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 struct 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_move_tail(<d->ltd_orphan_list,
251 spin_unlock(&lfsck_instance_lock);
253 down_write(<ds->ltd_rw_sem);
254 while (!list_empty(<ds->ltd_orphan)) {
255 ltd = list_entry(ltds->ltd_orphan.next,
256 struct lfsck_tgt_desc,
258 list_del_init(<d->ltd_orphan_list);
259 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, true);
260 /* Do not hold the semaphore for too long time. */
261 up_write(<ds->ltd_rw_sem);
265 down_write(<ds->ltd_rw_sem);
267 up_write(<ds->ltd_rw_sem);
270 ltds = &lfsck->li_mdt_descs;
271 head = &lfsck_mdt_orphan_list;
279 static inline struct lfsck_component *
280 __lfsck_component_find(struct lfsck_instance *lfsck, __u16 type,
281 struct list_head *list)
283 struct lfsck_component *com;
285 list_for_each_entry(com, list, lc_link) {
286 if (com->lc_type == type)
292 struct lfsck_component *
293 lfsck_component_find(struct lfsck_instance *lfsck, __u16 type)
295 struct lfsck_component *com;
297 spin_lock(&lfsck->li_lock);
298 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
302 com = __lfsck_component_find(lfsck, type,
303 &lfsck->li_list_double_scan);
307 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_idle);
311 lfsck_component_get(com);
312 spin_unlock(&lfsck->li_lock);
316 void lfsck_component_cleanup(const struct lu_env *env,
317 struct lfsck_component *com)
319 if (!list_empty(&com->lc_link))
320 list_del_init(&com->lc_link);
321 if (!list_empty(&com->lc_link_dir))
322 list_del_init(&com->lc_link_dir);
324 lfsck_component_put(env, com);
327 int lfsck_fid_alloc(const struct lu_env *env, struct lfsck_instance *lfsck,
328 struct lu_fid *fid, bool locked)
330 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
335 mutex_lock(&lfsck->li_mutex);
337 rc = seq_client_alloc_fid(env, lfsck->li_seq, fid);
339 bk->lb_last_fid = *fid;
340 /* We do not care about whether the subsequent sub-operations
341 * failed or not. The worst case is that one FID is lost that
342 * is not a big issue for the LFSCK since it is relative rare
343 * for LFSCK create. */
344 rc = lfsck_bookmark_store(env, lfsck);
348 mutex_unlock(&lfsck->li_mutex);
354 * Request the specified ibits lock for the given object.
356 * Before the LFSCK modifying on the namespace visible object,
357 * it needs to acquire related ibits ldlm lock.
359 * \param[in] env pointer to the thread context
360 * \param[in] lfsck pointer to the lfsck instance
361 * \param[in] obj pointer to the dt_object to be locked
362 * \param[out] lh pointer to the lock handle
363 * \param[in] ibits the bits for the ldlm lock to be acquired
364 * \param[in] mode the mode for the ldlm lock to be acquired
366 * \retval 0 for success
367 * \retval negative error number on failure
369 int lfsck_ibits_lock(const struct lu_env *env, struct lfsck_instance *lfsck,
370 struct dt_object *obj, struct lustre_handle *lh,
371 __u64 bits, ldlm_mode_t mode)
373 struct lfsck_thread_info *info = lfsck_env_info(env);
374 ldlm_policy_data_t *policy = &info->lti_policy;
375 struct ldlm_res_id *resid = &info->lti_resid;
376 __u64 flags = LDLM_FL_ATOMIC_CB;
379 LASSERT(lfsck->li_namespace != NULL);
381 memset(policy, 0, sizeof(*policy));
382 policy->l_inodebits.bits = bits;
383 fid_build_reg_res_name(lfsck_dto2fid(obj), resid);
384 rc = ldlm_cli_enqueue_local(lfsck->li_namespace, resid, LDLM_IBITS,
385 policy, mode, &flags, ldlm_blocking_ast,
386 ldlm_completion_ast, NULL, NULL, 0,
387 LVB_T_NONE, NULL, lh);
388 if (rc == ELDLM_OK) {
391 memset(lh, 0, sizeof(*lh));
399 * Release the the specified ibits lock.
401 * If the lock has been acquired before, release it
402 * and cleanup the handle. Otherwise, do nothing.
404 * \param[in] lh pointer to the lock handle
405 * \param[in] mode the mode for the ldlm lock to be released
407 void lfsck_ibits_unlock(struct lustre_handle *lh, ldlm_mode_t mode)
409 if (lustre_handle_is_used(lh)) {
410 ldlm_lock_decref(lh, mode);
411 memset(lh, 0, sizeof(*lh));
415 static const char dot[] = ".";
416 static const char dotdot[] = "..";
417 static const char dotlustre[] = ".lustre";
418 static const char lostfound[] = "lost+found";
420 static int lfsck_create_lpf_local(const struct lu_env *env,
421 struct lfsck_instance *lfsck,
422 struct dt_object *parent,
423 struct dt_object *child,
425 struct dt_object_format *dof,
428 struct dt_insert_rec *rec = &lfsck_env_info(env)->lti_dt_rec;
429 struct dt_device *dev = lfsck->li_bottom;
430 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
431 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
432 const struct lu_fid *cfid = lfsck_dto2fid(child);
433 struct thandle *th = NULL;
434 struct linkea_data ldata = { 0 };
435 struct lu_buf linkea_buf;
436 const struct lu_name *cname;
438 int len = sizeof(struct lfsck_bookmark);
442 rc = linkea_data_new(&ldata,
443 &lfsck_env_info(env)->lti_linkea_buf);
447 cname = lfsck_name_get_const(env, name, strlen(name));
448 rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
452 th = dt_trans_create(env, dev);
456 /* 1a. create child */
457 rc = dt_declare_create(env, child, la, NULL, dof, th);
461 /* 2a. increase child nlink */
462 rc = dt_declare_ref_add(env, child, th);
466 /* 3a. insert linkEA for child */
467 lfsck_buf_init(&linkea_buf, ldata.ld_buf->lb_buf,
468 ldata.ld_leh->leh_len);
469 rc = dt_declare_xattr_set(env, child, &linkea_buf,
470 XATTR_NAME_LINK, 0, th);
474 /* 4a. insert name into parent dir */
475 rec->rec_type = S_IFDIR;
477 rc = dt_declare_insert(env, parent, (const struct dt_rec *)rec,
478 (const struct dt_key *)name, th);
482 /* 5a. increase parent nlink */
483 rc = dt_declare_ref_add(env, parent, th);
487 /* 6a. update bookmark */
488 rc = dt_declare_record_write(env, bk_obj,
489 lfsck_buf_get(env, bk, len), 0, th);
493 rc = dt_trans_start_local(env, dev, th);
