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[] = {
101 const char *lfsck_status2names(enum lfsck_status status)
103 if (unlikely(status < 0 || status >= LS_MAX))
106 return lfsck_status_names[status];
109 static int lfsck_tgt_descs_init(struct lfsck_tgt_descs *ltds)
111 spin_lock_init(<ds->ltd_lock);
112 init_rwsem(<ds->ltd_rw_sem);
113 INIT_LIST_HEAD(<ds->ltd_orphan);
114 ltds->ltd_tgts_bitmap = CFS_ALLOCATE_BITMAP(BITS_PER_LONG);
115 if (ltds->ltd_tgts_bitmap == NULL)
121 static void lfsck_tgt_descs_fini(struct lfsck_tgt_descs *ltds)
123 struct lfsck_tgt_desc *ltd;
124 struct lfsck_tgt_desc *next;
127 down_write(<ds->ltd_rw_sem);
129 list_for_each_entry_safe(ltd, next, <ds->ltd_orphan,
131 list_del_init(<d->ltd_orphan_list);
135 if (unlikely(ltds->ltd_tgts_bitmap == NULL)) {
136 up_write(<ds->ltd_rw_sem);
141 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
142 ltd = LTD_TGT(ltds, idx);
143 if (likely(ltd != NULL)) {
144 LASSERT(list_empty(<d->ltd_layout_list));
145 LASSERT(list_empty(<d->ltd_layout_phase_list));
148 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, idx);
149 LTD_TGT(ltds, idx) = NULL;
154 LASSERTF(ltds->ltd_tgtnr == 0, "tgt count unmatched: %d\n",
157 for (idx = 0; idx < TGT_PTRS; idx++) {
158 if (ltds->ltd_tgts_idx[idx] != NULL) {
159 OBD_FREE_PTR(ltds->ltd_tgts_idx[idx]);
160 ltds->ltd_tgts_idx[idx] = NULL;
164 CFS_FREE_BITMAP(ltds->ltd_tgts_bitmap);
165 ltds->ltd_tgts_bitmap = NULL;
166 up_write(<ds->ltd_rw_sem);
169 static int __lfsck_add_target(const struct lu_env *env,
170 struct lfsck_instance *lfsck,
171 struct lfsck_tgt_desc *ltd,
172 bool for_ost, bool locked)
174 struct lfsck_tgt_descs *ltds;
175 __u32 index = ltd->ltd_index;
180 ltds = &lfsck->li_ost_descs;
182 ltds = &lfsck->li_mdt_descs;
185 down_write(<ds->ltd_rw_sem);
187 LASSERT(ltds->ltd_tgts_bitmap != NULL);
189 if (index >= ltds->ltd_tgts_bitmap->size) {
190 __u32 newsize = max((__u32)ltds->ltd_tgts_bitmap->size,
191 (__u32)BITS_PER_LONG);
192 cfs_bitmap_t *old_bitmap = ltds->ltd_tgts_bitmap;
193 cfs_bitmap_t *new_bitmap;
195 while (newsize < index + 1)
198 new_bitmap = CFS_ALLOCATE_BITMAP(newsize);
199 if (new_bitmap == NULL)
200 GOTO(unlock, rc = -ENOMEM);
202 if (ltds->ltd_tgtnr > 0)
203 cfs_bitmap_copy(new_bitmap, old_bitmap);
204 ltds->ltd_tgts_bitmap = new_bitmap;
205 CFS_FREE_BITMAP(old_bitmap);
208 if (cfs_bitmap_check(ltds->ltd_tgts_bitmap, index)) {
209 CERROR("%s: the device %s (%u) is registered already\n",
210 lfsck_lfsck2name(lfsck),
211 ltd->ltd_tgt->dd_lu_dev.ld_obd->obd_name, index);
212 GOTO(unlock, rc = -EEXIST);
215 if (ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK] == NULL) {
216 OBD_ALLOC_PTR(ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK]);
217 if (ltds->ltd_tgts_idx[index / TGT_PTRS_PER_BLOCK] == NULL)
218 GOTO(unlock, rc = -ENOMEM);
221 LTD_TGT(ltds, index) = ltd;
222 cfs_bitmap_set(ltds->ltd_tgts_bitmap, index);
225 GOTO(unlock, rc = 0);
229 up_write(<ds->ltd_rw_sem);
234 static int lfsck_add_target_from_orphan(const struct lu_env *env,
235 struct lfsck_instance *lfsck)
237 struct lfsck_tgt_descs *ltds = &lfsck->li_ost_descs;
238 struct lfsck_tgt_desc *ltd;
239 struct lfsck_tgt_desc *next;
240 struct list_head *head = &lfsck_ost_orphan_list;
245 spin_lock(&lfsck_instance_lock);
246 list_for_each_entry_safe(ltd, next, head, ltd_orphan_list) {
247 if (ltd->ltd_key == lfsck->li_bottom) {
248 list_del_init(<d->ltd_orphan_list);
249 list_add_tail(<d->ltd_orphan_list,
253 spin_unlock(&lfsck_instance_lock);
255 down_write(<ds->ltd_rw_sem);
256 while (!list_empty(<ds->ltd_orphan)) {
257 ltd = list_entry(ltds->ltd_orphan.next,
258 struct lfsck_tgt_desc,
260 list_del_init(<d->ltd_orphan_list);
261 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, true);
262 /* Do not hold the semaphore for too long time. */
263 up_write(<ds->ltd_rw_sem);
267 down_write(<ds->ltd_rw_sem);
269 up_write(<ds->ltd_rw_sem);
272 ltds = &lfsck->li_mdt_descs;
273 head = &lfsck_mdt_orphan_list;
281 static inline struct lfsck_component *
282 __lfsck_component_find(struct lfsck_instance *lfsck, __u16 type, cfs_list_t *list)
284 struct lfsck_component *com;
286 cfs_list_for_each_entry(com, list, lc_link) {
287 if (com->lc_type == type)
293 struct lfsck_component *
294 lfsck_component_find(struct lfsck_instance *lfsck, __u16 type)
296 struct lfsck_component *com;
298 spin_lock(&lfsck->li_lock);
299 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
303 com = __lfsck_component_find(lfsck, type,
304 &lfsck->li_list_double_scan);
308 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_idle);
312 lfsck_component_get(com);
313 spin_unlock(&lfsck->li_lock);
317 void lfsck_component_cleanup(const struct lu_env *env,
318 struct lfsck_component *com)
320 if (!cfs_list_empty(&com->lc_link))
321 cfs_list_del_init(&com->lc_link);
322 if (!cfs_list_empty(&com->lc_link_dir))
323 cfs_list_del_init(&com->lc_link_dir);
325 lfsck_component_put(env, com);
328 int lfsck_fid_alloc(const struct lu_env *env, struct lfsck_instance *lfsck,
329 struct lu_fid *fid, bool locked)
331 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
336 mutex_lock(&lfsck->li_mutex);
338 rc = seq_client_alloc_fid(env, lfsck->li_seq, fid);
340 bk->lb_last_fid = *fid;
341 /* We do not care about whether the subsequent sub-operations
342 * failed or not. The worst case is that one FID is lost that
343 * is not a big issue for the LFSCK since it is relative rare
344 * for LFSCK create. */
345 rc = lfsck_bookmark_store(env, lfsck);
349 mutex_unlock(&lfsck->li_mutex);
354 static const char dot[] = ".";
355 static const char dotdot[] = "..";
357 static int lfsck_create_lpf_local(const struct lu_env *env,
358 struct lfsck_instance *lfsck,
359 struct dt_object *parent,
360 struct dt_object *child,
362 struct dt_object_format *dof,
365 struct dt_device *dev = lfsck->li_bottom;
366 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
367 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
368 const struct lu_fid *cfid = lu_object_fid(&child->do_lu);
369 struct thandle *th = NULL;
371 int len = sizeof(struct lfsck_bookmark);
375 th = dt_trans_create(env, dev);
379 /* 1a. create child */
380 rc = dt_declare_create(env, child, la, NULL, dof, th);
384 /* 2a. increase child nlink */
385 rc = dt_declare_ref_add(env, child, th);
389 /* 3a. insert name into parent dir */
390 rc = dt_declare_insert(env, parent, (const struct dt_rec *)cfid,
391 (const struct dt_key *)name, th);
395 /* 4a. increase parent nlink */
396 rc = dt_declare_ref_add(env, parent, th);
400 /* 5a. update bookmark */
401 rc = dt_declare_record_write(env, bk_obj,
402 lfsck_buf_get(env, bk, len), 0, th);
406 rc = dt_trans_start_local(env, dev, th);
410 dt_write_lock(env, child, 0);
411 /* 1b.1 create child */
412 rc = dt_create(env, child, la, NULL, dof, th);
416 if (unlikely(!dt_try_as_dir(env, child)))
417 GOTO(unlock, rc = -ENOTDIR);
419 /* 1b.2 insert dot into child dir */
420 rc = dt_insert(env, child, (const struct dt_rec *)cfid,
421 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
425 /* 1b.3 insert dotdot into child dir */
426 rc = dt_insert(env, child, (const struct dt_rec *)&LU_LPF_FID,
427 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
431 /* 2b. increase child nlink */
432 rc = dt_ref_add(env, child, th);
433 dt_write_unlock(env, child);
437 /* 3b. insert name into parent dir */
438 rc = dt_insert(env, parent, (const struct dt_rec *)cfid,
439 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
443 dt_write_lock(env, parent, 0);
444 /* 4b. increase parent nlink */
445 rc = dt_ref_add(env, parent, th);
446 dt_write_unlock(env, parent);
450 bk->lb_lpf_fid = *cfid;
451 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
453 /* 5b. update bookmark */
454 rc = dt_record_write(env, bk_obj,
455 lfsck_buf_get(env, bk, len), &pos, th);
460 dt_write_unlock(env, child);
463 dt_trans_stop(env, dev, th);
468 static int lfsck_create_lpf_remote(const struct lu_env *env,
469 struct lfsck_instance *lfsck,
470 struct dt_object *parent,
471 struct dt_object *child,
473 struct dt_object_format *dof,
476 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
477 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
478 const struct lu_fid *cfid = lu_object_fid(&child->do_lu);
479 struct thandle *th = NULL;
480 struct dt_device *dev;
482 int len = sizeof(struct lfsck_bookmark);
486 /* Create .lustre/lost+found/MDTxxxx. */
488 /* XXX: Currently, cross-MDT create operation needs to create the child
489 * object firstly, then insert name into the parent directory. For
490 * this case, the child object resides on current MDT (local), but
491 * the parent ".lustre/lost+found" may be on remote MDT. It is not
492 * easy to contain all the sub-modifications orderly within single
495 * To avoid more inconsistency, we split the create operation into
498 * 1) create the child locally.
