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[] = "..";
356 static const char dotlustre[] = ".lustre";
357 static const char lostfound[] = "lost+found";
359 static int lfsck_create_lpf_local(const struct lu_env *env,
360 struct lfsck_instance *lfsck,
361 struct dt_object *parent,
362 struct dt_object *child,
364 struct dt_object_format *dof,
367 struct dt_insert_rec *rec = &lfsck_env_info(env)->lti_dt_rec;
368 struct dt_device *dev = lfsck->li_bottom;
369 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
370 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
371 const struct lu_fid *cfid = lfsck_dto2fid(child);
372 struct thandle *th = NULL;
373 struct linkea_data ldata = { 0 };
374 struct lu_buf linkea_buf;
375 const struct lu_name *cname;
377 int len = sizeof(struct lfsck_bookmark);
381 rc = linkea_data_new(&ldata,
382 &lfsck_env_info(env)->lti_linkea_buf);
386 cname = lfsck_name_get_const(env, name, strlen(name));
387 rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
391 th = dt_trans_create(env, dev);
395 /* 1a. create child */
396 rc = dt_declare_create(env, child, la, NULL, dof, th);
400 /* 2a. increase child nlink */
401 rc = dt_declare_ref_add(env, child, th);
405 /* 3a. insert linkEA for child */
406 linkea_buf.lb_buf = ldata.ld_buf->lb_buf;
407 linkea_buf.lb_len = ldata.ld_leh->leh_len;
408 rc = dt_declare_xattr_set(env, child, &linkea_buf,
409 XATTR_NAME_LINK, 0, th);
413 /* 4a. insert name into parent dir */
414 rec->rec_type = S_IFDIR;
416 rc = dt_declare_insert(env, parent, (const struct dt_rec *)rec,
417 (const struct dt_key *)name, th);
421 /* 5a. increase parent nlink */
422 rc = dt_declare_ref_add(env, parent, th);
426 /* 6a. update bookmark */
427 rc = dt_declare_record_write(env, bk_obj,
428 lfsck_buf_get(env, bk, len), 0, th);
432 rc = dt_trans_start_local(env, dev, th);
436 dt_write_lock(env, child, 0);
437 /* 1b.1. create child */
438 rc = dt_create(env, child, la, NULL, dof, th);
442 if (unlikely(!dt_try_as_dir(env, child)))
443 GOTO(unlock, rc = -ENOTDIR);
445 /* 1b.2. insert dot into child dir */
447 rc = dt_insert(env, child, (const struct dt_rec *)rec,
448 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
452 /* 1b.3. insert dotdot into child dir */
453 rec->rec_fid = &LU_LPF_FID;
454 rc = dt_insert(env, child, (const struct dt_rec *)rec,
455 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
459 /* 2b. increase child nlink */
460 rc = dt_ref_add(env, child, th);
464 /* 3b. insert linkEA for child. */
465 rc = dt_xattr_set(env, child, &linkea_buf,
466 XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
467 dt_write_unlock(env, child);
471 /* 4b. insert name into parent dir */
473 rc = dt_insert(env, parent, (const struct dt_rec *)rec,
474 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
478 dt_write_lock(env, parent, 0);
479 /* 5b. increase parent nlink */
480 rc = dt_ref_add(env, parent, th);
481 dt_write_unlock(env, parent);
485 bk->lb_lpf_fid = *cfid;
486 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
488 /* 6b. update bookmark */
489 rc = dt_record_write(env, bk_obj,
490 lfsck_buf_get(env, bk, len), &pos, th);
495 dt_write_unlock(env, child);
498 dt_trans_stop(env, dev, th);
503 static int lfsck_create_lpf_remote(const struct lu_env *env,
504 struct lfsck_instance *lfsck,
505 struct dt_object *parent,
506 struct dt_object *child,
508 struct dt_object_format *dof,
511 struct dt_insert_rec *rec = &lfsck_env_info(env)->lti_dt_rec;
512 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
513 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
514 const struct lu_fid *cfid = lfsck_dto2fid(child);
515 struct thandle *th = NULL;
516 struct linkea_data ldata = { 0 };
517 struct lu_buf linkea_buf;
518 const struct lu_name *cname;
519 struct dt_device *dev;
521 int len = sizeof(struct lfsck_bookmark);
525 rc = linkea_data_new(&ldata,
526 &lfsck_env_info(env)->lti_linkea_buf);
530 cname = lfsck_name_get_const(env, name, strlen(name));
531 rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
535 /* Create .lustre/lost+found/MDTxxxx. */
537 /* XXX: Currently, cross-MDT create operation needs to create the child
538 * object firstly, then insert name into the parent directory. For
539 * this case, the child object resides on current MDT (local), but
540 * the parent ".lustre/lost+found" may be on remote MDT. It is not
541 * easy to contain all the sub-modifications orderly within single
544 * To avoid more inconsistency, we split the create operation into
547 * 1) create the child and update the lfsck_bookmark::lb_lpf_fid
549 * 2) insert the name "MDTXXXX" in the parent ".lustre/lost+found"
552 * If 1) done, but 2) failed, then go ahead, the LFSCK will try to
553 * repair such inconsistency when LFSCK run next time. */
555 /* Transaction I: locally */
557 dev = lfsck->li_bottom;
558 th = dt_trans_create(env, dev);
562 /* 1a. create child */
563 rc = dt_declare_create(env, child, la, NULL, dof, th);
567 /* 2a. increase child nlink */
568 rc = dt_declare_ref_add(env, child, th);
572 /* 3a. insert linkEA for child */
573 linkea_buf.lb_buf = ldata.ld_buf->lb_buf;
574 linkea_buf.lb_len = ldata.ld_leh->leh_len;
575 rc = dt_declare_xattr_set(env, child, &linkea_buf,
576 XATTR_NAME_LINK, 0, th);
580 /* 4a. update bookmark */
581 rc = dt_declare_record_write(env, bk_obj,
582 lfsck_buf_get(env, bk, len), 0, th);
586 rc = dt_trans_start_local(env, dev, th);
590 dt_write_lock(env, child, 0);
591 /* 1b.1. create child */
592 rc = dt_create(env, child, la, NULL, dof, th);
596 if (unlikely(!dt_try_as_dir(env, child)))
597 GOTO(unlock, rc = -ENOTDIR);
599 /* 1b.2. insert dot into child dir */
600 rec->rec_type = S_IFDIR;
602 rc = dt_insert(env, child, (const struct dt_rec *)rec,
603 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
607 /* 1b.3. insert dotdot into child dir */
608 rec->rec_fid = &LU_LPF_FID;
609 rc = dt_insert(env, child, (const struct dt_rec *)rec,
610 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
614 /* 2b. increase child nlink */
615 rc = dt_ref_add(env, child, th);
619 /* 3b. insert linkEA for child */
620 rc = dt_xattr_set(env, child, &linkea_buf,
621 XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
625 bk->lb_lpf_fid = *cfid;
626 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
628 /* 4b. update bookmark */
629 rc = dt_record_write(env, bk_obj,