497 dt_write_lock(env, child, 0);
498 /* 1b.1. create child */
499 rc = dt_create(env, child, la, NULL, dof, th);
503 if (unlikely(!dt_try_as_dir(env, child)))
504 GOTO(unlock, rc = -ENOTDIR);
506 /* 1b.2. insert dot into child dir */
508 rc = dt_insert(env, child, (const struct dt_rec *)rec,
509 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
513 /* 1b.3. insert dotdot into child dir */
514 rec->rec_fid = &LU_LPF_FID;
515 rc = dt_insert(env, child, (const struct dt_rec *)rec,
516 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
520 /* 2b. increase child nlink */
521 rc = dt_ref_add(env, child, th);
525 /* 3b. insert linkEA for child. */
526 rc = dt_xattr_set(env, child, &linkea_buf,
527 XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
528 dt_write_unlock(env, child);
532 /* 4b. insert name into parent dir */
534 rc = dt_insert(env, parent, (const struct dt_rec *)rec,
535 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
539 dt_write_lock(env, parent, 0);
540 /* 5b. increase parent nlink */
541 rc = dt_ref_add(env, parent, th);
542 dt_write_unlock(env, parent);
546 bk->lb_lpf_fid = *cfid;
547 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
549 /* 6b. update bookmark */
550 rc = dt_record_write(env, bk_obj,
551 lfsck_buf_get(env, bk, len), &pos, th);
556 dt_write_unlock(env, child);
559 dt_trans_stop(env, dev, th);
564 static int lfsck_create_lpf_remote(const struct lu_env *env,
565 struct lfsck_instance *lfsck,
566 struct dt_object *parent,
567 struct dt_object *child,
569 struct dt_object_format *dof,
572 struct dt_insert_rec *rec = &lfsck_env_info(env)->lti_dt_rec;
573 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
574 struct dt_object *bk_obj = lfsck->li_bookmark_obj;
575 const struct lu_fid *cfid = lfsck_dto2fid(child);
576 struct thandle *th = NULL;
577 struct linkea_data ldata = { 0 };
578 struct lu_buf linkea_buf;
579 const struct lu_name *cname;
580 struct dt_device *dev;
582 int len = sizeof(struct lfsck_bookmark);
586 rc = linkea_data_new(&ldata,
587 &lfsck_env_info(env)->lti_linkea_buf);
591 cname = lfsck_name_get_const(env, name, strlen(name));
592 rc = linkea_add_buf(&ldata, cname, lfsck_dto2fid(parent));
596 /* Create .lustre/lost+found/MDTxxxx. */
598 /* XXX: Currently, cross-MDT create operation needs to create the child
599 * object firstly, then insert name into the parent directory. For
600 * this case, the child object resides on current MDT (local), but
601 * the parent ".lustre/lost+found" may be on remote MDT. It is not
602 * easy to contain all the sub-modifications orderly within single
605 * To avoid more inconsistency, we split the create operation into
608 * 1) create the child and update the lfsck_bookmark::lb_lpf_fid
610 * 2) insert the name "MDTXXXX" in the parent ".lustre/lost+found"
613 * If 1) done, but 2) failed, then go ahead, the LFSCK will try to
614 * repair such inconsistency when LFSCK run next time. */
616 /* Transaction I: locally */
618 dev = lfsck->li_bottom;
619 th = dt_trans_create(env, dev);
623 /* 1a. create child */
624 rc = dt_declare_create(env, child, la, NULL, dof, th);
628 /* 2a. increase child nlink */
629 rc = dt_declare_ref_add(env, child, th);
633 /* 3a. insert linkEA for child */
634 lfsck_buf_init(&linkea_buf, ldata.ld_buf->lb_buf,
635 ldata.ld_leh->leh_len);
636 rc = dt_declare_xattr_set(env, child, &linkea_buf,
637 XATTR_NAME_LINK, 0, th);
641 /* 4a. update bookmark */
642 rc = dt_declare_record_write(env, bk_obj,
643 lfsck_buf_get(env, bk, len), 0, th);
647 rc = dt_trans_start_local(env, dev, th);
651 dt_write_lock(env, child, 0);
652 /* 1b.1. create child */
653 rc = dt_create(env, child, la, NULL, dof, th);
657 if (unlikely(!dt_try_as_dir(env, child)))
658 GOTO(unlock, rc = -ENOTDIR);
660 /* 1b.2. insert dot into child dir */
661 rec->rec_type = S_IFDIR;
663 rc = dt_insert(env, child, (const struct dt_rec *)rec,
664 (const struct dt_key *)dot, th, BYPASS_CAPA, 1);
668 /* 1b.3. insert dotdot into child dir */
669 rec->rec_fid = &LU_LPF_FID;
670 rc = dt_insert(env, child, (const struct dt_rec *)rec,
671 (const struct dt_key *)dotdot, th, BYPASS_CAPA, 1);
675 /* 2b. increase child nlink */
676 rc = dt_ref_add(env, child, th);
680 /* 3b. insert linkEA for child */
681 rc = dt_xattr_set(env, child, &linkea_buf,
682 XATTR_NAME_LINK, 0, th, BYPASS_CAPA);
686 bk->lb_lpf_fid = *cfid;
687 lfsck_bookmark_cpu_to_le(&lfsck->li_bookmark_disk, bk);
689 /* 4b. update bookmark */
690 rc = dt_record_write(env, bk_obj,
691 lfsck_buf_get(env, bk, len), &pos, th);
693 dt_write_unlock(env, child);
694 dt_trans_stop(env, dev, th);
698 /* Transaction II: remotely */
700 dev = lfsck->li_next;
701 th = dt_trans_create(env, dev);
705 /* 5a. insert name into parent dir */
707 rc = dt_declare_insert(env, parent, (const struct dt_rec *)rec,
708 (const struct dt_key *)name, th);
712 /* 6a. increase parent nlink */
713 rc = dt_declare_ref_add(env, parent, th);
717 rc = dt_trans_start(env, dev, th);
721 /* 5b. insert name into parent dir */
722 rc = dt_insert(env, parent, (const struct dt_rec *)rec,
723 (const struct dt_key *)name, th, BYPASS_CAPA, 1);
727 dt_write_lock(env, parent, 0);
728 /* 6b. increase parent nlink */
729 rc = dt_ref_add(env, parent, th);
730 dt_write_unlock(env, parent);
735 dt_write_unlock(env, child);
737 dt_trans_stop(env, dev, th);
739 if (rc != 0 && dev == lfsck->li_next)
740 CDEBUG(D_LFSCK, "%s: partially created the object "DFID
741 "for orphans, but failed to insert the name %s "
742 "to the .lustre/lost+found/. Such inconsistency "
743 "will be repaired when LFSCK run next time: rc = %d\n",
744 lfsck_lfsck2name(lfsck), PFID(cfid), name, rc);
749 /* Do NOT create .lustre/lost+found/MDTxxxx when register the lfsck instance,
750 * because the MDT0 maybe not reaady for sequence allocation yet. We do that
