1 /* -*- mode: c; c-basic-offset: 8; -*- 2 * vim: noexpandtab sw=8 ts=8 sts=0: 3 * 4 * dlmdomain.c 5 * 6 * defines domain join / leave apis 7 * 8 * Copyright (C) 2004 Oracle. All rights reserved. 9 * 10 * This program is free software; you can redistribute it and/or 11 * modify it under the terms of the GNU General Public 12 * License as published by the Free Software Foundation; either 13 * version 2 of the License, or (at your option) any later version. 14 * 15 * This program is distributed in the hope that it will be useful, 16 * but WITHOUT ANY WARRANTY; without even the implied warranty of 17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU 18 * General Public License for more details. 19 * 20 * You should have received a copy of the GNU General Public 21 * License along with this program; if not, write to the 22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330, 23 * Boston, MA 021110-1307, USA. 24 * 25 */ 26 27 #include <linux/module.h> 28 #include <linux/types.h> 29 #include <linux/slab.h> 30 #include <linux/highmem.h> 31 #include <linux/init.h> 32 #include <linux/spinlock.h> 33 #include <linux/delay.h> 34 #include <linux/err.h> 35 #include <linux/debugfs.h> 36 37 #include "cluster/heartbeat.h" 38 #include "cluster/nodemanager.h" 39 #include "cluster/tcp.h" 40 41 #include "dlmapi.h" 42 #include "dlmcommon.h" 43 #include "dlmdomain.h" 44 #include "dlmdebug.h" 45 46 #include "dlmver.h" 47 48 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_DOMAIN) 49 #include "cluster/masklog.h" 50 51 /* 52 * ocfs2 node maps are array of long int, which limits to send them freely 53 * across the wire due to endianness issues. To workaround this, we convert 54 * long ints to byte arrays. Following 3 routines are helper functions to 55 * set/test/copy bits within those array of bytes 56 */ 57 static inline void byte_set_bit(u8 nr, u8 map[]) 58 { 59 map[nr >> 3] |= (1UL << (nr & 7)); 60 } 61 62 static inline int byte_test_bit(u8 nr, u8 map[]) 63 { 64 return ((1UL << (nr & 7)) & (map[nr >> 3])) != 0; 65 } 66 67 static inline void byte_copymap(u8 dmap[], unsigned long smap[], 68 unsigned int sz) 69 { 70 unsigned int nn; 71 72 if (!sz) 73 return; 74 75 memset(dmap, 0, ((sz + 7) >> 3)); 76 for (nn = 0 ; nn < sz; nn++) 77 if (test_bit(nn, smap)) 78 byte_set_bit(nn, dmap); 79 } 80 81 static void dlm_free_pagevec(void **vec, int pages) 82 { 83 while (pages--) 84 free_page((unsigned long)vec[pages]); 85 kfree(vec); 86 } 87 88 static void **dlm_alloc_pagevec(int pages) 89 { 90 void **vec = kmalloc(pages * sizeof(void *), GFP_KERNEL); 91 int i; 92 93 if (!vec) 94 return NULL; 95 96 for (i = 0; i < pages; i++) 97 if (!(vec[i] = (void *)__get_free_page(GFP_KERNEL))) 98 goto out_free; 99 100 mlog(0, "Allocated DLM hash pagevec; %d pages (%lu expected), %lu buckets per page\n", 101 pages, (unsigned long)DLM_HASH_PAGES, 102 (unsigned long)DLM_BUCKETS_PER_PAGE); 103 return vec; 104 out_free: 105 dlm_free_pagevec(vec, i); 106 return NULL; 107 } 108 109 /* 110 * 111 * spinlock lock ordering: if multiple locks are needed, obey this ordering: 112 * dlm_domain_lock 113 * struct dlm_ctxt->spinlock 114 * struct dlm_lock_resource->spinlock 115 * struct dlm_ctxt->master_lock 116 * struct dlm_ctxt->ast_lock 117 * dlm_master_list_entry->spinlock 118 * dlm_lock->spinlock 119 * 120 */ 121 122 DEFINE_SPINLOCK(dlm_domain_lock); 123 LIST_HEAD(dlm_domains); 124 static DECLARE_WAIT_QUEUE_HEAD(dlm_domain_events); 125 126 /* 127 * The supported protocol version for DLM communication. Running domains 128 * will have a negotiated version with the same major number and a minor 129 * number equal or smaller. The dlm_ctxt->dlm_locking_proto field should 130 * be used to determine what a running domain is actually using. 131 * 132 * New in version 1.1: 133 * - Message DLM_QUERY_REGION added to support global heartbeat 134 * - Message DLM_QUERY_NODEINFO added to allow online node removes 135 */ 136 static const struct dlm_protocol_version dlm_protocol = { 137 .pv_major = 1, 138 .pv_minor = 0, 139 }; 140 141 #define DLM_DOMAIN_BACKOFF_MS 200 142 143 static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data, 144 void **ret_data); 145 static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, 146 void **ret_data); 147 static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, 148 void **ret_data); 149 static int dlm_query_region_handler(struct o2net_msg *msg, u32 len, 150 void *data, void **ret_data); 151 static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, 152 void **ret_data); 153 static int dlm_protocol_compare(struct dlm_protocol_version *existing, 154 struct dlm_protocol_version *request); 155 156 static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm); 157 158 void __dlm_unhash_lockres(struct dlm_lock_resource *lockres) 159 { 160 if (!hlist_unhashed(&lockres->hash_node)) { 161 hlist_del_init(&lockres->hash_node); 162 dlm_lockres_put(lockres); 163 } 164 } 165 166 void __dlm_insert_lockres(struct dlm_ctxt *dlm, 167 struct dlm_lock_resource *res) 168 { 169 struct hlist_head *bucket; 170 struct qstr *q; 171 172 assert_spin_locked(&dlm->spinlock); 173 174 q = &res->lockname; 175 bucket = dlm_lockres_hash(dlm, q->hash); 176 177 /* get a reference for our hashtable */ 178 dlm_lockres_get(res); 179 180 hlist_add_head(&res->hash_node, bucket); 181 } 182 183 struct dlm_lock_resource * __dlm_lookup_lockres_full(struct dlm_ctxt *dlm, 184 const char *name, 185 unsigned int len, 186 unsigned int hash) 187 { 188 struct hlist_head *bucket; 189 struct hlist_node *list; 190 191 mlog_entry("%.*s\n", len, name); 192 193 assert_spin_locked(&dlm->spinlock); 194 195 bucket = dlm_lockres_hash(dlm, hash); 196 197 hlist_for_each(list, bucket) { 198 struct dlm_lock_resource *res = hlist_entry(list, 199 struct dlm_lock_resource, hash_node); 200 if (res->lockname.name[0] != name[0]) 201 continue; 202 if (unlikely(res->lockname.len != len)) 203 continue; 204 if (memcmp(res->lockname.name + 1, name + 1, len - 1)) 205 continue; 206 dlm_lockres_get(res); 207 return res; 208 } 209 return NULL; 210 } 211 212 /* intended to be called by functions which do not care about lock 213 * resources which are being purged (most net _handler functions). 214 * this will return NULL for any lock resource which is found but 215 * currently in the process of dropping its mastery reference. 216 * use __dlm_lookup_lockres_full when you need the lock resource 217 * regardless (e.g. dlm_get_lock_resource) */ 218 struct dlm_lock_resource * __dlm_lookup_lockres(struct dlm_ctxt *dlm, 219 const char *name, 220 unsigned int len, 221 unsigned int hash) 222 { 223 struct dlm_lock_resource *res = NULL; 224 225 mlog_entry("%.*s\n", len, name); 226 227 assert_spin_locked(&dlm->spinlock); 228 229 res = __dlm_lookup_lockres_full(dlm, name, len, hash); 230 if (res) { 231 spin_lock(&res->spinlock); 232 if (res->state & DLM_LOCK_RES_DROPPING_REF) { 233 spin_unlock(&res->spinlock); 234 dlm_lockres_put(res); 235 return NULL; 236 } 237 spin_unlock(&res->spinlock); 238 } 239 240 return res; 241 } 242 243 struct dlm_lock_resource * dlm_lookup_lockres(struct dlm_ctxt *dlm, 244 const char *name, 245 unsigned int len) 246 { 247 struct dlm_lock_resource *res; 248 unsigned int hash = dlm_lockid_hash(name, len); 249 250 spin_lock(&dlm->spinlock); 251 res = __dlm_lookup_lockres(dlm, name, len, hash); 252 spin_unlock(&dlm->spinlock); 253 return res; 254 } 255 256 static struct dlm_ctxt * __dlm_lookup_domain_full(const char *domain, int len) 257 { 258 struct dlm_ctxt *tmp = NULL; 259 struct list_head *iter; 260 261 assert_spin_locked(&dlm_domain_lock); 262 263 /* tmp->name here is always NULL terminated, 264 * but domain may not be! */ 265 list_for_each(iter, &dlm_domains) { 266 tmp = list_entry (iter, struct dlm_ctxt, list); 267 if (strlen(tmp->name) == len && 268 memcmp(tmp->name, domain, len)==0) 269 break; 270 tmp = NULL; 271 } 272 273 return tmp; 274 } 275 276 /* For null terminated domain strings ONLY */ 277 static struct dlm_ctxt * __dlm_lookup_domain(const char *domain) 278 { 279 assert_spin_locked(&dlm_domain_lock); 280 281 return __dlm_lookup_domain_full(domain, strlen(domain)); 282 } 283 284 285 /* returns true on one of two conditions: 286 * 1) the domain does not exist 287 * 2) the domain exists and it's state