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;
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 rc = dt_delete(env, parent, (struct dt_key *)lfsck_namespace_name,
741 handle, BYPASS_CAPA);
745 rc = child->do_ops->do_ref_del(env, child, handle);
747 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
748 rc = dt_insert(env, parent,
749 (const struct dt_rec*)fid,
750 (const struct dt_key *)lfsck_namespace_name,
751 handle, BYPASS_CAPA, 1);
757 rc = dt_destroy(env, child, handle);
759 lu_object_put(env, &child->do_lu);
766 dt_trans_stop(env, mdd->mdd_bottom, handle);
769 lu_object_put(env, &parent->do_lu);
773 static int mdd_lfsck_namespace_lookup(const struct lu_env *env,
774 struct lfsck_component *com,
775 const struct lu_fid *fid,
778 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
781 fid_cpu_to_be(key, fid);
782 rc = dt_lookup(env, com->lc_obj, (struct dt_rec *)flags,
783 (const struct dt_key *)key, BYPASS_CAPA);
787 static int mdd_lfsck_namespace_update(const struct lu_env *env,
788 struct lfsck_component *com,
789 const struct lu_fid *fid,
790 __u8 flags, bool force)
792 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
793 struct lu_fid *key = &mdd_env_info(env)->mti_fid;
794 struct thandle *handle;
795 struct dt_object *obj = com->lc_obj;
801 rc = mdd_lfsck_namespace_lookup(env, com, fid, &tf);
802 if (rc != 0 && rc != -ENOENT)
806 if (!force || flags == tf)
810 handle = dt_trans_create(env, mdd->mdd_bottom);
812 RETURN(PTR_ERR(handle));
814 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid,
819 handle = dt_trans_create(env, mdd->mdd_bottom);
821 RETURN(PTR_ERR(handle));
824 rc = dt_declare_insert(env, obj, (const struct dt_rec *)&flags,
825 (const struct dt_key *)fid, handle);
829 rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
833 fid_cpu_to_be(key, fid);
835 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
838 CERROR("%s: fail to insert "DFID", rc = %d\n",
839 mdd_lfsck2name(com->lc_lfsck), PFID(fid), rc);
844 rc = dt_insert(env, obj, (const struct dt_rec *)&flags,
845 (const struct dt_key *)key, handle, BYPASS_CAPA, 1);
850 dt_trans_stop(env, mdd->mdd_bottom, handle);
856 static int mdd_lfsck_namespace_reset(const struct lu_env *env,
857 struct lfsck_component *com, bool init)
859 struct mdd_thread_info *info = mdd_env_info(env);
860 struct lu_fid *fid = &info->mti_fid;
861 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
862 struct mdd_device *mdd = mdd_lfsck2mdd(com->lc_lfsck);
863 struct md_object *mdo;
864 struct dt_object *dto;
868 down_write(&com->lc_sem);
870 memset(ns, 0, sizeof(*ns));
872 __u32 count = ns->ln_success_count;
873 __u64 last_time = ns->ln_time_last_complete;
875 memset(ns, 0, sizeof(*ns));
876 ns->ln_success_count = count;
877 ns->ln_time_last_complete = last_time;
879 ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
880 ns->ln_status = LS_INIT;
882 rc = mdd_lfsck_namespace_unlink(env, mdd, com);
886 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
887 mdo = llo_store_create_index(env, &mdd->mdd_md_dev, mdd->mdd_bottom, "",
888 lfsck_namespace_name, fid,
891 GOTO(out, rc = PTR_ERR(mdo));
893 lu_object_put(env, &mdo->mo_lu);
894 dto = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name, fid);
896 GOTO(out, rc = PTR_ERR(dto));
899 rc = dto->do_ops->do_index_try(env, dto, &dt_lfsck_features);
903 rc = mdd_lfsck_namespace_store(env, com, true);
908 up_write(&com->lc_sem);
913 mdd_lfsck_namespace_fail(const struct lu_env *env, struct lfsck_component *com,
914 bool oit, bool new_checked)
916 struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
918 down_write(&com->lc_sem);
920 com->lc_new_checked++;
921 ns->ln_items_failed++;
922 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
923 mdd_lfsck_pos_fill(env, com->lc_lfsck,
924 &ns->ln_pos_first_inconsistent, oit, !oit);
925 up_write(&com->lc_sem);
928 static int mdd_lfsck_namespace_checkpoint(const struct lu_env *env,
929 struct lfsck_component *com,
932 struct md_lfsck *lfsck = com->lc_lfsck;
933 struct lfsck_namespace *ns =
934 (struct lfsck_namespace *)com->lc_file_ram;
937 if (com->lc_new_checked == 0 && !init)
940 down_write(&com->lc_sem);
942 ns->ln_pos_last_checkpoint = lfsck->ml_pos_current;
944 ns->ln_time_last_checkpoint = ns->ln_time_latest_start;
945 ns->ln_pos_latest_start = lfsck->ml_pos_current;
947 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
948 HALF_SEC - lfsck->ml_time_last_checkpoint);
949 ns->ln_time_last_checkpoint = cfs_time_current_sec();
950 ns->ln_items_checked += com->lc_new_checked;
951 com->lc_new_checked = 0;
