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 LASSERT(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. */
271 pos->lp_dir_cookie--;
273 fid_zero(&pos->lp_dir_parent);
274 pos->lp_dir_cookie = 0;
276 spin_unlock(&lfsck->ml_lock);
279 static inline void mdd_lfsck_pos_set_zero(struct lfsck_position *pos)
281 memset(pos, 0, sizeof(*pos));
284 static inline int mdd_lfsck_pos_is_zero(const struct lfsck_position *pos)
286 return pos->lp_oit_cookie == 0 && fid_is_zero(&pos->lp_dir_parent);
289 static inline int mdd_lfsck_pos_is_eq(const struct lfsck_position *pos1,
290 const struct lfsck_position *pos2)
292 if (pos1->lp_oit_cookie < pos2->lp_oit_cookie)
295 if (pos1->lp_oit_cookie > pos2->lp_oit_cookie)
298 if (fid_is_zero(&pos1->lp_dir_parent) &&
299 !fid_is_zero(&pos2->lp_dir_parent))
302 if (!fid_is_zero(&pos1->lp_dir_parent) &&
303 fid_is_zero(&pos2->lp_dir_parent))
306 if (fid_is_zero(&pos1->lp_dir_parent) &&
307 fid_is_zero(&pos2->lp_dir_parent))
310 LASSERT(lu_fid_eq(&pos1->lp_dir_parent, &pos2->lp_dir_parent));
312 if (pos1->lp_dir_cookie < pos2->lp_dir_cookie)
315 if (pos1->lp_dir_cookie > pos2->lp_dir_cookie)
321 static void mdd_lfsck_close_dir(const struct lu_env *env,
322 struct md_lfsck *lfsck)
324 struct dt_object *dir_obj = lfsck->ml_obj_dir;
325 const struct dt_it_ops *dir_iops = &dir_obj->do_index_ops->dio_it;
326 struct dt_it *dir_di = lfsck->ml_di_dir;
328 spin_lock(&lfsck->ml_lock);
329 lfsck->ml_di_dir = NULL;
330 spin_unlock(&lfsck->ml_lock);
332 dir_iops->put(env, dir_di);
333 dir_iops->fini(env, dir_di);
334 lfsck->ml_obj_dir = NULL;
335 lu_object_put(env, &dir_obj->do_lu);
338 static void __mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
340 lfsck->ml_bookmark_ram.lb_speed_limit = limit;
341 if (limit != LFSCK_SPEED_NO_LIMIT) {
342 if (limit > CFS_HZ) {
343 lfsck->ml_sleep_rate = limit / CFS_HZ;
344 lfsck->ml_sleep_jif = 1;
346 lfsck->ml_sleep_rate = 1;
347 lfsck->ml_sleep_jif = CFS_HZ / limit;
350 lfsck->ml_sleep_jif = 0;
351 lfsck->ml_sleep_rate = 0;
355 static void mdd_lfsck_control_speed(struct md_lfsck *lfsck)
357 struct ptlrpc_thread *thread = &lfsck->ml_thread;
358 struct l_wait_info lwi;
360 if (lfsck->ml_sleep_jif > 0 &&
361 lfsck->ml_new_scanned >= lfsck->ml_sleep_rate) {
362 spin_lock(&lfsck->ml_lock);
363 if (likely(lfsck->ml_sleep_jif > 0 &&
364 lfsck->ml_new_scanned >= lfsck->ml_sleep_rate)) {
365 lwi = LWI_TIMEOUT_INTR(lfsck->ml_sleep_jif, NULL,
366 LWI_ON_SIGNAL_NOOP, NULL);
367 spin_unlock(&lfsck->ml_lock);
369 l_wait_event(thread->t_ctl_waitq,
370 !thread_is_running(thread),
372 lfsck->ml_new_scanned = 0;
374 spin_unlock(&lfsck->ml_lock);
379 /* lfsck_bookmark file ops */
381 static void inline mdd_lfsck_bookmark_to_cpu(struct lfsck_bookmark *des,
382 struct lfsck_bookmark *src)
384 des->lb_magic = le32_to_cpu(src->lb_magic);
385 des->lb_version = le16_to_cpu(src->lb_version);
386 des->lb_param = le16_to_cpu(src->lb_param);
387 des->lb_speed_limit = le32_to_cpu(src->lb_speed_limit);
390 static void inline mdd_lfsck_bookmark_to_le(struct lfsck_bookmark *des,
391 struct lfsck_bookmark *src)
393 des->lb_magic = cpu_to_le32(src->lb_magic);
394 des->lb_version = cpu_to_le16(src->lb_version);
395 des->lb_param = cpu_to_le16(src->lb_param);
396 des->lb_speed_limit = cpu_to_le32(src->lb_speed_limit);
399 static int mdd_lfsck_bookmark_load(const struct lu_env *env,
400 struct md_lfsck *lfsck)
403 int len = sizeof(struct lfsck_bookmark);
406 rc = dt_record_read(env, lfsck->ml_bookmark_obj,
407 mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
410 struct lfsck_bookmark *bm = &lfsck->ml_bookmark_ram;
412 mdd_lfsck_bookmark_to_cpu(bm, &lfsck->ml_bookmark_disk);
413 if (bm->lb_magic != LFSCK_BOOKMARK_MAGIC) {
414 CWARN("%.16s: invalid lfsck_bookmark magic "
415 "0x%x != 0x%x\n", mdd_lfsck2name(lfsck),
416 bm->lb_magic, LFSCK_BOOKMARK_MAGIC);
417 /* Process it as new lfsck_bookmark. */
421 if (rc == -EFAULT && pos == 0)
422 /* return -ENODATA for empty lfsck_bookmark. */
425 CERROR("%.16s: fail to load lfsck_bookmark, "
426 "expected = %d, rc = %d\n",
427 mdd_lfsck2name(lfsck), len, rc);
432 static int mdd_lfsck_bookmark_store(const struct lu_env *env,
433 struct md_lfsck *lfsck)
435 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
436 struct thandle *handle;
437 struct dt_object *obj = lfsck->ml_bookmark_obj;
439 int len = sizeof(struct lfsck_bookmark);
443 mdd_lfsck_bookmark_to_le(&lfsck->ml_bookmark_disk,
444 &lfsck->ml_bookmark_ram);
445 handle = dt_trans_create(env, mdd->mdd_bottom);
446 if (IS_ERR(handle)) {
447 rc = PTR_ERR(handle);
448 CERROR("%.16s: fail to create trans for storing "
449 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
453 rc = dt_declare_record_write(env, obj, len, 0, handle);
455 CERROR("%.16s: fail to declare trans for storing "
456 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
460 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
462 CERROR("%.16s: fail to start trans for storing "
463 "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
467 rc = dt_record_write(env, obj,
468 mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
471 CERROR("%.16s: fail to store lfsck_bookmark, expected = %d, "
472 "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
477 dt_trans_stop(env, mdd->mdd_bottom, handle);
481 static int mdd_lfsck_bookmark_init(const struct lu_env *env,
482 struct md_lfsck *lfsck)
484 struct lfsck_bookmark *mb = &lfsck->ml_bookmark_ram;
487 memset(mb, 0, sizeof(mb));
488 mb->lb_magic = LFSCK_BOOKMARK_MAGIC;
489 mb->lb_version = LFSCK_VERSION_V2;
490 mutex_lock(&lfsck->ml_mutex);
491 rc = mdd_lfsck_bookmark_store(env, lfsck);
492 mutex_unlock(&lfsck->ml_mutex);
496 /* lfsck_namespace file ops */
498 static void inline mdd_lfsck_position_to_cpu(struct lfsck_position *des,
499 struct lfsck_position *src)
501 des->lp_oit_cookie = le64_to_cpu(src->lp_oit_cookie);
502 fid_le_to_cpu(&des->lp_dir_parent, &src->lp_dir_parent);
503 des->lp_dir_cookie = le64_to_cpu(src->lp_dir_cookie);
506 static void inline mdd_lfsck_position_to_le(struct lfsck_position *des,
507 struct lfsck_position *src)
509 des->lp_oit_cookie = cpu_to_le64(src->lp_oit_cookie);
510 fid_cpu_to_le(&des->lp_dir_parent, &src->lp_dir_parent);
511 des->lp_dir_cookie = cpu_to_le64(src->lp_dir_cookie);
514 static void inline mdd_lfsck_namespace_to_cpu(struct lfsck_namespace *des,
515 struct lfsck_namespace *src)
517 des->ln_magic = le32_to_cpu(src->ln_magic);
518 des->ln_status = le32_to_cpu(src->ln_status);
519 des->ln_flags = le32_to_cpu(src->ln_flags);
520 des->ln_success_count = le32_to_cpu(src->ln_success_count);
521 des->ln_run_time_phase1 = le32_to_cpu(src->ln_run_time_phase1);
522 des->ln_run_time_phase2 = le32_to_cpu(src->ln_run_time_phase2);
523 des->ln_time_last_complete = le64_to_cpu(src->ln_time_last_complete);
524 des->ln_time_latest_start = le64_to_cpu(src->ln_time_latest_start);
525 des->ln_time_last_checkpoint =
526 le64_to_cpu(src->ln_time_last_checkpoint);
527 mdd_lfsck_position_to_cpu(&des->ln_pos_latest_start,
528 &src->ln_pos_latest_start);
529 mdd_lfsck_position_to_cpu(&des->ln_pos_last_checkpoint,
530 &src->ln_pos_last_checkpoint);
531 mdd_lfsck_position_to_cpu(&des->ln_pos_first_inconsistent,
532 &src->ln_pos_first_inconsistent);
533 des->ln_items_checked = le64_to_cpu(src->ln_items_checked);
534 des->ln_items_repaired = le64_to_cpu(src->ln_items_repaired);
535 des->ln_items_failed = le64_to_cpu(src->ln_items_failed);
536 des->ln_dirs_checked = le64_to_cpu(src->ln_dirs_checked);
537 des->ln_mlinked_checked = le64_to_cpu(src->ln_mlinked_checked);
538 des->ln_objs_checked_phase2 = le64_to_cpu(src->ln_objs_checked_phase2);
539 des->ln_objs_repaired_phase2 =
540 le64_to_cpu(src->ln_objs_repaired_phase2);
541 des->ln_objs_failed_phase2 = le64_to_cpu(src->ln_objs_failed_phase2);
542 des->ln_objs_nlink_repaired = le64_to_cpu(src->ln_objs_nlink_repaired);
543 des->ln_objs_lost_found = le64_to_cpu(src->ln_objs_lost_found);
544 fid_le_to_cpu(&des->ln_fid_latest_scanned_phase2,
545 &src->ln_fid_latest_scanned_phase2);
548 static void inline mdd_lfsck_namespace_to_le(struct lfsck_namespace *des,
