[linux-2.6-block.git] / fs / dlm / recoverd.c

// SPDX-License-Identifier: GPL-2.0-only
/******************************************************************************
*******************************************************************************
**
**  Copyright (C) Sistina Software, Inc.  1997-2003  All rights reserved.
**  Copyright (C) 2004-2011 Red Hat, Inc.  All rights reserved.
**
**
*******************************************************************************
******************************************************************************/

#include "dlm_internal.h"
#include "lockspace.h"
#include "member.h"
#include "dir.h"
#include "ast.h"
#include "recover.h"
#include "lowcomms.h"
#include "lock.h"
#include "requestqueue.h"
#include "recoverd.h"


/* If the start for which we're re-enabling locking (seq) has been superseded
   by a newer stop (ls_recover_seq), we need to leave locking disabled.

   We suspend dlm_recv threads here to avoid the race where dlm_recv a) sees
   locking stopped and b) adds a message to the requestqueue, but dlm_recoverd
   enables locking and clears the requestqueue between a and b. */

static int enable_locking(struct dlm_ls *ls, uint64_t seq)
{
	int error = -EINTR;

	down_write(&ls->ls_recv_active);

	spin_lock(&ls->ls_recover_lock);
	if (ls->ls_recover_seq == seq) {
		set_bit(LSFL_RUNNING, &ls->ls_flags);
		/* unblocks processes waiting to enter the dlm */
		up_write(&ls->ls_in_recovery);
		clear_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
		error = 0;
	}
	spin_unlock(&ls->ls_recover_lock);

	up_write(&ls->ls_recv_active);
	return error;
}

static int ls_recover(struct dlm_ls *ls, struct dlm_recover *rv)
{
	unsigned long start;
	int error, neg = 0;

	log_rinfo(ls, "dlm_recover %llu", (unsigned long long)rv->seq);

	mutex_lock(&ls->ls_recoverd_active);

	dlm_callback_suspend(ls);

	dlm_clear_toss(ls);

	/*
	 * This list of root rsb's will be the basis of most of the recovery
	 * routines.
	 */

	dlm_create_root_list(ls);

	/*
	 * Add or remove nodes from the lockspace's ls_nodes list.
	 *
	 * Due to the fact that we must report all membership changes to lsops
	 * or midcomms layer, it is not permitted to abort ls_recover() until
	 * this is done.
	 */

	error = dlm_recover_members(ls, rv, &neg);
	if (error) {
		log_rinfo(ls, "dlm_recover_members error %d", error);
		goto fail;
	}

	dlm_recover_dir_nodeid(ls);

	ls->ls_recover_dir_sent_res = 0;
	ls->ls_recover_dir_sent_msg = 0;
	ls->ls_recover_locks_in = 0;

	dlm_set_recover_status(ls, DLM_RS_NODES);

	error = dlm_recover_members_wait(ls, rv->seq);
	if (error) {
		log_rinfo(ls, "dlm_recover_members_wait error %d", error);
		goto fail;
	}

	start = jiffies;

	/*
	 * Rebuild our own share of the directory by collecting from all other
	 * nodes their master rsb names that hash to us.
	 */

	error = dlm_recover_directory(ls, rv->seq);
	if (error) {
		log_rinfo(ls, "dlm_recover_directory error %d", error);
		goto fail;
	}

	dlm_set_recover_status(ls, DLM_RS_DIR);

	error = dlm_recover_directory_wait(ls, rv->seq);
	if (error) {
		log_rinfo(ls, "dlm_recover_directory_wait error %d", error);
		goto fail;
	}

	log_rinfo(ls, "dlm_recover_directory %u out %u messages",
		  ls->ls_recover_dir_sent_res, ls->ls_recover_dir_sent_msg);

	/*
	 * We may have outstanding operations that are waiting for a reply from
	 * a failed node.  Mark these to be resent after recovery.  Unlock and
	 * cancel ops can just be completed.
	 */

	dlm_recover_waiters_pre(ls);

	if (dlm_recovery_stopped(ls)) {
		error = -EINTR;
		goto fail;
	}

	if (neg || dlm_no_directory(ls)) {
		/*
		 * Clear lkb's for departed nodes.
		 */

		dlm_recover_purge(ls);

		/*
		 * Get new master nodeid's for rsb's that were mastered on
		 * departed nodes.
		 */

		error = dlm_recover_masters(ls, rv->seq);
		if (error) {
			log_rinfo(ls, "dlm_recover_masters error %d", error);
			goto fail;
		}

