fs/ocfs2/dlm/dlmmaster.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * dlmmod.c
   5  *
   6  * standalone DLM module
   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
  28 #include <linux/module.h>
  29 #include <linux/fs.h>
  30 #include <linux/types.h>
  31 #include <linux/slab.h>
  32 #include <linux/highmem.h>
  33 #include <linux/init.h>
  34 #include <linux/sysctl.h>
  35 #include <linux/random.h>
  36 #include <linux/blkdev.h>
  37 #include <linux/socket.h>
  38 #include <linux/inet.h>
  39 #include <linux/spinlock.h>
  40 #include <linux/delay.h>
  41
  42
  43 #include "cluster/heartbeat.h"
  44 #include "cluster/nodemanager.h"
  45 #include "cluster/tcp.h"
  46
  47 #include "dlmapi.h"
  48 #include "dlmcommon.h"
  49 #include "dlmdomain.h"
  50 #include "dlmdebug.h"
  51
  52 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_MASTER)
  53 #include "cluster/masklog.h"
  54
  55 static void dlm_mle_node_down(struct dlm_ctxt *dlm,
  56                               struct dlm_master_list_entry *mle,
  57                               struct o2nm_node *node,
  58                               int idx);
  59 static void dlm_mle_node_up(struct dlm_ctxt *dlm,
  60                             struct dlm_master_list_entry *mle,
  61                             struct o2nm_node *node,
  62                             int idx);
  63
  64 static void dlm_assert_master_worker(struct dlm_work_item *item, void *data);
  65 static int dlm_do_assert_master(struct dlm_ctxt *dlm,
  66                                 struct dlm_lock_resource *res,
  67                                 void *nodemap, u32 flags);
  68 static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data);
  69
  70 static inline int dlm_mle_equal(struct dlm_ctxt *dlm,
  71                                 struct dlm_master_list_entry *mle,
  72                                 const char *name,
  73                                 unsigned int namelen)
  74 {
  75         if (dlm != mle->dlm)
  76                 return 0;
  77
  78         if (namelen != mle->mnamelen ||
  79             memcmp(name, mle->mname, namelen) != 0)
  80                 return 0;
  81
  82         return 1;
  83 }
  84
  85 static struct kmem_cache *dlm_lockres_cache = NULL;
  86 static struct kmem_cache *dlm_lockname_cache = NULL;
  87 static struct kmem_cache *dlm_mle_cache = NULL;
  88
  89 static void dlm_mle_release(struct kref *kref);
  90 static void dlm_init_mle(struct dlm_master_list_entry *mle,
  91                         enum dlm_mle_type type,
  92                         struct dlm_ctxt *dlm,
  93                         struct dlm_lock_resource *res,
  94                         const char *name,
  95                         unsigned int namelen);
  96 static void dlm_put_mle(struct dlm_master_list_entry *mle);
  97 static void __dlm_put_mle(struct dlm_master_list_entry *mle);
  98 static int dlm_find_mle(struct dlm_ctxt *dlm,
  99                         struct dlm_master_list_entry **mle,
 100                         char *name, unsigned int namelen);
 101
 102 static int dlm_do_master_request(struct dlm_lock_resource *res,
 103                                  struct dlm_master_list_entry *mle, int to);
 104
 105
 106 static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
 107                                      struct dlm_lock_resource *res,
 108                                      struct dlm_master_list_entry *mle,
 109                                      int *blocked);
 110 static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
 111                                     struct dlm_lock_resource *res,
 112                                     struct dlm_master_list_entry *mle,
 113                                     int blocked);
 114 static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
 115                                  struct dlm_lock_resource *res,
 116                                  struct dlm_master_list_entry *mle,
 117                                  struct dlm_master_list_entry **oldmle,
 118                                  const char *name, unsigned int namelen,
 119                                  u8 new_master, u8 master);
 120
 121 static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
 122                                     struct dlm_lock_resource *res);
 123 static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
 124                                       struct dlm_lock_resource *res);
 125 static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
 126                                        struct dlm_lock_resource *res,
 127                                        u8 target);
 128 static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
 129                                        struct dlm_lock_resource *res);
 130
 131
 132 int dlm_is_host_down(int errno)
 133 {
 134         switch (errno) {
 135                 case -EBADF:
 136                 case -ECONNREFUSED:
 137                 case -ENOTCONN:
 138                 case -ECONNRESET:
 139                 case -EPIPE:
 140                 case -EHOSTDOWN:
 141                 case -EHOSTUNREACH:
 142                 case -ETIMEDOUT:
 143                 case -ECONNABORTED:
 144                 case -ENETDOWN:
 145                 case -ENETUNREACH:
 146                 case -ENETRESET:
 147                 case -ESHUTDOWN:
 148                 case -ENOPROTOOPT:
 149                 case -EINVAL:   /* if returned from our tcp code,
 150                                    this means there is no socket */
 151                         return 1;
 152         }
 153         return 0;
 154 }
 155
 156
 157 /*
 158  * MASTER LIST FUNCTIONS
 159  */
 160
 161
 162 /*
 163  * regarding master list entries and heartbeat callbacks:
 164  *
 165  * in order to avoid sleeping and allocation that occurs in
 166  * heartbeat, master list entries are simply attached to the
 167  * dlm's established heartbeat callbacks.  the mle is attached
 168  * when it is created, and since the dlm->spinlock is held at
 169  * that time, any heartbeat event will be properly discovered
 170  * by the mle.  the mle needs to be detached from the
 171  * dlm->mle_hb_events list as soon as heartbeat events are no
 172  * longer useful to the mle, and before the mle is freed.
 173  *
 174  * as a general rule, heartbeat events are no longer needed by
 175  * the mle once an "answer" regarding the lock master has been
 176  * received.
 177  */
 178 static inline void __dlm_mle_attach_hb_events(struct dlm_ctxt *dlm,
 179                                               struct dlm_master_list_entry *mle)
 180 {
 181         assert_spin_locked(&dlm->spinlock);
 182
 183         list_add_tail(&mle->hb_events, &dlm->mle_hb_events);
 184 }
 185
 186
 187 static inline void __dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
 188                                               struct dlm_master_list_entry *mle)
 189 {
 190         if (!list_empty(&mle->hb_events))
 191                 list_del_init(&mle->hb_events);
 192 }
 193
 194
 195 static inline void dlm_mle_detach_hb_events(struct dlm_ctxt *dlm,
 196                                             struct dlm_master_list_entry *mle)
 197 {
 198         spin_lock(&dlm->spinlock);
 199         __dlm_mle_detach_hb_events(dlm, mle);
 200         spin_unlock(&dlm->spinlock);
 201 }
 202
 203 static void dlm_get_mle_inuse(struct dlm_master_list_entry *mle)
 204 {
 205         struct dlm_ctxt *dlm;
 206         dlm = mle->dlm;
 207
 208         assert_spin_locked(&dlm->spinlock);
 209         assert_spin_locked(&dlm->master_lock);
 210         mle->inuse++;
 211         kref_get(&mle->mle_refs);
 212 }
 213
 214 static void dlm_put_mle_inuse(struct dlm_master_list_entry *mle)
 215 {
 216         struct dlm_ctxt *dlm;
 217         dlm = mle->dlm;
 218
 219         spin_lock(&dlm->spinlock);
 220         spin_lock(&dlm->master_lock);
 221         mle->inuse--;
 222         __dlm_put_mle(mle);
 223         spin_unlock(&dlm->master_lock);
 224         spin_unlock(&dlm->spinlock);
 225
 226 }
 227
 228 /* remove from list and free */
 229 static void __dlm_put_mle(struct dlm_master_list_entry *mle)
 230 {
 231         struct dlm_ctxt *dlm;
 232         dlm = mle->dlm;
 233
 234         assert_spin_locked(&dlm->spinlock);
 235         assert_spin_locked(&dlm->master_lock);
 236         if (!atomic_read(&mle->mle_refs.refcount)) {
 237                 /* this may or may not crash, but who cares.
 238                  * it's a BUG. */
 239                 mlog(ML_ERROR, "bad mle: %p\n", mle);
 240                 dlm_print_one_mle(mle);
 241                 BUG();
 242         } else
 243                 kref_put(&mle->mle_refs, dlm_mle_release);
 244 }
 245
 246
 247 /* must not have any spinlocks coming in */
 248 static void dlm_put_mle(struct dlm_master_list_entry *mle)
 249 {
 250         struct dlm_ctxt *dlm;
 251         dlm = mle->dlm;
 252
 253         spin_lock(&dlm->spinlock);
 254         spin_lock(&dlm->master_lock);
 255         __dlm_put_mle(mle);
 256         spin_unlock(&dlm->master_lock);
 257         spin_unlock(&dlm->spinlock);
 258 }
 259
 260 static inline void dlm_get_mle(struct dlm_master_list_entry *mle)
 261 {
 262         kref_get(&mle->mle_refs);
 263 }
 264
 265 static void dlm_init_mle(struct dlm_master_list_entry *mle,
 266                         enum dlm_mle_type type,
 267                         struct dlm_ctxt *dlm,
 268                         struct dlm_lock_resource *res,
 269                         const char *name,
 270                         unsigned int namelen)
 271 {
 272         assert_spin_locked(&dlm->spinlock);
 273
 274         mle->dlm = dlm;
 275         mle->type = type;
 276         INIT_HLIST_NODE(&mle->master_hash_node);
 277         INIT_LIST_HEAD(&mle->hb_events);
 278         memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
 279         spin_lock_init(&mle->spinlock);
 280         init_waitqueue_head(&mle->wq);
 281         atomic_set(&mle->woken, 0);
 282         kref_init(&mle->mle_refs);
 283         memset(mle->response_map, 0, sizeof(mle->response_map));
 284         mle->master = O2NM_MAX_NODES;
 285         mle->new_master = O2NM_MAX_NODES;
 286         mle->inuse = 0;
 287
 288         BUG_ON(mle->type != DLM_MLE_BLOCK &&
 289                mle->type != DLM_MLE_MASTER &&
 290                mle->type != DLM_MLE_MIGRATION);
 291
 292         if (mle->type == DLM_MLE_MASTER) {
 293                 BUG_ON(!res);
 294                 mle->mleres = res;
 295                 memcpy(mle->mname, res->lockname.name, res->lockname.len);
 296                 mle->mnamelen = res->lockname.len;
 297                 mle->mnamehash = res->lockname.hash;
 298         } else {
 299                 BUG_ON(!name);
 300                 mle->mleres = NULL;
 301                 memcpy(mle->mname, name, namelen);
 302                 mle->mnamelen = namelen;
 303                 mle->mnamehash = dlm_lockid_hash(name, namelen);
 304         }
 305
 306         atomic_inc(&dlm->mle_tot_count[mle->type]);
 307         atomic_inc(&dlm->mle_cur_count[mle->type]);
 308
 309         /* copy off the node_map and register hb callbacks on our copy */
 310         memcpy(mle->node_map, dlm->domain_map, sizeof(mle->node_map));
 311         memcpy(mle->vote_map, dlm->domain_map, sizeof(mle->vote_map));
 312         clear_bit(dlm->node_num, mle->vote_map);
 313         clear_bit(dlm->node_num, mle->node_map);
 314
 315         /* attach the mle to the domain node up/down events */
 316         __dlm_mle_attach_hb_events(dlm, mle);
 317 }
 318
 319 void __dlm_unlink_mle(struct dlm_ctxt *dlm, struct dlm_master_list_entry *mle)
 320 {
 321         assert_spin_locked(&dlm->spinlock);
 322         assert_spin_locked(&dlm->master_lock);
 323
 324         if (!hlist_unhashed(&mle->master_hash_node))
 325                 hlist_del_init(&mle->master_hash_node);
 326 }
 327
 328 void __dlm_insert_mle(struct dlm_ctxt *dlm, struct dlm_master_list_entry *mle)
 329 {
 330         struct hlist_head *bucket;
 331
 332         assert_spin_locked(&dlm->master_lock);
 333
 334         bucket = dlm_master_hash(dlm, mle->mnamehash);
 335         hlist_add_head(&mle->master_hash_node, bucket);
 336 }
 337
 338 /* returns 1 if found, 0 if not */
 339 static int dlm_find_mle(struct dlm_ctxt *dlm,
 340                         struct dlm_master_list_entry **mle,
 341                         char *name, unsigned int namelen)
 342 {
 343         struct dlm_master_list_entry *tmpmle;
 344         struct hlist_head *bucket;
 345         struct hlist_node *list;
 346         unsigned int hash;
 347
 348         assert_spin_locked(&dlm->master_lock);
 349
 350         hash = dlm_lockid_hash(name, namelen);
 351         bucket = dlm_master_hash(dlm, hash);
 352         hlist_for_each(list, bucket) {
 353                 tmpmle = hlist_entry(list, struct dlm_master_list_entry,
 354                                      master_hash_node);
 355                 if (!dlm_mle_equal(dlm, tmpmle, name, namelen))
 356                         continue;
 357                 dlm_get_mle(tmpmle);
 358                 *mle = tmpmle;
 359                 return 1;
 360         }
 361         return 0;
 362 }
 363
 364 void dlm_hb_event_notify_attached(struct dlm_ctxt *dlm, int idx, int node_up)
 365 {
 366         struct dlm_master_list_entry *mle;
 367
 368         assert_spin_locked(&dlm->spinlock);
 369
 370         list_for_each_entry(mle, &dlm->mle_hb_events, hb_events) {
 371                 if (node_up)
 372                         dlm_mle_node_up(dlm, mle, NULL, idx);
 373                 else
 374                         dlm_mle_node_down(dlm, mle, NULL, idx);
 375         }
 376 }
 377
 378 static void dlm_mle_node_down(struct dlm_ctxt *dlm,
 379                               struct dlm_master_list_entry *mle,
 380                               struct o2nm_node *node, int idx)
 381 {
 382         spin_lock(&mle->spinlock);
 383
 384         if (!test_bit(idx, mle->node_map))
 385                 mlog(0, "node %u already removed from nodemap!\n", idx);
 386         else
 387                 clear_bit(idx, mle->node_map);
 388
 389         spin_unlock(&mle->spinlock);
 390 }
 391
 392 static void dlm_mle_node_up(struct dlm_ctxt *dlm,
 393                             struct dlm_master_list_entry *mle,
 394                             struct o2nm_node *node, int idx)
 395 {
 396         spin_lock(&mle->spinlock);
 397
 398         if (test_bit(idx, mle->node_map))
 399                 mlog(0, "node %u already in node map!\n", idx);
 400         else
 401                 set_bit(idx, mle->node_map);
 402
 403         spin_unlock(&mle->spinlock);
 404 }
 405
 406
 407 int dlm_init_mle_cache(void)
 408 {
 409         dlm_mle_cache = kmem_cache_create("o2dlm_mle",
 410                                           sizeof(struct dlm_master_list_entry),
 411                                           0, SLAB_HWCACHE_ALIGN,
 412                                           NULL);
 413         if (dlm_mle_cache == NULL)
 414                 return -ENOMEM;
 415         return 0;
 416 }
 417
 418 void dlm_destroy_mle_cache(void)
 419 {
 420         if (dlm_mle_cache)
 421                 kmem_cache_destroy(dlm_mle_cache);
 422 }
 423
 424 static void dlm_mle_release(struct kref *kref)
 425 {
 426         struct dlm_master_list_entry *mle;
 427         struct dlm_ctxt *dlm;
 428
 429         mle = container_of(kref, struct dlm_master_list_entry, mle_refs);
 430         dlm = mle->dlm;
 431
 432         assert_spin_locked(&dlm->spinlock);
 433         assert_spin_locked(&dlm->master_lock);
 434
 435         mlog(0, "Releasing mle for %.*s, type %d\n", mle->mnamelen, mle->mname,
 436              mle->type);
 437
 438         /* remove from list if not already */
 439         __dlm_unlink_mle(dlm, mle);
 440
 441         /* detach the mle from the domain node up/down events */
 442         __dlm_mle_detach_hb_events(dlm, mle);
 443
 444         atomic_dec(&dlm->mle_cur_count[mle->type]);
 445
 446         /* NOTE: kfree under spinlock here.
