fs/ocfs2/super.c

   1 // SPDX-License-Identifier: GPL-2.0-or-later
   2 /* -*- mode: c; c-basic-offset: 8; -*-
   3  * vim: noexpandtab sw=8 ts=8 sts=0:
   4  *
   5  * super.c
   6  *
   7  * load/unload driver, mount/dismount volumes
   8  *
   9  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
  10  */
  11
  12 #include <linux/module.h>
  13 #include <linux/fs.h>
  14 #include <linux/types.h>
  15 #include <linux/slab.h>
  16 #include <linux/highmem.h>
  17 #include <linux/init.h>
  18 #include <linux/random.h>
  19 #include <linux/statfs.h>
  20 #include <linux/moduleparam.h>
  21 #include <linux/blkdev.h>
  22 #include <linux/socket.h>
  23 #include <linux/inet.h>
  24 #include <linux/parser.h>
  25 #include <linux/crc32.h>
  26 #include <linux/debugfs.h>
  27 #include <linux/mount.h>
  28 #include <linux/seq_file.h>
  29 #include <linux/quotaops.h>
  30 #include <linux/cleancache.h>
  31 #include <linux/signal.h>
  32
  33 #define CREATE_TRACE_POINTS
  34 #include "ocfs2_trace.h"
  35
  36 #include <cluster/masklog.h>
  37
  38 #include "ocfs2.h"
  39
  40 /* this should be the only file to include a version 1 header */
  41 #include "ocfs1_fs_compat.h"
  42
  43 #include "alloc.h"
  44 #include "aops.h"
  45 #include "blockcheck.h"
  46 #include "dlmglue.h"
  47 #include "export.h"
  48 #include "extent_map.h"
  49 #include "heartbeat.h"
  50 #include "inode.h"
  51 #include "journal.h"
  52 #include "localalloc.h"
  53 #include "namei.h"
  54 #include "slot_map.h"
  55 #include "super.h"
  56 #include "sysfile.h"
  57 #include "uptodate.h"
  58 #include "xattr.h"
  59 #include "quota.h"
  60 #include "refcounttree.h"
  61 #include "suballoc.h"
  62
  63 #include "buffer_head_io.h"
  64 #include "filecheck.h"
  65
  66 static struct kmem_cache *ocfs2_inode_cachep;
  67 struct kmem_cache *ocfs2_dquot_cachep;
  68 struct kmem_cache *ocfs2_qf_chunk_cachep;
  69
  70 static struct dentry *ocfs2_debugfs_root;
  71
  72 MODULE_AUTHOR("Oracle");
  73 MODULE_LICENSE("GPL");
  74 MODULE_DESCRIPTION("OCFS2 cluster file system");
  75
  76 struct mount_options
  77 {
  78         unsigned long   commit_interval;
  79         unsigned long   mount_opt;
  80         unsigned int    atime_quantum;
  81         unsigned short  slot;
  82         int             localalloc_opt;
  83         unsigned int    resv_level;
  84         int             dir_resv_level;
  85         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
  86 };
  87
  88 static int ocfs2_parse_options(struct super_block *sb, char *options,
  89                                struct mount_options *mopt,
  90                                int is_remount);
  91 static int ocfs2_check_set_options(struct super_block *sb,
  92                                    struct mount_options *options);
  93 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
  94 static void ocfs2_put_super(struct super_block *sb);
  95 static int ocfs2_mount_volume(struct super_block *sb);
  96 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
  97 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
  98 static int ocfs2_initialize_mem_caches(void);
  99 static void ocfs2_free_mem_caches(void);
 100 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 101
 102 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 103
 104 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 105
 106 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 107 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 108 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 109 static int ocfs2_check_volume(struct ocfs2_super *osb);
 110 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 111                                struct buffer_head *bh,
 112                                u32 sectsize,
 113                                struct ocfs2_blockcheck_stats *stats);
 114 static int ocfs2_initialize_super(struct super_block *sb,
 115                                   struct buffer_head *bh,
 116                                   int sector_size,
 117                                   struct ocfs2_blockcheck_stats *stats);
 118 static int ocfs2_get_sector(struct super_block *sb,
 119                             struct buffer_head **bh,
 120                             int block,
 121                             int sect_size);
 122 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 123 static void ocfs2_free_inode(struct inode *inode);
 124 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 125 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 126 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 127
 128 static struct dquot **ocfs2_get_dquots(struct inode *inode)
 129 {
 130         return OCFS2_I(inode)->i_dquot;
 131 }
 132
 133 static const struct super_operations ocfs2_sops = {
 134         .statfs         = ocfs2_statfs,
 135         .alloc_inode    = ocfs2_alloc_inode,
 136         .free_inode     = ocfs2_free_inode,
 137         .drop_inode     = ocfs2_drop_inode,
 138         .evict_inode    = ocfs2_evict_inode,
 139         .sync_fs        = ocfs2_sync_fs,
 140         .put_super      = ocfs2_put_super,
 141         .remount_fs     = ocfs2_remount,
 142         .show_options   = ocfs2_show_options,
 143         .quota_read     = ocfs2_quota_read,
 144         .quota_write    = ocfs2_quota_write,
 145         .get_dquots     = ocfs2_get_dquots,
 146 };
 147
 148 enum {
 149         Opt_barrier,
 150         Opt_err_panic,
 151         Opt_err_ro,
 152         Opt_intr,
 153         Opt_nointr,
 154         Opt_hb_none,
 155         Opt_hb_local,
 156         Opt_hb_global,
 157         Opt_data_ordered,
 158         Opt_data_writeback,
 159         Opt_atime_quantum,
 160         Opt_slot,
 161         Opt_commit,
 162         Opt_localalloc,
 163         Opt_localflocks,
 164         Opt_stack,
 165         Opt_user_xattr,
 166         Opt_nouser_xattr,
 167         Opt_inode64,
 168         Opt_acl,
 169         Opt_noacl,
 170         Opt_usrquota,
 171         Opt_grpquota,
 172         Opt_coherency_buffered,
 173         Opt_coherency_full,
 174         Opt_resv_level,
 175         Opt_dir_resv_level,
 176         Opt_journal_async_commit,
 177         Opt_err_cont,
 178         Opt_err,
 179 };
 180
 181 static const match_table_t tokens = {
 182         {Opt_barrier, "barrier=%u"},
 183         {Opt_err_panic, "errors=panic"},
 184         {Opt_err_ro, "errors=remount-ro"},
 185         {Opt_intr, "intr"},
 186         {Opt_nointr, "nointr"},
 187         {Opt_hb_none, OCFS2_HB_NONE},
 188         {Opt_hb_local, OCFS2_HB_LOCAL},
 189         {Opt_hb_global, OCFS2_HB_GLOBAL},
 190         {Opt_data_ordered, "data=ordered"},
 191         {Opt_data_writeback, "data=writeback"},
 192         {Opt_atime_quantum, "atime_quantum=%u"},
 193         {Opt_slot, "preferred_slot=%u"},
 194         {Opt_commit, "commit=%u"},
 195         {Opt_localalloc, "localalloc=%d"},
 196         {Opt_localflocks, "localflocks"},
 197         {Opt_stack, "cluster_stack=%s"},
 198         {Opt_user_xattr, "user_xattr"},
 199         {Opt_nouser_xattr, "nouser_xattr"},
 200         {Opt_inode64, "inode64"},
 201         {Opt_acl, "acl"},
 202         {Opt_noacl, "noacl"},
 203         {Opt_usrquota, "usrquota"},
 204         {Opt_grpquota, "grpquota"},
 205         {Opt_coherency_buffered, "coherency=buffered"},
 206         {Opt_coherency_full, "coherency=full"},
 207         {Opt_resv_level, "resv_level=%u"},
 208         {Opt_dir_resv_level, "dir_resv_level=%u"},
 209         {Opt_journal_async_commit, "journal_async_commit"},
 210         {Opt_err_cont, "errors=continue"},
 211         {Opt_err, NULL}
 212 };
 213
 214 #ifdef CONFIG_DEBUG_FS
 215 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 216 {
 217         struct ocfs2_cluster_connection *cconn = osb->cconn;
 218         struct ocfs2_recovery_map *rm = osb->recovery_map;
 219         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 220         int i, out = 0;
 221         unsigned long flags;
 222
 223         out += scnprintf(buf + out, len - out,
 224                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 225                         "Device", osb->dev_str, osb->uuid_str,
 226                         osb->fs_generation, osb->vol_label);
 227
 228         out += scnprintf(buf + out, len - out,
 229                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 230                         atomic_read(&osb->vol_state), osb->osb_flags);
 231
 232         out += scnprintf(buf + out, len - out,
 233                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 234                         osb->sb->s_blocksize, osb->s_clustersize);
 235
 236         out += scnprintf(buf + out, len - out,
 237                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 238                         "ROcompat: 0x%X\n",
 239                         "Features", osb->s_feature_compat,
 240                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 241
 242         out += scnprintf(buf + out, len - out,
 243                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 244                         osb->s_mount_opt, osb->s_atime_quantum);
 245
 246         if (cconn) {
 247                 out += scnprintf(buf + out, len - out,
 248                                 "%10s => Stack: %s  Name: %*s  "
 249                                 "Version: %d.%d\n", "Cluster",
 250                                 (*osb->osb_cluster_stack == '\0' ?
 251                                  "o2cb" : osb->osb_cluster_stack),
 252                                 cconn->cc_namelen, cconn->cc_name,
 253                                 cconn->cc_version.pv_major,
 254                                 cconn->cc_version.pv_minor);
 255         }
 256
 257         spin_lock_irqsave(&osb->dc_task_lock, flags);
 258         out += scnprintf(buf + out, len - out,
 259                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 260                         "WorkSeq: %lu\n", "DownCnvt",
 261                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 262                         osb->blocked_lock_count, osb->dc_wake_sequence,
 263                         osb->dc_work_sequence);
 264         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 265
 266         spin_lock(&osb->osb_lock);
 267         out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 268                         "Recovery",
 269                         (osb->recovery_thread_task ?
