ceph_common.c 14 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415416417418419420421422423424425426427428429430431432433434435436437438439440441442443444445446447448449450451452453454455456457458459460461462463464465466467468469470471472473474475476477478479480481482483484485486487488489490491492493494495496497498499500501502503504505506507508509510511512513514515516517518519520521522523524525526527528529530531532533534535536537538539540541542543544545546547548549550551552553554555556557558559560561562563564565566567568569570571572573574575576577578579580581582583584585586587588589590591592593594595596597598599600601602603604605606607608609610611612613614615616617618619620621622623
  1. #include <linux/ceph/ceph_debug.h>
  2. #include <linux/backing-dev.h>
  3. #include <linux/ctype.h>
  4. #include <linux/fs.h>
  5. #include <linux/inet.h>
  6. #include <linux/in6.h>
  7. #include <linux/key.h>
  8. #include <keys/ceph-type.h>
  9. #include <linux/module.h>
  10. #include <linux/mount.h>
  11. #include <linux/parser.h>
  12. #include <linux/sched.h>
  13. #include <linux/seq_file.h>
  14. #include <linux/slab.h>
  15. #include <linux/statfs.h>
  16. #include <linux/string.h>
  17. #include <linux/ceph/libceph.h>
  18. #include <linux/ceph/debugfs.h>
  19. #include <linux/ceph/decode.h>
  20. #include <linux/ceph/mon_client.h>
  21. #include <linux/ceph/auth.h>
  22. #include "crypto.h"
  23. /*
  24. * find filename portion of a path (/foo/bar/baz -> baz)
  25. */
  26. const char *ceph_file_part(const char *s, int len)
  27. {
  28. const char *e = s + len;
  29. while (e != s && *(e-1) != '/')
  30. e--;
  31. return e;
  32. }
  33. EXPORT_SYMBOL(ceph_file_part);
  34. const char *ceph_msg_type_name(int type)
  35. {
  36. switch (type) {
  37. case CEPH_MSG_SHUTDOWN: return "shutdown";
  38. case CEPH_MSG_PING: return "ping";
  39. case CEPH_MSG_AUTH: return "auth";
  40. case CEPH_MSG_AUTH_REPLY: return "auth_reply";
  41. case CEPH_MSG_MON_MAP: return "mon_map";
  42. case CEPH_MSG_MON_GET_MAP: return "mon_get_map";
  43. case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe";
  44. case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack";
  45. case CEPH_MSG_STATFS: return "statfs";
  46. case CEPH_MSG_STATFS_REPLY: return "statfs_reply";
  47. case CEPH_MSG_MDS_MAP: return "mds_map";
  48. case CEPH_MSG_CLIENT_SESSION: return "client_session";
  49. case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect";
  50. case CEPH_MSG_CLIENT_REQUEST: return "client_request";
  51. case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward";
  52. case CEPH_MSG_CLIENT_REPLY: return "client_reply";
  53. case CEPH_MSG_CLIENT_CAPS: return "client_caps";
  54. case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release";
  55. case CEPH_MSG_CLIENT_SNAP: return "client_snap";
  56. case CEPH_MSG_CLIENT_LEASE: return "client_lease";
  57. case CEPH_MSG_OSD_MAP: return "osd_map";
  58. case CEPH_MSG_OSD_OP: return "osd_op";
  59. case CEPH_MSG_OSD_OPREPLY: return "osd_opreply";
  60. case CEPH_MSG_WATCH_NOTIFY: return "watch_notify";
  61. default: return "unknown";
  62. }
  63. }
  64. EXPORT_SYMBOL(ceph_msg_type_name);
  65. /*
  66. * Initially learn our fsid, or verify an fsid matches.
