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, Intel Corporation.
26 * lustre/mdd/mdd_lfsck.c
28 * Top-level entry points into mdd module
30 * LFSCK controller, which scans the whole device through low layer
31 * iteration APIs, drives all lfsck compeonents, controls the speed.
33 * Author: Fan Yong <yong.fan@whamcloud.com>
37 # define EXPORT_SYMTAB
39 #define DEBUG_SUBSYSTEM S_MDS
41 #include <lustre/lustre_idl.h>
42 #include <lustre_fid.h>
43 #include <obd_support.h>
45 #include "mdd_internal.h"
46 #include "mdd_lfsck.h"
48 #define HALF_SEC (CFS_HZ >> 1)
49 #define LFSCK_CHECKPOINT_INTERVAL 60
50 #define MDS_DIR_DUMMY_START 0xffffffffffffffffULL
52 #define LFSCK_NAMEENTRY_DEAD 1 /* The object has been unlinked. */
53 #define LFSCK_NAMEENTRY_REMOVED 2 /* The entry has been removed. */
54 #define LFSCK_NAMEENTRY_RECREATED 3 /* The entry has been recreated. */
56 const char lfsck_bookmark_name[] = "lfsck_bookmark";
57 const char lfsck_namespace_name[] = "lfsck_namespace";
59 static const char *lfsck_status_names[] = {
71 static const char *lfsck_flags_names[] = {
78 static const char *lfsck_param_names[] = {
86 static inline struct mdd_device *mdd_lfsck2mdd(struct md_lfsck *lfsck)
88 return container_of0(lfsck, struct mdd_device, mdd_lfsck);
91 static inline char *mdd_lfsck2name(struct md_lfsck *lfsck)
93 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
95 return mdd2obd_dev(mdd)->obd_name;
98 static inline void mdd_lfsck_component_get(struct lfsck_component *com)
100 atomic_inc(&com->lc_ref);
103 static inline void mdd_lfsck_component_put(const struct lu_env *env,
104 struct lfsck_component *com)
106 if (atomic_dec_and_test(&com->lc_ref)) {
107 if (com->lc_obj != NULL)
108 lu_object_put(env, &com->lc_obj->do_lu);
109 if (com->lc_file_ram != NULL)
110 OBD_FREE(com->lc_file_ram, com->lc_file_size);
111 if (com->lc_file_disk != NULL)
112 OBD_FREE(com->lc_file_disk, com->lc_file_size);
117 static inline struct lfsck_component *
118 __mdd_lfsck_component_find(struct md_lfsck *lfsck, __u16 type, cfs_list_t *list)
120 struct lfsck_component *com;
122 cfs_list_for_each_entry(com, list, lc_link) {
123 if (com->lc_type == type)
129 static struct lfsck_component *
130 mdd_lfsck_component_find(struct md_lfsck *lfsck, __u16 type)
132 struct lfsck_component *com;
134 spin_lock(&lfsck->ml_lock);
135 com = __mdd_lfsck_component_find(lfsck, type, &lfsck->ml_list_scan);
139 com = __mdd_lfsck_component_find(lfsck, type,
140 &lfsck->ml_list_double_scan);
144 com = __mdd_lfsck_component_find(lfsck, type, &lfsck->ml_list_idle);
148 mdd_lfsck_component_get(com);
149 spin_unlock(&lfsck->ml_lock);
153 static void mdd_lfsck_component_cleanup(const struct lu_env *env,
154 struct lfsck_component *com)
156 if (!cfs_list_empty(&com->lc_link))
157 cfs_list_del_init(&com->lc_link);
158 if (!cfs_list_empty(&com->lc_link_dir))
159 cfs_list_del_init(&com->lc_link_dir);
161 mdd_lfsck_component_put(env, com);
164 static int lfsck_bits_dump(char **buf, int *len, int bits, const char *names[],
172 rc = snprintf(*buf, *len, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
178 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
181 rc = snprintf(*buf, *len, "%s%c", names[i],
182 bits != 0 ? ',' : '\n');
193 static int lfsck_time_dump(char **buf, int *len, __u64 time, const char *prefix)
198 rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
199 cfs_time_current_sec() - time);
201 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
210 static int lfsck_pos_dump(char **buf, int *len, struct lfsck_position *pos,
215 if (fid_is_zero(&pos->lp_dir_parent)) {
216 if (pos->lp_oit_cookie == 0)
217 rc = snprintf(*buf, *len, "%s: N/A, N/A, N/A\n",
220 rc = snprintf(*buf, *len, "%s: "LPU64", N/A, N/A\n",
221 prefix, pos->lp_oit_cookie);
223 rc = snprintf(*buf, *len, "%s: "LPU64", "DFID", "LPU64"\n",
224 prefix, pos->lp_oit_cookie,
225 PFID(&pos->lp_dir_parent), pos->lp_dir_cookie);
235 static void mdd_lfsck_pos_fill(const struct lu_env *env, struct md_lfsck *lfsck,
236 struct lfsck_position *pos, bool oit_processed,
239 const struct dt_it_ops *iops = &lfsck->ml_obj_oit->do_index_ops->dio_it;
241 spin_lock(&lfsck->ml_lock);
242 if (unlikely(lfsck->ml_di_oit == NULL)) {
243 spin_unlock(&lfsck->ml_lock);
244 memset(pos, 0, sizeof(*pos));
248 pos->lp_oit_cookie = iops->store(env, lfsck->ml_di_oit);
250 LASSERT(pos->lp_oit_cookie > 0);
253 pos->lp_oit_cookie--;
255 if (lfsck->ml_di_dir != NULL) {
256 struct dt_object *dto = lfsck->ml_obj_dir;
258 pos->lp_dir_parent = *lu_object_fid(&dto->do_lu);
259 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
262 LASSERT(pos->lp_dir_cookie != MDS_DIR_DUMMY_START);
264 if (pos->lp_dir_cookie >= MDS_DIR_END_OFF)
265 pos->lp_dir_cookie = MDS_DIR_END_OFF;
266 else if (!dir_processed)
267 /* For the dir which just to be processed,
268 * lp_dir_cookie will become MDS_DIR_DUMMY_START,
269 * which can be correctly handled by mdd_lfsck_prep. */
270 pos->lp_dir_cookie--;
272 fid_zero(&pos->lp_dir_parent);
273 pos->lp_dir_cookie = 0;
275 spin_unlock(&lfsck->ml_lock);
278 static inline void mdd_lfsck_pos_set_zero(struct lfsck_position *pos)
280 memset(pos, 0, sizeof(*pos));
283 static inline int mdd_lfsck_pos_is_zero(const struct lfsck_position *pos)
285 return pos->lp_oit_cookie == 0 && fid_is_zero(&pos->lp_dir_parent);
288 static inline int mdd_lfsck_pos_is_eq(const struct lfsck_position *pos1,
289 const struct lfsck_position *pos2)
291 if (pos1->lp_oit_cookie < pos2->lp_oit_cookie)
294 if (pos1->lp_oit_cookie > pos2->lp_oit_cookie)
297 if (fid_is_zero(&pos1->lp_dir_parent) &&
298 !fid_is_zero(&pos2->lp_dir_parent))
301 if (!fid_is_zero(&pos1->lp_dir_parent) &&
302 fid_is_zero(&pos2->lp_dir_parent))
305 if (fid_is_zero(&pos1->lp_dir_parent) &&
306 fid_is_zero(&pos2->lp_dir_parent))
309 LASSERT(lu_fid_eq(&pos1->lp_dir_parent, &pos2->lp_dir_parent));
311 if (pos1->lp_dir_cookie < pos2->lp_dir_cookie)
314 if (pos1->lp_dir_cookie > pos2->lp_dir_cookie)
320 static void mdd_lfsck_close_dir(const struct lu_env *env,
321 struct md_lfsck *lfsck)
323 struct dt_object *dir_obj = lfsck->ml_obj_dir;
324 const struct dt_it_ops *dir_iops = &dir_obj->do_index_ops->dio_it;
325 struct dt_it *dir_di = lfsck->ml_di_dir;
327 spin_lock(&lfsck->ml_lock);
328 lfsck->ml_di_dir = NULL;
329 spin_unlock(&lfsck->ml_lock);
331 dir_iops->put(env, dir_di);
332 dir_iops->fini(env, dir_di);
333 lfsck->ml_obj_dir = NULL;
334 lu_object_put(env, &dir_obj->do_lu);
337 static void __mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
339 lfsck->ml_bookmark_ram.lb_speed_limit = limit;
340 if (limit != LFSCK_SPEED_NO_LIMIT) {
341 if (limit > CFS_HZ) {
342 lfsck->ml_sleep_rate = limit / CFS_HZ;
343 lfsck->ml_sleep_jif = 1;
345 lfsck->ml_sleep_rate = 1;
346 lfsck->ml_sleep_jif = CFS_HZ / limit;
349 lfsck->ml_sleep_jif = 0;
350 lfsck->ml_sleep_rate = 0;
354 static void mdd_lfsck_control_speed(struct md_lfsck *lfsck)
356 struct ptlrpc_thread *thread = &lfsck->ml_thread;
357 struct l_wait_info lwi;
359 if (lfsck->ml_sleep_jif > 0 &&
360 lfsck->ml_new_scanned >= lfsck->ml_sleep_rate) {
361 spin_lock(&lfsck->ml_lock);
362 if (likely(lfsck->ml_sleep_jif > 0 &&
363 lfsck->ml_new_scanned >= lfsck->ml_sleep_rate)) {
364 lwi = LWI_TIMEOUT_INTR(lfsck->ml_sleep_jif, NULL,
365 LWI_ON_SIGNAL_NOOP, NULL);
366 spin_unlock(&lfsck->ml_lock);
368 l_wait_event(thread->t_ctl_waitq,
369 !thread_is_running(thread),
