fs/ocfs2/super.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /*
   3  * super.c
   4  *
   5  * load/unload driver, mount/dismount volumes
   6  *
   7  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   8  */
   9
  10 #include <linux/module.h>
  11 #include <linux/fs.h>
  12 #include <linux/types.h>
  13 #include <linux/slab.h>
  14 #include <linux/highmem.h>
  15 #include <linux/init.h>
  16 #include <linux/random.h>
  17 #include <linux/statfs.h>
  18 #include <linux/moduleparam.h>
  19 #include <linux/blkdev.h>
  20 #include <linux/socket.h>
  21 #include <linux/inet.h>
  22 #include <linux/parser.h>
  23 #include <linux/crc32.h>
  24 #include <linux/debugfs.h>
  25 #include <linux/mount.h>
  26 #include <linux/seq_file.h>
  27 #include <linux/quotaops.h>
  28 #include <linux/cleancache.h>
  29 #include <linux/signal.h>
  30
  31 #define CREATE_TRACE_POINTS
  32 #include "ocfs2_trace.h"
  33
  34 #include <cluster/masklog.h>
  35
  36 #include "ocfs2.h"
  37
  38 /* this should be the only file to include a version 1 header */
  39 #include "ocfs1_fs_compat.h"
  40
  41 #include "alloc.h"
  42 #include "aops.h"
  43 #include "blockcheck.h"
  44 #include "dlmglue.h"
  45 #include "export.h"
  46 #include "extent_map.h"
  47 #include "heartbeat.h"
  48 #include "inode.h"
  49 #include "journal.h"
  50 #include "localalloc.h"
  51 #include "namei.h"
  52 #include "slot_map.h"
  53 #include "super.h"
  54 #include "sysfile.h"
  55 #include "uptodate.h"
  56 #include "xattr.h"
  57 #include "quota.h"
  58 #include "refcounttree.h"
  59 #include "suballoc.h"
  60
  61 #include "buffer_head_io.h"
  62 #include "filecheck.h"
  63
  64 static struct kmem_cache *ocfs2_inode_cachep;
  65 struct kmem_cache *ocfs2_dquot_cachep;
  66 struct kmem_cache *ocfs2_qf_chunk_cachep;
  67
  68 static struct dentry *ocfs2_debugfs_root;
  69
  70 MODULE_AUTHOR("Oracle");
  71 MODULE_LICENSE("GPL");
  72 MODULE_DESCRIPTION("OCFS2 cluster file system");
  73
  74 struct mount_options
  75 {
  76         unsigned long   commit_interval;
  77         unsigned long   mount_opt;
  78         unsigned int    atime_quantum;
  79         unsigned short  slot;
  80         int             localalloc_opt;
  81         unsigned int    resv_level;
  82         int             dir_resv_level;
  83         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  84 };
  85
  86 static int ocfs2_parse_options(struct super_block *sb, char *options,
  87                                struct mount_options *mopt,
  88                                int is_remount);
  89 static int ocfs2_check_set_options(struct super_block *sb,
  90                                    struct mount_options *options);
  91 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
  92 static void ocfs2_put_super(struct super_block *sb);
  93 static int ocfs2_mount_volume(struct super_block *sb);
  94 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
  95 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
  96 static int ocfs2_initialize_mem_caches(void);
  97 static void ocfs2_free_mem_caches(void);
  98 static void ocfs2_delete_osb(struct ocfs2_super *osb);
  99
 100 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 101
 102 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 103
 104 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 105 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 106 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 107 static int ocfs2_check_volume(struct ocfs2_super *osb);
 108 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 109                                struct buffer_head *bh,
 110                                u32 sectsize,
 111                                struct ocfs2_blockcheck_stats *stats);
 112 static int ocfs2_initialize_super(struct super_block *sb,
 113                                   struct buffer_head *bh,
 114                                   int sector_size,
 115                                   struct ocfs2_blockcheck_stats *stats);
 116 static int ocfs2_get_sector(struct super_block *sb,
 117                             struct buffer_head **bh,
 118                             int block,
 119                             int sect_size);
 120 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 121 static void ocfs2_free_inode(struct inode *inode);
 122 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 123 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 124 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 125
 126 static struct dquot **ocfs2_get_dquots(struct inode *inode)
 127 {
 128         return OCFS2_I(inode)->i_dquot;
 129 }
 130
 131 static const struct super_operations ocfs2_sops = {
 132         .statfs         = ocfs2_statfs,
 133         .alloc_inode    = ocfs2_alloc_inode,
 134         .free_inode     = ocfs2_free_inode,
 135         .drop_inode     = ocfs2_drop_inode,
 136         .evict_inode    = ocfs2_evict_inode,
 137         .sync_fs        = ocfs2_sync_fs,
 138         .put_super      = ocfs2_put_super,
 139         .remount_fs     = ocfs2_remount,
 140         .show_options   = ocfs2_show_options,
 141         .quota_read     = ocfs2_quota_read,
 142         .quota_write    = ocfs2_quota_write,
 143         .get_dquots     = ocfs2_get_dquots,
 144 };
 145
 146 enum {
 147         Opt_barrier,
 148         Opt_err_panic,
 149         Opt_err_ro,
 150         Opt_intr,
 151         Opt_nointr,
 152         Opt_hb_none,
 153         Opt_hb_local,
 154         Opt_hb_global,
 155         Opt_data_ordered,
 156         Opt_data_writeback,
 157         Opt_atime_quantum,
 158         Opt_slot,
 159         Opt_commit,
 160         Opt_localalloc,
 161         Opt_localflocks,
 162         Opt_stack,
 163         Opt_user_xattr,
 164         Opt_nouser_xattr,
 165         Opt_inode64,
 166         Opt_acl,
 167         Opt_noacl,
 168         Opt_usrquota,
 169         Opt_grpquota,
 170         Opt_coherency_buffered,
 171         Opt_coherency_full,
 172         Opt_resv_level,
 173         Opt_dir_resv_level,
 174         Opt_journal_async_commit,
 175         Opt_err_cont,
 176         Opt_nocluster,
 177         Opt_err,
 178 };
 179
 180 static const match_table_t tokens = {
 181         {Opt_barrier, "barrier=%u"},
 182         {Opt_err_panic, "errors=panic"},
 183         {Opt_err_ro, "errors=remount-ro"},
 184         {Opt_intr, "intr"},
 185         {Opt_nointr, "nointr"},
 186         {Opt_hb_none, OCFS2_HB_NONE},
 187         {Opt_hb_local, OCFS2_HB_LOCAL},
 188         {Opt_hb_global, OCFS2_HB_GLOBAL},
 189         {Opt_data_ordered, "data=ordered"},
 190         {Opt_data_writeback, "data=writeback"},
 191         {Opt_atime_quantum, "atime_quantum=%u"},
 192         {Opt_slot, "preferred_slot=%u"},
 193         {Opt_commit, "commit=%u"},
 194         {Opt_localalloc, "localalloc=%d"},
 195         {Opt_localflocks, "localflocks"},
 196         {Opt_stack, "cluster_stack=%s"},
 197         {Opt_user_xattr, "user_xattr"},
 198         {Opt_nouser_xattr, "nouser_xattr"},
 199         {Opt_inode64, "inode64"},
 200         {Opt_acl, "acl"},
 201         {Opt_noacl, "noacl"},
 202         {Opt_usrquota, "usrquota"},
 203         {Opt_grpquota, "grpquota"},
 204         {Opt_coherency_buffered, "coherency=buffered"},
 205         {Opt_coherency_full, "coherency=full"},
 206         {Opt_resv_level, "resv_level=%u"},
 207         {Opt_dir_resv_level, "dir_resv_level=%u"},
 208         {Opt_journal_async_commit, "journal_async_commit"},
 209         {Opt_err_cont, "errors=continue"},
 210         {Opt_nocluster, "nocluster"},
 211         {Opt_err, NULL}
 212 };
 213
 214 #ifdef CONFIG_DEBUG_FS
 215 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 216 {
 217         struct ocfs2_cluster_connection *cconn = osb->cconn;
 218         struct ocfs2_recovery_map *rm = osb->recovery_map;
 219         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 220         int i, out = 0;
 221         unsigned long flags;
 222
 223         out += scnprintf(buf + out, len - out,
 224                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 225                         "Device", osb->dev_str, osb->uuid_str,
 226                         osb->fs_generation, osb->vol_label);
 227
 228         out += scnprintf(buf + out, len - out,
 229                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 230                         atomic_read(&osb->vol_state), osb->osb_flags);
 231
 232         out += scnprintf(buf + out, len - out,
 233                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 234                         osb->sb->s_blocksize, osb->s_clustersize);
 235
 236         out += scnprintf(buf + out, len - out,
 237                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 238                         "ROcompat: 0x%X\n",
 239                         "Features", osb->s_feature_compat,
 240                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 241
 242         out += scnprintf(buf + out, len - out,
 243                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 244                         osb->s_mount_opt, osb->s_atime_quantum);
 245
 246         if (cconn) {
 247                 out += scnprintf(buf + out, len - out,
 248                                 "%10s => Stack: %s  Name: %*s  "
 249                                 "Version: %d.%d\n", "Cluster",
 250                                 (*osb->osb_cluster_stack == '\0' ?
 251                                  "o2cb" : osb->osb_cluster_stack),
 252                                 cconn->cc_namelen, cconn->cc_name,
 253                                 cconn->cc_version.pv_major,
 254                                 cconn->cc_version.pv_minor);
 255         }
 256
 257         spin_lock_irqsave(&osb->dc_task_lock, flags);
 258         out += scnprintf(buf + out, len - out,
 259                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 260                         "WorkSeq: %lu\n", "DownCnvt",
 261                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 262                         osb->blocked_lock_count, osb->dc_wake_sequence,
 263                         osb->dc_work_sequence);
 264         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 265
 266         spin_lock(&osb->osb_lock);
 267         out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 268                         "Recovery",
 269                         (osb->recovery_thread_task ?
