1 2 #include <linux/ceph/ceph_debug.h> 3 #include <linux/backing-dev.h> 4 #include <linux/ctype.h> 5 #include <linux/fs.h> 6 #include <linux/inet.h> 7 #include <linux/in6.h> 8 #include <linux/key.h> 9 #include <keys/ceph-type.h> 10 #include <linux/module.h> 11 #include <linux/mount.h> 12 #include <linux/parser.h> 13 #include <linux/sched.h> 14 #include <linux/seq_file.h> 15 #include <linux/slab.h> 16 #include <linux/statfs.h> 17 #include <linux/string.h> 18 #include <linux/vmalloc.h> 19 #include <linux/nsproxy.h> 20 #include <net/net_namespace.h> 21 22 23 #include <linux/ceph/ceph_features.h> 24 #include <linux/ceph/libceph.h> 25 #include <linux/ceph/debugfs.h> 26 #include <linux/ceph/decode.h> 27 #include <linux/ceph/mon_client.h> 28 #include <linux/ceph/auth.h> 29 #include "crypto.h" 30 31 32 /* 33 * Module compatibility interface. For now it doesn't do anything, 34 * but its existence signals a certain level of functionality. 35 * 36 * The data buffer is used to pass information both to and from 37 * libceph. The return value indicates whether libceph determines 38 * it is compatible with the caller (from another kernel module), 39 * given the provided data. 40 * 41 * The data pointer can be null. 42 */ 43 bool libceph_compatible(void *data) 44 { 45 return true; 46 } 47 EXPORT_SYMBOL(libceph_compatible); 48 49 /* 50 * find filename portion of a path (/foo/bar/baz -> baz) 51 */ 52 const char *ceph_file_part(const char *s, int len) 53 { 54 const char *e = s + len; 55 56 while (e != s && *(e-1) != '/') 57 e--; 58 return e; 59 } 60 EXPORT_SYMBOL(ceph_file_part); 61 62 const char *ceph_msg_type_name(int type) 63 { 64 switch (type) { 65 case CEPH_MSG_SHUTDOWN: return "shutdown"; 66 case CEPH_MSG_PING: return "ping"; 67 case CEPH_MSG_AUTH: return "auth"; 68 case CEPH_MSG_AUTH_REPLY: return "auth_reply"; 69 case CEPH_MSG_MON_MAP: return "mon_map"; 70 case CEPH_MSG_MON_GET_MAP: return "mon_get_map"; 71 case CEPH_MSG_MON_SUBSCRIBE: return "mon_subscribe"; 72 case CEPH_MSG_MON_SUBSCRIBE_ACK: return "mon_subscribe_ack"; 73 case CEPH_MSG_STATFS: return "statfs"; 74 case CEPH_MSG_STATFS_REPLY: return "statfs_reply"; 75 case CEPH_MSG_MON_GET_VERSION: return "mon_get_version"; 76 case CEPH_MSG_MON_GET_VERSION_REPLY: return "mon_get_version_reply"; 77 case CEPH_MSG_MDS_MAP: return "mds_map"; 78 case CEPH_MSG_CLIENT_SESSION: return "client_session"; 79 case CEPH_MSG_CLIENT_RECONNECT: return "client_reconnect"; 80 case CEPH_MSG_CLIENT_REQUEST: return "client_request"; 81 case CEPH_MSG_CLIENT_REQUEST_FORWARD: return "client_request_forward"; 82 case CEPH_MSG_CLIENT_REPLY: return "client_reply"; 83 case CEPH_MSG_CLIENT_CAPS: return "client_caps"; 84 case CEPH_MSG_CLIENT_CAPRELEASE: return "client_cap_release"; 85 case CEPH_MSG_CLIENT_SNAP: return "client_snap"; 86 case CEPH_MSG_CLIENT_LEASE: return "client_lease"; 87 case CEPH_MSG_OSD_MAP: return "osd_map"; 88 case CEPH_MSG_OSD_OP: return "osd_op"; 89 case CEPH_MSG_OSD_OPREPLY: return "osd_opreply"; 90 case CEPH_MSG_WATCH_NOTIFY: return "watch_notify"; 91 default: return "unknown"; 92 } 93 } 94 EXPORT_SYMBOL(ceph_msg_type_name); 95 96 /* 97 * Initially learn our fsid, or verify an fsid matches. 