954 rc = mdd_lfsck_namespace_store(env, com, false);
956 up_write(&com->lc_sem);
960 static int mdd_lfsck_namespace_prep(const struct lu_env *env,
961 struct lfsck_component *com)
963 struct md_lfsck *lfsck = com->lc_lfsck;
964 struct lfsck_namespace *ns =
965 (struct lfsck_namespace *)com->lc_file_ram;
966 struct lfsck_position *pos = &com->lc_pos_start;
968 if (ns->ln_status == LS_COMPLETED) {
971 rc = mdd_lfsck_namespace_reset(env, com, false);
976 down_write(&com->lc_sem);
978 ns->ln_time_latest_start = cfs_time_current_sec();
980 spin_lock(&lfsck->ml_lock);
981 if (ns->ln_flags & LF_SCANNED_ONCE) {
982 if (!lfsck->ml_drop_dryrun ||
983 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
984 ns->ln_status = LS_SCANNING_PHASE2;
985 cfs_list_del_init(&com->lc_link);
986 cfs_list_add_tail(&com->lc_link,
987 &lfsck->ml_list_double_scan);
988 if (!cfs_list_empty(&com->lc_link_dir))
989 cfs_list_del_init(&com->lc_link_dir);
990 mdd_lfsck_pos_set_zero(pos);
992 ns->ln_status = LS_SCANNING_PHASE1;
993 ns->ln_run_time_phase1 = 0;
994 ns->ln_run_time_phase2 = 0;
995 ns->ln_items_checked = 0;
996 ns->ln_items_repaired = 0;
997 ns->ln_items_failed = 0;
998 ns->ln_dirs_checked = 0;
999 ns->ln_mlinked_checked = 0;
1000 ns->ln_objs_checked_phase2 = 0;
1001 ns->ln_objs_repaired_phase2 = 0;
1002 ns->ln_objs_failed_phase2 = 0;
1003 ns->ln_objs_nlink_repaired = 0;
1004 ns->ln_objs_lost_found = 0;
1005 fid_zero(&ns->ln_fid_latest_scanned_phase2);
1006 if (cfs_list_empty(&com->lc_link_dir))
1007 cfs_list_add_tail(&com->lc_link_dir,
1008 &lfsck->ml_list_dir);
1009 *pos = ns->ln_pos_first_inconsistent;
1012 ns->ln_status = LS_SCANNING_PHASE1;
1013 if (cfs_list_empty(&com->lc_link_dir))
1014 cfs_list_add_tail(&com->lc_link_dir,
1015 &lfsck->ml_list_dir);
1016 if (!lfsck->ml_drop_dryrun ||
1017 mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1018 *pos = ns->ln_pos_last_checkpoint;
1019 pos->lp_oit_cookie++;
1020 if (!fid_is_zero(&pos->lp_dir_parent)) {
1021 if (pos->lp_dir_cookie == MDS_DIR_END_OFF) {
1022 fid_zero(&pos->lp_dir_parent);
1024 pos->lp_dir_cookie++;
1028 *pos = ns->ln_pos_first_inconsistent;
1031 spin_unlock(&lfsck->ml_lock);
1033 up_write(&com->lc_sem);
1037 static int mdd_lfsck_namespace_exec_oit(const struct lu_env *env,
1038 struct lfsck_component *com,
1039 struct mdd_object *obj)
1041 down_write(&com->lc_sem);
1042 com->lc_new_checked++;
1043 if (S_ISDIR(mdd_object_type(obj)))
1044 ((struct lfsck_namespace *)com->lc_file_ram)->ln_dirs_checked++;
1045 up_write(&com->lc_sem);
1049 static int mdd_declare_lfsck_namespace_exec_dir(const struct lu_env *env,
1050 struct mdd_object *obj,
1051 struct thandle *handle)
1055 /* For destroying all invalid linkEA entries. */
1056 rc = mdo_declare_xattr_del(env, obj, XATTR_NAME_LINK, handle);
1060 /* For insert new linkEA entry. */
1061 rc = mdd_declare_links_add(env, obj, handle);
1065 static int mdd_lfsck_namespace_check_exist(const struct lu_env *env,
1066 struct md_lfsck *lfsck,
1067 struct mdd_object *obj,
1070 struct dt_object *dir = lfsck->ml_obj_dir;
1071 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
1075 if (unlikely(mdd_is_dead_obj(obj)))
1076 RETURN(LFSCK_NAMEENTRY_DEAD);
1078 rc = dt_lookup(env, dir, (struct dt_rec *)fid,
1079 (const struct dt_key *)name, BYPASS_CAPA);
1081 RETURN(LFSCK_NAMEENTRY_REMOVED);
1086 if (!lu_fid_eq(fid, mdo2fid(obj)))
1087 RETURN(LFSCK_NAMEENTRY_RECREATED);
1092 static int mdd_lfsck_namespace_exec_dir(const struct lu_env *env,
1093 struct lfsck_component *com,
1094 struct mdd_object *obj,
1095 struct lu_dirent *ent)
1097 struct mdd_thread_info *info = mdd_env_info(env);
1098 struct lu_attr *la = &info->mti_la;
1099 struct md_lfsck *lfsck = com->lc_lfsck;
1100 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1101 struct lfsck_namespace *ns =
1102 (struct lfsck_namespace *)com->lc_file_ram;
1103 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1104 struct mdd_link_data ldata = { 0 };
1105 const struct lu_fid *pfid =
1106 lu_object_fid(&lfsck->ml_obj_dir->do_lu);
1107 const struct lu_fid *cfid = mdo2fid(obj);
1108 const struct lu_name *cname;
1109 struct thandle *handle = NULL;
1110 bool repaired = false;
1111 bool locked = false;
1116 cname = mdd_name_get_const(env, ent->lde_name, ent->lde_namelen);
1117 down_write(&com->lc_sem);
1118 com->lc_new_checked++;
1120 if (ent->lde_attrs & LUDA_UPGRADE) {
1121 ns->ln_flags |= LF_UPGRADE;
1123 } else if (ent->lde_attrs & LUDA_REPAIR) {
1124 ns->ln_flags |= LF_INCONSISTENT;
1128 if (ent->lde_name[0] == '.' &&
1129 (ent->lde_namelen == 1 ||
1130 (ent->lde_namelen == 2 && ent->lde_name[1] == '.')))
