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 Whamcloud, Inc.
26 * lustre/osd-ldiskfs/osd_scrub.c
28 * Top-level entry points into osd module
30 * The OI scrub is used for rebuilding Object Index files when restores MDT from
33 * The otable based iterator scans ldiskfs inode table to feed up layer LFSCK.
35 * Author: Fan Yong <yong.fan@whamcloud.com>
39 # define EXPORT_SYMTAB
41 #define DEBUG_SUBSYSTEM S_MDS
43 #include <lustre/lustre_idl.h>
44 #include <lustre_disk.h>
45 #include <dt_object.h>
47 #include "osd_internal.h"
49 #include "osd_scrub.h"
51 #define HALF_SEC (CFS_HZ >> 1)
53 static inline struct osd_device *osd_scrub2dev(struct osd_scrub *scrub)
55 return container_of0(scrub, struct osd_device, od_scrub);
58 static inline struct super_block *osd_scrub2sb(struct osd_scrub *scrub)
60 return osd_sb(osd_scrub2dev(scrub));
63 static void osd_scrub_file_to_cpu(struct scrub_file *des,
64 struct scrub_file *src)
66 memcpy(des->sf_uuid, src->sf_uuid, 16);
67 des->sf_flags = le64_to_cpu(src->sf_flags);
68 des->sf_magic = le32_to_cpu(src->sf_magic);
69 des->sf_status = le16_to_cpu(src->sf_status);
70 des->sf_param = le16_to_cpu(src->sf_param);
71 des->sf_time_last_complete =
72 le64_to_cpu(src->sf_time_last_complete);
73 des->sf_time_latest_start =
74 le64_to_cpu(src->sf_time_latest_start);
75 des->sf_time_last_checkpoint =
76 le64_to_cpu(src->sf_time_last_checkpoint);
77 des->sf_pos_latest_start =
78 le64_to_cpu(src->sf_pos_latest_start);
79 des->sf_pos_last_checkpoint =
80 le64_to_cpu(src->sf_pos_last_checkpoint);
81 des->sf_pos_first_inconsistent =
82 le64_to_cpu(src->sf_pos_first_inconsistent);
83 des->sf_items_checked =
84 le64_to_cpu(src->sf_items_checked);
85 des->sf_items_updated =
86 le64_to_cpu(src->sf_items_updated);
87 des->sf_items_failed =
88 le64_to_cpu(src->sf_items_failed);
89 des->sf_items_updated_prior =
90 le64_to_cpu(src->sf_items_updated_prior);
91 des->sf_run_time = le32_to_cpu(src->sf_run_time);
92 des->sf_success_count = le32_to_cpu(src->sf_success_count);
93 des->sf_oi_count = le16_to_cpu(src->sf_oi_count);
94 memcpy(des->sf_oi_bitmap, src->sf_oi_bitmap, SCRUB_OI_BITMAP_SIZE);
97 static void osd_scrub_file_to_le(struct scrub_file *des,
98 struct scrub_file *src)
100 memcpy(des->sf_uuid, src->sf_uuid, 16);
101 des->sf_flags = cpu_to_le64(src->sf_flags);
102 des->sf_magic = cpu_to_le32(src->sf_magic);
103 des->sf_status = cpu_to_le16(src->sf_status);
104 des->sf_param = cpu_to_le16(src->sf_param);
105 des->sf_time_last_complete =
106 cpu_to_le64(src->sf_time_last_complete);
107 des->sf_time_latest_start =
108 cpu_to_le64(src->sf_time_latest_start);
109 des->sf_time_last_checkpoint =
110 cpu_to_le64(src->sf_time_last_checkpoint);
111 des->sf_pos_latest_start =
112 cpu_to_le64(src->sf_pos_latest_start);
113 des->sf_pos_last_checkpoint =
114 cpu_to_le64(src->sf_pos_last_checkpoint);
115 des->sf_pos_first_inconsistent =
116 cpu_to_le64(src->sf_pos_first_inconsistent);
117 des->sf_items_checked =
118 cpu_to_le64(src->sf_items_checked);
119 des->sf_items_updated =
120 cpu_to_le64(src->sf_items_updated);
121 des->sf_items_failed =
122 cpu_to_le64(src->sf_items_failed);
123 des->sf_items_updated_prior =
124 cpu_to_le64(src->sf_items_updated_prior);
125 des->sf_run_time = cpu_to_le32(src->sf_run_time);
126 des->sf_success_count = cpu_to_le32(src->sf_success_count);
127 des->sf_oi_count = cpu_to_le16(src->sf_oi_count);
128 memcpy(des->sf_oi_bitmap, src->sf_oi_bitmap, SCRUB_OI_BITMAP_SIZE);
131 static void osd_scrub_file_init(struct osd_scrub *scrub, __u8 *uuid)
133 struct scrub_file *sf = &scrub->os_file;
135 memset(sf, 0, sizeof(*sf));
136 memcpy(sf->sf_uuid, uuid, 16);
137 sf->sf_magic = SCRUB_MAGIC_V1;
138 sf->sf_status = SS_INIT;
141 void osd_scrub_file_reset(struct osd_scrub *scrub, __u8 *uuid, __u64 flags)
143 struct scrub_file *sf = &scrub->os_file;
145 CDEBUG(D_LFSCK, "Reset OI scrub file, flags = "LPX64"\n", flags);
146 memcpy(sf->sf_uuid, uuid, 16);
147 sf->sf_status = SS_INIT;
148 sf->sf_flags |= flags;
151 sf->sf_time_latest_start = 0;
152 sf->sf_time_last_checkpoint = 0;
153 sf->sf_pos_latest_start = 0;
154 sf->sf_pos_last_checkpoint = 0;
155 sf->sf_pos_first_inconsistent = 0;
156 sf->sf_items_checked = 0;
157 sf->sf_items_updated = 0;
158 sf->sf_items_failed = 0;
159 sf->sf_items_updated_prior = 0;
162 static int osd_scrub_file_load(struct osd_scrub *scrub)
165 char *name = LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name;
166 int len = sizeof(scrub->os_file_disk);
169 rc = osd_ldiskfs_read(scrub->os_inode, &scrub->os_file_disk, len, &pos);
171 struct scrub_file *sf = &scrub->os_file;
173 osd_scrub_file_to_cpu(sf, &scrub->os_file_disk);
174 if (sf->sf_magic != SCRUB_MAGIC_V1) {
175 CWARN("%.16s: invalid scrub magic 0x%x != 0x%x\n,",
176 name, sf->sf_magic, SCRUB_MAGIC_V1);
177 /* Process it as new scrub file. */
182 } else if (rc != 0) {
183 CERROR("%.16s: fail to load scrub file, expected = %d, "