 270                          task_pid_nr(osb->recovery_thread_task) : -1));
 271         if (rm->rm_used == 0)
 272                 out += scnprintf(buf + out, len - out, " None\n");
 273         else {
 274                 for (i = 0; i < rm->rm_used; i++)
 275                         out += scnprintf(buf + out, len - out, " %d",
 276                                         rm->rm_entries[i]);
 277                 out += scnprintf(buf + out, len - out, "\n");
 278         }
 279         spin_unlock(&osb->osb_lock);
 280
 281         out += scnprintf(buf + out, len - out,
 282                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
 283                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 284                         osb->osb_commit_interval);
 285
 286         out += scnprintf(buf + out, len - out,
 287                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 288                         "Journal", osb->journal->j_state,
 289                         osb->journal->j_trans_id,
 290                         atomic_read(&osb->journal->j_num_trans));
 291
 292         out += scnprintf(buf + out, len - out,
 293                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 294                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 295                         "Stats",
 296                         atomic_read(&osb->alloc_stats.bitmap_data),
 297                         atomic_read(&osb->alloc_stats.local_data),
 298                         atomic_read(&osb->alloc_stats.bg_allocs),
 299                         atomic_read(&osb->alloc_stats.moves),
 300                         atomic_read(&osb->alloc_stats.bg_extends));
 301
 302         out += scnprintf(buf + out, len - out,
 303                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 304                         "Default: %u bits\n",
 305                         "LocalAlloc", osb->local_alloc_state,
 306                         (unsigned long long)osb->la_last_gd,
 307                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 308
 309         spin_lock(&osb->osb_lock);
 310         out += scnprintf(buf + out, len - out,
 311                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 312                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 313                         osb->s_inode_steal_slot,
 314                         atomic_read(&osb->s_num_inodes_stolen),
 315                         osb->s_meta_steal_slot,
 316                         atomic_read(&osb->s_num_meta_stolen));
 317         spin_unlock(&osb->osb_lock);
 318
 319         out += scnprintf(buf + out, len - out, "OrphanScan => ");
 320         out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 321                         os->os_count, os->os_seqno);
 322         out += scnprintf(buf + out, len - out, " Last Scan: ");
 323         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 324                 out += scnprintf(buf + out, len - out, "Disabled\n");
 325         else
 326                 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 327                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
 328
 329         out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 330                         "Slots", "Num", "RecoGen");
 331         for (i = 0; i < osb->max_slots; ++i) {
 332                 out += scnprintf(buf + out, len - out,
 333                                 "%10s  %c %3d  %10d\n",
 334                                 " ",
 335                                 (i == osb->slot_num ? '*' : ' '),
 336                                 i, osb->slot_recovery_generations[i]);
 337         }
 338
 339         return out;
 340 }
 341
 342 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 343 {
 344         struct ocfs2_super *osb = inode->i_private;
 345         char *buf = NULL;
 346
 347         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 348         if (!buf)
 349                 goto bail;
 350
 351         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 352
 353         file->private_data = buf;
 354
 355         return 0;
 356 bail:
 357         return -ENOMEM;
 358 }
 359
 360 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 361 {
 362         kfree(file->private_data);
 363         return 0;
 364 }
 365
 366 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 367                                 size_t nbytes, loff_t *ppos)
 368 {
 369         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 370                                        i_size_read(file->f_mapping->host));
 371 }
 372 #else
 373 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 374 {
 375         return 0;
 376 }
 377 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 382                                 size_t nbytes, loff_t *ppos)
 383 {
 384         return 0;
 385 }
 386 #endif  /* CONFIG_DEBUG_FS */
 387
 388 static const struct file_operations ocfs2_osb_debug_fops = {
 389         .open =         ocfs2_osb_debug_open,
 390         .release =      ocfs2_debug_release,
 391         .read =         ocfs2_debug_read,
 392         .llseek =       generic_file_llseek,
 393 };
 394
 395 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 396 {
 397         int status;
 398         tid_t target;
 399         struct ocfs2_super *osb = OCFS2_SB(sb);
 400
 401         if (ocfs2_is_hard_readonly(osb))
 402                 return -EROFS;
 403
 404         if (wait) {
 405                 status = ocfs2_flush_truncate_log(osb);
 406                 if (status < 0)
 407                         mlog_errno(status);
 408         } else {
 409                 ocfs2_schedule_truncate_log_flush(osb, 0);
 410         }
 411
 412         if (jbd2_journal_start_commit(osb->journal->j_journal,
 413                                       &target)) {
 414                 if (wait)
 415                         jbd2_log_wait_commit(osb->journal->j_journal,
 416                                              target);
 417         }
 418         return 0;
 419 }
 420
 421 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 422 {
 423         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 424             && (ino == USER_QUOTA_SYSTEM_INODE
 425                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 426                 return 0;
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 428             && (ino == GROUP_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         return 1;
 432 }
 433
 434 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 435 {
 436         struct inode *new = NULL;
 437         int status = 0;
 438         int i;
 439
 440         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 441         if (IS_ERR(new)) {
 442                 status = PTR_ERR(new);
 443                 mlog_errno(status);
 444                 goto bail;
 445         }
 446         osb->root_inode = new;
 447
 448         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 449         if (IS_ERR(new)) {
 450                 status = PTR_ERR(new);
 451                 mlog_errno(status);
 452                 goto bail;
 453         }
 454         osb->sys_root_inode = new;
 455
 456         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 457              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 458                 if (!ocfs2_need_system_inode(osb, i))
 459                         continue;
 460                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 461                 if (!new) {
 462                         ocfs2_release_system_inodes(osb);
 463                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 464                         mlog_errno(status);
 465                         mlog(ML_ERROR, "Unable to load system inode %d, "
 466                              "possibly corrupt fs?", i);
 467                         goto bail;
 468                 }
 469                 // the array now has one ref, so drop this one
 470                 iput(new);
 471         }
 472
 473 bail:
 474         if (status)
 475                 mlog_errno(status);
 476         return status;
 477 }
 478
 479 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 480 {
 481         struct inode *new = NULL;
 482         int status = 0;
 483         int i;
 484
 485         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 486              i < NUM_SYSTEM_INODES;
 487              i++) {
 488                 if (!ocfs2_need_system_inode(osb, i))
 489                         continue;
 490                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 491                 if (!new) {
 492                         ocfs2_release_system_inodes(osb);
 493                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 494                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 495                              status, i, osb->slot_num);
 496                         goto bail;
 497                 }
 498                 /* the array now has one ref, so drop this one */
 499                 iput(new);
 500         }
 501
 502 bail:
 503         if (status)
 504                 mlog_errno(status);
 505         return status;
 506 }
 507
 508 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 509 {
 510         int i;
 511         struct inode *inode;
 512
 513         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 514                 inode = osb->global_system_inodes[i];
 515                 if (inode) {
 516                         iput(inode);
 517                         osb->global_system_inodes[i] = NULL;
 518                 }
 519         }
 520
 521         inode = osb->sys_root_inode;
 522         if (inode) {
 523                 iput(inode);
 524                 osb->sys_root_inode = NULL;
 525         }
 526
 527         inode = osb->root_inode;
 528         if (inode) {
 529                 iput(inode);
 530                 osb->root_inode = NULL;
 531         }
 532
 533         if (!osb->local_system_inodes)
 534                 return;
 535
 536         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 537                 if (osb->local_system_inodes[i]) {
 538                         iput(osb->local_system_inodes[i]);
 539                         osb->local_system_inodes[i] = NULL;
 540                 }
 541         }
 542
 543         kfree(osb->local_system_inodes);
 544         osb->local_system_inodes = NULL;
 545 }
 546
 547 /* We're allocating fs objects, use GFP_NOFS */
 548 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 549 {
 550         struct ocfs2_inode_info *oi;
 551
 552         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 553         if (!oi)
 554                 return NULL;
 555
 556         oi->i_sync_tid = 0;
 557         oi->i_datasync_tid = 0;
 558         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 559
 560         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 561         return &oi->vfs_inode;
 562 }
 563
 564 static void ocfs2_free_inode(struct inode *inode)
 565 {
 566         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 567 }
 568
 569 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 570                                                 unsigned int cbits)
 571 {
 572         unsigned int bytes = 1 << cbits;
 573         unsigned int trim = bytes;
 574         unsigned int bitshift = 32;
 575
 576         /*
 577          * i_size and all block offsets in ocfs2 are always 64 bits
 578          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 579          * 64 bit platforms, cluster size will be the limiting factor.
 580          */
 581
 582 #if BITS_PER_LONG == 32
 583         BUILD_BUG_ON(sizeof(sector_t) != 8);
 584         /*
 585          * We might be limited by page cache size.
 586          */
 587         if (bytes > PAGE_SIZE) {
 588                 bytes = PAGE_SIZE;
 589                 trim = 1;
 590                 /*
 591                  * Shift by 31 here so that we don't get larger than
 592                  * MAX_LFS_FILESIZE
 593                  */
 594                 bitshift = 31;
 595         }
 596 #endif
 597
 598         /*
 599          * Trim by a whole cluster when we can actually approach the
 600          * on-disk limits. Otherwise we can overflow i_clusters when
 601          * an extent start is at the max offset.
