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_nocluster,
 179         Opt_err,
 180 };
 181
 182 static const match_table_t tokens = {
 183         {Opt_barrier, "barrier=%u"},
 184         {Opt_err_panic, "errors=panic"},
 185         {Opt_err_ro, "errors=remount-ro"},
 186         {Opt_intr, "intr"},
 187         {Opt_nointr, "nointr"},
 188         {Opt_hb_none, OCFS2_HB_NONE},
 189         {Opt_hb_local, OCFS2_HB_LOCAL},
 190         {Opt_hb_global, OCFS2_HB_GLOBAL},
 191         {Opt_data_ordered, "data=ordered"},
 192         {Opt_data_writeback, "data=writeback"},
 193         {Opt_atime_quantum, "atime_quantum=%u"},
 194         {Opt_slot, "preferred_slot=%u"},
 195         {Opt_commit, "commit=%u"},
 196         {Opt_localalloc, "localalloc=%d"},
 197         {Opt_localflocks, "localflocks"},
 198         {Opt_stack, "cluster_stack=%s"},
 199         {Opt_user_xattr, "user_xattr"},
 200         {Opt_nouser_xattr, "nouser_xattr"},
 201         {Opt_inode64, "inode64"},
 202         {Opt_acl, "acl"},
 203         {Opt_noacl, "noacl"},
 204         {Opt_usrquota, "usrquota"},
 205         {Opt_grpquota, "grpquota"},
 206         {Opt_coherency_buffered, "coherency=buffered"},
 207         {Opt_coherency_full, "coherency=full"},
 208         {Opt_resv_level, "resv_level=%u"},
 209         {Opt_dir_resv_level, "dir_resv_level=%u"},
 210         {Opt_journal_async_commit, "journal_async_commit"},
 211         {Opt_err_cont, "errors=continue"},
 212         {Opt_nocluster, "nocluster"},
 213         {Opt_err, NULL}
 214 };
 215
 216 #ifdef CONFIG_DEBUG_FS
 217 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 218 {
 219         struct ocfs2_cluster_connection *cconn = osb->cconn;
 220         struct ocfs2_recovery_map *rm = osb->recovery_map;
 221         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 222         int i, out = 0;
 223         unsigned long flags;
 224
 225         out += scnprintf(buf + out, len - out,
 226                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 227                         "Device", osb->dev_str, osb->uuid_str,
 228                         osb->fs_generation, osb->vol_label);
 229
 230         out += scnprintf(buf + out, len - out,
 231                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 232                         atomic_read(&osb->vol_state), osb->osb_flags);
 233
 234         out += scnprintf(buf + out, len - out,
 235                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 236                         osb->sb->s_blocksize, osb->s_clustersize);
 237
 238         out += scnprintf(buf + out, len - out,
 239                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 240                         "ROcompat: 0x%X\n",
 241                         "Features", osb->s_feature_compat,
 242                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 243
 244         out += scnprintf(buf + out, len - out,
 245                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 246                         osb->s_mount_opt, osb->s_atime_quantum);
 247
 248         if (cconn) {
 249                 out += scnprintf(buf + out, len - out,
 250                                 "%10s => Stack: %s  Name: %*s  "
 251                                 "Version: %d.%d\n", "Cluster",
 252                                 (*osb->osb_cluster_stack == '\0' ?
 253                                  "o2cb" : osb->osb_cluster_stack),
 254                                 cconn->cc_namelen, cconn->cc_name,
 255                                 cconn->cc_version.pv_major,
 256                                 cconn->cc_version.pv_minor);
 257         }
 258
 259         spin_lock_irqsave(&osb->dc_task_lock, flags);
 260         out += scnprintf(buf + out, len - out,
 261                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 262                         "WorkSeq: %lu\n", "DownCnvt",
 263                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 264                         osb->blocked_lock_count, osb->dc_wake_sequence,
 265                         osb->dc_work_sequence);
 266         spin_unlock_irqrestore(&osb->dc_task_lock, flags);
 267
 268         spin_lock(&osb->osb_lock);
 269         out += scnprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 270                         "Recovery",
 271                         (osb->recovery_thread_task ?
 272                          task_pid_nr(osb->recovery_thread_task) : -1));
 273         if (rm->rm_used == 0)
 274                 out += scnprintf(buf + out, len - out, " None\n");
 275         else {
 276                 for (i = 0; i < rm->rm_used; i++)
 277                         out += scnprintf(buf + out, len - out, " %d",
 278                                         rm->rm_entries[i]);
 279                 out += scnprintf(buf + out, len - out, "\n");
 280         }
 281         spin_unlock(&osb->osb_lock);
 282
 283         out += scnprintf(buf + out, len - out,
 284                         "%10s => Pid: %d  Interval: %lu\n", "Commit",
 285                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 286                         osb->osb_commit_interval);
 287
 288         out += scnprintf(buf + out, len - out,
 289                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 290                         "Journal", osb->journal->j_state,
 291                         osb->journal->j_trans_id,
 292                         atomic_read(&osb->journal->j_num_trans));
 293
 294         out += scnprintf(buf + out, len - out,
 295                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 296                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 297                         "Stats",
 298                         atomic_read(&osb->alloc_stats.bitmap_data),
 299                         atomic_read(&osb->alloc_stats.local_data),
 300                         atomic_read(&osb->alloc_stats.bg_allocs),
 301                         atomic_read(&osb->alloc_stats.moves),
 302                         atomic_read(&osb->alloc_stats.bg_extends));
 303
 304         out += scnprintf(buf + out, len - out,
 305                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 306                         "Default: %u bits\n",
 307                         "LocalAlloc", osb->local_alloc_state,
 308                         (unsigned long long)osb->la_last_gd,
 309                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 310
 311         spin_lock(&osb->osb_lock);
 312         out += scnprintf(buf + out, len - out,
 313                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 314                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 315                         osb->s_inode_steal_slot,
 316                         atomic_read(&osb->s_num_inodes_stolen),
 317                         osb->s_meta_steal_slot,
 318                         atomic_read(&osb->s_num_meta_stolen));
 319         spin_unlock(&osb->osb_lock);
 320
 321         out += scnprintf(buf + out, len - out, "OrphanScan => ");
 322         out += scnprintf(buf + out, len - out, "Local: %u  Global: %u ",
 323                         os->os_count, os->os_seqno);
 324         out += scnprintf(buf + out, len - out, " Last Scan: ");
 325         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 326                 out += scnprintf(buf + out, len - out, "Disabled\n");
 327         else
 328                 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
 329                                 (unsigned long)(ktime_get_seconds() - os->os_scantime));
 330
 331         out += scnprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 332                         "Slots", "Num", "RecoGen");
 333         for (i = 0; i < osb->max_slots; ++i) {
 334                 out += scnprintf(buf + out, len - out,
 335                                 "%10s  %c %3d  %10d\n",
 336                                 " ",
 337                                 (i == osb->slot_num ? '*' : ' '),
 338                                 i, osb->slot_recovery_generations[i]);
 339         }
 340
 341         return out;