499 * 2) insert the name "MDTXXXX" in the parent ".lustre/lost+found"
500 * remotely and update the lfsck_bookmark::lb_lpf_fid locally.
502 * If 1) done but 2) failed, then the worst case is that we lose
503 * one object locally, which is not a big issue. (can be repaird
504 * by LFSCK phase III) */
508 dev = lfsck->li_bottom;
509 th = dt_trans_create(env, dev);
513 /* 1a. create child locally. */
514 rc = dt_declare_create(env, child, la, NULL, dof, th);
518 /* 2a. increase child nlink locally. */
519 rc = dt_declare_ref_add(env, child, th);
523 rc = dt_trans_start_local(env, dev, th);
527 dt_write_lock(env, child, 0);
528 /* 1b. create child locally. */
529 rc = dt_create(env, child, la, NULL, dof, th);
533 if (unlikely(!dt_try_as_dir(env, child)))
534 GOTO(unlock, rc = -ENOTDIR);
536 /* 2b.1 insert dot into child dir locally. */
537 rc = dt_insert(env, child, (const struct dt_rec *)cfid,
538 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
542 /* 2b.2 insert dotdot into child dir locally. */
543 rc = dt_insert(env, child, (const struct dt_rec *)&LU_LPF_FID,
544 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
548 /* 2b.3 increase child nlink locally. */
549 rc = dt_ref_add(env, child, th);
550 dt_write_unlock(env, child);
551 dt_trans_stop(env, dev, th);
555 /* Transaction II: */
557 dev = lfsck->li_next;
558 th = dt_trans_create(env, dev);
562 /* 3a. insert name into parent dir remotely. */
563 rc = dt_declare_insert(env, parent, (const struct dt_rec *)cfid,
564 (const struct dt_key *)name, th);
568 /* 4a. increase parent nlink remotely. */
569 rc = dt_declare_ref_add(env, parent, th);
573 /* 5a. decrease child nlink for dotdot locally if former remote
575 rc = dt_declare_ref_del(env, child, th);
579 /* 6a. decrease child nlink for dot locally if former remote
581 rc = dt_declare_ref_del(env, child, th);
585 /* 7a. destroy child locally if former remote update failed. */
586 rc = dt_declare_destroy(env, child, th);
590 /* 8a. update bookmark locally. */
591 rc = dt_declare_record_write(env, bk_obj,
592 lfsck_buf_get(env, bk, len), 0, th);
596 rc = dt_trans_start(env, dev, th);
600 /* 3b. insert name into parent dir remotely. */
601 rc = dt_insert(env, parent, (const struct dt_rec *)cfid,
602 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
604 dt_write_lock(env, parent, 0);
605 /* 4b. increase parent nlink remotely. */
606 rc = dt_ref_add(env, parent, th);
607 dt_write_unlock(env, parent);
610 /* 5b. decrease child nlink for dotdot locally. */
611 dt_ref_del(env, child, th);
612 /* 6b. decrease child nlink for dot locally. */
613 dt_ref_del(env, child, th);
614 /* 7b. destroy child locally. */
615 dt_destroy(env, child, th);
619 bk->lb_lpf_fid = *cfid;
620 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
622 /* 8b. update bookmark locally. */
623 rc = dt_record_write(env, bk_obj,
624 lfsck_buf_get(env, bk, len), &pos, th);
629 dt_write_unlock(env, child);
631 dt_trans_stop(env, dev, th);
636 /* Do NOT create .lustre/lost+found/MDTxxxx when register the lfsck instance,
637 * because the MDT0 maybe not reaady for sequence allocation yet. We do that
638 * only when it is required, such as orphan OST-objects repairing. */
639 int lfsck_create_lpf(const struct lu_env *env, struct lfsck_instance *lfsck)
641 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
642 struct lfsck_thread_info *info = lfsck_env_info(env);
643 struct lu_fid *cfid = &info->lti_fid2;
644 struct lu_attr *la = &info->lti_la;
645 struct dt_object_format *dof = &info->lti_dof;
646 struct dt_object *parent = NULL;
647 struct dt_object *child = NULL;
649 int node = lfsck_dev_idx(lfsck->li_bottom);
653 LASSERT(lfsck->li_master);
655 sprintf(name, "MDT%04x", node);
657 parent = lfsck_object_find_by_dev(env, lfsck->li_bottom,
660 struct lfsck_tgt_desc *ltd;
662 ltd = lfsck_tgt_get(&lfsck->li_mdt_descs, 0);
663 if (unlikely(ltd == NULL))
666 parent = lfsck_object_find_by_dev(env, ltd->ltd_tgt,
671 RETURN(PTR_ERR(parent));
673 if (unlikely(!dt_try_as_dir(env, parent)))
674 GOTO(out, rc = -ENOTDIR);
676 mutex_lock(&lfsck->li_mutex);
677 if (lfsck->li_lpf_obj != NULL)
678 GOTO(unlock, rc = 0);
680 if (fid_is_zero(&bk->lb_lpf_fid)) {
681 /* There is corner case that: in former LFSCK scanning we have
682 * created the .lustre/lost+found/MDTxxxx but failed to update
683 * the lfsck_bookmark::lb_lpf_fid successfully. So need lookup
684 * it from MDT0 firstly. */
685 rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
686 (const struct dt_key *)name, BYPASS_CAPA);
687 if (rc != 0 && rc != -ENOENT)
691 bk->lb_lpf_fid = *cfid;
692 rc = lfsck_bookmark_store(env, lfsck);
694 rc = lfsck_fid_alloc(env, lfsck, cfid, true);
699 *cfid = bk->lb_lpf_fid;
702 child = lfsck_object_find_by_dev(env, lfsck->li_bottom, cfid);
704 GOTO(unlock, rc = PTR_ERR(child));
706 if (dt_object_exists(child) != 0) {
707 if (unlikely(!dt_try_as_dir(env, child)))
708 GOTO(unlock, rc = -ENOTDIR);
710 lfsck->li_lpf_obj = child;
711 GOTO(unlock, rc = 0);
714 memset(la, 0, sizeof(*la));
715 la->la_atime = la->la_mtime = la->la_ctime = cfs_time_current_sec();
716 la->la_mode = S_IFDIR | S_IRWXU;
717 la->la_valid = LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE |
719 memset(dof, 0, sizeof(*dof));
720 dof->dof_type = dt_mode_to_dft(S_IFDIR);
723 rc = lfsck_create_lpf_local(env, lfsck, parent, child, la,
726 rc = lfsck_create_lpf_remote(env, lfsck, parent, child, la,
729 lfsck->li_lpf_obj = child;
734 mutex_unlock(&lfsck->li_mutex);
735 if (rc != 0 && child != NULL && !IS_ERR(child))
736 lu_object_put(env, &child->do_lu);
738 if (parent != NULL && !IS_ERR(parent))
739 lu_object_put(env, &parent->do_lu);
744 static int lfsck_fid_init(struct lfsck_instance *lfsck)
746 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
747 struct seq_server_site *ss;
752 ss = lu_site2seq(lfsck->li_bottom->dd_lu_dev.ld_site);
753 if (unlikely(ss == NULL))
756 OBD_ALLOC_PTR(lfsck->li_seq);
757 if (lfsck->li_seq == NULL)
760 OBD_ALLOC(prefix, MAX_OBD_NAME + 7);
762 GOTO(out, rc = -ENOMEM);
764 snprintf(prefix, MAX_OBD_NAME + 7, "lfsck-%s", lfsck_lfsck2name(lfsck));
765 rc = seq_client_init(lfsck->li_seq, NULL, LUSTRE_SEQ_METADATA, prefix,
767 OBD_FREE(prefix, MAX_OBD_NAME + 7);
771 if (fid_is_sane(&bk->lb_last_fid))
772 lfsck->li_seq->lcs_fid = bk->lb_last_fid;
777 OBD_FREE_PTR(lfsck->li_seq);
778 lfsck->li_seq = NULL;
783 static void lfsck_fid_fini(struct lfsck_instance *lfsck)
785 if (lfsck->li_seq != NULL) {
786 seq_client_fini(lfsck->li_seq);
787 OBD_FREE_PTR(lfsck->li_seq);
788 lfsck->li_seq = NULL;
792 void lfsck_instance_cleanup(const struct lu_env *env,
793 struct lfsck_instance *lfsck)
795 struct ptlrpc_thread *thread = &lfsck->li_thread;
796 struct lfsck_component *com;