371 lfsck->ml_new_scanned = 0;
373 spin_unlock(&lfsck->ml_lock);
378 /* lfsck_bookmark file ops */
380 static void inline mdd_lfsck_bookmark_to_cpu(struct lfsck_bookmark *des,
381 struct lfsck_bookmark *src)
383 des->lb_magic = le32_to_cpu(src->lb_magic);
384 des->lb_version = le16_to_cpu(src->lb_version);
385 des->lb_param = le16_to_cpu(src->lb_param);
386 des->lb_speed_limit = le32_to_cpu(src->lb_speed_limit);
389 static void inline mdd_lfsck_bookmark_to_le(struct lfsck_bookmark *des,
390 struct lfsck_bookmark *src)
392 des->lb_magic = cpu_to_le32(src->lb_magic);
393 des->lb_version = cpu_to_le16(src->lb_version);
394 des->lb_param = cpu_to_le16(src->lb_param);
395 des->lb_speed_limit = cpu_to_le32(src->lb_speed_limit);
398 static int mdd_lfsck_bookmark_load(const struct lu_env *env,
399 struct md_lfsck *lfsck)
402 int len = sizeof(struct lfsck_bookmark);
405 rc = dt_record_read(env, lfsck->ml_bookmark_obj,
406 mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
409 struct lfsck_bookmark *bm = &lfsck->ml_bookmark_ram;
411 mdd_lfsck_bookmark_to_cpu(bm, &lfsck->ml_bookmark_disk);
412 if (bm->lb_magic != LFSCK_BOOKMARK_MAGIC) {
413 CWARN("%.16s: invalid lfsck_bookmark magic "
414 "0x%x != 0x%x\n", mdd_lfsck2name(lfsck),
415 bm->lb_magic, LFSCK_BOOKMARK_MAGIC);
416 /* Process it as new lfsck_bookmark. */
420 if (rc == -EFAULT && pos == 0)
421 /* return -ENODATA for empty lfsck_bookmark. */
424 CERROR("%.16s: fail to load lfsck_bookmark, "
425 "expected = %d, rc = %d\n",
426 mdd_lfsck2name(lfsck), len, rc);
431 static int mdd_lfsck_bookmark_store(const struct lu_env *env,
432 struct md_lfsck *lfsck)
434 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
435 struct thandle *handle;
436 struct dt_object *obj = lfsck->ml_bookmark_obj;
438 int len = sizeof(struct lfsck_bookmark);
442 mdd_lfsck_bookmark_to_le(&lfsck->ml_bookmark_disk,
443 &lfsck->ml_bookmark_ram);
444 handle = dt_trans_create(env, mdd->mdd_bottom);
445 if (IS_ERR(handle)) {
446 rc = PTR_ERR(handle);
447 CERROR("%.16s: fail to create trans for storing "
448 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
452 rc = dt_declare_record_write(env, obj, len, 0, handle);
454 CERROR("%.16s: fail to declare trans for storing "
455 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
459 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
461 CERROR("%.16s: fail to start trans for storing "
462 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
466 rc = dt_record_write(env, obj,
467 mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
470 CERROR("%.16s: fail to store lfsck_bookmark, expected = %d, "
471 "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
476 dt_trans_stop(env, mdd->mdd_bottom, handle);
480 static int mdd_lfsck_bookmark_init(const struct lu_env *env,
481 struct md_lfsck *lfsck)
483 struct lfsck_bookmark *mb = &lfsck->ml_bookmark_ram;
486 memset(mb, 0, sizeof(*mb));
487 mb->lb_magic = LFSCK_BOOKMARK_MAGIC;
488 mb->lb_version = LFSCK_VERSION_V2;
489 mutex_lock(&lfsck->ml_mutex);
490 rc = mdd_lfsck_bookmark_store(env, lfsck);
491 mutex_unlock(&lfsck->ml_mutex);
495 /* lfsck_namespace file ops */
497 static void inline mdd_lfsck_position_to_cpu(struct lfsck_position *des,
498 struct lfsck_position *src)
500 des->lp_oit_cookie = le64_to_cpu(src->lp_oit_cookie);
501 fid_le_to_cpu(&des->lp_dir_parent, &src->lp_dir_parent);
502 des->lp_dir_cookie = le64_to_cpu(src->lp_dir_cookie);
505 static void inline mdd_lfsck_position_to_le(struct lfsck_position *des,
506 struct lfsck_position *src)
508 des->lp_oit_cookie = cpu_to_le64(src->lp_oit_cookie);
509 fid_cpu_to_le(&des->lp_dir_parent, &src->lp_dir_parent);
510 des->lp_dir_cookie = cpu_to_le64(src->lp_dir_cookie);
513 static void inline mdd_lfsck_namespace_to_cpu(struct lfsck_namespace *des,
514 struct lfsck_namespace *src)
516 des->ln_magic = le32_to_cpu(src->ln_magic);
517 des->ln_status = le32_to_cpu(src->ln_status);
518 des->ln_flags = le32_to_cpu(src->ln_flags);
519 des->ln_success_count = le32_to_cpu(src->ln_success_count);
520 des->ln_run_time_phase1 = le32_to_cpu(src->ln_run_time_phase1);
521 des->ln_run_time_phase2 = le32_to_cpu(src->ln_run_time_phase2);
522 des->ln_time_last_complete = le64_to_cpu(src->ln_time_last_complete);
523 des->ln_time_latest_start = le64_to_cpu(src->ln_time_latest_start);
524 des->ln_time_last_checkpoint =
525 le64_to_cpu(src->ln_time_last_checkpoint);
526 mdd_lfsck_position_to_cpu(&des->ln_pos_latest_start,
527 &src->ln_pos_latest_start);
528 mdd_lfsck_position_to_cpu(&des->ln_pos_last_checkpoint,
529 &src->ln_pos_last_checkpoint);
530 mdd_lfsck_position_to_cpu(&des->ln_pos_first_inconsistent,
531 &src->ln_pos_first_inconsistent);
532 des->ln_items_checked = le64_to_cpu(src->ln_items_checked);
533 des->ln_items_repaired = le64_to_cpu(src->ln_items_repaired);
534 des->ln_items_failed = le64_to_cpu(src->ln_items_failed);
535 des->ln_dirs_checked = le64_to_cpu(src->ln_dirs_checked);
536 des->ln_mlinked_checked = le64_to_cpu(src->ln_mlinked_checked);
537 des->ln_objs_checked_phase2 = le64_to_cpu(src->ln_objs_checked_phase2);
538 des->ln_objs_repaired_phase2 =
539 le64_to_cpu(src->ln_objs_repaired_phase2);
540 des->ln_objs_failed_phase2 = le64_to_cpu(src->ln_objs_failed_phase2);
541 des->ln_objs_nlink_repaired = le64_to_cpu(src->ln_objs_nlink_repaired);
542 des->ln_objs_lost_found = le64_to_cpu(src->ln_objs_lost_found);
543 fid_le_to_cpu(&des->ln_fid_latest_scanned_phase2,
544 &src->ln_fid_latest_scanned_phase2);
547 static void inline mdd_lfsck_namespace_to_le(struct lfsck_namespace *des,
548 struct lfsck_namespace *src)
550 des->ln_magic = cpu_to_le32(src->ln_magic);
551 des->ln_status = cpu_to_le32(src->ln_status);
552 des->ln_flags = cpu_to_le32(src->ln_flags);
553 des->ln_success_count = cpu_to_le32(src->ln_success_count);
554 des->ln_run_time_phase1 = cpu_to_le32(src->ln_run_time_phase1);
555 des->ln_run_time_phase2 = cpu_to_le32(src->ln_run_time_phase2);
556 des->ln_time_last_complete = cpu_to_le64(src->ln_time_last_complete);
557 des->ln_time_latest_start = cpu_to_le64(src->ln_time_latest_start);
558 des->ln_time_last_checkpoint =
559 cpu_to_le64(src->ln_time_last_checkpoint);
560 mdd_lfsck_position_to_le(&des->ln_pos_latest_start,
561 &src->ln_pos_latest_start);
562 mdd_lfsck_position_to_le(&des->ln_pos_last_checkpoint,
563 &src->ln_pos_last_checkpoint);
564 mdd_lfsck_position_to_le(&des->ln_pos_first_inconsistent,
565 &src->ln_pos_first_inconsistent);
566 des->ln_items_checked = cpu_to_le64(src->ln_items_checked);
567 des->ln_items_repaired = cpu_to_le64(src->ln_items_repaired);
568 des->ln_items_failed = cpu_to_le64(src->ln_items_failed);
569 des->ln_dirs_checked = cpu_to_le64(src->ln_dirs_checked);
570 des->ln_mlinked_checked = cpu_to_le64(src->ln_mlinked_checked);
571 des->ln_objs_checked_phase2 = cpu_to_le64(src->ln_objs_checked_phase2);
572 des->ln_objs_repaired_phase2 =
573 cpu_to_le64(src->ln_objs_repaired_phase2);
574 des->ln_objs_failed_phase2 = cpu_to_le64(src->ln_objs_failed_phase2);
575 des->ln_objs_nlink_repaired = cpu_to_le64(src->ln_objs_nlink_repaired);
576 des->ln_objs_lost_found = cpu_to_le64(src->ln_objs_lost_found);
577 fid_cpu_to_le(&des->ln_fid_latest_scanned_phase2,
578 &src->ln_fid_latest_scanned_phase2);
582 * \retval +ve: the lfsck_namespace is broken, the caller should reset it.
583 * \retval 0: succeed.
584 * \retval -ve: failed cases.
