lustre/mdd/mdd_lfsck.c

   1 /*
   2  * GPL HEADER START
   3  *
   4  * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
   5  *
   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.
   9
  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.
  15
  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
  19  *
  20  * GPL HEADER END
  21  */
  22 /*
  23  * Copyright (c) 2012, Intel Corporation.
  24  */
  25 /*
  26  * lustre/mdd/mdd_lfsck.c
  27  *
  28  * Top-level entry points into mdd module
  29  *
  30  * LFSCK controller, which scans the whole device through low layer
  31  * iteration APIs, drives all lfsck compeonents, controls the speed.
  32  *
  33  * Author: Fan Yong <yong.fan@whamcloud.com>
  34  */
  35
  36 #ifndef EXPORT_SYMTAB
  37 # define EXPORT_SYMTAB
  38 #endif
  39 #define DEBUG_SUBSYSTEM S_MDS
  40
  41 #include <lustre/lustre_idl.h>
  42 #include <lustre_fid.h>
  43 #include <obd_support.h>
  44
  45 #include "mdd_internal.h"
  46 #include "mdd_lfsck.h"
  47
  48 #define HALF_SEC                        (CFS_HZ >> 1)
  49 #define LFSCK_CHECKPOINT_INTERVAL       60
  50 #define MDS_DIR_DUMMY_START             0xffffffffffffffffULL
  51
  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. */
  55
  56 const char lfsck_bookmark_name[] = "lfsck_bookmark";
  57 const char lfsck_namespace_name[] = "lfsck_namespace";
  58
  59 static const char *lfsck_status_names[] = {
  60         "init",
  61         "scanning-phase1",
  62         "scanning-phase2",
  63         "completed",
  64         "failed",
  65         "stopped",
  66         "paused",
  67         "crashed",
  68         NULL
  69 };
  70
  71 static const char *lfsck_flags_names[] = {
  72         "scanned-once",
  73         "inconsistent",
  74         "upgrade",
  75         NULL
  76 };
  77
  78 static const char *lfsck_param_names[] = {
  79         "failout",
  80         "dryrun",
  81         NULL
  82 };
  83
  84 /* misc functions */
  85
  86 static inline struct mdd_device *mdd_lfsck2mdd(struct md_lfsck *lfsck)
  87 {
  88         return container_of0(lfsck, struct mdd_device, mdd_lfsck);
  89 }
  90
  91 static inline char *mdd_lfsck2name(struct md_lfsck *lfsck)
  92 {
  93         struct mdd_device *mdd = mdd_lfsck2mdd(lfsck);
  94
  95         return mdd2obd_dev(mdd)->obd_name;
  96 }
  97
  98 static inline void mdd_lfsck_component_get(struct lfsck_component *com)
  99 {
 100         atomic_inc(&com->lc_ref);
 101 }
 102
 103 static inline void mdd_lfsck_component_put(const struct lu_env *env,
 104                                            struct lfsck_component *com)
 105 {
 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);
 113                 OBD_FREE_PTR(com);
 114         }
 115 }
 116
 117 static inline struct lfsck_component *
 118 __mdd_lfsck_component_find(struct md_lfsck *lfsck, __u16 type, cfs_list_t *list)
 119 {
 120         struct lfsck_component *com;
 121
 122         cfs_list_for_each_entry(com, list, lc_link) {
 123                 if (com->lc_type == type)
 124                         return com;
 125         }
 126         return NULL;
 127 }
 128
 129 static struct lfsck_component *
 130 mdd_lfsck_component_find(struct md_lfsck *lfsck, __u16 type)
 131 {
 132         struct lfsck_component *com;
 133
 134         spin_lock(&lfsck->ml_lock);
 135         com = __mdd_lfsck_component_find(lfsck, type, &lfsck->ml_list_scan);
 136         if (com != NULL)
 137                 goto unlock;
 138
 139         com = __mdd_lfsck_component_find(lfsck, type,
 140                                          &lfsck->ml_list_double_scan);
 141         if (com != NULL)
 142                 goto unlock;
 143
 144         com = __mdd_lfsck_component_find(lfsck, type, &lfsck->ml_list_idle);
 145
 146 unlock:
 147         if (com != NULL)
 148                 mdd_lfsck_component_get(com);
 149         spin_unlock(&lfsck->ml_lock);
 150         return com;
 151 }
 152
 153 static void mdd_lfsck_component_cleanup(const struct lu_env *env,
 154                                         struct lfsck_component *com)
 155 {
 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);
 160
 161         mdd_lfsck_component_put(env, com);
 162 }
 163
 164 static int lfsck_bits_dump(char **buf, int *len, int bits, const char *names[],
 165                            const char *prefix)
 166 {
 167         int save = *len;
 168         int flag;
 169         int rc;
 170         int i;
 171
 172         rc = snprintf(*buf, *len, "%s:%c", prefix, bits != 0 ? ' ' : '\n');
 173         if (rc <= 0)
 174                 return -ENOSPC;
 175
 176         *buf += rc;
 177         *len -= rc;
 178         for (i = 0, flag = 1; bits != 0; i++, flag = 1 << i) {
 179                 if (flag & bits) {
 180                         bits &= ~flag;
 181                         rc = snprintf(*buf, *len, "%s%c", names[i],
 182                                       bits != 0 ? ',' : '\n');
 183                         if (rc <= 0)
 184                                 return -ENOSPC;
 185
 186                         *buf += rc;
 187                         *len -= rc;
 188                 }
 189         }
 190         return save - *len;
 191 }
 192
 193 static int lfsck_time_dump(char **buf, int *len, __u64 time, const char *prefix)
 194 {
 195         int rc;
 196
 197         if (time != 0)
 198                 rc = snprintf(*buf, *len, "%s: "LPU64" seconds\n", prefix,
 199                               cfs_time_current_sec() - time);
 200         else
 201                 rc = snprintf(*buf, *len, "%s: N/A\n", prefix);
 202         if (rc <= 0)
 203                 return -ENOSPC;
 204
 205         *buf += rc;
 206         *len -= rc;
 207         return rc;
 208 }
 209
 210 static int lfsck_pos_dump(char **buf, int *len, struct lfsck_position *pos,
 211                           const char *prefix)
 212 {
 213         int rc;
 214
 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",
 218                                       prefix);
 219                 else
 220                         rc = snprintf(*buf, *len, "%s: "LPU64", N/A, N/A\n",
 221                                       prefix, pos->lp_oit_cookie);
 222         } else {
 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);
 226         }
 227         if (rc <= 0)
 228                 return -ENOSPC;
 229
 230         *buf += rc;
 231         *len -= rc;
 232         return rc;
 233 }
 234
 235 static void mdd_lfsck_pos_fill(const struct lu_env *env, struct md_lfsck *lfsck,
 236                                struct lfsck_position *pos, bool oit_processed,
 237                                bool dir_processed)
 238 {
 239         const struct dt_it_ops *iops = &lfsck->ml_obj_oit->do_index_ops->dio_it;
 240
 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));
 245                 return;
 246         }
 247
 248         pos->lp_oit_cookie = iops->store(env, lfsck->ml_di_oit);
 249
 250         LASSERT(pos->lp_oit_cookie > 0);
 251
 252         if (!oit_processed)
 253                 pos->lp_oit_cookie--;
 254
 255         if (lfsck->ml_di_dir != NULL) {
 256                 struct dt_object *dto = lfsck->ml_obj_dir;
 257
 258                 pos->lp_dir_parent = *lu_object_fid(&dto->do_lu);
 259                 pos->lp_dir_cookie = dto->do_index_ops->dio_it.store(env,
 260                                                         lfsck->ml_di_dir);
 261
 262                 LASSERT(pos->lp_dir_cookie != MDS_DIR_DUMMY_START);
 263
 264                 if (pos->lp_dir_cookie == MDS_DIR_END_OFF)
 265                         LASSERT(dir_processed);
 266
 267                 /* For the dir which just to be processed,
 268                  * lp_dir_cookie will become MDS_DIR_DUMMY_START,
 269                  * which can be correctly handled by mdd_lfsck_prep. */
 270                 if (!dir_processed)
 271                         pos->lp_dir_cookie--;
 272         } else {
 273                 fid_zero(&pos->lp_dir_parent);
 274                 pos->lp_dir_cookie = 0;
 275         }
 276         spin_unlock(&lfsck->ml_lock);
 277 }
 278
 279 static inline void mdd_lfsck_pos_set_zero(struct lfsck_position *pos)
 280 {
 281         memset(pos, 0, sizeof(*pos));
 282 }
 283
 284 static inline int mdd_lfsck_pos_is_zero(const struct lfsck_position *pos)
 285 {
 286         return pos->lp_oit_cookie == 0 && fid_is_zero(&pos->lp_dir_parent);
 287 }
 288
 289 static inline int mdd_lfsck_pos_is_eq(const struct lfsck_position *pos1,
 290                                       const struct lfsck_position *pos2)
 291 {
 292         if (pos1->lp_oit_cookie < pos2->lp_oit_cookie)
 293                 return -1;
 294
 295         if (pos1->lp_oit_cookie > pos2->lp_oit_cookie)
 296                 return 1;
 297
 298         if (fid_is_zero(&pos1->lp_dir_parent) &&
 299             !fid_is_zero(&pos2->lp_dir_parent))
 300                 return -1;
 301
 302         if (!fid_is_zero(&pos1->lp_dir_parent) &&
 303             fid_is_zero(&pos2->lp_dir_parent))
 304                 return 1;
 305
 306         if (fid_is_zero(&pos1->lp_dir_parent) &&
 307             fid_is_zero(&pos2->lp_dir_parent))
 308                 return 0;
 309
 310         LASSERT(lu_fid_eq(&pos1->lp_dir_parent, &pos2->lp_dir_parent));
 311
 312         if (pos1->lp_dir_cookie < pos2->lp_dir_cookie)
 313                 return -1;
 314
 315         if (pos1->lp_dir_cookie > pos2->lp_dir_cookie)
 316                 return 1;
 317
 318         return 0;
 319 }
 320
 321 static void mdd_lfsck_close_dir(const struct lu_env *env,
 322                                 struct md_lfsck *lfsck)
 323 {
 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;
 327
 328         spin_lock(&lfsck->ml_lock);
 329         lfsck->ml_di_dir = NULL;
 330         spin_unlock(&lfsck->ml_lock);
 331
 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);
 336 }
 337
 338 static void __mdd_lfsck_set_speed(struct md_lfsck *lfsck, __u32 limit)
 339 {
 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;
 345                 } else {
 346                         lfsck->ml_sleep_rate = 1;
 347                         lfsck->ml_sleep_jif = CFS_HZ / limit;
 348                 }
 349         } else {
 350                 lfsck->ml_sleep_jif = 0;
 351                 lfsck->ml_sleep_rate = 0;
 352         }
 353 }
 354
 355 static void mdd_lfsck_control_speed(struct md_lfsck *lfsck)
 356 {
 357         struct ptlrpc_thread *thread = &lfsck->ml_thread;
 358         struct l_wait_info    lwi;
 359
 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);
 368
 369                         l_wait_event(thread->t_ctl_waitq,
 370                                      !thread_is_running(thread),
 371                                      &lwi);
 372                         lfsck->ml_new_scanned = 0;
 373                 } else {
 374                         spin_unlock(&lfsck->ml_lock);
 375                 }
 376         }
 377 }
 378
 379 /* lfsck_bookmark file ops */
 380
 381 static void inline mdd_lfsck_bookmark_to_cpu(struct lfsck_bookmark *des,
 382                                              struct lfsck_bookmark *src)
 383 {
 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);
 388 }
 389
 390 static void inline mdd_lfsck_bookmark_to_le(struct lfsck_bookmark *des,
 391                                             struct lfsck_bookmark *src)
 392 {
 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);
 397 }
 398
 399 static int mdd_lfsck_bookmark_load(const struct lu_env *env,
 400                                    struct md_lfsck *lfsck)
 401 {
 402         loff_t pos = 0;
 403         int    len = sizeof(struct lfsck_bookmark);
 404         int    rc;
 405
 406         rc = dt_record_read(env, lfsck->ml_bookmark_obj,
 407                             mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
 408                             &pos);
 409         if (rc == 0) {
 410                 struct lfsck_bookmark *bm = &lfsck->ml_bookmark_ram;
 411
 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. */
 418                         rc = -ENODATA;
 419                 }
 420         } else {
 421                 if (rc == -EFAULT && pos == 0)
 422                         /* return -ENODATA for empty lfsck_bookmark. */
 423                         rc = -ENODATA;
 424                 else
 425                         CERROR("%.16s: fail to load lfsck_bookmark, "
 426                                "expected = %d, rc = %d\n",
 427                                mdd_lfsck2name(lfsck), len, rc);
 428         }
 429         return rc;
 430 }
 431
 432 static int mdd_lfsck_bookmark_store(const struct lu_env *env,
 433                                     struct md_lfsck *lfsck)
 434 {
 435         struct mdd_device *mdd    = mdd_lfsck2mdd(lfsck);
 436         struct thandle    *handle;
 437         struct dt_object  *obj    = lfsck->ml_bookmark_obj;
 438         loff_t             pos    = 0;
 439         int                len    = sizeof(struct lfsck_bookmark);
 440         int                rc;
 441         ENTRY;
 442
 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);
 450                 RETURN(rc);
 451         }
 452
 453         rc = dt_declare_record_write(env, obj, len, 0, handle);
 454         if (rc != 0) {
 455                 CERROR("%.16s: fail to declare trans for storing "
 456                        "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
 457                 GOTO(out, rc);
 458         }
 459
 460         rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
 461         if (rc != 0) {
 462                 CERROR("%.16s: fail to start trans for storing "
 463                        "lfsck_bookmark: %d\n,", mdd_lfsck2name(lfsck), rc);
 464                 GOTO(out, rc);
 465         }
 466
 467         rc = dt_record_write(env, obj,
 468                              mdd_buf_get(env, &lfsck->ml_bookmark_disk, len),
 469                              &pos, handle);
 470         if (rc != 0)
 471                 CERROR("%.16s: fail to store lfsck_bookmark, expected = %d, "
 472                        "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
 473
 474         GOTO(out, rc);
 475
 476 out:
 477         dt_trans_stop(env, mdd->mdd_bottom, handle);
 478         return rc;
 479 }
 480
 481 static int mdd_lfsck_bookmark_init(const struct lu_env *env,
 482                                    struct md_lfsck *lfsck)
 483 {
 484         struct lfsck_bookmark *mb = &lfsck->ml_bookmark_ram;
 485         int rc;
 486
 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);
 493         return rc;
 494 }
 495
 496 /* lfsck_namespace file ops */
 497
 498 static void inline mdd_lfsck_position_to_cpu(struct lfsck_position *des,
 499                                              struct lfsck_position *src)
 500 {
 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);
 504 }
 505
 506 static void inline mdd_lfsck_position_to_le(struct lfsck_position *des,
 507                                              struct lfsck_position *src)
 508 {
 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);
 512 }
 513
 514 static void inline mdd_lfsck_namespace_to_cpu(struct lfsck_namespace *des,
 515                                               struct lfsck_namespace *src)
 516 {
 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);
 546 }
 547
 548 static void inline mdd_lfsck_namespace_to_le(struct lfsck_namespace *des,
 549                                              struct lfsck_namespace *src)
 550 {
 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);
 580 }
 581
 582 /**
 583  * \retval +ve: the lfsck_namespace is broken, the caller should reset it.