549 struct lfsck_namespace *src)
551 des->ln_magic = cpu_to_le32(src->ln_magic);
552 des->ln_status = cpu_to_le32(src->ln_status);
553 des->ln_flags = cpu_to_le32(src->ln_flags);
554 des->ln_success_count = cpu_to_le32(src->ln_success_count);
555 des->ln_run_time_phase1 = cpu_to_le32(src->ln_run_time_phase1);
556 des->ln_run_time_phase2 = cpu_to_le32(src->ln_run_time_phase2);
557 des->ln_time_last_complete = cpu_to_le64(src->ln_time_last_complete);
558 des->ln_time_latest_start = cpu_to_le64(src->ln_time_latest_start);
559 des->ln_time_last_checkpoint =
560 cpu_to_le64(src->ln_time_last_checkpoint);
561 mdd_lfsck_position_to_le(&des->ln_pos_latest_start,
562 &src->ln_pos_latest_start);
563 mdd_lfsck_position_to_le(&des->ln_pos_last_checkpoint,
564 &src->ln_pos_last_checkpoint);
565 mdd_lfsck_position_to_le(&des->ln_pos_first_inconsistent,
566 &src->ln_pos_first_inconsistent);
567 des->ln_items_checked = cpu_to_le64(src->ln_items_checked);
568 des->ln_items_repaired = cpu_to_le64(src->ln_items_repaired);
569 des->ln_items_failed = cpu_to_le64(src->ln_items_failed);
570 des->ln_dirs_checked = cpu_to_le64(src->ln_dirs_checked);
571 des->ln_mlinked_checked = cpu_to_le64(src->ln_mlinked_checked);
572 des->ln_objs_checked_phase2 = cpu_to_le64(src->ln_objs_checked_phase2);
573 des->ln_objs_repaired_phase2 =
574 cpu_to_le64(src->ln_objs_repaired_phase2);
575 des->ln_objs_failed_phase2 = cpu_to_le64(src->ln_objs_failed_phase2);
576 des->ln_objs_nlink_repaired = cpu_to_le64(src->ln_objs_nlink_repaired);
577 des->ln_objs_lost_found = cpu_to_le64(src->ln_objs_lost_found);
578 fid_cpu_to_le(&des->ln_fid_latest_scanned_phase2,
579 &src->ln_fid_latest_scanned_phase2);
583 * \retval +ve: the lfsck_namespace is broken, the caller should reset it.
584 * \retval 0: succeed.
585 * \retval -ve: failed cases.
587 static int mdd_lfsck_namespace_load(const struct lu_env *env,
588 struct lfsck_component *com)
590 int len = com->lc_file_size;
593 rc = dt_xattr_get(env, com->lc_obj,
594 mdd_buf_get(env, com->lc_file_disk, len),
595 XATTR_NAME_LFSCK_NAMESPACE, BYPASS_CAPA);
597 struct lfsck_namespace *ns = com->lc_file_ram;
599 mdd_lfsck_namespace_to_cpu(ns,
600 (struct lfsck_namespace *)com->lc_file_disk);
601 if (ns->ln_magic != LFSCK_NAMESPACE_MAGIC) {
602 CWARN("%.16s: invalid lfsck_namespace magic "
604 mdd_lfsck2name(com->lc_lfsck),
605 ns->ln_magic, LFSCK_NAMESPACE_MAGIC);
610 } else if (rc != -ENODATA) {
611 CERROR("%.16s: fail to load lfsck_namespace, expected = %d, "
612 "rc = %d\n", mdd_lfsck2name(com->lc_lfsck), len, rc);
619 static int mdd_lfsck_namespace_store(const struct lu_env *env,
620 struct lfsck_component *com, bool init)
622 struct dt_object *obj = com->lc_obj;
623 struct md_lfsck *lfsck = com->lc_lfsck;
624 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
625 struct thandle *handle;
626 int len = com->lc_file_size;
630 mdd_lfsck_namespace_to_le((struct lfsck_namespace *)com->lc_file_disk,
631 (struct lfsck_namespace *)com->lc_file_ram);
632 handle = dt_trans_create(env, mdd->mdd_bottom);
633 if (IS_ERR(handle)) {
634 rc = PTR_ERR(handle);
635 CERROR("%.16s: fail to create trans for storing "
636 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
640 rc = dt_declare_xattr_set(env, obj,
641 mdd_buf_get(env, com->lc_file_disk, len),
642 XATTR_NAME_LFSCK_NAMESPACE, 0, handle);
644 CERROR("%.16s: fail to declare trans for storing "
645 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
649 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
651 CERROR("%.16s: fail to start trans for storing "
652 "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
656 rc = dt_xattr_set(env, obj,
657 mdd_buf_get(env, com->lc_file_disk, len),
658 XATTR_NAME_LFSCK_NAMESPACE,
659 init ? LU_XATTR_CREATE : LU_XATTR_REPLACE,
660 handle, BYPASS_CAPA);
662 CERROR("%.16s: fail to store lfsck_namespace, len = %d, "
663 "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
668 dt_trans_stop(env, mdd->mdd_bottom, handle);
672 static int mdd_lfsck_namespace_init(const struct lu_env *env,
673 struct lfsck_component *com)
675 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
678 memset(ns, 0, sizeof(*ns));
679 ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
680 ns->ln_status = LS_INIT;
681 down_write(&com->lc_sem);
682 rc = mdd_lfsck_namespace_store(env, com, true);
683 up_write(&com->lc_sem);
687 static int mdd_declare_lfsck_namespace_unlink(const struct lu_env *env,
688 struct mdd_device *mdd,
692 struct thandle *handle)
696 rc = dt_declare_delete(env, p, (const struct dt_key *)name, handle);
700 rc = dt_declare_ref_del(env, c, handle);
704 rc = dt_declare_destroy(env, c, handle);
708 static int mdd_lfsck_namespace_unlink(const struct lu_env *env,
709 struct mdd_device *mdd,
710 struct lfsck_component *com)
712 struct mdd_thread_info *info = mdd_env_info(env);
713 struct lu_fid *fid = &info->mti_fid;
714 struct dt_object *child = com->lc_obj;
715 struct dt_object *parent;
716 struct thandle *handle;
721 parent = dt_store_resolve(env, mdd->mdd_bottom, "", fid);
723 RETURN(rc = PTR_ERR(parent));
725 if (!dt_try_as_dir(env, parent))
726 GOTO(out, rc = -ENOTDIR);
728 handle = dt_trans_create(env, mdd->mdd_bottom);
730 GOTO(out, rc = PTR_ERR(handle));
732 rc = mdd_declare_lfsck_namespace_unlink(env, mdd, parent, child,
733 lfsck_namespace_name, handle);
737 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
741 dt_write_lock(env, child, MOR_TGT_CHILD);
743 rc = dt_delete(env, parent, (struct dt_key *)lfsck_namespace_name,
744 handle, BYPASS_CAPA);
748 rc = child->do_ops->do_ref_del(env, child, handle);
750 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
751 rc = dt_insert(env, parent,
752 (const struct dt_rec*)fid,
753 (const struct dt_key *)lfsck_namespace_name,
754 handle, BYPASS_CAPA, 1);
760 rc = dt_destroy(env, child, handle);
766 dt_write_unlock(env, child);
769 lu_object_put(env, &child->do_lu);
773 dt_trans_stop(env, mdd->mdd_bottom, handle);
776 lu_object_put(env, &parent->do_lu);
780 static int mdd_lfsck_namespace_lookup(const struct lu_env *env,
781 struct lfsck_component *com,
782 const struct lu_fid *fid,
785 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
788 fid_cpu_to_be(key, fid);
789 rc = dt_lookup(env, com->lc_obj, (struct dt_rec *)flags,
790 (const struct dt_key *)key, BYPASS_CAPA);
794 static int mdd_lfsck_namespace_delete(const struct lu_env *env,
795 struct lfsck_component *com,
796 const struct lu_fid *fid)
798 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
799 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
800 struct thandle *handle;
801 struct dt_object *obj = com->lc_obj;
805 handle = dt_trans_create(env, mdd->mdd_bottom);
807 RETURN(PTR_ERR(handle));
809 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid, handle);
813 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
817 fid_cpu_to_be(key, fid);
818 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
824 dt_trans_stop(env, mdd->mdd_bottom, handle);
828 static int mdd_lfsck_namespace_update(const struct lu_env *env,
829 struct lfsck_component *com,
830 const struct lu_fid *fid,
831 __u8 flags, bool force)
833 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
834 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
835 struct thandle *handle;
836 struct dt_object *obj = com->lc_obj;
842 rc = mdd_lfsck_namespace_lookup(env, com, fid, &tf);
843 if (rc != 0 && rc != -ENOENT)
847 if (!force || flags == tf)
851 handle = dt_trans_create(env, mdd->mdd_bottom);
853 RETURN(PTR_ERR(handle));
855 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid,
860 handle = dt_trans_create(env, mdd->mdd_bottom);
862 RETURN(PTR_ERR(handle));
865 rc = dt_declare_insert(env, obj, (const struct dt_rec *)&flags,
866 (const struct dt_key *)fid, handle);
870 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
874 fid_cpu_to_be(key, fid);
876 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
879 CERROR("%s: fail to insert "DFID", rc = %d\n",
880 mdd_lfsck2name(com->lc_lfsck), PFID(fid), rc);
