X-Git-Url: https://git.whamcloud.com/?a=blobdiff_plain;f=lustre%2Flfsck%2Flfsck_layout.c;h=98e61d236cb556aa983afacf5d1f2123d3604060;hb=refs%2Fchanges%2F46%2F10046%2F7;hp=f77ee876515f80bb9c32395f0f1b2528ded2483b;hpb=77eea1985bb1655e58c8b7df00703b4f08b58ec7;p=fs%2Flustre-release.git

diff --git a/lustre/lfsck/lfsck_layout.c b/lustre/lfsck/lfsck_layout.c
index f77ee87..98e61d236 100644
--- a/lustre/lfsck/lfsck_layout.c
+++ b/lustre/lfsck/lfsck_layout.c
@@ -349,26 +349,47 @@ again:
 static int lfsck_layout_verify_header(struct lov_mds_md_v1 *lmm)
 {
 	__u32 magic;
-	__u32 patten;
+	__u32 pattern;
 
 	magic = le32_to_cpu(lmm->lmm_magic);
 	/* If magic crashed, keep it there. Sometime later, during OST-object
 	 * orphan handling, if some OST-object(s) back-point to it, it can be
 	 * verified and repaired. */
-	if (magic != LOV_MAGIC_V1 && magic != LOV_MAGIC_V3)
-		return -EINVAL;
+	if (magic != LOV_MAGIC_V1 && magic != LOV_MAGIC_V3) {
+		struct ost_id	oi;
+		int		rc;
+
+		lmm_oi_cpu_to_le(&oi, &lmm->lmm_oi);
+		if ((magic & LOV_MAGIC_MASK) == LOV_MAGIC_MAGIC)
+			rc = -EOPNOTSUPP;
+		else
+			rc = -EINVAL;
+
+		CDEBUG(D_LFSCK, "%s LOV EA magic %u on "DOSTID"\n",
+		       rc == -EINVAL ? "Unknown" : "Unsupported",
+		       magic, POSTID(&oi));
 
-	patten = le32_to_cpu(lmm->lmm_pattern);
+		return rc;
+	}
+
+	pattern = le32_to_cpu(lmm->lmm_pattern);
 	/* XXX: currently, we only support LOV_PATTERN_RAID0. */
-	if (patten != LOV_PATTERN_RAID0)
+	if (lov_pattern(pattern) != LOV_PATTERN_RAID0) {
+		struct ost_id oi;
+
+		lmm_oi_cpu_to_le(&oi, &lmm->lmm_oi);
+		CDEBUG(D_LFSCK, "Unsupported LOV EA pattern %u on "DOSTID"\n",
+		       pattern, POSTID(&oi));
+
 		return -EOPNOTSUPP;
+	}
 
 	return 0;
 }
 
 #define LFSCK_RBTREE_BITMAP_SIZE	PAGE_CACHE_SIZE
 #define LFSCK_RBTREE_BITMAP_WIDTH	(LFSCK_RBTREE_BITMAP_SIZE << 3)
-#define LFSCK_RBTREE_BITMAP_MASK	(LFSCK_RBTREE_BITMAP_SIZE - 1)
+#define LFSCK_RBTREE_BITMAP_MASK	(LFSCK_RBTREE_BITMAP_WIDTH - 1)
 
 struct lfsck_rbtree_node {
 	struct rb_node	 lrn_node;
@@ -392,7 +413,7 @@ static inline int lfsck_rbtree_cmp(struct lfsck_rbtree_node *lrn,
 	if (oid < lrn->lrn_first_oid)
 		return -1;
 
-	if (oid >= lrn->lrn_first_oid + LFSCK_RBTREE_BITMAP_WIDTH)
+	if (oid - lrn->lrn_first_oid >= LFSCK_RBTREE_BITMAP_WIDTH)
 		return 1;
 
 	return 0;
@@ -492,19 +513,19 @@ static struct lfsck_rbtree_node *
 lfsck_rbtree_insert(struct lfsck_layout_slave_data *llsd,
 		    struct lfsck_rbtree_node *lrn)
 {
-	struct rb_node		 **pos    = &(llsd->llsd_rb_root.rb_node);
+	struct rb_node		 **pos    = &llsd->llsd_rb_root.rb_node;
 	struct rb_node		  *parent = NULL;
 	struct lfsck_rbtree_node  *tmp;
 	int			   rc;
 
-	while (*pos) {
+	while (*pos != NULL) {
 		parent = *pos;
-		tmp = rb_entry(*pos, struct lfsck_rbtree_node, lrn_node);
+		tmp = rb_entry(parent, struct lfsck_rbtree_node, lrn_node);
 		rc = lfsck_rbtree_cmp(tmp, lrn->lrn_seq, lrn->lrn_first_oid);
 		if (rc < 0)
-			pos = &((*pos)->rb_left);
+			pos = &(*pos)->rb_left;
 		else if (rc > 0)
-			pos = &((*pos)->rb_right);
+			pos = &(*pos)->rb_right;
 		else
 			return tmp;
 	}
@@ -771,7 +792,8 @@ static int lfsck_layout_store(const struct lu_env *env,
 		RETURN(rc);
 	}
 
-	rc = dt_declare_record_write(env, obj, size, pos, handle);
+	rc = dt_declare_record_write(env, obj, lfsck_buf_get(env, lo, size),
+				     pos, handle);
 	if (rc != 0) {
 		CERROR("%s: fail to declare trans for storing lfsck_layout(1): "
 		       "rc = %d\n", lfsck_lfsck2name(lfsck), rc);
@@ -913,7 +935,10 @@ lfsck_layout_lastid_create(const struct lu_env *env,
 	if (rc != 0)
 		GOTO(stop, rc);
 
-	rc = dt_declare_record_write(env, obj, sizeof(lastid), pos, th);
+	rc = dt_declare_record_write(env, obj,
+				     lfsck_buf_get(env, &lastid,
+						   sizeof(lastid)),
+				     pos, th);
 	if (rc != 0)
 		GOTO(stop, rc);
 
@@ -1033,8 +1058,11 @@ lfsck_layout_lastid_store(const struct lu_env *env,
 			continue;
 		}
 
+		lastid = cpu_to_le64(lls->lls_lastid);
 		rc = dt_declare_record_write(env, lls->lls_lastid_obj,
-					     sizeof(lastid), pos, th);
+					     lfsck_buf_get(env, &lastid,
+							   sizeof(lastid)),
+					     pos, th);
 		if (rc != 0)
 			goto stop;
 
@@ -1042,7 +1070,6 @@ lfsck_layout_lastid_store(const struct lu_env *env,
 		if (rc != 0)
 			goto stop;
 
-		lastid = cpu_to_le64(lls->lls_lastid);
 		dt_write_lock(env, lls->lls_lastid_obj, 0);
 		rc = dt_record_write(env, lls->lls_lastid_obj,
 				     lfsck_buf_get(env, &lastid,
@@ -1138,6 +1165,17 @@ out:
 	return rc;
 }
 
+static void lfsck_layout_record_failure(const struct lu_env *env,
+						 struct lfsck_instance *lfsck,
+						 struct lfsck_layout *lo)
+{
+	lo->ll_objs_failed_phase1++;
+	if (unlikely(lo->ll_pos_first_inconsistent == 0))
+		lo->ll_pos_first_inconsistent =
+			lfsck->li_obj_oit->do_index_ops->dio_it.store(env,
+							lfsck->li_di_oit);
+}
+
 static int lfsck_layout_master_async_interpret(const struct lu_env *env,
 					       struct ptlrpc_request *req,
 					       void *args, int rc)
@@ -1562,7 +1600,6 @@ static int lfsck_layout_double_scan_result(const struct lu_env *env,
 	struct lfsck_bookmark	*bk    = &lfsck->li_bookmark_ram;
 
 	down_write(&com->lc_sem);
-
 	lo->ll_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
 				HALF_SEC - lfsck->li_time_last_checkpoint);
 	lo->ll_time_last_checkpoint = cfs_time_current_sec();
@@ -1586,15 +1623,7 @@ static int lfsck_layout_double_scan_result(const struct lu_env *env,
 		lo->ll_status = LS_FAILED;
 	}
 
-	if (lo->ll_status != LS_PAUSED) {
-		spin_lock(&lfsck->li_lock);
-		list_del_init(&com->lc_link);
-		list_add_tail(&com->lc_link, &lfsck->li_list_idle);
-		spin_unlock(&lfsck->li_lock);
-	}
-
 	rc = lfsck_layout_store(env, com);
-
 	up_write(&com->lc_sem);
 
 	return rc;
@@ -1645,27 +1674,1018 @@ static int lfsck_layout_trans_stop(const struct lu_env *env,
 {
 	int rc;
 
