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) 2013, Intel Corporation.
26 * lustre/lfsck/lfsck_internal.h
28 * Shared definitions and declarations for the LFSCK.
30 * Author: Fan, Yong <fan.yong@intel.com>
33 #ifndef _LFSCK_INTERNAL_H
34 # define _LFSCK_INTERNAL_H
36 #include <lustre/lustre_lfsck_user.h>
37 #include <lustre/lustre_user.h>
38 #include <lustre/lustre_idl.h>
39 #include <lustre_lfsck.h>
41 #include <lu_object.h>
42 #include <dt_object.h>
43 #include <md_object.h>
44 #include <lustre_net.h>
45 #include <lustre_dlm.h>
46 #include <lustre_fid.h>
47 #include <md_object.h>
49 #define HALF_SEC (HZ >> 1)
50 #define LFSCK_CHECKPOINT_INTERVAL 60
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. */
57 /* Finish the first cycle scanning. */
58 LF_SCANNED_ONCE = 0x00000001ULL,
60 /* There is some namespace inconsistency. */
61 LF_INCONSISTENT = 0x00000002ULL,
63 /* The device is upgraded from 1.8 format. */
64 LF_UPGRADE = 0x00000004ULL,
66 /* The server ever restarted during the LFSCK, and may miss to process
67 * some objects check/repair. */
68 LF_INCOMPLETE = 0x00000008ULL,
70 /* The LAST_ID (file) crashed. */
71 LF_CRASHED_LASTID = 0x00000010ULL,
74 struct lfsck_position {
75 /* low layer object table-based iteration position. */
78 /* parent FID for directory traversal. */
79 struct lu_fid lp_dir_parent;
81 /* namespace-based directory traversal position. */
85 struct lfsck_bookmark {
86 /* Magic number to detect that this struct contains valid data. */
89 /* For compatible with old versions. */
92 /* See 'enum lfsck_param_flags' */
95 /* How many items can be scanned at most per second. */
98 /* The windows size for async requests pipeline. */
99 __u16 lb_async_windows;
101 /* For 64-bits aligned. */
104 /* For future using. */
105 __u64 lb_reserved[6];
108 struct lfsck_namespace {
109 /* Magic number to detect that this struct contains valid data. */
112 /* See 'enum lfsck_status'. */
115 /* See 'enum lfsck_flags'. */
118 /* How many completed LFSCK runs on the device. */
119 __u32 ln_success_count;
121 /* How long the LFSCK phase1 has run in seconds. */
122 __u32 ln_run_time_phase1;
124 /* How long the LFSCK phase2 has run in seconds. */
125 __u32 ln_run_time_phase2;
127 /* Time for the last LFSCK completed in seconds since epoch. */
128 __u64 ln_time_last_complete;
130 /* Time for the latest LFSCK ran in seconds since epoch. */
131 __u64 ln_time_latest_start;
133 /* Time for the last LFSCK checkpoint in seconds since epoch. */
134 __u64 ln_time_last_checkpoint;
136 /* Position for the latest LFSCK started from. */
137 struct lfsck_position ln_pos_latest_start;
139 /* Position for the last LFSCK checkpoint. */
140 struct lfsck_position ln_pos_last_checkpoint;
142 /* Position for the first should be updated object. */
143 struct lfsck_position ln_pos_first_inconsistent;
145 /* How many items (including dir) have been checked. */
146 __u64 ln_items_checked;
148 /* How many items have been repaired. */
149 __u64 ln_items_repaired;
151 /* How many items failed to be processed. */
152 __u64 ln_items_failed;
154 /* How many directories have been traversed. */
155 __u64 ln_dirs_checked;