885 rc = dt_insert(env, obj, (const struct dt_rec *)&flags,
886 (const struct dt_key *)key, handle, BYPASS_CAPA, 1);
891 dt_trans_stop(env, mdd->mdd_bottom, handle);
896 * \retval +ve repaired
897 * \retval 0 no need to repair
898 * \retval -ve error cases
900 static int mdd_lfsck_namespace_double_scan_one(const struct lu_env *env,
901 struct lfsck_component *com,
902 struct mdd_object *child,
905 struct mdd_thread_info *info = mdd_env_info(env);
906 struct lu_attr *la = &info->mti_la;
907 struct lu_name *cname = &info->mti_name;
908 struct lu_fid *pfid = &info->mti_fid;
909 struct lu_fid *cfid = &info->mti_fid2;
910 struct md_lfsck *lfsck = com->lc_lfsck;
911 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
912 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
913 struct lfsck_namespace *ns =
914 (struct lfsck_namespace *)com->lc_file_ram;
915 struct linkea_data ldata = { 0 };
916 struct thandle *handle = NULL;
923 if (com->lc_journal) {
929 handle = mdd_trans_create(env, mdd);
931 RETURN(rc = PTR_ERR(handle));
933 rc = mdd_declare_links_add(env, child, handle, NULL);
937 rc = mdd_trans_start(env, mdd, handle);
941 mdd_write_lock(env, child, MOR_TGT_CHILD);
945 if (unlikely(mdd_is_dead_obj(child)))
948 rc = mdd_links_read(env, child, &ldata);
950 if ((bk->lb_param & LPF_DRYRUN) &&
951 (rc == -EINVAL || rc == -ENODATA))
957 rc = mdd_la_get(env, child, la, BYPASS_CAPA);
961 ldata.ld_lee = LINKEA_FIRST_ENTRY(ldata);
962 count = ldata.ld_leh->leh_reccount;
963 while (count-- > 0) {
964 struct mdd_object *parent = NULL;
965 struct dt_object *dir;
967 linkea_entry_unpack(ldata.ld_lee, &ldata.ld_reclen, cname,
969 if (!fid_is_sane(pfid))
972 parent = mdd_object_find(env, mdd, pfid);
975 else if (IS_ERR(parent))
976 GOTO(stop, rc = PTR_ERR(parent));
978 if (!mdd_object_exists(parent))
981 /* XXX: need more processing for remote object in the future. */
982 if (mdd_object_remote(parent)) {
983 mdd_object_put(env, parent);
984 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
988 dir = mdd_object_child(parent);
989 if (unlikely(!dt_try_as_dir(env, dir)))
992 /* To guarantee the 'name' is terminated with '0'. */
993 memcpy(info->mti_key, cname->ln_name, cname->ln_namelen);
994 info->mti_key[cname->ln_namelen] = 0;
995 cname->ln_name = info->mti_key;
996 rc = dt_lookup(env, dir, (struct dt_rec *)cfid,
997 (const struct dt_key *)cname->ln_name,
999 if (rc != 0 && rc != -ENOENT) {
1000 mdd_object_put(env, parent);
1005 if (lu_fid_eq(cfid, mdo2fid(child))) {
1006 mdd_object_put(env, parent);
1007 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
1014 if (ldata.ld_leh->leh_reccount > la->la_nlink)
1017 /* XXX: For the case of there is linkea entry, but without name
1018 * entry pointing to the object, and the object link count
1019 * isn't less than the count of name entries, then add the
1020 * name entry back to namespace.
1022 * It is out of LFSCK 1.5 scope, will implement it in the
1023 * future. Keep the linkEA entry. */
1024 mdd_object_put(env, parent);
1025 ldata.ld_lee = LINKEA_NEXT_ENTRY(ldata);
1030 mdd_object_put(env, parent);
1031 if (bk->lb_param & LPF_DRYRUN)
1034 CDEBUG(D_LFSCK, "Remove linkEA: "DFID"[%.*s], "DFID"\n",
1035 PFID(mdo2fid(child)), cname->ln_namelen, cname->ln_name,
1037 linkea_del_buf(&ldata, cname);
1042 if (!com->lc_journal) {
1043 com->lc_journal = 1;
1047 rc = mdd_links_write(env, child, &ldata, handle);
1054 mdd_write_unlock(env, child);
1057 mdd_trans_stop(env, mdd, rc, handle);
1059 if (rc == 0 && update) {
1060 ns->ln_objs_nlink_repaired++;
1066 /* namespace APIs */
1068 static int mdd_lfsck_namespace_reset(const struct lu_env *env,
1069 struct lfsck_component *com, bool init)
1071 struct mdd_thread_info *info = mdd_env_info(env);
1072 struct lu_fid *fid = &info->mti_fid;
1073 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
1074 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
1075 struct md_object *mdo;
1076 struct dt_object *dto;
1080 down_write(&com->lc_sem);
1082 memset(ns, 0, sizeof(*ns));
1084 __u32 count = ns->ln_success_count;
1085 __u64 last_time = ns->ln_time_last_complete;
1087 memset(ns, 0, sizeof(*ns));
1088 ns->ln_success_count = count;
1089 ns->ln_time_last_complete = last_time;
1091 ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
1092 ns->ln_status = LS_INIT;
1094 rc = mdd_lfsck_namespace_unlink(env, mdd, com);
1098 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
1099 mdo = llo_store_create_index(env, &mdd->mdd_md_dev, mdd->mdd_bottom, "",
1100 lfsck_namespace_name, fid,
1101 &dt_lfsck_features);
1103 GOTO(out, rc = PTR_ERR(mdo));
1105 lu_object_put(env, &mdo->mo_lu);
1106 dto = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name, fid);
1108 GOTO(out, rc = PTR_ERR(dto));
1111 rc = dto->do_ops->do_index_try(env, dto, &dt_lfsck_features);
1115 rc = mdd_lfsck_namespace_store(env, com, true);
1120 up_write(&com->lc_sem);
1125 mdd_lfsck_namespace_fail(const struct lu_env *env, struct lfsck_component *com,
1126 bool oit, bool new_checked)
1128 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
1130 down_write(&com->lc_sem);
1132 com->lc_new_checked++;
1133 ns->ln_items_failed++;
1134 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1135 mdd_lfsck_pos_fill(env, com->lc_lfsck,
1136 &ns->ln_pos_first_inconsistent, oit, !oit);
1137 up_write(&com->lc_sem);
1140 static int mdd_lfsck_namespace_checkpoint(const struct lu_env *env,
1141 struct lfsck_component *com,
1144 struct md_lfsck *lfsck = com->lc_lfsck;
1145 struct lfsck_namespace *ns =
1146 (struct lfsck_namespace *)com->lc_file_ram;
1149 if (com->lc_new_checked == 0 && !init)
1152 down_write(&com->lc_sem);
1154 ns->ln_pos_last_checkpoint = lfsck->ml_pos_current;
1156 ns->ln_time_last_checkpoint = ns->ln_time_latest_start;
1157 ns->ln_pos_latest_start = lfsck->ml_pos_current;
1159 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1160 HALF_SEC - lfsck->ml_time_last_checkpoint);
1161 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1162 ns->ln_items_checked += com->lc_new_checked;
1163 com->lc_new_checked = 0;
1166 rc = mdd_lfsck_namespace_store(env, com, false);
1168 up_write(&com->lc_sem);
1172 static int mdd_lfsck_namespace_prep(const struct lu_env *env,
1173 struct lfsck_component *com)
1175 struct md_lfsck *lfsck = com->lc_lfsck;
1176 struct lfsck_namespace *ns =
1177 (struct lfsck_namespace *)com->lc_file_ram;
1178 struct lfsck_position *pos = &com->lc_pos_start;
1180 if (ns->ln_status == LS_COMPLETED) {
1183 rc = mdd_lfsck_namespace_reset(env, com, false);
1188 down_write(&com->lc_sem);
1190 ns->ln_time_latest_start = cfs_time_current_sec();
1192 spin_lock(&lfsck->ml_lock);
1193 if (ns->ln_flags & LF_SCANNED_ONCE) {
1194 if (!lfsck->ml_drop_dryrun ||
1195 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1196 ns->ln_status = LS_SCANNING_PHASE2;
1197 cfs_list_del_init(&com->lc_link);
1198 cfs_list_add_tail(&com->lc_link,
1199 &lfsck->ml_list_double_scan);
1200 if (!cfs_list_empty(&com->lc_link_dir))
1201 cfs_list_del_init(&com->lc_link_dir);
1202 mdd_lfsck_pos_set_zero(pos);
1204 ns->ln_status = LS_SCANNING_PHASE1;
1205 ns->ln_run_time_phase1 = 0;
1206 ns->ln_run_time_phase2 = 0;
1207 ns->ln_items_checked = 0;
1208 ns->ln_items_repaired = 0;
1209 ns->ln_items_failed = 0;
1210 ns->ln_dirs_checked = 0;
1211 ns->ln_mlinked_checked = 0;
1212 ns->ln_objs_checked_phase2 = 0;
1213 ns->ln_objs_repaired_phase2 = 0;
1214 ns->ln_objs_failed_phase2 = 0;
1215 ns->ln_objs_nlink_repaired = 0;
1216 ns->ln_objs_lost_found = 0;
1217 fid_zero(&ns->ln_fid_latest_scanned_phase2);
1218 if (cfs_list_empty(&com->lc_link_dir))
1219 cfs_list_add_tail(&com->lc_link_dir,
1220 &lfsck->ml_list_dir);
1221 *pos = ns->ln_pos_first_inconsistent;
1224 ns->ln_status = LS_SCANNING_PHASE1;
1225 if (cfs_list_empty(&com->lc_link_dir))
1226 cfs_list_add_tail(&com->lc_link_dir,
1227 &lfsck->ml_list_dir);
1228 if (!lfsck->ml_drop_dryrun ||
1229 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1230 *pos = ns->ln_pos_last_checkpoint;
1231 pos->lp_oit_cookie++;
1232 if (!fid_is_zero(&pos->lp_dir_parent)) {
1233 if (pos->lp_dir_cookie == MDS_DIR_END_OFF) {
