kernel/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/highmem.h>
  32 #include <linux/init.h>
  33 #include <linux/sysctl.h>
  34 #include <linux/random.h>
  35 #include <linux/blkdev.h>
  36 #include <linux/socket.h>
  37 #include <linux/inet.h>
  38 #include <linux/timer.h>
  39 #include <linux/kthread.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
  51 #define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_THREAD)
  52 #include "cluster/masklog.h"
  53
  54 static int dlm_thread(void *data);
  55 static void dlm_flush_asts(struct dlm_ctxt *dlm);
  56
  57 #define dlm_lock_is_remote(dlm, lock)     ((lock)->ml.node != (dlm)->node_num)
  58
  59 /* will exit holding res->spinlock, but may drop in function */
  60 /* waits until flags are cleared on res->state */
  61 void __dlm_wait_on_lockres_flags(struct dlm_lock_resource *res, int flags)
  62 {
  63         DECLARE_WAITQUEUE(wait, current);
  64
  65         assert_spin_locked(&res->spinlock);
  66
  67         add_wait_queue(&res->wq, &wait);
  68 repeat:
  69         set_current_state(TASK_UNINTERRUPTIBLE);
  70         if (res->state & flags) {
  71                 spin_unlock(&res->spinlock);
  72                 schedule();
  73                 spin_lock(&res->spinlock);
  74                 goto repeat;
  75         }
  76         remove_wait_queue(&res->wq, &wait);
  77         __set_current_state(TASK_RUNNING);
  78 }
  79
  80 int __dlm_lockres_has_locks(struct dlm_lock_resource *res)
  81 {
  82         if (list_empty(&res->granted) &&
  83             list_empty(&res->converting) &&
  84             list_empty(&res->blocked))
  85                 return 0;
  86         return 1;
  87 }
  88
  89 /* "unused": the lockres has no locks, is not on the dirty list,
  90  * has no inflight locks (in the gap between mastery and acquiring
  91  * the first lock), and has no bits in its refmap.
  92  * truly ready to be freed. */
  93 int __dlm_lockres_unused(struct dlm_lock_resource *res)
  94 {
  95         int bit;
  96
  97         assert_spin_locked(&res->spinlock);
  98
  99         if (__dlm_lockres_has_locks(res))
 100                 return 0;
 101
 102         /* Locks are in the process of being created */
 103         if (res->inflight_locks)
 104                 return 0;
 105
 106         if (!list_empty(&res->dirty) || res->state & DLM_LOCK_RES_DIRTY)
 107                 return 0;
 108
 109         if (res->state & DLM_LOCK_RES_RECOVERING)
 110                 return 0;
 111
 112         /* Another node has this resource with this node as the master */
 113         bit = find_next_bit(res->refmap, O2NM_MAX_NODES, 0);
 114         if (bit < O2NM_MAX_NODES)
 115                 return 0;
 116
 117         return 1;
 118 }
 119
 120
 121 /* Call whenever you may have added or deleted something from one of
 122  * the lockres queue's. This will figure out whether it belongs on the
 123  * unused list or not and does the appropriate thing. */
 124 void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 125                               struct dlm_lock_resource *res)
 126 {
 127         assert_spin_locked(&dlm->spinlock);
 128         assert_spin_locked(&res->spinlock);
 129
 130         if (__dlm_lockres_unused(res)){
 131                 if (list_empty(&res->purge)) {
 132                         mlog(0, "%s: Adding res %.*s to purge list\n",
 133                              dlm->name, res->lockname.len, res->lockname.name);
 134
 135                         res->last_used = jiffies;
 136                         dlm_lockres_get(res);
 137                         list_add_tail(&res->purge, &dlm->purge_list);
 138                         dlm->purge_count++;
 139                 }
 140         } else if (!list_empty(&res->purge)) {
 141                 mlog(0, "%s: Removing res %.*s from purge list\n",
 142                      dlm->name, res->lockname.len, res->lockname.name);
 143
 144                 list_del_init(&res->purge);
 145                 dlm_lockres_put(res);
 146                 dlm->purge_count--;
 147         }
 148 }
 149
 150 void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
 151                             struct dlm_lock_resource *res)
 152 {
 153         spin_lock(&dlm->spinlock);
 154         spin_lock(&res->spinlock);
 155
 156         __dlm_lockres_calc_usage(dlm, res);
 157
 158         spin_unlock(&res->spinlock);
 159         spin_unlock(&dlm->spinlock);
 160 }
 161
 162 static void dlm_purge_lockres(struct dlm_ctxt *dlm,
 163                              struct dlm_lock_resource *res)
 164 {
 165         int master;
 166         int ret = 0;
 167
 168         assert_spin_locked(&dlm->spinlock);
 169         assert_spin_locked(&res->spinlock);
 170
 171         master = (res->owner == dlm->node_num);
 172
 173         mlog(0, "%s: Purging res %.*s, master %d\n", dlm->name,
 174              res->lockname.len, res->lockname.name, master);
