4 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
6 * This program is free software; you can redistribute it and/or modify
7 * it under the terms of the GNU General Public License version 2 only,
8 * as published by the Free Software Foundation.
10 * This program is distributed in the hope that it will be useful,
11 * but WITHOUT ANY WARRANTY; without even the implied warranty of
12 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 * GNU General Public License version 2 for more details. A copy is
14 * included in the COPYING file that accompanied this code.
16 * You should have received a copy of the GNU General Public License
17 * along with this program; if not, write to the Free Software
18 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
23 * Copyright (c) 2012, 2013, Intel Corporation.
26 * lustre/lfsck/lfsck_lib.c
28 * Author: Fan, Yong <fan.yong@intel.com>
31 #define DEBUG_SUBSYSTEM S_LFSCK
33 #include <libcfs/list.h>
34 #include <lu_object.h>
35 #include <dt_object.h>
36 #include <md_object.h>
37 #include <lustre_fld.h>
38 #include <lustre_lib.h>
39 #include <lustre_net.h>
40 #include <lustre_lfsck.h>
41 #include <lustre/lustre_lfsck_user.h>
43 #include "lfsck_internal.h"
45 /* define lfsck thread key */
46 LU_KEY_INIT(lfsck, struct lfsck_thread_info);
48 static void lfsck_key_fini(const struct lu_context *ctx,
49 struct lu_context_key *key, void *data)
51 struct lfsck_thread_info *info = data;
53 lu_buf_free(&info->lti_linkea_buf);
54 lu_buf_free(&info->lti_big_buf);
58 LU_CONTEXT_KEY_DEFINE(lfsck, LCT_MD_THREAD | LCT_DT_THREAD);
59 LU_KEY_INIT_GENERIC(lfsck);
61 static CFS_LIST_HEAD(lfsck_instance_list);
62 static struct list_head lfsck_ost_orphan_list;
63 static struct list_head lfsck_mdt_orphan_list;
64 static DEFINE_SPINLOCK(lfsck_instance_lock);
66 static const char *lfsck_status_names[] = {
68 [LS_SCANNING_PHASE1] = "scanning-phase1",
69 [LS_SCANNING_PHASE2] = "scanning-phase2",
70 [LS_COMPLETED] = "completed",
71 [LS_FAILED] = "failed",
72 [LS_STOPPED] = "stopped",
73 [LS_PAUSED] = "paused",
74 [LS_CRASHED] = "crashed",
75 [LS_PARTIAL] = "partial",
76 [LS_CO_FAILED] = "co-failed",
77 [LS_CO_STOPPED] = "co-stopped",
78 [LS_CO_PAUSED] = "co-paused"
81 const char *lfsck_flags_names[] = {
90 const char *lfsck_param_names[] = {
100 const char *lfsck_status2names(enum lfsck_status status)
102 if (unlikely(status < 0 || status >= LS_MAX))
105 return lfsck_status_names[status];
108 static int lfsck_tgt_descs_init(struct lfsck_tgt_descs *ltds)
110 spin_lock_init(<ds->ltd_lock);
111 init_rwsem(<ds->ltd_rw_sem);
112 INIT_LIST_HEAD(<ds->ltd_orphan);
113 ltds->ltd_tgts_bitmap = CFS_ALLOCATE_BITMAP(BITS_PER_LONG);
114 if (ltds->ltd_tgts_bitmap == NULL)
120 static void lfsck_tgt_descs_fini(struct lfsck_tgt_descs *ltds)
122 struct lfsck_tgt_desc *ltd;
123 struct lfsck_tgt_desc *next;
126 down_write(<ds->ltd_rw_sem);
128 list_for_each_entry_safe(ltd, next, <ds->ltd_orphan,
130 list_del_init(<d->ltd_orphan_list);
134 if (unlikely(ltds->ltd_tgts_bitmap == NULL)) {
135 up_write(<ds->ltd_rw_sem);
140 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
141 ltd = LTD_TGT(ltds, idx);
142 if (likely(ltd != NULL)) {
143 LASSERT(list_empty(<d->ltd_layout_list));
144 LASSERT(list_empty(<d->ltd_layout_phase_list));
147 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, idx);
148 LTD_TGT(ltds, idx) = NULL;
153 LASSERTF(ltds->ltd_tgtnr == 0, "tgt count unmatched: %d\n",
156 for (idx = 0; idx < TGT_PTRS; idx++) {
157 if (ltds->ltd_tgts_idx[idx] != NULL) {
158 OBD_FREE_PTR(ltds->ltd_tgts_idx[idx]);
159 ltds->ltd_tgts_idx[idx] = NULL;
163 CFS_FREE_BITMAP(ltds->ltd_tgts_bitmap);
164 ltds->ltd_tgts_bitmap = NULL;
165 up_write(<ds->ltd_rw_sem);
168 static int __lfsck_add_target(const struct lu_env *env,
169 struct lfsck_instance *lfsck,
170 struct lfsck_tgt_desc *ltd,
171 bool for_ost, bool locked)
173 struct lfsck_tgt_descs *ltds;
174 __u32 index = ltd->ltd_index;
179 ltds = &lfsck->li_ost_descs;
181 ltds = &lfsck->li_mdt_descs;
184 down_write(<ds->ltd_rw_sem);
186 LASSERT(ltds->ltd_tgts_bitmap != NULL);
188 if (index >= ltds->ltd_tgts_bitmap->size) {
189 __u32 newsize = max((__u32)ltds->ltd_tgts_bitmap->size,
190 (__u32)BITS_PER_LONG);
191 cfs_bitmap_t *old_bitmap = ltds->ltd_tgts_bitmap;
192 cfs_bitmap_t *new_bitmap;
194 while (newsize < index + 1)
197 new_bitmap = CFS_ALLOCATE_BITMAP(newsize);
198 if (new_bitmap == NULL)
199 GOTO(unlock, rc = -ENOMEM);
201 if (ltds->ltd_tgtnr > 0)
202 cfs_bitmap_copy(new_bitmap, old_bitmap);
203 ltds->ltd_tgts_bitmap = new_bitmap;
204 CFS_FREE_BITMAP(old_bitmap);
207 if (cfs_bitmap_check(ltds->ltd_tgts_bitmap, index)) {
208 CERROR("%s: the device %s (%u) is registered already\n",
209 lfsck_lfsck2name(lfsck),
210 ltd->ltd_tgt->dd_lu_dev.ld_obd->obd_name, index);
211 GOTO(unlock, rc = -EEXIST);
214 if (ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK] == NULL) {
215 OBD_ALLOC_PTR(ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK]);
216 if (ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK] == NULL)
217 GOTO(unlock, rc = -ENOMEM);
220 LTD_TGT(ltds, index) = ltd;
221 cfs_bitmap_set(ltds->ltd_tgts_bitmap, index);
224 GOTO(unlock, rc = 0);
228 up_write(<ds->ltd_rw_sem);
233 static int lfsck_add_target_from_orphan(const struct lu_env *env,
234 struct lfsck_instance *lfsck)
236 struct lfsck_tgt_descs *ltds = &lfsck->li_ost_descs;
237 struct lfsck_tgt_desc *ltd;
238 struct lfsck_tgt_desc *next;
239 struct list_head *head = &lfsck_ost_orphan_list;
244 spin_lock(&lfsck_instance_lock);
245 list_for_each_entry_safe(ltd, next, head, ltd_orphan_list) {
246 if (ltd->ltd_key == lfsck->li_bottom) {
247 list_del_init(<d->ltd_orphan_list);
248 list_add_tail(<d->ltd_orphan_list,
252 spin_unlock(&lfsck_instance_lock);
254 down_write(<ds->ltd_rw_sem);
255 while (!list_empty(<ds->ltd_orphan)) {
256 ltd = list_entry(ltds->ltd_orphan.next,
257 struct lfsck_tgt_desc,
259 list_del_init(<d->ltd_orphan_list);
260 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, true);
261 /* Do not hold the semaphore for too long time. */
262 up_write(<ds->ltd_rw_sem);
266 down_write(<ds->ltd_rw_sem);
268 up_write(<ds->ltd_rw_sem);
271 ltds = &lfsck->li_mdt_descs;
272 head = &lfsck_mdt_orphan_list;
280 static inline struct lfsck_component *
281 __lfsck_component_find(struct lfsck_instance *lfsck, __u16 type, cfs_list_t *list)
283 struct lfsck_component *com;
285 cfs_list_for_each_entry(com, list, lc_link) {
286 if (com->lc_type == type)
292 struct lfsck_component *
293 lfsck_component_find(struct lfsck_instance *lfsck, __u16 type)
295 struct lfsck_component *com;
297 spin_lock(&lfsck->li_lock);
298 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
302 com = __lfsck_component_find(lfsck, type,
303 &lfsck->li_list_double_scan);
307 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_idle);
311 lfsck_component_get(com);
312 spin_unlock(&lfsck->li_lock);
316 void lfsck_component_cleanup(const struct lu_env *env,
317 struct lfsck_component *com)
319 if (!cfs_list_empty(&com->lc_link))
320 cfs_list_del_init(&com->lc_link);
321 if (!cfs_list_empty(&com->lc_link_dir))
322 cfs_list_del_init(&com->lc_link_dir);
324 lfsck_component_put(env, com);
327 int lfsck_fid_alloc(const struct lu_env *env, struct lfsck_instance *lfsck,
328 struct lu_fid *fid, bool locked)
330 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