-	handle->th_result = result;
-	rc = dt_trans_stop(env, dev, handle);
-	if (rc > 0)
-		rc = 0;
-	else if (rc == 0)
-		rc = 1;
+	handle->th_result = result;
+	rc = dt_trans_stop(env, dev, handle);
+	if (rc > 0)
+		rc = 0;
+	else if (rc == 0)
+		rc = 1;
+
+	return rc;
+}
+
+/**
+ * Get the system default stripe size.
+ *
+ * \param[in] env	pointer to the thread context
+ * \param[in] lfsck	pointer to the lfsck instance
+ * \param[out] size	pointer to the default stripe size
+ *
+ * \retval		0 for success
+ * \retval		negative error number on failure
+ */
+static int lfsck_layout_get_def_stripesize(const struct lu_env *env,
+					   struct lfsck_instance *lfsck,
+					   __u32 *size)
+{
+	struct lov_user_md	*lum = &lfsck_env_info(env)->lti_lum;
+	struct dt_object	*root;
+	int			 rc;
+
+	root = dt_locate(env, lfsck->li_next, &lfsck->li_local_root_fid);
+	if (IS_ERR(root))
+		return PTR_ERR(root);
+
+	/* Get the default stripe size via xattr_get on the backend root. */
+	rc = dt_xattr_get(env, root, lfsck_buf_get(env, lum, sizeof(*lum)),
+			  XATTR_NAME_LOV, BYPASS_CAPA);
+	if (rc > 0) {
+		/* The lum->lmm_stripe_size is LE mode. The *size also
+		 * should be LE mode. So it is unnecessary to convert. */
+		*size = lum->lmm_stripe_size;
+		rc = 0;
+	} else if (unlikely(rc == 0)) {
+		rc = -EINVAL;
+	}
+
+	lfsck_object_put(env, root);
+
+	return rc;
+}
+
+/**
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_refill_lovea(const struct lu_env *env,
+				     struct thandle *handle,
+				     struct dt_object *parent,
+				     struct lu_fid *cfid,
+				     struct lu_buf *buf,
+				     struct lov_ost_data_v1 *slot,
+				     int fl, __u32 ost_idx)
+{
+	struct ost_id		*oi	= &lfsck_env_info(env)->lti_oi;
+	struct lov_mds_md_v1	*lmm	= buf->lb_buf;
+	int			 rc;
+
+	fid_to_ostid(cfid, oi);
+	ostid_cpu_to_le(oi, &slot->l_ost_oi);
+	slot->l_ost_gen = cpu_to_le32(0);
+	slot->l_ost_idx = cpu_to_le32(ost_idx);
+
+	if (le32_to_cpu(lmm->lmm_pattern) & LOV_PATTERN_F_HOLE) {
+		struct lov_ost_data_v1 *objs;
+		int			i;
+		__u16			count;
+
+		count = le16_to_cpu(lmm->lmm_stripe_count);
+		if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V1)
+			objs = &lmm->lmm_objects[0];
+		else
+			objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
+		for (i = 0; i < count; i++, objs++) {
+			if (objs != slot && lovea_slot_is_dummy(objs))
+				break;
+		}
+
+		/* If the @slot is the last dummy slot to be refilled,
+		 * then drop LOV_PATTERN_F_HOLE from lmm::lmm_pattern. */
+		if (i == count)
+			lmm->lmm_pattern &= ~cpu_to_le32(LOV_PATTERN_F_HOLE);
+	}
+
+	rc = dt_xattr_set(env, parent, buf, XATTR_NAME_LOV, fl, handle,
+			  BYPASS_CAPA);
+	if (rc == 0)
+		rc = 1;
+
+	return rc;
+}
+
+/**
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_extend_lovea(const struct lu_env *env,
+				     struct lfsck_instance *lfsck,
+				     struct thandle *handle,
+				     struct dt_object *parent,
+				     struct lu_fid *cfid,
+				     struct lu_buf *buf, int fl,
+				     __u32 ost_idx, __u32 ea_off, bool reset)
+{
+	struct lov_mds_md_v1	*lmm	= buf->lb_buf;
+	struct lov_ost_data_v1	*objs;
+	int			 rc;
+	__u16			 count;
+	ENTRY;
+
+	if (fl == LU_XATTR_CREATE || reset) {
+		__u32 pattern = LOV_PATTERN_RAID0;
+
+		count = ea_off + 1;
+		LASSERT(buf->lb_len == lov_mds_md_size(count, LOV_MAGIC_V1));
+
+		if (ea_off != 0 || reset)
+			pattern |= LOV_PATTERN_F_HOLE;
+
+		memset(lmm, 0, buf->lb_len);
+		lmm->lmm_magic = cpu_to_le32(LOV_MAGIC_V1);
+		lmm->lmm_pattern = cpu_to_le32(pattern);
+		fid_to_lmm_oi(lfsck_dto2fid(parent), &lmm->lmm_oi);
+		lmm_oi_cpu_to_le(&lmm->lmm_oi, &lmm->lmm_oi);
+
+		rc = lfsck_layout_get_def_stripesize(env, lfsck,
+						     &lmm->lmm_stripe_size);
+		if (rc != 0)
+			RETURN(rc);
+
+		objs = &lmm->lmm_objects[ea_off];
+	} else {
+		__u32	magic = le32_to_cpu(lmm->lmm_magic);
+		int	gap;
+
+		count = le16_to_cpu(lmm->lmm_stripe_count);
+		if (magic == LOV_MAGIC_V1)
+			objs = &lmm->lmm_objects[count];
+		else
+			objs = &((struct lov_mds_md_v3 *)lmm)->
+							lmm_objects[count];
+
+		gap = ea_off - count;
+		if (gap >= 0)
+			count = ea_off + 1;
+		LASSERT(buf->lb_len == lov_mds_md_size(count, magic));
+
+		if (gap > 0) {
+			memset(objs, 0, gap * sizeof(*objs));
+			lmm->lmm_pattern |= cpu_to_le32(LOV_PATTERN_F_HOLE);
+		}
+
+		lmm->lmm_layout_gen =
+			    cpu_to_le16(le16_to_cpu(lmm->lmm_layout_gen) + 1);
+		objs += gap;
+	}
+
+	lmm->lmm_stripe_count = cpu_to_le16(count);
+	rc = lfsck_layout_refill_lovea(env, handle, parent, cfid, buf, objs,
+				       fl, ost_idx);
+
+	RETURN(rc);
+}
+
+/**
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_update_pfid(const struct lu_env *env,
+				    struct lfsck_component *com,
+				    struct dt_object *parent,
+				    struct lu_fid *cfid,
+				    struct dt_device *cdev, __u32 ea_off)
+{
+	struct filter_fid	*pfid	= &lfsck_env_info(env)->lti_new_pfid;
+	struct dt_object	*child;
+	struct thandle		*handle;
+	const struct lu_fid	*tfid	= lu_object_fid(&parent->do_lu);
+	struct lu_buf		*buf;
+	int			 rc	= 0;
+	ENTRY;
+
+	child = lfsck_object_find_by_dev(env, cdev, cfid);
+	if (IS_ERR(child))
+		RETURN(PTR_ERR(child));
+
+	handle = dt_trans_create(env, cdev);
+	if (IS_ERR(handle))
+		GOTO(out, rc = PTR_ERR(handle));
+
+	pfid->ff_parent.f_seq = cpu_to_le64(tfid->f_seq);
+	pfid->ff_parent.f_oid = cpu_to_le32(tfid->f_oid);
+	/* Currently, the filter_fid::ff_parent::f_ver is not the real parent
+	 * MDT-object's FID::f_ver, instead it is the OST-object index in its
+	 * parent MDT-object's layout EA. */
+	pfid->ff_parent.f_stripe_idx = cpu_to_le32(ea_off);
+	buf = lfsck_buf_get(env, pfid, sizeof(struct filter_fid));
+
+	rc = dt_declare_xattr_set(env, child, buf, XATTR_NAME_FID, 0, handle);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_trans_start(env, cdev, handle);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_xattr_set(env, child, buf, XATTR_NAME_FID, 0, handle,
+			  BYPASS_CAPA);
+
+	GOTO(stop, rc = (rc == 0 ? 1 : rc));
+
+stop:
+	dt_trans_stop(env, cdev, handle);
+
+out:
+	lu_object_put(env, &child->do_lu);
+
+	return rc;
+}
+
+/**
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_recreate_parent(const struct lu_env *env,
+					struct lfsck_component *com,
+					struct lfsck_tgt_desc *ltd,
+					struct lu_orphan_rec *rec,
+					struct lu_fid *cfid,
+					const char *prefix,
+					const char *postfix,
+					__u32 ea_off)
+{
+	struct lfsck_thread_info	*info	= lfsck_env_info(env);
+	char				*name	= info->lti_key;
+	struct lu_attr			*la	= &info->lti_la;
+	struct dt_object_format 	*dof	= &info->lti_dof;
+	struct lfsck_instance		*lfsck	= com->lc_lfsck;
+	struct lu_fid			*pfid	= &rec->lor_fid;
+	struct lu_fid			*tfid	= &info->lti_fid3;
+	struct dt_device		*next	= lfsck->li_next;
+	struct dt_object		*pobj	= NULL;
+	struct dt_object		*cobj	= NULL;
+	struct thandle			*th	= NULL;
+	struct lu_buf			*pbuf	= NULL;
+	struct lu_buf			*ea_buf = &info->lti_big_buf;
+	struct lustre_handle		 lh	= { 0 };
+	int				 buflen = ea_buf->lb_len;
+	int				 idx	= 0;
+	int				 rc	= 0;
+	ENTRY;
+
+	/* Create .lustre/lost+found/MDTxxxx when needed. */
+	if (unlikely(lfsck->li_lpf_obj == NULL)) {
+		rc = lfsck_create_lpf(env, lfsck);
+		if (rc != 0)
+			RETURN(rc);
+	}
+
+	if (fid_is_zero(pfid)) {
+		struct filter_fid *ff = &info->lti_new_pfid;
+
+		rc = lfsck_fid_alloc(env, lfsck, pfid, false);
+		if (rc != 0)
+			RETURN(rc);
+
+		ff->ff_parent.f_seq = cpu_to_le64(pfid->f_seq);
+		ff->ff_parent.f_oid = cpu_to_le32(pfid->f_oid);
+		/* Currently, the filter_fid::ff_parent::f_ver is not the
+		 * real parent MDT-object's FID::f_ver, instead it is the
+		 * OST-object index in its parent MDT-object's layout EA. */
+		ff->ff_parent.f_stripe_idx = cpu_to_le32(ea_off);
+		pbuf = lfsck_buf_get(env, ff, sizeof(struct filter_fid));
+		cobj = lfsck_object_find_by_dev(env, ltd->ltd_tgt, cfid);
+		if (IS_ERR(cobj))
+			RETURN(PTR_ERR(cobj));
+	}
+
+	CDEBUG(D_LFSCK, "Re-create the lost MDT-object: parent "
+	       DFID", child "DFID", OST-index %u, stripe-index %u, "
+	       "prefix %s, postfix %s\n",
+	       PFID(pfid), PFID(cfid), ltd->ltd_index, ea_off, prefix, postfix);
+
+	pobj = lfsck_object_find_by_dev(env, lfsck->li_bottom, pfid);
+	if (IS_ERR(pobj))
+		GOTO(put, rc = PTR_ERR(pobj));
+
+	LASSERT(prefix != NULL);
+	LASSERT(postfix != NULL);
+
+	/** name rules:
+	 *
+	 *  1. Use the MDT-object's FID as the name with prefix and postfix.
+	 *
+	 *  1.1 prefix "C-":	More than one OST-objects claim the same
+	 *			MDT-object and the same slot in the layout EA.
+	 *			It may be created for dangling referenced MDT
+	 *			object or may be not.
+	 *  1.2 prefix "N-":	The orphan OST-object does not know which one
+	 *			is the real parent, so the LFSCK assign a new
+	 *			FID as its parent.
+	 *  1.3 prefix "R-":	The orphan OST-object know its parent FID but
+	 *			does not know the position in the namespace.
+	 *
+	 *  2. If there is name conflict, append more index for new name. */
+	sprintf(name, "%s"DFID"%s", prefix, PFID(pfid), postfix);
+	do {
+		rc = dt_lookup(env, lfsck->li_lpf_obj, (struct dt_rec *)tfid,
+			       (const struct dt_key *)name, BYPASS_CAPA);
+		if (rc != 0 && rc != -ENOENT)
+			GOTO(put, rc);
+
+		if (unlikely(rc == 0)) {
+			CWARN("%s: The name %s under lost+found has been used "