 602          */
 603         return (((unsigned long long)bytes) << bitshift) - trim;
 604 }
 605
 606 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 607 {
 608         int incompat_features;
 609         int ret = 0;
 610         struct mount_options parsed_options;
 611         struct ocfs2_super *osb = OCFS2_SB(sb);
 612         u32 tmp;
 613
 614         sync_filesystem(sb);
 615
 616         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 617             !ocfs2_check_set_options(sb, &parsed_options)) {
 618                 ret = -EINVAL;
 619                 goto out;
 620         }
 621
 622         tmp = OCFS2_MOUNT_NOCLUSTER;
 623         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 624                 ret = -EINVAL;
 625                 mlog(ML_ERROR, "Cannot change nocluster option on remount\n");
 626                 goto out;
 627         }
 628
 629         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 630                 OCFS2_MOUNT_HB_NONE;
 631         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 632                 ret = -EINVAL;
 633                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 634                 goto out;
 635         }
 636
 637         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 638             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 639                 ret = -EINVAL;
 640                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 641                 goto out;
 642         }
 643
 644         /* Probably don't want this on remount; it might
 645          * mess with other nodes */
 646         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 647             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 648                 ret = -EINVAL;
 649                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 650                 goto out;
 651         }
 652
 653         /* We're going to/from readonly mode. */
 654         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 655                 /* Disable quota accounting before remounting RO */
 656                 if (*flags & SB_RDONLY) {
 657                         ret = ocfs2_susp_quotas(osb, 0);
 658                         if (ret < 0)
 659                                 goto out;
 660                 }
 661                 /* Lock here so the check of HARD_RO and the potential
 662                  * setting of SOFT_RO is atomic. */
 663                 spin_lock(&osb->osb_lock);
 664                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 665                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 666                         ret = -EROFS;
 667                         goto unlock_osb;
 668                 }
 669
 670                 if (*flags & SB_RDONLY) {
 671                         sb->s_flags |= SB_RDONLY;
 672                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 673                 } else {
 674                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 675                                 mlog(ML_ERROR, "Cannot remount RDWR "
 676                                      "filesystem due to previous errors.\n");
 677                                 ret = -EROFS;
 678                                 goto unlock_osb;
 679                         }
 680                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 681                         if (incompat_features) {
 682                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 683                                      "of unsupported optional features "
 684                                      "(%x).\n", incompat_features);
 685                                 ret = -EINVAL;
 686                                 goto unlock_osb;
 687                         }
 688                         sb->s_flags &= ~SB_RDONLY;
 689                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 690                 }
 691                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 692 unlock_osb:
 693                 spin_unlock(&osb->osb_lock);
 694                 /* Enable quota accounting after remounting RW */
 695                 if (!ret && !(*flags & SB_RDONLY)) {
 696                         if (sb_any_quota_suspended(sb))
 697                                 ret = ocfs2_susp_quotas(osb, 1);
 698                         else
 699                                 ret = ocfs2_enable_quotas(osb);
 700                         if (ret < 0) {
 701                                 /* Return back changes... */
 702                                 spin_lock(&osb->osb_lock);
 703                                 sb->s_flags |= SB_RDONLY;
 704                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 705                                 spin_unlock(&osb->osb_lock);
 706                                 goto out;
 707                         }
 708                 }
 709         }
 710
 711         if (!ret) {
 712                 /* Only save off the new mount options in case of a successful
 713                  * remount. */
 714                 osb->s_mount_opt = parsed_options.mount_opt;
 715                 osb->s_atime_quantum = parsed_options.atime_quantum;
 716                 osb->preferred_slot = parsed_options.slot;
 717                 if (parsed_options.commit_interval)
 718                         osb->osb_commit_interval = parsed_options.commit_interval;
 719
 720                 if (!ocfs2_is_hard_readonly(osb))
 721                         ocfs2_set_journal_params(osb);
 722
 723                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 724                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 725                                                         SB_POSIXACL : 0);
 726         }
 727 out:
 728         return ret;
 729 }
 730
 731 static int ocfs2_sb_probe(struct super_block *sb,
 732                           struct buffer_head **bh,
 733                           int *sector_size,
 734                           struct ocfs2_blockcheck_stats *stats)
 735 {
 736         int status, tmpstat;
 737         struct ocfs1_vol_disk_hdr *hdr;
 738         struct ocfs2_dinode *di;
 739         int blksize;
 740
 741         *bh = NULL;
 742
 743         /* may be > 512 */
 744         *sector_size = bdev_logical_block_size(sb->s_bdev);
 745         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 746                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 747                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 748                 status = -EINVAL;
 749                 goto bail;
 750         }
 751
 752         /* Can this really happen? */
 753         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 754                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 755
 756         /* check block zero for old format */
 757         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 758         if (status < 0) {
 759                 mlog_errno(status);
 760                 goto bail;
 761         }
 762         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 763         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 764                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 765                      hdr->major_version, hdr->minor_version);
 766                 status = -EINVAL;
 767         }
 768         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 769                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 770                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 771                      hdr->signature);
 772                 status = -EINVAL;
 773         }
 774         brelse(*bh);
 775         *bh = NULL;
 776         if (status < 0) {
 777                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 778                      "upgraded before mounting with ocfs v2\n");
 779                 goto bail;
 780         }
 781
 782         /*
 783          * Now check at magic offset for 512, 1024, 2048, 4096
 784          * blocksizes.  4096 is the maximum blocksize because it is
 785          * the minimum clustersize.
 786          */
 787         status = -EINVAL;
 788         for (blksize = *sector_size;
 789              blksize <= OCFS2_MAX_BLOCKSIZE;
 790              blksize <<= 1) {
 791                 tmpstat = ocfs2_get_sector(sb, bh,
 792                                            OCFS2_SUPER_BLOCK_BLKNO,
 793                                            blksize);
 794                 if (tmpstat < 0) {
 795                         status = tmpstat;
 796                         mlog_errno(status);
 797                         break;
 798                 }
 799                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 800                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 801                 spin_lock_init(&stats->b_lock);
 802                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 803                 if (tmpstat < 0) {
 804                         brelse(*bh);
 805                         *bh = NULL;
 806                 }
 807                 if (tmpstat != -EAGAIN) {
 808                         status = tmpstat;
 809                         break;
 810                 }
 811         }
 812
 813 bail:
 814         return status;
 815 }
 816
 817 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 818 {
 819         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 820
 821         if (osb->s_mount_opt & hb_enabled) {
 822                 if (ocfs2_mount_local(osb)) {
 823                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 824                              "mounted device.\n");
 825                         return -EINVAL;
 826                 }
 827                 if (ocfs2_userspace_stack(osb)) {
 828                         mlog(ML_ERROR, "Userspace stack expected, but "
 829                              "o2cb heartbeat arguments passed to mount\n");
 830                         return -EINVAL;
 831                 }
 832                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 833                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 834                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 835                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 836                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 837                         return -EINVAL;
 838                 }
 839         }
 840
 841         if (!(osb->s_mount_opt & hb_enabled)) {
 842                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 843                     !ocfs2_userspace_stack(osb)) {
 844                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 845                              "a read-write clustered device.\n");
 846                         return -EINVAL;
 847                 }
 848         }
 849
 850         return 0;
 851 }
 852
 853 /*
 854  * If we're using a userspace stack, mount should have passed
 855  * a name that matches the disk.  If not, mount should not
 856  * have passed a stack.