751 * only when it is required, such as orphan OST-objects repairing. */
752 int lfsck_create_lpf(const struct lu_env *env, struct lfsck_instance *lfsck)
754 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
755 struct lfsck_thread_info *info = lfsck_env_info(env);
756 struct lu_fid *cfid = &info->lti_fid2;
757 struct lu_attr *la = &info->lti_la;
758 struct dt_object_format *dof = &info->lti_dof;
759 struct dt_object *parent = NULL;
760 struct dt_object *child = NULL;
761 struct lustre_handle lh = { 0 };
763 int node = lfsck_dev_idx(lfsck->li_bottom);
767 LASSERT(lfsck->li_master);
769 sprintf(name, "MDT%04x", node);
771 parent = lfsck_object_find_by_dev(env, lfsck->li_bottom,
774 struct lfsck_tgt_desc *ltd;
776 ltd = lfsck_tgt_get(&lfsck->li_mdt_descs, 0);
777 if (unlikely(ltd == NULL))
780 parent = lfsck_object_find_by_dev(env, ltd->ltd_tgt,
785 RETURN(PTR_ERR(parent));
787 if (lfsck->li_lpf_obj != NULL)
790 if (unlikely(!dt_try_as_dir(env, parent)))
791 GOTO(out, rc = -ENOTDIR);
793 rc = lfsck_ibits_lock(env, lfsck, parent, &lh,
794 MDS_INODELOCK_UPDATE, LCK_EX);
798 mutex_lock(&lfsck->li_mutex);
799 if (lfsck->li_lpf_obj != NULL)
800 GOTO(unlock, rc = 0);
802 if (fid_is_zero(&bk->lb_lpf_fid)) {
803 /* There is corner case that: in former LFSCK scanning we have
804 * created the .lustre/lost+found/MDTxxxx but failed to update
805 * the lfsck_bookmark::lb_lpf_fid successfully. So need lookup
806 * it from MDT0 firstly. */
807 rc = dt_lookup(env, parent, (struct dt_rec *)cfid,
808 (const struct dt_key *)name, BYPASS_CAPA);
809 if (rc != 0 && rc != -ENOENT)
813 bk->lb_lpf_fid = *cfid;
814 rc = lfsck_bookmark_store(env, lfsck);
816 rc = lfsck_fid_alloc(env, lfsck, cfid, true);
821 *cfid = bk->lb_lpf_fid;
824 child = lfsck_object_find_by_dev(env, lfsck->li_bottom, cfid);
826 GOTO(unlock, rc = PTR_ERR(child));
828 if (dt_object_exists(child) != 0) {
829 if (unlikely(!dt_try_as_dir(env, child)))
832 lfsck->li_lpf_obj = child;
837 memset(la, 0, sizeof(*la));
838 la->la_atime = la->la_mtime = la->la_ctime = cfs_time_current_sec();
839 la->la_mode = S_IFDIR | S_IRWXU;
840 la->la_valid = LA_ATIME | LA_MTIME | LA_CTIME | LA_MODE |
842 memset(dof, 0, sizeof(*dof));
843 dof->dof_type = dt_mode_to_dft(S_IFDIR);
846 rc = lfsck_create_lpf_local(env, lfsck, parent, child, la,
849 rc = lfsck_create_lpf_remote(env, lfsck, parent, child, la,
852 lfsck->li_lpf_obj = child;
857 mutex_unlock(&lfsck->li_mutex);
858 lfsck_ibits_unlock(&lh, LCK_EX);
859 if (rc != 0 && child != NULL && !IS_ERR(child))
860 lu_object_put(env, &child->do_lu);
862 if (parent != NULL && !IS_ERR(parent))
863 lu_object_put(env, &parent->do_lu);
868 static int lfsck_fid_init(struct lfsck_instance *lfsck)
870 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
871 struct seq_server_site *ss;
876 ss = lu_site2seq(lfsck->li_bottom->dd_lu_dev.ld_site);
877 if (unlikely(ss == NULL))
880 OBD_ALLOC_PTR(lfsck->li_seq);
881 if (lfsck->li_seq == NULL)
884 OBD_ALLOC(prefix, MAX_OBD_NAME + 7);
886 GOTO(out, rc = -ENOMEM);
888 snprintf(prefix, MAX_OBD_NAME + 7, "lfsck-%s", lfsck_lfsck2name(lfsck));
889 rc = seq_client_init(lfsck->li_seq, NULL, LUSTRE_SEQ_METADATA, prefix,
891 OBD_FREE(prefix, MAX_OBD_NAME + 7);
895 if (fid_is_sane(&bk->lb_last_fid))
896 lfsck->li_seq->lcs_fid = bk->lb_last_fid;
901 OBD_FREE_PTR(lfsck->li_seq);
902 lfsck->li_seq = NULL;
907 static void lfsck_fid_fini(struct lfsck_instance *lfsck)
909 if (lfsck->li_seq != NULL) {
910 seq_client_fini(lfsck->li_seq);
911 OBD_FREE_PTR(lfsck->li_seq);
912 lfsck->li_seq = NULL;
916 void lfsck_instance_cleanup(const struct lu_env *env,
917 struct lfsck_instance *lfsck)
919 struct ptlrpc_thread *thread = &lfsck->li_thread;
920 struct lfsck_component *com;
921 struct lfsck_component *next;
924 LASSERT(list_empty(&lfsck->li_link));
925 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
927 if (lfsck->li_obj_oit != NULL) {
928 lu_object_put_nocache(env, &lfsck->li_obj_oit->do_lu);
929 lfsck->li_obj_oit = NULL;
932 LASSERT(lfsck->li_obj_dir == NULL);
934 list_for_each_entry_safe(com, next, &lfsck->li_list_scan, lc_link) {
935 lfsck_component_cleanup(env, com);
938 LASSERT(list_empty(&lfsck->li_list_dir));
940 list_for_each_entry_safe(com, next, &lfsck->li_list_double_scan,
942 lfsck_component_cleanup(env, com);
945 list_for_each_entry_safe(com, next, &lfsck->li_list_idle, lc_link) {
946 lfsck_component_cleanup(env, com);
949 lfsck_tgt_descs_fini(&lfsck->li_ost_descs);
950 lfsck_tgt_descs_fini(&lfsck->li_mdt_descs);
952 if (lfsck->li_bookmark_obj != NULL) {
953 lu_object_put_nocache(env, &lfsck->li_bookmark_obj->do_lu);
954 lfsck->li_bookmark_obj = NULL;
957 if (lfsck->li_lpf_obj != NULL) {
958 lu_object_put(env, &lfsck->li_lpf_obj->do_lu);
959 lfsck->li_lpf_obj = NULL;
962 if (lfsck->li_los != NULL) {
963 local_oid_storage_fini(env, lfsck->li_los);
964 lfsck->li_los = NULL;
967 lfsck_fid_fini(lfsck);
972 static inline struct lfsck_instance *
973 __lfsck_instance_find(struct dt_device *key, bool ref, bool unlink)
975 struct lfsck_instance *lfsck;
977 list_for_each_entry(lfsck, &lfsck_instance_list, li_link) {
978 if (lfsck->li_bottom == key) {
980 lfsck_instance_get(lfsck);
982 list_del_init(&lfsck->li_link);
991 struct lfsck_instance *lfsck_instance_find(struct dt_device *key, bool ref,
994 struct lfsck_instance *lfsck;
996 spin_lock(&lfsck_instance_lock);
997 lfsck = __lfsck_instance_find(key, ref, unlink);
998 spin_unlock(&lfsck_instance_lock);
1003 static inline int lfsck_instance_add(struct lfsck_instance *lfsck)
1005 struct lfsck_instance *tmp;
1007 spin_lock(&lfsck_instance_lock);
1008 list_for_each_entry(tmp, &lfsck_instance_list, li_link) {
1009 if (lfsck->li_bottom == tmp->li_bottom) {
1010 spin_unlock(&lfsck_instance_lock);
1015 list_add_tail(&lfsck->li_link, &lfsck_instance_list);
1016 spin_unlock(&lfsck_instance_lock);
