dlmthread.c 20.1 KB
Newer Older
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41
/* -*- mode: c; c-basic-offset: 8; -*-
 * vim: noexpandtab sw=8 ts=8 sts=0:
 *
 * dlmthread.c
 *
 * standalone DLM module
 *
 * Copyright (C) 2004 Oracle.  All rights reserved.
 *
 * This program is free software; you can redistribute it and/or
 * modify it under the terms of the GNU General Public
 * License as published by the Free Software Foundation; either
 * version 2 of the License, or (at your option) any later version.
 *
 * This program is distributed in the hope that it will be useful,
 * but WITHOUT ANY WARRANTY; without even the implied warranty of
 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 * General Public License for more details.
 *
 * You should have received a copy of the GNU General Public
 * License along with this program; if not, write to the
 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
 * Boston, MA 021110-1307, USA.
 *
 */


#include <linux/module.h>
#include <linux/fs.h>
#include <linux/types.h>
#include <linux/slab.h>
#include <linux/highmem.h>
#include <linux/utsname.h>
#include <linux/init.h>
#include <linux/sysctl.h>
#include <linux/random.h>
#include <linux/blkdev.h>
#include <linux/socket.h>
#include <linux/inet.h>
#include <linux/timer.h>
#include <linux/kthread.h>
42
#include <linux/delay.h>
43 44 45 46 47 48 49 50 51 52 53 54 55 56


#include "cluster/heartbeat.h"
#include "cluster/nodemanager.h"
#include "cluster/tcp.h"

#include "dlmapi.h"
#include "dlmcommon.h"
#include "dlmdomain.h"

#define MLOG_MASK_PREFIX (ML_DLM|ML_DLM_THREAD)
#include "cluster/masklog.h"

static int dlm_thread(void *data);
K
Kurt Hackel 已提交
57 58
static void dlm_purge_lockres_now(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *lockres);
59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85

static void dlm_flush_asts(struct dlm_ctxt *dlm);

#define dlm_lock_is_remote(dlm, lock)     ((lock)->ml.node != (dlm)->node_num)

/* will exit holding res->spinlock, but may drop in function */
/* waits until flags are cleared on res->state */
void __dlm_wait_on_lockres_flags(struct dlm_lock_resource *res, int flags)
{
	DECLARE_WAITQUEUE(wait, current);

	assert_spin_locked(&res->spinlock);

	add_wait_queue(&res->wq, &wait);
repeat:
	set_current_state(TASK_UNINTERRUPTIBLE);
	if (res->state & flags) {
		spin_unlock(&res->spinlock);
		schedule();
		spin_lock(&res->spinlock);
		goto repeat;
	}
	remove_wait_queue(&res->wq, &wait);
	current->state = TASK_RUNNING;
}


86
int __dlm_lockres_unused(struct dlm_lock_resource *res)
87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108
{
	if (list_empty(&res->granted) &&
	    list_empty(&res->converting) &&
	    list_empty(&res->blocked) &&
	    list_empty(&res->dirty))
		return 1;
	return 0;
}


/* Call whenever you may have added or deleted something from one of
 * the lockres queue's. This will figure out whether it belongs on the
 * unused list or not and does the appropriate thing. */
void __dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
			      struct dlm_lock_resource *res)
{
	mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);

	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&res->spinlock);

	if (__dlm_lockres_unused(res)){
109 110 111 112 113 114 115 116 117 118 119 120 121 122
		/* For now, just keep any resource we master */
		if (res->owner == dlm->node_num)
		{
			if (!list_empty(&res->purge)) {
				mlog(0, "we master %s:%.*s, but it is on "
				     "the purge list.  Removing\n",
				     dlm->name, res->lockname.len,
				     res->lockname.name);
				list_del_init(&res->purge);
				dlm->purge_count--;
			}
			return;
		}

123 124 125 126 127 128 129
		if (list_empty(&res->purge)) {
			mlog(0, "putting lockres %.*s from purge list\n",
			     res->lockname.len, res->lockname.name);

			res->last_used = jiffies;
			list_add_tail(&res->purge, &dlm->purge_list);
			dlm->purge_count++;
K
Kurt Hackel 已提交
130 131 132 133 134 135 136 137 138 139 140 141