586 static int mdd_lfsck_namespace_load(const struct lu_env *env,
587 struct lfsck_component *com)
589 int len = com->lc_file_size;
592 rc = dt_xattr_get(env, com->lc_obj,
593 mdd_buf_get(env, com->lc_file_disk, len),
594 XATTR_NAME_LFSCK_NAMESPACE, BYPASS_CAPA);
596 struct lfsck_namespace *ns = com->lc_file_ram;
598 mdd_lfsck_namespace_to_cpu(ns,
599 (struct lfsck_namespace *)com->lc_file_disk);
600 if (ns->ln_magic != LFSCK_NAMESPACE_MAGIC) {
601 CWARN("%.16s: invalid lfsck_namespace magic "
603 mdd_lfsck2name(com->lc_lfsck),
604 ns->ln_magic, LFSCK_NAMESPACE_MAGIC);
609 } else if (rc != -ENODATA) {
610 CERROR("%.16s: fail to load lfsck_namespace, expected = %d, "
611 "rc = %d\n", mdd_lfsck2name(com->lc_lfsck), len, rc);
618 static int mdd_lfsck_namespace_store(const struct lu_env *env,
619 struct lfsck_component *com, bool init)
621 struct dt_object *obj = com->lc_obj;
622 struct md_lfsck *lfsck = com->lc_lfsck;
623 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
624 struct thandle *handle;
625 int len = com->lc_file_size;
629 mdd_lfsck_namespace_to_le((struct lfsck_namespace *)com->lc_file_disk,
630 (struct lfsck_namespace *)com->lc_file_ram);
631 handle = dt_trans_create(env, mdd->mdd_bottom);
632 if (IS_ERR(handle)) {
633 rc = PTR_ERR(handle);
634 CERROR("%.16s: fail to create trans for storing "
635 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
639 rc = dt_declare_xattr_set(env, obj,
640 mdd_buf_get(env, com->lc_file_disk, len),
641 XATTR_NAME_LFSCK_NAMESPACE, 0, handle);
643 CERROR("%.16s: fail to declare trans for storing "
644 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
648 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
650 CERROR("%.16s: fail to start trans for storing "
651 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
655 rc = dt_xattr_set(env, obj,
656 mdd_buf_get(env, com->lc_file_disk, len),
657 XATTR_NAME_LFSCK_NAMESPACE,
658 init ? LU_XATTR_CREATE : LU_XATTR_REPLACE,
659 handle, BYPASS_CAPA);
661 CERROR("%.16s: fail to store lfsck_namespace, len = %d, "
662 "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
667 dt_trans_stop(env, mdd->mdd_bottom, handle);
671 static int mdd_lfsck_namespace_init(const struct lu_env *env,
672 struct lfsck_component *com)
674 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
677 memset(ns, 0, sizeof(*ns));
678 ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
679 ns->ln_status = LS_INIT;
680 down_write(&com->lc_sem);
681 rc = mdd_lfsck_namespace_store(env, com, true);
682 up_write(&com->lc_sem);
686 static int mdd_declare_lfsck_namespace_unlink(const struct lu_env *env,
687 struct mdd_device *mdd,
691 struct thandle *handle)
695 rc = dt_declare_delete(env, p, (const struct dt_key *)name, handle);
699 rc = dt_declare_ref_del(env, c, handle);
703 rc = dt_declare_destroy(env, c, handle);
707 static int mdd_lfsck_namespace_unlink(const struct lu_env *env,
708 struct mdd_device *mdd,
709 struct lfsck_component *com)
711 struct mdd_thread_info *info = mdd_env_info(env);
712 struct lu_fid *fid = &info->mti_fid;
713 struct dt_object *child = com->lc_obj;
714 struct dt_object *parent;
715 struct thandle *handle;
720 parent = dt_store_resolve(env, mdd->mdd_bottom, "", fid);
722 RETURN(rc = PTR_ERR(parent));
724 if (!dt_try_as_dir(env, parent))
725 GOTO(out, rc = -ENOTDIR);
727 handle = dt_trans_create(env, mdd->mdd_bottom);
729 GOTO(out, rc = PTR_ERR(handle));
731 rc = mdd_declare_lfsck_namespace_unlink(env, mdd, parent, child,
732 lfsck_namespace_name, handle);
736 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
740 dt_write_lock(env, child, MOR_TGT_CHILD);
742 rc = dt_delete(env, parent, (struct dt_key *)lfsck_namespace_name,
743 handle, BYPASS_CAPA);
747 rc = child->do_ops->do_ref_del(env, child, handle);
749 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
750 rc = dt_insert(env, parent,
751 (const struct dt_rec*)fid,
752 (const struct dt_key *)lfsck_namespace_name,
753 handle, BYPASS_CAPA, 1);
759 rc = dt_destroy(env, child, handle);
765 dt_write_unlock(env, child);
768 lu_object_put(env, &child->do_lu);
772 dt_trans_stop(env, mdd->mdd_bottom, handle);
775 lu_object_put(env, &parent->do_lu);
779 static int mdd_lfsck_namespace_lookup(const struct lu_env *env,
780 struct lfsck_component *com,
781 const struct lu_fid *fid,
784 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
787 fid_cpu_to_be(key, fid);
788 rc = dt_lookup(env, com->lc_obj, (struct dt_rec *)flags,
789 (const struct dt_key *)key, BYPASS_CAPA);
793 static int mdd_lfsck_namespace_delete(const struct lu_env *env,
794 struct lfsck_component *com,
795 const struct lu_fid *fid)
797 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
798 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
799 struct thandle *handle;
800 struct dt_object *obj = com->lc_obj;
804 handle = dt_trans_create(env, mdd->mdd_bottom);
806 RETURN(PTR_ERR(handle));
808 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid, handle);
812 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
816 fid_cpu_to_be(key, fid);
817 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
823 dt_trans_stop(env, mdd->mdd_bottom, handle);
827 static int mdd_lfsck_namespace_update(const struct lu_env *env,
828 struct lfsck_component *com,
829 const struct lu_fid *fid,
830 __u8 flags, bool force)
832 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
833 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
834 struct thandle *handle;
835 struct dt_object *obj = com->lc_obj;
841 rc = mdd_lfsck_namespace_lookup(env, com, fid, &tf);
842 if (rc != 0 && rc != -ENOENT)
846 if (!force || flags == tf)
850 handle = dt_trans_create(env, mdd->mdd_bottom);
852 RETURN(PTR_ERR(handle));
854 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid,
859 handle = dt_trans_create(env, mdd->mdd_bottom);
861 RETURN(PTR_ERR(handle));
864 rc = dt_declare_insert(env, obj, (const struct dt_rec *)&flags,
865 (const struct dt_key *)fid, handle);
869 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
873 fid_cpu_to_be(key, fid);
875 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
878 CERROR("%s: fail to insert "DFID", rc = %d\n",
879 mdd_lfsck2name(com->lc_lfsck), PFID(fid), rc);
884 rc = dt_insert(env, obj, (const struct dt_rec *)&flags,
885 (const struct dt_key *)key, handle, BYPASS_CAPA, 1);
890 dt_trans_stop(env, mdd->mdd_bottom, handle);
895 * \retval +ve repaired
896 * \retval 0 no need to repair
897 * \retval -ve error cases
899 static int mdd_lfsck_namespace_double_scan_one(const struct lu_env *env,
900 struct lfsck_component *com,
901 struct mdd_object *child,
904 struct mdd_thread_info *info = mdd_env_info(env);
905 struct lu_attr *la = &info->mti_la;
906 struct lu_name *cname = &info->mti_name;
907 struct lu_fid *pfid = &info->mti_fid;
908 struct lu_fid *cfid = &info->mti_fid2;
909 struct md_lfsck *lfsck = com->lc_lfsck;
910 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
911 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
912 struct lfsck_namespace *ns =
913 (struct lfsck_namespace *)com->lc_file_ram;
914 struct linkea_data ldata = { 0 };
915 struct thandle *handle = NULL;
922 if (com->lc_journal) {
928 handle = mdd_trans_create(env, mdd);
930 RETURN(rc = PTR_ERR(handle));
932 rc = mdd_declare_links_add(env, child, handle, NULL);
936 rc = mdd_trans_start(env, mdd, handle);
940 mdd_write_lock(env, child, MOR_TGT_CHILD);
944 if (unlikely(mdd_is_dead_obj(child)))
947 rc = mdd_links_read(env, child, &ldata);
949 if ((bk->lb_param & LPF_DRYRUN) &&
950 (rc == -EINVAL || rc == -ENODATA))
956 rc = mdd_la_get(env, child, la, BYPASS_CAPA);
960 ldata.ld_lee = LINKEA_FIRST_ENTRY(ldata);
961 count = ldata.ld_leh->leh_reccount;
962 while (count-- > 0) {
963 struct mdd_object *parent = NULL;
964 struct dt_object *dir;
966 linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, cname,
968 if (!fid_is_sane(pfid))
971 parent = mdd_object_find(env, mdd, pfid);
974 else if (IS_ERR(parent))
975 GOTO(stop, rc = PTR_ERR(parent));
977 if (!mdd_object_exists(parent))
980 /* XXX: need more processing for remote object in the future. */
981 if (mdd_object_remote(parent)) {
982 mdd_object_put(env, parent);
983 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
987 dir = mdd_object_child(parent);
988 if (unlikely(!dt_try_as_dir(env, dir)))
991 /* To guarantee the 'name' is terminated with '0'. */
992 memcpy(info->mti_key, cname->ln_name, cname->ln_namelen);
993 info->mti_key[cname->ln_namelen] = 0;
994 cname->ln_name = info->mti_key;
995 rc = dt_lookup(env, dir, (struct dt_rec *)cfid,
996 (const struct dt_key *)cname->ln_name,
998 if (rc != 0 && rc != -ENOENT) {
999 mdd_object_put(env, parent);
1004 if (lu_fid_eq(cfid, mdo2fid(child))) {
1005 mdd_object_put(env, parent);
1006 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
1013 if (ldata.ld_leh->leh_reccount > la->la_nlink)
1016 /* XXX: For the case of there is linkea entry, but without name
1017 * entry pointing to the object, and the object link count
1018 * isn't less than the count of name entries, then add the
1019 * name entry back to namespace.