 175
 176         if (!master) {
 177                 res->state |= DLM_LOCK_RES_DROPPING_REF;
 178                 /* drop spinlock...  retake below */
 179                 spin_unlock(&res->spinlock);
 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                         if (!dlm_is_host_down(ret))
 191                                 BUG();
 192                 }
 193                 spin_lock(&dlm->spinlock);
 194                 spin_lock(&res->spinlock);
 195         }
 196
 197         if (!list_empty(&res->purge)) {
 198                 mlog(0, "%s: Removing res %.*s from purgelist, master %d\n",
 199                      dlm->name, res->lockname.len, res->lockname.name, master);
 200                 list_del_init(&res->purge);
 201                 dlm_lockres_put(res);
 202                 dlm->purge_count--;
 203         }
 204
 205         if (!__dlm_lockres_unused(res)) {
 206                 mlog(ML_ERROR, "%s: res %.*s in use after deref\n",
 207                      dlm->name, res->lockname.len, res->lockname.name);
 208                 __dlm_print_one_lock_resource(res);
 209                 BUG();
 210         }
 211
 212         __dlm_unhash_lockres(dlm, res);
 213
 214         spin_lock(&dlm->track_lock);
 215         if (!list_empty(&res->tracking))
 216                 list_del_init(&res->tracking);
 217         else {
 218                 mlog(ML_ERROR, "Resource %.*s not on the Tracking list\n",
 219                                 res->lockname.len, res->lockname.name);
 220                 __dlm_print_one_lock_resource(res);
 221         }
 222         spin_unlock(&dlm->track_lock);
 223
 224         /* lockres is not in the hash now.  drop the flag and wake up
 225          * any processes waiting in dlm_get_lock_resource. */
 226         if (!master) {
 227                 res->state &= ~DLM_LOCK_RES_DROPPING_REF;
 228                 spin_unlock(&res->spinlock);
 229                 wake_up(&res->wq);
 230         } else
 231                 spin_unlock(&res->spinlock);
 232 }
 233
 234 static void dlm_run_purge_list(struct dlm_ctxt *dlm,
 235                                int purge_now)
 236 {
 237         unsigned int run_max, unused;
 238         unsigned long purge_jiffies;
 239         struct dlm_lock_resource *lockres;
 240
 241         spin_lock(&dlm->spinlock);
 242         run_max = dlm->purge_count;
 243
 244         while(run_max && !list_empty(&dlm->purge_list)) {
 245                 run_max--;
 246
 247                 lockres = list_entry(dlm->purge_list.next,
 248                                      struct dlm_lock_resource, purge);
 249
 250                 spin_lock(&lockres->spinlock);
 251
 252                 purge_jiffies = lockres->last_used +
 253                         msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);
 254
 255                 /* Make sure that we want to be processing this guy at
 256                  * this time. */
 257                 if (!purge_now && time_after(purge_jiffies, jiffies)) {
 258                         /* Since resources are added to the purge list
 259                          * in tail order, we can stop at the first
 260                          * unpurgable resource -- anyone added after
 261                          * him will have a greater last_used value */
 262                         spin_unlock(&lockres->spinlock);
 263                         break;
 264                 }
 265
 266                 /* Status of the lockres *might* change so double
 267                  * check. If the lockres is unused, holding the dlm
 268                  * spinlock will prevent people from getting and more
 269                  * refs on it. */
 270                 unused = __dlm_lockres_unused(lockres);
 271                 if (!unused ||
 272                     (lockres->state & DLM_LOCK_RES_MIGRATING) ||
 273                     (lockres->inflight_assert_workers != 0)) {
 274                         mlog(0, "%s: res %.*s is in use or being remastered, "
 275                              "used %d, state %d, assert master workers %u\n",
 276                              dlm->name, lockres->lockname.len,
 277                              lockres->lockname.name,
 278                              !unused, lockres->state,
 279                              lockres->inflight_assert_workers);
 280                         list_move_tail(&lockres->purge, &dlm->purge_list);
 281                         spin_unlock(&lockres->spinlock);
 282                         continue;
 283                 }
 284
 285                 dlm_lockres_get(lockres);
 286
 287                 dlm_purge_lockres(dlm, lockres);
 288
 289                 dlm_lockres_put(lockres);
 290
 291                 /* Avoid adding any scheduling latencies */
 292                 cond_resched_lock(&dlm->spinlock);
 293         }
 294
 295         spin_unlock(&dlm->spinlock);