			/* if this node is not the owner, there is
			 * no way to keep track of who the owner could be.
			 * unhash it to avoid serious problems. */
			if (res->owner != dlm->node_num) {
				mlog(0, "%s:%.*s: doing immediate "
				     "purge of lockres owned by %u\n",
				     dlm->name, res->lockname.len,
				     res->lockname.name, res->owner);

				dlm_purge_lockres_now(dlm, res);
			}
142 143
		}
	} else if (!list_empty(&res->purge)) {
K
Kurt Hackel 已提交
144 145 146
		mlog(0, "removing lockres %.*s from purge list, "
		     "owner=%u\n", res->lockname.len, res->lockname.name,
		     res->owner);
147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197

		list_del_init(&res->purge);
		dlm->purge_count--;
	}
}

void dlm_lockres_calc_usage(struct dlm_ctxt *dlm,
			    struct dlm_lock_resource *res)
{
	mlog_entry("%.*s\n", res->lockname.len, res->lockname.name);
	spin_lock(&dlm->spinlock);
	spin_lock(&res->spinlock);

	__dlm_lockres_calc_usage(dlm, res);

	spin_unlock(&res->spinlock);
	spin_unlock(&dlm->spinlock);
}

/* TODO: Eventual API: Called with the dlm spinlock held, may drop it
 * to do migration, but will re-acquire before exit. */
void dlm_purge_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *lockres)
{
	int master;
	int ret;

	spin_lock(&lockres->spinlock);
	master = lockres->owner == dlm->node_num;
	spin_unlock(&lockres->spinlock);

	mlog(0, "purging lockres %.*s, master = %d\n", lockres->lockname.len,
	     lockres->lockname.name, master);

	/* Non master is the easy case -- no migration required, just
	 * quit. */
	if (!master)
		goto finish;

	/* Wheee! Migrate lockres here! */
	spin_unlock(&dlm->spinlock);
again:

	ret = dlm_migrate_lockres(dlm, lockres, O2NM_MAX_NODES);
	if (ret == -ENOTEMPTY) {
		mlog(ML_ERROR, "lockres %.*s still has local locks!\n",
		     lockres->lockname.len, lockres->lockname.name);

		BUG();
	} else if (ret < 0) {
		mlog(ML_NOTICE, "lockres %.*s: migrate failed, retrying\n",
		     lockres->lockname.len, lockres->lockname.name);
198
		msleep(100);
199 200 201 202 203 204 205 206 207 208 209 210 211
		goto again;
	}

	spin_lock(&dlm->spinlock);

finish:
	if (!list_empty(&lockres->purge)) {
		list_del_init(&lockres->purge);
		dlm->purge_count--;
	}
	__dlm_unhash_lockres(lockres);
}

K
Kurt Hackel 已提交
212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229
/* make an unused lockres go away immediately.
 * as soon as the dlm spinlock is dropped, this lockres
 * will not be found. kfree still happens on last put. */
static void dlm_purge_lockres_now(struct dlm_ctxt *dlm,
				  struct dlm_lock_resource *lockres)
{
	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&lockres->spinlock);

	BUG_ON(!__dlm_lockres_unused(lockres));

	if (!list_empty(&lockres->purge)) {
		list_del_init(&lockres->purge);
		dlm->purge_count--;
	}
	__dlm_unhash_lockres(lockres);
}

230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 251 252 253 254 255 256 257 258 259 260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283 284 285 286 287 288 289 290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306 307 308 309 310 311 312 313 314 315 316 317 318 319 320 321 322 323 324 325 326 327 328 329 330 331 332 333 334 335 336 337 338 339 340 341 342 343 344 345 346 347 348 349 350 351 352 353 354 355 356 357 358 359 360 361 362 363 364 365 366 367 368 369
static void dlm_run_purge_list(struct dlm_ctxt *dlm,
			       int purge_now)
{
	unsigned int run_max, unused;
	unsigned long purge_jiffies;
	struct dlm_lock_resource *lockres;

	spin_lock(&dlm->spinlock);
	run_max = dlm->purge_count;

	while(run_max && !list_empty(&dlm->purge_list)) {
		run_max--;

		lockres = list_entry(dlm->purge_list.next,
				     struct dlm_lock_resource, purge);

		/* Status of the lockres *might* change so double
		 * check. If the lockres is unused, holding the dlm
		 * spinlock will prevent people from getting and more
		 * refs on it -- there's no need to keep the lockres
		 * spinlock. */
		spin_lock(&lockres->spinlock);
		unused = __dlm_lockres_unused(lockres);
		spin_unlock(&lockres->spinlock);

		if (!unused)
			continue;

		purge_jiffies = lockres->last_used +
			msecs_to_jiffies(DLM_PURGE_INTERVAL_MS);