 447          * if this is bad, we can move this to a freelist. */
 448         kmem_cache_free(dlm_mle_cache, mle);
 449 }
 450
 451
 452 /*
 453  * LOCK RESOURCE FUNCTIONS
 454  */
 455
 456 int dlm_init_master_caches(void)
 457 {
 458         dlm_lockres_cache = kmem_cache_create("o2dlm_lockres",
 459                                               sizeof(struct dlm_lock_resource),
 460                                               0, SLAB_HWCACHE_ALIGN, NULL);
 461         if (!dlm_lockres_cache)
 462                 goto bail;
 463
 464         dlm_lockname_cache = kmem_cache_create("o2dlm_lockname",
 465                                                DLM_LOCKID_NAME_MAX, 0,
 466                                                SLAB_HWCACHE_ALIGN, NULL);
 467         if (!dlm_lockname_cache)
 468                 goto bail;
 469
 470         return 0;
 471 bail:
 472         dlm_destroy_master_caches();
 473         return -ENOMEM;
 474 }
 475
 476 void dlm_destroy_master_caches(void)
 477 {
 478         if (dlm_lockname_cache)
 479                 kmem_cache_destroy(dlm_lockname_cache);
 480
 481         if (dlm_lockres_cache)
 482                 kmem_cache_destroy(dlm_lockres_cache);
 483 }
 484
 485 static void dlm_lockres_release(struct kref *kref)
 486 {
 487         struct dlm_lock_resource *res;
 488         struct dlm_ctxt *dlm;
 489
 490         res = container_of(kref, struct dlm_lock_resource, refs);
 491         dlm = res->dlm;
 492
 493         /* This should not happen -- all lockres' have a name
 494          * associated with them at init time. */
 495         BUG_ON(!res->lockname.name);
 496
 497         mlog(0, "destroying lockres %.*s\n", res->lockname.len,
 498              res->lockname.name);
 499
 500         spin_lock(&dlm->track_lock);
 501         if (!list_empty(&res->tracking))
 502                 list_del_init(&res->tracking);
 503         else {
 504                 mlog(ML_ERROR, "Resource %.*s not on the Tracking list\n",
 505                      res->lockname.len, res->lockname.name);
 506                 dlm_print_one_lock_resource(res);
 507         }
 508         spin_unlock(&dlm->track_lock);
 509
 510         atomic_dec(&dlm->res_cur_count);
 511
 512         if (!hlist_unhashed(&res->hash_node) ||
 513             !list_empty(&res->granted) ||
 514             !list_empty(&res->converting) ||
 515             !list_empty(&res->blocked) ||
 516             !list_empty(&res->dirty) ||
 517             !list_empty(&res->recovering) ||
 518             !list_empty(&res->purge)) {
 519                 mlog(ML_ERROR,
 520                      "Going to BUG for resource %.*s."
 521                      "  We're on a list! [%c%c%c%c%c%c%c]\n",
 522                      res->lockname.len, res->lockname.name,
 523                      !hlist_unhashed(&res->hash_node) ? 'H' : ' ',
 524                      !list_empty(&res->granted) ? 'G' : ' ',
 525                      !list_empty(&res->converting) ? 'C' : ' ',
 526                      !list_empty(&res->blocked) ? 'B' : ' ',
 527                      !list_empty(&res->dirty) ? 'D' : ' ',
 528                      !list_empty(&res->recovering) ? 'R' : ' ',
 529                      !list_empty(&res->purge) ? 'P' : ' ');
 530
 531                 dlm_print_one_lock_resource(res);
 532         }
 533
 534         /* By the time we're ready to blow this guy away, we shouldn't
 535          * be on any lists. */
 536         BUG_ON(!hlist_unhashed(&res->hash_node));
 537         BUG_ON(!list_empty(&res->granted));
 538         BUG_ON(!list_empty(&res->converting));
 539         BUG_ON(!list_empty(&res->blocked));
 540         BUG_ON(!list_empty(&res->dirty));
 541         BUG_ON(!list_empty(&res->recovering));
 542         BUG_ON(!list_empty(&res->purge));
 543
 544         kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
 545
 546         kmem_cache_free(dlm_lockres_cache, res);
 547 }
 548
 549 void dlm_lockres_put(struct dlm_lock_resource *res)
 550 {
 551         kref_put(&res->refs, dlm_lockres_release);
 552 }
 553
 554 static void dlm_init_lockres(struct dlm_ctxt *dlm,
 555                              struct dlm_lock_resource *res,
 556                              const char *name, unsigned int namelen)
 557 {
 558         char *qname;
 559
 560         /* If we memset here, we lose our reference to the kmalloc'd
 561          * res->lockname.name, so be sure to init every field
 562          * correctly! */
 563
 564         qname = (char *) res->lockname.name;
 565         memcpy(qname, name, namelen);
 566
 567         res->lockname.len = namelen;
 568         res->lockname.hash = dlm_lockid_hash(name, namelen);
 569
 570         init_waitqueue_head(&res->wq);
 571         spin_lock_init(&res->spinlock);
 572         INIT_HLIST_NODE(&res->hash_node);
 573         INIT_LIST_HEAD(&res->granted);
 574         INIT_LIST_HEAD(&res->converting);
 575         INIT_LIST_HEAD(&res->blocked);
 576         INIT_LIST_HEAD(&res->dirty);
 577         INIT_LIST_HEAD(&res->recovering);
 578         INIT_LIST_HEAD(&res->purge);
 579         INIT_LIST_HEAD(&res->tracking);
 580         atomic_set(&res->asts_reserved, 0);
 581         res->migration_pending = 0;
 582         res->inflight_locks = 0;
 583
 584         res->dlm = dlm;
 585
 586         kref_init(&res->refs);
 587
 588         atomic_inc(&dlm->res_tot_count);
 589         atomic_inc(&dlm->res_cur_count);
 590
 591         /* just for consistency */
 592         spin_lock(&res->spinlock);
 593         dlm_set_lockres_owner(dlm, res, DLM_LOCK_RES_OWNER_UNKNOWN);
 594         spin_unlock(&res->spinlock);
 595
 596         res->state = DLM_LOCK_RES_IN_PROGRESS;
 597
 598         res->last_used = 0;
 599
 600         spin_lock(&dlm->spinlock);
 601         list_add_tail(&res->tracking, &dlm->tracking_list);
 602         spin_unlock(&dlm->spinlock);
 603
 604         memset(res->lvb, 0, DLM_LVB_LEN);
 605         memset(res->refmap, 0, sizeof(res->refmap));
 606 }
 607
 608 struct dlm_lock_resource *dlm_new_lockres(struct dlm_ctxt *dlm,
 609                                    const char *name,
 610                                    unsigned int namelen)
 611 {
 612         struct dlm_lock_resource *res = NULL;
 613
 614         res = kmem_cache_zalloc(dlm_lockres_cache, GFP_NOFS);
 615         if (!res)
 616                 goto error;
 617
 618         res->lockname.name = kmem_cache_zalloc(dlm_lockname_cache, GFP_NOFS);
 619         if (!res->lockname.name)
 620                 goto error;
 621
 622         dlm_init_lockres(dlm, res, name, namelen);
 623         return res;
 624
 625 error:
 626         if (res && res->lockname.name)
 627                 kmem_cache_free(dlm_lockname_cache, (void *)res->lockname.name);
 628
 629         if (res)
 630                 kmem_cache_free(dlm_lockres_cache, res);
 631         return NULL;
 632 }
 633
 634 void dlm_lockres_set_refmap_bit(struct dlm_ctxt *dlm,
 635                                 struct dlm_lock_resource *res, int bit)
 636 {
 637         assert_spin_locked(&res->spinlock);
 638
 639         mlog(0, "res %.*s, set node %u, %ps()\n", res->lockname.len,
 640              res->lockname.name, bit, __builtin_return_address(0));
 641
 642         set_bit(bit, res->refmap);
 643 }
 644
 645 void dlm_lockres_clear_refmap_bit(struct dlm_ctxt *dlm,
 646                                   struct dlm_lock_resource *res, int bit)
 647 {
 648         assert_spin_locked(&res->spinlock);
 649
 650         mlog(0, "res %.*s, clr node %u, %ps()\n", res->lockname.len,
 651              res->lockname.name, bit, __builtin_return_address(0));
 652
 653         clear_bit(bit, res->refmap);
 654 }
 655
 656
 657 void dlm_lockres_grab_inflight_ref(struct dlm_ctxt *dlm,
 658                                    struct dlm_lock_resource *res)
 659 {
 660         assert_spin_locked(&res->spinlock);
 661
 662         res->inflight_locks++;
 663
 664         mlog(0, "%s: res %.*s, inflight++: now %u, %ps()\n", dlm->name,
 665              res->lockname.len, res->lockname.name, res->inflight_locks,
 666              __builtin_return_address(0));
 667 }
 668
 669 void dlm_lockres_drop_inflight_ref(struct dlm_ctxt *dlm,
 670                                    struct dlm_lock_resource *res)
 671 {
 672         assert_spin_locked(&res->spinlock);
 673
 674         BUG_ON(res->inflight_locks == 0);
 675
 676         res->inflight_locks--;
 677
 678         mlog(0, "%s: res %.*s, inflight--: now %u, %ps()\n", dlm->name,
 679              res->lockname.len, res->lockname.name, res->inflight_locks,
 680              __builtin_return_address(0));
 681
 682         wake_up(&res->wq);
 683 }
 684
 685 /*
 686  * lookup a lock resource by name.
 687  * may already exist in the hashtable.
 688  * lockid is null terminated
 689  *
 690  * if not, allocate enough for the lockres and for
 691  * the temporary structure used in doing the mastering.
 692  *
 693  * also, do a lookup in the dlm->master_list to see
 694  * if another node has begun mastering the same lock.
 695  * if so, there should be a block entry in there
 696  * for this name, and we should *not* attempt to master
 697  * the lock here.   need to wait around for that node
 698  * to assert_master (or die).
 699  *
 700  */
 701 struct dlm_lock_resource * dlm_get_lock_resource(struct dlm_ctxt *dlm,
 702                                           const char *lockid,
 703                                           int namelen,
 704                                           int flags)
 705 {
 706         struct dlm_lock_resource *tmpres=NULL, *res=NULL;
 707         struct dlm_master_list_entry *mle = NULL;
 708         struct dlm_master_list_entry *alloc_mle = NULL;
 709         int blocked = 0;
 710         int ret, nodenum;
 711         struct dlm_node_iter iter;
 712         unsigned int hash;
 713         int tries = 0;
 714         int bit, wait_on_recovery = 0;
 715
 716         BUG_ON(!lockid);
 717
 718         hash = dlm_lockid_hash(lockid, namelen);
 719
 720         mlog(0, "get lockres %s (len %d)\n", lockid, namelen);
 721
 722 lookup:
 723         spin_lock(&dlm->spinlock);
 724         tmpres = __dlm_lookup_lockres_full(dlm, lockid, namelen, hash);
 725         if (tmpres) {
 726                 spin_unlock(&dlm->spinlock);
 727                 spin_lock(&tmpres->spinlock);
 728                 /* Wait on the thread that is mastering the resource */
 729                 if (tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN) {
 730                         __dlm_wait_on_lockres(tmpres);
 731                         BUG_ON(tmpres->owner == DLM_LOCK_RES_OWNER_UNKNOWN);
 732                         spin_unlock(&tmpres->spinlock);
 733                         dlm_lockres_put(tmpres);
 734                         tmpres = NULL;
 735                         goto lookup;
 736                 }
 737
 738                 /* Wait on the resource purge to complete before continuing */
 739                 if (tmpres->state & DLM_LOCK_RES_DROPPING_REF) {
 740                         BUG_ON(tmpres->owner == dlm->node_num);
 741                         __dlm_wait_on_lockres_flags(tmpres,
 742                                                     DLM_LOCK_RES_DROPPING_REF);
 743                         spin_unlock(&tmpres->spinlock);
 744                         dlm_lockres_put(tmpres);
 745                         tmpres = NULL;
 746                         goto lookup;
 747                 }
 748
 749                 /* Grab inflight ref to pin the resource */
 750                 dlm_lockres_grab_inflight_ref(dlm, tmpres);
 751
 752                 spin_unlock(&tmpres->spinlock);
 753                 if (res)
 754                         dlm_lockres_put(res);
 755                 res = tmpres;
 756                 goto leave;
 757         }
 758
 759         if (!res) {
 760                 spin_unlock(&dlm->spinlock);
 761                 mlog(0, "allocating a new resource\n");
 762                 /* nothing found and we need to allocate one. */
 763                 alloc_mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
 764                 if (!alloc_mle)
 765                         goto leave;
 766                 res = dlm_new_lockres(dlm, lockid, namelen);
 767                 if (!res)
 768                         goto leave;
 769                 goto lookup;
 770         }
 771
 772         mlog(0, "no lockres found, allocated our own: %p\n", res);
 773
 774         if (flags & LKM_LOCAL) {
 775                 /* caller knows it's safe to assume it's not mastered elsewhere
 776                  * DONE!  return right away */
 777                 spin_lock(&res->spinlock);
 778                 dlm_change_lockres_owner(dlm, res, dlm->node_num);
 779                 __dlm_insert_lockres(dlm, res);
 780                 dlm_lockres_grab_inflight_ref(dlm, res);
 781                 spin_unlock(&res->spinlock);
 782                 spin_unlock(&dlm->spinlock);
 783                 /* lockres still marked IN_PROGRESS */
 784                 goto wake_waiters;
 785         }
 786
 787         /* check master list to see if another node has started mastering it */
 788         spin_lock(&dlm->master_lock);
 789
 790         /* if we found a block, wait for lock to be mastered by another node */
 791         blocked = dlm_find_mle(dlm, &mle, (char *)lockid, namelen);
 792         if (blocked) {
 793                 int mig;
 794                 if (mle->type == DLM_MLE_MASTER) {
 795                         mlog(ML_ERROR, "master entry for nonexistent lock!\n");
 796                         BUG();
 797                 }
 798                 mig = (mle->type == DLM_MLE_MIGRATION);
 799                 /* if there is a migration in progress, let the migration
 800                  * finish before continuing.  we can wait for the absence
 801                  * of the MIGRATION mle: either the migrate finished or
 802                  * one of the nodes died and the mle was cleaned up.
 803                  * if there is a BLOCK here, but it already has a master
 804                  * set, we are too late.  the master does not have a ref
 805                  * for us in the refmap.  detach the mle and drop it.