 342 }
 343
 344 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 345 {
 346         struct ocfs2_super *osb = inode->i_private;
 347         char *buf = NULL;
 348
 349         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 350         if (!buf)
 351                 goto bail;
 352
 353         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 354
 355         file->private_data = buf;
 356
 357         return 0;
 358 bail:
 359         return -ENOMEM;
 360 }
 361
 362 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 363 {
 364         kfree(file->private_data);
 365         return 0;
 366 }
 367
 368 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 369                                 size_t nbytes, loff_t *ppos)
 370 {
 371         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 372                                        i_size_read(file->f_mapping->host));
 373 }
 374 #else
 375 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 376 {
 377         return 0;
 378 }
 379 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 380 {
 381         return 0;
 382 }
 383 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 384                                 size_t nbytes, loff_t *ppos)
 385 {
 386         return 0;
 387 }
 388 #endif  /* CONFIG_DEBUG_FS */
 389
 390 static const struct file_operations ocfs2_osb_debug_fops = {
 391         .open =         ocfs2_osb_debug_open,
 392         .release =      ocfs2_debug_release,
 393         .read =         ocfs2_debug_read,
 394         .llseek =       generic_file_llseek,
 395 };
 396
 397 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 398 {
 399         int status;
 400         tid_t target;
 401         struct ocfs2_super *osb = OCFS2_SB(sb);
 402
 403         if (ocfs2_is_hard_readonly(osb))
 404                 return -EROFS;
 405
 406         if (wait) {
 407                 status = ocfs2_flush_truncate_log(osb);
 408                 if (status < 0)
 409                         mlog_errno(status);
 410         } else {
 411                 ocfs2_schedule_truncate_log_flush(osb, 0);
 412         }
 413
 414         if (jbd2_journal_start_commit(osb->journal->j_journal,
 415                                       &target)) {
 416                 if (wait)
 417                         jbd2_log_wait_commit(osb->journal->j_journal,
 418                                              target);
 419         }
 420         return 0;
 421 }
 422
 423 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 424 {
 425         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 426             && (ino == USER_QUOTA_SYSTEM_INODE
 427                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 428                 return 0;
 429         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 430             && (ino == GROUP_QUOTA_SYSTEM_INODE
 431                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 432                 return 0;
 433         return 1;
 434 }
 435
 436 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 437 {
 438         struct inode *new = NULL;
 439         int status = 0;
 440         int i;
 441
 442         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 443         if (IS_ERR(new)) {
 444                 status = PTR_ERR(new);
 445                 mlog_errno(status);
 446                 goto bail;
 447         }
 448         osb->root_inode = new;
 449
 450         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 451         if (IS_ERR(new)) {
 452                 status = PTR_ERR(new);
 453                 mlog_errno(status);
 454                 goto bail;
 455         }
 456         osb->sys_root_inode = new;
 457
 458         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 459              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 460                 if (!ocfs2_need_system_inode(osb, i))
 461                         continue;
 462                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 463                 if (!new) {
 464                         ocfs2_release_system_inodes(osb);
 465                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 466                         mlog_errno(status);
 467                         mlog(ML_ERROR, "Unable to load system inode %d, "
 468                              "possibly corrupt fs?", i);
 469                         goto bail;
 470                 }
 471                 // the array now has one ref, so drop this one
 472                 iput(new);
 473         }
 474
 475 bail:
 476         if (status)
 477                 mlog_errno(status);
 478         return status;
 479 }
 480
 481 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 482 {
 483         struct inode *new = NULL;
 484         int status = 0;
 485         int i;
 486
 487         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 488              i < NUM_SYSTEM_INODES;
 489              i++) {
 490                 if (!ocfs2_need_system_inode(osb, i))
 491                         continue;
 492                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 493                 if (!new) {
 494                         ocfs2_release_system_inodes(osb);
 495                         status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
 496                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 497                              status, i, osb->slot_num);
 498                         goto bail;
 499                 }
 500                 /* the array now has one ref, so drop this one */
 501                 iput(new);
 502         }
 503
 504 bail:
 505         if (status)
 506                 mlog_errno(status);
 507         return status;
 508 }
 509
 510 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 511 {
 512         int i;
 513         struct inode *inode;
 514
 515         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 516                 inode = osb->global_system_inodes[i];
 517                 if (inode) {
 518                         iput(inode);
 519                         osb->global_system_inodes[i] = NULL;
 520                 }
 521         }
 522
 523         inode = osb->sys_root_inode;
 524         if (inode) {
 525                 iput(inode);
 526                 osb->sys_root_inode = NULL;
 527         }
 528
 529         inode = osb->root_inode;
 530         if (inode) {
 531                 iput(inode);
 532                 osb->root_inode = NULL;
 533         }
 534
 535         if (!osb->local_system_inodes)
 536                 return;
 537
 538         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 539                 if (osb->local_system_inodes[i]) {
 540                         iput(osb->local_system_inodes[i]);
 541                         osb->local_system_inodes[i] = NULL;
 542                 }
 543         }
 544
 545         kfree(osb->local_system_inodes);
 546         osb->local_system_inodes = NULL;
 547 }
 548
 549 /* We're allocating fs objects, use GFP_NOFS */
 550 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 551 {
 552         struct ocfs2_inode_info *oi;
 553
 554         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 555         if (!oi)
 556                 return NULL;
 557
 558         oi->i_sync_tid = 0;
 559         oi->i_datasync_tid = 0;
 560         memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
 561
 562         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 563         return &oi->vfs_inode;
 564 }
 565
 566 static void ocfs2_free_inode(struct inode *inode)
 567 {
 568         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 569 }
 570
 571 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 572                                                 unsigned int cbits)
 573 {
 574         unsigned int bytes = 1 << cbits;
 575         unsigned int trim = bytes;
 576         unsigned int bitshift = 32;
 577
 578         /*
 579          * i_size and all block offsets in ocfs2 are always 64 bits
 580          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 581          * 64 bit platforms, cluster size will be the limiting factor.
