
  583. for (pri = TIPC_MAX_LINK_PRI; pri >= 0; pri--) {
  584. if (!bp_temp[pri].primary)
  585. continue;
  586. bp_curr->primary = bp_temp[pri].primary;
  587. if (bp_temp[pri].secondary) {
  588. if (tipc_nmap_equal(&bp_temp[pri].primary->nodes,
  589. &bp_temp[pri].secondary->nodes)) {
  590. bp_curr->secondary = bp_temp[pri].secondary;
  591. } else {
  592. bp_curr++;
  593. bp_curr->primary = bp_temp[pri].secondary;
  594. }
  595. }
  596. bp_curr++;
  597. }
  598. spin_unlock_bh(&bc_lock);
  599. }
  600. int tipc_bclink_stats(char *buf, const u32 buf_size)
  601. {
  602. int ret;
  603. struct tipc_stats *s;
  604. if (!bcl)
  605. return 0;
  606. spin_lock_bh(&bc_lock);
  607. s = &bcl->stats;
  608. ret = tipc_snprintf(buf, buf_size, "Link <%s>\n"
  609. " Window:%u packets\n",
  610. bcl->name, bcl->queue_limit[0]);
  611. ret += tipc_snprintf(buf + ret, buf_size - ret,
  612. " RX packets:%u fragments:%u/%u bundles:%u/%u\n",
  613. s->recv_info, s->recv_fragments,
  614. s->recv_fragmented, s->recv_bundles,
  615. s->recv_bundled);
  616. ret += tipc_snprintf(buf + ret, buf_size - ret,
  617. " TX packets:%u fragments:%u/%u bundles:%u/%u\n",
  618. s->sent_info, s->sent_fragments,
  619. s->sent_fragmented, s->sent_bundles,
  620. s->sent_bundled);
  621. ret += tipc_snprintf(buf + ret, buf_size - ret,
  622. " RX naks:%u defs:%u dups:%u\n",
  623. s->recv_nacks, s->deferred_recv, s->duplicates);
  624. ret += tipc_snprintf(buf + ret, buf_size - ret,
  625. " TX naks:%u acks:%u dups:%u\n",
  626. s->sent_nacks, s->sent_acks, s->retransmitted);
  627. ret += tipc_snprintf(buf + ret, buf_size - ret,
  628. " Congestion link:%u Send queue max:%u avg:%u\n",
  629. s->link_congs, s->max_queue_sz,
  630. s->queue_sz_counts ?
  631. (s->accu_queue_sz / s->queue_sz_counts) : 0);
  632. spin_unlock_bh(&bc_lock);
  633. return ret;
  634. }
  635. int tipc_bclink_reset_stats(void)
  636. {
  637. if (!bcl)
  638. return -ENOPROTOOPT;
  639. spin_lock_bh(&bc_lock);
  640. memset(&bcl->stats, 0, sizeof(bcl->stats));
  641. spin_unlock_bh(&bc_lock);
  642. return 0;
  643. }
  644. int tipc_bclink_set_queue_limits(u32 limit)
  645. {
  646. if (!bcl)
  647. return -ENOPROTOOPT;
  648. if ((limit < TIPC_MIN_LINK_WIN) || (limit > TIPC_MAX_LINK_WIN))
  649. return -EINVAL;
  650. spin_lock_bh(&bc_lock);
  651. tipc_link_set_queue_limits(bcl, limit);
  652. spin_unlock_bh(&bc_lock);
  653. return 0;
  654. }
  655. void tipc_bclink_init(void)
  656. {
  657. bcbearer->bearer.media = &bcbearer->media;
  658. bcbearer->media.send_msg = tipc_bcbearer_send;
  659. sprintf(bcbearer->media.name, "tipc-broadcast");
  660. INIT_LIST_HEAD(&bcl->waiting_ports);
  661. bcl->next_out_no = 1;
  662. spin_lock_init(&bclink->node.lock);
  663. bcl->owner = &bclink->node;
  664. bcl->max_pkt = MAX_PKT_DEFAULT_MCAST;
  665. tipc_link_set_queue_limits(bcl, BCLINK_WIN_DEFAULT);
  666. spin_lock_init(&bcbearer->bearer.lock);
  667. bcl->b_ptr = &bcbearer->bearer;
  668. bcl->state = WORKING_WORKING;