 857  */
 858 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 859                                         struct mount_options *mopt)
 860 {
 861         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 862                 mlog(ML_ERROR,
 863                      "cluster stack passed to mount, but this filesystem "
 864                      "does not support it\n");
 865                 return -EINVAL;
 866         }
 867
 868         if (ocfs2_userspace_stack(osb) &&
 869             !(osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
 870             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 871                     OCFS2_STACK_LABEL_LEN)) {
 872                 mlog(ML_ERROR,
 873                      "cluster stack passed to mount (\"%s\") does not "
 874                      "match the filesystem (\"%s\")\n",
 875                      mopt->cluster_stack,
 876                      osb->osb_cluster_stack);
 877                 return -EINVAL;
 878         }
 879
 880         return 0;
 881 }
 882
 883 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 884 {
 885         int type;
 886         struct super_block *sb = osb->sb;
 887         unsigned int feature[OCFS2_MAXQUOTAS] = {
 888                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 889                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 890         int status = 0;
 891
 892         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 893                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 894                         continue;
 895                 if (unsuspend)
 896                         status = dquot_resume(sb, type);
 897                 else {
 898                         struct ocfs2_mem_dqinfo *oinfo;
 899
 900                         /* Cancel periodic syncing before suspending */
 901                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 902                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 903                         status = dquot_suspend(sb, type);
 904                 }
 905                 if (status < 0)
 906                         break;
 907         }
 908         if (status < 0)
 909                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 910                      "remount (error = %d).\n", status);
 911         return status;
 912 }
 913
 914 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 915 {
 916         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 917         struct super_block *sb = osb->sb;
 918         unsigned int feature[OCFS2_MAXQUOTAS] = {
 919                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 920                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 921         unsigned int ino[OCFS2_MAXQUOTAS] = {
 922                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
 923                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 924         int status;
 925         int type;
 926
 927         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 928         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 929                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 930                         continue;
 931                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 932                                                         osb->slot_num);
 933                 if (!inode[type]) {
 934                         status = -ENOENT;
 935                         goto out_quota_off;
 936                 }
 937                 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 938                                                 DQUOT_USAGE_ENABLED);
 939                 if (status < 0)
 940                         goto out_quota_off;
 941         }
 942
 943         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 944                 iput(inode[type]);
 945         return 0;
 946 out_quota_off:
 947         ocfs2_disable_quotas(osb);
 948         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 949                 iput(inode[type]);
 950         mlog_errno(status);
 951         return status;
 952 }
 953
 954 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 955 {
 956         int type;
 957         struct inode *inode;
 958         struct super_block *sb = osb->sb;
 959         struct ocfs2_mem_dqinfo *oinfo;
 960
 961         /* We mostly ignore errors in this function because there's not much
 962          * we can do when we see them */
 963         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 964                 if (!sb_has_quota_loaded(sb, type))
 965                         continue;
 966                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 967                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 968                 inode = igrab(sb->s_dquot.files[type]);
 969                 /* Turn off quotas. This will remove all dquot structures from
 970                  * memory and so they will be automatically synced to global
 971                  * quota files */
 972                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 973                                         DQUOT_LIMITS_ENABLED);
 974                 iput(inode);
 975         }
 976 }
 977
 978 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 979 {
 980         struct dentry *root;
 981         int status, sector_size;
 982         struct mount_options parsed_options;
 983         struct inode *inode = NULL;
 984         struct ocfs2_super *osb = NULL;
 985         struct buffer_head *bh = NULL;
 986         char nodestr[12];
 987         struct ocfs2_blockcheck_stats stats;
 988
 989         trace_ocfs2_fill_super(sb, data, silent);
 990
 991         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 992                 status = -EINVAL;
 993                 goto read_super_error;
 994         }
 995
 996         /* probe for superblock */
 997         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
 998         if (status < 0) {
 999                 mlog(ML_ERROR, "superblock probe failed!\n");
1000                 goto read_super_error;
1001         }
1002
1003         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1004         osb = OCFS2_SB(sb);
1005         if (status < 0) {
1006                 mlog_errno(status);
1007                 goto read_super_error;
1008         }
1009         brelse(bh);
1010         bh = NULL;
1011
1012         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1013                 status = -EINVAL;
1014                 goto read_super_error;
1015         }
1016         osb->s_mount_opt = parsed_options.mount_opt;
1017         osb->s_atime_quantum = parsed_options.atime_quantum;
1018         osb->preferred_slot = parsed_options.slot;
1019         osb->osb_commit_interval = parsed_options.commit_interval;
1020
1021         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1022         osb->osb_resv_level = parsed_options.resv_level;
1023         osb->osb_dir_resv_level = parsed_options.resv_level;
1024         if (parsed_options.dir_resv_level == -1)
1025                 osb->osb_dir_resv_level = parsed_options.resv_level;
1026         else
1027                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1028
1029         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1030         if (status)
1031                 goto read_super_error;
1032
1033         sb->s_magic = OCFS2_SUPER_MAGIC;
1034
1035         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1036                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1037
1038         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1039          * heartbeat=none */
1040         if (bdev_read_only(sb->s_bdev)) {
1041                 if (!sb_rdonly(sb)) {
1042                         status = -EACCES;
1043                         mlog(ML_ERROR, "Readonly device detected but readonly "
1044                              "mount was not specified.\n");
1045                         goto read_super_error;
1046                 }
1047
1048                 /* You should not be able to start a local heartbeat
1049                  * on a readonly device. */
1050                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1051                         status = -EROFS;
1052                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1053                              "device.\n");
1054                         goto read_super_error;
1055                 }
1056
1057                 status = ocfs2_check_journals_nolocks(osb);
1058                 if (status < 0) {
1059                         if (status == -EROFS)
1060                                 mlog(ML_ERROR, "Recovery required on readonly "
1061                                      "file system, but write access is "
1062                                      "unavailable.\n");
1063                         else
1064                                 mlog_errno(status);
1065                         goto read_super_error;
1066                 }
1067
1068                 ocfs2_set_ro_flag(osb, 1);
1069
1070                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1071                        "Cluster services will not be used for this mount. "
1072                        "Recovery will be skipped.\n", osb->dev_str);
1073         }
1074
1075         if (!ocfs2_is_hard_readonly(osb)) {
1076                 if (sb_rdonly(sb))
1077                         ocfs2_set_ro_flag(osb, 0);
1078         }
1079
1080         status = ocfs2_verify_heartbeat(osb);
1081         if (status < 0) {
1082                 mlog_errno(status);
1083                 goto read_super_error;
1084         }
1085
1086         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1087                                                  ocfs2_debugfs_root);
1088
1089         debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1090                             osb, &ocfs2_osb_debug_fops);
1091
1092         if (ocfs2_meta_ecc(osb))
1093                 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1094                                                         osb->osb_debug_root);
1095
1096         status = ocfs2_mount_volume(sb);
1097         if (status < 0)
1098                 goto read_super_error;
1099
1100         if (osb->root_inode)
1101                 inode = igrab(osb->root_inode);
1102
1103         if (!inode) {
1104                 status = -EIO;
1105                 mlog_errno(status);
1106                 goto read_super_error;
1107         }
1108
1109         root = d_make_root(inode);
1110         if (!root) {
1111                 status = -ENOMEM;
1112                 mlog_errno(status);
1113                 goto read_super_error;
1114         }
1115
1116         sb->s_root = root;
1117
1118         ocfs2_complete_mount_recovery(osb);
1119
1120         osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1121                                                 &ocfs2_kset->kobj);
1122         if (!osb->osb_dev_kset) {
1123                 status = -ENOMEM;
1124                 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1125                 goto read_super_error;
1126         }
1127
1128         /* Create filecheck sysfs related directories/files at
1129          * /sys/fs/ocfs2/<devname>/filecheck */
1130         if (ocfs2_filecheck_create_sysfs(osb)) {
1131                 status = -ENOMEM;
1132                 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1133                         "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1134                 goto read_super_error;
1135         }
1136
1137         if (ocfs2_mount_local(osb))
1138                 snprintf(nodestr, sizeof(nodestr), "local");
1139         else
1140                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1141
1142         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1143                "with %s data mode.\n",
1144                osb->dev_str, nodestr, osb->slot_num,
1145                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1146                "ordered");
1147
1148         if ((osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
1149            !(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT))
1150                 printk(KERN_NOTICE "ocfs2: The shared device (%s) is mounted "
1151                        "without cluster aware mode.\n", osb->dev_str);
1152
1153         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1154         wake_up(&osb->osb_mount_event);
1155
1156         /* Now we can initialize quotas because we can afford to wait
1157          * for cluster locks recovery now. That also means that truncation
1158          * log recovery can happen but that waits for proper quota setup */
1159         if (!sb_rdonly(sb)) {
1160                 status = ocfs2_enable_quotas(osb);
1161                 if (status < 0) {
1162                         /* We have to err-out specially here because
1163                          * s_root is already set */
1164                         mlog_errno(status);
1165                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1166                         wake_up(&osb->osb_mount_event);
1167                         return status;
1168                 }
1169         }
1170
1171         ocfs2_complete_quota_recovery(osb);
1172
1173         /* Now we wake up again for processes waiting for quotas */
1174         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1175         wake_up(&osb->osb_mount_event);
1176
1177         /* Start this when the mount is almost sure of being successful */
1178         ocfs2_orphan_scan_start(osb);
1179
1180         return status;
1181
1182 read_super_error:
1183         brelse(bh);
1184
1185         if (status)
1186                 mlog_errno(status);
1187
1188         if (osb) {
1189                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1190                 wake_up(&osb->osb_mount_event);
1191                 ocfs2_dismount_volume(sb, 1);
1192         }
1193
1194         return status;
1195 }
1196
1197 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1198                         int flags,
1199                         const char *dev_name,
1200                         void *data)
1201 {
1202         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1203 }
1204
1205 static struct file_system_type ocfs2_fs_type = {
1206         .owner          = THIS_MODULE,
1207         .name           = "ocfs2",
1208         .mount          = ocfs2_mount,
1209         .kill_sb        = kill_block_super,
1210         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1211         .next           = NULL
1212 };
1213 MODULE_ALIAS_FS("ocfs2");
1214
1215 static int ocfs2_check_set_options(struct super_block *sb,
1216                                    struct mount_options *options)
1217 {
1218         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1219             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1220                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1221                 mlog(ML_ERROR, "User quotas were requested, but this "
1222                      "filesystem does not have the feature enabled.\n");
1223                 return 0;
1224         }
1225         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1226             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1227                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1228                 mlog(ML_ERROR, "Group quotas were requested, but this "
1229                      "filesystem does not have the feature enabled.\n");
1230                 return 0;
1231         }
1232         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1233             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1234                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1235                      "feature is not enabled\n");
1236                 return 0;
1237         }
1238         /* No ACL setting specified? Use XATTR feature... */
1239         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1240                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1241                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1242                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1243                 else
1244                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1245         }
1246         return 1;
1247 }
1248
1249 static int ocfs2_parse_options(struct super_block *sb,
1250                                char *options,
1251                                struct mount_options *mopt,
1252                                int is_remount)
1253 {
1254         int status, user_stack = 0;
1255         char *p;
1256         u32 tmp;
1257         int token, option;
1258         substring_t args[MAX_OPT_ARGS];
1259
1260         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1261
1262         mopt->commit_interval = 0;
1263         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1264         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1265         mopt->slot = OCFS2_INVALID_SLOT;
1266         mopt->localalloc_opt = -1;
1267         mopt->cluster_stack[0] = '\0';
1268         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1269         mopt->dir_resv_level = -1;
1270
1271         if (!options) {
1272                 status = 1;
1273                 goto bail;
1274         }
1275
1276         while ((p = strsep(&options, ",")) != NULL) {
1277                 if (!*p)
1278                         continue;
1279
1280                 token = match_token(p, tokens, args);
1281                 switch (token) {
1282                 case Opt_hb_local:
1283                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1284                         break;
1285                 case Opt_hb_none:
1286                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1287                         break;
1288                 case Opt_hb_global:
1289                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1290                         break;
1291                 case Opt_barrier:
1292                         if (match_int(&args[0], &option)) {
1293                                 status = 0;
1294                                 goto bail;
1295                         }
1296                         if (option)
1297                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1298                         else
1299                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1300                         break;
1301                 case Opt_intr:
1302                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1303                         break;
1304                 case Opt_nointr:
1305                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1306                         break;
1307                 case Opt_err_panic:
1308                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1309                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1310                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1311                         break;
1312                 case Opt_err_ro:
1313                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1314                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1315                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1316                         break;
1317                 case Opt_err_cont:
1318                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1319                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1320                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1321                         break;
1322                 case Opt_data_ordered:
1323                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1324                         break;
1325                 case Opt_data_writeback:
1326                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1327                         break;
1328                 case Opt_user_xattr:
1329                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1330                         break;
1331                 case Opt_nouser_xattr:
1332                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1333                         break;
1334                 case Opt_atime_quantum:
1335                         if (match_int(&args[0], &option)) {
1336                                 status = 0;
1337                                 goto bail;
1338                         }
1339                         if (option >= 0)
1340                                 mopt->atime_quantum = option;
1341                         break;
1342                 case Opt_slot:
1343                         if (match_int(&args[0], &option)) {
1344                                 status = 0;
1345                                 goto bail;
1346                         }
1347                         if (option)
1348                                 mopt->slot = (u16)option;
1349                         break;
1350                 case Opt_commit:
1351                         if (match_int(&args[0], &option)) {
1352                                 status = 0;
1353                                 goto bail;
1354                         }
1355                         if (option < 0)
1356                                 return 0;
1357                         if (option == 0)
1358                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1359                         mopt->commit_interval = HZ * option;
1360                         break;
1361                 case Opt_localalloc:
1362                         if (match_int(&args[0], &option)) {
1363                                 status = 0;
1364                                 goto bail;
1365                         }
1366                         if (option >= 0)
1367                                 mopt->localalloc_opt = option;
1368                         break;
1369                 case Opt_localflocks:
1370                         /*
1371                          * Changing this during remount could race
1372                          * flock() requests, or "unbalance" existing
1373                          * ones (e.g., a lock is taken in one mode but
1374                          * dropped in the other). If users care enough
1375                          * to flip locking modes during remount, we
1376                          * could add a "local" flag to individual
1377                          * flock structures for proper tracking of
1378                          * state.
