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