184 "rc = %d\n", name, len, rc);
188 /* return -ENOENT for empty scrub file case. */
195 int osd_scrub_file_store(struct osd_scrub *scrub)
197 struct osd_device *dev;
200 int len = sizeof(scrub->os_file_disk);
204 dev = container_of0(scrub, struct osd_device, od_scrub);
205 credits = osd_dto_credits_noquota[DTO_WRITE_BASE] +
206 osd_dto_credits_noquota[DTO_WRITE_BLOCK];
207 jh = ldiskfs_journal_start_sb(osd_sb(dev), credits);
210 CERROR("%.16s: fail to start trans for scrub store, rc = %d\n",
211 LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,rc);
215 osd_scrub_file_to_le(&scrub->os_file_disk, &scrub->os_file);
216 rc = osd_ldiskfs_write_record(scrub->os_inode, &scrub->os_file_disk,
218 ldiskfs_journal_stop(jh);
220 CERROR("%.16s: fail to store scrub file, expected = %d, "
222 LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,
224 scrub->os_time_last_checkpoint = cfs_time_current();
225 scrub->os_time_next_checkpoint = scrub->os_time_last_checkpoint +
226 cfs_time_seconds(SCRUB_CHECKPOINT_INTERVAL);
230 static int osd_scrub_prep(struct osd_device *dev)
232 struct osd_scrub *scrub = &dev->od_scrub;
233 struct ptlrpc_thread *thread = &scrub->os_thread;
234 struct scrub_file *sf = &scrub->os_file;
235 __u32 flags = scrub->os_start_flags;
239 cfs_down_write(&scrub->os_rwsem);
240 if (flags & SS_SET_FAILOUT)
241 sf->sf_param |= SP_FAILOUT;
243 if (flags & SS_CLEAR_FAILOUT)
244 sf->sf_param &= ~SP_FAILOUT;
246 if (flags & SS_RESET)
247 osd_scrub_file_reset(scrub,
248 LDISKFS_SB(osd_sb(dev))->s_es->s_uuid, sf->sf_flags);
250 if (flags & SS_AUTO) {
251 scrub->os_full_speed = 1;
252 sf->sf_flags |= SF_AUTO;
254 scrub->os_full_speed = 0;
257 if (sf->sf_flags & (SF_RECREATED | SF_INCONSISTENT))
258 scrub->os_full_speed = 1;
260 scrub->os_in_prior = 0;
261 scrub->os_waiting = 0;
262 scrub->os_new_checked = 0;
263 if (sf->sf_pos_last_checkpoint != 0)
264 sf->sf_pos_latest_start = sf->sf_pos_last_checkpoint + 1;
266 sf->sf_pos_latest_start = LDISKFS_FIRST_INO(osd_sb(dev));
268 scrub->os_pos_current = sf->sf_pos_latest_start;
269 sf->sf_status = SS_SCANNING;
270 sf->sf_time_latest_start = cfs_time_current_sec();
271 sf->sf_time_last_checkpoint = sf->sf_time_latest_start;
272 rc = osd_scrub_file_store(scrub);
274 cfs_spin_lock(&scrub->os_lock);
275 thread_set_flags(thread, SVC_RUNNING);
276 cfs_spin_unlock(&scrub->os_lock);
277 cfs_waitq_broadcast(&thread->t_ctl_waitq);
279 cfs_up_write(&scrub->os_rwsem);
285 osd_scrub_error(struct osd_device *dev, struct osd_inode_id *lid, int rc)
287 struct osd_scrub *scrub = &dev->od_scrub;
288 struct scrub_file *sf = &scrub->os_file;
290 cfs_down_write(&scrub->os_rwsem);
291 scrub->os_new_checked++;
292 sf->sf_items_failed++;
293 if (sf->sf_pos_first_inconsistent == 0 ||
294 sf->sf_pos_first_inconsistent > lid->oii_ino)
295 sf->sf_pos_first_inconsistent = lid->oii_ino;
296 cfs_up_write(&scrub->os_rwsem);
297 return sf->sf_param & SP_FAILOUT ? rc : 0;
301 osd_scrub_check_update(struct osd_thread_info *info, struct osd_device *dev,
302 struct osd_idmap_cache *oic)
304 struct osd_scrub *scrub = &dev->od_scrub;
305 struct scrub_file *sf = &scrub->os_file;
306 struct osd_inode_id *lid2 = &info->oti_id;
307 struct lu_fid *oi_fid = &info->oti_fid;
308 struct osd_inode_id *oi_id = &info->oti_id2;
310 struct osd_inconsistent_item *oii = NULL;
311 struct inode *inode = NULL;
312 struct lu_fid *fid = &oic->oic_fid;
313 struct osd_inode_id *lid = &oic->oic_lid;
314 struct iam_container *bag;
315 struct iam_path_descr *ipd;
316 int ops = DTO_INDEX_UPDATE;
321 if (scrub->os_in_prior)
322 oii = cfs_list_entry(oic, struct osd_inconsistent_item,
325 cfs_down_write(&scrub->os_rwsem);
326 scrub->os_new_checked++;
327 if (lid->oii_ino < sf->sf_pos_latest_start && oii == NULL)
330 if (oii != NULL && oii->oii_insert)
333 rc = osd_oi_lookup(info, dev, fid, lid2);
339 inode = osd_iget(info, dev, lid);
342 /* Someone removed the inode. */
343 if (rc == -ENOENT || rc == -ESTALE)
348 /* Prevent the inode to be unlinked during OI scrub. */
349 cfs_mutex_lock(&inode->i_mutex);
350 if (unlikely(inode->i_nlink == 0)) {
351 cfs_mutex_unlock(&inode->i_mutex);
356 ops = DTO_INDEX_INSERT;
357 idx = osd_oi_fid2idx(dev, fid);
358 if (unlikely(!ldiskfs_test_bit(idx, sf->sf_oi_bitmap)))
359 ldiskfs_set_bit(idx, sf->sf_oi_bitmap);
360 sf->sf_flags |= SF_RECREATED;
361 } else if (osd_id_eq(lid, lid2)) {
365 sf->sf_flags |= SF_INCONSISTENT;
366 fid_cpu_to_be(oi_fid, fid);
367 osd_id_pack(oi_id, &oic->oic_lid);
368 jh = ldiskfs_journal_start_sb(osd_sb(dev),
369 osd_dto_credits_noquota[ops]);
372 CERROR("%.16s: fail to start trans for scrub store, rc = %d\n",
373 LDISKFS_SB(osd_sb(dev))->s_es->s_volume_name, rc);
377 bag = &osd_fid2oi(dev, fid)->oi_dir.od_container;
378 ipd = osd_idx_ipd_get(info->oti_env, bag);
379 if (unlikely(ipd == NULL)) {
380 ldiskfs_journal_stop(jh);
381 CERROR("%.16s: fail to get ipd for scrub store\n",