335 mutex_lock(&lfsck->li_mutex);
337 rc = seq_client_alloc_fid(env, lfsck->li_seq, fid);
339 bk->lb_last_fid = *fid;
340 /* We do not care about whether the subsequent sub-operations
341 * failed or not. The worst case is that one FID is lost that
342 * is not a big issue for the LFSCK since it is relative rare
343 * for LFSCK create. */
344 rc = lfsck_bookmark_store(env, lfsck);
348 mutex_unlock(&lfsck->li_mutex);
353 static const char dot[] = ".";
354 static const char dotdot[] = "..";
356 static int lfsck_create_lpf_local(const struct lu_env *env,
357 struct lfsck_instance *lfsck,
358 struct dt_object *parent,
359 struct dt_object *child,
361 struct dt_object_format *dof,
364 struct dt_device *dev = lfsck->li_bottom;
365 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
366 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
367 const struct lu_fid *cfid = lu_object_fid(&child->do_lu);
368 struct thandle *th = NULL;
370 int len = sizeof(struct lfsck_bookmark);
374 th = dt_trans_create(env, dev);
378 /* 1a. create child */
379 rc = dt_declare_create(env, child, la, NULL, dof, th);
383 /* 2a. increase child nlink */
384 rc = dt_declare_ref_add(env, child, th);
388 /* 3a. insert name into parent dir */
389 rc = dt_declare_insert(env, parent, (const struct dt_rec *)cfid,
390 (const struct dt_key *)name, th);
394 /* 4a. increase parent nlink */
395 rc = dt_declare_ref_add(env, parent, th);
399 /* 5a. update bookmark */
400 rc = dt_declare_record_write(env, bk_obj,
401 lfsck_buf_get(env, bk, len), 0, th);
405 rc = dt_trans_start_local(env, dev, th);
409 dt_write_lock(env, child, 0);
410 /* 1b.1 create child */
411 rc = dt_create(env, child, la, NULL, dof, th);
415 if (unlikely(!dt_try_as_dir(env, child)))
416 GOTO(unlock, rc = -ENOTDIR);
418 /* 1b.2 insert dot into child dir */
419 rc = dt_insert(env, child, (const struct dt_rec *)cfid,
420 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
424 /* 1b.3 insert dotdot into child dir */
425 rc = dt_insert(env, child, (const struct dt_rec *)&LU_LPF_FID,
426 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
430 /* 2b. increase child nlink */
431 rc = dt_ref_add(env, child, th);
432 dt_write_unlock(env, child);
436 /* 3b. insert name into parent dir */
437 rc = dt_insert(env, parent, (const struct dt_rec *)cfid,
438 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
442 dt_write_lock(env, parent, 0);
443 /* 4b. increase parent nlink */
444 rc = dt_ref_add(env, parent, th);
445 dt_write_unlock(env, parent);
449 bk->lb_lpf_fid = *cfid;
450 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
452 /* 5b. update bookmark */
453 rc = dt_record_write(env, bk_obj,
454 lfsck_buf_get(env, bk, len), &pos, th);
459 dt_write_unlock(env, child);
462 dt_trans_stop(env, dev, th);
467 static int lfsck_create_lpf_remote(const struct lu_env *env,
468 struct lfsck_instance *lfsck,
469 struct dt_object *parent,
470 struct dt_object *child,
472 struct dt_object_format *dof,
475 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
476 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
477 const struct lu_fid *cfid = lu_object_fid(&child->do_lu);
478 struct thandle *th = NULL;
479 struct dt_device *dev;
481 int len = sizeof(struct lfsck_bookmark);
485 /* Create .lustre/lost+found/MDTxxxx. */
487 /* XXX: Currently, cross-MDT create operation needs to create the child
488 * object firstly, then insert name into the parent directory. For
489 * this case, the child object resides on current MDT (local), but
490 * the parent ".lustre/lost+found" may be on remote MDT. It is not
491 * easy to contain all the sub-modifications orderly within single
494 * To avoid more inconsistency, we split the create operation into
497 * 1) create the child locally.
498 * 2) insert the name "MDTXXXX" in the parent ".lustre/lost+found"
499 * remotely and update the lfsck_bookmark::lb_lpf_fid locally.
501 * If 1) done but 2) failed, then the worst case is that we lose
502 * one object locally, which is not a big issue. (can be repaird
503 * by LFSCK phase III) */
507 dev = lfsck->li_bottom;
508 th = dt_trans_create(env, dev);
512 /* 1a. create child locally. */
513 rc = dt_declare_create(env, child, la, NULL, dof, th);
517 /* 2a. increase child nlink locally. */
518 rc = dt_declare_ref_add(env, child, th);
522 rc = dt_trans_start_local(env, dev, th);
526 dt_write_lock(env, child, 0);
527 /* 1b. create child locally. */
528 rc = dt_create(env, child, la, NULL, dof, th);
532 if (unlikely(!dt_try_as_dir(env, child)))
533 GOTO(unlock, rc = -ENOTDIR);
535 /* 2b.1 insert dot into child dir locally. */
536 rc = dt_insert(env, child, (const struct dt_rec *)cfid,
537 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
541 /* 2b.2 insert dotdot into child dir locally. */
542 rc = dt_insert(env, child, (const struct dt_rec *)&LU_LPF_FID,
543 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
547 /* 2b.3 increase child nlink locally. */
548 rc = dt_ref_add(env, child, th);
549 dt_write_unlock(env, child);
550 dt_trans_stop(env, dev, th);
554 /* Transaction II: */
556 dev = lfsck->li_next;
557 th = dt_trans_create(env, dev);
561 /* 3a. insert name into parent dir remotely. */
562 rc = dt_declare_insert(env, parent, (const struct dt_rec *)cfid,
563 (const struct dt_key *)name, th);
567 /* 4a. increase parent nlink remotely. */
568 rc = dt_declare_ref_add(env, parent, th);
572 /* 5a. decrease child nlink for dotdot locally if former remote
574 rc = dt_declare_ref_del(env, child, th);
578 /* 6a. decrease child nlink for dot locally if former remote
580 rc = dt_declare_ref_del(env, child, th);
584 /* 7a. destroy child locally if former remote update failed. */
585 rc = dt_declare_destroy(env, child, th);
589 /* 8a. update bookmark locally. */
590 rc = dt_declare_record_write(env, bk_obj,
591 lfsck_buf_get(env, bk, len), 0, th);
595 rc = dt_trans_start(env, dev, th);
599 /* 3b. insert name into parent dir remotely. */
600 rc = dt_insert(env, parent, (const struct dt_rec *)cfid,
601 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
603 dt_write_lock(env, parent, 0);
604 /* 4b. increase parent nlink remotely. */
605 rc = dt_ref_add(env, parent, th);
606 dt_write_unlock(env, parent);
609 /* 5b. decrease child nlink for dotdot locally. */
610 dt_ref_del(env, child, th);
611 /* 6b. decrease child nlink for dot locally. */
612 dt_ref_del(env, child, th);
613 /* 7b. destroy child locally. */
614 dt_destroy(env, child, th);
618 bk->lb_lpf_fid = *cfid;
619 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
621 /* 8b. update bookmark locally. */
622 rc = dt_record_write(env, bk_obj,
623 lfsck_buf_get(env, bk, len), &pos, th);
628 dt_write_unlock(env, child);
630 dt_trans_stop(env, dev, th);
635 /* Do NOT create .lustre/lost+found/MDTxxxx when register the lfsck instance,
636 * because the MDT0 maybe not reaady for sequence allocation yet. We do that
637 * only when it is required, such as orphan OST-objects repairing. */
638 int lfsck_create_lpf(const struct lu_env *env, struct lfsck_instance *lfsck)
640 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
641 struct lfsck_thread_info *info = lfsck_env_info(env);
642 struct lu_fid *cfid = &info->lti_fid2;
643 struct lu_attr *la = &info->lti_la;
644 struct dt_object_format *dof = &info->lti_dof;
645 struct dt_object *parent = NULL;
646 struct dt_object *child = NULL;