1020 int lfsck_bits_dump(struct seq_file *m, int bits, const char *names[],
1025 bool newline = (bits != 0 ? false : true);
1027 seq_printf(m, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
1029 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
1032 if (names[i] != NULL) {
1036 seq_printf(m, "%s%c", names[i],
1037 newline ? '\n' : ',');
1043 seq_printf(m, "\n");
1047 int lfsck_time_dump(struct seq_file *m, __u64 time, const char *prefix)
1050 seq_printf(m, "%s: "LPU64" seconds\n", prefix,
1051 cfs_time_current_sec() - time);
1053 seq_printf(m, "%s: N/A\n", prefix);
1057 int lfsck_pos_dump(struct seq_file *m, struct lfsck_position *pos,
1060 if (fid_is_zero(&pos->lp_dir_parent)) {
1061 if (pos->lp_oit_cookie == 0)
1062 seq_printf(m, "%s: N/A, N/A, N/A\n",
1065 seq_printf(m, "%s: "LPU64", N/A, N/A\n",
1066 prefix, pos->lp_oit_cookie);
1068 seq_printf(m, "%s: "LPU64", "DFID", "LPX64"\n",
1069 prefix, pos->lp_oit_cookie,
1070 PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
1075 void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
1076 struct lfsck_position *pos, bool init)
1078 const struct dt_it_ops *iops = &lfsck->li_obj_oit->do_index_ops->dio_it;
1080 if (unlikely(lfsck->li_di_oit == NULL)) {
1081 memset(pos, 0, sizeof(*pos));
1085 pos->lp_oit_cookie = iops->store(env, lfsck->li_di_oit);
1086 if (!lfsck->li_current_oit_processed && !init)
1087 pos->lp_oit_cookie--;
1089 LASSERT(pos->lp_oit_cookie > 0);
1091 if (lfsck->li_di_dir != NULL) {
1092 struct dt_object *dto = lfsck->li_obj_dir;
1094 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
1097 if (pos->lp_dir_cookie >= MDS_DIR_END_OFF) {
1098 fid_zero(&pos->lp_dir_parent);
1099 pos->lp_dir_cookie = 0;
1101 pos->lp_dir_parent = *lfsck_dto2fid(dto);
1104 fid_zero(&pos->lp_dir_parent);
1105 pos->lp_dir_cookie = 0;
1109 bool __lfsck_set_speed(struct lfsck_instance *lfsck, __u32 limit)
1113 if (limit != LFSCK_SPEED_NO_LIMIT) {
1115 lfsck->li_sleep_rate = limit / HZ;
1116 lfsck->li_sleep_jif = 1;
1118 lfsck->li_sleep_rate = 1;
1119 lfsck->li_sleep_jif = HZ / limit;
1122 lfsck->li_sleep_jif = 0;
1123 lfsck->li_sleep_rate = 0;
1126 if (lfsck->li_bookmark_ram.lb_speed_limit != limit) {
1127 lfsck->li_bookmark_ram.lb_speed_limit = limit;
1134 void lfsck_control_speed(struct lfsck_instance *lfsck)
1136 struct ptlrpc_thread *thread = &lfsck->li_thread;
1137 struct l_wait_info lwi;
1139 if (lfsck->li_sleep_jif > 0 &&
1140 lfsck->li_new_scanned >= lfsck->li_sleep_rate) {
1141 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1142 LWI_ON_SIGNAL_NOOP, NULL);
1144 l_wait_event(thread->t_ctl_waitq,
1145 !thread_is_running(thread),
1147 lfsck->li_new_scanned = 0;
1151 void lfsck_control_speed_by_self(struct lfsck_component *com)
1153 struct lfsck_instance *lfsck = com->lc_lfsck;
1154 struct ptlrpc_thread *thread = &lfsck->li_thread;
1155 struct l_wait_info lwi;
1157 if (lfsck->li_sleep_jif > 0 &&
1158 com->lc_new_scanned >= lfsck->li_sleep_rate) {
1159 lwi = LWI_TIMEOUT_INTR(lfsck->li_sleep_jif, NULL,
1160 LWI_ON_SIGNAL_NOOP, NULL);
1162 l_wait_event(thread->t_ctl_waitq,
1163 !thread_is_running(thread),
1165 com->lc_new_scanned = 0;
1169 struct lfsck_thread_args *lfsck_thread_args_init(struct lfsck_instance *lfsck,
1170 struct lfsck_component *com,
1171 struct lfsck_start_param *lsp)
1173 struct lfsck_thread_args *lta;
1178 return ERR_PTR(-ENOMEM);
1180 rc = lu_env_init(<a->lta_env, LCT_MD_THREAD | LCT_DT_THREAD);
1186 lta->lta_lfsck = lfsck_instance_get(lfsck);
1188 lta->lta_com = lfsck_component_get(com);
1195 void lfsck_thread_args_fini(struct lfsck_thread_args *lta)
1197 if (lta->lta_com != NULL)
1198 lfsck_component_put(<a->lta_env, lta->lta_com);
1199 lfsck_instance_put(<a->lta_env, lta->lta_lfsck);
1200 lu_env_fini(<a->lta_env);
1204 static void lfsck_interpret(const struct lu_env *env,
1205 struct lfsck_instance *lfsck,
1206 struct ptlrpc_request *req, void *args, int result)
1208 struct lfsck_async_interpret_args *laia = args;
1209 struct lfsck_component *com;
1211 LASSERT(laia->laia_com == NULL);
1212 LASSERT(laia->laia_shared);
1214 spin_lock(&lfsck->li_lock);
1215 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1216 if (com->lc_ops->lfsck_interpret != NULL) {
1217 laia->laia_com = com;
1218 com->lc_ops->lfsck_interpret(env, req, laia, result);
1222 list_for_each_entry(com, &lfsck->li_list_double_scan, lc_link) {
1223 if (com->lc_ops->lfsck_interpret != NULL) {
1224 laia->laia_com = com;
1225 com->lc_ops->lfsck_interpret(env, req, laia, result);
1228 spin_unlock(&lfsck->li_lock);
1231 static int lfsck_stop_notify(const struct lu_env *env,
1232 struct lfsck_instance *lfsck,
1233 struct lfsck_tgt_descs *ltds,
1234 struct lfsck_tgt_desc *ltd, __u16 type)
1236 struct ptlrpc_request_set *set;
1237 struct lfsck_component *com;
1241 spin_lock(&lfsck->li_lock);
1242 com = __lfsck_component_find(lfsck, type, &lfsck->li_list_scan);
1244 com = __lfsck_component_find(lfsck, type,
1245 &lfsck->li_list_double_scan);
1247 lfsck_component_get(com);
1248 spin_unlock(&lfsck->li_lock);
1251 if (com->lc_ops->lfsck_stop_notify != NULL) {
1252 set = ptlrpc_prep_set();
1254 lfsck_component_put(env, com);
1259 rc = com->lc_ops->lfsck_stop_notify(env, com, ltds,
1262 rc = ptlrpc_set_wait(set);
1264 ptlrpc_set_destroy(set);
1267 lfsck_component_put(env, com);
1273 static int lfsck_async_interpret(const struct lu_env *env,
1274 struct ptlrpc_request *req,
1277 struct lfsck_async_interpret_args *laia = args;
1278 struct lfsck_instance *lfsck;
1280 lfsck = container_of0(laia->laia_ltds, struct lfsck_instance,
1282 lfsck_interpret(env, lfsck, req, laia, rc);
1283 lfsck_tgt_put(laia->laia_ltd);
1284 if (rc != 0 && laia->laia_result != -EALREADY)
1285 laia->laia_result = rc;
1290 int lfsck_async_request(const struct lu_env *env, struct obd_export *exp,
1291 struct lfsck_request *lr,
1292 struct ptlrpc_request_set *set,