382 LDISKFS_SB(osd_sb(dev))->s_es->s_volume_name);
383 GOTO(out, rc = -ENOMEM);
386 if (ops == DTO_INDEX_UPDATE)
387 rc = iam_update(jh, bag, (const struct iam_key *)oi_fid,
388 (struct iam_rec *)oi_id, ipd);
390 rc = iam_insert(jh, bag, (const struct iam_key *)oi_fid,
391 (struct iam_rec *)oi_id, ipd);
392 osd_ipd_put(info->oti_env, bag, ipd);
393 ldiskfs_journal_stop(jh);
395 if (scrub->os_in_prior)
396 sf->sf_items_updated_prior++;
398 sf->sf_items_updated++;
405 sf->sf_items_failed++;
406 if (sf->sf_pos_first_inconsistent == 0 ||
407 sf->sf_pos_first_inconsistent > lid->oii_ino)
408 sf->sf_pos_first_inconsistent = lid->oii_ino;
411 if (ops == DTO_INDEX_INSERT) {
412 cfs_mutex_unlock(&inode->i_mutex);
415 cfs_up_write(&scrub->os_rwsem);
418 LASSERT(!cfs_list_empty(&oii->oii_list));
420 cfs_spin_lock(&scrub->os_lock);
421 cfs_list_del_init(&oii->oii_list);
422 cfs_spin_unlock(&scrub->os_lock);
425 RETURN(sf->sf_param & SP_FAILOUT ? rc : 0);
428 static int do_osd_scrub_checkpoint(struct osd_scrub *scrub)
430 struct scrub_file *sf = &scrub->os_file;
434 cfs_down_write(&scrub->os_rwsem);
435 sf->sf_items_checked += scrub->os_new_checked;
436 scrub->os_new_checked = 0;
437 sf->sf_pos_last_checkpoint = scrub->os_pos_current;
438 sf->sf_time_last_checkpoint = cfs_time_current_sec();
439 sf->sf_run_time += cfs_duration_sec(cfs_time_current() + HALF_SEC -
440 scrub->os_time_last_checkpoint);
441 rc = osd_scrub_file_store(scrub);
442 cfs_up_write(&scrub->os_rwsem);
447 static inline int osd_scrub_checkpoint(struct osd_scrub *scrub)
449 if (unlikely(cfs_time_beforeq(scrub->os_time_next_checkpoint,
450 cfs_time_current()) &&
451 scrub->os_new_checked > 0))
452 return do_osd_scrub_checkpoint(scrub);
456 static void osd_scrub_post(struct osd_scrub *scrub, int result)
458 struct scrub_file *sf = &scrub->os_file;
461 cfs_down_write(&scrub->os_rwsem);
462 cfs_spin_lock(&scrub->os_lock);
463 thread_set_flags(&scrub->os_thread, SVC_STOPPING);
464 cfs_spin_unlock(&scrub->os_lock);
465 if (scrub->os_new_checked > 0) {
466 sf->sf_items_checked += scrub->os_new_checked;
467 scrub->os_new_checked = 0;
468 sf->sf_pos_last_checkpoint = scrub->os_pos_current;
470 sf->sf_time_last_checkpoint = cfs_time_current_sec();
472 sf->sf_status = SS_COMPLETED;
473 memset(sf->sf_oi_bitmap, 0, SCRUB_OI_BITMAP_SIZE);
474 sf->sf_flags &= ~(SF_RECREATED | SF_INCONSISTENT | SF_AUTO);
475 sf->sf_time_last_complete = sf->sf_time_last_checkpoint;
476 sf->sf_success_count++;
477 } else if (result == 0) {
478 if (scrub->os_paused)
479 sf->sf_status = SS_PAUSED;
481 sf->sf_status = SS_STOPPED;
483 sf->sf_status = SS_FAILED;
485 sf->sf_run_time += cfs_duration_sec(cfs_time_current() + HALF_SEC -
486 scrub->os_time_last_checkpoint);
487 result = osd_scrub_file_store(scrub);
489 CERROR("%.16s: fail to osd_scrub_post, rc = %d\n",
490 LDISKFS_SB(osd_scrub2sb(scrub))->s_es->s_volume_name,
492 cfs_up_write(&scrub->os_rwsem);
497 #define SCRUB_NEXT_BREAK 1 /* exit current loop and process next group */
498 #define SCRUB_NEXT_CONTINUE 2 /* skip current object and process next bit */
499 #define SCRUB_NEXT_EXIT 3 /* exit all the loops */
500 #define SCRUB_NEXT_WAIT 4 /* wait for free cache slot */
501 #define SCRUB_NEXT_CRASH 5 /* simulate system crash during OI scrub */
502 #define SCRUB_NEXT_FATAL 6 /* simulate failure during OI scrub */
504 struct osd_iit_param {
505 struct super_block *sb;
506 struct buffer_head *bitmap;
512 typedef int (*osd_iit_next_policy)(struct osd_thread_info *info,
513 struct osd_device *dev,
514 struct osd_iit_param *param,
515 struct osd_idmap_cache **oic,
518 typedef int (*osd_iit_exec_policy)(struct osd_thread_info *info,
519 struct osd_device *dev,
520 struct osd_iit_param *param,
521 struct osd_idmap_cache *oic,
522 int *noslot, int rc);
524 static inline int osd_scrub_has_window(struct osd_scrub *scrub,
525 struct osd_otable_cache *ooc)
527 return scrub->os_pos_current < ooc->ooc_pos_preload + SCRUB_WINDOW_SIZE;
530 static int osd_iit_next(struct osd_iit_param *param, __u32 *pos)
532 param->offset = ldiskfs_find_next_bit(param->bitmap->b_data,
533 LDISKFS_INODES_PER_GROUP(param->sb), param->offset);
534 if (param->offset >= LDISKFS_INODES_PER_GROUP(param->sb)) {
535 *pos = 1 + (param->bg+1) * LDISKFS_INODES_PER_GROUP(param->sb);
536 return SCRUB_NEXT_BREAK;
538 *pos = param->gbase + param->offset;
543 static int osd_iit_iget(struct osd_thread_info *info, struct osd_device *dev,
544 struct lu_fid *fid, struct osd_inode_id *lid, __u32 pos,
545 struct super_block *sb, struct inode **pinode)
550 osd_id_gen(lid, pos, OSD_OII_NOGEN);
551 inode = osd_iget_fid(info, dev, lid, fid);
554 /* The inode may be removed after bitmap searching, or the
555 * file is new created without inode initialized yet. */
556 if (rc == -ENOENT || rc == -ESTALE)
557 return SCRUB_NEXT_CONTINUE;
559 CERROR("%.16s: fail to read inode, ino# = %u, rc = %d\n",