98 */ 99 int ceph_check_fsid(struct ceph_client *client, struct ceph_fsid *fsid) 100 { 101 if (client->have_fsid) { 102 if (ceph_fsid_compare(&client->fsid, fsid)) { 103 pr_err("bad fsid, had %pU got %pU", 104 &client->fsid, fsid); 105 return -1; 106 } 107 } else { 108 memcpy(&client->fsid, fsid, sizeof(*fsid)); 109 } 110 return 0; 111 } 112 EXPORT_SYMBOL(ceph_check_fsid); 113 114 static int strcmp_null(const char *s1, const char *s2) 115 { 116 if (!s1 && !s2) 117 return 0; 118 if (s1 && !s2) 119 return -1; 120 if (!s1 && s2) 121 return 1; 122 return strcmp(s1, s2); 123 } 124 125 int ceph_compare_options(struct ceph_options *new_opt, 126 struct ceph_client *client) 127 { 128 struct ceph_options *opt1 = new_opt; 129 struct ceph_options *opt2 = client->options; 130 int ofs = offsetof(struct ceph_options, mon_addr); 131 int i; 132 int ret; 133 134 ret = memcmp(opt1, opt2, ofs); 135 if (ret) 136 return ret; 137 138 ret = strcmp_null(opt1->name, opt2->name); 139 if (ret) 140 return ret; 141 142 if (opt1->key && !opt2->key) 143 return -1; 144 if (!opt1->key && opt2->key) 145 return 1; 146 if (opt1->key && opt2->key) { 147 if (opt1->key->type != opt2->key->type) 148 return -1; 149 if (opt1->key->created.tv_sec != opt2->key->created.tv_sec) 150 return -1; 151 if (opt1->key->created.tv_nsec != opt2->key->created.tv_nsec) 152 return -1; 153 if (opt1->key->len != opt2->key->len) 154 return -1; 155 if (opt1->key->key && !opt2->key->key) 156 return -1; 157 if (!opt1->key->key && opt2->key->key) 158 return 1; 159 if (opt1->key->key && opt2->key->key) { 160 ret = memcmp(opt1->key->key, opt2->key->key, opt1->key->len); 161 if (ret) 162 return ret; 163 } 164 } 165 166 /* any matching mon ip implies a match */ 167 for (i = 0; i < opt1->num_mon; i++) { 168 if (ceph_monmap_contains(client->monc.monmap, 169 &opt1->mon_addr[i])) 170 return 0; 171 } 172 return -1; 173 } 174 EXPORT_SYMBOL(ceph_compare_options); 175 176 void *ceph_kvmalloc(size_t size, gfp_t flags) 177 { 178 if (size <= (PAGE_SIZE << PAGE_ALLOC_COSTLY_ORDER)) { 179 void *ptr = kmalloc(size, flags | __GFP_NOWARN); 180 if (ptr) 181 return ptr; 182 } 183 184 return __vmalloc(size, flags | __GFP_HIGHMEM, PAGE_KERNEL); 185 } 186 187 void ceph_kvfree(const void *ptr) 188 { 189 if (is_vmalloc_addr(ptr)) 190 vfree(ptr); 191 else 192 kfree(ptr); 193 } 194 195 196 static int parse_fsid(const char *str, struct ceph_fsid *fsid) 197 { 198 int i = 0; 199 char tmp[3]; 200 int err = -EINVAL; 201 int d; 202 203 dout("parse_fsid '%s'\n", str); 204 tmp[2] = 0; 205 while (*str && i < 16) { 206 if (ispunct(*str)) { 207 str++; 208 continue; 209 } 210 if (!isxdigit(str[0]) || !isxdigit(str[1])) 211 break; 212 tmp[0] = str[0]; 213 tmp[1] = str[1]; 214 if (sscanf(tmp, "%x", &d) < 1) 215 break; 216 fsid->fsid[i] = d & 0xff; 217 i++; 218 str += 2; 219 } 220 221 if (i == 16) 222 err = 0; 223 dout("parse_fsid ret %d got fsid %pU", err, fsid); 224 return err; 225 } 226 227 /* 228 * ceph options 229 */ 230 enum { 231 Opt_osdtimeout, 232 Opt_osdkeepalivetimeout, 233 Opt_mount_timeout, 234 Opt_osd_idle_ttl, 235 Opt_last_int, 236 /* int args above */ 237 Opt_fsid, 238 Opt_name, 239 Opt_secret, 240 Opt_key, 241 Opt_ip, 242 Opt_last_string, 243 /* string args above */ 244 Opt_share, 245 Opt_noshare, 246 Opt_crc, 247 Opt_nocrc, 248 }; 249 250 static match_table_t opt_tokens = { 251 {Opt_osdtimeout, "osdtimeout=%d"}, 252 {Opt_osdkeepalivetimeout, "osdkeepalive=%d"}, 253 {Opt_mount_timeout, "mount_timeout=%d"}, 254 {Opt_osd_idle_ttl, "osd_idle_ttl=%d"}, 255 /* int args above */ 256 {Opt_fsid, "fsid=%s"}, 