		/* Make sure that we want to be processing this guy at
		 * this time. */
		if (!purge_now && time_after(purge_jiffies, jiffies)) {
			/* Since resources are added to the purge list
			 * in tail order, we can stop at the first
			 * unpurgable resource -- anyone added after
			 * him will have a greater last_used value */
			break;
		}

		list_del_init(&lockres->purge);
		dlm->purge_count--;

		/* This may drop and reacquire the dlm spinlock if it
		 * has to do migration. */
		mlog(0, "calling dlm_purge_lockres!\n");
		dlm_purge_lockres(dlm, lockres);
		mlog(0, "DONE calling dlm_purge_lockres!\n");

		/* Avoid adding any scheduling latencies */
		cond_resched_lock(&dlm->spinlock);
	}

	spin_unlock(&dlm->spinlock);
}

static void dlm_shuffle_lists(struct dlm_ctxt *dlm,
			      struct dlm_lock_resource *res)
{
	struct dlm_lock *lock, *target;
	struct list_head *iter;
	struct list_head *head;
	int can_grant = 1;

	//mlog(0, "res->lockname.len=%d\n", res->lockname.len);
	//mlog(0, "res->lockname.name=%p\n", res->lockname.name);
	//mlog(0, "shuffle res %.*s\n", res->lockname.len,
	//	  res->lockname.name);

	/* because this function is called with the lockres
	 * spinlock, and because we know that it is not migrating/
	 * recovering/in-progress, it is fine to reserve asts and
	 * basts right before queueing them all throughout */
	assert_spin_locked(&res->spinlock);
	BUG_ON((res->state & (DLM_LOCK_RES_MIGRATING|
			      DLM_LOCK_RES_RECOVERING|
			      DLM_LOCK_RES_IN_PROGRESS)));

converting:
	if (list_empty(&res->converting))
		goto blocked;
	mlog(0, "res %.*s has locks on a convert queue\n", res->lockname.len,
	     res->lockname.name);

	target = list_entry(res->converting.next, struct dlm_lock, list);
	if (target->ml.convert_type == LKM_IVMODE) {
		mlog(ML_ERROR, "%.*s: converting a lock with no "
		     "convert_type!\n", res->lockname.len, res->lockname.name);
		BUG();
	}
	head = &res->granted;
	list_for_each(iter, head) {
		lock = list_entry(iter, struct dlm_lock, list);
		if (lock==target)
			continue;
		if (!dlm_lock_compatible(lock->ml.type,
					 target->ml.convert_type)) {
			can_grant = 0;
			/* queue the BAST if not already */
			if (lock->ml.highest_blocked == LKM_IVMODE) {
				__dlm_lockres_reserve_ast(res);
				dlm_queue_bast(dlm, lock);
			}
			/* update the highest_blocked if needed */
			if (lock->ml.highest_blocked < target->ml.convert_type)
				lock->ml.highest_blocked =
					target->ml.convert_type;
		}
	}
	head = &res->converting;
	list_for_each(iter, head) {
		lock = list_entry(iter, struct dlm_lock, list);
		if (lock==target)
			continue;
		if (!dlm_lock_compatible(lock->ml.type,
					 target->ml.convert_type)) {
			can_grant = 0;
			if (lock->ml.highest_blocked == LKM_IVMODE) {
				__dlm_lockres_reserve_ast(res);
				dlm_queue_bast(dlm, lock);
			}
			if (lock->ml.highest_blocked < target->ml.convert_type)
				lock->ml.highest_blocked =
					target->ml.convert_type;
		}
	}

	/* we can convert the lock */
	if (can_grant) {
		spin_lock(&target->spinlock);
		BUG_ON(target->ml.highest_blocked != LKM_IVMODE);

		mlog(0, "calling ast for converting lock: %.*s, have: %d, "
		     "granting: %d, node: %u\n", res->lockname.len,
		     res->lockname.name, target->ml.type,
		     target->ml.convert_type, target->ml.node);

		target->ml.type = target->ml.convert_type;
		target->ml.convert_type = LKM_IVMODE;
A
Akinobu Mita 已提交
370
		list_move_tail(&target->list, &res->granted);
371 372 373 374 375 376 377 378 379 380 381 382 383 384 385 386 387 388 389 390 391 392 393 394 395 396 397 398 399 400 401 402 403 404 405 406 407 408 409 410 411 412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430

		BUG_ON(!target->lksb);
		target->lksb->status = DLM_NORMAL;

		spin_unlock(&target->spinlock);

