sync.c 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484
  1. /*
  2. * Copyright (c) 2005-2006 MontaVista Software, Inc.
  3. *
  4. * All rights reserved.
  5. *
  6. * Author: Steven Dake (sdake@mvista.com)
  7. *
  8. * This software licensed under BSD license, the text of which follows:
  9. *
  10. * Redistribution and use in source and binary forms, with or without
  11. * modification, are permitted provided that the following conditions are met:
  12. *
  13. * - Redistributions of source code must retain the above copyright notice,
  14. * this list of conditions and the following disclaimer.
  15. * - Redistributions in binary form must reproduce the above copyright notice,
  16. * this list of conditions and the following disclaimer in the documentation
  17. * and/or other materials provided with the distribution.
  18. * - Neither the name of the MontaVista Software, Inc. nor the names of its
  19. * contributors may be used to endorse or promote products derived from this
  20. * software without specific prior written permission.
  21. *
  22. * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
  23. * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  24. * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  25. * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE
  26. * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
  27. * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
  28. * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
  29. * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
  30. * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  31. * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF
  32. * THE POSSIBILITY OF SUCH DAMAGE.
  33. */
  34. #include <sys/types.h>
  35. #include <sys/socket.h>
  36. #include <sys/un.h>
  37. #include <sys/ioctl.h>
  38. #include <netinet/in.h>
  39. #include <sys/uio.h>
  40. #include <unistd.h>
  41. #include <fcntl.h>
  42. #include <stdlib.h>
  43. #include <stdio.h>
  44. #include <errno.h>
  45. #include <signal.h>
  46. #include <time.h>
  47. #include <unistd.h>
  48. #include <netinet/in.h>
  49. #include <arpa/inet.h>
  50. #include "../include/saAis.h"
  51. #include "main.h"
  52. #include "sync.h"
  53. #include "totempg.h"
  54. #include "totemip.h"
  55. #include "totem.h"
  56. #include "vsf.h"
  57. #include "swab.h"
  58. #include "../lcr/lcr_ifact.h"
  59. #include "print.h"
  60. #define MESSAGE_REQ_SYNC_BARRIER 0
  61. struct barrier_data {
  62. unsigned int nodeid;
  63. int completed;
  64. };
  65. static struct memb_ring_id *sync_ring_id;
  66. static int vsf_none = 0;
  67. static int (*sync_callbacks_retrieve) (int sync_id, struct sync_callbacks *callack);
  68. static struct sync_callbacks sync_callbacks;
  69. static int sync_processing = 0;
  70. static int sync_callback_loaded = 0;
  71. static void (*sync_synchronization_completed) (void);
  72. static int sync_recovery_index = 0;
  73. static void *sync_callback_token_handle = 0;
  74. static struct barrier_data barrier_data_confchg[PROCESSOR_COUNT_MAX];
  75. static int barrier_data_confchg_entries;
  76. static struct barrier_data barrier_data_process[PROCESSOR_COUNT_MAX];
  77. static struct openais_vsf_iface_ver0 *vsf_iface;
  78. static int sync_barrier_send (struct memb_ring_id *ring_id);
  79. static int sync_start_process (enum totem_callback_token_type type, void *data);
  80. static void sync_service_init (struct memb_ring_id *ring_id);
  81. static int sync_service_process (enum totem_callback_token_type type, void *data);
  82. static void sync_deliver_fn (
  83. unsigned int nodeid,