630 lfsck_buf_get(env, bk, len), &pos, th);
632 dt_write_unlock(env, child);
633 dt_trans_stop(env, dev, th);
637 /* Transaction II: remotely */
639 dev = lfsck->li_next;
640 th = dt_trans_create(env, dev);
644 /* 5a. insert name into parent dir */
646 rc = dt_declare_insert(env, parent, (const struct dt_rec *)rec,
647 (const struct dt_key *)name, th);
651 /* 6a. increase parent nlink */
652 rc = dt_declare_ref_add(env, parent, th);
656 rc = dt_trans_start(env, dev, th);
660 /* 5b. insert name into parent dir */
661 rc = dt_insert(env, parent, (const struct dt_rec *)rec,
662 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
666 dt_write_lock(env, parent, 0);
667 /* 6b. increase parent nlink */
668 rc = dt_ref_add(env, parent, th);
669 dt_write_unlock(env, parent);
674 dt_write_unlock(env, child);
676 dt_trans_stop(env, dev, th);
678 if (rc != 0 && dev == lfsck->li_next)
679 CDEBUG(D_LFSCK, "%s: partially created the object "DFID
680 "for orphans, but failed to insert the name %s "
681 "to the .lustre/lost+found/. Such inconsistency "
682 "will be repaired when LFSCK run next time: rc = %d\n",
683 lfsck_lfsck2name(lfsck), PFID(cfid), name, rc);
688 /* Do NOT create .lustre/lost+found/MDTxxxx when register the lfsck instance,
689 * because the MDT0 maybe not reaady for sequence allocation yet. We do that
690 * only when it is required, such as orphan OST-objects repairing. */
691 int lfsck_create_lpf(const struct lu_env *env, struct lfsck_instance *lfsck)
693 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
694 struct lfsck_thread_info *info = lfsck_env_info(env);
695 struct lu_fid *cfid = &info->lti_fid2;
696 struct lu_attr *la = &info->lti_la;
697 struct dt_object_format *dof = &info->lti_dof;
698 struct dt_object *parent = NULL;
699 struct dt_object *child = NULL;
701 int node = lfsck_dev_idx(lfsck->li_bottom);
705 LASSERT(lfsck->li_master);
707 sprintf(name, "MDT%04x", node);
709 parent = lfsck_object_find_by_dev(env, lfsck->li_bottom,
712 struct lfsck_tgt_desc *ltd;
714 ltd = lfsck_tgt_get(&lfsck->li_mdt_descs, 0);
715 if (unlikely(ltd == NULL))
718 parent = lfsck_object_find_by_dev(env, ltd->ltd_tgt,
723 RETURN(PTR_ERR(parent));
725 if (unlikely(!dt_try_as_dir(env, parent)))
726 GOTO(out, rc = -ENOTDIR);
728 mutex_lock(&lfsck->li_mutex);
729 if (lfsck->li_lpf_obj != NULL)
730 GOTO(unlock, rc = 0);
732 if (fid_is_zero(&bk->lb_lpf_fid)) {
733 /* There is corner case that: in former LFSCK scanning we have
734 * created the .lustre/lost+found/MDTxxxx but failed to update
735 * the lfsck_bookmark::lb_lpf_fid successfully. So need lookup
736 * it from MDT0 firstly. */
737 rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
738 (const struct dt_key *)name, BYPASS_CAPA);
739 if (rc != 0 && rc != -ENOENT)
743 bk->lb_lpf_fid = *cfid;
744 rc = lfsck_bookmark_store(env, lfsck);
746 rc = lfsck_fid_alloc(env, lfsck, cfid, true);
751 *cfid = bk->lb_lpf_fid;
754 child = lfsck_object_find_by_dev(env, lfsck->li_bottom, cfid);
756 GOTO(unlock, rc = PTR_ERR(child));
758 if (dt_object_exists(child) != 0) {
759 if (unlikely(!dt_try_as_dir(env, child)))
762 lfsck->li_lpf_obj = child;
767 memset(la, 0, sizeof(*la));
768 la->la_atime = la->la_mtime = la->la_ctime = cfs_time_current_sec();
769 la->la_mode = S_IFDIR | S_IRWXU;
770 la->la_valid = LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE |
772 memset(dof, 0, sizeof(*dof));
773 dof->dof_type = dt_mode_to_dft(S_IFDIR);
776 rc = lfsck_create_lpf_local(env, lfsck, parent, child, la,
779 rc = lfsck_create_lpf_remote(env, lfsck, parent, child, la,
782 lfsck->li_lpf_obj = child;
787 mutex_unlock(&lfsck->li_mutex);
788 if (rc != 0 && child != NULL && !IS_ERR(child))
789 lu_object_put(env, &child->do_lu);
791 if (parent != NULL && !IS_ERR(parent))
792 lu_object_put(env, &parent->do_lu);
797 static int lfsck_fid_init(struct lfsck_instance *lfsck)
799 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
800 struct seq_server_site *ss;
805 ss = lu_site2seq(lfsck->li_bottom->dd_lu_dev.ld_site);
806 if (unlikely(ss == NULL))
809 OBD_ALLOC_PTR(lfsck->li_seq);
810 if (lfsck->li_seq == NULL)
813 OBD_ALLOC(prefix, MAX_OBD_NAME + 7);
815 GOTO(out, rc = -ENOMEM);
817 snprintf(prefix, MAX_OBD_NAME + 7, "lfsck-%s", lfsck_lfsck2name(lfsck));
818 rc = seq_client_init(lfsck->li_seq, NULL, LUSTRE_SEQ_METADATA, prefix,
820 OBD_FREE(prefix, MAX_OBD_NAME + 7);
824 if (fid_is_sane(&bk->lb_last_fid))
825 lfsck->li_seq->lcs_fid = bk->lb_last_fid;
830 OBD_FREE_PTR(lfsck->li_seq);
831 lfsck->li_seq = NULL;
836 static void lfsck_fid_fini(struct lfsck_instance *lfsck)
838 if (lfsck->li_seq != NULL) {
839 seq_client_fini(lfsck->li_seq);
840 OBD_FREE_PTR(lfsck->li_seq);
841 lfsck->li_seq = NULL;
845 void lfsck_instance_cleanup(const struct lu_env *env,
846 struct lfsck_instance *lfsck)
848 struct ptlrpc_thread *thread = &lfsck->li_thread;
849 struct lfsck_component *com;
852 LASSERT(list_empty(&lfsck->li_link));
853 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
855 if (lfsck->li_obj_oit != NULL) {
856 lu_object_put_nocache(env, &lfsck->li_obj_oit->do_lu);
857 lfsck->li_obj_oit = NULL;
860 LASSERT(lfsck->li_obj_dir == NULL);
862 while (!cfs_list_empty(&lfsck->li_list_scan)) {
863 com = cfs_list_entry(lfsck->li_list_scan.next,
864 struct lfsck_component,
866 lfsck_component_cleanup(env, com);
869 LASSERT(cfs_list_empty(&lfsck->li_list_dir));
871 while (!cfs_list_empty(&lfsck->li_list_double_scan)) {
872 com = cfs_list_entry(lfsck->li_list_double_scan.next,
873 struct lfsck_component,
875 lfsck_component_cleanup(env, com);
878 while (!cfs_list_empty(&lfsck->li_list_idle)) {
879 com = cfs_list_entry(lfsck->li_list_idle.next,
880 struct lfsck_component,
882 lfsck_component_cleanup(env, com);
885 lfsck_tgt_descs_fini(&lfsck->li_ost_descs);
886 lfsck_tgt_descs_fini(&lfsck->li_mdt_descs);
888 if (lfsck->li_bookmark_obj != NULL) {
889 lu_object_put_nocache(env, &lfsck->li_bookmark_obj->do_lu);
890 lfsck->li_bookmark_obj = NULL;
893 if (lfsck->li_lpf_obj != NULL) {
894 lu_object_put(env, &lfsck->li_lpf_obj->do_lu);
895 lfsck->li_lpf_obj = NULL;
898 if (lfsck->li_los != NULL) {
899 local_oid_storage_fini(env, lfsck->li_los);
900 lfsck->li_los = NULL;
903 lfsck_fid_fini(lfsck);
908 static inline struct lfsck_instance *
909 __lfsck_instance_find(struct dt_device *key, bool ref, bool unlink)
911 struct lfsck_instance *lfsck;