 584  * \retval 0: succeed.
 585  * \retval -ve: failed cases.
 586  */
 587 static int mdd_lfsck_namespace_load(const struct lu_env *env,
 588                                     struct lfsck_component *com)
 589 {
 590         int len = com->lc_file_size;
 591         int rc;
 592
 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);
 596         if (rc == len) {
 597                 struct lfsck_namespace *ns = com->lc_file_ram;
 598
 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 "
 603                               "0x%x != 0x%x\n",
 604                               mdd_lfsck2name(com->lc_lfsck),
 605                               ns->ln_magic, LFSCK_NAMESPACE_MAGIC);
 606                         rc = 1;
 607                 } else {
 608                         rc = 0;
 609                 }
 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);
 613                 if (rc >= 0)
 614                         rc = 1;
 615         }
 616         return rc;
 617 }
 618
 619 static int mdd_lfsck_namespace_store(const struct lu_env *env,
 620                                      struct lfsck_component *com, bool init)
 621 {
 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;
 627         int                rc;
 628         ENTRY;
 629
 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);
 637                 RETURN(rc);
 638         }
 639
 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);
 643         if (rc != 0) {
 644                 CERROR("%.16s: fail to declare trans for storing "
 645                        "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
 646                 GOTO(out, rc);
 647         }
 648
 649         rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
 650         if (rc != 0) {
 651                 CERROR("%.16s: fail to start trans for storing "
 652                        "lfsck_namespace: %d\n,", mdd_lfsck2name(lfsck), rc);
 653                 GOTO(out, rc);
 654         }
 655
 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);
 661         if (rc != 0)
 662                 CERROR("%.16s: fail to store lfsck_namespace, len = %d, "
 663                        "rc = %d\n", mdd_lfsck2name(lfsck), len, rc);
 664
 665         GOTO(out, rc);
 666
 667 out:
 668         dt_trans_stop(env, mdd->mdd_bottom, handle);
 669         return rc;
 670 }
 671
 672 static int mdd_lfsck_namespace_init(const struct lu_env *env,
 673                                     struct lfsck_component *com)
 674 {
 675         struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
 676         int rc;
 677
 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);
 684         return rc;
 685 }
 686
 687 static int mdd_declare_lfsck_namespace_unlink(const struct lu_env *env,
 688                                               struct mdd_device *mdd,
 689                                               struct dt_object *p,
 690                                               struct dt_object *c,
 691                                               const char *name,
 692                                               struct thandle *handle)
 693 {
 694         int rc;
 695
 696         rc = dt_declare_delete(env, p, (const struct dt_key *)name, handle);
 697         if (rc != 0)
 698                 return rc;
 699
 700         rc = dt_declare_ref_del(env, c, handle);
 701         if (rc != 0)
 702                 return rc;
 703
 704         rc = dt_declare_destroy(env, c, handle);
 705         return rc;
 706 }
 707
 708 static int mdd_lfsck_namespace_unlink(const struct lu_env *env,
 709                                       struct mdd_device *mdd,
 710                                       struct lfsck_component *com)
 711 {
 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;
 717         bool                     locked = false;
 718         int                      rc;
 719         ENTRY;
 720
 721         parent = dt_store_resolve(env, mdd->mdd_bottom, "", fid);
 722         if (IS_ERR(parent))
 723                 RETURN(rc = PTR_ERR(parent));
 724
 725         if (!dt_try_as_dir(env, parent))
 726                 GOTO(out, rc = -ENOTDIR);
 727
 728         handle = dt_trans_create(env, mdd->mdd_bottom);
 729         if (IS_ERR(handle))
 730                 GOTO(out, rc = PTR_ERR(handle));
 731
 732         rc = mdd_declare_lfsck_namespace_unlink(env, mdd, parent, child,
 733                                                 lfsck_namespace_name, handle);
 734         if (rc != 0)
 735                 GOTO(stop, rc);
 736
 737         rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
 738         if (rc != 0)
 739                 GOTO(stop, rc);
 740
 741         dt_write_lock(env, child, MOR_TGT_CHILD);
 742         locked = true;
 743         rc = dt_delete(env, parent, (struct dt_key *)lfsck_namespace_name,
 744                        handle, BYPASS_CAPA);
 745         if (rc != 0)
 746                 GOTO(stop, rc);
 747
 748         rc = child->do_ops->do_ref_del(env, child, handle);
 749         if (rc != 0) {
 750                 lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
 751                 rc = dt_insert(env, parent,
 752                                (const struct dt_rec*)fid,
 753                                (const struct dt_key *)lfsck_namespace_name,
 754                                handle, BYPASS_CAPA, 1);
 755
 756                 GOTO(stop, rc);
 757         }
 758
 759
 760         rc = dt_destroy(env, child, handle);
 761
 762         GOTO(stop, rc);
 763
 764 stop:
 765         if (locked)
 766                 dt_write_unlock(env, child);
 767
 768         if (rc == 0) {
 769                 lu_object_put(env, &child->do_lu);
 770                 com->lc_obj = NULL;
 771         }
 772
 773         dt_trans_stop(env, mdd->mdd_bottom, handle);
 774
 775 out:
 776         lu_object_put(env, &parent->do_lu);
 777         return rc;
 778 }
 779
 780 static int mdd_lfsck_namespace_lookup(const struct lu_env *env,
 781                                       struct lfsck_component *com,
 782                                       const struct lu_fid *fid,
 783                                       __u8 *flags)
 784 {
 785         struct lu_fid *key = &mdd_env_info(env)->mti_fid;
 786         int            rc;
 787
 788         fid_cpu_to_be(key, fid);
 789         rc = dt_lookup(env, com->lc_obj, (struct dt_rec *)flags,
 790                        (const struct dt_key *)key, BYPASS_CAPA);
 791         return rc;
 792 }
 793
 794 static int mdd_lfsck_namespace_delete(const struct lu_env *env,
 795                                       struct lfsck_component *com,
 796                                       const struct lu_fid *fid)
 797 {
 798         struct mdd_device *mdd    = mdd_lfsck2mdd(com->lc_lfsck);
 799         struct lu_fid     *key    = &mdd_env_info(env)->mti_fid;
 800         struct thandle    *handle;
 801         struct dt_object *obj     = com->lc_obj;
 802         int               rc;
 803         ENTRY;
 804
 805         handle = dt_trans_create(env, mdd->mdd_bottom);
 806         if (IS_ERR(handle))
 807                 RETURN(PTR_ERR(handle));
 808
 809         rc = dt_declare_delete(env, obj, (const struct dt_key *)fid, handle);
 810         if (rc != 0)
 811                 GOTO(out, rc);
 812
 813         rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
 814         if (rc != 0)
 815                 GOTO(out, rc);
 816
 817         fid_cpu_to_be(key, fid);
 818         rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
 819                        BYPASS_CAPA);
 820
 821         GOTO(out, rc);
 822
 823 out:
 824         dt_trans_stop(env, mdd->mdd_bottom, handle);
 825         return rc;
 826 }
 827
 828 static int mdd_lfsck_namespace_update(const struct lu_env *env,
 829                                       struct lfsck_component *com,
 830                                       const struct lu_fid *fid,
 831                                       __u8 flags, bool force)
 832 {
 833         struct mdd_device *mdd    = mdd_lfsck2mdd(com->lc_lfsck);
 834         struct lu_fid     *key    = &mdd_env_info(env)->mti_fid;
 835         struct thandle    *handle;
 836         struct dt_object *obj     = com->lc_obj;
 837         int               rc;
 838         bool              exist   = false;
 839         __u8              tf;
 840         ENTRY;
 841
 842         rc = mdd_lfsck_namespace_lookup(env, com, fid, &tf);
 843         if (rc != 0 && rc != -ENOENT)
 844                 RETURN(rc);
 845
 846         if (rc == 0) {
 847                 if (!force || flags == tf)
 848                         RETURN(0);
 849
 850                 exist = true;
 851                 handle = dt_trans_create(env, mdd->mdd_bottom);
 852                 if (IS_ERR(handle))
 853                         RETURN(PTR_ERR(handle));
 854
 855                 rc = dt_declare_delete(env, obj, (const struct dt_key *)fid,
 856                                        handle);
 857                 if (rc != 0)
 858                         GOTO(out, rc);
 859         } else {
 860                 handle = dt_trans_create(env, mdd->mdd_bottom);
 861                 if (IS_ERR(handle))
 862                         RETURN(PTR_ERR(handle));
 863         }
 864
 865         rc = dt_declare_insert(env, obj, (const struct dt_rec *)&flags,
 866                                (const struct dt_key *)fid, handle);
 867         if (rc != 0)
 868                 GOTO(out, rc);
 869
 870         rc = dt_trans_start_local(env, mdd->mdd_bottom, handle);
 871         if (rc != 0)
 872                 GOTO(out, rc);
 873
 874         fid_cpu_to_be(key, fid);
 875         if (exist) {
 876                 rc = dt_delete(env, obj, (const struct dt_key *)key, handle,
 877                                BYPASS_CAPA);
 878                 if (rc != 0) {
 879                         CERROR("%s: fail to insert "DFID", rc = %d\n",
 880                                mdd_lfsck2name(com->lc_lfsck), PFID(fid), rc);
 881                         GOTO(out, rc);
 882                 }
 883         }
 884
 885         rc = dt_insert(env, obj, (const struct dt_rec *)&flags,
 886                        (const struct dt_key *)key, handle, BYPASS_CAPA, 1);
 887
 888         GOTO(out, rc);
 889
 890 out:
 891         dt_trans_stop(env, mdd->mdd_bottom, handle);
 892         return rc;
 893 }
 894
 895 /**
 896  * \retval +ve  repaired
 897  * \retval 0    no need to repair
 898  * \retval -ve  error cases
 899  */
 900 static int mdd_lfsck_namespace_double_scan_one(const struct lu_env *env,
 901                                                struct lfsck_component *com,
 902                                                struct mdd_object *child,
 903                                                __u8 flags)
 904 {
 905         struct mdd_thread_info  *info     = mdd_env_info(env);
 906         struct lu_attr          *la       = &info->mti_la;
 907         struct lu_name          *cname    = &info->mti_name;
 908         struct lu_fid           *pfid     = &info->mti_fid;
 909         struct lu_fid           *cfid     = &info->mti_fid2;
 910         struct md_lfsck         *lfsck    = com->lc_lfsck;
 911         struct mdd_device       *mdd      = mdd_lfsck2mdd(lfsck);
 912         struct lfsck_bookmark   *bk       = &lfsck->ml_bookmark_ram;
 913         struct lfsck_namespace  *ns       =
 914                                 (struct lfsck_namespace *)com->lc_file_ram;
 915         struct mdd_link_data     ldata    = { 0 };
 916         struct thandle          *handle   = NULL;
 917         bool                     locked   = false;
 918         bool                     update   = false;
 919         int                      count;
 920         int                      rc;
 921         ENTRY;
 922
 923         if (com->lc_journal) {
 924
 925 again:
 926                 LASSERT(!locked);
 927
 928                 com->lc_journal = 1;
 929                 handle = mdd_trans_create(env, mdd);
 930                 if (IS_ERR(handle))
 931                         RETURN(rc = PTR_ERR(handle));
 932
 933                 rc = mdd_declare_links_add(env, child, handle);;
 934                 if (rc != 0)
 935                         GOTO(stop, rc);
 936
 937                 rc = mdd_trans_start(env, mdd, handle);
 938                 if (rc != 0)
 939                         GOTO(stop, rc);
 940
 941                 mdd_write_lock(env, child, MOR_TGT_CHILD);
 942                 locked = true;
 943         }
 944
 945         if (unlikely(mdd_is_dead_obj(child)))
 946                 GOTO(stop, rc = 0);
 947
 948         rc = mdd_links_read(env, child, &ldata);
 949         if (rc != 0) {
 950                 if ((bk->lb_param & LPF_DRYRUN) &&
 951                     (rc == -EINVAL || rc == -ENODATA))
 952                         rc = 1;
 953
 954                 GOTO(stop, rc);
 955         }
 956
 957         rc = mdd_la_get(env, child, la, BYPASS_CAPA);
 958         if (rc != 0)
 959                 GOTO(stop, rc);
 960
 961         ldata.ml_lee = (struct link_ea_entry *)(ldata.ml_leh + 1);
 962         count = ldata.ml_leh->leh_reccount;
 963         while (count-- > 0) {
 964                 struct mdd_object *parent = NULL;
 965                 struct dt_object *dir;
 966
 967                 mdd_lee_unpack(ldata.ml_lee, &ldata.ml_reclen, cname, pfid);
 968                 if (!fid_is_sane(pfid))
 969                         goto shrink;
 970
 971                 parent = mdd_object_find(env, mdd, pfid);
 972                 if (parent == NULL)
 973                         goto shrink;
 974                 else if (IS_ERR(parent))
 975                         GOTO(stop, rc = PTR_ERR(parent));
 976
 977                 if (!mdd_object_exists(parent))
 978                         goto shrink;
 979
 980                 /* XXX: need more processing for remote object in the future. */
 981                 if (mdd_object_remote(parent)) {
 982                         mdd_object_put(env, parent);
 983                         ldata.ml_lee = (struct link_ea_entry *)
 984                                        ((char *)ldata.ml_lee + ldata.ml_reclen);
 985                         continue;
 986                 }
 987
 988                 dir = mdd_object_child(parent);
 989                 if (unlikely(!dt_try_as_dir(env, dir)))
 990                         goto shrink;
 991
 992                 /* To guarantee the 'name' is terminated with '0'. */
 993                 memcpy(info->mti_key, cname->ln_name, cname->ln_namelen);
 994                 info->mti_key[cname->ln_namelen] = 0;
 995                 cname->ln_name = info->mti_key;
 996                 rc = dt_lookup(env, dir, (struct dt_rec *)cfid,
 997                                (const struct dt_key *)cname->ln_name,
 998                                BYPASS_CAPA);
 999                 if (rc != 0 && rc != -ENOENT) {
1000                         mdd_object_put(env, parent);
1001                         GOTO(stop, rc);
1002                 }
1003
1004                 if (rc == 0) {
1005                         if (lu_fid_eq(cfid, mdo2fid(child))) {
1006                                 mdd_object_put(env, parent);
1007                                 ldata.ml_lee = (struct link_ea_entry *)
1008                                        ((char *)ldata.ml_lee + ldata.ml_reclen);
1009                                 continue;
1010                         }
1011
1012                         goto shrink;
1013                 }
1014
1015                 if (ldata.ml_leh->leh_reccount > la->la_nlink)
1016                         goto shrink;
1017
1018                 /* XXX: For the case of there is linkea entry, but without name
1019                  *      entry pointing to the object, and the object link count
1020                  *      isn't less than the count of name entries, then add the
1021                  *      name entry back to namespace.