1234 fid_zero(&pos->lp_dir_parent);
1236 pos->lp_dir_cookie++;
1240 *pos = ns->ln_pos_first_inconsistent;
1243 spin_unlock(&lfsck->ml_lock);
1245 up_write(&com->lc_sem);
1249 static int mdd_lfsck_namespace_exec_oit(const struct lu_env *env,
1250 struct lfsck_component *com,
1251 struct mdd_object *obj)
1253 down_write(&com->lc_sem);
1254 com->lc_new_checked++;
1255 if (S_ISDIR(mdd_object_type(obj)))
1256 ((struct lfsck_namespace *)com->lc_file_ram)->ln_dirs_checked++;
1257 up_write(&com->lc_sem);
1261 static int mdd_declare_lfsck_namespace_exec_dir(const struct lu_env *env,
1262 struct mdd_object *obj,
1263 struct thandle *handle)
1267 /* For destroying all invalid linkEA entries. */
1268 rc = mdo_declare_xattr_del(env, obj, XATTR_NAME_LINK, handle);
1272 /* For insert new linkEA entry. */
1273 rc = mdd_declare_links_add(env, obj, handle, NULL);
1277 static int mdd_lfsck_namespace_check_exist(const struct lu_env *env,
1278 struct md_lfsck *lfsck,
1279 struct mdd_object *obj,
1282 struct dt_object *dir = lfsck->ml_obj_dir;
1283 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
1287 if (unlikely(mdd_is_dead_obj(obj)))
1288 RETURN(LFSCK_NAMEENTRY_DEAD);
1290 rc = dt_lookup(env, dir, (struct dt_rec *)fid,
1291 (const struct dt_key *)name, BYPASS_CAPA);
1293 RETURN(LFSCK_NAMEENTRY_REMOVED);
1298 if (!lu_fid_eq(fid, mdo2fid(obj)))
1299 RETURN(LFSCK_NAMEENTRY_RECREATED);
1304 static int mdd_lfsck_namespace_exec_dir(const struct lu_env *env,
1305 struct lfsck_component *com,
1306 struct mdd_object *obj,
1307 struct lu_dirent *ent)
1309 struct mdd_thread_info *info = mdd_env_info(env);
1310 struct lu_attr *la = &info->mti_la;
1311 struct md_lfsck *lfsck = com->lc_lfsck;
1312 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1313 struct lfsck_namespace *ns =
1314 (struct lfsck_namespace *)com->lc_file_ram;
1315 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1316 struct linkea_data ldata = { 0 };
1317 const struct lu_fid *pfid =
1318 lu_object_fid(&lfsck->ml_obj_dir->do_lu);
1319 const struct lu_fid *cfid = mdo2fid(obj);
1320 const struct lu_name *cname;
1321 struct thandle *handle = NULL;
1322 bool repaired = false;
1323 bool locked = false;
1328 cname = mdd_name_get_const(env, ent->lde_name, ent->lde_namelen);
1329 down_write(&com->lc_sem);
1330 com->lc_new_checked++;
1332 if (ent->lde_attrs & LUDA_UPGRADE) {
1333 ns->ln_flags |= LF_UPGRADE;
1335 } else if (ent->lde_attrs & LUDA_REPAIR) {
1336 ns->ln_flags |= LF_INCONSISTENT;
1340 if (ent->lde_name[0] == '.' &&
1341 (ent->lde_namelen == 1 ||
1342 (ent->lde_namelen == 2 && ent->lde_name[1] == '.')))
1345 if (!(bk->lb_param & LPF_DRYRUN) &&
1346 (com->lc_journal || repaired)) {
1351 com->lc_journal = 1;
1352 handle = mdd_trans_create(env, mdd);
1354 GOTO(out, rc = PTR_ERR(handle));
1356 rc = mdd_declare_lfsck_namespace_exec_dir(env, obj, handle);
1360 rc = mdd_trans_start(env, mdd, handle);
1364 mdd_write_lock(env, obj, MOR_TGT_CHILD);
1368 rc = mdd_lfsck_namespace_check_exist(env, lfsck, obj, ent->lde_name);
1372 rc = mdd_links_read(env, obj, &ldata);
1374 count = ldata.ld_leh->leh_reccount;
1375 rc = linkea_links_find(&ldata, cname, pfid);
1377 /* For dir, if there are more than one linkea entries,
1378 * then remove all the other redundant linkea entries.*/
1379 if (unlikely(count > 1 &&
1380 S_ISDIR(mdd_object_type(obj))))
1387 ns->ln_flags |= LF_INCONSISTENT;
1388 if (bk->lb_param & LPF_DRYRUN) {
1393 /*For dir, remove the unmatched linkea entry directly.*/
1394 if (S_ISDIR(mdd_object_type(obj))) {
1395 if (!com->lc_journal)
1398 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK,
1399 handle, BYPASS_CAPA);
1408 } else if (unlikely(rc == -EINVAL)) {
1409 ns->ln_flags |= LF_INCONSISTENT;
1410 if (bk->lb_param & LPF_DRYRUN) {
1416 if (!com->lc_journal)
1419 /* The magic crashed, we are not sure whether there are more
1420 * corrupt data in the linkea, so remove all linkea entries. */
1421 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK, handle,
1427 } else if (rc == -ENODATA) {
1428 ns->ln_flags |= LF_UPGRADE;
1429 if (bk->lb_param & LPF_DRYRUN) {
1436 rc = linkea_data_new(&ldata, &mdd_env_info(env)->mti_link_buf);
1441 if (!com->lc_journal)
1444 rc = linkea_add_buf(&ldata, cname, pfid);
1448 rc = mdd_links_write(env, obj, &ldata, handle);
1452 count = ldata.ld_leh->leh_reccount;
1461 rc = mdd_la_get(env, obj, la, BYPASS_CAPA);
1466 (la->la_nlink == 1 || S_ISDIR(mdd_object_type(obj))))
1467 /* Usually, it is for single linked object or dir, do nothing.*/
1470 /* Following modification will be in another transaction. */
1471 if (handle != NULL) {
1472 LASSERT(mdd_write_locked(env, obj));
1474 mdd_write_unlock(env, obj);
1477 mdd_trans_stop(env, mdd, 0, handle);
1481 ns->ln_mlinked_checked++;
1482 rc = mdd_lfsck_namespace_update(env, com, cfid,
1483 count != la->la_nlink ? LLF_UNMATCH_NLINKS : 0, false);
1489 mdd_write_unlock(env, obj);
1492 mdd_trans_stop(env, mdd, rc, handle);
1496 ns->ln_items_failed++;
1497 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1498 mdd_lfsck_pos_fill(env, lfsck,
1499 &ns->ln_pos_first_inconsistent,
1501 if (!(bk->lb_param & LPF_FAILOUT))
1505 ns->ln_items_repaired++;
1507 com->lc_journal = 0;
1510 up_write(&com->lc_sem);
1514 static int mdd_lfsck_namespace_post(const struct lu_env *env,
1515 struct lfsck_component *com,
1518 struct md_lfsck *lfsck = com->lc_lfsck;
1519 struct lfsck_namespace *ns =
1520 (struct lfsck_namespace *)com->lc_file_ram;
1523 down_write(&com->lc_sem);
1525 spin_lock(&lfsck->ml_lock);
1527 ns->ln_status = LS_SCANNING_PHASE2;
1528 ns->ln_flags |= LF_SCANNED_ONCE;
1529 ns->ln_flags &= ~LF_UPGRADE;
1530 cfs_list_del_init(&com->lc_link);
1531 cfs_list_del_init(&com->lc_link_dir);
1532 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_double_scan);
1533 } else if (result == 0) {
1534 if (lfsck->ml_paused) {
1535 ns->ln_status = LS_PAUSED;
1537 ns->ln_status = LS_STOPPED;
1538 cfs_list_del_init(&com->lc_link);
1539 cfs_list_del_init(&com->lc_link_dir);
1540 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1543 ns->ln_status = LS_FAILED;
1544 cfs_list_del_init(&com->lc_link);
1545 cfs_list_del_init(&com->lc_link_dir);
1546 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1548 spin_unlock(&lfsck->ml_lock);
1550 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1551 HALF_SEC - lfsck->ml_time_last_checkpoint);
1552 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1553 ns->ln_items_checked += com->lc_new_checked;
1554 com->lc_new_checked = 0;
1556 rc = mdd_lfsck_namespace_store(env, com, false);
1558 up_write(&com->lc_sem);
1563 mdd_lfsck_namespace_dump(const struct lu_env *env, struct lfsck_component *com,
1566 struct md_lfsck *lfsck = com->lc_lfsck;
1567 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1568 struct lfsck_namespace *ns =
1569 (struct lfsck_namespace *)com->lc_file_ram;
1574 down_read(&com->lc_sem);
1575 rc = snprintf(buf, len,
1576 "name: lfsck_namespace\n"
1582 lfsck_status_names[ns->ln_status]);
1588 rc = lfsck_bits_dump(&buf, &len, ns->ln_flags, lfsck_flags_names,
1593 rc = lfsck_bits_dump(&buf, &len, bk->lb_param, lfsck_param_names,
1598 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_complete,
1599 "time_since_last_completed");
1603 rc = lfsck_time_dump(&buf, &len, ns->ln_time_latest_start,
1604 "time_since_latest_start");
1608 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_checkpoint,
1609 "time_since_last_checkpoint");
1613 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_latest_start,
1614 "latest_start_position");
1618 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_last_checkpoint,
1619 "last_checkpoint_position");
1623 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_first_inconsistent,
1624 "first_failure_position");
1628 if (ns->ln_status == LS_SCANNING_PHASE1) {
1629 struct lfsck_position pos;
1630 cfs_duration_t duration = cfs_time_current() -
1631 lfsck->ml_time_last_checkpoint;
1632 __u64 checked = ns->ln_items_checked + com->lc_new_checked;
1633 __u64 speed = checked;
1634 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1635 __u32 rtime = ns->ln_run_time_phase1 +