 296 }
 297
 298 static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
 299                               struct dlm_lock_resource *res)
 300 {
 301         struct dlm_lock *lock, *target;
 302         int can_grant = 1;
 303
 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          */
 310         assert_spin_locked(&dlm->ast_lock);
 311         assert_spin_locked(&res->spinlock);
 312         BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
 313                               DLM_LOCK_RES_RECOVERING|
 314                               DLM_LOCK_RES_IN_PROGRESS)));
 315
 316 converting:
 317         if (list_empty(&res->converting))
 318                 goto blocked;
 319         mlog(0, "%s: res %.*s has locks on the convert queue\n", dlm->name,
 320              res->lockname.len, res->lockname.name);
 321
 322         target = list_entry(res->converting.next, struct dlm_lock, list);
 323         if (target->ml.convert_type == LKM_IVMODE) {
 324                 mlog(ML_ERROR, "%s: res %.*s converting lock to invalid mode\n",
 325                      dlm->name, res->lockname.len, res->lockname.name);
 326                 BUG();
 327         }
 328         list_for_each_entry(lock, &res->granted, 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
 346         list_for_each_entry(lock, &res->converting, list) {
 347                 if (lock==target)
 348                         continue;
 349                 if (!dlm_lock_compatible(lock->ml.type,
 350                                          target->ml.convert_type)) {
 351                         can_grant = 0;
 352                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 353                                 __dlm_lockres_reserve_ast(res);
 354                                 __dlm_queue_bast(dlm, lock);
 355                         }
 356                         if (lock->ml.highest_blocked < target->ml.convert_type)
 357                                 lock->ml.highest_blocked =
 358                                         target->ml.convert_type;
 359                 }
 360         }
 361
 362         /* we can convert the lock */
 363         if (can_grant) {
 364                 spin_lock(&target->spinlock);
 365                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 366
 367                 mlog(0, "%s: res %.*s, AST for Converting lock %u:%llu, type "
 368                      "%d => %d, node %u\n", dlm->name, res->lockname.len,
 369                      res->lockname.name,
 370                      dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)),
 371                      dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)),
 372                      target->ml.type,
 373                      target->ml.convert_type, target->ml.node);
 374
 375                 target->ml.type = target->ml.convert_type;
 376                 target->ml.convert_type = LKM_IVMODE;
 377                 list_move_tail(&target->list, &res->granted);
 378
 379                 BUG_ON(!target->lksb);
 380                 target->lksb->status = DLM_NORMAL;
 381
 382                 spin_unlock(&target->spinlock);
 383
 384                 __dlm_lockres_reserve_ast(res);
 385                 __dlm_queue_ast(dlm, target);
 386                 /* go back and check for more */
 387                 goto converting;
 388         }
 389
 390 blocked:
 391         if (list_empty(&res->blocked))
 392                 goto leave;
 393         target = list_entry(res->blocked.next, struct dlm_lock, list);
 394
 395         list_for_each_entry(lock, &res->granted, 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         list_for_each_entry(lock, &res->converting, list) {
 410                 if (lock==target)
 411                         continue;
 412                 if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
 413                         can_grant = 0;
 414                         if (lock->ml.highest_blocked == LKM_IVMODE) {
 415                                 __dlm_lockres_reserve_ast(res);
 416                                 __dlm_queue_bast(dlm, lock);
 417                         }
 418                         if (lock->ml.highest_blocked < target->ml.type)
 419                                 lock->ml.highest_blocked = target->ml.type;
 420                 }
 421         }
 422
 423         /* we can grant the blocked lock (only
 424          * possible if converting list empty) */
 425         if (can_grant) {
 426                 spin_lock(&target->spinlock);
 427                 BUG_ON(target->ml.highest_blocked != LKM_IVMODE);
 428
 429                 mlog(0, "%s: res %.*s, AST for Blocked lock %u:%llu, type %d, "
 430                      "node %u\n", dlm->name, res->lockname.len,
 431                      res->lockname.name,
 432                      dlm_get_lock_cookie_node(be64_to_cpu(target->ml.cookie)),