1021 * It is out of LFSCK 1.5 scope, will implement it in the
1022 * future. Keep the linkEA entry. */
1023 mdd_object_put(env, parent);
1024 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
1029 mdd_object_put(env, parent);
1030 if (bk->lb_param & LPF_DRYRUN)
1033 CDEBUG(D_LFSCK, "Remove linkEA: "DFID"[%.*s], "DFID"\n",
1034 PFID(mdo2fid(child)), cname->ln_namelen, cname->ln_name,
1036 linkea_del_buf(&ldata, cname);
1041 if (!com->lc_journal) {
1042 com->lc_journal = 1;
1046 rc = mdd_links_write(env, child, &ldata, handle);
1053 mdd_write_unlock(env, child);
1056 mdd_trans_stop(env, mdd, rc, handle);
1058 if (rc == 0 && update) {
1059 ns->ln_objs_nlink_repaired++;
1065 /* namespace APIs */
1067 static int mdd_lfsck_namespace_reset(const struct lu_env *env,
1068 struct lfsck_component *com, bool init)
1070 struct mdd_thread_info *info = mdd_env_info(env);
1071 struct lu_fid *fid = &info->mti_fid;
1072 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
1073 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
1074 struct md_object *mdo;
1075 struct dt_object *dto;
1079 down_write(&com->lc_sem);
1081 memset(ns, 0, sizeof(*ns));
1083 __u32 count = ns->ln_success_count;
1084 __u64 last_time = ns->ln_time_last_complete;
1086 memset(ns, 0, sizeof(*ns));
1087 ns->ln_success_count = count;
1088 ns->ln_time_last_complete = last_time;
1090 ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
1091 ns->ln_status = LS_INIT;
1093 rc = mdd_lfsck_namespace_unlink(env, mdd, com);
1097 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
1098 mdo = llo_store_create_index(env, &mdd->mdd_md_dev, mdd->mdd_bottom, "",
1099 lfsck_namespace_name, fid,
1100 &dt_lfsck_features);
1102 GOTO(out, rc = PTR_ERR(mdo));
1104 lu_object_put(env, &mdo->mo_lu);
1105 dto = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name, fid);
1107 GOTO(out, rc = PTR_ERR(dto));
1110 rc = dto->do_ops->do_index_try(env, dto, &dt_lfsck_features);
1114 rc = mdd_lfsck_namespace_store(env, com, true);
1119 up_write(&com->lc_sem);
1124 mdd_lfsck_namespace_fail(const struct lu_env *env, struct lfsck_component *com,
1125 bool oit, bool new_checked)
1127 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
1129 down_write(&com->lc_sem);
1131 com->lc_new_checked++;
1132 ns->ln_items_failed++;
1133 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1134 mdd_lfsck_pos_fill(env, com->lc_lfsck,
1135 &ns->ln_pos_first_inconsistent, oit, !oit);
1136 up_write(&com->lc_sem);
1139 static int mdd_lfsck_namespace_checkpoint(const struct lu_env *env,
1140 struct lfsck_component *com,
1143 struct md_lfsck *lfsck = com->lc_lfsck;
1144 struct lfsck_namespace *ns =
1145 (struct lfsck_namespace *)com->lc_file_ram;
1148 if (com->lc_new_checked == 0 && !init)
1151 down_write(&com->lc_sem);
1153 ns->ln_pos_last_checkpoint = lfsck->ml_pos_current;
1155 ns->ln_time_last_checkpoint = ns->ln_time_latest_start;
1156 ns->ln_pos_latest_start = lfsck->ml_pos_current;
1158 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1159 HALF_SEC - lfsck->ml_time_last_checkpoint);
1160 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1161 ns->ln_items_checked += com->lc_new_checked;
1162 com->lc_new_checked = 0;
1165 rc = mdd_lfsck_namespace_store(env, com, false);
1167 up_write(&com->lc_sem);
1171 static int mdd_lfsck_namespace_prep(const struct lu_env *env,
1172 struct lfsck_component *com)
1174 struct md_lfsck *lfsck = com->lc_lfsck;
1175 struct lfsck_namespace *ns =
1176 (struct lfsck_namespace *)com->lc_file_ram;
1177 struct lfsck_position *pos = &com->lc_pos_start;
1179 if (ns->ln_status == LS_COMPLETED) {
1182 rc = mdd_lfsck_namespace_reset(env, com, false);
1187 down_write(&com->lc_sem);
1189 ns->ln_time_latest_start = cfs_time_current_sec();
1191 spin_lock(&lfsck->ml_lock);
1192 if (ns->ln_flags & LF_SCANNED_ONCE) {
1193 if (!lfsck->ml_drop_dryrun ||
1194 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1195 ns->ln_status = LS_SCANNING_PHASE2;
1196 cfs_list_del_init(&com->lc_link);
1197 cfs_list_add_tail(&com->lc_link,
1198 &lfsck->ml_list_double_scan);
1199 if (!cfs_list_empty(&com->lc_link_dir))
1200 cfs_list_del_init(&com->lc_link_dir);
1201 mdd_lfsck_pos_set_zero(pos);
1203 ns->ln_status = LS_SCANNING_PHASE1;
1204 ns->ln_run_time_phase1 = 0;
1205 ns->ln_run_time_phase2 = 0;
1206 ns->ln_items_checked = 0;
1207 ns->ln_items_repaired = 0;
1208 ns->ln_items_failed = 0;
1209 ns->ln_dirs_checked = 0;
1210 ns->ln_mlinked_checked = 0;
1211 ns->ln_objs_checked_phase2 = 0;
1212 ns->ln_objs_repaired_phase2 = 0;
1213 ns->ln_objs_failed_phase2 = 0;
1214 ns->ln_objs_nlink_repaired = 0;
1215 ns->ln_objs_lost_found = 0;
1216 fid_zero(&ns->ln_fid_latest_scanned_phase2);
1217 if (cfs_list_empty(&com->lc_link_dir))
1218 cfs_list_add_tail(&com->lc_link_dir,
1219 &lfsck->ml_list_dir);
1220 *pos = ns->ln_pos_first_inconsistent;
1223 ns->ln_status = LS_SCANNING_PHASE1;
1224 if (cfs_list_empty(&com->lc_link_dir))
1225 cfs_list_add_tail(&com->lc_link_dir,
1226 &lfsck->ml_list_dir);
1227 if (!lfsck->ml_drop_dryrun ||
1228 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1229 *pos = ns->ln_pos_last_checkpoint;
1230 pos->lp_oit_cookie++;
1231 if (!fid_is_zero(&pos->lp_dir_parent)) {
1232 if (pos->lp_dir_cookie == MDS_DIR_END_OFF) {
1233 fid_zero(&pos->lp_dir_parent);
1235 pos->lp_dir_cookie++;
1239 *pos = ns->ln_pos_first_inconsistent;
1242 spin_unlock(&lfsck->ml_lock);
1244 up_write(&com->lc_sem);
1248 static int mdd_lfsck_namespace_exec_oit(const struct lu_env *env,
1249 struct lfsck_component *com,
1250 struct mdd_object *obj)
1252 down_write(&com->lc_sem);
1253 com->lc_new_checked++;
1254 if (S_ISDIR(mdd_object_type(obj)))
1255 ((struct lfsck_namespace *)com->lc_file_ram)->ln_dirs_checked++;
1256 up_write(&com->lc_sem);
1260 static int mdd_declare_lfsck_namespace_exec_dir(const struct lu_env *env,
1261 struct mdd_object *obj,
1262 struct thandle *handle)
1266 /* For destroying all invalid linkEA entries. */
1267 rc = mdo_declare_xattr_del(env, obj, XATTR_NAME_LINK, handle);
1271 /* For insert new linkEA entry. */
1272 rc = mdd_declare_links_add(env, obj, handle, NULL);
1276 static int mdd_lfsck_namespace_check_exist(const struct lu_env *env,
1277 struct md_lfsck *lfsck,
1278 struct mdd_object *obj,
1281 struct dt_object *dir = lfsck->ml_obj_dir;
1282 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
1286 if (unlikely(mdd_is_dead_obj(obj)))
1287 RETURN(LFSCK_NAMEENTRY_DEAD);
1289 rc = dt_lookup(env, dir, (struct dt_rec *)fid,
1290 (const struct dt_key *)name, BYPASS_CAPA);
1292 RETURN(LFSCK_NAMEENTRY_REMOVED);
1297 if (!lu_fid_eq(fid, mdo2fid(obj)))
1298 RETURN(LFSCK_NAMEENTRY_RECREATED);
1303 static int mdd_lfsck_namespace_exec_dir(const struct lu_env *env,
1304 struct lfsck_component *com,
1305 struct mdd_object *obj,
1306 struct lu_dirent *ent)
1308 struct mdd_thread_info *info = mdd_env_info(env);
1309 struct lu_attr *la = &info->mti_la;
1310 struct md_lfsck *lfsck = com->lc_lfsck;
1311 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1312 struct lfsck_namespace *ns =
1313 (struct lfsck_namespace *)com->lc_file_ram;
1314 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1315 struct linkea_data ldata = { 0 };
1316 const struct lu_fid *pfid =
1317 lu_object_fid(&lfsck->ml_obj_dir->do_lu);
1318 const struct lu_fid *cfid = mdo2fid(obj);
1319 const struct lu_name *cname;
1320 struct thandle *handle = NULL;
1321 bool repaired = false;
1322 bool locked = false;
1327 cname = mdd_name_get_const(env, ent->lde_name, ent->lde_namelen);
1328 down_write(&com->lc_sem);
1329 com->lc_new_checked++;
1331 if (ent->lde_attrs & LUDA_UPGRADE) {
1332 ns->ln_flags |= LF_UPGRADE;
1334 } else if (ent->lde_attrs & LUDA_REPAIR) {
1335 ns->ln_flags |= LF_INCONSISTENT;
1339 if (ent->lde_name[0] == '.' &&
1340 (ent->lde_namelen == 1 ||
1341 (ent->lde_namelen == 2 && ent->lde_name[1] == '.')))