  67. */
  68. int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid)
  69. {
  70. if (client->have_fsid) {
  71. if (ceph_fsid_compare(&client->fsid, fsid)) {
  72. pr_err("bad fsid, had %pU got %pU",
  73. &client->fsid, fsid);
  74. return -1;
  75. }
  76. } else {
  77. pr_info("client%lld fsid %pU\n", ceph_client_id(client), fsid);
  78. memcpy(&client->fsid, fsid, sizeof(*fsid));
  79. ceph_debugfs_client_init(client);
  80. client->have_fsid = true;
  81. }
  82. return 0;
  83. }
  84. EXPORT_SYMBOL(ceph_check_fsid);
  85. static int strcmp_null(const char *s1, const char *s2)
  86. {
  87. if (!s1 && !s2)
  88. return 0;
  89. if (s1 && !s2)
  90. return -1;
  91. if (!s1 && s2)
  92. return 1;
  93. return strcmp(s1, s2);
  94. }
  95. int ceph_compare_options(struct ceph_options *new_opt,
  96. struct ceph_client *client)
  97. {
  98. struct ceph_options *opt1 = new_opt;
  99. struct ceph_options *opt2 = client->options;
  100. int ofs = offsetof(struct ceph_options, mon_addr);
  101. int i;
  102. int ret;
  103. ret = memcmp(opt1, opt2, ofs);
  104. if (ret)
  105. return ret;
  106. ret = strcmp_null(opt1->name, opt2->name);
  107. if (ret)
  108. return ret;
  109. if (opt1->key && !opt2->key)
  110. return -1;
  111. if (!opt1->key && opt2->key)
  112. return 1;
  113. if (opt1->key && opt2->key) {
  114. if (opt1->key->type != opt2->key->type)
  115. return -1;
  116. if (opt1->key->created.tv_sec != opt2->key->created.tv_sec)
  117. return -1;
  118. if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec)
  119. return -1;
  120. if (opt1->key->len != opt2->key->len)
  121. return -1;
  122. if (opt1->key->key && !opt2->key->key)
  123. return -1;
  124. if (!opt1->key->key && opt2->key->key)
  125. return 1;
  126. if (opt1->key->key && opt2->key->key) {
  127. ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len);
  128. if (ret)
  129. return ret;
  130. }
  131. }
  132. /* any matching mon ip implies a match */
  133. for (i = 0; i < opt1->num_mon; i++) {
  134. if (ceph_monmap_contains(client->monc.monmap,
  135. &opt1->mon_addr[i]))
  136. return 0;
  137. }
  138. return -1;
  139. }
  140. EXPORT_SYMBOL(ceph_compare_options);
  141. static int parse_fsid(const char *str, struct ceph_fsid *fsid)
  142. {
  143. int i = 0;
  144. char tmp[3];
  145. int err = -EINVAL;
  146. int d;
  147. dout("parse_fsid '%s'\n", str);
  148. tmp[2] = 0;
  149. while (*str && i < 16) {
  150. if (ispunct(*str)) {
  151. str++;
  152. continue;
  153. }
  154. if (!isxdigit(str[0]) || !isxdigit(str[1]))
  155. break;
  156. tmp[0] = str[0];
  157. tmp[1] = str[1];
  158. if (sscanf(tmp, "%x", &d) < 1)
  159. break;
  160. fsid->fsid[i] = d & 0xff;
  161. i++;
  162. str += 2;
  163. }
  164. if (i == 16)
  165. err = 0;
  166. dout("parse_fsid ret %d got fsid %pU", err, fsid);
  167. return err;
  168. }
  169. /*
  170. * ceph options
  171. */
  172. enum {
  173. Opt_osdtimeout,
  174. Opt_osdkeepalivetimeout,
  175. Opt_mount_timeout,
  176. Opt_osd_idle_ttl,
  177. Opt_last_int,
  178. /* int args above */
  179. Opt_fsid,
  180. Opt_name,
  181. Opt_secret,
  182. Opt_key,
  183. Opt_ip,
  184. Opt_last_string,
  185. /* string args above */
  186. Opt_noshare,
  187. Opt_nocrc,
  188. };
  189. static match_table_t opt_tokens = {
  190. {Opt_osdtimeout, "osdtimeout=%d"},
  191. {Opt_osdkeepalivetimeout, "osdkeepalive=%d"},
  192. {Opt_mount_timeout, "mount_timeout=%d"},
  193. {Opt_osd_idle_ttl, "osd_idle_ttl=%d"},
  194. /* int args above */
  195. {Opt_fsid, "fsid=%s"},
  196. {Opt_name, "name=%s"},