1636 cfs_duration_sec(duration + HALF_SEC);
1639 do_div(new_checked, duration);
1641 do_div(speed, rtime);
1642 rc = snprintf(buf, len,
1643 "checked_phase1: "LPU64"\n"
1644 "checked_phase2: "LPU64"\n"
1645 "updated_phase1: "LPU64"\n"
1646 "updated_phase2: "LPU64"\n"
1647 "failed_phase1: "LPU64"\n"
1648 "failed_phase2: "LPU64"\n"
1650 "M-linked: "LPU64"\n"
1651 "nlinks_repaired: "LPU64"\n"
1652 "lost_found: "LPU64"\n"
1653 "success_count: %u\n"
1654 "run_time_phase1: %u seconds\n"
1655 "run_time_phase2: %u seconds\n"
1656 "average_speed_phase1: "LPU64" items/sec\n"
1657 "average_speed_phase2: N/A\n"
1658 "real-time_speed_phase1: "LPU64" items/sec\n"
1659 "real-time_speed_phase2: N/A\n",
1661 ns->ln_objs_checked_phase2,
1662 ns->ln_items_repaired,
1663 ns->ln_objs_repaired_phase2,
1664 ns->ln_items_failed,
1665 ns->ln_objs_failed_phase2,
1666 ns->ln_dirs_checked,
1667 ns->ln_mlinked_checked,
1668 ns->ln_objs_nlink_repaired,
1669 ns->ln_objs_lost_found,
1670 ns->ln_success_count,
1672 ns->ln_run_time_phase2,
1680 mdd_lfsck_pos_fill(env, lfsck, &pos, true, true);
1681 rc = lfsck_pos_dump(&buf, &len, &pos, "current_position");
1684 } else if (ns->ln_status == LS_SCANNING_PHASE2) {
1685 cfs_duration_t duration = cfs_time_current() -
1686 lfsck->ml_time_last_checkpoint;
1687 __u64 checked = ns->ln_objs_checked_phase2 +
1688 com->lc_new_checked;
1689 __u64 speed1 = ns->ln_items_checked;
1690 __u64 speed2 = checked;
1691 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1692 __u32 rtime = ns->ln_run_time_phase2 +
1693 cfs_duration_sec(duration + HALF_SEC);
1696 do_div(new_checked, duration);
1697 if (ns->ln_run_time_phase1 != 0)
1698 do_div(speed1, ns->ln_run_time_phase1);
1700 do_div(speed2, rtime);
1701 rc = snprintf(buf, len,
1702 "checked_phase1: "LPU64"\n"
1703 "checked_phase2: "LPU64"\n"
1704 "updated_phase1: "LPU64"\n"
1705 "updated_phase2: "LPU64"\n"
1706 "failed_phase1: "LPU64"\n"
1707 "failed_phase2: "LPU64"\n"
1709 "M-linked: "LPU64"\n"
1710 "nlinks_repaired: "LPU64"\n"
1711 "lost_found: "LPU64"\n"
1712 "success_count: %u\n"
1713 "run_time_phase1: %u seconds\n"
1714 "run_time_phase2: %u seconds\n"
1715 "average_speed_phase1: "LPU64" items/sec\n"
1716 "average_speed_phase2: "LPU64" objs/sec\n"
1717 "real-time_speed_phase1: N/A\n"
1718 "real-time_speed_phase2: "LPU64" objs/sec\n"
1719 "current_position: "DFID"\n",
1720 ns->ln_items_checked,
1722 ns->ln_items_repaired,
1723 ns->ln_objs_repaired_phase2,
1724 ns->ln_items_failed,
1725 ns->ln_objs_failed_phase2,
1726 ns->ln_dirs_checked,
1727 ns->ln_mlinked_checked,
1728 ns->ln_objs_nlink_repaired,
1729 ns->ln_objs_lost_found,
1730 ns->ln_success_count,
1731 ns->ln_run_time_phase1,
1736 PFID(&ns->ln_fid_latest_scanned_phase2));
1743 __u64 speed1 = ns->ln_items_checked;
1744 __u64 speed2 = ns->ln_objs_checked_phase2;
1746 if (ns->ln_run_time_phase1 != 0)
1747 do_div(speed1, ns->ln_run_time_phase1);
1748 if (ns->ln_run_time_phase2 != 0)
1749 do_div(speed2, ns->ln_run_time_phase2);
1750 rc = snprintf(buf, len,
1751 "checked_phase1: "LPU64"\n"
1752 "checked_phase2: "LPU64"\n"
1753 "updated_phase1: "LPU64"\n"
1754 "updated_phase2: "LPU64"\n"
1755 "failed_phase1: "LPU64"\n"
1756 "failed_phase2: "LPU64"\n"
1758 "M-linked: "LPU64"\n"
1759 "nlinks_repaired: "LPU64"\n"
1760 "lost_found: "LPU64"\n"
1761 "success_count: %u\n"
1762 "run_time_phase1: %u seconds\n"
1763 "run_time_phase2: %u seconds\n"
1764 "average_speed_phase1: "LPU64" items/sec\n"
1765 "average_speed_phase2: "LPU64" objs/sec\n"
1766 "real-time_speed_phase1: N/A\n"
1767 "real-time_speed_phase2: N/A\n"
1768 "current_position: N/A\n",
1769 ns->ln_items_checked,
1770 ns->ln_objs_checked_phase2,
1771 ns->ln_items_repaired,
1772 ns->ln_objs_repaired_phase2,
1773 ns->ln_items_failed,
1774 ns->ln_objs_failed_phase2,
1775 ns->ln_dirs_checked,
1776 ns->ln_mlinked_checked,
1777 ns->ln_objs_nlink_repaired,
1778 ns->ln_objs_lost_found,
1779 ns->ln_success_count,
1780 ns->ln_run_time_phase1,
1781 ns->ln_run_time_phase2,
1793 up_read(&com->lc_sem);
1797 static int mdd_lfsck_namespace_double_scan(const struct lu_env *env,
1798 struct lfsck_component *com)
1800 struct md_lfsck *lfsck = com->lc_lfsck;
1801 struct ptlrpc_thread *thread = &lfsck->ml_thread;
1802 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1803 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1804 struct lfsck_namespace *ns =
1805 (struct lfsck_namespace *)com->lc_file_ram;
1806 struct dt_object *obj = com->lc_obj;
1807 const struct dt_it_ops *iops = &obj->do_index_ops->dio_it;
1808 struct mdd_object *target;
1816 lfsck->ml_new_scanned = 0;
1817 lfsck->ml_time_last_checkpoint = cfs_time_current();
1818 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1819 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1821 di = iops->init(env, obj, 0, BYPASS_CAPA);
1823 RETURN(PTR_ERR(di));
1825 fid_cpu_to_be(&fid, &ns->ln_fid_latest_scanned_phase2);
1826 rc = iops->get(env, di, (const struct dt_key *)&fid);
1830 /* Skip the start one, which either has been processed or non-exist. */
1831 rc = iops->next(env, di);
1835 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_DOUBLESCAN))
1839 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY3) &&
1841 struct l_wait_info lwi;
1843 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
1845 l_wait_event(thread->t_ctl_waitq,
1846 !thread_is_running(thread),
1850 key = iops->key(env, di);
1851 fid_be_to_cpu(&fid, (const struct lu_fid *)key);
1852 target = mdd_object_find(env, mdd, &fid);
1853 down_write(&com->lc_sem);
1854 if (target == NULL) {
1857 } else if (IS_ERR(target)) {
1858 rc = PTR_ERR(target);
1862 /* XXX: need more processing for remote object in the future. */
1863 if (!mdd_object_exists(target) || mdd_object_remote(target))
1866 rc = iops->rec(env, di, (struct dt_rec *)&flags, 0);
1868 rc = mdd_lfsck_namespace_double_scan_one(env, com,
1872 mdd_object_put(env, target);
1875 lfsck->ml_new_scanned++;
1876 com->lc_new_checked++;
1877 ns->ln_fid_latest_scanned_phase2 = fid;
1879 ns->ln_objs_repaired_phase2++;
1881 ns->ln_objs_failed_phase2++;
1882 up_write(&com->lc_sem);
1884 if ((rc == 0) || ((rc > 0) && !(bk->lb_param & LPF_DRYRUN))) {
1885 mdd_lfsck_namespace_delete(env, com, &fid);
1886 } else if (rc < 0) {
1887 flags |= LLF_REPAIR_FAILED;
1888 mdd_lfsck_namespace_update(env, com, &fid, flags, true);
1891 if (rc < 0 && bk->lb_param & LPF_FAILOUT)
1894 if (likely(cfs_time_beforeq(cfs_time_current(),
1895 lfsck->ml_time_next_checkpoint)) ||
1896 com->lc_new_checked == 0)
1899 down_write(&com->lc_sem);
1900 ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1901 HALF_SEC - lfsck->ml_time_last_checkpoint);
1902 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1903 ns->ln_objs_checked_phase2 += com->lc_new_checked;
1904 com->lc_new_checked = 0;
1905 rc = mdd_lfsck_namespace_store(env, com, false);
1906 up_write(&com->lc_sem);
1910 lfsck->ml_time_last_checkpoint = cfs_time_current();
1911 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1912 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1915 mdd_lfsck_control_speed(lfsck);
1916 if (unlikely(!thread_is_running(thread)))
1919 rc = iops->next(env, di);
1928 iops->fini(env, di);
1929 down_write(&com->lc_sem);
1931 ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1932 HALF_SEC - lfsck->ml_time_last_checkpoint);
1933 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1934 ns->ln_objs_checked_phase2 += com->lc_new_checked;
1935 com->lc_new_checked = 0;
1938 com->lc_journal = 0;
1939 ns->ln_status = LS_COMPLETED;
1940 if (!(bk->lb_param & LPF_DRYRUN))
1942 ~(LF_SCANNED_ONCE | LF_INCONSISTENT | LF_UPGRADE);
1943 ns->ln_time_last_complete = ns->ln_time_last_checkpoint;
1944 ns->ln_success_count++;
1945 } else if (rc == 0) {
1946 if (lfsck->ml_paused)
1947 ns->ln_status = LS_PAUSED;
1949 ns->ln_status = LS_STOPPED;
1951 ns->ln_status = LS_FAILED;
1954 if (ns->ln_status != LS_PAUSED) {
1955 spin_lock(&lfsck->ml_lock);
1956 cfs_list_del_init(&com->lc_link);
1957 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1958 spin_unlock(&lfsck->ml_lock);