1022                  *
1023                  *      It is out of LFSCK 1.5 scope, will implement it in the
1024                  *      future. Keep the linkEA entry. */
1025                 mdd_object_put(env, parent);
1026                 ldata.ml_lee = (struct link_ea_entry *)
1027                                ((char *)ldata.ml_lee + ldata.ml_reclen);
1028                 continue;
1029
1030 shrink:
1031                 if (parent != NULL)
1032                         mdd_object_put(env, parent);
1033                 if (bk->lb_param & LPF_DRYRUN)
1034                         RETURN(1);
1035
1036                 CDEBUG(D_LFSCK, "Remove linkEA: "DFID"[%.*s], "DFID"\n",
1037                        PFID(mdo2fid(child)), cname->ln_namelen, cname->ln_name,
1038                        PFID(pfid));
1039                 mdd_links_del_buf(env, &ldata, cname);
1040                 update = true;
1041         }
1042
1043         if (update) {
1044                 if (!com->lc_journal) {
1045                         com->lc_journal = 1;
1046                         goto again;
1047                 }
1048
1049                 rc = mdd_links_write(env, child, &ldata, handle);
1050         }
1051
1052         GOTO(stop, rc);
1053
1054 stop:
1055         if (locked)
1056                 mdd_write_unlock(env, child);
1057
1058         if (handle != NULL)
1059                 mdd_trans_stop(env, mdd, rc, handle);
1060
1061         if (rc == 0 && update) {
1062                 ns->ln_objs_nlink_repaired++;
1063                 rc = 1;
1064         }
1065         return rc;
1066 }
1067
1068 /* namespace APIs */
1069
1070 static int mdd_lfsck_namespace_reset(const struct lu_env *env,
1071                                      struct lfsck_component *com, bool init)
1072 {
1073         struct mdd_thread_info  *info = mdd_env_info(env);
1074         struct lu_fid           *fid  = &info->mti_fid;
1075         struct lfsck_namespace  *ns   = (struct lfsck_namespace *)com->lc_file_ram;
1076         struct mdd_device       *mdd  = mdd_lfsck2mdd(com->lc_lfsck);
1077         struct md_object        *mdo;
1078         struct dt_object        *dto;
1079         int                      rc;
1080         ENTRY;
1081
1082         down_write(&com->lc_sem);
1083         if (init) {
1084                 memset(ns, 0, sizeof(*ns));
1085         } else {
1086                 __u32 count = ns->ln_success_count;
1087                 __u64 last_time = ns->ln_time_last_complete;
1088
1089                 memset(ns, 0, sizeof(*ns));
1090                 ns->ln_success_count = count;
1091                 ns->ln_time_last_complete = last_time;
1092         }
1093         ns->ln_magic = LFSCK_NAMESPACE_MAGIC;
1094         ns->ln_status = LS_INIT;
1095
1096         rc = mdd_lfsck_namespace_unlink(env, mdd, com);
1097         if (rc != 0)
1098                 GOTO(out, rc);
1099
1100         lu_local_obj_fid(fid, LFSCK_NAMESPACE_OID);
1101         mdo = llo_store_create_index(env, &mdd->mdd_md_dev, mdd->mdd_bottom, "",
1102                                      lfsck_namespace_name, fid,
1103                                      &dt_lfsck_features);
1104         if (IS_ERR(mdo))
1105                 GOTO(out, rc = PTR_ERR(mdo));
1106
1107         lu_object_put(env, &mdo->mo_lu);
1108         dto = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name, fid);
1109         if (IS_ERR(dto))
1110                 GOTO(out, rc = PTR_ERR(dto));
1111
1112         com->lc_obj = dto;
1113         rc = dto->do_ops->do_index_try(env, dto, &dt_lfsck_features);
1114         if (rc != 0)
1115                 GOTO(out, rc);
1116
1117         rc = mdd_lfsck_namespace_store(env, com, true);
1118
1119         GOTO(out, rc);
1120
1121 out:
1122         up_write(&com->lc_sem);
1123         return rc;
1124 }
1125
1126 static void
1127 mdd_lfsck_namespace_fail(const struct lu_env *env, struct lfsck_component *com,
1128                          bool oit, bool new_checked)
1129 {
1130         struct lfsck_namespace *ns = (struct lfsck_namespace *)com->lc_file_ram;
1131
1132         down_write(&com->lc_sem);
1133         if (new_checked)
1134                 com->lc_new_checked++;
1135         ns->ln_items_failed++;
1136         if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1137                 mdd_lfsck_pos_fill(env, com->lc_lfsck,
1138                                    &ns->ln_pos_first_inconsistent, oit, !oit);
1139         up_write(&com->lc_sem);
1140 }
1141
1142 static int mdd_lfsck_namespace_checkpoint(const struct lu_env *env,
1143                                           struct lfsck_component *com,
1144                                           bool init)
1145 {
1146         struct md_lfsck         *lfsck = com->lc_lfsck;
1147         struct lfsck_namespace  *ns    =
1148                                 (struct lfsck_namespace *)com->lc_file_ram;
1149         int                      rc;
1150
1151         if (com->lc_new_checked == 0 && !init)
1152                 return 0;
1153
1154         down_write(&com->lc_sem);
1155
1156         ns->ln_pos_last_checkpoint = lfsck->ml_pos_current;
1157         if (init) {
1158                 ns->ln_time_last_checkpoint = ns->ln_time_latest_start;
1159                 ns->ln_pos_latest_start = lfsck->ml_pos_current;
1160         } else {
1161                 ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1162                                 HALF_SEC - lfsck->ml_time_last_checkpoint);
1163                 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1164                 ns->ln_items_checked += com->lc_new_checked;
1165                 com->lc_new_checked = 0;
1166         }
1167
1168         rc = mdd_lfsck_namespace_store(env, com, false);
1169
1170         up_write(&com->lc_sem);
1171         return rc;
1172 }
1173
1174 static int mdd_lfsck_namespace_prep(const struct lu_env *env,
1175                                     struct lfsck_component *com)
1176 {
1177         struct md_lfsck         *lfsck  = com->lc_lfsck;
1178         struct lfsck_namespace  *ns     =
1179                                 (struct lfsck_namespace *)com->lc_file_ram;
1180         struct lfsck_position   *pos    = &com->lc_pos_start;
1181
1182         if (ns->ln_status == LS_COMPLETED) {
1183                 int rc;
1184
1185                 rc = mdd_lfsck_namespace_reset(env, com, false);
1186                 if (rc != 0)
1187                         return rc;
1188         }
1189
1190         down_write(&com->lc_sem);
1191
1192         ns->ln_time_latest_start = cfs_time_current_sec();
1193
1194         spin_lock(&lfsck->ml_lock);
1195         if (ns->ln_flags & LF_SCANNED_ONCE) {
1196                 if (!lfsck->ml_drop_dryrun ||
1197                     mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1198                         ns->ln_status = LS_SCANNING_PHASE2;
1199                         cfs_list_del_init(&com->lc_link);
1200                         cfs_list_add_tail(&com->lc_link,
1201                                           &lfsck->ml_list_double_scan);
1202                         if (!cfs_list_empty(&com->lc_link_dir))
1203                                 cfs_list_del_init(&com->lc_link_dir);
1204                         mdd_lfsck_pos_set_zero(pos);
1205                 } else {
1206                         ns->ln_status = LS_SCANNING_PHASE1;
1207                         ns->ln_run_time_phase1 = 0;
1208                         ns->ln_run_time_phase2 = 0;
1209                         ns->ln_items_checked = 0;
1210                         ns->ln_items_repaired = 0;
1211                         ns->ln_items_failed = 0;
1212                         ns->ln_dirs_checked = 0;
1213                         ns->ln_mlinked_checked = 0;
1214                         ns->ln_objs_checked_phase2 = 0;
1215                         ns->ln_objs_repaired_phase2 = 0;
1216                         ns->ln_objs_failed_phase2 = 0;
1217                         ns->ln_objs_nlink_repaired = 0;
1218                         ns->ln_objs_lost_found = 0;
1219                         fid_zero(&ns->ln_fid_latest_scanned_phase2);
1220                         if (cfs_list_empty(&com->lc_link_dir))
1221                                 cfs_list_add_tail(&com->lc_link_dir,
1222                                                   &lfsck->ml_list_dir);
1223                         *pos = ns->ln_pos_first_inconsistent;
1224                 }
1225         } else {
1226                 ns->ln_status = LS_SCANNING_PHASE1;
1227                 if (cfs_list_empty(&com->lc_link_dir))
1228                         cfs_list_add_tail(&com->lc_link_dir,
1229                                           &lfsck->ml_list_dir);
1230                 if (!lfsck->ml_drop_dryrun ||
1231                     mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent)) {
1232                         *pos = ns->ln_pos_last_checkpoint;
1233                         pos->lp_oit_cookie++;
1234                         if (!fid_is_zero(&pos->lp_dir_parent)) {
1235                                 if (pos->lp_dir_cookie == MDS_DIR_END_OFF) {
1236                                         fid_zero(&pos->lp_dir_parent);
1237                                 } else {
1238                                         pos->lp_dir_cookie++;
1239                                 }
1240                         }
1241                 } else {
1242                         *pos = ns->ln_pos_first_inconsistent;
1243                 }
1244         }
1245         spin_unlock(&lfsck->ml_lock);
1246
1247         up_write(&com->lc_sem);
1248         return 0;
1249 }
1250
1251 static int mdd_lfsck_namespace_exec_oit(const struct lu_env *env,
1252                                         struct lfsck_component *com,
1253                                         struct mdd_object *obj)
1254 {
1255         down_write(&com->lc_sem);
1256         com->lc_new_checked++;
1257         if (S_ISDIR(mdd_object_type(obj)))
1258                 ((struct lfsck_namespace *)com->lc_file_ram)->ln_dirs_checked++;
1259         up_write(&com->lc_sem);
1260         return 0;
1261 }
1262
1263 static int mdd_declare_lfsck_namespace_exec_dir(const struct lu_env *env,
1264                                                 struct mdd_object *obj,
1265                                                 struct thandle *handle)
1266 {
1267         int rc;
1268
1269         /* For destroying all invalid linkEA entries. */
1270         rc = mdo_declare_xattr_del(env, obj, XATTR_NAME_LINK, handle);
1271         if (rc != 0)
1272                 return rc;
1273
1274         /* For insert new linkEA entry. */
1275         rc = mdd_declare_links_add(env, obj, handle);
1276         return rc;
1277 }
1278
1279 static int mdd_lfsck_namespace_check_exist(const struct lu_env *env,
1280                                            struct md_lfsck *lfsck,
1281                                            struct mdd_object *obj,
1282                                            const char *name)
1283 {
1284         struct dt_object *dir = lfsck->ml_obj_dir;
1285         struct lu_fid    *fid = &mdd_env_info(env)->mti_fid;
1286         int               rc;
1287         ENTRY;
1288
1289         if (unlikely(mdd_is_dead_obj(obj)))
1290                 RETURN(LFSCK_NAMEENTRY_DEAD);
1291
1292         rc = dt_lookup(env, dir, (struct dt_rec *)fid,
1293                        (const struct dt_key *)name, BYPASS_CAPA);
1294         if (rc == -ENOENT)
1295                 RETURN(LFSCK_NAMEENTRY_REMOVED);
1296
1297         if (rc < 0)
1298                 RETURN(rc);
1299
1300         if (!lu_fid_eq(fid, mdo2fid(obj)))
1301                 RETURN(LFSCK_NAMEENTRY_RECREATED);
1302
1303         RETURN(0);
1304 }
1305
1306 static int mdd_lfsck_namespace_exec_dir(const struct lu_env *env,
1307                                         struct lfsck_component *com,
1308                                         struct mdd_object *obj,
1309                                         struct lu_dirent *ent)
1310 {
1311         struct mdd_thread_info     *info     = mdd_env_info(env);
1312         struct lu_attr             *la       = &info->mti_la;
1313         struct md_lfsck            *lfsck    = com->lc_lfsck;
1314         struct lfsck_bookmark      *bk       = &lfsck->ml_bookmark_ram;
1315         struct lfsck_namespace     *ns       =
1316                                 (struct lfsck_namespace *)com->lc_file_ram;
1317         struct mdd_device          *mdd      = mdd_lfsck2mdd(lfsck);
1318         struct mdd_link_data        ldata    = { 0 };
1319         const struct lu_fid        *pfid     =
1320                                 lu_object_fid(&lfsck->ml_obj_dir->do_lu);
1321         const struct lu_fid        *cfid     = mdo2fid(obj);
1322         const struct lu_name       *cname;
1323         struct thandle             *handle   = NULL;
1324         bool                        repaired = false;
1325         bool                        locked   = false;
1326         int                         count    = 0;
1327         int                         rc;
1328         ENTRY;
1329
1330         cname = mdd_name_get_const(env, ent->lde_name, ent->lde_namelen);
1331         down_write(&com->lc_sem);
1332         com->lc_new_checked++;
1333
1334         if (ent->lde_attrs & LUDA_UPGRADE) {
1335                 ns->ln_flags |= LF_UPGRADE;
1336                 repaired = true;
1337         } else if (ent->lde_attrs & LUDA_REPAIR) {
1338                 ns->ln_flags |= LF_INCONSISTENT;
1339                 repaired = true;
1340         }
1341
1342         if (ent->lde_name[0] == '.' &&
1343             (ent->lde_namelen == 1 ||
1344              (ent->lde_namelen == 2 && ent->lde_name[1] == '.')))