1133 if (!(bk->lb_param & LPF_DRYRUN) &&
1134 (com->lc_journal || repaired)) {
1139 com->lc_journal = 1;
1140 handle = mdd_trans_create(env, mdd);
1142 GOTO(out, rc = PTR_ERR(handle));
1144 rc = mdd_declare_lfsck_namespace_exec_dir(env, obj, handle);
1148 rc = mdd_trans_start(env, mdd, handle);
1152 mdd_write_lock(env, obj, MOR_TGT_CHILD);
1156 rc = mdd_lfsck_namespace_check_exist(env, lfsck, obj, ent->lde_name);
1160 rc = mdd_links_read(env, obj, &ldata);
1162 count = ldata.ml_leh->leh_reccount;
1163 rc = mdd_links_find(env, obj, &ldata, cname, pfid);
1165 /* For dir, if there are more than one linkea entries,
1166 * then remove all the other redundant linkea entries.*/
1167 if (unlikely(count > 1 &&
1168 S_ISDIR(mdd_object_type(obj))))
1175 ns->ln_flags |= LF_INCONSISTENT;
1176 if (bk->lb_param & LPF_DRYRUN) {
1181 /*For dir, remove the unmatched linkea entry directly.*/
1182 if (S_ISDIR(mdd_object_type(obj))) {
1183 if (!com->lc_journal)
1186 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK,
1187 handle, BYPASS_CAPA);
1196 } else if (unlikely(rc == -EINVAL)) {
1197 ns->ln_flags |= LF_INCONSISTENT;
1198 if (bk->lb_param & LPF_DRYRUN) {
1204 if (!com->lc_journal)
1207 /* The magic crashed, we are not sure whether there are more
1208 * corrupt data in the linkea, so remove all linkea entries. */
1209 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK, handle,
1215 } else if (rc == -ENODATA) {
1216 ns->ln_flags |= LF_UPGRADE;
1217 if (bk->lb_param & LPF_DRYRUN) {
1224 rc = mdd_links_new(env, &ldata);
1229 if (!com->lc_journal)
1232 rc = mdd_links_add_buf(env, &ldata, cname, pfid);
1236 rc = mdd_links_write(env, obj, &ldata, handle);
1240 count = ldata.ml_leh->leh_reccount;
1249 rc = mdd_la_get(env, obj, la, BYPASS_CAPA);
1254 (la->la_nlink == 1 || S_ISDIR(mdd_object_type(obj))))
1255 /* Usually, it is for single linked object or dir, do nothing.*/
1258 /* Following modification will be in another transaction. */
1259 if (handle != NULL) {
1260 LASSERT(mdd_write_locked(env, obj));
1262 mdd_write_unlock(env, obj);
1265 mdd_trans_stop(env, mdd, 0, handle);
1269 ns->ln_mlinked_checked++;
1270 rc = mdd_lfsck_namespace_update(env, com, cfid,
1271 count != la->la_nlink ? LLF_UNMATCH_NLINKS : 0, false);
1277 mdd_write_unlock(env, obj);
1280 mdd_trans_stop(env, mdd, rc, handle);
1284 ns->ln_items_failed++;
1285 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1286 mdd_lfsck_pos_fill(env, lfsck,
1287 &ns->ln_pos_first_inconsistent,
1289 if (!(bk->lb_param & LPF_FAILOUT))
1293 ns->ln_items_repaired++;
1295 com->lc_journal = 0;
1298 up_write(&com->lc_sem);
1302 static int mdd_lfsck_namespace_post(const struct lu_env *env,
1303 struct lfsck_component *com,
1306 struct md_lfsck *lfsck = com->lc_lfsck;
1307 struct lfsck_namespace *ns =
1308 (struct lfsck_namespace *)com->lc_file_ram;
1311 down_write(&com->lc_sem);
1313 spin_lock(&lfsck->ml_lock);
1315 ns->ln_status = LS_SCANNING_PHASE2;
1316 ns->ln_flags |= LF_SCANNED_ONCE;
1317 ns->ln_flags &= ~LF_UPGRADE;
1318 cfs_list_del_init(&com->lc_link);
1319 cfs_list_del_init(&com->lc_link_dir);
1320 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_double_scan);
1321 } else if (result == 0) {
1322 if (lfsck->ml_paused) {
1323 ns->ln_status = LS_PAUSED;
1325 ns->ln_status = LS_STOPPED;
1326 cfs_list_del_init(&com->lc_link);
1327 cfs_list_del_init(&com->lc_link_dir);
1328 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1331 ns->ln_status = LS_FAILED;
1332 cfs_list_del_init(&com->lc_link);
1333 cfs_list_del_init(&com->lc_link_dir);
1334 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1336 spin_unlock(&lfsck->ml_lock);
1338 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1339 HALF_SEC - lfsck->ml_time_last_checkpoint);
1340 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1341 ns->ln_items_checked += com->lc_new_checked;
1342 com->lc_new_checked = 0;
1344 rc = mdd_lfsck_namespace_store(env, com, false);
1346 up_write(&com->lc_sem);
1351 mdd_lfsck_namespace_dump(const struct lu_env *env, struct lfsck_component *com,
1354 struct md_lfsck *lfsck = com->lc_lfsck;
1355 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1356 struct lfsck_namespace *ns =
1357 (struct lfsck_namespace *)com->lc_file_ram;
1362 down_read(&com->lc_sem);
1363 rc = snprintf(buf, len,
1364 "name: lfsck_namespace\n"
1370 lfsck_status_names[ns->ln_status]);
1376 rc = lfsck_bits_dump(&buf, &len, ns->ln_flags, lfsck_flags_names,
1381 rc = lfsck_bits_dump(&buf, &len, bk->lb_param, lfsck_param_names,
1386 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_complete,
1387 "time_since_last_completed");
1391 rc = lfsck_time_dump(&buf, &len, ns->ln_time_latest_start,
1392 "time_since_latest_start");
1396 rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_checkpoint,
1397 "time_since_last_checkpoint");
1401 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_latest_start,
1402 "latest_start_position");
1406 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_last_checkpoint,
1407 "last_checkpoint_position");
1411 rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_first_inconsistent,
1412 "first_failure_position");
1416 if (ns->ln_status == LS_SCANNING_PHASE1) {
1417 struct lfsck_position pos;
1418 cfs_duration_t duration = cfs_time_current() -
1419 lfsck->ml_time_last_checkpoint;
1420 __u64 checked = ns->ln_items_checked + com->lc_new_checked;
1421 __u64 speed = checked;
1422 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1423 __u32 rtime = ns->ln_run_time_phase1 +
1424 cfs_duration_sec(duration + HALF_SEC);
1427 do_div(new_checked, duration);
1429 do_div(speed, rtime);
1430 rc = snprintf(buf, len,
1431 "checked_phase1: "LPU64"\n"
1432 "checked_phase2: "LPU64"\n"
1433 "updated_phase1: "LPU64"\n"
1434 "updated_phase2: "LPU64"\n"
1435 "failed_phase1: "LPU64"\n"
1436 "failed_phase2: "LPU64"\n"
1438 "M-linked: "LPU64"\n"
1439 "nlinks_repaired: "LPU64"\n"
1440 "lost_found: "LPU64"\n"
1441 "success_count: %u\n"
1442 "run_time_phase1: %u seconds\n"
1443 "run_time_phase2: %u seconds\n"
1444 "average_speed_phase1: "LPU64" items/sec\n"
1445 "average_speed_phase2: N/A\n"
1446 "real-time_speed_phase1: "LPU64" items/sec\n"
1447 "real-time_speed_phase2: N/A\n",
1449 ns->ln_objs_checked_phase2,
1450 ns->ln_items_repaired,