1293 ptlrpc_interpterer_t interpreter,
1294 void *args, int request)
1296 struct lfsck_async_interpret_args *laia;
1297 struct ptlrpc_request *req;
1298 struct lfsck_request *tmp;
1299 struct req_format *format;
1304 format = &RQF_LFSCK_NOTIFY;
1307 format = &RQF_LFSCK_QUERY;
1310 CDEBUG(D_LFSCK, "%s: unknown async request %d: rc = %d\n",
1311 exp->exp_obd->obd_name, request, -EINVAL);
1315 req = ptlrpc_request_alloc(class_exp2cliimp(exp), format);
1319 rc = ptlrpc_request_pack(req, LUSTRE_OBD_VERSION, request);
1321 ptlrpc_request_free(req);
1326 tmp = req_capsule_client_get(&req->rq_pill, &RMF_LFSCK_REQUEST);
1328 ptlrpc_request_set_replen(req);
1330 laia = ptlrpc_req_async_args(req);
1331 *laia = *(struct lfsck_async_interpret_args *)args;
1332 if (laia->laia_com != NULL)
1333 lfsck_component_get(laia->laia_com);
1334 req->rq_interpret_reply = interpreter;
1335 ptlrpc_set_add_req(set, req);
1340 /* external interfaces */
1342 int lfsck_get_speed(struct seq_file *m, struct dt_device *key)
1345 struct lfsck_instance *lfsck;
1349 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1353 lfsck = lfsck_instance_find(key, true, false);
1354 if (likely(lfsck != NULL)) {
1355 seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_speed_limit);
1356 lfsck_instance_put(&env, lfsck);
1365 EXPORT_SYMBOL(lfsck_get_speed);
1367 int lfsck_set_speed(struct dt_device *key, int val)
1370 struct lfsck_instance *lfsck;
1374 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1378 lfsck = lfsck_instance_find(key, true, false);
1379 if (likely(lfsck != NULL)) {
1380 mutex_lock(&lfsck->li_mutex);
1381 if (__lfsck_set_speed(lfsck, val))
1382 rc = lfsck_bookmark_store(&env, lfsck);
1383 mutex_unlock(&lfsck->li_mutex);
1384 lfsck_instance_put(&env, lfsck);
1393 EXPORT_SYMBOL(lfsck_set_speed);
1395 int lfsck_get_windows(struct seq_file *m, struct dt_device *key)
1398 struct lfsck_instance *lfsck;
1402 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1406 lfsck = lfsck_instance_find(key, true, false);
1407 if (likely(lfsck != NULL)) {
1408 seq_printf(m, "%u\n", lfsck->li_bookmark_ram.lb_async_windows);
1409 lfsck_instance_put(&env, lfsck);
1418 EXPORT_SYMBOL(lfsck_get_windows);
1420 int lfsck_set_windows(struct dt_device *key, int val)
1423 struct lfsck_instance *lfsck;
1427 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1431 lfsck = lfsck_instance_find(key, true, false);
1432 if (likely(lfsck != NULL)) {
1433 if (val > LFSCK_ASYNC_WIN_MAX) {
1434 CWARN("%s: Too large async window size, which "
1435 "may cause memory issues. The valid range "
1436 "is [0 - %u]. If you do not want to restrict "
1437 "the window size for async requests pipeline, "
1438 "just set it as 0.\n",
1439 lfsck_lfsck2name(lfsck), LFSCK_ASYNC_WIN_MAX);
1441 } else if (lfsck->li_bookmark_ram.lb_async_windows != val) {
1442 mutex_lock(&lfsck->li_mutex);
1443 lfsck->li_bookmark_ram.lb_async_windows = val;
1444 rc = lfsck_bookmark_store(&env, lfsck);
1445 mutex_unlock(&lfsck->li_mutex);
1447 lfsck_instance_put(&env, lfsck);
1456 EXPORT_SYMBOL(lfsck_set_windows);
1458 int lfsck_dump(struct seq_file *m, struct dt_device *key, enum lfsck_type type)
1461 struct lfsck_instance *lfsck;
1462 struct lfsck_component *com;
1466 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
1470 lfsck = lfsck_instance_find(key, true, false);
1471 if (likely(lfsck != NULL)) {
1472 com = lfsck_component_find(lfsck, type);
1473 if (likely(com != NULL)) {
1474 rc = com->lc_ops->lfsck_dump(&env, com, m);
1475 lfsck_component_put(&env, com);
1480 lfsck_instance_put(&env, lfsck);
1489 EXPORT_SYMBOL(lfsck_dump);
1491 static int lfsck_stop_all(const struct lu_env *env,
1492 struct lfsck_instance *lfsck,
1493 struct lfsck_stop *stop)
1495 struct lfsck_thread_info *info = lfsck_env_info(env);
1496 struct lfsck_request *lr = &info->lti_lr;
1497 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1498 struct ptlrpc_request_set *set;
1499 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1500 struct lfsck_tgt_desc *ltd;
1501 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1507 LASSERT(stop->ls_flags & LPF_BROADCAST);
1509 set = ptlrpc_prep_set();
1510 if (unlikely(set == NULL))
1513 memset(lr, 0, sizeof(*lr));
1514 lr->lr_event = LE_STOP;
1515 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1516 lr->lr_status = stop->ls_status;
1517 lr->lr_version = bk->lb_version;
1518 lr->lr_active = LFSCK_TYPES_ALL;
1519 lr->lr_param = stop->ls_flags;
1521 laia->laia_com = NULL;
1522 laia->laia_ltds = ltds;
1524 laia->laia_result = 0;
1525 laia->laia_shared = 1;
1527 down_read(<ds->ltd_rw_sem);
1528 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1529 ltd = lfsck_tgt_get(ltds, idx);
1530 LASSERT(ltd != NULL);
1532 laia->laia_ltd = ltd;
1533 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1534 lfsck_async_interpret, laia,
1537 lfsck_interpret(env, lfsck, NULL, laia, rc);
1539 CERROR("%s: cannot notify MDT %x for LFSCK stop: "
1540 "rc = %d\n", lfsck_lfsck2name(lfsck), idx, rc);
1544 up_read(<ds->ltd_rw_sem);
1546 rc = ptlrpc_set_wait(set);
1547 ptlrpc_set_destroy(set);
1550 rc = laia->laia_result;
1552 if (rc == -EALREADY)
1556 CERROR("%s: fail to stop LFSCK on some MDTs: rc = %d\n",
1557 lfsck_lfsck2name(lfsck), rc);
1559 RETURN(rc != 0 ? rc : rc1);
1562 static int lfsck_start_all(const struct lu_env *env,
1563 struct lfsck_instance *lfsck,
1564 struct lfsck_start *start)
1566 struct lfsck_thread_info *info = lfsck_env_info(env);
1567 struct lfsck_request *lr = &info->lti_lr;
1568 struct lfsck_async_interpret_args *laia = &info->lti_laia;
1569 struct ptlrpc_request_set *set;
1570 struct lfsck_tgt_descs *ltds = &lfsck->li_mdt_descs;
1571 struct lfsck_tgt_desc *ltd;
1572 struct lfsck_bookmark *bk = &lfsck->li_bookmark_ram;
1577 LASSERT(start->ls_flags & LPF_BROADCAST);
1579 set = ptlrpc_prep_set();