+			      "by the "DFID". Try to increase the FID version "
+			      "for the new file name.\n",
+			      lfsck_lfsck2name(lfsck), name, PFID(tfid));
+			sprintf(name, "%s"DFID"%s-%d", prefix, PFID(pfid),
+				postfix, ++idx);
+		}
+	} while (rc == 0);
+
+	memset(la, 0, sizeof(*la));
+	la->la_uid = rec->lor_uid;
+	la->la_gid = rec->lor_gid;
+	la->la_mode = S_IFREG | S_IRUSR;
+	la->la_valid = LA_MODE | LA_UID | LA_GID;
+
+	memset(dof, 0, sizeof(*dof));
+	dof->dof_type = dt_mode_to_dft(S_IFREG);
+
+	rc = lov_mds_md_size(ea_off + 1, LOV_MAGIC_V1);
+	if (buflen < rc) {
+		lu_buf_realloc(ea_buf, rc);
+		buflen = ea_buf->lb_len;
+		if (ea_buf->lb_buf == NULL)
+			GOTO(put, rc = -ENOMEM);
+	} else {
+		ea_buf->lb_len = rc;
+	}
+
+	/* Hold update lock on the .lustre/lost+found/MDTxxxx/.
+	 *
+	 * XXX: Currently, we do not grab the PDO lock as normal create cases,
+	 *	because creating MDT-object for orphan OST-object is rare, we
+	 *	do not much care about the performance. It can be improved in
+	 *	the future when needed. */
+	rc = lfsck_layout_lock(env, com, lfsck->li_lpf_obj, &lh,
+			       MDS_INODELOCK_UPDATE);
+	if (rc != 0)
+		GOTO(put, rc);
+
+	th = dt_trans_create(env, next);
+	if (IS_ERR(th))
+		GOTO(unlock, rc = PTR_ERR(th));
+
+	/* 1a. Update OST-object's parent information remotely.
+	 *
+	 * If other subsequent modifications failed, then next LFSCK scanning
+	 * will process the OST-object as orphan again with known parent FID. */
+	if (cobj != NULL) {
+		rc = dt_declare_xattr_set(env, cobj, pbuf, XATTR_NAME_FID, 0, th);
+		if (rc != 0)
+			GOTO(stop, rc);
+	}
+
+	/* 2a. Create the MDT-object locally. */
+	rc = dt_declare_create(env, pobj, la, NULL, dof, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	/* 3a. Add layout EA for the MDT-object. */
+	rc = dt_declare_xattr_set(env, pobj, ea_buf, XATTR_NAME_LOV,
+				  LU_XATTR_CREATE, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	/* 4a. Insert the MDT-object to .lustre/lost+found/MDTxxxx/ */
+	rc = dt_declare_insert(env, lfsck->li_lpf_obj,
+			       (const struct dt_rec *)pfid,
+			       (const struct dt_key *)name, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_trans_start(env, next, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	/* 1b. Update OST-object's parent information remotely. */
+	if (cobj != NULL) {
+		rc = dt_xattr_set(env, cobj, pbuf, XATTR_NAME_FID, 0, th,
+				  BYPASS_CAPA);
+		if (rc != 0)
+			GOTO(stop, rc);
+	}
+
+	dt_write_lock(env, pobj, 0);
+	/* 2b. Create the MDT-object locally. */
+	rc = dt_create(env, pobj, la, NULL, dof, th);
+	if (rc == 0)
+		/* 3b. Add layout EA for the MDT-object. */
+		rc = lfsck_layout_extend_lovea(env, lfsck, th, pobj, cfid,
+					       ea_buf, LU_XATTR_CREATE,
+					       ltd->ltd_index, ea_off, false);
+	dt_write_unlock(env, pobj);
+	if (rc < 0)
+		GOTO(stop, rc);
+
+	/* 4b. Insert the MDT-object to .lustre/lost+found/MDTxxxx/ */
+	rc = dt_insert(env, lfsck->li_lpf_obj,
+		       (const struct dt_rec *)pfid,
+		       (const struct dt_key *)name, th, BYPASS_CAPA, 1);
+
+	GOTO(stop, rc);
+
+stop:
+	dt_trans_stop(env, next, th);
+
+unlock:
+	lfsck_layout_unlock(&lh);
+
+put:
+	if (cobj != NULL && !IS_ERR(cobj))
+		lu_object_put(env, &cobj->do_lu);
+	if (pobj != NULL && !IS_ERR(pobj))
+		lu_object_put(env, &pobj->do_lu);
+	ea_buf->lb_len = buflen;
+
+	return rc >= 0 ? 1 : rc;
+}
+
+static int lfsck_layout_master_conditional_destroy(const struct lu_env *env,
+						   struct lfsck_component *com,
+						   const struct lu_fid *fid,
+						   __u32 index)
+{
+	struct lfsck_thread_info *info	= lfsck_env_info(env);
+	struct lfsck_request	 *lr	= &info->lti_lr;
+	struct lfsck_instance	 *lfsck = com->lc_lfsck;
+	struct lfsck_tgt_desc	 *ltd;
+	struct ptlrpc_request	 *req;
+	struct lfsck_request	 *tmp;
+	struct obd_export	 *exp;
+	int			  rc	= 0;
+	ENTRY;
+
+	ltd = lfsck_tgt_get(&lfsck->li_ost_descs, index);
+	if (unlikely(ltd == NULL))
+		RETURN(-ENXIO);
+
+	exp = ltd->ltd_exp;
+	if (!(exp_connect_flags(exp) & OBD_CONNECT_LFSCK))
+		GOTO(put, rc = -EOPNOTSUPP);
+
+	req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_LFSCK_NOTIFY);
+	if (req == NULL)
+		GOTO(put, rc = -ENOMEM);
+
+	rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, LFSCK_NOTIFY);
+	if (rc != 0) {
+		ptlrpc_request_free(req);
+
+		GOTO(put, rc);
+	}
+
+	memset(lr, 0, sizeof(*lr));
+	lr->lr_event = LE_CONDITIONAL_DESTROY;
+	lr->lr_active = LT_LAYOUT;
+	lr->lr_fid = *fid;
+
+	tmp = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
+	*tmp = *lr;
+	ptlrpc_request_set_replen(req);
+
+	rc = ptlrpc_queue_wait(req);
+	ptlrpc_req_finished(req);
+
+	GOTO(put, rc);
+
+put:
+	lfsck_tgt_put(ltd);
+
+	return rc;
+}
+
+static int lfsck_layout_slave_conditional_destroy(const struct lu_env *env,
+						  struct lfsck_component *com,
+						  struct lfsck_request *lr)
+{
+	struct lfsck_thread_info	*info	= lfsck_env_info(env);
+	struct lu_attr			*la	= &info->lti_la;
+	ldlm_policy_data_t		*policy = &info->lti_policy;
+	struct ldlm_res_id		*resid	= &info->lti_resid;
+	struct lfsck_instance		*lfsck	= com->lc_lfsck;
+	struct dt_device		*dev	= lfsck->li_bottom;
+	struct lu_fid			*fid	= &lr->lr_fid;
+	struct dt_object		*obj;
+	struct thandle			*th	= NULL;
+	struct lustre_handle		 lh	= { 0 };
+	__u64				 flags	= 0;
+	int				 rc	= 0;
+	ENTRY;
+
+	obj = lfsck_object_find_by_dev(env, dev, fid);
+	if (IS_ERR(obj))
+		RETURN(PTR_ERR(obj));
+
+	dt_read_lock(env, obj, 0);
+	if (dt_object_exists(obj) == 0) {
+		dt_read_unlock(env, obj);
+
+		GOTO(put, rc = -ENOENT);
+	}
+
+	/* Get obj's attr without lock firstly. */
+	rc = dt_attr_get(env, obj, la, BYPASS_CAPA);
+	dt_read_unlock(env, obj);
+	if (rc != 0)
+		GOTO(put, rc);
+
+	if (likely(la->la_ctime != 0 || la->la_mode & S_ISUID))
+		GOTO(put, rc = -ETXTBSY);
+
+	/* Acquire extent lock on [0, EOF] to sync with all possible written. */
+	LASSERT(lfsck->li_namespace != NULL);
+
+	memset(policy, 0, sizeof(*policy));
+	policy->l_extent.end = OBD_OBJECT_EOF;
+	ost_fid_build_resid(fid, resid);
+	rc = ldlm_cli_enqueue_local(lfsck->li_namespace, resid, LDLM_EXTENT,
+				    policy, LCK_EX, &flags, ldlm_blocking_ast,
+				    ldlm_completion_ast, NULL, NULL, 0,
+				    LVB_T_NONE, NULL, &lh);
+	if (rc != ELDLM_OK)
+		GOTO(put, rc = -EIO);
+
+	dt_write_lock(env, obj, 0);
+	/* Get obj's attr within lock again. */
+	rc = dt_attr_get(env, obj, la, BYPASS_CAPA);
+	if (rc != 0)
+		GOTO(unlock, rc);
+
+	if (la->la_ctime != 0)
+		GOTO(unlock, rc = -ETXTBSY);
+
+	th = dt_trans_create(env, dev);
+	if (IS_ERR(th))
+		GOTO(unlock, rc = PTR_ERR(th));
+
+	rc = dt_declare_ref_del(env, obj, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_declare_destroy(env, obj, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_trans_start_local(env, dev, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_ref_del(env, obj, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_destroy(env, obj, th);
+	if (rc == 0)
+		CDEBUG(D_LFSCK, "Destroy the empty OST-object "DFID" which "
+		       "was created for reparing dangling referenced case. "
+		       "But the original missed OST-object is found now.\n",
+		       PFID(fid));
+
+	GOTO(stop, rc);
+
+stop:
+	dt_trans_stop(env, dev, th);
+
+unlock:
+	dt_write_unlock(env, obj);
+	ldlm_lock_decref(&lh, LCK_EX);
+
+put:
+	lu_object_put(env, &obj->do_lu);
+
+	return rc;
+}
+
+/**
+ * Some OST-object has occupied the specified layout EA slot.
+ * Such OST-object may be generated by the LFSCK when repair
+ * dangling referenced MDT-object, which can be indicated by
+ * attr::la_ctime == 0 but without S_ISUID in la_mode. If it
+ * is true and such OST-object has not been modified yet, we
+ * will replace it with the orphan OST-object; otherwise the
+ * LFSCK will create new MDT-object to reference the orphan.
+ *
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_conflict_create(const struct lu_env *env,
+					struct lfsck_component *com,
+					struct lfsck_tgt_desc *ltd,
+					struct lu_orphan_rec *rec,
+					struct dt_object *parent,
+					struct lu_fid *cfid,
+					struct lu_buf *ea_buf,
+					struct lov_ost_data_v1 *slot,
+					__u32 ea_off, __u32 ori_len)
+{
+	struct lfsck_thread_info *info		= lfsck_env_info(env);
+	struct lu_fid		 *cfid2		= &info->lti_fid2;
+	struct ost_id		 *oi		= &info->lti_oi;
+	char			 *postfix	= info->lti_tmpbuf;
+	struct lov_mds_md_v1	 *lmm		= ea_buf->lb_buf;
+	struct dt_device	 *dev		= com->lc_lfsck->li_bottom;
+	struct thandle		 *th		= NULL;
+	struct lustre_handle	  lh		= { 0 };
+	__u32			  ost_idx2	= le32_to_cpu(slot->l_ost_idx);
+	int			  rc		= 0;
+	ENTRY;
+
+	ostid_le_to_cpu(&slot->l_ost_oi, oi);
+	ostid_to_fid(cfid2, oi, ost_idx2);
+
+	CDEBUG(D_LFSCK, "Handle layout EA conflict: parent "DFID
+	       ", cur-child "DFID" on the OST %u, orphan-child "
+	       DFID" on the OST %u, stripe-index %u\n",
+	       PFID(lfsck_dto2fid(parent)), PFID(cfid2), ost_idx2,
+	       PFID(cfid), ltd->ltd_index, ea_off);
+
+	/* Hold layout lock on the parent to prevent others to access. */
+	rc = lfsck_layout_lock(env, com, parent, &lh,
+			       MDS_INODELOCK_LAYOUT | MDS_INODELOCK_XATTR);
+	if (rc != 0)
+		GOTO(out, rc);
+
+	rc = lfsck_layout_master_conditional_destroy(env, com, cfid2, ost_idx2);
+
+	/* If the conflict OST-obejct is not created for fixing dangling
+	 * referenced MDT-object in former LFSCK check/repair, or it has