 806                  * either way, go back to the top and start over. */
 807                 if (mig || mle->master != O2NM_MAX_NODES) {
 808                         BUG_ON(mig && mle->master == dlm->node_num);
 809                         /* we arrived too late.  the master does not
 810                          * have a ref for us. retry. */
 811                         mlog(0, "%s:%.*s: late on %s\n",
 812                              dlm->name, namelen, lockid,
 813                              mig ?  "MIGRATION" : "BLOCK");
 814                         spin_unlock(&dlm->master_lock);
 815                         spin_unlock(&dlm->spinlock);
 816
 817                         /* master is known, detach */
 818                         if (!mig)
 819                                 dlm_mle_detach_hb_events(dlm, mle);
 820                         dlm_put_mle(mle);
 821                         mle = NULL;
 822                         /* this is lame, but we can't wait on either
 823                          * the mle or lockres waitqueue here */
 824                         if (mig)
 825                                 msleep(100);
 826                         goto lookup;
 827                 }
 828         } else {
 829                 /* go ahead and try to master lock on this node */
 830                 mle = alloc_mle;
 831                 /* make sure this does not get freed below */
 832                 alloc_mle = NULL;
 833                 dlm_init_mle(mle, DLM_MLE_MASTER, dlm, res, NULL, 0);
 834                 set_bit(dlm->node_num, mle->maybe_map);
 835                 __dlm_insert_mle(dlm, mle);
 836
 837                 /* still holding the dlm spinlock, check the recovery map
 838                  * to see if there are any nodes that still need to be
 839                  * considered.  these will not appear in the mle nodemap
 840                  * but they might own this lockres.  wait on them. */
 841                 bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
 842                 if (bit < O2NM_MAX_NODES) {
 843                         mlog(0, "%s: res %.*s, At least one node (%d) "
 844                              "to recover before lock mastery can begin\n",
 845                              dlm->name, namelen, (char *)lockid, bit);
 846                         wait_on_recovery = 1;
 847                 }
 848         }
 849
 850         /* at this point there is either a DLM_MLE_BLOCK or a
 851          * DLM_MLE_MASTER on the master list, so it's safe to add the
 852          * lockres to the hashtable.  anyone who finds the lock will
 853          * still have to wait on the IN_PROGRESS. */
 854
 855         /* finally add the lockres to its hash bucket */
 856         __dlm_insert_lockres(dlm, res);
 857
 858         /* Grab inflight ref to pin the resource */
 859         spin_lock(&res->spinlock);
 860         dlm_lockres_grab_inflight_ref(dlm, res);
 861         spin_unlock(&res->spinlock);
 862
 863         /* get an extra ref on the mle in case this is a BLOCK
 864          * if so, the creator of the BLOCK may try to put the last
 865          * ref at this time in the assert master handler, so we
 866          * need an extra one to keep from a bad ptr deref. */
 867         dlm_get_mle_inuse(mle);
 868         spin_unlock(&dlm->master_lock);
 869         spin_unlock(&dlm->spinlock);
 870
 871 redo_request:
 872         while (wait_on_recovery) {
 873                 /* any cluster changes that occurred after dropping the
 874                  * dlm spinlock would be detectable be a change on the mle,
 875                  * so we only need to clear out the recovery map once. */
 876                 if (dlm_is_recovery_lock(lockid, namelen)) {
 877                         mlog(0, "%s: Recovery map is not empty, but must "
 878                              "master $RECOVERY lock now\n", dlm->name);
 879                         if (!dlm_pre_master_reco_lockres(dlm, res))
 880                                 wait_on_recovery = 0;
 881                         else {
 882                                 mlog(0, "%s: waiting 500ms for heartbeat state "
 883                                     "change\n", dlm->name);
 884                                 msleep(500);
 885                         }
 886                         continue;
 887                 }
 888
 889                 dlm_kick_recovery_thread(dlm);
 890                 msleep(1000);
 891                 dlm_wait_for_recovery(dlm);
 892
 893                 spin_lock(&dlm->spinlock);
 894                 bit = find_next_bit(dlm->recovery_map, O2NM_MAX_NODES, 0);
 895                 if (bit < O2NM_MAX_NODES) {
 896                         mlog(0, "%s: res %.*s, At least one node (%d) "
 897                              "to recover before lock mastery can begin\n",
 898                              dlm->name, namelen, (char *)lockid, bit);
 899                         wait_on_recovery = 1;
 900                 } else
 901                         wait_on_recovery = 0;
 902                 spin_unlock(&dlm->spinlock);
 903
 904                 if (wait_on_recovery)
 905                         dlm_wait_for_node_recovery(dlm, bit, 10000);
 906         }
 907
 908         /* must wait for lock to be mastered elsewhere */
 909         if (blocked)
 910                 goto wait;
 911
 912         ret = -EINVAL;
 913         dlm_node_iter_init(mle->vote_map, &iter);
 914         while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
 915                 ret = dlm_do_master_request(res, mle, nodenum);
 916                 if (ret < 0)
 917                         mlog_errno(ret);
 918                 if (mle->master != O2NM_MAX_NODES) {
 919                         /* found a master ! */
 920                         if (mle->master <= nodenum)
 921                                 break;
 922                         /* if our master request has not reached the master
 923                          * yet, keep going until it does.  this is how the
 924                          * master will know that asserts are needed back to
 925                          * the lower nodes. */
 926                         mlog(0, "%s: res %.*s, Requests only up to %u but "
 927                              "master is %u, keep going\n", dlm->name, namelen,
 928                              lockid, nodenum, mle->master);
 929                 }
 930         }
 931
 932 wait:
 933         /* keep going until the response map includes all nodes */
 934         ret = dlm_wait_for_lock_mastery(dlm, res, mle, &blocked);
 935         if (ret < 0) {
 936                 wait_on_recovery = 1;
 937                 mlog(0, "%s: res %.*s, Node map changed, redo the master "
 938                      "request now, blocked=%d\n", dlm->name, res->lockname.len,
 939                      res->lockname.name, blocked);
 940                 if (++tries > 20) {
 941                         mlog(ML_ERROR, "%s: res %.*s, Spinning on "
 942                              "dlm_wait_for_lock_mastery, blocked = %d\n",
 943                              dlm->name, res->lockname.len,
 944                              res->lockname.name, blocked);
 945                         dlm_print_one_lock_resource(res);
 946                         dlm_print_one_mle(mle);
 947                         tries = 0;
 948                 }
 949                 goto redo_request;
 950         }
 951
 952         mlog(0, "%s: res %.*s, Mastered by %u\n", dlm->name, res->lockname.len,
 953              res->lockname.name, res->owner);
 954         /* make sure we never continue without this */
 955         BUG_ON(res->owner == O2NM_MAX_NODES);
 956
 957         /* master is known, detach if not already detached */
 958         dlm_mle_detach_hb_events(dlm, mle);
 959         dlm_put_mle(mle);
 960         /* put the extra ref */
 961         dlm_put_mle_inuse(mle);
 962
 963 wake_waiters:
 964         spin_lock(&res->spinlock);
 965         res->state &= ~DLM_LOCK_RES_IN_PROGRESS;
 966         spin_unlock(&res->spinlock);
 967         wake_up(&res->wq);
 968
 969 leave:
 970         /* need to free the unused mle */
 971         if (alloc_mle)
 972                 kmem_cache_free(dlm_mle_cache, alloc_mle);
 973
 974         return res;
 975 }
 976
 977
 978 #define DLM_MASTERY_TIMEOUT_MS   5000
 979
 980 static int dlm_wait_for_lock_mastery(struct dlm_ctxt *dlm,
 981                                      struct dlm_lock_resource *res,
 982                                      struct dlm_master_list_entry *mle,
 983                                      int *blocked)
 984 {
 985         u8 m;
 986         int ret, bit;
 987         int map_changed, voting_done;
 988         int assert, sleep;
 989
 990 recheck:
 991         ret = 0;
 992         assert = 0;
 993
 994         /* check if another node has already become the owner */
 995         spin_lock(&res->spinlock);
 996         if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
 997                 mlog(0, "%s:%.*s: owner is suddenly %u\n", dlm->name,
 998                      res->lockname.len, res->lockname.name, res->owner);
 999                 spin_unlock(&res->spinlock);
1000                 /* this will cause the master to re-assert across
1001                  * the whole cluster, freeing up mles */
1002                 if (res->owner != dlm->node_num) {
1003                         ret = dlm_do_master_request(res, mle, res->owner);
1004                         if (ret < 0) {
1005                                 /* give recovery a chance to run */
1006                                 mlog(ML_ERROR, "link to %u went down?: %d\n", res->owner, ret);
1007                                 msleep(500);
1008                                 goto recheck;
1009                         }
1010                 }
1011                 ret = 0;
1012                 goto leave;
1013         }
1014         spin_unlock(&res->spinlock);
1015
1016         spin_lock(&mle->spinlock);
1017         m = mle->master;
1018         map_changed = (memcmp(mle->vote_map, mle->node_map,
1019                               sizeof(mle->vote_map)) != 0);
1020         voting_done = (memcmp(mle->vote_map, mle->response_map,
1021                              sizeof(mle->vote_map)) == 0);
1022
1023         /* restart if we hit any errors */
1024         if (map_changed) {
1025                 int b;
1026                 mlog(0, "%s: %.*s: node map changed, restarting\n",
1027                      dlm->name, res->lockname.len, res->lockname.name);
1028                 ret = dlm_restart_lock_mastery(dlm, res, mle, *blocked);
1029                 b = (mle->type == DLM_MLE_BLOCK);
1030                 if ((*blocked && !b) || (!*blocked && b)) {
1031                         mlog(0, "%s:%.*s: status change: old=%d new=%d\n",
1032                              dlm->name, res->lockname.len, res->lockname.name,
1033                              *blocked, b);
1034                         *blocked = b;
1035                 }
1036                 spin_unlock(&mle->spinlock);
1037                 if (ret < 0) {
1038                         mlog_errno(ret);
1039                         goto leave;
1040                 }
1041                 mlog(0, "%s:%.*s: restart lock mastery succeeded, "
1042                      "rechecking now\n", dlm->name, res->lockname.len,
1043                      res->lockname.name);
1044                 goto recheck;
1045         } else {
1046                 if (!voting_done) {
1047                         mlog(0, "map not changed and voting not done "
1048                              "for %s:%.*s\n", dlm->name, res->lockname.len,
1049                              res->lockname.name);
1050                 }
1051         }
1052
1053         if (m != O2NM_MAX_NODES) {
1054                 /* another node has done an assert!
1055                  * all done! */
1056                 sleep = 0;
1057         } else {
1058                 sleep = 1;
1059                 /* have all nodes responded? */
1060                 if (voting_done && !*blocked) {
1061                         bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
1062                         if (dlm->node_num <= bit) {
1063                                 /* my node number is lowest.
1064                                  * now tell other nodes that I am
1065                                  * mastering this. */
1066                                 mle->master = dlm->node_num;
1067                                 /* ref was grabbed in get_lock_resource
1068                                  * will be dropped in dlmlock_master */
1069                                 assert = 1;
1070                                 sleep = 0;
1071                         }
1072                         /* if voting is done, but we have not received
1073                          * an assert master yet, we must sleep */
1074                 }
1075         }
1076
1077         spin_unlock(&mle->spinlock);
1078
1079         /* sleep if we haven't finished voting yet */
1080         if (sleep) {
1081                 unsigned long timeo = msecs_to_jiffies(DLM_MASTERY_TIMEOUT_MS);
1082
1083                 /*
1084                 if (atomic_read(&mle->mle_refs.refcount) < 2)
1085                         mlog(ML_ERROR, "mle (%p) refs=%d, name=%.*s\n", mle,
1086                         atomic_read(&mle->mle_refs.refcount),
1087                         res->lockname.len, res->lockname.name);
1088                 */
1089                 atomic_set(&mle->woken, 0);
1090                 (void)wait_event_timeout(mle->wq,
1091                                          (atomic_read(&mle->woken) == 1),
1092                                          timeo);
1093                 if (res->owner == O2NM_MAX_NODES) {
1094                         mlog(0, "%s:%.*s: waiting again\n", dlm->name,
1095                              res->lockname.len, res->lockname.name);
1096                         goto recheck;
1097                 }
1098                 mlog(0, "done waiting, master is %u\n", res->owner);
1099                 ret = 0;
1100                 goto leave;
1101         }
1102
1103         ret = 0;   /* done */
1104         if (assert) {
1105                 m = dlm->node_num;
1106                 mlog(0, "about to master %.*s here, this=%u\n",
1107                      res->lockname.len, res->lockname.name, m);
1108                 ret = dlm_do_assert_master(dlm, res, mle->vote_map, 0);
1109                 if (ret) {
1110                         /* This is a failure in the network path,
1111                          * not in the response to the assert_master
1112                          * (any nonzero response is a BUG on this node).
1113                          * Most likely a socket just got disconnected
1114                          * due to node death. */
1115                         mlog_errno(ret);
1116                 }
1117                 /* no longer need to restart lock mastery.
1118                  * all living nodes have been contacted. */
1119                 ret = 0;
1120         }
1121
1122         /* set the lockres owner */
1123         spin_lock(&res->spinlock);
1124         /* mastery reference obtained either during
1125          * assert_master_handler or in get_lock_resource */
1126         dlm_change_lockres_owner(dlm, res, m);
1127         spin_unlock(&res->spinlock);
1128
1129 leave:
1130         return ret;
1131 }
1132
1133 struct dlm_bitmap_diff_iter
1134 {
1135         int curnode;
1136         unsigned long *orig_bm;
1137         unsigned long *cur_bm;
1138         unsigned long diff_bm[BITS_TO_LONGS(O2NM_MAX_NODES)];
1139 };
1140
1141 enum dlm_node_state_change
1142 {
1143         NODE_DOWN = -1,
1144         NODE_NO_CHANGE = 0,
1145         NODE_UP
1146 };
1147
1148 static void dlm_bitmap_diff_iter_init(struct dlm_bitmap_diff_iter *iter,
1149                                       unsigned long *orig_bm,
1150                                       unsigned long *cur_bm)
1151 {
1152         unsigned long p1, p2;
1153         int i;
1154
1155         iter->curnode = -1;
1156         iter->orig_bm = orig_bm;
1157         iter->cur_bm = cur_bm;
1158
1159         for (i = 0; i < BITS_TO_LONGS(O2NM_MAX_NODES); i++) {
1160                 p1 = *(iter->orig_bm + i);
1161                 p2 = *(iter->cur_bm + i);
1162                 iter->diff_bm[i] = (p1 & ~p2) | (p2 & ~p1);
1163         }
1164 }
1165
1166 static int dlm_bitmap_diff_iter_next(struct dlm_bitmap_diff_iter *iter,
1167                                      enum dlm_node_state_change *state)
1168 {
1169         int bit;
1170
1171         if (iter->curnode >= O2NM_MAX_NODES)
1172                 return -ENOENT;
1173
1174         bit = find_next_bit(iter->diff_bm, O2NM_MAX_NODES,
1175                             iter->curnode+1);
1176         if (bit >= O2NM_MAX_NODES) {
1177                 iter->curnode = O2NM_MAX_NODES;
1178                 return -ENOENT;
1179         }
1180
1181         /* if it was there in the original then this node died */
1182         if (test_bit(bit, iter->orig_bm))
1183                 *state = NODE_DOWN;
1184         else
1185                 *state = NODE_UP;
1186
1187         iter->curnode = bit;
1188         return bit;
1189 }
1190
1191
1192 static int dlm_restart_lock_mastery(struct dlm_ctxt *dlm,
1193                                     struct dlm_lock_resource *res,
1194                                     struct dlm_master_list_entry *mle,
1195                                     int blocked)
1196 {
1197         struct dlm_bitmap_diff_iter bdi;
1198         enum dlm_node_state_change sc;
1199         int node;
1200         int ret = 0;
1201
1202         mlog(0, "something happened such that the "
1203              "master process may need to be restarted!\n");
1204
1205         assert_spin_locked(&mle->spinlock);
1206
1207         dlm_bitmap_diff_iter_init(&bdi, mle->vote_map, mle->node_map);
1208         node = dlm_bitmap_diff_iter_next(&bdi, &sc);
1209         while (node >= 0) {
1210                 if (sc == NODE_UP) {
1211                         /* a node came up.  clear any old vote from
1212                          * the response map and set it in the vote map
1213                          * then restart the mastery. */
1214                         mlog(ML_NOTICE, "node %d up while restarting\n", node);
1215
1216                         /* redo the master request, but only for the new node */
1217                         mlog(0, "sending request to new node\n");
1218                         clear_bit(node, mle->response_map);
1219                         set_bit(node, mle->vote_map);
1220                 } else {
1221                         mlog(ML_ERROR, "node down! %d\n", node);
1222                         if (blocked) {
1223                                 int lowest = find_next_bit(mle->maybe_map,
1224                                                        O2NM_MAX_NODES, 0);
1225
1226                                 /* act like it was never there */
1227                                 clear_bit(node, mle->maybe_map);
1228
1229                                 if (node == lowest) {
1230                                         mlog(0, "expected master %u died"
1231                                             " while this node was blocked "
1232                                             "waiting on it!\n", node);
1233                                         lowest = find_next_bit(mle->maybe_map,
1234                                                         O2NM_MAX_NODES,
1235                                                         lowest+1);
1236                                         if (lowest < O2NM_MAX_NODES) {
1237                                                 mlog(0, "%s:%.*s:still "
1238                                                      "blocked. waiting on %u "
1239                                                      "now\n", dlm->name,
1240                                                      res->lockname.len,
1241                                                      res->lockname.name,
1242                                                      lowest);
1243                                         } else {
1244                                                 /* mle is an MLE_BLOCK, but
1245                                                  * there is now nothing left to
1246                                                  * block on.  we need to return
1247                                                  * all the way back out and try
1248                                                  * again with an MLE_MASTER.
1249                                                  * dlm_do_local_recovery_cleanup
1250                                                  * has already run, so the mle
1251                                                  * refcount is ok */
1252                                                 mlog(0, "%s:%.*s: no "
1253                                                      "longer blocking. try to "
1254                                                      "master this here\n",
1255                                                      dlm->name,
1256                                                      res->lockname.len,
1257                                                      res->lockname.name);
1258                                                 mle->type = DLM_MLE_MASTER;
1259                                                 mle->mleres = res;
1260                                         }
1261                                 }
1262                         }
1263
1264                         /* now blank out everything, as if we had never
1265                          * contacted anyone */
1266                         memset(mle->maybe_map, 0, sizeof(mle->maybe_map));
1267                         memset(mle->response_map, 0, sizeof(mle->response_map));
1268                         /* reset the vote_map to the current node_map */
1269                         memcpy(mle->vote_map, mle->node_map,
1270                                sizeof(mle->node_map));
1271                         /* put myself into the maybe map */
1272                         if (mle->type != DLM_MLE_BLOCK)
1273                                 set_bit(dlm->node_num, mle->maybe_map);
1274                 }
1275                 ret = -EAGAIN;
1276                 node = dlm_bitmap_diff_iter_next(&bdi, &sc);
1277         }
1278         return ret;
1279 }
1280
1281
1282 /*
1283  * DLM_MASTER_REQUEST_MSG
1284  *
1285  * returns: 0 on success,
1286  *          -errno on a network error
1287  *
1288  * on error, the caller should assume the target node is "dead"
1289  *
1290  */
1291
1292 static int dlm_do_master_request(struct dlm_lock_resource *res,
1293                                  struct dlm_master_list_entry *mle, int to)
1294 {
1295         struct dlm_ctxt *dlm = mle->dlm;
1296         struct dlm_master_request request;
1297         int ret, response=0, resend;
1298
1299         memset(&request, 0, sizeof(request));
1300         request.node_idx = dlm->node_num;
1301
1302         BUG_ON(mle->type == DLM_MLE_MIGRATION);
1303
1304         request.namelen = (u8)mle->mnamelen;
1305         memcpy(request.name, mle->mname, request.namelen);
1306
1307 again:
1308         ret = o2net_send_message(DLM_MASTER_REQUEST_MSG, dlm->key, &request,
1309                                  sizeof(request), to, &response);
1310         if (ret < 0)  {
1311                 if (ret == -ESRCH) {
1312                         /* should never happen */
1313                         mlog(ML_ERROR, "TCP stack not ready!\n");
1314                         BUG();
1315                 } else if (ret == -EINVAL) {
1316                         mlog(ML_ERROR, "bad args passed to o2net!\n");
1317                         BUG();
1318                 } else if (ret == -ENOMEM) {
1319                         mlog(ML_ERROR, "out of memory while trying to send "
1320                              "network message!  retrying\n");
1321                         /* this is totally crude */
1322                         msleep(50);
1323                         goto again;
1324                 } else if (!dlm_is_host_down(ret)) {
1325                         /* not a network error. bad. */
1326                         mlog_errno(ret);
1327                         mlog(ML_ERROR, "unhandled error!");
1328                         BUG();
1329                 }
1330                 /* all other errors should be network errors,
1331                  * and likely indicate node death */
1332                 mlog(ML_ERROR, "link to %d went down!\n", to);
1333                 goto out;
1334         }
1335
1336         ret = 0;
1337         resend = 0;
1338         spin_lock(&mle->spinlock);
1339         switch (response) {
1340                 case DLM_MASTER_RESP_YES:
1341                         set_bit(to, mle->response_map);
1342                         mlog(0, "node %u is the master, response=YES\n", to);
1343                         mlog(0, "%s:%.*s: master node %u now knows I have a "
1344                              "reference\n", dlm->name, res->lockname.len,
1345                              res->lockname.name, to);
1346                         mle->master = to;
1347                         break;
1348                 case DLM_MASTER_RESP_NO:
1349                         mlog(0, "node %u not master, response=NO\n", to);
1350                         set_bit(to, mle->response_map);
1351                         break;
1352                 case DLM_MASTER_RESP_MAYBE:
1353                         mlog(0, "node %u not master, response=MAYBE\n", to);
1354                         set_bit(to, mle->response_map);
1355                         set_bit(to, mle->maybe_map);
1356                         break;
1357                 case DLM_MASTER_RESP_ERROR:
1358                         mlog(0, "node %u hit an error, resending\n", to);
1359                         resend = 1;
1360                         response = 0;
1361                         break;
1362                 default:
1363                         mlog(ML_ERROR, "bad response! %u\n", response);
1364                         BUG();
1365         }
1366         spin_unlock(&mle->spinlock);
1367         if (resend) {
1368                 /* this is also totally crude */
1369                 msleep(50);
1370                 goto again;
1371         }
1372
1373 out:
1374         return ret;
1375 }
1376
1377 /*
1378  * locks that can be taken here:
1379  * dlm->spinlock
1380  * res->spinlock
1381  * mle->spinlock
1382  * dlm->master_list
1383  *
1384  * if possible, TRIM THIS DOWN!!!