 582          */
 583
 584 #if BITS_PER_LONG == 32
 585         BUILD_BUG_ON(sizeof(sector_t) != 8);
 586         /*
 587          * We might be limited by page cache size.
 588          */
 589         if (bytes > PAGE_SIZE) {
 590                 bytes = PAGE_SIZE;
 591                 trim = 1;
 592                 /*
 593                  * Shift by 31 here so that we don't get larger than
 594                  * MAX_LFS_FILESIZE
 595                  */
 596                 bitshift = 31;
 597         }
 598 #endif
 599
 600         /*
 601          * Trim by a whole cluster when we can actually approach the
 602          * on-disk limits. Otherwise we can overflow i_clusters when
 603          * an extent start is at the max offset.
 604          */
 605         return (((unsigned long long)bytes) << bitshift) - trim;
 606 }
 607
 608 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 609 {
 610         int incompat_features;
 611         int ret = 0;
 612         struct mount_options parsed_options;
 613         struct ocfs2_super *osb = OCFS2_SB(sb);
 614         u32 tmp;
 615
 616         sync_filesystem(sb);
 617
 618         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 619             !ocfs2_check_set_options(sb, &parsed_options)) {
 620                 ret = -EINVAL;
 621                 goto out;
 622         }
 623
 624         tmp = OCFS2_MOUNT_NOCLUSTER;
 625         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 626                 ret = -EINVAL;
 627                 mlog(ML_ERROR, "Cannot change nocluster option on remount\n");
 628                 goto out;
 629         }
 630
 631         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 632                 OCFS2_MOUNT_HB_NONE;
 633         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 634                 ret = -EINVAL;
 635                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 636                 goto out;
 637         }
 638
 639         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 640             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 641                 ret = -EINVAL;
 642                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 643                 goto out;
 644         }
 645
 646         /* Probably don't want this on remount; it might
 647          * mess with other nodes */
 648         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 649             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 650                 ret = -EINVAL;
 651                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 652                 goto out;
 653         }
 654
 655         /* We're going to/from readonly mode. */
 656         if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
 657                 /* Disable quota accounting before remounting RO */
 658                 if (*flags & SB_RDONLY) {
 659                         ret = ocfs2_susp_quotas(osb, 0);
 660                         if (ret < 0)
 661                                 goto out;
 662                 }
 663                 /* Lock here so the check of HARD_RO and the potential
 664                  * setting of SOFT_RO is atomic. */
 665                 spin_lock(&osb->osb_lock);
 666                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 667                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 668                         ret = -EROFS;
 669                         goto unlock_osb;
 670                 }
 671
 672                 if (*flags & SB_RDONLY) {
 673                         sb->s_flags |= SB_RDONLY;
 674                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 675                 } else {
 676                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 677                                 mlog(ML_ERROR, "Cannot remount RDWR "
 678                                      "filesystem due to previous errors.\n");
 679                                 ret = -EROFS;
 680                                 goto unlock_osb;
 681                         }
 682                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 683                         if (incompat_features) {
 684                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 685                                      "of unsupported optional features "
 686                                      "(%x).\n", incompat_features);
 687                                 ret = -EINVAL;
 688                                 goto unlock_osb;
 689                         }
 690                         sb->s_flags &= ~SB_RDONLY;
 691                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 692                 }
 693                 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
 694 unlock_osb:
 695                 spin_unlock(&osb->osb_lock);
 696                 /* Enable quota accounting after remounting RW */
 697                 if (!ret && !(*flags & SB_RDONLY)) {
 698                         if (sb_any_quota_suspended(sb))
 699                                 ret = ocfs2_susp_quotas(osb, 1);
 700                         else
 701                                 ret = ocfs2_enable_quotas(osb);
 702                         if (ret < 0) {
 703                                 /* Return back changes... */
 704                                 spin_lock(&osb->osb_lock);
 705                                 sb->s_flags |= SB_RDONLY;
 706                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 707                                 spin_unlock(&osb->osb_lock);
 708                                 goto out;
 709                         }
 710                 }
 711         }
 712
 713         if (!ret) {
 714                 /* Only save off the new mount options in case of a successful
 715                  * remount. */
 716                 osb->s_mount_opt = parsed_options.mount_opt;
 717                 osb->s_atime_quantum = parsed_options.atime_quantum;
 718                 osb->preferred_slot = parsed_options.slot;
 719                 if (parsed_options.commit_interval)
 720                         osb->osb_commit_interval = parsed_options.commit_interval;
 721
 722                 if (!ocfs2_is_hard_readonly(osb))
 723                         ocfs2_set_journal_params(osb);
 724
 725                 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
 726                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 727                                                         SB_POSIXACL : 0);
 728         }
 729 out:
 730         return ret;
 731 }
 732
 733 static int ocfs2_sb_probe(struct super_block *sb,
 734                           struct buffer_head **bh,
 735                           int *sector_size,
 736                           struct ocfs2_blockcheck_stats *stats)
 737 {
 738         int status, tmpstat;
 739         struct ocfs1_vol_disk_hdr *hdr;
 740         struct ocfs2_dinode *di;
 741         int blksize;
 742
 743         *bh = NULL;
 744
 745         /* may be > 512 */
 746         *sector_size = bdev_logical_block_size(sb->s_bdev);
 747         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 748                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 749                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 750                 status = -EINVAL;
 751                 goto bail;
 752         }
 753
 754         /* Can this really happen? */
 755         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 756                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 757
 758         /* check block zero for old format */
 759         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 760         if (status < 0) {
 761                 mlog_errno(status);
 762                 goto bail;
 763         }
 764         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 765         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 766                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 767                      hdr->major_version, hdr->minor_version);
 768                 status = -EINVAL;
 769         }
 770         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 771                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 772                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 773                      hdr->signature);
 774                 status = -EINVAL;
 775         }
 776         brelse(*bh);
 777         *bh = NULL;
 778         if (status < 0) {
 779                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 780                      "upgraded before mounting with ocfs v2\n");
 781                 goto bail;
 782         }
 783
 784         /*
 785          * Now check at magic offset for 512, 1024, 2048, 4096
 786          * blocksizes.  4096 is the maximum blocksize because it is
 787          * the minimum clustersize.