1344 if (!(bk->lb_param & LPF_DRYRUN) &&
1345 (com->lc_journal || repaired)) {
1350 com->lc_journal = 1;
1351 handle = mdd_trans_create(env, mdd);
1353 GOTO(out, rc = PTR_ERR(handle));
1355 rc = mdd_declare_lfsck_namespace_exec_dir(env, obj, handle);
1359 rc = mdd_trans_start(env, mdd, handle);
1363 mdd_write_lock(env, obj, MOR_TGT_CHILD);
1367 rc = mdd_lfsck_namespace_check_exist(env, lfsck, obj, ent->lde_name);
1371 rc = mdd_links_read(env, obj, &ldata);
1373 count = ldata.ld_leh->leh_reccount;
1374 rc = linkea_links_find(&ldata, cname, pfid);
1376 /* For dir, if there are more than one linkea entries,
1377 * then remove all the other redundant linkea entries.*/
1378 if (unlikely(count > 1 &&
1379 S_ISDIR(mdd_object_type(obj))))
1386 ns->ln_flags |= LF_INCONSISTENT;
1387 if (bk->lb_param & LPF_DRYRUN) {
1392 /*For dir, remove the unmatched linkea entry directly.*/
1393 if (S_ISDIR(mdd_object_type(obj))) {
1394 if (!com->lc_journal)
1397 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK,
1398 handle, BYPASS_CAPA);
1407 } else if (unlikely(rc == -EINVAL)) {
1408 ns->ln_flags |= LF_INCONSISTENT;
1409 if (bk->lb_param & LPF_DRYRUN) {
1415 if (!com->lc_journal)
1418 /* The magic crashed, we are not sure whether there are more
1419 * corrupt data in the linkea, so remove all linkea entries. */
1420 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK, handle,
1426 } else if (rc == -ENODATA) {
1427 ns->ln_flags |= LF_UPGRADE;
1428 if (bk->lb_param & LPF_DRYRUN) {
1435 rc = linkea_data_new(&ldata, &mdd_env_info(env)->mti_link_buf);
1440 if (!com->lc_journal)
1443 rc = linkea_add_buf(&ldata, cname, pfid);
1447 rc = mdd_links_write(env, obj, &ldata, handle);
1451 count = ldata.ld_leh->leh_reccount;
1460 rc = mdd_la_get(env, obj, la, BYPASS_CAPA);
1465 (la->la_nlink == 1 || S_ISDIR(mdd_object_type(obj))))
1466 /* Usually, it is for single linked object or dir, do nothing.*/
1469 /* Following modification will be in another transaction. */
1470 if (handle != NULL) {
1471 LASSERT(mdd_write_locked(env, obj));
1473 mdd_write_unlock(env, obj);
1476 mdd_trans_stop(env, mdd, 0, handle);
1480 ns->ln_mlinked_checked++;
1481 rc = mdd_lfsck_namespace_update(env, com, cfid,
1482 count != la->la_nlink ? LLF_UNMATCH_NLINKS : 0, false);
1488 mdd_write_unlock(env, obj);
1491 mdd_trans_stop(env, mdd, rc, handle);
1495 ns->ln_items_failed++;
1496 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1497 mdd_lfsck_pos_fill(env, lfsck,
1498 &ns->ln_pos_first_inconsistent,
1500 if (!(bk->lb_param & LPF_FAILOUT))
1504 ns->ln_items_repaired++;
1506 com->lc_journal = 0;
1509 up_write(&com->lc_sem);
1513 static int mdd_lfsck_namespace_post(const struct lu_env *env,
1514 struct lfsck_component *com,
1517 struct md_lfsck *lfsck = com->lc_lfsck;
1518 struct lfsck_namespace *ns =
1519 (struct lfsck_namespace *)com->lc_file_ram;
1522 down_write(&com->lc_sem);
1524 spin_lock(&lfsck->ml_lock);
1526 ns->ln_status = LS_SCANNING_PHASE2;
1527 ns->ln_flags |= LF_SCANNED_ONCE;
1528 ns->ln_flags &= ~LF_UPGRADE;
1529 cfs_list_del_init(&com->lc_link);
1530 cfs_list_del_init(&com->lc_link_dir);
1531 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_double_scan);
1532 } else if (result == 0) {
1533 if (lfsck->ml_paused) {
1534 ns->ln_status = LS_PAUSED;
1536 ns->ln_status = LS_STOPPED;
1537 cfs_list_del_init(&com->lc_link);
1538 cfs_list_del_init(&com->lc_link_dir);
1539 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1542 ns->ln_status = LS_FAILED;
1543 cfs_list_del_init(&com->lc_link);
1544 cfs_list_del_init(&com->lc_link_dir);
1545 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1547 spin_unlock(&lfsck->ml_lock);
1549 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1550 HALF_SEC - lfsck->ml_time_last_checkpoint);
1551 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1552 ns->ln_items_checked += com->lc_new_checked;
1553 com->lc_new_checked = 0;
1555 rc = mdd_lfsck_namespace_store(env, com, false);
1557 up_write(&com->lc_sem);
1562 mdd_lfsck_namespace_dump(const struct lu_env *env, struct lfsck_component *com,
1565 struct md_lfsck *lfsck = com->lc_lfsck;
1566 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1567 struct lfsck_namespace *ns =
1568 (struct lfsck_namespace *)com->lc_file_ram;
1573 down_read(&com->lc_sem);
1574 rc = snprintf(buf, len,
1575 "name: lfsck_namespace\n"
1581 lfsck_status_names[ns->ln_status]);
1587 rc = lfsck_bits_dump(&buf, &len, ns->ln_flags, lfsck_flags_names,
1592 rc = lfsck_bits_dump(&buf, &len, bk->lb_param, lfsck_param_names,
1597 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_complete,
1598 "time_since_last_completed");
1602 rc = lfsck_time_dump(&buf, &len, ns->ln_time_latest_start,
1603 "time_since_latest_start");
1607 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_checkpoint,
1608 "time_since_last_checkpoint");
1612 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_latest_start,
1613 "latest_start_position");
1617 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_last_checkpoint,
1618 "last_checkpoint_position");
1622 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_first_inconsistent,
1623 "first_failure_position");
1627 if (ns->ln_status == LS_SCANNING_PHASE1) {
1628 struct lfsck_position pos;
1629 cfs_duration_t duration = cfs_time_current() -
1630 lfsck->ml_time_last_checkpoint;
1631 __u64 checked = ns->ln_items_checked + com->lc_new_checked;
1632 __u64 speed = checked;
1633 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1634 __u32 rtime = ns->ln_run_time_phase1 +
1635 cfs_duration_sec(duration + HALF_SEC);
1638 do_div(new_checked, duration);
1640 do_div(speed, rtime);
1641 rc = snprintf(buf, len,
1642 "checked_phase1: "LPU64"\n"
1643 "checked_phase2: "LPU64"\n"
1644 "updated_phase1: "LPU64"\n"
1645 "updated_phase2: "LPU64"\n"
1646 "failed_phase1: "LPU64"\n"
1647 "failed_phase2: "LPU64"\n"
1649 "M-linked: "LPU64"\n"
1650 "nlinks_repaired: "LPU64"\n"
1651 "lost_found: "LPU64"\n"
1652 "success_count: %u\n"
1653 "run_time_phase1: %u seconds\n"
1654 "run_time_phase2: %u seconds\n"
1655 "average_speed_phase1: "LPU64" items/sec\n"
1656 "average_speed_phase2: N/A\n"
1657 "real-time_speed_phase1: "LPU64" items/sec\n"
1658 "real-time_speed_phase2: N/A\n",
1660 ns->ln_objs_checked_phase2,
1661 ns->ln_items_repaired,
1662 ns->ln_objs_repaired_phase2,
1663 ns->ln_items_failed,
1664 ns->ln_objs_failed_phase2,
1665 ns->ln_dirs_checked,
1666 ns->ln_mlinked_checked,
1667 ns->ln_objs_nlink_repaired,
1668 ns->ln_objs_lost_found,
1669 ns->ln_success_count,
1671 ns->ln_run_time_phase2,
1679 mdd_lfsck_pos_fill(env, lfsck, &pos, true, true);
1680 rc = lfsck_pos_dump(&buf, &len, &pos, "current_position");
1683 } else if (ns->ln_status == LS_SCANNING_PHASE2) {
1684 cfs_duration_t duration = cfs_time_current() -
1685 lfsck->ml_time_last_checkpoint;
1686 __u64 checked = ns->ln_objs_checked_phase2 +
1687 com->lc_new_checked;
1688 __u64 speed1 = ns->ln_items_checked;
1689 __u64 speed2 = checked;
1690 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1691 __u32 rtime = ns->ln_run_time_phase2 +
1692 cfs_duration_sec(duration + HALF_SEC);
1695 do_div(new_checked, duration);
1696 if (ns->ln_run_time_phase1 != 0)
1697 do_div(speed1, ns->ln_run_time_phase1);
1699 do_div(speed2, rtime);
1700 rc = snprintf(buf, len,
1701 "checked_phase1: "LPU64"\n"
1702 "checked_phase2: "LPU64"\n"
1703 "updated_phase1: "LPU64"\n"
1704 "updated_phase2: "LPU64"\n"
1705 "failed_phase1: "LPU64"\n"
1706 "failed_phase2: "LPU64"\n"
1708 "M-linked: "LPU64"\n"
1709 "nlinks_repaired: "LPU64"\n"
1710 "lost_found: "LPU64"\n"
1711 "success_count: %u\n"
1712 "run_time_phase1: %u seconds\n"
1713 "run_time_phase2: %u seconds\n"
1714 "average_speed_phase1: "LPU64" items/sec\n"
1715 "average_speed_phase2: "LPU64" objs/sec\n"
1716 "real-time_speed_phase1: N/A\n"
1717 "real-time_speed_phase2: "LPU64" objs/sec\n"
1718 "current_position: "DFID"\n",
1719 ns->ln_items_checked,
1721 ns->ln_items_repaired,
1722 ns->ln_objs_repaired_phase2,
1723 ns->ln_items_failed,
1724 ns->ln_objs_failed_phase2,
1725 ns->ln_dirs_checked,
1726 ns->ln_mlinked_checked,
1727 ns->ln_objs_nlink_repaired,
1728 ns->ln_objs_lost_found,
1729 ns->ln_success_count,
1730 ns->ln_run_time_phase1,
1735 PFID(&ns->ln_fid_latest_scanned_phase2));
1742 __u64 speed1 = ns->ln_items_checked;
1743 __u64 speed2 = ns->ln_objs_checked_phase2;
1745 if (ns->ln_run_time_phase1 != 0)
1746 do_div(speed1, ns->ln_run_time_phase1);
1747 if (ns->ln_run_time_phase2 != 0)
1748 do_div(speed2, ns->ln_run_time_phase2);
1749 rc = snprintf(buf, len,
1750 "checked_phase1: "LPU64"\n"
1751 "checked_phase2: "LPU64"\n"
1752 "updated_phase1: "LPU64"\n"
1753 "updated_phase2: "LPU64"\n"
1754 "failed_phase1: "LPU64"\n"
1755 "failed_phase2: "LPU64"\n"
1757 "M-linked: "LPU64"\n"
1758 "nlinks_repaired: "LPU64"\n"
1759 "lost_found: "LPU64"\n"
