X-Git-Url: https://git.whamcloud.com/?p=fs%2Flustre-release.git;a=blobdiff_plain;f=lustre%2Fmdd%2Fmdd_lfsck.c;h=5fc99148f1b87c3a1bb46baf5a1260f0e540c705;hp=16842fd43e3dce4f4e7b30d2f917bb66be1271cd;hb=e5a6d30afa60105a924f236a953b9c3c51510539;hpb=9fb46705ae86aa2c0ac29427f0ff24f923560eb7

diff --git a/lustre/mdd/mdd_lfsck.c b/lustre/mdd/mdd_lfsck.c
index 16842fd..5fc9914 100644
--- a/lustre/mdd/mdd_lfsck.c
+++ b/lustre/mdd/mdd_lfsck.c
@@ -20,7 +20,7 @@
  * GPL HEADER END
  */
 /*
- * Copyright (c) 2012 Whamcloud, Inc.
+ * Copyright (c) 2012, Intel Corporation.
  */
 /*
  * lustre/mdd/mdd_lfsck.c
@@ -40,21 +40,169 @@
 
 #include <lustre/lustre_idl.h>
 #include <lustre_fid.h>
+#include <obd_support.h>
 
 #include "mdd_internal.h"
+#include "mdd_lfsck.h"
+
+#define HALF_SEC			(CFS_HZ >> 1)
+#define LFSCK_CHECKPOINT_INTERVAL	60
+#define MDS_DIR_DUMMY_START		0xffffffffffffffffULL
+
+const char lfsck_bookmark_name[] = "lfsck_bookmark";
+
+/* misc functions */
+
+static inline struct mdd_device *mdd_lfsck2mdd(struct md_lfsck *lfsck)
+{
+	return container_of0(lfsck, struct mdd_device, mdd_lfsck);
+}
 
 static inline char *mdd_lfsck2name(struct md_lfsck *lfsck)
 {
-	struct mdd_device *mdd;
+	struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
 
-	mdd = container_of0(lfsck, struct mdd_device, mdd_lfsck);
 	return mdd2obd_dev(mdd)->obd_name;
 }
 
-void mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
+static inline void mdd_lfsck_component_put(const struct lu_env *env,
+					   struct lfsck_component *com)
+{
+	if (atomic_dec_and_test(&com->lc_ref)) {
+		if (com->lc_obj != NULL)
+			lu_object_put(env, &com->lc_obj->do_lu);
+		if (com->lc_file_ram != NULL)
+			OBD_FREE(com->lc_file_ram, com->lc_file_size);
+		if (com->lc_file_disk != NULL)
+			OBD_FREE(com->lc_file_disk, com->lc_file_size);
+		OBD_FREE_PTR(com);
+	}
+}
+
+static inline struct lfsck_component *
+__mdd_lfsck_component_find(struct md_lfsck *lfsck, __u16 type, cfs_list_t *list)
 {
+	struct lfsck_component *com;
+
+	cfs_list_for_each_entry(com, list, lc_link) {
+		if (com->lc_type == type)
+			return com;
+	}
+	return NULL;
+}
+
+static void mdd_lfsck_component_cleanup(const struct lu_env *env,
+					struct lfsck_component *com)
+{
+	if (!cfs_list_empty(&com->lc_link))
+		cfs_list_del_init(&com->lc_link);
+	if (!cfs_list_empty(&com->lc_link_dir))
+		cfs_list_del_init(&com->lc_link_dir);
+
+	mdd_lfsck_component_put(env, com);
+}
+
+static void mdd_lfsck_pos_fill(const struct lu_env *env, struct md_lfsck *lfsck,
+			       struct lfsck_position *pos, bool oit_processed,
+			       bool dir_processed)
+{
+	const struct dt_it_ops *iops = &lfsck->ml_obj_oit->do_index_ops->dio_it;
+
 	spin_lock(&lfsck->ml_lock);
-	lfsck->ml_speed_limit = limit;
+	if (unlikely(lfsck->ml_di_oit == NULL)) {
+		spin_unlock(&lfsck->ml_lock);
+		memset(pos, 0, sizeof(*pos));
+		return;
+	}
+
+	pos->lp_oit_cookie = iops->store(env, lfsck->ml_di_oit);
+
+	LASSERT(pos->lp_oit_cookie > 0);
+
+	if (!oit_processed)
+		pos->lp_oit_cookie--;
+
+	if (lfsck->ml_di_dir != NULL) {
+		struct dt_object *dto = lfsck->ml_obj_dir;
+
+		pos->lp_dir_parent = *lu_object_fid(&dto->do_lu);
+		pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
+							lfsck->ml_di_dir);
+
+		LASSERT(pos->lp_dir_cookie != MDS_DIR_DUMMY_START);
+
+		if (pos->lp_dir_cookie == MDS_DIR_END_OFF)
+			LASSERT(dir_processed);
+
+		/* For the dir which just to be processed,
+		 * lp_dir_cookie will become MDS_DIR_DUMMY_START,
+		 * which can be correctly handled by mdd_lfsck_prep. */
+		if (!dir_processed)
+			pos->lp_dir_cookie--;
+	} else {
+		fid_zero(&pos->lp_dir_parent);
+		pos->lp_dir_cookie = 0;
+	}
+	spin_unlock(&lfsck->ml_lock);
+}
+
+static inline int mdd_lfsck_pos_is_zero(const struct lfsck_position *pos)
+{
+	return pos->lp_oit_cookie == 0 && fid_is_zero(&pos->lp_dir_parent);
+}
+
+static inline int mdd_lfsck_pos_is_eq(const struct lfsck_position *pos1,
+				      const struct lfsck_position *pos2)
+{
+	if (pos1->lp_oit_cookie < pos2->lp_oit_cookie)
+		return -1;
+
+	if (pos1->lp_oit_cookie > pos2->lp_oit_cookie)
+		return 1;
+
+	if (fid_is_zero(&pos1->lp_dir_parent) &&
+	    !fid_is_zero(&pos2->lp_dir_parent))
+		return -1;
+
+	if (!fid_is_zero(&pos1->lp_dir_parent) &&
+	    fid_is_zero(&pos2->lp_dir_parent))
+		return 1;
+
+	if (fid_is_zero(&pos1->lp_dir_parent) &&
+	    fid_is_zero(&pos2->lp_dir_parent))
+		return 0;
+
+	LASSERT(lu_fid_eq(&pos1->lp_dir_parent, &pos2->lp_dir_parent));
+
+	if (pos1->lp_dir_cookie < pos2->lp_dir_cookie)
+		return -1;
+
+	if (pos1->lp_dir_cookie > pos2->lp_dir_cookie)
+		return 1;
+
+	return 0;
+}
+
+static void mdd_lfsck_close_dir(const struct lu_env *env,
+				struct md_lfsck *lfsck)
+{
+	struct dt_object	*dir_obj  = lfsck->ml_obj_dir;
+	const struct dt_it_ops	*dir_iops = &dir_obj->do_index_ops->dio_it;
+	struct dt_it		*dir_di   = lfsck->ml_di_dir;
+
+	spin_lock(&lfsck->ml_lock);
+	lfsck->ml_di_dir = NULL;
+	spin_unlock(&lfsck->ml_lock);
+
+	dir_iops->put(env, dir_di);
+	dir_iops->fini(env, dir_di);
+	lfsck->ml_obj_dir = NULL;
+	lu_object_put(env, &dir_obj->do_lu);
+}
+
+static void __mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
+{
+	lfsck->ml_bookmark_ram.lb_speed_limit = limit;
 	if (limit != LFSCK_SPEED_NO_LIMIT) {
 		if (limit > CFS_HZ) {
 			lfsck->ml_sleep_rate = limit / CFS_HZ;
@@ -67,7 +215,6 @@ void mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
 		lfsck->ml_sleep_jif = 0;
 		lfsck->ml_sleep_rate = 0;
 	}
-	spin_unlock(&lfsck->ml_lock);
 }
 