648 int node = lfsck_dev_idx(lfsck->li_bottom);
652 LASSERT(lfsck->li_master);
654 sprintf(name, "MDT%04x", node);
656 parent = lfsck_object_find_by_dev(env, lfsck->li_bottom,
659 struct lfsck_tgt_desc *ltd;
661 ltd = lfsck_tgt_get(&lfsck->li_mdt_descs, 0);
662 if (unlikely(ltd == NULL))
665 parent = lfsck_object_find_by_dev(env, ltd->ltd_tgt,
670 RETURN(PTR_ERR(parent));
672 if (unlikely(!dt_try_as_dir(env, parent)))
673 GOTO(out, rc = -ENOTDIR);
675 mutex_lock(&lfsck->li_mutex);
676 if (lfsck->li_lpf_obj != NULL)
677 GOTO(unlock, rc = 0);
679 if (fid_is_zero(&bk->lb_lpf_fid)) {
680 /* There is corner case that: in former LFSCK scanning we have
681 * created the .lustre/lost+found/MDTxxxx but failed to update
682 * the lfsck_bookmark::lb_lpf_fid successfully. So need lookup
683 * it from MDT0 firstly. */
684 rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
685 (const struct dt_key *)name, BYPASS_CAPA);
686 if (rc != 0 && rc != -ENOENT)
690 bk->lb_lpf_fid = *cfid;
691 rc = lfsck_bookmark_store(env, lfsck);
693 rc = lfsck_fid_alloc(env, lfsck, cfid, true);
698 *cfid = bk->lb_lpf_fid;
701 child = lfsck_object_find_by_dev(env, lfsck->li_bottom, cfid);
703 GOTO(unlock, rc = PTR_ERR(child));
705 if (dt_object_exists(child) != 0) {
706 if (unlikely(!dt_try_as_dir(env, child)))
707 GOTO(unlock, rc = -ENOTDIR);
709 lfsck->li_lpf_obj = child;
710 GOTO(unlock, rc = 0);
713 memset(la, 0, sizeof(*la));
714 la->la_atime = la->la_mtime = la->la_ctime = cfs_time_current_sec();
715 la->la_mode = S_IFDIR | S_IRWXU;
716 la->la_valid = LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE |
718 memset(dof, 0, sizeof(*dof));
719 dof->dof_type = dt_mode_to_dft(S_IFDIR);
722 rc = lfsck_create_lpf_local(env, lfsck, parent, child, la,
725 rc = lfsck_create_lpf_remote(env, lfsck, parent, child, la,
728 lfsck->li_lpf_obj = child;
733 mutex_unlock(&lfsck->li_mutex);
734 if (rc != 0 && child != NULL && !IS_ERR(child))
735 lu_object_put(env, &child->do_lu);
737 if (parent != NULL && !IS_ERR(parent))
738 lu_object_put(env, &parent->do_lu);
743 static int lfsck_fid_init(struct lfsck_instance *lfsck)
745 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
746 struct seq_server_site *ss;
751 ss = lu_site2seq(lfsck->li_bottom->dd_lu_dev.ld_site);
752 if (unlikely(ss == NULL))
755 OBD_ALLOC_PTR(lfsck->li_seq);
756 if (lfsck->li_seq == NULL)
759 OBD_ALLOC(prefix, MAX_OBD_NAME + 7);
761 GOTO(out, rc = -ENOMEM);
763 snprintf(prefix, MAX_OBD_NAME + 7, "lfsck-%s", lfsck_lfsck2name(lfsck));
764 rc = seq_client_init(lfsck->li_seq, NULL, LUSTRE_SEQ_METADATA, prefix,
766 OBD_FREE(prefix, MAX_OBD_NAME + 7);
770 if (fid_is_sane(&bk->lb_last_fid))
771 lfsck->li_seq->lcs_fid = bk->lb_last_fid;
776 OBD_FREE_PTR(lfsck->li_seq);
777 lfsck->li_seq = NULL;
782 static void lfsck_fid_fini(struct lfsck_instance *lfsck)
784 if (lfsck->li_seq != NULL) {
785 seq_client_fini(lfsck->li_seq);
786 OBD_FREE_PTR(lfsck->li_seq);
787 lfsck->li_seq = NULL;
791 void lfsck_instance_cleanup(const struct lu_env *env,
792 struct lfsck_instance *lfsck)
794 struct ptlrpc_thread *thread = &lfsck->li_thread;
795 struct lfsck_component *com;
798 LASSERT(list_empty(&lfsck->li_link));
799 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
801 if (lfsck->li_obj_oit != NULL) {
802 lu_object_put_nocache(env, &lfsck->li_obj_oit->do_lu);
803 lfsck->li_obj_oit = NULL;
806 LASSERT(lfsck->li_obj_dir == NULL);
808 while (!cfs_list_empty(&lfsck->li_list_scan)) {
809 com = cfs_list_entry(lfsck->li_list_scan.next,
810 struct lfsck_component,
812 lfsck_component_cleanup(env, com);
815 LASSERT(cfs_list_empty(&lfsck->li_list_dir));
817 while (!cfs_list_empty(&lfsck->li_list_double_scan)) {
818 com = cfs_list_entry(lfsck->li_list_double_scan.next,
819 struct lfsck_component,
821 lfsck_component_cleanup(env, com);
824 while (!cfs_list_empty(&lfsck->li_list_idle)) {
825 com = cfs_list_entry(lfsck->li_list_idle.next,
826 struct lfsck_component,
828 lfsck_component_cleanup(env, com);
831 lfsck_tgt_descs_fini(&lfsck->li_ost_descs);
832 lfsck_tgt_descs_fini(&lfsck->li_mdt_descs);
834 if (lfsck->li_bookmark_obj != NULL) {
835 lu_object_put_nocache(env, &lfsck->li_bookmark_obj->do_lu);
836 lfsck->li_bookmark_obj = NULL;
839 if (lfsck->li_lpf_obj != NULL) {
840 lu_object_put(env, &lfsck->li_lpf_obj->do_lu);
841 lfsck->li_lpf_obj = NULL;
844 if (lfsck->li_los != NULL) {
845 local_oid_storage_fini(env, lfsck->li_los);
846 lfsck->li_los = NULL;
849 lfsck_fid_fini(lfsck);
854 static inline struct lfsck_instance *
855 __lfsck_instance_find(struct dt_device *key, bool ref, bool unlink)
857 struct lfsck_instance *lfsck;
859 cfs_list_for_each_entry(lfsck, &lfsck_instance_list, li_link) {
860 if (lfsck->li_bottom == key) {
862 lfsck_instance_get(lfsck);
864 list_del_init(&lfsck->li_link);
873 struct lfsck_instance *lfsck_instance_find(struct dt_device *key, bool ref,
876 struct lfsck_instance *lfsck;
878 spin_lock(&lfsck_instance_lock);
879 lfsck = __lfsck_instance_find(key, ref, unlink);
880 spin_unlock(&lfsck_instance_lock);
885 static inline int lfsck_instance_add(struct lfsck_instance *lfsck)
887 struct lfsck_instance *tmp;
889 spin_lock(&lfsck_instance_lock);
890 cfs_list_for_each_entry(tmp, &lfsck_instance_list, li_link) {
891 if (lfsck->li_bottom == tmp->li_bottom) {
892 spin_unlock(&lfsck_instance_lock);
897 cfs_list_add_tail(&lfsck->li_link, &lfsck_instance_list);
898 spin_unlock(&lfsck_instance_lock);
902 int lfsck_bits_dump(char **buf, int *len, int bits, const char *names[],
909 bool newline = (bits != 0 ? false : true);
911 rc = snprintf(*buf, *len, "%s:%c", prefix, newline ? '\n' : ' ');
917 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
920 if (names[i] != NULL) {
924 rc = snprintf(*buf, *len, "%s%c", names[i],
925 newline ? '\n' : ',');
936 rc = snprintf(*buf, *len, "\n");
947 int lfsck_time_dump(char **buf, int *len, __u64 time, const char *prefix)
952 rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
953 cfs_time_current_sec() - time);
955 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
964 int lfsck_pos_dump(char **buf, int *len, struct lfsck_position *pos,
969 if (fid_is_zero(&pos->lp_dir_parent)) {
970 if (pos->lp_oit_cookie == 0)
971 rc = snprintf(*buf, *len, "%s: N/A, N/A, N/A\n",
974 rc = snprintf(*buf, *len, "%s: "LPU64", N/A, N/A\n",
975 prefix, pos->lp_oit_cookie);
977 rc = snprintf(*buf, *len, "%s: "LPU64", "DFID", "LPU64"\n",
978 prefix, pos->lp_oit_cookie,
979 PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
989 void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
990 struct lfsck_position *pos, bool init)
992 const struct dt_it_ops *iops = &lfsck->li_obj_oit->do_index_ops->dio_it;
994 if (unlikely(lfsck->li_di_oit == NULL)) {
995 memset(pos, 0, sizeof(*pos));
999 pos->lp_oit_cookie = iops->store(env, lfsck->li_di_oit);
1000 if (!lfsck->li_current_oit_processed && !init)
1001 pos->lp_oit_cookie--;
1003 LASSERT(pos->lp_oit_cookie > 0);
1005 if (lfsck->li_di_dir != NULL) {
1006 struct dt_object *dto = lfsck->li_obj_dir;
1008 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
1011 if (pos->lp_dir_cookie >= MDS_DIR_END_OFF) {
1012 fid_zero(&pos->lp_dir_parent);
1013 pos->lp_dir_cookie = 0;
1015 pos->lp_dir_parent = *lfsck_dto2fid(dto);
1018 fid_zero(&pos->lp_dir_parent);
1019 pos->lp_dir_cookie = 0;
1023 static void __lfsck_set_speed(struct lfsck_instance *lfsck, __u32 limit)