 788          */
 789         status = -EINVAL;
 790         for (blksize = *sector_size;
 791              blksize <= OCFS2_MAX_BLOCKSIZE;
 792              blksize <<= 1) {
 793                 tmpstat = ocfs2_get_sector(sb, bh,
 794                                            OCFS2_SUPER_BLOCK_BLKNO,
 795                                            blksize);
 796                 if (tmpstat < 0) {
 797                         status = tmpstat;
 798                         mlog_errno(status);
 799                         break;
 800                 }
 801                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 802                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 803                 spin_lock_init(&stats->b_lock);
 804                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 805                 if (tmpstat < 0) {
 806                         brelse(*bh);
 807                         *bh = NULL;
 808                 }
 809                 if (tmpstat != -EAGAIN) {
 810                         status = tmpstat;
 811                         break;
 812                 }
 813         }
 814
 815 bail:
 816         return status;
 817 }
 818
 819 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 820 {
 821         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 822
 823         if (osb->s_mount_opt & hb_enabled) {
 824                 if (ocfs2_mount_local(osb)) {
 825                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 826                              "mounted device.\n");
 827                         return -EINVAL;
 828                 }
 829                 if (ocfs2_userspace_stack(osb)) {
 830                         mlog(ML_ERROR, "Userspace stack expected, but "
 831                              "o2cb heartbeat arguments passed to mount\n");
 832                         return -EINVAL;
 833                 }
 834                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 835                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 836                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 837                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 838                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 839                         return -EINVAL;
 840                 }
 841         }
 842
 843         if (!(osb->s_mount_opt & hb_enabled)) {
 844                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 845                     !ocfs2_userspace_stack(osb)) {
 846                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 847                              "a read-write clustered device.\n");
 848                         return -EINVAL;
 849                 }
 850         }
 851
 852         return 0;
 853 }
 854
 855 /*
 856  * If we're using a userspace stack, mount should have passed
 857  * a name that matches the disk.  If not, mount should not
 858  * have passed a stack.
 859  */
 860 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 861                                         struct mount_options *mopt)
 862 {
 863         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 864                 mlog(ML_ERROR,
 865                      "cluster stack passed to mount, but this filesystem "
 866                      "does not support it\n");
 867                 return -EINVAL;
 868         }
 869
 870         if (ocfs2_userspace_stack(osb) &&
 871             !(osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
 872             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 873                     OCFS2_STACK_LABEL_LEN)) {
 874                 mlog(ML_ERROR,
 875                      "cluster stack passed to mount (\"%s\") does not "
 876                      "match the filesystem (\"%s\")\n",
 877                      mopt->cluster_stack,
 878                      osb->osb_cluster_stack);
 879                 return -EINVAL;
 880         }
 881
 882         return 0;
 883 }
 884
 885 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 886 {
 887         int type;
 888         struct super_block *sb = osb->sb;
 889         unsigned int feature[OCFS2_MAXQUOTAS] = {
 890                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 891                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 892         int status = 0;
 893
 894         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 895                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 896                         continue;
 897                 if (unsuspend)
 898                         status = dquot_resume(sb, type);
 899                 else {
 900                         struct ocfs2_mem_dqinfo *oinfo;
 901
 902                         /* Cancel periodic syncing before suspending */
 903                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 904                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 905                         status = dquot_suspend(sb, type);
 906                 }
 907                 if (status < 0)
 908                         break;
 909         }
 910         if (status < 0)
 911                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 912                      "remount (error = %d).\n", status);
 913         return status;
 914 }
 915
 916 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 917 {
 918         struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
 919         struct super_block *sb = osb->sb;
 920         unsigned int feature[OCFS2_MAXQUOTAS] = {
 921                                         OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 922                                         OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 923         unsigned int ino[OCFS2_MAXQUOTAS] = {
 924                                         LOCAL_USER_QUOTA_SYSTEM_INODE,
 925                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 926         int status;
 927         int type;
 928
 929         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 930         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 931                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 932                         continue;
 933                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 934                                                         osb->slot_num);
 935                 if (!inode[type]) {
 936                         status = -ENOENT;
 937                         goto out_quota_off;
 938                 }
 939                 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
 940                                                 DQUOT_USAGE_ENABLED);
 941                 if (status < 0)
 942                         goto out_quota_off;
 943         }
 944
 945         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 946                 iput(inode[type]);
 947         return 0;
 948 out_quota_off:
 949         ocfs2_disable_quotas(osb);
 950         for (type = 0; type < OCFS2_MAXQUOTAS; type++)
 951                 iput(inode[type]);
 952         mlog_errno(status);
 953         return status;
 954 }
 955
 956 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 957 {
 958         int type;
 959         struct inode *inode;
 960         struct super_block *sb = osb->sb;
 961         struct ocfs2_mem_dqinfo *oinfo;
 962
 963         /* We mostly ignore errors in this function because there's not much
 964          * we can do when we see them */
 965         for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
 966                 if (!sb_has_quota_loaded(sb, type))
 967                         continue;
 968                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 969                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 970                 inode = igrab(sb->s_dquot.files[type]);
 971                 /* Turn off quotas. This will remove all dquot structures from
 972                  * memory and so they will be automatically synced to global
 973                  * quota files */
 974                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 975                                         DQUOT_LIMITS_ENABLED);
 976                 if (!inode)
 977                         continue;
 978                 iput(inode);
 979         }
 980 }
 981
 982 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 983 {
 984         struct dentry *root;
 985         int status, sector_size;
 986         struct mount_options parsed_options;
 987         struct inode *inode = NULL;