799 LASSERT(list_empty(&lfsck->li_link));
800 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
802 if (lfsck->li_obj_oit != NULL) {
803 lu_object_put_nocache(env, &lfsck->li_obj_oit->do_lu);
804 lfsck->li_obj_oit = NULL;
807 LASSERT(lfsck->li_obj_dir == NULL);
809 while (!cfs_list_empty(&lfsck->li_list_scan)) {
810 com = cfs_list_entry(lfsck->li_list_scan.next,
811 struct lfsck_component,
813 lfsck_component_cleanup(env, com);
816 LASSERT(cfs_list_empty(&lfsck->li_list_dir));
818 while (!cfs_list_empty(&lfsck->li_list_double_scan)) {
819 com = cfs_list_entry(lfsck->li_list_double_scan.next,
820 struct lfsck_component,
822 lfsck_component_cleanup(env, com);
825 while (!cfs_list_empty(&lfsck->li_list_idle)) {
826 com = cfs_list_entry(lfsck->li_list_idle.next,
827 struct lfsck_component,
829 lfsck_component_cleanup(env, com);
832 lfsck_tgt_descs_fini(&lfsck->li_ost_descs);
833 lfsck_tgt_descs_fini(&lfsck->li_mdt_descs);
835 if (lfsck->li_bookmark_obj != NULL) {
836 lu_object_put_nocache(env, &lfsck->li_bookmark_obj->do_lu);
837 lfsck->li_bookmark_obj = NULL;
840 if (lfsck->li_lpf_obj != NULL) {
841 lu_object_put(env, &lfsck->li_lpf_obj->do_lu);
842 lfsck->li_lpf_obj = NULL;
845 if (lfsck->li_los != NULL) {
846 local_oid_storage_fini(env, lfsck->li_los);
847 lfsck->li_los = NULL;
850 lfsck_fid_fini(lfsck);
855 static inline struct lfsck_instance *
856 __lfsck_instance_find(struct dt_device *key, bool ref, bool unlink)
858 struct lfsck_instance *lfsck;
860 cfs_list_for_each_entry(lfsck, &lfsck_instance_list, li_link) {
861 if (lfsck->li_bottom == key) {
863 lfsck_instance_get(lfsck);
865 list_del_init(&lfsck->li_link);
874 struct lfsck_instance *lfsck_instance_find(struct dt_device *key, bool ref,
877 struct lfsck_instance *lfsck;
879 spin_lock(&lfsck_instance_lock);
880 lfsck = __lfsck_instance_find(key, ref, unlink);
881 spin_unlock(&lfsck_instance_lock);
886 static inline int lfsck_instance_add(struct lfsck_instance *lfsck)
888 struct lfsck_instance *tmp;
890 spin_lock(&lfsck_instance_lock);
891 cfs_list_for_each_entry(tmp, &lfsck_instance_list, li_link) {
892 if (lfsck->li_bottom == tmp->li_bottom) {
893 spin_unlock(&lfsck_instance_lock);
898 cfs_list_add_tail(&lfsck->li_link, &lfsck_instance_list);
899 spin_unlock(&lfsck_instance_lock);
903 int lfsck_bits_dump(char **buf, int *len, int bits, const char *names[],
910 bool newline = (bits != 0 ? false : true);
912 rc = snprintf(*buf, *len, "%s:%c", prefix, newline ? '\n' : ' ');
918 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
921 if (names[i] != NULL) {
925 rc = snprintf(*buf, *len, "%s%c", names[i],
926 newline ? '\n' : ',');
937 rc = snprintf(*buf, *len, "\n");
948 int lfsck_time_dump(char **buf, int *len, __u64 time, const char *prefix)
953 rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
954 cfs_time_current_sec() - time);
956 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
965 int lfsck_pos_dump(char **buf, int *len, struct lfsck_position *pos,
970 if (fid_is_zero(&pos->lp_dir_parent)) {
971 if (pos->lp_oit_cookie == 0)
972 rc = snprintf(*buf, *len, "%s: N/A, N/A, N/A\n",
975 rc = snprintf(*buf, *len, "%s: "LPU64", N/A, N/A\n",
976 prefix, pos->lp_oit_cookie);
978 rc = snprintf(*buf, *len, "%s: "LPU64", "DFID", "LPU64"\n",
979 prefix, pos->lp_oit_cookie,
980 PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
990 void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
991 struct lfsck_position *pos, bool init)
993 const struct dt_it_ops *iops = &lfsck->li_obj_oit->do_index_ops->dio_it;
995 if (unlikely(lfsck->li_di_oit == NULL)) {
996 memset(pos, 0, sizeof(*pos));
1000 pos->lp_oit_cookie = iops->store(env, lfsck->li_di_oit);
1001 if (!lfsck->li_current_oit_processed && !init)
1002 pos->lp_oit_cookie--;
1004 LASSERT(pos->lp_oit_cookie > 0);
1006 if (lfsck->li_di_dir != NULL) {
1007 struct dt_object *dto = lfsck->li_obj_dir;
1009 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
1012 if (pos->lp_dir_cookie >= MDS_DIR_END_OFF) {
1013 fid_zero(&pos->lp_dir_parent);
1014 pos->lp_dir_cookie = 0;
1016 pos->lp_dir_parent = *lfsck_dto2fid(dto);
1019 fid_zero(&pos->lp_dir_parent);
1020 pos->lp_dir_cookie = 0;
1024 bool __lfsck_set_speed(struct lfsck_instance *lfsck, __u32 limit)
1028 if (limit != LFSCK_SPEED_NO_LIMIT) {
1030 lfsck->li_sleep_rate = limit / HZ;
1031 lfsck->li_sleep_jif = 1;
1033 lfsck->li_sleep_rate = 1;
1034 lfsck->li_sleep_jif = HZ / limit;
1037 lfsck->li_sleep_jif = 0;
1038 lfsck->li_sleep_rate = 0;
1041 if (lfsck->li_bookmark_ram.lb_speed_limit != limit) {
1042 lfsck->li_bookmark_ram.lb_speed_limit = limit;
1049 void lfsck_control_speed(struct lfsck_instance *lfsck)
1051 struct ptlrpc_thread *thread = &lfsck->li_thread;
1052 struct l_wait_info lwi;
1054 if (lfsck->li_sleep_jif > 0 &&
1055 lfsck->li_new_scanned >= lfsck->li_sleep_rate) {
1056 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1057 LWI_ON_SIGNAL_NOOP, NULL);
1059 l_wait_event(thread->t_ctl_waitq,
1060 !thread_is_running(thread),
1062 lfsck->li_new_scanned = 0;
1066 void lfsck_control_speed_by_self(struct lfsck_component *com)
1068 struct lfsck_instance *lfsck = com->lc_lfsck;
1069 struct ptlrpc_thread *thread = &lfsck->li_thread;
1070 struct l_wait_info lwi;
1072 if (lfsck->li_sleep_jif > 0 &&
1073 com->lc_new_scanned >= lfsck->li_sleep_rate) {
1074 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1075 LWI_ON_SIGNAL_NOOP, NULL);
1077 l_wait_event(thread->t_ctl_waitq,
1078 !thread_is_running(thread),
1080 com->lc_new_scanned = 0;
1084 static int lfsck_parent_fid(const struct lu_env *env, struct dt_object *obj,
1087 if (unlikely(!S_ISDIR(lfsck_object_type(obj)) ||
1088 !dt_try_as_dir(env, obj)))
1091 return dt_lookup(env, obj, (struct dt_rec *)fid,
1092 (const struct dt_key *)"..", BYPASS_CAPA);
1095 static int lfsck_needs_scan_dir(const struct lu_env *env,
1096 struct lfsck_instance *lfsck,
1097 struct dt_object *obj)
1099 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
1103 if (!lfsck->li_master || !S_ISDIR(lfsck_object_type(obj)) ||
1104 cfs_list_empty(&lfsck->li_list_dir))
1108 /* XXX: Currently, we do not scan the "/REMOTE_PARENT_DIR",
1109 * which is the agent directory to manage the objects
1110 * which name entries reside on remote MDTs. Related
1111 * consistency verification will be processed in LFSCK
1113 if (lu_fid_eq(lfsck_dto2fid(obj), &lfsck->li_global_root_fid)) {
1115 lfsck_object_put(env, obj);
1119 /* No need to check .lustre and its children. */
1120 if (fid_seq_is_dot_lustre(fid_seq(lfsck_dto2fid(obj)))) {
1122 lfsck_object_put(env, obj);
1126 dt_read_lock(env, obj, MOR_TGT_CHILD);
1127 if (unlikely(lfsck_is_dead_obj(obj))) {
1128 dt_read_unlock(env, obj);
1130 lfsck_object_put(env, obj);
1134 rc = dt_xattr_get(env, obj,