1025 lfsck->li_bookmark_ram.lb_speed_limit = limit;
1026 if (limit != LFSCK_SPEED_NO_LIMIT) {
1028 lfsck->li_sleep_rate = limit / HZ;
1029 lfsck->li_sleep_jif = 1;
1031 lfsck->li_sleep_rate = 1;
1032 lfsck->li_sleep_jif = HZ / limit;
1035 lfsck->li_sleep_jif = 0;
1036 lfsck->li_sleep_rate = 0;
1040 void lfsck_control_speed(struct lfsck_instance *lfsck)
1042 struct ptlrpc_thread *thread = &lfsck->li_thread;
1043 struct l_wait_info lwi;
1045 if (lfsck->li_sleep_jif > 0 &&
1046 lfsck->li_new_scanned >= lfsck->li_sleep_rate) {
1047 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1048 LWI_ON_SIGNAL_NOOP, NULL);
1050 l_wait_event(thread->t_ctl_waitq,
1051 !thread_is_running(thread),
1053 lfsck->li_new_scanned = 0;
1057 void lfsck_control_speed_by_self(struct lfsck_component *com)
1059 struct lfsck_instance *lfsck = com->lc_lfsck;
1060 struct ptlrpc_thread *thread = &lfsck->li_thread;
1061 struct l_wait_info lwi;
1063 if (lfsck->li_sleep_jif > 0 &&
1064 com->lc_new_scanned >= lfsck->li_sleep_rate) {
1065 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1066 LWI_ON_SIGNAL_NOOP, NULL);
1068 l_wait_event(thread->t_ctl_waitq,
1069 !thread_is_running(thread),
1071 com->lc_new_scanned = 0;
1075 static int lfsck_parent_fid(const struct lu_env *env, struct dt_object *obj,
1078 if (unlikely(!S_ISDIR(lfsck_object_type(obj)) ||
1079 !dt_try_as_dir(env, obj)))
1082 return dt_lookup(env, obj, (struct dt_rec *)fid,
1083 (const struct dt_key *)"..", BYPASS_CAPA);
1086 static int lfsck_needs_scan_dir(const struct lu_env *env,
1087 struct lfsck_instance *lfsck,
1088 struct dt_object *obj)
1090 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
1094 if (!lfsck->li_master || !S_ISDIR(lfsck_object_type(obj)) ||
1095 cfs_list_empty(&lfsck->li_list_dir))
1099 /* XXX: Currently, we do not scan the "/REMOTE_PARENT_DIR",
1100 * which is the agent directory to manage the objects
1101 * which name entries reside on remote MDTs. Related
1102 * consistency verification will be processed in LFSCK
1104 if (lu_fid_eq(lfsck_dto2fid(obj), &lfsck->li_global_root_fid)) {
1106 lfsck_object_put(env, obj);
1110 /* .lustre doesn't contain "real" user objects, no need lfsck */
1111 if (fid_is_dot_lustre(lfsck_dto2fid(obj))) {
1113 lfsck_object_put(env, obj);
1117 dt_read_lock(env, obj, MOR_TGT_CHILD);
1118 if (unlikely(lfsck_is_dead_obj(obj))) {
1119 dt_read_unlock(env, obj);
1121 lfsck_object_put(env, obj);
1125 rc = dt_xattr_get(env, obj,
1126 lfsck_buf_get(env, NULL, 0), XATTR_NAME_LINK,
1128 dt_read_unlock(env, obj);
1131 lfsck_object_put(env, obj);
1135 if (rc < 0 && rc != -ENODATA) {
1137 lfsck_object_put(env, obj);
1141 rc = lfsck_parent_fid(env, obj, fid);
1143 lfsck_object_put(env, obj);
1147 if (unlikely(lu_fid_eq(fid, &lfsck->li_local_root_fid)))
1150 obj = lfsck_object_find(env, lfsck, fid);
1153 else if (IS_ERR(obj))
1154 return PTR_ERR(obj);
1156 if (!dt_object_exists(obj)) {
1157 lfsck_object_put(env, obj);
1161 /* Currently, only client visible directory can be remote. */
1162 if (dt_object_remote(obj)) {
1163 lfsck_object_put(env, obj);
1172 struct lfsck_thread_args *lfsck_thread_args_init(struct lfsck_instance *lfsck,
1173 struct lfsck_component *com,
1174 struct lfsck_start_param *lsp)
1176 struct lfsck_thread_args *lta;
1181 return ERR_PTR(-ENOMEM);
1183 rc = lu_env_init(<a->lta_env, LCT_MD_THREAD | LCT_DT_THREAD);
1189 lta->lta_lfsck = lfsck_instance_get(lfsck);
1191 lta->lta_com = lfsck_component_get(com);
1198 void lfsck_thread_args_fini(struct lfsck_thread_args *lta)
1200 if (lta->lta_com != NULL)
1201 lfsck_component_put(<a->lta_env, lta->lta_com);
1202 lfsck_instance_put(<a->lta_env, lta->lta_lfsck);
1203 lu_env_fini(<a->lta_env);
1207 /* LFSCK wrap functions */
1209 void lfsck_fail(const struct lu_env *env, struct lfsck_instance *lfsck,
1212 struct lfsck_component *com;
1214 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1215 com->lc_ops->lfsck_fail(env, com, new_checked);
1219 int lfsck_checkpoint(const struct lu_env *env, struct lfsck_instance *lfsck)
1221 struct lfsck_component *com;
1225 if (likely(cfs_time_beforeq(cfs_time_current(),
1226 lfsck->li_time_next_checkpoint)))
1229 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1230 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1231 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
1236 lfsck->li_time_last_checkpoint = cfs_time_current();
1237 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1238 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1239 return rc1 != 0 ? rc1 : rc;
1242 int lfsck_prep(const struct lu_env *env, struct lfsck_instance *lfsck,
1243 struct lfsck_start_param *lsp)
1245 struct dt_object *obj = NULL;
1246 struct lfsck_component *com;
1247 struct lfsck_component *next;
1248 struct lfsck_position *pos = NULL;
1249 const struct dt_it_ops *iops =
1250 &lfsck->li_obj_oit->do_index_ops->dio_it;
1255 LASSERT(lfsck->li_obj_dir == NULL);
1256 LASSERT(lfsck->li_di_dir == NULL);
1258 lfsck->li_current_oit_processed = 0;
1259 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1260 com->lc_new_checked = 0;
1261 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1262 com->lc_journal = 0;
1264 rc = com->lc_ops->lfsck_prep(env, com, lsp);
1268 if ((pos == NULL) ||
1269 (!lfsck_pos_is_zero(&com->lc_pos_start) &&
1270 lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
1271 pos = &com->lc_pos_start;
1274 /* Init otable-based iterator. */
1276 rc = iops->load(env, lfsck->li_di_oit, 0);
1278 lfsck->li_oit_over = 1;
1285 rc = iops->load(env, lfsck->li_di_oit, pos->lp_oit_cookie);
1289 lfsck->li_oit_over = 1;
1291 if (!lfsck->li_master || fid_is_zero(&pos->lp_dir_parent))
1294 /* Find the directory for namespace-based traverse. */
1295 obj = lfsck_object_find(env, lfsck, &pos->lp_dir_parent);
1298 else if (IS_ERR(obj))
1299 RETURN(PTR_ERR(obj));
1301 /* XXX: Currently, skip remote object, the consistency for
1302 * remote object will be processed in LFSCK phase III. */
1303 if (!dt_object_exists(obj) || dt_object_remote(obj) ||
1304 unlikely(!S_ISDIR(lfsck_object_type(obj))))
1307 if (unlikely(!dt_try_as_dir(env, obj)))
1308 GOTO(out, rc = -ENOTDIR);
1310 /* Init the namespace-based directory traverse. */
1311 iops = &obj->do_index_ops->dio_it;
1312 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1314 GOTO(out, rc = PTR_ERR(di));
1316 LASSERT(pos->lp_dir_cookie < MDS_DIR_END_OFF);
1318 rc = iops->load(env, di, pos->lp_dir_cookie);
1319 if ((rc == 0) || (rc > 0 && pos->lp_dir_cookie > 0))
1320 rc = iops->next(env, di);
1326 iops->fini(env, di);
1330 lfsck->li_obj_dir = lfsck_object_get(obj);
1331 lfsck->li_cookie_dir = iops->store(env, di);
1332 spin_lock(&lfsck->li_lock);
1333 lfsck->li_di_dir = di;
1334 spin_unlock(&lfsck->li_lock);
1340 lfsck_object_put(env, obj);
1343 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1345 com->lc_ops->lfsck_post(env, com, rc, true);
1351 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, true);
1352 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1353 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
1358 lfsck->li_time_last_checkpoint = cfs_time_current();
1359 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1360 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1364 int lfsck_exec_oit(const struct lu_env *env, struct lfsck_instance *lfsck,
1365 struct dt_object *obj)
1367 struct lfsck_component *com;