560 LDISKFS_SB(sb)->s_es->s_volume_name, pos, rc);
568 static int osd_scrub_next(struct osd_thread_info *info, struct osd_device *dev,
569 struct osd_iit_param *param,
570 struct osd_idmap_cache **oic, int noslot)
572 struct osd_scrub *scrub = &dev->od_scrub;
573 struct ptlrpc_thread *thread = &scrub->os_thread;
575 struct osd_inode_id *lid;
579 if (OBD_FAIL_CHECK(OBD_FAIL_OSD_SCRUB_DELAY) && cfs_fail_val > 0) {
580 struct l_wait_info lwi;
582 lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val), NULL, NULL);
583 l_wait_event(thread->t_ctl_waitq,
584 !cfs_list_empty(&scrub->os_inconsistent_items) ||
585 !thread_is_running(thread),
589 if (OBD_FAIL_CHECK(OBD_FAIL_OSD_SCRUB_CRASH)) {
590 cfs_spin_lock(&scrub->os_lock);
591 thread_set_flags(thread, SVC_STOPPING);
592 cfs_spin_unlock(&scrub->os_lock);
593 return SCRUB_NEXT_CRASH;
596 if (OBD_FAIL_CHECK(OBD_FAIL_OSD_SCRUB_FATAL))
597 return SCRUB_NEXT_FATAL;
599 if (unlikely(!thread_is_running(thread)))
600 return SCRUB_NEXT_EXIT;
602 if (!cfs_list_empty(&scrub->os_inconsistent_items)) {
603 struct osd_inconsistent_item *oii;
605 oii = cfs_list_entry(scrub->os_inconsistent_items.next,
606 struct osd_inconsistent_item, oii_list);
607 *oic = &oii->oii_cache;
608 scrub->os_in_prior = 1;
613 return SCRUB_NEXT_WAIT;
615 rc = osd_iit_next(param, &scrub->os_pos_current);
619 *oic = &scrub->os_oic;
620 fid = &(*oic)->oic_fid;
621 lid = &(*oic)->oic_lid;
622 rc = osd_iit_iget(info, dev, fid, lid,
623 scrub->os_pos_current, param->sb, &inode);
627 if (!fid_is_norm(fid) || inode->i_state & I_LUSTRE_NOSCRUB)
628 rc = SCRUB_NEXT_CONTINUE;
633 static int osd_preload_next(struct osd_thread_info *info,
634 struct osd_device *dev, struct osd_iit_param *param,
635 struct osd_idmap_cache **oic, int noslot)
637 struct osd_otable_cache *ooc = &dev->od_otable_it->ooi_cache;
638 struct osd_scrub *scrub;
639 struct ptlrpc_thread *thread;
643 rc = osd_iit_next(param, &ooc->ooc_pos_preload);
647 scrub = &dev->od_scrub;
648 thread = &scrub->os_thread;
649 if (thread_is_running(thread) &&
650 ooc->ooc_pos_preload >= scrub->os_pos_current)
651 return SCRUB_NEXT_EXIT;
653 rc = osd_iit_iget(info, dev,
654 &ooc->ooc_cache[ooc->ooc_producer_idx].oic_fid,
655 &ooc->ooc_cache[ooc->ooc_producer_idx].oic_lid,
656 ooc->ooc_pos_preload, param->sb, &inode);
657 /* If succeed, it needs to move forward; otherwise up layer LFSCK may
658 * ignore the failure, so it still need to skip the inode next time. */
659 ooc->ooc_pos_preload = param->gbase + ++(param->offset);
665 static int osd_scrub_exec(struct osd_thread_info *info, struct osd_device *dev,
666 struct osd_iit_param *param,
667 struct osd_idmap_cache *oic, int *noslot, int rc)
669 struct l_wait_info lwi = { 0 };
670 struct osd_scrub *scrub = &dev->od_scrub;
671 struct ptlrpc_thread *thread = &scrub->os_thread;
672 struct osd_otable_it *it = dev->od_otable_it;
673 struct osd_otable_cache *ooc = it ? &it->ooi_cache : NULL;
676 case SCRUB_NEXT_CONTINUE:
678 case SCRUB_NEXT_WAIT:
682 LASSERTF(rc <= 0, "unexpected rc = %d\n", rc);
685 rc = osd_scrub_error(dev, &oic->oic_lid, rc);
687 rc = osd_scrub_check_update(info, dev, oic);
691 rc = osd_scrub_checkpoint(scrub);
693 CERROR("%.16s: fail to checkpoint, pos = %u, rc = %d\n",
694 LDISKFS_SB(param->sb)->s_es->s_volume_name,
695 scrub->os_pos_current, rc);
696 /* Continue, as long as the scrub itself can go ahead. */
699 if (scrub->os_in_prior) {
700 scrub->os_in_prior = 0;
705 scrub->os_pos_current = param->gbase + ++(param->offset);
706 if (it != NULL && it->ooi_waiting &&
707 ooc->ooc_pos_preload < scrub->os_pos_current) {
709 cfs_waitq_broadcast(&thread->t_ctl_waitq);
712 if (scrub->os_full_speed || rc == SCRUB_NEXT_CONTINUE)
716 if (osd_scrub_has_window(scrub, ooc)) {
721 scrub->os_waiting = 1;
722 l_wait_event(thread->t_ctl_waitq,
723 osd_scrub_has_window(scrub, ooc) ||
724 !cfs_list_empty(&scrub->os_inconsistent_items) ||
725 !thread_is_running(thread),
727 scrub->os_waiting = 0;
729 if (osd_scrub_has_window(scrub, ooc))
736 static int osd_preload_exec(struct osd_thread_info *info,
737 struct osd_device *dev, struct osd_iit_param *param,
738 struct osd_idmap_cache *oic, int *noslot, int rc)
740 struct osd_otable_cache *ooc = &dev->od_otable_it->ooi_cache;
743 ooc->ooc_cached_items++;
744 ooc->ooc_producer_idx = (ooc->ooc_producer_idx + 1) &
745 ~OSD_OTABLE_IT_CACHE_MASK;
747 return rc > 0 ? 0 : rc;
750 #define SCRUB_IT_ALL 1
751 #define SCRUB_IT_CRASH 2
753 static int osd_inode_iteration(struct osd_thread_info *info,
754 struct osd_device *dev, __u32 max, int preload)
756 osd_iit_next_policy next;
757 osd_iit_exec_policy exec;
760 struct osd_iit_param param;
767 struct osd_scrub *scrub = &dev->od_scrub;
769 next = osd_scrub_next;
770 exec = osd_scrub_exec;
771 pos = &scrub->os_pos_current;
772 count = &scrub->os_new_checked;
774 struct osd_otable_cache *ooc = &dev->od_otable_it->ooi_cache;
776 next = osd_preload_next;
