fs/ocfs2/dlm/dlmthread.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * dlmthread.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/timer.h>
  40 #include <linux/kthread.h>
  41 #include <linux/delay.h>
  42
  43
  44 #include "cluster/heartbeat.h"
  45 #include "cluster/nodemanager.h"
  46 #include "cluster/tcp.h"
  47
  48 #include "dlmapi.h"
  49 #include "dlmcommon.h"
  50 #include "dlmdomain.h"
  51
  52 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_THREAD)
  53 #include "cluster/masklog.h"
  54
  55 static int dlm_thread(void *data);
  56 static void dlm_flush_asts(struct dlm_ctxt *dlm);
  57
  58 #define dlm_lock_is_remote(dlm, lock)     ((lock)->ml.node != (dlm)->node_num)
  59
  60 /* will exit holding res->spinlock, but may drop in function */
  61 /* waits until flags are cleared on res->state */
  62 void __dlm_wait_on_lockres_flags(struct dlm_lock_resource *res, int flags)
  63 {
  64         DECLARE_WAITQUEUE(wait, current);
  65
  66         assert_spin_locked(&res->spinlock);
  67
  68         add_wait_queue(&res->wq, &wait);
  69 repeat:
  70         set_current_state(TASK_UNINTERRUPTIBLE);
  71         if (res->state & flags) {
  72                 spin_unlock(&res->spinlock);
  73                 schedule();
  74                 spin_lock(&res->spinlock);
  75                 goto repeat;
  76         }
  77         remove_wait_queue(&res->wq, &wait);
  78         __set_current_state(TASK_RUNNING);
  79 }
  80
  81 int __dlm_lockres_has_locks(struct dlm_lock_resource *res)
  82 {
  83         if (list_empty(&res->granted) &&
  84             list_empty(&res->converting) &&
  85             list_empty(&res->blocked))
  86                 return 0;
  87         return 1;
  88 }
  89
  90 /* "unused": the lockres has no locks, is not on the dirty list,
  91  * has no inflight locks (in the gap between mastery and acquiring
  92  * the first lock), and has no bits in its refmap.
  93  * truly ready to be freed. */
  94 int __dlm_lockres_unused(struct dlm_lock_resource *res)
  95 {
  96         int bit;
  97
  98         if (__dlm_lockres_has_locks(res))
  99                 return 0;
 100
 101         if (!list_empty(&res->dirty) || res->state & DLM_LOCK_RES_DIRTY)
 102                 return 0;
 103
 104         if (res->state & DLM_LOCK_RES_RECOVERING)
 105                 return 0;
 106
 107         bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
 108         if (bit < O2NM_MAX_NODES)
 109                 return 0;
 110
 111         /*
 112          * since the bit for dlm->node_num is not set, inflight_locks better
 113          * be zero
 114          */
 115         BUG_ON(res->inflight_locks != 0);
 116         return 1;
 117 }
 118
 119
 120 /* Call whenever you may have added or deleted something from one of
 121  * the lockres queue's. This will figure out whether it belongs on the
 122  * unused list or not and does the appropriate thing. */
 123 void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 124                               struct dlm_lock_resource *res)
 125 {
 126         mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
 127
 128         assert_spin_locked(&dlm->spinlock);
 129         assert_spin_locked(&res->spinlock);
 130
 131         if (__dlm_lockres_unused(res)){
 132                 if (list_empty(&res->purge)) {
 133                         mlog(0, "putting lockres %.*s:%p onto purge list\n",
 134                              res->lockname.len, res->lockname.name, res);
 135
 136                         res->last_used = jiffies;
 137                         dlm_lockres_get(res);
 138                         list_add_tail(&res->purge, &dlm->purge_list);
 139                         dlm->purge_count++;
 140                 }
 141         } else if (!list_empty(&res->purge)) {
 142                 mlog(0, "removing lockres %.*s:%p from purge list, owner=%u\n",
 143                      res->lockname.len, res->lockname.name, res, res->owner);
 144
 145                 list_del_init(&res->purge);
 146                 dlm_lockres_put(res);
 147                 dlm->purge_count--;
 148         }
 149 }
 150
 151 void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 152                             struct dlm_lock_resource *res)
 153 {
 154         mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