1451 ns->ln_objs_repaired_phase2,
1452 ns->ln_items_failed,
1453 ns->ln_objs_failed_phase2,
1454 ns->ln_dirs_checked,
1455 ns->ln_mlinked_checked,
1456 ns->ln_objs_nlink_repaired,
1457 ns->ln_objs_lost_found,
1458 ns->ln_success_count,
1460 ns->ln_run_time_phase2,
1468 mdd_lfsck_pos_fill(env, lfsck, &pos, true, true);
1469 rc = lfsck_pos_dump(&buf, &len, &pos, "current_position");
1472 } else if (ns->ln_status == LS_SCANNING_PHASE2) {
1473 cfs_duration_t duration = cfs_time_current() -
1474 lfsck->ml_time_last_checkpoint;
1475 __u64 checked = ns->ln_objs_checked_phase2 +
1476 com->lc_new_checked;
1477 __u64 speed1 = ns->ln_items_checked;
1478 __u64 speed2 = checked;
1479 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1480 __u32 rtime = ns->ln_run_time_phase2 +
1481 cfs_duration_sec(duration + HALF_SEC);
1484 do_div(new_checked, duration);
1485 if (ns->ln_run_time_phase1 != 0)
1486 do_div(speed1, ns->ln_run_time_phase1);
1488 do_div(speed2, rtime);
1489 rc = snprintf(buf, len,
1490 "checked_phase1: "LPU64"\n"
1491 "checked_phase2: "LPU64"\n"
1492 "updated_phase1: "LPU64"\n"
1493 "updated_phase2: "LPU64"\n"
1494 "failed_phase1: "LPU64"\n"
1495 "failed_phase2: "LPU64"\n"
1497 "M-linked: "LPU64"\n"
1498 "nlinks_repaired: "LPU64"\n"
1499 "lost_found: "LPU64"\n"
1500 "success_count: %u\n"
1501 "run_time_phase1: %u seconds\n"
1502 "run_time_phase2: %u seconds\n"
1503 "average_speed_phase1: "LPU64" items/sec\n"
1504 "average_speed_phase2: "LPU64" objs/sec\n"
1505 "real-time_speed_phase1: N/A\n"
1506 "real-time_speed_phase2: "LPU64" objs/sec\n"
1507 "current_position: "DFID"\n",
1508 ns->ln_items_checked,
1510 ns->ln_items_repaired,
1511 ns->ln_objs_repaired_phase2,
1512 ns->ln_items_failed,
1513 ns->ln_objs_failed_phase2,
1514 ns->ln_dirs_checked,
1515 ns->ln_mlinked_checked,
1516 ns->ln_objs_nlink_repaired,
1517 ns->ln_objs_lost_found,
1518 ns->ln_success_count,
1519 ns->ln_run_time_phase1,
1524 PFID(&ns->ln_fid_latest_scanned_phase2));
1531 __u64 speed1 = ns->ln_items_checked;
1532 __u64 speed2 = ns->ln_objs_checked_phase2;
1534 if (ns->ln_run_time_phase1 != 0)
1535 do_div(speed1, ns->ln_run_time_phase1);
1536 if (ns->ln_run_time_phase2 != 0)
1537 do_div(speed2, ns->ln_run_time_phase2);
1538 rc = snprintf(buf, len,
1539 "checked_phase1: "LPU64"\n"
1540 "checked_phase2: "LPU64"\n"
1541 "updated_phase1: "LPU64"\n"
1542 "updated_phase2: "LPU64"\n"
1543 "failed_phase1: "LPU64"\n"
1544 "failed_phase2: "LPU64"\n"
1546 "M-linked: "LPU64"\n"
1547 "nlinks_repaired: "LPU64"\n"
1548 "lost_found: "LPU64"\n"
1549 "success_count: %u\n"
1550 "run_time_phase1: %u seconds\n"
1551 "run_time_phase2: %u seconds\n"
1552 "average_speed_phase1: "LPU64" items/sec\n"
1553 "average_speed_phase2: "LPU64" objs/sec\n"
1554 "real-time_speed_phase1: N/A\n"
1555 "real-time_speed_phase2: N/A\n"
1556 "current_position: N/A\n",
1557 ns->ln_items_checked,
1558 ns->ln_objs_checked_phase2,
1559 ns->ln_items_repaired,
1560 ns->ln_objs_repaired_phase2,
1561 ns->ln_items_failed,
1562 ns->ln_objs_failed_phase2,
1563 ns->ln_dirs_checked,
1564 ns->ln_mlinked_checked,
1565 ns->ln_objs_nlink_repaired,
1566 ns->ln_objs_lost_found,
1567 ns->ln_success_count,
1568 ns->ln_run_time_phase1,
1569 ns->ln_run_time_phase2,
1581 up_read(&com->lc_sem);
1585 /* XXX: to be implemented in other patch. */
1586 static int mdd_lfsck_namespace_double_scan(const struct lu_env *env,
1587 struct lfsck_component *com)
1589 struct md_lfsck *lfsck = com->lc_lfsck;
1590 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
1591 struct lfsck_namespace *ns =
1592 (struct lfsck_namespace *)com->lc_file_ram;
1595 down_write(&com->lc_sem);
1597 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1598 com->lc_new_checked = 0;
1599 com->lc_journal = 0;
1601 ns->ln_status = LS_COMPLETED;
1602 if (!(bk->lb_param & LPF_DRYRUN))
1604 ~(LF_SCANNED_ONCE | LF_INCONSISTENT | LF_UPGRADE);
1605 ns->ln_time_last_complete = ns->ln_time_last_checkpoint;
1606 ns->ln_success_count++;
1608 spin_lock(&lfsck->ml_lock);
1609 cfs_list_del_init(&com->lc_link);
1610 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1611 spin_unlock(&lfsck->ml_lock);
1613 rc = mdd_lfsck_namespace_store(env, com, false);
1615 up_write(&com->lc_sem);
1619 static struct lfsck_operations mdd_lfsck_namespace_ops = {
1620 .lfsck_reset = mdd_lfsck_namespace_reset,
1621 .lfsck_fail = mdd_lfsck_namespace_fail,
1622 .lfsck_checkpoint = mdd_lfsck_namespace_checkpoint,
1623 .lfsck_prep = mdd_lfsck_namespace_prep,
1624 .lfsck_exec_oit = mdd_lfsck_namespace_exec_oit,
1625 .lfsck_exec_dir = mdd_lfsck_namespace_exec_dir,
1626 .lfsck_post = mdd_lfsck_namespace_post,
1627 .lfsck_dump = mdd_lfsck_namespace_dump,
1628 .lfsck_double_scan = mdd_lfsck_namespace_double_scan,
1631 /* LFSCK component setup/cleanup functions */
1633 static int mdd_lfsck_namespace_setup(const struct lu_env *env,
1634 struct md_lfsck *lfsck)
1636 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1637 struct lfsck_component *com;
1638 struct lfsck_namespace *ns;
1639 struct dt_object *obj;
1647 CFS_INIT_LIST_HEAD(&com->lc_link);
1648 CFS_INIT_LIST_HEAD(&com->lc_link_dir);
1649 init_rwsem(&com->lc_sem);
1650 atomic_set(&com->lc_ref, 1);
1651 com->lc_lfsck = lfsck;
1652 com->lc_type = LT_NAMESPACE;
1653 com->lc_ops = &mdd_lfsck_namespace_ops;
1654 com->lc_file_size = sizeof(struct lfsck_namespace);
1655 OBD_ALLOC(com->lc_file_ram, com->lc_file_size);
1656 if (com->lc_file_ram == NULL)
1657 GOTO(out, rc = -ENOMEM);
1659 OBD_ALLOC(com->lc_file_disk, com->lc_file_size);
1660 if (com->lc_file_disk == NULL)
1661 GOTO(out, rc = -ENOMEM);
1663 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name,
1664 &mdd_env_info(env)->mti_fid);
1666 GOTO(out, rc = PTR_ERR(obj));
1669 rc = obj->do_ops->do_index_try(env, obj, &dt_lfsck_features);
1673 rc = mdd_lfsck_namespace_load(env, com);
1675 rc = mdd_lfsck_namespace_reset(env, com, true);
1676 else if (rc == -ENODATA)
1677 rc = mdd_lfsck_namespace_init(env, com);
1681 ns = (struct lfsck_namespace *)com->lc_file_ram;
1682 switch (ns->ln_status) {
1687 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1690 CERROR("%s: unknown status: %u\n",
1691 mdd_lfsck2name(lfsck), ns->ln_status);
1693 case LS_SCANNING_PHASE1:
1694 case LS_SCANNING_PHASE2:
1695 /* No need to store the status to disk right now.