257 {Opt_name, "name=%s"}, 258 {Opt_secret, "secret=%s"}, 259 {Opt_key, "key=%s"}, 260 {Opt_ip, "ip=%s"}, 261 /* string args above */ 262 {Opt_share, "share"}, 263 {Opt_noshare, "noshare"}, 264 {Opt_crc, "crc"}, 265 {Opt_nocrc, "nocrc"}, 266 {-1, NULL} 267 }; 268 269 void ceph_destroy_options(struct ceph_options *opt) 270 { 271 dout("destroy_options %p\n", opt); 272 kfree(opt->name); 273 if (opt->key) { 274 ceph_crypto_key_destroy(opt->key); 275 kfree(opt->key); 276 } 277 kfree(opt->mon_addr); 278 kfree(opt); 279 } 280 EXPORT_SYMBOL(ceph_destroy_options); 281 282 /* get secret from key store */ 283 static int get_secret(struct ceph_crypto_key *dst, const char *name) { 284 struct key *ukey; 285 int key_err; 286 int err = 0; 287 struct ceph_crypto_key *ckey; 288 289 ukey = request_key(&key_type_ceph, name, NULL); 290 if (!ukey || IS_ERR(ukey)) { 291 /* request_key errors don't map nicely to mount(2) 292 errors; don't even try, but still printk */ 293 key_err = PTR_ERR(ukey); 294 switch (key_err) { 295 case -ENOKEY: 296 pr_warning("ceph: Mount failed due to key not found: %s\n", name); 297 break; 298 case -EKEYEXPIRED: 299 pr_warning("ceph: Mount failed due to expired key: %s\n", name); 300 break; 301 case -EKEYREVOKED: 302 pr_warning("ceph: Mount failed due to revoked key: %s\n", name); 303 break; 304 default: 305 pr_warning("ceph: Mount failed due to unknown key error" 306 " %d: %s\n", key_err, name); 307 } 308 err = -EPERM; 309 goto out; 310 } 311 312 ckey = ukey->payload.data; 313 err = ceph_crypto_key_clone(dst, ckey); 314 if (err) 315 goto out_key; 316 /* pass through, err is 0 */ 317 318 out_key: 319 key_put(ukey); 320 out: 321 return err; 322 } 323 324 struct ceph_options * 325 ceph_parse_options(char *options, const char *dev_name, 326 const char *dev_name_end, 327 int (*parse_extra_token)(char *c, void *private), 328 void *private) 329 { 330 struct ceph_options *opt; 331 const char *c; 332 int err = -ENOMEM; 333 substring_t argstr[MAX_OPT_ARGS]; 334 335 if (current->nsproxy->net_ns != &init_net) 336 return ERR_PTR(-EINVAL); 337 338 opt = kzalloc(sizeof(*opt), GFP_KERNEL); 339 if (!opt) 340 return ERR_PTR(-ENOMEM); 341 opt->mon_addr = kcalloc(CEPH_MAX_MON, sizeof(*opt->mon_addr), 342 GFP_KERNEL); 343 if (!opt->mon_addr) 344 goto out; 345 346 dout("parse_options %p options '%s' dev_name '%s'\n", opt, options, 347 dev_name); 348 349 /* start with defaults */ 350 opt->flags = CEPH_OPT_DEFAULT; 351 opt->osd_keepalive_timeout = CEPH_OSD_KEEPALIVE_DEFAULT; 352 opt->mount_timeout = CEPH_MOUNT_TIMEOUT_DEFAULT; /* seconds */ 353 opt->osd_idle_ttl = CEPH_OSD_IDLE_TTL_DEFAULT; /* seconds */ 354 355 /* get mon ip(s) */ 356 /* ip1[:port1][,ip2[:port2]...] */ 357 err = ceph_parse_ips(dev_name, dev_name_end, opt->mon_addr, 358 CEPH_MAX_MON, &opt->num_mon); 359 if (err < 0) 360 goto out; 361 362 /* parse mount options */ 363 while ((c = strsep(&options, ",")) != NULL) { 364 int token, intval, ret; 365 if (!