1135 lfsck_buf_get(env, NULL, 0), XATTR_NAME_LINK,
1137 dt_read_unlock(env, obj);
1140 lfsck_object_put(env, obj);
1144 if (rc < 0 && rc != -ENODATA) {
1146 lfsck_object_put(env, obj);
1150 rc = lfsck_parent_fid(env, obj, fid);
1152 lfsck_object_put(env, obj);
1156 if (unlikely(lu_fid_eq(fid, &lfsck->li_local_root_fid)))
1159 obj = lfsck_object_find(env, lfsck, fid);
1162 else if (IS_ERR(obj))
1163 return PTR_ERR(obj);
1165 if (!dt_object_exists(obj)) {
1166 lfsck_object_put(env, obj);
1170 if (dt_object_remote(obj)) {
1171 /* .lustre/lost+found/MDTxxx can be remote directory. */
1172 if (fid_seq_is_dot_lustre(fid_seq(lfsck_dto2fid(obj))))
1175 /* Other remote directory should be client
1176 * visible and need to be checked. */
1178 lfsck_object_put(env, obj);
1187 struct lfsck_thread_args *lfsck_thread_args_init(struct lfsck_instance *lfsck,
1188 struct lfsck_component *com,
1189 struct lfsck_start_param *lsp)
1191 struct lfsck_thread_args *lta;
1196 return ERR_PTR(-ENOMEM);
1198 rc = lu_env_init(<a->lta_env, LCT_MD_THREAD | LCT_DT_THREAD);
1204 lta->lta_lfsck = lfsck_instance_get(lfsck);
1206 lta->lta_com = lfsck_component_get(com);
1213 void lfsck_thread_args_fini(struct lfsck_thread_args *lta)
1215 if (lta->lta_com != NULL)
1216 lfsck_component_put(<a->lta_env, lta->lta_com);
1217 lfsck_instance_put(<a->lta_env, lta->lta_lfsck);
1218 lu_env_fini(<a->lta_env);
1222 /* LFSCK wrap functions */
1224 void lfsck_fail(const struct lu_env *env, struct lfsck_instance *lfsck,
1227 struct lfsck_component *com;
1229 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1230 com->lc_ops->lfsck_fail(env, com, new_checked);
1234 int lfsck_checkpoint(const struct lu_env *env, struct lfsck_instance *lfsck)
1236 struct lfsck_component *com;
1240 if (likely(cfs_time_beforeq(cfs_time_current(),
1241 lfsck->li_time_next_checkpoint)))
1244 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1245 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1246 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
1251 lfsck->li_time_last_checkpoint = cfs_time_current();
1252 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1253 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1254 return rc1 != 0 ? rc1 : rc;
1257 int lfsck_prep(const struct lu_env *env, struct lfsck_instance *lfsck,
1258 struct lfsck_start_param *lsp)
1260 struct dt_object *obj = NULL;
1261 struct lfsck_component *com;
1262 struct lfsck_component *next;
1263 struct lfsck_position *pos = NULL;
1264 const struct dt_it_ops *iops =
1265 &lfsck->li_obj_oit->do_index_ops->dio_it;
1270 LASSERT(lfsck->li_obj_dir == NULL);
1271 LASSERT(lfsck->li_di_dir == NULL);
1273 lfsck->li_current_oit_processed = 0;
1274 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1275 com->lc_new_checked = 0;
1276 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1277 com->lc_journal = 0;
1279 rc = com->lc_ops->lfsck_prep(env, com, lsp);
1283 if ((pos == NULL) ||
1284 (!lfsck_pos_is_zero(&com->lc_pos_start) &&
1285 lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
1286 pos = &com->lc_pos_start;
1289 /* Init otable-based iterator. */
1291 rc = iops->load(env, lfsck->li_di_oit, 0);
1293 lfsck->li_oit_over = 1;
1300 rc = iops->load(env, lfsck->li_di_oit, pos->lp_oit_cookie);
1304 lfsck->li_oit_over = 1;
1306 if (!lfsck->li_master || fid_is_zero(&pos->lp_dir_parent))
1309 /* Find the directory for namespace-based traverse. */
1310 obj = lfsck_object_find(env, lfsck, &pos->lp_dir_parent);
1313 else if (IS_ERR(obj))
1314 RETURN(PTR_ERR(obj));
1316 /* XXX: Currently, skip remote object, the consistency for
1317 * remote object will be processed in LFSCK phase III. */
1318 if (!dt_object_exists(obj) || dt_object_remote(obj) ||
1319 unlikely(!S_ISDIR(lfsck_object_type(obj))))
1322 if (unlikely(!dt_try_as_dir(env, obj)))
1323 GOTO(out, rc = -ENOTDIR);
1325 /* Init the namespace-based directory traverse. */
1326 iops = &obj->do_index_ops->dio_it;
1327 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1329 GOTO(out, rc = PTR_ERR(di));
1331 LASSERT(pos->lp_dir_cookie < MDS_DIR_END_OFF);
1333 rc = iops->load(env, di, pos->lp_dir_cookie);
1334 if ((rc == 0) || (rc > 0 && pos->lp_dir_cookie > 0))
1335 rc = iops->next(env, di);
1341 iops->fini(env, di);
1345 lfsck->li_obj_dir = lfsck_object_get(obj);
1346 lfsck->li_cookie_dir = iops->store(env, di);
1347 spin_lock(&lfsck->li_lock);
1348 lfsck->li_di_dir = di;
1349 spin_unlock(&lfsck->li_lock);
1355 lfsck_object_put(env, obj);
1358 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1360 com->lc_ops->lfsck_post(env, com, rc, true);
1366 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, true);
1367 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1368 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
1373 lfsck->li_time_last_checkpoint = cfs_time_current();
1374 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1375 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1379 int lfsck_exec_oit(const struct lu_env *env, struct lfsck_instance *lfsck,
1380 struct dt_object *obj)
1382 struct lfsck_component *com;
1383 const struct dt_it_ops *iops;
1388 LASSERT(lfsck->li_obj_dir == NULL);
1390 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1391 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
1396 rc = lfsck_needs_scan_dir(env, lfsck, obj);
1400 if (unlikely(!dt_try_as_dir(env, obj)))
1401 GOTO(out, rc = -ENOTDIR);
1403 iops = &obj->do_index_ops->dio_it;
1404 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1406 GOTO(out, rc = PTR_ERR(di));
1408 rc = iops->load(env, di, 0);
1410 rc = iops->next(env, di);
1416 iops->fini(env, di);
1420 lfsck->li_obj_dir = lfsck_object_get(obj);
1421 lfsck->li_cookie_dir = iops->store(env, di);
1422 spin_lock(&lfsck->li_lock);
1423 lfsck->li_di_dir = di;
1424 spin_unlock(&lfsck->li_lock);
1430 lfsck_fail(env, lfsck, false);
1431 return (rc > 0 ? 0 : rc);
1434 int lfsck_exec_dir(const struct lu_env *env, struct lfsck_instance *lfsck,
1435 struct dt_object *obj, struct lu_dirent *ent)
1437 struct lfsck_component *com;
1440 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1441 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
1448 int lfsck_post(const struct lu_env *env, struct lfsck_instance *lfsck,
1451 struct lfsck_component *com;
1452 struct lfsck_component *next;
1456 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1457 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1458 rc = com->lc_ops->lfsck_post(env, com, result, false);
1463 lfsck->li_time_last_checkpoint = cfs_time_current();
1464 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1465 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1467 /* Ignore some component post failure to make other can go ahead. */