1379                          */
1380                         if (!is_remount)
1381                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1382                         break;
1383                 case Opt_stack:
1384                         /* Check both that the option we were passed
1385                          * is of the right length and that it is a proper
1386                          * string of the right length.
1387                          */
1388                         if (((args[0].to - args[0].from) !=
1389                              OCFS2_STACK_LABEL_LEN) ||
1390                             (strnlen(args[0].from,
1391                                      OCFS2_STACK_LABEL_LEN) !=
1392                              OCFS2_STACK_LABEL_LEN)) {
1393                                 mlog(ML_ERROR,
1394                                      "Invalid cluster_stack option\n");
1395                                 status = 0;
1396                                 goto bail;
1397                         }
1398                         memcpy(mopt->cluster_stack, args[0].from,
1399                                OCFS2_STACK_LABEL_LEN);
1400                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1401                         /*
1402                          * Open code the memcmp here as we don't have
1403                          * an osb to pass to
1404                          * ocfs2_userspace_stack().
1405                          */
1406                         if (memcmp(mopt->cluster_stack,
1407                                    OCFS2_CLASSIC_CLUSTER_STACK,
1408                                    OCFS2_STACK_LABEL_LEN))
1409                                 user_stack = 1;
1410                         break;
1411                 case Opt_inode64:
1412                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1413                         break;
1414                 case Opt_usrquota:
1415                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1416                         break;
1417                 case Opt_grpquota:
1418                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1419                         break;
1420                 case Opt_coherency_buffered:
1421                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1422                         break;
1423                 case Opt_coherency_full:
1424                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1425                         break;
1426                 case Opt_acl:
1427                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1428                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1429                         break;
1430                 case Opt_noacl:
1431                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1432                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1433                         break;
1434                 case Opt_resv_level:
1435                         if (is_remount)
1436                                 break;
1437                         if (match_int(&args[0], &option)) {
1438                                 status = 0;
1439                                 goto bail;
1440                         }
1441                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1442                             option < OCFS2_MAX_RESV_LEVEL)
1443                                 mopt->resv_level = option;
1444                         break;
1445                 case Opt_dir_resv_level:
1446                         if (is_remount)
1447                                 break;
1448                         if (match_int(&args[0], &option)) {
1449                                 status = 0;
1450                                 goto bail;
1451                         }
1452                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1453                             option < OCFS2_MAX_RESV_LEVEL)
1454                                 mopt->dir_resv_level = option;
1455                         break;
1456                 case Opt_journal_async_commit:
1457                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1458                         break;
1459                 case Opt_nocluster:
1460                         mopt->mount_opt |= OCFS2_MOUNT_NOCLUSTER;
1461                         break;
1462                 default:
1463                         mlog(ML_ERROR,
1464                              "Unrecognized mount option \"%s\" "
1465                              "or missing value\n", p);
1466                         status = 0;
1467                         goto bail;
1468                 }
1469         }
1470
1471         if (user_stack == 0) {
1472                 /* Ensure only one heartbeat mode */
1473                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1474                                          OCFS2_MOUNT_HB_GLOBAL |
1475                                          OCFS2_MOUNT_HB_NONE);
1476                 if (hweight32(tmp) != 1) {
1477                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1478                         status = 0;
1479                         goto bail;
1480                 }
1481         }
1482
1483         status = 1;
1484
1485 bail:
1486         return status;
1487 }
1488
1489 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1490 {
1491         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1492         unsigned long opts = osb->s_mount_opt;
1493         unsigned int local_alloc_megs;
1494
1495         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1496                 seq_printf(s, ",_netdev");
1497                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1498                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1499                 else
1500                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1501         } else
1502                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1503
1504         if (opts & OCFS2_MOUNT_NOINTR)
1505                 seq_printf(s, ",nointr");
1506
1507         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1508                 seq_printf(s, ",data=writeback");
1509         else
1510                 seq_printf(s, ",data=ordered");
1511
1512         if (opts & OCFS2_MOUNT_BARRIER)
1513                 seq_printf(s, ",barrier=1");
1514
1515         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1516                 seq_printf(s, ",errors=panic");
1517         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1518                 seq_printf(s, ",errors=continue");
1519         else
1520                 seq_printf(s, ",errors=remount-ro");
1521
1522         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1523                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1524
1525         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1526
1527         if (osb->osb_commit_interval)
1528                 seq_printf(s, ",commit=%u",
1529                            (unsigned) (osb->osb_commit_interval / HZ));
1530
1531         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1532         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1533                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1534
1535         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1536                 seq_printf(s, ",localflocks,");
1537
1538         if (osb->osb_cluster_stack[0])
1539                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1540                                   OCFS2_STACK_LABEL_LEN);
1541         if (opts & OCFS2_MOUNT_USRQUOTA)
1542                 seq_printf(s, ",usrquota");
1543         if (opts & OCFS2_MOUNT_GRPQUOTA)
1544                 seq_printf(s, ",grpquota");
1545
1546         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1547                 seq_printf(s, ",coherency=buffered");
1548         else
1549                 seq_printf(s, ",coherency=full");
1550
1551         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1552                 seq_printf(s, ",nouser_xattr");
1553         else
1554                 seq_printf(s, ",user_xattr");
1555
1556         if (opts & OCFS2_MOUNT_INODE64)
1557                 seq_printf(s, ",inode64");
1558
1559         if (opts & OCFS2_MOUNT_POSIX_ACL)
1560                 seq_printf(s, ",acl");
1561         else
1562                 seq_printf(s, ",noacl");
1563
1564         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1565                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1566
1567         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1568                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1569
1570         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1571                 seq_printf(s, ",journal_async_commit");
1572
1573         if (opts & OCFS2_MOUNT_NOCLUSTER)
1574                 seq_printf(s, ",nocluster");
1575
1576         return 0;
1577 }
1578
1579 static int __init ocfs2_init(void)
1580 {
1581         int status;
1582
1583         status = init_ocfs2_uptodate_cache();
1584         if (status < 0)
1585                 goto out1;
1586
1587         status = ocfs2_initialize_mem_caches();
1588         if (status < 0)
1589                 goto out2;
1590
1591         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1592
1593         ocfs2_set_locking_protocol();
1594
1595         status = register_quota_format(&ocfs2_quota_format);
1596         if (status < 0)
1597                 goto out3;
1598         status = register_filesystem(&ocfs2_fs_type);
1599         if (!status)
1600                 return 0;
1601
1602         unregister_quota_format(&ocfs2_quota_format);
1603 out3:
1604         debugfs_remove(ocfs2_debugfs_root);
1605         ocfs2_free_mem_caches();
1606 out2:
1607         exit_ocfs2_uptodate_cache();
1608 out1:
1609         mlog_errno(status);
1610         return status;
1611 }
1612
1613 static void __exit ocfs2_exit(void)
1614 {
1615         unregister_quota_format(&ocfs2_quota_format);
1616
1617         debugfs_remove(ocfs2_debugfs_root);
1618
1619         ocfs2_free_mem_caches();
1620
1621         unregister_filesystem(&ocfs2_fs_type);
1622
1623         exit_ocfs2_uptodate_cache();
1624 }
1625
1626 static void ocfs2_put_super(struct super_block *sb)
1627 {
1628         trace_ocfs2_put_super(sb);
1629
1630         ocfs2_sync_blockdev(sb);
1631         ocfs2_dismount_volume(sb, 0);
1632 }
1633