 270                          task_pid_nr(osb->recovery_thread_task) : -1));
 271         if (rm->rm_used == 0)
 272                 out += scnprintf(buf + out, len - out, " None\n");
 273         else {
 274                 for (i = 0; i < rm->rm_used; i++)
 275                         out += scnprintf(buf + out, len - out, " %d",
 276                                         rm->rm_entries[i]);
 277                 out += scnprintf(buf + out, len - out, "\n");
 278         }
 279         spin_unlock(&osb->osb_lock);
 280
 281         out += scnprintf(buf + out, len - out,
 282                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
 283                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 284                         osb->osb_commit_interval);
 285
 286         out += scnprintf(buf + out, len - out,
 287                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 288                         "Journal", osb->journal->j_state,
 289                         osb->journal->j_trans_id,
 290                         atomic_read(&osb->journal->j_num_trans));
 291
 292         out += scnprintf(buf + out, len - out,
 293                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 294                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 295                         "Stats",
 296                         atomic_read(&osb->alloc_stats.bitmap_data),
 297                         atomic_read(&osb->alloc_stats.local_data),
 298                         atomic_read(&osb->alloc_stats.bg_allocs),
 299                         atomic_read(&osb->alloc_stats.moves),
 300                         atomic_read(&osb->alloc_stats.bg_extends));
 301
 302         out += scnprintf(buf + out, len - out,
 303                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 304                         "Default: %u bits\n",
 305                         "LocalAlloc", osb->local_alloc_state,
 306                         (unsigned long long)osb->la_last_gd,
 307                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 308
 309         spin_lock(&osb->osb_lock);
 310         out += scnprintf(buf + out, len - out,
 311                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 312                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 313                         osb->s_inode_steal_slot,
 314                         atomic_read(&osb->s_num_inodes_stolen),
 315                         osb->s_meta_steal_slot,
 316                         atomic_read(&osb->s_num_meta_stolen));
 317         spin_unlock(&osb->osb_lock);
 318
 319         out += scnprintf(buf + out, len - out, "OrphanScan => ");
 320         out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 321                         os->os_count, os->os_seqno);
 322         out += scnprintf(buf + out, len - out, " Last Scan: ");
 323         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 324                 out += scnprintf(buf + out, len - out, "Disabled\n");
 325         else
 326                 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 327                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
 328
 329         out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 330                         "Slots", "Num", "RecoGen");
 331         for (i = 0; i < osb->max_slots; ++i) {
 332                 out += scnprintf(buf + out, len - out,
 333                                 "%10s  %c %3d  %10d\n",
 334                                 " ",
 335                                 (i == osb->slot_num ? '*' : ' '),
 336                                 i, osb->slot_recovery_generations[i]);
 337         }
 338
 339         return out;
 340 }
 341
 342 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 343 {
 344         struct ocfs2_super *osb = inode->i_private;
 345         char *buf = NULL;
 346
 347         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 348         if (!buf)
 349                 goto bail;
 350
 351         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 352
 353         file->private_data = buf;
 354
 355         return 0;
 356 bail:
 357         return -ENOMEM;
 358 }
 359
 360 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 361 {
 362         kfree(file->private_data);
 363         return 0;
 364 }
 365
 366 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 367                                 size_t nbytes, loff_t *ppos)
 368 {
 369         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 370                                        i_size_read(file->f_mapping->host));
 371 }
 372 #else
 373 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 374 {
 375         return 0;
 376 }
 377 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 382                                 size_t nbytes, loff_t *ppos)
 383 {
 384         return 0;
 385 }
 386 #endif  /* CONFIG_DEBUG_FS */
 387
 388 static const struct file_operations ocfs2_osb_debug_fops = {
 389         .open =         ocfs2_osb_debug_open,
 390         .release =      ocfs2_debug_release,
 391         .read =         ocfs2_debug_read,
 392         .llseek =       generic_file_llseek,
 393 };
 394
 395 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 396 {
 397         int status;
 398         tid_t target;
 399         struct ocfs2_super *osb = OCFS2_SB(sb);
 400
 401         if (ocfs2_is_hard_readonly(osb))
 402                 return -EROFS;
 403
 404         if (wait) {
 405                 status = ocfs2_flush_truncate_log(osb);
 406                 if (status < 0)
 407                         mlog_errno(status);
 408         } else {
 409                 ocfs2_schedule_truncate_log_flush(osb, 0);
 410         }
 411
 412         if (jbd2_journal_start_commit(osb->journal->j_journal,
 413                                       &target)) {
 414                 if (wait)
 415                         jbd2_log_wait_commit(osb->journal->j_journal,
 416                                              target);
 417         }
 418         return 0;
 419 }
 420
 421 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 422 {
 423         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 424             && (ino == USER_QUOTA_SYSTEM_INODE
 425                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 426                 return 0;
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 428             && (ino == GROUP_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         return 1;
 432 }
 433
 434 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 435 {
 436         struct inode *new = NULL;
 437         int status = 0;
 438         int i;
 439
 440         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 441         if (IS_ERR(new)) {
 442                 status = PTR_ERR(new);
 443                 mlog_errno(status);
 444                 goto bail;
 445         }
 446         osb->root_inode = new;
 447
 448         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 449         if (IS_ERR(new)) {
 450                 status = PTR_ERR(new);
 451                 mlog_errno(status);
 452                 goto bail;
 453         }
 454         osb->sys_root_inode = new;
 455
 456         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 457              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 458                 if (!ocfs2_need_system_inode(osb, i))
 459                         continue;
 460                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 461                 if (!new) {
 462                         ocfs2_release_system_inodes(osb);
 463                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 464                         mlog_errno(status);
 465                         mlog(ML_ERROR, "Unable to load system inode %d, "
 466                              "possibly corrupt fs?", i);
 467                         goto bail;
 468                 }
 469                 // the array now has one ref, so drop this one
 470                 iput(new);
 471         }
 472
 473 bail:
 474         if (status)
 475                 mlog_errno(status);
 476         return status;
 477 }
 478
 479 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 480 {
 481         struct inode *new = NULL;
 482         int status = 0;
 483         int i;
 484
 485         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 486              i < NUM_SYSTEM_INODES;
 487              i++) {
 488                 if (!ocfs2_need_system_inode(osb, i))
 489                         continue;
 490                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 491                 if (!new) {
 492                         ocfs2_release_system_inodes(osb);
 493                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 494                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 495                              status, i, osb->slot_num);
 496                         goto bail;
 497                 }
 498                 /* the array now has one ref, so drop this one */
 499                 iput(new);
 500         }
 501
 502 bail:
 503         if (status)
 504                 mlog_errno(status);
 505         return status;
 506 }
 507
 508 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 509 {
 510         int i;
 511         struct inode *inode;
 512
 513         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 514                 inode = osb->global_system_inodes[i];
 515                 if (inode) {
 516                         iput(inode);
 517                         osb->global_system_inodes[i] = NULL;
 518                 }
 519         }
 520
 521         inode = osb->sys_root_inode;
 522         if (inode) {
 523                 iput(inode);
 524                 osb->sys_root_inode = NULL;
 525         }
 526
 527         inode = osb->root_inode;
 528         if (inode) {
 529                 iput(inode);
 530                 osb->root_inode = NULL;
 531         }
 532
 533         if (!osb->local_system_inodes)
 534                 return;
 535
 536         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 537                 if (osb->local_system_inodes[i]) {
 538                         iput(osb->local_system_inodes[i]);
 539                         osb->local_system_inodes[i] = NULL;
 540                 }
 541         }
 542
 543         kfree(osb->local_system_inodes);
 544         osb->local_system_inodes = NULL;
 545 }
 546
 547 /* We're allocating fs objects, use GFP_NOFS */
 548 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 549 {
 550         struct ocfs2_inode_info *oi;
 551
 552         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 553         if (!oi)
 554                 return NULL;
 555
 556         oi->i_sync_tid = 0;
 557         oi->i_datasync_tid = 0;
 558         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 559
 560         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 561         return &oi->vfs_inode;
 562 }
 563
 564 static void ocfs2_free_inode(struct inode *inode)
 565 {
 566         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 567 }
 568
 569 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 570                                                 unsigned int cbits)
 571 {
 572         unsigned int bytes = 1 << cbits;
 573         unsigned int trim = bytes;
 574         unsigned int bitshift = 32;
 575
 576         /*
 577          * i_size and all block offsets in ocfs2 are always 64 bits
 578          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 579          * 64 bit platforms, cluster size will be the limiting factor.
 580          */
 581
 582 #if BITS_PER_LONG == 32
 583         BUILD_BUG_ON(sizeof(sector_t) != 8);
 584         /*
 585          * We might be limited by page cache size.
 586          */
 587         if (bytes > PAGE_SIZE) {
 588                 bytes = PAGE_SIZE;
 589                 trim = 1;
 590                 /*
 591                  * Shift by 31 here so that we don't get larger than
 592                  * MAX_LFS_FILESIZE
 593                  */
 594                 bitshift = 31;
 595         }
 596 #endif
 597
 598         /*
 599          * Trim by a whole cluster when we can actually approach the
 600          * on-disk limits. Otherwise we can overflow i_clusters when
 601          * an extent start is at the max offset.