913 cfs_list_for_each_entry(lfsck, &lfsck_instance_list, li_link) {
914 if (lfsck->li_bottom == key) {
916 lfsck_instance_get(lfsck);
918 list_del_init(&lfsck->li_link);
927 struct lfsck_instance *lfsck_instance_find(struct dt_device *key, bool ref,
930 struct lfsck_instance *lfsck;
932 spin_lock(&lfsck_instance_lock);
933 lfsck = __lfsck_instance_find(key, ref, unlink);
934 spin_unlock(&lfsck_instance_lock);
939 static inline int lfsck_instance_add(struct lfsck_instance *lfsck)
941 struct lfsck_instance *tmp;
943 spin_lock(&lfsck_instance_lock);
944 cfs_list_for_each_entry(tmp, &lfsck_instance_list, li_link) {
945 if (lfsck->li_bottom == tmp->li_bottom) {
946 spin_unlock(&lfsck_instance_lock);
951 cfs_list_add_tail(&lfsck->li_link, &lfsck_instance_list);
952 spin_unlock(&lfsck_instance_lock);
956 int lfsck_bits_dump(struct seq_file *m, int bits, const char *names[],
961 bool newline = (bits != 0 ? false : true);
963 seq_printf(m, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
965 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
968 if (names[i] != NULL) {
972 seq_printf(m, "%s%c", names[i],
973 newline ? '\n' : ',');
983 int lfsck_time_dump(struct seq_file *m, __u64 time, const char *prefix)
986 seq_printf(m, "%s: "LPU64" seconds\n", prefix,
987 cfs_time_current_sec() - time);
989 seq_printf(m, "%s: N/A\n", prefix);
993 int lfsck_pos_dump(struct seq_file *m, struct lfsck_position *pos,
996 if (fid_is_zero(&pos->lp_dir_parent)) {
997 if (pos->lp_oit_cookie == 0)
998 seq_printf(m, "%s: N/A, N/A, N/A\n",
1001 seq_printf(m, "%s: "LPU64", N/A, N/A\n",
1002 prefix, pos->lp_oit_cookie);
1004 seq_printf(m, "%s: "LPU64", "DFID", "LPX64"\n",
1005 prefix, pos->lp_oit_cookie,
1006 PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
1011 void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
1012 struct lfsck_position *pos, bool init)
1014 const struct dt_it_ops *iops = &lfsck->li_obj_oit->do_index_ops->dio_it;
1016 if (unlikely(lfsck->li_di_oit == NULL)) {
1017 memset(pos, 0, sizeof(*pos));
1021 pos->lp_oit_cookie = iops->store(env, lfsck->li_di_oit);
1022 if (!lfsck->li_current_oit_processed && !init)
1023 pos->lp_oit_cookie--;
1025 LASSERT(pos->lp_oit_cookie > 0);
1027 if (lfsck->li_di_dir != NULL) {
1028 struct dt_object *dto = lfsck->li_obj_dir;
1030 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
1033 if (pos->lp_dir_cookie >= MDS_DIR_END_OFF) {
1034 fid_zero(&pos->lp_dir_parent);
1035 pos->lp_dir_cookie = 0;
1037 pos->lp_dir_parent = *lfsck_dto2fid(dto);
1040 fid_zero(&pos->lp_dir_parent);
1041 pos->lp_dir_cookie = 0;
1045 bool __lfsck_set_speed(struct lfsck_instance *lfsck, __u32 limit)
1049 if (limit != LFSCK_SPEED_NO_LIMIT) {
1051 lfsck->li_sleep_rate = limit / HZ;
1052 lfsck->li_sleep_jif = 1;
1054 lfsck->li_sleep_rate = 1;
1055 lfsck->li_sleep_jif = HZ / limit;
1058 lfsck->li_sleep_jif = 0;
1059 lfsck->li_sleep_rate = 0;
1062 if (lfsck->li_bookmark_ram.lb_speed_limit != limit) {
1063 lfsck->li_bookmark_ram.lb_speed_limit = limit;
1070 void lfsck_control_speed(struct lfsck_instance *lfsck)
1072 struct ptlrpc_thread *thread = &lfsck->li_thread;
1073 struct l_wait_info lwi;
1075 if (lfsck->li_sleep_jif > 0 &&
1076 lfsck->li_new_scanned >= lfsck->li_sleep_rate) {
1077 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1078 LWI_ON_SIGNAL_NOOP, NULL);
1080 l_wait_event(thread->t_ctl_waitq,
1081 !thread_is_running(thread),
1083 lfsck->li_new_scanned = 0;
1087 void lfsck_control_speed_by_self(struct lfsck_component *com)
1089 struct lfsck_instance *lfsck = com->lc_lfsck;
1090 struct ptlrpc_thread *thread = &lfsck->li_thread;
1091 struct l_wait_info lwi;
1093 if (lfsck->li_sleep_jif > 0 &&
1094 com->lc_new_scanned >= lfsck->li_sleep_rate) {
1095 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1096 LWI_ON_SIGNAL_NOOP, NULL);
1098 l_wait_event(thread->t_ctl_waitq,
1099 !thread_is_running(thread),
1101 com->lc_new_scanned = 0;
1105 static int lfsck_parent_fid(const struct lu_env *env, struct dt_object *obj,
1108 if (unlikely(!S_ISDIR(lfsck_object_type(obj)) ||
1109 !dt_try_as_dir(env, obj)))
1112 return dt_lookup(env, obj, (struct dt_rec *)fid,
1113 (const struct dt_key *)"..", BYPASS_CAPA);
1116 static int lfsck_needs_scan_dir(const struct lu_env *env,
1117 struct lfsck_instance *lfsck,
1118 struct dt_object *obj)
1120 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
1124 if (!lfsck->li_master || !S_ISDIR(lfsck_object_type(obj)) ||
1125 cfs_list_empty(&lfsck->li_list_dir))
1129 /* XXX: Currently, we do not scan the "/REMOTE_PARENT_DIR",
1130 * which is the agent directory to manage the objects
1131 * which name entries reside on remote MDTs. Related
1132 * consistency verification will be processed in LFSCK
1134 if (lu_fid_eq(lfsck_dto2fid(obj), &lfsck->li_global_root_fid)) {
1136 lfsck_object_put(env, obj);
1140 /* No need to check .lustre and its children. */
1141 if (fid_seq_is_dot_lustre(fid_seq(lfsck_dto2fid(obj)))) {
1143 lfsck_object_put(env, obj);
1147 dt_read_lock(env, obj, MOR_TGT_CHILD);
1148 if (unlikely(lfsck_is_dead_obj(obj))) {
1149 dt_read_unlock(env, obj);
1151 lfsck_object_put(env, obj);
1155 rc = dt_xattr_get(env, obj,
1156 lfsck_buf_get(env, NULL, 0), XATTR_NAME_LINK,
1158 dt_read_unlock(env, obj);
1161 lfsck_object_put(env, obj);
1165 if (rc < 0 && rc != -ENODATA) {
1167 lfsck_object_put(env, obj);
1171 rc = lfsck_parent_fid(env, obj, fid);
1173 lfsck_object_put(env, obj);
1177 if (unlikely(lu_fid_eq(fid, &lfsck->li_local_root_fid)))
1180 obj = lfsck_object_find(env, lfsck, fid);
1183 else if (IS_ERR(obj))
1184 return PTR_ERR(obj);
1186 if (!dt_object_exists(obj)) {
1187 lfsck_object_put(env, obj);
1191 if (dt_object_remote(obj)) {
1192 /* .lustre/lost+found/MDTxxx can be remote directory. */
1193 if (fid_seq_is_dot_lustre(fid_seq(lfsck_dto2fid(obj))))
1196 /* Other remote directory should be client
1197 * visible and need to be checked. */
1199 lfsck_object_put(env, obj);
1208 struct lfsck_thread_args *lfsck_thread_args_init(struct lfsck_instance *lfsck,
1209 struct lfsck_component *com,
1210 struct lfsck_start_param *lsp)
1212 struct lfsck_thread_args *lta;
1217 return ERR_PTR(-ENOMEM);
1219 rc = lu_env_init(<a->lta_env, LCT_MD_THREAD | LCT_DT_THREAD);
1225 lta->lta_lfsck = lfsck_instance_get(lfsck);
1227 lta->lta_com = lfsck_component_get(com);
1234 void lfsck_thread_args_fini(struct lfsck_thread_args *lta)