+	 * been modified by others, then we cannot destroy it. Re-create
+	 * a new MDT-object for the orphan OST-object. */
+	if (rc == -ETXTBSY) {
+		/* No need the layout lock on the original parent. */
+		lfsck_layout_unlock(&lh);
+		ea_buf->lb_len = ori_len;
+
+		fid_zero(&rec->lor_fid);
+		snprintf(postfix, LFSCK_TMPBUF_LEN, "-"DFID"-%x",
+			 PFID(lu_object_fid(&parent->do_lu)), ea_off);
+		rc = lfsck_layout_recreate_parent(env, com, ltd, rec, cfid,
+						  "C-", postfix, ea_off);
+
+		RETURN(rc);
+	}
+
+	if (rc != 0 && rc != -ENOENT)
+		GOTO(unlock, rc);
+
+	th = dt_trans_create(env, dev);
+	if (IS_ERR(th))
+		GOTO(unlock, rc = PTR_ERR(th));
+
+	rc = dt_declare_xattr_set(env, parent, ea_buf, XATTR_NAME_LOV,
+				  LU_XATTR_REPLACE, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_trans_start_local(env, dev, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	dt_write_lock(env, parent, 0);
+	lmm->lmm_layout_gen = cpu_to_le16(le16_to_cpu(lmm->lmm_layout_gen) + 1);
+	rc = lfsck_layout_refill_lovea(env, th, parent, cfid, ea_buf, slot,
+				       LU_XATTR_REPLACE, ltd->ltd_index);
+	dt_write_unlock(env, parent);
+
+	GOTO(stop, rc);
+
+stop:
+	dt_trans_stop(env, dev, th);
+
+unlock:
+	lfsck_layout_unlock(&lh);
+
+out:
+	ea_buf->lb_len = ori_len;
+
+	return rc >= 0 ? 1 : rc;
+}
+
+/**
+ * \retval	 +1: repaired
+ * \retval	  0: did nothing
+ * \retval	-ve: on error
+ */
+static int lfsck_layout_recreate_lovea(const struct lu_env *env,
+				       struct lfsck_component *com,
+				       struct lfsck_tgt_desc *ltd,
+				       struct lu_orphan_rec *rec,
+				       struct dt_object *parent,
+				       struct lu_fid *cfid,
+				       __u32 ost_idx, __u32 ea_off)
+{
+	struct lfsck_thread_info *info		= lfsck_env_info(env);
+	struct lu_buf		 *buf		= &info->lti_big_buf;
+	struct lu_fid		 *fid		= &info->lti_fid2;
+	struct ost_id		 *oi		= &info->lti_oi;
+	struct lfsck_instance	 *lfsck 	= com->lc_lfsck;
+	struct dt_device	 *dt		= lfsck->li_bottom;
+	struct lfsck_bookmark	 *bk		= &lfsck->li_bookmark_ram;
+	struct thandle		  *handle	= NULL;
+	size_t			  buflen	= buf->lb_len;
+	struct lov_mds_md_v1	 *lmm;
+	struct lov_ost_data_v1   *objs;
+	struct lustre_handle	  lh		= { 0 };
+	__u32			  magic;
+	int			  fl		= 0;
+	int			  rc		= 0;
+	int			  rc1;
+	int			  i;
+	__u16			  count;
+	bool			  locked	= false;
+	ENTRY;
+
+	CDEBUG(D_LFSCK, "Re-create the crashed layout EA: parent "
+	       DFID", child "DFID", OST-index %u, stripe-index %u\n",
+	       PFID(lfsck_dto2fid(parent)), PFID(cfid), ost_idx, ea_off);
+
+	rc = lfsck_layout_lock(env, com, parent, &lh,
+			       MDS_INODELOCK_LAYOUT | MDS_INODELOCK_XATTR);
+	if (rc != 0)
+		RETURN(rc);
+
+again:
+	if (locked) {
+		dt_write_unlock(env, parent);
+		locked = false;
+	}
+
+	if (handle != NULL) {
+		dt_trans_stop(env, dt, handle);
+		handle = NULL;
+	}
+
+	if (rc < 0)
+		GOTO(unlock_layout, rc);
+
+	if (buf->lb_len < rc) {
+		lu_buf_realloc(buf, rc);
+		buflen = buf->lb_len;
+		if (buf->lb_buf == NULL)
+			GOTO(unlock_layout, rc = -ENOMEM);
+	}
+
+	if (!(bk->lb_param & LPF_DRYRUN)) {
+		handle = dt_trans_create(env, dt);
+		if (IS_ERR(handle))
+			GOTO(unlock_layout, rc = PTR_ERR(handle));
+
+		rc = dt_declare_xattr_set(env, parent, buf, XATTR_NAME_LOV,
+					  fl, handle);
+		if (rc != 0)
+			GOTO(stop, rc);
+
+		rc = dt_trans_start_local(env, dt, handle);
+		if (rc != 0)
+			GOTO(stop, rc);
+	}
+
+	dt_write_lock(env, parent, 0);
+	locked = true;
+	rc = dt_xattr_get(env, parent, buf, XATTR_NAME_LOV, BYPASS_CAPA);
+	if (rc == -ERANGE) {
+		rc = dt_xattr_get(env, parent, &LU_BUF_NULL, XATTR_NAME_LOV,
+				  BYPASS_CAPA);
+		LASSERT(rc != 0);
+		goto again;
+	} else if (rc == -ENODATA || rc == 0) {
+		rc = lov_mds_md_size(ea_off + 1, LOV_MAGIC_V1);
+		/* If the declared is not big enough, re-try. */
+		if (buf->lb_len < rc)
+			goto again;
+
+		fl = LU_XATTR_CREATE;
+	} else if (rc < 0) {
+		GOTO(unlock_parent, rc);
+	} else if (unlikely(buf->lb_len == 0)) {
+		goto again;
+	} else {
+		fl = LU_XATTR_REPLACE;
+	}
+
+	if (fl == LU_XATTR_CREATE) {
+		if (bk->lb_param & LPF_DRYRUN)
+			GOTO(unlock_parent, rc = 1);
+
+		LASSERT(buf->lb_len >= rc);
+
+		buf->lb_len = rc;
+		rc = lfsck_layout_extend_lovea(env, lfsck, handle, parent, cfid,
+					       buf, fl, ost_idx, ea_off, false);
+
+		GOTO(unlock_parent, rc);
+	}
+
+	lmm = buf->lb_buf;
+	rc1 = lfsck_layout_verify_header(lmm);
+
+	/* If the LOV EA crashed, the rebuild it. */
+	if (rc1 == -EINVAL) {
+		if (bk->lb_param & LPF_DRYRUN)
+			GOTO(unlock_parent, rc = 1);
+
+		LASSERT(buf->lb_len >= rc);
+
+		buf->lb_len = rc;
+		memset(lmm, 0, buf->lb_len);
+		rc = lfsck_layout_extend_lovea(env, lfsck, handle, parent, cfid,
+					       buf, fl, ost_idx, ea_off, true);
+
+		GOTO(unlock_parent, rc);
+	}
+
+	/* For other unknown magic/pattern, keep the current LOV EA. */
+	if (rc1 != 0)
+		GOTO(unlock_parent, rc = rc1);
+
+	/* Currently, we only support LOV_MAGIC_V1/LOV_MAGIC_V3 which has
+	 * been verified in lfsck_layout_verify_header() already. If some
+	 * new magic introduced in the future, then layout LFSCK needs to
+	 * be updated also. */
+	magic = le32_to_cpu(lmm->lmm_magic);
+	if (magic == LOV_MAGIC_V1) {
+		objs = &lmm->lmm_objects[0];
+	} else {
+		LASSERT(magic == LOV_MAGIC_V3);
+		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
+	}
+
+	count = le16_to_cpu(lmm->lmm_stripe_count);
+	if (count == 0)
+		GOTO(unlock_parent, rc = -EINVAL);
+	LASSERT(count > 0);
+
+	/* Exceed the current end of MDT-object layout EA. Then extend it. */
+	if (count <= ea_off) {
+		if (bk->lb_param & LPF_DRYRUN)
+			GOTO(unlock_parent, rc = 1);
+
+		rc = lov_mds_md_size(ea_off + 1, magic);
+		/* If the declared is not big enough, re-try. */
+		if (buf->lb_len < rc)
+			goto again;
+
+		buf->lb_len = rc;
+		rc = lfsck_layout_extend_lovea(env, lfsck, handle, parent, cfid,
+					       buf, fl, ost_idx, ea_off, false);
+
+		GOTO(unlock_parent, rc);
+	}
+
+	LASSERTF(rc > 0, "invalid rc = %d\n", rc);
+
+	buf->lb_len = rc;
+	for (i = 0; i < count; i++, objs++) {
+		/* The MDT-object was created via lfsck_layout_recover_create()
+		 * by others before, and we fill the dummy layout EA. */
+		if (lovea_slot_is_dummy(objs)) {
+			if (i != ea_off)
+				continue;
+
+			if (bk->lb_param & LPF_DRYRUN)
+				GOTO(unlock_parent, rc = 1);
+
+			lmm->lmm_layout_gen =
+			    cpu_to_le16(le16_to_cpu(lmm->lmm_layout_gen) + 1);
+			rc = lfsck_layout_refill_lovea(env, handle, parent,
+						       cfid, buf, objs, fl,
+						       ost_idx);
+			GOTO(unlock_parent, rc);
+		}
+
+		ostid_le_to_cpu(&objs->l_ost_oi, oi);
+		ostid_to_fid(fid, oi, le32_to_cpu(objs->l_ost_idx));
+		/* It should be rare case, the slot is there, but the LFSCK
+		 * does not handle it during the first-phase cycle scanning. */
+		if (unlikely(lu_fid_eq(fid, cfid))) {
+			if (i == ea_off) {
+				GOTO(unlock_parent, rc = 0);
+			} else {
+				/* Rare case that the OST-object index
+				 * does not match the parent MDT-object
+				 * layout EA. We trust the later one. */
+				if (bk->lb_param & LPF_DRYRUN)
+					GOTO(unlock_parent, rc = 1);
+
+				dt_write_unlock(env, parent);
+				if (handle != NULL)
+					dt_trans_stop(env, dt, handle);
+				lfsck_layout_unlock(&lh);
+				buf->lb_len = buflen;
+				rc = lfsck_layout_update_pfid(env, com, parent,
+							cfid, ltd->ltd_tgt, i);
+
+				RETURN(rc);
+			}
+		}
+	}
+
+	/* The MDT-object exists, but related layout EA slot is occupied
+	 * by others. */
+	if (bk->lb_param & LPF_DRYRUN)
+		GOTO(unlock_parent, rc = 1);
+
+	dt_write_unlock(env, parent);
+	if (handle != NULL)
+		dt_trans_stop(env, dt, handle);
+	lfsck_layout_unlock(&lh);
+	if (le32_to_cpu(lmm->lmm_magic) == LOV_MAGIC_V1)
+		objs = &lmm->lmm_objects[ea_off];
+	else
+		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[ea_off];
+	rc = lfsck_layout_conflict_create(env, com, ltd, rec, parent, cfid,
+					  buf, objs, ea_off, buflen);
+
+	RETURN(rc);
+
+unlock_parent:
+	if (locked)
+		dt_write_unlock(env, parent);
+
+stop:
+	if (handle != NULL)
+		dt_trans_stop(env, dt, handle);
+
+unlock_layout:
+	lfsck_layout_unlock(&lh);
+	buf->lb_len = buflen;
+
+	return rc;
+}
+
+static int lfsck_layout_scan_orphan_one(const struct lu_env *env,
+					struct lfsck_component *com,
+					struct lfsck_tgt_desc *ltd,
+					struct lu_orphan_rec *rec,
+					struct lu_fid *cfid)
+{
+	struct lfsck_layout	*lo	= com->lc_file_ram;
+	struct lu_fid		*pfid	= &rec->lor_fid;
+	struct dt_object	*parent = NULL;
+	__u32			 ea_off = pfid->f_stripe_idx;
+	int			 rc	= 0;
+	ENTRY;
+
+	if (!fid_is_sane(cfid))
+		GOTO(out, rc = -EINVAL);
+
+	if (fid_is_zero(pfid)) {
+		rc = lfsck_layout_recreate_parent(env, com, ltd, rec, cfid,
+						  "N-", "", ea_off);
+		GOTO(out, rc);
+	}
+
+	pfid->f_ver = 0;
+	if (!fid_is_sane(pfid))
+		GOTO(out, rc = -EINVAL);
+
+	parent = lfsck_object_find_by_dev(env, com->lc_lfsck->li_bottom, pfid);
+	if (IS_ERR(parent))
+		GOTO(out, rc = PTR_ERR(parent));
+
+	if (unlikely(dt_object_remote(parent) != 0))
+		GOTO(put, rc = -EXDEV);
+
+	if (dt_object_exists(parent) == 0) {
+		lu_object_put(env, &parent->do_lu);
+		rc = lfsck_layout_recreate_parent(env, com, ltd, rec, cfid,
+						  "R-", "", ea_off);
+		GOTO(out, rc);
+	}
+
+	if (!S_ISREG(lu_object_attr(&parent->do_lu)))
+		GOTO(put, rc = -EISDIR);
 