  84. struct iovec *iovec,
  85. int iov_len,
  86. int endian_conversion_required);
  87. static void sync_confchg_fn (
  88. enum totem_configuration_type configuration_type,
  89. unsigned int *member_list, int member_list_entries,
  90. unsigned int *left_list, int left_list_entries,
  91. unsigned int *joined_list, int joined_list_entries,
  92. struct memb_ring_id *ring_id);
  93. static void sync_primary_callback_fn (
  94. unsigned int *view_list,
  95. int view_list_entries,
  96. int primary_designated,
  97. struct memb_ring_id *ring_id);
  98. static struct totempg_group sync_group = {
  99. .group = "sync",
  100. .group_len = 4
  101. };
  102. static totempg_groups_handle sync_group_handle;
  103. struct req_exec_sync_barrier_start {
  104. struct req_header header;
  105. struct memb_ring_id ring_id;
  106. };
  107. /*
  108. * Send a barrier data structure
  109. */
  110. static int sync_barrier_send (struct memb_ring_id *ring_id)
  111. {
  112. struct req_exec_sync_barrier_start req_exec_sync_barrier_start;
  113. struct iovec iovec;
  114. int res;
  115. req_exec_sync_barrier_start.header.size = sizeof (struct req_exec_sync_barrier_start);
  116. req_exec_sync_barrier_start.header.id = MESSAGE_REQ_SYNC_BARRIER;
  117. memcpy (&req_exec_sync_barrier_start.ring_id, ring_id,
  118. sizeof (struct memb_ring_id));
  119. iovec.iov_base = (char *)&req_exec_sync_barrier_start;
  120. iovec.iov_len = sizeof (req_exec_sync_barrier_start);
  121. res = totempg_groups_mcast_joined (sync_group_handle, &iovec, 1, TOTEMPG_AGREED);
  122. return (res);
  123. }
  124. void sync_start_init (struct memb_ring_id *ring_id)
  125. {
  126. totempg_callback_token_create (
  127. &sync_callback_token_handle,
  128. TOTEM_CALLBACK_TOKEN_SENT,
  129. 0, /* don't delete after callback */
  130. sync_start_process,
  131. (void *)ring_id);
  132. }
  133. static void sync_service_init (struct memb_ring_id *ring_id)
  134. {
  135. sync_callbacks.sync_init ();
  136. totempg_callback_token_destroy (&sync_callback_token_handle);
  137. /*
  138. * Create the token callback for the processing
  139. */
  140. totempg_callback_token_create (
  141. &sync_callback_token_handle,
  142. TOTEM_CALLBACK_TOKEN_SENT,
  143. 0, /* don't delete after callback */
  144. sync_service_process,
  145. (void *)ring_id);
  146. }
  147. static int sync_start_process (enum totem_callback_token_type type, void *data)
  148. {
  149. int res;
  150. struct memb_ring_id *ring_id = (struct memb_ring_id *)data;
  151. res = sync_barrier_send (ring_id);
  152. if (res == 0) {
  153. /*
  154. * Delete the token callback for the barrier
  155. */
  156. totempg_callback_token_destroy (&sync_callback_token_handle);
  157. }
  158. return (0);
  159. }
  160. void sync_callbacks_load (void)
  161. {
  162. int res;
  163. // TODO rewrite this to get rid of the for (;;)
  164. for (;;) {
  165. res = sync_callbacks_retrieve (sync_recovery_index, &sync_callbacks);
  166. /*
  167. * No more service handlers have sync callbacks at this tie
  168. ` */
  169. if (res == -1) {
  170. sync_processing = 0;
  171. break;
  172. }
  173. sync_recovery_index += 1;
  174. if (sync_callbacks.sync_init != NULL) {
  175. sync_callback_loaded = 1;
  176. break;
  177. }
  178. }
  179. }
  180. static int sync_service_process (enum totem_callback_token_type type, void *data)
  181. {
  182. int res;
  183. struct memb_ring_id *ring_id = (struct memb_ring_id *)data;
  184. /*
  185. * If process operation not from this ring id, then ignore it and stop
  186. * processing
  187. */
  188. if (memcmp (ring_id, sync_ring_id, sizeof (struct memb_ring_id)) != 0) {