1236 if (lta->lta_com != NULL)
1237 lfsck_component_put(<a->lta_env, lta->lta_com);
1238 lfsck_instance_put(<a->lta_env, lta->lta_lfsck);
1239 lu_env_fini(<a->lta_env);
1243 /* LFSCK wrap functions */
1245 void lfsck_fail(const struct lu_env *env, struct lfsck_instance *lfsck,
1248 struct lfsck_component *com;
1250 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1251 com->lc_ops->lfsck_fail(env, com, new_checked);
1255 int lfsck_checkpoint(const struct lu_env *env, struct lfsck_instance *lfsck)
1257 struct lfsck_component *com;
1261 if (likely(cfs_time_beforeq(cfs_time_current(),
1262 lfsck->li_time_next_checkpoint)))
1265 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1266 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1267 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
1272 lfsck->li_time_last_checkpoint = cfs_time_current();
1273 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1274 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1275 return rc1 != 0 ? rc1 : rc;
1278 int lfsck_prep(const struct lu_env *env, struct lfsck_instance *lfsck,
1279 struct lfsck_start_param *lsp)
1281 struct dt_object *obj = NULL;
1282 struct lfsck_component *com;
1283 struct lfsck_component *next;
1284 struct lfsck_position *pos = NULL;
1285 const struct dt_it_ops *iops =
1286 &lfsck->li_obj_oit->do_index_ops->dio_it;
1291 LASSERT(lfsck->li_obj_dir == NULL);
1292 LASSERT(lfsck->li_di_dir == NULL);
1294 lfsck->li_current_oit_processed = 0;
1295 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1296 com->lc_new_checked = 0;
1297 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1298 com->lc_journal = 0;
1300 rc = com->lc_ops->lfsck_prep(env, com, lsp);
1304 if ((pos == NULL) ||
1305 (!lfsck_pos_is_zero(&com->lc_pos_start) &&
1306 lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
1307 pos = &com->lc_pos_start;
1310 /* Init otable-based iterator. */
1312 rc = iops->load(env, lfsck->li_di_oit, 0);
1314 lfsck->li_oit_over = 1;
1321 rc = iops->load(env, lfsck->li_di_oit, pos->lp_oit_cookie);
1325 lfsck->li_oit_over = 1;
1327 if (!lfsck->li_master || fid_is_zero(&pos->lp_dir_parent))
1330 /* Find the directory for namespace-based traverse. */
1331 obj = lfsck_object_find(env, lfsck, &pos->lp_dir_parent);
1334 else if (IS_ERR(obj))
1335 RETURN(PTR_ERR(obj));
1337 /* XXX: Currently, skip remote object, the consistency for
1338 * remote object will be processed in LFSCK phase III. */
1339 if (!dt_object_exists(obj) || dt_object_remote(obj) ||
1340 unlikely(!S_ISDIR(lfsck_object_type(obj))))
1343 if (unlikely(!dt_try_as_dir(env, obj)))
1344 GOTO(out, rc = -ENOTDIR);
1346 /* Init the namespace-based directory traverse. */
1347 iops = &obj->do_index_ops->dio_it;
1348 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1350 GOTO(out, rc = PTR_ERR(di));
1352 LASSERT(pos->lp_dir_cookie < MDS_DIR_END_OFF);
1354 rc = iops->load(env, di, pos->lp_dir_cookie);
1355 if ((rc == 0) || (rc > 0 && pos->lp_dir_cookie > 0))
1356 rc = iops->next(env, di);
1362 iops->fini(env, di);
1366 lfsck->li_obj_dir = lfsck_object_get(obj);
1367 lfsck->li_cookie_dir = iops->store(env, di);
1368 spin_lock(&lfsck->li_lock);
1369 lfsck->li_di_dir = di;
1370 spin_unlock(&lfsck->li_lock);
1376 lfsck_object_put(env, obj);
1379 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1381 com->lc_ops->lfsck_post(env, com, rc, true);
1387 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, true);
1388 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1389 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
1394 lfsck->li_time_last_checkpoint = cfs_time_current();
1395 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1396 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1400 int lfsck_exec_oit(const struct lu_env *env, struct lfsck_instance *lfsck,
1401 struct dt_object *obj)
1403 struct lfsck_component *com;
1404 const struct dt_it_ops *iops;
1409 LASSERT(lfsck->li_obj_dir == NULL);
1411 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1412 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
1417 rc = lfsck_needs_scan_dir(env, lfsck, obj);
1421 if (unlikely(!dt_try_as_dir(env, obj)))
1422 GOTO(out, rc = -ENOTDIR);
1424 iops = &obj->do_index_ops->dio_it;
1425 di = iops->init(env, obj, lfsck->li_args_dir, BYPASS_CAPA);
1427 GOTO(out, rc = PTR_ERR(di));
1429 rc = iops->load(env, di, 0);
1431 rc = iops->next(env, di);
1437 iops->fini(env, di);
1441 lfsck->li_obj_dir = lfsck_object_get(obj);
1442 lfsck->li_cookie_dir = iops->store(env, di);
1443 spin_lock(&lfsck->li_lock);
1444 lfsck->li_di_dir = di;
1445 spin_unlock(&lfsck->li_lock);
1451 lfsck_fail(env, lfsck, false);
1452 return (rc > 0 ? 0 : rc);
1455 int lfsck_exec_dir(const struct lu_env *env, struct lfsck_instance *lfsck,
1456 struct dt_object *obj, struct lu_dirent *ent)
1458 struct lfsck_component *com;
1461 cfs_list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1462 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
1469 int lfsck_post(const struct lu_env *env, struct lfsck_instance *lfsck,
1472 struct lfsck_component *com;
1473 struct lfsck_component *next;
1477 lfsck_pos_fill(env, lfsck, &lfsck->li_pos_current, false);
1478 cfs_list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
1479 rc = com->lc_ops->lfsck_post(env, com, result, false);
1484 lfsck->li_time_last_checkpoint = cfs_time_current();
1485 lfsck->li_time_next_checkpoint = lfsck->li_time_last_checkpoint +
1486 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1488 /* Ignore some component post failure to make other can go ahead. */
1492 static void lfsck_interpret(const struct lu_env *env,
1493 struct lfsck_instance *lfsck,
1494 struct ptlrpc_request *req, void *args, int result)
1496 struct lfsck_async_interpret_args *laia = args;
1497 struct lfsck_component *com;
1499 LASSERT(laia->laia_com == NULL);
1500 LASSERT(laia->laia_shared);
1502 spin_lock(&lfsck->li_lock);
1503 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1504 if (com->lc_ops->lfsck_interpret != NULL) {
1505 laia->laia_com = com;
1506 com->lc_ops->lfsck_interpret(env, req, laia, result);
1510 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1511 if (com->lc_ops->lfsck_interpret != NULL) {
1512 laia->laia_com = com;
1513 com->lc_ops->lfsck_interpret(env, req, laia, result);
1516 spin_unlock(&lfsck->li_lock);
1519 int lfsck_double_scan(const struct lu_env *env, struct lfsck_instance *lfsck)