1471 static void lfsck_interpret(const struct lu_env *env,
1472 struct lfsck_instance *lfsck,
1473 struct ptlrpc_request *req, void *args, int result)
1475 struct lfsck_async_interpret_args *laia = args;
1476 struct lfsck_component *com;
1478 LASSERT(laia->laia_com == NULL);
1479 LASSERT(laia->laia_shared);
1481 spin_lock(&lfsck->li_lock);
1482 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1483 if (com->lc_ops->lfsck_interpret != NULL) {
1484 laia->laia_com = com;
1485 com->lc_ops->lfsck_interpret(env, req, laia, result);
1489 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1490 if (com->lc_ops->lfsck_interpret != NULL) {
1491 laia->laia_com = com;
1492 com->lc_ops->lfsck_interpret(env, req, laia, result);
1495 spin_unlock(&lfsck->li_lock);
1498 int lfsck_double_scan(const struct lu_env *env, struct lfsck_instance *lfsck)
1500 struct lfsck_component *com;
1501 struct lfsck_component *next;
1502 struct l_wait_info lwi = { 0 };
1506 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1507 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1508 com->lc_journal = 0;
1510 rc = com->lc_ops->lfsck_double_scan(env, com);
1515 l_wait_event(lfsck->li_thread.t_ctl_waitq,
1516 atomic_read(&lfsck->li_double_scan_count) == 0,
1519 if (lfsck->li_status != LS_PAUSED &&
1520 lfsck->li_status != LS_CO_PAUSED) {
1521 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1523 spin_lock(&lfsck->li_lock);
1524 list_del_init(&com->lc_link);
1525 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1526 spin_unlock(&lfsck->li_lock);
1530 return rc1 != 0 ? rc1 : rc;
1533 static int lfsck_stop_notify(const struct lu_env *env,
1534 struct lfsck_instance *lfsck,
1535 struct lfsck_tgt_descs *ltds,
1536 struct lfsck_tgt_desc *ltd, __u16 type)
1538 struct ptlrpc_request_set *set;
1539 struct lfsck_component *com;
1543 spin_lock(&lfsck->li_lock);
1544 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
1546 com = __lfsck_component_find(lfsck, type,
1547 &lfsck->li_list_double_scan);
1549 lfsck_component_get(com);
1550 spin_lock(&lfsck->li_lock);
1553 if (com->lc_ops->lfsck_stop_notify != NULL) {
1554 set = ptlrpc_prep_set();
1556 lfsck_component_put(env, com);
1561 rc = com->lc_ops->lfsck_stop_notify(env, com, ltds,
1564 rc = ptlrpc_set_wait(set);
1566 ptlrpc_set_destroy(set);
1569 lfsck_component_put(env, com);
1575 void lfsck_quit(const struct lu_env *env, struct lfsck_instance *lfsck)
1577 struct lfsck_component *com;
1578 struct lfsck_component *next;
1580 list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1582 if (com->lc_ops->lfsck_quit != NULL)
1583 com->lc_ops->lfsck_quit(env, com);
1585 spin_lock(&lfsck->li_lock);
1586 list_del_init(&com->lc_link);
1587 list_del_init(&com->lc_link_dir);
1588 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1589 spin_unlock(&lfsck->li_lock);
1592 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1594 if (com->lc_ops->lfsck_quit != NULL)
1595 com->lc_ops->lfsck_quit(env, com);
1597 spin_lock(&lfsck->li_lock);
1598 list_del_init(&com->lc_link);
1599 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1600 spin_unlock(&lfsck->li_lock);
1604 static int lfsck_async_interpret(const struct lu_env *env,
1605 struct ptlrpc_request *req,
1608 struct lfsck_async_interpret_args *laia = args;
1609 struct lfsck_instance *lfsck;
1611 lfsck = container_of0(laia->laia_ltds, struct lfsck_instance,
1613 lfsck_interpret(env, lfsck, req, laia, rc);
1614 lfsck_tgt_put(laia->laia_ltd);
1615 if (rc != 0 && laia->laia_result != -EALREADY)
1616 laia->laia_result = rc;
1621 int lfsck_async_request(const struct lu_env *env, struct obd_export *exp,
1622 struct lfsck_request *lr,
1623 struct ptlrpc_request_set *set,
1624 ptlrpc_interpterer_t interpreter,
1625 void *args, int request)
1627 struct lfsck_async_interpret_args *laia;
1628 struct ptlrpc_request *req;
1629 struct lfsck_request *tmp;
1630 struct req_format *format;
1635 format = &RQF_LFSCK_NOTIFY;
1638 format = &RQF_LFSCK_QUERY;
1641 CERROR("%s: unknown async request: opc = %d\n",
1642 exp->exp_obd->obd_name, request);
1646 req = ptlrpc_request_alloc(class_exp2cliimp(exp), format);
1650 rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, request);
1652 ptlrpc_request_free(req);
1657 tmp = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
1659 ptlrpc_request_set_replen(req);
1661 laia = ptlrpc_req_async_args(req);
1662 *laia = *(struct lfsck_async_interpret_args *)args;
1663 if (laia->laia_com != NULL)
1664 lfsck_component_get(laia->laia_com);
1665 req->rq_interpret_reply = interpreter;
1666 ptlrpc_set_add_req(set, req);
1671 /* external interfaces */
1673 int lfsck_get_speed(struct dt_device *key, void *buf, int len)
1676 struct lfsck_instance *lfsck;
1680 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1684 lfsck = lfsck_instance_find(key, true, false);
1685 if (likely(lfsck != NULL)) {
1686 rc = snprintf(buf, len, "%u\n",
1687 lfsck->li_bookmark_ram.lb_speed_limit);
1688 lfsck_instance_put(&env, lfsck);
1697 EXPORT_SYMBOL(lfsck_get_speed);
1699 int lfsck_set_speed(struct dt_device *key, int val)
1702 struct lfsck_instance *lfsck;
1706 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1710 lfsck = lfsck_instance_find(key, true, false);
1711 if (likely(lfsck != NULL)) {
1712 mutex_lock(&lfsck->li_mutex);
1713 if (__lfsck_set_speed(lfsck, val))
1714 rc = lfsck_bookmark_store(&env, lfsck);
1715 mutex_unlock(&lfsck->li_mutex);
1716 lfsck_instance_put(&env, lfsck);
1725 EXPORT_SYMBOL(lfsck_set_speed);
1727 int lfsck_get_windows(struct dt_device *key, void *buf, int len)
1730 struct lfsck_instance *lfsck;
1734 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1738 lfsck = lfsck_instance_find(key, true, false);
1739 if (likely(lfsck != NULL)) {
1740 rc = snprintf(buf, len, "%u\n",
1741 lfsck->li_bookmark_ram.lb_async_windows);
1742 lfsck_instance_put(&env, lfsck);
1751 EXPORT_SYMBOL(lfsck_get_windows);
1753 int lfsck_set_windows(struct dt_device *key, int val)
1756 struct lfsck_instance *lfsck;
1760 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1764 lfsck = lfsck_instance_find(key, true, false);
1765 if (likely(lfsck != NULL)) {
1766 if (val > LFSCK_ASYNC_WIN_MAX) {
1767 CWARN("%s: Too large async window size, which "
1768 "may cause memory issues. The valid range "
1769 "is [0 - %u]. If you do not want to restrict "
1770 "the window size for async requests pipeline, "
1771 "just set it as 0.\n",
1772 lfsck_lfsck2name(lfsck), LFSCK_ASYNC_WIN_MAX);
1774 } else if (lfsck->li_bookmark_ram.lb_async_windows != val) {
1775 mutex_lock(&lfsck->li_mutex);
1776 lfsck->li_bookmark_ram.lb_async_windows = val;
1777 rc = lfsck_bookmark_store(&env, lfsck);
1778 mutex_unlock(&lfsck->li_mutex);