 static void mdd_lfsck_control_speed(struct md_lfsck *lfsck)
@@ -94,15 +241,624 @@ static void mdd_lfsck_control_speed(struct md_lfsck *lfsck)
 	}
 }
 
+/* lfsck_bookmark file ops */
+
+static void inline mdd_lfsck_bookmark_to_cpu(struct lfsck_bookmark *des,
+					     struct lfsck_bookmark *src)
+{
+	des->lb_magic = le32_to_cpu(src->lb_magic);
+	des->lb_version = le16_to_cpu(src->lb_version);
+	des->lb_param = le16_to_cpu(src->lb_param);
+	des->lb_speed_limit = le32_to_cpu(src->lb_speed_limit);
+}
+
+static void inline mdd_lfsck_bookmark_to_le(struct lfsck_bookmark *des,
+					    struct lfsck_bookmark *src)
+{
+	des->lb_magic = cpu_to_le32(src->lb_magic);
+	des->lb_version = cpu_to_le16(src->lb_version);
+	des->lb_param = cpu_to_le16(src->lb_param);
+	des->lb_speed_limit = cpu_to_le32(src->lb_speed_limit);
+}
+
+static int mdd_lfsck_bookmark_load(const struct lu_env *env,
+				   struct md_lfsck *lfsck)
+{
+	loff_t pos = 0;
+	int    len = sizeof(struct lfsck_bookmark);
+	int    rc;
+
+	rc = dt_record_read(env, lfsck->ml_bookmark_obj,
+			    mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
+			    &pos);
+	if (rc == 0) {
+		struct lfsck_bookmark *bm = &lfsck->ml_bookmark_ram;
+
+		mdd_lfsck_bookmark_to_cpu(bm, &lfsck->ml_bookmark_disk);
+		if (bm->lb_magic != LFSCK_BOOKMARK_MAGIC) {
+			CWARN("%.16s: invalid lfsck_bookmark magic "
+			      "0x%x != 0x%x\n", mdd_lfsck2name(lfsck),
+			      bm->lb_magic, LFSCK_BOOKMARK_MAGIC);
+			/* Process it as new lfsck_bookmark. */
+			rc = -ENODATA;
+		}
+	} else {
+		if (rc == -EFAULT && pos == 0)
+			/* return -ENODATA for empty lfsck_bookmark. */
+			rc = -ENODATA;
+		else
+			CERROR("%.16s: fail to load lfsck_bookmark, "
+			       "expected = %d, rc = %d\n",
+			       mdd_lfsck2name(lfsck), len, rc);
+	}
+	return rc;
+}
+
+static int mdd_lfsck_bookmark_store(const struct lu_env *env,
+				    struct md_lfsck *lfsck)
+{
+	struct mdd_device *mdd    = mdd_lfsck2mdd(lfsck);
+	struct thandle    *handle;
+	struct dt_object  *obj    = lfsck->ml_bookmark_obj;
+	loff_t		   pos    = 0;
+	int		   len    = sizeof(struct lfsck_bookmark);
+	int		   rc;
+	ENTRY;
+
+	mdd_lfsck_bookmark_to_le(&lfsck->ml_bookmark_disk,
+				 &lfsck->ml_bookmark_ram);
+	handle = dt_trans_create(env, mdd->mdd_bottom);
+	if (IS_ERR(handle)) {
+		rc = PTR_ERR(handle);
+		CERROR("%.16s: fail to create trans for storing "
+		       "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
+		RETURN(rc);
+	}
+
+	rc = dt_declare_record_write(env, obj, len, 0, handle);
+	if (rc != 0) {
+		CERROR("%.16s: fail to declare trans for storing "
+		       "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
+		GOTO(out, rc);
+	}
+
+	rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
+	if (rc != 0) {
+		CERROR("%.16s: fail to start trans for storing "
+		       "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
+		GOTO(out, rc);
+	}
+
+	rc = dt_record_write(env, obj,
+			     mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
+			     &pos, handle);
+	if (rc != 0)
+		CERROR("%.16s: fail to store lfsck_bookmark, expected = %d, "
+		       "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
+
+	GOTO(out, rc);
+
+out:
+	dt_trans_stop(env, mdd->mdd_bottom, handle);
+	return rc;
+}
+
+static int mdd_lfsck_bookmark_init(const struct lu_env *env,
+				   struct md_lfsck *lfsck)
+{
+	struct lfsck_bookmark *mb = &lfsck->ml_bookmark_ram;
+	int rc;
+
+	memset(mb, 0, sizeof(mb));
+	mb->lb_magic = LFSCK_BOOKMARK_MAGIC;
+	mb->lb_version = LFSCK_VERSION_V1;
+	mutex_lock(&lfsck->ml_mutex);
+	rc = mdd_lfsck_bookmark_store(env, lfsck);
+	mutex_unlock(&lfsck->ml_mutex);
+	return rc;
+}
+
+/* helper functions for framework */
+
+static int object_is_client_visible(const struct lu_env *env,
+				    struct mdd_device *mdd,
+				    struct mdd_object *obj)
+{
+	struct lu_fid *fid   = &mdd_env_info(env)->mti_fid;
+	int	       depth = 0;
+	int	       rc;
+
+	LASSERT(S_ISDIR(mdd_object_type(obj)));
+
+	while (1) {
+		if (mdd_is_root(mdd, mdo2fid(obj))) {
+			if (depth > 0)
+				mdd_object_put(env, obj);
+			return 1;
+		}
+
+		mdd_read_lock(env, obj, MOR_TGT_CHILD);
+		if (unlikely(mdd_is_dead_obj(obj))) {