		__dlm_lockres_reserve_ast(res);
		dlm_queue_ast(dlm, target);
		/* go back and check for more */
		goto converting;
	}

blocked:
	if (list_empty(&res->blocked))
		goto leave;
	target = list_entry(res->blocked.next, struct dlm_lock, list);

	head = &res->granted;
	list_for_each(iter, head) {
		lock = list_entry(iter, struct dlm_lock, list);
		if (lock==target)
			continue;
		if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
			can_grant = 0;
			if (lock->ml.highest_blocked == LKM_IVMODE) {
				__dlm_lockres_reserve_ast(res);
				dlm_queue_bast(dlm, lock);
			}
			if (lock->ml.highest_blocked < target->ml.type)
				lock->ml.highest_blocked = target->ml.type;
		}
	}

	head = &res->converting;
	list_for_each(iter, head) {
		lock = list_entry(iter, struct dlm_lock, list);
		if (lock==target)
			continue;
		if (!dlm_lock_compatible(lock->ml.type, target->ml.type)) {
			can_grant = 0;
			if (lock->ml.highest_blocked == LKM_IVMODE) {
				__dlm_lockres_reserve_ast(res);
				dlm_queue_bast(dlm, lock);
			}
			if (lock->ml.highest_blocked < target->ml.type)
				lock->ml.highest_blocked = target->ml.type;
		}
	}

	/* we can grant the blocked lock (only
	 * possible if converting list empty) */
	if (can_grant) {
		spin_lock(&target->spinlock);
		BUG_ON(target->ml.highest_blocked != LKM_IVMODE);

		mlog(0, "calling ast for blocked lock: %.*s, granting: %d, "
		     "node: %u\n", res->lockname.len, res->lockname.name,
		     target->ml.type, target->ml.node);

		// target->ml.type is already correct
A
Akinobu Mita 已提交
431
		list_move_tail(&target->list, &res->granted);
432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471

		BUG_ON(!target->lksb);
		target->lksb->status = DLM_NORMAL;

		spin_unlock(&target->spinlock);

		__dlm_lockres_reserve_ast(res);
		dlm_queue_ast(dlm, target);
		/* go back and check for more */
		goto converting;
	}

leave:
	return;
}

/* must have NO locks when calling this with res !=NULL * */
void dlm_kick_thread(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
{
	mlog_entry("dlm=%p, res=%p\n", dlm, res);
	if (res) {
		spin_lock(&dlm->spinlock);
		spin_lock(&res->spinlock);
		__dlm_dirty_lockres(dlm, res);
		spin_unlock(&res->spinlock);
		spin_unlock(&dlm->spinlock);
	}
	wake_up(&dlm->dlm_thread_wq);
}

void __dlm_dirty_lockres(struct dlm_ctxt *dlm, struct dlm_lock_resource *res)
{
	mlog_entry("dlm=%p, res=%p\n", dlm, res);

	assert_spin_locked(&dlm->spinlock);
	assert_spin_locked(&res->spinlock);

	/* don't shuffle secondary queues */
	if ((res->owner == dlm->node_num) &&
	    !(res->state & DLM_LOCK_RES_DIRTY)) {
472 473
		/* ref for dirty_list */
		dlm_lockres_get(res);
474 475 476 477 478 479 480 481 482 483 484 485 486 487 488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539 540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595 596 597 598 599 600 601 602 603 604 605 606 607 608 609 610 611 612 613 614 615 616 617 618 619 620 621 622 623 624 625 626 627 628 629 630 631 632 633 634 635 636 637 638 639 640 641 642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657
		list_add_tail(&res->dirty, &dlm->dirty_list);
		res->state |= DLM_LOCK_RES_DIRTY;
	}
}


/* Launch the NM thread for the mounted volume */
int dlm_launch_thread(struct dlm_ctxt *dlm)
{
	mlog(0, "starting dlm thread...\n");

	dlm->dlm_thread_task = kthread_run(dlm_thread, dlm, "dlm_thread");
	if (IS_ERR(dlm->dlm_thread_task)) {
		mlog_errno(PTR_ERR(dlm->dlm_thread_task));
		dlm->dlm_thread_task = NULL;
		return -EINVAL;
	}

	return 0;
}

void dlm_complete_thread(struct dlm_ctxt *dlm)
{
	if (dlm->dlm_thread_task) {
		mlog(ML_KTHREAD, "waiting for dlm thread to exit\n");
		kthread_stop(dlm->dlm_thread_task);
		dlm->dlm_thread_task = NULL;
	}
}