*c) 366 continue; 367 err = -EINVAL; 368 token = match_token((char *)c, opt_tokens, argstr); 369 if (token < 0 && parse_extra_token) { 370 /* extra? */ 371 err = parse_extra_token((char *)c, private); 372 if (err < 0) { 373 pr_err("bad option at '%s'\n", c); 374 goto out; 375 } 376 continue; 377 } 378 if (token < Opt_last_int) { 379 ret = match_int(&argstr[0], &intval); 380 if (ret < 0) { 381 pr_err("bad mount option arg (not int) " 382 "at '%s'\n", c); 383 continue; 384 } 385 dout("got int token %d val %d\n", token, intval); 386 } else if (token > Opt_last_int && token < Opt_last_string) { 387 dout("got string token %d val %s\n", token, 388 argstr[0].from); 389 } else { 390 dout("got token %d\n", token); 391 } 392 switch (token) { 393 case Opt_ip: 394 err = ceph_parse_ips(argstr[0].from, 395 argstr[0].to, 396 &opt->my_addr, 397 1, NULL); 398 if (err < 0) 399 goto out; 400 opt->flags |= CEPH_OPT_MYIP; 401 break; 402 403 case Opt_fsid: 404 err = parse_fsid(argstr[0].from, &opt->fsid); 405 if (err == 0) 406 opt->flags |= CEPH_OPT_FSID; 407 break; 408 case Opt_name: 409 opt->name = kstrndup(argstr[0].from, 410 argstr[0].to-argstr[0].from, 411 GFP_KERNEL); 412 break; 413 case Opt_secret: 414 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL); 415 if (!opt->key) { 416 err = -ENOMEM; 417 goto out; 418 } 419 err = ceph_crypto_key_unarmor(opt->key, argstr[0].from); 420 if (err < 0) 421 goto out; 422 break; 423 case Opt_key: 424 opt->key = kzalloc(sizeof(*opt->key), GFP_KERNEL); 425 if (!opt->key) { 426 err = -ENOMEM; 427 goto out; 428 } 429 err = get_secret(opt->key, argstr[0].from); 430 if (err < 0) 431 goto out; 432 break; 433 434 /* misc */ 435 case Opt_osdtimeout: 436 pr_warning("ignoring deprecated osdtimeout option\n"); 437 break; 438 case Opt_osdkeepalivetimeout: 439 opt->osd_keepalive_timeout = intval; 440 break; 441 case Opt_osd_idle_ttl: 442 opt->osd_idle_ttl = intval; 443 break; 444 case Opt_mount_timeout: 445 opt->mount_timeout = intval; 446 break; 447 448 case Opt_share: 449 opt->flags &= ~CEPH_OPT_NOSHARE; 450 break; 451 case Opt_noshare: 452 opt->flags |= CEPH_OPT_NOSHARE; 453 break; 454 455 case Opt_crc: 456 opt->flags &= ~CEPH_OPT_NOCRC; 457 break; 458 case Opt_nocrc: 459 opt->flags |= CEPH_OPT_NOCRC; 460 break; 461 462 default: 463 BUG_ON(token); 464 } 465 } 466 467 /* success */ 468 return opt; 469 470 out: 471 ceph_destroy_options(opt); 472 return ERR_PTR(err); 473 } 474 EXPORT_SYMBOL(ceph_parse_options); 475 476 u64 ceph_client_id(struct ceph_client *client) 477 { 478 return client->monc.auth->global_id; 479 } 480 EXPORT_SYMBOL(ceph_client_id); 481 482 /* 483 * create a fresh client instance 484 */ 485 struct ceph_client *ceph_create_client(struct ceph_options *opt, void *private, 486 u64 supported_features, 487 u64 required_features) 488 { 489 struct ceph_client *client; 490 struct ceph_entity_addr *myaddr = NULL; 491 int err = -ENOMEM; 492 493 client = kzalloc(sizeof(*client), GFP_KERNEL); 494 if (client == NULL) 495 return ERR_PTR(-ENOMEM); 496 497 client->private = private; 498 client->options = opt; 499 500 mutex_init(&client->mount_mutex); 501 init_waitqueue_head(&client->auth_wq); 502 client->auth_err = 0; 503 504 client->extra_mon_dispatch = NULL; 505 client->supported_features = CEPH_FEATURES_SUPPORTED_DEFAULT | 506 supported_features; 507 client->required_features = CEPH_FEATURES_REQUIRED_DEFAULT | 508 required_features; 509 510 /* msgr */ 511 if (ceph_test_opt(client, MYIP)) 512 myaddr = &client->options->my_addr; 513 ceph_messenger_init(&client->msgr, myaddr, 514 client->supported_features, 515 client->required_features, 516 ceph_test_opt(client, NOCRC)); 517 518 /* subsystems */ 519 err = ceph_monc_init(&client->monc, client); 520 if (err < 0) 521 goto fail; 522 err = ceph_osdc_init(&client->osdc, client); 523 