git.oblomov.eu Git - linux-2.6/blob - 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/utsname.h>
  34 #include <linux/init.h>
  35 #include <linux/sysctl.h>
  36 #include <linux/random.h>
  37 #include <linux/blkdev.h>
  38 #include <linux/socket.h>
  39 #include <linux/inet.h>
  40 #include <linux/timer.h>
  41 #include <linux/kthread.h>
  42 #include <linux/delay.h>
  43
  44
  45 #include "cluster/heartbeat.h"
  46 #include "cluster/nodemanager.h"
  47 #include "cluster/tcp.h"
  48
  49 #include "dlmapi.h"
  50 #include "dlmcommon.h"
  51 #include "dlmdomain.h"
  52
  53 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_THREAD)
  54 #include "cluster/masklog.h"
  55
  56 static int dlm_thread(void *data);
  57 static void dlm_flush_asts(struct dlm_ctxt *dlm);
  58
  59 #define dlm_lock_is_remote(dlm, lock)     ((lock)->ml.node != (dlm)->node_num)
  60
  61 /* will exit holding res->spinlock, but may drop in function */
  62 /* waits until flags are cleared on res->state */
  63 void __dlm_wait_on_lockres_flags(struct dlm_lock_resource *res, int flags)
  64 {
  65         DECLARE_WAITQUEUE(wait, current);
  66
  67         assert_spin_locked(&res->spinlock);
  68
  69         add_wait_queue(&res->wq, &wait);
  70 repeat:
  71         set_current_state(TASK_UNINTERRUPTIBLE);
  72         if (res->state & flags) {
  73                 spin_unlock(&res->spinlock);
  74                 schedule();
  75                 spin_lock(&res->spinlock);
  76                 goto repeat;
  77         }
  78         remove_wait_queue(&res->wq, &wait);
  79         __set_current_state(TASK_RUNNING);
  80 }
  81
  82 int __dlm_lockres_has_locks(struct dlm_lock_resource *res)
  83 {
  84         if (list_empty(&res->granted) &&
  85             list_empty(&res->converting) &&
  86             list_empty(&res->blocked))
  87                 return 0;
  88         return 1;
  89 }
  90
  91 /* "unused": the lockres has no locks, is not on the dirty list,
  92  * has no inflight locks (in the gap between mastery and acquiring
  93  * the first lock), and has no bits in its refmap.
  94  * truly ready to be freed. */
  95 int __dlm_lockres_unused(struct dlm_lock_resource *res)
  96 {
  97         if (!__dlm_lockres_has_locks(res) &&
  98             (list_empty(&res->dirty) && !(res->state & DLM_LOCK_RES_DIRTY))) {
  99                 /* try not to scan the bitmap unless the first two
 100                  * conditions are already true */
 101                 int bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
 102                 if (bit >= O2NM_MAX_NODES) {
 103                         /* since the bit for dlm->node_num is not
 104                          * set, inflight_locks better be zero */
 105                         BUG_ON(res->inflight_locks != 0);
 106                         return 1;
 107                 }
 108         }
 109         return 0;
 110 }
 111
 112
 113 /* Call whenever you may have added or deleted something from one of
 114  * the lockres queue's. This will figure out whether it belongs on the
 115  * unused list or not and does the appropriate thing. */
 116 void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 117                               struct dlm_lock_resource *res)
 118 {
 119         mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
 120
 121         assert_spin_locked(&dlm->spinlock);
 122         assert_spin_locked(&res->spinlock);
 123
 124         if (__dlm_lockres_unused(res)){
 125                 if (list_empty(&res->purge)) {
 126                         mlog(0, "putting lockres %.*s:%p onto purge list\n",
 127                              res->lockname.len, res->lockname.name, res);
 128
 129                         res->last_used = jiffies;
 130                         dlm_lockres_get(res);
 131                         list_add_tail(&res->purge, &dlm->purge_list);
 132                         dlm->purge_count++;
 133                 }
 134         } else if (!list_empty(&res->purge)) {
 135                 mlog(0, "removing lockres %.*s:%p from purge list, owner=%u\n",
 136                      res->lockname.len, res->lockname.name, res, res->owner);
 137
 138                 list_del_init(&res->purge);
 139                 dlm_lockres_put(res);
 140                 dlm->purge_count--;