1345                 GOTO(out, rc = 0);
1346
1347         if (!(bk->lb_param & LPF_DRYRUN) &&
1348             (com->lc_journal || repaired)) {
1349
1350 again:
1351                 LASSERT(!locked);
1352
1353                 com->lc_journal = 1;
1354                 handle = mdd_trans_create(env, mdd);
1355                 if (IS_ERR(handle))
1356                         GOTO(out, rc = PTR_ERR(handle));
1357
1358                 rc = mdd_declare_lfsck_namespace_exec_dir(env, obj, handle);
1359                 if (rc != 0)
1360                         GOTO(stop, rc);
1361
1362                 rc = mdd_trans_start(env, mdd, handle);
1363                 if (rc != 0)
1364                         GOTO(stop, rc);
1365
1366                 mdd_write_lock(env, obj, MOR_TGT_CHILD);
1367                 locked = true;
1368         }
1369
1370         rc = mdd_lfsck_namespace_check_exist(env, lfsck, obj, ent->lde_name);
1371         if (rc != 0)
1372                 GOTO(stop, rc);
1373
1374         rc = mdd_links_read(env, obj, &ldata);
1375         if (rc == 0) {
1376                 count = ldata.ml_leh->leh_reccount;
1377                 rc = mdd_links_find(env, obj, &ldata, cname, pfid);
1378                 if (rc == 0) {
1379                         /* For dir, if there are more than one linkea entries,
1380                          * then remove all the other redundant linkea entries.*/
1381                         if (unlikely(count > 1 &&
1382                                      S_ISDIR(mdd_object_type(obj))))
1383                                 goto unmatch;
1384
1385                         goto record;
1386                 } else {
1387
1388 unmatch:
1389                         ns->ln_flags |= LF_INCONSISTENT;
1390                         if (bk->lb_param & LPF_DRYRUN) {
1391                                 repaired = true;
1392                                 goto record;
1393                         }
1394
1395                         /*For dir, remove the unmatched linkea entry directly.*/
1396                         if (S_ISDIR(mdd_object_type(obj))) {
1397                                 if (!com->lc_journal)
1398                                         goto again;
1399
1400                                 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK,
1401                                                    handle, BYPASS_CAPA);
1402                                 if (rc != 0)
1403                                         GOTO(stop, rc);
1404
1405                                 goto nodata;
1406                         } else {
1407                                 goto add;
1408                         }
1409                 }
1410         } else if (unlikely(rc == -EINVAL)) {
1411                 ns->ln_flags |= LF_INCONSISTENT;
1412                 if (bk->lb_param & LPF_DRYRUN) {
1413                         count = 1;
1414                         repaired = true;
1415                         goto record;
1416                 }
1417
1418                 if (!com->lc_journal)
1419                         goto again;
1420
1421                 /* The magic crashed, we are not sure whether there are more
1422                  * corrupt data in the linkea, so remove all linkea entries. */
1423                 rc = mdo_xattr_del(env, obj, XATTR_NAME_LINK, handle,
1424                                    BYPASS_CAPA);
1425                 if (rc != 0)
1426                         GOTO(stop, rc);
1427
1428                 goto nodata;
1429         } else if (rc == -ENODATA) {
1430                 ns->ln_flags |= LF_UPGRADE;
1431                 if (bk->lb_param & LPF_DRYRUN) {
1432                         count = 1;
1433                         repaired = true;
1434                         goto record;
1435                 }
1436
1437 nodata:
1438                 rc = mdd_links_new(env, &ldata);
1439                 if (rc != 0)
1440                         GOTO(stop, rc);
1441
1442 add:
1443                 if (!com->lc_journal)
1444                         goto again;
1445
1446                 rc = mdd_links_add_buf(env, &ldata, cname, pfid);
1447                 if (rc != 0)
1448                         GOTO(stop, rc);
1449
1450                 rc = mdd_links_write(env, obj, &ldata, handle);
1451                 if (rc != 0)
1452                         GOTO(stop, rc);
1453
1454                 count = ldata.ml_leh->leh_reccount;
1455                 repaired = true;
1456         } else {
1457                 GOTO(stop, rc);
1458         }
1459
1460 record:
1461         LASSERT(count > 0);
1462
1463         rc = mdd_la_get(env, obj, la, BYPASS_CAPA);
1464         if (rc != 0)
1465                 GOTO(stop, rc);
1466
1467         if ((count == 1) &&
1468             (la->la_nlink == 1 || S_ISDIR(mdd_object_type(obj))))
1469                 /* Usually, it is for single linked object or dir, do nothing.*/
1470                 GOTO(stop, rc);
1471
1472         /* Following modification will be in another transaction.  */
1473         if (handle != NULL) {
1474                 LASSERT(mdd_write_locked(env, obj));
1475
1476                 mdd_write_unlock(env, obj);
1477                 locked = false;
1478
1479                 mdd_trans_stop(env, mdd, 0, handle);
1480                 handle = NULL;
1481         }
1482
1483         ns->ln_mlinked_checked++;
1484         rc = mdd_lfsck_namespace_update(env, com, cfid,
1485                         count != la->la_nlink ? LLF_UNMATCH_NLINKS : 0, false);
1486
1487         GOTO(out, rc);
1488
1489 stop:
1490         if (locked)
1491                 mdd_write_unlock(env, obj);
1492
1493         if (handle != NULL)
1494                 mdd_trans_stop(env, mdd, rc, handle);
1495
1496 out:
1497         if (rc < 0) {
1498                 ns->ln_items_failed++;
1499                 if (mdd_lfsck_pos_is_zero(&ns->ln_pos_first_inconsistent))
1500                         mdd_lfsck_pos_fill(env, lfsck,
1501                                            &ns->ln_pos_first_inconsistent,
1502                                            true, false);
1503                 if (!(bk->lb_param & LPF_FAILOUT))
1504                         rc = 0;
1505         } else {
1506                 if (repaired)
1507                         ns->ln_items_repaired++;
1508                 else
1509                         com->lc_journal = 0;
1510                 rc = 0;
1511         }
1512         up_write(&com->lc_sem);
1513         return rc;
1514 }
1515
1516 static int mdd_lfsck_namespace_post(const struct lu_env *env,
1517                                     struct lfsck_component *com,
1518                                     int result)
1519 {
1520         struct md_lfsck         *lfsck = com->lc_lfsck;
1521         struct lfsck_namespace  *ns    =
1522                                 (struct lfsck_namespace *)com->lc_file_ram;
1523         int                      rc;
1524
1525         down_write(&com->lc_sem);
1526
1527         spin_lock(&lfsck->ml_lock);
1528         if (result > 0) {
1529                 ns->ln_status = LS_SCANNING_PHASE2;
1530                 ns->ln_flags |= LF_SCANNED_ONCE;
1531                 ns->ln_flags &= ~LF_UPGRADE;
1532                 cfs_list_del_init(&com->lc_link);
1533                 cfs_list_del_init(&com->lc_link_dir);
1534                 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_double_scan);
1535         } else if (result == 0) {
1536                 if (lfsck->ml_paused) {
1537                         ns->ln_status = LS_PAUSED;
1538                 } else {
1539                         ns->ln_status = LS_STOPPED;
1540                         cfs_list_del_init(&com->lc_link);
1541                         cfs_list_del_init(&com->lc_link_dir);
1542                         cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1543                 }
1544         } else {
1545                 ns->ln_status = LS_FAILED;
1546                 cfs_list_del_init(&com->lc_link);
1547                 cfs_list_del_init(&com->lc_link_dir);
1548                 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1549         }
1550         spin_unlock(&lfsck->ml_lock);
1551
1552         ns->ln_run_time_phase1 += cfs_duration_sec(cfs_time_current() +
1553                                 HALF_SEC - lfsck->ml_time_last_checkpoint);
1554         ns->ln_time_last_checkpoint = cfs_time_current_sec();
1555         ns->ln_items_checked += com->lc_new_checked;
1556         com->lc_new_checked = 0;
1557
1558         rc = mdd_lfsck_namespace_store(env, com, false);
1559
1560         up_write(&com->lc_sem);
1561         return rc;
1562 }
1563
1564 static int
1565 mdd_lfsck_namespace_dump(const struct lu_env *env, struct lfsck_component *com,
1566                          char *buf, int len)
1567 {
1568         struct md_lfsck         *lfsck = com->lc_lfsck;
1569         struct lfsck_bookmark   *bk    = &lfsck->ml_bookmark_ram;
1570         struct lfsck_namespace  *ns    =
1571                                 (struct lfsck_namespace *)com->lc_file_ram;
1572         int                      save  = len;
1573         int                      ret   = -ENOSPC;
1574         int                      rc;
1575
1576         down_read(&com->lc_sem);
1577         rc = snprintf(buf, len,
1578                       "name: lfsck_namespace\n"
1579                       "magic: 0x%x\n"
1580                       "version: %d\n"
1581                       "status: %s\n",
1582                       ns->ln_magic,
1583                       bk->lb_version,
1584                       lfsck_status_names[ns->ln_status]);
1585         if (rc <= 0)
1586                 goto out;
1587
1588         buf += rc;
1589         len -= rc;
1590         rc = lfsck_bits_dump(&buf, &len, ns->ln_flags, lfsck_flags_names,
1591                              "flags");
1592         if (rc < 0)
1593                 goto out;
1594
1595         rc = lfsck_bits_dump(&buf, &len, bk->lb_param, lfsck_param_names,
1596                              "param");
1597         if (rc < 0)
1598                 goto out;
1599
1600         rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_complete,
1601                              "time_since_last_completed");
1602         if (rc < 0)
1603                 goto out;
1604
1605         rc = lfsck_time_dump(&buf, &len, ns->ln_time_latest_start,
1606                              "time_since_latest_start");
1607         if (rc < 0)
1608                 goto out;
1609
1610         rc = lfsck_time_dump(&buf, &len, ns->ln_time_last_checkpoint,
1611                              "time_since_last_checkpoint");
1612         if (rc < 0)
1613                 goto out;
1614
1615         rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_latest_start,
1616                             "latest_start_position");
1617         if (rc < 0)
1618                 goto out;
1619
1620         rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_last_checkpoint,
1621                             "last_checkpoint_position");
1622         if (rc < 0)
1623                 goto out;
1624
1625         rc = lfsck_pos_dump(&buf, &len, &ns->ln_pos_first_inconsistent,
1626                             "first_failure_position");
1627         if (rc < 0)
1628                 goto out;
1629
1630         if (ns->ln_status == LS_SCANNING_PHASE1) {
1631                 struct lfsck_position pos;
1632                 cfs_duration_t duration = cfs_time_current() -
1633                                           lfsck->ml_time_last_checkpoint;
1634                 __u64 checked = ns->ln_items_checked + com->lc_new_checked;
1635                 __u64 speed = checked;
1636                 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1637                 __u32 rtime = ns->ln_run_time_phase1 +
1638                               cfs_duration_sec(duration + HALF_SEC);
1639
1640                 if (duration != 0)
1641                         do_div(new_checked, duration);
1642                 if (rtime != 0)
1643                         do_div(speed, rtime);
1644                 rc = snprintf(buf, len,
1645                               "checked_phase1: "LPU64"\n"
1646                               "checked_phase2: "LPU64"\n"
1647                               "updated_phase1: "LPU64"\n"
1648                               "updated_phase2: "LPU64"\n"
1649                               "failed_phase1: "LPU64"\n"
1650                               "failed_phase2: "LPU64"\n"
1651                               "dirs: "LPU64"\n"
1652                               "M-linked: "LPU64"\n"
1653                               "nlinks_repaired: "LPU64"\n"
1654                               "lost_found: "LPU64"\n"
1655                               "success_count: %u\n"
1656                               "run_time_phase1: %u seconds\n"
1657                               "run_time_phase2: %u seconds\n"
1658                               "average_speed_phase1: "LPU64" items/sec\n"
1659                               "average_speed_phase2: N/A\n"
1660                               "real-time_speed_phase1: "LPU64" items/sec\n"
1661                               "real-time_speed_phase2: N/A\n",
1662                               checked,
1663                               ns->ln_objs_checked_phase2,
1664                               ns->ln_items_repaired,
1665                               ns->ln_objs_repaired_phase2,
1666                               ns->ln_items_failed,
1667                               ns->ln_objs_failed_phase2,
1668                               ns->ln_dirs_checked,
1669                               ns->ln_mlinked_checked,
1670                               ns->ln_objs_nlink_repaired,
1671                               ns->ln_objs_lost_found,
1672                               ns->ln_success_count,
1673                               rtime,
1674                               ns->ln_run_time_phase2,
1675                               speed,
1676                               new_checked);
1677                 if (rc <= 0)
1678                         goto out;
1679
1680                 buf += rc;
1681                 len -= rc;
1682                 mdd_lfsck_pos_fill(env, lfsck, &pos, true, true);
1683                 rc = lfsck_pos_dump(&buf, &len, &pos, "current_position");
1684                 if (rc <= 0)
1685                         goto out;
1686         } else if (ns->ln_status == LS_SCANNING_PHASE2) {
1687                 cfs_duration_t duration = cfs_time_current() -
1688                                           lfsck->ml_time_last_checkpoint;
1689                 __u64 checked = ns->ln_objs_checked_phase2 +
1690                                 com->lc_new_checked;
1691                 __u64 speed1 = ns->ln_items_checked;
1692                 __u64 speed2 = checked;
1693                 __u64 new_checked = com->lc_new_checked * CFS_HZ;
1694                 __u32 rtime = ns->ln_run_time_phase2 +
1695                               cfs_duration_sec(duration + HALF_SEC);
1696
1697                 if (duration != 0)
1698                         do_div(new_checked, duration);
1699                 if (ns->ln_run_time_phase1 != 0)
1700                         do_div(speed1, ns->ln_run_time_phase1);
1701                 if (rtime != 0)
1702                         do_div(speed2, rtime);
1703                 rc = snprintf(buf, len,
1704                               "checked_phase1: "LPU64"\n"
1705                               "checked_phase2: "LPU64"\n"