  189. return (0);
  190. }
  191. /*
  192. * If process returns 0, then its time to activate
  193. * and start the next service's synchronization
  194. */
  195. res = sync_callbacks.sync_process ();
  196. if (res != 0) {
  197. return (0);
  198. }
  199. /*
  200. * This sync is complete so activate and start next service sync
  201. */
  202. sync_callbacks.sync_activate ();
  203. totempg_callback_token_destroy (&sync_callback_token_handle);
  204. log_printf (LOG_LEVEL_NOTICE,"Synchronization actions done for (%s)\n",sync_callbacks.name);
  205. sync_callbacks_load();
  206. /*
  207. * if sync service found, execute it
  208. */
  209. if (sync_processing && sync_callbacks.sync_init) {
  210. sync_start_init (ring_id);
  211. }
  212. else {
  213. sync_callback_loaded = 0;
  214. }
  215. return (0);
  216. }
  217. int sync_register (
  218. int (*callbacks_retrieve) (int sync_id, struct sync_callbacks *callack),
  219. void (*synchronization_completed) (void),
  220. char *vsf_type)
  221. {
  222. unsigned int res;
  223. unsigned int vsf_handle;
  224. void *vsf_iface_p;
  225. char openais_vsf_type[1024];
  226. log_init ("SYNC");
  227. res = totempg_groups_initialize (
  228. &sync_group_handle,
  229. sync_deliver_fn,
  230. sync_confchg_fn);
  231. if (res == -1) {
  232. log_printf (LOG_LEVEL_ERROR,
  233. "Couldn't initialize groups interface.\n");
  234. return (-1);
  235. }
  236. res = totempg_groups_join (
  237. sync_group_handle,
  238. &sync_group,
  239. 1);
  240. if (res == -1) {
  241. log_printf (LOG_LEVEL_ERROR, "Couldn't join group.\n");
  242. return (-1);
  243. }
  244. if (strcmp (vsf_type, "none") == 0) {
  245. log_printf (LOG_LEVEL_NOTICE,
  246. "Not using a virtual synchrony filter.\n");
  247. vsf_none = 1;
  248. } else {
  249. vsf_none = 0;
  250. sprintf (openais_vsf_type, "openais_vsf_%s", vsf_type);
  251. res = lcr_ifact_reference (
  252. &vsf_handle,
  253. openais_vsf_type,
  254. 0,
  255. &vsf_iface_p,
  256. 0);
  257. if (res == -1) {
  258. log_printf (LOG_LEVEL_NOTICE,
  259. "Couldn't load virtual synchrony filter %s\n",
  260. vsf_type);
  261. return (-1);
  262. }
  263. log_printf (LOG_LEVEL_NOTICE,
  264. "Using virtual synchrony filter %s\n", openais_vsf_type);
  265. vsf_iface = (struct openais_vsf_iface_ver0 *)vsf_iface_p;
  266. vsf_iface->init (sync_primary_callback_fn);
  267. }
  268. sync_callbacks_retrieve = callbacks_retrieve;
  269. sync_synchronization_completed = synchronization_completed;
  270. return (0);
  271. }
  272. static void sync_primary_callback_fn (
  273. unsigned int *view_list,
  274. int view_list_entries,
  275. int primary_designated,
  276. struct memb_ring_id *ring_id)
  277. {
  278. int i;
  279. if (primary_designated) {
  280. log_printf (LOG_LEVEL_NOTICE, "This node is within the primary component and will provide service.\n");
  281. } else {
  282. log_printf (LOG_LEVEL_NOTICE, "This node is within the non-primary component and will NOT provide any services.\n");
  283. return;
  284. }
  285. /*
  286. * Execute configuration change for synchronization service
  287. */
  288. sync_processing = 1;
  289. totempg_callback_token_destroy (&sync_callback_token_handle);
  290. sync_recovery_index = 0;
  291. sync_callback_loaded = 0;
  292. memset (&barrier_data_confchg, 0, sizeof (barrier_data_confchg));
  293. for (i = 0; i < view_list_entries; i++) {
  294. barrier_data_confchg[i].nodeid = view_list[i];
  295. barrier_data_confchg[i].completed = 0;
  296. }
  297. memcpy (barrier_data_process, barrier_data_confchg,
  298. sizeof (barrier_data_confchg));