 155         spin_lock(&dlm->spinlock);
 156         spin_lock(&res->spinlock);
 157
 158         __dlm_lockres_calc_usage(dlm, res);
 159
 160         spin_unlock(&res->spinlock);
 161         spin_unlock(&dlm->spinlock);
 162 }
 163
 164 static void dlm_purge_lockres(struct dlm_ctxt *dlm,
 165                              struct dlm_lock_resource *res)
 166 {
 167         int master;
 168         int ret = 0;
 169
 170         assert_spin_locked(&dlm->spinlock);
 171         assert_spin_locked(&res->spinlock);
 172
 173         master = (res->owner == dlm->node_num);
 174
 175
 176         mlog(0, "purging lockres %.*s, master = %d\n", res->lockname.len,
 177              res->lockname.name, master);
 178
 179         if (!master) {
 180                 res->state |= DLM_LOCK_RES_DROPPING_REF;
 181                 /* drop spinlock...  retake below */
 182                 spin_unlock(&res->spinlock);
 183                 spin_unlock(&dlm->spinlock);
 184
 185                 spin_lock(&res->spinlock);
 186                 /* This ensures that clear refmap is sent after the set */
 187                 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
 188                 spin_unlock(&res->spinlock);
 189
 190                 /* clear our bit from the master's refmap, ignore errors */
 191                 ret = dlm_drop_lockres_ref(dlm, res);
 192                 if (ret < 0) {
 193                         mlog_errno(ret);
 194                         if (!dlm_is_host_down(ret))
 195                                 BUG();
 196                 }
 197                 mlog(0, "%s:%.*s: dlm_deref_lockres returned %d\n",
 198                      dlm->name, res->lockname.len, res->lockname.name, ret);
 199                 spin_lock(&dlm->spinlock);
 200                 spin_lock(&res->spinlock);
 201         }
 202
 203         if (!list_empty(&res->purge)) {
 204                 mlog(0, "removing lockres %.*s:%p from purgelist, "
 205                      "master = %d\n", res->lockname.len, res->lockname.name,
 206                      res, master);
 207                 list_del_init(&res->purge);
 208                 dlm_lockres_put(res);
 209                 dlm->purge_count--;
 210         }
 211
 212         if (!__dlm_lockres_unused(res)) {
 213                 mlog(ML_ERROR, "found lockres %s:%.*s: in use after deref\n",
 214                      dlm->name, res->lockname.len, res->lockname.name);
 215                 __dlm_print_one_lock_resource(res);
 216                 BUG();
 217         }
 218
 219         __dlm_unhash_lockres(res);
 220
 221         /* lockres is not in the hash now.  drop the flag and wake up
 222          * any processes waiting in dlm_get_lock_resource. */
 223         if (!master) {
 224                 res->state &= ~DLM_LOCK_RES_DROPPING_REF;
 225                 spin_unlock(&res->spinlock);
 226                 wake_up(&res->wq);
 227         } else
 228                 spin_unlock(&res->spinlock);
 229 }
 230
 231 static void dlm_run_purge_list(struct dlm_ctxt *dlm,
 232                                int purge_now)
 233 {
 234         unsigned int run_max, unused;
 235         unsigned long purge_jiffies;
 236         struct dlm_lock_resource *lockres;
 237
 238         spin_lock(&dlm->spinlock);
 239         run_max = dlm->purge_count;
 240
 241         while(run_max && !list_empty(&dlm->purge_list)) {
 242                 run_max--;
 243
 244                 lockres = list_entry(dlm->purge_list.next,
 245                                      struct dlm_lock_resource, purge);
 246
 247                 spin_lock(&lockres->spinlock);
 248
 249                 purge_jiffies = lockres->last_used +
 250                         msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);
 251
 252                 /* Make sure that we want to be processing this guy at
 253                  * this time. */
 254                 if (!purge_now && time_after(purge_jiffies, jiffies)) {
 255                         /* Since resources are added to the purge list
 256                          * in tail order, we can stop at the first
 257                          * unpurgable resource -- anyone added after
 258                          * him will have a greater last_used value */
 259                         spin_unlock(&lockres->spinlock);