  197. {Opt_secret, "secret=%s"},
  198. {Opt_key, "key=%s"},
  199. {Opt_ip, "ip=%s"},
  200. /* string args above */
  201. {Opt_noshare, "noshare"},
  202. {Opt_nocrc, "nocrc"},
  203. {-1, NULL}
  204. };
  205. void ceph_destroy_options(struct ceph_options *opt)
  206. {
  207. dout("destroy_options %p\n", opt);
  208. kfree(opt->name);
  209. if (opt->key) {
  210. ceph_crypto_key_destroy(opt->key);
  211. kfree(opt->key);
  212. }
  213. kfree(opt);
  214. }
  215. EXPORT_SYMBOL(ceph_destroy_options);
  216. /* get secret from key store */
  217. static int get_secret(struct ceph_crypto_key *dst, const char *name) {
  218. struct key *ukey;
  219. int key_err;
  220. int err = 0;
  221. struct ceph_crypto_key *ckey;
  222. ukey = request_key(&key_type_ceph, name, NULL);
  223. if (!ukey || IS_ERR(ukey)) {
  224. /* request_key errors don't map nicely to mount(2)
  225. errors; don't even try, but still printk */
  226. key_err = PTR_ERR(ukey);
  227. switch (key_err) {
  228. case -ENOKEY:
  229. pr_warning("ceph: Mount failed due to key not found: %s\n", name);
  230. break;
  231. case -EKEYEXPIRED:
  232. pr_warning("ceph: Mount failed due to expired key: %s\n", name);
  233. break;
  234. case -EKEYREVOKED:
  235. pr_warning("ceph: Mount failed due to revoked key: %s\n", name);
  236. break;
  237. default:
  238. pr_warning("ceph: Mount failed due to unknown key error"
  239. " %d: %s\n", key_err, name);
  240. }
  241. err = -EPERM;
  242. goto out;
  243. }
  244. ckey = ukey->payload.data;
  245. err = ceph_crypto_key_clone(dst, ckey);
  246. if (err)
  247. goto out_key;
  248. /* pass through, err is 0 */
  249. out_key:
  250. key_put(ukey);
  251. out:
  252. return err;
  253. }
  254. int ceph_parse_options(struct ceph_options **popt, char *options,
  255. const char *dev_name, const char *dev_name_end,
  256. int (*parse_extra_token)(char *c, void *private),
  257. void *private)
  258. {
  259. struct ceph_options *opt;
  260. const char *c;
  261. int err = -ENOMEM;
  262. substring_t argstr[MAX_OPT_ARGS];
  263. opt = kzalloc(sizeof(*opt), GFP_KERNEL);
  264. if (!opt)
  265. return err;
  266. opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr),
  267. GFP_KERNEL);
  268. if (!opt->mon_addr)
  269. goto out;
  270. dout("parse_options %p options '%s' dev_name '%s'\n", opt, options,
  271. dev_name);
  272. /* start with defaults */
  273. opt->flags = CEPH_OPT_DEFAULT;
  274. opt->osd_timeout = CEPH_OSD_TIMEOUT_DEFAULT;
  275. opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT;
  276. opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */
  277. opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */
  278. /* get mon ip(s) */
  279. /* ip1[:port1][,ip2[:port2]...] */
  280. err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr,
  281. CEPH_MAX_MON, &opt->num_mon);
  282. if (err < 0)
  283. goto out;
  284. /* parse mount options */
  285. while ((c = strsep(&options, ",")) != NULL) {
  286. int token, intval, ret;
  287. if (!*c)
  288. continue;
  289. err = -EINVAL;
  290. token = match_token((char *)c, opt_tokens, argstr);
  291. if (token < 0 && parse_extra_token) {
  292. /* extra? */
  293. err = parse_extra_token((char *)c, private);
  294. if (err < 0) {
  295. pr_err("bad option at '%s'\n", c);
  296. goto out;
  297. }
  298. continue;
  299. }
  300. if (token < Opt_last_int) {
  301. ret = match_int(&argstr[0], &intval);
  302. if (ret < 0) {
  303. pr_err("bad mount option arg (not int) "
  304. "at '%s'\n", c);
  305. continue;
  306. }
  307. dout("got int token %d val %d\n", token, intval);