157 /* How many multiple-linked objects have been checked. */
158 __u64 ln_mlinked_checked;
160 /* How many objects have been double scanned. */
161 __u64 ln_objs_checked_phase2;
163 /* How many objects have been reparied during double scan. */
164 __u64 ln_objs_repaired_phase2;
166 /* How many objects failed to be processed during double scan. */
167 __u64 ln_objs_failed_phase2;
169 /* How many objects with nlink fixed. */
170 __u64 ln_objs_nlink_repaired;
172 /* How many objects were lost before, but found back now. */
173 __u64 ln_objs_lost_found;
175 /* The latest object has been processed (failed) during double scan. */
176 struct lu_fid ln_fid_latest_scanned_phase2;
178 /* For further using. 256-bytes aligned now. */
179 __u64 ln_reserved[2];
182 enum lfsck_layout_inconsistency_type {
185 LLIT_UNMATCHED_PAIR = 2,
186 LLIT_MULTIPLE_REFERENCED = 3,
188 LLIT_INCONSISTENT_OWNER = 5,
190 LLIT_MAX = LLIT_OTHERS
193 struct lfsck_layout {
194 /* Magic number to detect that this struct contains valid data. */
197 /* See 'enum lfsck_status'. */
200 /* See 'enum lfsck_flags'. */
203 /* How many completed LFSCK runs on the device. */
204 __u32 ll_success_count;
206 /* How long the LFSCK phase1 has run in seconds. */
207 __u32 ll_run_time_phase1;
209 /* How long the LFSCK phase2 has run in seconds. */
210 __u32 ll_run_time_phase2;
212 /* Time for the last LFSCK completed in seconds since epoch. */
213 __u64 ll_time_last_complete;
215 /* Time for the latest LFSCK ran in seconds since epoch. */
216 __u64 ll_time_latest_start;
218 /* Time for the last LFSCK checkpoint in seconds since epoch. */
219 __u64 ll_time_last_checkpoint;
221 /* Position for the latest LFSCK started from. */
222 __u64 ll_pos_latest_start;
224 /* Position for the last LFSCK checkpoint. */
225 __u64 ll_pos_last_checkpoint;
227 /* Position for the first should be updated object. */
228 __u64 ll_pos_first_inconsistent;
230 /* How many objects have been checked. */
231 __u64 ll_objs_checked_phase1;
233 /* How many objects failed to be processed. */
234 __u64 ll_objs_failed_phase1;
236 /* How many objects have been double scanned. */
237 __u64 ll_objs_checked_phase2;
239 /* How many objects failed to be processed during double scan. */
240 __u64 ll_objs_failed_phase2;
242 /* kinds of inconsistency have been repaired.
243 * ll_objs_repaired[type - 1] is the count for the given @type. */
244 __u64 ll_objs_repaired[LLIT_MAX];
246 /* How many objects have been skipped because of related
247 * MDT(s)/OST(s) do not participate in the LFSCK */
248 __u64 ll_objs_skipped;
250 /* For further using. 256-bytes aligned now. */
251 __u64 ll_reserved[12];
254 struct lfsck_component;
256 struct lfsck_operations {
257 int (*lfsck_reset)(const struct lu_env *env,
258 struct lfsck_component *com,
261 void (*lfsck_fail)(const struct lu_env *env,
262 struct lfsck_component *com,
265 int (*lfsck_checkpoint)(const struct lu_env *env,
266 struct lfsck_component *com,
269 int (*lfsck_prep)(const struct lu_env *env,
270 struct lfsck_component *com);
272 int (*lfsck_exec_oit)(const struct lu_env *env,
273 struct lfsck_component *com,
274 struct dt_object *obj);