1961 rc = mdd_lfsck_namespace_store(env, com, false);
1963 up_write(&com->lc_sem);
1967 static struct lfsck_operations mdd_lfsck_namespace_ops = {
1968 .lfsck_reset = mdd_lfsck_namespace_reset,
1969 .lfsck_fail = mdd_lfsck_namespace_fail,
1970 .lfsck_checkpoint = mdd_lfsck_namespace_checkpoint,
1971 .lfsck_prep = mdd_lfsck_namespace_prep,
1972 .lfsck_exec_oit = mdd_lfsck_namespace_exec_oit,
1973 .lfsck_exec_dir = mdd_lfsck_namespace_exec_dir,
1974 .lfsck_post = mdd_lfsck_namespace_post,
1975 .lfsck_dump = mdd_lfsck_namespace_dump,
1976 .lfsck_double_scan = mdd_lfsck_namespace_double_scan,
1979 /* LFSCK component setup/cleanup functions */
1981 static int mdd_lfsck_namespace_setup(const struct lu_env *env,
1982 struct md_lfsck *lfsck)
1984 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1985 struct lfsck_component *com;
1986 struct lfsck_namespace *ns;
1987 struct dt_object *obj;
1995 CFS_INIT_LIST_HEAD(&com->lc_link);
1996 CFS_INIT_LIST_HEAD(&com->lc_link_dir);
1997 init_rwsem(&com->lc_sem);
1998 atomic_set(&com->lc_ref, 1);
1999 com->lc_lfsck = lfsck;
2000 com->lc_type = LT_NAMESPACE;
2001 com->lc_ops = &mdd_lfsck_namespace_ops;
2002 com->lc_file_size = sizeof(struct lfsck_namespace);
2003 OBD_ALLOC(com->lc_file_ram, com->lc_file_size);
2004 if (com->lc_file_ram == NULL)
2005 GOTO(out, rc = -ENOMEM);
2007 OBD_ALLOC(com->lc_file_disk, com->lc_file_size);
2008 if (com->lc_file_disk == NULL)
2009 GOTO(out, rc = -ENOMEM);
2011 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name,
2012 &mdd_env_info(env)->mti_fid);
2014 GOTO(out, rc = PTR_ERR(obj));
2017 rc = obj->do_ops->do_index_try(env, obj, &dt_lfsck_features);
2021 rc = mdd_lfsck_namespace_load(env, com);
2023 rc = mdd_lfsck_namespace_reset(env, com, true);
2024 else if (rc == -ENODATA)
2025 rc = mdd_lfsck_namespace_init(env, com);
2029 ns = (struct lfsck_namespace *)com->lc_file_ram;
2030 switch (ns->ln_status) {
2035 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
2038 CERROR("%s: unknown status: %u\n",
2039 mdd_lfsck2name(lfsck), ns->ln_status);
2041 case LS_SCANNING_PHASE1:
2042 case LS_SCANNING_PHASE2:
2043 /* No need to store the status to disk right now.
2044 * If the system crashed before the status stored,
2045 * it will be loaded back when next time. */
2046 ns->ln_status = LS_CRASHED;
2050 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_scan);
2051 cfs_list_add_tail(&com->lc_link_dir, &lfsck->ml_list_dir);
2059 mdd_lfsck_component_cleanup(env, com);
2063 /* helper functions for framework */
2065 static int object_is_client_visible(const struct lu_env *env,
2066 struct mdd_device *mdd,
2067 struct mdd_object *obj)
2069 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
2073 LASSERT(S_ISDIR(mdd_object_type(obj)));
2076 if (mdd_is_root(mdd, mdo2fid(obj))) {
2078 mdd_object_put(env, obj);
2082 mdd_read_lock(env, obj, MOR_TGT_CHILD);
2083 if (unlikely(mdd_is_dead_obj(obj))) {
2084 mdd_read_unlock(env, obj);
2086 mdd_object_put(env, obj);
2090 rc = dt_xattr_get(env, mdd_object_child(obj),
2091 mdd_buf_get(env, NULL, 0), XATTR_NAME_LINK,
2093 mdd_read_unlock(env, obj);
2096 mdd_object_put(env, obj);
2100 if (rc < 0 && rc != -ENODATA) {
2102 mdd_object_put(env, obj);
2106 rc = mdd_parent_fid(env, obj, fid);
2108 mdd_object_put(env, obj);
2112 if (unlikely(lu_fid_eq(fid, &mdd->mdd_local_root_fid)))
2115 obj = mdd_object_find(env, mdd, fid);
2118 else if (IS_ERR(obj))
2119 return PTR_ERR(obj);
2121 if (!mdd_object_exists(obj)) {
2122 mdd_object_put(env, obj);
2126 /* Currently, only client visible directory can be remote. */
2127 if (mdd_object_remote(obj)) {
2128 mdd_object_put(env, obj);
2137 static void mdd_lfsck_unpack_ent(struct lu_dirent *ent)
2139 fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid);
2140 ent->lde_hash = le64_to_cpu(ent->lde_hash);
2141 ent->lde_reclen = le16_to_cpu(ent->lde_reclen);
2142 ent->lde_namelen = le16_to_cpu(ent->lde_namelen);
2143 ent->lde_attrs = le32_to_cpu(ent->lde_attrs);
2145 /* Make sure the name is terminated with '0'.
2146 * The data (type) after ent::lde_name maybe
2147 * broken, but we do not care. */
2148 ent->lde_name[ent->lde_namelen] = 0;
2151 /* LFSCK wrap functions */
2153 static void mdd_lfsck_fail(const struct lu_env *env, struct md_lfsck *lfsck,
2154 bool oit, bool new_checked)
2156 struct lfsck_component *com;
2158 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2159 com->lc_ops->lfsck_fail(env, com, oit, new_checked);
2163 static int mdd_lfsck_checkpoint(const struct lu_env *env,
2164 struct md_lfsck *lfsck, bool oit)
2166 struct lfsck_component *com;
2169 if (likely(cfs_time_beforeq(cfs_time_current(),
2170 lfsck->ml_time_next_checkpoint)))
2173 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, oit, !oit);
2174 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2175 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
2180 lfsck->ml_time_last_checkpoint = cfs_time_current();
2181 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2182 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2186 static int mdd_lfsck_prep(struct lu_env *env, struct md_lfsck *lfsck)
2188 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2189 struct mdd_object *obj = NULL;
2190 struct dt_object *dt_obj;
2191 struct lfsck_component *com;
2192 struct lfsck_component *next;
2193 struct lfsck_position *pos = NULL;
2194 const struct dt_it_ops *iops =
2195 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2200 LASSERT(lfsck->ml_obj_dir == NULL);
2201 LASSERT(lfsck->ml_di_dir == NULL);
2203 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2204 com->lc_new_checked = 0;
2205 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2206 com->lc_journal = 0;
2208 rc = com->lc_ops->lfsck_prep(env, com);
2212 if ((pos == NULL) ||
2213 (!mdd_lfsck_pos_is_zero(&com->lc_pos_start) &&
2214 mdd_lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
2215 pos = &com->lc_pos_start;
2218 /* Init otable-based iterator. */
2220 rc = iops->load(env, lfsck->ml_di_oit, 0);
2222 lfsck->ml_oit_over = 1;
2229 rc = iops->load(env, lfsck->ml_di_oit, pos->lp_oit_cookie);
2233 lfsck->ml_oit_over = 1;
2235 if (fid_is_zero(&pos->lp_dir_parent))
2238 /* Find the directory for namespace-based traverse. */
2239 obj = mdd_object_find(env, mdd, &pos->lp_dir_parent);
2242 else if (IS_ERR(obj))
2243 RETURN(PTR_ERR(obj));
2245 /* XXX: need more processing for remote object in the future. */
2246 if (!mdd_object_exists(obj) || mdd_object_remote(obj) ||
2247 unlikely(!S_ISDIR(mdd_object_type(obj))))
2250 if (unlikely(mdd_is_dead_obj(obj)))
2253 dt_obj = mdd_object_child(obj);
2254 if (unlikely(!dt_try_as_dir(env, dt_obj)))
2255 GOTO(out, rc = -ENOTDIR);
2257 /* Init the namespace-based directory traverse. */
2258 iops = &dt_obj->do_index_ops->dio_it;
2259 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2261 GOTO(out, rc = PTR_ERR(di));
2263 rc = iops->load(env, di, pos->lp_dir_cookie);
2265 rc = iops->next(env, di);
2271 iops->fini(env, di);
2275 lfsck->ml_obj_dir = dt_obj;
2276 spin_lock(&lfsck->ml_lock);
2277 lfsck->ml_di_dir = di;
2278 spin_unlock(&lfsck->ml_lock);
2285 mdd_object_put(env, obj);
2288 return (rc > 0 ? 0 : rc);
2290 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, false, false);
2291 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2292 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
2297 lfsck->ml_time_last_checkpoint = cfs_time_current();
2298 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2299 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2303 static int mdd_lfsck_exec_oit(const struct lu_env *env, struct md_lfsck *lfsck,
2304 struct mdd_object *obj)
2306 struct lfsck_component *com;
2307 struct dt_object *dt_obj;
2308 const struct dt_it_ops *iops;
2313 LASSERT(lfsck->ml_obj_dir == NULL);
2315 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2316 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
2321 if (!S_ISDIR(mdd_object_type(obj)) ||
2322 cfs_list_empty(&lfsck->ml_list_dir))
2325 rc = object_is_client_visible(env, mdd_lfsck2mdd(lfsck), obj);