 260                         break;
 261                 }
 262
 263                 /* Status of the lockres *might* change so double
 264                  * check. If the lockres is unused, holding the dlm
 265                  * spinlock will prevent people from getting and more
 266                  * refs on it. */
 267                 unused = __dlm_lockres_unused(lockres);
 268                 if (!unused ||
 269                     (lockres->state & DLM_LOCK_RES_MIGRATING)) {
 270                         mlog(0, "lockres %s:%.*s: is in use or "
 271                              "being remastered, used %d, state %d\n",
 272                              dlm->name, lockres->lockname.len,
 273                              lockres->lockname.name, !unused, lockres->state);
 274                         list_move_tail(&dlm->purge_list, &lockres->purge);
 275                         spin_unlock(&lockres->spinlock);
 276                         continue;
 277                 }
 278
 279                 dlm_lockres_get(lockres);
 280
 281                 dlm_purge_lockres(dlm, lockres);
 282
 283                 dlm_lockres_put(lockres);
 284
 285                 /* Avoid adding any scheduling latencies */
 286                 cond_resched_lock(&dlm->spinlock);
 287         }
 288
 289         spin_unlock(&dlm->spinlock);
 290 }
 291
 292 static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
 293                               struct dlm_lock_resource *res)
 294 {
 295         struct dlm_lock *lock, *target;
 296         struct list_head *iter;
 297         struct list_head *head;
 298         int can_grant = 1;
 299
 300         //mlog(0, "res->lockname.len=%d\n", res->lockname.len);
 301         //mlog(0, "res->lockname.name=%p\n", res->lockname.name);
 302         //mlog(0, "shuffle res %.*s\n", res->lockname.len,
 303         //        res->lockname.name);
 304
 305         /* because this function is called with the lockres
 306          * spinlock, and because we know that it is not migrating/
 307          * recovering/in-progress, it is fine to reserve asts and
 308          * basts right before queueing them all throughout */
 309         assert_spin_locked(&res->spinlock);
 310         BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
 311                               DLM_LOCK_RES_RECOVERING|
 312                               DLM_LOCK_RES_IN_PROGRESS)));
 313
 314 converting:
 315         if (list_empty(&res->converting))
 316                 goto blocked;
 317         mlog(0, "res %.*s has locks on a convert queue\n", res->lockname.len,
 318              res->lockname.name);
 319
 320         target = list_entry(res->converting.next, struct dlm_lock, list);
 321         if (target->ml.convert_type == LKM_IVMODE) {
 322                 mlog(ML_ERROR, "%.*s: converting a lock with no "
 323                      "convert_type!\n", res->lockname.len, res->lockname.name);
 324                 BUG();
 325         }
 326         head = &res->granted;
 327         list_for_each(iter, head) {
 328                 lock = list_entry(iter, struct dlm_lock, list);
 329                 if (lock==target)
 330                         continue;
 331                 if (!dlm_lock_compatible(lock->ml.type,
 332                                          target->ml.convert_type)) {
 333                         can_grant = 0;
 334                         /* queue the BAST if not already */
 335                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 336                                 __dlm_lockres_reserve_ast(res);
 337                                 dlm_queue_bast(dlm, lock);
 338                         }
 339                         /* update the highest_blocked if needed */
 340                         if (lock->ml.highest_blocked < target->ml.convert_type)
 341                                 lock->ml.highest_blocked =
 342                                         target->ml.convert_type;
 343                 }
 344         }
 345         head = &res->converting;
 346         list_for_each(iter, head) {
 347                 lock = list_entry(iter, struct dlm_lock, list);
 348                 if (lock==target)
 349                         continue;
 350                 if (!dlm_lock_compatible(lock->ml.type,