1385  */
1386 int dlm_master_request_handler(struct o2net_msg *msg, u32 len, void *data,
1387                                void **ret_data)
1388 {
1389         u8 response = DLM_MASTER_RESP_MAYBE;
1390         struct dlm_ctxt *dlm = data;
1391         struct dlm_lock_resource *res = NULL;
1392         struct dlm_master_request *request = (struct dlm_master_request *) msg->buf;
1393         struct dlm_master_list_entry *mle = NULL, *tmpmle = NULL;
1394         char *name;
1395         unsigned int namelen, hash;
1396         int found, ret;
1397         int set_maybe;
1398         int dispatch_assert = 0;
1399
1400         if (!dlm_grab(dlm))
1401                 return DLM_MASTER_RESP_NO;
1402
1403         if (!dlm_domain_fully_joined(dlm)) {
1404                 response = DLM_MASTER_RESP_NO;
1405                 goto send_response;
1406         }
1407
1408         name = request->name;
1409         namelen = request->namelen;
1410         hash = dlm_lockid_hash(name, namelen);
1411
1412         if (namelen > DLM_LOCKID_NAME_MAX) {
1413                 response = DLM_IVBUFLEN;
1414                 goto send_response;
1415         }
1416
1417 way_up_top:
1418         spin_lock(&dlm->spinlock);
1419         res = __dlm_lookup_lockres(dlm, name, namelen, hash);
1420         if (res) {
1421                 spin_unlock(&dlm->spinlock);
1422
1423                 /* take care of the easy cases up front */
1424                 spin_lock(&res->spinlock);
1425                 if (res->state & (DLM_LOCK_RES_RECOVERING|
1426                                   DLM_LOCK_RES_MIGRATING)) {
1427                         spin_unlock(&res->spinlock);
1428                         mlog(0, "returning DLM_MASTER_RESP_ERROR since res is "
1429                              "being recovered/migrated\n");
1430                         response = DLM_MASTER_RESP_ERROR;
1431                         if (mle)
1432                                 kmem_cache_free(dlm_mle_cache, mle);
1433                         goto send_response;
1434                 }
1435
1436                 if (res->owner == dlm->node_num) {
1437                         dlm_lockres_set_refmap_bit(dlm, res, request->node_idx);
1438                         spin_unlock(&res->spinlock);
1439                         response = DLM_MASTER_RESP_YES;
1440                         if (mle)
1441                                 kmem_cache_free(dlm_mle_cache, mle);
1442
1443                         /* this node is the owner.
1444                          * there is some extra work that needs to
1445                          * happen now.  the requesting node has
1446                          * caused all nodes up to this one to
1447                          * create mles.  this node now needs to
1448                          * go back and clean those up. */
1449                         dispatch_assert = 1;
1450                         goto send_response;
1451                 } else if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
1452                         spin_unlock(&res->spinlock);
1453                         // mlog(0, "node %u is the master\n", res->owner);
1454                         response = DLM_MASTER_RESP_NO;
1455                         if (mle)
1456                                 kmem_cache_free(dlm_mle_cache, mle);
1457                         goto send_response;
1458                 }
1459
1460                 /* ok, there is no owner.  either this node is
1461                  * being blocked, or it is actively trying to
1462                  * master this lock. */
1463                 if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
1464                         mlog(ML_ERROR, "lock with no owner should be "
1465                              "in-progress!\n");
1466                         BUG();
1467                 }
1468
1469                 // mlog(0, "lockres is in progress...\n");
1470                 spin_lock(&dlm->master_lock);
1471                 found = dlm_find_mle(dlm, &tmpmle, name, namelen);
1472                 if (!found) {
1473                         mlog(ML_ERROR, "no mle found for this lock!\n");
1474                         BUG();
1475                 }
1476                 set_maybe = 1;
1477                 spin_lock(&tmpmle->spinlock);
1478                 if (tmpmle->type == DLM_MLE_BLOCK) {
1479                         // mlog(0, "this node is waiting for "
1480                         // "lockres to be mastered\n");
1481                         response = DLM_MASTER_RESP_NO;
1482                 } else if (tmpmle->type == DLM_MLE_MIGRATION) {
1483                         mlog(0, "node %u is master, but trying to migrate to "
1484                              "node %u.\n", tmpmle->master, tmpmle->new_master);
1485                         if (tmpmle->master == dlm->node_num) {
1486                                 mlog(ML_ERROR, "no owner on lockres, but this "
1487                                      "node is trying to migrate it to %u?!\n",
1488                                      tmpmle->new_master);
1489                                 BUG();
1490                         } else {
1491                                 /* the real master can respond on its own */
1492                                 response = DLM_MASTER_RESP_NO;
1493                         }
1494                 } else if (tmpmle->master != DLM_LOCK_RES_OWNER_UNKNOWN) {
1495                         set_maybe = 0;
1496                         if (tmpmle->master == dlm->node_num) {
1497                                 response = DLM_MASTER_RESP_YES;
1498                                 /* this node will be the owner.
1499                                  * go back and clean the mles on any
1500                                  * other nodes */
1501                                 dispatch_assert = 1;
1502                                 dlm_lockres_set_refmap_bit(dlm, res,
1503                                                            request->node_idx);
1504                         } else
1505                                 response = DLM_MASTER_RESP_NO;
1506                 } else {
1507                         // mlog(0, "this node is attempting to "
1508                         // "master lockres\n");
1509                         response = DLM_MASTER_RESP_MAYBE;
1510                 }
1511                 if (set_maybe)
1512                         set_bit(request->node_idx, tmpmle->maybe_map);
1513                 spin_unlock(&tmpmle->spinlock);
1514
1515                 spin_unlock(&dlm->master_lock);
1516                 spin_unlock(&res->spinlock);
1517
1518                 /* keep the mle attached to heartbeat events */
1519                 dlm_put_mle(tmpmle);
1520                 if (mle)
1521                         kmem_cache_free(dlm_mle_cache, mle);
1522                 goto send_response;
1523         }
1524
1525         /*
1526          * lockres doesn't exist on this node
1527          * if there is an MLE_BLOCK, return NO
1528          * if there is an MLE_MASTER, return MAYBE
1529          * otherwise, add an MLE_BLOCK, return NO
1530          */
1531         spin_lock(&dlm->master_lock);
1532         found = dlm_find_mle(dlm, &tmpmle, name, namelen);
1533         if (!found) {
1534                 /* this lockid has never been seen on this node yet */
1535                 // mlog(0, "no mle found\n");
1536                 if (!mle) {
1537                         spin_unlock(&dlm->master_lock);
1538                         spin_unlock(&dlm->spinlock);
1539
1540                         mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
1541                         if (!mle) {
1542                                 response = DLM_MASTER_RESP_ERROR;
1543                                 mlog_errno(-ENOMEM);
1544                                 goto send_response;
1545                         }
1546                         goto way_up_top;
1547                 }
1548
1549                 // mlog(0, "this is second time thru, already allocated, "
1550                 // "add the block.\n");
1551                 dlm_init_mle(mle, DLM_MLE_BLOCK, dlm, NULL, name, namelen);
1552                 set_bit(request->node_idx, mle->maybe_map);
1553                 __dlm_insert_mle(dlm, mle);
1554                 response = DLM_MASTER_RESP_NO;
1555         } else {
1556                 // mlog(0, "mle was found\n");
1557                 set_maybe = 1;
1558                 spin_lock(&tmpmle->spinlock);
1559                 if (tmpmle->master == dlm->node_num) {
1560                         mlog(ML_ERROR, "no lockres, but an mle with this node as master!\n");
1561                         BUG();
1562                 }
1563                 if (tmpmle->type == DLM_MLE_BLOCK)
1564                         response = DLM_MASTER_RESP_NO;
1565                 else if (tmpmle->type == DLM_MLE_MIGRATION) {
1566                         mlog(0, "migration mle was found (%u->%u)\n",
1567                              tmpmle->master, tmpmle->new_master);
1568                         /* real master can respond on its own */
1569                         response = DLM_MASTER_RESP_NO;
1570                 } else
1571                         response = DLM_MASTER_RESP_MAYBE;
1572                 if (set_maybe)
1573                         set_bit(request->node_idx, tmpmle->maybe_map);
1574                 spin_unlock(&tmpmle->spinlock);
1575         }
1576         spin_unlock(&dlm->master_lock);
1577         spin_unlock(&dlm->spinlock);
1578
1579         if (found) {
1580                 /* keep the mle attached to heartbeat events */
1581                 dlm_put_mle(tmpmle);
1582         }
1583 send_response:
1584         /*
1585          * __dlm_lookup_lockres() grabbed a reference to this lockres.
1586          * The reference is released by dlm_assert_master_worker() under
1587          * the call to dlm_dispatch_assert_master().  If
1588          * dlm_assert_master_worker() isn't called, we drop it here.
1589          */
1590         if (dispatch_assert) {
1591                 if (response != DLM_MASTER_RESP_YES)
1592                         mlog(ML_ERROR, "invalid response %d\n", response);
1593                 if (!res) {
1594                         mlog(ML_ERROR, "bad lockres while trying to assert!\n");
1595                         BUG();
1596                 }
1597                 mlog(0, "%u is the owner of %.*s, cleaning everyone else\n",
1598                              dlm->node_num, res->lockname.len, res->lockname.name);
1599                 ret = dlm_dispatch_assert_master(dlm, res, 0, request->node_idx,
1600                                                  DLM_ASSERT_MASTER_MLE_CLEANUP);
1601                 if (ret < 0) {
1602                         mlog(ML_ERROR, "failed to dispatch assert master work\n");
1603                         response = DLM_MASTER_RESP_ERROR;
1604                         dlm_lockres_put(res);
1605                 }
1606         } else {
1607                 if (res)
1608                         dlm_lockres_put(res);
1609         }
1610
1611         dlm_put(dlm);
1612         return response;
1613 }
1614
1615 /*
1616  * DLM_ASSERT_MASTER_MSG
1617  */
1618
1619
1620 /*
1621  * NOTE: this can be used for debugging
1622  * can periodically run all locks owned by this node
1623  * and re-assert across the cluster...
1624  */
1625 static int dlm_do_assert_master(struct dlm_ctxt *dlm,
1626                                 struct dlm_lock_resource *res,
1627                                 void *nodemap, u32 flags)
1628 {
1629         struct dlm_assert_master assert;
1630         int to, tmpret;
1631         struct dlm_node_iter iter;
1632         int ret = 0;
1633         int reassert;
1634         const char *lockname = res->lockname.name;
1635         unsigned int namelen = res->lockname.len;
1636
1637         BUG_ON(namelen > O2NM_MAX_NAME_LEN);
1638
1639         spin_lock(&res->spinlock);
1640         res->state |= DLM_LOCK_RES_SETREF_INPROG;
1641         spin_unlock(&res->spinlock);
1642
1643 again:
1644         reassert = 0;
1645
1646         /* note that if this nodemap is empty, it returns 0 */
1647         dlm_node_iter_init(nodemap, &iter);
1648         while ((to = dlm_node_iter_next(&iter)) >= 0) {
1649                 int r = 0;
1650                 struct dlm_master_list_entry *mle = NULL;
1651
1652                 mlog(0, "sending assert master to %d (%.*s)\n", to,
1653                      namelen, lockname);
1654                 memset(&assert, 0, sizeof(assert));
1655                 assert.node_idx = dlm->node_num;
1656                 assert.namelen = namelen;
1657                 memcpy(assert.name, lockname, namelen);
1658                 assert.flags = cpu_to_be32(flags);
1659
1660                 tmpret = o2net_send_message(DLM_ASSERT_MASTER_MSG, dlm->key,
1661                                             &assert, sizeof(assert), to, &r);
1662                 if (tmpret < 0) {
1663                         mlog(ML_ERROR, "Error %d when sending message %u (key "
1664                              "0x%x) to node %u\n", tmpret,
1665                              DLM_ASSERT_MASTER_MSG, dlm->key, to);
1666                         if (!dlm_is_host_down(tmpret)) {
1667                                 mlog(ML_ERROR, "unhandled error=%d!\n", tmpret);
1668                                 BUG();
1669                         }
1670                         /* a node died.  finish out the rest of the nodes. */
1671                         mlog(0, "link to %d went down!\n", to);
1672                         /* any nonzero status return will do */
1673                         ret = tmpret;
1674                         r = 0;
1675                 } else if (r < 0) {
1676                         /* ok, something horribly messed.  kill thyself. */
1677                         mlog(ML_ERROR,"during assert master of %.*s to %u, "
1678                              "got %d.\n", namelen, lockname, to, r);
1679                         spin_lock(&dlm->spinlock);
1680                         spin_lock(&dlm->master_lock);
1681                         if (dlm_find_mle(dlm, &mle, (char *)lockname,
1682                                          namelen)) {
1683                                 dlm_print_one_mle(mle);
1684                                 __dlm_put_mle(mle);
1685                         }
1686                         spin_unlock(&dlm->master_lock);
1687                         spin_unlock(&dlm->spinlock);
1688                         BUG();
1689                 }
1690
1691                 if (r & DLM_ASSERT_RESPONSE_REASSERT &&
1692                     !(r & DLM_ASSERT_RESPONSE_MASTERY_REF)) {
1693                                 mlog(ML_ERROR, "%.*s: very strange, "
1694                                      "master MLE but no lockres on %u\n",
1695                                      namelen, lockname, to);
1696                 }
1697
1698                 if (r & DLM_ASSERT_RESPONSE_REASSERT) {
1699                         mlog(0, "%.*s: node %u create mles on other "
1700                              "nodes and requests a re-assert\n",
1701                              namelen, lockname, to);
1702                         reassert = 1;
1703                 }
1704                 if (r & DLM_ASSERT_RESPONSE_MASTERY_REF) {
1705                         mlog(0, "%.*s: node %u has a reference to this "
1706                              "lockres, set the bit in the refmap\n",
1707                              namelen, lockname, to);
1708                         spin_lock(&res->spinlock);
1709                         dlm_lockres_set_refmap_bit(dlm, res, to);
1710                         spin_unlock(&res->spinlock);
1711                 }
1712         }
1713
1714         if (reassert)
1715                 goto again;
1716
1717         spin_lock(&res->spinlock);
1718         res->state &= ~DLM_LOCK_RES_SETREF_INPROG;
1719         spin_unlock(&res->spinlock);
1720         wake_up(&res->wq);
1721
1722         return ret;
1723 }
1724
1725 /*
1726  * locks that can be taken here:
1727  * dlm->spinlock
1728  * res->spinlock
1729  * mle->spinlock
1730  * dlm->master_list
1731  *
1732  * if possible, TRIM THIS DOWN!!!