if (err < 0) 524 goto fail_monc; 525 526 return client; 527 528 fail_monc: 529 ceph_monc_stop(&client->monc); 530 fail: 531 kfree(client); 532 return ERR_PTR(err); 533 } 534 EXPORT_SYMBOL(ceph_create_client); 535 536 void ceph_destroy_client(struct ceph_client *client) 537 { 538 dout("destroy_client %p\n", client); 539 540 atomic_set(&client->msgr.stopping, 1); 541 542 /* unmount */ 543 ceph_osdc_stop(&client->osdc); 544 545 ceph_monc_stop(&client->monc); 546 547 ceph_debugfs_client_cleanup(client); 548 549 ceph_destroy_options(client->options); 550 551 kfree(client); 552 dout("destroy_client %p done\n", client); 553 } 554 EXPORT_SYMBOL(ceph_destroy_client); 555 556 /* 557 * true if we have the mon map (and have thus joined the cluster) 558 */ 559 static int have_mon_and_osd_map(struct ceph_client *client) 560 { 561 return client->monc.monmap && client->monc.monmap->epoch && 562 client->osdc.osdmap && client->osdc.osdmap->epoch; 563 } 564 565 /* 566 * mount: join the ceph cluster, and open root directory. 567 */ 568 int __ceph_open_session(struct ceph_client *client, unsigned long started) 569 { 570 int err; 571 unsigned long timeout = client->options->mount_timeout * HZ; 572 573 /* open session, and wait for mon and osd maps */ 574 err = ceph_monc_open_session(&client->monc); 575 if (err < 0) 576 return err; 577 578 while (!have_mon_and_osd_map(client)) { 579 err = -EIO; 580 if (timeout && time_after_eq(jiffies, started + timeout)) 581 return err; 582 583 /* wait */ 584 dout("mount waiting for mon_map\n"); 585 err = wait_event_interruptible_timeout(client->auth_wq, 586 have_mon_and_osd_map(client) || (client->auth_err < 0), 587 timeout); 588 if (err == -EINTR || err == -ERESTARTSYS) 589 return err; 590 if (client->auth_err < 0) 591 return client->auth_err; 592 } 593 594 return 0; 595 } 596 EXPORT_SYMBOL(__ceph_open_session); 597 598 599 int ceph_open_session(struct ceph_client *client) 600 { 601 int ret; 602 unsigned long started = jiffies; /* note the start time */ 603 604 dout("open_session start\n"); 605 mutex_lock(&client->mount_mutex); 606 607 ret = __ceph_open_session(client, started); 608 609 mutex_unlock(&client->mount_mutex); 610 return ret; 611 } 612 EXPORT_SYMBOL(ceph_open_session); 613 614 615 static int __init init_ceph_lib(void) 616 { 617 int ret = 0; 618 619 ret = ceph_debugfs_init(); 620 if (ret < 0) 621 goto out; 622 623 ret = ceph_crypto_init(); 624 if (ret < 0) 625 goto out_debugfs; 626 627 ret = ceph_msgr_init(); 628 if (ret < 0) 629 goto out_crypto; 630 631 ret = ceph_osdc_setup(); 632 if (ret < 0) 633 goto out_msgr; 634 635 pr_info("loaded (mon/osd proto %d/%d)\n", 636 CEPH_MONC_PROTOCOL, CEPH_OSDC_PROTOCOL); 637 638 return 0; 639 640 out_msgr: 641 ceph_msgr_exit(); 642 out_crypto: 643 ceph_crypto_shutdown(); 644 out_debugfs: 645 ceph_debugfs_cleanup(); 646 out: 647 return ret; 648 } 649 650 static void __exit exit_ceph_lib(void) 651 { 652 dout("exit_ceph_lib\n"); 653 ceph_osdc_cleanup(); 654 ceph_msgr_exit(); 655 ceph_crypto_shutdown(); 656 ceph_debugfs_cleanup(); 657 } 658 659 module_init(init_ceph_lib); 660 module_exit(exit_ceph_lib); 661 662 MODULE_AUTHOR("Sage Weil <sage@newdream.net>"); 663 MODULE_AUTHOR("Yehuda Sadeh <yehuda@hq.newdream.net>"); 664 MODULE_AUTHOR("Patience Warnick <patience@newdream.net>"); 665 MODULE_DESCRIPTION("Ceph filesystem for Linux"); 666 MODULE_LICENSE("GPL"); 667