1760 "success_count: %u\n"
1761 "run_time_phase1: %u seconds\n"
1762 "run_time_phase2: %u seconds\n"
1763 "average_speed_phase1: "LPU64" items/sec\n"
1764 "average_speed_phase2: "LPU64" objs/sec\n"
1765 "real-time_speed_phase1: N/A\n"
1766 "real-time_speed_phase2: N/A\n"
1767 "current_position: N/A\n",
1768 ns->ln_items_checked,
1769 ns->ln_objs_checked_phase2,
1770 ns->ln_items_repaired,
1771 ns->ln_objs_repaired_phase2,
1772 ns->ln_items_failed,
1773 ns->ln_objs_failed_phase2,
1774 ns->ln_dirs_checked,
1775 ns->ln_mlinked_checked,
1776 ns->ln_objs_nlink_repaired,
1777 ns->ln_objs_lost_found,
1778 ns->ln_success_count,
1779 ns->ln_run_time_phase1,
1780 ns->ln_run_time_phase2,
1792 up_read(&com->lc_sem);
1796 static int mdd_lfsck_namespace_double_scan(const struct lu_env *env,
1797 struct lfsck_component *com)
1799 struct md_lfsck *lfsck = com->lc_lfsck;
1800 struct ptlrpc_thread *thread = &lfsck->ml_thread;
1801 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1802 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1803 struct lfsck_namespace *ns =
1804 (struct lfsck_namespace *)com->lc_file_ram;
1805 struct dt_object *obj = com->lc_obj;
1806 const struct dt_it_ops *iops = &obj->do_index_ops->dio_it;
1807 struct mdd_object *target;
1815 lfsck->ml_new_scanned = 0;
1816 lfsck->ml_time_last_checkpoint = cfs_time_current();
1817 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1818 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1820 di = iops->init(env, obj, 0, BYPASS_CAPA);
1822 RETURN(PTR_ERR(di));
1824 fid_cpu_to_be(&fid, &ns->ln_fid_latest_scanned_phase2);
1825 rc = iops->get(env, di, (const struct dt_key *)&fid);
1829 /* Skip the start one, which either has been processed or non-exist. */
1830 rc = iops->next(env, di);
1834 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_DOUBLESCAN))
1838 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY3) &&
1840 struct l_wait_info lwi;
1842 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
1844 l_wait_event(thread->t_ctl_waitq,
1845 !thread_is_running(thread),
1849 key = iops->key(env, di);
1850 fid_be_to_cpu(&fid, (const struct lu_fid *)key);
1851 target = mdd_object_find(env, mdd, &fid);
1852 down_write(&com->lc_sem);
1853 if (target == NULL) {
1856 } else if (IS_ERR(target)) {
1857 rc = PTR_ERR(target);
1861 /* XXX: need more processing for remote object in the future. */
1862 if (!mdd_object_exists(target) || mdd_object_remote(target))
1865 rc = iops->rec(env, di, (struct dt_rec *)&flags, 0);
1867 rc = mdd_lfsck_namespace_double_scan_one(env, com,
1871 mdd_object_put(env, target);
1874 lfsck->ml_new_scanned++;
1875 com->lc_new_checked++;
1876 ns->ln_fid_latest_scanned_phase2 = fid;
1878 ns->ln_objs_repaired_phase2++;
1880 ns->ln_objs_failed_phase2++;
1881 up_write(&com->lc_sem);
1883 if ((rc == 0) || ((rc > 0) && !(bk->lb_param & LPF_DRYRUN))) {
1884 mdd_lfsck_namespace_delete(env, com, &fid);
1885 } else if (rc < 0) {
1886 flags |= LLF_REPAIR_FAILED;
1887 mdd_lfsck_namespace_update(env, com, &fid, flags, true);
1890 if (rc < 0 && bk->lb_param & LPF_FAILOUT)
1893 if (likely(cfs_time_beforeq(cfs_time_current(),
1894 lfsck->ml_time_next_checkpoint)) ||
1895 com->lc_new_checked == 0)
1898 down_write(&com->lc_sem);
1899 ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1900 HALF_SEC - lfsck->ml_time_last_checkpoint);
1901 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1902 ns->ln_objs_checked_phase2 += com->lc_new_checked;
1903 com->lc_new_checked = 0;
1904 rc = mdd_lfsck_namespace_store(env, com, false);
1905 up_write(&com->lc_sem);
1909 lfsck->ml_time_last_checkpoint = cfs_time_current();
1910 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1911 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1914 mdd_lfsck_control_speed(lfsck);
1915 if (unlikely(!thread_is_running(thread)))
1918 rc = iops->next(env, di);
1927 iops->fini(env, di);
1928 down_write(&com->lc_sem);
1930 ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1931 HALF_SEC - lfsck->ml_time_last_checkpoint);
1932 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1933 ns->ln_objs_checked_phase2 += com->lc_new_checked;
1934 com->lc_new_checked = 0;
1937 com->lc_journal = 0;
1938 ns->ln_status = LS_COMPLETED;
1939 if (!(bk->lb_param & LPF_DRYRUN))
1941 ~(LF_SCANNED_ONCE | LF_INCONSISTENT | LF_UPGRADE);
1942 ns->ln_time_last_complete = ns->ln_time_last_checkpoint;
1943 ns->ln_success_count++;
1944 } else if (rc == 0) {
1945 if (lfsck->ml_paused)
1946 ns->ln_status = LS_PAUSED;
1948 ns->ln_status = LS_STOPPED;
1950 ns->ln_status = LS_FAILED;
1953 if (ns->ln_status != LS_PAUSED) {
1954 spin_lock(&lfsck->ml_lock);
1955 cfs_list_del_init(&com->lc_link);
1956 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1957 spin_unlock(&lfsck->ml_lock);
1960 rc = mdd_lfsck_namespace_store(env, com, false);
1962 up_write(&com->lc_sem);
1966 static struct lfsck_operations mdd_lfsck_namespace_ops = {
1967 .lfsck_reset = mdd_lfsck_namespace_reset,
1968 .lfsck_fail = mdd_lfsck_namespace_fail,
1969 .lfsck_checkpoint = mdd_lfsck_namespace_checkpoint,
1970 .lfsck_prep = mdd_lfsck_namespace_prep,
1971 .lfsck_exec_oit = mdd_lfsck_namespace_exec_oit,
1972 .lfsck_exec_dir = mdd_lfsck_namespace_exec_dir,
1973 .lfsck_post = mdd_lfsck_namespace_post,
1974 .lfsck_dump = mdd_lfsck_namespace_dump,
1975 .lfsck_double_scan = mdd_lfsck_namespace_double_scan,
1978 /* LFSCK component setup/cleanup functions */
1980 static int mdd_lfsck_namespace_setup(const struct lu_env *env,
1981 struct md_lfsck *lfsck)
1983 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1984 struct lfsck_component *com;
1985 struct lfsck_namespace *ns;
1986 struct dt_object *obj;
1994 CFS_INIT_LIST_HEAD(&com->lc_link);
1995 CFS_INIT_LIST_HEAD(&com->lc_link_dir);
1996 init_rwsem(&com->lc_sem);
1997 atomic_set(&com->lc_ref, 1);
1998 com->lc_lfsck = lfsck;
1999 com->lc_type = LT_NAMESPACE;
2000 com->lc_ops = &mdd_lfsck_namespace_ops;
2001 com->lc_file_size = sizeof(struct lfsck_namespace);
2002 OBD_ALLOC(com->lc_file_ram, com->lc_file_size);
2003 if (com->lc_file_ram == NULL)
2004 GOTO(out, rc = -ENOMEM);
2006 OBD_ALLOC(com->lc_file_disk, com->lc_file_size);
2007 if (com->lc_file_disk == NULL)
2008 GOTO(out, rc = -ENOMEM);
2010 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name,
2011 &mdd_env_info(env)->mti_fid);
2013 GOTO(out, rc = PTR_ERR(obj));
2016 rc = obj->do_ops->do_index_try(env, obj, &dt_lfsck_features);
2020 rc = mdd_lfsck_namespace_load(env, com);
2022 rc = mdd_lfsck_namespace_reset(env, com, true);
2023 else if (rc == -ENODATA)
2024 rc = mdd_lfsck_namespace_init(env, com);
2028 ns = (struct lfsck_namespace *)com->lc_file_ram;
2029 switch (ns->ln_status) {
2034 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
2037 CERROR("%s: unknown status: %u\n",
2038 mdd_lfsck2name(lfsck), ns->ln_status);
2040 case LS_SCANNING_PHASE1:
2041 case LS_SCANNING_PHASE2:
2042 /* No need to store the status to disk right now.
2043 * If the system crashed before the status stored,
2044 * it will be loaded back when next time. */
2045 ns->ln_status = LS_CRASHED;
2049 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_scan);
2050 cfs_list_add_tail(&com->lc_link_dir, &lfsck->ml_list_dir);
2058 mdd_lfsck_component_cleanup(env, com);
2062 /* helper functions for framework */
2064 static int object_is_client_visible(const struct lu_env *env,
2065 struct mdd_device *mdd,
2066 struct mdd_object *obj)
2068 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
2072 LASSERT(S_ISDIR(mdd_object_type(obj)));
2075 if (mdd_is_root(mdd, mdo2fid(obj))) {
2077 mdd_object_put(env, obj);
2081 mdd_read_lock(env, obj, MOR_TGT_CHILD);
2082 if (unlikely(mdd_is_dead_obj(obj))) {
2083 mdd_read_unlock(env, obj);
2085 mdd_object_put(env, obj);
2089 rc = dt_xattr_get(env, mdd_object_child(obj),
2090 mdd_buf_get(env, NULL, 0), XATTR_NAME_LINK,
2092 mdd_read_unlock(env, obj);
2095 mdd_object_put(env, obj);
2099 if (rc < 0 && rc != -ENODATA) {
2101 mdd_object_put(env, obj);
2105 rc = mdd_parent_fid(env, obj, fid);
2107 mdd_object_put(env, obj);
2111 if (unlikely(lu_fid_eq(fid, &mdd->mdd_local_root_fid)))
2114 obj = mdd_object_find(env, mdd, fid);
2117 else if (IS_ERR(obj))
2118 return PTR_ERR(obj);
2120 if (!mdd_object_exists(obj)) {
2121 mdd_object_put(env, obj);
2125 /* Currently, only client visible directory can be remote. */
2126 if (mdd_object_remote(obj)) {
2127 mdd_object_put(env, obj);
2136 static void mdd_lfsck_unpack_ent(struct lu_dirent *ent)
2138 fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid);
2139 ent->lde_hash = le64_to_cpu(ent->lde_hash);
2140 ent->lde_reclen = le16_to_cpu(ent->lde_reclen);
2141 ent->lde_namelen = le16_to_cpu(ent->lde_namelen);
2142 ent->lde_attrs = le32_to_cpu(ent->lde_attrs);
2144 /* Make sure the name is terminated with '0'.