is "joined" */ 288 static int dlm_wait_on_domain_helper(const char *domain) 289 { 290 int ret = 0; 291 struct dlm_ctxt *tmp = NULL; 292 293 spin_lock(&dlm_domain_lock); 294 295 tmp = __dlm_lookup_domain(domain); 296 if (!tmp) 297 ret = 1; 298 else if (tmp->dlm_state == DLM_CTXT_JOINED) 299 ret = 1; 300 301 spin_unlock(&dlm_domain_lock); 302 return ret; 303 } 304 305 static void dlm_free_ctxt_mem(struct dlm_ctxt *dlm) 306 { 307 dlm_destroy_debugfs_subroot(dlm); 308 309 if (dlm->lockres_hash) 310 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES); 311 312 if (dlm->master_hash) 313 dlm_free_pagevec((void **)dlm->master_hash, DLM_HASH_PAGES); 314 315 if (dlm->name) 316 kfree(dlm->name); 317 318 kfree(dlm); 319 } 320 321 /* A little strange - this function will be called while holding 322 * dlm_domain_lock and is expected to be holding it on the way out. We 323 * will however drop and reacquire it multiple times */ 324 static void dlm_ctxt_release(struct kref *kref) 325 { 326 struct dlm_ctxt *dlm; 327 328 dlm = container_of(kref, struct dlm_ctxt, dlm_refs); 329 330 BUG_ON(dlm->num_joins); 331 BUG_ON(dlm->dlm_state == DLM_CTXT_JOINED); 332 333 /* we may still be in the list if we hit an error during join. */ 334 list_del_init(&dlm->list); 335 336 spin_unlock(&dlm_domain_lock); 337 338 mlog(0, "freeing memory from domain %s\n", dlm->name); 339 340 wake_up(&dlm_domain_events); 341 342 dlm_free_ctxt_mem(dlm); 343 344 spin_lock(&dlm_domain_lock); 345 } 346 347 void dlm_put(struct dlm_ctxt *dlm) 348 { 349 spin_lock(&dlm_domain_lock); 350 kref_put(&dlm->dlm_refs, dlm_ctxt_release); 351 spin_unlock(&dlm_domain_lock); 352 } 353 354 static void __dlm_get(struct dlm_ctxt *dlm) 355 { 356 kref_get(&dlm->dlm_refs); 357 } 358 359 /* given a questionable reference to a dlm object, gets a reference if 360 * it can find it in the list, otherwise returns NULL in which case 361 * you shouldn't trust your pointer. */ 362 struct dlm_ctxt *dlm_grab(struct dlm_ctxt *dlm) 363 { 364 struct list_head *iter; 365 struct dlm_ctxt *target = NULL; 366 367 spin_lock(&dlm_domain_lock); 368 369 list_for_each(iter, &dlm_domains) { 370 target = list_entry (iter, struct dlm_ctxt, list); 371 372 if (target == dlm) { 373 __dlm_get(target); 374 break; 375 } 376 377 target = NULL; 378 } 379 380 spin_unlock(&dlm_domain_lock); 381 382 return target; 383 } 384 385 int dlm_domain_fully_joined(struct dlm_ctxt *dlm) 386 { 387 int ret; 388 389 spin_lock(&dlm_domain_lock); 390 ret = (dlm->dlm_state == DLM_CTXT_JOINED) || 391 (dlm->dlm_state == DLM_CTXT_IN_SHUTDOWN); 392 spin_unlock(&dlm_domain_lock); 393 394 return ret; 395 } 396 397 static void dlm_destroy_dlm_worker(struct dlm_ctxt *dlm) 398 { 399 if (dlm->dlm_worker) { 400 flush_workqueue(dlm->dlm_worker); 401 destroy_workqueue(dlm->dlm_worker); 402 dlm->dlm_worker = NULL; 403 } 404 } 405 406 static void dlm_complete_dlm_shutdown(struct dlm_ctxt *dlm) 407 { 408 dlm_unregister_domain_handlers(dlm); 409 dlm_debug_shutdown(dlm); 410 dlm_complete_thread(dlm); 411 dlm_complete_recovery_thread(dlm); 412 dlm_destroy_dlm_worker(dlm); 413 414 /* We've left the domain. Now we can take ourselves out of the 415 * list and allow the kref stuff to help us free the 416 * memory. */ 417 spin_lock(&dlm_domain_lock); 418 list_del_init(&dlm->list); 419 spin_unlock(&dlm_domain_lock); 420 421 /* Wake up anyone waiting for us to remove this domain */ 422 wake_up(&dlm_domain_events); 423 } 424 425 static int dlm_migrate_all_locks(struct dlm_ctxt *dlm) 426 { 427 int i, num, n, ret = 0; 428 struct dlm_lock_resource *res; 429 struct hlist_node *iter; 430 struct hlist_head *bucket; 431 int dropped; 432 433 mlog(0, "Migrating locks from domain %s\n", dlm->name); 434 435 num = 0; 436 spin_lock(&dlm->spinlock); 437 for (i = 0; i < DLM_HASH_BUCKETS; i++) { 438 redo_bucket: 439 n = 0; 440 bucket = dlm_lockres_hash(dlm, i); 441 iter = bucket->first; 442 while (iter) { 443 n++; 444 res = hlist_entry(iter, struct dlm_lock_resource, 445 hash_node); 446 dlm_lockres_get(res); 447 /* migrate, if necessary. this will drop the dlm 448 * spinlock and retake it if it does migration. */ 449 dropped = dlm_empty_lockres(dlm, res); 450 451 spin_lock(&res->spinlock); 452 __dlm_lockres_calc_usage(dlm, res); 453 iter = res->hash_node.next; 454 spin_unlock(&res->spinlock); 455 456 dlm_lockres_put(res); 457 458 if (dropped) 459 goto redo_bucket; 460 } 461 cond_resched_lock(&dlm->spinlock); 462 num += n; 463 mlog(0, "%s: touched %d lockreses in bucket %d " 464 "(tot=%d)\n", dlm->name, n, i, num); 465 } 466 spin_unlock(&dlm->spinlock); 467 wake_up(&dlm->dlm_thread_wq); 468 469 /* let the dlm thread take care of purging, keep scanning until 470 * nothing remains in the hash */ 471 if (num) { 472 mlog(0, "%s: %d lock resources in hash last pass\n", 473 dlm->name, num); 474 ret = -EAGAIN; 475 } 476 mlog(0, "DONE Migrating locks from domain %s\n", dlm->name); 477 return ret; 478 } 479 480 static int dlm_no_joining_node(struct dlm_ctxt *dlm) 481 { 482 int ret; 483 484 spin_lock(&dlm->spinlock); 485 ret = dlm->joining_node == DLM_LOCK_RES_OWNER_UNKNOWN; 486 spin_unlock(&dlm->spinlock); 487 488 return ret; 489 } 490 491 static void dlm_mark_domain_leaving(struct dlm_ctxt *dlm) 492 { 493 /* Yikes, a double spinlock! I need domain_lock for the dlm 494 * state and the dlm spinlock for join state... Sorry! */ 495 again: 496 spin_lock(&dlm_domain_lock); 497 spin_lock(&dlm->spinlock); 498 499 if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) { 500 mlog(0, "Node %d is joining, we wait on it.\n", 501 dlm->joining_node); 502 spin_unlock(&dlm->spinlock); 503 spin_unlock(&dlm_domain_lock); 504 505 wait_event(dlm->dlm_join_events, dlm_no_joining_node(dlm)); 506 goto again; 507 } 508 509 dlm->dlm_state = DLM_CTXT_LEAVING; 510 spin_unlock(&dlm->spinlock); 511 spin_unlock(&dlm_domain_lock); 512 } 513 514 static void __dlm_print_nodes(struct dlm_ctxt *dlm) 515 { 516 int node = -1; 517 518 assert_spin_locked(&dlm->spinlock); 519 520 printk(KERN_NOTICE "o2dlm: Nodes in domain %s: ", dlm->name); 521 522 while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES, 523 node + 1)) < O2NM_MAX_NODES) { 524 printk("%d ", node); 525 } 526 printk("\n"); 527 } 528 529 static int dlm_exit_domain_handler(struct o2net_msg *msg, u32 len, void *data, 530 void **ret_data) 531 { 532 struct dlm_ctxt *dlm = data; 533 unsigned int node; 534 struct dlm_exit_domain *exit_msg = (struct dlm_exit_domain *) msg->buf; 535 536 mlog_entry("%p %u %p", msg, len, data); 537 538 if (!dlm_grab(dlm)) 539 return 0; 540 541 node = exit_msg->node_idx; 542 543 printk(KERN_NOTICE "o2dlm: Node %u leaves domain %s\n", node, dlm->name); 544 545 spin_lock(&dlm->spinlock); 546 clear_bit(node, dlm->domain_map); 547 __dlm_print_nodes(dlm); 548 549 /* notify anything attached to the heartbeat events */ 550 dlm_hb_event_notify_attached(dlm, node, 0); 551 552 spin_unlock(&dlm->spinlock); 553 554 dlm_put(dlm); 555 556 return 0; 557 } 558 559 static int dlm_send_one_domain_exit(struct dlm_ctxt *dlm, 560 unsigned int node) 561 { 562 int status; 563 struct dlm_exit_domain leave_msg; 564 565 mlog(0, "Asking node %u if we can leave the domain %s me = %u\n", 566 node, dlm->name, dlm->node_num); 567 568 memset(&leave_msg, 0, sizeof(leave_msg)); 569 leave_msg.node_idx = dlm->node_num; 570 571 status = o2net_send_message(DLM_EXIT_DOMAIN_MSG, dlm->key, 572 &leave_msg, sizeof(leave_msg), node, 573 NULL); 574 if (status < 0) 575 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " 576 "node %u\n", status, DLM_EXIT_DOMAIN_MSG, dlm->key, node); 577 mlog(0, "status return %d from o2net_send_message\n", status); 578 579 return status; 580 } 581 582 583 static void dlm_leave_domain(struct dlm_ctxt *dlm) 584 { 585 int node, clear_node, status; 586 587 /* At this point we've migrated away all our locks and won't 588 * accept mastership of new ones. The dlm is responsible for 589 * almost nothing now. We make sure not to confuse any joining 590 * nodes and then commence shutdown procedure. */ 591 592 spin_lock(&dlm->spinlock); 593 /* Clear ourselves from the domain map */ 594 clear_bit(dlm->node_num, dlm->domain_map); 595 while ((node = find_next_bit(dlm->domain_map, O2NM_MAX_NODES, 596 0)) < O2NM_MAX_NODES) { 597 /* Drop the dlm spinlock. This is safe