 141         }
 142 }
 143
 144 void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 145                             struct dlm_lock_resource *res)
 146 {
 147         mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
 148         spin_lock(&dlm->spinlock);
 149         spin_lock(&res->spinlock);
 150
 151         __dlm_lockres_calc_usage(dlm, res);
 152
 153         spin_unlock(&res->spinlock);
 154         spin_unlock(&dlm->spinlock);
 155 }
 156
 157 static int dlm_purge_lockres(struct dlm_ctxt *dlm,
 158                              struct dlm_lock_resource *res)
 159 {
 160         int master;
 161         int ret = 0;
 162
 163         spin_lock(&res->spinlock);
 164         if (!__dlm_lockres_unused(res)) {
 165                 spin_unlock(&res->spinlock);
 166                 mlog(0, "%s:%.*s: tried to purge but not unused\n",
 167                      dlm->name, res->lockname.len, res->lockname.name);
 168                 return -ENOTEMPTY;
 169         }
 170         master = (res->owner == dlm->node_num);
 171         if (!master)
 172                 res->state |= DLM_LOCK_RES_DROPPING_REF;
 173         spin_unlock(&res->spinlock);
 174
 175         mlog(0, "purging lockres %.*s, master = %d\n", res->lockname.len,
 176              res->lockname.name, master);
 177
 178         if (!master) {
 179                 /* drop spinlock...  retake below */
 180                 spin_unlock(&dlm->spinlock);
 181
 182                 spin_lock(&res->spinlock);
 183                 /* This ensures that clear refmap is sent after the set */
 184                 __dlm_wait_on_lockres_flags(res, DLM_LOCK_RES_SETREF_INPROG);
 185                 spin_unlock(&res->spinlock);
 186
 187                 /* clear our bit from the master's refmap, ignore errors */
 188                 ret = dlm_drop_lockres_ref(dlm, res);
 189                 if (ret < 0) {
 190                         mlog_errno(ret);
 191                         if (!dlm_is_host_down(ret))
 192                                 BUG();
 193                 }
 194                 mlog(0, "%s:%.*s: dlm_deref_lockres returned %d\n",
 195                      dlm->name, res->lockname.len, res->lockname.name, ret);
 196                 spin_lock(&dlm->spinlock);
 197         }
 198
 199         if (!list_empty(&res->purge)) {
 200                 mlog(0, "removing lockres %.*s:%p from purgelist, "
 201                      "master = %d\n", res->lockname.len, res->lockname.name,
 202                      res, master);
 203                 list_del_init(&res->purge);
 204                 dlm_lockres_put(res);
 205                 dlm->purge_count--;
 206         }
 207         __dlm_unhash_lockres(res);
 208
 209         /* lockres is not in the hash now.  drop the flag and wake up
 210          * any processes waiting in dlm_get_lock_resource. */
 211         if (!master) {
 212                 spin_lock(&res->spinlock);
 213                 res->state &= ~DLM_LOCK_RES_DROPPING_REF;
 214                 spin_unlock(&res->spinlock);
 215                 wake_up(&res->wq);
 216         }
 217         return 0;
 218 }
 219
 220 static void dlm_run_purge_list(struct dlm_ctxt *dlm,
 221                                int purge_now)
 222 {
 223         unsigned int run_max, unused;
 224         unsigned long purge_jiffies;
 225         struct dlm_lock_resource *lockres;
 226
 227         spin_lock(&dlm->spinlock);
 228         run_max = dlm->purge_count;
 229
 230         while(run_max && !list_empty(&dlm->purge_list)) {
 231                 run_max--;
 232
 233                 lockres = list_entry(dlm->purge_list.next,
 234                                      struct dlm_lock_resource, purge);
 235
 236                 /* Status of the lockres *might* change so double
 237                  * check. If the lockres is unused, holding the dlm
 238                  * spinlock will prevent people from getting and more
 239                  * refs on it -- there's no need to keep the lockres
 240                  * spinlock. */
 241                 spin_lock(&lockres->spinlock);
 242                 unused = __dlm_lockres_unused(lockres);
 243                 spin_unlock(&lockres->spinlock);
 244
 245                 if (!unused)
 246                         continue;