276 int (*lfsck_exec_dir)(const struct lu_env *env,
277 struct lfsck_component *com,
278 struct dt_object *obj,
279 struct lu_dirent *ent);
281 int (*lfsck_post)(const struct lu_env *env,
282 struct lfsck_component *com,
286 int (*lfsck_dump)(const struct lu_env *env,
287 struct lfsck_component *com,
291 int (*lfsck_double_scan)(const struct lu_env *env,
292 struct lfsck_component *com);
294 void (*lfsck_data_release)(const struct lu_env *env,
295 struct lfsck_component *com);
297 void (*lfsck_quit)(const struct lu_env *env,
298 struct lfsck_component *com);
301 #define TGT_PTRS 256 /* number of pointers at 1st level */
302 #define TGT_PTRS_PER_BLOCK 256 /* number of pointers at 2nd level */
304 struct lfsck_tgt_desc {
305 struct list_head ltd_orphan_list;
306 struct dt_device *ltd_tgt;
307 struct dt_device *ltd_key;
308 struct obd_export *ltd_exp;
309 struct list_head ltd_layout_list;
314 struct lfsck_tgt_desc_idx {
315 struct lfsck_tgt_desc *ldi_tgts[TGT_PTRS_PER_BLOCK];
318 struct lfsck_tgt_descs {
319 /* list of known TGTs */
320 struct lfsck_tgt_desc_idx *ltd_tgts_idx[TGT_PTRS];
322 /* bitmap of TGTs available */
323 cfs_bitmap_t *ltd_tgts_bitmap;
325 /* for lfsck_tgt_desc::ltd_xxx_list */
328 /* for tgts table accessing and changes */
329 struct rw_semaphore ltd_rw_sem;
331 /* Temporary list for orphan targets. */
332 struct list_head ltd_orphan;
334 /* number of registered TGTs */
338 #define LTD_TGT(ltd, index) \
339 ((ltd)->ltd_tgts_idx[(index) / TGT_PTRS_PER_BLOCK]->\
340 ldi_tgts[(index) % TGT_PTRS_PER_BLOCK])
342 #define OST_TGT(lfsck, index) LTD_TGT(&lfsck->li_ost_descs, index)
343 #define MDT_TGT(lfsck, index) LTD_TGT(&lfsck->li_mdt_descs, index)
345 struct lfsck_component {
346 /* into lfsck_instance::li_list_(scan,double_scan,idle} */
349 /* into lfsck_instance::li_list_dir */
350 cfs_list_t lc_link_dir;
352 struct rw_semaphore lc_sem;
355 struct lfsck_position lc_pos_start;
356 struct lfsck_instance *lc_lfsck;
357 struct dt_object *lc_obj;
358 struct lfsck_operations *lc_ops;
363 /* The time for last checkpoint, jiffies */
364 cfs_time_t lc_time_last_checkpoint;
366 /* The time for next checkpoint, jiffies */
367 cfs_time_t lc_time_next_checkpoint;
371 /* How many objects have been checked since last checkpoint. */
372 __u32 lc_new_checked;
374 /* How many objects have been scanned since last sleep. */
375 __u32 lc_new_scanned;
377 unsigned int lc_journal:1;
381 struct lfsck_instance {
382 struct mutex li_mutex;
385 /* Link into the lfsck_instance_list. */
388 /* For the components in (first) scanning via otable-based iteration. */
389 cfs_list_t li_list_scan;
391 /* For the components in scanning via directory traversal. Because
392 * directory traversal cannot guarantee all the object be scanned,
393 * so the component in the li_list_dir must be in li_list_scan. */
394 cfs_list_t li_list_dir;
396 /* For the components in double scanning. */
397 cfs_list_t li_list_double_scan;
399 /* For the components those are not scanning now. */
400 cfs_list_t li_list_idle;
403 atomic_t li_double_scan_count;
404 struct ptlrpc_thread li_thread;
406 /* The time for last checkpoint, jiffies */
407 cfs_time_t li_time_last_checkpoint;