777 exec = osd_preload_exec;
778 pos = &ooc->ooc_pos_preload;
779 count = &ooc->ooc_cached_items;
781 param.sb = osd_sb(dev);
782 limit = le32_to_cpu(LDISKFS_SB(param.sb)->s_es->s_inodes_count);
784 while (*pos <= limit && *count < max) {
785 struct osd_idmap_cache *oic = NULL;
787 param.bg = (*pos - 1) / LDISKFS_INODES_PER_GROUP(param.sb);
788 param.offset = (*pos - 1) % LDISKFS_INODES_PER_GROUP(param.sb);
789 param.gbase = 1 + param.bg * LDISKFS_INODES_PER_GROUP(param.sb);
790 param.bitmap = ldiskfs_read_inode_bitmap(param.sb, param.bg);
791 if (param.bitmap == NULL) {
792 CERROR("%.16s: fail to read bitmap for %u, "
793 "scrub will stop, urgent mode\n",
794 LDISKFS_SB(param.sb)->s_es->s_volume_name,
799 while (param.offset < LDISKFS_INODES_PER_GROUP(param.sb) &&
801 rc = next(info, dev, ¶m, &oic, noslot);
803 case SCRUB_NEXT_BREAK:
805 case SCRUB_NEXT_EXIT:
806 brelse(param.bitmap);
808 case SCRUB_NEXT_CRASH:
809 brelse(param.bitmap);
810 RETURN(SCRUB_IT_CRASH);
811 case SCRUB_NEXT_FATAL:
812 brelse(param.bitmap);
816 rc = exec(info, dev, ¶m, oic, &noslot, rc);
818 brelse(param.bitmap);
824 brelse(param.bitmap);
828 RETURN(SCRUB_IT_ALL);
832 static int osd_scrub_main(void *args)
835 struct osd_device *dev = (struct osd_device *)args;
836 struct osd_scrub *scrub = &dev->od_scrub;
837 struct ptlrpc_thread *thread = &scrub->os_thread;
838 struct super_block *sb = osd_sb(dev);
842 cfs_daemonize("OI_scrub");
843 rc = lu_env_init(&env, LCT_DT_THREAD);
845 CERROR("%.16s: OI scrub, fail to init env, rc = %d\n",
846 LDISKFS_SB(sb)->s_es->s_volume_name, rc);
850 rc = osd_scrub_prep(dev);
852 CERROR("%.16s: OI scrub, fail to scrub prep, rc = %d\n",
853 LDISKFS_SB(sb)->s_es->s_volume_name, rc);
857 if (!scrub->os_full_speed) {
858 struct l_wait_info lwi = { 0 };
859 struct osd_otable_it *it = dev->od_otable_it;
860 struct osd_otable_cache *ooc = &it->ooi_cache;
862 l_wait_event(thread->t_ctl_waitq,
863 it->ooi_user_ready || !thread_is_running(thread),
865 if (unlikely(!thread_is_running(thread)))
868 LASSERT(scrub->os_pos_current >= ooc->ooc_pos_preload);
869 scrub->os_pos_current = ooc->ooc_pos_preload;
872 CDEBUG(D_LFSCK, "OI scrub: flags = 0x%x, pos = %u\n",
873 scrub->os_start_flags, scrub->os_pos_current);
875 rc = osd_inode_iteration(osd_oti_get(&env), dev, ~0U, 0);
876 if (unlikely(rc == SCRUB_IT_CRASH))
877 GOTO(out, rc = -EINVAL);
881 osd_scrub_post(scrub, rc);
882 CDEBUG(D_LFSCK, "OI scrub: stop, rc = %d, pos = %u\n",
883 rc, scrub->os_pos_current);
886 while (!cfs_list_empty(&scrub->os_inconsistent_items)) {
887 struct osd_inconsistent_item *oii;
889 oii = cfs_list_entry(scrub->os_inconsistent_items.next,
890 struct osd_inconsistent_item, oii_list);
891 cfs_list_del_init(&oii->oii_list);
897 cfs_spin_lock(&scrub->os_lock);
898 thread_set_flags(thread, SVC_STOPPED);
899 cfs_waitq_broadcast(&thread->t_ctl_waitq);
900 cfs_spin_unlock(&scrub->os_lock);
904 static int do_osd_scrub_start(struct osd_device *dev, __u32 flags)
906 struct osd_scrub *scrub = &dev->od_scrub;
907 struct ptlrpc_thread *thread = &scrub->os_thread;
908 struct l_wait_info lwi = { 0 };
913 /* os_lock: sync status between stop and scrub thread */
914 cfs_spin_lock(&scrub->os_lock);
915 if (thread_is_running(thread)) {
916 cfs_spin_unlock(&scrub->os_lock);
918 } else if (unlikely(thread_is_stopping(thread))) {
919 cfs_spin_unlock(&scrub->os_lock);
920 l_wait_event(thread->t_ctl_waitq,
921 thread_is_stopped(thread),
925 cfs_spin_unlock(&scrub->os_lock);
927 scrub->os_start_flags = flags;
928 thread_set_flags(thread, 0);
929 rc = cfs_create_thread(osd_scrub_main, dev, 0);
931 CERROR("%.16s: cannot start iteration thread, rc = %d\n",
932 LDISKFS_SB(osd_sb(dev))->s_es->s_volume_name, rc);
936 l_wait_event(thread->t_ctl_waitq,
937 thread_is_running(thread) || thread_is_stopped(thread),
943 int osd_scrub_start(struct osd_device *dev)
945 __u32 flags = SS_AUTO;
949 if (dev->od_scrub.os_file.sf_status == SS_COMPLETED)
952 /* od_otable_mutex: prevent curcurrent start/stop */
953 cfs_mutex_lock(&dev->od_otable_mutex);
954 rc = do_osd_scrub_start(dev, flags);
955 cfs_mutex_unlock(&dev->od_otable_mutex);
957 RETURN(rc == -EALREADY ? 0 : rc);
960 static void do_osd_scrub_stop(struct osd_scrub *scrub)
962 struct ptlrpc_thread *thread = &scrub->os_thread;
963 struct l_wait_info lwi = { 0 };
965 /* os_lock: sync status between stop and scrub thread */
966 cfs_spin_lock(&scrub->os_lock);
967 if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
968 thread_set_flags(thread, SVC_STOPPING);
969 cfs_spin_unlock(&scrub->os_lock);
970 cfs_waitq_broadcast(&thread->t_ctl_waitq);
971 l_wait_event(thread->t_ctl_waitq,
972 thread_is_stopped(thread),
974 /* Do not skip the last lock/unlock, which can guarantee that
975 * the caller cannot return until the OI scrub thread exit. */
976 cfs_spin_lock(&scrub->os_lock);
978 cfs_spin_unlock(&scrub->os_lock);