  669. strlcpy(bcl->name, tipc_bclink_name, TIPC_MAX_LINK_NAME);
  670. }
  671. void tipc_bclink_stop(void)
  672. {
  673. spin_lock_bh(&bc_lock);
  674. tipc_link_stop(bcl);
  675. spin_unlock_bh(&bc_lock);
  676. memset(bclink, 0, sizeof(*bclink));
  677. memset(bcbearer, 0, sizeof(*bcbearer));
  678. }
  679. /**
  680. * tipc_nmap_add - add a node to a node map
  681. */
  682. void tipc_nmap_add(struct tipc_node_map *nm_ptr, u32 node)
  683. {
  684. int n = tipc_node(node);
  685. int w = n / WSIZE;
  686. u32 mask = (1 << (n % WSIZE));
  687. if ((nm_ptr->map[w] & mask) == 0) {
  688. nm_ptr->count++;
  689. nm_ptr->map[w] |= mask;
  690. }
  691. }
  692. /**
  693. * tipc_nmap_remove - remove a node from a node map
  694. */
  695. void tipc_nmap_remove(struct tipc_node_map *nm_ptr, u32 node)
  696. {
  697. int n = tipc_node(node);
  698. int w = n / WSIZE;
  699. u32 mask = (1 << (n % WSIZE));
  700. if ((nm_ptr->map[w] & mask) != 0) {
  701. nm_ptr->map[w] &= ~mask;
  702. nm_ptr->count--;
  703. }
  704. }
  705. /**
  706. * tipc_nmap_diff - find differences between node maps
  707. * @nm_a: input node map A
  708. * @nm_b: input node map B
  709. * @nm_diff: output node map A-B (i.e. nodes of A that are not in B)
  710. */
  711. static void tipc_nmap_diff(struct tipc_node_map *nm_a,
  712. struct tipc_node_map *nm_b,
  713. struct tipc_node_map *nm_diff)
  714. {
  715. int stop = ARRAY_SIZE(nm_a->map);
  716. int w;
  717. int b;
  718. u32 map;
  719. memset(nm_diff, 0, sizeof(*nm_diff));
  720. for (w = 0; w < stop; w++) {
  721. map = nm_a->map[w] ^ (nm_a->map[w] & nm_b->map[w]);
  722. nm_diff->map[w] = map;
  723. if (map != 0) {
  724. for (b = 0 ; b < WSIZE; b++) {
  725. if (map & (1 << b))
  726. nm_diff->count++;
  727. }
  728. }
  729. }
  730. }
  731. /**
  732. * tipc_port_list_add - add a port to a port list, ensuring no duplicates
  733. */
  734. void tipc_port_list_add(struct tipc_port_list *pl_ptr, u32 port)
  735. {
  736. struct tipc_port_list *item = pl_ptr;
  737. int i;
  738. int item_sz = PLSIZE;
  739. int cnt = pl_ptr->count;
  740. for (; ; cnt -= item_sz, item = item->next) {
  741. if (cnt < PLSIZE)
  742. item_sz = cnt;
  743. for (i = 0; i < item_sz; i++)
  744. if (item->ports[i] == port)
  745. return;
  746. if (i < PLSIZE) {
  747. item->ports[i] = port;
  748. pl_ptr->count++;
  749. return;
  750. }
  751. if (!item->next) {
  752. item->next = kmalloc(sizeof(*item), GFP_ATOMIC);
  753. if (!item->next) {
  754. pr_warn("Incomplete multicast delivery, no memory\n");
  755. return;
  756. }
  757. item->next->next = NULL;
  758. }
  759. }
  760. }
  761. /**
  762. * tipc_port_list_free - free dynamically created entries in port_list chain
  763. *
  764. */
  765. void tipc_port_list_free(struct tipc_port_list *pl_ptr)
  766. {
  767. struct tipc_port_list *item;
  768. struct tipc_port_list *next;
  769. for (item = pl_ptr->next; item; item = next) {
  770. next = item->next;
  771. kfree(item);
  772. }
  773. }