 351                                          target->ml.convert_type)) {
 352                         can_grant = 0;
 353                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 354                                 __dlm_lockres_reserve_ast(res);
 355                                 dlm_queue_bast(dlm, lock);
 356                         }
 357                         if (lock->ml.highest_blocked < target->ml.convert_type)
 358                                 lock->ml.highest_blocked =
 359                                         target->ml.convert_type;
 360                 }
 361         }
 362
 363         /* we can convert the lock */
 364         if (can_grant) {
 365                 spin_lock(&target->spinlock);
 366                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 367
 368                 mlog(0, "calling ast for converting lock: %.*s, have: %d, "
 369                      "granting: %d, node: %u\n", res->lockname.len,
 370                      res->lockname.name, target->ml.type,
 371                      target->ml.convert_type, target->ml.node);
 372
 373                 target->ml.type = target->ml.convert_type;
 374                 target->ml.convert_type = LKM_IVMODE;
 375                 list_move_tail(&target->list, &res->granted);
 376
 377                 BUG_ON(!target->lksb);
 378                 target->lksb->status = DLM_NORMAL;
 379
 380                 spin_unlock(&target->spinlock);
 381
 382                 __dlm_lockres_reserve_ast(res);
 383                 dlm_queue_ast(dlm, target);
 384                 /* go back and check for more */
 385                 goto converting;
 386         }
 387
 388 blocked:
 389         if (list_empty(&res->blocked))
 390                 goto leave;
 391         target = list_entry(res->blocked.next, struct dlm_lock, list);
 392
 393         head = &res->granted;
 394         list_for_each(iter, head) {
 395                 lock = list_entry(iter, struct dlm_lock, list);
 396                 if (lock==target)
 397                         continue;
 398                 if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
 399                         can_grant = 0;
 400                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 401                                 __dlm_lockres_reserve_ast(res);
 402                                 dlm_queue_bast(dlm, lock);
 403                         }
 404                         if (lock->ml.highest_blocked < target->ml.type)
 405                                 lock->ml.highest_blocked = target->ml.type;
 406                 }
 407         }
 408
 409         head = &res->converting;
 410         list_for_each(iter, head) {
 411                 lock = list_entry(iter, struct dlm_lock, list);
 412                 if (lock==target)
 413                         continue;
 414                 if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
 415                         can_grant = 0;
 416                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 417                                 __dlm_lockres_reserve_ast(res);
 418                                 dlm_queue_bast(dlm, lock);
 419                         }
 420                         if (lock->ml.highest_blocked < target->ml.type)
 421                                 lock->ml.highest_blocked = target->ml.type;
 422                 }
 423         }
 424
 425         /* we can grant the blocked lock (only
 426          * possible if converting list empty) */
 427         if (can_grant) {
 428                 spin_lock(&target->spinlock);
 429                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 430
 431                 mlog(0, "calling ast for blocked lock: %.*s, granting: %d, "
 432                      "node: %u\n", res->lockname.len, res->lockname.name,
 433                      target->ml.type, target->ml.node);
 434
 435                 // target->ml.type is already correct
 436                 list_move_tail(&target->list, &res->granted);
 437
 438                 BUG_ON(!target->lksb);
 439                 target->lksb->status = DLM_NORMAL;
 440
 441                 spin_unlock(&target->spinlock);
 442
 443                 __dlm_lockres_reserve_ast(res);
 444                 dlm_queue_ast(dlm, target);
 445                 /* go back and check for more */