1521 struct lfsck_component *com;
1522 struct lfsck_component *next;
1523 struct l_wait_info lwi = { 0 };
1527 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1528 if (lfsck->li_bookmark_ram.lb_param & LPF_DRYRUN)
1529 com->lc_journal = 0;
1531 rc = com->lc_ops->lfsck_double_scan(env, com);
1536 l_wait_event(lfsck->li_thread.t_ctl_waitq,
1537 atomic_read(&lfsck->li_double_scan_count) == 0,
1540 if (lfsck->li_status != LS_PAUSED &&
1541 lfsck->li_status != LS_CO_PAUSED) {
1542 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1544 spin_lock(&lfsck->li_lock);
1545 list_del_init(&com->lc_link);
1546 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1547 spin_unlock(&lfsck->li_lock);
1551 return rc1 != 0 ? rc1 : rc;
1554 static int lfsck_stop_notify(const struct lu_env *env,
1555 struct lfsck_instance *lfsck,
1556 struct lfsck_tgt_descs *ltds,
1557 struct lfsck_tgt_desc *ltd, __u16 type)
1559 struct ptlrpc_request_set *set;
1560 struct lfsck_component *com;
1564 spin_lock(&lfsck->li_lock);
1565 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
1567 com = __lfsck_component_find(lfsck, type,
1568 &lfsck->li_list_double_scan);
1570 lfsck_component_get(com);
1571 spin_unlock(&lfsck->li_lock);
1574 if (com->lc_ops->lfsck_stop_notify != NULL) {
1575 set = ptlrpc_prep_set();
1577 lfsck_component_put(env, com);
1582 rc = com->lc_ops->lfsck_stop_notify(env, com, ltds,
1585 rc = ptlrpc_set_wait(set);
1587 ptlrpc_set_destroy(set);
1590 lfsck_component_put(env, com);
1596 void lfsck_quit(const struct lu_env *env, struct lfsck_instance *lfsck)
1598 struct lfsck_component *com;
1599 struct lfsck_component *next;
1601 list_for_each_entry_safe(com, next, &lfsck->li_list_scan,
1603 if (com->lc_ops->lfsck_quit != NULL)
1604 com->lc_ops->lfsck_quit(env, com);
1606 spin_lock(&lfsck->li_lock);
1607 list_del_init(&com->lc_link);
1608 list_del_init(&com->lc_link_dir);
1609 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1610 spin_unlock(&lfsck->li_lock);
1613 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
1615 if (com->lc_ops->lfsck_quit != NULL)
1616 com->lc_ops->lfsck_quit(env, com);
1618 spin_lock(&lfsck->li_lock);
1619 list_del_init(&com->lc_link);
1620 list_add_tail(&com->lc_link, &lfsck->li_list_idle);
1621 spin_unlock(&lfsck->li_lock);
1625 static int lfsck_async_interpret(const struct lu_env *env,
1626 struct ptlrpc_request *req,
1629 struct lfsck_async_interpret_args *laia = args;
1630 struct lfsck_instance *lfsck;
1632 lfsck = container_of0(laia->laia_ltds, struct lfsck_instance,
1634 lfsck_interpret(env, lfsck, req, laia, rc);
1635 lfsck_tgt_put(laia->laia_ltd);
1636 if (rc != 0 && laia->laia_result != -EALREADY)
1637 laia->laia_result = rc;
1642 int lfsck_async_request(const struct lu_env *env, struct obd_export *exp,
1643 struct lfsck_request *lr,
1644 struct ptlrpc_request_set *set,
1645 ptlrpc_interpterer_t interpreter,
1646 void *args, int request)
1648 struct lfsck_async_interpret_args *laia;
1649 struct ptlrpc_request *req;
1650 struct lfsck_request *tmp;
1651 struct req_format *format;
1656 format = &RQF_LFSCK_NOTIFY;
1659 format = &RQF_LFSCK_QUERY;
1662 CDEBUG(D_LFSCK, "%s: unknown async request %d: rc = %d\n",
1663 exp->exp_obd->obd_name, request, -EINVAL);
1667 req = ptlrpc_request_alloc(class_exp2cliimp(exp), format);
1671 rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, request);
1673 ptlrpc_request_free(req);
1678 tmp = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
1680 ptlrpc_request_set_replen(req);
1682 laia = ptlrpc_req_async_args(req);
1683 *laia = *(struct lfsck_async_interpret_args *)args;
1684 if (laia->laia_com != NULL)
1685 lfsck_component_get(laia->laia_com);
1686 req->rq_interpret_reply = interpreter;
1687 ptlrpc_set_add_req(set, req);
1692 /* external interfaces */
1694 int lfsck_get_speed(struct seq_file *m, struct dt_device *key)
1697 struct lfsck_instance *lfsck;
1701 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1705 lfsck = lfsck_instance_find(key, true, false);
1706 if (likely(lfsck != NULL)) {
1707 seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_speed_limit);
1708 lfsck_instance_put(&env, lfsck);
1717 EXPORT_SYMBOL(lfsck_get_speed);
1719 int lfsck_set_speed(struct dt_device *key, int val)
1722 struct lfsck_instance *lfsck;
1726 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1730 lfsck = lfsck_instance_find(key, true, false);
1731 if (likely(lfsck != NULL)) {
1732 mutex_lock(&lfsck->li_mutex);
1733 if (__lfsck_set_speed(lfsck, val))
1734 rc = lfsck_bookmark_store(&env, lfsck);
1735 mutex_unlock(&lfsck->li_mutex);
1736 lfsck_instance_put(&env, lfsck);
1745 EXPORT_SYMBOL(lfsck_set_speed);
1747 int lfsck_get_windows(struct seq_file *m, struct dt_device *key)
1750 struct lfsck_instance *lfsck;
1754 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1758 lfsck = lfsck_instance_find(key, true, false);
1759 if (likely(lfsck != NULL)) {
1760 seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_async_windows);
1761 lfsck_instance_put(&env, lfsck);
1770 EXPORT_SYMBOL(lfsck_get_windows);
1772 int lfsck_set_windows(struct dt_device *key, int val)
1775 struct lfsck_instance *lfsck;
1779 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1783 lfsck = lfsck_instance_find(key, true, false);
1784 if (likely(lfsck != NULL)) {
1785 if (val > LFSCK_ASYNC_WIN_MAX) {
1786 CWARN("%s: Too large async window size, which "
1787 "may cause memory issues. The valid range "
1788 "is [0 - %u]. If you do not want to restrict "
1789 "the window size for async requests pipeline, "
1790 "just set it as 0.\n",
1791 lfsck_lfsck2name(lfsck), LFSCK_ASYNC_WIN_MAX);
1793 } else if (lfsck->li_bookmark_ram.lb_async_windows != val) {
1794 mutex_lock(&lfsck->li_mutex);
1795 lfsck->li_bookmark_ram.lb_async_windows = val;
1796 rc = lfsck_bookmark_store(&env, lfsck);
1797 mutex_unlock(&lfsck->li_mutex);
1799 lfsck_instance_put(&env, lfsck);
1808 EXPORT_SYMBOL(lfsck_set_windows);
1810 int lfsck_dump(struct seq_file *m, struct dt_device *key, enum lfsck_type type)
1813 struct lfsck_instance *lfsck;
1814 struct lfsck_component *com;
1818 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1822 lfsck = lfsck_instance_find(key, true, false);
1823 if (likely(lfsck != NULL)) {
1824 com = lfsck_component_find(lfsck, type);
1825 if (likely(com != NULL)) {
1826 rc = com->lc_ops->lfsck_dump(&env, com, m);