1580 if (unlikely(set == NULL))
1583 memset(lr, 0, sizeof(*lr));
1584 lr->lr_event = LE_START;
1585 lr->lr_index = lfsck_dev_idx(lfsck->li_bottom);
1586 lr->lr_speed = bk->lb_speed_limit;
1587 lr->lr_version = bk->lb_version;
1588 lr->lr_active = start->ls_active;
1589 lr->lr_param = start->ls_flags;
1590 lr->lr_async_windows = bk->lb_async_windows;
1591 lr->lr_valid = LSV_SPEED_LIMIT | LSV_ERROR_HANDLE | LSV_DRYRUN |
1594 laia->laia_com = NULL;
1595 laia->laia_ltds = ltds;
1597 laia->laia_result = 0;
1598 laia->laia_shared = 1;
1600 down_read(<ds->ltd_rw_sem);
1601 cfs_foreach_bit(ltds->ltd_tgts_bitmap, idx) {
1602 ltd = lfsck_tgt_get(ltds, idx);
1603 LASSERT(ltd != NULL);
1605 laia->laia_ltd = ltd;
1606 ltd->ltd_layout_done = 0;
1607 rc = lfsck_async_request(env, ltd->ltd_exp, lr, set,
1608 lfsck_async_interpret, laia,
1611 lfsck_interpret(env, lfsck, NULL, laia, rc);
1613 CERROR("%s: cannot notify MDT %x for LFSCK "
1614 "start, failout: rc = %d\n",
1615 lfsck_lfsck2name(lfsck), idx, rc);
1619 up_read(<ds->ltd_rw_sem);
1622 ptlrpc_set_destroy(set);
1627 rc = ptlrpc_set_wait(set);
1628 ptlrpc_set_destroy(set);
1631 rc = laia->laia_result;
1634 struct lfsck_stop *stop = &info->lti_stop;
1636 CERROR("%s: cannot start LFSCK on some MDTs, "
1637 "stop all: rc = %d\n",
1638 lfsck_lfsck2name(lfsck), rc);
1639 if (rc != -EALREADY) {
1640 stop->ls_status = LS_FAILED;
1641 stop->ls_flags = LPF_ALL_TGT | LPF_BROADCAST;
1642 lfsck_stop_all(env, lfsck, stop);
1649 int lfsck_start(const struct lu_env *env, struct dt_device *key,
1650 struct lfsck_start_param *lsp)
1652 struct lfsck_start *start = lsp->lsp_start;
1653 struct lfsck_instance *lfsck;
1654 struct lfsck_bookmark *bk;
1655 struct ptlrpc_thread *thread;
1656 struct lfsck_component *com;
1657 struct l_wait_info lwi = { 0 };
1658 struct lfsck_thread_args *lta;
1659 struct task_struct *task;
1666 lfsck = lfsck_instance_find(key, true, false);
1667 if (unlikely(lfsck == NULL))
1670 /* System is not ready, try again later. */
1671 if (unlikely(lfsck->li_namespace == NULL))
1672 GOTO(put, rc = -EAGAIN);
1674 /* start == NULL means auto trigger paused LFSCK. */
1675 if ((start == NULL) &&
1676 (list_empty(&lfsck->li_list_scan) ||
1677 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
1680 bk = &lfsck->li_bookmark_ram;
1681 thread = &lfsck->li_thread;
1682 mutex_lock(&lfsck->li_mutex);
1683 spin_lock(&lfsck->li_lock);
1684 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
1686 if (unlikely(start == NULL)) {
1687 spin_unlock(&lfsck->li_lock);
1691 while (start->ls_active != 0) {
1692 if (!(type & start->ls_active)) {
1697 com = __lfsck_component_find(lfsck, type,
1698 &lfsck->li_list_scan);
1700 com = __lfsck_component_find(lfsck, type,
1701 &lfsck->li_list_double_scan);
1707 if (com->lc_ops->lfsck_join != NULL) {
1708 rc = com->lc_ops->lfsck_join( env, com, lsp);
1709 if (rc != 0 && rc != -EALREADY)
1712 start->ls_active &= ~type;
1715 spin_unlock(&lfsck->li_lock);
1718 spin_unlock(&lfsck->li_lock);
1720 lfsck->li_status = 0;
1721 lfsck->li_oit_over = 0;
1722 lfsck->li_start_unplug = 0;
1723 lfsck->li_drop_dryrun = 0;
1724 lfsck->li_new_scanned = 0;
1726 /* For auto trigger. */
1730 if (start->ls_flags & LPF_BROADCAST && !lfsck->li_master) {
1731 CERROR("%s: only allow to specify '-A | -o' via MDS\n",
1732 lfsck_lfsck2name(lfsck));
1734 GOTO(out, rc = -EPERM);
1737 start->ls_version = bk->lb_version;
1739 if (start->ls_active != 0) {
1740 struct lfsck_component *next;
1742 if (start->ls_active == LFSCK_TYPES_ALL)
1743 start->ls_active = LFSCK_TYPES_SUPPORTED;
1745 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
1746 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
1747 GOTO(out, rc = -ENOTSUPP);
1750 list_for_each_entry_safe(com, next,
1751 &lfsck->li_list_scan, lc_link) {
1752 if (!(com->lc_type & start->ls_active)) {
1753 rc = com->lc_ops->lfsck_post(env, com, 0,
1760 while (start->ls_active != 0) {
1761 if (type & start->ls_active) {
1762 com = __lfsck_component_find(lfsck, type,
1763 &lfsck->li_list_idle);
1765 /* The component status will be updated
1766 * when its prep() is called later by
1767 * the LFSCK main engine. */
1768 list_move_tail(&com->lc_link,
1769 &lfsck->li_list_scan);
1770 start->ls_active &= ~type;
1776 if (list_empty(&lfsck->li_list_scan)) {
1777 /* The speed limit will be used to control both the LFSCK and
1778 * low layer scrub (if applied), need to be handled firstly. */
1779 if (start->ls_valid & LSV_SPEED_LIMIT) {
1780 if (__lfsck_set_speed(lfsck, start->ls_speed_limit)) {
1781 rc = lfsck_bookmark_store(env, lfsck);
1790 if (start->ls_flags & LPF_RESET)
1791 flags |= DOIF_RESET;
1793 rc = lfsck_set_param(env, lfsck, start, !!(flags & DOIF_RESET));
1797 list_for_each_entry(com, &lfsck->li_list_scan, lc_link) {
1798 start->ls_active |= com->lc_type;
1799 if (flags & DOIF_RESET) {
1800 rc = com->lc_ops->lfsck_reset(env, com, false);
1807 lfsck->li_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
1808 if (bk->lb_param & LPF_DRYRUN)
1809 lfsck->li_args_dir |= LUDA_VERIFY_DRYRUN;
1811 if (start != NULL && start->ls_valid & LSV_ERROR_HANDLE) {
1812 valid |= DOIV_ERROR_HANDLE;
1813 if (start->ls_flags & LPF_FAILOUT)
1814 flags |= DOIF_FAILOUT;
1817 if (start != NULL && start->ls_valid & LSV_DRYRUN) {
1818 valid |= DOIV_DRYRUN;
1819 if (start->ls_flags & LPF_DRYRUN)
1820 flags |= DOIF_DRYRUN;
1823 if (!list_empty(&lfsck->li_list_scan))
1824 flags |= DOIF_OUTUSED;
1826 lfsck->li_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
1827 thread_set_flags(thread, 0);
1828 lta = lfsck_thread_args_init(lfsck, NULL, lsp);
1830 GOTO(out, rc = PTR_ERR(lta));
1832 __lfsck_set_speed(lfsck, bk->lb_speed_limit);
1833 task = kthread_run(lfsck_master_engine, lta, "lfsck");
1836 CERROR("%s: cannot start LFSCK thread: rc = %d\n",
1837 lfsck_lfsck2name(lfsck), rc);
1838 lfsck_thread_args_fini(lta);