1634 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1635 {
1636         struct ocfs2_super *osb;
1637         u32 numbits, freebits;
1638         int status;
1639         struct ocfs2_dinode *bm_lock;
1640         struct buffer_head *bh = NULL;
1641         struct inode *inode = NULL;
1642
1643         trace_ocfs2_statfs(dentry->d_sb, buf);
1644
1645         osb = OCFS2_SB(dentry->d_sb);
1646
1647         inode = ocfs2_get_system_file_inode(osb,
1648                                             GLOBAL_BITMAP_SYSTEM_INODE,
1649                                             OCFS2_INVALID_SLOT);
1650         if (!inode) {
1651                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1652                 status = -EIO;
1653                 goto bail;
1654         }
1655
1656         status = ocfs2_inode_lock(inode, &bh, 0);
1657         if (status < 0) {
1658                 mlog_errno(status);
1659                 goto bail;
1660         }
1661
1662         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1663
1664         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1665         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1666
1667         buf->f_type = OCFS2_SUPER_MAGIC;
1668         buf->f_bsize = dentry->d_sb->s_blocksize;
1669         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1670         buf->f_blocks = ((sector_t) numbits) *
1671                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1672         buf->f_bfree = ((sector_t) freebits) *
1673                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1674         buf->f_bavail = buf->f_bfree;
1675         buf->f_files = numbits;
1676         buf->f_ffree = freebits;
1677         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1678                                 & 0xFFFFFFFFUL;
1679         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1680                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1681
1682         brelse(bh);
1683
1684         ocfs2_inode_unlock(inode, 0);
1685         status = 0;
1686 bail:
1687         iput(inode);
1688
1689         if (status)
1690                 mlog_errno(status);
1691
1692         return status;
1693 }
1694
1695 static void ocfs2_inode_init_once(void *data)
1696 {
1697         struct ocfs2_inode_info *oi = data;
1698
1699         oi->ip_flags = 0;
1700         oi->ip_open_count = 0;
1701         spin_lock_init(&oi->ip_lock);
1702         ocfs2_extent_map_init(&oi->vfs_inode);
1703         INIT_LIST_HEAD(&oi->ip_io_markers);
1704         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1705         oi->ip_dir_start_lookup = 0;
1706         init_rwsem(&oi->ip_alloc_sem);
1707         init_rwsem(&oi->ip_xattr_sem);
1708         mutex_init(&oi->ip_io_mutex);
1709
1710         oi->ip_blkno = 0ULL;
1711         oi->ip_clusters = 0;
1712         oi->ip_next_orphan = NULL;
1713
1714         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1715
1716         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1717         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1718         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1719
1720         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1721                                   &ocfs2_inode_caching_ops);
1722
1723         inode_init_once(&oi->vfs_inode);
1724 }
1725
1726 static int ocfs2_initialize_mem_caches(void)
1727 {
1728         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1729                                        sizeof(struct ocfs2_inode_info),
1730                                        0,
1731                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1732                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1733                                        ocfs2_inode_init_once);
1734         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1735                                         sizeof(struct ocfs2_dquot),
1736                                         0,
1737                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1738                                                 SLAB_MEM_SPREAD),
1739                                         NULL);
1740         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1741                                         sizeof(struct ocfs2_quota_chunk),
1742                                         0,
1743                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1744                                         NULL);
1745         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1746             !ocfs2_qf_chunk_cachep) {
1747                 kmem_cache_destroy(ocfs2_inode_cachep);
1748                 kmem_cache_destroy(ocfs2_dquot_cachep);
1749                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1750                 return -ENOMEM;
1751         }
1752
1753         return 0;
1754 }
1755
1756 static void ocfs2_free_mem_caches(void)
1757 {
1758         /*
1759          * Make sure all delayed rcu free inodes are flushed before we
1760          * destroy cache.
1761          */
1762         rcu_barrier();
1763         kmem_cache_destroy(ocfs2_inode_cachep);
1764         ocfs2_inode_cachep = NULL;
1765
1766         kmem_cache_destroy(ocfs2_dquot_cachep);
1767         ocfs2_dquot_cachep = NULL;
1768
1769         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1770         ocfs2_qf_chunk_cachep = NULL;
1771 }
1772
1773 static int ocfs2_get_sector(struct super_block *sb,
1774                             struct buffer_head **bh,
1775                             int block,
1776                             int sect_size)
1777 {
1778         if (!sb_set_blocksize(sb, sect_size)) {
1779                 mlog(ML_ERROR, "unable to set blocksize\n");
1780                 return -EIO;
1781         }
1782
1783         *bh = sb_getblk(sb, block);
1784         if (!*bh) {
1785                 mlog_errno(-ENOMEM);
1786                 return -ENOMEM;
1787         }
1788         lock_buffer(*bh);
1789         if (!buffer_dirty(*bh))
1790                 clear_buffer_uptodate(*bh);
1791         unlock_buffer(*bh);
1792         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1793         wait_on_buffer(*bh);
1794         if (!buffer_uptodate(*bh)) {
1795                 mlog_errno(-EIO);
1796                 brelse(*bh);
1797                 *bh = NULL;
1798                 return -EIO;
1799         }
1800
1801         return 0;
1802 }
1803
1804 static int ocfs2_mount_volume(struct super_block *sb)
1805 {
1806         int status = 0;
1807         int unlock_super = 0;
1808         struct ocfs2_super *osb = OCFS2_SB(sb);
1809
1810         if (ocfs2_is_hard_readonly(osb))
1811                 goto leave;
1812
1813         mutex_init(&osb->obs_trim_fs_mutex);
1814
1815         status = ocfs2_dlm_init(osb);
1816         if (status < 0) {
1817                 mlog_errno(status);
1818                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1819                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1820                         " disk does not match the running cluster name.\n");
1821                 goto leave;
1822         }
1823
1824         status = ocfs2_super_lock(osb, 1);
1825         if (status < 0) {
1826                 mlog_errno(status);
1827                 goto leave;
1828         }
1829         unlock_super = 1;
1830
1831         /* This will load up the node map and add ourselves to it. */
1832         status = ocfs2_find_slot(osb);
1833         if (status < 0) {
1834                 mlog_errno(status);
1835                 goto leave;
1836         }
1837
1838         /* load all node-local system inodes */
1839         status = ocfs2_init_local_system_inodes(osb);
1840         if (status < 0) {
1841                 mlog_errno(status);
1842                 goto leave;
1843         }
1844
1845         status = ocfs2_check_volume(osb);
1846         if (status < 0) {
1847                 mlog_errno(status);
1848                 goto leave;
1849         }
1850
1851         status = ocfs2_truncate_log_init(osb);
1852         if (status < 0)
1853                 mlog_errno(status);
1854
1855 leave:
1856         if (unlock_super)
1857                 ocfs2_super_unlock(osb, 1);
1858
1859         return status;
1860 }
1861
1862 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1863 {
1864         int tmp, hangup_needed = 0;
1865         struct ocfs2_super *osb = NULL;
1866         char nodestr[12];
1867
1868         trace_ocfs2_dismount_volume(sb);
1869
1870         BUG_ON(!sb);
1871         osb = OCFS2_SB(sb);
1872         BUG_ON(!osb);
1873
1874         /* Remove file check sysfs related directores/files,
1875          * and wait for the pending file check operations */
1876         ocfs2_filecheck_remove_sysfs(osb);
1877
1878         kset_unregister(osb->osb_dev_kset);
1879
1880         /* Orphan scan should be stopped as early as possible */
1881         ocfs2_orphan_scan_stop(osb);
1882
1883         ocfs2_disable_quotas(osb);
1884
1885         /* All dquots should be freed by now */
1886         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1887         /* Wait for worker to be done with the work structure in osb */
1888         cancel_work_sync(&osb->dquot_drop_work);
1889
1890         ocfs2_shutdown_local_alloc(osb);
1891
1892         ocfs2_truncate_log_shutdown(osb);
1893
1894         /* This will disable recovery and flush any recovery work. */
1895         ocfs2_recovery_exit(osb);
1896
1897         ocfs2_sync_blockdev(sb);
1898
1899         ocfs2_purge_refcount_trees(osb);
1900
1901         /* No cluster connection means we've failed during mount, so skip
1902          * all the steps which depended on that to complete. */
1903         if (osb->cconn) {
1904                 tmp = ocfs2_super_lock(osb, 1);
1905                 if (tmp < 0) {
1906                         mlog_errno(tmp);
1907                         return;
1908                 }
1909         }
1910
1911         if (osb->slot_num != OCFS2_INVALID_SLOT)
1912                 ocfs2_put_slot(osb);
1913
1914         if (osb->cconn)
1915                 ocfs2_super_unlock(osb, 1);
1916
1917         ocfs2_release_system_inodes(osb);
1918
1919         ocfs2_journal_shutdown(osb);
1920
1921         /*
1922          * If we're dismounting due to mount error, mount.ocfs2 will clean
1923          * up heartbeat.  If we're a local mount, there is no heartbeat.
1924          * If we failed before we got a uuid_str yet, we can't stop
1925          * heartbeat.  Otherwise, do it.