wrt the domain_map. 598 * -nodes cannot be added now as the 599 * query_join_handlers knows to respond with OK_NO_MAP 600 * -we catch the right network errors if a node is 601 * removed from the map while we're sending him the 602 * exit message. */ 603 spin_unlock(&dlm->spinlock); 604 605 clear_node = 1; 606 607 status = dlm_send_one_domain_exit(dlm, node); 608 if (status < 0 && 609 status != -ENOPROTOOPT && 610 status != -ENOTCONN) { 611 mlog(ML_NOTICE, "Error %d sending domain exit message " 612 "to node %d\n", status, node); 613 614 /* Not sure what to do here but lets sleep for 615 * a bit in case this was a transient 616 * error... */ 617 msleep(DLM_DOMAIN_BACKOFF_MS); 618 clear_node = 0; 619 } 620 621 spin_lock(&dlm->spinlock); 622 /* If we're not clearing the node bit then we intend 623 * to loop back around to try again. */ 624 if (clear_node) 625 clear_bit(node, dlm->domain_map); 626 } 627 spin_unlock(&dlm->spinlock); 628 } 629 630 int dlm_joined(struct dlm_ctxt *dlm) 631 { 632 int ret = 0; 633 634 spin_lock(&dlm_domain_lock); 635 636 if (dlm->dlm_state == DLM_CTXT_JOINED) 637 ret = 1; 638 639 spin_unlock(&dlm_domain_lock); 640 641 return ret; 642 } 643 644 int dlm_shutting_down(struct dlm_ctxt *dlm) 645 { 646 int ret = 0; 647 648 spin_lock(&dlm_domain_lock); 649 650 if (dlm->dlm_state == DLM_CTXT_IN_SHUTDOWN) 651 ret = 1; 652 653 spin_unlock(&dlm_domain_lock); 654 655 return ret; 656 } 657 658 void dlm_unregister_domain(struct dlm_ctxt *dlm) 659 { 660 int leave = 0; 661 struct dlm_lock_resource *res; 662 663 spin_lock(&dlm_domain_lock); 664 BUG_ON(dlm->dlm_state != DLM_CTXT_JOINED); 665 BUG_ON(!dlm->num_joins); 666 667 dlm->num_joins--; 668 if (!dlm->num_joins) { 669 /* We mark it "in shutdown" now so new register 670 * requests wait until we've completely left the 671 * domain. Don't use DLM_CTXT_LEAVING yet as we still 672 * want new domain joins to communicate with us at 673 * least until we've completed migration of our 674 * resources. */ 675 dlm->dlm_state = DLM_CTXT_IN_SHUTDOWN; 676 leave = 1; 677 } 678 spin_unlock(&dlm_domain_lock); 679 680 if (leave) { 681 mlog(0, "shutting down domain %s\n", dlm->name); 682 683 /* We changed dlm state, notify the thread */ 684 dlm_kick_thread(dlm, NULL); 685 686 while (dlm_migrate_all_locks(dlm)) { 687 /* Give dlm_thread time to purge the lockres' */ 688 msleep(500); 689 mlog(0, "%s: more migration to do\n", dlm->name); 690 } 691 692 /* This list should be empty. If not, print remaining lockres */ 693 if (!list_empty(&dlm->tracking_list)) { 694 mlog(ML_ERROR, "Following lockres' are still on the " 695 "tracking list:\n"); 696 list_for_each_entry(res, &dlm->tracking_list, tracking) 697 dlm_print_one_lock_resource(res); 698 } 699 700 dlm_mark_domain_leaving(dlm); 701 dlm_leave_domain(dlm); 702 dlm_force_free_mles(dlm); 703 dlm_complete_dlm_shutdown(dlm); 704 } 705 dlm_put(dlm); 706 } 707 EXPORT_SYMBOL_GPL(dlm_unregister_domain); 708 709 static int dlm_query_join_proto_check(char *proto_type, int node, 710 struct dlm_protocol_version *ours, 711 struct dlm_protocol_version *request) 712 { 713 int rc; 714 struct dlm_protocol_version proto = *request; 715 716 if (!dlm_protocol_compare(ours, &proto)) { 717 mlog(0, 718 "node %u wanted to join with %s locking protocol " 719 "%u.%u, we respond with %u.%u\n", 720 node, proto_type, 721 request->pv_major, 722 request->pv_minor, 723 proto.pv_major, proto.pv_minor); 724 request->pv_minor = proto.pv_minor; 725 rc = 0; 726 } else { 727 mlog(ML_NOTICE, 728 "Node %u wanted to join with %s locking " 729 "protocol %u.%u, but we have %u.%u, disallowing\n", 730 node, proto_type, 731 request->pv_major, 732 request->pv_minor, 733 ours->pv_major, 734 ours->pv_minor); 735 rc = 1; 736 } 737 738 return rc; 739 } 740 741 /* 742 * struct dlm_query_join_packet is made up of four one-byte fields. They 743 * are effectively in big-endian order already. However, little-endian 744 * machines swap them before putting the packet on the wire (because 745 * query_join's response is a status, and that status is treated as a u32 746 * on the wire). Thus, a big-endian and little-endian machines will treat 747 * this structure differently. 748 * 749 * The solution is to have little-endian machines swap the structure when 750 * converting from the structure to the u32 representation. This will 751 * result in the structure having the correct format on the wire no matter 752 * the host endian format. 753 */ 754 static void dlm_query_join_packet_to_wire(struct dlm_query_join_packet *packet, 755 u32 *wire) 756 { 757 union dlm_query_join_response response; 758 759 response.packet = *packet; 760 *wire = cpu_to_be32(response.intval); 761 } 762 763 static void dlm_query_join_wire_to_packet(u32 wire, 764 struct dlm_query_join_packet *packet) 765 { 766 union dlm_query_join_response response; 767 768 response.intval = cpu_to_be32(wire); 769 *packet = response.packet; 770 } 771 772 static int dlm_query_join_handler(struct o2net_msg *msg, u32 len, void *data, 773 void **ret_data) 774 { 775 struct dlm_query_join_request *query; 776 struct dlm_query_join_packet packet = { 777 .code = JOIN_DISALLOW, 778 }; 779 struct dlm_ctxt *dlm = NULL; 780 u32 response; 781 u8 nodenum; 782 783 query = (struct dlm_query_join_request *) msg->buf; 784 785 mlog(0, "node %u wants to join domain %s\n", query->node_idx, 786 query->domain); 787 788 /* 789 * If heartbeat doesn't consider the node live, tell it 790 * to back off and try again. This gives heartbeat a chance 791 * to catch up. 792 */ 793 if (!o2hb_check_node_heartbeating(query->node_idx)) { 794 mlog(0, "node %u is not in our live map yet\n", 795 query->node_idx); 796 797 packet.code = JOIN_DISALLOW; 798 goto respond; 799 } 800 801 packet.code = JOIN_OK_NO_MAP; 802 803 spin_lock(&dlm_domain_lock); 804 dlm = __dlm_lookup_domain_full(query->domain, query->name_len); 805 if (!dlm) 806 goto unlock_respond; 807 808 /* 809 * There is a small window where the joining node may not see the 810 * node(s) that just left but still part of the cluster. DISALLOW 811 * join request if joining node has different node map. 812 */ 813 nodenum=0; 814 while (nodenum < O2NM_MAX_NODES) { 815 if (test_bit(nodenum, dlm->domain_map)) { 816 if (!byte_test_bit(nodenum, query->node_map)) { 817 mlog(0, "disallow join as node %u does not " 818 "have node %u in its nodemap\n", 819 query->node_idx, nodenum); 820 packet.code = JOIN_DISALLOW; 821 goto unlock_respond; 822 } 823 } 824 nodenum++; 825 } 826 827 /* Once the dlm ctxt is marked as leaving then we don't want 828 * to be put in someone's domain map. 829 * Also, explicitly disallow joining at certain troublesome 830 * times (ie. during recovery). */ 831 if (dlm && dlm->dlm_state != DLM_CTXT_LEAVING) { 832 int bit = query->node_idx; 833 spin_lock(&dlm->spinlock); 834 835 if (dlm->dlm_state == DLM_CTXT_NEW && 836 dlm->joining_node == DLM_LOCK_RES_OWNER_UNKNOWN) { 837 /*If this is a brand new context and we 838 * haven't started our join process yet, then 839 * the other node won the race. */ 840 packet.code = JOIN_OK_NO_MAP; 841 } else if (dlm->joining_node != DLM_LOCK_RES_OWNER_UNKNOWN) { 842 /* Disallow parallel joins. */ 843 packet.code = JOIN_DISALLOW; 844 } else if (dlm->reco.state & DLM_RECO_STATE_ACTIVE) { 845 mlog(0, "node %u trying to join, but recovery " 846 "is ongoing.\n", bit); 847 packet.code = JOIN_DISALLOW; 848 } else if (test_bit(bit, dlm->recovery_map)) { 849 mlog(0, "node %u trying to join, but it " 850 "still needs recovery.\n", bit); 851 packet.code = JOIN_DISALLOW; 852 } else if (test_bit(bit, dlm->domain_map)) { 853 mlog(0, "node %u trying to join, but it " 854 "is still in the domain! needs recovery?