409 /* The time for next checkpoint, jiffies */
410 cfs_time_t li_time_next_checkpoint;
412 lfsck_out_notify li_out_notify;
413 void *li_out_notify_data;
414 struct dt_device *li_next;
415 struct dt_device *li_bottom;
416 struct ldlm_namespace *li_namespace;
417 struct local_oid_storage *li_los;
418 struct lu_fid li_local_root_fid; /* backend root "/" */
419 struct lu_fid li_global_root_fid; /* /ROOT */
420 struct dt_object *li_bookmark_obj;
421 struct lfsck_bookmark li_bookmark_ram;
422 struct lfsck_bookmark li_bookmark_disk;
423 struct lfsck_position li_pos_current;
425 /* Obj for otable-based iteration */
426 struct dt_object *li_obj_oit;
428 /* Obj for directory traversal */
429 struct dt_object *li_obj_dir;
431 /* It for otable-based iteration */
432 struct dt_it *li_di_oit;
434 /* It for directory traversal */
435 struct dt_it *li_di_dir;
437 /* Description of OST */
438 struct lfsck_tgt_descs li_ost_descs;
440 /* Description of MDT */
441 struct lfsck_tgt_descs li_mdt_descs;
443 /* namespace-based directory traversal position. */
446 /* Arguments for low layer otable-based iteration. */
449 /* Arugments for namespace-based directory traversal. */
452 /* Schedule for every N objects. */
455 /* Sleep N jiffies for each schedule. */
458 /* How many objects have been scanned since last sleep. */
459 __u32 li_new_scanned;
461 unsigned int li_paused:1, /* The lfsck is paused. */
462 li_oit_over:1, /* oit is finished. */
463 li_drop_dryrun:1, /* Ever dryrun, not now. */
464 li_master:1, /* Master instance or not. */
465 li_current_oit_processed:1;
468 enum lfsck_linkea_flags {
469 /* The linkea entries does not match the object nlinks. */
470 LLF_UNMATCH_NLINKS = 0x01,
472 /* Fail to repair the multiple-linked objects during the double scan. */
473 LLF_REPAIR_FAILED = 0x02,
476 struct lfsck_thread_args {
477 struct lu_env lta_env;
478 struct lfsck_instance *lta_lfsck;
479 struct lfsck_component *lta_com;
482 struct lfsck_thread_info {
483 struct lu_name lti_name;
484 struct lu_buf lti_buf;
485 struct lu_buf lti_linkea_buf;
486 struct lu_fid lti_fid;
487 struct lu_fid lti_fid2;
488 struct lu_attr lti_la;
489 struct ost_id lti_oi;
491 struct lustre_mdt_attrs lti_lma;
492 /* old LMA for compatibility */
493 char lti_lma_old[LMA_OLD_SIZE];
495 struct dt_object_format lti_dof;
496 /* lti_ent and lti_key must be conjoint,
497 * then lti_ent::lde_name will be lti_key. */
498 struct lu_dirent lti_ent;
499 char lti_key[NAME_MAX + 16];
500 struct lfsck_request lti_lr;
504 const char *lfsck_status2names(enum lfsck_status status);
505 void lfsck_component_cleanup(const struct lu_env *env,
506 struct lfsck_component *com);
507 void lfsck_instance_cleanup(const struct lu_env *env,
508 struct lfsck_instance *lfsck);
509 int lfsck_bits_dump(char **buf, int *len, int bits, const char *names[],
511 int lfsck_time_dump(char **buf, int *len, __u64 time, const char *prefix);
512 int lfsck_pos_dump(char **buf, int *len, struct lfsck_position *pos,
514 void lfsck_pos_fill(const struct lu_env *env, struct lfsck_instance *lfsck,
515 struct lfsck_position *pos, bool init);
516 void lfsck_control_speed(struct lfsck_instance *lfsck);