+			mdd_read_unlock(env, obj);
+			if (depth > 0)
+				mdd_object_put(env, obj);
+			return 0;
+		}
+
+		rc = dt_xattr_get(env, mdd_object_child(obj),
+				  mdd_buf_get(env, NULL, 0), XATTR_NAME_LINK,
+				  BYPASS_CAPA);
+		mdd_read_unlock(env, obj);
+		if (rc >= 0) {
+			if (depth > 0)
+				mdd_object_put(env, obj);
+			return 1;
+		}
+
+		if (rc < 0 && rc != -ENODATA) {
+			if (depth > 0)
+				mdd_object_put(env, obj);
+			return rc;
+		}
+
+		rc = mdd_parent_fid(env, obj, fid);
+		if (depth > 0)
+			mdd_object_put(env, obj);
+		if (rc != 0)
+			return rc;
+
+		if (unlikely(lu_fid_eq(fid, &mdd->mdd_local_root_fid)))
+			return 0;
+
+		obj = mdd_object_find(env, mdd, fid);
+		if (obj == NULL)
+			return 0;
+		else if (IS_ERR(obj))
+			return PTR_ERR(obj);
+
+		/* XXX: need more processing for remote object in the future. */
+		if (!mdd_object_exists(obj) || mdd_object_remote(obj)) {
+			mdd_object_put(env, obj);
+			return 0;
+		}
+
+		depth++;
+	}
+	return 0;
+}
+
+static void mdd_lfsck_unpack_ent(struct lu_dirent *ent)
+{
+	fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid);
+	ent->lde_hash = le64_to_cpu(ent->lde_hash);
+	ent->lde_reclen = le16_to_cpu(ent->lde_reclen);
+	ent->lde_namelen = le16_to_cpu(ent->lde_namelen);
+	ent->lde_attrs = le32_to_cpu(ent->lde_attrs);
+
+	/* Make sure the name is terminated with '0'.
+	 * The data (type) after ent::lde_name maybe
+	 * broken, but we do not care. */
+	ent->lde_name[ent->lde_namelen] = 0;
+}
+
+/* LFSCK wrap functions */
+
+static void mdd_lfsck_fail(const struct lu_env *env, struct md_lfsck *lfsck,
+			   bool oit, bool new_checked)
+{
+	struct lfsck_component *com;
+
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		com->lc_ops->lfsck_fail(env, com, oit, new_checked);
+	}
+}
+
+static int mdd_lfsck_checkpoint(const struct lu_env *env,
+				struct md_lfsck *lfsck, bool oit)
+{
+	struct lfsck_component *com;
+	int			rc;
+
+	if (likely(cfs_time_beforeq(cfs_time_current(),
+				    lfsck->ml_time_next_checkpoint)))
+		return 0;
+
+	mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, oit, !oit);
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		rc = com->lc_ops->lfsck_checkpoint(env, com, false);
+		if (rc != 0)
+			return rc;;
+	}
+
+	lfsck->ml_time_last_checkpoint = cfs_time_current();
+	lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
+				cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
+	return 0;
+}
+
+static int mdd_lfsck_prep(struct lu_env *env, struct md_lfsck *lfsck)
+{
+	struct mdd_device      *mdd	= mdd_lfsck2mdd(lfsck);
+	struct mdd_object      *obj	= NULL;
+	struct dt_object       *dt_obj;
+	struct lfsck_component *com;
+	struct lfsck_component *next;
+	struct lfsck_position  *pos	= NULL;
+	const struct dt_it_ops *iops	=
+				&lfsck->ml_obj_oit->do_index_ops->dio_it;
+	struct dt_it	       *di;
+	int			rc;
+	ENTRY;
+
+	LASSERT(lfsck->ml_obj_dir == NULL);
+	LASSERT(lfsck->ml_di_dir == NULL);
+
+	cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
+		com->lc_new_checked = 0;
+		if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
+			com->lc_journal = 0;
+
+		rc = com->lc_ops->lfsck_prep(env, com);
+		if (rc != 0)
+			RETURN(rc);
+
+		if ((pos == NULL) ||
+		    (!mdd_lfsck_pos_is_zero(&com->lc_pos_start) &&
+		     mdd_lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
+			pos = &com->lc_pos_start;
+	}
+
+	/* Init otable-based iterator. */
+	if (pos == NULL) {
+		rc = iops->load(env, lfsck->ml_di_oit, 0);
+		GOTO(out, rc = (rc >= 0 ? 0 : rc));
+	}
+
+	rc = iops->load(env, lfsck->ml_di_oit, pos->lp_oit_cookie);
+	if (rc < 0)
+		GOTO(out, rc);
+
+	if (fid_is_zero(&pos->lp_dir_parent))
+		GOTO(out, rc = 0);
+
+	/* Find the directory for namespace-based traverse. */
+	obj = mdd_object_find(env, mdd, &pos->lp_dir_parent);
+	if (obj == NULL)
+		GOTO(out, rc = 0);
+	else if (IS_ERR(obj))
+		RETURN(PTR_ERR(obj));
+
+	/* XXX: need more processing for remote object in the future. */
+	if (!mdd_object_exists(obj) || mdd_object_remote(obj) ||
+	    unlikely(!S_ISDIR(mdd_object_type(obj))))
+		GOTO(out, rc = 0);
+
+	if (unlikely(mdd_is_dead_obj(obj)))
+		GOTO(out, rc = 0);
+
+	dt_obj = mdd_object_child(obj);