\n", 855 bit); 856 packet.code = JOIN_DISALLOW; 857 } else { 858 /* Alright we're fully a part of this domain 859 * so we keep some state as to who's joining 860 * and indicate to him that needs to be fixed 861 * up. */ 862 863 /* Make sure we speak compatible locking protocols. */ 864 if (dlm_query_join_proto_check("DLM", bit, 865 &dlm->dlm_locking_proto, 866 &query->dlm_proto)) { 867 packet.code = JOIN_PROTOCOL_MISMATCH; 868 } else if (dlm_query_join_proto_check("fs", bit, 869 &dlm->fs_locking_proto, 870 &query->fs_proto)) { 871 packet.code = JOIN_PROTOCOL_MISMATCH; 872 } else { 873 packet.dlm_minor = query->dlm_proto.pv_minor; 874 packet.fs_minor = query->fs_proto.pv_minor; 875 packet.code = JOIN_OK; 876 __dlm_set_joining_node(dlm, query->node_idx); 877 } 878 } 879 880 spin_unlock(&dlm->spinlock); 881 } 882 unlock_respond: 883 spin_unlock(&dlm_domain_lock); 884 885 respond: 886 mlog(0, "We respond with %u\n", packet.code); 887 888 dlm_query_join_packet_to_wire(&packet, &response); 889 return response; 890 } 891 892 static int dlm_assert_joined_handler(struct o2net_msg *msg, u32 len, void *data, 893 void **ret_data) 894 { 895 struct dlm_assert_joined *assert; 896 struct dlm_ctxt *dlm = NULL; 897 898 assert = (struct dlm_assert_joined *) msg->buf; 899 900 mlog(0, "node %u asserts join on domain %s\n", assert->node_idx, 901 assert->domain); 902 903 spin_lock(&dlm_domain_lock); 904 dlm = __dlm_lookup_domain_full(assert->domain, assert->name_len); 905 /* XXX should we consider no dlm ctxt an error? */ 906 if (dlm) { 907 spin_lock(&dlm->spinlock); 908 909 /* Alright, this node has officially joined our 910 * domain. Set him in the map and clean up our 911 * leftover join state. */ 912 BUG_ON(dlm->joining_node != assert->node_idx); 913 set_bit(assert->node_idx, dlm->domain_map); 914 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); 915 916 printk(KERN_NOTICE "o2dlm: Node %u joins domain %s\n", 917 assert->node_idx, dlm->name); 918 __dlm_print_nodes(dlm); 919 920 /* notify anything attached to the heartbeat events */ 921 dlm_hb_event_notify_attached(dlm, assert->node_idx, 1); 922 923 spin_unlock(&dlm->spinlock); 924 } 925 spin_unlock(&dlm_domain_lock); 926 927 return 0; 928 } 929 930 static int dlm_match_regions(struct dlm_ctxt *dlm, 931 struct dlm_query_region *qr) 932 { 933 char *local = NULL, *remote = qr->qr_regions; 934 char *l, *r; 935 int localnr, i, j, foundit; 936 int status = 0; 937 938 if (!o2hb_global_heartbeat_active()) { 939 if (qr->qr_numregions) { 940 mlog(ML_ERROR, "Domain %s: Joining node %d has global " 941 "heartbeat enabled but local node %d does not\n", 942 qr->qr_domain, qr->qr_node, dlm->node_num); 943 status = -EINVAL; 944 } 945 goto bail; 946 } 947 948 if (o2hb_global_heartbeat_active() && !qr->qr_numregions) { 949 mlog(ML_ERROR, "Domain %s: Local node %d has global " 950 "heartbeat enabled but joining node %d does not\n", 951 qr->qr_domain, dlm->node_num, qr->qr_node); 952 status = -EINVAL; 953 goto bail; 954 } 955 956 r = remote; 957 for (i = 0; i < qr->qr_numregions; ++i) { 958 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, r); 959 r += O2HB_MAX_REGION_NAME_LEN; 960 } 961 962 local = kmalloc(sizeof(qr->qr_regions), GFP_KERNEL); 963 if (!local) { 964 status = -ENOMEM; 965 goto bail; 966 } 967 968 localnr = o2hb_get_all_regions(local, O2NM_MAX_REGIONS); 969 970 /* compare local regions with remote */ 971 l = local; 972 for (i = 0; i < localnr; ++i) { 973 foundit = 0; 974 r = remote; 975 for (j = 0; j <= qr->qr_numregions; ++j) { 976 if (!memcmp(l, r, O2HB_MAX_REGION_NAME_LEN)) { 977 foundit = 1; 978 break; 979 } 980 r += O2HB_MAX_REGION_NAME_LEN; 981 } 982 if (!foundit) { 983 status = -EINVAL; 984 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered " 985 "in local node %d but not in joining node %d\n", 986 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, l, 987 dlm->node_num, qr->qr_node); 988 goto bail; 989 } 990 l += O2HB_MAX_REGION_NAME_LEN; 991 } 992 993 /* compare remote with local regions */ 994 r = remote; 995 for (i = 0; i < qr->qr_numregions; ++i) { 996 foundit = 0; 997 l = local; 998 for (j = 0; j < localnr; ++j) { 999 if (!memcmp(r, l, O2HB_MAX_REGION_NAME_LEN)) { 1000 foundit = 1; 1001 break; 1002 } 1003 l += O2HB_MAX_REGION_NAME_LEN; 1004 } 1005 if (!foundit) { 1006 status = -EINVAL; 1007 mlog(ML_ERROR, "Domain %s: Region '%.*s' registered " 1008 "in joining node %d but not in local node %d\n", 1009 qr->qr_domain, O2HB_MAX_REGION_NAME_LEN, r, 1010 qr->qr_node, dlm->node_num); 1011 goto bail; 1012 } 1013 r += O2HB_MAX_REGION_NAME_LEN; 1014 } 1015 1016 bail: 1017 kfree(local); 1018 1019 return status; 1020 } 1021 1022 static int dlm_send_regions(struct dlm_ctxt *dlm, unsigned long *node_map) 1023 { 1024 struct dlm_query_region *qr = NULL; 1025 int status, ret = 0, i; 1026 char *p; 1027 1028 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES) 1029 goto bail; 1030 1031 qr = kzalloc(sizeof(struct dlm_query_region), GFP_KERNEL); 1032 if (!qr) { 1033 ret = -ENOMEM; 1034 mlog_errno(ret); 1035 goto bail; 1036 } 1037 1038 qr->qr_node = dlm->node_num; 1039 qr->qr_namelen = strlen(dlm->name); 1040 memcpy(qr->qr_domain, dlm->name, qr->qr_namelen); 1041 /* if local hb, the numregions will be zero */ 1042 if (o2hb_global_heartbeat_active()) 1043 qr->qr_numregions = o2hb_get_all_regions(qr->qr_regions, 1044 O2NM_MAX_REGIONS); 1045 1046 p = qr->qr_regions; 1047 for (i = 0; i < qr->qr_numregions; ++i, p += O2HB_MAX_REGION_NAME_LEN) 1048 mlog(0, "Region %.*s\n", O2HB_MAX_REGION_NAME_LEN, p); 1049 1050 i = -1; 1051 while ((i = find_next_bit(node_map, O2NM_MAX_NODES, 1052 i + 1)) < O2NM_MAX_NODES) { 1053 if (i == dlm->node_num) 1054 continue; 1055 1056 mlog(0, "Sending regions to node %d\n", i); 1057 1058 ret = o2net_send_message(DLM_QUERY_REGION, DLM_MOD_KEY, qr, 1059 sizeof(struct dlm_query_region), 1060 i, &status); 1061 if (ret >= 0) 1062 ret = status; 1063 if (ret) { 1064 mlog(ML_ERROR, "Region mismatch %d, node %d\n", 1065 ret, i); 1066 break; 1067 } 1068 } 1069 1070 bail: 1071 kfree(qr); 1072 return ret; 1073 } 1074 1075 static int dlm_query_region_handler(struct o2net_msg *msg, u32 len, 1076 void *data, void **ret_data) 1077 { 1078 struct dlm_query_region *qr; 1079 struct dlm_ctxt *dlm = NULL; 1080 int status = 0; 1081 int locked = 0; 1082 1083 qr = (struct dlm_query_region *) msg->buf; 1084 1085 mlog(0, "Node %u queries hb regions on domain %s\n", qr->qr_node, 1086 qr->qr_domain); 1087 1088 status = -EINVAL; 1089 1090 spin_lock(&dlm_domain_lock); 1091 dlm = __dlm_lookup_domain_full(qr->qr_domain, qr->qr_namelen); 1092 if (!dlm) { 1093 mlog(ML_ERROR, "Node %d queried hb regions on domain %s " 1094 "before join domain\n", qr->qr_node, qr->qr_domain); 1095 goto bail; 1096 } 1097 1098 spin_lock(&dlm->spinlock); 1099 locked = 1; 1100 if (dlm->joining_node != qr->qr_node) { 1101 mlog(ML_ERROR, "Node %d queried hb regions on domain %s " 1102 "but joining node is %d\n", qr->qr_node, qr->qr_domain, 1103 dlm->joining_node); 1104 goto bail; 1105 } 1106 1107 /* Support for global heartbeat was added in 1.1 */ 1108 if (dlm->dlm_locking_proto.pv_major == 1 && 1109 dlm->dlm_locking_proto.pv_minor == 0) { 1110 mlog(ML_ERROR, "Node %d queried hb regions on domain %s " 1111 "but active dlm protocol is %d.%d\n", qr->qr_node, 1112 qr->qr_domain, dlm->dlm_locking_proto.pv_major, 1113 dlm->dlm_locking_proto.pv_minor); 1114 goto bail; 1115 } 1116 1117 status = dlm_match_regions(dlm, qr); 1118 1119 bail: 1120 if (locked) 1121 spin_unlock(&dlm->spinlock); 1122 spin_unlock(&dlm_domain_lock); 1123 1124 return status; 1125 } 1126 1127 static int dlm_match_nodes(struct dlm_ctxt *dlm, struct dlm_query_nodeinfo *qn) 1128 { 1129 struct o2nm_node *local; 1130 struct dlm_node_info *remote; 1131 int i, j; 1132 int status = 0; 1133 1134 for (j = 0; j < qn->qn_numnodes; ++j) 1135 mlog(0, "Node %3d, %pI4:%u\n", qn->qn_nodes[j].ni_nodenum, 1136 &(qn->qn_nodes[j].ni_ipv4_address), 1137 ntohs(qn->qn_nodes[j].ni_ipv4_port)); 1138 1139 for (i = 0; i < O2NM_MAX_NODES && !status; ++i) { 1140 local = o2nm_get_node_by_num(i); 1141 remote = NULL; 1142 for (j = 0; j < qn->qn_numnodes; ++j) { 1143 if (qn->qn_nodes[j].ni_nodenum == i) { 1144 remote = &(qn->qn_nodes[j]); 1145 break; 1146 } 1147 } 1148 1149 if (!local && !remote) 1150 continue; 1151 1152 if ((local && !remote) || (!local && remote)) 1153 status = -EINVAL; 1154 1155 if (!status && 1156 ((remote->ni_nodenum != local->nd_num) || 1157 (remote->ni_ipv4_port != local->nd_ipv4_port) || 1158 (remote->ni_ipv4_address != local->nd_ipv4_address))) 1159 status = -EINVAL; 1160 1161 if (status) { 1162 if (remote && !local) 1163 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) " 1164 "registered in joining node %d but not in " 1165 "local node %d\n", qn->qn_domain, 1166 remote->ni_nodenum, 1167 &(remote->ni_ipv4_address), 1168 ntohs(remote->ni_ipv4_port), 1169 