 602          */
 603         return (((unsigned long long)bytes) << bitshift) - trim;
 604 }
 605
 606 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 607 {
 608         int incompat_features;
 609         int ret = 0;
 610         struct mount_options parsed_options;
 611         struct ocfs2_super *osb = OCFS2_SB(sb);
 612         u32 tmp;
 613
 614         sync_filesystem(sb);
 615
 616         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 617             !ocfs2_check_set_options(sb, &parsed_options)) {
 618                 ret = -EINVAL;
 619                 goto out;
 620         }
 621
 622         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 623                 OCFS2_MOUNT_HB_NONE;
 624         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 625                 ret = -EINVAL;
 626                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 627                 goto out;
 628         }
 629
 630         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 631             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 632                 ret = -EINVAL;
 633                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 634                 goto out;
 635         }
 636
 637         /* Probably don't want this on remount; it might
 638          * mess with other nodes */
 639         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 640             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 641                 ret = -EINVAL;
 642                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 643                 goto out;
 644         }
 645
 646         /* We're going to/from readonly mode. */
 647         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 648                 /* Disable quota accounting before remounting RO */
 649                 if (*flags & SB_RDONLY) {
 650                         ret = ocfs2_susp_quotas(osb, 0);
 651                         if (ret < 0)
 652                                 goto out;
 653                 }
 654                 /* Lock here so the check of HARD_RO and the potential
 655                  * setting of SOFT_RO is atomic. */
 656                 spin_lock(&osb->osb_lock);
 657                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 658                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 659                         ret = -EROFS;
 660                         goto unlock_osb;
 661                 }
 662
 663                 if (*flags & SB_RDONLY) {
 664                         sb->s_flags |= SB_RDONLY;
 665                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 666                 } else {
 667                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 668                                 mlog(ML_ERROR, "Cannot remount RDWR "
 669                                      "filesystem due to previous errors.\n");
 670                                 ret = -EROFS;
 671                                 goto unlock_osb;
 672                         }
 673                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 674                         if (incompat_features) {
 675                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 676                                      "of unsupported optional features "
 677                                      "(%x).\n", incompat_features);
 678                                 ret = -EINVAL;
 679                                 goto unlock_osb;
 680                         }
 681                         sb->s_flags &= ~SB_RDONLY;
 682                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 683                 }
 684                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 685 unlock_osb:
 686                 spin_unlock(&osb->osb_lock);
 687                 /* Enable quota accounting after remounting RW */
 688                 if (!ret && !(*flags & SB_RDONLY)) {
 689                         if (sb_any_quota_suspended(sb))
 690                                 ret = ocfs2_susp_quotas(osb, 1);
 691                         else
 692                                 ret = ocfs2_enable_quotas(osb);
 693                         if (ret < 0) {
 694                                 /* Return back changes... */
 695                                 spin_lock(&osb->osb_lock);
 696                                 sb->s_flags |= SB_RDONLY;
 697                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 698                                 spin_unlock(&osb->osb_lock);
 699                                 goto out;
 700                         }
 701                 }
 702         }
 703
 704         if (!ret) {
 705                 /* Only save off the new mount options in case of a successful
 706                  * remount. */
 707                 osb->s_mount_opt = parsed_options.mount_opt;
 708                 osb->s_atime_quantum = parsed_options.atime_quantum;
 709                 osb->preferred_slot = parsed_options.slot;
 710                 if (parsed_options.commit_interval)
 711                         osb->osb_commit_interval = parsed_options.commit_interval;
 712
 713                 if (!ocfs2_is_hard_readonly(osb))
 714                         ocfs2_set_journal_params(osb);
 715
 716                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 717                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 718                                                         SB_POSIXACL : 0);
 719         }
 720 out:
 721         return ret;
 722 }
 723
 724 static int ocfs2_sb_probe(struct super_block *sb,
 725                           struct buffer_head **bh,
 726                           int *sector_size,
 727                           struct ocfs2_blockcheck_stats *stats)
 728 {
 729         int status, tmpstat;
 730         struct ocfs1_vol_disk_hdr *hdr;
 731         struct ocfs2_dinode *di;
 732         int blksize;
 733
 734         *bh = NULL;
 735
 736         /* may be > 512 */
 737         *sector_size = bdev_logical_block_size(sb->s_bdev);
 738         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 739                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 740                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 741                 status = -EINVAL;
 742                 goto bail;
 743         }
 744
 745         /* Can this really happen? */
 746         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 747                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 748
 749         /* check block zero for old format */
 750         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 751         if (status < 0) {
 752                 mlog_errno(status);
 753                 goto bail;
 754         }
 755         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 756         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 757                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 758                      hdr->major_version, hdr->minor_version);
 759                 status = -EINVAL;
 760         }
 761         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 762                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 763                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 764                      hdr->signature);
 765                 status = -EINVAL;
 766         }
 767         brelse(*bh);
 768         *bh = NULL;
 769         if (status < 0) {
 770                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 771                      "upgraded before mounting with ocfs v2\n");
 772                 goto bail;
 773         }
 774
 775         /*
 776          * Now check at magic offset for 512, 1024, 2048, 4096
 777          * blocksizes.  4096 is the maximum blocksize because it is
 778          * the minimum clustersize.
 779          */
 780         status = -EINVAL;
 781         for (blksize = *sector_size;
 782              blksize <= OCFS2_MAX_BLOCKSIZE;
 783              blksize <<= 1) {
 784                 tmpstat = ocfs2_get_sector(sb, bh,
 785                                            OCFS2_SUPER_BLOCK_BLKNO,
 786                                            blksize);
 787                 if (tmpstat < 0) {
 788                         status = tmpstat;
 789                         mlog_errno(status);
 790                         break;
 791                 }
 792                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 793                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 794                 spin_lock_init(&stats->b_lock);
 795                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 796                 if (tmpstat < 0) {
 797                         brelse(*bh);
 798                         *bh = NULL;
 799                 }
 800                 if (tmpstat != -EAGAIN) {
 801                         status = tmpstat;
 802                         break;
 803                 }
 804         }
 805
 806 bail:
 807         return status;
 808 }
 809
 810 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 811 {
 812         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 813
 814         if (osb->s_mount_opt & hb_enabled) {
 815                 if (ocfs2_mount_local(osb)) {
 816                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 817                              "mounted device.\n");
 818                         return -EINVAL;
 819                 }
 820                 if (ocfs2_userspace_stack(osb)) {
 821                         mlog(ML_ERROR, "Userspace stack expected, but "
 822                              "o2cb heartbeat arguments passed to mount\n");
 823                         return -EINVAL;
 824                 }
 825                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 826                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 827                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 828                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 829                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 830                         return -EINVAL;
 831                 }
 832         }
 833
 834         if (!(osb->s_mount_opt & hb_enabled)) {
 835                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 836                     !ocfs2_userspace_stack(osb)) {
 837                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 838                              "a read-write clustered device.\n");
 839                         return -EINVAL;
 840                 }
 841         }
 842
 843         return 0;
 844 }
 845
 846 /*
 847  * If we're using a userspace stack, mount should have passed
 848  * a name that matches the disk.  If not, mount should not
 849  * have passed a stack.