 446                 goto converting;
 447         }
 448
 449 leave:
 450         return;
 451 }
 452
 453 /* must have NO locks when calling this with res !=NULL * */
 454 void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 455 {
 456         mlog_entry("dlm=%p, res=%p\n", dlm, res);
 457         if (res) {
 458                 spin_lock(&dlm->spinlock);
 459                 spin_lock(&res->spinlock);
 460                 __dlm_dirty_lockres(dlm, res);
 461                 spin_unlock(&res->spinlock);
 462                 spin_unlock(&dlm->spinlock);
 463         }
 464         wake_up(&dlm->dlm_thread_wq);
 465 }
 466
 467 void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 468 {
 469         mlog_entry("dlm=%p, res=%p\n", dlm, res);
 470
 471         assert_spin_locked(&dlm->spinlock);
 472         assert_spin_locked(&res->spinlock);
 473
 474         /* don't shuffle secondary queues */
 475         if ((res->owner == dlm->node_num)) {
 476                 if (res->state & (DLM_LOCK_RES_MIGRATING |
 477                                   DLM_LOCK_RES_BLOCK_DIRTY))
 478                     return;
 479
 480                 if (list_empty(&res->dirty)) {
 481                         /* ref for dirty_list */
 482                         dlm_lockres_get(res);
 483                         list_add_tail(&res->dirty, &dlm->dirty_list);
 484                         res->state |= DLM_LOCK_RES_DIRTY;
 485                 }
 486         }
 487 }
 488
 489
 490 /* Launch the NM thread for the mounted volume */
 491 int dlm_launch_thread(struct dlm_ctxt *dlm)
 492 {
 493         mlog(0, "starting dlm thread...\n");
 494
 495         dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm_thread");
 496         if (IS_ERR(dlm->dlm_thread_task)) {
 497                 mlog_errno(PTR_ERR(dlm->dlm_thread_task));
 498                 dlm->dlm_thread_task = NULL;
 499                 return -EINVAL;
 500         }
 501
 502         return 0;
 503 }
 504
 505 void dlm_complete_thread(struct dlm_ctxt *dlm)
 506 {
 507         if (dlm->dlm_thread_task) {
 508                 mlog(ML_KTHREAD, "waiting for dlm thread to exit\n");
 509                 kthread_stop(dlm->dlm_thread_task);
 510                 dlm->dlm_thread_task = NULL;
 511         }
 512 }
 513
 514 static int dlm_dirty_list_empty(struct dlm_ctxt *dlm)
 515 {
 516         int empty;
 517
 518         spin_lock(&dlm->spinlock);
 519         empty = list_empty(&dlm->dirty_list);
 520         spin_unlock(&dlm->spinlock);
 521
 522         return empty;
 523 }
 524
 525 static void dlm_flush_asts(struct dlm_ctxt *dlm)
 526 {
 527         int ret;
 528         struct dlm_lock *lock;
 529         struct dlm_lock_resource *res;
 530         u8 hi;
 531
 532         spin_lock(&dlm->ast_lock);
 533         while (!list_empty(&dlm->pending_asts)) {
 534                 lock = list_entry(dlm->pending_asts.next,
 535                                   struct dlm_lock, ast_list);
 536                 /* get an extra ref on lock */
 537                 dlm_lock_get(lock);
 538                 res = lock->lockres;
 539                 mlog(0, "delivering an ast for this lockres\n");
 540
 541                 BUG_ON(!lock->ast_pending);
 542
 543                 /* remove from list (including ref) */
 544                 list_del_init(&lock->ast_list);
 545                 dlm_lock_put(lock);
 546                 spin_unlock(&dlm->ast_lock);
 547
 548                 if (lock->ml.node != dlm->node_num) {
 549                         ret = dlm_do_remote_ast(dlm, res, lock);
 550                         if (ret < 0)
 551                                 mlog_errno(ret);
 552                 } else
 553                         dlm_do_local_ast(dlm, res, lock);
 554
 555                 spin_lock(&dlm->ast_lock);
 556
 557                 /* possible that another ast was queued while
 558                  * we were delivering the last one */
 559                 if (!list_empty(&lock->ast_list)) {
 560                         mlog(0, "aha another ast got queued while "
 561                              "we were finishing the last one.  will "
 562                              "keep the ast_pending flag set.\n");
 563                 } else
 564                         lock->ast_pending = 0;
 565
 566                 /* drop the extra ref.