+	if (unlikely(!dt_try_as_dir(env, dt_obj)))
+		GOTO(out, rc = -ENOTDIR);
+
+	/* Init the namespace-based directory traverse. */
+	iops = &dt_obj->do_index_ops->dio_it;
+	di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
+	if (IS_ERR(di))
+		GOTO(out, rc = PTR_ERR(di));
+
+	rc = iops->load(env, di, pos->lp_dir_cookie);
+	if (rc == 0)
+		rc = iops->next(env, di);
+	else if (rc > 0)
+		rc = 0;
+
+	if (rc != 0) {
+		iops->put(env, di);
+		iops->fini(env, di);
+		GOTO(out, rc);
+	}
+
+	lfsck->ml_obj_dir = dt_obj;
+	spin_lock(&lfsck->ml_lock);
+	lfsck->ml_di_dir = di;
+	spin_unlock(&lfsck->ml_lock);
+	obj = NULL;
+
+	GOTO(out, rc = 0);
+
+out:
+	if (obj != NULL)
+		mdd_object_put(env, obj);
+
+	if (rc != 0)
+		return (rc > 0 ? 0 : rc);
+
+	mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, false, false);
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		rc = com->lc_ops->lfsck_checkpoint(env, com, true);
+		if (rc != 0)
+			break;
+	}
+
+	lfsck->ml_time_last_checkpoint = cfs_time_current();
+	lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
+				cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
+	return rc;
+}
+
+static int mdd_lfsck_exec_oit(const struct lu_env *env, struct md_lfsck *lfsck,
+			      struct mdd_object *obj)
+{
+	struct lfsck_component *com;
+	struct dt_object       *dt_obj;
+	const struct dt_it_ops *iops;
+	struct dt_it	       *di;
+	int			rc;
+	ENTRY;
+
+	LASSERT(lfsck->ml_obj_dir == NULL);
+
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
+		if (rc != 0)
+			RETURN(rc);
+	}
+
+	if (!S_ISDIR(mdd_object_type(obj)) ||
+	    cfs_list_empty(&lfsck->ml_list_dir))
+	       RETURN(0);
+
+	rc = object_is_client_visible(env, mdd_lfsck2mdd(lfsck), obj);
+	if (rc <= 0)
+		GOTO(out, rc);
+
+	if (unlikely(mdd_is_dead_obj(obj)))
+		GOTO(out, rc = 0);
+
+	dt_obj = mdd_object_child(obj);
+	if (unlikely(!dt_try_as_dir(env, dt_obj)))
+		GOTO(out, rc = -ENOTDIR);
+
+	iops = &dt_obj->do_index_ops->dio_it;
+	di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
+	if (IS_ERR(di))
+		GOTO(out, rc = PTR_ERR(di));
+
+	rc = iops->load(env, di, 0);
+	if (rc == 0)
+		rc = iops->next(env, di);
+	else if (rc > 0)
+		rc = 0;
+
+	if (rc != 0) {
+		iops->put(env, di);
+		iops->fini(env, di);
+		GOTO(out, rc);
+	}
+
+	mdd_object_get(obj);
+	lfsck->ml_obj_dir = dt_obj;
+	spin_lock(&lfsck->ml_lock);
+	lfsck->ml_di_dir = di;
+	spin_unlock(&lfsck->ml_lock);
+
+	GOTO(out, rc = 0);
+
+out:
+	if (rc < 0)
+		mdd_lfsck_fail(env, lfsck, false, false);
+	return (rc > 0 ? 0 : rc);
+}
+
+static int mdd_lfsck_exec_dir(const struct lu_env *env, struct md_lfsck *lfsck,
+			      struct mdd_object *obj, struct lu_dirent *ent)
+{
+	struct lfsck_component *com;
+	int			rc;
+
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
+		if (rc != 0)
+			return rc;
+	}
+	return 0;
+}
+
+static int mdd_lfsck_post(const struct lu_env *env, struct md_lfsck *lfsck,
+			  int result)
+{
+	struct lfsck_component *com;
+	struct lfsck_component *next;
+	int			rc;
+
+	mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, true, true);
+	cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
+		rc = com->lc_ops->lfsck_post(env, com, result);
+		if (rc != 0)
+			return rc;
+	}
+
+	lfsck->ml_time_last_checkpoint = cfs_time_current();
+	lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
+				cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
+	return result;
+}
+
+static int mdd_lfsck_double_scan(const struct lu_env *env,
+				 struct md_lfsck *lfsck)
+{
+	struct lfsck_component *com;
+	struct lfsck_component *next;
+	int			rc;
+
+	cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_double_scan,
+				     lc_link) {
+		if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
+			com->lc_journal = 0;
+
+		rc = com->lc_ops->lfsck_double_scan(env, com);
+		if (rc != 0)
+			return rc;
+	}
+	return 0;
+}
+
+/* LFSCK engines */
+
+static int mdd_lfsck_dir_engine(const struct lu_env *env,
+				struct md_lfsck *lfsck)
+{
+	struct mdd_thread_info	*info	= mdd_env_info(env);
+	struct mdd_device	*mdd	= mdd_lfsck2mdd(lfsck);
+	const struct dt_it_ops	*iops	=
+			&lfsck->ml_obj_dir->do_index_ops->dio_it;