qn->qn_nodenum, dlm->node_num); 1170 if (local && !remote) 1171 mlog(ML_ERROR, "Domain %s: Node %d (%pI4:%u) " 1172 "registered in local node %d but not in " 1173 "joining node %d\n", qn->qn_domain, 1174 local->nd_num, &(local->nd_ipv4_address), 1175 ntohs(local->nd_ipv4_port), 1176 dlm->node_num, qn->qn_nodenum); 1177 BUG_ON((!local && !remote)); 1178 } 1179 1180 if (local) 1181 o2nm_node_put(local); 1182 } 1183 1184 return status; 1185 } 1186 1187 static int dlm_send_nodeinfo(struct dlm_ctxt *dlm, unsigned long *node_map) 1188 { 1189 struct dlm_query_nodeinfo *qn = NULL; 1190 struct o2nm_node *node; 1191 int ret = 0, status, count, i; 1192 1193 if (find_next_bit(node_map, O2NM_MAX_NODES, 0) >= O2NM_MAX_NODES) 1194 goto bail; 1195 1196 qn = kzalloc(sizeof(struct dlm_query_nodeinfo), GFP_KERNEL); 1197 if (!qn) { 1198 ret = -ENOMEM; 1199 mlog_errno(ret); 1200 goto bail; 1201 } 1202 1203 for (i = 0, count = 0; i < O2NM_MAX_NODES; ++i) { 1204 node = o2nm_get_node_by_num(i); 1205 if (!node) 1206 continue; 1207 qn->qn_nodes[count].ni_nodenum = node->nd_num; 1208 qn->qn_nodes[count].ni_ipv4_port = node->nd_ipv4_port; 1209 qn->qn_nodes[count].ni_ipv4_address = node->nd_ipv4_address; 1210 mlog(0, "Node %3d, %pI4:%u\n", node->nd_num, 1211 &(node->nd_ipv4_address), ntohs(node->nd_ipv4_port)); 1212 ++count; 1213 o2nm_node_put(node); 1214 } 1215 1216 qn->qn_nodenum = dlm->node_num; 1217 qn->qn_numnodes = count; 1218 qn->qn_namelen = strlen(dlm->name); 1219 memcpy(qn->qn_domain, dlm->name, qn->qn_namelen); 1220 1221 i = -1; 1222 while ((i = find_next_bit(node_map, O2NM_MAX_NODES, 1223 i + 1)) < O2NM_MAX_NODES) { 1224 if (i == dlm->node_num) 1225 continue; 1226 1227 mlog(0, "Sending nodeinfo to node %d\n", i); 1228 1229 ret = o2net_send_message(DLM_QUERY_NODEINFO, DLM_MOD_KEY, 1230 qn, sizeof(struct dlm_query_nodeinfo), 1231 i, &status); 1232 if (ret >= 0) 1233 ret = status; 1234 if (ret) { 1235 mlog(ML_ERROR, "node mismatch %d, node %d\n", ret, i); 1236 break; 1237 } 1238 } 1239 1240 bail: 1241 kfree(qn); 1242 return ret; 1243 } 1244 1245 static int dlm_query_nodeinfo_handler(struct o2net_msg *msg, u32 len, 1246 void *data, void **ret_data) 1247 { 1248 struct dlm_query_nodeinfo *qn; 1249 struct dlm_ctxt *dlm = NULL; 1250 int locked = 0, status = -EINVAL; 1251 1252 qn = (struct dlm_query_nodeinfo *) msg->buf; 1253 1254 mlog(0, "Node %u queries nodes on domain %s\n", qn->qn_nodenum, 1255 qn->qn_domain); 1256 1257 spin_lock(&dlm_domain_lock); 1258 dlm = __dlm_lookup_domain_full(qn->qn_domain, qn->qn_namelen); 1259 if (!dlm) { 1260 mlog(ML_ERROR, "Node %d queried nodes on domain %s before " 1261 "join domain\n", qn->qn_nodenum, qn->qn_domain); 1262 goto bail; 1263 } 1264 1265 spin_lock(&dlm->spinlock); 1266 locked = 1; 1267 if (dlm->joining_node != qn->qn_nodenum) { 1268 mlog(ML_ERROR, "Node %d queried nodes on domain %s but " 1269 "joining node is %d\n", qn->qn_nodenum, qn->qn_domain, 1270 dlm->joining_node); 1271 goto bail; 1272 } 1273 1274 /* Support for node query was added in 1.1 */ 1275 if (dlm->dlm_locking_proto.pv_major == 1 && 1276 dlm->dlm_locking_proto.pv_minor == 0) { 1277 mlog(ML_ERROR, "Node %d queried nodes on domain %s " 1278 "but active dlm protocol is %d.%d\n", qn->qn_nodenum, 1279 qn->qn_domain, dlm->dlm_locking_proto.pv_major, 1280 dlm->dlm_locking_proto.pv_minor); 1281 goto bail; 1282 } 1283 1284 status = dlm_match_nodes(dlm, qn); 1285 1286 bail: 1287 if (locked) 1288 spin_unlock(&dlm->spinlock); 1289 spin_unlock(&dlm_domain_lock); 1290 1291 return status; 1292 } 1293 1294 static int dlm_cancel_join_handler(struct o2net_msg *msg, u32 len, void *data, 1295 void **ret_data) 1296 { 1297 struct dlm_cancel_join *cancel; 1298 struct dlm_ctxt *dlm = NULL; 1299 1300 cancel = (struct dlm_cancel_join *) msg->buf; 1301 1302 mlog(0, "node %u cancels join on domain %s\n", cancel->node_idx, 1303 cancel->domain); 1304 1305 spin_lock(&dlm_domain_lock); 1306 dlm = __dlm_lookup_domain_full(cancel->domain, cancel->name_len); 1307 1308 if (dlm) { 1309 spin_lock(&dlm->spinlock); 1310 1311 /* Yikes, this guy wants to cancel his join. No 1312 * problem, we simply cleanup our join state. */ 1313 BUG_ON(dlm->joining_node != cancel->node_idx); 1314 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); 1315 1316 spin_unlock(&dlm->spinlock); 1317 } 1318 spin_unlock(&dlm_domain_lock); 1319 1320 return 0; 1321 } 1322 1323 static int dlm_send_one_join_cancel(struct dlm_ctxt *dlm, 1324 unsigned int node) 1325 { 1326 int status; 1327 struct dlm_cancel_join cancel_msg; 1328 1329 memset(&cancel_msg, 0, sizeof(cancel_msg)); 1330 cancel_msg.node_idx = dlm->node_num; 1331 cancel_msg.name_len = strlen(dlm->name); 1332 memcpy(cancel_msg.domain, dlm->name, cancel_msg.name_len); 1333 1334 status = o2net_send_message(DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY, 1335 &cancel_msg, sizeof(cancel_msg), node, 1336 NULL); 1337 if (status < 0) { 1338 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " 1339 "node %u\n", status, DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY, 1340 node); 1341 goto bail; 1342 } 1343 1344 bail: 1345 return status; 1346 } 1347 1348 /* map_size should be in bytes. */ 1349 static int dlm_send_join_cancels(struct dlm_ctxt *dlm, 1350 unsigned long *node_map, 1351 unsigned int map_size) 1352 { 1353 int status, tmpstat; 1354 unsigned int node; 1355 1356 if (map_size != (BITS_TO_LONGS(O2NM_MAX_NODES) * 1357 sizeof(unsigned long))) { 1358 mlog(ML_ERROR, 1359 "map_size %u != BITS_TO_LONGS(O2NM_MAX_NODES) %u\n", 1360 map_size, (unsigned)BITS_TO_LONGS(O2NM_MAX_NODES)); 1361 return -EINVAL; 1362 } 1363 1364 status = 0; 1365 node = -1; 1366 while ((node = find_next_bit(node_map, O2NM_MAX_NODES, 1367 node + 1)) < O2NM_MAX_NODES) { 1368 if (node == dlm->node_num) 1369 continue; 1370 1371 tmpstat = dlm_send_one_join_cancel(dlm, node); 1372 if (tmpstat) { 1373 mlog(ML_ERROR, "Error return %d cancelling join on " 1374 "node %d\n", tmpstat, node); 1375 if (!status) 1376 status = tmpstat; 1377 } 1378 } 1379 1380 if (status) 1381 mlog_errno(status); 1382 return status; 1383 } 1384 1385 static int dlm_request_join(struct dlm_ctxt *dlm, 1386 int node, 1387 enum dlm_query_join_response_code *response) 1388 { 1389 int status; 1390 struct dlm_query_join_request join_msg; 1391 struct dlm_query_join_packet packet; 1392 u32 join_resp; 1393 1394 mlog(0, "querying node %d\n", node); 1395 1396 memset(&join_msg, 0, sizeof(join_msg)); 1397 join_msg.node_idx = dlm->node_num; 1398 join_msg.name_len = strlen(dlm->name); 1399 memcpy(join_msg.domain, dlm->name, join_msg.name_len); 1400 join_msg.dlm_proto = dlm->dlm_locking_proto; 1401 join_msg.fs_proto = dlm->fs_locking_proto; 1402 1403 /* copy live node map to join message */ 1404 byte_copymap(join_msg.node_map, dlm->live_nodes_map, O2NM_MAX_NODES); 1405 1406 status = o2net_send_message(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, &join_msg, 1407 sizeof(join_msg), node, &join_resp); 1408 if (status < 0 && status != -ENOPROTOOPT) { 1409 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " 1410 "node %u\n", status, DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, 1411 node); 1412 goto bail; 1413 } 1414 dlm_query_join_wire_to_packet(join_resp, &packet); 1415 1416 /* -ENOPROTOOPT from the net code means the other side isn't 1417 listening for our message type -- that's fine, it means 1418 his dlm isn't up, so we can consider him a 'yes' but not 1419 joined into the domain. */ 1420 if (status == -ENOPROTOOPT) { 1421 status = 0; 1422 *response = JOIN_OK_NO_MAP; 1423 } else if (packet.code == JOIN_DISALLOW || 1424 packet.code == JOIN_OK_NO_MAP) { 1425 *response = packet.code; 1426 } else if (packet.code == JOIN_PROTOCOL_MISMATCH) { 1427 mlog(ML_NOTICE, 1428 "This node requested DLM locking protocol %u.%u and " 1429 "filesystem locking protocol %u.%u. At least one of " 1430 "the protocol versions on node %d is not compatible, " 1431 "disconnecting\n", 1432 dlm->dlm_locking_proto.pv_major, 1433 dlm->dlm_locking_proto.pv_minor, 1434 dlm->fs_locking_proto.pv_major, 1435 dlm->fs_locking_proto.pv_minor, 1436 node); 1437 status = -EPROTO; 1438 *response = packet.code; 1439 } else if (packet.code == JOIN_OK) { 1440 *response = packet.code; 1441 /* Use the same locking protocol as the remote node */ 1442 dlm->dlm_locking_proto.pv_minor = packet.dlm_minor; 1443 dlm->fs_locking_proto.pv_minor = packet.fs_minor; 1444 mlog(0, 1445 "Node %d responds JOIN_OK with DLM locking protocol " 1446 "%u.