  308. } else if (token > Opt_last_int && token < Opt_last_string) {
  309. dout("got string token %d val %s\n", token,
  310. argstr[0].from);
  311. } else {
  312. dout("got token %d\n", token);
  313. }
  314. switch (token) {
  315. case Opt_ip:
  316. err = ceph_parse_ips(argstr[0].from,
  317. argstr[0].to,
  318. &opt->my_addr,
  319. 1, NULL);
  320. if (err < 0)
  321. goto out;
  322. opt->flags |= CEPH_OPT_MYIP;
  323. break;
  324. case Opt_fsid:
  325. err = parse_fsid(argstr[0].from, &opt->fsid);
  326. if (err == 0)
  327. opt->flags |= CEPH_OPT_FSID;
  328. break;
  329. case Opt_name:
  330. opt->name = kstrndup(argstr[0].from,
  331. argstr[0].to-argstr[0].from,
  332. GFP_KERNEL);
  333. break;
  334. case Opt_secret:
  335. opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
  336. if (!opt->key) {
  337. err = -ENOMEM;
  338. goto out;
  339. }
  340. err = ceph_crypto_key_unarmor(opt->key, argstr[0].from);
  341. if (err < 0)
  342. goto out;
  343. break;
  344. case Opt_key:
  345. opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL);
  346. if (!opt->key) {
  347. err = -ENOMEM;
  348. goto out;
  349. }
  350. err = get_secret(opt->key, argstr[0].from);
  351. if (err < 0)
  352. goto out;
  353. break;
  354. /* misc */
  355. case Opt_osdtimeout:
  356. opt->osd_timeout = intval;
  357. break;
  358. case Opt_osdkeepalivetimeout:
  359. opt->osd_keepalive_timeout = intval;
  360. break;
  361. case Opt_osd_idle_ttl:
  362. opt->osd_idle_ttl = intval;
  363. break;
  364. case Opt_mount_timeout:
  365. opt->mount_timeout = intval;
  366. break;
  367. case Opt_noshare:
  368. opt->flags |= CEPH_OPT_NOSHARE;
  369. break;
  370. case Opt_nocrc:
  371. opt->flags |= CEPH_OPT_NOCRC;
  372. break;
  373. default:
  374. BUG_ON(token);
  375. }
  376. }
  377. /* success */
  378. *popt = opt;
  379. return 0;
  380. out:
  381. ceph_destroy_options(opt);
  382. return err;
  383. }
  384. EXPORT_SYMBOL(ceph_parse_options);
  385. u64 ceph_client_id(struct ceph_client *client)
  386. {
  387. return client->monc.auth->global_id;
  388. }
  389. EXPORT_SYMBOL(ceph_client_id);
  390. /*
  391. * create a fresh client instance
  392. */
  393. struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private)
  394. {
  395. struct ceph_client *client;
  396. int err = -ENOMEM;
  397. client = kzalloc(sizeof(*client), GFP_KERNEL);
  398. if (client == NULL)
  399. return ERR_PTR(-ENOMEM);
  400. client->private = private;
  401. client->options = opt;
  402. mutex_init(&client->mount_mutex);
  403. init_waitqueue_head(&client->auth_wq);
  404. client->auth_err = 0;
  405. client->extra_mon_dispatch = NULL;
  406. client->supported_features = CEPH_FEATURE_SUPPORTED_DEFAULT;
  407. client->required_features = CEPH_FEATURE_REQUIRED_DEFAULT;
  408. client->msgr = NULL;
  409. /* subsystems */
  410. err = ceph_monc_init(&client->monc, client);
  411. if (err < 0)
  412. goto fail;
  413. err = ceph_osdc_init(&client->osdc, client);
  414. if (err < 0)
  415. goto fail_monc;
  416. return client;
  417. fail_monc:
  418. ceph_monc_stop(&client->monc);
  419. fail:
  420. kfree(client);
  421. return ERR_PTR(err);
  422. }
  423. EXPORT_SYMBOL(ceph_create_client);
  424. void ceph_destroy_client(struct ceph_client *client)
  425. {
  426. dout("destroy_client %p\n", client);
  427. /* unmount */
  428. ceph_osdc_stop(&client->osdc);
  429. /*
  430. * make sure osd connections close out before destroying the
  431. * auth module, which is needed to free those connections'
  432. * ceph_authorizers.