517 void lfsck_control_speed_by_self(struct lfsck_component *com);
518 int lfsck_reset(const struct lu_env *env, struct lfsck_instance *lfsck,
520 struct lfsck_thread_args *lfsck_thread_args_init(struct lfsck_instance *lfsck,
521 struct lfsck_component *com);
522 void lfsck_thread_args_fini(struct lfsck_thread_args *lta);
523 void lfsck_fail(const struct lu_env *env, struct lfsck_instance *lfsck,
525 int lfsck_checkpoint(const struct lu_env *env, struct lfsck_instance *lfsck);
526 int lfsck_prep(const struct lu_env *env, struct lfsck_instance *lfsck);
527 int lfsck_exec_oit(const struct lu_env *env, struct lfsck_instance *lfsck,
528 struct dt_object *obj);
529 int lfsck_exec_dir(const struct lu_env *env, struct lfsck_instance *lfsck,
530 struct dt_object *obj, struct lu_dirent *ent);
531 int lfsck_post(const struct lu_env *env, struct lfsck_instance *lfsck,
533 int lfsck_double_scan(const struct lu_env *env, struct lfsck_instance *lfsck);
534 void lfsck_quit(const struct lu_env *env, struct lfsck_instance *lfsck);
537 int lfsck_master_engine(void *args);
539 /* lfsck_bookmark.c */
540 int lfsck_bookmark_store(const struct lu_env *env,
541 struct lfsck_instance *lfsck);
542 int lfsck_bookmark_setup(const struct lu_env *env,
543 struct lfsck_instance *lfsck);
545 /* lfsck_namespace.c */
546 int lfsck_namespace_setup(const struct lu_env *env,
547 struct lfsck_instance *lfsck);
550 int lfsck_layout_setup(const struct lu_env *env, struct lfsck_instance *lfsck);
552 extern const char *lfsck_flags_names[];
553 extern const char *lfsck_param_names[];
554 extern struct lu_context_key lfsck_thread_key;
556 static inline struct lfsck_thread_info *
557 lfsck_env_info(const struct lu_env *env)
559 struct lfsck_thread_info *info;
561 info = lu_context_key_get(&env->le_ctx, &lfsck_thread_key);
562 LASSERT(info != NULL);
566 static inline const struct lu_name *
567 lfsck_name_get_const(const struct lu_env *env, const void *area, ssize_t len)
569 struct lu_name *lname;
571 lname = &lfsck_env_info(env)->lti_name;
572 lname->ln_name = area;
573 lname->ln_namelen = len;
577 static inline struct lu_buf *
578 lfsck_buf_get(const struct lu_env *env, void *area, ssize_t len)
582 buf = &lfsck_env_info(env)->lti_buf;
588 static inline const struct lu_buf *
589 lfsck_buf_get_const(const struct lu_env *env, const void *area, ssize_t len)
593 buf = &lfsck_env_info(env)->lti_buf;
594 buf->lb_buf = (void *)area;
599 static inline char *lfsck_lfsck2name(struct lfsck_instance *lfsck)
601 return lfsck->li_bottom->dd_lu_dev.ld_obd->obd_name;
604 static inline const struct lu_fid *lfsck_dto2fid(const struct dt_object *obj)
606 return lu_object_fid(&obj->do_lu);
609 static inline void lfsck_pos_set_zero(struct lfsck_position *pos)
611 memset(pos, 0, sizeof(*pos));
614 static inline int lfsck_pos_is_zero(const struct lfsck_position *pos)
616 return pos->lp_oit_cookie == 0 && fid_is_zero(&pos->lp_dir_parent);
619 static inline int lfsck_pos_is_eq(const struct lfsck_position *pos1,
620 const struct lfsck_position *pos2)
622 if (pos1->lp_oit_cookie < pos2->lp_oit_cookie)
625 if (pos1->lp_oit_cookie > pos2->lp_oit_cookie)