1706                               "updated_phase1: "LPU64"\n"
1707                               "updated_phase2: "LPU64"\n"
1708                               "failed_phase1: "LPU64"\n"
1709                               "failed_phase2: "LPU64"\n"
1710                               "dirs: "LPU64"\n"
1711                               "M-linked: "LPU64"\n"
1712                               "nlinks_repaired: "LPU64"\n"
1713                               "lost_found: "LPU64"\n"
1714                               "success_count: %u\n"
1715                               "run_time_phase1: %u seconds\n"
1716                               "run_time_phase2: %u seconds\n"
1717                               "average_speed_phase1: "LPU64" items/sec\n"
1718                               "average_speed_phase2: "LPU64" objs/sec\n"
1719                               "real-time_speed_phase1: N/A\n"
1720                               "real-time_speed_phase2: "LPU64" objs/sec\n"
1721                               "current_position: "DFID"\n",
1722                               ns->ln_items_checked,
1723                               checked,
1724                               ns->ln_items_repaired,
1725                               ns->ln_objs_repaired_phase2,
1726                               ns->ln_items_failed,
1727                               ns->ln_objs_failed_phase2,
1728                               ns->ln_dirs_checked,
1729                               ns->ln_mlinked_checked,
1730                               ns->ln_objs_nlink_repaired,
1731                               ns->ln_objs_lost_found,
1732                               ns->ln_success_count,
1733                               ns->ln_run_time_phase1,
1734                               rtime,
1735                               speed1,
1736                               speed2,
1737                               new_checked,
1738                               PFID(&ns->ln_fid_latest_scanned_phase2));
1739                 if (rc <= 0)
1740                         goto out;
1741
1742                 buf += rc;
1743                 len -= rc;
1744         } else {
1745                 __u64 speed1 = ns->ln_items_checked;
1746                 __u64 speed2 = ns->ln_objs_checked_phase2;
1747
1748                 if (ns->ln_run_time_phase1 != 0)
1749                         do_div(speed1, ns->ln_run_time_phase1);
1750                 if (ns->ln_run_time_phase2 != 0)
1751                         do_div(speed2, ns->ln_run_time_phase2);
1752                 rc = snprintf(buf, len,
1753                               "checked_phase1: "LPU64"\n"
1754                               "checked_phase2: "LPU64"\n"
1755                               "updated_phase1: "LPU64"\n"
1756                               "updated_phase2: "LPU64"\n"
1757                               "failed_phase1: "LPU64"\n"
1758                               "failed_phase2: "LPU64"\n"
1759                               "dirs: "LPU64"\n"
1760                               "M-linked: "LPU64"\n"
1761                               "nlinks_repaired: "LPU64"\n"
1762                               "lost_found: "LPU64"\n"
1763                               "success_count: %u\n"
1764                               "run_time_phase1: %u seconds\n"
1765                               "run_time_phase2: %u seconds\n"
1766                               "average_speed_phase1: "LPU64" items/sec\n"
1767                               "average_speed_phase2: "LPU64" objs/sec\n"
1768                               "real-time_speed_phase1: N/A\n"
1769                               "real-time_speed_phase2: N/A\n"
1770                               "current_position: N/A\n",
1771                               ns->ln_items_checked,
1772                               ns->ln_objs_checked_phase2,
1773                               ns->ln_items_repaired,
1774                               ns->ln_objs_repaired_phase2,
1775                               ns->ln_items_failed,
1776                               ns->ln_objs_failed_phase2,
1777                               ns->ln_dirs_checked,
1778                               ns->ln_mlinked_checked,
1779                               ns->ln_objs_nlink_repaired,
1780                               ns->ln_objs_lost_found,
1781                               ns->ln_success_count,
1782                               ns->ln_run_time_phase1,
1783                               ns->ln_run_time_phase2,
1784                               speed1,
1785                               speed2);
1786                 if (rc <= 0)
1787                         goto out;
1788
1789                 buf += rc;
1790                 len -= rc;
1791         }
1792         ret = save - len;
1793
1794 out:
1795         up_read(&com->lc_sem);
1796         return ret;
1797 }
1798
1799 static int mdd_lfsck_namespace_double_scan(const struct lu_env *env,
1800                                            struct lfsck_component *com)
1801 {
1802         struct md_lfsck         *lfsck  = com->lc_lfsck;
1803         struct ptlrpc_thread    *thread = &lfsck->ml_thread;
1804         struct mdd_device       *mdd    = mdd_lfsck2mdd(lfsck);
1805         struct lfsck_bookmark   *bk     = &lfsck->ml_bookmark_ram;
1806         struct lfsck_namespace  *ns     =
1807                                 (struct lfsck_namespace *)com->lc_file_ram;
1808         struct dt_object        *obj    = com->lc_obj;
1809         const struct dt_it_ops  *iops   = &obj->do_index_ops->dio_it;
1810         struct mdd_object       *target;
1811         struct dt_it            *di;
1812         struct dt_key           *key;
1813         struct lu_fid            fid;
1814         int                      rc;
1815         __u8                     flags;
1816         ENTRY;
1817
1818         lfsck->ml_new_scanned = 0;
1819         lfsck->ml_time_last_checkpoint = cfs_time_current();
1820         lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1821                                 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1822
1823         di = iops->init(env, obj, 0, BYPASS_CAPA);
1824         if (IS_ERR(di))
1825                 RETURN(PTR_ERR(di));
1826
1827         fid_cpu_to_be(&fid, &ns->ln_fid_latest_scanned_phase2);
1828         rc = iops->get(env, di, (const struct dt_key *)&fid);
1829         if (rc < 0)
1830                 GOTO(fini, rc);
1831
1832         /* Skip the start one, which either has been processed or non-exist. */
1833         rc = iops->next(env, di);
1834         if (rc != 0)
1835                 GOTO(put, rc);
1836
1837         if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_DOUBLESCAN))
1838                 GOTO(put, rc = 0);
1839
1840         do {
1841                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY3) &&
1842                     cfs_fail_val > 0) {
1843                         struct l_wait_info lwi;
1844
1845                         lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
1846                                           NULL, NULL);
1847                         l_wait_event(thread->t_ctl_waitq,
1848                                      !thread_is_running(thread),
1849                                      &lwi);
1850                 }
1851
1852                 key = iops->key(env, di);
1853                 fid_be_to_cpu(&fid, (const struct lu_fid *)key);
1854                 target = mdd_object_find(env, mdd, &fid);
1855                 down_write(&com->lc_sem);
1856                 if (target == NULL) {
1857                         rc = 0;
1858                         goto checkpoint;
1859                 } else if (IS_ERR(target)) {
1860                         rc = PTR_ERR(target);
1861                         goto checkpoint;
1862                 }
1863
1864                 /* XXX: need more processing for remote object in the future. */
1865                 if (!mdd_object_exists(target) || mdd_object_remote(target))
1866                         goto obj_put;
1867
1868                 rc = iops->rec(env, di, (struct dt_rec *)&flags, 0);
1869                 if (rc == 0)
1870                         rc = mdd_lfsck_namespace_double_scan_one(env, com,
1871                                                                  target, flags);
1872
1873 obj_put:
1874                 mdd_object_put(env, target);
1875
1876 checkpoint:
1877                 lfsck->ml_new_scanned++;
1878                 com->lc_new_checked++;
1879                 ns->ln_fid_latest_scanned_phase2 = fid;
1880                 if (rc > 0)
1881                         ns->ln_objs_repaired_phase2++;
1882                 else if (rc < 0)
1883                         ns->ln_objs_failed_phase2++;
1884                 up_write(&com->lc_sem);
1885
1886                 if ((rc == 0) || ((rc > 0) && !(bk->lb_param & LPF_DRYRUN))) {
1887                         mdd_lfsck_namespace_delete(env, com, &fid);
1888                 } else if (rc < 0) {
1889                         flags |= LLF_REPAIR_FAILED;
1890                         mdd_lfsck_namespace_update(env, com, &fid, flags, true);
1891                 }
1892
1893                 if (rc < 0 && bk->lb_param & LPF_FAILOUT)
1894                         GOTO(put, rc);
1895
1896                 if (likely(cfs_time_beforeq(cfs_time_current(),
1897                                             lfsck->ml_time_next_checkpoint)) ||
1898                     com->lc_new_checked == 0)
1899                         goto speed;
1900
1901                 down_write(&com->lc_sem);
1902                 ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1903                                 HALF_SEC - lfsck->ml_time_last_checkpoint);
1904                 ns->ln_time_last_checkpoint = cfs_time_current_sec();
1905                 ns->ln_objs_checked_phase2 += com->lc_new_checked;
1906                 com->lc_new_checked = 0;
1907                 rc = mdd_lfsck_namespace_store(env, com, false);
1908                 up_write(&com->lc_sem);
1909                 if (rc != 0)
1910                         GOTO(put, rc);
1911
1912                 lfsck->ml_time_last_checkpoint = cfs_time_current();
1913                 lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
1914                                 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
1915
1916 speed:
1917                 mdd_lfsck_control_speed(lfsck);
1918                 if (unlikely(!thread_is_running(thread)))
1919                         GOTO(put, rc = 0);
1920
1921                 rc = iops->next(env, di);
1922         } while (rc == 0);
1923
1924         GOTO(put, rc);
1925
1926 put:
1927         iops->put(env, di);
1928
1929 fini:
1930         iops->fini(env, di);
1931         down_write(&com->lc_sem);
1932
1933         ns->ln_run_time_phase2 += cfs_duration_sec(cfs_time_current() +
1934                                 HALF_SEC - lfsck->ml_time_last_checkpoint);
1935         ns->ln_time_last_checkpoint = cfs_time_current_sec();
1936         ns->ln_objs_checked_phase2 += com->lc_new_checked;
1937         com->lc_new_checked = 0;
1938
1939         if (rc > 0) {
1940                 com->lc_journal = 0;
1941                 ns->ln_status = LS_COMPLETED;
1942                 if (!(bk->lb_param & LPF_DRYRUN))
1943                         ns->ln_flags &=
1944                         ~(LF_SCANNED_ONCE | LF_INCONSISTENT | LF_UPGRADE);
1945                 ns->ln_time_last_complete = ns->ln_time_last_checkpoint;
1946                 ns->ln_success_count++;
1947         } else if (rc == 0) {
1948                 if (lfsck->ml_paused)
1949                         ns->ln_status = LS_PAUSED;
1950                 else
1951                         ns->ln_status = LS_STOPPED;
1952         } else {
1953                 ns->ln_status = LS_FAILED;
1954         }
1955
1956         if (ns->ln_status != LS_PAUSED) {
1957                 spin_lock(&lfsck->ml_lock);
1958                 cfs_list_del_init(&com->lc_link);
1959                 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
1960                 spin_unlock(&lfsck->ml_lock);
1961         }
1962
1963         rc = mdd_lfsck_namespace_store(env, com, false);
1964
1965         up_write(&com->lc_sem);
1966         return rc;
1967 }
1968
1969 static struct lfsck_operations mdd_lfsck_namespace_ops = {
1970         .lfsck_reset            = mdd_lfsck_namespace_reset,
1971         .lfsck_fail             = mdd_lfsck_namespace_fail,
1972         .lfsck_checkpoint       = mdd_lfsck_namespace_checkpoint,
1973         .lfsck_prep             = mdd_lfsck_namespace_prep,
1974         .lfsck_exec_oit         = mdd_lfsck_namespace_exec_oit,
1975         .lfsck_exec_dir         = mdd_lfsck_namespace_exec_dir,
1976         .lfsck_post             = mdd_lfsck_namespace_post,
1977         .lfsck_dump             = mdd_lfsck_namespace_dump,
1978         .lfsck_double_scan      = mdd_lfsck_namespace_double_scan,
1979 };
1980
1981 /* LFSCK component setup/cleanup functions */
1982
1983 static int mdd_lfsck_namespace_setup(const struct lu_env *env,
1984                                      struct md_lfsck *lfsck)
1985 {
1986         struct mdd_device      *mdd = mdd_lfsck2mdd(lfsck);
1987         struct lfsck_component *com;
1988         struct lfsck_namespace *ns;
1989         struct dt_object       *obj;
1990         int                     rc;
1991         ENTRY;
1992
1993         OBD_ALLOC_PTR(com);
1994         if (com == NULL)
1995                 RETURN(-ENOMEM);
1996
1997         CFS_INIT_LIST_HEAD(&com->lc_link);
1998         CFS_INIT_LIST_HEAD(&com->lc_link_dir);
1999         init_rwsem(&com->lc_sem);
2000         atomic_set(&com->lc_ref, 1);
2001         com->lc_lfsck = lfsck;
2002         com->lc_type = LT_NAMESPACE;
2003         com->lc_ops = &mdd_lfsck_namespace_ops;
2004         com->lc_file_size = sizeof(struct lfsck_namespace);
2005         OBD_ALLOC(com->lc_file_ram, com->lc_file_size);
2006         if (com->lc_file_ram == NULL)
2007                 GOTO(out, rc = -ENOMEM);
2008
2009         OBD_ALLOC(com->lc_file_disk, com->lc_file_size);
2010         if (com->lc_file_disk == NULL)
2011                 GOTO(out, rc = -ENOMEM);
2012
2013         obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_namespace_name,
2014                             &mdd_env_info(env)->mti_fid);
2015         if (IS_ERR(obj))
2016                 GOTO(out, rc = PTR_ERR(obj));
2017
2018         com->lc_obj = obj;
2019         rc = obj->do_ops->do_index_try(env, obj, &dt_lfsck_features);
2020         if (rc != 0)
2021                 GOTO(out, rc);
2022
2023         rc = mdd_lfsck_namespace_load(env, com);
2024         if (rc > 0)
2025                 rc = mdd_lfsck_namespace_reset(env, com, true);
2026         else if (rc == -ENODATA)
2027                 rc = mdd_lfsck_namespace_init(env, com);
2028         if (rc != 0)
2029                 GOTO(out, rc);
2030
2031         ns = (struct lfsck_namespace *)com->lc_file_ram;
2032         switch (ns->ln_status) {
2033         case LS_INIT:
2034         case LS_COMPLETED:
2035         case LS_FAILED:
2036         case LS_STOPPED:
2037                 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_idle);
2038                 break;
2039         default:
2040                 CERROR("%s: unknown status: %u\n",
2041                        mdd_lfsck2name(lfsck), ns->ln_status);
2042                 /* fall through */
2043         case LS_SCANNING_PHASE1:
2044         case LS_SCANNING_PHASE2:
2045                 /* No need to store the status to disk right now.