1368 const struct dt_it_ops *iops;
1373 LASSERT(lfsck->li_obj_dir == NULL);
1375 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1376 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
1381 rc = lfsck_needs_scan_dir(env, lfsck, obj);
1385 if (unlikely(!dt_try_as_dir(env, obj)))
1386 GOTO(out, rc = -ENOTDIR);
1388 iops = &obj->do_index_ops->dio_it;
1389 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1391 GOTO(out, rc = PTR_ERR(di));
1393 rc = iops->load(env, di, 0);
1395 rc = iops->next(env, di);
1401 iops->fini(env, di);
1405 lfsck->li_obj_dir = lfsck_object_get(obj);
1406 lfsck->li_cookie_dir = iops->store(env, di);
1407 spin_lock(&lfsck->li_lock);
1408 lfsck->li_di_dir = di;
1409 spin_unlock(&lfsck->li_lock);
1415 lfsck_fail(env, lfsck, false);
1416 return (rc > 0 ? 0 : rc);
1419 int lfsck_exec_dir(const struct lu_env *env, struct lfsck_instance *lfsck,
1420 struct dt_object *obj, struct lu_dirent *ent)
1422 struct lfsck_component *com;
1425 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1426 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
1433 int lfsck_post(const struct lu_env *env, struct lfsck_instance *lfsck,
1436 struct lfsck_component *com;
1437 struct lfsck_component *next;
1441 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1442 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1443 rc = com->lc_ops->lfsck_post(env, com, result, false);
1448 lfsck->li_time_last_checkpoint = cfs_time_current();
1449 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1450 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1452 /* Ignore some component post failure to make other can go ahead. */
1456 static void lfsck_interpret(const struct lu_env *env,
1457 struct lfsck_instance *lfsck,
1458 struct ptlrpc_request *req, void *args, int result)
1460 struct lfsck_async_interpret_args *laia = args;
1461 struct lfsck_component *com;
1463 LASSERT(laia->laia_com == NULL);
1464 LASSERT(laia->laia_shared);
1466 spin_lock(&lfsck->li_lock);
1467 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1468 if (com->lc_ops->lfsck_interpret != NULL) {
1469 laia->laia_com = com;
1470 com->lc_ops->lfsck_interpret(env, req, laia, result);
1474 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1475 if (com->lc_ops->lfsck_interpret != NULL) {
1476 laia->laia_com = com;
1477 com->lc_ops->lfsck_interpret(env, req, laia, result);
1480 spin_unlock(&lfsck->li_lock);
1483 int lfsck_double_scan(const struct lu_env *env, struct lfsck_instance *lfsck)
1485 struct lfsck_component *com;
1486 struct lfsck_component *next;
1487 struct l_wait_info lwi = { 0 };
1491 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1492 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1493 com->lc_journal = 0;
1495 rc = com->lc_ops->lfsck_double_scan(env, com);
1500 l_wait_event(lfsck->li_thread.t_ctl_waitq,
1501 atomic_read(&lfsck->li_double_scan_count) == 0,
1504 if (lfsck->li_status != LS_PAUSED &&
1505 lfsck->li_status != LS_CO_PAUSED) {
1506 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1508 spin_lock(&lfsck->li_lock);
1509 list_del_init(&com->lc_link);
1510 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1511 spin_unlock(&lfsck->li_lock);
1515 return rc1 != 0 ? rc1 : rc;
1518 static int lfsck_stop_notify(const struct lu_env *env,
1519 struct lfsck_instance *lfsck,
1520 struct lfsck_tgt_descs *ltds,
1521 struct lfsck_tgt_desc *ltd, __u16 type)
1523 struct ptlrpc_request_set *set;
1524 struct lfsck_component *com;
1528 spin_lock(&lfsck->li_lock);
1529 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
1531 com = __lfsck_component_find(lfsck, type,
1532 &lfsck->li_list_double_scan);
1534 lfsck_component_get(com);
1535 spin_lock(&lfsck->li_lock);
1538 if (com->lc_ops->lfsck_stop_notify != NULL) {
1539 set = ptlrpc_prep_set();
1541 lfsck_component_put(env, com);
1546 rc = com->lc_ops->lfsck_stop_notify(env, com, ltds,
1549 rc = ptlrpc_set_wait(set);
1551 ptlrpc_set_destroy(set);
1554 lfsck_component_put(env, com);
1560 void lfsck_quit(const struct lu_env *env, struct lfsck_instance *lfsck)
1562 struct lfsck_component *com;
1563 struct lfsck_component *next;
1565 list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1567 if (com->lc_ops->lfsck_quit != NULL)
1568 com->lc_ops->lfsck_quit(env, com);
1570 spin_lock(&lfsck->li_lock);
1571 list_del_init(&com->lc_link);
1572 list_del_init(&com->lc_link_dir);
1573 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1574 spin_unlock(&lfsck->li_lock);
1577 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1579 if (com->lc_ops->lfsck_quit != NULL)
1580 com->lc_ops->lfsck_quit(env, com);
1582 spin_lock(&lfsck->li_lock);
1583 list_del_init(&com->lc_link);
1584 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1585 spin_unlock(&lfsck->li_lock);
1589 static int lfsck_async_interpret(const struct lu_env *env,
1590 struct ptlrpc_request *req,
1593 struct lfsck_async_interpret_args *laia = args;
1594 struct lfsck_instance *lfsck;
1596 lfsck = container_of0(laia->laia_ltds, struct lfsck_instance,
1598 lfsck_interpret(env, lfsck, req, laia, rc);
1599 lfsck_tgt_put(laia->laia_ltd);
1600 if (rc != 0 && laia->laia_result != -EALREADY)
1601 laia->laia_result = rc;
1606 int lfsck_async_request(const struct lu_env *env, struct obd_export *exp,
1607 struct lfsck_request *lr,
1608 struct ptlrpc_request_set *set,
1609 ptlrpc_interpterer_t interpreter,
1610 void *args, int request)
1612 struct lfsck_async_interpret_args *laia;
1613 struct ptlrpc_request *req;
1614 struct lfsck_request *tmp;
1615 struct req_format *format;
1618 if (!(exp_connect_flags(exp) & OBD_CONNECT_LFSCK))
1623 format = &RQF_LFSCK_NOTIFY;
1626 format = &RQF_LFSCK_QUERY;
1629 CERROR("%s: unknown async request: opc = %d\n",
1630 exp->exp_obd->obd_name, request);
1634 req = ptlrpc_request_alloc(class_exp2cliimp(exp), format);
1638 rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, request);
1640 ptlrpc_request_free(req);
1645 tmp = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
1647 ptlrpc_request_set_replen(req);
1649 laia = ptlrpc_req_async_args(req);
1650 *laia = *(struct lfsck_async_interpret_args *)args;
1651 if (laia->laia_com != NULL)
1652 lfsck_component_get(laia->laia_com);
1653 req->rq_interpret_reply = interpreter;
1654 ptlrpc_set_add_req(set, req);
1659 /* external interfaces */
1661 int lfsck_get_speed(struct dt_device *key, void *buf, int len)
1664 struct lfsck_instance *lfsck;
1668 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1672 lfsck = lfsck_instance_find(key, true, false);
1673 if (likely(lfsck != NULL)) {
1674 rc = snprintf(buf, len, "%u\n",
1675 lfsck->li_bookmark_ram.lb_speed_limit);
1676 lfsck_instance_put(&env, lfsck);
1685 EXPORT_SYMBOL(lfsck_get_speed);
1687 int lfsck_set_speed(struct dt_device *key, int val)
1690 struct lfsck_instance *lfsck;
1694 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1698 lfsck = lfsck_instance_find(key, true, false);
1699 if (likely(lfsck != NULL)) {
1700 mutex_lock(&lfsck->li_mutex);
1701 __lfsck_set_speed(lfsck, val);
1702 rc = lfsck_bookmark_store(&env, lfsck);
1703 mutex_unlock(&lfsck->li_mutex);
1704 lfsck_instance_put(&env, lfsck);
1713 EXPORT_SYMBOL(lfsck_set_speed);
1715 int lfsck_get_windows(struct dt_device *key, void *buf, int len)
1718 struct lfsck_instance *lfsck;
1722 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1726 lfsck = lfsck_instance_find(key, true, false);