		/*
		 * Send our locks on remastered rsb's to the new masters.
		 */

		error = dlm_recover_locks(ls, rv->seq);
		if (error) {
			log_rinfo(ls, "dlm_recover_locks error %d", error);
			goto fail;
		}

		dlm_set_recover_status(ls, DLM_RS_LOCKS);

		error = dlm_recover_locks_wait(ls, rv->seq);
		if (error) {
			log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
			goto fail;
		}

		log_rinfo(ls, "dlm_recover_locks %u in",
			  ls->ls_recover_locks_in);

		/*
		 * Finalize state in master rsb's now that all locks can be
		 * checked.  This includes conversion resolution and lvb
		 * settings.
		 */

		dlm_recover_rsbs(ls);
	} else {
		/*
		 * Other lockspace members may be going through the "neg" steps
		 * while also adding us to the lockspace, in which case they'll
		 * be doing the recover_locks (RS_LOCKS) barrier.
		 */
		dlm_set_recover_status(ls, DLM_RS_LOCKS);

		error = dlm_recover_locks_wait(ls, rv->seq);
		if (error) {
			log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
			goto fail;
		}
	}

	dlm_release_root_list(ls);

	/*
	 * Purge directory-related requests that are saved in requestqueue.
	 * All dir requests from before recovery are invalid now due to the dir
	 * rebuild and will be resent by the requesting nodes.
	 */

	dlm_purge_requestqueue(ls);

	dlm_set_recover_status(ls, DLM_RS_DONE);

	error = dlm_recover_done_wait(ls, rv->seq);
	if (error) {
		log_rinfo(ls, "dlm_recover_done_wait error %d", error);
		goto fail;
	}

	dlm_clear_members_gone(ls);

	dlm_callback_resume(ls);

	error = enable_locking(ls, rv->seq);
	if (error) {
		log_rinfo(ls, "enable_locking error %d", error);
		goto fail;
	}

	error = dlm_process_requestqueue(ls);
	if (error) {
		log_rinfo(ls, "dlm_process_requestqueue error %d", error);
		goto fail;
	}

	error = dlm_recover_waiters_post(ls);
	if (error) {
		log_rinfo(ls, "dlm_recover_waiters_post error %d", error);
		goto fail;
	}

	dlm_recover_grant(ls);

	log_rinfo(ls, "dlm_recover %llu generation %u done: %u ms",
		  (unsigned long long)rv->seq, ls->ls_generation,
		  jiffies_to_msecs(jiffies - start));
	mutex_unlock(&ls->ls_recoverd_active);

	return 0;

 fail:
	dlm_release_root_list(ls);
	mutex_unlock(&ls->ls_recoverd_active);

	return error;
}

/* The dlm_ls_start() that created the rv we take here may already have been
   stopped via dlm_ls_stop(); in that case we need to leave the RECOVERY_STOP
   flag set. */

static void do_ls_recovery(struct dlm_ls *ls)
{
	struct dlm_recover *rv = NULL;
	int error;

	spin_lock(&ls->ls_recover_lock);
	rv = ls->ls_recover_args;
	ls->ls_recover_args = NULL;
	if (rv && ls->ls_recover_seq == rv->seq)
		clear_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
	spin_unlock(&ls->ls_recover_lock);

	if (rv) {
		error = ls_recover(ls, rv);
		switch (error) {
		case 0:
			ls->ls_recovery_result = 0;
			complete(&ls->ls_recovery_done);

			dlm_lsop_recover_done(ls);
			break;
		case -EINTR:
			/* if recovery was interrupted -EINTR we wait for the next
			 * ls_recover() iteration until it hopefully succeeds.
			 */
			log_rinfo(ls, "%s %llu interrupted and should be queued to run again",
				  __func__, (unsigned long long)rv->seq);
			break;
		default:
			log_rinfo(ls, "%s %llu error %d", __func__,
				  (unsigned long long)rv->seq, error);