-	return rc;
-}
+	rc = lfsck_layout_recreate_lovea(env, com, ltd, rec, parent, cfid,
+					 ltd->ltd_index, ea_off);
 
-static int lfsck_layout_scan_orphan_one(const struct lu_env *env,
-					struct lfsck_component *com,
-					struct lfsck_tgt_desc *ltd,
-					struct lu_orphan_rec *rec,
-					struct lu_fid *cfid)
-{
-	struct lfsck_layout		*lo	= com->lc_file_ram;
-	int				 rc	= 0;
+	GOTO(put, rc);
 
-	/* XXX: To be extended in other patch. */
+put:
+	if (rc <= 0)
+		lu_object_put(env, &parent->do_lu);
+	else
+		/* The layout EA is changed, need to be reloaded next time. */
+		lu_object_put_nocache(env, &parent->do_lu);
 
+out:
 	down_write(&com->lc_sem);
 	com->lc_new_scanned++;
 	com->lc_new_checked++;
@@ -1738,6 +2758,18 @@ static int lfsck_layout_scan_orphan(const struct lu_env *env,
 		struct dt_key		*key;
 		struct lu_orphan_rec	*rec = &info->lti_rec;
 
+		if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY3) &&
+		    cfs_fail_val > 0) {
+			struct ptlrpc_thread	*thread = &lfsck->li_thread;
+			struct l_wait_info	 lwi;
+
+			lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
+					  NULL, NULL);
+			l_wait_event(thread->t_ctl_waitq,
+				     !thread_is_running(thread),
+				     &lwi);
+		}
+
 		key = iops->key(env, di);
 		com->lc_fid_latest_scanned_phase2 = *(struct lu_fid *)key;
 		rc = iops->rec(env, di, (struct dt_rec *)rec, 0);
@@ -1767,16 +2799,24 @@ put:
 	return rc > 0 ? 0 : rc;
 }
 
-/* For the MDT-object with dangling reference, we need to re-create
- * the missed OST-object with the known FID/owner information. */
-static int lfsck_layout_recreate_ostobj(const struct lu_env *env,
+/* For the MDT-object with dangling reference, we need to repare the
+ * inconsistency according to the LFSCK sponsor's requirement:
+ *
+ * 1) Keep the inconsistency there and report the inconsistency case,
+ *    then give the chance to the application to find related issues,
+ *    and the users can make the decision about how to handle it with
+ *    more human knownledge. (by default)
+ *
+ * 2) Re-create the missed OST-object with the FID/owner information. */
+static int lfsck_layout_repair_dangling(const struct lu_env *env,
 					struct lfsck_component *com,
 					struct lfsck_layout_req *llr,
-					struct lu_attr *la)
+					const struct lu_attr *pla)
 {
 	struct lfsck_thread_info	*info	= lfsck_env_info(env);
 	struct filter_fid		*pfid	= &info->lti_new_pfid;
 	struct dt_allocation_hint	*hint	= &info->lti_hint;
+	struct lu_attr			*cla    = &info->lti_la2;
 	struct dt_object		*parent = llr->llr_parent->llo_obj;
 	struct dt_object		*child  = llr->llr_child;
 	struct dt_device		*dev	= lfsck_obj2dt_dev(child);
@@ -1785,12 +2825,30 @@ static int lfsck_layout_recreate_ostobj(const struct lu_env *env,
 	struct lu_buf			*buf;
 	struct lustre_handle		 lh	= { 0 };
 	int				 rc;
+	bool				 create;
 	ENTRY;
 
-	CDEBUG(D_LFSCK, "Repair dangling reference for: parent "DFID
-	       ", child "DFID", OST-index %u, stripe-index %u, owner %u:%u\n",
+	if (com->lc_lfsck->li_bookmark_ram.lb_param & LPF_CREATE_OSTOBJ)
+		create = true;
+	else
+		create = false;
+
+	CDEBUG(D_LFSCK, "Found dangling reference for: parent "DFID
+	       ", child "DFID", OST-index %u, stripe-index %u, owner %u:%u. %s",
 	       PFID(lfsck_dto2fid(parent)), PFID(lfsck_dto2fid(child)),
-	       llr->llr_ost_idx, llr->llr_lov_idx, la->la_uid, la->la_gid);
+	       llr->llr_ost_idx, llr->llr_lov_idx, pla->la_uid, pla->la_gid,
+	       create ? "Create the lost OST-object as required.\n" :
+			"Keep the MDT-object there by default.\n");
+
+	if (!create)
+		RETURN(1);
+
+	memset(cla, 0, sizeof(*cla));
+	cla->la_uid = pla->la_uid;
+	cla->la_gid = pla->la_gid;
+	cla->la_mode = S_IFREG | 0666;
+	cla->la_valid = LA_TYPE | LA_MODE | LA_UID | LA_GID |
+			LA_ATIME | LA_MTIME | LA_CTIME;
 
 	rc = lfsck_layout_lock(env, com, parent, &lh,
 			       MDS_INODELOCK_LAYOUT | MDS_INODELOCK_XATTR);
@@ -1805,10 +2863,13 @@ static int lfsck_layout_recreate_ostobj(const struct lu_env *env,
 	hint->dah_mode = 0;
 	pfid->ff_parent.f_seq = cpu_to_le64(tfid->f_seq);
 	pfid->ff_parent.f_oid = cpu_to_le32(tfid->f_oid);
-	pfid->ff_parent.f_ver = cpu_to_le32(llr->llr_lov_idx);
+	/* Currently, the filter_fid::ff_parent::f_ver is not the real parent
+	 * MDT-object's FID::f_ver, instead it is the OST-object index in its
+	 * parent MDT-object's layout EA. */
+	pfid->ff_parent.f_stripe_idx = cpu_to_le32(llr->llr_lov_idx);
 	buf = lfsck_buf_get(env, pfid, sizeof(struct filter_fid));
 
-	rc = dt_declare_create(env, child, la, hint, NULL, handle);
+	rc = dt_declare_create(env, child, cla, hint, NULL, handle);
 	if (rc != 0)
 		GOTO(stop, rc);
 
@@ -1825,7 +2886,7 @@ static int lfsck_layout_recreate_ostobj(const struct lu_env *env,
 	if (unlikely(lu_object_is_dying(parent->do_lu.lo_header)))
 		GOTO(unlock2, rc = 1);
 
-	rc = dt_create(env, child, la, hint, NULL, handle);
+	rc = dt_create(env, child, cla, hint, NULL, handle);
 	if (rc != 0)
 		GOTO(unlock2, rc);
 