 247
 248                 purge_jiffies = lockres->last_used +
 249                         msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);
 250
 251                 /* Make sure that we want to be processing this guy at
 252                  * this time. */
 253                 if (!purge_now && time_after(purge_jiffies, jiffies)) {
 254                         /* Since resources are added to the purge list
 255                          * in tail order, we can stop at the first
 256                          * unpurgable resource -- anyone added after
 257                          * him will have a greater last_used value */
 258                         break;
 259                 }
 260
 261                 dlm_lockres_get(lockres);
 262
 263                 /* This may drop and reacquire the dlm spinlock if it
 264                  * has to do migration. */
 265                 if (dlm_purge_lockres(dlm, lockres))
 266                         BUG();
 267
 268                 dlm_lockres_put(lockres);
 269
 270                 /* Avoid adding any scheduling latencies */
 271                 cond_resched_lock(&dlm->spinlock);
 272         }
 273
 274         spin_unlock(&dlm->spinlock);
 275 }
 276
 277 static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
 278                               struct dlm_lock_resource *res)
 279 {
 280         struct dlm_lock *lock, *target;
 281         struct list_head *iter;
 282         struct list_head *head;
 283         int can_grant = 1;
 284
 285         //mlog(0, "res->lockname.len=%d\n", res->lockname.len);
 286         //mlog(0, "res->lockname.name=%p\n", res->lockname.name);
 287         //mlog(0, "shuffle res %.*s\n", res->lockname.len,
 288         //        res->lockname.name);
 289
 290         /* because this function is called with the lockres
 291          * spinlock, and because we know that it is not migrating/
 292          * recovering/in-progress, it is fine to reserve asts and
 293          * basts right before queueing them all throughout */
 294         assert_spin_locked(&res->spinlock);
 295         BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
 296                               DLM_LOCK_RES_RECOVERING|
 297                               DLM_LOCK_RES_IN_PROGRESS)));
 298
 299 converting:
 300         if (list_empty(&res->converting))
 301                 goto blocked;
 302         mlog(0, "res %.*s has locks on a convert queue\n", res->lockname.len,
 303              res->lockname.name);
 304
 305         target = list_entry(res->converting.next, struct dlm_lock, list);
 306         if (target->ml.convert_type == LKM_IVMODE) {
 307                 mlog(ML_ERROR, "%.*s: converting a lock with no "
 308                      "convert_type!\n", res->lockname.len, res->lockname.name);
 309                 BUG();
 310         }
 311         head = &res->granted;
 312         list_for_each(iter, head) {
 313                 lock = list_entry(iter, struct dlm_lock, list);
 314                 if (lock==target)
 315                         continue;
 316                 if (!dlm_lock_compatible(lock->ml.type,
 317                                          target->ml.convert_type)) {
 318                         can_grant = 0;
 319                         /* queue the BAST if not already */
 320                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 321                                 __dlm_lockres_reserve_ast(res);
 322                                 dlm_queue_bast(dlm, lock);
 323                         }
 324                         /* update the highest_blocked if needed */
 325                         if (lock->ml.highest_blocked < target->ml.convert_type)
 326                                 lock->ml.highest_blocked =
 327                                         target->ml.convert_type;
 328                 }
 329         }
 330         head = &res->converting;
 331         list_for_each(iter, head) {
 332                 lock = list_entry(iter, struct dlm_lock, list);
 333                 if (lock==target)
 334                         continue;
 335                 if (!dlm_lock_compatible(lock->ml.type,
 336                                          target->ml.convert_type)) {
 337                         can_grant = 0;