			/* let new_lockspace() get aware of critical error */
			ls->ls_recovery_result = error;
			complete(&ls->ls_recovery_done);
			break;
		}

		kfree(rv->nodes);
		kfree(rv);
	}
}

static int dlm_recoverd(void *arg)
{
	struct dlm_ls *ls;

	ls = dlm_find_lockspace_local(arg);
	if (!ls) {
		log_print("dlm_recoverd: no lockspace %p", arg);
		return -1;
	}

	down_write(&ls->ls_in_recovery);
	set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
	wake_up(&ls->ls_recover_lock_wait);

	while (1) {
		/*
		 * We call kthread_should_stop() after set_current_state().
		 * This is because it works correctly if kthread_stop() is
		 * called just before set_current_state().
		 */
		set_current_state(TASK_INTERRUPTIBLE);
		if (kthread_should_stop()) {
			set_current_state(TASK_RUNNING);
			break;
		}
		if (!test_bit(LSFL_RECOVER_WORK, &ls->ls_flags) &&
		    !test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
			if (kthread_should_stop())
				break;
			schedule();
		}
		set_current_state(TASK_RUNNING);

		if (test_and_clear_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
			down_write(&ls->ls_in_recovery);
			set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
			wake_up(&ls->ls_recover_lock_wait);
		}

		if (test_and_clear_bit(LSFL_RECOVER_WORK, &ls->ls_flags))
			do_ls_recovery(ls);
	}

	if (test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags))
		up_write(&ls->ls_in_recovery);

	dlm_put_lockspace(ls);
	return 0;
}

int dlm_recoverd_start(struct dlm_ls *ls)
{
	struct task_struct *p;
	int error = 0;

	p = kthread_run(dlm_recoverd, ls, "dlm_recoverd");
	if (IS_ERR(p))
		error = PTR_ERR(p);
	else
                ls->ls_recoverd_task = p;
	return error;
}

void dlm_recoverd_stop(struct dlm_ls *ls)
{
	kthread_stop(ls->ls_recoverd_task);
}

void dlm_recoverd_suspend(struct dlm_ls *ls)
{
	wake_up(&ls->ls_wait_general);
	mutex_lock(&ls->ls_recoverd_active);
}