1733  */
1734 int dlm_assert_master_handler(struct o2net_msg *msg, u32 len, void *data,
1735                               void **ret_data)
1736 {
1737         struct dlm_ctxt *dlm = data;
1738         struct dlm_master_list_entry *mle = NULL;
1739         struct dlm_assert_master *assert = (struct dlm_assert_master *)msg->buf;
1740         struct dlm_lock_resource *res = NULL;
1741         char *name;
1742         unsigned int namelen, hash;
1743         u32 flags;
1744         int master_request = 0, have_lockres_ref = 0;
1745         int ret = 0;
1746
1747         if (!dlm_grab(dlm))
1748                 return 0;
1749
1750         name = assert->name;
1751         namelen = assert->namelen;
1752         hash = dlm_lockid_hash(name, namelen);
1753         flags = be32_to_cpu(assert->flags);
1754
1755         if (namelen > DLM_LOCKID_NAME_MAX) {
1756                 mlog(ML_ERROR, "Invalid name length!");
1757                 goto done;
1758         }
1759
1760         spin_lock(&dlm->spinlock);
1761
1762         if (flags)
1763                 mlog(0, "assert_master with flags: %u\n", flags);
1764
1765         /* find the MLE */
1766         spin_lock(&dlm->master_lock);
1767         if (!dlm_find_mle(dlm, &mle, name, namelen)) {
1768                 /* not an error, could be master just re-asserting */
1769                 mlog(0, "just got an assert_master from %u, but no "
1770                      "MLE for it! (%.*s)\n", assert->node_idx,
1771                      namelen, name);
1772         } else {
1773                 int bit = find_next_bit (mle->maybe_map, O2NM_MAX_NODES, 0);
1774                 if (bit >= O2NM_MAX_NODES) {
1775                         /* not necessarily an error, though less likely.
1776                          * could be master just re-asserting. */
1777                         mlog(0, "no bits set in the maybe_map, but %u "
1778                              "is asserting! (%.*s)\n", assert->node_idx,
1779                              namelen, name);
1780                 } else if (bit != assert->node_idx) {
1781                         if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
1782                                 mlog(0, "master %u was found, %u should "
1783                                      "back off\n", assert->node_idx, bit);
1784                         } else {
1785                                 /* with the fix for bug 569, a higher node
1786                                  * number winning the mastery will respond
1787                                  * YES to mastery requests, but this node
1788                                  * had no way of knowing.  let it pass. */
1789                                 mlog(0, "%u is the lowest node, "
1790                                      "%u is asserting. (%.*s)  %u must "
1791                                      "have begun after %u won.\n", bit,
1792                                      assert->node_idx, namelen, name, bit,
1793                                      assert->node_idx);
1794                         }
1795                 }
1796                 if (mle->type == DLM_MLE_MIGRATION) {
1797                         if (flags & DLM_ASSERT_MASTER_MLE_CLEANUP) {
1798                                 mlog(0, "%s:%.*s: got cleanup assert"
1799                                      " from %u for migration\n",
1800                                      dlm->name, namelen, name,
1801                                      assert->node_idx);
1802                         } else if (!(flags & DLM_ASSERT_MASTER_FINISH_MIGRATION)) {
1803                                 mlog(0, "%s:%.*s: got unrelated assert"
1804                                      " from %u for migration, ignoring\n",
1805                                      dlm->name, namelen, name,
1806                                      assert->node_idx);
1807                                 __dlm_put_mle(mle);
1808                                 spin_unlock(&dlm->master_lock);
1809                                 spin_unlock(&dlm->spinlock);
1810                                 goto done;
1811                         }
1812                 }
1813         }
1814         spin_unlock(&dlm->master_lock);
1815
1816         /* ok everything checks out with the MLE
1817          * now check to see if there is a lockres */
1818         res = __dlm_lookup_lockres(dlm, name, namelen, hash);
1819         if (res) {
1820                 spin_lock(&res->spinlock);
1821                 if (res->state & DLM_LOCK_RES_RECOVERING)  {
1822                         mlog(ML_ERROR, "%u asserting but %.*s is "
1823                              "RECOVERING!\n", assert->node_idx, namelen, name);
1824                         goto kill;
1825                 }
1826                 if (!mle) {
1827                         if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN &&
1828                             res->owner != assert->node_idx) {
1829                                 mlog(ML_ERROR, "DIE! Mastery assert from %u, "
1830                                      "but current owner is %u! (%.*s)\n",
1831                                      assert->node_idx, res->owner, namelen,
1832                                      name);
1833                                 __dlm_print_one_lock_resource(res);
1834                                 BUG();
1835                         }
1836                 } else if (mle->type != DLM_MLE_MIGRATION) {
1837                         if (res->owner != DLM_LOCK_RES_OWNER_UNKNOWN) {
1838                                 /* owner is just re-asserting */
1839                                 if (res->owner == assert->node_idx) {
1840                                         mlog(0, "owner %u re-asserting on "
1841                                              "lock %.*s\n", assert->node_idx,
1842                                              namelen, name);
1843                                         goto ok;
1844                                 }
1845                                 mlog(ML_ERROR, "got assert_master from "
1846                                      "node %u, but %u is the owner! "
1847                                      "(%.*s)\n", assert->node_idx,
1848                                      res->owner, namelen, name);
1849                                 goto kill;
1850                         }
1851                         if (!(res->state & DLM_LOCK_RES_IN_PROGRESS)) {
1852                                 mlog(ML_ERROR, "got assert from %u, but lock "
1853                                      "with no owner should be "
1854                                      "in-progress! (%.*s)\n",
1855                                      assert->node_idx,
1856                                      namelen, name);
1857                                 goto kill;
1858                         }
1859                 } else /* mle->type == DLM_MLE_MIGRATION */ {
1860                         /* should only be getting an assert from new master */
1861                         if (assert->node_idx != mle->new_master) {
1862                                 mlog(ML_ERROR, "got assert from %u, but "
1863                                      "new master is %u, and old master "
1864                                      "was %u (%.*s)\n",
1865                                      assert->node_idx, mle->new_master,
1866                                      mle->master, namelen, name);
1867                                 goto kill;
1868                         }
1869
1870                 }
1871 ok:
1872                 spin_unlock(&res->spinlock);
1873         }
1874
1875         // mlog(0, "woo!  got an assert_master from node %u!\n",
1876         //           assert->node_idx);
1877         if (mle) {
1878                 int extra_ref = 0;
1879                 int nn = -1;
1880                 int rr, err = 0;
1881
1882                 spin_lock(&mle->spinlock);
1883                 if (mle->type == DLM_MLE_BLOCK || mle->type == DLM_MLE_MIGRATION)
1884                         extra_ref = 1;
1885                 else {
1886                         /* MASTER mle: if any bits set in the response map
1887                          * then the calling node needs to re-assert to clear
1888                          * up nodes that this node contacted */
1889                         while ((nn = find_next_bit (mle->response_map, O2NM_MAX_NODES,
1890                                                     nn+1)) < O2NM_MAX_NODES) {
1891                                 if (nn != dlm->node_num && nn != assert->node_idx)
1892                                         master_request = 1;
1893                         }
1894                 }
1895                 mle->master = assert->node_idx;
1896                 atomic_set(&mle->woken, 1);
1897                 wake_up(&mle->wq);
1898                 spin_unlock(&mle->spinlock);
1899
1900                 if (res) {
1901                         int wake = 0;
1902                         spin_lock(&res->spinlock);
1903                         if (mle->type == DLM_MLE_MIGRATION) {
1904                                 mlog(0, "finishing off migration of lockres %.*s, "
1905                                         "from %u to %u\n",
1906                                         res->lockname.len, res->lockname.name,
1907                                         dlm->node_num, mle->new_master);
1908                                 res->state &= ~DLM_LOCK_RES_MIGRATING;
1909                                 wake = 1;
1910                                 dlm_change_lockres_owner(dlm, res, mle->new_master);
1911                                 BUG_ON(res->state & DLM_LOCK_RES_DIRTY);
1912                         } else {
1913                                 dlm_change_lockres_owner(dlm, res, mle->master);
1914                         }
1915                         spin_unlock(&res->spinlock);
1916                         have_lockres_ref = 1;
1917                         if (wake)
1918                                 wake_up(&res->wq);
1919                 }
1920
1921                 /* master is known, detach if not already detached.
1922                  * ensures that only one assert_master call will happen
1923                  * on this mle. */
1924                 spin_lock(&dlm->master_lock);
1925
1926                 rr = atomic_read(&mle->mle_refs.refcount);
1927                 if (mle->inuse > 0) {
1928                         if (extra_ref && rr < 3)
1929                                 err = 1;
1930                         else if (!extra_ref && rr < 2)
1931                                 err = 1;
1932                 } else {
1933                         if (extra_ref && rr < 2)
1934                                 err = 1;
1935                         else if (!extra_ref && rr < 1)
1936                                 err = 1;
1937                 }
1938                 if (err) {
1939                         mlog(ML_ERROR, "%s:%.*s: got assert master from %u "
1940                              "that will mess up this node, refs=%d, extra=%d, "
1941                              "inuse=%d\n", dlm->name, namelen, name,
1942                              assert->node_idx, rr, extra_ref, mle->inuse);
1943                         dlm_print_one_mle(mle);
1944                 }
1945                 __dlm_unlink_mle(dlm, mle);
1946                 __dlm_mle_detach_hb_events(dlm, mle);
1947                 __dlm_put_mle(mle);
1948                 if (extra_ref) {
1949                         /* the assert master message now balances the extra
1950                          * ref given by the master / migration request message.
1951                          * if this is the last put, it will be removed
1952                          * from the list. */
1953                         __dlm_put_mle(mle);
1954                 }
1955                 spin_unlock(&dlm->master_lock);
1956         } else if (res) {
1957                 if (res->owner != assert->node_idx) {
1958                         mlog(0, "assert_master from %u, but current "
1959                              "owner is %u (%.*s), no mle\n", assert->node_idx,
1960                              res->owner, namelen, name);
1961                 }
1962         }
1963         spin_unlock(&dlm->spinlock);
1964
1965 done:
1966         ret = 0;
1967         if (res) {
1968                 spin_lock(&res->spinlock);
1969                 res->state |= DLM_LOCK_RES_SETREF_INPROG;
1970                 spin_unlock(&res->spinlock);
1971                 *ret_data = (void *)res;
1972         }
1973         dlm_put(dlm);
1974         if (master_request) {
1975                 mlog(0, "need to tell master to reassert\n");
1976                 /* positive. negative would shoot down the node. */
1977                 ret |= DLM_ASSERT_RESPONSE_REASSERT;
1978                 if (!have_lockres_ref) {
1979                         mlog(ML_ERROR, "strange, got assert from %u, MASTER "
1980                              "mle present here for %s:%.*s, but no lockres!\n",
1981                              assert->node_idx, dlm->name, namelen, name);
1982                 }
1983         }
1984         if (have_lockres_ref) {
1985                 /* let the master know we have a reference to the lockres */
1986                 ret |= DLM_ASSERT_RESPONSE_MASTERY_REF;
1987                 mlog(0, "%s:%.*s: got assert from %u, need a ref\n",
1988                      dlm->name, namelen, name, assert->node_idx);
1989         }
1990         return ret;
1991
1992 kill:
1993         /* kill the caller! */
1994         mlog(ML_ERROR, "Bad message received from another node.  Dumping state "
1995              "and killing the other node now!  This node is OK and can continue.\n");
1996         __dlm_print_one_lock_resource(res);
1997         spin_unlock(&res->spinlock);
1998         spin_unlock(&dlm->spinlock);
1999         *ret_data = (void *)res;
2000         dlm_put(dlm);
2001         return -EINVAL;
2002 }
2003
2004 void dlm_assert_master_post_handler(int status, void *data, void *ret_data)
2005 {
2006         struct dlm_lock_resource *res = (struct dlm_lock_resource *)ret_data;
2007
2008         if (ret_data) {
2009                 spin_lock(&res->spinlock);
2010                 res->state &= ~DLM_LOCK_RES_SETREF_INPROG;
2011                 spin_unlock(&res->spinlock);
2012                 wake_up(&res->wq);
2013                 dlm_lockres_put(res);
2014         }
2015         return;
2016 }
2017
2018 int dlm_dispatch_assert_master(struct dlm_ctxt *dlm,
2019                                struct dlm_lock_resource *res,
2020                                int ignore_higher, u8 request_from, u32 flags)
2021 {
2022         struct dlm_work_item *item;
2023         item = kzalloc(sizeof(*item), GFP_NOFS);
2024         if (!item)
2025                 return -ENOMEM;
2026
2027
2028         /* queue up work for dlm_assert_master_worker */
2029         dlm_grab(dlm);  /* get an extra ref for the work item */
2030         dlm_init_work_item(dlm, item, dlm_assert_master_worker, NULL);
2031         item->u.am.lockres = res; /* already have a ref */
2032         /* can optionally ignore node numbers higher than this node */
2033         item->u.am.ignore_higher = ignore_higher;
2034         item->u.am.request_from = request_from;
2035         item->u.am.flags = flags;
2036
2037         if (ignore_higher)
2038                 mlog(0, "IGNORE HIGHER: %.*s\n", res->lockname.len,
2039                      res->lockname.name);
2040
2041         spin_lock(&dlm->work_lock);
2042         list_add_tail(&item->list, &dlm->work_list);
2043         spin_unlock(&dlm->work_lock);
2044
2045         queue_work(dlm->dlm_worker, &dlm->dispatched_work);
2046         return 0;
2047 }
2048
2049 static void dlm_assert_master_worker(struct dlm_work_item *item, void *data)
2050 {
2051         struct dlm_ctxt *dlm = data;
2052         int ret = 0;
2053         struct dlm_lock_resource *res;
2054         unsigned long nodemap[BITS_TO_LONGS(O2NM_MAX_NODES)];
2055         int ignore_higher;
2056         int bit;
2057         u8 request_from;
2058         u32 flags;
2059
2060         dlm = item->dlm;
2061         res = item->u.am.lockres;
2062         ignore_higher = item->u.am.ignore_higher;
2063         request_from = item->u.am.request_from;
2064         flags = item->u.am.flags;
2065
2066         spin_lock(&dlm->spinlock);
2067         memcpy(nodemap, dlm->domain_map, sizeof(nodemap));
2068         spin_unlock(&dlm->spinlock);
2069
2070         clear_bit(dlm->node_num, nodemap);
2071         if (ignore_higher) {
2072                 /* if is this just to clear up mles for nodes below
2073                  * this node, do not send the message to the original
2074                  * caller or any node number higher than this */
2075                 clear_bit(request_from, nodemap);
2076                 bit = dlm->node_num;
2077                 while (1) {
2078                         bit = find_next_bit(nodemap, O2NM_MAX_NODES,
2079                                             bit+1);
2080                         if (bit >= O2NM_MAX_NODES)
2081                                 break;
2082                         clear_bit(bit, nodemap);
2083                 }
2084         }
2085
2086         /*
2087          * If we're migrating this lock to someone else, we are no
2088          * longer allowed to assert out own mastery.  OTOH, we need to
2089          * prevent migration from starting while we're still asserting
2090          * our dominance.  The reserved ast delays migration.
2091          */
2092         spin_lock(&res->spinlock);
2093         if (res->state & DLM_LOCK_RES_MIGRATING) {
2094                 mlog(0, "Someone asked us to assert mastery, but we're "
2095                      "in the middle of migration.  Skipping assert, "
2096                      "the new master will handle that.\n");
2097                 spin_unlock(&res->spinlock);
2098                 goto put;
2099         } else
2100                 __dlm_lockres_reserve_ast(res);
2101         spin_unlock(&res->spinlock);
2102
2103         /* this call now finishes out the nodemap
2104          * even if one or more nodes die */
2105         mlog(0, "worker about to master %.*s here, this=%u\n",
2106                      res->lockname.len, res->lockname.name, dlm->node_num);
2107         ret = dlm_do_assert_master(dlm, res, nodemap, flags);
2108         if (ret < 0) {
2109                 /* no need to restart, we are done */
2110                 if (!dlm_is_host_down(ret))
2111                         mlog_errno(ret);
2112         }
2113
2114         /* Ok, we've asserted ourselves.  Let's let migration start. */
2115         dlm_lockres_release_ast(dlm, res);
2116
2117 put:
2118         dlm_lockres_put(res);
2119
2120         mlog(0, "finished with dlm_assert_master_worker\n");
2121 }
2122
2123 /* SPECIAL CASE for the $RECOVERY lock used by the recovery thread.
