fs/ocfs2/dlm/dlmdomain.c

   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 = 1,
 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_ATOMIC);
 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);