static int dlm_dirty_list_empty(struct dlm_ctxt *dlm)
{
	int empty;

	spin_lock(&dlm->spinlock);
	empty = list_empty(&dlm->dirty_list);
	spin_unlock(&dlm->spinlock);

	return empty;
}

static void dlm_flush_asts(struct dlm_ctxt *dlm)
{
	int ret;
	struct dlm_lock *lock;
	struct dlm_lock_resource *res;
	u8 hi;

	spin_lock(&dlm->ast_lock);
	while (!list_empty(&dlm->pending_asts)) {
		lock = list_entry(dlm->pending_asts.next,
				  struct dlm_lock, ast_list);
		/* get an extra ref on lock */
		dlm_lock_get(lock);
		res = lock->lockres;
		mlog(0, "delivering an ast for this lockres\n");

		BUG_ON(!lock->ast_pending);

		/* remove from list (including ref) */
		list_del_init(&lock->ast_list);
		dlm_lock_put(lock);
		spin_unlock(&dlm->ast_lock);

		if (lock->ml.node != dlm->node_num) {
			ret = dlm_do_remote_ast(dlm, res, lock);
			if (ret < 0)
				mlog_errno(ret);
		} else
			dlm_do_local_ast(dlm, res, lock);

		spin_lock(&dlm->ast_lock);

		/* possible that another ast was queued while
		 * we were delivering the last one */
		if (!list_empty(&lock->ast_list)) {
			mlog(0, "aha another ast got queued while "
			     "we were finishing the last one.  will "
			     "keep the ast_pending flag set.\n");
		} else
			lock->ast_pending = 0;

		/* drop the extra ref.
		 * this may drop it completely. */
		dlm_lock_put(lock);
		dlm_lockres_release_ast(dlm, res);
	}

	while (!list_empty(&dlm->pending_basts)) {
		lock = list_entry(dlm->pending_basts.next,
				  struct dlm_lock, bast_list);
		/* get an extra ref on lock */
		dlm_lock_get(lock);
		res = lock->lockres;

		BUG_ON(!lock->bast_pending);

		/* get the highest blocked lock, and reset */
		spin_lock(&lock->spinlock);
		BUG_ON(lock->ml.highest_blocked <= LKM_IVMODE);
		hi = lock->ml.highest_blocked;
		lock->ml.highest_blocked = LKM_IVMODE;
		spin_unlock(&lock->spinlock);

		/* remove from list (including ref) */
		list_del_init(&lock->bast_list);
		dlm_lock_put(lock);
		spin_unlock(&dlm->ast_lock);

		mlog(0, "delivering a bast for this lockres "
		     "(blocked = %d\n", hi);

		if (lock->ml.node != dlm->node_num) {
			ret = dlm_send_proxy_bast(dlm, res, lock, hi);
			if (ret < 0)
				mlog_errno(ret);
		} else
			dlm_do_local_bast(dlm, res, lock, hi);

		spin_lock(&dlm->ast_lock);

		/* possible that another bast was queued while
		 * we were delivering the last one */
		if (!list_empty(&lock->bast_list)) {
			mlog(0, "aha another bast got queued while "
			     "we were finishing the last one.  will "
			     "keep the bast_pending flag set.\n");
		} else
			lock->bast_pending = 0;

		/* drop the extra ref.
		 * this may drop it completely. */
		dlm_lock_put(lock);
		dlm_lockres_release_ast(dlm, res);
	}
	wake_up(&dlm->ast_wq);
	spin_unlock(&dlm->ast_lock);
}


#define DLM_THREAD_TIMEOUT_MS (4 * 1000)
#define DLM_THREAD_MAX_DIRTY  100
#define DLM_THREAD_MAX_ASTS   10

static int dlm_thread(void *data)
{
	struct dlm_lock_resource *res;
	struct dlm_ctxt *dlm = data;
	unsigned long timeout = msecs_to_jiffies(DLM_THREAD_TIMEOUT_MS);

	mlog(0, "dlm thread running for %s...\n", dlm->name);

	while (!kthread_should_stop()) {
		int n = DLM_THREAD_MAX_DIRTY;

		/* dlm_shutting_down is very point-in-time, but that
		 * doesn't matter as we'll just loop back around if we
		 * get false on the leading edge of a state
		 * transition. */
		dlm_run_purge_list(dlm, dlm_shutting_down(dlm));