2145 * The data (type) after ent::lde_name maybe
2146 * broken, but we do not care. */
2147 ent->lde_name[ent->lde_namelen] = 0;
2150 /* LFSCK wrap functions */
2152 static void mdd_lfsck_fail(const struct lu_env *env, struct md_lfsck *lfsck,
2153 bool oit, bool new_checked)
2155 struct lfsck_component *com;
2157 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2158 com->lc_ops->lfsck_fail(env, com, oit, new_checked);
2162 static int mdd_lfsck_checkpoint(const struct lu_env *env,
2163 struct md_lfsck *lfsck, bool oit)
2165 struct lfsck_component *com;
2168 if (likely(cfs_time_beforeq(cfs_time_current(),
2169 lfsck->ml_time_next_checkpoint)))
2172 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, oit, !oit);
2173 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2174 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
2179 lfsck->ml_time_last_checkpoint = cfs_time_current();
2180 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2181 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2185 static int mdd_lfsck_prep(struct lu_env *env, struct md_lfsck *lfsck)
2187 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2188 struct mdd_object *obj = NULL;
2189 struct dt_object *dt_obj;
2190 struct lfsck_component *com;
2191 struct lfsck_component *next;
2192 struct lfsck_position *pos = NULL;
2193 const struct dt_it_ops *iops =
2194 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2199 LASSERT(lfsck->ml_obj_dir == NULL);
2200 LASSERT(lfsck->ml_di_dir == NULL);
2202 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2203 com->lc_new_checked = 0;
2204 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2205 com->lc_journal = 0;
2207 rc = com->lc_ops->lfsck_prep(env, com);
2211 if ((pos == NULL) ||
2212 (!mdd_lfsck_pos_is_zero(&com->lc_pos_start) &&
2213 mdd_lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
2214 pos = &com->lc_pos_start;
2217 /* Init otable-based iterator. */
2219 rc = iops->load(env, lfsck->ml_di_oit, 0);
2221 lfsck->ml_oit_over = 1;
2228 rc = iops->load(env, lfsck->ml_di_oit, pos->lp_oit_cookie);
2232 lfsck->ml_oit_over = 1;
2234 if (fid_is_zero(&pos->lp_dir_parent))
2237 /* Find the directory for namespace-based traverse. */
2238 obj = mdd_object_find(env, mdd, &pos->lp_dir_parent);
2241 else if (IS_ERR(obj))
2242 RETURN(PTR_ERR(obj));
2244 /* XXX: need more processing for remote object in the future. */
2245 if (!mdd_object_exists(obj) || mdd_object_remote(obj) ||
2246 unlikely(!S_ISDIR(mdd_object_type(obj))))
2249 if (unlikely(mdd_is_dead_obj(obj)))
2252 dt_obj = mdd_object_child(obj);
2253 if (unlikely(!dt_try_as_dir(env, dt_obj)))
2254 GOTO(out, rc = -ENOTDIR);
2256 /* Init the namespace-based directory traverse. */
2257 iops = &dt_obj->do_index_ops->dio_it;
2258 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2260 GOTO(out, rc = PTR_ERR(di));
2262 rc = iops->load(env, di, pos->lp_dir_cookie);
2264 rc = iops->next(env, di);
2270 iops->fini(env, di);
2274 lfsck->ml_obj_dir = dt_obj;
2275 spin_lock(&lfsck->ml_lock);
2276 lfsck->ml_di_dir = di;
2277 spin_unlock(&lfsck->ml_lock);
2284 mdd_object_put(env, obj);
2287 return (rc > 0 ? 0 : rc);
2289 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, false, false);
2290 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2291 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
2296 lfsck->ml_time_last_checkpoint = cfs_time_current();
2297 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2298 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2302 static int mdd_lfsck_exec_oit(const struct lu_env *env, struct md_lfsck *lfsck,
2303 struct mdd_object *obj)
2305 struct lfsck_component *com;
2306 struct dt_object *dt_obj;
2307 const struct dt_it_ops *iops;
2312 LASSERT(lfsck->ml_obj_dir == NULL);
2314 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2315 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
2320 if (!S_ISDIR(mdd_object_type(obj)) ||
2321 cfs_list_empty(&lfsck->ml_list_dir))
2324 rc = object_is_client_visible(env, mdd_lfsck2mdd(lfsck), obj);
2328 if (unlikely(mdd_is_dead_obj(obj)))
2331 dt_obj = mdd_object_child(obj);
2332 if (unlikely(!dt_try_as_dir(env, dt_obj)))
2333 GOTO(out, rc = -ENOTDIR);
2335 iops = &dt_obj->do_index_ops->dio_it;
2336 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2338 GOTO(out, rc = PTR_ERR(di));
2340 rc = iops->load(env, di, 0);
2342 rc = iops->next(env, di);
2348 iops->fini(env, di);
2352 mdd_object_get(obj);
2353 lfsck->ml_obj_dir = dt_obj;
2354 spin_lock(&lfsck->ml_lock);
2355 lfsck->ml_di_dir = di;
2356 spin_unlock(&lfsck->ml_lock);
2362 mdd_lfsck_fail(env, lfsck, false, false);
2363 return (rc > 0 ? 0 : rc);
2366 static int mdd_lfsck_exec_dir(const struct lu_env *env, struct md_lfsck *lfsck,
2367 struct mdd_object *obj, struct lu_dirent *ent)
2369 struct lfsck_component *com;
2372 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2373 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
2380 static int mdd_lfsck_post(const struct lu_env *env, struct md_lfsck *lfsck,
2383 struct lfsck_component *com;
2384 struct lfsck_component *next;
2387 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, true, true);
2388 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2389 rc = com->lc_ops->lfsck_post(env, com, result);
2394 lfsck->ml_time_last_checkpoint = cfs_time_current();
2395 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2396 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2400 static int mdd_lfsck_double_scan(const struct lu_env *env,
2401 struct md_lfsck *lfsck)
2403 struct lfsck_component *com;
2404 struct lfsck_component *next;
2407 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_double_scan,
2409 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2410 com->lc_journal = 0;
2412 rc = com->lc_ops->lfsck_double_scan(env, com);
2421 static int mdd_lfsck_dir_engine(const struct lu_env *env,
2422 struct md_lfsck *lfsck)
2424 struct mdd_thread_info *info = mdd_env_info(env);
2425 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2426 const struct dt_it_ops *iops =
2427 &lfsck->ml_obj_dir->do_index_ops->dio_it;
2428 struct dt_it *di = lfsck->ml_di_dir;
2429 struct lu_dirent *ent = &info->mti_ent;
2430 struct lu_fid *fid = &info->mti_fid;
2431 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2432 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2437 struct mdd_object *child;
2439 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY2) &&
2441 struct l_wait_info lwi;
2443 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2445 l_wait_event(thread->t_ctl_waitq,
2446 !thread_is_running(thread),
2450 lfsck->ml_new_scanned++;
2451 rc = iops->rec(env, di, (struct dt_rec *)ent,
2452 lfsck->ml_args_dir);
2454 mdd_lfsck_fail(env, lfsck, false, true);
2455 if (bk->lb_param & LPF_FAILOUT)
2461 mdd_lfsck_unpack_ent(ent);
2462 if (ent->lde_attrs & LUDA_IGNORE)
2465 *fid = ent->lde_fid;
2466 child = mdd_object_find(env, mdd, fid);
2467 if (child == NULL) {
2469 } else if (IS_ERR(child)) {
2470 mdd_lfsck_fail(env, lfsck, false, true);
2471 if (bk->lb_param & LPF_FAILOUT)
2472 RETURN(PTR_ERR(child));
2477 /* XXX: need more processing for remote object in the future. */
2478 if (mdd_object_exists(child) && !mdd_object_remote(child))
2479 rc = mdd_lfsck_exec_dir(env, lfsck, child, ent);
2480 mdd_object_put(env, child);
2481 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2485 rc = mdd_lfsck_checkpoint(env, lfsck, false);
2486 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2490 mdd_lfsck_control_speed(lfsck);
2491 if (unlikely(!thread_is_running(thread)))
2494 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL2)) {
2495 spin_lock(&lfsck->ml_lock);
2496 thread_set_flags(thread, SVC_STOPPING);
2497 spin_unlock(&lfsck->ml_lock);
2501 rc = iops->next(env, di);
2504 if (rc > 0 && !lfsck->ml_oit_over)
2505 mdd_lfsck_close_dir(env, lfsck);
2510 static int mdd_lfsck_oit_engine(const struct lu_env *env,
2511 struct md_lfsck *lfsck)
2513 struct mdd_thread_info *info = mdd_env_info(env);
2514 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2515 const struct dt_it_ops *iops =
2516 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2517 struct dt_it *di = lfsck->ml_di_oit;
2518 struct lu_fid *fid = &info->mti_fid;
2519 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2520 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2525 struct mdd_object *target;
2527 if (lfsck->ml_di_dir != NULL) {
2528 rc = mdd_lfsck_dir_engine(env, lfsck);
2533 if (unlikely(lfsck->ml_oit_over))
2536 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY1) &&
2538 struct l_wait_info lwi;
2540 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2542 l_wait_event(thread->t_ctl_waitq,
2543 !thread_is_running(thread),
2547 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2550 lfsck->ml_new_scanned++;
2551 rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
2553 mdd_lfsck_fail(env, lfsck, true, true);
2554 if (bk->lb_param & LPF_FAILOUT)
2560 target = mdd_object_find(env, mdd, fid);
2561 if (target == NULL) {
2563 } else if (IS_ERR(target)) {
2564 mdd_lfsck_fail(env, lfsck, true, true);
2565 if (bk->lb_param & LPF_FAILOUT)
2566 RETURN(PTR_ERR(target));
2571 /* XXX: In fact, low layer otable-based iteration should not
2572 * return agent object. But before LU-2646 resolved, we
2573 * need more processing for agent object. */
2574 if (mdd_object_exists(target) && !mdd_object_remote(target))
2575 rc = mdd_lfsck_exec_oit(env, lfsck, target);
2576 mdd_object_put(env, target);
2577 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2581 rc = mdd_lfsck_checkpoint(env, lfsck, true);
2582 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2586 mdd_lfsck_control_speed(lfsck);
2588 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL1)) {
2589 spin_lock(&lfsck->ml_lock);
2590 thread_set_flags(thread, SVC_STOPPING);
2591 spin_unlock(&lfsck->ml_lock);
2595 rc = iops->next(env, di);
2597 lfsck->ml_oit_over = 1;
2599 if (unlikely(!thread_is_running(thread)))
2601 } while (rc == 0 || lfsck->ml_di_dir != NULL);
2606 static int mdd_lfsck_main(void *args)