1780 lfsck_instance_put(&env, lfsck);
1789 EXPORT_SYMBOL(lfsck_set_windows);
1791 int lfsck_dump(struct dt_device *key, void *buf, int len, enum lfsck_type type)
1794 struct lfsck_instance *lfsck;
1795 struct lfsck_component *com;
1799 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1803 lfsck = lfsck_instance_find(key, true, false);
1804 if (likely(lfsck != NULL)) {
1805 com = lfsck_component_find(lfsck, type);
1806 if (likely(com != NULL)) {
1807 rc = com->lc_ops->lfsck_dump(&env, com, buf, len);
1808 lfsck_component_put(&env, com);
1813 lfsck_instance_put(&env, lfsck);
1822 EXPORT_SYMBOL(lfsck_dump);
1824 static int lfsck_stop_all(const struct lu_env *env,
1825 struct lfsck_instance *lfsck,
1826 struct lfsck_stop *stop)
1828 struct lfsck_thread_info *info = lfsck_env_info(env);
1829 struct lfsck_request *lr = &info->lti_lr;
1830 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1831 struct ptlrpc_request_set *set;
1832 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1833 struct lfsck_tgt_desc *ltd;
1834 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1840 LASSERT(stop->ls_flags & LPF_BROADCAST);
1842 set = ptlrpc_prep_set();
1843 if (unlikely(set == NULL)) {
1844 CERROR("%s: cannot allocate memory for stop LFSCK on "
1845 "all targets\n", lfsck_lfsck2name(lfsck));
1850 memset(lr, 0, sizeof(*lr));
1851 lr->lr_event = LE_STOP;
1852 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1853 lr->lr_status = stop->ls_status;
1854 lr->lr_version = bk->lb_version;
1855 lr->lr_active = LFSCK_TYPES_ALL;
1856 lr->lr_param = stop->ls_flags;
1858 laia->laia_com = NULL;
1859 laia->laia_ltds = ltds;
1861 laia->laia_result = 0;
1862 laia->laia_shared = 1;
1864 down_read(<ds->ltd_rw_sem);
1865 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1866 ltd = lfsck_tgt_get(ltds, idx);
1867 LASSERT(ltd != NULL);
1869 laia->laia_ltd = ltd;
1870 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1871 lfsck_async_interpret, laia,
1874 lfsck_interpret(env, lfsck, NULL, laia, rc);
1876 CWARN("%s: cannot notify MDT %x for LFSCK stop: "
1877 "rc = %d\n", lfsck_lfsck2name(lfsck), idx, rc);
1881 up_read(<ds->ltd_rw_sem);
1883 rc = ptlrpc_set_wait(set);
1884 ptlrpc_set_destroy(set);
1887 rc = laia->laia_result;
1889 if (rc == -EALREADY)
1893 CWARN("%s: fail to stop LFSCK on some MDTs: rc = %d\n",
1894 lfsck_lfsck2name(lfsck), rc);
1896 RETURN(rc != 0 ? rc : rc1);
1899 static int lfsck_start_all(const struct lu_env *env,
1900 struct lfsck_instance *lfsck,
1901 struct lfsck_start *start)
1903 struct lfsck_thread_info *info = lfsck_env_info(env);
1904 struct lfsck_request *lr = &info->lti_lr;
1905 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1906 struct ptlrpc_request_set *set;
1907 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1908 struct lfsck_tgt_desc *ltd;
1909 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1914 LASSERT(start->ls_flags & LPF_BROADCAST);
1916 set = ptlrpc_prep_set();
1917 if (unlikely(set == NULL)) {
1918 if (bk->lb_param & LPF_FAILOUT) {
1919 CERROR("%s: cannot allocate memory for start LFSCK on "
1920 "all targets, failout.\n",
1921 lfsck_lfsck2name(lfsck));
1925 CWARN("%s: cannot allocate memory for start LFSCK on "
1926 "all targets, partly scan.\n",
1927 lfsck_lfsck2name(lfsck));
1933 memset(lr, 0, sizeof(*lr));
1934 lr->lr_event = LE_START;
1935 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1936 lr->lr_speed = bk->lb_speed_limit;
1937 lr->lr_version = bk->lb_version;
1938 lr->lr_active = start->ls_active;
1939 lr->lr_param = start->ls_flags;
1940 lr->lr_async_windows = bk->lb_async_windows;
1941 lr->lr_valid = LSV_SPEED_LIMIT | LSV_ERROR_HANDLE | LSV_DRYRUN |
1944 laia->laia_com = NULL;
1945 laia->laia_ltds = ltds;
1947 laia->laia_result = 0;
1948 laia->laia_shared = 1;
1950 down_read(<ds->ltd_rw_sem);
1951 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1952 ltd = lfsck_tgt_get(ltds, idx);
1953 LASSERT(ltd != NULL);
1955 laia->laia_ltd = ltd;
1956 ltd->ltd_layout_done = 0;
1957 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1958 lfsck_async_interpret, laia,
1961 lfsck_interpret(env, lfsck, NULL, laia, rc);
1963 if (bk->lb_param & LPF_FAILOUT) {
1964 CERROR("%s: cannot notify MDT %x for LFSCK "
1965 "start, failout: rc = %d\n",
1966 lfsck_lfsck2name(lfsck), idx, rc);
1969 CWARN("%s: cannot notify MDT %x for LFSCK "
1970 "start, partly scan: rc = %d\n",
1971 lfsck_lfsck2name(lfsck), idx, rc);
1976 up_read(<ds->ltd_rw_sem);
1979 ptlrpc_set_destroy(set);
1984 rc = ptlrpc_set_wait(set);
1985 ptlrpc_set_destroy(set);
1988 rc = laia->laia_result;
1991 if (bk->lb_param & LPF_FAILOUT) {
1992 struct lfsck_stop *stop = &info->lti_stop;
1994 CERROR("%s: cannot start LFSCK on some MDTs, "
1995 "stop all: rc = %d\n",
1996 lfsck_lfsck2name(lfsck), rc);
1997 if (rc != -EALREADY) {
1998 stop->ls_status = LS_FAILED;
1999 stop->ls_flags = LPF_ALL_TGT | LPF_BROADCAST;
2000 lfsck_stop_all(env, lfsck, stop);
2003 CWARN("%s: cannot start LFSCK on some MDTs, "
2004 "partly scan: rc = %d\n",
2005 lfsck_lfsck2name(lfsck), rc);
2013 int lfsck_start(const struct lu_env *env, struct dt_device *key,
2014 struct lfsck_start_param *lsp)
2016 struct lfsck_start *start = lsp->lsp_start;
2017 struct lfsck_instance *lfsck;
2018 struct lfsck_bookmark *bk;
2019 struct ptlrpc_thread *thread;
2020 struct lfsck_component *com;
2021 struct l_wait_info lwi = { 0 };
2022 struct lfsck_thread_args *lta;
2023 struct task_struct *task;
2030 lfsck = lfsck_instance_find(key, true, false);
2031 if (unlikely(lfsck == NULL))
2034 /* System is not ready, try again later. */
2035 if (unlikely(lfsck->li_namespace == NULL))
2036 GOTO(put, rc = -EAGAIN);
2038 /* start == NULL means auto trigger paused LFSCK. */
2039 if ((start == NULL) &&
2040 (cfs_list_empty(&lfsck->li_list_scan) ||
2041 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
2044 bk = &lfsck->li_bookmark_ram;
2045 thread = &lfsck->li_thread;
2046 mutex_lock(&lfsck->li_mutex);
2047 spin_lock(&lfsck->li_lock);
2048 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2050 if (unlikely(start == NULL)) {
2051 spin_unlock(&lfsck->li_lock);
2055 while (start->ls_active != 0) {
2056 if (!(type & start->ls_active)) {
2061 com = __lfsck_component_find(lfsck, type,
2062 &lfsck->li_list_scan);
2064 com = __lfsck_component_find(lfsck, type,
2065 &lfsck->li_list_double_scan);
2071 if (com->lc_ops->lfsck_join != NULL) {
2072 rc = com->lc_ops->lfsck_join( env, com, lsp);
2073 if (rc != 0 && rc != -EALREADY)
2076 start->ls_active &= ~type;
2079 spin_unlock(&lfsck->li_lock);
2082 spin_unlock(&lfsck->li_lock);
2084 lfsck->li_status = 0;
2085 lfsck->li_oit_over = 0;
2086 lfsck->li_start_unplug = 0;
2087 lfsck->li_drop_dryrun = 0;
2088 lfsck->li_new_scanned = 0;
2090 /* For auto trigger. */