void dlm_recoverd_resume(struct dlm_ls *ls)
{
	mutex_unlock(&ls->ls_recoverd_active);
}
Commit	Line	Data
2522fe45	1	// SPDX-License-Identifier: GPL-2.0-only
e7fd4179 DT	2	/******************************************************************************
	3	*******************************************************************************
	4	**
	5	** Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
60f98d18	6	** Copyright (C) 2004-2011 Red Hat, Inc. All rights reserved.
e7fd4179	7	**
e7fd4179 DT	8	**
	9	*******************************************************************************
	10	******************************************************************************/
	11
	12	#include "dlm_internal.h"
	13	#include "lockspace.h"
	14	#include "member.h"
	15	#include "dir.h"
	16	#include "ast.h"
	17	#include "recover.h"
	18	#include "lowcomms.h"
	19	#include "lock.h"
	20	#include "requestqueue.h"
	21	#include "recoverd.h"
	22
	23
	24	/* If the start for which we're re-enabling locking (seq) has been superseded
c36258b5 DT	25	by a newer stop (ls_recover_seq), we need to leave locking disabled.
	26
	27	We suspend dlm_recv threads here to avoid the race where dlm_recv a) sees
	28	locking stopped and b) adds a message to the requestqueue, but dlm_recoverd
	29	enables locking and clears the requestqueue between a and b. */
e7fd4179 DT	30
	31	static int enable_locking(struct dlm_ls *ls, uint64_t seq)
	32	{
	33	int error = -EINTR;
	34
c36258b5 DT	35	down_write(&ls->ls_recv_active);
c36258b5 DT	36
e7fd4179 DT	37	spin_lock(&ls->ls_recover_lock);
	38	if (ls->ls_recover_seq == seq) {
	39	set_bit(LSFL_RUNNING, &ls->ls_flags);
c36258b5	40	/* unblocks processes waiting to enter the dlm */
e7fd4179	41	up_write(&ls->ls_in_recovery);
475f230c	42	clear_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
e7fd4179 DT	43	error = 0;
	44	}
	45	spin_unlock(&ls->ls_recover_lock);
c36258b5 DT	46
c36258b5 DT	47	up_write(&ls->ls_recv_active);
e7fd4179 DT	48	return error;
	49	}
	50
	51	static int ls_recover(struct dlm_ls ls, struct dlm_recover rv)
	52	{
	53	unsigned long start;
	54	int error, neg = 0;
	55
075f0177	56	log_rinfo(ls, "dlm_recover %llu", (unsigned long long)rv->seq);
e7fd4179	57
90135925	58	mutex_lock(&ls->ls_recoverd_active);
e7fd4179	59
23e8e1aa	60	dlm_callback_suspend(ls);
e7fd4179	61
c04fecb4	62	dlm_clear_toss(ls);
e7fd4179 DT	63
e7fd4179 DT	64	/*
85f0379a DT	65	* This list of root rsb's will be the basis of most of the recovery
85f0379a DT	66	* routines.
e7fd4179 DT	67	*/
e7fd4179 DT	68
85f0379a	69	dlm_create_root_list(ls);
e7fd4179 DT	70
	71	/*
	72	* Add or remove nodes from the lockspace's ls_nodes list.
ca8031d9 AA	73	*
	74	* Due to the fact that we must report all membership changes to lsops
	75	* or midcomms layer, it is not permitted to abort ls_recover() until
	76	* this is done.
e7fd4179 DT	77	*/
	78
	79	error = dlm_recover_members(ls, rv, &neg);
	80	if (error) {
075f0177	81	log_rinfo(ls, "dlm_recover_members error %d", error);
e7fd4179 DT	82	goto fail;
e7fd4179 DT	83	}
f95a34c6	84
c04fecb4 DT	85	dlm_recover_dir_nodeid(ls);
	86
	87	ls->ls_recover_dir_sent_res = 0;
	88	ls->ls_recover_dir_sent_msg = 0;
4875647a DT	89	ls->ls_recover_locks_in = 0;
4875647a DT	90
f95a34c6 DT	91	dlm_set_recover_status(ls, DLM_RS_NODES);
f95a34c6 DT	92
c4f4e135	93	error = dlm_recover_members_wait(ls, rv->seq);
f95a34c6	94	if (error) {
075f0177	95	log_rinfo(ls, "dlm_recover_members_wait error %d", error);
f95a34c6 DT	96	goto fail;
	97	}
	98
e7fd4179 DT	99	start = jiffies;
	100
	101	/*
	102	* Rebuild our own share of the directory by collecting from all other
	103	* nodes their master rsb names that hash to us.
	104	*/
	105
c4f4e135	106	error = dlm_recover_directory(ls, rv->seq);
e7fd4179	107	if (error) {
075f0177	108	log_rinfo(ls, "dlm_recover_directory error %d", error);
e7fd4179 DT	109	goto fail;
	110	}
	111
f95a34c6	112	dlm_set_recover_status(ls, DLM_RS_DIR);
e7fd4179	113
c4f4e135	114	error = dlm_recover_directory_wait(ls, rv->seq);