+	struct dt_it		*di	= lfsck->ml_di_dir;
+	struct lu_dirent	*ent	= &info->mti_ent;
+	struct lu_fid		*fid	= &info->mti_fid;
+	struct lfsck_bookmark	*bk	= &lfsck->ml_bookmark_ram;
+	struct ptlrpc_thread	*thread = &lfsck->ml_thread;
+	int			 rc;
+	ENTRY;
+
+	do {
+		struct mdd_object *child;
+
+		lfsck->ml_new_scanned++;
+		rc = iops->rec(env, di, (struct dt_rec *)ent,
+			       lfsck->ml_args_dir);
+		if (rc != 0) {
+			mdd_lfsck_fail(env, lfsck, false, true);
+			if (bk->lb_param & LPF_FAILOUT)
+				RETURN(rc);
+			else
+				goto checkpoint;
+		}
+
+		mdd_lfsck_unpack_ent(ent);
+		if (ent->lde_attrs & LUDA_IGNORE)
+			goto checkpoint;
+
+		*fid = ent->lde_fid;
+		child = mdd_object_find(env, mdd, fid);
+		if (child == NULL) {
+			goto checkpoint;
+		} else if (IS_ERR(child)) {
+			mdd_lfsck_fail(env, lfsck, false, true);
+			if (bk->lb_param & LPF_FAILOUT)
+				RETURN(PTR_ERR(child));
+			else
+				goto checkpoint;
+		}
+
+		/* XXX: need more processing for remote object in the future. */
+		if (mdd_object_exists(child) && !mdd_object_remote(child))
+			rc = mdd_lfsck_exec_dir(env, lfsck, child, ent);
+		mdd_object_put(env, child);
+		if (rc != 0 && bk->lb_param & LPF_FAILOUT)
+			RETURN(rc);
+
+checkpoint:
+		rc = mdd_lfsck_checkpoint(env, lfsck, false);
+		if (rc != 0 && bk->lb_param & LPF_FAILOUT)
+			RETURN(rc);
+
+		/* Rate control. */
+		mdd_lfsck_control_speed(lfsck);
+		if (unlikely(!thread_is_running(thread)))
+			RETURN(0);
+
+		rc = iops->next(env, di);
+	} while (rc == 0);
+
+	if (rc > 0 && !lfsck->ml_oit_over)
+		mdd_lfsck_close_dir(env, lfsck);
+
+	RETURN(rc);
+}
+
+static int mdd_lfsck_oit_engine(const struct lu_env *env,
+				struct md_lfsck *lfsck)
+{
+	struct mdd_thread_info	*info	= mdd_env_info(env);
+	struct mdd_device	*mdd	= mdd_lfsck2mdd(lfsck);
+	const struct dt_it_ops	*iops	=
+				&lfsck->ml_obj_oit->do_index_ops->dio_it;
+	struct dt_it		*di	= lfsck->ml_di_oit;
+	struct lu_fid		*fid	= &info->mti_fid;
+	struct lfsck_bookmark	*bk	= &lfsck->ml_bookmark_ram;
+	struct ptlrpc_thread	*thread = &lfsck->ml_thread;
+	int			 rc;
+	ENTRY;
+
+	do {
+		struct mdd_object *target;
+
+		if (lfsck->ml_di_dir != NULL) {
+			rc = mdd_lfsck_dir_engine(env, lfsck);
+			if (rc <= 0)
+				RETURN(rc);
+		}
+
+		if (unlikely(lfsck->ml_oit_over))
+			RETURN(1);
+
+		lfsck->ml_new_scanned++;
+		rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
+		if (rc != 0) {
+			mdd_lfsck_fail(env, lfsck, true, true);
+			if (bk->lb_param & LPF_FAILOUT)
+				RETURN(rc);
+			else
+				goto checkpoint;
+		}
+
+		target = mdd_object_find(env, mdd, fid);
+		if (target == NULL) {
+			goto checkpoint;
+		} else if (IS_ERR(target)) {
+			mdd_lfsck_fail(env, lfsck, true, true);
+			if (bk->lb_param & LPF_FAILOUT)
+				RETURN(PTR_ERR(target));
+			else
+				goto checkpoint;
+		}
+
+		/* XXX: In fact, low layer otable-based iteration should not
+		 * 	return agent object. But before LU-2646 resolved, we
+		 * 	need more processing for agent object. */
+		if (mdd_object_exists(target) && !mdd_object_remote(target))
+			rc = mdd_lfsck_exec_oit(env, lfsck, target);
+		mdd_object_put(env, target);
+		if (rc != 0 && bk->lb_param & LPF_FAILOUT)
+			RETURN(rc);
+
+checkpoint:
+		rc = mdd_lfsck_checkpoint(env, lfsck, true);
+		if (rc != 0 && bk->lb_param & LPF_FAILOUT)
+			RETURN(rc);
+
+		/* Rate control. */
+		mdd_lfsck_control_speed(lfsck);
+
+		rc = iops->next(env, di);
+		if (rc > 0)
+			lfsck->ml_oit_over = 1;
+
+		if (unlikely(!thread_is_running(thread)))
+			RETURN(0);
+	} while (rc == 0 || lfsck->ml_di_dir != NULL);
+
+	RETURN(rc);
+}
+
 static int mdd_lfsck_main(void *args)
 {
 	struct lu_env		 env;
-	struct md_lfsck		*lfsck  = (struct md_lfsck *)args;
-	struct ptlrpc_thread	*thread = &lfsck->ml_thread;
-	struct dt_object	*obj    = lfsck->ml_it_obj;
-	const struct dt_it_ops	*iops   = &obj->do_index_ops->dio_it;
-	struct dt_it		*di;
-	struct lu_fid		*fid;
+	struct md_lfsck		*lfsck    = (struct md_lfsck *)args;
+	struct ptlrpc_thread	*thread   = &lfsck->ml_thread;
+	struct dt_object	*oit_obj  = lfsck->ml_obj_oit;
+	const struct dt_it_ops	*oit_iops = &oit_obj->do_index_ops->dio_it;
+	struct dt_it		*oit_di;
 	int			 rc;
 	ENTRY;
 