1926          */
1927         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1928             !ocfs2_is_hard_readonly(osb))
1929                 hangup_needed = 1;
1930
1931         if (osb->cconn)
1932                 ocfs2_dlm_shutdown(osb, hangup_needed);
1933
1934         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1935         debugfs_remove_recursive(osb->osb_debug_root);
1936
1937         if (hangup_needed)
1938                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1939
1940         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1941
1942         if (ocfs2_mount_local(osb))
1943                 snprintf(nodestr, sizeof(nodestr), "local");
1944         else
1945                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1946
1947         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1948                osb->dev_str, nodestr);
1949
1950         ocfs2_delete_osb(osb);
1951         kfree(osb);
1952         sb->s_dev = 0;
1953         sb->s_fs_info = NULL;
1954 }
1955
1956 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1957                                 unsigned uuid_bytes)
1958 {
1959         int i, ret;
1960         char *ptr;
1961
1962         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1963
1964         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1965         if (osb->uuid_str == NULL)
1966                 return -ENOMEM;
1967
1968         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1969                 /* print with null */
1970                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1971                 if (ret != 2) /* drop super cleans up */
1972                         return -EINVAL;
1973                 /* then only advance past the last char */
1974                 ptr += 2;
1975         }
1976
1977         return 0;
1978 }
1979
1980 /* Make sure entire volume is addressable by our journal.  Requires
1981    osb_clusters_at_boot to be valid and for the journal to have been
1982    initialized by ocfs2_journal_init(). */
1983 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1984 {
1985         int status = 0;
1986         u64 max_block =
1987                 ocfs2_clusters_to_blocks(osb->sb,
1988                                          osb->osb_clusters_at_boot) - 1;
1989
1990         /* 32-bit block number is always OK. */
1991         if (max_block <= (u32)~0ULL)
1992                 goto out;
1993
1994         /* Volume is "huge", so see if our journal is new enough to
1995            support it. */
1996         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1997                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1998               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1999                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2000                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2001                      "Enable the 'block64' journal option with tunefs.ocfs2");
2002                 status = -EFBIG;
2003                 goto out;
2004         }
2005
2006  out:
2007         return status;
2008 }
2009
2010 static int ocfs2_initialize_super(struct super_block *sb,
2011                                   struct buffer_head *bh,
2012                                   int sector_size,
2013                                   struct ocfs2_blockcheck_stats *stats)
2014 {
2015         int status;
2016         int i, cbits, bbits;
2017         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2018         struct inode *inode = NULL;
2019         struct ocfs2_super *osb;
2020         u64 total_blocks;
2021
2022         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2023         if (!osb) {
2024                 status = -ENOMEM;
2025                 mlog_errno(status);
2026                 goto bail;
2027         }
2028
2029         sb->s_fs_info = osb;
2030         sb->s_op = &ocfs2_sops;
2031         sb->s_d_op = &ocfs2_dentry_ops;
2032         sb->s_export_op = &ocfs2_export_ops;
2033         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2034         sb->dq_op = &ocfs2_quota_operations;
2035         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2036         sb->s_xattr = ocfs2_xattr_handlers;
2037         sb->s_time_gran = 1;
2038         sb->s_flags |= SB_NOATIME;
2039         /* this is needed to support O_LARGEFILE */
2040         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2041         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2042         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2043         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2044                sizeof(di->id2.i_super.s_uuid));
2045
2046         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2047
2048         for (i = 0; i < 3; i++)
2049                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2050         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2051
2052         osb->sb = sb;
2053         osb->s_sectsize_bits = blksize_bits(sector_size);
2054         BUG_ON(!osb->s_sectsize_bits);
2055
2056         spin_lock_init(&osb->dc_task_lock);
2057         init_waitqueue_head(&osb->dc_event);
2058         osb->dc_work_sequence = 0;
2059         osb->dc_wake_sequence = 0;
2060         INIT_LIST_HEAD(&osb->blocked_lock_list);
2061         osb->blocked_lock_count = 0;
2062         spin_lock_init(&osb->osb_lock);
2063         spin_lock_init(&osb->osb_xattr_lock);
2064         ocfs2_init_steal_slots(osb);
2065
2066         mutex_init(&osb->system_file_mutex);
2067
2068         atomic_set(&osb->alloc_stats.moves, 0);
2069         atomic_set(&osb->alloc_stats.local_data, 0);
2070         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2071         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2072         atomic_set(&osb->alloc_stats.bg_extends, 0);
2073
2074         /* Copy the blockcheck stats from the superblock probe */
2075         osb->osb_ecc_stats = *stats;
2076
2077         ocfs2_init_node_maps(osb);
2078
2079         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2080                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2081
2082         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2083         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2084                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2085                      osb->max_slots);
2086                 status = -EINVAL;
2087                 goto bail;
2088         }
2089
2090         ocfs2_orphan_scan_init(osb);
2091
2092         status = ocfs2_recovery_init(osb);
2093         if (status) {
2094                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2095                 mlog_errno(status);
2096                 goto bail;
2097         }
2098
2099         init_waitqueue_head(&osb->checkpoint_event);
2100
2101         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2102
2103         osb->slot_num = OCFS2_INVALID_SLOT;
2104
2105         osb->s_xattr_inline_size = le16_to_cpu(
2106                                         di->id2.i_super.s_xattr_inline_size);
2107
2108         osb->local_alloc_state = OCFS2_LA_UNUSED;
2109         osb->local_alloc_bh = NULL;
2110         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2111
2112         init_waitqueue_head(&osb->osb_mount_event);
2113
2114         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2115         if (status) {
2116                 mlog_errno(status);
2117                 goto bail;
2118         }
2119
2120         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2121         if (!osb->vol_label) {
2122                 mlog(ML_ERROR, "unable to alloc vol label\n");
2123                 status = -ENOMEM;
2124                 goto bail;
2125         }
2126
2127         osb->slot_recovery_generations =
2128                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2129                         GFP_KERNEL);
2130         if (!osb->slot_recovery_generations) {
2131                 status = -ENOMEM;
2132                 mlog_errno(status);
2133                 goto bail;
2134         }
2135
2136         init_waitqueue_head(&osb->osb_wipe_event);
2137         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2138                                         sizeof(*osb->osb_orphan_wipes),
2139                                         GFP_KERNEL);
2140         if (!osb->osb_orphan_wipes) {
2141                 status = -ENOMEM;
2142                 mlog_errno(status);
2143                 goto bail;
2144         }
2145
2146         osb->osb_rf_lock_tree = RB_ROOT;
2147
2148         osb->s_feature_compat =
2149                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2150         osb->s_feature_ro_compat =
2151                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2152         osb->s_feature_incompat =
2153                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2154
2155         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2156                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2157                      "optional features (%x).\n", i);
2158                 status = -EINVAL;
2159                 goto bail;
2160         }
2161         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2162                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2163                      "unsupported optional features (%x).\n", i);
2164                 status = -EINVAL;
2165                 goto bail;
2166         }
2167
2168         if (ocfs2_clusterinfo_valid(osb)) {
2169                 /*
2170                  * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2171                  * terminated, so make sure no overflow happens here by using
2172                  * memcpy. Destination strings will always be null terminated
2173                  * because osb is allocated using kzalloc.
2174                  */
2175                 osb->osb_stackflags =
2176                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2177                 memcpy(osb->osb_cluster_stack,
2178                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2179                        OCFS2_STACK_LABEL_LEN);
2180                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2181                         mlog(ML_ERROR,
2182                              "couldn't mount because of an invalid "
2183                              "cluster stack label (%s) \n",
2184                              osb->osb_cluster_stack);
2185                         status = -EINVAL;
2186                         goto bail;
2187                 }
2188                 memcpy(osb->osb_cluster_name,
2189                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2190                         OCFS2_CLUSTER_NAME_LEN);
2191         } else {
2192                 /* The empty string is identical with classic tools that
2193                  * don't know about s_cluster_info. */
2194                 osb->osb_cluster_stack[0] = '\0';
2195         }
2196
2197         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2198
2199         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2200         init_llist_head(&osb->dquot_drop_list);
2201
2202         /* get some pseudo constants for clustersize bits */
2203         osb->s_clustersize_bits =
2204                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2205         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2206
2207         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2208             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2209                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2210                      osb->s_clustersize);
2211                 status = -EINVAL;
2212                 goto bail;
2213         }
2214
2215         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2216                                                 le32_to_cpu(di->i_clusters));
2217
2218         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2219                                            total_blocks);
2220         if (status) {
2221                 mlog(ML_ERROR, "Volume too large "
2222                      "to mount safely on this system");
2223                 status = -EFBIG;
2224                 goto bail;
2225         }
2226
2227         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2228                                  sizeof(di->id2.i_super.s_uuid))) {
2229                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2230                 status = -ENOMEM;
2231                 goto bail;
2232         }
2233
2234         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2235                 OCFS2_MAX_VOL_LABEL_LEN);
2236         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2237         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2238         osb->first_cluster_group_blkno =
2239                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2240         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2241         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2242         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2243                                      (unsigned long long)osb->root_blkno,
2244                                      (unsigned long long)osb->system_dir_blkno,
2245                                      osb->s_clustersize_bits);
2246
2247         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2248         if (!osb->osb_dlm_debug) {
2249                 status = -ENOMEM;
2250                 mlog_errno(status);
2251                 goto bail;
2252         }
2253
2254         atomic_set(&osb->vol_state, VOLUME_INIT);
2255
2256         /* load root, system_dir, and all global system inodes */
2257         status = ocfs2_init_global_system_inodes(osb);
2258         if (status < 0) {
2259                 mlog_errno(status);
2260                 goto bail;