e7fd4179	115	if (error) {
075f0177	116	log_rinfo(ls, "dlm_recover_directory_wait error %d", error);
e7fd4179 DT	117	goto fail;
	118	}
	119
075f0177	120	log_rinfo(ls, "dlm_recover_directory %u out %u messages",
c04fecb4 DT	121	ls->ls_recover_dir_sent_res, ls->ls_recover_dir_sent_msg);
c04fecb4 DT	122
e7fd4179 DT	123	/*
	124	* We may have outstanding operations that are waiting for a reply from
	125	* a failed node. Mark these to be resent after recovery. Unlock and
	126	* cancel ops can just be completed.
	127	*/
	128
	129	dlm_recover_waiters_pre(ls);
	130
e10249b1	131	if (dlm_recovery_stopped(ls)) {
aee742c9	132	error = -EINTR;
e7fd4179	133	goto fail;
aee742c9	134	}
e7fd4179 DT	135
	136	if (neg \|\| dlm_no_directory(ls)) {
	137	/*
	138	* Clear lkb's for departed nodes.
	139	*/
	140
4875647a	141	dlm_recover_purge(ls);
e7fd4179 DT	142
	143	/*
	144	* Get new master nodeid's for rsb's that were mastered on
	145	* departed nodes.
	146	*/
	147
c4f4e135	148	error = dlm_recover_masters(ls, rv->seq);
e7fd4179	149	if (error) {
075f0177	150	log_rinfo(ls, "dlm_recover_masters error %d", error);
e7fd4179 DT	151	goto fail;
	152	}
	153
	154	/*
	155	* Send our locks on remastered rsb's to the new masters.
	156	*/
	157
c4f4e135	158	error = dlm_recover_locks(ls, rv->seq);
e7fd4179	159	if (error) {
075f0177	160	log_rinfo(ls, "dlm_recover_locks error %d", error);
e7fd4179 DT	161	goto fail;
	162	}
	163
f95a34c6 DT	164	dlm_set_recover_status(ls, DLM_RS_LOCKS);
f95a34c6 DT	165
c4f4e135	166	error = dlm_recover_locks_wait(ls, rv->seq);
e7fd4179	167	if (error) {
075f0177	168	log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
e7fd4179 DT	169	goto fail;
	170	}
	171
075f0177	172	log_rinfo(ls, "dlm_recover_locks %u in",
4875647a DT	173	ls->ls_recover_locks_in);
4875647a DT	174
e7fd4179 DT	175	/*
	176	* Finalize state in master rsb's now that all locks can be
	177	* checked. This includes conversion resolution and lvb
	178	* settings.
	179	*/
	180
	181	dlm_recover_rsbs(ls);
91c0dc93 DT	182	} else {
	183	/*
	184	* Other lockspace members may be going through the "neg" steps
	185	* while also adding us to the lockspace, in which case they'll
4b77f2c9	186	* be doing the recover_locks (RS_LOCKS) barrier.
91c0dc93 DT	187	*/
91c0dc93 DT	188	dlm_set_recover_status(ls, DLM_RS_LOCKS);
4b77f2c9	189
c4f4e135	190	error = dlm_recover_locks_wait(ls, rv->seq);
4b77f2c9	191	if (error) {
075f0177	192	log_rinfo(ls, "dlm_recover_locks_wait error %d", error);
4b77f2c9 DT	193	goto fail;
4b77f2c9 DT	194	}
e7fd4179 DT	195	}
	196
	197	dlm_release_root_list(ls);
	198
2896ee37 DT	199	/*
	200	* Purge directory-related requests that are saved in requestqueue.
	201	* All dir requests from before recovery are invalid now due to the dir
	202	* rebuild and will be resent by the requesting nodes.
	203	*/
	204
	205	dlm_purge_requestqueue(ls);
	206
e7fd4179	207	dlm_set_recover_status(ls, DLM_RS_DONE);
f95a34c6	208
c4f4e135	209	error = dlm_recover_done_wait(ls, rv->seq);
e7fd4179	210	if (error) {
075f0177	211	log_rinfo(ls, "dlm_recover_done_wait error %d", error);
e7fd4179 DT	212	goto fail;
	213	}
	214
	215	dlm_clear_members_gone(ls);
	216
23e8e1aa DT	217	dlm_callback_resume(ls);
23e8e1aa DT	218
e7fd4179 DT	219	error = enable_locking(ls, rv->seq);
e7fd4179 DT	220	if (error) {
075f0177	221	log_rinfo(ls, "enable_locking error %d", error);
e7fd4179 DT	222	goto fail;
	223	}
	224
	225	error = dlm_process_requestqueue(ls);
	226	if (error) {
075f0177	227	log_rinfo(ls, "dlm_process_requestqueue error %d", error);
e7fd4179 DT	228	goto fail;
	229	}
	230
	231	error = dlm_recover_waiters_post(ls);
	232	if (error) {
075f0177	233	log_rinfo(ls, "dlm_recover_waiters_post error %d", error);
e7fd4179 DT	234	goto fail;
	235	}
	236
4875647a	237	dlm_recover_grant(ls);
e7fd4179	238
075f0177	239	log_rinfo(ls, "dlm_recover %llu generation %u done: %u ms",
60f98d18	240	(unsigned long long)rv->seq, ls->ls_generation,
e7fd4179	241	jiffies_to_msecs(jiffies - start));