@@ -114,82 +870,69 @@ static int mdd_lfsck_main(void *args)
 		GOTO(noenv, rc);
 	}
 
-	di = iops->init(&env, obj, lfsck->ml_args, BYPASS_CAPA);
-	if (IS_ERR(di)) {
-		rc = PTR_ERR(di);
+	oit_di = oit_iops->init(&env, oit_obj, lfsck->ml_args_oit, BYPASS_CAPA);
+	if (IS_ERR(oit_di)) {
+		rc = PTR_ERR(oit_di);
 		CERROR("%s: LFSCK, fail to init iteration, rc = %d\n",
 		       mdd_lfsck2name(lfsck), rc);
 		GOTO(fini_env, rc);
 	}
 
-	CDEBUG(D_LFSCK, "LFSCK: flags = 0x%x, pid = %d\n",
-	       lfsck->ml_args, cfs_curproc_pid());
+	spin_lock(&lfsck->ml_lock);
+	lfsck->ml_di_oit = oit_di;
+	spin_unlock(&lfsck->ml_lock);
+	rc = mdd_lfsck_prep(&env, lfsck);
+	if (rc != 0)
+		GOTO(fini_oit, rc);
 
-	/* XXX: Prepare before wakeup the sponsor.
-	 *      Each lfsck component should call iops->get() API with
-	 *      every bookmark, then low layer module can decide the
-	 *      start point for current iteration. */
+	CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = 0x%x, dir_flags = 0x%x, "
+	       "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
+	       ", pid = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
+	       lfsck->ml_pos_current.lp_oit_cookie,
+	       lfsck->ml_pos_current.lp_dir_cookie,
+	       PFID(&lfsck->ml_pos_current.lp_dir_parent),
+	       cfs_curproc_pid());
 
 	spin_lock(&lfsck->ml_lock);
 	thread_set_flags(thread, SVC_RUNNING);
 	spin_unlock(&lfsck->ml_lock);
 	cfs_waitq_broadcast(&thread->t_ctl_waitq);
 
-	/* Call iops->load() to finish the choosing start point. */
-	rc = iops->load(&env, di, 0);
-	if (rc != 0)
-		GOTO(out, rc);
-
-	CDEBUG(D_LFSCK, "LFSCK: iteration start: pos = %s\n",
-	       (char *)iops->key(&env, di));
+	if (!cfs_list_empty(&lfsck->ml_list_scan) ||
+	    cfs_list_empty(&lfsck->ml_list_double_scan))
+		rc = mdd_lfsck_oit_engine(&env, lfsck);
+	else
+		rc = 1;
 
-	lfsck->ml_new_scanned = 0;
-	fid = &mdd_env_info(&env)->mti_fid;
-	while (rc == 0) {
-		iops->rec(&env, di, (struct dt_rec *)fid, 0);
+	CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = 0x%x, dir_flags = 0x%x, "
+	       "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
+	       ", pid = %d, rc = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
+	       lfsck->ml_pos_current.lp_oit_cookie,
+	       lfsck->ml_pos_current.lp_dir_cookie,
+	       PFID(&lfsck->ml_pos_current.lp_dir_parent),
+	       cfs_curproc_pid(), rc);
 
-		/* XXX: here, perform LFSCK when some LFSCK component(s)
-		 *      introduced in the future. */
-		lfsck->ml_new_scanned++;
+	if (lfsck->ml_paused && cfs_list_empty(&lfsck->ml_list_scan))
+		oit_iops->put(&env, oit_di);
 
-		/* XXX: here, make checkpoint when some LFSCK component(s)
-		 *      introduced in the future. */
+	rc = mdd_lfsck_post(&env, lfsck, rc);
+	if (lfsck->ml_di_dir != NULL)
+		mdd_lfsck_close_dir(&env, lfsck);
 
-		/* Rate control. */
-		mdd_lfsck_control_speed(lfsck);
-		if (unlikely(!thread_is_running(thread)))
-			GOTO(out, rc = 0);
+fini_oit:
+	spin_lock(&lfsck->ml_lock);
+	lfsck->ml_di_oit = NULL;
+	spin_unlock(&lfsck->ml_lock);
 
-		rc = iops->next(&env, di);
+	oit_iops->fini(&env, oit_di);
+	if (rc == 1) {
+		if (!cfs_list_empty(&lfsck->ml_list_double_scan))
+			rc = mdd_lfsck_double_scan(&env, lfsck);
+		else
+			rc = 0;
 	}
 
-	GOTO(out, rc);
-
-out:
-	if (lfsck->ml_paused) {
-		/* XXX: It is hack here: if the lfsck is still running when MDS
-		 *	umounts, it should be restarted automatically after MDS
-		 *	remounts up.
-		 *
-		 *	To support that, we need to record the lfsck status in
-		 *	the lfsck on-disk bookmark file. But now, there is not
-		 *	lfsck component under the lfsck framework. To avoid to
-		 *	introduce unnecessary bookmark incompatibility issues,
-		 *	we write nothing to the lfsck bookmark file now.
-		 *
-		 *	Instead, we will reuse dt_it_ops::put() method to notify
-		 *	low layer iterator to process such case.
-		 *
-		 * 	It is just temporary solution, and will be replaced when
-		 * 	some lfsck component is introduced in the future. */
-		iops->put(&env, di);
-		CDEBUG(D_LFSCK, "LFSCK: iteration pasued: pos = %s, rc = %d\n",
-		       (char *)iops->key(&env, di), rc);
-	} else {
-		CDEBUG(D_LFSCK, "LFSCK: iteration stop: pos = %s, rc = %d\n",
-		       (char *)iops->key(&env, di), rc);
-	}
-	iops->fini(&env, di);
+	/* XXX: Purge the pinned objects in the future. */
 
 fini_env:
 	lu_env_fini(&env);
@@ -202,49 +945,165 @@ noenv:
 	return rc;
 }
 
+/* external interfaces */
+
+int mdd_lfsck_set_speed(const struct lu_env *env, struct md_lfsck *lfsck,
+			__u32 limit)
+{
+	int rc;
+
+	mutex_lock(&lfsck->ml_mutex);
+	__mdd_lfsck_set_speed(lfsck, limit);
+	rc = mdd_lfsck_bookmark_store(env, lfsck);
+	mutex_unlock(&lfsck->ml_mutex);
+	return rc;
+}
+
 int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
 		    struct lfsck_start *start)
 {
-	struct ptlrpc_thread *thread  = &lfsck->ml_thread;
-	struct l_wait_info    lwi     = { 0 };
-	int		      rc      = 0;
-	__u16		      valid   = 0;
-	__u16		      flags   = 0;
+	struct lfsck_bookmark  *bk     = &lfsck->ml_bookmark_ram;
+	struct ptlrpc_thread   *thread = &lfsck->ml_thread;
+	struct lfsck_component *com;
+	struct l_wait_info      lwi    = { 0 };
+	bool			dirty  = false;
+	int			rc     = 0;
+	__u16			valid  = 0;
+	__u16			flags  = 0;
 	ENTRY;
 