2046                  * If the system crashed before the status stored,
2047                  * it will be loaded back when next time. */
2048                 ns->ln_status = LS_CRASHED;
2049                 /* fall through */
2050         case LS_PAUSED:
2051         case LS_CRASHED:
2052                 cfs_list_add_tail(&com->lc_link, &lfsck->ml_list_scan);
2053                 cfs_list_add_tail(&com->lc_link_dir, &lfsck->ml_list_dir);
2054                 break;
2055         }
2056
2057         GOTO(out, rc = 0);
2058
2059 out:
2060         if (rc != 0)
2061                 mdd_lfsck_component_cleanup(env, com);
2062         return rc;
2063 }
2064
2065 /* helper functions for framework */
2066
2067 static int object_is_client_visible(const struct lu_env *env,
2068                                     struct mdd_device *mdd,
2069                                     struct mdd_object *obj)
2070 {
2071         struct lu_fid *fid   = &mdd_env_info(env)->mti_fid;
2072         int            depth = 0;
2073         int            rc;
2074
2075         LASSERT(S_ISDIR(mdd_object_type(obj)));
2076
2077         while (1) {
2078                 if (mdd_is_root(mdd, mdo2fid(obj))) {
2079                         if (depth > 0)
2080                                 mdd_object_put(env, obj);
2081                         return 1;
2082                 }
2083
2084                 mdd_read_lock(env, obj, MOR_TGT_CHILD);
2085                 if (unlikely(mdd_is_dead_obj(obj))) {
2086                         mdd_read_unlock(env, obj);
2087                         if (depth > 0)
2088                                 mdd_object_put(env, obj);
2089                         return 0;
2090                 }
2091
2092                 rc = dt_xattr_get(env, mdd_object_child(obj),
2093                                   mdd_buf_get(env, NULL, 0), XATTR_NAME_LINK,
2094                                   BYPASS_CAPA);
2095                 mdd_read_unlock(env, obj);
2096                 if (rc >= 0) {
2097                         if (depth > 0)
2098                                 mdd_object_put(env, obj);
2099                         return 1;
2100                 }
2101
2102                 if (rc < 0 && rc != -ENODATA) {
2103                         if (depth > 0)
2104                                 mdd_object_put(env, obj);
2105                         return rc;
2106                 }
2107
2108                 rc = mdd_parent_fid(env, obj, fid);
2109                 if (depth > 0)
2110                         mdd_object_put(env, obj);
2111                 if (rc != 0)
2112                         return rc;
2113
2114                 if (unlikely(lu_fid_eq(fid, &mdd->mdd_local_root_fid)))
2115                         return 0;
2116
2117                 obj = mdd_object_find(env, mdd, fid);
2118                 if (obj == NULL)
2119                         return 0;
2120                 else if (IS_ERR(obj))
2121                         return PTR_ERR(obj);
2122
2123                 /* XXX: need more processing for remote object in the future. */
2124                 if (!mdd_object_exists(obj) || mdd_object_remote(obj)) {
2125                         mdd_object_put(env, obj);
2126                         return 0;
2127                 }
2128
2129                 depth++;
2130         }
2131         return 0;
2132 }
2133
2134 static void mdd_lfsck_unpack_ent(struct lu_dirent *ent)
2135 {
2136         fid_le_to_cpu(&ent->lde_fid, &ent->lde_fid);
2137         ent->lde_hash = le64_to_cpu(ent->lde_hash);
2138         ent->lde_reclen = le16_to_cpu(ent->lde_reclen);
2139         ent->lde_namelen = le16_to_cpu(ent->lde_namelen);
2140         ent->lde_attrs = le32_to_cpu(ent->lde_attrs);
2141
2142         /* Make sure the name is terminated with '0'.
2143          * The data (type) after ent::lde_name maybe
2144          * broken, but we do not care. */
2145         ent->lde_name[ent->lde_namelen] = 0;
2146 }
2147
2148 /* LFSCK wrap functions */
2149
2150 static void mdd_lfsck_fail(const struct lu_env *env, struct md_lfsck *lfsck,
2151                            bool oit, bool new_checked)
2152 {
2153         struct lfsck_component *com;
2154
2155         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2156                 com->lc_ops->lfsck_fail(env, com, oit, new_checked);
2157         }
2158 }
2159
2160 static int mdd_lfsck_checkpoint(const struct lu_env *env,
2161                                 struct md_lfsck *lfsck, bool oit)
2162 {
2163         struct lfsck_component *com;
2164         int                     rc;
2165
2166         if (likely(cfs_time_beforeq(cfs_time_current(),
2167                                     lfsck->ml_time_next_checkpoint)))
2168                 return 0;
2169
2170         mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, oit, !oit);
2171         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2172                 rc = com->lc_ops->lfsck_checkpoint(env, com, false);
2173                 if (rc != 0)
2174                         return rc;;
2175         }
2176
2177         lfsck->ml_time_last_checkpoint = cfs_time_current();
2178         lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2179                                 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2180         return 0;
2181 }
2182
2183 static int mdd_lfsck_prep(struct lu_env *env, struct md_lfsck *lfsck)
2184 {
2185         struct mdd_device      *mdd     = mdd_lfsck2mdd(lfsck);
2186         struct mdd_object      *obj     = NULL;
2187         struct dt_object       *dt_obj;
2188         struct lfsck_component *com;
2189         struct lfsck_component *next;
2190         struct lfsck_position  *pos     = NULL;
2191         const struct dt_it_ops *iops    =
2192                                 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2193         struct dt_it           *di;
2194         int                     rc;
2195         ENTRY;
2196
2197         LASSERT(lfsck->ml_obj_dir == NULL);
2198         LASSERT(lfsck->ml_di_dir == NULL);
2199
2200         cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2201                 com->lc_new_checked = 0;
2202                 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2203                         com->lc_journal = 0;
2204
2205                 rc = com->lc_ops->lfsck_prep(env, com);
2206                 if (rc != 0)
2207                         RETURN(rc);
2208
2209                 if ((pos == NULL) ||
2210                     (!mdd_lfsck_pos_is_zero(&com->lc_pos_start) &&
2211                      mdd_lfsck_pos_is_eq(pos, &com->lc_pos_start) > 0))
2212                         pos = &com->lc_pos_start;
2213         }
2214
2215         /* Init otable-based iterator. */
2216         if (pos == NULL) {
2217                 rc = iops->load(env, lfsck->ml_di_oit, 0);
2218                 GOTO(out, rc = (rc >= 0 ? 0 : rc));
2219         }
2220
2221         rc = iops->load(env, lfsck->ml_di_oit, pos->lp_oit_cookie);
2222         if (rc < 0)
2223                 GOTO(out, rc);
2224
2225         if (fid_is_zero(&pos->lp_dir_parent))
2226                 GOTO(out, rc = 0);
2227
2228         /* Find the directory for namespace-based traverse. */
2229         obj = mdd_object_find(env, mdd, &pos->lp_dir_parent);
2230         if (obj == NULL)
2231                 GOTO(out, rc = 0);
2232         else if (IS_ERR(obj))
2233                 RETURN(PTR_ERR(obj));
2234
2235         /* XXX: need more processing for remote object in the future. */
2236         if (!mdd_object_exists(obj) || mdd_object_remote(obj) ||
2237             unlikely(!S_ISDIR(mdd_object_type(obj))))
2238                 GOTO(out, rc = 0);
2239
2240         if (unlikely(mdd_is_dead_obj(obj)))
2241                 GOTO(out, rc = 0);
2242
2243         dt_obj = mdd_object_child(obj);
2244         if (unlikely(!dt_try_as_dir(env, dt_obj)))
2245                 GOTO(out, rc = -ENOTDIR);
2246
2247         /* Init the namespace-based directory traverse. */
2248         iops = &dt_obj->do_index_ops->dio_it;
2249         di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2250         if (IS_ERR(di))
2251                 GOTO(out, rc = PTR_ERR(di));
2252
2253         rc = iops->load(env, di, pos->lp_dir_cookie);
2254         if (rc == 0)
2255                 rc = iops->next(env, di);
2256         else if (rc > 0)
2257                 rc = 0;
2258
2259         if (rc != 0) {
2260                 iops->put(env, di);
2261                 iops->fini(env, di);
2262                 GOTO(out, rc);
2263         }
2264
2265         lfsck->ml_obj_dir = dt_obj;
2266         spin_lock(&lfsck->ml_lock);
2267         lfsck->ml_di_dir = di;
2268         spin_unlock(&lfsck->ml_lock);
2269         obj = NULL;
2270
2271         GOTO(out, rc = 0);
2272
2273 out:
2274         if (obj != NULL)
2275                 mdd_object_put(env, obj);
2276
2277         if (rc != 0)
2278                 return (rc > 0 ? 0 : rc);
2279
2280         mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, false, false);
2281         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2282                 rc = com->lc_ops->lfsck_checkpoint(env, com, true);
2283                 if (rc != 0)
2284                         break;
2285         }
2286
2287         lfsck->ml_time_last_checkpoint = cfs_time_current();
2288         lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2289                                 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2290         return rc;
2291 }
2292
2293 static int mdd_lfsck_exec_oit(const struct lu_env *env, struct md_lfsck *lfsck,
2294                               struct mdd_object *obj)
2295 {
2296         struct lfsck_component *com;
2297         struct dt_object       *dt_obj;
2298         const struct dt_it_ops *iops;
2299         struct dt_it           *di;
2300         int                     rc;
2301         ENTRY;
2302
2303         LASSERT(lfsck->ml_obj_dir == NULL);
2304
2305         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2306                 rc = com->lc_ops->lfsck_exec_oit(env, com, obj);
2307                 if (rc != 0)
2308                         RETURN(rc);
2309         }
2310
2311         if (!S_ISDIR(mdd_object_type(obj)) ||
2312             cfs_list_empty(&lfsck->ml_list_dir))
2313                RETURN(0);
2314
2315         rc = object_is_client_visible(env, mdd_lfsck2mdd(lfsck), obj);
2316         if (rc <= 0)
2317                 GOTO(out, rc);
2318
2319         if (unlikely(mdd_is_dead_obj(obj)))
2320                 GOTO(out, rc = 0);
2321
2322         dt_obj = mdd_object_child(obj);
2323         if (unlikely(!dt_try_as_dir(env, dt_obj)))
2324                 GOTO(out, rc = -ENOTDIR);
2325
2326         iops = &dt_obj->do_index_ops->dio_it;
2327         di = iops->init(env, dt_obj, lfsck->ml_args_dir, BYPASS_CAPA);
2328         if (IS_ERR(di))
2329                 GOTO(out, rc = PTR_ERR(di));
2330
2331         rc = iops->load(env, di, 0);
2332         if (rc == 0)
2333                 rc = iops->next(env, di);
2334         else if (rc > 0)
2335                 rc = 0;
2336
2337         if (rc != 0) {
2338                 iops->put(env, di);
2339                 iops->fini(env, di);
2340                 GOTO(out, rc);
2341         }
2342
2343         mdd_object_get(obj);
2344         lfsck->ml_obj_dir = dt_obj;
2345         spin_lock(&lfsck->ml_lock);
2346         lfsck->ml_di_dir = di;
2347         spin_unlock(&lfsck->ml_lock);
2348
2349         GOTO(out, rc = 0);
2350
2351 out:
2352         if (rc < 0)
2353                 mdd_lfsck_fail(env, lfsck, false, false);
2354         return (rc > 0 ? 0 : rc);
2355 }
2356
2357 static int mdd_lfsck_exec_dir(const struct lu_env *env, struct md_lfsck *lfsck,
2358                               struct mdd_object *obj, struct lu_dirent *ent)
2359 {
2360         struct lfsck_component *com;
2361         int                     rc;
2362
2363         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2364                 rc = com->lc_ops->lfsck_exec_dir(env, com, obj, ent);
2365                 if (rc != 0)
2366                         return rc;
2367         }
2368         return 0;
2369 }
2370
2371 static int mdd_lfsck_post(const struct lu_env *env, struct md_lfsck *lfsck,
2372                           int result)
2373 {
2374         struct lfsck_component *com;
2375         struct lfsck_component *next;
2376         int                     rc;
2377
2378         mdd_lfsck_pos_fill(env, lfsck, &lfsck->ml_pos_current, true, true);
2379         cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_scan, lc_link) {
2380                 rc = com->lc_ops->lfsck_post(env, com, result);
2381                 if (rc != 0)
2382                         return rc;
2383         }
2384
2385         lfsck->ml_time_last_checkpoint = cfs_time_current();
2386         lfsck->ml_time_next_checkpoint = lfsck->ml_time_last_checkpoint +
2387                                 cfs_time_seconds(LFSCK_CHECKPOINT_INTERVAL);
2388         return result;
2389 }
2390