2124  * We cannot wait for node recovery to complete to begin mastering this
2125  * lockres because this lockres is used to kick off recovery! ;-)
2126  * So, do a pre-check on all living nodes to see if any of those nodes
2127  * think that $RECOVERY is currently mastered by a dead node.  If so,
2128  * we wait a short time to allow that node to get notified by its own
2129  * heartbeat stack, then check again.  All $RECOVERY lock resources
2130  * mastered by dead nodes are purged when the hearbeat callback is
2131  * fired, so we can know for sure that it is safe to continue once
2132  * the node returns a live node or no node.  */
2133 static int dlm_pre_master_reco_lockres(struct dlm_ctxt *dlm,
2134                                        struct dlm_lock_resource *res)
2135 {
2136         struct dlm_node_iter iter;
2137         int nodenum;
2138         int ret = 0;
2139         u8 master = DLM_LOCK_RES_OWNER_UNKNOWN;
2140
2141         spin_lock(&dlm->spinlock);
2142         dlm_node_iter_init(dlm->domain_map, &iter);
2143         spin_unlock(&dlm->spinlock);
2144
2145         while ((nodenum = dlm_node_iter_next(&iter)) >= 0) {
2146                 /* do not send to self */
2147                 if (nodenum == dlm->node_num)
2148                         continue;
2149                 ret = dlm_do_master_requery(dlm, res, nodenum, &master);
2150                 if (ret < 0) {
2151                         mlog_errno(ret);
2152                         if (!dlm_is_host_down(ret))
2153                                 BUG();
2154                         /* host is down, so answer for that node would be
2155                          * DLM_LOCK_RES_OWNER_UNKNOWN.  continue. */
2156                         ret = 0;
2157                 }
2158
2159                 if (master != DLM_LOCK_RES_OWNER_UNKNOWN) {
2160                         /* check to see if this master is in the recovery map */
2161                         spin_lock(&dlm->spinlock);
2162                         if (test_bit(master, dlm->recovery_map)) {
2163                                 mlog(ML_NOTICE, "%s: node %u has not seen "
2164                                      "node %u go down yet, and thinks the "
2165                                      "dead node is mastering the recovery "
2166                                      "lock.  must wait.\n", dlm->name,
2167                                      nodenum, master);
2168                                 ret = -EAGAIN;
2169                         }
2170                         spin_unlock(&dlm->spinlock);
2171                         mlog(0, "%s: reco lock master is %u\n", dlm->name,
2172                              master);
2173                         break;
2174                 }
2175         }
2176         return ret;
2177 }
2178
2179 /*
2180  * DLM_DEREF_LOCKRES_MSG
2181  */
2182
2183 int dlm_drop_lockres_ref(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
2184 {
2185         struct dlm_deref_lockres deref;
2186         int ret = 0, r;
2187         const char *lockname;
2188         unsigned int namelen;
2189
2190         lockname = res->lockname.name;
2191         namelen = res->lockname.len;
2192         BUG_ON(namelen > O2NM_MAX_NAME_LEN);
2193
2194         memset(&deref, 0, sizeof(deref));
2195         deref.node_idx = dlm->node_num;
2196         deref.namelen = namelen;
2197         memcpy(deref.name, lockname, namelen);
2198
2199         ret = o2net_send_message(DLM_DEREF_LOCKRES_MSG, dlm->key,
2200                                  &deref, sizeof(deref), res->owner, &r);
2201         if (ret < 0)
2202                 mlog(ML_ERROR, "%s: res %.*s, error %d send DEREF to node %u\n",
2203                      dlm->name, namelen, lockname, ret, res->owner);
2204         else if (r < 0) {
2205                 /* BAD.  other node says I did not have a ref. */
2206                 mlog(ML_ERROR, "%s: res %.*s, DEREF to node %u got %d\n",
2207                      dlm->name, namelen, lockname, res->owner, r);
2208                 dlm_print_one_lock_resource(res);
2209                 BUG();
2210         }
2211         return ret;
2212 }
2213
2214 int dlm_deref_lockres_handler(struct o2net_msg *msg, u32 len, void *data,
2215                               void **ret_data)
2216 {
2217         struct dlm_ctxt *dlm = data;
2218         struct dlm_deref_lockres *deref = (struct dlm_deref_lockres *)msg->buf;
2219         struct dlm_lock_resource *res = NULL;
2220         char *name;
2221         unsigned int namelen;
2222         int ret = -EINVAL;
2223         u8 node;
2224         unsigned int hash;
2225         struct dlm_work_item *item;
2226         int cleared = 0;
2227         int dispatch = 0;
2228
2229         if (!dlm_grab(dlm))
2230                 return 0;
2231
2232         name = deref->name;
2233         namelen = deref->namelen;
2234         node = deref->node_idx;
2235
2236         if (namelen > DLM_LOCKID_NAME_MAX) {
2237                 mlog(ML_ERROR, "Invalid name length!");
2238                 goto done;
2239         }
2240         if (deref->node_idx >= O2NM_MAX_NODES) {
2241                 mlog(ML_ERROR, "Invalid node number: %u\n", node);
2242                 goto done;
2243         }
2244
2245         hash = dlm_lockid_hash(name, namelen);
2246
2247         spin_lock(&dlm->spinlock);
2248         res = __dlm_lookup_lockres_full(dlm, name, namelen, hash);
2249         if (!res) {
2250                 spin_unlock(&dlm->spinlock);
2251                 mlog(ML_ERROR, "%s:%.*s: bad lockres name\n",
2252                      dlm->name, namelen, name);
2253                 goto done;
2254         }
2255         spin_unlock(&dlm->spinlock);
2256
2257         spin_lock(&res->spinlock);
2258         if (res->state & DLM_LOCK_RES_SETREF_INPROG)
2259                 dispatch = 1;
2260         else {
2261                 BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF);
2262                 if (test_bit(node, res->refmap)) {
2263                         dlm_lockres_clear_refmap_bit(dlm, res, node);
2264                         cleared = 1;
2265                 }
2266         }
2267         spin_unlock(&res->spinlock);
2268
2269         if (!dispatch) {
2270                 if (cleared)
2271                         dlm_lockres_calc_usage(dlm, res);
2272                 else {
2273                         mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2274                         "but it is already dropped!\n", dlm->name,
2275                         res->lockname.len, res->lockname.name, node);
2276                         dlm_print_one_lock_resource(res);
2277                 }
2278                 ret = 0;
2279                 goto done;
2280         }
2281
2282         item = kzalloc(sizeof(*item), GFP_NOFS);
2283         if (!item) {
2284                 ret = -ENOMEM;
2285                 mlog_errno(ret);
2286                 goto done;
2287         }
2288
2289         dlm_init_work_item(dlm, item, dlm_deref_lockres_worker, NULL);
2290         item->u.dl.deref_res = res;
2291         item->u.dl.deref_node = node;
2292
2293         spin_lock(&dlm->work_lock);
2294         list_add_tail(&item->list, &dlm->work_list);
2295         spin_unlock(&dlm->work_lock);
2296
2297         queue_work(dlm->dlm_worker, &dlm->dispatched_work);
2298         return 0;
2299
2300 done:
2301         if (res)
2302                 dlm_lockres_put(res);
2303         dlm_put(dlm);
2304
2305         return ret;
2306 }
2307
2308 static void dlm_deref_lockres_worker(struct dlm_work_item *item, void *data)
2309 {
2310         struct dlm_ctxt *dlm;
2311         struct dlm_lock_resource *res;
2312         u8 node;
2313         u8 cleared = 0;
2314
2315         dlm = item->dlm;
2316         res = item->u.dl.deref_res;
2317         node = item->u.dl.deref_node;
2318
2319         spin_lock(&res->spinlock);
2320         BUG_ON(res->state & DLM_LOCK_RES_DROPPING_REF);
2321         if (test_bit(node, res->refmap)) {
2322                 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
2323                 dlm_lockres_clear_refmap_bit(dlm, res, node);
2324                 cleared = 1;
2325         }
2326         spin_unlock(&res->spinlock);
2327
2328         if (cleared) {
2329                 mlog(0, "%s:%.*s node %u ref dropped in dispatch\n",
2330                      dlm->name, res->lockname.len, res->lockname.name, node);
2331                 dlm_lockres_calc_usage(dlm, res);
2332         } else {
2333                 mlog(ML_ERROR, "%s:%.*s: node %u trying to drop ref "
2334                      "but it is already dropped!\n", dlm->name,
2335                      res->lockname.len, res->lockname.name, node);
2336                 dlm_print_one_lock_resource(res);
2337         }
2338
2339         dlm_lockres_put(res);
2340 }
2341
2342 /*
2343  * A migrateable resource is one that is :
2344  * 1. locally mastered, and,
2345  * 2. zero local locks, and,
2346  * 3. one or more non-local locks, or, one or more references
2347  * Returns 1 if yes, 0 if not.
2348  */
2349 static int dlm_is_lockres_migrateable(struct dlm_ctxt *dlm,
2350                                       struct dlm_lock_resource *res)
2351 {
2352         enum dlm_lockres_list idx;
2353         int nonlocal = 0, node_ref;
2354         struct list_head *queue;
2355         struct dlm_lock *lock;
2356         u64 cookie;
2357
2358         assert_spin_locked(&res->spinlock);
2359
2360         if (res->owner != dlm->node_num)
2361                 return 0;
2362
2363         for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) {
2364                 queue = dlm_list_idx_to_ptr(res, idx);
2365                 list_for_each_entry(lock, queue, list) {
2366                         if (lock->ml.node != dlm->node_num) {
2367                                 nonlocal++;
2368                                 continue;
2369                         }
2370                         cookie = be64_to_cpu(lock->ml.cookie);
2371                         mlog(0, "%s: Not migrateable res %.*s, lock %u:%llu on "
2372                              "%s list\n", dlm->name, res->lockname.len,
2373                              res->lockname.name,
2374                              dlm_get_lock_cookie_node(cookie),
2375                              dlm_get_lock_cookie_seq(cookie),
2376                              dlm_list_in_text(idx));
2377                         return 0;
2378                 }
2379         }
2380
2381         if (!nonlocal) {
2382                 node_ref = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
2383                 if (node_ref >= O2NM_MAX_NODES)
2384                         return 0;
2385         }
2386
2387         mlog(0, "%s: res %.*s, Migrateable\n", dlm->name, res->lockname.len,
2388              res->lockname.name);
2389
2390         return 1;
2391 }
2392
2393 /*
2394  * DLM_MIGRATE_LOCKRES
2395  */
2396
2397
2398 static int dlm_migrate_lockres(struct dlm_ctxt *dlm,
2399                                struct dlm_lock_resource *res, u8 target)
2400 {
2401         struct dlm_master_list_entry *mle = NULL;
2402         struct dlm_master_list_entry *oldmle = NULL;
2403         struct dlm_migratable_lockres *mres = NULL;
2404         int ret = 0;
2405         const char *name;
2406         unsigned int namelen;
2407         int mle_added = 0;
2408         int wake = 0;
2409
2410         if (!dlm_grab(dlm))
2411                 return -EINVAL;
2412
2413         BUG_ON(target == O2NM_MAX_NODES);
2414
2415         name = res->lockname.name;
2416         namelen = res->lockname.len;
2417
2418         mlog(0, "%s: Migrating %.*s to node %u\n", dlm->name, namelen, name,
2419              target);
2420
2421         /* preallocate up front. if this fails, abort */
2422         ret = -ENOMEM;
2423         mres = (struct dlm_migratable_lockres *) __get_free_page(GFP_NOFS);
2424         if (!mres) {
2425                 mlog_errno(ret);
2426                 goto leave;
2427         }
2428
2429         mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
2430         if (!mle) {
2431                 mlog_errno(ret);
2432                 goto leave;
2433         }
2434         ret = 0;
2435
2436         /*
2437          * clear any existing master requests and
2438          * add the migration mle to the list
2439          */
2440         spin_lock(&dlm->spinlock);
2441         spin_lock(&dlm->master_lock);
2442         ret = dlm_add_migration_mle(dlm, res, mle, &oldmle, name,
2443                                     namelen, target, dlm->node_num);
2444         spin_unlock(&dlm->master_lock);
2445         spin_unlock(&dlm->spinlock);
2446
2447         if (ret == -EEXIST) {
2448                 mlog(0, "another process is already migrating it\n");
2449                 goto fail;
2450         }
2451         mle_added = 1;
2452
2453         /*
2454          * set the MIGRATING flag and flush asts
2455          * if we fail after this we need to re-dirty the lockres
2456          */
2457         if (dlm_mark_lockres_migrating(dlm, res, target) < 0) {
2458                 mlog(ML_ERROR, "tried to migrate %.*s to %u, but "
2459                      "the target went down.\n", res->lockname.len,
2460                      res->lockname.name, target);
2461                 spin_lock(&res->spinlock);
2462                 res->state &= ~DLM_LOCK_RES_MIGRATING;
2463                 wake = 1;
2464                 spin_unlock(&res->spinlock);
2465                 ret = -EINVAL;
2466         }
2467
2468 fail:
2469         if (oldmle) {
2470                 /* master is known, detach if not already detached */
2471                 dlm_mle_detach_hb_events(dlm, oldmle);
2472                 dlm_put_mle(oldmle);
2473         }
2474
2475         if (ret < 0) {
2476                 if (mle_added) {
2477                         dlm_mle_detach_hb_events(dlm, mle);
2478                         dlm_put_mle(mle);
2479                 } else if (mle) {
2480                         kmem_cache_free(dlm_mle_cache, mle);
2481                         mle = NULL;
2482                 }
2483                 goto leave;
2484         }
2485
2486         /*
2487          * at this point, we have a migration target, an mle
2488          * in the master list, and the MIGRATING flag set on
2489          * the lockres
2490          */
2491
2492         /* now that remote nodes are spinning on the MIGRATING flag,
2493          * ensure that all assert_master work is flushed. */
2494         flush_workqueue(dlm->dlm_worker);
2495
2496         /* get an extra reference on the mle.
2497          * otherwise the assert_master from the new
2498          * master will destroy this.
2499          * also, make sure that all callers of dlm_get_mle
2500          * take both dlm->spinlock and dlm->master_lock */
2501         spin_lock(&dlm->spinlock);
2502         spin_lock(&dlm->master_lock);
2503         dlm_get_mle_inuse(mle);
2504         spin_unlock(&dlm->master_lock);
2505         spin_unlock(&dlm->spinlock);
2506
2507         /* notify new node and send all lock state */
2508         /* call send_one_lockres with migration flag.
2509          * this serves as notice to the target node that a
2510          * migration is starting. */
2511         ret = dlm_send_one_lockres(dlm, res, mres, target,
2512                                    DLM_MRES_MIGRATION);
2513
2514         if (ret < 0) {
2515                 mlog(0, "migration to node %u failed with %d\n",
2516                      target, ret);
2517                 /* migration failed, detach and clean up mle */
2518                 dlm_mle_detach_hb_events(dlm, mle);
2519                 dlm_put_mle(mle);
2520                 dlm_put_mle_inuse(mle);
2521                 spin_lock(&res->spinlock);
2522                 res->state &= ~DLM_LOCK_RES_MIGRATING;
2523                 wake = 1;
2524                 spin_unlock(&res->spinlock);
2525                 if (dlm_is_host_down(ret))
2526                         dlm_wait_for_node_death(dlm, target,
2527                                                 DLM_NODE_DEATH_WAIT_MAX);
2528                 goto leave;
2529         }
2530
2531         /* at this point, the target sends a message to all nodes,
2532          * (using dlm_do_migrate_request).  this node is skipped since
2533          * we had to put an mle in the list to begin the process.  this
2534          * node now waits for target to do an assert master.  this node
2535          * will be the last one notified, ensuring that the migration
2536          * is complete everywhere.  if the target dies while this is
2537          * going on, some nodes could potentially see the target as the
2538          * master, so it is important that my recovery finds the migration
2539          * mle and sets the master to UNKNOWN. */
2540
2541
2542         /* wait for new node to assert master */
2543         while (1) {
2544                 ret = wait_event_interruptible_timeout(mle->wq,
2545                                         (atomic_read(&mle->woken) == 1),
2546                                         msecs_to_jiffies(5000));
2547
2548                 if (ret >= 0) {
2549                         if (atomic_read(&mle->woken) == 1 ||
2550                             res->owner == target)
2551                                 break;
2552
2553                         mlog(0, "%s:%.*s: timed out during migration\n",
2554                              dlm->name, res->lockname.len, res->lockname.name);
2555                         /* avoid hang during shutdown when migrating lockres
2556                          * to a node which also goes down */
2557                         if (dlm_is_node_dead(dlm, target)) {
2558                                 mlog(0, "%s:%.*s: expected migration "
2559                                      "target %u is no longer up, restarting\n",
2560                                      dlm->name, res->lockname.len,
2561                                      res->lockname.name, target);
2562                                 ret = -EINVAL;
2563                                 /* migration failed, detach and clean up mle */
2564                                 dlm_mle_detach_hb_events(dlm, mle);
2565                                 dlm_put_mle(mle);
2566                                 dlm_put_mle_inuse(mle);
2567                                 spin_lock(&res->spinlock);
2568                                 res->state &= ~DLM_LOCK_RES_MIGRATING;
2569                                 wake = 1;
2570                                 spin_unlock(&res->spinlock);
2571                                 goto leave;
2572                         }
2573                 } else
2574                         mlog(0, "%s:%.*s: caught signal during migration\n",
2575                              dlm->name, res->lockname.len, res->lockname.name);
2576         }
2577
2578         /* all done, set the owner, clear the flag */
2579         spin_lock(&res->spinlock);
2580         dlm_set_lockres_owner(dlm, res, target);
2581         res->state &= ~DLM_LOCK_RES_MIGRATING;
2582         dlm_remove_nonlocal_locks(dlm, res);
2583         spin_unlock(&res->spinlock);
2584         wake_up(&res->wq);
2585
2586         /* master is known, detach if not already detached */
2587         dlm_mle_detach_hb_events(dlm, mle);
2588         dlm_put_mle_inuse(mle);
2589         ret = 0;
2590
2591         dlm_lockres_calc_usage(dlm, res);
2592
2593 leave:
2594         /* re-dirty the lockres if we failed */
2595         if (ret < 0)
2596                 dlm_kick_thread(dlm, res);
2597
2598         /* wake up waiters if the MIGRATING flag got set
2599          * but migration failed */
2600         if (wake)
2601                 wake_up(&res->wq);
2602
2603         if (mres)
2604                 free_page((unsigned long)mres);
2605
2606         dlm_put(dlm);
2607
2608         mlog(0, "%s: Migrating %.*s to %u, returns %d\n", dlm->name, namelen,
2609              name, target, ret);
2610         return ret;
2611 }
2612
2613 #define DLM_MIGRATION_RETRY_MS  100
2614
2615 /*
2616  * Should be called only after beginning the domain leave process.