%u and fs locking protocol %u.%u\n", 1447 node, 1448 dlm->dlm_locking_proto.pv_major, 1449 dlm->dlm_locking_proto.pv_minor, 1450 dlm->fs_locking_proto.pv_major, 1451 dlm->fs_locking_proto.pv_minor); 1452 } else { 1453 status = -EINVAL; 1454 mlog(ML_ERROR, "invalid response %d from node %u\n", 1455 packet.code, node); 1456 } 1457 1458 mlog(0, "status %d, node %d response is %d\n", status, node, 1459 *response); 1460 1461 bail: 1462 return status; 1463 } 1464 1465 static int dlm_send_one_join_assert(struct dlm_ctxt *dlm, 1466 unsigned int node) 1467 { 1468 int status; 1469 struct dlm_assert_joined assert_msg; 1470 1471 mlog(0, "Sending join assert to node %u\n", node); 1472 1473 memset(&assert_msg, 0, sizeof(assert_msg)); 1474 assert_msg.node_idx = dlm->node_num; 1475 assert_msg.name_len = strlen(dlm->name); 1476 memcpy(assert_msg.domain, dlm->name, assert_msg.name_len); 1477 1478 status = o2net_send_message(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY, 1479 &assert_msg, sizeof(assert_msg), node, 1480 NULL); 1481 if (status < 0) 1482 mlog(ML_ERROR, "Error %d when sending message %u (key 0x%x) to " 1483 "node %u\n", status, DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY, 1484 node); 1485 1486 return status; 1487 } 1488 1489 static void dlm_send_join_asserts(struct dlm_ctxt *dlm, 1490 unsigned long *node_map) 1491 { 1492 int status, node, live; 1493 1494 status = 0; 1495 node = -1; 1496 while ((node = find_next_bit(node_map, O2NM_MAX_NODES, 1497 node + 1)) < O2NM_MAX_NODES) { 1498 if (node == dlm->node_num) 1499 continue; 1500 1501 do { 1502 /* It is very important that this message be 1503 * received so we spin until either the node 1504 * has died or it gets the message. */ 1505 status = dlm_send_one_join_assert(dlm, node); 1506 1507 spin_lock(&dlm->spinlock); 1508 live = test_bit(node, dlm->live_nodes_map); 1509 spin_unlock(&dlm->spinlock); 1510 1511 if (status) { 1512 mlog(ML_ERROR, "Error return %d asserting " 1513 "join on node %d\n", status, node); 1514 1515 /* give us some time between errors... */ 1516 if (live) 1517 msleep(DLM_DOMAIN_BACKOFF_MS); 1518 } 1519 } while (status && live); 1520 } 1521 } 1522 1523 struct domain_join_ctxt { 1524 unsigned long live_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1525 unsigned long yes_resp_map[BITS_TO_LONGS(O2NM_MAX_NODES)]; 1526 }; 1527 1528 static int dlm_should_restart_join(struct dlm_ctxt *dlm, 1529 struct domain_join_ctxt *ctxt, 1530 enum dlm_query_join_response_code response) 1531 { 1532 int ret; 1533 1534 if (response == JOIN_DISALLOW) { 1535 mlog(0, "Latest response of disallow -- should restart\n"); 1536 return 1; 1537 } 1538 1539 spin_lock(&dlm->spinlock); 1540 /* For now, we restart the process if the node maps have 1541 * changed at all */ 1542 ret = memcmp(ctxt->live_map, dlm->live_nodes_map, 1543 sizeof(dlm->live_nodes_map)); 1544 spin_unlock(&dlm->spinlock); 1545 1546 if (ret) 1547 mlog(0, "Node maps changed -- should restart\n"); 1548 1549 return ret; 1550 } 1551 1552 static int dlm_try_to_join_domain(struct dlm_ctxt *dlm) 1553 { 1554 int status = 0, tmpstat, node; 1555 struct domain_join_ctxt *ctxt; 1556 enum dlm_query_join_response_code response = JOIN_DISALLOW; 1557 1558 mlog_entry("%p", dlm); 1559 1560 ctxt = kzalloc(sizeof(*ctxt), GFP_KERNEL); 1561 if (!ctxt) { 1562 status = -ENOMEM; 1563 mlog_errno(status); 1564 goto bail; 1565 } 1566 1567 /* group sem locking should work for us here -- we're already 1568 * registered for heartbeat events so filling this should be 1569 * atomic wrt getting those handlers called. */ 1570 o2hb_fill_node_map(dlm->live_nodes_map, sizeof(dlm->live_nodes_map)); 1571 1572 spin_lock(&dlm->spinlock); 1573 memcpy(ctxt->live_map, dlm->live_nodes_map, sizeof(ctxt->live_map)); 1574 1575 __dlm_set_joining_node(dlm, dlm->node_num); 1576 1577 spin_unlock(&dlm->spinlock); 1578 1579 node = -1; 1580 while ((node = find_next_bit(ctxt->live_map, O2NM_MAX_NODES, 1581 node + 1)) < O2NM_MAX_NODES) { 1582 if (node == dlm->node_num) 1583 continue; 1584 1585 status = dlm_request_join(dlm, node, &response); 1586 if (status < 0) { 1587 mlog_errno(status); 1588 goto bail; 1589 } 1590 1591 /* Ok, either we got a response or the node doesn't have a 1592 * dlm up. */ 1593 if (response == JOIN_OK) 1594 set_bit(node, ctxt->yes_resp_map); 1595 1596 if (dlm_should_restart_join(dlm, ctxt, response)) { 1597 status = -EAGAIN; 1598 goto bail; 1599 } 1600 } 1601 1602 mlog(0, "Yay, done querying nodes!\n"); 1603 1604 /* Yay, everyone agree's we can join the domain. My domain is 1605 * comprised of all nodes who were put in the 1606 * yes_resp_map. Copy that into our domain map and send a join 1607 * assert message to clean up everyone elses state. */ 1608 spin_lock(&dlm->spinlock); 1609 memcpy(dlm->domain_map, ctxt->yes_resp_map, 1610 sizeof(ctxt->yes_resp_map)); 1611 set_bit(dlm->node_num, dlm->domain_map); 1612 spin_unlock(&dlm->spinlock); 1613 1614 /* Support for global heartbeat and node info was added in 1.1 */ 1615 if (dlm_protocol.pv_major > 1 || dlm_protocol.pv_minor > 0) { 1616 status = dlm_send_nodeinfo(dlm, ctxt->yes_resp_map); 1617 if (status) { 1618 mlog_errno(status); 1619 goto bail; 1620 } 1621 status = dlm_send_regions(dlm, ctxt->yes_resp_map); 1622 if (status) { 1623 mlog_errno(status); 1624 goto bail; 1625 } 1626 } 1627 1628 dlm_send_join_asserts(dlm, ctxt->yes_resp_map); 1629 1630 /* Joined state *must* be set before the joining node 1631 * information, otherwise the query_join handler may read no 1632 * current joiner but a state of NEW and tell joining nodes 1633 * we're not in the domain. */ 1634 spin_lock(&dlm_domain_lock); 1635 dlm->dlm_state = DLM_CTXT_JOINED; 1636 dlm->num_joins++; 1637 spin_unlock(&dlm_domain_lock); 1638 1639 bail: 1640 spin_lock(&dlm->spinlock); 1641 __dlm_set_joining_node(dlm, DLM_LOCK_RES_OWNER_UNKNOWN); 1642 if (!status) 1643 __dlm_print_nodes(dlm); 1644 spin_unlock(&dlm->spinlock); 1645 1646 if (ctxt) { 1647 /* Do we need to send a cancel message to any nodes? */ 1648 if (status < 0) { 1649 tmpstat = dlm_send_join_cancels(dlm, 1650 ctxt->yes_resp_map, 1651 sizeof(ctxt->yes_resp_map)); 1652 if (tmpstat < 0) 1653 mlog_errno(tmpstat); 1654 } 1655 kfree(ctxt); 1656 } 1657 1658 mlog(0, "returning %d\n", status); 1659 return status; 1660 } 1661 1662 static void dlm_unregister_domain_handlers(struct dlm_ctxt *dlm) 1663 { 1664 o2hb_unregister_callback(NULL, &dlm->dlm_hb_up); 1665 o2hb_unregister_callback(NULL, &dlm->dlm_hb_down); 1666 o2net_unregister_handler_list(&dlm->dlm_domain_handlers); 1667 } 1668 1669 static int dlm_register_domain_handlers(struct dlm_ctxt *dlm) 1670 { 1671 int status; 1672 1673 mlog(0, "registering handlers.\n"); 1674 1675 o2hb_setup_callback(&dlm->dlm_hb_down, O2HB_NODE_DOWN_CB, 1676 dlm_hb_node_down_cb, dlm, DLM_HB_NODE_DOWN_PRI); 1677 status = o2hb_register_callback(NULL, &dlm->dlm_hb_down); 1678 if (status) 1679 goto bail; 1680 1681 o2hb_setup_callback(&dlm->dlm_hb_up, O2HB_NODE_UP_CB, 1682 dlm_hb_node_up_cb, dlm, DLM_HB_NODE_UP_PRI); 1683 status = o2hb_register_callback(NULL, &dlm->dlm_hb_up); 1684 if (status) 1685 goto bail; 1686 1687 status = o2net_register_handler(DLM_MASTER_REQUEST_MSG, dlm->key, 1688 sizeof(struct dlm_master_request), 1689 dlm_master_request_handler, 1690 dlm, NULL, &dlm->dlm_domain_handlers); 1691 if (status) 1692 goto bail; 1693 1694 status = o2net_register_handler(DLM_ASSERT_MASTER_MSG, dlm->key, 1695 sizeof(struct dlm_assert_master), 1696 dlm_assert_master_handler, 1697 dlm, dlm_assert_master_post_handler, 1698 &dlm->dlm_domain_handlers); 1699 if (status) 1700 goto bail; 1701 1702 status = o2net_register_handler(DLM_CREATE_LOCK_MSG, dlm->key, 1703 sizeof(struct dlm_create_lock), 1704 dlm_create_lock_handler, 1705 dlm, NULL, &dlm->dlm_domain_handlers); 1706 if (status) 1707 goto bail; 1708 1709 status = o2net_register_handler(DLM_CONVERT_LOCK_MSG, dlm->key, 1710 DLM_CONVERT_LOCK_MAX_LEN, 1711 dlm_convert_lock_handler, 1712 dlm, NULL, &dlm->dlm_domain_handlers); 1713 if (status) 1714 goto bail; 1715 1716 status = o2net_register_handler(DLM_UNLOCK_LOCK_MSG, dlm->key, 