1696 * If the system crashed before the status stored,
1697 * it will be loaded back when next time. */
1698 ns->ln_status = LS_CRASHED;
1702 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_scan);
1703 cfs_list_add_tail(&com->lc_link_dir, &lfsck->ml_list_dir);
1711 mdd_lfsck_component_cleanup(env, com);
1715 /* helper functions for framework */
1717 static int object_is_client_visible(const struct lu_env *env,
1718 struct mdd_device *mdd,
1719 struct mdd_object *obj)
1721 struct lu_fid *fid = &mdd_env_info(env)->mti_fid;
1725 LASSERT(S_ISDIR(mdd_object_type(obj)));
1728 if (mdd_is_root(mdd, mdo2fid(obj))) {
1730 mdd_object_put(env, obj);
1734 mdd_read_lock(env, obj, MOR_TGT_CHILD);
1735 if (unlikely(mdd_is_dead_obj(obj))) {
1736 mdd_read_unlock(env, obj);
1738 mdd_object_put(env, obj);
1742 rc = dt_xattr_get(env, mdd_object_child(obj),
1743 mdd_buf_get(env, NULL, 0), XATTR_NAME_LINK,
1745 mdd_read_unlock(env, obj);
1748 mdd_object_put(env, obj);
1752 if (rc < 0 && rc != -ENODATA) {
1754 mdd_object_put(env, obj);
1758 rc = mdd_parent_fid(env, obj, fid);
1760 mdd_object_put(env, obj);
1764 if (unlikely(lu_fid_eq(fid, &mdd->mdd_local_root_fid)))
1767 obj = mdd_object_find(env, mdd, fid);
1770 else if (IS_ERR(obj))
1771 return PTR_ERR(obj);
1773 /* XXX: need more processing for remote object in the future. */
1774 if (!mdd_object_exists(obj) || mdd_object_remote(obj)) {
1775 mdd_object_put(env, obj);
1784 static void mdd_lfsck_unpack_ent(struct lu_dirent *ent)
1786 fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid);
1787 ent->lde_hash = le64_to_cpu(ent->lde_hash);
1788 ent->lde_reclen = le16_to_cpu(ent->lde_reclen);
1789 ent->lde_namelen = le16_to_cpu(ent->lde_namelen);
1790 ent->lde_attrs = le32_to_cpu(ent->lde_attrs);
1792 /* Make sure the name is terminated with '0'.
1793 * The data (type) after ent::lde_name maybe
1794 * broken, but we do not care. */
1795 ent->lde_name[ent->lde_namelen] = 0;
1798 /* LFSCK wrap functions */
1800 static void mdd_lfsck_fail(const struct lu_env *env, struct md_lfsck *lfsck,
1801 bool oit, bool new_checked)
1803 struct lfsck_component *com;
1805 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
1806 com->lc_ops->lfsck_fail(env, com, oit, new_checked);
1810 static int mdd_lfsck_checkpoint(const struct lu_env *env,
1811 struct md_lfsck *lfsck, bool oit)
1813 struct lfsck_component *com;
1816 if (likely(cfs_time_beforeq(cfs_time_current(),
1817 lfsck->ml_time_next_checkpoint)))
1820 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, oit, !oit);
1821 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
1822 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
1827 lfsck->ml_time_last_checkpoint = cfs_time_current();
1828 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1829 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1833 static int mdd_lfsck_prep(struct lu_env *env, struct md_lfsck *lfsck)
1835 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
1836 struct mdd_object *obj = NULL;
1837 struct dt_object *dt_obj;
1838 struct lfsck_component *com;
1839 struct lfsck_component *next;
1840 struct lfsck_position *pos = NULL;
1841 const struct dt_it_ops *iops =
1842 &lfsck->ml_obj_oit->do_index_ops->dio_it;
1847 LASSERT(lfsck->ml_obj_dir == NULL);
1848 LASSERT(lfsck->ml_di_dir == NULL);
1850 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
1851 com->lc_new_checked = 0;
1852 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
1853 com->lc_journal = 0;
1855 rc = com->lc_ops->lfsck_prep(env, com);
1859 if ((pos == NULL) ||
1860 (!mdd_lfsck_pos_is_zero(&com->lc_pos_start) &&
1861 mdd_lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
1862 pos = &com->lc_pos_start;
1865 /* Init otable-based iterator. */
1867 rc = iops->load(env, lfsck->ml_di_oit, 0);
1868 GOTO(out, rc = (rc >= 0 ? 0 : rc));
1871 rc = iops->load(env, lfsck->ml_di_oit, pos->lp_oit_cookie);
1875 if (fid_is_zero(&pos->lp_dir_parent))
1878 /* Find the directory for namespace-based traverse. */
1879 obj = mdd_object_find(env, mdd, &pos->lp_dir_parent);
1882 else if (IS_ERR(obj))
1883 RETURN(PTR_ERR(obj));
1885 /* XXX: need more processing for remote object in the future. */
1886 if (!mdd_object_exists(obj) || mdd_object_remote(obj) ||
1887 unlikely(!S_ISDIR(mdd_object_type(obj))))
1890 if (unlikely(mdd_is_dead_obj(obj)))
1893 dt_obj = mdd_object_child(obj);
1894 if (unlikely(!dt_try_as_dir(env, dt_obj)))
1895 GOTO(out, rc = -ENOTDIR);
1897 /* Init the namespace-based directory traverse. */
1898 iops = &dt_obj->do_index_ops->dio_it;
1899 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
1901 GOTO(out, rc = PTR_ERR(di));
1903 rc = iops->load(env, di, pos->lp_dir_cookie);
1905 rc = iops->next(env, di);
1911 iops->fini(env, di);
1915 lfsck->ml_obj_dir = dt_obj;
1916 spin_lock(&lfsck->ml_lock);
1917 lfsck->ml_di_dir = di;
1918 spin_unlock(&lfsck->ml_lock);
1925 mdd_object_put(env, obj);
1928 return (rc > 0 ? 0 : rc);
1930 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, false, false);
1931 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
1932 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
1937 lfsck->ml_time_last_checkpoint = cfs_time_current();
1938 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1939 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1943 static int mdd_lfsck_exec_oit(const struct lu_env *env, struct md_lfsck *lfsck,
1944 struct mdd_object *obj)
1946 struct lfsck_component *com;