981 static void osd_scrub_stop(struct osd_device *dev)
983 /* od_otable_mutex: prevent curcurrent start/stop */
984 cfs_mutex_lock(&dev->od_otable_mutex);
985 dev->od_scrub.os_paused = 1;
986 do_osd_scrub_stop(&dev->od_scrub);
987 cfs_mutex_unlock(&dev->od_otable_mutex);
990 static const char osd_scrub_name[] = "OI_scrub";
992 int osd_scrub_setup(const struct lu_env *env, struct osd_device *dev)
994 struct osd_thread_info *info = osd_oti_get(env);
995 struct osd_scrub *scrub = &dev->od_scrub;
996 struct lvfs_run_ctxt *ctxt = &scrub->os_ctxt;
997 struct scrub_file *sf = &scrub->os_file;
998 struct osd_inode_id *id = &scrub->os_oic.oic_lid;
999 struct super_block *sb = osd_sb(dev);
1000 struct ldiskfs_super_block *es = LDISKFS_SB(sb)->s_es;
1001 struct inode *inode;
1002 struct lvfs_run_ctxt saved;
1009 memset(scrub, 0, sizeof(*scrub));
1010 OBD_SET_CTXT_MAGIC(ctxt);
1011 ctxt->pwdmnt = dev->od_mnt;
1012 ctxt->pwd = dev->od_mnt->mnt_root;
1013 ctxt->fs = get_ds();
1015 cfs_waitq_init(&scrub->os_thread.t_ctl_waitq);
1016 cfs_init_rwsem(&scrub->os_rwsem);
1017 cfs_spin_lock_init(&scrub->os_lock);
1018 CFS_INIT_LIST_HEAD(&scrub->os_inconsistent_items);
1019 if (get_mount_flags(dev->od_mount->lmi_sb) & LMD_FLG_NOSCRUB)
1020 scrub->os_no_scrub = 1;
1022 push_ctxt(&saved, ctxt, NULL);
1023 filp = filp_open(osd_scrub_name, O_RDWR | O_CREAT, 0644);
1025 RETURN(PTR_ERR(filp));
1027 scrub->os_inode = igrab(filp->f_dentry->d_inode);
1028 filp_close(filp, 0);
1029 pop_ctxt(&saved, ctxt, NULL);
1031 rc = osd_scrub_file_load(scrub);
1032 if (rc == -ENOENT) {
1033 osd_scrub_file_init(scrub, es->s_uuid);
1036 } else if (rc != 0) {
1039 if (memcmp(sf->sf_uuid, es->s_uuid, 16) != 0) {
1040 osd_scrub_file_reset(scrub, es->s_uuid,SF_INCONSISTENT);
1042 } else if (sf->sf_status == SS_SCANNING) {
1043 sf->sf_status = SS_CRASHED;
1048 if (sf->sf_pos_last_checkpoint != 0)
1049 scrub->os_pos_current = sf->sf_pos_last_checkpoint + 1;
1051 scrub->os_pos_current = LDISKFS_FIRST_INO(sb);
1054 rc = osd_scrub_file_store(scrub);
1059 /* Initialize OI files. */
1060 rc = osd_oi_init(info, dev);
1065 rc = __osd_oi_lookup(info, dev, &LU_DOT_LUSTRE_FID, id);
1067 inode = osd_iget(info, dev, id);
1068 if (IS_ERR(inode)) {
1069 rc = PTR_ERR(inode);
1070 /* It is restored from old 2.x backup. */
1071 if (rc == -ENOENT || rc == -ESTALE) {
1072 osd_scrub_file_reset(scrub, es->s_uuid,
1074 rc = osd_scrub_file_store(scrub);
1079 } else if (rc == -ENOENT) {
1084 if (rc == 0 && !scrub->os_no_scrub &&
1085 ((sf->sf_status == SS_PAUSED) ||
1086 (sf->sf_status == SS_CRASHED &&
1087 sf->sf_flags & (SF_RECREATED | SF_INCONSISTENT | SF_AUTO)) ||
1088 (sf->sf_status == SS_INIT &&
1089 sf->sf_flags & (SF_RECREATED | SF_INCONSISTENT))))
1090 rc = osd_scrub_start(dev);
1095 void osd_scrub_cleanup(const struct lu_env *env, struct osd_device *dev)
1097 struct osd_scrub *scrub = &dev->od_scrub;
1099 LASSERT(dev->od_otable_it == NULL);
1101 if (scrub->os_inode != NULL) {
1102 osd_scrub_stop(dev);
1103 iput(scrub->os_inode);
1104 scrub->os_inode = NULL;
1106 if (dev->od_oi_table != NULL)
1107 osd_oi_fini(osd_oti_get(env), dev);
1110 static struct dt_it *osd_otable_it_init(const struct lu_env *env,
1111 struct dt_object *dt, __u32 attr,
1112 struct lustre_capa *capa)
1114 enum dt_otable_it_flags flags = attr >> DT_OTABLE_IT_FLAGS_SHIFT;
1115 enum dt_otable_it_valid valid = attr & ~DT_OTABLE_IT_FLAGS_MASK;
1116 struct osd_device *dev = osd_dev(dt->do_lu.lo_dev);
1117 struct osd_scrub *scrub = &dev->od_scrub;
1118 struct osd_otable_it *it;
1123 /* od_otable_mutex: prevent curcurrent init/fini */
1124 cfs_mutex_lock(&dev->od_otable_mutex);
1125 if (dev->od_otable_it != NULL)
1126 GOTO(out, it = ERR_PTR(-EALREADY));
1130 GOTO(out, it = ERR_PTR(-ENOMEM));
1132 dev->od_otable_it = it;
1134 it->ooi_cache.ooc_consumer_idx = -1;
1135 if (flags & DOIF_OUTUSED)
1136 it->ooi_used_outside = 1;
1138 if (flags & DOIF_RESET)
1141 if (valid & DOIV_ERROR_HANDLE) {
1142 if (flags & DOIF_FAILOUT)
1143 start |= SS_SET_FAILOUT;
1145 start |= SS_CLEAR_FAILOUT;
1148 rc = do_osd_scrub_start(dev, start);
1149 if (rc == -EALREADY) {
1150 it->ooi_cache.ooc_pos_preload = scrub->os_pos_current - 1;
1151 } else if (rc < 0) {
1152 dev->od_otable_it = NULL;
1154 GOTO(out, it = ERR_PTR(-EALREADY));
1156 it->ooi_cache.ooc_pos_preload = scrub->os_pos_current;
1162 cfs_mutex_unlock(&dev->od_otable_mutex);
1163 return (struct dt_it *)it;
1166 static void osd_otable_it_fini(const struct lu_env *env, struct dt_it *di)
1168 struct osd_otable_it *it = (struct osd_otable_it *)di;
1169 struct osd_device *dev = it->ooi_dev;
1171 /* od_otable_mutex: prevent curcurrent init/fini */
1172 cfs_mutex_lock(&dev->od_otable_mutex);
1173 do_osd_scrub_stop(&dev->od_scrub);
1174 LASSERT(dev->od_otable_it == it);
1176 dev->od_otable_it = NULL;
1177 cfs_mutex_unlock(&dev->od_otable_mutex);
1182 * XXX: Temporary used to notify otable iteration to be paused.