 338                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 339                                 __dlm_lockres_reserve_ast(res);
 340                                 dlm_queue_bast(dlm, lock);
 341                         }
 342                         if (lock->ml.highest_blocked < target->ml.convert_type)
 343                                 lock->ml.highest_blocked =
 344                                         target->ml.convert_type;
 345                 }
 346         }
 347
 348         /* we can convert the lock */
 349         if (can_grant) {
 350                 spin_lock(&target->spinlock);
 351                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 352
 353                 mlog(0, "calling ast for converting lock: %.*s, have: %d, "
 354                      "granting: %d, node: %u\n", res->lockname.len,
 355                      res->lockname.name, target->ml.type,
 356                      target->ml.convert_type, target->ml.node);
 357
 358                 target->ml.type = target->ml.convert_type;
 359                 target->ml.convert_type = LKM_IVMODE;
 360                 list_move_tail(&target->list, &res->granted);
 361
 362                 BUG_ON(!target->lksb);
 363                 target->lksb->status = DLM_NORMAL;
 364
 365                 spin_unlock(&target->spinlock);
 366
 367                 __dlm_lockres_reserve_ast(res);
 368                 dlm_queue_ast(dlm, target);
 369                 /* go back and check for more */
 370                 goto converting;
 371         }
 372
 373 blocked:
 374         if (list_empty(&res->blocked))
 375                 goto leave;
 376         target = list_entry(res->blocked.next, struct dlm_lock, list);
 377
 378         head = &res->granted;
 379         list_for_each(iter, head) {
 380                 lock = list_entry(iter, struct dlm_lock, list);
 381                 if (lock==target)
 382                         continue;
 383                 if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
 384                         can_grant = 0;
 385                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 386                                 __dlm_lockres_reserve_ast(res);
 387                                 dlm_queue_bast(dlm, lock);
 388                         }
 389                         if (lock->ml.highest_blocked < target->ml.type)
 390                                 lock->ml.highest_blocked = target->ml.type;
 391                 }
 392         }
 393
 394         head = &res->converting;
 395         list_for_each(iter, head) {
 396                 lock = list_entry(iter, struct dlm_lock, list);
 397                 if (lock==target)
 398                         continue;
 399                 if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
 400                         can_grant = 0;
 401                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 402                                 __dlm_lockres_reserve_ast(res);
 403                                 dlm_queue_bast(dlm, lock);
 404                         }
 405                         if (lock->ml.highest_blocked < target->ml.type)
 406                                 lock->ml.highest_blocked = target->ml.type;
 407                 }
 408         }
 409
 410         /* we can grant the blocked lock (only
 411          * possible if converting list empty) */
 412         if (can_grant) {
 413                 spin_lock(&target->spinlock);
 414                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 415
 416                 mlog(0, "calling ast for blocked lock: %.*s, granting: %d, "
 417                      "node: %u\n", res->lockname.len, res->lockname.name,
 418                      target->ml.type, target->ml.node);
 419
 420                 // target->ml.type is already correct
 421                 list_move_tail(&target->list, &res->granted);
 422
 423                 BUG_ON(!target->lksb);
 424                 target->lksb->status = DLM_NORMAL;
 425
 426                 spin_unlock(&target->spinlock);
 427
 428                 __dlm_lockres_reserve_ast(res);
 429                 dlm_queue_ast(dlm, target);
 430                 /* go back and check for more */
 431                 goto converting;
 432         }