 988         struct ocfs2_super *osb = NULL;
 989         struct buffer_head *bh = NULL;
 990         char nodestr[12];
 991         struct ocfs2_blockcheck_stats stats;
 992
 993         trace_ocfs2_fill_super(sb, data, silent);
 994
 995         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
 996                 status = -EINVAL;
 997                 goto read_super_error;
 998         }
 999
1000         /* probe for superblock */
1001         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1002         if (status < 0) {
1003                 mlog(ML_ERROR, "superblock probe failed!\n");
1004                 goto read_super_error;
1005         }
1006
1007         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1008         osb = OCFS2_SB(sb);
1009         if (status < 0) {
1010                 mlog_errno(status);
1011                 goto read_super_error;
1012         }
1013         brelse(bh);
1014         bh = NULL;
1015
1016         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1017                 status = -EINVAL;
1018                 goto read_super_error;
1019         }
1020         osb->s_mount_opt = parsed_options.mount_opt;
1021         osb->s_atime_quantum = parsed_options.atime_quantum;
1022         osb->preferred_slot = parsed_options.slot;
1023         osb->osb_commit_interval = parsed_options.commit_interval;
1024
1025         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1026         osb->osb_resv_level = parsed_options.resv_level;
1027         osb->osb_dir_resv_level = parsed_options.resv_level;
1028         if (parsed_options.dir_resv_level == -1)
1029                 osb->osb_dir_resv_level = parsed_options.resv_level;
1030         else
1031                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1032
1033         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1034         if (status)
1035                 goto read_super_error;
1036
1037         sb->s_magic = OCFS2_SUPER_MAGIC;
1038
1039         sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1040                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1041
1042         /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1043          * heartbeat=none */
1044         if (bdev_read_only(sb->s_bdev)) {
1045                 if (!sb_rdonly(sb)) {
1046                         status = -EACCES;
1047                         mlog(ML_ERROR, "Readonly device detected but readonly "
1048                              "mount was not specified.\n");
1049                         goto read_super_error;
1050                 }
1051
1052                 /* You should not be able to start a local heartbeat
1053                  * on a readonly device. */
1054                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1055                         status = -EROFS;
1056                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1057                              "device.\n");
1058                         goto read_super_error;
1059                 }
1060
1061                 status = ocfs2_check_journals_nolocks(osb);
1062                 if (status < 0) {
1063                         if (status == -EROFS)
1064                                 mlog(ML_ERROR, "Recovery required on readonly "
1065                                      "file system, but write access is "
1066                                      "unavailable.\n");
1067                         else
1068                                 mlog_errno(status);
1069                         goto read_super_error;
1070                 }
1071
1072                 ocfs2_set_ro_flag(osb, 1);
1073
1074                 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1075                        "Cluster services will not be used for this mount. "
1076                        "Recovery will be skipped.\n", osb->dev_str);
1077         }
1078
1079         if (!ocfs2_is_hard_readonly(osb)) {
1080                 if (sb_rdonly(sb))
1081                         ocfs2_set_ro_flag(osb, 0);
1082         }
1083
1084         status = ocfs2_verify_heartbeat(osb);
1085         if (status < 0) {
1086                 mlog_errno(status);
1087                 goto read_super_error;
1088         }
1089
1090         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1091                                                  ocfs2_debugfs_root);
1092
1093         debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1094                             osb, &ocfs2_osb_debug_fops);
1095
1096         if (ocfs2_meta_ecc(osb))
1097                 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1098                                                         osb->osb_debug_root);
1099
1100         status = ocfs2_mount_volume(sb);
1101         if (status < 0)
1102                 goto read_super_error;
1103
1104         if (osb->root_inode)
1105                 inode = igrab(osb->root_inode);
1106
1107         if (!inode) {
1108                 status = -EIO;
1109                 mlog_errno(status);
1110                 goto read_super_error;
1111         }
1112
1113         root = d_make_root(inode);
1114         if (!root) {
1115                 status = -ENOMEM;
1116                 mlog_errno(status);
1117                 goto read_super_error;
1118         }
1119
1120         sb->s_root = root;
1121
1122         ocfs2_complete_mount_recovery(osb);
1123
1124         osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1125                                                 &ocfs2_kset->kobj);
1126         if (!osb->osb_dev_kset) {
1127                 status = -ENOMEM;
1128                 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1129                 goto read_super_error;
1130         }
1131
1132         /* Create filecheck sysfs related directories/files at
1133          * /sys/fs/ocfs2/<devname>/filecheck */
1134         if (ocfs2_filecheck_create_sysfs(osb)) {
1135                 status = -ENOMEM;
1136                 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1137                         "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1138                 goto read_super_error;
1139         }
1140
1141         if (ocfs2_mount_local(osb))
1142                 snprintf(nodestr, sizeof(nodestr), "local");
1143         else
1144                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1145
1146         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1147                "with %s data mode.\n",
1148                osb->dev_str, nodestr, osb->slot_num,
1149                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1150                "ordered");
1151
1152         if ((osb->s_mount_opt & OCFS2_MOUNT_NOCLUSTER) &&
1153            !(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_LOCAL_MOUNT))
1154                 printk(KERN_NOTICE "ocfs2: The shared device (%s) is mounted "
1155                        "without cluster aware mode.\n", osb->dev_str);
1156
1157         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1158         wake_up(&osb->osb_mount_event);
1159
1160         /* Now we can initialize quotas because we can afford to wait
1161          * for cluster locks recovery now. That also means that truncation
1162          * log recovery can happen but that waits for proper quota setup */
1163         if (!sb_rdonly(sb)) {
1164                 status = ocfs2_enable_quotas(osb);
1165                 if (status < 0) {
1166                         /* We have to err-out specially here because
1167                          * s_root is already set */
1168                         mlog_errno(status);
1169                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1170                         wake_up(&osb->osb_mount_event);
1171                         return status;
1172                 }
1173         }
1174
1175         ocfs2_complete_quota_recovery(osb);
1176
1177         /* Now we wake up again for processes waiting for quotas */
1178         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1179         wake_up(&osb->osb_mount_event);
1180
1181         /* Start this when the mount is almost sure of being successful */
1182         ocfs2_orphan_scan_start(osb);
1183
1184         return status;
1185
1186 read_super_error:
1187         brelse(bh);
1188
1189         if (status)
1190                 mlog_errno(status);
1191
1192         if (osb) {
1193                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1194                 wake_up(&osb->osb_mount_event);
1195                 ocfs2_dismount_volume(sb, 1);
1196         }
1197
1198         return status;
1199 }
1200