@@ -1883,9 +2944,10 @@ static int lfsck_layout_repair_unmatched_pair(const struct lu_env *env,
 
 	pfid->ff_parent.f_seq = cpu_to_le64(tfid->f_seq);
 	pfid->ff_parent.f_oid = cpu_to_le32(tfid->f_oid);
-	/* The ff_parent->f_ver is not the real parent fid->f_ver. Instead,
-	 * it is the OST-object index in the parent MDT-object layout. */
-	pfid->ff_parent.f_ver = cpu_to_le32(llr->llr_lov_idx);
+	/* Currently, the filter_fid::ff_parent::f_ver is not the real parent
+	 * MDT-object's FID::f_ver, instead it is the OST-object index in its
+	 * parent MDT-object's layout EA. */
+	pfid->ff_parent.f_stripe_idx = cpu_to_le32(llr->llr_lov_idx);
 	buf = lfsck_buf_get(env, pfid, sizeof(struct filter_fid));
 
 	rc = dt_declare_xattr_set(env, child, buf, XATTR_NAME_FID, 0, handle);
@@ -2011,10 +3073,6 @@ static int lfsck_layout_repair_multiple_references(const struct lu_env *env,
 		GOTO(unlock2, rc = 0);
 
 	lmm = buf->lb_buf;
-	rc = lfsck_layout_verify_header(lmm);
-	if (rc != 0)
-		GOTO(unlock2, rc);
-
 	/* Someone change layout during the LFSCK, no need to repair then. */
 	if (le16_to_cpu(lmm->lmm_layout_gen) != llr->llr_parent->llo_gen)
 		GOTO(unlock2, rc = 0);
@@ -2029,7 +3087,7 @@ static int lfsck_layout_repair_multiple_references(const struct lu_env *env,
 	 * be updated also. */
 	magic = le32_to_cpu(lmm->lmm_magic);
 	if (magic == LOV_MAGIC_V1) {
-		objs = &(lmm->lmm_objects[0]);
+		objs = &lmm->lmm_objects[0];
 	} else {
 		LASSERT(magic == LOV_MAGIC_V3);
 		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
@@ -2201,17 +3259,9 @@ static int lfsck_layout_check_parent(const struct lu_env *env,
 		GOTO(out, rc);
 
 	lmm = buf->lb_buf;
-	rc = lfsck_layout_verify_header(lmm);
-	if (rc != 0)
-		GOTO(out, rc);
-
-	/* Currently, we only support LOV_MAGIC_V1/LOV_MAGIC_V3 which has
-	 * been verified in lfsck_layout_verify_header() already. If some
-	 * new magic introduced in the future, then layout LFSCK needs to
-	 * be updated also. */
 	magic = le32_to_cpu(lmm->lmm_magic);
 	if (magic == LOV_MAGIC_V1) {
-		objs = &(lmm->lmm_objects[0]);
+		objs = &lmm->lmm_objects[0];
 	} else {
 		LASSERT(magic == LOV_MAGIC_V3);
 		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
@@ -2222,6 +3272,9 @@ static int lfsck_layout_check_parent(const struct lu_env *env,
 		struct lu_fid		*tfid	= &info->lti_fid2;
 		struct ost_id		*oi	= &info->lti_oi;
 
+		if (lovea_slot_is_dummy(objs))
+			continue;
+
 		ostid_le_to_cpu(&objs->l_ost_oi, oi);
 		ostid_to_fid(tfid, oi, le32_to_cpu(objs->l_ost_idx));
 		if (lu_fid_eq(cfid, tfid)) {
@@ -2294,10 +3347,10 @@ static int lfsck_layout_assistant_handle_one(const struct lu_env *env,
 		fid_zero(pfid);
 	} else {
 		fid_le_to_cpu(pfid, &pea->ff_parent);
-		/* OST-object does not save parent FID::f_ver, instead,
-		 * the OST-object index in the parent MDT-object layout
-		 * EA reuses the pfid->f_ver. */
-		idx = pfid->f_ver;
+		/* Currently, the filter_fid::ff_parent::f_ver is not the
+		 * real parent MDT-object's FID::f_ver, instead it is the
+		 * OST-object index in its parent MDT-object's layout EA. */
+		idx = pfid->f_stripe_idx;
 		pfid->f_ver = 0;
 	}
 
@@ -2328,13 +3381,7 @@ repair:
 
 	switch (type) {
 	case LLIT_DANGLING:
-		memset(cla, 0, sizeof(*cla));
-		cla->la_uid = pla->la_uid;
-		cla->la_gid = pla->la_gid;
-		cla->la_mode = S_IFREG | 0666;
-		cla->la_valid = LA_TYPE | LA_MODE | LA_UID | LA_GID |
-				LA_ATIME | LA_MTIME | LA_CTIME;
-		rc = lfsck_layout_recreate_ostobj(env, com, llr, cla);
+		rc = lfsck_layout_repair_dangling(env, com, llr, pla);
 		break;
 	case LLIT_UNMATCHED_PAIR:
 		rc = lfsck_layout_repair_unmatched_pair(env, com, llr, pla);
@@ -2356,23 +3403,33 @@ repair:
 out:
 	down_write(&com->lc_sem);
 	if (rc < 0) {
-		/* If cannot touch the target server,
-		 * mark the LFSCK as INCOMPLETE. */
-		if (rc == -ENOTCONN || rc == -ESHUTDOWN || rc == -ETIMEDOUT ||
-		    rc == -EHOSTDOWN || rc == -EHOSTUNREACH) {
+		struct lfsck_layout_master_data *llmd = com->lc_data;
+
+		if (unlikely(llmd->llmd_exit)) {
+			rc = 0;
+		} else if (rc == -ENOTCONN || rc == -ESHUTDOWN ||
+			   rc == -ETIMEDOUT || rc == -EHOSTDOWN ||
+			   rc == -EHOSTUNREACH) {
+			/* If cannot touch the target server,
+			 * mark the LFSCK as INCOMPLETE. */
 			CERROR("%s: Fail to talk with OST %x: rc = %d.\n",
 			       lfsck_lfsck2name(lfsck), llr->llr_ost_idx, rc);
 			lo->ll_flags |= LF_INCOMPLETE;
 			lo->ll_objs_skipped++;
 			rc = 0;
 		} else {
-			lo->ll_objs_failed_phase1++;
+			lfsck_layout_record_failure(env, lfsck, lo);
 		}
 	} else if (rc > 0) {
 		LASSERTF(type > LLIT_NONE && type <= LLIT_MAX,
 			 "unknown type = %d\n", type);
 
 		lo->ll_objs_repaired[type - 1]++;
+		if (bk->lb_param & LPF_DRYRUN &&
+		    unlikely(lo->ll_pos_first_inconsistent == 0))
+			lo->ll_pos_first_inconsistent =
+			lfsck->li_obj_oit->do_index_ops->dio_it.store(env,
+							lfsck->li_di_oit);
 	}
 	up_write(&com->lc_sem);
 
@@ -2401,7 +3458,7 @@ static int lfsck_layout_assistant(void *args)
 	memset(lr, 0, sizeof(*lr));
 	lr->lr_event = LE_START;
 	lr->lr_valid = LSV_SPEED_LIMIT | LSV_ERROR_HANDLE | LSV_DRYRUN |
-		       LSV_ASYNC_WINDOWS;
+		       LSV_ASYNC_WINDOWS | LSV_CREATE_OSTOBJ;
 	lr->lr_speed = bk->lb_speed_limit;
 	lr->lr_version = bk->lb_version;
 	lr->lr_param = bk->lb_param;
@@ -2426,7 +3483,8 @@ static int lfsck_layout_assistant(void *args)
 		while (!list_empty(&llmd->llmd_req_list)) {
 			bool wakeup = false;
 
-			if (unlikely(llmd->llmd_exit))
+			if (unlikely(llmd->llmd_exit ||
+				     !thread_is_running(mthread)))
 				GOTO(cleanup1, rc = llmd->llmd_post_result);
 
 			llr = list_entry(llmd->llmd_req_list.next,
@@ -2440,11 +3498,15 @@ static int lfsck_layout_assistant(void *args)
 			rc = lfsck_layout_assistant_handle_one(env, com, llr);
 			spin_lock(&llmd->llmd_lock);
 			list_del_init(&llr->llr_list);
-			if (bk->lb_async_windows != 0 &&
-			    llmd->llmd_prefetched >= bk->lb_async_windows)
-				wakeup = true;
-
 			llmd->llmd_prefetched--;
+			/* Wake up the main engine thread only when the list
+			 * is empty or half of the prefetched items have been
+			 * handled to avoid too frequent thread schedule. */
+			if (llmd->llmd_prefetched == 0 ||
+			    (bk->lb_async_windows != 0 &&
+			     bk->lb_async_windows / 2 ==
+			     llmd->llmd_prefetched))
+				wakeup = true;
 			spin_unlock(&llmd->llmd_lock);
 			if (wakeup)
 				wake_up_all(&mthread->t_ctl_waitq);
@@ -2454,9 +3516,6 @@ static int lfsck_layout_assistant(void *args)
 				GOTO(cleanup1, rc);
 		}
 
-		/* Wakeup the master engine if it is waiting in checkpoint. */
-		wake_up_all(&mthread->t_ctl_waitq);
-
 		l_wait_event(athread->t_ctl_waitq,
 			     !lfsck_layout_req_empty(llmd) ||
 			     llmd->llmd_exit ||
@@ -2500,6 +3559,9 @@ static int lfsck_layout_assistant(void *args)
 				com->lc_time_last_checkpoint +
 				cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
 
+			/* flush all async updating before handling orphan. */
+			dt_sync(env, lfsck->li_next);
+
 			while (llmd->llmd_in_double_scan) {
 				struct lfsck_tgt_descs	*ltds =
 							&lfsck->li_ost_descs;
@@ -2638,7 +3700,7 @@ cleanup2:
 	/* Under force exit case, some requests may be just freed without
 	 * verification, those objects should be re-handled when next run.
 	 * So not update the on-disk tracing file under such case. */
-	if (!llmd->llmd_exit)
+	if (llmd->llmd_in_double_scan && !llmd->llmd_exit)
 		rc1 = lfsck_layout_double_scan_result(env, com, rc);
 
 fini:
@@ -2887,6 +3949,209 @@ lfsck_layout_slave_notify_master(const struct lu_env *env,
 	RETURN_EXIT;
 }
 