2329 if (unlikely(mdd_is_dead_obj(obj)))
2332 dt_obj = mdd_object_child(obj);
2333 if (unlikely(!dt_try_as_dir(env, dt_obj)))
2334 GOTO(out, rc = -ENOTDIR);
2336 iops = &dt_obj->do_index_ops->dio_it;
2337 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2339 GOTO(out, rc = PTR_ERR(di));
2341 rc = iops->load(env, di, 0);
2343 rc = iops->next(env, di);
2349 iops->fini(env, di);
2353 mdd_object_get(obj);
2354 lfsck->ml_obj_dir = dt_obj;
2355 spin_lock(&lfsck->ml_lock);
2356 lfsck->ml_di_dir = di;
2357 spin_unlock(&lfsck->ml_lock);
2363 mdd_lfsck_fail(env, lfsck, false, false);
2364 return (rc > 0 ? 0 : rc);
2367 static int mdd_lfsck_exec_dir(const struct lu_env *env, struct md_lfsck *lfsck,
2368 struct mdd_object *obj, struct lu_dirent *ent)
2370 struct lfsck_component *com;
2373 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2374 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
2381 static int mdd_lfsck_post(const struct lu_env *env, struct md_lfsck *lfsck,
2384 struct lfsck_component *com;
2385 struct lfsck_component *next;
2388 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, true, true);
2389 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2390 rc = com->lc_ops->lfsck_post(env, com, result);
2395 lfsck->ml_time_last_checkpoint = cfs_time_current();
2396 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2397 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2401 static int mdd_lfsck_double_scan(const struct lu_env *env,
2402 struct md_lfsck *lfsck)
2404 struct lfsck_component *com;
2405 struct lfsck_component *next;
2408 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_double_scan,
2410 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2411 com->lc_journal = 0;
2413 rc = com->lc_ops->lfsck_double_scan(env, com);
2422 static int mdd_lfsck_dir_engine(const struct lu_env *env,
2423 struct md_lfsck *lfsck)
2425 struct mdd_thread_info *info = mdd_env_info(env);
2426 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2427 const struct dt_it_ops *iops =
2428 &lfsck->ml_obj_dir->do_index_ops->dio_it;
2429 struct dt_it *di = lfsck->ml_di_dir;
2430 struct lu_dirent *ent = &info->mti_ent;
2431 struct lu_fid *fid = &info->mti_fid;
2432 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2433 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2438 struct mdd_object *child;
2440 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY2) &&
2442 struct l_wait_info lwi;
2444 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2446 l_wait_event(thread->t_ctl_waitq,
2447 !thread_is_running(thread),
2451 lfsck->ml_new_scanned++;
2452 rc = iops->rec(env, di, (struct dt_rec *)ent,
2453 lfsck->ml_args_dir);
2455 mdd_lfsck_fail(env, lfsck, false, true);
2456 if (bk->lb_param & LPF_FAILOUT)
2462 mdd_lfsck_unpack_ent(ent);
2463 if (ent->lde_attrs & LUDA_IGNORE)
2466 *fid = ent->lde_fid;
2467 child = mdd_object_find(env, mdd, fid);
2468 if (child == NULL) {
2470 } else if (IS_ERR(child)) {
2471 mdd_lfsck_fail(env, lfsck, false, true);
2472 if (bk->lb_param & LPF_FAILOUT)
2473 RETURN(PTR_ERR(child));
2478 /* XXX: need more processing for remote object in the future. */
2479 if (mdd_object_exists(child) && !mdd_object_remote(child))
2480 rc = mdd_lfsck_exec_dir(env, lfsck, child, ent);
2481 mdd_object_put(env, child);
2482 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2486 rc = mdd_lfsck_checkpoint(env, lfsck, false);
2487 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2491 mdd_lfsck_control_speed(lfsck);
2492 if (unlikely(!thread_is_running(thread)))
2495 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL2)) {
2496 spin_lock(&lfsck->ml_lock);
2497 thread_set_flags(thread, SVC_STOPPING);
2498 spin_unlock(&lfsck->ml_lock);
2502 rc = iops->next(env, di);
2505 if (rc > 0 && !lfsck->ml_oit_over)
2506 mdd_lfsck_close_dir(env, lfsck);
2511 static int mdd_lfsck_oit_engine(const struct lu_env *env,
2512 struct md_lfsck *lfsck)
2514 struct mdd_thread_info *info = mdd_env_info(env);
2515 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2516 const struct dt_it_ops *iops =
2517 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2518 struct dt_it *di = lfsck->ml_di_oit;
2519 struct lu_fid *fid = &info->mti_fid;
2520 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2521 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2526 struct mdd_object *target;
2528 if (lfsck->ml_di_dir != NULL) {
2529 rc = mdd_lfsck_dir_engine(env, lfsck);
2534 if (unlikely(lfsck->ml_oit_over))
2537 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY1) &&
2539 struct l_wait_info lwi;
2541 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2543 l_wait_event(thread->t_ctl_waitq,
2544 !thread_is_running(thread),
2548 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2551 lfsck->ml_new_scanned++;
2552 rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
2554 mdd_lfsck_fail(env, lfsck, true, true);
2555 if (bk->lb_param & LPF_FAILOUT)
2561 target = mdd_object_find(env, mdd, fid);
2562 if (target == NULL) {
2564 } else if (IS_ERR(target)) {
2565 mdd_lfsck_fail(env, lfsck, true, true);
2566 if (bk->lb_param & LPF_FAILOUT)
2567 RETURN(PTR_ERR(target));
2572 /* XXX: In fact, low layer otable-based iteration should not
2573 * return agent object. But before LU-2646 resolved, we
2574 * need more processing for agent object. */
2575 if (mdd_object_exists(target) && !mdd_object_remote(target))
2576 rc = mdd_lfsck_exec_oit(env, lfsck, target);
2577 mdd_object_put(env, target);
2578 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2582 rc = mdd_lfsck_checkpoint(env, lfsck, true);
2583 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2587 mdd_lfsck_control_speed(lfsck);
2589 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL1)) {
2590 spin_lock(&lfsck->ml_lock);
2591 thread_set_flags(thread, SVC_STOPPING);
2592 spin_unlock(&lfsck->ml_lock);
2596 rc = iops->next(env, di);
2598 lfsck->ml_oit_over = 1;
2600 if (unlikely(!thread_is_running(thread)))
2602 } while (rc == 0 || lfsck->ml_di_dir != NULL);
2607 static int mdd_lfsck_main(void *args)
2610 struct md_lfsck *lfsck = (struct md_lfsck *)args;
2611 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2612 struct dt_object *oit_obj = lfsck->ml_obj_oit;
2613 const struct dt_it_ops *oit_iops = &oit_obj->do_index_ops->dio_it;
2614 struct dt_it *oit_di;
2618 cfs_daemonize("lfsck");
2619 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
2621 CERROR("%s: LFSCK, fail to init env, rc = %d\n",
2622 mdd_lfsck2name(lfsck), rc);
2626 oit_di = oit_iops->init(&env, oit_obj, lfsck->ml_args_oit, BYPASS_CAPA);
2627 if (IS_ERR(oit_di)) {
2628 rc = PTR_ERR(oit_di);
2629 CERROR("%s: LFSCK, fail to init iteration, rc = %d\n",
2630 mdd_lfsck2name(lfsck), rc);
2634 spin_lock(&lfsck->ml_lock);
2635 lfsck->ml_di_oit = oit_di;
2636 spin_unlock(&lfsck->ml_lock);
2637 rc = mdd_lfsck_prep(&env, lfsck);
2641 CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = 0x%x, dir_flags = 0x%x, "
2642 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2643 ", pid = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2644 lfsck->ml_pos_current.lp_oit_cookie,
2645 lfsck->ml_pos_current.lp_dir_cookie,
2646 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2649 spin_lock(&lfsck->ml_lock);
2650 thread_set_flags(thread, SVC_RUNNING);
2651 spin_unlock(&lfsck->ml_lock);
2652 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2654 if (!cfs_list_empty(&lfsck->ml_list_scan) ||
2655 cfs_list_empty(&lfsck->ml_list_double_scan))
2656 rc = mdd_lfsck_oit_engine(&env, lfsck);
2660 CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = 0x%x, dir_flags = 0x%x, "
2661 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2662 ", pid = %d, rc = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2663 lfsck->ml_pos_current.lp_oit_cookie,
2664 lfsck->ml_pos_current.lp_dir_cookie,
2665 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2666 cfs_curproc_pid(), rc);
2668 if (lfsck->ml_paused && cfs_list_empty(&lfsck->ml_list_scan))
2669 oit_iops->put(&env, oit_di);
2671 if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2672 rc = mdd_lfsck_post(&env, lfsck, rc);