 433
 434 leave:
 435         return;
 436 }
 437
 438 /* must have NO locks when calling this with res !=NULL * */
 439 void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 440 {
 441         mlog_entry("dlm=%p, res=%p\n", dlm, res);
 442         if (res) {
 443                 spin_lock(&dlm->spinlock);
 444                 spin_lock(&res->spinlock);
 445                 __dlm_dirty_lockres(dlm, res);
 446                 spin_unlock(&res->spinlock);
 447                 spin_unlock(&dlm->spinlock);
 448         }
 449         wake_up(&dlm->dlm_thread_wq);
 450 }
 451
 452 void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 453 {
 454         mlog_entry("dlm=%p, res=%p\n", dlm, res);
 455
 456         assert_spin_locked(&dlm->spinlock);
 457         assert_spin_locked(&res->spinlock);
 458
 459         /* don't shuffle secondary queues */
 460         if ((res->owner == dlm->node_num)) {
 461                 if (res->state & (DLM_LOCK_RES_MIGRATING |
 462                                   DLM_LOCK_RES_BLOCK_DIRTY))
 463                     return;
 464
 465                 if (list_empty(&res->dirty)) {
 466                         /* ref for dirty_list */
 467                         dlm_lockres_get(res);
 468                         list_add_tail(&res->dirty, &dlm->dirty_list);
 469                         res->state |= DLM_LOCK_RES_DIRTY;
 470                 }
 471         }
 472 }
 473
 474
 475 /* Launch the NM thread for the mounted volume */
 476 int dlm_launch_thread(struct dlm_ctxt *dlm)
 477 {
 478         mlog(0, "starting dlm thread...\n");
 479
 480         dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm_thread");
 481         if (IS_ERR(dlm->dlm_thread_task)) {
 482                 mlog_errno(PTR_ERR(dlm->dlm_thread_task));
 483                 dlm->dlm_thread_task = NULL;
 484                 return -EINVAL;
 485         }
 486
 487         return 0;
 488 }
 489
 490 void dlm_complete_thread(struct dlm_ctxt *dlm)
 491 {
 492         if (dlm->dlm_thread_task) {
 493                 mlog(ML_KTHREAD, "waiting for dlm thread to exit\n");
 494                 kthread_stop(dlm->dlm_thread_task);
 495                 dlm->dlm_thread_task = NULL;
 496         }
 497 }
 498
 499 static int dlm_dirty_list_empty(struct dlm_ctxt *dlm)
 500 {
 501         int empty;
 502
 503         spin_lock(&dlm->spinlock);
 504         empty = list_empty(&dlm->dirty_list);
 505         spin_unlock(&dlm->spinlock);
 506
 507         return empty;
 508 }
 509
 510 static void dlm_flush_asts(struct dlm_ctxt *dlm)
 511 {
 512         int ret;
 513         struct dlm_lock *lock;
 514         struct dlm_lock_resource *res;
 515         u8 hi;
 516
 517         spin_lock(&dlm->ast_lock);
 518         while (!list_empty(&dlm->pending_asts)) {
 519                 lock = list_entry(dlm->pending_asts.next,
 520                                   struct dlm_lock, ast_list);
 521                 /* get an extra ref on lock */
 522                 dlm_lock_get(lock);
 523                 res = lock->lockres;
 524                 mlog(0, "delivering an ast for this lockres\n");
 525
 526                 BUG_ON(!lock->ast_pending);
 527
 528                 /* remove from list (including ref) */
 529                 list_del_init(&lock->ast_list);
 530                 dlm_lock_put(lock);
 531                 spin_unlock(&dlm->ast_lock);
 532
 533                 if (lock->ml.node != dlm->node_num) {
 534                         ret = dlm_do_remote_ast(dlm, res, lock);
 535                         if (ret < 0)
 536                                 mlog_errno(ret);
 537                 } else
 538                         dlm_do_local_ast(dlm, res, lock);
 539
 540                 spin_lock(&dlm->ast_lock);
 541
 542                 /* possible that another ast was queued while
 543                  * we were delivering the last one */
 544                 if (!list_empty(&lock->ast_list)) {
 545                         mlog(0, "aha another ast got queued while "
 546                              "we were finishing the last one.  will "
 547                              "keep the ast_pending flag set.\n");
 548                 } else
 549                         lock->ast_pending = 0;
 550
 551                 /* drop the extra ref.