-	if (lfsck->ml_it_obj == NULL)
+	if (lfsck->ml_obj_oit == NULL)
 		RETURN(-ENOTSUPP);
 
+	/* start == NULL means auto trigger paused LFSCK. */
+	if (start == NULL && cfs_list_empty(&lfsck->ml_list_scan))
+		RETURN(0);
+
 	mutex_lock(&lfsck->ml_mutex);
 	spin_lock(&lfsck->ml_lock);
-	if (thread_is_running(thread)) {
+	if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
 		spin_unlock(&lfsck->ml_lock);
 		mutex_unlock(&lfsck->ml_mutex);
 		RETURN(-EALREADY);
 	}
 
 	spin_unlock(&lfsck->ml_lock);
-	if (start->ls_valid & LSV_SPEED_LIMIT)
-		mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
+
+	lfsck->ml_paused = 0;
+	lfsck->ml_oit_over = 0;
+	lfsck->ml_drop_dryrun = 0;
+	lfsck->ml_new_scanned = 0;
+
+	/* For auto trigger. */
+	if (start == NULL)
+		goto trigger;
+
+	start->ls_version = bk->lb_version;
+	if (start->ls_valid & LSV_SPEED_LIMIT) {
+		__mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
+		dirty = true;
+	}
 
 	if (start->ls_valid & LSV_ERROR_HANDLE) {
 		valid |= DOIV_ERROR_HANDLE;
 		if (start->ls_flags & LPF_FAILOUT)
 			flags |= DOIF_FAILOUT;
+
+		if ((start->ls_flags & LPF_FAILOUT) &&
+		    !(bk->lb_param & LPF_FAILOUT)) {
+			bk->lb_param |= LPF_FAILOUT;
+			dirty = true;
+		} else if (!(start->ls_flags & LPF_FAILOUT) &&
+			   (bk->lb_param & LPF_FAILOUT)) {
+			bk->lb_param &= ~LPF_FAILOUT;
+			dirty = true;
+		}
+	}
+
+	if (start->ls_valid & LSV_DRYRUN) {
+		if ((start->ls_flags & LPF_DRYRUN) &&
+		    !(bk->lb_param & LPF_DRYRUN)) {
+			bk->lb_param |= LPF_DRYRUN;
+			dirty = true;
+		} else if (!(start->ls_flags & LPF_DRYRUN) &&
+			   (bk->lb_param & LPF_DRYRUN)) {
+			bk->lb_param &= ~LPF_DRYRUN;
+			lfsck->ml_drop_dryrun = 1;
+			dirty = true;
+		}
 	}
 
-	/* XXX: 1. low layer does not care 'dryrun'.
-	 *      2. will process 'ls_active' when introduces LFSCK for layout
-	 *	   consistency, DNE consistency, and so on in the future. */
-	start->ls_active = 0;
+	if (dirty) {
+		rc = mdd_lfsck_bookmark_store(env, lfsck);
+		if (rc != 0)
+			GOTO(out, rc);
+	}
 
 	if (start->ls_flags & LPF_RESET)
 		flags |= DOIF_RESET;
 
-	if (start->ls_active != 0)
+	if (start->ls_active != 0) {
+		struct lfsck_component *next;
+		__u16 type = 1;
+
+		if (start->ls_active == LFSCK_TYPES_ALL)
+			start->ls_active = LFSCK_TYPES_SUPPORTED;
+
+		if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
+			start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
+			GOTO(out, rc = -ENOTSUPP);
+		}
+
+		cfs_list_for_each_entry_safe(com, next,
+					     &lfsck->ml_list_scan, lc_link) {
+			if (!(com->lc_type & start->ls_active)) {
+				rc = com->lc_ops->lfsck_post(env, com, 0);
+				if (rc != 0)
+					GOTO(out, rc);
+			}
+		}
+
+		while (start->ls_active != 0) {
+			if (type & start->ls_active) {
+				com = __mdd_lfsck_component_find(lfsck, type,
+							&lfsck->ml_list_idle);
+				if (com != NULL) {
+					/* The component status will be updated
+					 * when its prep() is called later by
+					 * the LFSCK main engine. */
+					cfs_list_del_init(&com->lc_link);
+					cfs_list_add_tail(&com->lc_link,
+							  &lfsck->ml_list_scan);
+				}
+				start->ls_active &= ~type;
+			}
+			type <<= 1;
+		}
+	}
+
+	cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
+		start->ls_active |= com->lc_type;
+		if (flags & DOIF_RESET) {
+			rc = com->lc_ops->lfsck_reset(env, com, false);
+			if (rc != 0)
+				GOTO(out, rc);
+		}
+	}
+
+trigger:
+	lfsck->ml_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
+	if (bk->lb_param & LPF_DRYRUN)
+		lfsck->ml_args_dir |= LUDA_VERIFY_DRYRUN;
+
+	if (bk->lb_param & LPF_FAILOUT) {
+		valid |= DOIV_ERROR_HANDLE;
+		flags |= DOIF_FAILOUT;
+	}
+
+	if (!cfs_list_empty(&lfsck->ml_list_scan))
 		flags |= DOIF_OUTUSED;
 
-	lfsck->ml_args = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
+	lfsck->ml_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
 	thread_set_flags(thread, 0);
 	rc = cfs_create_thread(mdd_lfsck_main, lfsck, 0);
 	if (rc < 0)
@@ -255,12 +1114,16 @@ int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
 			     thread_is_running(thread) ||
 			     thread_is_stopped(thread),
 			     &lwi);
-	mutex_unlock(&lfsck->ml_mutex);
 