  299. barrier_data_confchg_entries = view_list_entries;
  300. sync_start_init (sync_ring_id);
  301. }
  302. static struct memb_ring_id deliver_ring_id;
  303. void sync_endian_convert (struct req_exec_sync_barrier_start *req_exec_sync_barrier_start)
  304. {
  305. totemip_copy_endian_convert(&req_exec_sync_barrier_start->ring_id.rep,
  306. &req_exec_sync_barrier_start->ring_id.rep);
  307. req_exec_sync_barrier_start->ring_id.seq = swab64 (req_exec_sync_barrier_start->ring_id.seq);
  308. }
  309. static void sync_deliver_fn (
  310. unsigned int nodeid,
  311. struct iovec *iovec,
  312. int iov_len,
  313. int endian_conversion_required)
  314. {
  315. struct req_exec_sync_barrier_start *req_exec_sync_barrier_start =
  316. (struct req_exec_sync_barrier_start *)iovec[0].iov_base;
  317. int i;
  318. if (endian_conversion_required) {
  319. sync_endian_convert (req_exec_sync_barrier_start);
  320. }
  321. int barrier_completed = 1;
  322. memcpy (&deliver_ring_id, &req_exec_sync_barrier_start->ring_id,
  323. sizeof (struct memb_ring_id));
  324. /*
  325. * Is this barrier from this configuration, if not, ignore it
  326. */
  327. if (memcmp (&req_exec_sync_barrier_start->ring_id, sync_ring_id,
  328. sizeof (struct memb_ring_id)) != 0) {
  329. return;
  330. }
  331. /*
  332. * Set completion for source_addr's address
  333. */
  334. for (i = 0; i < barrier_data_confchg_entries; i++) {
  335. if (nodeid == barrier_data_process[i].nodeid) {
  336. barrier_data_process[i].completed = 1;
  337. log_printf (LOG_LEVEL_DEBUG,"Barrier Start Recieved From %d\n", barrier_data_process[i].nodeid);
  338. break;
  339. }
  340. }
  341. /*
  342. * Test if barrier is complete
  343. */
  344. for (i = 0; i < barrier_data_confchg_entries; i++) {
  345. log_printf (LOG_LEVEL_DEBUG,"Barrier completion status for nodeid %d = %d. \n",
  346. barrier_data_process[i].nodeid,
  347. barrier_data_process[i].completed);
  348. if (barrier_data_process[i].completed == 0) {
  349. barrier_completed = 0;
  350. }
  351. }
  352. if (barrier_completed) {
  353. log_printf (LOG_LEVEL_NOTICE,
  354. "Synchronization barrier completed\n");
  355. }
  356. /*
  357. * Start synchronization if the barrier has completed
  358. */
  359. if (barrier_completed) {
  360. memcpy (barrier_data_process, barrier_data_confchg,
  361. sizeof (barrier_data_confchg));
  362. if ( sync_callback_loaded == 0 ) {
  363. sync_callbacks_load();
  364. }
  365. /*
  366. * if sync service found, execute it
  367. */
  368. if (sync_processing && sync_callbacks.sync_init) {
  369. log_printf (LOG_LEVEL_NOTICE,"Synchronization actions starting for (%s)\n",sync_callbacks.name);
  370. sync_service_init (&deliver_ring_id);
  371. }
  372. }
  373. return;
  374. }
  375. static void sync_confchg_fn (
  376. enum totem_configuration_type configuration_type,
  377. unsigned int *member_list, int member_list_entries,
  378. unsigned int *left_list, int left_list_entries,
  379. unsigned int *joined_list, int joined_list_entries,
  380. struct memb_ring_id *ring_id)
  381. {
  382. sync_ring_id = ring_id;
  383. /*
  384. * If no virtual synchrony filter configured, then start
  385. * synchronization process
  386. */
  387. if (vsf_none == 1) {
  388. sync_primary_callback_fn (
  389. member_list,
  390. member_list_entries,
  391. 1,
  392. ring_id);
  393. }
  394. }
  395. int sync_in_process (void)
  396. {
  397. return (sync_processing);
  398. }
  399. int sync_primary_designated (void)
  400. {
  401. if (vsf_none == 1) {
  402. return (1);
  403. } else {
  404. return (vsf_iface->primary());
  405. }
  406. }