1827 lfsck_component_put(&env, com);
1832 lfsck_instance_put(&env, lfsck);
1841 EXPORT_SYMBOL(lfsck_dump);
1843 static int lfsck_stop_all(const struct lu_env *env,
1844 struct lfsck_instance *lfsck,
1845 struct lfsck_stop *stop)
1847 struct lfsck_thread_info *info = lfsck_env_info(env);
1848 struct lfsck_request *lr = &info->lti_lr;
1849 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1850 struct ptlrpc_request_set *set;
1851 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1852 struct lfsck_tgt_desc *ltd;
1853 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1859 LASSERT(stop->ls_flags & LPF_BROADCAST);
1861 set = ptlrpc_prep_set();
1862 if (unlikely(set == NULL))
1865 memset(lr, 0, sizeof(*lr));
1866 lr->lr_event = LE_STOP;
1867 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1868 lr->lr_status = stop->ls_status;
1869 lr->lr_version = bk->lb_version;
1870 lr->lr_active = LFSCK_TYPES_ALL;
1871 lr->lr_param = stop->ls_flags;
1873 laia->laia_com = NULL;
1874 laia->laia_ltds = ltds;
1876 laia->laia_result = 0;
1877 laia->laia_shared = 1;
1879 down_read(<ds->ltd_rw_sem);
1880 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1881 ltd = lfsck_tgt_get(ltds, idx);
1882 LASSERT(ltd != NULL);
1884 laia->laia_ltd = ltd;
1885 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1886 lfsck_async_interpret, laia,
1889 lfsck_interpret(env, lfsck, NULL, laia, rc);
1891 CERROR("%s: cannot notify MDT %x for LFSCK stop: "
1892 "rc = %d\n", lfsck_lfsck2name(lfsck), idx, rc);
1896 up_read(<ds->ltd_rw_sem);
1898 rc = ptlrpc_set_wait(set);
1899 ptlrpc_set_destroy(set);
1902 rc = laia->laia_result;
1904 if (rc == -EALREADY)
1908 CERROR("%s: fail to stop LFSCK on some MDTs: rc = %d\n",
1909 lfsck_lfsck2name(lfsck), rc);
1911 RETURN(rc != 0 ? rc : rc1);
1914 static int lfsck_start_all(const struct lu_env *env,
1915 struct lfsck_instance *lfsck,
1916 struct lfsck_start *start)
1918 struct lfsck_thread_info *info = lfsck_env_info(env);
1919 struct lfsck_request *lr = &info->lti_lr;
1920 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1921 struct ptlrpc_request_set *set;
1922 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1923 struct lfsck_tgt_desc *ltd;
1924 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1929 LASSERT(start->ls_flags & LPF_BROADCAST);
1931 set = ptlrpc_prep_set();
1932 if (unlikely(set == NULL))
1935 memset(lr, 0, sizeof(*lr));
1936 lr->lr_event = LE_START;
1937 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1938 lr->lr_speed = bk->lb_speed_limit;
1939 lr->lr_version = bk->lb_version;
1940 lr->lr_active = start->ls_active;
1941 lr->lr_param = start->ls_flags;
1942 lr->lr_async_windows = bk->lb_async_windows;
1943 lr->lr_valid = LSV_SPEED_LIMIT | LSV_ERROR_HANDLE | LSV_DRYRUN |
1946 laia->laia_com = NULL;
1947 laia->laia_ltds = ltds;
1949 laia->laia_result = 0;
1950 laia->laia_shared = 1;
1952 down_read(<ds->ltd_rw_sem);
1953 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1954 ltd = lfsck_tgt_get(ltds, idx);
1955 LASSERT(ltd != NULL);
1957 laia->laia_ltd = ltd;
1958 ltd->ltd_layout_done = 0;
1959 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1960 lfsck_async_interpret, laia,
1963 lfsck_interpret(env, lfsck, NULL, laia, rc);
1965 CERROR("%s: cannot notify MDT %x for LFSCK "
1966 "start, failout: rc = %d\n",
1967 lfsck_lfsck2name(lfsck), idx, rc);
1971 up_read(<ds->ltd_rw_sem);
1974 ptlrpc_set_destroy(set);
1979 rc = ptlrpc_set_wait(set);
1980 ptlrpc_set_destroy(set);
1983 rc = laia->laia_result;
1986 struct lfsck_stop *stop = &info->lti_stop;
1988 CERROR("%s: cannot start LFSCK on some MDTs, "
1989 "stop all: rc = %d\n",
1990 lfsck_lfsck2name(lfsck), rc);
1991 if (rc != -EALREADY) {
1992 stop->ls_status = LS_FAILED;
1993 stop->ls_flags = LPF_ALL_TGT | LPF_BROADCAST;
1994 lfsck_stop_all(env, lfsck, stop);
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;
2011 struct task_struct *task;
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 if (unlikely(start == NULL)) {
2039 spin_unlock(&lfsck->li_lock);
2043 while (start->ls_active != 0) {
2044 if (!(type & start->ls_active)) {
2049 com = __lfsck_component_find(lfsck, type,
2050 &lfsck->li_list_scan);
2052 com = __lfsck_component_find(lfsck, type,
2053 &lfsck->li_list_double_scan);
2059 if (com->lc_ops->lfsck_join != NULL) {
2060 rc = com->lc_ops->lfsck_join( env, com, lsp);
2061 if (rc != 0 && rc != -EALREADY)
2064 start->ls_active &= ~type;
2067 spin_unlock(&lfsck->li_lock);
2070 spin_unlock(&lfsck->li_lock);
2072 lfsck->li_status = 0;
2073 lfsck->li_oit_over = 0;
2074 lfsck->li_start_unplug = 0;
2075 lfsck->li_drop_dryrun = 0;
2076 lfsck->li_new_scanned = 0;
2078 /* For auto trigger. */
2082 if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
2083 CERROR("%s: only allow to specify '-A | -o' via MDS\n",
2084 lfsck_lfsck2name(lfsck));
2086 GOTO(out, rc = -EPERM);
2089 start->ls_version = bk->lb_version;
2091 if (start->ls_active != 0) {
2092 struct lfsck_component *next;
2094 if (start->ls_active == LFSCK_TYPES_ALL)
2095 start->ls_active = LFSCK_TYPES_SUPPORTED;
2097 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2098 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2099 GOTO(out, rc = -ENOTSUPP);
2102 list_for_each_entry_safe(com, next,
2103 &lfsck->li_list_scan, lc_link) {
2104 if (!(com->lc_type & start->ls_active)) {
2105 rc = com->lc_ops->lfsck_post(env, com, 0,
2112 while (start->ls_active != 0) {
2113 if (type & start->ls_active) {
2114 com = __lfsck_component_find(lfsck, type,
2115 &lfsck->li_list_idle);
2117 /* The component status will be updated
2118 * when its prep() is called later by
2119 * the LFSCK main engine. */
2120 list_del_init(&com->lc_link);
2121 list_add_tail(&com->lc_link,
2122 &lfsck->li_list_scan);
2124 start->ls_active &= ~type;
2130 if (list_empty(&lfsck->li_list_scan)) {
2131 /* The speed limit will be used to control both the LFSCK and
2132 * low layer scrub (if applied), need to be handled firstly. */
2133 if (start->ls_valid & LSV_SPEED_LIMIT) {
2134 if (__lfsck_set_speed(lfsck, start->ls_speed_limit)) {
2135 rc = lfsck_bookmark_store(env, lfsck);
2144 if (start->ls_flags & LPF_RESET)
2145 flags |= DOIF_RESET;
2147 rc = lfsck_set_param(env, lfsck, start, !!(flags & DOIF_RESET));
2151 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
2152 start->ls_active |= com->lc_type;