-	RETURN(rc < 0 ? rc : 0);
+	GOTO(out, rc = 0);
+
+out:
+	mutex_unlock(&lfsck->ml_mutex);
+	return (rc < 0 ? rc : 0);
 }
 
-int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck)
+int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck,
+		   bool pause)
 {
 	struct ptlrpc_thread *thread = &lfsck->ml_thread;
 	struct l_wait_info    lwi    = { 0 };
@@ -274,7 +1137,14 @@ int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck)
 		RETURN(-EALREADY);
 	}
 
+	if (pause)
+		lfsck->ml_paused = 1;
 	thread_set_flags(thread, SVC_STOPPING);
+	/* The LFSCK thread may be sleeping on low layer wait queue,
+	 * wake it up. */
+	if (likely(lfsck->ml_di_oit != NULL))
+		lfsck->ml_obj_oit->do_index_ops->dio_it.put(env,
+							    lfsck->ml_di_oit);
 	spin_unlock(&lfsck->ml_lock);
 
 	cfs_waitq_broadcast(&thread->t_ctl_waitq);
@@ -286,8 +1156,6 @@ int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck)
 	RETURN(0);
 }
 
-const char lfsck_bookmark_name[] = "lfsck_bookmark";
-
 static const struct lu_fid lfsck_it_fid = { .f_seq = FID_SEQ_LOCAL_FILE,
 					    .f_oid = OTABLE_IT_OID,
 					    .f_ver = 0 };
@@ -297,50 +1165,90 @@ int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd)
 	struct md_lfsck  *lfsck = &mdd->mdd_lfsck;
 	struct dt_object *obj;
 	int		  rc;
+	ENTRY;
 
-	memset(lfsck, 0, sizeof(*lfsck));
-	lfsck->ml_version = LFSCK_VERSION_V1;
-	cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
+	LASSERT(!lfsck->ml_initialized);
+
+	lfsck->ml_initialized = 1;
 	mutex_init(&lfsck->ml_mutex);
 	spin_lock_init(&lfsck->ml_lock);
+	CFS_INIT_LIST_HEAD(&lfsck->ml_list_scan);
+	CFS_INIT_LIST_HEAD(&lfsck->ml_list_dir);
+	CFS_INIT_LIST_HEAD(&lfsck->ml_list_double_scan);
+	CFS_INIT_LIST_HEAD(&lfsck->ml_list_idle);
+	cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
 
-	obj = dt_store_open(env, mdd->mdd_child, "", lfsck_bookmark_name,
-			    &mdd_env_info(env)->mti_fid);
-	if (IS_ERR(obj))
-		return PTR_ERR(obj);
-
-	lfsck->ml_bookmark_obj = obj;
-
-	obj = dt_locate(env, mdd->mdd_child, &lfsck_it_fid);
+	obj = dt_locate(env, mdd->mdd_bottom, &lfsck_it_fid);
 	if (IS_ERR(obj))
-		return PTR_ERR(obj);
+		RETURN(PTR_ERR(obj));
 
+	lfsck->ml_obj_oit = obj;
 	rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
 	if (rc != 0) {
-		lu_object_put(env, &obj->do_lu);
 		if (rc == -ENOTSUPP)
 			rc = 0;
-		return rc;
+
+		RETURN(rc);
 	}
 
-	lfsck->ml_it_obj = obj;
+	obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_bookmark_name,
+			    &mdd_env_info(env)->mti_fid);
+	if (IS_ERR(obj))
+		RETURN(PTR_ERR(obj));
 
-	return 0;
+	lfsck->ml_bookmark_obj = obj;
+	rc = mdd_lfsck_bookmark_load(env, lfsck);
+	if (rc == -ENODATA)
+		rc = mdd_lfsck_bookmark_init(env, lfsck);
+
+	/* XXX: LFSCK components initialization to be added here. */
+
+	RETURN(rc);
 }
 
 void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd)
 {
-	struct md_lfsck *lfsck = &mdd->mdd_lfsck;
+	struct md_lfsck		*lfsck  = &mdd->mdd_lfsck;
+	struct ptlrpc_thread	*thread = &lfsck->ml_thread;
+	struct lfsck_component	*com;
 
-	if (lfsck->ml_it_obj != NULL) {
-		lfsck->ml_paused = 1;
-		mdd_lfsck_stop(env, lfsck);
-		lu_object_put(env, &lfsck->ml_it_obj->do_lu);
-		lfsck->ml_it_obj = NULL;
+	if (!lfsck->ml_initialized)
+		return;
+
+	LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
+
+	if (lfsck->ml_obj_oit != NULL) {
+		lu_object_put(env, &lfsck->ml_obj_oit->do_lu);
+		lfsck->ml_obj_oit = NULL;
 	}
 
+	LASSERT(lfsck->ml_obj_dir == NULL);
+
 	if (lfsck->ml_bookmark_obj != NULL) {
 		lu_object_put(env, &lfsck->ml_bookmark_obj->do_lu);
 		lfsck->ml_bookmark_obj = NULL;
 	}
+
+	while (!cfs_list_empty(&lfsck->ml_list_scan)) {
+		com = cfs_list_entry(lfsck->ml_list_scan.next,
+				     struct lfsck_component,
+				     lc_link);
+		mdd_lfsck_component_cleanup(env, com);
+	}
+
+	LASSERT(cfs_list_empty(&lfsck->ml_list_dir));
+
+	while (!cfs_list_empty(&lfsck->ml_list_double_scan)) {
+		com = cfs_list_entry(lfsck->ml_list_double_scan.next,
+				     struct lfsck_component,
+				     lc_link);
+		mdd_lfsck_component_cleanup(env, com);
+	}
+
+	while (!cfs_list_empty(&lfsck->ml_list_idle)) {
+		com = cfs_list_entry(lfsck->ml_list_idle.next,
+				     struct lfsck_component,
+				     lc_link);
+		mdd_lfsck_component_cleanup(env, com);
+	}
 }