628 if (fid_is_zero(&pos1->lp_dir_parent) &&
629 !fid_is_zero(&pos2->lp_dir_parent))
632 if (!fid_is_zero(&pos1->lp_dir_parent) &&
633 fid_is_zero(&pos2->lp_dir_parent))
636 if (fid_is_zero(&pos1->lp_dir_parent) &&
637 fid_is_zero(&pos2->lp_dir_parent))
640 LASSERT(lu_fid_eq(&pos1->lp_dir_parent, &pos2->lp_dir_parent));
642 if (pos1->lp_dir_cookie < pos2->lp_dir_cookie)
645 if (pos1->lp_dir_cookie > pos2->lp_dir_cookie)
651 static void inline lfsck_position_le_to_cpu(struct lfsck_position *des,
652 struct lfsck_position *src)
654 des->lp_oit_cookie = le64_to_cpu(src->lp_oit_cookie);
655 fid_le_to_cpu(&des->lp_dir_parent, &src->lp_dir_parent);
656 des->lp_dir_cookie = le64_to_cpu(src->lp_dir_cookie);
659 static void inline lfsck_position_cpu_to_le(struct lfsck_position *des,
660 struct lfsck_position *src)
662 des->lp_oit_cookie = cpu_to_le64(src->lp_oit_cookie);
663 fid_cpu_to_le(&des->lp_dir_parent, &src->lp_dir_parent);
664 des->lp_dir_cookie = cpu_to_le64(src->lp_dir_cookie);
667 static inline umode_t lfsck_object_type(const struct dt_object *obj)
669 return lu_object_attr(&obj->do_lu);
672 static inline int lfsck_is_dead_obj(const struct dt_object *obj)
674 struct lu_object_header *loh = obj->do_lu.lo_header;
676 return !!test_bit(LU_OBJECT_HEARD_BANSHEE, &loh->loh_flags);
679 static inline struct dt_object *lfsck_object_find(const struct lu_env *env,
680 struct lfsck_instance *lfsck,
681 const struct lu_fid *fid)
683 return lu2dt(lu_object_find_slice(env, dt2lu_dev(lfsck->li_next),
687 static inline struct dt_object *lfsck_object_get(struct dt_object *obj)
689 lu_object_get(&obj->do_lu);
693 static inline void lfsck_object_put(const struct lu_env *env,
694 struct dt_object *obj)
696 lu_object_put(env, &obj->do_lu);
699 static inline struct lfsck_tgt_desc *lfsck_tgt_get(struct lfsck_tgt_descs *ltds,
702 struct lfsck_tgt_desc *ltd;
704 ltd = LTD_TGT(ltds, index);
706 atomic_inc(<d->ltd_ref);
711 static inline void lfsck_tgt_put(struct lfsck_tgt_desc *ltd)
713 if (atomic_dec_and_test(<d->ltd_ref))
717 static inline struct lfsck_component *
718 lfsck_component_get(struct lfsck_component *com)
720 atomic_inc(&com->lc_ref);
725 static inline void lfsck_component_put(const struct lu_env *env,
726 struct lfsck_component *com)
728 if (atomic_dec_and_test(&com->lc_ref)) {
729 if (com->lc_obj != NULL)
730 lu_object_put_nocache(env, &com->lc_obj->do_lu);
731 if (com->lc_file_ram != NULL)
732 OBD_FREE(com->lc_file_ram, com->lc_file_size);
733 if (com->lc_file_disk != NULL)
734 OBD_FREE(com->lc_file_disk, com->lc_file_size);
735 if (com->lc_data != NULL) {
736 LASSERT(com->lc_ops->lfsck_data_release != NULL);
738 com->lc_ops->lfsck_data_release(env, com);
745 static inline struct lfsck_instance *
746 lfsck_instance_get(struct lfsck_instance *lfsck)
748 atomic_inc(&lfsck->li_ref);
753 static inline void lfsck_instance_put(const struct lu_env *env,
754 struct lfsck_instance *lfsck)
756 if (atomic_dec_and_test(&lfsck->li_ref))
757 lfsck_instance_cleanup(env, lfsck);
760 static inline mdsno_t lfsck_dev_idx(struct dt_device *dev)
762 return dev->dd_lu_dev.ld_site->ld_seq_site->ss_node_id;
765 #endif /* _LFSCK_INTERNAL_H */