 850  */
 851 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 852                                         struct mount_options *mopt)
 853 {
 854         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 855                 mlog(ML_ERROR,
 856                      "cluster stack passed to mount, but this filesystem "
 857                      "does not support it\n");
 858                 return -EINVAL;
 859         }
 860
 861         if (ocfs2_userspace_stack(osb) &&
 862             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 863                     OCFS2_STACK_LABEL_LEN)) {
 864                 mlog(ML_ERROR,
 865                      "cluster stack passed to mount (\"%s\") does not "
 866                      "match the filesystem (\"%s\")\n",
 867                      mopt->cluster_stack,
 868                      osb->osb_cluster_stack);
 869                 return -EINVAL;
 870         }
 871
 872         return 0;
 873 }
 874
 875 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 876 {
 877         int type;
 878         struct super_block *sb = osb->sb;
 879         unsigned int feature[OCFS2_MAXQUOTAS] = {
 880                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 881                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 882         int status = 0;
 883
 884         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 885                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 886                         continue;
 887                 if (unsuspend)
 888                         status = dquot_resume(sb, type);
 889                 else {
 890                         struct ocfs2_mem_dqinfo *oinfo;
 891
 892                         /* Cancel periodic syncing before suspending */
 893                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 894                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 895                         status = dquot_suspend(sb, type);
 896                 }
 897                 if (status < 0)
 898                         break;
 899         }
 900         if (status < 0)
 901                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 902                      "remount (error = %d).\n", status);
 903         return status;
 904 }
 905
 906 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 907 {
 908         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 909         struct super_block *sb = osb->sb;
 910         unsigned int feature[OCFS2_MAXQUOTAS] = {
 911                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 912                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 913         unsigned int ino[OCFS2_MAXQUOTAS] = {
 914                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
 915                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 916         int status;
 917         int type;
 918
 919         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 920         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 921                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 922                         continue;
 923                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 924                                                         osb->slot_num);
 925                 if (!inode[type]) {
 926                         status = -ENOENT;
 927                         goto out_quota_off;
 928                 }
 929                 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 930                                                 DQUOT_USAGE_ENABLED);
 931                 if (status < 0)
 932                         goto out_quota_off;
 933         }
 934
 935         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 936                 iput(inode[type]);
 937         return 0;
 938 out_quota_off:
 939         ocfs2_disable_quotas(osb);
 940         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 941                 iput(inode[type]);
 942         mlog_errno(status);
 943         return status;
 944 }
 945
 946 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 947 {
 948         int type;
 949         struct inode *inode;
 950         struct super_block *sb = osb->sb;
 951         struct ocfs2_mem_dqinfo *oinfo;
 952
 953         /* We mostly ignore errors in this function because there's not much
 954          * we can do when we see them */
 955         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 956                 if (!sb_has_quota_loaded(sb, type))
 957                         continue;
 958                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 959                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 960                 inode = igrab(sb->s_dquot.files[type]);
 961                 /* Turn off quotas. This will remove all dquot structures from
 962                  * memory and so they will be automatically synced to global
 963                  * quota files */
 964                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 965                                         DQUOT_LIMITS_ENABLED);
 966                 if (!inode)
 967                         continue;
 968                 iput(inode);
 969         }
 970 }
 971
 972 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 973 {
 974         struct dentry *root;
 975         int status, sector_size;
 976         struct mount_options parsed_options;
 977         struct inode *inode = NULL;
 978         struct ocfs2_super *osb = NULL;
 979         struct buffer_head *bh = NULL;
 980         char nodestr[12];
 981         struct ocfs2_blockcheck_stats stats;
 982
 983         trace_ocfs2_fill_super(sb, data, silent);
 984
 985         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 986                 status = -EINVAL;
 987                 goto read_super_error;
 988         }
 989
 990         /* probe for superblock */
 991         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
 992         if (status < 0) {
 993                 mlog(ML_ERROR, "superblock probe failed!\n");
 994                 goto read_super_error;
 995         }
 996
 997         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
 998         osb = OCFS2_SB(sb);
 999         if (status < 0) {
1000                 mlog_errno(status);
1001                 goto read_super_error;
1002         }
1003         brelse(bh);
1004         bh = NULL;
1005
1006         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1007                 status = -EINVAL;
1008                 goto read_super_error;
1009         }
1010         osb->s_mount_opt = parsed_options.mount_opt;
1011         osb->s_atime_quantum = parsed_options.atime_quantum;
1012         osb->preferred_slot = parsed_options.slot;
1013         osb->osb_commit_interval = parsed_options.commit_interval;
1014
1015         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1016         osb->osb_resv_level = parsed_options.resv_level;
1017         osb->osb_dir_resv_level = parsed_options.resv_level;
1018         if (parsed_options.dir_resv_level == -1)
1019                 osb->osb_dir_resv_level = parsed_options.resv_level;
1020         else
1021                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1022
1023         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1024         if (status)
1025                 goto read_super_error;
1026
1027         sb->s_magic = OCFS2_SUPER_MAGIC;
1028
1029         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1030                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1031
1032         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1033          * heartbeat=none */
1034         if (bdev_read_only(sb->s_bdev)) {
1035                 if (!sb_rdonly(sb)) {
1036                         status = -EACCES;
1037                         mlog(ML_ERROR, "Readonly device detected but readonly "
1038                              "mount was not specified.\n");
1039                         goto read_super_error;
1040                 }
1041
1042                 /* You should not be able to start a local heartbeat
1043                  * on a readonly device. */
1044                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1045                         status = -EROFS;
1046                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1047                              "device.\n");
1048                         goto read_super_error;
1049                 }
1050
1051                 status = ocfs2_check_journals_nolocks(osb);
1052                 if (status < 0) {
1053                         if (status == -EROFS)
1054                                 mlog(ML_ERROR, "Recovery required on readonly "
1055                                      "file system, but write access is "
1056                                      "unavailable.\n");
1057                         else
1058                                 mlog_errno(status);
1059                         goto read_super_error;
1060                 }
1061
1062                 ocfs2_set_ro_flag(osb, 1);
1063
1064                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1065                        "Cluster services will not be used for this mount. "
1066                        "Recovery will be skipped.\n", osb->dev_str);
1067         }
1068
1069         if (!ocfs2_is_hard_readonly(osb)) {
1070                 if (sb_rdonly(sb))
1071                         ocfs2_set_ro_flag(osb, 0);
1072         }
1073
1074         status = ocfs2_verify_heartbeat(osb);
1075         if (status < 0) {
1076                 mlog_errno(status);
1077                 goto read_super_error;
1078         }
1079
1080         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1081                                                  ocfs2_debugfs_root);
1082
1083         debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1084                             osb, &ocfs2_osb_debug_fops);
1085
1086         if (ocfs2_meta_ecc(osb))
1087                 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1088                                                         osb->osb_debug_root);
1089
1090         status = ocfs2_mount_volume(sb);
1091         if (status < 0)
1092                 goto read_super_error;
1093
1094         if (osb->root_inode)
1095                 inode = igrab(osb->root_inode);
1096
1097         if (!inode) {
1098                 status = -EIO;
1099                 mlog_errno(status);
1100                 goto read_super_error;
1101         }
1102
1103         osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1104                                                 &ocfs2_kset->kobj);
1105         if (!osb->osb_dev_kset) {
1106                 status = -ENOMEM;
1107                 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1108                 goto read_super_error;
1109         }
1110
1111         /* Create filecheck sysfs related directories/files at
1112          * /sys/fs/ocfs2/<devname>/filecheck */
1113         if (ocfs2_filecheck_create_sysfs(osb)) {
1114                 status = -ENOMEM;
1115                 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1116                         "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1117                 goto read_super_error;
1118         }
1119
1120         root = d_make_root(inode);
1121         if (!root) {
1122                 status = -ENOMEM;
1123                 mlog_errno(status);
1124                 goto read_super_error;
1125         }
1126
1127         sb->s_root = root;
1128
1129         ocfs2_complete_mount_recovery(osb);
1130
1131         if (ocfs2_mount_local(osb))
1132                 snprintf(nodestr, sizeof(nodestr), "local");
1133         else
1134                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1135
1136         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1137                "with %s data mode.\n",
1138                osb->dev_str, nodestr, osb->slot_num,
1139                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1140                "ordered");
1141
1142         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1143         wake_up(&osb->osb_mount_event);
1144
1145         /* Now we can initialize quotas because we can afford to wait
1146          * for cluster locks recovery now. That also means that truncation
1147          * log recovery can happen but that waits for proper quota setup */
1148         if (!sb_rdonly(sb)) {
1149                 status = ocfs2_enable_quotas(osb);
1150                 if (status < 0) {
1151                         /* We have to err-out specially here because
1152                          * s_root is already set */
1153                         mlog_errno(status);
1154                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1155                         wake_up(&osb->osb_mount_event);
1156                         return status;
1157                 }
1158         }
1159
1160         ocfs2_complete_quota_recovery(osb);
1161
1162         /* Now we wake up again for processes waiting for quotas */
1163         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1164         wake_up(&osb->osb_mount_event);
1165
1166         /* Start this when the mount is almost sure of being successful */
1167         ocfs2_orphan_scan_start(osb);
1168
1169         return status;
1170
1171 read_super_error:
1172         brelse(bh);
1173
1174         if (status)
1175                 mlog_errno(status);
1176
1177         if (osb) {
1178                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1179                 wake_up(&osb->osb_mount_event);
1180                 ocfs2_dismount_volume(sb, 1);
1181         }
1182
1183         return status;
1184 }
1185
1186 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1187                         int flags,
1188                         const char *dev_name,
1189                         void *data)
1190 {
1191         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1192 }
1193
1194 static struct file_system_type ocfs2_fs_type = {
1195         .owner          = THIS_MODULE,
1196         .name           = "ocfs2",
1197         .mount          = ocfs2_mount,
1198         .kill_sb        = kill_block_super,
1199         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1200         .next           = NULL
1201 };
1202 MODULE_ALIAS_FS("ocfs2");
1203
1204 static int ocfs2_check_set_options(struct super_block *sb,
1205                                    struct mount_options *options)
1206 {
1207         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1208             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1209                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1210                 mlog(ML_ERROR, "User quotas were requested, but this "
1211                      "filesystem does not have the feature enabled.\n");
1212                 return 0;
1213         }
1214         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1215             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1216                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1217                 mlog(ML_ERROR, "Group quotas were requested, but this "
1218                      "filesystem does not have the feature enabled.\n");
1219                 return 0;
1220         }
1221         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1222             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1223                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1224                      "feature is not enabled\n");
1225                 return 0;
1226         }
1227         /* No ACL setting specified? Use XATTR feature... */
1228         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1229                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1230                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1231                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1232                 else
1233                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1234         }
1235         return 1;
1236 }
1237
1238 static int ocfs2_parse_options(struct super_block *sb,
1239                                char *options,
1240                                struct mount_options *mopt,
1241                                int is_remount)
1242 {
1243         int status, user_stack = 0;
1244         char *p;
1245         u32 tmp;
1246         int token, option;
1247         substring_t args[MAX_OPT_ARGS];
1248
1249         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1250
1251         mopt->commit_interval = 0;
1252         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1253         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1254         mopt->slot = OCFS2_INVALID_SLOT;
1255         mopt->localalloc_opt = -1;
1256         mopt->cluster_stack[0] = '\0';
1257         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1258         mopt->dir_resv_level = -1;
1259
1260         if (!options) {
1261                 status = 1;
1262                 goto bail;
1263         }
1264
1265         while ((p = strsep(&options, ",")) != NULL) {
1266                 if (!*p)
1267                         continue;
1268
1269                 token = match_token(p, tokens, args);
1270                 switch (token) {
1271                 case Opt_hb_local:
1272                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1273                         break;
1274                 case Opt_hb_none:
1275                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1276                         break;
1277                 case Opt_hb_global:
1278                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1279                         break;
1280                 case Opt_barrier:
1281                         if (match_int(&args[0], &option)) {
1282                                 status = 0;
1283                                 goto bail;
1284                         }
1285                         if (option)
1286                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1287                         else
1288                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1289                         break;
1290                 case Opt_intr:
1291                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1292                         break;
1293                 case Opt_nointr:
1294                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1295                         break;
1296                 case Opt_err_panic:
1297                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1298                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1299                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1300                         break;
1301                 case Opt_err_ro:
1302                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1303                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1304                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1305                         break;
1306                 case Opt_err_cont:
1307                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1308                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1309                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1310                         break;
1311                 case Opt_data_ordered:
1312                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1313                         break;
1314                 case Opt_data_writeback:
1315                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1316                         break;
1317                 case Opt_user_xattr:
1318                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1319                         break;
1320                 case Opt_nouser_xattr:
1321                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1322                         break;
1323                 case Opt_atime_quantum:
1324                         if (match_int(&args[0], &option)) {
1325                                 status = 0;
1326                                 goto bail;
1327                         }
1328                         if (option >= 0)
1329                                 mopt->atime_quantum = option;
1330                         break;
1331                 case Opt_slot:
1332                         if (match_int(&args[0], &option)) {
1333                                 status = 0;
1334                                 goto bail;
1335                         }
1336                         if (option)
1337                                 mopt->slot = (u16)option;
1338                         break;
1339                 case Opt_commit:
1340                         if (match_int(&args[0], &option)) {
1341                                 status = 0;
1342                                 goto bail;
1343                         }
1344                         if (option < 0)
1345                                 return 0;
1346                         if (option == 0)
1347                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1348                         mopt->commit_interval = HZ * option;
1349                         break;
1350                 case Opt_localalloc:
1351                         if (match_int(&args[0], &option)) {
1352                                 status = 0;
1353                                 goto bail;
1354                         }
1355                         if (option >= 0)
1356                                 mopt->localalloc_opt = option;
1357                         break;
1358                 case Opt_localflocks:
1359                         /*
1360                          * Changing this during remount could race
1361                          * flock() requests, or "unbalance" existing
1362                          * ones (e.g., a lock is taken in one mode but
1363                          * dropped in the other). If users care enough
1364                          * to flip locking modes during remount, we
1365                          * could add a "local" flag to individual
1366                          * flock structures for proper tracking of
1367                          * state.