2673 if (lfsck->ml_di_dir != NULL)
2674 mdd_lfsck_close_dir(&env, lfsck);
2677 spin_lock(&lfsck->ml_lock);
2678 lfsck->ml_di_oit = NULL;
2679 spin_unlock(&lfsck->ml_lock);
2681 oit_iops->fini(&env, oit_di);
2683 if (!cfs_list_empty(&lfsck->ml_list_double_scan))
2684 rc = mdd_lfsck_double_scan(&env, lfsck);
2689 /* XXX: Purge the pinned objects in the future. */
2695 spin_lock(&lfsck->ml_lock);
2696 thread_set_flags(thread, SVC_STOPPED);
2697 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2698 spin_unlock(&lfsck->ml_lock);
2702 /* external interfaces */
2704 int mdd_lfsck_set_speed(const struct lu_env *env, struct md_lfsck *lfsck,
2709 mutex_lock(&lfsck->ml_mutex);
2710 __mdd_lfsck_set_speed(lfsck, limit);
2711 rc = mdd_lfsck_bookmark_store(env, lfsck);
2712 mutex_unlock(&lfsck->ml_mutex);
2716 int mdd_lfsck_dump(const struct lu_env *env, struct md_lfsck *lfsck,
2717 __u16 type, char *buf, int len)
2719 struct lfsck_component *com;
2722 if (!lfsck->ml_initialized)
2725 com = mdd_lfsck_component_find(lfsck, type);
2729 rc = com->lc_ops->lfsck_dump(env, com, buf, len);
2730 mdd_lfsck_component_put(env, com);
2734 int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
2735 struct lfsck_start *start)
2737 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2738 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2739 struct lfsck_component *com;
2740 struct l_wait_info lwi = { 0 };
2747 if (lfsck->ml_obj_oit == NULL)
2750 /* start == NULL means auto trigger paused LFSCK. */
2751 if ((start == NULL) &&
2752 (cfs_list_empty(&lfsck->ml_list_scan) ||
2753 OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
2756 mutex_lock(&lfsck->ml_mutex);
2757 spin_lock(&lfsck->ml_lock);
2758 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2759 spin_unlock(&lfsck->ml_lock);
2760 mutex_unlock(&lfsck->ml_mutex);
2764 spin_unlock(&lfsck->ml_lock);
2766 lfsck->ml_paused = 0;
2767 lfsck->ml_oit_over = 0;
2768 lfsck->ml_drop_dryrun = 0;
2769 lfsck->ml_new_scanned = 0;
2771 /* For auto trigger. */
2775 start->ls_version = bk->lb_version;
2776 if (start->ls_valid & LSV_SPEED_LIMIT) {
2777 __mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
2781 if (start->ls_valid & LSV_ERROR_HANDLE) {
2782 valid |= DOIV_ERROR_HANDLE;
2783 if (start->ls_flags & LPF_FAILOUT)
2784 flags |= DOIF_FAILOUT;
2786 if ((start->ls_flags & LPF_FAILOUT) &&
2787 !(bk->lb_param & LPF_FAILOUT)) {
2788 bk->lb_param |= LPF_FAILOUT;
2790 } else if (!(start->ls_flags & LPF_FAILOUT) &&
2791 (bk->lb_param & LPF_FAILOUT)) {
2792 bk->lb_param &= ~LPF_FAILOUT;
2797 if (start->ls_valid & LSV_DRYRUN) {
2798 if ((start->ls_flags & LPF_DRYRUN) &&
2799 !(bk->lb_param & LPF_DRYRUN)) {
2800 bk->lb_param |= LPF_DRYRUN;
2802 } else if (!(start->ls_flags & LPF_DRYRUN) &&
2803 (bk->lb_param & LPF_DRYRUN)) {
2804 bk->lb_param &= ~LPF_DRYRUN;
2805 lfsck->ml_drop_dryrun = 1;
2811 rc = mdd_lfsck_bookmark_store(env, lfsck);
2816 if (start->ls_flags & LPF_RESET)
2817 flags |= DOIF_RESET;
2819 if (start->ls_active != 0) {
2820 struct lfsck_component *next;
2823 if (start->ls_active == LFSCK_TYPES_ALL)
2824 start->ls_active = LFSCK_TYPES_SUPPORTED;
2826 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2827 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2828 GOTO(out, rc = -ENOTSUPP);
2831 cfs_list_for_each_entry_safe(com, next,
2832 &lfsck->ml_list_scan, lc_link) {
2833 if (!(com->lc_type & start->ls_active)) {
2834 rc = com->lc_ops->lfsck_post(env, com, 0);
2840 while (start->ls_active != 0) {
2841 if (type & start->ls_active) {
2842 com = __mdd_lfsck_component_find(lfsck, type,
2843 &lfsck->ml_list_idle);
2845 /* The component status will be updated
2846 * when its prep() is called later by
2847 * the LFSCK main engine. */
2848 cfs_list_del_init(&com->lc_link);
2849 cfs_list_add_tail(&com->lc_link,
2850 &lfsck->ml_list_scan);
2852 start->ls_active &= ~type;
2858 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2859 start->ls_active |= com->lc_type;
2860 if (flags & DOIF_RESET) {
2861 rc = com->lc_ops->lfsck_reset(env, com, false);
2868 lfsck->ml_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2869 if (bk->lb_param & LPF_DRYRUN)
2870 lfsck->ml_args_dir |= LUDA_VERIFY_DRYRUN;
2872 if (bk->lb_param & LPF_FAILOUT) {
2873 valid |= DOIV_ERROR_HANDLE;
2874 flags |= DOIF_FAILOUT;
2877 if (!cfs_list_empty(&lfsck->ml_list_scan))
2878 flags |= DOIF_OUTUSED;
2880 lfsck->ml_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2881 thread_set_flags(thread, 0);
2882 rc = cfs_create_thread(mdd_lfsck_main, lfsck, 0);
2884 CERROR("%s: cannot start LFSCK thread, rc = %d\n",
2885 mdd_lfsck2name(lfsck), rc);
2887 l_wait_event(thread->t_ctl_waitq,
2888 thread_is_running(thread) ||
2889 thread_is_stopped(thread),
2895 mutex_unlock(&lfsck->ml_mutex);
2896 return (rc < 0 ? rc : 0);
2899 int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck,
2902 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2903 struct l_wait_info lwi = { 0 };
2906 if (!lfsck->ml_initialized)
2909 mutex_lock(&lfsck->ml_mutex);
2910 spin_lock(&lfsck->ml_lock);
2911 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2912 spin_unlock(&lfsck->ml_lock);
2913 mutex_unlock(&lfsck->ml_mutex);
2918 lfsck->ml_paused = 1;
2919 thread_set_flags(thread, SVC_STOPPING);
2920 /* The LFSCK thread may be sleeping on low layer wait queue,
2922 if (likely(lfsck->ml_di_oit != NULL))
2923 lfsck->ml_obj_oit->do_index_ops->dio_it.put(env,
2925 spin_unlock(&lfsck->ml_lock);
2927 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2928 l_wait_event(thread->t_ctl_waitq,
2929 thread_is_stopped(thread),
2931 mutex_unlock(&lfsck->ml_mutex);
2936 static const struct lu_fid lfsck_it_fid = { .f_seq = FID_SEQ_LOCAL_FILE,
2937 .f_oid = OTABLE_IT_OID,
2940 int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd)
2942 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2943 struct dt_object *obj;
2947 LASSERT(!lfsck->ml_initialized);
2949 lfsck->ml_initialized = 1;
2950 mutex_init(&lfsck->ml_mutex);
2951 spin_lock_init(&lfsck->ml_lock);
2952 CFS_INIT_LIST_HEAD(&lfsck->ml_list_scan);
2953 CFS_INIT_LIST_HEAD(&lfsck->ml_list_dir);
2954 CFS_INIT_LIST_HEAD(&lfsck->ml_list_double_scan);
2955 CFS_INIT_LIST_HEAD(&lfsck->ml_list_idle);
2956 cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
2958 obj = dt_locate(env, mdd->mdd_bottom, &lfsck_it_fid);
2960 RETURN(PTR_ERR(obj));
2962 lfsck->ml_obj_oit = obj;
2963 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2965 if (rc == -ENOTSUPP)
2971 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_bookmark_name,
2972 &mdd_env_info(env)->mti_fid);
2974 RETURN(PTR_ERR(obj));
2976 lfsck->ml_bookmark_obj = obj;
2977 rc = mdd_lfsck_bookmark_load(env, lfsck);
2979 rc = mdd_lfsck_bookmark_init(env, lfsck);
2983 rc = mdd_lfsck_namespace_setup(env, lfsck);
2984 /* XXX: LFSCK components initialization to be added here. */
2989 void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd)
2991 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2992 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2993 struct lfsck_component *com;
2995 if (!lfsck->ml_initialized)
2998 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
3000 if (lfsck->ml_obj_oit != NULL) {
3001 lu_object_put(env, &lfsck->ml_obj_oit->do_lu);
3002 lfsck->ml_obj_oit = NULL;
3005 LASSERT(lfsck->ml_obj_dir == NULL);
3007 if (lfsck->ml_bookmark_obj != NULL) {
3008 lu_object_put(env, &lfsck->ml_bookmark_obj->do_lu);
3009 lfsck->ml_bookmark_obj = NULL;
3012 while (!cfs_list_empty(&lfsck->ml_list_scan)) {
3013 com = cfs_list_entry(lfsck->ml_list_scan.next,
3014 struct lfsck_component,
3016 mdd_lfsck_component_cleanup(env, com);
3019 LASSERT(cfs_list_empty(&lfsck->ml_list_dir));
3021 while (!cfs_list_empty(&lfsck->ml_list_double_scan)) {
3022 com = cfs_list_entry(lfsck->ml_list_double_scan.next,
3023 struct lfsck_component,
3025 mdd_lfsck_component_cleanup(env, com);
3028 while (!cfs_list_empty(&lfsck->ml_list_idle)) {
3029 com = cfs_list_entry(lfsck->ml_list_idle.next,
3030 struct lfsck_component,
3032 mdd_lfsck_component_cleanup(env, com);