 433                      dlm_get_lock_cookie_seq(be64_to_cpu(target->ml.cookie)),
 434                      target->ml.type, target->ml.node);
 435
 436                 /* target->ml.type is already correct */
 437                 list_move_tail(&target->list, &res->granted);
 438
 439                 BUG_ON(!target->lksb);
 440                 target->lksb->status = DLM_NORMAL;
 441
 442                 spin_unlock(&target->spinlock);
 443
 444                 __dlm_lockres_reserve_ast(res);
 445                 __dlm_queue_ast(dlm, target);
 446                 /* go back and check for more */
 447                 goto converting;
 448         }
 449
 450 leave:
 451         return;
 452 }
 453
 454 /* must have NO locks when calling this with res !=NULL * */
 455 void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
 456 {
 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         assert_spin_locked(&dlm->spinlock);
 470         assert_spin_locked(&res->spinlock);
 471
 472         /* don't shuffle secondary queues */
 473         if ((res->owner == dlm->node_num)) {
 474                 if (res->state & (DLM_LOCK_RES_MIGRATING |
 475                                   DLM_LOCK_RES_BLOCK_DIRTY))
 476                     return;
 477
 478                 if (list_empty(&res->dirty)) {
 479                         /* ref for dirty_list */
 480                         dlm_lockres_get(res);
 481                         list_add_tail(&res->dirty, &dlm->dirty_list);
 482                         res->state |= DLM_LOCK_RES_DIRTY;
 483                 }
 484         }
 485
 486         mlog(0, "%s: res %.*s\n", dlm->name, res->lockname.len,
 487              res->lockname.name);
 488 }
 489
 490
 491 /* Launch the NM thread for the mounted volume */
 492 int dlm_launch_thread(struct dlm_ctxt *dlm)
 493 {
 494         mlog(0, "Starting dlm_thread...\n");
 495
 496         dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm-%s",
 497                         dlm->name);
 498         if (IS_ERR(dlm->dlm_thread_task)) {
 499                 mlog_errno(PTR_ERR(dlm->dlm_thread_task));
 500                 dlm->dlm_thread_task = NULL;
 501                 return -EINVAL;
 502         }
 503
 504         return 0;
 505 }
 506
 507 void dlm_complete_thread(struct dlm_ctxt *dlm)
 508 {
 509         if (dlm->dlm_thread_task) {
 510                 mlog(ML_KTHREAD, "Waiting for dlm thread to exit\n");
 511                 kthread_stop(dlm->dlm_thread_task);
 512                 dlm->dlm_thread_task = NULL;
 513         }
 514 }
 515
 516 static int dlm_dirty_list_empty(struct dlm_ctxt *dlm)
 517 {
 518         int empty;
 519
 520         spin_lock(&dlm->spinlock);
 521         empty = list_empty(&dlm->dirty_list);
 522         spin_unlock(&dlm->spinlock);
 523
 524         return empty;
 525 }
 526
 527 static void dlm_flush_asts(struct dlm_ctxt *dlm)
 528 {
 529         int ret;
 530         struct dlm_lock *lock;
 531         struct dlm_lock_resource *res;
 532         u8 hi;
 533
 534         spin_lock(&dlm->ast_lock);
 535         while (!list_empty(&dlm->pending_asts)) {
 536                 lock = list_entry(dlm->pending_asts.next,
 537                                   struct dlm_lock, ast_list);
 538                 /* get an extra ref on lock */
 539                 dlm_lock_get(lock);
 540                 res = lock->lockres;
 541                 mlog(0, "%s: res %.*s, Flush AST for lock %u:%llu, type %d, "
 542                      "node %u\n", dlm->name, res->lockname.len,
 543                      res->lockname.name,
 544                      dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
 545                      dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
 546                      lock->ml.type, lock->ml.node);
 547
 548                 BUG_ON(!lock->ast_pending);
 549
 550                 /* remove from list (including ref) */
 551                 list_del_init(&lock->ast_list);
 552                 dlm_lock_put(lock);
 553                 spin_unlock(&dlm->ast_lock);
 554
 555                 if (lock->ml.node != dlm->node_num) {
 556                         ret = dlm_do_remote_ast(dlm, res, lock);
 557                         if (ret < 0)
 558                                 mlog_errno(ret);
 559                 } else
 560                         dlm_do_local_ast(dlm, res, lock);
 561
 562                 spin_lock(&dlm->ast_lock);
 563
 564                 /* possible that another ast was queued while
 565                  * we were delivering the last one */
 566                 if (!list_empty(&lock->ast_list)) {
 567                         mlog(0, "%s: res %.*s, AST queued while flushing last "
 568                              "one\n", dlm->name, res->lockname.len,
 569                              res->lockname.name);
 570                 } else
 571                         lock->ast_pending = 0;
 572
 573                 /* drop the extra ref.