1368                          */
1369                         if (!is_remount)
1370                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1371                         break;
1372                 case Opt_stack:
1373                         /* Check both that the option we were passed
1374                          * is of the right length and that it is a proper
1375                          * string of the right length.
1376                          */
1377                         if (((args[0].to - args[0].from) !=
1378                              OCFS2_STACK_LABEL_LEN) ||
1379                             (strnlen(args[0].from,
1380                                      OCFS2_STACK_LABEL_LEN) !=
1381                              OCFS2_STACK_LABEL_LEN)) {
1382                                 mlog(ML_ERROR,
1383                                      "Invalid cluster_stack option\n");
1384                                 status = 0;
1385                                 goto bail;
1386                         }
1387                         memcpy(mopt->cluster_stack, args[0].from,
1388                                OCFS2_STACK_LABEL_LEN);
1389                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1390                         /*
1391                          * Open code the memcmp here as we don't have
1392                          * an osb to pass to
1393                          * ocfs2_userspace_stack().
1394                          */
1395                         if (memcmp(mopt->cluster_stack,
1396                                    OCFS2_CLASSIC_CLUSTER_STACK,
1397                                    OCFS2_STACK_LABEL_LEN))
1398                                 user_stack = 1;
1399                         break;
1400                 case Opt_inode64:
1401                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1402                         break;
1403                 case Opt_usrquota:
1404                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1405                         break;
1406                 case Opt_grpquota:
1407                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1408                         break;
1409                 case Opt_coherency_buffered:
1410                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1411                         break;
1412                 case Opt_coherency_full:
1413                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1414                         break;
1415                 case Opt_acl:
1416                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1417                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1418                         break;
1419                 case Opt_noacl:
1420                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1421                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1422                         break;
1423                 case Opt_resv_level:
1424                         if (is_remount)
1425                                 break;
1426                         if (match_int(&args[0], &option)) {
1427                                 status = 0;
1428                                 goto bail;
1429                         }
1430                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1431                             option < OCFS2_MAX_RESV_LEVEL)
1432                                 mopt->resv_level = option;
1433                         break;
1434                 case Opt_dir_resv_level:
1435                         if (is_remount)
1436                                 break;
1437                         if (match_int(&args[0], &option)) {
1438                                 status = 0;
1439                                 goto bail;
1440                         }
1441                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1442                             option < OCFS2_MAX_RESV_LEVEL)
1443                                 mopt->dir_resv_level = option;
1444                         break;
1445                 case Opt_journal_async_commit:
1446                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1447                         break;
1448                 default:
1449                         mlog(ML_ERROR,
1450                              "Unrecognized mount option \"%s\" "
1451                              "or missing value\n", p);
1452                         status = 0;
1453                         goto bail;
1454                 }
1455         }
1456
1457         if (user_stack == 0) {
1458                 /* Ensure only one heartbeat mode */
1459                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1460                                          OCFS2_MOUNT_HB_GLOBAL |
1461                                          OCFS2_MOUNT_HB_NONE);
1462                 if (hweight32(tmp) != 1) {
1463                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1464                         status = 0;
1465                         goto bail;
1466                 }
1467         }
1468
1469         status = 1;
1470
1471 bail:
1472         return status;
1473 }
1474
1475 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1476 {
1477         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1478         unsigned long opts = osb->s_mount_opt;
1479         unsigned int local_alloc_megs;
1480
1481         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1482                 seq_printf(s, ",_netdev");
1483                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1484                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1485                 else
1486                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1487         } else
1488                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1489
1490         if (opts & OCFS2_MOUNT_NOINTR)
1491                 seq_printf(s, ",nointr");
1492
1493         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1494                 seq_printf(s, ",data=writeback");
1495         else
1496                 seq_printf(s, ",data=ordered");
1497
1498         if (opts & OCFS2_MOUNT_BARRIER)
1499                 seq_printf(s, ",barrier=1");
1500
1501         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1502                 seq_printf(s, ",errors=panic");
1503         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1504                 seq_printf(s, ",errors=continue");
1505         else
1506                 seq_printf(s, ",errors=remount-ro");
1507
1508         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1509                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1510
1511         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1512
1513         if (osb->osb_commit_interval)
1514                 seq_printf(s, ",commit=%u",
1515                            (unsigned) (osb->osb_commit_interval / HZ));
1516
1517         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1518         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1519                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1520
1521         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1522                 seq_printf(s, ",localflocks,");
1523
1524         if (osb->osb_cluster_stack[0])
1525                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1526                                   OCFS2_STACK_LABEL_LEN);
1527         if (opts & OCFS2_MOUNT_USRQUOTA)
1528                 seq_printf(s, ",usrquota");
1529         if (opts & OCFS2_MOUNT_GRPQUOTA)
1530                 seq_printf(s, ",grpquota");
1531
1532         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1533                 seq_printf(s, ",coherency=buffered");
1534         else
1535                 seq_printf(s, ",coherency=full");
1536
1537         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1538                 seq_printf(s, ",nouser_xattr");
1539         else
1540                 seq_printf(s, ",user_xattr");
1541
1542         if (opts & OCFS2_MOUNT_INODE64)
1543                 seq_printf(s, ",inode64");
1544
1545         if (opts & OCFS2_MOUNT_POSIX_ACL)
1546                 seq_printf(s, ",acl");
1547         else
1548                 seq_printf(s, ",noacl");
1549
1550         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1551                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1552
1553         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1554                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1555
1556         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1557                 seq_printf(s, ",journal_async_commit");
1558
1559         return 0;
1560 }
1561
1562 static int __init ocfs2_init(void)
1563 {
1564         int status;
1565
1566         status = init_ocfs2_uptodate_cache();
1567         if (status < 0)
1568                 goto out1;
1569
1570         status = ocfs2_initialize_mem_caches();
1571         if (status < 0)
1572                 goto out2;
1573
1574         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1575
1576         ocfs2_set_locking_protocol();
1577
1578         status = register_quota_format(&ocfs2_quota_format);
1579         if (status < 0)
1580                 goto out3;
1581         status = register_filesystem(&ocfs2_fs_type);
1582         if (!status)
1583                 return 0;
1584
1585         unregister_quota_format(&ocfs2_quota_format);
1586 out3:
1587         debugfs_remove(ocfs2_debugfs_root);
1588         ocfs2_free_mem_caches();
1589 out2:
1590         exit_ocfs2_uptodate_cache();
1591 out1:
1592         mlog_errno(status);
1593         return status;
1594 }
1595
1596 static void __exit ocfs2_exit(void)
1597 {
1598         unregister_quota_format(&ocfs2_quota_format);
1599
1600         debugfs_remove(ocfs2_debugfs_root);
1601
1602         ocfs2_free_mem_caches();
1603
1604         unregister_filesystem(&ocfs2_fs_type);
1605
1606         exit_ocfs2_uptodate_cache();
1607 }
1608
1609 static void ocfs2_put_super(struct super_block *sb)
1610 {
1611         trace_ocfs2_put_super(sb);
1612
1613         ocfs2_sync_blockdev(sb);
1614         ocfs2_dismount_volume(sb, 0);
1615 }
1616
1617 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1618 {
1619         struct ocfs2_super *osb;
1620         u32 numbits, freebits;
1621         int status;
1622         struct ocfs2_dinode *bm_lock;
1623         struct buffer_head *bh = NULL;
1624         struct inode *inode = NULL;
1625
1626         trace_ocfs2_statfs(dentry->d_sb, buf);
1627
1628         osb = OCFS2_SB(dentry->d_sb);
1629
1630         inode = ocfs2_get_system_file_inode(osb,
1631                                             GLOBAL_BITMAP_SYSTEM_INODE,
1632                                             OCFS2_INVALID_SLOT);
1633         if (!inode) {
1634                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1635                 status = -EIO;
1636                 goto bail;
1637         }
1638
1639         status = ocfs2_inode_lock(inode, &bh, 0);
1640         if (status < 0) {
1641                 mlog_errno(status);
1642                 goto bail;
1643         }
1644
1645         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1646
1647         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1648         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1649
1650         buf->f_type = OCFS2_SUPER_MAGIC;
1651         buf->f_bsize = dentry->d_sb->s_blocksize;
1652         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1653         buf->f_blocks = ((sector_t) numbits) *
1654                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1655         buf->f_bfree = ((sector_t) freebits) *
1656                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1657         buf->f_bavail = buf->f_bfree;
1658         buf->f_files = numbits;
1659         buf->f_ffree = freebits;
1660         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1661                                 & 0xFFFFFFFFUL;
1662         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1663                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1664
1665         brelse(bh);
1666
1667         ocfs2_inode_unlock(inode, 0);
1668         status = 0;
1669 bail:
1670         iput(inode);
1671
1672         if (status)
1673                 mlog_errno(status);
1674
1675         return status;
1676 }
1677
1678 static void ocfs2_inode_init_once(void *data)
1679 {
1680         struct ocfs2_inode_info *oi = data;
1681
1682         oi->ip_flags = 0;
1683         oi->ip_open_count = 0;
1684         spin_lock_init(&oi->ip_lock);
1685         ocfs2_extent_map_init(&oi->vfs_inode);
1686         INIT_LIST_HEAD(&oi->ip_io_markers);
1687         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1688         oi->ip_dir_start_lookup = 0;
1689         init_rwsem(&oi->ip_alloc_sem);
1690         init_rwsem(&oi->ip_xattr_sem);
1691         mutex_init(&oi->ip_io_mutex);
1692
1693         oi->ip_blkno = 0ULL;
1694         oi->ip_clusters = 0;
1695         oi->ip_next_orphan = NULL;
1696
1697         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1698
1699         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1700         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1701         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1702
1703         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1704                                   &ocfs2_inode_caching_ops);
1705
1706         inode_init_once(&oi->vfs_inode);
1707 }
1708
1709 static int ocfs2_initialize_mem_caches(void)
1710 {
1711         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1712                                        sizeof(struct ocfs2_inode_info),
1713                                        0,
1714                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1715                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1716                                        ocfs2_inode_init_once);
1717         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1718                                         sizeof(struct ocfs2_dquot),
1719                                         0,
1720                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1721                                                 SLAB_MEM_SPREAD),
1722                                         NULL);
1723         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1724                                         sizeof(struct ocfs2_quota_chunk),
1725                                         0,
1726                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1727                                         NULL);
1728         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1729             !ocfs2_qf_chunk_cachep) {
1730                 kmem_cache_destroy(ocfs2_inode_cachep);
1731                 kmem_cache_destroy(ocfs2_dquot_cachep);
1732                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1733                 return -ENOMEM;
1734         }
1735
1736         return 0;
1737 }
1738
1739 static void ocfs2_free_mem_caches(void)
1740 {
1741         /*
1742          * Make sure all delayed rcu free inodes are flushed before we
1743          * destroy cache.