1184 static void osd_otable_it_put(const struct lu_env *env, struct dt_it *di)
1186 struct osd_device *dev = ((struct osd_otable_it *)di)->ooi_dev;
1188 /* od_otable_mutex: prevent curcurrent init/fini */
1189 cfs_mutex_lock(&dev->od_otable_mutex);
1190 dev->od_scrub.os_paused = 1;
1191 cfs_mutex_unlock(&dev->od_otable_mutex);
1195 * Set the OSD layer iteration start position as the specified key.
1197 * The LFSCK out of OSD layer does not know the detail of the key, so if there
1198 * are several keys, they cannot be compared out of OSD, so call "::get()" for
1199 * each key, and OSD will select the smallest one by itself.
1201 static int osd_otable_it_get(const struct lu_env *env,
1202 struct dt_it *di, const struct dt_key *key)
1204 struct osd_otable_it *it = (struct osd_otable_it *)di;
1205 struct osd_otable_cache *ooc = &it->ooi_cache;
1206 const char *str = (const char *)key;
1210 /* Forbid to set iteration position after iteration started. */
1211 if (it->ooi_user_ready)
1217 if (sscanf(str, "%u", &ino) <= 0)
1220 /* Skip the one that has been processed last time. */
1221 if (ooc->ooc_pos_preload > ++ino)
1222 ooc->ooc_pos_preload = ino;
1227 static int osd_otable_it_preload(const struct lu_env *env,
1228 struct osd_otable_it *it)
1230 struct osd_device *dev = it->ooi_dev;
1231 struct osd_scrub *scrub = &dev->od_scrub;
1232 struct osd_otable_cache *ooc = &it->ooi_cache;
1236 rc = osd_inode_iteration(osd_oti_get(env), dev,
1237 OSD_OTABLE_IT_CACHE_SIZE, 1);
1238 if (rc == SCRUB_IT_ALL)
1239 it->ooi_all_cached = 1;
1241 CDEBUG(D_LFSCK, "OSD pre-loaded: max = %u, preload = %u, rc = %d\n",
1242 le32_to_cpu(LDISKFS_SB(osd_sb(dev))->s_es->s_inodes_count),
1243 ooc->ooc_pos_preload, rc);
1245 if (scrub->os_waiting && osd_scrub_has_window(scrub, ooc)) {
1246 scrub->os_waiting = 0;
1247 cfs_waitq_broadcast(&scrub->os_thread.t_ctl_waitq);
1250 RETURN(rc < 0 ? rc : ooc->ooc_cached_items);
1253 static int osd_otable_it_next(const struct lu_env *env, struct dt_it *di)
1255 struct osd_otable_it *it = (struct osd_otable_it *)di;
1256 struct osd_device *dev = it->ooi_dev;
1257 struct osd_scrub *scrub = &dev->od_scrub;
1258 struct osd_otable_cache *ooc = &it->ooi_cache;
1259 struct ptlrpc_thread *thread = &scrub->os_thread;
1260 struct l_wait_info lwi = { 0 };
1264 LASSERT(it->ooi_user_ready);
1267 if (!thread_is_running(thread) && !it->ooi_used_outside)
1270 if (ooc->ooc_cached_items > 0) {
1271 ooc->ooc_cached_items--;
1272 ooc->ooc_consumer_idx = (ooc->ooc_consumer_idx + 1) &
1273 ~OSD_OTABLE_IT_CACHE_MASK;
1277 if (it->ooi_all_cached) {
1278 l_wait_event(thread->t_ctl_waitq,
1279 !thread_is_running(thread),
1284 it->ooi_waiting = 1;
1285 l_wait_event(thread->t_ctl_waitq,
1286 ooc->ooc_pos_preload < scrub->os_pos_current ||
1287 !thread_is_running(thread),
1289 it->ooi_waiting = 0;
1291 if (!thread_is_running(thread) && !it->ooi_used_outside)
1294 rc = osd_otable_it_preload(env, it);
1301 static struct dt_key *osd_otable_it_key(const struct lu_env *env,
1302 const struct dt_it *di)
1304 struct osd_otable_it *it = (struct osd_otable_it *)di;
1305 struct osd_otable_cache *ooc = &it->ooi_cache;
1307 sprintf(it->ooi_key, "%u",
1308 ooc->ooc_cache[ooc->ooc_consumer_idx].oic_lid.oii_ino);
1309 return (struct dt_key *)it->ooi_key;
1312 static int osd_otable_it_key_size(const struct lu_env *env,
1313 const struct dt_it *di)
1315 return sizeof(((struct osd_otable_it *)di)->ooi_key);
1318 static int osd_otable_it_rec(const struct lu_env *env, const struct dt_it *di,
1319 struct dt_rec *rec, __u32 attr)
1321 struct osd_otable_it *it = (struct osd_otable_it *)di;
1322 struct osd_otable_cache *ooc = &it->ooi_cache;
1324 *(struct lu_fid *)rec = ooc->ooc_cache[ooc->ooc_consumer_idx].oic_fid;
1328 static int osd_otable_it_load(const struct lu_env *env,
1329 const struct dt_it *di, __u64 hash)
1331 struct osd_otable_it *it = (struct osd_otable_it *)di;
1332 struct osd_device *dev = it->ooi_dev;
1333 struct osd_otable_cache *ooc = &it->ooi_cache;
1334 struct osd_scrub *scrub = &dev->od_scrub;
1336 if (it->ooi_user_ready)
1339 if (ooc->ooc_pos_preload < LDISKFS_FIRST_INO(osd_sb(dev)))
1340 ooc->ooc_pos_preload = LDISKFS_FIRST_INO(osd_sb(dev));
1341 it->ooi_user_ready = 1;
1342 if (!scrub->os_full_speed)
1343 cfs_waitq_broadcast(&scrub->os_thread.t_ctl_waitq);
1345 /* Unplug OSD layer iteration by the first next() call. */
1346 return osd_otable_it_next(env, (struct dt_it *)it);
1349 const struct dt_index_operations osd_otable_ops = {
1351 .init = osd_otable_it_init,