 552                  * this may drop it completely. */
 553                 dlm_lock_put(lock);
 554                 dlm_lockres_release_ast(dlm, res);
 555         }
 556
 557         while (!list_empty(&dlm->pending_basts)) {
 558                 lock = list_entry(dlm->pending_basts.next,
 559                                   struct dlm_lock, bast_list);
 560                 /* get an extra ref on lock */
 561                 dlm_lock_get(lock);
 562                 res = lock->lockres;
 563
 564                 BUG_ON(!lock->bast_pending);
 565
 566                 /* get the highest blocked lock, and reset */
 567                 spin_lock(&lock->spinlock);
 568                 BUG_ON(lock->ml.highest_blocked <= LKM_IVMODE);
 569                 hi = lock->ml.highest_blocked;
 570                 lock->ml.highest_blocked = LKM_IVMODE;
 571                 spin_unlock(&lock->spinlock);
 572
 573                 /* remove from list (including ref) */
 574                 list_del_init(&lock->bast_list);
 575                 dlm_lock_put(lock);
 576                 spin_unlock(&dlm->ast_lock);
 577
 578                 mlog(0, "delivering a bast for this lockres "
 579                      "(blocked = %d\n", hi);
 580
 581                 if (lock->ml.node != dlm->node_num) {
 582                         ret = dlm_send_proxy_bast(dlm, res, lock, hi);
 583                         if (ret < 0)
 584                                 mlog_errno(ret);
 585                 } else
 586                         dlm_do_local_bast(dlm, res, lock, hi);
 587
 588                 spin_lock(&dlm->ast_lock);
 589
 590                 /* possible that another bast was queued while
 591                  * we were delivering the last one */
 592                 if (!list_empty(&lock->bast_list)) {
 593                         mlog(0, "aha another bast got queued while "
 594                              "we were finishing the last one.  will "
 595                              "keep the bast_pending flag set.\n");
 596                 } else
 597                         lock->bast_pending = 0;
 598
 599                 /* drop the extra ref.
 600                  * this may drop it completely. */
 601                 dlm_lock_put(lock);
 602                 dlm_lockres_release_ast(dlm, res);
 603         }
 604         wake_up(&dlm->ast_wq);
 605         spin_unlock(&dlm->ast_lock);
 606 }
 607
 608
 609 #define DLM_THREAD_TIMEOUT_MS (4 * 1000)
 610 #define DLM_THREAD_MAX_DIRTY  100
 611 #define DLM_THREAD_MAX_ASTS   10
 612
 613 static int dlm_thread(void *data)
 614 {
 615         struct dlm_lock_resource *res;
 616         struct dlm_ctxt *dlm = data;
 617         unsigned long timeout = msecs_to_jiffies(DLM_THREAD_TIMEOUT_MS);
 618
 619         mlog(0, "dlm thread running for %s...\n", dlm->name);
 620
 621         while (!kthread_should_stop()) {
 622                 int n = DLM_THREAD_MAX_DIRTY;
 623
 624                 /* dlm_shutting_down is very point-in-time, but that
 625                  * doesn't matter as we'll just loop back around if we
 626                  * get false on the leading edge of a state
 627                  * transition. */
 628                 dlm_run_purge_list(dlm, dlm_shutting_down(dlm));
 629
 630                 /* We really don't want to hold dlm->spinlock while
 631                  * calling dlm_shuffle_lists on each lockres that
 632                  * needs to have its queues adjusted and AST/BASTs
 633                  * run.  So let's pull each entry off the dirty_list
 634                  * and drop dlm->spinlock ASAP.  Once off the list,
 635                  * res->spinlock needs to be taken again to protect
 636                  * the queues while calling dlm_shuffle_lists.  */
 637                 spin_lock(&dlm->spinlock);
 638                 while (!list_empty(&dlm->dirty_list)) {
 639                         int delay = 0;
 640                         res = list_entry(dlm->dirty_list.next,
 641                                          struct dlm_lock_resource, dirty);
 642
 643                         /* peel a lockres off, remove it from the list,
 644                          * unset the dirty flag and drop the dlm lock */
 645                         BUG_ON(!res);
 646                         dlm_lockres_get(res);
 647
 648                         spin_lock(&res->spinlock);
 649                         /* We clear the DLM_LOCK_RES_DIRTY state once we shuffle lists below */
 650                         list_del_init(&res->dirty);
 651                         spin_unlock(&res->spinlock);