2261         }
2262
2263         /*
2264          * global bitmap
2265          */
2266         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2267                                             OCFS2_INVALID_SLOT);
2268         if (!inode) {
2269                 status = -EINVAL;
2270                 mlog_errno(status);
2271                 goto bail;
2272         }
2273
2274         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2275         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2276         iput(inode);
2277
2278         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2279                                  osb->s_feature_incompat) * 8;
2280
2281         status = ocfs2_init_slot_info(osb);
2282         if (status < 0) {
2283                 mlog_errno(status);
2284                 goto bail;
2285         }
2286         cleancache_init_shared_fs(sb);
2287
2288         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2289         if (!osb->ocfs2_wq) {
2290                 status = -ENOMEM;
2291                 mlog_errno(status);
2292         }
2293
2294 bail:
2295         return status;
2296 }
2297
2298 /*
2299  * will return: -EAGAIN if it is ok to keep searching for superblocks
2300  *              -EINVAL if there is a bad superblock
2301  *              0 on success
2302  */
2303 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2304                                struct buffer_head *bh,
2305                                u32 blksz,
2306                                struct ocfs2_blockcheck_stats *stats)
2307 {
2308         int status = -EAGAIN;
2309
2310         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2311                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2312                 /* We have to do a raw check of the feature here */
2313                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2314                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2315                         status = ocfs2_block_check_validate(bh->b_data,
2316                                                             bh->b_size,
2317                                                             &di->i_check,
2318                                                             stats);
2319                         if (status)
2320                                 goto out;
2321                 }
2322                 status = -EINVAL;
2323                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2324                         mlog(ML_ERROR, "found superblock with incorrect block "
2325                              "size: found %u, should be %u\n",
2326                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2327                                blksz);
2328                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2329                            OCFS2_MAJOR_REV_LEVEL ||
2330                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2331                            OCFS2_MINOR_REV_LEVEL) {
2332                         mlog(ML_ERROR, "found superblock with bad version: "
2333                              "found %u.%u, should be %u.%u\n",
2334                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2335                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2336                              OCFS2_MAJOR_REV_LEVEL,
2337                              OCFS2_MINOR_REV_LEVEL);
2338                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2339                         mlog(ML_ERROR, "bad block number on superblock: "
2340                              "found %llu, should be %llu\n",
2341                              (unsigned long long)le64_to_cpu(di->i_blkno),
2342                              (unsigned long long)bh->b_blocknr);
2343                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2344                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2345                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2346                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2347                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2348                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2349                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2350                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2351                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2352                         mlog(ML_ERROR,
2353                              "Superblock slots found greater than file system "
2354                              "maximum: found %u, max %u\n",
2355                              le16_to_cpu(di->id2.i_super.s_max_slots),
2356                              OCFS2_MAX_SLOTS);
2357                 } else {
2358                         /* found it! */
2359                         status = 0;
2360                 }
2361         }
2362
2363 out:
2364         if (status && status != -EAGAIN)
2365                 mlog_errno(status);
2366         return status;
2367 }
2368
2369 static int ocfs2_check_volume(struct ocfs2_super *osb)
2370 {
2371         int status;
2372         int dirty;
2373         int local;
2374         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2375                                                   * recover
2376                                                   * ourselves. */
2377
2378         /* Init our journal object. */
2379         status = ocfs2_journal_init(osb, &dirty);
2380         if (status < 0) {
2381                 mlog(ML_ERROR, "Could not initialize journal!\n");
2382                 goto finally;
2383         }
2384
2385         /* Now that journal has been initialized, check to make sure
2386            entire volume is addressable. */
2387         status = ocfs2_journal_addressable(osb);
2388         if (status)
2389                 goto finally;
2390
2391         /* If the journal was unmounted cleanly then we don't want to
2392          * recover anything. Otherwise, journal_load will do that
2393          * dirty work for us :) */
2394         if (!dirty) {
2395                 status = ocfs2_journal_wipe(osb->journal, 0);
2396                 if (status < 0) {
2397                         mlog_errno(status);
2398                         goto finally;
2399                 }
2400         } else {
2401                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2402                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2403         }
2404
2405         local = ocfs2_mount_local(osb);
2406
2407         /* will play back anything left in the journal. */
2408         status = ocfs2_journal_load(osb->journal, local, dirty);
2409         if (status < 0) {
2410                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2411                 goto finally;
2412         }
2413
2414         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2415                 jbd2_journal_set_features(osb->journal->j_journal,
2416                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2417                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2418         else
2419                 jbd2_journal_clear_features(osb->journal->j_journal,
2420                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2421                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2422
2423         if (dirty) {
2424                 /* recover my local alloc if we didn't unmount cleanly. */
2425                 status = ocfs2_begin_local_alloc_recovery(osb,
2426                                                           osb->slot_num,
2427                                                           &local_alloc);
2428                 if (status < 0) {
2429                         mlog_errno(status);
2430                         goto finally;
2431                 }
2432                 /* we complete the recovery process after we've marked
2433                  * ourselves as mounted. */
2434         }
2435
2436         status = ocfs2_load_local_alloc(osb);
2437         if (status < 0) {
2438                 mlog_errno(status);
2439                 goto finally;
2440         }
2441
2442         if (dirty) {
2443                 /* Recovery will be completed after we've mounted the
2444                  * rest of the volume. */
2445                 osb->local_alloc_copy = local_alloc;
2446                 local_alloc = NULL;
2447         }
2448
2449         /* go through each journal, trylock it and if you get the
2450          * lock, and it's marked as dirty, set the bit in the recover
2451          * map and launch a recovery thread for it. */
2452         status = ocfs2_mark_dead_nodes(osb);
2453         if (status < 0) {
2454                 mlog_errno(status);
2455                 goto finally;
2456         }
2457
2458         status = ocfs2_compute_replay_slots(osb);
2459         if (status < 0)
2460                 mlog_errno(status);
2461
2462 finally:
2463         kfree(local_alloc);
2464
2465         if (status)
2466                 mlog_errno(status);
2467         return status;
2468 }
2469
2470 /*
2471  * The routine gets called from dismount or close whenever a dismount on
2472  * volume is requested and the osb open count becomes 1.
2473  * It will remove the osb from the global list and also free up all the
2474  * initialized resources and fileobject.
2475  */
2476 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2477 {
2478         /* This function assumes that the caller has the main osb resource */
2479
2480         /* ocfs2_initializer_super have already created this workqueue */
2481         if (osb->ocfs2_wq)
2482                 destroy_workqueue(osb->ocfs2_wq);
2483
2484         ocfs2_free_slot_info(osb);
2485
2486         kfree(osb->osb_orphan_wipes);
2487         kfree(osb->slot_recovery_generations);
2488         kfree(osb->local_alloc_copy);
2489         kfree(osb->uuid_str);
2490         kfree(osb->vol_label);
2491         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2492         memset(osb, 0, sizeof(struct ocfs2_super));
2493 }
2494
2495 /* Depending on the mount option passed, perform one of the following:
2496  * Put OCFS2 into a readonly state (default)
2497  * Return EIO so that only the process errs
2498  * Fix the error as if fsck.ocfs2 -y
2499  * panic
2500  */
2501 static int ocfs2_handle_error(struct super_block *sb)
2502 {
2503         struct ocfs2_super *osb = OCFS2_SB(sb);
2504         int rv = 0;
2505
2506         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2507         pr_crit("On-disk corruption discovered. "
2508                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2509
2510         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2511                 panic("OCFS2: (device %s): panic forced after error\n",
2512                       sb->s_id);
2513         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2514                 pr_crit("OCFS2: Returning error to the calling process.\n");
2515                 rv = -EIO;
2516         } else { /* default option */
2517                 rv = -EROFS;
2518                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2519                         return rv;
2520
2521                 pr_crit("OCFS2: File system is now read-only.\n");
2522                 sb->s_flags |= SB_RDONLY;
2523                 ocfs2_set_ro_flag(osb, 0);
2524         }
2525
2526         return rv;
2527 }
2528
2529 int __ocfs2_error(struct super_block *sb, const char *function,
2530                   const char *fmt, ...)
2531 {
2532         struct va_format vaf;
2533         va_list args;
2534
2535         va_start(args, fmt);
2536         vaf.fmt = fmt;
2537         vaf.va = &args;
2538
2539         /* Not using mlog here because we want to show the actual
2540          * function the error came from. */
2541         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2542                sb->s_id, function, &vaf);
2543
2544         va_end(args);
2545
2546         return ocfs2_handle_error(sb);
2547 }
2548
2549 /* Handle critical errors. This is intentionally more drastic than
2550  * ocfs2_handle_error, so we only use for things like journal errors,
2551  * etc. */
2552 void __ocfs2_abort(struct super_block *sb, const char *function,
2553                    const char *fmt, ...)
2554 {
2555         struct va_format vaf;
2556         va_list args;
2557
2558         va_start(args, fmt);
2559
2560         vaf.fmt = fmt;
2561         vaf.va = &args;
2562
2563         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2564                sb->s_id, function, &vaf);
2565
2566         va_end(args);
2567
2568         /* We don't have the cluster support yet to go straight to
2569          * hard readonly in here. Until then, we want to keep
2570          * ocfs2_abort() so that we can at least mark critical
2571          * errors.
2572          *
2573          * TODO: This should abort the journal and alert other nodes
2574          * that our slot needs recovery. */
2575
2576         /* Force a panic(). This stinks, but it's better than letting
2577          * things continue without having a proper hard readonly
2578          * here. */
2579         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2580                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2581         ocfs2_handle_error(sb);
2582 }
2583
2584 /*
2585  * Void signal blockers, because in-kernel sigprocmask() only fails
2586  * when SIG_* is wrong.
2587  */
2588 void ocfs2_block_signals(sigset_t *oldset)
2589 {
2590         int rc;
2591         sigset_t blocked;
2592
2593         sigfillset(&blocked);
2594         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2595         BUG_ON(rc);
2596 }
2597
2598 void ocfs2_unblock_signals(sigset_t *oldset)
2599 {
2600         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2601         BUG_ON(rc);
2602 }
2603
2604 module_init(ocfs2_init);
2605 module_exit(ocfs2_exit);