1843 l_wait_event(thread->t_ctl_waitq,
1844 thread_is_running(thread) ||
1845 thread_is_stopped(thread),
1847 if (start == NULL || !(start->ls_flags & LPF_BROADCAST)) {
1848 lfsck->li_start_unplug = 1;
1849 wake_up_all(&thread->t_ctl_waitq);
1854 /* release lfsck::li_mutex to avoid deadlock. */
1855 mutex_unlock(&lfsck->li_mutex);
1856 rc = lfsck_start_all(env, lfsck, start);
1858 spin_lock(&lfsck->li_lock);
1859 if (thread_is_stopped(thread)) {
1860 spin_unlock(&lfsck->li_lock);
1862 lfsck->li_status = LS_FAILED;
1863 lfsck->li_flags = 0;
1864 thread_set_flags(thread, SVC_STOPPING);
1865 spin_unlock(&lfsck->li_lock);
1867 lfsck->li_start_unplug = 1;
1868 wake_up_all(&thread->t_ctl_waitq);
1869 l_wait_event(thread->t_ctl_waitq,
1870 thread_is_stopped(thread),
1874 lfsck->li_start_unplug = 1;
1875 wake_up_all(&thread->t_ctl_waitq);
1881 mutex_unlock(&lfsck->li_mutex);
1884 lfsck_instance_put(env, lfsck);
1886 return rc < 0 ? rc : 0;
1888 EXPORT_SYMBOL(lfsck_start);
1890 int lfsck_stop(const struct lu_env *env, struct dt_device *key,
1891 struct lfsck_stop *stop)
1893 struct lfsck_instance *lfsck;
1894 struct ptlrpc_thread *thread;
1895 struct l_wait_info lwi = { 0 };
1900 lfsck = lfsck_instance_find(key, true, false);
1901 if (unlikely(lfsck == NULL))
1904 thread = &lfsck->li_thread;
1905 /* release lfsck::li_mutex to avoid deadlock. */
1906 if (stop != NULL && stop->ls_flags & LPF_BROADCAST) {
1907 if (!lfsck->li_master) {
1908 CERROR("%s: only allow to specify '-A' via MDS\n",
1909 lfsck_lfsck2name(lfsck));
1911 GOTO(out, rc = -EPERM);
1914 rc1 = lfsck_stop_all(env, lfsck, stop);
1917 mutex_lock(&lfsck->li_mutex);
1918 spin_lock(&lfsck->li_lock);
1919 /* no error if LFSCK is already stopped, or was never started */
1920 if (thread_is_init(thread) || thread_is_stopped(thread)) {
1921 spin_unlock(&lfsck->li_lock);
1926 lfsck->li_status = stop->ls_status;
1927 lfsck->li_flags = stop->ls_flags;
1929 lfsck->li_status = LS_STOPPED;
1930 lfsck->li_flags = 0;
1933 thread_set_flags(thread, SVC_STOPPING);
1934 spin_unlock(&lfsck->li_lock);
1936 wake_up_all(&thread->t_ctl_waitq);
1937 l_wait_event(thread->t_ctl_waitq,
1938 thread_is_stopped(thread),
1944 mutex_unlock(&lfsck->li_mutex);
1945 lfsck_instance_put(env, lfsck);
1947 return rc != 0 ? rc : rc1;
1949 EXPORT_SYMBOL(lfsck_stop);
1951 int lfsck_in_notify(const struct lu_env *env, struct dt_device *key,
1952 struct lfsck_request *lr)
1954 int rc = -EOPNOTSUPP;
1957 switch (lr->lr_event) {
1959 struct lfsck_start *start = &lfsck_env_info(env)->lti_start;
1960 struct lfsck_start_param lsp;
1962 memset(start, 0, sizeof(*start));
1963 start->ls_valid = lr->lr_valid;
1964 start->ls_speed_limit = lr->lr_speed;
1965 start->ls_version = lr->lr_version;
1966 start->ls_active = lr->lr_active;
1967 start->ls_flags = lr->lr_param & ~LPF_BROADCAST;
1968 start->ls_async_windows = lr->lr_async_windows;
1970 lsp.lsp_start = start;
1971 lsp.lsp_index = lr->lr_index;
1972 lsp.lsp_index_valid = 1;
1973 rc = lfsck_start(env, key, &lsp);
1977 struct lfsck_stop *stop = &lfsck_env_info(env)->lti_stop;
1979 memset(stop, 0, sizeof(*stop));
1980 stop->ls_status = lr->lr_status;
1981 stop->ls_flags = lr->lr_param & ~LPF_BROADCAST;
1982 rc = lfsck_stop(env, key, stop);
1985 case LE_PHASE1_DONE:
1986 case LE_PHASE2_DONE:
1987 case LE_FID_ACCESSED:
1989 case LE_CONDITIONAL_DESTROY:
1990 case LE_PAIRS_VERIFY: {
1991 struct lfsck_instance *lfsck;
1992 struct lfsck_component *com;
1994 lfsck = lfsck_instance_find(key, true, false);
1995 if (unlikely(lfsck == NULL))
1998 com = lfsck_component_find(lfsck, lr->lr_active);
1999 if (likely(com != NULL)) {
2000 rc = com->lc_ops->lfsck_in_notify(env, com, lr);
2001 lfsck_component_put(env, com);
2004 lfsck_instance_put(env, lfsck);
2013 EXPORT_SYMBOL(lfsck_in_notify);
2015 int lfsck_query(const struct lu_env *env, struct dt_device *key,
2016 struct lfsck_request *lr)
2018 struct lfsck_instance *lfsck;
2019 struct lfsck_component *com;
2023 lfsck = lfsck_instance_find(key, true, false);
2024 if (unlikely(lfsck == NULL))
2027 com = lfsck_component_find(lfsck, lr->lr_active);
2028 if (likely(com != NULL)) {
2029 rc = com->lc_ops->lfsck_query(env, com);
2030 lfsck_component_put(env, com);
2035 lfsck_instance_put(env, lfsck);
2039 EXPORT_SYMBOL(lfsck_query);
2041 int lfsck_register_namespace(const struct lu_env *env, struct dt_device *key,
2042 struct ldlm_namespace *ns)
2044 struct lfsck_instance *lfsck;
2047 lfsck = lfsck_instance_find(key, true, false);
2048 if (likely(lfsck != NULL)) {
2049 lfsck->li_namespace = ns;
2050 lfsck_instance_put(env, lfsck);
2056 EXPORT_SYMBOL(lfsck_register_namespace);
2058 int lfsck_register(const struct lu_env *env, struct dt_device *key,
2059 struct dt_device *next, struct obd_device *obd,
2060 lfsck_out_notify notify, void *notify_data, bool master)
2062 struct lfsck_instance *lfsck;
2063 struct dt_object *root = NULL;
2064 struct dt_object *obj = NULL;
2065 struct lu_fid *fid = &lfsck_env_info(env)->lti_fid;
2069 lfsck = lfsck_instance_find(key, false, false);
2070 if (unlikely(lfsck != NULL))
2073 OBD_ALLOC_PTR(lfsck);
2077 mutex_init(&lfsck->li_mutex);
2078 spin_lock_init(&lfsck->li_lock);
2079 INIT_LIST_HEAD(&lfsck->li_link);
2080 INIT_LIST_HEAD(&lfsck->li_list_scan);
2081 INIT_LIST_HEAD(&lfsck->li_list_dir);
2082 INIT_LIST_HEAD(&lfsck->li_list_double_scan);
2083 INIT_LIST_HEAD(&lfsck->li_list_idle);
2084 atomic_set(&lfsck->li_ref, 1);
2085 atomic_set(&lfsck->li_double_scan_count, 0);
2086 init_waitqueue_head(&lfsck->li_thread.t_ctl_waitq);
2087 lfsck->li_out_notify = notify;
2088 lfsck->li_out_notify_data = notify_data;
2089 lfsck->li_next = next;
2090 lfsck->li_bottom = key;
2091 lfsck->li_obd = obd;
2093 rc = lfsck_tgt_descs_init(&lfsck->li_ost_descs);
2097 rc = lfsck_tgt_descs_init(&lfsck->li_mdt_descs);
2101 fid->f_seq = FID_SEQ_LOCAL_NAME;