2153 if (flags & DOIF_RESET) {
2154 rc = com->lc_ops->lfsck_reset(env, com, false);
2161 lfsck->li_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2162 if (bk->lb_param & LPF_DRYRUN)
2163 lfsck->li_args_dir |= LUDA_VERIFY_DRYRUN;
2165 if (start != NULL && start->ls_valid & LSV_ERROR_HANDLE) {
2166 valid |= DOIV_ERROR_HANDLE;
2167 if (start->ls_flags & LPF_FAILOUT)
2168 flags |= DOIF_FAILOUT;
2171 if (start != NULL && start->ls_valid & LSV_DRYRUN) {
2172 valid |= DOIV_DRYRUN;
2173 if (start->ls_flags & LPF_DRYRUN)
2174 flags |= DOIF_DRYRUN;
2177 if (!list_empty(&lfsck->li_list_scan))
2178 flags |= DOIF_OUTUSED;
2180 lfsck->li_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2181 thread_set_flags(thread, 0);
2182 lta = lfsck_thread_args_init(lfsck, NULL, lsp);
2184 GOTO(out, rc = PTR_ERR(lta));
2186 __lfsck_set_speed(lfsck, bk->lb_speed_limit);
2187 task = kthread_run(lfsck_master_engine, lta, "lfsck");
2190 CERROR("%s: cannot start LFSCK thread: rc = %d\n",
2191 lfsck_lfsck2name(lfsck), rc);
2192 lfsck_thread_args_fini(lta);
2197 l_wait_event(thread->t_ctl_waitq,
2198 thread_is_running(thread) ||
2199 thread_is_stopped(thread),
2201 if (start == NULL || !(start->ls_flags & LPF_BROADCAST)) {
2202 lfsck->li_start_unplug = 1;
2203 wake_up_all(&thread->t_ctl_waitq);
2208 /* release lfsck::li_mutex to avoid deadlock. */
2209 mutex_unlock(&lfsck->li_mutex);
2210 rc = lfsck_start_all(env, lfsck, start);
2212 spin_lock(&lfsck->li_lock);
2213 if (thread_is_stopped(thread)) {
2214 spin_unlock(&lfsck->li_lock);
2216 lfsck->li_status = LS_FAILED;
2217 lfsck->li_flags = 0;
2218 thread_set_flags(thread, SVC_STOPPING);
2219 spin_unlock(&lfsck->li_lock);
2221 lfsck->li_start_unplug = 1;
2222 wake_up_all(&thread->t_ctl_waitq);
2223 l_wait_event(thread->t_ctl_waitq,
2224 thread_is_stopped(thread),
2228 lfsck->li_start_unplug = 1;
2229 wake_up_all(&thread->t_ctl_waitq);
2235 mutex_unlock(&lfsck->li_mutex);
2238 lfsck_instance_put(env, lfsck);
2240 return rc < 0 ? rc : 0;
2242 EXPORT_SYMBOL(lfsck_start);
2244 int lfsck_stop(const struct lu_env *env, struct dt_device *key,
2245 struct lfsck_stop *stop)
2247 struct lfsck_instance *lfsck;
2248 struct ptlrpc_thread *thread;
2249 struct l_wait_info lwi = { 0 };
2254 lfsck = lfsck_instance_find(key, true, false);
2255 if (unlikely(lfsck == NULL))
2258 thread = &lfsck->li_thread;
2259 /* release lfsck::li_mutex to avoid deadlock. */
2260 if (stop != NULL && stop->ls_flags & LPF_BROADCAST) {
2261 if (!lfsck->li_master) {
2262 CERROR("%s: only allow to specify '-A' via MDS\n",
2263 lfsck_lfsck2name(lfsck));
2265 GOTO(out, rc = -EPERM);
2268 rc1 = lfsck_stop_all(env, lfsck, stop);
2271 mutex_lock(&lfsck->li_mutex);
2272 spin_lock(&lfsck->li_lock);
2273 /* no error if LFSCK is already stopped, or was never started */
2274 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2275 spin_unlock(&lfsck->li_lock);
2280 lfsck->li_status = stop->ls_status;
2281 lfsck->li_flags = stop->ls_flags;
2283 lfsck->li_status = LS_STOPPED;
2284 lfsck->li_flags = 0;
2287 thread_set_flags(thread, SVC_STOPPING);
2288 spin_unlock(&lfsck->li_lock);
2290 wake_up_all(&thread->t_ctl_waitq);
2291 l_wait_event(thread->t_ctl_waitq,
2292 thread_is_stopped(thread),
2298 mutex_unlock(&lfsck->li_mutex);
2299 lfsck_instance_put(env, lfsck);
2301 return rc != 0 ? rc : rc1;
2303 EXPORT_SYMBOL(lfsck_stop);
2305 int lfsck_in_notify(const struct lu_env *env, struct dt_device *key,
2306 struct lfsck_request *lr)
2308 int rc = -EOPNOTSUPP;
2311 switch (lr->lr_event) {
2313 struct lfsck_start *start = &lfsck_env_info(env)->lti_start;
2314 struct lfsck_start_param lsp;
2316 memset(start, 0, sizeof(*start));
2317 start->ls_valid = lr->lr_valid;
2318 start->ls_speed_limit = lr->lr_speed;
2319 start->ls_version = lr->lr_version;
2320 start->ls_active = lr->lr_active;
2321 start->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2322 start->ls_async_windows = lr->lr_async_windows;
2324 lsp.lsp_start = start;
2325 lsp.lsp_index = lr->lr_index;
2326 lsp.lsp_index_valid = 1;
2327 rc = lfsck_start(env, key, &lsp);
2331 struct lfsck_stop *stop = &lfsck_env_info(env)->lti_stop;
2333 memset(stop, 0, sizeof(*stop));
2334 stop->ls_status = lr->lr_status;
2335 stop->ls_flags = lr->lr_param & ~LPF_BROADCAST;
2336 rc = lfsck_stop(env, key, stop);
2339 case LE_PHASE1_DONE:
2340 case LE_PHASE2_DONE:
2341 case LE_FID_ACCESSED:
2343 case LE_CONDITIONAL_DESTROY:
2344 case LE_PAIRS_VERIFY: {
2345 struct lfsck_instance *lfsck;
2346 struct lfsck_component *com;
2348 lfsck = lfsck_instance_find(key, true, false);
2349 if (unlikely(lfsck == NULL))
2352 com = lfsck_component_find(lfsck, lr->lr_active);
2353 if (likely(com != NULL)) {
2354 rc = com->lc_ops->lfsck_in_notify(env, com, lr);
2355 lfsck_component_put(env, com);
2358 lfsck_instance_put(env, lfsck);
2367 EXPORT_SYMBOL(lfsck_in_notify);
2369 int lfsck_query(const struct lu_env *env, struct dt_device *key,
2370 struct lfsck_request *lr)
2372 struct lfsck_instance *lfsck;
2373 struct lfsck_component *com;
2377 lfsck = lfsck_instance_find(key, true, false);
2378 if (unlikely(lfsck == NULL))
2381 com = lfsck_component_find(lfsck, lr->lr_active);
2382 if (likely(com != NULL)) {
2383 rc = com->lc_ops->lfsck_query(env, com);
2384 lfsck_component_put(env, com);
2389 lfsck_instance_put(env, lfsck);
2393 EXPORT_SYMBOL(lfsck_query);
2395 int lfsck_register_namespace(const struct lu_env *env, struct dt_device *key,
2396 struct ldlm_namespace *ns)
2398 struct lfsck_instance *lfsck;
2401 lfsck = lfsck_instance_find(key, true, false);
2402 if (likely(lfsck != NULL)) {
2403 lfsck->li_namespace = ns;
2404 lfsck_instance_put(env, lfsck);
2410 EXPORT_SYMBOL(lfsck_register_namespace);
2412 int lfsck_register(const struct lu_env *env, struct dt_device *key,
2413 struct dt_device *next, struct obd_device *obd,
2414 lfsck_out_notify notify, void *notify_data, bool master)
2416 struct lfsck_instance *lfsck;
2417 struct dt_object *root = NULL;
2418 struct dt_object *obj = NULL;
2419 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
2423 lfsck = lfsck_instance_find(key, false, false);
2424 if (unlikely(lfsck != NULL))
2427 OBD_ALLOC_PTR(lfsck);
2431 mutex_init(&lfsck->li_mutex);
2432 spin_lock_init(&lfsck->li_lock);
2433 CFS_INIT_LIST_HEAD(&lfsck->li_link);
2434 CFS_INIT_LIST_HEAD(&lfsck->li_list_scan);
2435 CFS_INIT_LIST_HEAD(&lfsck->li_list_dir);