1201 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1202                         int flags,
1203                         const char *dev_name,
1204                         void *data)
1205 {
1206         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1207 }
1208
1209 static struct file_system_type ocfs2_fs_type = {
1210         .owner          = THIS_MODULE,
1211         .name           = "ocfs2",
1212         .mount          = ocfs2_mount,
1213         .kill_sb        = kill_block_super,
1214         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1215         .next           = NULL
1216 };
1217 MODULE_ALIAS_FS("ocfs2");
1218
1219 static int ocfs2_check_set_options(struct super_block *sb,
1220                                    struct mount_options *options)
1221 {
1222         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1223             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1224                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1225                 mlog(ML_ERROR, "User quotas were requested, but this "
1226                      "filesystem does not have the feature enabled.\n");
1227                 return 0;
1228         }
1229         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1230             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1231                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1232                 mlog(ML_ERROR, "Group quotas were requested, but this "
1233                      "filesystem does not have the feature enabled.\n");
1234                 return 0;
1235         }
1236         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1237             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1238                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1239                      "feature is not enabled\n");
1240                 return 0;
1241         }
1242         /* No ACL setting specified? Use XATTR feature... */
1243         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1244                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1245                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1246                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1247                 else
1248                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1249         }
1250         return 1;
1251 }
1252
1253 static int ocfs2_parse_options(struct super_block *sb,
1254                                char *options,
1255                                struct mount_options *mopt,
1256                                int is_remount)
1257 {
1258         int status, user_stack = 0;
1259         char *p;
1260         u32 tmp;
1261         int token, option;
1262         substring_t args[MAX_OPT_ARGS];
1263
1264         trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1265
1266         mopt->commit_interval = 0;
1267         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1268         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1269         mopt->slot = OCFS2_INVALID_SLOT;
1270         mopt->localalloc_opt = -1;
1271         mopt->cluster_stack[0] = '\0';
1272         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1273         mopt->dir_resv_level = -1;
1274
1275         if (!options) {
1276                 status = 1;
1277                 goto bail;
1278         }
1279
1280         while ((p = strsep(&options, ",")) != NULL) {
1281                 if (!*p)
1282                         continue;
1283
1284                 token = match_token(p, tokens, args);
1285                 switch (token) {
1286                 case Opt_hb_local:
1287                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1288                         break;
1289                 case Opt_hb_none:
1290                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1291                         break;
1292                 case Opt_hb_global:
1293                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1294                         break;
1295                 case Opt_barrier:
1296                         if (match_int(&args[0], &option)) {
1297                                 status = 0;
1298                                 goto bail;
1299                         }
1300                         if (option)
1301                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1302                         else
1303                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1304                         break;
1305                 case Opt_intr:
1306                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1307                         break;
1308                 case Opt_nointr:
1309                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1310                         break;
1311                 case Opt_err_panic:
1312                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1313                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1314                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1315                         break;
1316                 case Opt_err_ro:
1317                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1318                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1319                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1320                         break;
1321                 case Opt_err_cont:
1322                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1323                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1324                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1325                         break;
1326                 case Opt_data_ordered:
1327                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1328                         break;
1329                 case Opt_data_writeback:
1330                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1331                         break;
1332                 case Opt_user_xattr:
1333                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1334                         break;
1335                 case Opt_nouser_xattr:
1336                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1337                         break;
1338                 case Opt_atime_quantum:
1339                         if (match_int(&args[0], &option)) {
1340                                 status = 0;
1341                                 goto bail;
1342                         }
1343                         if (option >= 0)
1344                                 mopt->atime_quantum = option;
1345                         break;
1346                 case Opt_slot:
1347                         if (match_int(&args[0], &option)) {
1348                                 status = 0;
1349                                 goto bail;
1350                         }
1351                         if (option)
1352                                 mopt->slot = (u16)option;
1353                         break;
1354                 case Opt_commit:
1355                         if (match_int(&args[0], &option)) {
1356                                 status = 0;
1357                                 goto bail;
1358                         }
1359                         if (option < 0)
1360                                 return 0;
1361                         if (option == 0)
1362                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1363                         mopt->commit_interval = HZ * option;
1364                         break;
1365                 case Opt_localalloc:
1366                         if (match_int(&args[0], &option)) {
1367                                 status = 0;
1368                                 goto bail;
1369                         }
1370                         if (option >= 0)
1371                                 mopt->localalloc_opt = option;
1372                         break;
1373                 case Opt_localflocks:
1374                         /*
1375                          * Changing this during remount could race
1376                          * flock() requests, or "unbalance" existing
1377                          * ones (e.g., a lock is taken in one mode but
1378                          * dropped in the other). If users care enough
1379                          * to flip locking modes during remount, we
1380                          * could add a "local" flag to individual
1381                          * flock structures for proper tracking of
1382                          * state.