2104 rc = local_oid_storage_init(env, key, fid, &lfsck->li_los);
2108 rc = dt_root_get(env, key, fid);
2112 root = dt_locate(env, key, fid);
2114 GOTO(out, rc = PTR_ERR(root));
2116 if (unlikely(!dt_try_as_dir(env, root)))
2117 GOTO(out, rc = -ENOTDIR);
2119 lfsck->li_local_root_fid = *fid;
2121 lfsck->li_master = 1;
2122 if (lfsck_dev_idx(key) == 0) {
2123 struct lu_fid *pfid = &lfsck_env_info(env)->lti_fid2;
2124 const struct lu_name *cname;
2126 rc = dt_lookup(env, root,
2127 (struct dt_rec *)(&lfsck->li_global_root_fid),
2128 (const struct dt_key *)"ROOT", BYPASS_CAPA);
2132 obj = dt_locate(env, key, &lfsck->li_global_root_fid);
2134 GOTO(out, rc = PTR_ERR(obj));
2136 rc = dt_lookup(env, obj, (struct dt_rec *)fid,
2137 (const struct dt_key *)dotlustre, BYPASS_CAPA);
2141 lu_object_put(env, &obj->do_lu);
2142 obj = dt_locate(env, key, fid);
2144 GOTO(out, rc = PTR_ERR(obj));
2146 cname = lfsck_name_get_const(env, dotlustre,
2148 rc = lfsck_verify_linkea(env, key, obj, cname,
2149 &lfsck->li_global_root_fid);
2154 rc = dt_lookup(env, obj, (struct dt_rec *)fid,
2155 (const struct dt_key *)lostfound,
2160 lu_object_put(env, &obj->do_lu);
2161 obj = dt_locate(env, key, fid);
2163 GOTO(out, rc = PTR_ERR(obj));
2165 cname = lfsck_name_get_const(env, lostfound,
2167 rc = lfsck_verify_linkea(env, key, obj, cname, pfid);
2171 lu_object_put(env, &obj->do_lu);
2176 fid->f_seq = FID_SEQ_LOCAL_FILE;
2177 fid->f_oid = OTABLE_IT_OID;
2179 obj = dt_locate(env, key, fid);
2181 GOTO(out, rc = PTR_ERR(obj));
2183 lu_object_get(&obj->do_lu);
2184 lfsck->li_obj_oit = obj;
2185 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2189 rc = lfsck_bookmark_setup(env, lfsck);
2194 rc = lfsck_fid_init(lfsck);
2198 rc = lfsck_namespace_setup(env, lfsck);
2203 rc = lfsck_layout_setup(env, lfsck);
2207 /* XXX: more LFSCK components initialization to be added here. */
2209 rc = lfsck_instance_add(lfsck);
2211 rc = lfsck_add_target_from_orphan(env, lfsck);
2213 if (obj != NULL && !IS_ERR(obj))
2214 lu_object_put(env, &obj->do_lu);
2215 if (root != NULL && !IS_ERR(root))
2216 lu_object_put(env, &root->do_lu);
2218 lfsck_instance_cleanup(env, lfsck);
2221 EXPORT_SYMBOL(lfsck_register);
2223 void lfsck_degister(const struct lu_env *env, struct dt_device *key)
2225 struct lfsck_instance *lfsck;
2227 lfsck = lfsck_instance_find(key, false, true);
2229 lfsck_instance_put(env, lfsck);
2231 EXPORT_SYMBOL(lfsck_degister);
2233 int lfsck_add_target(const struct lu_env *env, struct dt_device *key,
2234 struct dt_device *tgt, struct obd_export *exp,
2235 __u32 index, bool for_ost)
2237 struct lfsck_instance *lfsck;
2238 struct lfsck_tgt_desc *ltd;
2249 INIT_LIST_HEAD(<d->ltd_orphan_list);
2250 INIT_LIST_HEAD(<d->ltd_layout_list);
2251 INIT_LIST_HEAD(<d->ltd_layout_phase_list);
2252 atomic_set(<d->ltd_ref, 1);
2253 ltd->ltd_index = index;
2255 spin_lock(&lfsck_instance_lock);
2256 lfsck = __lfsck_instance_find(key, true, false);
2257 if (lfsck == NULL) {
2259 list_add_tail(<d->ltd_orphan_list,
2260 &lfsck_ost_orphan_list);
2262 list_add_tail(<d->ltd_orphan_list,
2263 &lfsck_mdt_orphan_list);
2264 spin_unlock(&lfsck_instance_lock);
2268 spin_unlock(&lfsck_instance_lock);
2270 rc = __lfsck_add_target(env, lfsck, ltd, for_ost, false);
2274 lfsck_instance_put(env, lfsck);
2278 EXPORT_SYMBOL(lfsck_add_target);
2280 void lfsck_del_target(const struct lu_env *env, struct dt_device *key,
2281 struct dt_device *tgt, __u32 index, bool for_ost)
2283 struct lfsck_instance *lfsck;
2284 struct lfsck_tgt_descs *ltds;
2285 struct lfsck_tgt_desc *ltd;
2286 struct list_head *head;
2289 head = &lfsck_ost_orphan_list;
2291 head = &lfsck_mdt_orphan_list;
2293 spin_lock(&lfsck_instance_lock);
2294 list_for_each_entry(ltd, head, ltd_orphan_list) {
2295 if (ltd->ltd_tgt == tgt) {
2296 list_del_init(<d->ltd_orphan_list);
2297 spin_unlock(&lfsck_instance_lock);
2305 lfsck = __lfsck_instance_find(key, true, false);
2306 spin_unlock(&lfsck_instance_lock);
2307 if (unlikely(lfsck == NULL))
2311 ltds = &lfsck->li_ost_descs;
2313 ltds = &lfsck->li_mdt_descs;
2315 down_write(<ds->ltd_rw_sem);
2316 LASSERT(ltds->ltd_tgts_bitmap != NULL);
2318 if (unlikely(index >= ltds->ltd_tgts_bitmap->size))
2321 ltd = LTD_TGT(ltds, index);
2322 if (unlikely(ltd == NULL))
2325 LASSERT(ltds->ltd_tgtnr > 0);
2328 cfs_bitmap_clear(ltds->ltd_tgts_bitmap, index);
2329 LTD_TGT(ltds, index) = NULL;
2334 head = &lfsck->li_ost_descs.ltd_orphan;
2336 head = &lfsck->li_mdt_descs.ltd_orphan;
2338 list_for_each_entry(ltd, head, ltd_orphan_list) {
2339 if (ltd->ltd_tgt == tgt) {
2340 list_del_init(<d->ltd_orphan_list);
2346 up_write(<ds->ltd_rw_sem);
2348 spin_lock(<ds->ltd_lock);
2350 spin_unlock(<ds->ltd_lock);
2351 lfsck_stop_notify(env, lfsck, ltds, ltd, LFSCK_TYPE_LAYOUT);
2355 lfsck_instance_put(env, lfsck);
2357 EXPORT_SYMBOL(lfsck_del_target);
2359 static int __init lfsck_init(void)
2363 INIT_LIST_HEAD(&lfsck_instance_list);
2364 INIT_LIST_HEAD(&lfsck_ost_orphan_list);
2365 INIT_LIST_HEAD(&lfsck_mdt_orphan_list);
2366 lfsck_key_init_generic(&lfsck_thread_key, NULL);
2367 rc = lu_context_key_register(&lfsck_thread_key);
2369 tgt_register_lfsck_in_notify(lfsck_in_notify);
2370 tgt_register_lfsck_query(lfsck_query);
2376 static void __exit lfsck_exit(void)
2378 struct lfsck_tgt_desc *ltd;
2379 struct lfsck_tgt_desc *next;
2381 LASSERT(list_empty(&lfsck_instance_list));
2383 list_for_each_entry_safe(ltd, next, &lfsck_ost_orphan_list,
2385 list_del_init(<d->ltd_orphan_list);
2389 list_for_each_entry_safe(ltd, next, &lfsck_mdt_orphan_list,
2391 list_del_init(<d->ltd_orphan_list);
2395 lu_context_key_degister(&lfsck_thread_key);
2398 MODULE_AUTHOR("Intel Corporation <http://www.intel.com/>");
2399 MODULE_DESCRIPTION("LFSCK");
2400 MODULE_LICENSE("GPL");
2402 cfs_module(lfsck, LUSTRE_VERSION_STRING, lfsck_init, lfsck_exit);