2436 CFS_INIT_LIST_HEAD(&lfsck->li_list_double_scan);
2437 CFS_INIT_LIST_HEAD(&lfsck->li_list_idle);
2438 atomic_set(&lfsck->li_ref, 1);
2439 atomic_set(&lfsck->li_double_scan_count, 0);
2440 init_waitqueue_head(&lfsck->li_thread.t_ctl_waitq);
2441 lfsck->li_out_notify = notify;
2442 lfsck->li_out_notify_data = notify_data;
2443 lfsck->li_next = next;
2444 lfsck->li_bottom = key;
2445 lfsck->li_obd = obd;
2447 rc = lfsck_tgt_descs_init(&lfsck->li_ost_descs);
2451 rc = lfsck_tgt_descs_init(&lfsck->li_mdt_descs);
2455 fid->f_seq = FID_SEQ_LOCAL_NAME;
2458 rc = local_oid_storage_init(env, key, fid, &lfsck->li_los);
2462 rc = dt_root_get(env, key, fid);
2466 root = dt_locate(env, key, fid);
2468 GOTO(out, rc = PTR_ERR(root));
2470 if (unlikely(!dt_try_as_dir(env, root)))
2471 GOTO(out, rc = -ENOTDIR);
2473 lfsck->li_local_root_fid = *fid;
2475 lfsck->li_master = 1;
2476 if (lfsck_dev_idx(key) == 0) {
2477 struct lu_fid *pfid = &lfsck_env_info(env)->lti_fid2;
2478 const struct lu_name *cname;
2480 rc = dt_lookup(env, root,
2481 (struct dt_rec *)(&lfsck->li_global_root_fid),
2482 (const struct dt_key *)"ROOT", BYPASS_CAPA);
2486 obj = dt_locate(env, key, &lfsck->li_global_root_fid);
2488 GOTO(out, rc = PTR_ERR(obj));
2490 rc = dt_lookup(env, obj, (struct dt_rec *)fid,
2491 (const struct dt_key *)dotlustre, BYPASS_CAPA);
2495 lu_object_put(env, &obj->do_lu);
2496 obj = dt_locate(env, key, fid);
2498 GOTO(out, rc = PTR_ERR(obj));
2500 cname = lfsck_name_get_const(env, dotlustre,
2502 rc = lfsck_verify_linkea(env, key, obj, cname,
2503 &lfsck->li_global_root_fid);
2508 rc = dt_lookup(env, obj, (struct dt_rec *)fid,
2509 (const struct dt_key *)lostfound,
2514 lu_object_put(env, &obj->do_lu);
2515 obj = dt_locate(env, key, fid);
2517 GOTO(out, rc = PTR_ERR(obj));
2519 cname = lfsck_name_get_const(env, lostfound,
2521 rc = lfsck_verify_linkea(env, key, obj, cname, pfid);
2525 lu_object_put(env, &obj->do_lu);
2530 fid->f_seq = FID_SEQ_LOCAL_FILE;
2531 fid->f_oid = OTABLE_IT_OID;
2533 obj = dt_locate(env, key, fid);
2535 GOTO(out, rc = PTR_ERR(obj));
2537 lu_object_get(&obj->do_lu);
2538 lfsck->li_obj_oit = obj;
2539 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2543 rc = lfsck_bookmark_setup(env, lfsck);
2548 rc = lfsck_fid_init(lfsck);
2552 rc = lfsck_namespace_setup(env, lfsck);
2557 rc = lfsck_layout_setup(env, lfsck);
2561 /* XXX: more LFSCK components initialization to be added here. */
2563 rc = lfsck_instance_add(lfsck);
2565 rc = lfsck_add_target_from_orphan(env, lfsck);
2567 if (obj != NULL && !IS_ERR(obj))
2568 lu_object_put(env, &obj->do_lu);
2569 if (root != NULL && !IS_ERR(root))
2570 lu_object_put(env, &root->do_lu);
2572 lfsck_instance_cleanup(env, lfsck);
2575 EXPORT_SYMBOL(lfsck_register);
2577 void lfsck_degister(const struct lu_env *env, struct dt_device *key)
2579 struct lfsck_instance *lfsck;
2581 lfsck = lfsck_instance_find(key, false, true);
2583 lfsck_instance_put(env, lfsck);
2585 EXPORT_SYMBOL(lfsck_degister);
2587 int lfsck_add_target(const struct lu_env *env, struct dt_device *key,
2588 struct dt_device *tgt, struct obd_export *exp,
2589 __u32 index, bool for_ost)
2591 struct lfsck_instance *lfsck;
2592 struct lfsck_tgt_desc *ltd;
2603 INIT_LIST_HEAD(<d->ltd_orphan_list);
2604 INIT_LIST_HEAD(<d->ltd_layout_list);
2605 INIT_LIST_HEAD(<d->ltd_layout_phase_list);
2606 atomic_set(<d->ltd_ref, 1);
2607 ltd->ltd_index = index;
2609 spin_lock(&lfsck_instance_lock);
2610 lfsck = __lfsck_instance_find(key, true, false);
2611 if (lfsck == NULL) {
2613 list_add_tail(<d->ltd_orphan_list,
2614 &lfsck_ost_orphan_list);
2616 list_add_tail(<d->ltd_orphan_list,
2617 &lfsck_mdt_orphan_list);
2618 spin_unlock(&lfsck_instance_lock);
2622 spin_unlock(&lfsck_instance_lock);
2624 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, false);
2628 lfsck_instance_put(env, lfsck);
2632 EXPORT_SYMBOL(lfsck_add_target);
2634 void lfsck_del_target(const struct lu_env *env, struct dt_device *key,
2635 struct dt_device *tgt, __u32 index, bool for_ost)
2637 struct lfsck_instance *lfsck;
2638 struct lfsck_tgt_descs *ltds;
2639 struct lfsck_tgt_desc *ltd;
2640 struct list_head *head;
2643 head = &lfsck_ost_orphan_list;
2645 head = &lfsck_mdt_orphan_list;
2647 spin_lock(&lfsck_instance_lock);
2648 list_for_each_entry(ltd, head, ltd_orphan_list) {
2649 if (ltd->ltd_tgt == tgt) {
2650 list_del_init(<d->ltd_orphan_list);
2651 spin_unlock(&lfsck_instance_lock);
2659 lfsck = __lfsck_instance_find(key, true, false);
2660 spin_unlock(&lfsck_instance_lock);
2661 if (unlikely(lfsck == NULL))
2665 ltds = &lfsck->li_ost_descs;
2667 ltds = &lfsck->li_mdt_descs;
2669 down_write(<ds->ltd_rw_sem);
2670 LASSERT(ltds->ltd_tgts_bitmap != NULL);
2672 if (unlikely(index >= ltds->ltd_tgts_bitmap->size))
2675 ltd = LTD_TGT(ltds, index);
2676 if (unlikely(ltd == NULL))
2679 LASSERT(ltds->ltd_tgtnr > 0);
2682 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, index);
2683 LTD_TGT(ltds, index) = NULL;
2688 head = &lfsck->li_ost_descs.ltd_orphan;
2690 head = &lfsck->li_ost_descs.ltd_orphan;
2692 list_for_each_entry(ltd, head, ltd_orphan_list) {
2693 if (ltd->ltd_tgt == tgt) {
2694 list_del_init(<d->ltd_orphan_list);
2700 up_write(<ds->ltd_rw_sem);
2702 spin_lock(<ds->ltd_lock);
2704 spin_unlock(<ds->ltd_lock);
2705 lfsck_stop_notify(env, lfsck, ltds, ltd, LFSCK_TYPE_LAYOUT);
2709 lfsck_instance_put(env, lfsck);
2711 EXPORT_SYMBOL(lfsck_del_target);
2713 static int __init lfsck_init(void)
2717 INIT_LIST_HEAD(&lfsck_ost_orphan_list);
2718 INIT_LIST_HEAD(&lfsck_mdt_orphan_list);
2719 lfsck_key_init_generic(&lfsck_thread_key, NULL);
2720 rc = lu_context_key_register(&lfsck_thread_key);
2722 tgt_register_lfsck_in_notify(lfsck_in_notify);
2723 tgt_register_lfsck_query(lfsck_query);
2729 static void __exit lfsck_exit(void)
2731 struct lfsck_tgt_desc *ltd;
2732 struct lfsck_tgt_desc *next;
2734 LASSERT(cfs_list_empty(&lfsck_instance_list));
2736 list_for_each_entry_safe(ltd, next, &lfsck_ost_orphan_list,
2738 list_del_init(<d->ltd_orphan_list);
2742 list_for_each_entry_safe(ltd, next, &lfsck_mdt_orphan_list,
2744 list_del_init(<d->ltd_orphan_list);
2748 lu_context_key_degister(&lfsck_thread_key);
2751 MODULE_AUTHOR("Intel Corporation <http://www.intel.com/>");
2752 MODULE_DESCRIPTION("LFSCK");
2753 MODULE_LICENSE("GPL");
2755 cfs_module(lfsck, LUSTRE_VERSION_STRING, lfsck_init, lfsck_exit);