90135925	242	mutex_unlock(&ls->ls_recoverd_active);
e7fd4179 DT	243
	244	return 0;
	245
	246	fail:
	247	dlm_release_root_list(ls);
90135925	248	mutex_unlock(&ls->ls_recoverd_active);
682bb91b	249
e7fd4179 DT	250	return error;
	251	}
	252
2cdc98aa DT	253	/* The dlm_ls_start() that created the rv we take here may already have been
	254	stopped via dlm_ls_stop(); in that case we need to leave the RECOVERY_STOP
	255	flag set. */
	256
e7fd4179 DT	257	static void do_ls_recovery(struct dlm_ls *ls)
	258	{
	259	struct dlm_recover *rv = NULL;
3182599f	260	int error;
e7fd4179 DT	261
	262	spin_lock(&ls->ls_recover_lock);
	263	rv = ls->ls_recover_args;
	264	ls->ls_recover_args = NULL;
2cdc98aa	265	if (rv && ls->ls_recover_seq == rv->seq)
475f230c	266	clear_bit(LSFL_RECOVER_STOP, &ls->ls_flags);
e7fd4179 DT	267	spin_unlock(&ls->ls_recover_lock);
	268
	269	if (rv) {
3182599f AA	270	error = ls_recover(ls, rv);
	271	switch (error) {
	272	case 0:
	273	ls->ls_recovery_result = 0;
	274	complete(&ls->ls_recovery_done);
	275
	276	dlm_lsop_recover_done(ls);
	277	break;
	278	case -EINTR:
	279	/* if recovery was interrupted -EINTR we wait for the next
	280	* ls_recover() iteration until it hopefully succeeds.
	281	*/
	282	log_rinfo(ls, "%s %llu interrupted and should be queued to run again",
	283	__func__, (unsigned long long)rv->seq);
	284	break;
	285	default:
	286	log_rinfo(ls, "%s %llu error %d", __func__,
	287	(unsigned long long)rv->seq, error);
	288
	289	/* let new_lockspace() get aware of critical error */
	290	ls->ls_recovery_result = error;
	291	complete(&ls->ls_recovery_done);
	292	break;
	293	}
	294
60f98d18	295	kfree(rv->nodes);
e7fd4179 DT	296	kfree(rv);
	297	}
	298	}
	299
	300	static int dlm_recoverd(void *arg)
	301	{
	302	struct dlm_ls *ls;
	303
	304	ls = dlm_find_lockspace_local(arg);
5f88f1ea DT	305	if (!ls) {
	306	log_print("dlm_recoverd: no lockspace %p", arg);
	307	return -1;
	308	}
e7fd4179	309
475f230c DT	310	down_write(&ls->ls_in_recovery);
	311	set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
	312	wake_up(&ls->ls_recover_lock_wait);
	313
e412f920	314	while (1) {
	315	/*
	316	* We call kthread_should_stop() after set_current_state().
	317	* This is because it works correctly if kthread_stop() is
	318	* called just before set_current_state().
	319	*/
e7fd4179	320	set_current_state(TASK_INTERRUPTIBLE);
e412f920	321	if (kthread_should_stop()) {
	322	set_current_state(TASK_RUNNING);
	323	break;
	324	}
475f230c	325	if (!test_bit(LSFL_RECOVER_WORK, &ls->ls_flags) &&
9e1b0211 GJ	326	!test_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
	327	if (kthread_should_stop())
	328	break;
e7fd4179	329	schedule();
9e1b0211	330	}
e7fd4179 DT	331	set_current_state(TASK_RUNNING);
e7fd4179 DT	332
475f230c DT	333	if (test_and_clear_bit(LSFL_RECOVER_DOWN, &ls->ls_flags)) {
	334	down_write(&ls->ls_in_recovery);
	335	set_bit(LSFL_RECOVER_LOCK, &ls->ls_flags);
	336	wake_up(&ls->ls_recover_lock_wait);
	337	}
	338
	339	if (test_and_clear_bit(LSFL_RECOVER_WORK, &ls->ls_flags))
e7fd4179 DT	340	do_ls_recovery(ls);
	341	}
	342
475f230c DT	343	if (test_bit(LSFL_RECOVER_LOCK, &ls->ls_flags))
	344	up_write(&ls->ls_in_recovery);
	345
e7fd4179 DT	346	dlm_put_lockspace(ls);
	347	return 0;
	348	}
	349
e7fd4179 DT	350	int dlm_recoverd_start(struct dlm_ls *ls)
	351	{
	352	struct task_struct *p;
	353	int error = 0;
	354
	355	p = kthread_run(dlm_recoverd, ls, "dlm_recoverd");
	356	if (IS_ERR(p))
	357	error = PTR_ERR(p);
	358	else
	359	ls->ls_recoverd_task = p;
	360	return error;
	361	}
	362
	363	void dlm_recoverd_stop(struct dlm_ls *ls)
	364	{
	365	kthread_stop(ls->ls_recoverd_task);
	366	}
	367
	368	void dlm_recoverd_suspend(struct dlm_ls *ls)
	369	{
f6db1b8e	370	wake_up(&ls->ls_wait_general);
90135925	371	mutex_lock(&ls->ls_recoverd_active);
e7fd4179 DT	372	}
	373
	374	void dlm_recoverd_resume(struct dlm_ls *ls)
	375	{
90135925	376	mutex_unlock(&ls->ls_recoverd_active);
e7fd4179 DT	377	}
e7fd4179 DT	378