1947 struct dt_object *dt_obj;
1948 const struct dt_it_ops *iops;
1953 LASSERT(lfsck->ml_obj_dir == NULL);
1955 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
1956 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
1961 if (!S_ISDIR(mdd_object_type(obj)) ||
1962 cfs_list_empty(&lfsck->ml_list_dir))
1965 rc = object_is_client_visible(env, mdd_lfsck2mdd(lfsck), obj);
1969 if (unlikely(mdd_is_dead_obj(obj)))
1972 dt_obj = mdd_object_child(obj);
1973 if (unlikely(!dt_try_as_dir(env, dt_obj)))
1974 GOTO(out, rc = -ENOTDIR);
1976 iops = &dt_obj->do_index_ops->dio_it;
1977 di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
1979 GOTO(out, rc = PTR_ERR(di));
1981 rc = iops->load(env, di, 0);
1983 rc = iops->next(env, di);
1989 iops->fini(env, di);
1993 mdd_object_get(obj);
1994 lfsck->ml_obj_dir = dt_obj;
1995 spin_lock(&lfsck->ml_lock);
1996 lfsck->ml_di_dir = di;
1997 spin_unlock(&lfsck->ml_lock);
2003 mdd_lfsck_fail(env, lfsck, false, false);
2004 return (rc > 0 ? 0 : rc);
2007 static int mdd_lfsck_exec_dir(const struct lu_env *env, struct md_lfsck *lfsck,
2008 struct mdd_object *obj, struct lu_dirent *ent)
2010 struct lfsck_component *com;
2013 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2014 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
2021 static int mdd_lfsck_post(const struct lu_env *env, struct md_lfsck *lfsck,
2024 struct lfsck_component *com;
2025 struct lfsck_component *next;
2028 mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, true, true);
2029 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2030 rc = com->lc_ops->lfsck_post(env, com, result);
2035 lfsck->ml_time_last_checkpoint = cfs_time_current();
2036 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2037 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2041 static int mdd_lfsck_double_scan(const struct lu_env *env,
2042 struct md_lfsck *lfsck)
2044 struct lfsck_component *com;
2045 struct lfsck_component *next;
2048 cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_double_scan,
2050 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2051 com->lc_journal = 0;
2053 rc = com->lc_ops->lfsck_double_scan(env, com);
2062 static int mdd_lfsck_dir_engine(const struct lu_env *env,
2063 struct md_lfsck *lfsck)
2065 struct mdd_thread_info *info = mdd_env_info(env);
2066 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2067 const struct dt_it_ops *iops =
2068 &lfsck->ml_obj_dir->do_index_ops->dio_it;
2069 struct dt_it *di = lfsck->ml_di_dir;
2070 struct lu_dirent *ent = &info->mti_ent;
2071 struct lu_fid *fid = &info->mti_fid;
2072 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2073 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2078 struct mdd_object *child;
2080 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY2) &&
2082 struct l_wait_info lwi;
2084 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2086 l_wait_event(thread->t_ctl_waitq,
2087 !thread_is_running(thread),
2091 lfsck->ml_new_scanned++;
2092 rc = iops->rec(env, di, (struct dt_rec *)ent,
2093 lfsck->ml_args_dir);
2095 mdd_lfsck_fail(env, lfsck, false, true);
2096 if (bk->lb_param & LPF_FAILOUT)
2102 mdd_lfsck_unpack_ent(ent);
2103 if (ent->lde_attrs & LUDA_IGNORE)
2106 *fid = ent->lde_fid;
2107 child = mdd_object_find(env, mdd, fid);
2108 if (child == NULL) {
2110 } else if (IS_ERR(child)) {
2111 mdd_lfsck_fail(env, lfsck, false, true);
2112 if (bk->lb_param & LPF_FAILOUT)
2113 RETURN(PTR_ERR(child));
2118 /* XXX: need more processing for remote object in the future. */
2119 if (mdd_object_exists(child) && !mdd_object_remote(child))
2120 rc = mdd_lfsck_exec_dir(env, lfsck, child, ent);
2121 mdd_object_put(env, child);
2122 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2126 rc = mdd_lfsck_checkpoint(env, lfsck, false);
2127 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2131 mdd_lfsck_control_speed(lfsck);
2132 if (unlikely(!thread_is_running(thread)))
2135 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL2)) {
2136 spin_lock(&lfsck->ml_lock);
2137 thread_set_flags(thread, SVC_STOPPING);
2138 spin_unlock(&lfsck->ml_lock);
2142 rc = iops->next(env, di);
2145 if (rc > 0 && !lfsck->ml_oit_over)
2146 mdd_lfsck_close_dir(env, lfsck);
2151 static int mdd_lfsck_oit_engine(const struct lu_env *env,
2152 struct md_lfsck *lfsck)
2154 struct mdd_thread_info *info = mdd_env_info(env);
2155 struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
2156 const struct dt_it_ops *iops =
2157 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2158 struct dt_it *di = lfsck->ml_di_oit;
2159 struct lu_fid *fid = &info->mti_fid;
2160 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2161 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2166 struct mdd_object *target;
2168 if (lfsck->ml_di_dir != NULL) {
2169 rc = mdd_lfsck_dir_engine(env, lfsck);
2174 if (unlikely(lfsck->ml_oit_over))
2177 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY1) &&
2179 struct l_wait_info lwi;
2181 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2183 l_wait_event(thread->t_ctl_waitq,
2184 !thread_is_running(thread),
2188 lfsck->ml_new_scanned++;
2189 rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
2191 mdd_lfsck_fail(env, lfsck, true, true);
2192 if (bk->lb_param & LPF_FAILOUT)
2198 target = mdd_object_find(env, mdd, fid);
2199 if (target == NULL) {
2201 } else if (IS_ERR(target)) {
2202 mdd_lfsck_fail(env, lfsck, true, true);
2203 if (bk->lb_param & LPF_FAILOUT)