1744          */
1745         rcu_barrier();
1746         kmem_cache_destroy(ocfs2_inode_cachep);
1747         ocfs2_inode_cachep = NULL;
1748
1749         kmem_cache_destroy(ocfs2_dquot_cachep);
1750         ocfs2_dquot_cachep = NULL;
1751
1752         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1753         ocfs2_qf_chunk_cachep = NULL;
1754 }
1755
1756 static int ocfs2_get_sector(struct super_block *sb,
1757                             struct buffer_head **bh,
1758                             int block,
1759                             int sect_size)
1760 {
1761         if (!sb_set_blocksize(sb, sect_size)) {
1762                 mlog(ML_ERROR, "unable to set blocksize\n");
1763                 return -EIO;
1764         }
1765
1766         *bh = sb_getblk(sb, block);
1767         if (!*bh) {
1768                 mlog_errno(-ENOMEM);
1769                 return -ENOMEM;
1770         }
1771         lock_buffer(*bh);
1772         if (!buffer_dirty(*bh))
1773                 clear_buffer_uptodate(*bh);
1774         unlock_buffer(*bh);
1775         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1776         wait_on_buffer(*bh);
1777         if (!buffer_uptodate(*bh)) {
1778                 mlog_errno(-EIO);
1779                 brelse(*bh);
1780                 *bh = NULL;
1781                 return -EIO;
1782         }
1783
1784         return 0;
1785 }
1786
1787 static int ocfs2_mount_volume(struct super_block *sb)
1788 {
1789         int status = 0;
1790         int unlock_super = 0;
1791         struct ocfs2_super *osb = OCFS2_SB(sb);
1792
1793         if (ocfs2_is_hard_readonly(osb))
1794                 goto leave;
1795
1796         mutex_init(&osb->obs_trim_fs_mutex);
1797
1798         status = ocfs2_dlm_init(osb);
1799         if (status < 0) {
1800                 mlog_errno(status);
1801                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1802                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1803                         " disk does not match the running cluster name.\n");
1804                 goto leave;
1805         }
1806
1807         status = ocfs2_super_lock(osb, 1);
1808         if (status < 0) {
1809                 mlog_errno(status);
1810                 goto leave;
1811         }
1812         unlock_super = 1;
1813
1814         /* This will load up the node map and add ourselves to it. */
1815         status = ocfs2_find_slot(osb);
1816         if (status < 0) {
1817                 mlog_errno(status);
1818                 goto leave;
1819         }
1820
1821         /* load all node-local system inodes */
1822         status = ocfs2_init_local_system_inodes(osb);
1823         if (status < 0) {
1824                 mlog_errno(status);
1825                 goto leave;
1826         }
1827
1828         status = ocfs2_check_volume(osb);
1829         if (status < 0) {
1830                 mlog_errno(status);
1831                 goto leave;
1832         }
1833
1834         status = ocfs2_truncate_log_init(osb);
1835         if (status < 0)
1836                 mlog_errno(status);
1837
1838 leave:
1839         if (unlock_super)
1840                 ocfs2_super_unlock(osb, 1);
1841
1842         return status;
1843 }
1844
1845 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1846 {
1847         int tmp, hangup_needed = 0;
1848         struct ocfs2_super *osb = NULL;
1849         char nodestr[12];
1850
1851         trace_ocfs2_dismount_volume(sb);
1852
1853         BUG_ON(!sb);
1854         osb = OCFS2_SB(sb);
1855         BUG_ON(!osb);
1856
1857         /* Remove file check sysfs related directores/files,
1858          * and wait for the pending file check operations */
1859         ocfs2_filecheck_remove_sysfs(osb);
1860
1861         kset_unregister(osb->osb_dev_kset);
1862
1863         /* Orphan scan should be stopped as early as possible */
1864         ocfs2_orphan_scan_stop(osb);
1865
1866         ocfs2_disable_quotas(osb);
1867
1868         /* All dquots should be freed by now */
1869         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1870         /* Wait for worker to be done with the work structure in osb */
1871         cancel_work_sync(&osb->dquot_drop_work);
1872
1873         ocfs2_shutdown_local_alloc(osb);
1874
1875         ocfs2_truncate_log_shutdown(osb);
1876
1877         /* This will disable recovery and flush any recovery work. */
1878         ocfs2_recovery_exit(osb);
1879
1880         ocfs2_journal_shutdown(osb);
1881
1882         ocfs2_sync_blockdev(sb);
1883
1884         ocfs2_purge_refcount_trees(osb);
1885
1886         /* No cluster connection means we've failed during mount, so skip
1887          * all the steps which depended on that to complete. */
1888         if (osb->cconn) {
1889                 tmp = ocfs2_super_lock(osb, 1);
1890                 if (tmp < 0) {
1891                         mlog_errno(tmp);
1892                         return;
1893                 }
1894         }
1895
1896         if (osb->slot_num != OCFS2_INVALID_SLOT)
1897                 ocfs2_put_slot(osb);
1898
1899         if (osb->cconn)
1900                 ocfs2_super_unlock(osb, 1);
1901
1902         ocfs2_release_system_inodes(osb);
1903
1904         /*
1905          * If we're dismounting due to mount error, mount.ocfs2 will clean
1906          * up heartbeat.  If we're a local mount, there is no heartbeat.
1907          * If we failed before we got a uuid_str yet, we can't stop
1908          * heartbeat.  Otherwise, do it.