 652                         spin_unlock(&dlm->spinlock);
 653                         /* Drop dirty_list ref */
 654                         dlm_lockres_put(res);
 655
 656                         /* lockres can be re-dirtied/re-added to the
 657                          * dirty_list in this gap, but that is ok */
 658
 659                         spin_lock(&res->spinlock);
 660                         if (res->owner != dlm->node_num) {
 661                                 __dlm_print_one_lock_resource(res);
 662                                 mlog(ML_ERROR, "inprog:%s, mig:%s, reco:%s, dirty:%s\n",
 663                                      res->state & DLM_LOCK_RES_IN_PROGRESS ? "yes" : "no",
 664                                      res->state & DLM_LOCK_RES_MIGRATING ? "yes" : "no",
 665                                      res->state & DLM_LOCK_RES_RECOVERING ? "yes" : "no",
 666                                      res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
 667                         }
 668                         BUG_ON(res->owner != dlm->node_num);
 669
 670                         /* it is now ok to move lockreses in these states
 671                          * to the dirty list, assuming that they will only be
 672                          * dirty for a short while. */
 673                         BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
 674                         if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
 675                                           DLM_LOCK_RES_RECOVERING)) {
 676                                 /* move it to the tail and keep going */
 677                                 res->state &= ~DLM_LOCK_RES_DIRTY;
 678                                 spin_unlock(&res->spinlock);
 679                                 mlog(0, "delaying list shuffling for in-"
 680                                      "progress lockres %.*s, state=%d\n",
 681                                      res->lockname.len, res->lockname.name,
 682                                      res->state);
 683                                 delay = 1;
 684                                 goto in_progress;
 685                         }
 686
 687                         /* at this point the lockres is not migrating/
 688                          * recovering/in-progress.  we have the lockres
 689                          * spinlock and do NOT have the dlm lock.
 690                          * safe to reserve/queue asts and run the lists. */
 691
 692                         mlog(0, "calling dlm_shuffle_lists with dlm=%s, "
 693                              "res=%.*s\n", dlm->name,
 694                              res->lockname.len, res->lockname.name);
 695
 696                         /* called while holding lockres lock */
 697                         dlm_shuffle_lists(dlm, res);
 698                         res->state &= ~DLM_LOCK_RES_DIRTY;
 699                         spin_unlock(&res->spinlock);
 700
 701                         dlm_lockres_calc_usage(dlm, res);
 702
 703 in_progress:
 704
 705                         spin_lock(&dlm->spinlock);
 706                         /* if the lock was in-progress, stick
 707                          * it on the back of the list */
 708                         if (delay) {
 709                                 spin_lock(&res->spinlock);
 710                                 __dlm_dirty_lockres(dlm, res);
 711                                 spin_unlock(&res->spinlock);
 712                         }
 713                         dlm_lockres_put(res);
 714
 715                         /* unlikely, but we may need to give time to
 716                          * other tasks */
 717                         if (!--n) {
 718                                 mlog(0, "throttling dlm_thread\n");
 719                                 break;
 720                         }
 721                 }
 722
 723                 spin_unlock(&dlm->spinlock);
 724                 dlm_flush_asts(dlm);
 725
 726                 /* yield and continue right away if there is more work to do */
 727                 if (!n) {
 728                         cond_resched();
 729                         continue;
 730                 }
 731
 732                 wait_event_interruptible_timeout(dlm->dlm_thread_wq,
 733                                                  !dlm_dirty_list_empty(dlm) ||
 734                                                  kthread_should_stop(),
 735                                                  timeout);
 736         }
 737
 738         mlog(0, "quitting DLM thread\n");
 739         return 0;
 740 }