1727 if (likely(lfsck != NULL)) {
1728 rc = snprintf(buf, len, "%u\n",
1729 lfsck->li_bookmark_ram.lb_async_windows);
1730 lfsck_instance_put(&env, lfsck);
1739 EXPORT_SYMBOL(lfsck_get_windows);
1741 int lfsck_set_windows(struct dt_device *key, int val)
1744 struct lfsck_instance *lfsck;
1748 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1752 lfsck = lfsck_instance_find(key, true, false);
1753 if (likely(lfsck != NULL)) {
1754 if (val > LFSCK_ASYNC_WIN_MAX) {
1755 CERROR("%s: Too large async windows size, which "
1756 "may cause memory issues. The valid range "
1757 "is [0 - %u]. If you do not want to restrict "
1758 "the windows size for async requests pipeline, "
1759 "just set it as 0.\n",
1760 lfsck_lfsck2name(lfsck), LFSCK_ASYNC_WIN_MAX);
1762 } else if (lfsck->li_bookmark_ram.lb_async_windows != val) {
1763 mutex_lock(&lfsck->li_mutex);
1764 lfsck->li_bookmark_ram.lb_async_windows = val;
1765 rc = lfsck_bookmark_store(&env, lfsck);
1766 mutex_unlock(&lfsck->li_mutex);
1768 lfsck_instance_put(&env, lfsck);
1777 EXPORT_SYMBOL(lfsck_set_windows);
1779 int lfsck_dump(struct dt_device *key, void *buf, int len, enum lfsck_type type)
1782 struct lfsck_instance *lfsck;
1783 struct lfsck_component *com;
1787 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1791 lfsck = lfsck_instance_find(key, true, false);
1792 if (likely(lfsck != NULL)) {
1793 com = lfsck_component_find(lfsck, type);
1794 if (likely(com != NULL)) {
1795 rc = com->lc_ops->lfsck_dump(&env, com, buf, len);
1796 lfsck_component_put(&env, com);
1801 lfsck_instance_put(&env, lfsck);
1810 EXPORT_SYMBOL(lfsck_dump);
1812 static int lfsck_stop_all(const struct lu_env *env,
1813 struct lfsck_instance *lfsck,
1814 struct lfsck_stop *stop)
1816 struct lfsck_thread_info *info = lfsck_env_info(env);
1817 struct lfsck_request *lr = &info->lti_lr;
1818 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1819 struct ptlrpc_request_set *set;
1820 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1821 struct lfsck_tgt_desc *ltd;
1822 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1828 LASSERT(stop->ls_flags & LPF_BROADCAST);
1830 set = ptlrpc_prep_set();
1831 if (unlikely(set == NULL)) {
1832 CERROR("%s: cannot allocate memory for stop LFSCK on "
1833 "all targets\n", lfsck_lfsck2name(lfsck));
1838 memset(lr, 0, sizeof(*lr));
1839 lr->lr_event = LE_STOP;
1840 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1841 lr->lr_status = stop->ls_status;
1842 lr->lr_version = bk->lb_version;
1843 lr->lr_active = LFSCK_TYPES_ALL;
1844 lr->lr_param = stop->ls_flags;
1846 laia->laia_com = NULL;
1847 laia->laia_ltds = ltds;
1849 laia->laia_result = 0;
1850 laia->laia_shared = 1;
1852 down_read(<ds->ltd_rw_sem);
1853 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1854 ltd = lfsck_tgt_get(ltds, idx);
1855 LASSERT(ltd != NULL);
1857 laia->laia_ltd = ltd;
1858 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1859 lfsck_async_interpret, laia,
1862 lfsck_interpret(env, lfsck, NULL, laia, rc);
1864 CWARN("%s: cannot notify MDT %x for LFSCK stop: "
1865 "rc = %d\n", lfsck_lfsck2name(lfsck), idx, rc);
1869 up_read(<ds->ltd_rw_sem);
1871 rc = ptlrpc_set_wait(set);
1872 ptlrpc_set_destroy(set);
1875 rc = laia->laia_result;
1877 if (rc == -EALREADY)
1881 CWARN("%s: fail to stop LFSCK on some MDTs: rc = %d\n",
1882 lfsck_lfsck2name(lfsck), rc);
1884 RETURN(rc != 0 ? rc : rc1);
1887 static int lfsck_start_all(const struct lu_env *env,
1888 struct lfsck_instance *lfsck,
1889 struct lfsck_start *start)
1891 struct lfsck_thread_info *info = lfsck_env_info(env);
1892 struct lfsck_request *lr = &info->lti_lr;
1893 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1894 struct ptlrpc_request_set *set;
1895 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1896 struct lfsck_tgt_desc *ltd;
1897 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1902 LASSERT(start->ls_flags & LPF_BROADCAST);
1904 set = ptlrpc_prep_set();
1905 if (unlikely(set == NULL)) {
1906 if (bk->lb_param & LPF_FAILOUT) {
1907 CERROR("%s: cannot allocate memory for start LFSCK on "
1908 "all targets, failout.\n",
1909 lfsck_lfsck2name(lfsck));
1913 CWARN("%s: cannot allocate memory for start LFSCK on "
1914 "all targets, partly scan.\n",
1915 lfsck_lfsck2name(lfsck));
1921 memset(lr, 0, sizeof(*lr));
1922 lr->lr_event = LE_START;
1923 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1924 lr->lr_speed = bk->lb_speed_limit;
1925 lr->lr_version = bk->lb_version;
1926 lr->lr_active = start->ls_active;
1927 lr->lr_param = start->ls_flags;
1928 lr->lr_async_windows = bk->lb_async_windows;
1929 lr->lr_valid = LSV_SPEED_LIMIT | LSV_ERROR_HANDLE | LSV_DRYRUN |
1932 laia->laia_com = NULL;
1933 laia->laia_ltds = ltds;
1935 laia->laia_result = 0;
1936 laia->laia_shared = 1;
1938 down_read(<ds->ltd_rw_sem);
1939 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1940 ltd = lfsck_tgt_get(ltds, idx);
1941 LASSERT(ltd != NULL);
1943 laia->laia_ltd = ltd;
1944 ltd->ltd_layout_done = 0;
1945 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1946 lfsck_async_interpret, laia,
1949 lfsck_interpret(env, lfsck, NULL, laia, rc);
1951 if (bk->lb_param & LPF_FAILOUT) {
1952 CERROR("%s: cannot notify MDT %x for LFSCK "
1953 "start, failout: rc = %d\n",
1954 lfsck_lfsck2name(lfsck), idx, rc);
1957 CWARN("%s: cannot notify MDT %x for LFSCK "
1958 "start, partly scan: rc = %d\n",
1959 lfsck_lfsck2name(lfsck), idx, rc);
1964 up_read(<ds->ltd_rw_sem);
1967 ptlrpc_set_destroy(set);
1972 rc = ptlrpc_set_wait(set);
1973 ptlrpc_set_destroy(set);
1976 rc = laia->laia_result;
1979 if (bk->lb_param & LPF_FAILOUT) {
1980 struct lfsck_stop *stop = &info->lti_stop;
1982 CERROR("%s: cannot start LFSCK on some MDTs, "
1983 "stop all: rc = %d\n",
1984 lfsck_lfsck2name(lfsck), rc);
1985 if (rc != -EALREADY) {
1986 stop->ls_status = LS_FAILED;
1987 stop->ls_flags = LPF_ALL_TGT | LPF_BROADCAST;
1988 lfsck_stop_all(env, lfsck, stop);
1991 CWARN("%s: cannot start LFSCK on some MDTs, "
1992 "partly scan: rc = %d\n",
1993 lfsck_lfsck2name(lfsck), rc);
2001 int lfsck_start(const struct lu_env *env, struct dt_device *key,
2002 struct lfsck_start_param *lsp)
2004 struct lfsck_start *start = lsp->lsp_start;
2005 struct lfsck_instance *lfsck;
2006 struct lfsck_bookmark *bk;
2007 struct ptlrpc_thread *thread;
2008 struct lfsck_component *com;
2009 struct l_wait_info lwi = { 0 };
2010 struct lfsck_thread_args *lta;
2018 lfsck = lfsck_instance_find(key, true, false);
2019 if (unlikely(lfsck == NULL))
2022 /* System is not ready, try again later. */
2023 if (unlikely(lfsck->li_namespace == NULL))
2024 GOTO(put, rc = -EAGAIN);
2026 /* start == NULL means auto trigger paused LFSCK. */
2027 if ((start == NULL) &&
2028 (cfs_list_empty(&lfsck->li_list_scan) ||
2029 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
2032 bk = &lfsck->li_bookmark_ram;
2033 thread = &lfsck->li_thread;
2034 mutex_lock(&lfsck->li_mutex);
2035 spin_lock(&lfsck->li_lock);
2036 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2038 while (start->ls_active != 0) {
2039 if (!(type & start->ls_active)) {
2044 com = __lfsck_component_find(lfsck, type,
2045 &lfsck->li_list_scan);
2047 com = __lfsck_component_find(lfsck, type,
2048 &lfsck->li_list_double_scan);
2054 if (com->lc_ops->lfsck_join != NULL) {
2055 rc = com->lc_ops->lfsck_join( env, com, lsp);