1909          */
1910         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1911             !ocfs2_is_hard_readonly(osb))
1912                 hangup_needed = 1;
1913
1914         if (osb->cconn)
1915                 ocfs2_dlm_shutdown(osb, hangup_needed);
1916
1917         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1918         debugfs_remove_recursive(osb->osb_debug_root);
1919
1920         if (hangup_needed)
1921                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1922
1923         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1924
1925         if (ocfs2_mount_local(osb))
1926                 snprintf(nodestr, sizeof(nodestr), "local");
1927         else
1928                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1929
1930         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1931                osb->dev_str, nodestr);
1932
1933         ocfs2_delete_osb(osb);
1934         kfree(osb);
1935         sb->s_dev = 0;
1936         sb->s_fs_info = NULL;
1937 }
1938
1939 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1940                                 unsigned uuid_bytes)
1941 {
1942         int i, ret;
1943         char *ptr;
1944
1945         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1946
1947         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1948         if (osb->uuid_str == NULL)
1949                 return -ENOMEM;
1950
1951         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1952                 /* print with null */
1953                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1954                 if (ret != 2) /* drop super cleans up */
1955                         return -EINVAL;
1956                 /* then only advance past the last char */
1957                 ptr += 2;
1958         }
1959
1960         return 0;
1961 }
1962
1963 /* Make sure entire volume is addressable by our journal.  Requires
1964    osb_clusters_at_boot to be valid and for the journal to have been
1965    initialized by ocfs2_journal_init(). */
1966 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1967 {
1968         int status = 0;
1969         u64 max_block =
1970                 ocfs2_clusters_to_blocks(osb->sb,
1971                                          osb->osb_clusters_at_boot) - 1;
1972
1973         /* 32-bit block number is always OK. */
1974         if (max_block <= (u32)~0ULL)
1975                 goto out;
1976
1977         /* Volume is "huge", so see if our journal is new enough to
1978            support it. */
1979         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1980                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1981               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1982                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
1983                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1984                      "Enable the 'block64' journal option with tunefs.ocfs2");
1985                 status = -EFBIG;
1986                 goto out;
1987         }
1988
1989  out:
1990         return status;
1991 }
1992
1993 static int ocfs2_initialize_super(struct super_block *sb,
1994                                   struct buffer_head *bh,
1995                                   int sector_size,
1996                                   struct ocfs2_blockcheck_stats *stats)
1997 {
1998         int status;
1999         int i, cbits, bbits;
2000         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2001         struct inode *inode = NULL;
2002         struct ocfs2_journal *journal;
2003         struct ocfs2_super *osb;
2004         u64 total_blocks;
2005
2006         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2007         if (!osb) {
2008                 status = -ENOMEM;
2009                 mlog_errno(status);
2010                 goto bail;
2011         }
2012
2013         sb->s_fs_info = osb;
2014         sb->s_op = &ocfs2_sops;
2015         sb->s_d_op = &ocfs2_dentry_ops;
2016         sb->s_export_op = &ocfs2_export_ops;
2017         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2018         sb->dq_op = &ocfs2_quota_operations;
2019         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2020         sb->s_xattr = ocfs2_xattr_handlers;
2021         sb->s_time_gran = 1;
2022         sb->s_flags |= SB_NOATIME;
2023         /* this is needed to support O_LARGEFILE */
2024         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2025         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2026         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2027         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2028                sizeof(di->id2.i_super.s_uuid));
2029
2030         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2031
2032         for (i = 0; i < 3; i++)
2033                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2034         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2035
2036         osb->sb = sb;
2037         osb->s_sectsize_bits = blksize_bits(sector_size);
2038         BUG_ON(!osb->s_sectsize_bits);
2039
2040         spin_lock_init(&osb->dc_task_lock);
2041         init_waitqueue_head(&osb->dc_event);
2042         osb->dc_work_sequence = 0;
2043         osb->dc_wake_sequence = 0;
2044         INIT_LIST_HEAD(&osb->blocked_lock_list);
2045         osb->blocked_lock_count = 0;
2046         spin_lock_init(&osb->osb_lock);
2047         spin_lock_init(&osb->osb_xattr_lock);
2048         ocfs2_init_steal_slots(osb);
2049
2050         mutex_init(&osb->system_file_mutex);
2051
2052         atomic_set(&osb->alloc_stats.moves, 0);
2053         atomic_set(&osb->alloc_stats.local_data, 0);
2054         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2055         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2056         atomic_set(&osb->alloc_stats.bg_extends, 0);
2057
2058         /* Copy the blockcheck stats from the superblock probe */
2059         osb->osb_ecc_stats = *stats;
2060
2061         ocfs2_init_node_maps(osb);
2062
2063         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2064                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2065
2066         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2067         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2068                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2069                      osb->max_slots);
2070                 status = -EINVAL;
2071                 goto bail;
2072         }
2073
2074         ocfs2_orphan_scan_init(osb);
2075
2076         status = ocfs2_recovery_init(osb);
2077         if (status) {
2078                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2079                 mlog_errno(status);
2080                 goto bail;
2081         }
2082
2083         init_waitqueue_head(&osb->checkpoint_event);
2084
2085         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2086
2087         osb->slot_num = OCFS2_INVALID_SLOT;
2088
2089         osb->s_xattr_inline_size = le16_to_cpu(
2090                                         di->id2.i_super.s_xattr_inline_size);
2091
2092         osb->local_alloc_state = OCFS2_LA_UNUSED;
2093         osb->local_alloc_bh = NULL;
2094         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2095
2096         init_waitqueue_head(&osb->osb_mount_event);
2097
2098         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2099         if (status) {
2100                 mlog_errno(status);
2101                 goto bail;
2102         }
2103
2104         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2105         if (!osb->vol_label) {
2106                 mlog(ML_ERROR, "unable to alloc vol label\n");
2107                 status = -ENOMEM;
2108                 goto bail;
2109         }
2110
2111         osb->slot_recovery_generations =
2112                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2113                         GFP_KERNEL);
2114         if (!osb->slot_recovery_generations) {
2115                 status = -ENOMEM;
2116                 mlog_errno(status);
2117                 goto bail;
2118         }
2119
2120         init_waitqueue_head(&osb->osb_wipe_event);
2121         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2122                                         sizeof(*osb->osb_orphan_wipes),
2123                                         GFP_KERNEL);
2124         if (!osb->osb_orphan_wipes) {
2125                 status = -ENOMEM;
2126                 mlog_errno(status);
2127                 goto bail;
2128         }
2129
2130         osb->osb_rf_lock_tree = RB_ROOT;
2131
2132         osb->s_feature_compat =
2133                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2134         osb->s_feature_ro_compat =
2135                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2136         osb->s_feature_incompat =
2137                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2138
2139         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2140                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2141                      "optional features (%x).\n", i);
2142                 status = -EINVAL;
2143                 goto bail;
2144         }
2145         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2146                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2147                      "unsupported optional features (%x).\n", i);
2148                 status = -EINVAL;
2149                 goto bail;
2150         }
2151
2152         if (ocfs2_clusterinfo_valid(osb)) {
2153                 /*
2154                  * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2155                  * terminated, so make sure no overflow happens here by using
2156                  * memcpy. Destination strings will always be null terminated
2157                  * because osb is allocated using kzalloc.
2158                  */
2159                 osb->osb_stackflags =
2160                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2161                 memcpy(osb->osb_cluster_stack,
2162                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2163                        OCFS2_STACK_LABEL_LEN);
2164                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2165                         mlog(ML_ERROR,
2166                              "couldn't mount because of an invalid "
2167                              "cluster stack label (%s) \n",
2168                              osb->osb_cluster_stack);
2169                         status = -EINVAL;
2170                         goto bail;
2171                 }
2172                 memcpy(osb->osb_cluster_name,
2173                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2174                         OCFS2_CLUSTER_NAME_LEN);
2175         } else {
2176                 /* The empty string is identical with classic tools that
2177                  * don't know about s_cluster_info. */
2178                 osb->osb_cluster_stack[0] = '\0';
2179         }
2180
2181         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2182
2183         /* FIXME
2184          * This should be done in ocfs2_journal_init(), but unknown
2185          * ordering issues will cause the filesystem to crash.
2186          * If anyone wants to figure out what part of the code
2187          * refers to osb->journal before ocfs2_journal_init() is run,
2188          * be my guest.
2189          */
2190         /* initialize our journal structure */
2191
2192         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2193         if (!journal) {
2194                 mlog(ML_ERROR, "unable to alloc journal\n");
2195                 status = -ENOMEM;
2196                 goto bail;
2197         }
2198         osb->journal = journal;
2199         journal->j_osb = osb;
2200
2201         atomic_set(&journal->j_num_trans, 0);
2202         init_rwsem(&journal->j_trans_barrier);
2203         init_waitqueue_head(&journal->j_checkpointed);
2204         spin_lock_init(&journal->j_lock);
2205         journal->j_trans_id = (unsigned long) 1;
2206         INIT_LIST_HEAD(&journal->j_la_cleanups);
2207         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2208         journal->j_state = OCFS2_JOURNAL_FREE;
2209
2210         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2211         init_llist_head(&osb->dquot_drop_list);
2212
2213         /* get some pseudo constants for clustersize bits */
2214         osb->s_clustersize_bits =
2215                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2216         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2217
2218         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2219             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2220                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2221                      osb->s_clustersize);
2222                 status = -EINVAL;
2223                 goto bail;
2224         }
2225
2226         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2227                                                 le32_to_cpu(di->i_clusters));
2228
2229         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2230                                            total_blocks);
2231         if (status) {
2232                 mlog(ML_ERROR, "Volume too large "
2233                      "to mount safely on this system");
2234                 status = -EFBIG;
2235                 goto bail;
2236         }
2237
2238         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2239                                  sizeof(di->id2.i_super.s_uuid))) {
2240                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2241                 status = -ENOMEM;
2242                 goto bail;
2243         }
2244
2245         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2246                 OCFS2_MAX_VOL_LABEL_LEN);
2247         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2248         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2249         osb->first_cluster_group_blkno =
2250                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2251         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2252         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2253         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2254                                      (unsigned long long)osb->root_blkno,
2255                                      (unsigned long long)osb->system_dir_blkno,
2256                                      osb->s_clustersize_bits);
2257
2258         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2259         if (!osb->osb_dlm_debug) {