2391 static int mdd_lfsck_double_scan(const struct lu_env *env,
2392                                  struct md_lfsck *lfsck)
2393 {
2394         struct lfsck_component *com;
2395         struct lfsck_component *next;
2396         int                     rc;
2397
2398         cfs_list_for_each_entry_safe(com, next, &lfsck->ml_list_double_scan,
2399                                      lc_link) {
2400                 if (lfsck->ml_bookmark_ram.lb_param & LPF_DRYRUN)
2401                         com->lc_journal = 0;
2402
2403                 rc = com->lc_ops->lfsck_double_scan(env, com);
2404                 if (rc != 0)
2405                         return rc;
2406         }
2407         return 0;
2408 }
2409
2410 /* LFSCK engines */
2411
2412 static int mdd_lfsck_dir_engine(const struct lu_env *env,
2413                                 struct md_lfsck *lfsck)
2414 {
2415         struct mdd_thread_info  *info   = mdd_env_info(env);
2416         struct mdd_device       *mdd    = mdd_lfsck2mdd(lfsck);
2417         const struct dt_it_ops  *iops   =
2418                         &lfsck->ml_obj_dir->do_index_ops->dio_it;
2419         struct dt_it            *di     = lfsck->ml_di_dir;
2420         struct lu_dirent        *ent    = &info->mti_ent;
2421         struct lu_fid           *fid    = &info->mti_fid;
2422         struct lfsck_bookmark   *bk     = &lfsck->ml_bookmark_ram;
2423         struct ptlrpc_thread    *thread = &lfsck->ml_thread;
2424         int                      rc;
2425         ENTRY;
2426
2427         do {
2428                 struct mdd_object *child;
2429
2430                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY2) &&
2431                     cfs_fail_val > 0) {
2432                         struct l_wait_info lwi;
2433
2434                         lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2435                                           NULL, NULL);
2436                         l_wait_event(thread->t_ctl_waitq,
2437                                      !thread_is_running(thread),
2438                                      &lwi);
2439                 }
2440
2441                 lfsck->ml_new_scanned++;
2442                 rc = iops->rec(env, di, (struct dt_rec *)ent,
2443                                lfsck->ml_args_dir);
2444                 if (rc != 0) {
2445                         mdd_lfsck_fail(env, lfsck, false, true);
2446                         if (bk->lb_param & LPF_FAILOUT)
2447                                 RETURN(rc);
2448                         else
2449                                 goto checkpoint;
2450                 }
2451
2452                 mdd_lfsck_unpack_ent(ent);
2453                 if (ent->lde_attrs & LUDA_IGNORE)
2454                         goto checkpoint;
2455
2456                 *fid = ent->lde_fid;
2457                 child = mdd_object_find(env, mdd, fid);
2458                 if (child == NULL) {
2459                         goto checkpoint;
2460                 } else if (IS_ERR(child)) {
2461                         mdd_lfsck_fail(env, lfsck, false, true);
2462                         if (bk->lb_param & LPF_FAILOUT)
2463                                 RETURN(PTR_ERR(child));
2464                         else
2465                                 goto checkpoint;
2466                 }
2467
2468                 /* XXX: need more processing for remote object in the future. */
2469                 if (mdd_object_exists(child) && !mdd_object_remote(child))
2470                         rc = mdd_lfsck_exec_dir(env, lfsck, child, ent);
2471                 mdd_object_put(env, child);
2472                 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2473                         RETURN(rc);
2474
2475 checkpoint:
2476                 rc = mdd_lfsck_checkpoint(env, lfsck, false);
2477                 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2478                         RETURN(rc);
2479
2480                 /* Rate control. */
2481                 mdd_lfsck_control_speed(lfsck);
2482                 if (unlikely(!thread_is_running(thread)))
2483                         RETURN(0);
2484
2485                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL2)) {
2486                         spin_lock(&lfsck->ml_lock);
2487                         thread_set_flags(thread, SVC_STOPPING);
2488                         spin_unlock(&lfsck->ml_lock);
2489                         RETURN(-EINVAL);
2490                 }
2491
2492                 rc = iops->next(env, di);
2493         } while (rc == 0);
2494
2495         if (rc > 0 && !lfsck->ml_oit_over)
2496                 mdd_lfsck_close_dir(env, lfsck);
2497
2498         RETURN(rc);
2499 }
2500
2501 static int mdd_lfsck_oit_engine(const struct lu_env *env,
2502                                 struct md_lfsck *lfsck)
2503 {
2504         struct mdd_thread_info  *info   = mdd_env_info(env);
2505         struct mdd_device       *mdd    = mdd_lfsck2mdd(lfsck);
2506         const struct dt_it_ops  *iops   =
2507                                 &lfsck->ml_obj_oit->do_index_ops->dio_it;
2508         struct dt_it            *di     = lfsck->ml_di_oit;
2509         struct lu_fid           *fid    = &info->mti_fid;
2510         struct lfsck_bookmark   *bk     = &lfsck->ml_bookmark_ram;
2511         struct ptlrpc_thread    *thread = &lfsck->ml_thread;
2512         int                      rc;
2513         ENTRY;
2514
2515         do {
2516                 struct mdd_object *target;
2517
2518                 if (lfsck->ml_di_dir != NULL) {
2519                         rc = mdd_lfsck_dir_engine(env, lfsck);
2520                         if (rc <= 0)
2521                                 RETURN(rc);
2522                 }
2523
2524                 if (unlikely(lfsck->ml_oit_over))
2525                         RETURN(1);
2526
2527                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_DELAY1) &&
2528                     cfs_fail_val > 0) {
2529                         struct l_wait_info lwi;
2530
2531                         lwi = LWI_TIMEOUT(cfs_time_seconds(cfs_fail_val),
2532                                           NULL, NULL);
2533                         l_wait_event(thread->t_ctl_waitq,
2534                                      !thread_is_running(thread),
2535                                      &lwi);
2536                 }
2537
2538                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2539                         RETURN(0);
2540
2541                 lfsck->ml_new_scanned++;
2542                 rc = iops->rec(env, di, (struct dt_rec *)fid, 0);
2543                 if (rc != 0) {
2544                         mdd_lfsck_fail(env, lfsck, true, true);
2545                         if (bk->lb_param & LPF_FAILOUT)
2546                                 RETURN(rc);
2547                         else
2548                                 goto checkpoint;
2549                 }
2550
2551                 target = mdd_object_find(env, mdd, fid);
2552                 if (target == NULL) {
2553                         goto checkpoint;
2554                 } else if (IS_ERR(target)) {
2555                         mdd_lfsck_fail(env, lfsck, true, true);
2556                         if (bk->lb_param & LPF_FAILOUT)
2557                                 RETURN(PTR_ERR(target));
2558                         else
2559                                 goto checkpoint;
2560                 }
2561
2562                 /* XXX: In fact, low layer otable-based iteration should not
2563                  *      return agent object. But before LU-2646 resolved, we
2564                  *      need more processing for agent object. */
2565                 if (mdd_object_exists(target) && !mdd_object_remote(target))
2566                         rc = mdd_lfsck_exec_oit(env, lfsck, target);
2567                 mdd_object_put(env, target);
2568                 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2569                         RETURN(rc);
2570
2571 checkpoint:
2572                 rc = mdd_lfsck_checkpoint(env, lfsck, true);
2573                 if (rc != 0 && bk->lb_param & LPF_FAILOUT)
2574                         RETURN(rc);
2575
2576                 /* Rate control. */
2577                 mdd_lfsck_control_speed(lfsck);
2578
2579                 if (OBD_FAIL_CHECK(OBD_FAIL_LFSCK_FATAL1)) {
2580                         spin_lock(&lfsck->ml_lock);
2581                         thread_set_flags(thread, SVC_STOPPING);
2582                         spin_unlock(&lfsck->ml_lock);
2583                         RETURN(-EINVAL);
2584                 }
2585
2586                 rc = iops->next(env, di);
2587                 if (rc > 0)
2588                         lfsck->ml_oit_over = 1;
2589
2590                 if (unlikely(!thread_is_running(thread)))
2591                         RETURN(0);
2592         } while (rc == 0 || lfsck->ml_di_dir != NULL);
2593
2594         RETURN(rc);
2595 }
2596
2597 static int mdd_lfsck_main(void *args)
2598 {
2599         struct lu_env            env;
2600         struct md_lfsck         *lfsck    = (struct md_lfsck *)args;
2601         struct ptlrpc_thread    *thread   = &lfsck->ml_thread;
2602         struct dt_object        *oit_obj  = lfsck->ml_obj_oit;
2603         const struct dt_it_ops  *oit_iops = &oit_obj->do_index_ops->dio_it;
2604         struct dt_it            *oit_di;
2605         int                      rc;
2606         ENTRY;
2607
2608         cfs_daemonize("lfsck");
2609         rc = lu_env_init(&env, LCT_MD_THREAD | LCT_DT_THREAD);
2610         if (rc != 0) {
2611                 CERROR("%s: LFSCK, fail to init env, rc = %d\n",
2612                        mdd_lfsck2name(lfsck), rc);
2613                 GOTO(noenv, rc);
2614         }
2615
2616         oit_di = oit_iops->init(&env, oit_obj, lfsck->ml_args_oit, BYPASS_CAPA);
2617         if (IS_ERR(oit_di)) {
2618                 rc = PTR_ERR(oit_di);
2619                 CERROR("%s: LFSCK, fail to init iteration, rc = %d\n",
2620                        mdd_lfsck2name(lfsck), rc);
2621                 GOTO(fini_env, rc);
2622         }
2623
2624         spin_lock(&lfsck->ml_lock);
2625         lfsck->ml_di_oit = oit_di;
2626         spin_unlock(&lfsck->ml_lock);
2627         rc = mdd_lfsck_prep(&env, lfsck);
2628         if (rc != 0)
2629                 GOTO(fini_oit, rc);
2630
2631         CDEBUG(D_LFSCK, "LFSCK entry: oit_flags = 0x%x, dir_flags = 0x%x, "
2632                "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2633                ", pid = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2634                lfsck->ml_pos_current.lp_oit_cookie,
2635                lfsck->ml_pos_current.lp_dir_cookie,
2636                PFID(&lfsck->ml_pos_current.lp_dir_parent),
2637                cfs_curproc_pid());
2638
2639         spin_lock(&lfsck->ml_lock);
2640         thread_set_flags(thread, SVC_RUNNING);
2641         spin_unlock(&lfsck->ml_lock);
2642         cfs_waitq_broadcast(&thread->t_ctl_waitq);
2643
2644         if (!cfs_list_empty(&lfsck->ml_list_scan) ||
2645             cfs_list_empty(&lfsck->ml_list_double_scan))
2646                 rc = mdd_lfsck_oit_engine(&env, lfsck);
2647         else
2648                 rc = 1;
2649
2650         CDEBUG(D_LFSCK, "LFSCK exit: oit_flags = 0x%x, dir_flags = 0x%x, "
2651                "oit_cookie = "LPU64", dir_cookie = "LPU64", parent = "DFID
2652                ", pid = %d, rc = %d\n", lfsck->ml_args_oit, lfsck->ml_args_dir,
2653                lfsck->ml_pos_current.lp_oit_cookie,
2654                lfsck->ml_pos_current.lp_dir_cookie,
2655                PFID(&lfsck->ml_pos_current.lp_dir_parent),
2656                cfs_curproc_pid(), rc);
2657
2658         if (lfsck->ml_paused && cfs_list_empty(&lfsck->ml_list_scan))
2659                 oit_iops->put(&env, oit_di);
2660
2661         if (!OBD_FAIL_CHECK(OBD_FAIL_LFSCK_CRASH))
2662                 rc = mdd_lfsck_post(&env, lfsck, rc);
2663         if (lfsck->ml_di_dir != NULL)
2664                 mdd_lfsck_close_dir(&env, lfsck);
2665
2666 fini_oit:
2667         spin_lock(&lfsck->ml_lock);
2668         lfsck->ml_di_oit = NULL;
2669         spin_unlock(&lfsck->ml_lock);
2670
2671         oit_iops->fini(&env, oit_di);
2672         if (rc == 1) {
2673                 if (!cfs_list_empty(&lfsck->ml_list_double_scan))
2674                         rc = mdd_lfsck_double_scan(&env, lfsck);
2675                 else
2676                         rc = 0;
2677         }
2678
2679         /* XXX: Purge the pinned objects in the future. */
2680
2681 fini_env:
2682         lu_env_fini(&env);
2683
2684 noenv:
2685         spin_lock(&lfsck->ml_lock);
2686         thread_set_flags(thread, SVC_STOPPED);
2687         cfs_waitq_broadcast(&thread->t_ctl_waitq);
2688         spin_unlock(&lfsck->ml_lock);
2689         return rc;
2690 }
2691
2692 /* external interfaces */
2693
2694 int mdd_lfsck_set_speed(const struct lu_env *env, struct md_lfsck *lfsck,
2695                         __u32 limit)
2696 {
2697         int rc;
2698