2617  * There should not be any remaining locks on nonlocal lock resources,
2618  * and there should be no local locks left on locally mastered resources.
2619  *
2620  * Called with the dlm spinlock held, may drop it to do migration, but
2621  * will re-acquire before exit.
2622  *
2623  * Returns: 1 if dlm->spinlock was dropped/retaken, 0 if never dropped
2624  */
2625 int dlm_empty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
2626 {
2627         int ret;
2628         int lock_dropped = 0;
2629         u8 target = O2NM_MAX_NODES;
2630
2631         assert_spin_locked(&dlm->spinlock);
2632
2633         spin_lock(&res->spinlock);
2634         if (dlm_is_lockres_migrateable(dlm, res))
2635                 target = dlm_pick_migration_target(dlm, res);
2636         spin_unlock(&res->spinlock);
2637
2638         if (target == O2NM_MAX_NODES)
2639                 goto leave;
2640
2641         /* Wheee! Migrate lockres here! Will sleep so drop spinlock. */
2642         spin_unlock(&dlm->spinlock);
2643         lock_dropped = 1;
2644         ret = dlm_migrate_lockres(dlm, res, target);
2645         if (ret)
2646                 mlog(0, "%s: res %.*s, Migrate to node %u failed with %d\n",
2647                      dlm->name, res->lockname.len, res->lockname.name,
2648                      target, ret);
2649         spin_lock(&dlm->spinlock);
2650 leave:
2651         return lock_dropped;
2652 }
2653
2654 int dlm_lock_basts_flushed(struct dlm_ctxt *dlm, struct dlm_lock *lock)
2655 {
2656         int ret;
2657         spin_lock(&dlm->ast_lock);
2658         spin_lock(&lock->spinlock);
2659         ret = (list_empty(&lock->bast_list) && !lock->bast_pending);
2660         spin_unlock(&lock->spinlock);
2661         spin_unlock(&dlm->ast_lock);
2662         return ret;
2663 }
2664
2665 static int dlm_migration_can_proceed(struct dlm_ctxt *dlm,
2666                                      struct dlm_lock_resource *res,
2667                                      u8 mig_target)
2668 {
2669         int can_proceed;
2670         spin_lock(&res->spinlock);
2671         can_proceed = !!(res->state & DLM_LOCK_RES_MIGRATING);
2672         spin_unlock(&res->spinlock);
2673
2674         /* target has died, so make the caller break out of the
2675          * wait_event, but caller must recheck the domain_map */
2676         spin_lock(&dlm->spinlock);
2677         if (!test_bit(mig_target, dlm->domain_map))
2678                 can_proceed = 1;
2679         spin_unlock(&dlm->spinlock);
2680         return can_proceed;
2681 }
2682
2683 static int dlm_lockres_is_dirty(struct dlm_ctxt *dlm,
2684                                 struct dlm_lock_resource *res)
2685 {
2686         int ret;
2687         spin_lock(&res->spinlock);
2688         ret = !!(res->state & DLM_LOCK_RES_DIRTY);
2689         spin_unlock(&res->spinlock);
2690         return ret;
2691 }
2692
2693
2694 static int dlm_mark_lockres_migrating(struct dlm_ctxt *dlm,
2695                                        struct dlm_lock_resource *res,
2696                                        u8 target)
2697 {
2698         int ret = 0;
2699
2700         mlog(0, "dlm_mark_lockres_migrating: %.*s, from %u to %u\n",
2701                res->lockname.len, res->lockname.name, dlm->node_num,
2702                target);
2703         /* need to set MIGRATING flag on lockres.  this is done by
2704          * ensuring that all asts have been flushed for this lockres. */
2705         spin_lock(&res->spinlock);
2706         BUG_ON(res->migration_pending);
2707         res->migration_pending = 1;
2708         /* strategy is to reserve an extra ast then release
2709          * it below, letting the release do all of the work */
2710         __dlm_lockres_reserve_ast(res);
2711         spin_unlock(&res->spinlock);
2712
2713         /* now flush all the pending asts */
2714         dlm_kick_thread(dlm, res);
2715         /* before waiting on DIRTY, block processes which may
2716          * try to dirty the lockres before MIGRATING is set */
2717         spin_lock(&res->spinlock);
2718         BUG_ON(res->state & DLM_LOCK_RES_BLOCK_DIRTY);
2719         res->state |= DLM_LOCK_RES_BLOCK_DIRTY;
2720         spin_unlock(&res->spinlock);
2721         /* now wait on any pending asts and the DIRTY state */
2722         wait_event(dlm->ast_wq, !dlm_lockres_is_dirty(dlm, res));
2723         dlm_lockres_release_ast(dlm, res);
2724
2725         mlog(0, "about to wait on migration_wq, dirty=%s\n",
2726                res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
2727         /* if the extra ref we just put was the final one, this
2728          * will pass thru immediately.  otherwise, we need to wait
2729          * for the last ast to finish. */
2730 again:
2731         ret = wait_event_interruptible_timeout(dlm->migration_wq,
2732                    dlm_migration_can_proceed(dlm, res, target),
2733                    msecs_to_jiffies(1000));
2734         if (ret < 0) {
2735                 mlog(0, "woken again: migrating? %s, dead? %s\n",
2736                        res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
2737                        test_bit(target, dlm->domain_map) ? "no":"yes");
2738         } else {
2739                 mlog(0, "all is well: migrating? %s, dead? %s\n",
2740                        res->state & DLM_LOCK_RES_MIGRATING ? "yes":"no",
2741                        test_bit(target, dlm->domain_map) ? "no":"yes");
2742         }
2743         if (!dlm_migration_can_proceed(dlm, res, target)) {
2744                 mlog(0, "trying again...\n");
2745                 goto again;
2746         }
2747
2748         ret = 0;
2749         /* did the target go down or die? */
2750         spin_lock(&dlm->spinlock);
2751         if (!test_bit(target, dlm->domain_map)) {
2752                 mlog(ML_ERROR, "aha. migration target %u just went down\n",
2753                      target);
2754                 ret = -EHOSTDOWN;
2755         }
2756         spin_unlock(&dlm->spinlock);
2757
2758         /*
2759          * if target is down, we need to clear DLM_LOCK_RES_BLOCK_DIRTY for
2760          * another try; otherwise, we are sure the MIGRATING state is there,
2761          * drop the unneded state which blocked threads trying to DIRTY
2762          */
2763         spin_lock(&res->spinlock);
2764         BUG_ON(!(res->state & DLM_LOCK_RES_BLOCK_DIRTY));
2765         res->state &= ~DLM_LOCK_RES_BLOCK_DIRTY;
2766         if (!ret)
2767                 BUG_ON(!(res->state & DLM_LOCK_RES_MIGRATING));
2768         spin_unlock(&res->spinlock);
2769
2770         /*
2771          * at this point:
2772          *
2773          *   o the DLM_LOCK_RES_MIGRATING flag is set if target not down
2774          *   o there are no pending asts on this lockres
2775          *   o all processes trying to reserve an ast on this
2776          *     lockres must wait for the MIGRATING flag to clear
2777          */
2778         return ret;
2779 }
2780
2781 /* last step in the migration process.
2782  * original master calls this to free all of the dlm_lock
2783  * structures that used to be for other nodes. */
2784 static void dlm_remove_nonlocal_locks(struct dlm_ctxt *dlm,
2785                                       struct dlm_lock_resource *res)
2786 {
2787         struct list_head *queue = &res->granted;
2788         int i, bit;
2789         struct dlm_lock *lock, *next;
2790
2791         assert_spin_locked(&res->spinlock);
2792
2793         BUG_ON(res->owner == dlm->node_num);
2794
2795         for (i=0; i<3; i++) {
2796                 list_for_each_entry_safe(lock, next, queue, list) {
2797                         if (lock->ml.node != dlm->node_num) {
2798                                 mlog(0, "putting lock for node %u\n",
2799                                      lock->ml.node);
2800                                 /* be extra careful */
2801                                 BUG_ON(!list_empty(&lock->ast_list));
2802                                 BUG_ON(!list_empty(&lock->bast_list));
2803                                 BUG_ON(lock->ast_pending);
2804                                 BUG_ON(lock->bast_pending);
2805                                 dlm_lockres_clear_refmap_bit(dlm, res,
2806                                                              lock->ml.node);
2807                                 list_del_init(&lock->list);
2808                                 dlm_lock_put(lock);
2809                                 /* In a normal unlock, we would have added a
2810                                  * DLM_UNLOCK_FREE_LOCK action. Force it. */
2811                                 dlm_lock_put(lock);
2812                         }
2813                 }
2814                 queue++;
2815         }
2816         bit = 0;
2817         while (1) {
2818                 bit = find_next_bit(res->refmap, O2NM_MAX_NODES, bit);
2819                 if (bit >= O2NM_MAX_NODES)
2820                         break;
2821                 /* do not clear the local node reference, if there is a
2822                  * process holding this, let it drop the ref itself */
2823                 if (bit != dlm->node_num) {
2824                         mlog(0, "%s:%.*s: node %u had a ref to this "
2825                              "migrating lockres, clearing\n", dlm->name,
2826                              res->lockname.len, res->lockname.name, bit);
2827                         dlm_lockres_clear_refmap_bit(dlm, res, bit);
2828                 }
2829                 bit++;
2830         }
2831 }
2832
2833 /*
2834  * Pick a node to migrate the lock resource to. This function selects a
2835  * potential target based first on the locks and then on refmap. It skips
2836  * nodes that are in the process of exiting the domain.
2837  */
2838 static u8 dlm_pick_migration_target(struct dlm_ctxt *dlm,
2839                                     struct dlm_lock_resource *res)
2840 {
2841         enum dlm_lockres_list idx;
2842         struct list_head *queue = &res->granted;
2843         struct dlm_lock *lock;
2844         int noderef;
2845         u8 nodenum = O2NM_MAX_NODES;
2846
2847         assert_spin_locked(&dlm->spinlock);
2848         assert_spin_locked(&res->spinlock);
2849
2850         /* Go through all the locks */
2851         for (idx = DLM_GRANTED_LIST; idx <= DLM_BLOCKED_LIST; idx++) {
2852                 queue = dlm_list_idx_to_ptr(res, idx);
2853                 list_for_each_entry(lock, queue, list) {
2854                         if (lock->ml.node == dlm->node_num)
2855                                 continue;
2856                         if (test_bit(lock->ml.node, dlm->exit_domain_map))
2857                                 continue;
2858                         nodenum = lock->ml.node;
2859                         goto bail;
2860                 }
2861         }
2862
2863         /* Go thru the refmap */
2864         noderef = -1;
2865         while (1) {
2866                 noderef = find_next_bit(res->refmap, O2NM_MAX_NODES,
2867                                         noderef + 1);
2868                 if (noderef >= O2NM_MAX_NODES)
2869                         break;
2870                 if (noderef == dlm->node_num)
2871                         continue;
2872                 if (test_bit(noderef, dlm->exit_domain_map))
2873                         continue;
2874                 nodenum = noderef;
2875                 goto bail;
2876         }
2877
2878 bail:
2879         return nodenum;
2880 }
2881
2882 /* this is called by the new master once all lockres
2883  * data has been received */
2884 static int dlm_do_migrate_request(struct dlm_ctxt *dlm,
2885                                   struct dlm_lock_resource *res,
2886                                   u8 master, u8 new_master,
2887                                   struct dlm_node_iter *iter)
2888 {
2889         struct dlm_migrate_request migrate;
2890         int ret, skip, status = 0;
2891         int nodenum;
2892
2893         memset(&migrate, 0, sizeof(migrate));
2894         migrate.namelen = res->lockname.len;
2895         memcpy(migrate.name, res->lockname.name, migrate.namelen);
2896         migrate.new_master = new_master;
2897         migrate.master = master;
2898
2899         ret = 0;
2900
2901         /* send message to all nodes, except the master and myself */
2902         while ((nodenum = dlm_node_iter_next(iter)) >= 0) {
2903                 if (nodenum == master ||
2904                     nodenum == new_master)
2905                         continue;
2906
2907                 /* We could race exit domain. If exited, skip. */
2908                 spin_lock(&dlm->spinlock);
2909                 skip = (!test_bit(nodenum, dlm->domain_map));
2910                 spin_unlock(&dlm->spinlock);
2911                 if (skip) {
2912                         clear_bit(nodenum, iter->node_map);
2913                         continue;
2914                 }
2915
2916                 ret = o2net_send_message(DLM_MIGRATE_REQUEST_MSG, dlm->key,
2917                                          &migrate, sizeof(migrate), nodenum,
2918                                          &status);
2919                 if (ret < 0) {
2920                         mlog(ML_ERROR, "%s: res %.*s, Error %d send "
2921                              "MIGRATE_REQUEST to node %u\n", dlm->name,
2922                              migrate.namelen, migrate.name, ret, nodenum);
2923                         if (!dlm_is_host_down(ret)) {
2924                                 mlog(ML_ERROR, "unhandled error=%d!\n", ret);
2925                                 BUG();
2926                         }
2927                         clear_bit(nodenum, iter->node_map);
2928                         ret = 0;
2929                 } else if (status < 0) {
2930                         mlog(0, "migrate request (node %u) returned %d!\n",
2931                              nodenum, status);
2932                         ret = status;
2933                 } else if (status == DLM_MIGRATE_RESPONSE_MASTERY_REF) {
2934                         /* during the migration request we short-circuited
2935                          * the mastery of the lockres.  make sure we have
2936                          * a mastery ref for nodenum */
2937                         mlog(0, "%s:%.*s: need ref for node %u\n",
2938                              dlm->name, res->lockname.len, res->lockname.name,
2939                              nodenum);
2940                         spin_lock(&res->spinlock);
2941                         dlm_lockres_set_refmap_bit(dlm, res, nodenum);
2942                         spin_unlock(&res->spinlock);
2943                 }
2944         }
2945
2946         if (ret < 0)
2947                 mlog_errno(ret);
2948
2949         mlog(0, "returning ret=%d\n", ret);
2950         return ret;
2951 }
2952
2953
2954 /* if there is an existing mle for this lockres, we now know who the master is.
2955  * (the one who sent us *this* message) we can clear it up right away.