		/* We really don't want to hold dlm->spinlock while
		 * calling dlm_shuffle_lists on each lockres that
		 * needs to have its queues adjusted and AST/BASTs
		 * run.  So let's pull each entry off the dirty_list
		 * and drop dlm->spinlock ASAP.  Once off the list,
		 * res->spinlock needs to be taken again to protect
		 * the queues while calling dlm_shuffle_lists.  */
		spin_lock(&dlm->spinlock);
		while (!list_empty(&dlm->dirty_list)) {
			int delay = 0;
			res = list_entry(dlm->dirty_list.next,
					 struct dlm_lock_resource, dirty);

			/* peel a lockres off, remove it from the list,
			 * unset the dirty flag and drop the dlm lock */
			BUG_ON(!res);
			dlm_lockres_get(res);

			spin_lock(&res->spinlock);
			res->state &= ~DLM_LOCK_RES_DIRTY;
			list_del_init(&res->dirty);
			spin_unlock(&res->spinlock);
			spin_unlock(&dlm->spinlock);
658 659
			/* Drop dirty_list ref */
			dlm_lockres_put(res);
660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684 685 686 687 688 689 690 691 692 693 694 695

		 	/* lockres can be re-dirtied/re-added to the
			 * dirty_list in this gap, but that is ok */

			spin_lock(&res->spinlock);
			if (res->owner != dlm->node_num) {
				__dlm_print_one_lock_resource(res);
				mlog(ML_ERROR, "inprog:%s, mig:%s, reco:%s, dirty:%s\n",
				     res->state & DLM_LOCK_RES_IN_PROGRESS ? "yes" : "no",
				     res->state & DLM_LOCK_RES_MIGRATING ? "yes" : "no",
				     res->state & DLM_LOCK_RES_RECOVERING ? "yes" : "no",
				     res->state & DLM_LOCK_RES_DIRTY ? "yes" : "no");
			}
			BUG_ON(res->owner != dlm->node_num);

			/* it is now ok to move lockreses in these states
			 * to the dirty list, assuming that they will only be
			 * dirty for a short while. */
			if (res->state & (DLM_LOCK_RES_IN_PROGRESS |
					  DLM_LOCK_RES_MIGRATING |
					  DLM_LOCK_RES_RECOVERING)) {
				/* move it to the tail and keep going */
				spin_unlock(&res->spinlock);
				mlog(0, "delaying list shuffling for in-"
				     "progress lockres %.*s, state=%d\n",
				     res->lockname.len, res->lockname.name,
				     res->state);
				delay = 1;
				goto in_progress;
			}

			/* at this point the lockres is not migrating/
			 * recovering/in-progress.  we have the lockres
			 * spinlock and do NOT have the dlm lock.
			 * safe to reserve/queue asts and run the lists. */

696 697 698
			mlog(0, "calling dlm_shuffle_lists with dlm=%s, "
			     "res=%.*s\n", dlm->name,
			     res->lockname.len, res->lockname.name);
699 700 701 702 703 704 705 706 707 708 709 710 711

			/* called while holding lockres lock */
			dlm_shuffle_lists(dlm, res);
			spin_unlock(&res->spinlock);

			dlm_lockres_calc_usage(dlm, res);

in_progress:

			spin_lock(&dlm->spinlock);
			/* if the lock was in-progress, stick
			 * it on the back of the list */
			if (delay) {
712 713
				/* ref for dirty_list */
				dlm_lockres_get(res);
714 715 716 717 718 719 720 721 722 723 724 725 726 727 728 729 730 731 732 733 734 735 736 737 738 739 740 741 742 743 744 745 746
				spin_lock(&res->spinlock);
				list_add_tail(&res->dirty, &dlm->dirty_list);
				res->state |= DLM_LOCK_RES_DIRTY;
				spin_unlock(&res->spinlock);
			}
			dlm_lockres_put(res);

			/* unlikely, but we may need to give time to
			 * other tasks */
			if (!--n) {
				mlog(0, "throttling dlm_thread\n");
				break;
			}
		}

		spin_unlock(&dlm->spinlock);
		dlm_flush_asts(dlm);

		/* yield and continue right away if there is more work to do */
		if (!n) {
			yield();
			continue;
		}

		wait_event_interruptible_timeout(dlm->dlm_thread_wq,
						 !dlm_dirty_list_empty(dlm) ||
						 kthread_should_stop(),
						 timeout);
	}

	mlog(0, "quitting DLM thread\n");
	return 0;
}