1717 DLM_UNLOCK_LOCK_MAX_LEN, 1718 dlm_unlock_lock_handler, 1719 dlm, NULL, &dlm->dlm_domain_handlers); 1720 if (status) 1721 goto bail; 1722 1723 status = o2net_register_handler(DLM_PROXY_AST_MSG, dlm->key, 1724 DLM_PROXY_AST_MAX_LEN, 1725 dlm_proxy_ast_handler, 1726 dlm, NULL, &dlm->dlm_domain_handlers); 1727 if (status) 1728 goto bail; 1729 1730 status = o2net_register_handler(DLM_EXIT_DOMAIN_MSG, dlm->key, 1731 sizeof(struct dlm_exit_domain), 1732 dlm_exit_domain_handler, 1733 dlm, NULL, &dlm->dlm_domain_handlers); 1734 if (status) 1735 goto bail; 1736 1737 status = o2net_register_handler(DLM_DEREF_LOCKRES_MSG, dlm->key, 1738 sizeof(struct dlm_deref_lockres), 1739 dlm_deref_lockres_handler, 1740 dlm, NULL, &dlm->dlm_domain_handlers); 1741 if (status) 1742 goto bail; 1743 1744 status = o2net_register_handler(DLM_MIGRATE_REQUEST_MSG, dlm->key, 1745 sizeof(struct dlm_migrate_request), 1746 dlm_migrate_request_handler, 1747 dlm, NULL, &dlm->dlm_domain_handlers); 1748 if (status) 1749 goto bail; 1750 1751 status = o2net_register_handler(DLM_MIG_LOCKRES_MSG, dlm->key, 1752 DLM_MIG_LOCKRES_MAX_LEN, 1753 dlm_mig_lockres_handler, 1754 dlm, NULL, &dlm->dlm_domain_handlers); 1755 if (status) 1756 goto bail; 1757 1758 status = o2net_register_handler(DLM_MASTER_REQUERY_MSG, dlm->key, 1759 sizeof(struct dlm_master_requery), 1760 dlm_master_requery_handler, 1761 dlm, NULL, &dlm->dlm_domain_handlers); 1762 if (status) 1763 goto bail; 1764 1765 status = o2net_register_handler(DLM_LOCK_REQUEST_MSG, dlm->key, 1766 sizeof(struct dlm_lock_request), 1767 dlm_request_all_locks_handler, 1768 dlm, NULL, &dlm->dlm_domain_handlers); 1769 if (status) 1770 goto bail; 1771 1772 status = o2net_register_handler(DLM_RECO_DATA_DONE_MSG, dlm->key, 1773 sizeof(struct dlm_reco_data_done), 1774 dlm_reco_data_done_handler, 1775 dlm, NULL, &dlm->dlm_domain_handlers); 1776 if (status) 1777 goto bail; 1778 1779 status = o2net_register_handler(DLM_BEGIN_RECO_MSG, dlm->key, 1780 sizeof(struct dlm_begin_reco), 1781 dlm_begin_reco_handler, 1782 dlm, NULL, &dlm->dlm_domain_handlers); 1783 if (status) 1784 goto bail; 1785 1786 status = o2net_register_handler(DLM_FINALIZE_RECO_MSG, dlm->key, 1787 sizeof(struct dlm_finalize_reco), 1788 dlm_finalize_reco_handler, 1789 dlm, NULL, &dlm->dlm_domain_handlers); 1790 if (status) 1791 goto bail; 1792 1793 bail: 1794 if (status) 1795 dlm_unregister_domain_handlers(dlm); 1796 1797 return status; 1798 } 1799 1800 static int dlm_join_domain(struct dlm_ctxt *dlm) 1801 { 1802 int status; 1803 unsigned int backoff; 1804 unsigned int total_backoff = 0; 1805 1806 BUG_ON(!dlm); 1807 1808 mlog(0, "Join domain %s\n", dlm->name); 1809 1810 status = dlm_register_domain_handlers(dlm); 1811 if (status) { 1812 mlog_errno(status); 1813 goto bail; 1814 } 1815 1816 status = dlm_debug_init(dlm); 1817 if (status < 0) { 1818 mlog_errno(status); 1819 goto bail; 1820 } 1821 1822 status = dlm_launch_thread(dlm); 1823 if (status < 0) { 1824 mlog_errno(status); 1825 goto bail; 1826 } 1827 1828 status = dlm_launch_recovery_thread(dlm); 1829 if (status < 0) { 1830 mlog_errno(status); 1831 goto bail; 1832 } 1833 1834 dlm->dlm_worker = create_singlethread_workqueue("dlm_wq"); 1835 if (!dlm->dlm_worker) { 1836 status = -ENOMEM; 1837 mlog_errno(status); 1838 goto bail; 1839 } 1840 1841 do { 1842 status = dlm_try_to_join_domain(dlm); 1843 1844 /* If we're racing another node to the join, then we 1845 * need to back off temporarily and let them 1846 * complete. */ 1847 #define DLM_JOIN_TIMEOUT_MSECS 90000 1848 if (status == -EAGAIN) { 1849 if (signal_pending(current)) { 1850 status = -ERESTARTSYS; 1851 goto bail; 1852 } 1853 1854 if (total_backoff > 1855 msecs_to_jiffies(DLM_JOIN_TIMEOUT_MSECS)) { 1856 status = -ERESTARTSYS; 1857 mlog(ML_NOTICE, "Timed out joining dlm domain " 1858 "%s after %u msecs\n", dlm->name, 1859 jiffies_to_msecs(total_backoff)); 1860 goto bail; 1861 } 1862 1863 /* 1864 * <chip> After you! 1865 * <dale> No, after you! 1866 * <chip> I insist! 1867 * <dale> But you first! 1868 * ... 1869 */ 1870 backoff = (unsigned int)(jiffies & 0x3); 1871 backoff *= DLM_DOMAIN_BACKOFF_MS; 1872 total_backoff += backoff; 1873 mlog(0, "backoff %d\n", backoff); 1874 msleep(backoff); 1875 } 1876 } while (status == -EAGAIN); 1877 1878 if (status < 0) { 1879 mlog_errno(status); 1880 goto bail; 1881 } 1882 1883 status = 0; 1884 bail: 1885 wake_up(&dlm_domain_events); 1886 1887 if (status) { 1888 dlm_unregister_domain_handlers(dlm); 1889 dlm_debug_shutdown(dlm); 1890 dlm_complete_thread(dlm); 1891 dlm_complete_recovery_thread(dlm); 1892 dlm_destroy_dlm_worker(dlm); 1893 } 1894 1895 return status; 1896 } 1897 1898 static struct dlm_ctxt *dlm_alloc_ctxt(const char *domain, 1899 u32 key) 1900 { 1901 int i; 1902 int ret; 1903 struct dlm_ctxt *dlm = NULL; 1904 1905 dlm = kzalloc(sizeof(*dlm), GFP_KERNEL); 1906 if (!dlm) { 1907 mlog_errno(-ENOMEM); 1908 goto leave; 1909 } 1910 1911 dlm->name = kstrdup(domain, GFP_KERNEL); 1912 if (dlm->name == NULL) { 1913 mlog_errno(-ENOMEM); 1914 kfree(dlm); 1915 dlm = NULL; 1916 goto leave; 1917 } 1918 1919 dlm->lockres_hash = (struct hlist_head **)dlm_alloc_pagevec(DLM_HASH_PAGES); 1920 if (!dlm->lockres_hash) { 1921 mlog_errno(-ENOMEM); 1922 kfree(dlm->name); 1923 kfree(dlm); 1924 dlm = NULL; 1925 goto leave; 1926 } 1927 1928 for (i = 0; i < DLM_HASH_BUCKETS; i++) 1929 INIT_HLIST_HEAD(dlm_lockres_hash(dlm, i)); 1930 1931 dlm->master_hash = (struct hlist_head **) 1932 dlm_alloc_pagevec(DLM_HASH_PAGES); 1933 if (!dlm->master_hash) { 1934 mlog_errno(-ENOMEM); 1935 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES); 1936 kfree(dlm->name); 1937 kfree(dlm); 1938 dlm = NULL; 1939 goto leave; 1940 } 1941 1942 for (i = 0; i < DLM_HASH_BUCKETS; i++) 1943 INIT_HLIST_HEAD(dlm_master_hash(dlm, i)); 1944 1945 dlm->key = key; 1946 dlm->node_num = o2nm_this_node(); 1947 1948 ret = dlm_create_debugfs_subroot(dlm); 1949 if (ret < 0) { 1950 dlm_free_pagevec((void **)dlm->master_hash, DLM_HASH_PAGES); 1951 dlm_free_pagevec((void **)dlm->lockres_hash, DLM_HASH_PAGES); 1952 kfree(dlm->name); 1953 kfree(dlm); 1954 dlm = NULL; 1955 goto leave; 1956 } 1957 1958 spin_lock_init(&dlm->spinlock); 1959 spin_lock_init(&dlm->master_lock); 1960 spin_lock_init(&dlm->ast_lock); 1961 spin_lock_init(&dlm->track_lock); 1962 INIT_LIST_HEAD(&dlm->list); 1963 INIT_LIST_HEAD(&dlm->dirty_list); 1964 INIT_LIST_HEAD(&dlm->reco.resources); 1965 INIT_LIST_HEAD(&dlm->reco.received); 1966 INIT_LIST_HEAD(&dlm->reco.node_data); 1967 INIT_LIST_HEAD(&dlm->purge_list); 1968 INIT_LIST_HEAD(&dlm->dlm_domain_handlers); 1969 INIT_LIST_HEAD(&dlm->tracking_list); 1970 dlm->reco.state = 0; 1971 1972 INIT_LIST_HEAD(&dlm->pending_asts); 1973 INIT_LIST_HEAD(&dlm->pending_basts); 1974 1975 mlog(0, "dlm->recovery_map=%p, &(dlm->recovery_map[0])=%p\n", 1976 dlm->recovery_map, &(dlm->recovery_map[0])); 1977 1978 memset(dlm->recovery_map, 0, sizeof(dlm->recovery_map)); 1979 memset(dlm->live_nodes_map, 0, sizeof(dlm->live_nodes_map)); 1980 memset(dlm->domain_map, 0, sizeof(dlm->domain_map)); 1981 1982 dlm->dlm_thread_task = NULL; 1983 dlm->dlm_reco_thread_task = NULL; 1984 dlm->dlm_worker = NULL; 1985 init_waitqueue_head(&dlm->dlm_thread_wq); 1986 init_waitqueue_head(&dlm->dlm_reco_thread_wq); 1987 init_waitqueue_head(&dlm->reco.event); 1988 init_waitqueue_head(&dlm->ast_wq); 1989 init_waitqueue_head(&dlm->migration_wq); 1990 INIT_LIST_HEAD(&dlm->mle_hb_events); 1991 1992 dlm->joining_node = DLM_LOCK_RES_OWNER_UNKNOWN; 1993 init_waitqueue_head(&dlm->dlm_join_events); 1994 1995 dlm->reco.new_master = O2NM_INVALID_NODE_NUM; 1996 dlm->reco.dead_node = O2NM_INVALID_NODE_NUM; 1997 1998 atomic_set(&dlm->res_tot_count, 0); 1999 atomic_set(&dlm->res_cur_count, 0); 2000 for (i = 0; i < DLM_MLE_NUM_TYPES; ++i) { 2001 atomic_set(&dlm->mle_tot_count[i], 0); 2002 atomic_set(&dlm->mle_cur_count[i], 0); 2003 } 2004 2005 spin_lock_init(&dlm->work_lock); 2006 INIT_LIST_HEAD(&dlm->work_list); 2007 INIT_WORK(&dlm->dispatched_work, dlm_dispatch_work); 2008 2009 kref_init(&dlm->dlm_refs); 2010 dlm->dlm_state = DLM_CTXT_NEW; 2011 2012 INIT_LIST_HEAD(&dlm->dlm_eviction_callbacks); 2013 2014 mlog(0, "context init: refcount %u\n", 2015 atomic_read(&dlm->dlm_refs.refcount)); 2016 2017 leave: 2018 return dlm; 2019 } 2020 2021 /* 2022 * Compare a requested locking protocol version against the current one. 2023 * 2024 * If the major numbers are different, they are incompatible. 