2204 RETURN(PTR_ERR(target));
2209 /* XXX: In fact, low layer otable-based iteration should not
2210 * return agent object. But before LU-2646 resolved, we
2211 * need more processing for agent object. */
2212 if (mdd_object_exists(target) && !mdd_object_remote(target))
2213 rc = mdd_lfsck_exec_oit(env, lfsck, target);
2214 mdd_object_put(env, target);
2215 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2219 rc = mdd_lfsck_checkpoint(env, lfsck, true);
2220 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2224 mdd_lfsck_control_speed(lfsck);
2226 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL1)) {
2227 spin_lock(&lfsck->ml_lock);
2228 thread_set_flags(thread, SVC_STOPPING);
2229 spin_unlock(&lfsck->ml_lock);
2233 rc = iops->next(env, di);
2235 lfsck->ml_oit_over = 1;
2237 if (unlikely(!thread_is_running(thread)))
2239 } while (rc == 0 || lfsck->ml_di_dir != NULL);
2244 static int mdd_lfsck_main(void *args)
2247 struct md_lfsck *lfsck = (struct md_lfsck *)args;
2248 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2249 struct dt_object *oit_obj = lfsck->ml_obj_oit;
2250 const struct dt_it_ops *oit_iops = &oit_obj->do_index_ops->dio_it;
2251 struct dt_it *oit_di;
2255 cfs_daemonize("lfsck");
2256 rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
2258 CERROR("%s: LFSCK, fail to init env, rc = %d\n",
2259 mdd_lfsck2name(lfsck), rc);
2263 oit_di = oit_iops->init(&env, oit_obj, lfsck->ml_args_oit, BYPASS_CAPA);
2264 if (IS_ERR(oit_di)) {
2265 rc = PTR_ERR(oit_di);
2266 CERROR("%s: LFSCK, fail to init iteration, rc = %d\n",
2267 mdd_lfsck2name(lfsck), rc);
2271 spin_lock(&lfsck->ml_lock);
2272 lfsck->ml_di_oit = oit_di;
2273 spin_unlock(&lfsck->ml_lock);
2274 rc = mdd_lfsck_prep(&env, lfsck);
2278 CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = 0x%x, dir_flags = 0x%x, "
2279 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2280 ", pid = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2281 lfsck->ml_pos_current.lp_oit_cookie,
2282 lfsck->ml_pos_current.lp_dir_cookie,
2283 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2286 spin_lock(&lfsck->ml_lock);
2287 thread_set_flags(thread, SVC_RUNNING);
2288 spin_unlock(&lfsck->ml_lock);
2289 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2291 if (!cfs_list_empty(&lfsck->ml_list_scan) ||
2292 cfs_list_empty(&lfsck->ml_list_double_scan))
2293 rc = mdd_lfsck_oit_engine(&env, lfsck);
2297 CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = 0x%x, dir_flags = 0x%x, "
2298 "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2299 ", pid = %d, rc = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2300 lfsck->ml_pos_current.lp_oit_cookie,
2301 lfsck->ml_pos_current.lp_dir_cookie,
2302 PFID(&lfsck->ml_pos_current.lp_dir_parent),
2303 cfs_curproc_pid(), rc);
2305 if (lfsck->ml_paused && cfs_list_empty(&lfsck->ml_list_scan))
2306 oit_iops->put(&env, oit_di);
2308 rc = mdd_lfsck_post(&env, lfsck, rc);
2309 if (lfsck->ml_di_dir != NULL)
2310 mdd_lfsck_close_dir(&env, lfsck);
2313 spin_lock(&lfsck->ml_lock);
2314 lfsck->ml_di_oit = NULL;
2315 spin_unlock(&lfsck->ml_lock);
2317 oit_iops->fini(&env, oit_di);
2319 if (!cfs_list_empty(&lfsck->ml_list_double_scan))
2320 rc = mdd_lfsck_double_scan(&env, lfsck);
2325 /* XXX: Purge the pinned objects in the future. */
2331 spin_lock(&lfsck->ml_lock);
2332 thread_set_flags(thread, SVC_STOPPED);
2333 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2334 spin_unlock(&lfsck->ml_lock);
2338 /* external interfaces */
2340 int mdd_lfsck_set_speed(const struct lu_env *env, struct md_lfsck *lfsck,
2345 mutex_lock(&lfsck->ml_mutex);
2346 __mdd_lfsck_set_speed(lfsck, limit);
2347 rc = mdd_lfsck_bookmark_store(env, lfsck);
2348 mutex_unlock(&lfsck->ml_mutex);
2352 int mdd_lfsck_dump(const struct lu_env *env, struct md_lfsck *lfsck,
2353 __u16 type, char *buf, int len)
2355 struct lfsck_component *com;
2358 if (!lfsck->ml_initialized)
2361 com = mdd_lfsck_component_find(lfsck, type);
2365 rc = com->lc_ops->lfsck_dump(env, com, buf, len);
2366 mdd_lfsck_component_put(env, com);
2370 int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
2371 struct lfsck_start *start)
2373 struct lfsck_bookmark *bk = &lfsck->ml_bookmark_ram;
2374 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2375 struct lfsck_component *com;
2376 struct l_wait_info lwi = { 0 };
2383 if (lfsck->ml_obj_oit == NULL)
2386 /* start == NULL means auto trigger paused LFSCK. */
2387 if (start == NULL && cfs_list_empty(&lfsck->ml_list_scan))
2390 mutex_lock(&lfsck->ml_mutex);
2391 spin_lock(&lfsck->ml_lock);
2392 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2393 spin_unlock(&lfsck->ml_lock);
2394 mutex_unlock(&lfsck->ml_mutex);
2398 spin_unlock(&lfsck->ml_lock);
2400 lfsck->ml_paused = 0;
2401 lfsck->ml_oit_over = 0;
2402 lfsck->ml_drop_dryrun = 0;
2403 lfsck->ml_new_scanned = 0;
2405 /* For auto trigger. */
2409 start->ls_version = bk->lb_version;
2410 if (start->ls_valid & LSV_SPEED_LIMIT) {
2411 __mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
2415 if (start->ls_valid & LSV_ERROR_HANDLE) {
2416 valid |= DOIV_ERROR_HANDLE;
2417 if (start->ls_flags & LPF_FAILOUT)
2418 flags |= DOIF_FAILOUT;
2420 if ((start->ls_flags & LPF_FAILOUT) &&
2421 !(bk->lb_param & LPF_FAILOUT)) {
2422 bk->lb_param |= LPF_FAILOUT;
2424 } else if (!(start->ls_flags & LPF_FAILOUT) &&
2425 (bk->lb_param & LPF_FAILOUT)) {
2426 bk->lb_param &= ~LPF_FAILOUT;