2956  * since the process that put the mle on the list still has a reference to it,
2957  * we can unhash it now, set the master and wake the process.  as a result,
2958  * we will have no mle in the list to start with.  now we can add an mle for
2959  * the migration and this should be the only one found for those scanning the
2960  * list.  */
2961 int dlm_migrate_request_handler(struct o2net_msg *msg, u32 len, void *data,
2962                                 void **ret_data)
2963 {
2964         struct dlm_ctxt *dlm = data;
2965         struct dlm_lock_resource *res = NULL;
2966         struct dlm_migrate_request *migrate = (struct dlm_migrate_request *) msg->buf;
2967         struct dlm_master_list_entry *mle = NULL, *oldmle = NULL;
2968         const char *name;
2969         unsigned int namelen, hash;
2970         int ret = 0;
2971
2972         if (!dlm_grab(dlm))
2973                 return -EINVAL;
2974
2975         name = migrate->name;
2976         namelen = migrate->namelen;
2977         hash = dlm_lockid_hash(name, namelen);
2978
2979         /* preallocate.. if this fails, abort */
2980         mle = kmem_cache_alloc(dlm_mle_cache, GFP_NOFS);
2981
2982         if (!mle) {
2983                 ret = -ENOMEM;
2984                 goto leave;
2985         }
2986
2987         /* check for pre-existing lock */
2988         spin_lock(&dlm->spinlock);
2989         res = __dlm_lookup_lockres(dlm, name, namelen, hash);
2990         if (res) {
2991                 spin_lock(&res->spinlock);
2992                 if (res->state & DLM_LOCK_RES_RECOVERING) {
2993                         /* if all is working ok, this can only mean that we got
2994                         * a migrate request from a node that we now see as
2995                         * dead.  what can we do here?  drop it to the floor? */
2996                         spin_unlock(&res->spinlock);
2997                         mlog(ML_ERROR, "Got a migrate request, but the "
2998                              "lockres is marked as recovering!");
2999                         kmem_cache_free(dlm_mle_cache, mle);
3000                         ret = -EINVAL; /* need a better solution */
3001                         goto unlock;
3002                 }
3003                 res->state |= DLM_LOCK_RES_MIGRATING;
3004                 spin_unlock(&res->spinlock);
3005         }
3006
3007         spin_lock(&dlm->master_lock);
3008         /* ignore status.  only nonzero status would BUG. */
3009         ret = dlm_add_migration_mle(dlm, res, mle, &oldmle,
3010                                     name, namelen,
3011                                     migrate->new_master,
3012                                     migrate->master);
3013
3014         spin_unlock(&dlm->master_lock);
3015 unlock:
3016         spin_unlock(&dlm->spinlock);
3017
3018         if (oldmle) {
3019                 /* master is known, detach if not already detached */
3020                 dlm_mle_detach_hb_events(dlm, oldmle);
3021                 dlm_put_mle(oldmle);
3022         }
3023
3024         if (res)
3025                 dlm_lockres_put(res);
3026 leave:
3027         dlm_put(dlm);
3028         return ret;
3029 }
3030
3031 /* must be holding dlm->spinlock and dlm->master_lock
3032  * when adding a migration mle, we can clear any other mles
3033  * in the master list because we know with certainty that
3034  * the master is "master".  so we remove any old mle from
3035  * the list after setting it's master field, and then add
3036  * the new migration mle.  this way we can hold with the rule
3037  * of having only one mle for a given lock name at all times. */
3038 static int dlm_add_migration_mle(struct dlm_ctxt *dlm,
3039                                  struct dlm_lock_resource *res,
3040                                  struct dlm_master_list_entry *mle,
3041                                  struct dlm_master_list_entry **oldmle,
3042                                  const char *name, unsigned int namelen,
3043                                  u8 new_master, u8 master)
3044 {
3045         int found;
3046         int ret = 0;
3047
3048         *oldmle = NULL;
3049
3050         assert_spin_locked(&dlm->spinlock);
3051         assert_spin_locked(&dlm->master_lock);
3052
3053         /* caller is responsible for any ref taken here on oldmle */
3054         found = dlm_find_mle(dlm, oldmle, (char *)name, namelen);
3055         if (found) {
3056                 struct dlm_master_list_entry *tmp = *oldmle;
3057                 spin_lock(&tmp->spinlock);
3058                 if (tmp->type == DLM_MLE_MIGRATION) {
3059                         if (master == dlm->node_num) {
3060                                 /* ah another process raced me to it */
3061                                 mlog(0, "tried to migrate %.*s, but some "
3062                                      "process beat me to it\n",
3063                                      namelen, name);
3064                                 ret = -EEXIST;
3065                         } else {
3066                                 /* bad.  2 NODES are trying to migrate! */
3067                                 mlog(ML_ERROR, "migration error  mle: "
3068                                      "master=%u new_master=%u // request: "
3069                                      "master=%u new_master=%u // "
3070                                      "lockres=%.*s\n",
3071                                      tmp->master, tmp->new_master,
3072                                      master, new_master,
3073                                      namelen, name);
3074                                 BUG();
3075                         }
3076                 } else {
3077                         /* this is essentially what assert_master does */
3078                         tmp->master = master;
3079                         atomic_set(&tmp->woken, 1);
3080                         wake_up(&tmp->wq);
3081                         /* remove it so that only one mle will be found */
3082                         __dlm_unlink_mle(dlm, tmp);
3083                         __dlm_mle_detach_hb_events(dlm, tmp);
3084                         ret = DLM_MIGRATE_RESPONSE_MASTERY_REF;
3085                         mlog(0, "%s:%.*s: master=%u, newmaster=%u, "
3086                             "telling master to get ref for cleared out mle "
3087                             "during migration\n", dlm->name, namelen, name,
3088                             master, new_master);
3089                 }
3090                 spin_unlock(&tmp->spinlock);
3091         }
3092
3093         /* now add a migration mle to the tail of the list */
3094         dlm_init_mle(mle, DLM_MLE_MIGRATION, dlm, res, name, namelen);
3095         mle->new_master = new_master;
3096         /* the new master will be sending an assert master for this.
3097          * at that point we will get the refmap reference */
3098         mle->master = master;
3099         /* do this for consistency with other mle types */
3100         set_bit(new_master, mle->maybe_map);
3101         __dlm_insert_mle(dlm, mle);
3102
3103         return ret;
3104 }
3105
3106 /*
3107  * Sets the owner of the lockres, associated to the mle, to UNKNOWN
3108  */
3109 static struct dlm_lock_resource *dlm_reset_mleres_owner(struct dlm_ctxt *dlm,
3110                                         struct dlm_master_list_entry *mle)
3111 {
3112         struct dlm_lock_resource *res;
3113
3114         /* Find the lockres associated to the mle and set its owner to UNK */
3115         res = __dlm_lookup_lockres(dlm, mle->mname, mle->mnamelen,
3116                                    mle->mnamehash);
3117         if (res) {
3118                 spin_unlock(&dlm->master_lock);
3119
3120                 /* move lockres onto recovery list */
3121                 spin_lock(&res->spinlock);
3122                 dlm_set_lockres_owner(dlm, res, DLM_LOCK_RES_OWNER_UNKNOWN);
3123                 dlm_move_lockres_to_recovery_list(dlm, res);
3124                 spin_unlock(&res->spinlock);
3125                 dlm_lockres_put(res);
3126
3127                 /* about to get rid of mle, detach from heartbeat */
3128                 __dlm_mle_detach_hb_events(dlm, mle);
3129
3130                 /* dump the mle */
3131                 spin_lock(&dlm->master_lock);
3132                 __dlm_put_mle(mle);
3133                 spin_unlock(&dlm->master_lock);
3134         }
3135
3136         return res;
3137 }
3138
3139 static void dlm_clean_migration_mle(struct dlm_ctxt *dlm,
3140                                     struct dlm_master_list_entry *mle)
3141 {
3142         __dlm_mle_detach_hb_events(dlm, mle);
3143
3144         spin_lock(&mle->spinlock);
3145         __dlm_unlink_mle(dlm, mle);
3146         atomic_set(&mle->woken, 1);
3147         spin_unlock(&mle->spinlock);
3148
3149         wake_up(&mle->wq);
3150 }
3151
3152 static void dlm_clean_block_mle(struct dlm_ctxt *dlm,
3153                                 struct dlm_master_list_entry *mle, u8 dead_node)
3154 {
3155         int bit;
3156
3157         BUG_ON(mle->type != DLM_MLE_BLOCK);
3158
3159         spin_lock(&mle->spinlock);
3160         bit = find_next_bit(mle->maybe_map, O2NM_MAX_NODES, 0);
3161         if (bit != dead_node) {
3162                 mlog(0, "mle found, but dead node %u would not have been "
3163                      "master\n", dead_node);
3164                 spin_unlock(&mle->spinlock);
3165         } else {
3166                 /* Must drop the refcount by one since the assert_master will
3167                  * never arrive. This may result in the mle being unlinked and
3168                  * freed, but there may still be a process waiting in the
3169                  * dlmlock path which is fine. */
3170                 mlog(0, "node %u was expected master\n", dead_node);
3171                 atomic_set(&mle->woken, 1);
3172                 spin_unlock(&mle->spinlock);
3173                 wake_up(&mle->wq);
3174
3175                 /* Do not need events any longer, so detach from heartbeat */
3176                 __dlm_mle_detach_hb_events(dlm, mle);
3177                 __dlm_put_mle(mle);
3178         }
3179 }
3180
3181 void dlm_clean_master_list(struct dlm_ctxt *dlm, u8 dead_node)
3182 {
3183         struct dlm_master_list_entry *mle;
3184         struct dlm_lock_resource *res;
3185         struct hlist_head *bucket;
3186         struct hlist_node *list;
3187         unsigned int i;
3188
3189         mlog(0, "dlm=%s, dead node=%u\n", dlm->name, dead_node);
3190 top:
3191         assert_spin_locked(&dlm->spinlock);
3192
3193         /* clean the master list */
3194         spin_lock(&dlm->master_lock);
3195         for (i = 0; i < DLM_HASH_BUCKETS; i++) {
3196                 bucket = dlm_master_hash(dlm, i);
3197                 hlist_for_each(list, bucket) {
3198                         mle = hlist_entry(list, struct dlm_master_list_entry,
3199                                           master_hash_node);
3200
3201                         BUG_ON(mle->type != DLM_MLE_BLOCK &&
3202                                mle->type != DLM_MLE_MASTER &&
3203                                mle->type != DLM_MLE_MIGRATION);
3204
3205                         /* MASTER mles are initiated locally. The waiting
3206                          * process will notice the node map change shortly.
3207                          * Let that happen as normal. */
3208                         if (mle->type == DLM_MLE_MASTER)
3209                                 continue;
3210
3211                         /* BLOCK mles are initiated by other nodes. Need to
3212                          * clean up if the dead node would have been the
3213                          * master. */
3214                         if (mle->type == DLM_MLE_BLOCK) {
3215                                 dlm_clean_block_mle(dlm, mle, dead_node);
3216                                 continue;
3217                         }
3218
3219                         /* Everything else is a MIGRATION mle */
3220
3221                         /* The rule for MIGRATION mles is that the master
3222                          * becomes UNKNOWN if *either* the original or the new
3223                          * master dies. All UNKNOWN lockres' are sent to
3224                          * whichever node becomes the recovery master. The new
3225                          * master is responsible for determining if there is
3226                          * still a master for this lockres, or if he needs to
3227                          * take over mastery. Either way, this node should
3228                          * expect another message to resolve this. */
3229
3230                         if (mle->master != dead_node &&
3231                             mle->new_master != dead_node)
3232                                 continue;
3233
3234                         /* If we have reached this point, this mle needs to be
3235                          * removed from the list and freed. */
3236                         dlm_clean_migration_mle(dlm, mle);
3237
3238                         mlog(0, "%s: node %u died during migration from "
3239                              "%u to %u!\n", dlm->name, dead_node, mle->master,
3240                              mle->new_master);
3241
3242                         /* If we find a lockres associated with the mle, we've
3243                          * hit this rare case that messes up our lock ordering.
3244                          * If so, we need to drop the master lock so that we can
3245                          * take the lockres lock, meaning that we will have to
3246                          * restart from the head of list. */
3247                         res = dlm_reset_mleres_owner(dlm, mle);
3248                         if (res)
3249                                 /* restart */
3250                                 goto top;
3251
3252                         /* This may be the last reference */
3253                         __dlm_put_mle(mle);
3254                 }
3255         }
3256         spin_unlock(&dlm->master_lock);
3257 }
3258
3259 int dlm_finish_migration(struct dlm_ctxt *dlm, struct dlm_lock_resource *res,
3260                          u8 old_master)
3261 {
3262         struct dlm_node_iter iter;
3263         int ret = 0;
3264
3265         spin_lock(&dlm->spinlock);
3266         dlm_node_iter_init(dlm->domain_map, &iter);
3267         clear_bit(old_master, iter.node_map);
3268         clear_bit(dlm->node_num, iter.node_map);
3269         spin_unlock(&dlm->spinlock);
3270
3271         /* ownership of the lockres is changing.  account for the
3272          * mastery reference here since old_master will briefly have
3273          * a reference after the migration completes */
3274         spin_lock(&res->spinlock);
3275         dlm_lockres_set_refmap_bit(dlm, res, old_master);
3276         spin_unlock(&res->spinlock);
3277
3278         mlog(0, "now time to do a migrate request to other nodes\n");
3279         ret = dlm_do_migrate_request(dlm, res, old_master,
3280                                      dlm->node_num, &iter);
3281         if (ret < 0) {
3282                 mlog_errno(ret);
3283                 goto leave;
3284         }
3285
3286         mlog(0, "doing assert master of %.*s to all except the original node\n",
3287              res->lockname.len, res->lockname.name);
3288         /* this call now finishes out the nodemap
3289          * even if one or more nodes die */
3290         ret = dlm_do_assert_master(dlm, res, iter.node_map,
3291                                    DLM_ASSERT_MASTER_FINISH_MIGRATION);
3292         if (ret < 0) {
3293                 /* no longer need to retry.  all living nodes contacted. */
3294                 mlog_errno(ret);
3295                 ret = 0;
3296         }
3297
3298         memset(iter.node_map, 0, sizeof(iter.node_map));
3299         set_bit(old_master, iter.node_map);
3300         mlog(0, "doing assert master of %.*s back to %u\n",
3301              res->lockname.len, res->lockname.name, old_master);
3302         ret = dlm_do_assert_master(dlm, res, iter.node_map,
3303                                    DLM_ASSERT_MASTER_FINISH_MIGRATION);
3304         if (ret < 0) {
3305                 mlog(0, "assert master to original master failed "
3306                      "with %d.\n", ret);
3307                 /* the only nonzero status here would be because of
3308                  * a dead original node.  we're done. */
3309                 ret = 0;
3310         }
3311
3312         /* all done, set the owner, clear the flag */
3313         spin_lock(&res->spinlock);
3314         dlm_set_lockres_owner(dlm, res, dlm->node_num);
3315         res->state &= ~DLM_LOCK_RES_MIGRATING;
3316         spin_unlock(&res->spinlock);
3317         /* re-dirty it on the new master */
3318         dlm_kick_thread(dlm, res);
3319         wake_up(&res->wq);
3320 leave:
3321         return ret;
3322 }
3323
3324 /*
3325  * LOCKRES AST REFCOUNT
3326  * this is integral to migration
3327  */
3328
3329 /* for future intent to call an ast, reserve one ahead of time.
3330  * this should be called only after waiting on the lockres
3331  * with dlm_wait_on_lockres, and while still holding the
3332  * spinlock after the call. */
3333 void __dlm_lockres_reserve_ast(struct dlm_lock_resource *res)
3334 {
3335         assert_spin_locked(&res->spinlock);
3336         if (res->state & DLM_LOCK_RES_MIGRATING) {
3337                 __dlm_print_one_lock_resource(res);
3338         }
3339         BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
3340
3341         atomic_inc(&res->asts_reserved);
3342 }
3343
3344 /*
3345  * used to drop the reserved ast, either because it went unused,
3346  * or because the ast/bast was actually called.
3347  *
3348  * also, if there is a pending migration on this lockres,
3349  * and this was the last pending ast on the lockres,
3350  * atomically set the MIGRATING flag before we drop the lock.
3351  * this is how we ensure that migration can proceed with no
3352  * asts in progress.  note that it is ok if the state of the
3353  * queues is such that a lock should be granted in the future
3354  * or that a bast should be fired, because the new master will
3355  * shuffle the lists on this lockres as soon as it is migrated.
3356  */
3357 void dlm_lockres_release_ast(struct dlm_ctxt *dlm,
3358                              struct dlm_lock_resource *res)
3359 {
3360         if (!atomic_dec_and_lock(&res->asts_reserved, &res->spinlock))
3361                 return;
3362
3363         if (!res->migration_pending) {
3364                 spin_unlock(&res->spinlock);
3365                 return;
3366         }
3367
3368         BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
3369         res->migration_pending = 0;
3370         res->state |= DLM_LOCK_RES_MIGRATING;
3371         spin_unlock(&res->spinlock);
3372         wake_up(&res->wq);
3373         wake_up(&dlm->migration_wq);
3374 }
3375
3376 void dlm_force_free_mles(struct dlm_ctxt *dlm)
3377 {
3378         int i;
3379         struct hlist_head *bucket;
3380         struct dlm_master_list_entry *mle;
3381         struct hlist_node *tmp, *list;
3382
3383         /*
3384          * We notified all other nodes that we are exiting the domain and
3385          * marked the dlm state to DLM_CTXT_LEAVING. If any mles are still
3386          * around we force free them and wake any processes that are waiting
3387          * on the mles
3388          */
3389         spin_lock(&dlm->spinlock);
3390         spin_lock(&dlm->master_lock);
3391
3392         BUG_ON(dlm->dlm_state != DLM_CTXT_LEAVING);
3393         BUG_ON((find_next_bit(dlm->domain_map, O2NM_MAX_NODES, 0) < O2NM_MAX_NODES));
3394
3395         for (i = 0; i < DLM_HASH_BUCKETS; i++) {
3396                 bucket = dlm_master_hash(dlm, i);
3397                 hlist_for_each_safe(list, tmp, bucket) {
3398                         mle = hlist_entry(list, struct dlm_master_list_entry,
3399                                           master_hash_node);
3400                         if (mle->type != DLM_MLE_BLOCK) {
3401                                 mlog(ML_ERROR, "bad mle: %p\n", mle);
3402                                 dlm_print_one_mle(mle);
3403                         }
3404                         atomic_set(&mle->woken, 1);
3405                         wake_up(&mle->wq);
3406
3407                         __dlm_unlink_mle(dlm, mle);
3408                         __dlm_mle_detach_hb_events(dlm, mle);
3409                         __dlm_put_mle(mle);
3410                 }
3411         }
3412         spin_unlock(&dlm->master_lock);
3413         spin_unlock(&dlm->spinlock);
3414 }