2699         mutex_lock(&lfsck->ml_mutex);
2700         __mdd_lfsck_set_speed(lfsck, limit);
2701         rc = mdd_lfsck_bookmark_store(env, lfsck);
2702         mutex_unlock(&lfsck->ml_mutex);
2703         return rc;
2704 }
2705
2706 int mdd_lfsck_dump(const struct lu_env *env, struct md_lfsck *lfsck,
2707                    __u16 type, char *buf, int len)
2708 {
2709         struct lfsck_component *com;
2710         int                     rc;
2711
2712         if (!lfsck->ml_initialized)
2713                 return -ENODEV;
2714
2715         com = mdd_lfsck_component_find(lfsck, type);
2716         if (com == NULL)
2717                 return -ENOTSUPP;
2718
2719         rc = com->lc_ops->lfsck_dump(env, com, buf, len);
2720         mdd_lfsck_component_put(env, com);
2721         return rc;
2722 }
2723
2724 int mdd_lfsck_start(const struct lu_env *env, struct md_lfsck *lfsck,
2725                     struct lfsck_start *start)
2726 {
2727         struct lfsck_bookmark  *bk     = &lfsck->ml_bookmark_ram;
2728         struct ptlrpc_thread   *thread = &lfsck->ml_thread;
2729         struct lfsck_component *com;
2730         struct l_wait_info      lwi    = { 0 };
2731         bool                    dirty  = false;
2732         int                     rc     = 0;
2733         __u16                   valid  = 0;
2734         __u16                   flags  = 0;
2735         ENTRY;
2736
2737         if (lfsck->ml_obj_oit == NULL)
2738                 RETURN(-ENOTSUPP);
2739
2740         /* start == NULL means auto trigger paused LFSCK. */
2741         if ((start == NULL) &&
2742             (cfs_list_empty(&lfsck->ml_list_scan) ||
2743              OBD_FAIL_CHECK(OBD_FAIL_LFSCK_NO_AUTO)))
2744                 RETURN(0);
2745
2746         mutex_lock(&lfsck->ml_mutex);
2747         spin_lock(&lfsck->ml_lock);
2748         if (!thread_is_init(thread) && !thread_is_stopped(thread)) {
2749                 spin_unlock(&lfsck->ml_lock);
2750                 mutex_unlock(&lfsck->ml_mutex);
2751                 RETURN(-EALREADY);
2752         }
2753
2754         spin_unlock(&lfsck->ml_lock);
2755
2756         lfsck->ml_paused = 0;
2757         lfsck->ml_oit_over = 0;
2758         lfsck->ml_drop_dryrun = 0;
2759         lfsck->ml_new_scanned = 0;
2760
2761         /* For auto trigger. */
2762         if (start == NULL)
2763                 goto trigger;
2764
2765         start->ls_version = bk->lb_version;
2766         if (start->ls_valid & LSV_SPEED_LIMIT) {
2767                 __mdd_lfsck_set_speed(lfsck, start->ls_speed_limit);
2768                 dirty = true;
2769         }
2770
2771         if (start->ls_valid & LSV_ERROR_HANDLE) {
2772                 valid |= DOIV_ERROR_HANDLE;
2773                 if (start->ls_flags & LPF_FAILOUT)
2774                         flags |= DOIF_FAILOUT;
2775
2776                 if ((start->ls_flags & LPF_FAILOUT) &&
2777                     !(bk->lb_param & LPF_FAILOUT)) {
2778                         bk->lb_param |= LPF_FAILOUT;
2779                         dirty = true;
2780                 } else if (!(start->ls_flags & LPF_FAILOUT) &&
2781                            (bk->lb_param & LPF_FAILOUT)) {
2782                         bk->lb_param &= ~LPF_FAILOUT;
2783                         dirty = true;
2784                 }
2785         }
2786
2787         if (start->ls_valid & LSV_DRYRUN) {
2788                 if ((start->ls_flags & LPF_DRYRUN) &&
2789                     !(bk->lb_param & LPF_DRYRUN)) {
2790                         bk->lb_param |= LPF_DRYRUN;
2791                         dirty = true;
2792                 } else if (!(start->ls_flags & LPF_DRYRUN) &&
2793                            (bk->lb_param & LPF_DRYRUN)) {
2794                         bk->lb_param &= ~LPF_DRYRUN;
2795                         lfsck->ml_drop_dryrun = 1;
2796                         dirty = true;
2797                 }
2798         }
2799
2800         if (dirty) {
2801                 rc = mdd_lfsck_bookmark_store(env, lfsck);
2802                 if (rc != 0)
2803                         GOTO(out, rc);
2804         }
2805
2806         if (start->ls_flags & LPF_RESET)
2807                 flags |= DOIF_RESET;
2808
2809         if (start->ls_active != 0) {
2810                 struct lfsck_component *next;
2811                 __u16 type = 1;
2812
2813                 if (start->ls_active == LFSCK_TYPES_ALL)
2814                         start->ls_active = LFSCK_TYPES_SUPPORTED;
2815
2816                 if (start->ls_active & ~LFSCK_TYPES_SUPPORTED) {
2817                         start->ls_active &= ~LFSCK_TYPES_SUPPORTED;
2818                         GOTO(out, rc = -ENOTSUPP);
2819                 }
2820
2821                 cfs_list_for_each_entry_safe(com, next,
2822                                              &lfsck->ml_list_scan, lc_link) {
2823                         if (!(com->lc_type & start->ls_active)) {
2824                                 rc = com->lc_ops->lfsck_post(env, com, 0);
2825                                 if (rc != 0)
2826                                         GOTO(out, rc);
2827                         }
2828                 }
2829
2830                 while (start->ls_active != 0) {
2831                         if (type & start->ls_active) {
2832                                 com = __mdd_lfsck_component_find(lfsck, type,
2833                                                         &lfsck->ml_list_idle);
2834                                 if (com != NULL) {
2835                                         /* The component status will be updated
2836                                          * when its prep() is called later by
2837                                          * the LFSCK main engine. */
2838                                         cfs_list_del_init(&com->lc_link);
2839                                         cfs_list_add_tail(&com->lc_link,
2840                                                           &lfsck->ml_list_scan);
2841                                 }
2842                                 start->ls_active &= ~type;
2843                         }
2844                         type <<= 1;
2845                 }
2846         }
2847
2848         cfs_list_for_each_entry(com, &lfsck->ml_list_scan, lc_link) {
2849                 start->ls_active |= com->lc_type;
2850                 if (flags & DOIF_RESET) {
2851                         rc = com->lc_ops->lfsck_reset(env, com, false);
2852                         if (rc != 0)
2853                                 GOTO(out, rc);
2854                 }
2855         }
2856
2857 trigger:
2858         lfsck->ml_args_dir = LUDA_64BITHASH | LUDA_VERIFY;
2859         if (bk->lb_param & LPF_DRYRUN)
2860                 lfsck->ml_args_dir |= LUDA_VERIFY_DRYRUN;
2861
2862         if (bk->lb_param & LPF_FAILOUT) {
2863                 valid |= DOIV_ERROR_HANDLE;
2864                 flags |= DOIF_FAILOUT;
2865         }
2866
2867         if (!cfs_list_empty(&lfsck->ml_list_scan))
2868                 flags |= DOIF_OUTUSED;
2869
2870         lfsck->ml_args_oit = (flags << DT_OTABLE_IT_FLAGS_SHIFT) | valid;
2871         thread_set_flags(thread, 0);
2872         rc = cfs_create_thread(mdd_lfsck_main, lfsck, 0);
2873         if (rc < 0)
2874                 CERROR("%s: cannot start LFSCK thread, rc = %d\n",
2875                        mdd_lfsck2name(lfsck), rc);
2876         else
2877                 l_wait_event(thread->t_ctl_waitq,
2878                              thread_is_running(thread) ||
2879                              thread_is_stopped(thread),
2880                              &lwi);
2881
2882         GOTO(out, rc = 0);
2883
2884 out:
2885         mutex_unlock(&lfsck->ml_mutex);
2886         return (rc < 0 ? rc : 0);
2887 }
2888
2889 int mdd_lfsck_stop(const struct lu_env *env, struct md_lfsck *lfsck,
2890                    bool pause)
2891 {
2892         struct ptlrpc_thread *thread = &lfsck->ml_thread;
2893         struct l_wait_info    lwi    = { 0 };
2894         ENTRY;
2895
2896         if (!lfsck->ml_initialized)
2897                 RETURN(0);
2898
2899         mutex_lock(&lfsck->ml_mutex);
2900         spin_lock(&lfsck->ml_lock);
2901         if (thread_is_init(thread) || thread_is_stopped(thread)) {
2902                 spin_unlock(&lfsck->ml_lock);
2903                 mutex_unlock(&lfsck->ml_mutex);
2904                 RETURN(-EALREADY);
2905         }
2906
2907         if (pause)
2908                 lfsck->ml_paused = 1;
2909         thread_set_flags(thread, SVC_STOPPING);
2910         /* The LFSCK thread may be sleeping on low layer wait queue,
2911          * wake it up. */
2912         if (likely(lfsck->ml_di_oit != NULL))
2913                 lfsck->ml_obj_oit->do_index_ops->dio_it.put(env,
2914                                                             lfsck->ml_di_oit);
2915         spin_unlock(&lfsck->ml_lock);
2916
2917         cfs_waitq_broadcast(&thread->t_ctl_waitq);
2918         l_wait_event(thread->t_ctl_waitq,
2919                      thread_is_stopped(thread),
2920                      &lwi);
2921         mutex_unlock(&lfsck->ml_mutex);
2922
2923         RETURN(0);
2924 }
2925
2926 static const struct lu_fid lfsck_it_fid = { .f_seq = FID_SEQ_LOCAL_FILE,
2927                                             .f_oid = OTABLE_IT_OID,
2928                                             .f_ver = 0 };
2929
2930 int mdd_lfsck_setup(const struct lu_env *env, struct mdd_device *mdd)
2931 {
2932         struct md_lfsck  *lfsck = &mdd->mdd_lfsck;
2933         struct dt_object *obj;
2934         int               rc;
2935         ENTRY;
2936
2937         LASSERT(!lfsck->ml_initialized);
2938
2939         lfsck->ml_initialized = 1;
2940         mutex_init(&lfsck->ml_mutex);
2941         spin_lock_init(&lfsck->ml_lock);
2942         CFS_INIT_LIST_HEAD(&lfsck->ml_list_scan);
2943         CFS_INIT_LIST_HEAD(&lfsck->ml_list_dir);
2944         CFS_INIT_LIST_HEAD(&lfsck->ml_list_double_scan);
2945         CFS_INIT_LIST_HEAD(&lfsck->ml_list_idle);
2946         cfs_waitq_init(&lfsck->ml_thread.t_ctl_waitq);
2947
2948         obj = dt_locate(env, mdd->mdd_bottom, &lfsck_it_fid);
2949         if (IS_ERR(obj))
2950                 RETURN(PTR_ERR(obj));
2951
2952         lfsck->ml_obj_oit = obj;
2953         rc = obj->do_ops->do_index_try(env, obj, &dt_otable_features);
2954         if (rc != 0) {
2955                 if (rc == -ENOTSUPP)
2956                         rc = 0;
2957
2958                 RETURN(rc);
2959         }
2960
2961         obj = dt_store_open(env, mdd->mdd_bottom, "", lfsck_bookmark_name,
2962                             &mdd_env_info(env)->mti_fid);
2963         if (IS_ERR(obj))
2964                 RETURN(PTR_ERR(obj));
2965
2966         lfsck->ml_bookmark_obj = obj;
2967         rc = mdd_lfsck_bookmark_load(env, lfsck);
2968         if (rc == -ENODATA)
2969                 rc = mdd_lfsck_bookmark_init(env, lfsck);
2970         if (rc != 0)
2971                 RETURN(rc);
2972
2973         rc = mdd_lfsck_namespace_setup(env, lfsck);
2974         /* XXX: LFSCK components initialization to be added here. */
2975
2976         RETURN(rc);
2977 }
2978
2979 void mdd_lfsck_cleanup(const struct lu_env *env, struct mdd_device *mdd)
2980 {
2981         struct md_lfsck         *lfsck  = &mdd->mdd_lfsck;
2982         struct ptlrpc_thread    *thread = &lfsck->ml_thread;
2983         struct lfsck_component  *com;
2984
2985         if (!lfsck->ml_initialized)
2986                 return;
2987
2988         LASSERT(thread_is_init(thread) || thread_is_stopped(thread));
2989
2990         if (lfsck->ml_obj_oit != NULL) {
2991                 lu_object_put(env, &lfsck->ml_obj_oit->do_lu);
2992                 lfsck->ml_obj_oit = NULL;
2993         }
2994
2995         LASSERT(lfsck->ml_obj_dir == NULL);
2996
2997         if (lfsck->ml_bookmark_obj != NULL) {
2998                 lu_object_put(env, &lfsck->ml_bookmark_obj->do_lu);
2999                 lfsck->ml_bookmark_obj = NULL;
3000         }
3001
3002         while (!cfs_list_empty(&lfsck->ml_list_scan)) {
3003                 com = cfs_list_entry(lfsck->ml_list_scan.next,
3004                                      struct lfsck_component,
3005                                      lc_link);
3006                 mdd_lfsck_component_cleanup(env, com);
3007         }
3008
3009         LASSERT(cfs_list_empty(&lfsck->ml_list_dir));
3010
3011         while (!cfs_list_empty(&lfsck->ml_list_double_scan)) {
3012                 com = cfs_list_entry(lfsck->ml_list_double_scan.next,
3013                                      struct lfsck_component,
3014                                      lc_link);
3015                 mdd_lfsck_component_cleanup(env, com);
3016         }
3017
3018         while (!cfs_list_empty(&lfsck->ml_list_idle)) {
3019                 com = cfs_list_entry(lfsck->ml_list_idle.next,
3020                                      struct lfsck_component,
3021                                      lc_link);
3022                 mdd_lfsck_component_cleanup(env, com);
3023         }
3024 }