1383                          */
1384                         if (!is_remount)
1385                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1386                         break;
1387                 case Opt_stack:
1388                         /* Check both that the option we were passed
1389                          * is of the right length and that it is a proper
1390                          * string of the right length.
1391                          */
1392                         if (((args[0].to - args[0].from) !=
1393                              OCFS2_STACK_LABEL_LEN) ||
1394                             (strnlen(args[0].from,
1395                                      OCFS2_STACK_LABEL_LEN) !=
1396                              OCFS2_STACK_LABEL_LEN)) {
1397                                 mlog(ML_ERROR,
1398                                      "Invalid cluster_stack option\n");
1399                                 status = 0;
1400                                 goto bail;
1401                         }
1402                         memcpy(mopt->cluster_stack, args[0].from,
1403                                OCFS2_STACK_LABEL_LEN);
1404                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1405                         /*
1406                          * Open code the memcmp here as we don't have
1407                          * an osb to pass to
1408                          * ocfs2_userspace_stack().
1409                          */
1410                         if (memcmp(mopt->cluster_stack,
1411                                    OCFS2_CLASSIC_CLUSTER_STACK,
1412                                    OCFS2_STACK_LABEL_LEN))
1413                                 user_stack = 1;
1414                         break;
1415                 case Opt_inode64:
1416                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1417                         break;
1418                 case Opt_usrquota:
1419                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1420                         break;
1421                 case Opt_grpquota:
1422                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1423                         break;
1424                 case Opt_coherency_buffered:
1425                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1426                         break;
1427                 case Opt_coherency_full:
1428                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1429                         break;
1430                 case Opt_acl:
1431                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1432                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1433                         break;
1434                 case Opt_noacl:
1435                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1436                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1437                         break;
1438                 case Opt_resv_level:
1439                         if (is_remount)
1440                                 break;
1441                         if (match_int(&args[0], &option)) {
1442                                 status = 0;
1443                                 goto bail;
1444                         }
1445                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1446                             option < OCFS2_MAX_RESV_LEVEL)
1447                                 mopt->resv_level = option;
1448                         break;
1449                 case Opt_dir_resv_level:
1450                         if (is_remount)
1451                                 break;
1452                         if (match_int(&args[0], &option)) {
1453                                 status = 0;
1454                                 goto bail;
1455                         }
1456                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1457                             option < OCFS2_MAX_RESV_LEVEL)
1458                                 mopt->dir_resv_level = option;
1459                         break;
1460                 case Opt_journal_async_commit:
1461                         mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1462                         break;
1463                 case Opt_nocluster:
1464                         mopt->mount_opt |= OCFS2_MOUNT_NOCLUSTER;
1465                         break;
1466                 default:
1467                         mlog(ML_ERROR,
1468                              "Unrecognized mount option \"%s\" "
1469                              "or missing value\n", p);
1470                         status = 0;
1471                         goto bail;
1472                 }
1473         }
1474
1475         if (user_stack == 0) {
1476                 /* Ensure only one heartbeat mode */
1477                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1478                                          OCFS2_MOUNT_HB_GLOBAL |
1479                                          OCFS2_MOUNT_HB_NONE);
1480                 if (hweight32(tmp) != 1) {
1481                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1482                         status = 0;
1483                         goto bail;
1484                 }
1485         }
1486
1487         status = 1;
1488
1489 bail:
1490         return status;
1491 }
1492
1493 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1494 {
1495         struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1496         unsigned long opts = osb->s_mount_opt;
1497         unsigned int local_alloc_megs;
1498
1499         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1500                 seq_printf(s, ",_netdev");
1501                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1502                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1503                 else
1504                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1505         } else
1506                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1507
1508         if (opts & OCFS2_MOUNT_NOINTR)
1509                 seq_printf(s, ",nointr");
1510
1511         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1512                 seq_printf(s, ",data=writeback");
1513         else
1514                 seq_printf(s, ",data=ordered");
1515
1516         if (opts & OCFS2_MOUNT_BARRIER)
1517                 seq_printf(s, ",barrier=1");
1518
1519         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1520                 seq_printf(s, ",errors=panic");
1521         else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1522                 seq_printf(s, ",errors=continue");
1523         else
1524                 seq_printf(s, ",errors=remount-ro");
1525
1526         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1527                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1528
1529         seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1530
1531         if (osb->osb_commit_interval)
1532                 seq_printf(s, ",commit=%u",
1533                            (unsigned) (osb->osb_commit_interval / HZ));
1534
1535         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1536         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1537                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1538
1539         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1540                 seq_printf(s, ",localflocks,");
1541
1542         if (osb->osb_cluster_stack[0])
1543                 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1544                                   OCFS2_STACK_LABEL_LEN);