  433. */
  434. ceph_msgr_flush();
  435. ceph_monc_stop(&client->monc);
  436. ceph_debugfs_client_cleanup(client);
  437. if (client->msgr)
  438. ceph_messenger_destroy(client->msgr);
  439. ceph_destroy_options(client->options);
  440. kfree(client);
  441. dout("destroy_client %p done\n", client);
  442. }
  443. EXPORT_SYMBOL(ceph_destroy_client);
  444. /*
  445. * true if we have the mon map (and have thus joined the cluster)
  446. */
  447. static int have_mon_and_osd_map(struct ceph_client *client)
  448. {
  449. return client->monc.monmap && client->monc.monmap->epoch &&
  450. client->osdc.osdmap && client->osdc.osdmap->epoch;
  451. }
  452. /*
  453. * mount: join the ceph cluster, and open root directory.
  454. */
  455. int __ceph_open_session(struct ceph_client *client, unsigned long started)
  456. {
  457. struct ceph_entity_addr *myaddr = NULL;
  458. int err;
  459. unsigned long timeout = client->options->mount_timeout * HZ;
  460. /* initialize the messenger */
  461. if (client->msgr == NULL) {
  462. if (ceph_test_opt(client, MYIP))
  463. myaddr = &client->options->my_addr;
  464. client->msgr = ceph_messenger_create(myaddr,
  465. client->supported_features,
  466. client->required_features);
  467. if (IS_ERR(client->msgr)) {
  468. client->msgr = NULL;
  469. return PTR_ERR(client->msgr);
  470. }
  471. client->msgr->nocrc = ceph_test_opt(client, NOCRC);
  472. }
  473. /* open session, and wait for mon and osd maps */
  474. err = ceph_monc_open_session(&client->monc);
  475. if (err < 0)
  476. return err;
  477. while (!have_mon_and_osd_map(client)) {
  478. err = -EIO;
  479. if (timeout && time_after_eq(jiffies, started + timeout))
  480. return err;
  481. /* wait */
  482. dout("mount waiting for mon_map\n");
  483. err = wait_event_interruptible_timeout(client->auth_wq,
  484. have_mon_and_osd_map(client) || (client->auth_err < 0),
  485. timeout);
  486. if (err == -EINTR || err == -ERESTARTSYS)
  487. return err;
  488. if (client->auth_err < 0)
  489. return client->auth_err;
  490. }
  491. return 0;
  492. }
  493. EXPORT_SYMBOL(__ceph_open_session);
  494. int ceph_open_session(struct ceph_client *client)
  495. {
  496. int ret;
  497. unsigned long started = jiffies; /* note the start time */
  498. dout("open_session start\n");
  499. mutex_lock(&client->mount_mutex);
  500. ret = __ceph_open_session(client, started);
  501. mutex_unlock(&client->mount_mutex);
  502. return ret;
  503. }
  504. EXPORT_SYMBOL(ceph_open_session);
  505. static int __init init_ceph_lib(void)
  506. {
  507. int ret = 0;
  508. ret = ceph_debugfs_init();
  509. if (ret < 0)
  510. goto out;
  511. ret = ceph_crypto_init();
  512. if (ret < 0)
  513. goto out_debugfs;
  514. ret = ceph_msgr_init();
  515. if (ret < 0)
  516. goto out_crypto;
  517. pr_info("loaded (mon/osd proto %d/%d, osdmap %d/%d %d/%d)\n",
  518. CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL,
  519. CEPH_OSDMAP_VERSION, CEPH_OSDMAP_VERSION_EXT,
  520. CEPH_OSDMAP_INC_VERSION, CEPH_OSDMAP_INC_VERSION_EXT);
  521. return 0;
  522. out_crypto:
  523. ceph_crypto_shutdown();
  524. out_debugfs:
  525. ceph_debugfs_cleanup();
  526. out:
  527. return ret;
  528. }
  529. static void __exit exit_ceph_lib(void)
  530. {
  531. dout("exit_ceph_lib\n");
  532. ceph_msgr_exit();
  533. ceph_crypto_shutdown();
  534. ceph_debugfs_cleanup();
  535. }
  536. module_init(init_ceph_lib);
  537. module_exit(exit_ceph_lib);
  538. MODULE_AUTHOR("Sage Weil <sage@newdream.net>");
  539. MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>");
  540. MODULE_AUTHOR("Patience Warnick <patience@newdream.net>");
  541. MODULE_DESCRIPTION("Ceph filesystem for Linux");
  542. MODULE_LICENSE("GPL");