2056 if (rc != 0 && rc != -EALREADY)
2059 start->ls_active &= ~type;
2062 spin_unlock(&lfsck->li_lock);
2065 spin_unlock(&lfsck->li_lock);
2067 lfsck->li_status = 0;
2068 lfsck->li_oit_over = 0;
2069 lfsck->li_start_unplug = 0;
2070 lfsck->li_drop_dryrun = 0;
2071 lfsck->li_new_scanned = 0;
2073 /* For auto trigger. */
2077 if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
2078 CERROR("%s: only allow to specify '-A | -o' via MDS\n",
2079 lfsck_lfsck2name(lfsck));
2081 GOTO(out, rc = -EPERM);
2084 start->ls_version = bk->lb_version;
2085 if (start->ls_valid & LSV_SPEED_LIMIT) {
2086 __lfsck_set_speed(lfsck, start->ls_speed_limit);
2090 if (start->ls_valid & LSV_ASYNC_WINDOWS &&
2091 bk->lb_async_windows != start->ls_async_windows) {
2092 bk->lb_async_windows = start->ls_async_windows;
2096 if (start->ls_valid & LSV_ERROR_HANDLE) {
2097 valid |= DOIV_ERROR_HANDLE;
2098 if (start->ls_flags & LPF_FAILOUT)
2099 flags |= DOIF_FAILOUT;
2101 if ((start->ls_flags & LPF_FAILOUT) &&
2102 !(bk->lb_param & LPF_FAILOUT)) {
2103 bk->lb_param |= LPF_FAILOUT;
2105 } else if (!(start->ls_flags & LPF_FAILOUT) &&
2106 (bk->lb_param & LPF_FAILOUT)) {
2107 bk->lb_param &= ~LPF_FAILOUT;
2112 if (start->ls_valid & LSV_DRYRUN) {
2113 valid |= DOIV_DRYRUN;
2114 if (start->ls_flags & LPF_DRYRUN)
2115 flags |= DOIF_DRYRUN;
2117 if ((start->ls_flags & LPF_DRYRUN) &&
2118 !(bk->lb_param & LPF_DRYRUN)) {
2119 bk->lb_param |= LPF_DRYRUN;
2121 } else if (!(start->ls_flags & LPF_DRYRUN) &&
2122 (bk->lb_param & LPF_DRYRUN)) {
2123 bk->lb_param &= ~LPF_DRYRUN;
2124 lfsck->li_drop_dryrun = 1;
2129 if (bk->lb_param & LPF_ALL_TGT &&
2130 !(start->ls_flags & LPF_ALL_TGT)) {
2131 bk->lb_param &= ~LPF_ALL_TGT;
2133 } else if (!(bk->lb_param & LPF_ALL_TGT) &&
2134 start->ls_flags & LPF_ALL_TGT) {
2135 bk->lb_param |= LPF_ALL_TGT;
2139 if (bk->lb_param & LPF_ORPHAN &&
2140 !(start->ls_flags & LPF_ORPHAN)) {
2141 bk->lb_param &= ~LPF_ORPHAN;
2143 } else if (!(bk->lb_param & LPF_ORPHAN) &&
2144 start->ls_flags & LPF_ORPHAN) {
2145 bk->lb_param |= LPF_ORPHAN;
2150 rc = lfsck_bookmark_store(env, lfsck);
2155 if (start->ls_flags & LPF_RESET)
2156 flags |= DOIF_RESET;
2158 if (start->ls_active != 0) {
2159 struct lfsck_component *next;
2161 if (start->ls_active == LFSCK_TYPES_ALL)
2162 start->ls_active = LFSCK_TYPES_SUPPORTED;
2164 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2165 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2166 GOTO(out, rc = -ENOTSUPP);
2169 cfs_list_for_each_entry_safe(com, next,
2170 &lfsck->li_list_scan, lc_link) {
2171 if (!(com->lc_type & start->ls_active)) {
2172 rc = com->lc_ops->lfsck_post(env, com, 0,
2179 while (start->ls_active != 0) {
2180 if (type & start->ls_active) {
2181 com = __lfsck_component_find(lfsck, type,
2182 &lfsck->li_list_idle);
2184 /* The component status will be updated
2185 * when its prep() is called later by
2186 * the LFSCK main engine. */
2187 cfs_list_del_init(&com->lc_link);
2188 cfs_list_add_tail(&com->lc_link,
2189 &lfsck->li_list_scan);
2191 start->ls_active &= ~type;
2197 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
2198 start->ls_active |= com->lc_type;
2199 if (flags & DOIF_RESET) {
2200 rc = com->lc_ops->lfsck_reset(env, com, false);
2207 lfsck->li_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2208 if (bk->lb_param & LPF_DRYRUN) {
2209 lfsck->li_args_dir |= LUDA_VERIFY_DRYRUN;
2210 valid |= DOIV_DRYRUN;
2211 flags |= DOIF_DRYRUN;
2214 if (bk->lb_param & LPF_FAILOUT) {
2215 valid |= DOIV_ERROR_HANDLE;
2216 flags |= DOIF_FAILOUT;
2219 if (!cfs_list_empty(&lfsck->li_list_scan))
2220 flags |= DOIF_OUTUSED;
2222 lfsck->li_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2223 thread_set_flags(thread, 0);
2224 lta = lfsck_thread_args_init(lfsck, NULL, lsp);
2226 GOTO(out, rc = PTR_ERR(lta));
2228 rc = PTR_ERR(kthread_run(lfsck_master_engine, lta, "lfsck"));
2229 if (IS_ERR_VALUE(rc)) {
2230 CERROR("%s: cannot start LFSCK thread: rc = %ld\n",
2231 lfsck_lfsck2name(lfsck), rc);
2232 lfsck_thread_args_fini(lta);
2237 l_wait_event(thread->t_ctl_waitq,
2238 thread_is_running(thread) ||
2239 thread_is_stopped(thread),
2241 if (start == NULL || !(start->ls_flags & LPF_BROADCAST)) {
2242 lfsck->li_start_unplug = 1;
2243 wake_up_all(&thread->t_ctl_waitq);
2248 /* release lfsck::li_mutex to avoid deadlock. */
2249 mutex_unlock(&lfsck->li_mutex);
2250 rc = lfsck_start_all(env, lfsck, start);
2252 spin_lock(&lfsck->li_lock);
2253 if (thread_is_stopped(thread)) {
2254 spin_unlock(&lfsck->li_lock);
2256 lfsck->li_status = LS_FAILED;
2257 lfsck->li_flags = 0;
2258 thread_set_flags(thread, SVC_STOPPING);
2259 spin_unlock(&lfsck->li_lock);
2261 lfsck->li_start_unplug = 1;
2262 wake_up_all(&thread->t_ctl_waitq);
2263 l_wait_event(thread->t_ctl_waitq,
2264 thread_is_stopped(thread),
2268 lfsck->li_start_unplug = 1;
2269 wake_up_all(&thread->t_ctl_waitq);
2275 mutex_unlock(&lfsck->li_mutex);
2278 lfsck_instance_put(env, lfsck);
2280 return rc < 0 ? rc : 0;
2282 EXPORT_SYMBOL(lfsck_start);
2284 int lfsck_stop(const struct lu_env *env, struct dt_device *key,
2285 struct lfsck_stop *stop)
2287 struct lfsck_instance *lfsck;
2288 struct ptlrpc_thread *thread;
2289 struct l_wait_info lwi = { 0 };
2294 lfsck = lfsck_instance_find(key, true, false);
2295 if (unlikely(lfsck == NULL))
2298 thread = &lfsck->li_thread;
2299 /* release lfsck::li_mutex to avoid deadlock. */
2300 if (stop != NULL && stop->ls_flags & LPF_BROADCAST) {
2301 if (!lfsck->li_master) {
2302 CERROR("%s: only allow to specify '-A' via MDS\n",
2303 lfsck_lfsck2name(lfsck));
2305 GOTO(out, rc = -EPERM);
2308 rc1 = lfsck_stop_all(env, lfsck, stop);
2311 mutex_lock(&lfsck->li_mutex);
2312 spin_lock(&lfsck->li_lock);
2313 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2314 spin_unlock(&lfsck->li_lock);
2315 GOTO(out, rc = -EALREADY);
2319 lfsck->li_status = stop->ls_status;
2320 lfsck->li_flags = stop->ls_flags;
2322 lfsck->li_status = LS_STOPPED;
2323 lfsck->li_flags = 0;
2326 thread_set_flags(thread, SVC_STOPPING);
2327 spin_unlock(&lfsck->li_lock);
2329 wake_up_all(&thread->t_ctl_waitq);
2330 l_wait_event(thread->t_ctl_waitq,
2331 thread_is_stopped(thread),
2337 mutex_unlock(&lfsck->li_mutex);
2338 lfsck_instance_put(env, lfsck);
2340 return rc != 0 ? rc : rc1;
2342 EXPORT_SYMBOL(lfsck_stop);
2344 int lfsck_in_notify(const struct lu_env *env, struct dt_device *key,
2345 struct lfsck_request *lr)
2347 int rc = -EOPNOTSUPP;
2350 switch (lr->lr_event) {
2352 struct lfsck_start *start = &lfsck_env_info(env)->lti_start;
2353 struct lfsck_start_param lsp;
2355 memset(start, 0, sizeof(*start));
2356 start->ls_valid = lr->lr_valid;
2357 start->ls_speed_limit = lr->lr_speed;
2358 start->ls_version = lr->lr_version;
2359 start->ls_active = lr->lr_active;
2360 start->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2361 start->ls_async_windows = lr->lr_async_windows;
2363 lsp.lsp_start = start;
2364 lsp.lsp_index = lr->lr_index;
2365 lsp.lsp_index_valid = 1;
2366 rc = lfsck_start(env, key, &lsp);
2370 struct lfsck_stop *stop = &lfsck_env_info(env)->lti_stop;
2372 memset(stop, 0, sizeof(*stop));
2373 stop->ls_status = lr->lr_status;
2374 stop->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2375 rc = lfsck_stop(env, key, stop);
2378 case LE_PHASE1_DONE:
2379 case LE_PHASE2_DONE:
2380 case LE_FID_ACCESSED:
2382 case LE_CONDITIONAL_DESTROY:
2383 case LE_PAIRS_VERIFY: {