2431 if (start->ls_valid & LSV_DRYRUN) {
2432 if ((start->ls_flags & LPF_DRYRUN) &&
2433 !(bk->lb_param & LPF_DRYRUN)) {
2434 bk->lb_param |= LPF_DRYRUN;
2436 } else if (!(start->ls_flags & LPF_DRYRUN) &&
2437 (bk->lb_param & LPF_DRYRUN)) {
2438 bk->lb_param &= ~LPF_DRYRUN;
2439 lfsck->ml_drop_dryrun = 1;
2445 rc = mdd_lfsck_bookmark_store(env, lfsck);
2450 if (start->ls_flags & LPF_RESET)
2451 flags |= DOIF_RESET;
2453 if (start->ls_active != 0) {
2454 struct lfsck_component *next;
2457 if (start->ls_active == LFSCK_TYPES_ALL)
2458 start->ls_active = LFSCK_TYPES_SUPPORTED;
2460 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2461 start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2462 GOTO(out, rc = -ENOTSUPP);
2465 cfs_list_for_each_entry_safe(com, next,
2466 &lfsck->ml_list_scan, lc_link) {
2467 if (!(com->lc_type & start->ls_active)) {
2468 rc = com->lc_ops->lfsck_post(env, com, 0);
2474 while (start->ls_active != 0) {
2475 if (type & start->ls_active) {
2476 com = __mdd_lfsck_component_find(lfsck, type,
2477 &lfsck->ml_list_idle);
2479 /* The component status will be updated
2480 * when its prep() is called later by
2481 * the LFSCK main engine. */
2482 cfs_list_del_init(&com->lc_link);
2483 cfs_list_add_tail(&com->lc_link,
2484 &lfsck->ml_list_scan);
2486 start->ls_active &= ~type;
2492 cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2493 start->ls_active |= com->lc_type;
2494 if (flags & DOIF_RESET) {
2495 rc = com->lc_ops->lfsck_reset(env, com, false);
2502 lfsck->ml_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2503 if (bk->lb_param & LPF_DRYRUN)
2504 lfsck->ml_args_dir |= LUDA_VERIFY_DRYRUN;
2506 if (bk->lb_param & LPF_FAILOUT) {
2507 valid |= DOIV_ERROR_HANDLE;
2508 flags |= DOIF_FAILOUT;
2511 if (!cfs_list_empty(&lfsck->ml_list_scan))
2512 flags |= DOIF_OUTUSED;
2514 lfsck->ml_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2515 thread_set_flags(thread, 0);
2516 rc = cfs_create_thread(mdd_lfsck_main, lfsck, 0);
2518 CERROR("%s: cannot start LFSCK thread, rc = %d\n",
2519 mdd_lfsck2name(lfsck), rc);
2521 l_wait_event(thread->t_ctl_waitq,
2522 thread_is_running(thread) ||
2523 thread_is_stopped(thread),
2529 mutex_unlock(&lfsck->ml_mutex);
2530 return (rc < 0 ? rc : 0);
2533 int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck,
2536 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2537 struct l_wait_info lwi = { 0 };
2540 mutex_lock(&lfsck->ml_mutex);
2541 spin_lock(&lfsck->ml_lock);
2542 if (thread_is_init(thread) || thread_is_stopped(thread)) {
2543 spin_unlock(&lfsck->ml_lock);
2544 mutex_unlock(&lfsck->ml_mutex);
2549 lfsck->ml_paused = 1;
2550 thread_set_flags(thread, SVC_STOPPING);
2551 /* The LFSCK thread may be sleeping on low layer wait queue,
2553 if (likely(lfsck->ml_di_oit != NULL))
2554 lfsck->ml_obj_oit->do_index_ops->dio_it.put(env,
2556 spin_unlock(&lfsck->ml_lock);
2558 cfs_waitq_broadcast(&thread->t_ctl_waitq);
2559 l_wait_event(thread->t_ctl_waitq,
2560 thread_is_stopped(thread),
2562 mutex_unlock(&lfsck->ml_mutex);
2567 static const struct lu_fid lfsck_it_fid = { .f_seq = FID_SEQ_LOCAL_FILE,
2568 .f_oid = OTABLE_IT_OID,
2571 int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd)
2573 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2574 struct dt_object *obj;
2578 LASSERT(!lfsck->ml_initialized);
2580 lfsck->ml_initialized = 1;
2581 mutex_init(&lfsck->ml_mutex);
2582 spin_lock_init(&lfsck->ml_lock);
2583 CFS_INIT_LIST_HEAD(&lfsck->ml_list_scan);
2584 CFS_INIT_LIST_HEAD(&lfsck->ml_list_dir);
2585 CFS_INIT_LIST_HEAD(&lfsck->ml_list_double_scan);
2586 CFS_INIT_LIST_HEAD(&lfsck->ml_list_idle);
2587 cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
2589 obj = dt_locate(env, mdd->mdd_bottom, &lfsck_it_fid);
2591 RETURN(PTR_ERR(obj));
2593 lfsck->ml_obj_oit = obj;
2594 rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2596 if (rc == -ENOTSUPP)
2602 obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_bookmark_name,
2603 &mdd_env_info(env)->mti_fid);
2605 RETURN(PTR_ERR(obj));
2607 lfsck->ml_bookmark_obj = obj;
2608 rc = mdd_lfsck_bookmark_load(env, lfsck);
2610 rc = mdd_lfsck_bookmark_init(env, lfsck);
2614 rc = mdd_lfsck_namespace_setup(env, lfsck);
2615 /* XXX: LFSCK components initialization to be added here. */
2620 void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd)
2622 struct md_lfsck *lfsck = &mdd->mdd_lfsck;
2623 struct ptlrpc_thread *thread = &lfsck->ml_thread;
2624 struct lfsck_component *com;
2626 if (!lfsck->ml_initialized)
2629 LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
2631 if (lfsck->ml_obj_oit != NULL) {
2632 lu_object_put(env, &lfsck->ml_obj_oit->do_lu);
2633 lfsck->ml_obj_oit = NULL;
2636 LASSERT(lfsck->ml_obj_dir == NULL);
2638 if (lfsck->ml_bookmark_obj != NULL) {
2639 lu_object_put(env, &lfsck->ml_bookmark_obj->do_lu);
2640 lfsck->ml_bookmark_obj = NULL;
2643 while (!cfs_list_empty(&lfsck->ml_list_scan)) {
2644 com = cfs_list_entry(lfsck->ml_list_scan.next,
2645 struct lfsck_component,
2647 mdd_lfsck_component_cleanup(env, com);
2650 LASSERT(cfs_list_empty(&lfsck->ml_list_dir));
2652 while (!cfs_list_empty(&lfsck->ml_list_double_scan)) {
2653 com = cfs_list_entry(lfsck->ml_list_double_scan.next,
2654 struct lfsck_component,
2656 mdd_lfsck_component_cleanup(env, com);
2659 while (!cfs_list_empty(&lfsck->ml_list_idle)) {
2660 com = cfs_list_entry(lfsck->ml_list_idle.next,
2661 struct lfsck_component,
2663 mdd_lfsck_component_cleanup(env, com);