2025 * If the current minor is greater than the request, they are incompatible. 2026 * If the current minor is less than or equal to the request, they are 2027 * compatible, and the requester should run at the current minor version. 2028 */ 2029 static int dlm_protocol_compare(struct dlm_protocol_version *existing, 2030 struct dlm_protocol_version *request) 2031 { 2032 if (existing->pv_major != request->pv_major) 2033 return 1; 2034 2035 if (existing->pv_minor > request->pv_minor) 2036 return 1; 2037 2038 if (existing->pv_minor < request->pv_minor) 2039 request->pv_minor = existing->pv_minor; 2040 2041 return 0; 2042 } 2043 2044 /* 2045 * dlm_register_domain: one-time setup per "domain". 2046 * 2047 * The filesystem passes in the requested locking version via proto. 2048 * If registration was successful, proto will contain the negotiated 2049 * locking protocol. 2050 */ 2051 struct dlm_ctxt * dlm_register_domain(const char *domain, 2052 u32 key, 2053 struct dlm_protocol_version *fs_proto) 2054 { 2055 int ret; 2056 struct dlm_ctxt *dlm = NULL; 2057 struct dlm_ctxt *new_ctxt = NULL; 2058 2059 if (strlen(domain) >= O2NM_MAX_NAME_LEN) { 2060 ret = -ENAMETOOLONG; 2061 mlog(ML_ERROR, "domain name length too long\n"); 2062 goto leave; 2063 } 2064 2065 if (!o2hb_check_local_node_heartbeating()) { 2066 mlog(ML_ERROR, "the local node has not been configured, or is " 2067 "not heartbeating\n"); 2068 ret = -EPROTO; 2069 goto leave; 2070 } 2071 2072 mlog(0, "register called for domain \"%s\"\n", domain); 2073 2074 retry: 2075 dlm = NULL; 2076 if (signal_pending(current)) { 2077 ret = -ERESTARTSYS; 2078 mlog_errno(ret); 2079 goto leave; 2080 } 2081 2082 spin_lock(&dlm_domain_lock); 2083 2084 dlm = __dlm_lookup_domain(domain); 2085 if (dlm) { 2086 if (dlm->dlm_state != DLM_CTXT_JOINED) { 2087 spin_unlock(&dlm_domain_lock); 2088 2089 mlog(0, "This ctxt is not joined yet!\n"); 2090 wait_event_interruptible(dlm_domain_events, 2091 dlm_wait_on_domain_helper( 2092 domain)); 2093 goto retry; 2094 } 2095 2096 if (dlm_protocol_compare(&dlm->fs_locking_proto, fs_proto)) { 2097 spin_unlock(&dlm_domain_lock); 2098 mlog(ML_ERROR, 2099 "Requested locking protocol version is not " 2100 "compatible with already registered domain " 2101 "\"%s\"\n", domain); 2102 ret = -EPROTO; 2103 goto leave; 2104 } 2105 2106 __dlm_get(dlm); 2107 dlm->num_joins++; 2108 2109 spin_unlock(&dlm_domain_lock); 2110 2111 ret = 0; 2112 goto leave; 2113 } 2114 2115 /* doesn't exist */ 2116 if (!new_ctxt) { 2117 spin_unlock(&dlm_domain_lock); 2118 2119 new_ctxt = dlm_alloc_ctxt(domain, key); 2120 if (new_ctxt) 2121 goto retry; 2122 2123 ret = -ENOMEM; 2124 mlog_errno(ret); 2125 goto leave; 2126 } 2127 2128 /* a little variable switch-a-roo here... */ 2129 dlm = new_ctxt; 2130 new_ctxt = NULL; 2131 2132 /* add the new domain */ 2133 list_add_tail(&dlm->list, &dlm_domains); 2134 spin_unlock(&dlm_domain_lock); 2135 2136 /* 2137 * Pass the locking protocol version into the join. If the join 2138 * succeeds, it will have the negotiated protocol set. 2139 */ 2140 dlm->dlm_locking_proto = dlm_protocol; 2141 dlm->fs_locking_proto = *fs_proto; 2142 2143 ret = dlm_join_domain(dlm); 2144 if (ret) { 2145 mlog_errno(ret); 2146 dlm_put(dlm); 2147 goto leave; 2148 } 2149 2150 /* Tell the caller what locking protocol we negotiated */ 2151 *fs_proto = dlm->fs_locking_proto; 2152 2153 ret = 0; 2154 leave: 2155 if (new_ctxt) 2156 dlm_free_ctxt_mem(new_ctxt); 2157 2158 if (ret < 0) 2159 dlm = ERR_PTR(ret); 2160 2161 return dlm; 2162 } 2163 EXPORT_SYMBOL_GPL(dlm_register_domain); 2164 2165 static LIST_HEAD(dlm_join_handlers); 2166 2167 static void dlm_unregister_net_handlers(void) 2168 { 2169 o2net_unregister_handler_list(&dlm_join_handlers); 2170 } 2171 2172 static int dlm_register_net_handlers(void) 2173 { 2174 int status = 0; 2175 2176 status = o2net_register_handler(DLM_QUERY_JOIN_MSG, DLM_MOD_KEY, 2177 sizeof(struct dlm_query_join_request), 2178 dlm_query_join_handler, 2179 NULL, NULL, &dlm_join_handlers); 2180 if (status) 2181 goto bail; 2182 2183 status = o2net_register_handler(DLM_ASSERT_JOINED_MSG, DLM_MOD_KEY, 2184 sizeof(struct dlm_assert_joined), 2185 dlm_assert_joined_handler, 2186 NULL, NULL, &dlm_join_handlers); 2187 if (status) 2188 goto bail; 2189 2190 status = o2net_register_handler(DLM_CANCEL_JOIN_MSG, DLM_MOD_KEY, 2191 sizeof(struct dlm_cancel_join), 2192 dlm_cancel_join_handler, 2193 NULL, NULL, &dlm_join_handlers); 2194 if (status) 2195 goto bail; 2196 2197 status = o2net_register_handler(DLM_QUERY_REGION, DLM_MOD_KEY, 2198 sizeof(struct dlm_query_region), 2199 dlm_query_region_handler, 2200 NULL, NULL, &dlm_join_handlers); 2201 2202 if (status) 2203 goto bail; 2204 2205 status = o2net_register_handler(DLM_QUERY_NODEINFO, DLM_MOD_KEY, 2206 sizeof(struct dlm_query_nodeinfo), 2207 dlm_query_nodeinfo_handler, 2208 NULL, NULL, &dlm_join_handlers); 2209 bail: 2210 if (status < 0) 2211 dlm_unregister_net_handlers(); 2212 2213 return status; 2214 } 2215 2216 /* Domain eviction callback handling. 2217 * 2218 * The file system requires notification of node death *before* the 2219 * dlm completes it's recovery work, otherwise it may be able to 2220 * acquire locks on resources requiring recovery. Since the dlm can 2221 * evict a node from it's domain *before* heartbeat fires, a similar 2222 * mechanism is required. */ 2223 2224 /* Eviction is not expected to happen often, so a per-domain lock is 2225 * not necessary. Eviction callbacks are allowed to sleep for short 2226 * periods of time. */ 2227 static DECLARE_RWSEM(dlm_callback_sem); 2228 2229 void dlm_fire_domain_eviction_callbacks(struct dlm_ctxt *dlm, 2230 int node_num) 2231 { 2232 struct list_head *iter; 2233 struct dlm_eviction_cb *cb; 2234 2235 down_read(&dlm_callback_sem); 2236 list_for_each(iter, &dlm->dlm_eviction_callbacks) { 2237 cb = list_entry(iter, struct dlm_eviction_cb, ec_item); 2238 2239 cb->ec_func(node_num, cb->ec_data); 2240 } 2241 up_read(&dlm_callback_sem); 2242 } 2243 2244 void dlm_setup_eviction_cb(struct dlm_eviction_cb *cb, 2245 dlm_eviction_func *f, 2246 void *data) 2247 { 2248 INIT_LIST_HEAD(&cb->ec_item); 2249 cb->ec_func = f; 2250 cb->ec_data = data; 2251 } 2252 EXPORT_SYMBOL_GPL(dlm_setup_eviction_cb); 2253 2254 void dlm_register_eviction_cb(struct dlm_ctxt *dlm, 2255 struct dlm_eviction_cb *cb) 2256 { 2257 down_write(&dlm_callback_sem); 2258 list_add_tail(&cb->ec_item, &dlm->dlm_eviction_callbacks); 2259 up_write(&dlm_callback_sem); 2260 } 2261 EXPORT_SYMBOL_GPL(dlm_register_eviction_cb); 2262 2263 void dlm_unregister_eviction_cb(struct dlm_eviction_cb *cb) 2264 { 2265 down_write(&dlm_callback_sem); 2266 list_del_init(&cb->ec_item); 2267 up_write(&dlm_callback_sem); 2268 } 2269 EXPORT_SYMBOL_GPL(dlm_unregister_eviction_cb); 2270 2271 static int __init dlm_init(void) 2272 { 2273 int status; 2274 2275 dlm_print_version(); 2276 2277 status = dlm_init_mle_cache(); 2278 if (status) { 2279 mlog(ML_ERROR, "Could not create o2dlm_mle slabcache\n"); 2280 goto error; 2281 } 2282 2283 status = dlm_init_master_caches(); 2284 if (status) { 2285 mlog(ML_ERROR, "Could not create o2dlm_lockres and " 2286 "o2dlm_lockname slabcaches\n"); 2287 goto error; 2288 } 2289 2290 status = dlm_init_lock_cache(); 2291 if (status) { 2292 mlog(ML_ERROR, "Count not create o2dlm_lock slabcache\n"); 2293 goto error; 2294 } 2295 2296 status = dlm_register_net_handlers(); 2297 if (status) { 2298 mlog(ML_ERROR, "Unable to register network handlers\n"); 2299 goto error; 2300 } 2301 2302 status = dlm_create_debugfs_root(); 2303 if (status) 2304 goto error; 2305 2306 return 0; 2307 error: 2308 dlm_unregister_net_handlers(); 2309 dlm_destroy_lock_cache(); 2310 dlm_destroy_master_caches(); 2311 dlm_destroy_mle_cache(); 2312 return -1; 2313 } 2314 2315 static void __exit dlm_exit (void) 2316 { 2317 dlm_destroy_debugfs_root(); 2318 dlm_unregister_net_handlers(); 2319 dlm_destroy_lock_cache(); 2320 dlm_destroy_master_caches(); 2321 dlm_destroy_mle_cache(); 2322 } 2323 2324 MODULE_AUTHOR("Oracle"); 2325 MODULE_LICENSE("GPL"); 2326 2327 module_init(dlm_init); 2328 module_exit(dlm_exit); 2329