+/*
+ * \ret -ENODATA: unrecognized stripe
+ * \ret = 0     : recognized stripe
+ * \ret < 0     : other failures
+ */
+static int lfsck_layout_master_check_pairs(const struct lu_env *env,
+					   struct lfsck_component *com,
+					   struct lu_fid *cfid,
+					   struct lu_fid *pfid)
+{
+	struct lfsck_thread_info	*info	= lfsck_env_info(env);
+	struct lu_buf			*buf	= &info->lti_big_buf;
+	struct ost_id			*oi     = &info->lti_oi;
+	struct dt_object		*obj;
+	struct lov_mds_md_v1		*lmm;
+	struct lov_ost_data_v1		*objs;
+	__u32				 idx	= pfid->f_stripe_idx;
+	__u32				 magic;
+	int				 rc	= 0;
+	int				 i;
+	__u16				 count;
+	ENTRY;
+
+	pfid->f_ver = 0;
+	obj = lfsck_object_find_by_dev(env, com->lc_lfsck->li_bottom, pfid);
+	if (IS_ERR(obj))
+		RETURN(PTR_ERR(obj));
+
+	dt_read_lock(env, obj, 0);
+	if (unlikely(!dt_object_exists(obj)))
+		GOTO(unlock, rc = -ENOENT);
+
+	rc = lfsck_layout_get_lovea(env, obj, buf, NULL);
+	if (rc < 0)
+		GOTO(unlock, rc);
+
+	if (rc == 0)
+		GOTO(unlock, rc = -ENODATA);
+
+	lmm = buf->lb_buf;
+	rc = lfsck_layout_verify_header(lmm);
+	if (rc != 0)
+		GOTO(unlock, rc);
+
+	/* Currently, we only support LOV_MAGIC_V1/LOV_MAGIC_V3 which has
+	 * been verified in lfsck_layout_verify_header() already. If some
+	 * new magic introduced in the future, then layout LFSCK needs to
+	 * be updated also. */
+	magic = le32_to_cpu(lmm->lmm_magic);
+	if (magic == LOV_MAGIC_V1) {
+		objs = &lmm->lmm_objects[0];
+	} else {
+		LASSERT(magic == LOV_MAGIC_V3);
+		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
+	}
+
+	fid_to_ostid(cfid, oi);
+	count = le16_to_cpu(lmm->lmm_stripe_count);
+	for (i = 0; i < count; i++, objs++) {
+		struct ost_id oi2;
+
+		ostid_le_to_cpu(&objs->l_ost_oi, &oi2);
+		if (memcmp(oi, &oi2, sizeof(*oi)) == 0)
+			GOTO(unlock, rc = (i != idx ? -ENODATA : 0));
+	}
+
+	GOTO(unlock, rc = -ENODATA);
+
+unlock:
+	dt_read_unlock(env, obj);
+	lu_object_put(env, &obj->do_lu);
+
+	return rc;
+}
+
+/*
+ * The LFSCK-on-OST will ask the LFSCK-on-MDT to check whether the given
+ * MDT-object/OST-object pairs match or not to aviod transfer MDT-object
+ * layout EA from MDT to OST. On one hand, the OST no need to understand
+ * the layout EA structure; on the other hand, it may cause trouble when
+ * transfer large layout EA from MDT to OST via normal OUT RPC.
+ *
+ * \ret > 0: unrecognized stripe
+ * \ret = 0: recognized stripe
+ * \ret < 0: other failures
+ */
+static int lfsck_layout_slave_check_pairs(const struct lu_env *env,
+					  struct lfsck_component *com,
+					  struct lu_fid *cfid,
+					  struct lu_fid *pfid)
+{
+	struct lfsck_instance	 *lfsck	 = com->lc_lfsck;
+	struct obd_device	 *obd	 = lfsck->li_obd;
+	struct seq_server_site	 *ss	 =
+			lu_site2seq(lfsck->li_bottom->dd_lu_dev.ld_site);
+	struct obd_export	 *exp	 = NULL;
+	struct ptlrpc_request	 *req	 = NULL;
+	struct lfsck_request	 *lr;
+	struct lu_seq_range	  range	 = { 0 };
+	int			  rc	 = 0;
+	ENTRY;
+
+	if (unlikely(fid_is_idif(pfid)))
+		RETURN(1);
+
+	fld_range_set_any(&range);
+	rc = fld_server_lookup(env, ss->ss_server_fld, fid_seq(pfid), &range);
+	if (rc != 0)
+		RETURN(rc == -ENOENT ? 1 : rc);
+
+	if (unlikely(!fld_range_is_mdt(&range)))
+		RETURN(1);
+
+	exp = lustre_find_lwp_by_index(obd->obd_name, range.lsr_index);
+	if (unlikely(exp == NULL))
+		RETURN(1);
+
+	if (!(exp_connect_flags(exp) & OBD_CONNECT_LFSCK))
+		GOTO(out, rc = -EOPNOTSUPP);
+
+	req = ptlrpc_request_alloc(class_exp2cliimp(exp), &RQF_LFSCK_NOTIFY);
+	if (req == NULL)
+		GOTO(out, rc = -ENOMEM);
+
+	rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, LFSCK_NOTIFY);
+	if (rc != 0) {
+		ptlrpc_request_free(req);
+
+		GOTO(out, rc);
+	}
+
+	lr = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
+	memset(lr, 0, sizeof(*lr));
+	lr->lr_event = LE_PAIRS_VERIFY;
+	lr->lr_active = LT_LAYOUT;
+	lr->lr_fid = *cfid; /* OST-object itself FID. */
+	lr->lr_fid2 = *pfid; /* The claimed parent FID. */
+
+	ptlrpc_request_set_replen(req);
+	rc = ptlrpc_queue_wait(req);
+	ptlrpc_req_finished(req);
+
+	if (rc == -ENOENT || rc == -ENODATA)
+		rc = 1;
+
+	GOTO(out, rc);
+
+out:
+	if (exp != NULL)
+		class_export_put(exp);
+
+	return rc;
+}
+
+static int lfsck_layout_slave_repair_pfid(const struct lu_env *env,
+					  struct lfsck_component *com,
+					  struct lfsck_request *lr)
+{
+	struct lfsck_thread_info	*info	= lfsck_env_info(env);
+	struct filter_fid		*ff	= &info->lti_new_pfid;
+	struct lu_buf			*buf;
+	struct dt_device		*dev	= com->lc_lfsck->li_bottom;
+	struct dt_object		*obj;
+	struct thandle			*th	= NULL;
+	int				 rc	= 0;
+	ENTRY;
+
+	obj = lfsck_object_find_by_dev(env, dev, &lr->lr_fid);
+	if (IS_ERR(obj))
+		RETURN(PTR_ERR(obj));
+
+	fid_cpu_to_le(&ff->ff_parent, &lr->lr_fid2);
+	buf = lfsck_buf_get(env, ff, sizeof(*ff));
+	dt_write_lock(env, obj, 0);
+	if (unlikely(!dt_object_exists(obj)))
+		GOTO(unlock, rc = 0);
+
+	th = dt_trans_create(env, dev);
+	if (IS_ERR(th))
+		GOTO(unlock, rc = PTR_ERR(th));
+
+	rc = dt_declare_xattr_set(env, obj, buf, XATTR_NAME_FID, 0, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_trans_start_local(env, dev, th);
+	if (rc != 0)
+		GOTO(stop, rc);
+
+	rc = dt_xattr_set(env, obj, buf, XATTR_NAME_FID, 0, th, BYPASS_CAPA);
+
+	GOTO(stop, rc);
+
+stop:
+	dt_trans_stop(env, dev, th);
+
+unlock:
+	dt_write_unlock(env, obj);
+	lu_object_put(env, &obj->do_lu);
+
+	return rc;
+}
+
 /* layout APIs */
 
 static int lfsck_layout_reset(const struct lu_env *env,
@@ -2924,14 +4189,7 @@ static void lfsck_layout_fail(const struct lu_env *env,
 	down_write(&com->lc_sem);
 	if (new_checked)
 		com->lc_new_checked++;
-	lo->ll_objs_failed_phase1++;
-	if (lo->ll_pos_first_inconsistent == 0) {
-		struct lfsck_instance *lfsck = com->lc_lfsck;
-
-		lo->ll_pos_first_inconsistent =
-			lfsck->li_obj_oit->do_index_ops->dio_it.store(env,
-							lfsck->li_di_oit);
-	}
+	lfsck_layout_record_failure(env, com->lc_lfsck, lo);
 	up_write(&com->lc_sem);
 }
 
@@ -3025,6 +4283,9 @@ static int lfsck_layout_prep(const struct lu_env *env,
 		int rc;
 
 		rc = lfsck_layout_reset(env, com, false);
+		if (rc == 0)
+			rc = lfsck_set_param(env, lfsck, start, true);
+
 		if (rc != 0)
 			return rc;
 	}
@@ -3075,13 +4336,26 @@ static int lfsck_layout_slave_prep(const struct lu_env *env,
 				   struct lfsck_start_param *lsp)
 {
 	struct lfsck_layout_slave_data	*llsd	= com->lc_data;
+	struct lfsck_instance		*lfsck	= com->lc_lfsck;
+	struct lfsck_layout		*lo	= com->lc_file_ram;
 	struct lfsck_start		*start  = lsp->lsp_start;
 	int				 rc;
 
 	rc = lfsck_layout_prep(env, com, start);
-	if (rc != 0 || !lsp->lsp_index_valid)
+	if (rc != 0)
 		return rc;
 
+	if (lo->ll_flags & LF_CRASHED_LASTID &&
+	    list_empty(&llsd->llsd_master_list)) {
+		LASSERT(lfsck->li_out_notify != NULL);
+
+		lfsck->li_out_notify(env, lfsck->li_out_notify_data,
+				     LE_LASTID_REBUILDING);
+	}
+
+	if (!lsp->lsp_index_valid)
+		return 0;
+
 	rc = lfsck_layout_llst_add(llsd, lsp->lsp_index);
 	if (rc == 0 && start != NULL && start->ls_flags & LPF_ORPHAN) {
 		LASSERT(!llsd->llsd_rbtree_valid);
@@ -3103,7 +4377,8 @@ static int lfsck_layout_master_prep(const struct lu_env *env,
 	struct ptlrpc_thread		*mthread = &lfsck->li_thread;
 	struct ptlrpc_thread		*athread = &llmd->llmd_thread;
 	struct lfsck_thread_args	*lta;
-	long				 rc;
+	struct task_struct		*task;
+	int				 rc;
 	ENTRY;
 
 	rc = lfsck_layout_prep(env, com, lsp->lsp_start);
@@ -3122,10 +4397,11 @@ static int lfsck_layout_master_prep(const struct lu_env *env,
 	if (IS_ERR(lta))
 		RETURN(PTR_ERR(lta));
 
-	rc = PTR_ERR(kthread_run(lfsck_layout_assistant, lta, "lfsck_layout"));
-	if (IS_ERR_VALUE(rc)) {
+	task = kthread_run(lfsck_layout_assistant, lta, "lfsck_layout");
+	if (IS_ERR(task)) {
+		rc = PTR_ERR(task);
 		CERROR("%s: Cannot start LFSCK layout assistant thread: "
-		       "rc = %ld\n", lfsck_lfsck2name(lfsck), rc);
+		       "rc = %d\n", lfsck_lfsck2name(lfsck), rc);
 		lfsck_thread_args_fini(lta);
 	} else {
 		struct l_wait_info lwi = { 0 };
@@ -3178,7 +4454,7 @@ static int lfsck_layout_scan_stripes(const struct lu_env *env,
 	 * be updated also. */
 	magic = le32_to_cpu(lmm->lmm_magic);
 	if (magic == LOV_MAGIC_V1) {
-		objs = &(lmm->lmm_objects[0]);
+		objs = &lmm->lmm_objects[0];
 	} else {
 		LASSERT(magic == LOV_MAGIC_V3);
 		objs = &((struct lov_mds_md_v3 *)lmm)->lmm_objects[0];
@@ -3194,6 +4470,9 @@ static int lfsck_layout_scan_stripes(const struct lu_env *env,
 					le32_to_cpu(objs->l_ost_idx);
 		bool			 wakeup = false;
 