2260                 status = -ENOMEM;
2261                 mlog_errno(status);
2262                 goto bail;
2263         }
2264
2265         atomic_set(&osb->vol_state, VOLUME_INIT);
2266
2267         /* load root, system_dir, and all global system inodes */
2268         status = ocfs2_init_global_system_inodes(osb);
2269         if (status < 0) {
2270                 mlog_errno(status);
2271                 goto bail;
2272         }
2273
2274         /*
2275          * global bitmap
2276          */
2277         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2278                                             OCFS2_INVALID_SLOT);
2279         if (!inode) {
2280                 status = -EINVAL;
2281                 mlog_errno(status);
2282                 goto bail;
2283         }
2284
2285         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2286         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2287         iput(inode);
2288
2289         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2290                                  osb->s_feature_incompat) * 8;
2291
2292         status = ocfs2_init_slot_info(osb);
2293         if (status < 0) {
2294                 mlog_errno(status);
2295                 goto bail;
2296         }
2297         cleancache_init_shared_fs(sb);
2298
2299         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2300         if (!osb->ocfs2_wq) {
2301                 status = -ENOMEM;
2302                 mlog_errno(status);
2303         }
2304
2305 bail:
2306         return status;
2307 }
2308
2309 /*
2310  * will return: -EAGAIN if it is ok to keep searching for superblocks
2311  *              -EINVAL if there is a bad superblock
2312  *              0 on success
2313  */
2314 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2315                                struct buffer_head *bh,
2316                                u32 blksz,
2317                                struct ocfs2_blockcheck_stats *stats)
2318 {
2319         int status = -EAGAIN;
2320
2321         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2322                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2323                 /* We have to do a raw check of the feature here */
2324                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2325                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2326                         status = ocfs2_block_check_validate(bh->b_data,
2327                                                             bh->b_size,
2328                                                             &di->i_check,
2329                                                             stats);
2330                         if (status)
2331                                 goto out;
2332                 }
2333                 status = -EINVAL;
2334                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2335                         mlog(ML_ERROR, "found superblock with incorrect block "
2336                              "size: found %u, should be %u\n",
2337                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2338                                blksz);
2339                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2340                            OCFS2_MAJOR_REV_LEVEL ||
2341                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2342                            OCFS2_MINOR_REV_LEVEL) {
2343                         mlog(ML_ERROR, "found superblock with bad version: "
2344                              "found %u.%u, should be %u.%u\n",
2345                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2346                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2347                              OCFS2_MAJOR_REV_LEVEL,
2348                              OCFS2_MINOR_REV_LEVEL);
2349                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2350                         mlog(ML_ERROR, "bad block number on superblock: "
2351                              "found %llu, should be %llu\n",
2352                              (unsigned long long)le64_to_cpu(di->i_blkno),
2353                              (unsigned long long)bh->b_blocknr);
2354                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2355                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2356                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2357                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2358                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2359                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2360                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2361                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2362                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2363                         mlog(ML_ERROR,
2364                              "Superblock slots found greater than file system "
2365                              "maximum: found %u, max %u\n",
2366                              le16_to_cpu(di->id2.i_super.s_max_slots),
2367                              OCFS2_MAX_SLOTS);
2368                 } else {
2369                         /* found it! */
2370                         status = 0;
2371                 }
2372         }
2373
2374 out:
2375         if (status && status != -EAGAIN)
2376                 mlog_errno(status);
2377         return status;
2378 }
2379
2380 static int ocfs2_check_volume(struct ocfs2_super *osb)
2381 {
2382         int status;
2383         int dirty;
2384         int local;
2385         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2386                                                   * recover
2387                                                   * ourselves. */
2388
2389         /* Init our journal object. */
2390         status = ocfs2_journal_init(osb->journal, &dirty);
2391         if (status < 0) {
2392                 mlog(ML_ERROR, "Could not initialize journal!\n");
2393                 goto finally;
2394         }
2395
2396         /* Now that journal has been initialized, check to make sure
2397            entire volume is addressable. */
2398         status = ocfs2_journal_addressable(osb);
2399         if (status)
2400                 goto finally;
2401
2402         /* If the journal was unmounted cleanly then we don't want to
2403          * recover anything. Otherwise, journal_load will do that
2404          * dirty work for us :) */
2405         if (!dirty) {
2406                 status = ocfs2_journal_wipe(osb->journal, 0);
2407                 if (status < 0) {
2408                         mlog_errno(status);
2409                         goto finally;
2410                 }
2411         } else {
2412                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2413                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2414         }
2415
2416         local = ocfs2_mount_local(osb);
2417
2418         /* will play back anything left in the journal. */
2419         status = ocfs2_journal_load(osb->journal, local, dirty);
2420         if (status < 0) {
2421                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2422                 goto finally;
2423         }
2424
2425         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2426                 jbd2_journal_set_features(osb->journal->j_journal,
2427                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2428                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2429         else
2430                 jbd2_journal_clear_features(osb->journal->j_journal,
2431                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2432                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2433
2434         if (dirty) {
2435                 /* recover my local alloc if we didn't unmount cleanly. */
2436                 status = ocfs2_begin_local_alloc_recovery(osb,
2437                                                           osb->slot_num,
2438                                                           &local_alloc);
2439                 if (status < 0) {
2440                         mlog_errno(status);
2441                         goto finally;
2442                 }
2443                 /* we complete the recovery process after we've marked
2444                  * ourselves as mounted. */
2445         }
2446
2447         status = ocfs2_load_local_alloc(osb);
2448         if (status < 0) {
2449                 mlog_errno(status);
2450                 goto finally;
2451         }
2452
2453         if (dirty) {
2454                 /* Recovery will be completed after we've mounted the
2455                  * rest of the volume. */
2456                 osb->local_alloc_copy = local_alloc;
2457                 local_alloc = NULL;
2458         }
2459
2460         /* go through each journal, trylock it and if you get the
2461          * lock, and it's marked as dirty, set the bit in the recover
2462          * map and launch a recovery thread for it. */
2463         status = ocfs2_mark_dead_nodes(osb);
2464         if (status < 0) {
2465                 mlog_errno(status);
2466                 goto finally;
2467         }
2468
2469         status = ocfs2_compute_replay_slots(osb);
2470         if (status < 0)
2471                 mlog_errno(status);
2472
2473 finally:
2474         kfree(local_alloc);
2475
2476         if (status)
2477                 mlog_errno(status);
2478         return status;
2479 }
2480
2481 /*
2482  * The routine gets called from dismount or close whenever a dismount on
2483  * volume is requested and the osb open count becomes 1.
2484  * It will remove the osb from the global list and also free up all the
2485  * initialized resources and fileobject.
2486  */
2487 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2488 {
2489         /* This function assumes that the caller has the main osb resource */
2490
2491         /* ocfs2_initializer_super have already created this workqueue */
2492         if (osb->ocfs2_wq)
2493                 destroy_workqueue(osb->ocfs2_wq);
2494
2495         ocfs2_free_slot_info(osb);
2496
2497         kfree(osb->osb_orphan_wipes);
2498         kfree(osb->slot_recovery_generations);
2499         /* FIXME
2500          * This belongs in journal shutdown, but because we have to
2501          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2502          * we free it here.
2503          */
2504         kfree(osb->journal);
2505         kfree(osb->local_alloc_copy);
2506         kfree(osb->uuid_str);
2507         kfree(osb->vol_label);
2508         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2509         memset(osb, 0, sizeof(struct ocfs2_super));
2510 }
2511
2512 /* Depending on the mount option passed, perform one of the following:
2513  * Put OCFS2 into a readonly state (default)
2514  * Return EIO so that only the process errs
2515  * Fix the error as if fsck.ocfs2 -y
2516  * panic
2517  */
2518 static int ocfs2_handle_error(struct super_block *sb)
2519 {
2520         struct ocfs2_super *osb = OCFS2_SB(sb);
2521         int rv = 0;
2522
2523         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2524         pr_crit("On-disk corruption discovered. "
2525                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2526
2527         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2528                 panic("OCFS2: (device %s): panic forced after error\n",
2529                       sb->s_id);
2530         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2531                 pr_crit("OCFS2: Returning error to the calling process.\n");
2532                 rv = -EIO;
2533         } else { /* default option */
2534                 rv = -EROFS;
2535                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2536                         return rv;
2537
2538                 pr_crit("OCFS2: File system is now read-only.\n");
2539                 sb->s_flags |= SB_RDONLY;
2540                 ocfs2_set_ro_flag(osb, 0);
2541         }
2542
2543         return rv;
2544 }
2545
2546 int __ocfs2_error(struct super_block *sb, const char *function,
2547                   const char *fmt, ...)
2548 {
2549         struct va_format vaf;
2550         va_list args;
2551
2552         va_start(args, fmt);
2553         vaf.fmt = fmt;
2554         vaf.va = &args;
2555
2556         /* Not using mlog here because we want to show the actual
2557          * function the error came from. */
2558         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2559                sb->s_id, function, &vaf);
2560
2561         va_end(args);
2562
2563         return ocfs2_handle_error(sb);
2564 }
2565
2566 /* Handle critical errors. This is intentionally more drastic than
2567  * ocfs2_handle_error, so we only use for things like journal errors,
2568  * etc. */
2569 void __ocfs2_abort(struct super_block *sb, const char *function,
2570                    const char *fmt, ...)
2571 {
2572         struct va_format vaf;
2573         va_list args;
2574
2575         va_start(args, fmt);
2576
2577         vaf.fmt = fmt;
2578         vaf.va = &args;
2579
2580         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2581                sb->s_id, function, &vaf);
2582
2583         va_end(args);
2584
2585         /* We don't have the cluster support yet to go straight to
2586          * hard readonly in here. Until then, we want to keep
2587          * ocfs2_abort() so that we can at least mark critical
2588          * errors.
2589          *
2590          * TODO: This should abort the journal and alert other nodes
2591          * that our slot needs recovery. */
2592
2593         /* Force a panic(). This stinks, but it's better than letting
2594          * things continue without having a proper hard readonly
2595          * here. */
2596         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2597                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2598         ocfs2_handle_error(sb);
2599 }
2600
2601 /*
2602  * Void signal blockers, because in-kernel sigprocmask() only fails
2603  * when SIG_* is wrong.
2604  */
2605 void ocfs2_block_signals(sigset_t *oldset)
2606 {
2607         int rc;
2608         sigset_t blocked;
2609
2610         sigfillset(&blocked);
2611         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2612         BUG_ON(rc);
2613 }
2614
2615 void ocfs2_unblock_signals(sigset_t *oldset)
2616 {
2617         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2618         BUG_ON(rc);
2619 }
2620
2621 module_init(ocfs2_init);
2622 module_exit(ocfs2_exit);