2384 struct lfsck_instance *lfsck;
2385 struct lfsck_component *com;
2387 lfsck = lfsck_instance_find(key, true, false);
2388 if (unlikely(lfsck == NULL))
2391 com = lfsck_component_find(lfsck, lr->lr_active);
2392 if (likely(com != NULL)) {
2393 rc = com->lc_ops->lfsck_in_notify(env, com, lr);
2394 lfsck_component_put(env, com);
2397 lfsck_instance_put(env, lfsck);
2406 EXPORT_SYMBOL(lfsck_in_notify);
2408 int lfsck_query(const struct lu_env *env, struct dt_device *key,
2409 struct lfsck_request *lr)
2411 struct lfsck_instance *lfsck;
2412 struct lfsck_component *com;
2416 lfsck = lfsck_instance_find(key, true, false);
2417 if (unlikely(lfsck == NULL))
2420 com = lfsck_component_find(lfsck, lr->lr_active);
2421 if (likely(com != NULL)) {
2422 rc = com->lc_ops->lfsck_query(env, com);
2423 lfsck_component_put(env, com);
2428 lfsck_instance_put(env, lfsck);
2432 EXPORT_SYMBOL(lfsck_query);
2434 int lfsck_register_namespace(const struct lu_env *env, struct dt_device *key,
2435 struct ldlm_namespace *ns)
2437 struct lfsck_instance *lfsck;
2440 lfsck = lfsck_instance_find(key, true, false);
2441 if (likely(lfsck != NULL)) {
2442 lfsck->li_namespace = ns;
2443 lfsck_instance_put(env, lfsck);
2449 EXPORT_SYMBOL(lfsck_register_namespace);
2451 int lfsck_register(const struct lu_env *env, struct dt_device *key,
2452 struct dt_device *next, struct obd_device *obd,
2453 lfsck_out_notify notify, void *notify_data, bool master)
2455 struct lfsck_instance *lfsck;
2456 struct dt_object *root = NULL;
2457 struct dt_object *obj;
2458 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
2462 lfsck = lfsck_instance_find(key, false, false);
2463 if (unlikely(lfsck != NULL))
2466 OBD_ALLOC_PTR(lfsck);
2470 mutex_init(&lfsck->li_mutex);
2471 spin_lock_init(&lfsck->li_lock);
2472 CFS_INIT_LIST_HEAD(&lfsck->li_link);
2473 CFS_INIT_LIST_HEAD(&lfsck->li_list_scan);
2474 CFS_INIT_LIST_HEAD(&lfsck->li_list_dir);
2475 CFS_INIT_LIST_HEAD(&lfsck->li_list_double_scan);
2476 CFS_INIT_LIST_HEAD(&lfsck->li_list_idle);
2477 atomic_set(&lfsck->li_ref, 1);
2478 atomic_set(&lfsck->li_double_scan_count, 0);
2479 init_waitqueue_head(&lfsck->li_thread.t_ctl_waitq);
2480 lfsck->li_out_notify = notify;
2481 lfsck->li_out_notify_data = notify_data;
2482 lfsck->li_next = next;
2483 lfsck->li_bottom = key;
2484 lfsck->li_obd = obd;
2486 rc = lfsck_tgt_descs_init(&lfsck->li_ost_descs);
2490 rc = lfsck_tgt_descs_init(&lfsck->li_mdt_descs);
2494 fid->f_seq = FID_SEQ_LOCAL_NAME;
2497 rc = local_oid_storage_init(env, lfsck->li_bottom, fid, &lfsck->li_los);
2501 rc = dt_root_get(env, key, fid);
2505 root = dt_locate(env, lfsck->li_bottom, fid);
2507 GOTO(out, rc = PTR_ERR(root));
2509 if (unlikely(!dt_try_as_dir(env, root)))
2510 GOTO(out, rc = -ENOTDIR);
2512 lfsck->li_local_root_fid = *fid;
2514 lfsck->li_master = 1;
2515 if (lfsck_dev_idx(lfsck->li_bottom) == 0) {
2516 rc = dt_lookup(env, root,
2517 (struct dt_rec *)(&lfsck->li_global_root_fid),
2518 (const struct dt_key *)"ROOT", BYPASS_CAPA);
2524 fid->f_seq = FID_SEQ_LOCAL_FILE;
2525 fid->f_oid = OTABLE_IT_OID;
2527 obj = dt_locate(env, lfsck->li_bottom, fid);
2529 GOTO(out, rc = PTR_ERR(obj));
2531 lfsck->li_obj_oit = obj;
2532 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2534 if (rc == -ENOTSUPP)
2540 rc = lfsck_bookmark_setup(env, lfsck);
2545 rc = lfsck_fid_init(lfsck);
2549 rc = lfsck_namespace_setup(env, lfsck);
2554 rc = lfsck_layout_setup(env, lfsck);
2558 /* XXX: more LFSCK components initialization to be added here. */
2561 rc = lfsck_instance_add(lfsck);
2563 rc = lfsck_add_target_from_orphan(env, lfsck);
2565 if (root != NULL && !IS_ERR(root))
2566 lu_object_put(env, &root->do_lu);
2568 lfsck_instance_cleanup(env, lfsck);
2571 EXPORT_SYMBOL(lfsck_register);
2573 void lfsck_degister(const struct lu_env *env, struct dt_device *key)
2575 struct lfsck_instance *lfsck;
2577 lfsck = lfsck_instance_find(key, false, true);
2579 lfsck_instance_put(env, lfsck);
2581 EXPORT_SYMBOL(lfsck_degister);
2583 int lfsck_add_target(const struct lu_env *env, struct dt_device *key,
2584 struct dt_device *tgt, struct obd_export *exp,
2585 __u32 index, bool for_ost)
2587 struct lfsck_instance *lfsck;
2588 struct lfsck_tgt_desc *ltd;
2599 INIT_LIST_HEAD(<d->ltd_orphan_list);
2600 INIT_LIST_HEAD(<d->ltd_layout_list);
2601 INIT_LIST_HEAD(<d->ltd_layout_phase_list);
2602 atomic_set(<d->ltd_ref, 1);
2603 ltd->ltd_index = index;
2605 spin_lock(&lfsck_instance_lock);
2606 lfsck = __lfsck_instance_find(key, true, false);
2607 if (lfsck == NULL) {
2609 list_add_tail(<d->ltd_orphan_list,
2610 &lfsck_ost_orphan_list);
2612 list_add_tail(<d->ltd_orphan_list,
2613 &lfsck_mdt_orphan_list);
2614 spin_unlock(&lfsck_instance_lock);
2618 spin_unlock(&lfsck_instance_lock);
2620 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, false);
2624 lfsck_instance_put(env, lfsck);
2628 EXPORT_SYMBOL(lfsck_add_target);
2630 void lfsck_del_target(const struct lu_env *env, struct dt_device *key,
2631 struct dt_device *tgt, __u32 index, bool for_ost)
2633 struct lfsck_instance *lfsck;
2634 struct lfsck_tgt_descs *ltds;
2635 struct lfsck_tgt_desc *ltd = NULL;
2636 struct list_head *head;
2639 head = &lfsck_ost_orphan_list;
2641 head = &lfsck_mdt_orphan_list;
2643 spin_lock(&lfsck_instance_lock);
2644 list_for_each_entry(ltd, head, ltd_orphan_list) {
2645 if (ltd->ltd_tgt == tgt) {
2646 list_del_init(<d->ltd_orphan_list);
2647 spin_unlock(&lfsck_instance_lock);
2654 lfsck = __lfsck_instance_find(key, true, false);
2655 spin_unlock(&lfsck_instance_lock);
2656 if (unlikely(lfsck == NULL))
2660 ltds = &lfsck->li_ost_descs;
2662 ltds = &lfsck->li_mdt_descs;
2664 down_write(<ds->ltd_rw_sem);
2665 LASSERT(ltds->ltd_tgts_bitmap != NULL);
2667 if (unlikely(index >= ltds->ltd_tgts_bitmap->size))
2670 ltd = LTD_TGT(ltds, index);
2671 if (unlikely(ltd == NULL))
2674 LASSERT(ltds->ltd_tgtnr > 0);
2677 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, index);
2678 LTD_TGT(ltds, index) = NULL;
2683 head = &lfsck->li_ost_descs.ltd_orphan;
2685 head = &lfsck->li_ost_descs.ltd_orphan;
2687 list_for_each_entry(ltd, head, ltd_orphan_list) {
2688 if (ltd->ltd_tgt == tgt) {
2689 list_del_init(<d->ltd_orphan_list);
2695 up_write(<ds->ltd_rw_sem);
2697 spin_lock(<ds->ltd_lock);
2699 spin_unlock(<ds->ltd_lock);
2700 lfsck_stop_notify(env, lfsck, ltds, ltd, LT_LAYOUT);
2704 lfsck_instance_put(env, lfsck);
2706 EXPORT_SYMBOL(lfsck_del_target);
2708 static int __init lfsck_init(void)
2712 INIT_LIST_HEAD(&lfsck_ost_orphan_list);
2713 INIT_LIST_HEAD(&lfsck_mdt_orphan_list);
2714 lfsck_key_init_generic(&lfsck_thread_key, NULL);
2715 rc = lu_context_key_register(&lfsck_thread_key);
2717 tgt_register_lfsck_in_notify(lfsck_in_notify);
2718 tgt_register_lfsck_query(lfsck_query);
2724 static void __exit lfsck_exit(void)
2726 struct lfsck_tgt_desc *ltd;
2727 struct lfsck_tgt_desc *next;
2729 LASSERT(cfs_list_empty(&lfsck_instance_list));
2731 list_for_each_entry_safe(ltd, next, &lfsck_ost_orphan_list,
2733 list_del_init(<d->ltd_orphan_list);
2737 list_for_each_entry_safe(ltd, next, &lfsck_mdt_orphan_list,
2739 list_del_init(<d->ltd_orphan_list);
2743 lu_context_key_degister(&lfsck_thread_key);
2746 MODULE_AUTHOR("Intel Corporation <http://www.intel.com/>");
2747 MODULE_DESCRIPTION("LFSCK");
2748 MODULE_LICENSE("GPL");
2750 cfs_module(lfsck, LUSTRE_VERSION_STRING, lfsck_init, lfsck_exit);