+		if (unlikely(lovea_slot_is_dummy(objs)))
+			continue;
+
 		l_wait_event(mthread->t_ctl_waitq,
 			     bk->lb_async_windows == 0 ||
 			     llmd->llmd_prefetched < bk->lb_async_windows ||
@@ -3267,7 +4546,7 @@ next:
 		down_write(&com->lc_sem);
 		com->lc_new_checked++;
 		if (rc < 0)
-			lo->ll_objs_failed_phase1++;
+			lfsck_layout_record_failure(env, lfsck, lo);
 		up_write(&com->lc_sem);
 
 		if (cobj != NULL && !IS_ERR(cobj))
@@ -3337,6 +4616,8 @@ again:
 	buf->lb_len = rc;
 	lmm = buf->lb_buf;
 	rc = lfsck_layout_verify_header(lmm);
+	/* If the LOV EA crashed, then it is possible to be rebuilt later
+	 * when handle orphan OST-objects. */
 	if (rc != 0)
 		GOTO(out, rc);
 
@@ -3414,7 +4695,7 @@ out:
 		down_write(&com->lc_sem);
 		com->lc_new_checked++;
 		if (rc < 0)
-			lo->ll_objs_failed_phase1++;
+			lfsck_layout_record_failure(env, lfsck, lo);
 		up_write(&com->lc_sem);
 	}
 	buf->lb_len = buflen;
@@ -3438,6 +4719,17 @@ static int lfsck_layout_slave_exec_oit(const struct lu_env *env,
 
 	LASSERT(llsd != NULL);
 
+	if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY5) &&
+	    cfs_fail_val == lfsck_dev_idx(lfsck->li_bottom)) {
+		struct l_wait_info	 lwi = LWI_TIMEOUT(cfs_time_seconds(1),
+							   NULL, NULL);
+		struct ptlrpc_thread	*thread = &lfsck->li_thread;
+
+		l_wait_event(thread->t_ctl_waitq,
+			     !thread_is_running(thread),
+			     &lwi);
+	}
+
 	lfsck_rbtree_update_bitmap(env, com, fid, false);
 
 	down_write(&com->lc_sem);
@@ -3750,7 +5042,8 @@ static int lfsck_layout_dump(const struct lu_env *env,
 		const struct dt_it_ops *iops;
 		cfs_duration_t duration = cfs_time_current() -
 					  lfsck->li_time_last_checkpoint;
-		__u64 checked = lo->ll_objs_checked_phase1 + com->lc_new_checked;
+		__u64 checked = lo->ll_objs_checked_phase1 +
+				com->lc_new_checked;
 		__u64 speed = checked;
 		__u64 new_checked = com->lc_new_checked * HZ;
 		__u32 rtime = lo->ll_run_time_phase1 +
@@ -3801,31 +5094,36 @@ static int lfsck_layout_dump(const struct lu_env *env,
 	} else if (lo->ll_status == LS_SCANNING_PHASE2) {
 		cfs_duration_t duration = cfs_time_current() -
 					  lfsck->li_time_last_checkpoint;
-		__u64 checked = lo->ll_objs_checked_phase1 + com->lc_new_checked;
-		__u64 speed = checked;
+		__u64 checked = lo->ll_objs_checked_phase2 +
+				com->lc_new_checked;
+		__u64 speed1 = lo->ll_objs_checked_phase1;
+		__u64 speed2 = checked;
 		__u64 new_checked = com->lc_new_checked * HZ;
-		__u32 rtime = lo->ll_run_time_phase1 +
+		__u32 rtime = lo->ll_run_time_phase2 +
 			      cfs_duration_sec(duration + HALF_SEC);
 
 		if (duration != 0)
 			do_div(new_checked, duration);
+		if (lo->ll_run_time_phase1 != 0)
+			do_div(speed1, lo->ll_run_time_phase1);
 		if (rtime != 0)
-			do_div(speed, rtime);
+			do_div(speed2, rtime);
 		rc = snprintf(buf, len,
 			      "checked_phase1: "LPU64"\n"
 			      "checked_phase2: "LPU64"\n"
 			      "run_time_phase1: %u seconds\n"
 			      "run_time_phase2: %u seconds\n"
 			      "average_speed_phase1: "LPU64" items/sec\n"
-			      "average_speed_phase2: N/A\n"
-			      "real-time_speed_phase1: "LPU64" items/sec\n"
-			      "real-time_speed_phase2: N/A\n"
+			      "average_speed_phase2: "LPU64" items/sec\n"
+			      "real-time_speed_phase1: N/A\n"
+			      "real-time_speed_phase2: "LPU64" items/sec\n"
 			      "current_position: "DFID"\n",
+			      lo->ll_objs_checked_phase1,
 			      checked,
-			      lo->ll_objs_checked_phase2,
+			      lo->ll_run_time_phase1,
 			      rtime,
-			      lo->ll_run_time_phase2,
-			      speed,
+			      speed1,
+			      speed2,
 			      new_checked,
 			      PFID(&com->lc_fid_latest_scanned_phase2));
 		if (rc <= 0)
@@ -4070,6 +5368,15 @@ static int lfsck_layout_master_in_notify(const struct lu_env *env,
 	bool				 fail  = false;
 	ENTRY;
 
+	if (lr->lr_event == LE_PAIRS_VERIFY) {
+		int rc;
+
+		rc = lfsck_layout_master_check_pairs(env, com, &lr->lr_fid,
+						     &lr->lr_fid2);
+
+		RETURN(rc);
+	}
+
 	if (lr->lr_event != LE_PHASE1_DONE &&
 	    lr->lr_event != LE_PHASE2_DONE &&
 	    lr->lr_event != LE_PEER_EXIT)
@@ -4084,7 +5391,7 @@ static int lfsck_layout_master_in_notify(const struct lu_env *env,
 	if (ltd == NULL) {
 		spin_unlock(&ltds->ltd_lock);
 
-		RETURN(-ENODEV);
+		RETURN(-ENXIO);
 	}
 
 	list_del_init(&ltd->ltd_layout_phase_list);
@@ -4158,20 +5465,59 @@ static int lfsck_layout_slave_in_notify(const struct lu_env *env,
 	struct lfsck_instance		 *lfsck = com->lc_lfsck;
 	struct lfsck_layout_slave_data	 *llsd  = com->lc_data;
 	struct lfsck_layout_slave_target *llst;
+	int				  rc;
 	ENTRY;
 
-	if (lr->lr_event == LE_FID_ACCESSED) {
+	switch (lr->lr_event) {
+	case LE_FID_ACCESSED:
 		lfsck_rbtree_update_bitmap(env, com, &lr->lr_fid, true);
-
 		RETURN(0);
-	}
+	case LE_CONDITIONAL_DESTROY:
+		rc = lfsck_layout_slave_conditional_destroy(env, com, lr);
+		RETURN(rc);
+	case LE_PAIRS_VERIFY: {
+		lr->lr_status = LPVS_INIT;
+		/* Firstly, if the MDT-object which is claimed via OST-object
+		 * local stored PFID xattr recognizes the OST-object, then it
+		 * must be that the client given PFID is wrong. */
+		rc = lfsck_layout_slave_check_pairs(env, com, &lr->lr_fid,
+						    &lr->lr_fid3);
+		if (rc <= 0)
+			RETURN(0);
+
+		lr->lr_status = LPVS_INCONSISTENT;
+		/* The OST-object local stored PFID xattr is stale. We need to
+		 * check whether the MDT-object that is claimed via the client
+		 * given PFID information recognizes the OST-object or not. If
+		 * matches, then need to update the OST-object's PFID xattr. */
+		rc = lfsck_layout_slave_check_pairs(env, com, &lr->lr_fid,
+						    &lr->lr_fid2);
+		/* For rc < 0 case:
+		 * We are not sure whether the client given PFID information
+		 * is correct or not, do nothing to avoid improper fixing.
+		 *
+		 * For rc > 0 case:
+		 * The client given PFID information is also invalid, we can
+		 * NOT fix the OST-object inconsistency.
+		 */
+		if (rc != 0)
+			RETURN(rc);
 
-	if (lr->lr_event != LE_PHASE2_DONE && lr->lr_event != LE_PEER_EXIT)
+		lr->lr_status = LPVS_INCONSISTENT_TOFIX;
+		rc = lfsck_layout_slave_repair_pfid(env, com, lr);
+
+		RETURN(rc);
+	}
+	case LE_PHASE2_DONE:
+	case LE_PEER_EXIT:
+		break;
+	default:
 		RETURN(-EINVAL);
+	}
 
 	llst = lfsck_layout_llst_find_and_del(llsd, lr->lr_index, true);
 	if (llst == NULL)
-		RETURN(-ENODEV);
+		RETURN(-ENXIO);
 
 	lfsck_layout_llst_put(llst);
 	if (list_empty(&llsd->llsd_master_list))
@@ -4603,7 +5949,7 @@ static struct dt_it *lfsck_orphan_it_init(const struct lu_env *env,
 
 	lfsck = lfsck_instance_find(dev, true, false);
 	if (unlikely(lfsck == NULL))
-		RETURN(ERR_PTR(-ENODEV));
+		RETURN(ERR_PTR(-ENXIO));
 
 	com = lfsck_component_find(lfsck, LT_LAYOUT);
 	if (unlikely(com == NULL))
@@ -4619,10 +5965,10 @@ static struct dt_it *lfsck_orphan_it_init(const struct lu_env *env,
 
 	it->loi_llst = lfsck_layout_llst_find_and_del(llsd, attr, false);
 	if (it->loi_llst == NULL)
-		GOTO(out, rc = -ENODEV);
+		GOTO(out, rc = -ENXIO);
 
 	if (dev->dd_record_fid_accessed) {
-		/* The first iteratino against the rbtree, scan the whole rbtree
+		/* The first iteration against the rbtree, scan the whole rbtree
 		 * to remove the nodes which do NOT need to be handled. */
 		write_lock(&llsd->llsd_rb_lock);
 		if (dev->dd_record_fid_accessed) {
@@ -4863,10 +6209,10 @@ again1:
 		GOTO(out, rc = -EINVAL);
 
 	fid_le_to_cpu(&rec->lor_fid, &pfid->ff_parent);
-	/* In fact, the ff_parent::f_ver is not the real parent FID::f_ver,
-	 * instead, it is the OST-object index in its parent MDT-object
-	 * layout EA. */
-	save = rec->lor_fid.f_ver;
+	/* Currently, the filter_fid::ff_parent::f_ver is not the real parent
+	 * MDT-object's FID::f_ver, instead it is the OST-object index in its
+	 * parent MDT-object's layout EA. */
+	save = rec->lor_fid.f_stripe_idx;
 	rec->lor_fid.f_ver = 0;
 	rc = lfsck_fid_match_idx(env, lfsck, &rec->lor_fid, idx);
 	/* If the orphan OST-object does not claim the MDT, then next.
@@ -4880,7 +6226,7 @@ again1:
 		goto again1;
 	}
 
-	rec->lor_fid.f_ver = save;
+	rec->lor_fid.f_stripe_idx = save;
 	rec->lor_uid = la->la_uid;
 	rec->lor_gid = la->la_gid;