 567                  * this may drop it completely. */
 568                 dlm_lock_put(lock);
 569                 dlm_lockres_release_ast(dlm, res);
 570         }
 571
 572         while (!list_empty(&dlm->pending_basts)) {
 573                 lock = list_entry(dlm->pending_basts.next,
 574                                   struct dlm_lock, bast_list);
 575                 /* get an extra ref on lock */
 576                 dlm_lock_get(lock);
 577                 res = lock->lockres;
 578
 579                 BUG_ON(!lock->bast_pending);
 580
 581                 /* get the highest blocked lock, and reset */
 582                 spin_lock(&lock->spinlock);
 583                 BUG_ON(lock->ml.highest_blocked <= LKM_IVMODE);
 584                 hi = lock->ml.highest_blocked;
 585                 lock->ml.highest_blocked = LKM_IVMODE;
 586                 spin_unlock(&lock->spinlock);
 587
 588                 /* remove from list (including ref) */
 589                 list_del_init(&lock->bast_list);
 590                 dlm_lock_put(lock);
 591                 spin_unlock(&dlm->ast_lock);
 592
 593                 mlog(0, "delivering a bast for this lockres "
 594                      "(blocked = %d\n", hi);
 595
 596                 if (lock->ml.node != dlm->node_num) {
 597                         ret = dlm_send_proxy_bast(dlm, res, lock, hi);
 598                         if (ret < 0)
 599                                 mlog_errno(ret);
 600                 } else
 601                         dlm_do_local_bast(dlm, res, lock, hi);
 602
 603                 spin_lock(&dlm->ast_lock);
 604
 605                 /* possible that another bast was queued while
 606                  * we were delivering the last one */
 607                 if (!list_empty(&lock->bast_list)) {
 608                         mlog(0, "aha another bast got queued while "
 609                              "we were finishing the last one.  will "
 610                              "keep the bast_pending flag set.\n");
 611                 } else
 612                         lock->bast_pending = 0;
 613
 614                 /* drop the extra ref.
 615                  * this may drop it completely. */
 616                 dlm_lock_put(lock);
 617                 dlm_lockres_release_ast(dlm, res);
 618         }
 619         wake_up(&dlm->ast_wq);
 620         spin_unlock(&dlm->ast_lock);
 621 }
 622
 623
 624 #define DLM_THREAD_TIMEOUT_MS (4 * 1000)
 625 #define DLM_THREAD_MAX_DIRTY  100
 626 #define DLM_THREAD_MAX_ASTS   10
 627
 628 static int dlm_thread(void *data)
 629 {
 630         struct dlm_lock_resource *res;
 631         struct dlm_ctxt *dlm = data;
 632         unsigned long timeout = msecs_to_jiffies(DLM_THREAD_TIMEOUT_MS);
 633
 634         mlog(0, "dlm thread running for %s...\n", dlm->name);
 635
 636         while (!kthread_should_stop()) {
 637                 int n = DLM_THREAD_MAX_DIRTY;
 638
 639                 /* dlm_shutting_down is very point-in-time, but that
 640                  * doesn't matter as we'll just loop back around if we
 641                  * get false on the leading edge of a state
 642                  * transition. */
 643                 dlm_run_purge_list(dlm, dlm_shutting_down(dlm));
 644
 645                 /* We really don't want to hold dlm->spinlock while
 646                  * calling dlm_shuffle_lists on each lockres that
 647                  * needs to have its queues adjusted and AST/BASTs
 648                  * run.  So let's pull each entry off the dirty_list
 649                  * and drop dlm->spinlock ASAP.  Once off the list,
 650                  * res->spinlock needs to be taken again to protect
 651                  * the queues while calling dlm_shuffle_lists.  */
 652                 spin_lock(&dlm->spinlock);
 653                 while (!list_empty(&dlm->dirty_list)) {
 654                         int delay = 0;
 655                         res = list_entry(dlm->dirty_list.next,
 656                                          struct dlm_lock_resource, dirty);
 657
 658                         /* peel a lockres off, remove it from the list,
 659                          * unset the dirty flag and drop the dlm lock */
 660                         BUG_ON(!res);
 661                         dlm_lockres_get(res);
 662
 663                         spin_lock(&res->spinlock);
 664                         /* We clear the DLM_LOCK_RES_DIRTY state once we shuffle lists below */
 665                         list_del_init(&res->dirty);
 666                         spin_unlock(&res->spinlock);