1352 .fini = osd_otable_it_fini,
1353 .put = osd_otable_it_put,
1354 .get = osd_otable_it_get,
1355 .next = osd_otable_it_next,
1356 .key = osd_otable_it_key,
1357 .key_size = osd_otable_it_key_size,
1358 .rec = osd_otable_it_rec,
1359 .load = osd_otable_it_load,
1363 int osd_oii_insert(struct osd_device *dev, struct osd_idmap_cache *oic,
1366 struct osd_inconsistent_item *oii;
1367 struct osd_scrub *scrub = &dev->od_scrub;
1368 struct ptlrpc_thread *thread = &scrub->os_thread;
1373 if (unlikely(oii == NULL))
1376 CFS_INIT_LIST_HEAD(&oii->oii_list);
1377 oii->oii_cache = *oic;
1378 oii->oii_insert = insert;
1380 cfs_spin_lock(&scrub->os_lock);
1381 if (unlikely(!thread_is_running(thread))) {
1382 cfs_spin_unlock(&scrub->os_lock);
1387 if (cfs_list_empty(&scrub->os_inconsistent_items))
1389 cfs_list_add_tail(&oii->oii_list, &scrub->os_inconsistent_items);
1390 cfs_spin_unlock(&scrub->os_lock);
1393 cfs_waitq_broadcast(&thread->t_ctl_waitq);
1398 int osd_oii_lookup(struct osd_device *dev, const struct lu_fid *fid,
1399 struct osd_inode_id *id)
1401 struct osd_scrub *scrub = &dev->od_scrub;
1402 struct osd_inconsistent_item *oii;
1405 cfs_spin_lock(&scrub->os_lock);
1406 cfs_list_for_each_entry(oii, &scrub->os_inconsistent_items, oii_list) {
1407 if (lu_fid_eq(fid, &oii->oii_cache.oic_fid)) {
1408 *id = oii->oii_cache.oic_lid;
1409 cfs_spin_unlock(&scrub->os_lock);
1413 cfs_spin_unlock(&scrub->os_lock);
1418 static const char *scrub_status_names[] = {
1429 static const char *scrub_flags_names[] = {
1436 static const char *scrub_param_names[] = {
1441 static int scrub_bits_dump(char **buf, int *len, int bits, const char *names[],
1449 rc = snprintf(*buf, *len, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
1455 for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
1458 rc = snprintf(*buf, *len, "%s%c", names[i],
1459 bits != 0 ? ',' : '\n');
1470 static int scrub_time_dump(char **buf, int *len, __u64 time, const char *prefix)
1475 rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
1476 cfs_time_current_sec() - time);
1478 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
1487 static int scrub_pos_dump(char **buf, int *len, __u64 pos, const char *prefix)
1492 rc = snprintf(*buf, *len, "%s: "LPU64"\n", prefix, pos);
1494 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
1503 int osd_scrub_dump(struct osd_device *dev, char *buf, int len)
1505 struct osd_scrub *scrub = &dev->od_scrub;
1506 struct scrub_file *sf = &scrub->os_file;
1513 cfs_down_read(&scrub->os_rwsem);
1514 rc = snprintf(buf, len,
1519 sf->sf_magic, (int)sf->sf_oi_count,
1520 scrub_status_names[sf->sf_status]);
1526 rc = scrub_bits_dump(&buf, &len, sf->sf_flags, scrub_flags_names,
1531 rc = scrub_bits_dump(&buf, &len, sf->sf_param, scrub_param_names,
1536 rc = scrub_time_dump(&buf, &len, sf->sf_time_last_complete,
1537 "time_since_last_completed");
1541 rc = scrub_time_dump(&buf, &len, sf->sf_time_latest_start,
1542 "time_since_latest_start");
1546 rc = scrub_time_dump(&buf, &len, sf->sf_time_last_checkpoint,
1547 "time_since_last_checkpoint");
1551 rc = scrub_pos_dump(&buf, &len, sf->sf_pos_latest_start,
1552 "latest_start_position");
1556 rc = scrub_pos_dump(&buf, &len, sf->sf_pos_last_checkpoint,
1557 "last_checkpoint_position");
1561 rc = scrub_pos_dump(&buf, &len, sf->sf_pos_first_inconsistent,
1562 "first_failure_position");
1566 checked = sf->sf_items_checked + scrub->os_new_checked;
1567 rc = snprintf(buf, len,
1568 "checked: "LPU64"\n"
1569 "updated: "LPU64"\n"
1571 "prior_updated: "LPU64"\n"
1572 "success_count: %u\n",
1573 checked, sf->sf_items_updated, sf->sf_items_failed,
1574 sf->sf_items_updated_prior, sf->sf_success_count);
1581 if (thread_is_running(&scrub->os_thread)) {
1582 cfs_duration_t duration = cfs_time_current() -
1583 scrub->os_time_last_checkpoint;
1584 __u64 new_checked = scrub->os_new_checked * CFS_HZ;
1585 __u32 rtime = sf->sf_run_time +
1586 cfs_duration_sec(duration + HALF_SEC);
1589 do_div(new_checked, duration);
1591 do_div(speed, rtime);
1592 rc = snprintf(buf, len,
1593 "run_time: %u seconds\n"
1594 "average_speed: "LPU64" objects/sec\n"
1595 "real-time_speed: "LPU64" objects/sec\n"
1596 "current_position: %u\n",
1597 rtime, speed, new_checked, scrub->os_pos_current);
1599 if (sf->sf_run_time != 0)
1600 do_div(speed, sf->sf_run_time);
1601 rc = snprintf(buf, len,
1602 "run_time: %u seconds\n"
1603 "average_speed: "LPU64" objects/sec\n"
1604 "real-time_speed: N/A\n"
1605 "current_position: N/A\n",
1606 sf->sf_run_time, speed);
1616 cfs_up_read(&scrub->os_rwsem);