2609 struct md_lfsck *lfsck = (struct md_lfsck *)args;
2610 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2611 struct dt_object *oit_obj = lfsck->ml_obj_oit;
2612 const struct dt_it_ops *oit_iops = &oit_obj->do_index_ops->dio_it;
2613 struct dt_it *oit_di;
2617 cfs_daemonize("lfsck");
2618 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
2620 CERROR("%s: LFSCK, fail to init env, rc = %d\n",
2621 mdd_lfsck2name(lfsck), rc);
2625 oit_di = oit_iops->init(&env, oit_obj, lfsck->ml_args_oit, BYPASS_CAPA);
2626 if (IS_ERR(oit_di)) {
2627 rc = PTR_ERR(oit_di);
2628 CERROR("%s: LFSCK, fail to init iteration, rc = %d\n",
2629 mdd_lfsck2name(lfsck), rc);
2633 spin_lock(&lfsck->ml_lock);
2634 lfsck->ml_di_oit = oit_di;
2635 spin_unlock(&lfsck->ml_lock);
2636 rc = mdd_lfsck_prep(&env, lfsck);
2640 CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = 0x%x, dir_flags = 0x%x, "
2641 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2642 ", pid = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2643 lfsck->ml_pos_current.lp_oit_cookie,
2644 lfsck->ml_pos_current.lp_dir_cookie,
2645 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2648 spin_lock(&lfsck->ml_lock);
2649 thread_set_flags(thread, SVC_RUNNING);
2650 spin_unlock(&lfsck->ml_lock);
2651 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2653 if (!cfs_list_empty(&lfsck->ml_list_scan) ||
2654 cfs_list_empty(&lfsck->ml_list_double_scan))
2655 rc = mdd_lfsck_oit_engine(&env, lfsck);
2659 CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = 0x%x, dir_flags = 0x%x, "
2660 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2661 ", pid = %d, rc = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2662 lfsck->ml_pos_current.lp_oit_cookie,
2663 lfsck->ml_pos_current.lp_dir_cookie,
2664 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2665 cfs_curproc_pid(), rc);
2667 if (lfsck->ml_paused && cfs_list_empty(&lfsck->ml_list_scan))
2668 oit_iops->put(&env, oit_di);
2670 if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2671 rc = mdd_lfsck_post(&env, lfsck, rc);
2672 if (lfsck->ml_di_dir != NULL)
2673 mdd_lfsck_close_dir(&env, lfsck);
2676 spin_lock(&lfsck->ml_lock);
2677 lfsck->ml_di_oit = NULL;
2678 spin_unlock(&lfsck->ml_lock);
2680 oit_iops->fini(&env, oit_di);
2682 if (!cfs_list_empty(&lfsck->ml_list_double_scan))
2683 rc = mdd_lfsck_double_scan(&env, lfsck);
2688 /* XXX: Purge the pinned objects in the future. */
2694 spin_lock(&lfsck->ml_lock);
2695 thread_set_flags(thread, SVC_STOPPED);
2696 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2697 spin_unlock(&lfsck->ml_lock);
2701 /* external interfaces */
2703 int mdd_lfsck_set_speed(const struct lu_env *env, struct md_lfsck *lfsck,
2708 mutex_lock(&lfsck->ml_mutex);
2709 __mdd_lfsck_set_speed(lfsck, limit);
2710 rc = mdd_lfsck_bookmark_store(env, lfsck);
2711 mutex_unlock(&lfsck->ml_mutex);
2715 int mdd_lfsck_dump(const struct lu_env *env, struct md_lfsck *lfsck,
2716 __u16 type, char *buf, int len)
2718 struct lfsck_component *com;
2721 if (!lfsck->ml_initialized)
2724 com = mdd_lfsck_component_find(lfsck, type);
2728 rc = com->lc_ops->lfsck_dump(env, com, buf, len);
2729 mdd_lfsck_component_put(env, com);
2733 int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
2734 struct lfsck_start *start)
2736 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2737 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2738 struct lfsck_component *com;
2739 struct l_wait_info lwi = { 0 };
2746 if (lfsck->ml_obj_oit == NULL)
2749 /* start == NULL means auto trigger paused LFSCK. */
2750 if ((start == NULL) &&
2751 (cfs_list_empty(&lfsck->ml_list_scan) ||
2752 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
2755 mutex_lock(&lfsck->ml_mutex);
2756 spin_lock(&lfsck->ml_lock);
2757 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2758 spin_unlock(&lfsck->ml_lock);
2759 mutex_unlock(&lfsck->ml_mutex);
2763 spin_unlock(&lfsck->ml_lock);
2765 lfsck->ml_paused = 0;
2766 lfsck->ml_oit_over = 0;
2767 lfsck->ml_drop_dryrun = 0;
2768 lfsck->ml_new_scanned = 0;
2770 /* For auto trigger. */
2774 start->ls_version = bk->lb_version;
2775 if (start->ls_valid & LSV_SPEED_LIMIT) {
2776 __mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
2780 if (start->ls_valid & LSV_ERROR_HANDLE) {
2781 valid |= DOIV_ERROR_HANDLE;
2782 if (start->ls_flags & LPF_FAILOUT)
2783 flags |= DOIF_FAILOUT;
2785 if ((start->ls_flags & LPF_FAILOUT) &&
2786 !(bk->lb_param & LPF_FAILOUT)) {
2787 bk->lb_param |= LPF_FAILOUT;
2789 } else if (!(start->ls_flags & LPF_FAILOUT) &&
2790 (bk->lb_param & LPF_FAILOUT)) {
2791 bk->lb_param &= ~LPF_FAILOUT;
2796 if (start->ls_valid & LSV_DRYRUN) {
2797 if ((start->ls_flags & LPF_DRYRUN) &&
2798 !(bk->lb_param & LPF_DRYRUN)) {
2799 bk->lb_param |= LPF_DRYRUN;
2801 } else if (!(start->ls_flags & LPF_DRYRUN) &&
2802 (bk->lb_param & LPF_DRYRUN)) {
2803 bk->lb_param &= ~LPF_DRYRUN;
2804 lfsck->ml_drop_dryrun = 1;
2810 rc = mdd_lfsck_bookmark_store(env, lfsck);
2815 if (start->ls_flags & LPF_RESET)
2816 flags |= DOIF_RESET;
2818 if (start->ls_active != 0) {
2819 struct lfsck_component *next;
2822 if (start->ls_active == LFSCK_TYPES_ALL)
2823 start->ls_active = LFSCK_TYPES_SUPPORTED;
2825 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2826 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2827 GOTO(out, rc = -ENOTSUPP);
2830 cfs_list_for_each_entry_safe(com, next,
2831 &lfsck->ml_list_scan, lc_link) {
2832 if (!(com->lc_type & start->ls_active)) {
2833 rc = com->lc_ops->lfsck_post(env, com, 0);
2839 while (start->ls_active != 0) {
2840 if (type & start->ls_active) {
2841 com = __mdd_lfsck_component_find(lfsck, type,
2842 &lfsck->ml_list_idle);
2844 /* The component status will be updated
2845 * when its prep() is called later by
2846 * the LFSCK main engine. */
2847 cfs_list_del_init(&com->lc_link);
2848 cfs_list_add_tail(&com->lc_link,
2849 &lfsck->ml_list_scan);
2851 start->ls_active &= ~type;
2857 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2858 start->ls_active |= com->lc_type;
2859 if (flags & DOIF_RESET) {
2860 rc = com->lc_ops->lfsck_reset(env, com, false);
2867 lfsck->ml_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2868 if (bk->lb_param & LPF_DRYRUN)
2869 lfsck->ml_args_dir |= LUDA_VERIFY_DRYRUN;
2871 if (bk->lb_param & LPF_FAILOUT) {
2872 valid |= DOIV_ERROR_HANDLE;
2873 flags |= DOIF_FAILOUT;
2876 if (!cfs_list_empty(&lfsck->ml_list_scan))
2877 flags |= DOIF_OUTUSED;
2879 lfsck->ml_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2880 thread_set_flags(thread, 0);
2881 rc = cfs_create_thread(mdd_lfsck_main, lfsck, 0);
2883 CERROR("%s: cannot start LFSCK thread, rc = %d\n",
2884 mdd_lfsck2name(lfsck), rc);
2886 l_wait_event(thread->t_ctl_waitq,
2887 thread_is_running(thread) ||
2888 thread_is_stopped(thread),
2894 mutex_unlock(&lfsck->ml_mutex);
2895 return (rc < 0 ? rc : 0);
2898 int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck,
2901 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2902 struct l_wait_info lwi = { 0 };
2905 if (!lfsck->ml_initialized)
2908 mutex_lock(&lfsck->ml_mutex);
2909 spin_lock(&lfsck->ml_lock);
2910 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2911 spin_unlock(&lfsck->ml_lock);
2912 mutex_unlock(&lfsck->ml_mutex);
2917 lfsck->ml_paused = 1;
2918 thread_set_flags(thread, SVC_STOPPING);
2919 /* The LFSCK thread may be sleeping on low layer wait queue,
2921 if (likely(lfsck->ml_di_oit != NULL))
2922 lfsck->ml_obj_oit->do_index_ops->dio_it.put(env,
2924 spin_unlock(&lfsck->ml_lock);
2926 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2927 l_wait_event(thread->t_ctl_waitq,
2928 thread_is_stopped(thread),
2930 mutex_unlock(&lfsck->ml_mutex);
2935 static const struct lu_fid lfsck_it_fid = { .f_seq = FID_SEQ_LOCAL_FILE,
2936 .f_oid = OTABLE_IT_OID,
2939 int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd)
2941 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2942 struct dt_object *obj;
2946 LASSERT(!lfsck->ml_initialized);
2948 lfsck->ml_initialized = 1;
2949 mutex_init(&lfsck->ml_mutex);
2950 spin_lock_init(&lfsck->ml_lock);
2951 CFS_INIT_LIST_HEAD(&lfsck->ml_list_scan);
2952 CFS_INIT_LIST_HEAD(&lfsck->ml_list_dir);
2953 CFS_INIT_LIST_HEAD(&lfsck->ml_list_double_scan);
2954 CFS_INIT_LIST_HEAD(&lfsck->ml_list_idle);
2955 cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
2957 obj = dt_locate(env, mdd->mdd_bottom, &lfsck_it_fid);
2959 RETURN(PTR_ERR(obj));
2961 lfsck->ml_obj_oit = obj;
2962 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2964 if (rc == -ENOTSUPP)
2970 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_bookmark_name,
2971 &mdd_env_info(env)->mti_fid);
2973 RETURN(PTR_ERR(obj));
2975 lfsck->ml_bookmark_obj = obj;
2976 rc = mdd_lfsck_bookmark_load(env, lfsck);
2978 rc = mdd_lfsck_bookmark_init(env, lfsck);
2982 rc = mdd_lfsck_namespace_setup(env, lfsck);
2983 /* XXX: LFSCK components initialization to be added here. */
2988 void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd)
2990 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2991 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2992 struct lfsck_component *com;
2994 if (!lfsck->ml_initialized)
2997 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
2999 if (lfsck->ml_obj_oit != NULL) {
3000 lu_object_put(env, &lfsck->ml_obj_oit->do_lu);
3001 lfsck->ml_obj_oit = NULL;
3004 LASSERT(lfsck->ml_obj_dir == NULL);
3006 if (lfsck->ml_bookmark_obj != NULL) {
3007 lu_object_put(env, &lfsck->ml_bookmark_obj->do_lu);
3008 lfsck->ml_bookmark_obj = NULL;
3011 while (!cfs_list_empty(&lfsck->ml_list_scan)) {
3012 com = cfs_list_entry(lfsck->ml_list_scan.next,
3013 struct lfsck_component,
3015 mdd_lfsck_component_cleanup(env, com);
3018 LASSERT(cfs_list_empty(&lfsck->ml_list_dir));
3020 while (!cfs_list_empty(&lfsck->ml_list_double_scan)) {
3021 com = cfs_list_entry(lfsck->ml_list_double_scan.next,
3022 struct lfsck_component,
3024 mdd_lfsck_component_cleanup(env, com);
3027 while (!cfs_list_empty(&lfsck->ml_list_idle)) {
3028 com = cfs_list_entry(lfsck->ml_list_idle.next,
3029 struct lfsck_component,
3031 mdd_lfsck_component_cleanup(env, com);