2094 if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
2095 CERROR("%s: only allow to specify '-A | -o' via MDS\n",
2096 lfsck_lfsck2name(lfsck));
2098 GOTO(out, rc = -EPERM);
2101 start->ls_version = bk->lb_version;
2103 if (start->ls_active != 0) {
2104 struct lfsck_component *next;
2106 if (start->ls_active == LFSCK_TYPES_ALL)
2107 start->ls_active = LFSCK_TYPES_SUPPORTED;
2109 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2110 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2111 GOTO(out, rc = -ENOTSUPP);
2114 list_for_each_entry_safe(com, next,
2115 &lfsck->li_list_scan, lc_link) {
2116 if (!(com->lc_type & start->ls_active)) {
2117 rc = com->lc_ops->lfsck_post(env, com, 0,
2124 while (start->ls_active != 0) {
2125 if (type & start->ls_active) {
2126 com = __lfsck_component_find(lfsck, type,
2127 &lfsck->li_list_idle);
2129 /* The component status will be updated
2130 * when its prep() is called later by
2131 * the LFSCK main engine. */
2132 list_del_init(&com->lc_link);
2133 list_add_tail(&com->lc_link,
2134 &lfsck->li_list_scan);
2136 start->ls_active &= ~type;
2142 if (list_empty(&lfsck->li_list_scan)) {
2143 /* The speed limit will be used to control both the LFSCK and
2144 * low layer scrub (if applied), need to be handled firstly. */
2145 if (start->ls_valid & LSV_SPEED_LIMIT) {
2146 if (__lfsck_set_speed(lfsck, start->ls_speed_limit)) {
2147 rc = lfsck_bookmark_store(env, lfsck);
2156 if (start->ls_flags & LPF_RESET)
2157 flags |= DOIF_RESET;
2159 rc = lfsck_set_param(env, lfsck, start, !!(flags & DOIF_RESET));
2163 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
2164 start->ls_active |= com->lc_type;
2165 if (flags & DOIF_RESET) {
2166 rc = com->lc_ops->lfsck_reset(env, com, false);
2173 lfsck->li_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2174 if (bk->lb_param & LPF_DRYRUN)
2175 lfsck->li_args_dir |= LUDA_VERIFY_DRYRUN;
2177 if (start != NULL && start->ls_valid & LSV_ERROR_HANDLE) {
2178 valid |= DOIV_ERROR_HANDLE;
2179 if (start->ls_flags & LPF_FAILOUT)
2180 flags |= DOIF_FAILOUT;
2183 if (start != NULL && start->ls_valid & LSV_DRYRUN) {
2184 valid |= DOIV_DRYRUN;
2185 if (start->ls_flags & LPF_DRYRUN)
2186 flags |= DOIF_DRYRUN;
2189 if (!list_empty(&lfsck->li_list_scan))
2190 flags |= DOIF_OUTUSED;
2192 lfsck->li_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2193 thread_set_flags(thread, 0);
2194 lta = lfsck_thread_args_init(lfsck, NULL, lsp);
2196 GOTO(out, rc = PTR_ERR(lta));
2198 __lfsck_set_speed(lfsck, bk->lb_speed_limit);
2199 task = kthread_run(lfsck_master_engine, lta, "lfsck");
2202 CERROR("%s: cannot start LFSCK thread: rc = %d\n",
2203 lfsck_lfsck2name(lfsck), rc);
2204 lfsck_thread_args_fini(lta);
2209 l_wait_event(thread->t_ctl_waitq,
2210 thread_is_running(thread) ||
2211 thread_is_stopped(thread),
2213 if (start == NULL || !(start->ls_flags & LPF_BROADCAST)) {
2214 lfsck->li_start_unplug = 1;
2215 wake_up_all(&thread->t_ctl_waitq);
2220 /* release lfsck::li_mutex to avoid deadlock. */
2221 mutex_unlock(&lfsck->li_mutex);
2222 rc = lfsck_start_all(env, lfsck, start);
2224 spin_lock(&lfsck->li_lock);
2225 if (thread_is_stopped(thread)) {
2226 spin_unlock(&lfsck->li_lock);
2228 lfsck->li_status = LS_FAILED;
2229 lfsck->li_flags = 0;
2230 thread_set_flags(thread, SVC_STOPPING);
2231 spin_unlock(&lfsck->li_lock);
2233 lfsck->li_start_unplug = 1;
2234 wake_up_all(&thread->t_ctl_waitq);
2235 l_wait_event(thread->t_ctl_waitq,
2236 thread_is_stopped(thread),
2240 lfsck->li_start_unplug = 1;
2241 wake_up_all(&thread->t_ctl_waitq);
2247 mutex_unlock(&lfsck->li_mutex);
2250 lfsck_instance_put(env, lfsck);
2252 return rc < 0 ? rc : 0;
2254 EXPORT_SYMBOL(lfsck_start);
2256 int lfsck_stop(const struct lu_env *env, struct dt_device *key,
2257 struct lfsck_stop *stop)
2259 struct lfsck_instance *lfsck;
2260 struct ptlrpc_thread *thread;
2261 struct l_wait_info lwi = { 0 };
2266 lfsck = lfsck_instance_find(key, true, false);
2267 if (unlikely(lfsck == NULL))
2270 thread = &lfsck->li_thread;
2271 /* release lfsck::li_mutex to avoid deadlock. */
2272 if (stop != NULL && stop->ls_flags & LPF_BROADCAST) {
2273 if (!lfsck->li_master) {
2274 CERROR("%s: only allow to specify '-A' via MDS\n",
2275 lfsck_lfsck2name(lfsck));
2277 GOTO(out, rc = -EPERM);
2280 rc1 = lfsck_stop_all(env, lfsck, stop);
2283 mutex_lock(&lfsck->li_mutex);
2284 spin_lock(&lfsck->li_lock);
2285 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2286 spin_unlock(&lfsck->li_lock);
2287 GOTO(out, rc = -EALREADY);
2291 lfsck->li_status = stop->ls_status;
2292 lfsck->li_flags = stop->ls_flags;
2294 lfsck->li_status = LS_STOPPED;
2295 lfsck->li_flags = 0;
2298 thread_set_flags(thread, SVC_STOPPING);
2299 spin_unlock(&lfsck->li_lock);
2301 wake_up_all(&thread->t_ctl_waitq);
2302 l_wait_event(thread->t_ctl_waitq,
2303 thread_is_stopped(thread),
2309 mutex_unlock(&lfsck->li_mutex);
2310 lfsck_instance_put(env, lfsck);
2312 return rc != 0 ? rc : rc1;
2314 EXPORT_SYMBOL(lfsck_stop);
2316 int lfsck_in_notify(const struct lu_env *env, struct dt_device *key,
2317 struct lfsck_request *lr)
2319 int rc = -EOPNOTSUPP;
2322 switch (lr->lr_event) {
2324 struct lfsck_start *start = &lfsck_env_info(env)->lti_start;
2325 struct lfsck_start_param lsp;
2327 memset(start, 0, sizeof(*start));
2328 start->ls_valid = lr->lr_valid;
2329 start->ls_speed_limit = lr->lr_speed;
2330 start->ls_version = lr->lr_version;
2331 start->ls_active = lr->lr_active;
2332 start->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2333 start->ls_async_windows = lr->lr_async_windows;
2335 lsp.lsp_start = start;
2336 lsp.lsp_index = lr->lr_index;
2337 lsp.lsp_index_valid = 1;
2338 rc = lfsck_start(env, key, &lsp);
2342 struct lfsck_stop *stop = &lfsck_env_info(env)->lti_stop;
2344 memset(stop, 0, sizeof(*stop));
2345 stop->ls_status = lr->lr_status;
2346 stop->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2347 rc = lfsck_stop(env, key, stop);
2350 case LE_PHASE1_DONE:
2351 case LE_PHASE2_DONE:
2352 case LE_FID_ACCESSED:
2354 case LE_CONDITIONAL_DESTROY:
2355 case LE_PAIRS_VERIFY: {
2356 struct lfsck_instance *lfsck;
2357 struct lfsck_component *com;
2359 lfsck = lfsck_instance_find(key, true, false);
2360 if (unlikely(lfsck == NULL))
2363 com = lfsck_component_find(lfsck, lr->lr_active);
2364 if (likely(com != NULL)) {
2365 rc = com->lc_ops->lfsck_in_notify(env, com, lr);
2366 lfsck_component_put(env, com);
2369 lfsck_instance_put(env, lfsck);
2378 EXPORT_SYMBOL(lfsck_in_notify);
2380 int lfsck_query(const struct lu_env *env, struct dt_device *key,
2381 struct lfsck_request *lr)
2383 struct lfsck_instance *lfsck;
2384 struct lfsck_component *com;
2388 lfsck = lfsck_instance_find(key, true, false);