 574                  * this may drop it completely. */
 575                 dlm_lock_put(lock);
 576                 dlm_lockres_release_ast(dlm, res);
 577         }
 578
 579         while (!list_empty(&dlm->pending_basts)) {
 580                 lock = list_entry(dlm->pending_basts.next,
 581                                   struct dlm_lock, bast_list);
 582                 /* get an extra ref on lock */
 583                 dlm_lock_get(lock);
 584                 res = lock->lockres;
 585
 586                 BUG_ON(!lock->bast_pending);
 587
 588                 /* get the highest blocked lock, and reset */
 589                 spin_lock(&lock->spinlock);
 590                 BUG_ON(lock->ml.highest_blocked <= LKM_IVMODE);
 591                 hi = lock->ml.highest_blocked;
 592                 lock->ml.highest_blocked = LKM_IVMODE;
 593                 spin_unlock(&lock->spinlock);
 594
 595                 /* remove from list (including ref) */
 596                 list_del_init(&lock->bast_list);
 597                 dlm_lock_put(lock);
 598                 spin_unlock(&dlm->ast_lock);
 599
 600                 mlog(0, "%s: res %.*s, Flush BAST for lock %u:%llu, "
 601                      "blocked %d, node %u\n",
 602                      dlm->name, res->lockname.len, res->lockname.name,
 603                      dlm_get_lock_cookie_node(be64_to_cpu(lock->ml.cookie)),
 604                      dlm_get_lock_cookie_seq(be64_to_cpu(lock->ml.cookie)),
 605                      hi, lock->ml.node);
 606
 607                 if (lock->ml.node != dlm->node_num) {
 608                         ret = dlm_send_proxy_bast(dlm, res, lock, hi);
 609                         if (ret < 0)
 610                                 mlog_errno(ret);
 611                 } else
 612                         dlm_do_local_bast(dlm, res, lock, hi);
 613
 614                 spin_lock(&dlm->ast_lock);
 615
 616                 /* possible that another bast was queued while
 617                  * we were delivering the last one */
 618                 if (!list_empty(&lock->bast_list)) {
 619                         mlog(0, "%s: res %.*s, BAST queued while flushing last "
 620                              "one\n", dlm->name, res->lockname.len,
 621                              res->lockname.name);
 622                 } else
 623                         lock->bast_pending = 0;
 624
 625                 /* drop the extra ref.
 626                  * this may drop it completely. */
 627                 dlm_lock_put(lock);
 628                 dlm_lockres_release_ast(dlm, res);
 629         }
 630         wake_up(&dlm->ast_wq);
 631         spin_unlock(&dlm->ast_lock);
 632 }
 633
 634
 635 #define DLM_THREAD_TIMEOUT_MS (4 * 1000)
 636 #define DLM_THREAD_MAX_DIRTY  100
 637 #define DLM_THREAD_MAX_ASTS   10
 638
 639 static int dlm_thread(void *data)
 640 {
 641         struct dlm_lock_resource *res;
 642         struct dlm_ctxt *dlm = data;
 643         unsigned long timeout = msecs_to_jiffies(DLM_THREAD_TIMEOUT_MS);
 644
 645         mlog(0, "dlm thread running for %s...\n", dlm->name);
 646
 647         while (!kthread_should_stop()) {
 648                 int n = DLM_THREAD_MAX_DIRTY;
 649
 650                 /* dlm_shutting_down is very point-in-time, but that
 651                  * doesn't matter as we'll just loop back around if we
 652                  * get false on the leading edge of a state
 653                  * transition. */
 654                 dlm_run_purge_list(dlm, dlm_shutting_down(dlm));
 655
 656                 /* We really don't want to hold dlm->spinlock while
 657                  * calling dlm_shuffle_lists on each lockres that
 658                  * needs to have its queues adjusted and AST/BASTs
 659                  * run.  So let's pull each entry off the dirty_list
 660                  * and drop dlm->spinlock ASAP.  Once off the list,
 661                  * res->spinlock needs to be taken again to protect
 662                  * the queues while calling dlm_shuffle_lists.  */
 663                 spin_lock(&dlm->spinlock);
 664                 while (!list_empty(&dlm->dirty_list)) {
 665                         int delay = 0;
 666                         res = list_entry(dlm->dirty_list.next,
 667                                          struct dlm_lock_resource, dirty);
 668
 669                         /* peel a lockres off, remove it from the list,
 670                          * unset the dirty flag and drop the dlm lock */
 671                         BUG_ON(!res);
 672                         dlm_lockres_get(res);
 673
 674                         spin_lock(&res->spinlock);
 675                         /* We clear the DLM_LOCK_RES_DIRTY state once we shuffle lists below */