 667                         spin_unlock(&dlm->spinlock);
 668                         /* Drop dirty_list ref */
 669                         dlm_lockres_put(res);
 670
 671                         /* lockres can be re-dirtied/re-added to the
 672                          * dirty_list in this gap, but that is ok */
 673
 674                         spin_lock(&res->spinlock);
 675                         if (res->owner != dlm->node_num) {
 676                                 __dlm_print_one_lock_resource(res);
 677                                 mlog(ML_ERROR, "inprog:%s, mig:%s, reco:%s, dirty:%s\n",
 678                                      res->state & DLM_LOCK_RES_IN_PROGRESS ? "yes" : "no",
 679                                      res->state & DLM_LOCK_RES_MIGRATING ? "yes" : "no",
 680                                      res->state & DLM_LOCK_RES_RECOVERING ? "yes" : "no",
 681                                      res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
 682                         }
 683                         BUG_ON(res->owner != dlm->node_num);
 684
 685                         /* it is now ok to move lockreses in these states
 686                          * to the dirty list, assuming that they will only be
 687                          * dirty for a short while. */
 688                         BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
 689                         if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
 690                                           DLM_LOCK_RES_RECOVERING)) {
 691                                 /* move it to the tail and keep going */
 692                                 res->state &= ~DLM_LOCK_RES_DIRTY;
 693                                 spin_unlock(&res->spinlock);
 694                                 mlog(0, "delaying list shuffling for in-"
 695                                      "progress lockres %.*s, state=%d\n",
 696                                      res->lockname.len, res->lockname.name,
 697                                      res->state);
 698                                 delay = 1;
 699                                 goto in_progress;
 700                         }
 701
 702                         /* at this point the lockres is not migrating/
 703                          * recovering/in-progress.  we have the lockres
 704                          * spinlock and do NOT have the dlm lock.
 705                          * safe to reserve/queue asts and run the lists. */
 706
 707                         mlog(0, "calling dlm_shuffle_lists with dlm=%s, "
 708                              "res=%.*s\n", dlm->name,
 709                              res->lockname.len, res->lockname.name);
 710
 711                         /* called while holding lockres lock */
 712                         dlm_shuffle_lists(dlm, res);
 713                         res->state &= ~DLM_LOCK_RES_DIRTY;
 714                         spin_unlock(&res->spinlock);
 715
 716                         dlm_lockres_calc_usage(dlm, res);
 717
 718 in_progress:
 719
 720                         spin_lock(&dlm->spinlock);
 721                         /* if the lock was in-progress, stick
 722                          * it on the back of the list */
 723                         if (delay) {
 724                                 spin_lock(&res->spinlock);
 725                                 __dlm_dirty_lockres(dlm, res);
 726                                 spin_unlock(&res->spinlock);
 727                         }
 728                         dlm_lockres_put(res);
 729
 730                         /* unlikely, but we may need to give time to
 731                          * other tasks */
 732                         if (!--n) {
 733                                 mlog(0, "throttling dlm_thread\n");
 734                                 break;
 735                         }
 736                 }
 737
 738                 spin_unlock(&dlm->spinlock);
 739                 dlm_flush_asts(dlm);
 740
 741                 /* yield and continue right away if there is more work to do */
 742                 if (!n) {
 743                         cond_resched();
 744                         continue;
 745                 }
 746
 747                 wait_event_interruptible_timeout(dlm->dlm_thread_wq,
 748                                                  !dlm_dirty_list_empty(dlm) ||
 749                                                  kthread_should_stop(),
 750                                                  timeout);
 751         }
 752
 753         mlog(0, "quitting DLM thread\n");
 754         return 0;
 755 }