2389 if (unlikely(lfsck == NULL))
2392 com = lfsck_component_find(lfsck, lr->lr_active);
2393 if (likely(com != NULL)) {
2394 rc = com->lc_ops->lfsck_query(env, com);
2395 lfsck_component_put(env, com);
2400 lfsck_instance_put(env, lfsck);
2404 EXPORT_SYMBOL(lfsck_query);
2406 int lfsck_register_namespace(const struct lu_env *env, struct dt_device *key,
2407 struct ldlm_namespace *ns)
2409 struct lfsck_instance *lfsck;
2412 lfsck = lfsck_instance_find(key, true, false);
2413 if (likely(lfsck != NULL)) {
2414 lfsck->li_namespace = ns;
2415 lfsck_instance_put(env, lfsck);
2421 EXPORT_SYMBOL(lfsck_register_namespace);
2423 int lfsck_register(const struct lu_env *env, struct dt_device *key,
2424 struct dt_device *next, struct obd_device *obd,
2425 lfsck_out_notify notify, void *notify_data, bool master)
2427 struct lfsck_instance *lfsck;
2428 struct dt_object *root = NULL;
2429 struct dt_object *obj;
2430 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
2434 lfsck = lfsck_instance_find(key, false, false);
2435 if (unlikely(lfsck != NULL))
2438 OBD_ALLOC_PTR(lfsck);
2442 mutex_init(&lfsck->li_mutex);
2443 spin_lock_init(&lfsck->li_lock);
2444 CFS_INIT_LIST_HEAD(&lfsck->li_link);
2445 CFS_INIT_LIST_HEAD(&lfsck->li_list_scan);
2446 CFS_INIT_LIST_HEAD(&lfsck->li_list_dir);
2447 CFS_INIT_LIST_HEAD(&lfsck->li_list_double_scan);
2448 CFS_INIT_LIST_HEAD(&lfsck->li_list_idle);
2449 atomic_set(&lfsck->li_ref, 1);
2450 atomic_set(&lfsck->li_double_scan_count, 0);
2451 init_waitqueue_head(&lfsck->li_thread.t_ctl_waitq);
2452 lfsck->li_out_notify = notify;
2453 lfsck->li_out_notify_data = notify_data;
2454 lfsck->li_next = next;
2455 lfsck->li_bottom = key;
2456 lfsck->li_obd = obd;
2458 rc = lfsck_tgt_descs_init(&lfsck->li_ost_descs);
2462 rc = lfsck_tgt_descs_init(&lfsck->li_mdt_descs);
2466 fid->f_seq = FID_SEQ_LOCAL_NAME;
2469 rc = local_oid_storage_init(env, lfsck->li_bottom, fid, &lfsck->li_los);
2473 rc = dt_root_get(env, key, fid);
2477 root = dt_locate(env, lfsck->li_bottom, fid);
2479 GOTO(out, rc = PTR_ERR(root));
2481 if (unlikely(!dt_try_as_dir(env, root)))
2482 GOTO(out, rc = -ENOTDIR);
2484 lfsck->li_local_root_fid = *fid;
2486 lfsck->li_master = 1;
2487 if (lfsck_dev_idx(lfsck->li_bottom) == 0) {
2488 rc = dt_lookup(env, root,
2489 (struct dt_rec *)(&lfsck->li_global_root_fid),
2490 (const struct dt_key *)"ROOT", BYPASS_CAPA);
2496 fid->f_seq = FID_SEQ_LOCAL_FILE;
2497 fid->f_oid = OTABLE_IT_OID;
2499 obj = dt_locate(env, lfsck->li_bottom, fid);
2501 GOTO(out, rc = PTR_ERR(obj));
2503 lfsck->li_obj_oit = obj;
2504 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2506 if (rc == -ENOTSUPP)
2512 rc = lfsck_bookmark_setup(env, lfsck);
2517 rc = lfsck_fid_init(lfsck);
2521 rc = lfsck_namespace_setup(env, lfsck);
2526 rc = lfsck_layout_setup(env, lfsck);
2530 /* XXX: more LFSCK components initialization to be added here. */
2533 rc = lfsck_instance_add(lfsck);
2535 rc = lfsck_add_target_from_orphan(env, lfsck);
2537 if (root != NULL && !IS_ERR(root))
2538 lu_object_put(env, &root->do_lu);
2540 lfsck_instance_cleanup(env, lfsck);
2543 EXPORT_SYMBOL(lfsck_register);
2545 void lfsck_degister(const struct lu_env *env, struct dt_device *key)
2547 struct lfsck_instance *lfsck;
2549 lfsck = lfsck_instance_find(key, false, true);
2551 lfsck_instance_put(env, lfsck);
2553 EXPORT_SYMBOL(lfsck_degister);
2555 int lfsck_add_target(const struct lu_env *env, struct dt_device *key,
2556 struct dt_device *tgt, struct obd_export *exp,
2557 __u32 index, bool for_ost)
2559 struct lfsck_instance *lfsck;
2560 struct lfsck_tgt_desc *ltd;
2571 INIT_LIST_HEAD(<d->ltd_orphan_list);
2572 INIT_LIST_HEAD(<d->ltd_layout_list);
2573 INIT_LIST_HEAD(<d->ltd_layout_phase_list);
2574 atomic_set(<d->ltd_ref, 1);
2575 ltd->ltd_index = index;
2577 spin_lock(&lfsck_instance_lock);
2578 lfsck = __lfsck_instance_find(key, true, false);
2579 if (lfsck == NULL) {
2581 list_add_tail(<d->ltd_orphan_list,
2582 &lfsck_ost_orphan_list);
2584 list_add_tail(<d->ltd_orphan_list,
2585 &lfsck_mdt_orphan_list);
2586 spin_unlock(&lfsck_instance_lock);
2590 spin_unlock(&lfsck_instance_lock);
2592 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, false);
2596 lfsck_instance_put(env, lfsck);
2600 EXPORT_SYMBOL(lfsck_add_target);
2602 void lfsck_del_target(const struct lu_env *env, struct dt_device *key,
2603 struct dt_device *tgt, __u32 index, bool for_ost)
2605 struct lfsck_instance *lfsck;
2606 struct lfsck_tgt_descs *ltds;
2607 struct lfsck_tgt_desc *ltd = NULL;
2608 struct list_head *head;
2611 head = &lfsck_ost_orphan_list;
2613 head = &lfsck_mdt_orphan_list;
2615 spin_lock(&lfsck_instance_lock);
2616 list_for_each_entry(ltd, head, ltd_orphan_list) {
2617 if (ltd->ltd_tgt == tgt) {
2618 list_del_init(<d->ltd_orphan_list);
2619 spin_unlock(&lfsck_instance_lock);
2626 lfsck = __lfsck_instance_find(key, true, false);
2627 spin_unlock(&lfsck_instance_lock);
2628 if (unlikely(lfsck == NULL))
2632 ltds = &lfsck->li_ost_descs;
2634 ltds = &lfsck->li_mdt_descs;
2636 down_write(<ds->ltd_rw_sem);
2637 LASSERT(ltds->ltd_tgts_bitmap != NULL);
2639 if (unlikely(index >= ltds->ltd_tgts_bitmap->size))
2642 ltd = LTD_TGT(ltds, index);
2643 if (unlikely(ltd == NULL))
2646 LASSERT(ltds->ltd_tgtnr > 0);
2649 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, index);
2650 LTD_TGT(ltds, index) = NULL;
2655 head = &lfsck->li_ost_descs.ltd_orphan;
2657 head = &lfsck->li_ost_descs.ltd_orphan;
2659 list_for_each_entry(ltd, head, ltd_orphan_list) {
2660 if (ltd->ltd_tgt == tgt) {
2661 list_del_init(<d->ltd_orphan_list);
2667 up_write(<ds->ltd_rw_sem);
2669 spin_lock(<ds->ltd_lock);
2671 spin_unlock(<ds->ltd_lock);
2672 lfsck_stop_notify(env, lfsck, ltds, ltd, LT_LAYOUT);
2676 lfsck_instance_put(env, lfsck);
2678 EXPORT_SYMBOL(lfsck_del_target);
2680 static int __init lfsck_init(void)
2684 INIT_LIST_HEAD(&lfsck_ost_orphan_list);
2685 INIT_LIST_HEAD(&lfsck_mdt_orphan_list);
2686 lfsck_key_init_generic(&lfsck_thread_key, NULL);
2687 rc = lu_context_key_register(&lfsck_thread_key);
2689 tgt_register_lfsck_in_notify(lfsck_in_notify);
2690 tgt_register_lfsck_query(lfsck_query);
2696 static void __exit lfsck_exit(void)
2698 struct lfsck_tgt_desc *ltd;
2699 struct lfsck_tgt_desc *next;
2701 LASSERT(cfs_list_empty(&lfsck_instance_list));
2703 list_for_each_entry_safe(ltd, next, &lfsck_ost_orphan_list,
2705 list_del_init(<d->ltd_orphan_list);
2709 list_for_each_entry_safe(ltd, next, &lfsck_mdt_orphan_list,
2711 list_del_init(<d->ltd_orphan_list);
2715 lu_context_key_degister(&lfsck_thread_key);
2718 MODULE_AUTHOR("Intel Corporation <http://www.intel.com/>");
2719 MODULE_DESCRIPTION("LFSCK");
2720 MODULE_LICENSE("GPL");
2722 cfs_module(lfsck, LUSTRE_VERSION_STRING, lfsck_init, lfsck_exit);