 676                         list_del_init(&res->dirty);
 677                         spin_unlock(&res->spinlock);
 678                         spin_unlock(&dlm->spinlock);
 679                         /* Drop dirty_list ref */
 680                         dlm_lockres_put(res);
 681
 682                         /* lockres can be re-dirtied/re-added to the
 683                          * dirty_list in this gap, but that is ok */
 684
 685                         spin_lock(&dlm->ast_lock);
 686                         spin_lock(&res->spinlock);
 687                         if (res->owner != dlm->node_num) {
 688                                 __dlm_print_one_lock_resource(res);
 689                                 mlog(ML_ERROR, "%s: inprog %d, mig %d, reco %d,"
 690                                      " dirty %d\n", dlm->name,
 691                                      !!(res->state & DLM_LOCK_RES_IN_PROGRESS),
 692                                      !!(res->state & DLM_LOCK_RES_MIGRATING),
 693                                      !!(res->state & DLM_LOCK_RES_RECOVERING),
 694                                      !!(res->state & DLM_LOCK_RES_DIRTY));
 695                         }
 696                         BUG_ON(res->owner != dlm->node_num);
 697
 698                         /* it is now ok to move lockreses in these states
 699                          * to the dirty list, assuming that they will only be
 700                          * dirty for a short while. */
 701                         BUG_ON(res->state & DLM_LOCK_RES_MIGRATING);
 702                         if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
 703                                           DLM_LOCK_RES_RECOVERING)) {
 704                                 /* move it to the tail and keep going */
 705                                 res->state &= ~DLM_LOCK_RES_DIRTY;
 706                                 spin_unlock(&res->spinlock);
 707                                 spin_unlock(&dlm->ast_lock);
 708                                 mlog(0, "%s: res %.*s, inprogress, delay list "
 709                                      "shuffle, state %d\n", dlm->name,
 710                                      res->lockname.len, res->lockname.name,
 711                                      res->state);
 712                                 delay = 1;
 713                                 goto in_progress;
 714                         }
 715
 716                         /* at this point the lockres is not migrating/
 717                          * recovering/in-progress.  we have the lockres
 718                          * spinlock and do NOT have the dlm lock.
 719                          * safe to reserve/queue asts and run the lists. */
 720
 721                         /* called while holding lockres lock */
 722                         dlm_shuffle_lists(dlm, res);
 723                         res->state &= ~DLM_LOCK_RES_DIRTY;
 724                         spin_unlock(&res->spinlock);
 725                         spin_unlock(&dlm->ast_lock);
 726
 727                         dlm_lockres_calc_usage(dlm, res);
 728
 729 in_progress:
 730
 731                         spin_lock(&dlm->spinlock);
 732                         /* if the lock was in-progress, stick
 733                          * it on the back of the list */
 734                         if (delay) {
 735                                 spin_lock(&res->spinlock);
 736                                 __dlm_dirty_lockres(dlm, res);
 737                                 spin_unlock(&res->spinlock);
 738                         }
 739                         dlm_lockres_put(res);
 740
 741                         /* unlikely, but we may need to give time to
 742                          * other tasks */
 743                         if (!--n) {
 744                                 mlog(0, "%s: Throttling dlm thread\n",
 745                                      dlm->name);
 746                                 break;
 747                         }
 748                 }
 749
 750                 spin_unlock(&dlm->spinlock);
 751                 dlm_flush_asts(dlm);
 752
 753                 /* yield and continue right away if there is more work to do */
 754                 if (!n) {
 755                         cond_resched();
 756                         continue;
 757                 }
 758
 759                 wait_event_interruptible_timeout(dlm->dlm_thread_wq,
 760                                                  !dlm_dirty_list_empty(dlm) ||
 761                                                  kthread_should_stop(),
 762                                                  timeout);
 763         }
 764
 765         mlog(0, "quitting DLM thread\n");
 766         return 0;
 767 }