1545         if (opts & OCFS2_MOUNT_USRQUOTA)
1546                 seq_printf(s, ",usrquota");
1547         if (opts & OCFS2_MOUNT_GRPQUOTA)
1548                 seq_printf(s, ",grpquota");
1549
1550         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1551                 seq_printf(s, ",coherency=buffered");
1552         else
1553                 seq_printf(s, ",coherency=full");
1554
1555         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1556                 seq_printf(s, ",nouser_xattr");
1557         else
1558                 seq_printf(s, ",user_xattr");
1559
1560         if (opts & OCFS2_MOUNT_INODE64)
1561                 seq_printf(s, ",inode64");
1562
1563         if (opts & OCFS2_MOUNT_POSIX_ACL)
1564                 seq_printf(s, ",acl");
1565         else
1566                 seq_printf(s, ",noacl");
1567
1568         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1569                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1570
1571         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1572                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1573
1574         if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1575                 seq_printf(s, ",journal_async_commit");
1576
1577         if (opts & OCFS2_MOUNT_NOCLUSTER)
1578                 seq_printf(s, ",nocluster");
1579
1580         return 0;
1581 }
1582
1583 static int __init ocfs2_init(void)
1584 {
1585         int status;
1586
1587         status = init_ocfs2_uptodate_cache();
1588         if (status < 0)
1589                 goto out1;
1590
1591         status = ocfs2_initialize_mem_caches();
1592         if (status < 0)
1593                 goto out2;
1594
1595         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1596
1597         ocfs2_set_locking_protocol();
1598
1599         status = register_quota_format(&ocfs2_quota_format);
1600         if (status < 0)
1601                 goto out3;
1602         status = register_filesystem(&ocfs2_fs_type);
1603         if (!status)
1604                 return 0;
1605
1606         unregister_quota_format(&ocfs2_quota_format);
1607 out3:
1608         debugfs_remove(ocfs2_debugfs_root);
1609         ocfs2_free_mem_caches();
1610 out2:
1611         exit_ocfs2_uptodate_cache();
1612 out1:
1613         mlog_errno(status);
1614         return status;
1615 }
1616
1617 static void __exit ocfs2_exit(void)
1618 {
1619         unregister_quota_format(&ocfs2_quota_format);
1620
1621         debugfs_remove(ocfs2_debugfs_root);
1622
1623         ocfs2_free_mem_caches();
1624
1625         unregister_filesystem(&ocfs2_fs_type);
1626
1627         exit_ocfs2_uptodate_cache();
1628 }
1629
1630 static void ocfs2_put_super(struct super_block *sb)
1631 {
1632         trace_ocfs2_put_super(sb);
1633
1634         ocfs2_sync_blockdev(sb);
1635         ocfs2_dismount_volume(sb, 0);
1636 }
1637
1638 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1639 {
1640         struct ocfs2_super *osb;
1641         u32 numbits, freebits;
1642         int status;
1643         struct ocfs2_dinode *bm_lock;
1644         struct buffer_head *bh = NULL;
1645         struct inode *inode = NULL;
1646
1647         trace_ocfs2_statfs(dentry->d_sb, buf);
1648
1649         osb = OCFS2_SB(dentry->d_sb);
1650
1651         inode = ocfs2_get_system_file_inode(osb,
1652                                             GLOBAL_BITMAP_SYSTEM_INODE,
1653                                             OCFS2_INVALID_SLOT);
1654         if (!inode) {
1655                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1656                 status = -EIO;
1657                 goto bail;
1658         }
1659
1660         status = ocfs2_inode_lock(inode, &bh, 0);
1661         if (status < 0) {
1662                 mlog_errno(status);
1663                 goto bail;
1664         }
1665
1666         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1667
1668         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1669         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1670
1671         buf->f_type = OCFS2_SUPER_MAGIC;
1672         buf->f_bsize = dentry->d_sb->s_blocksize;
1673         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1674         buf->f_blocks = ((sector_t) numbits) *
1675                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1676         buf->f_bfree = ((sector_t) freebits) *
1677                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1678         buf->f_bavail = buf->f_bfree;
1679         buf->f_files = numbits;
1680         buf->f_ffree = freebits;
1681         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1682                                 & 0xFFFFFFFFUL;
1683         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1684                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1685
1686         brelse(bh);
1687
1688         ocfs2_inode_unlock(inode, 0);
1689         status = 0;
1690 bail:
1691         iput(inode);
1692
1693         if (status)
1694                 mlog_errno(status);
1695
1696         return status;
1697 }
1698
1699 static void ocfs2_inode_init_once(void *data)
1700 {
1701         struct ocfs2_inode_info *oi = data;
1702
1703         oi->ip_flags = 0;
1704         oi->ip_open_count = 0;
1705         spin_lock_init(&oi->ip_lock);
1706         ocfs2_extent_map_init(&oi->vfs_inode);
1707         INIT_LIST_HEAD(&oi->ip_io_markers);
1708         INIT_LIST_HEAD(&oi->ip_unwritten_list);
1709         oi->ip_dir_start_lookup = 0;
1710         init_rwsem(&oi->ip_alloc_sem);
1711         init_rwsem(&oi->ip_xattr_sem);
1712         mutex_init(&oi->ip_io_mutex);
1713
1714         oi->ip_blkno = 0ULL;
1715         oi->ip_clusters = 0;
1716         oi->ip_next_orphan = NULL;
1717
1718         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1719
1720         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1721         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1722         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1723
1724         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1725                                   &ocfs2_inode_caching_ops);
1726
1727         inode_init_once(&oi->vfs_inode);
1728 }
1729
1730 static int ocfs2_initialize_mem_caches(void)
1731 {
1732         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1733                                        sizeof(struct ocfs2_inode_info),
1734                                        0,
1735                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1736                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1737                                        ocfs2_inode_init_once);
1738         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1739                                         sizeof(struct ocfs2_dquot),
1740                                         0,
1741                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1742                                                 SLAB_MEM_SPREAD),
1743                                         NULL);
1744         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1745                                         sizeof(struct ocfs2_quota_chunk),
1746                                         0,
1747                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1748                                         NULL);
1749         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1750             !ocfs2_qf_chunk_cachep) {
1751                 kmem_cache_destroy(ocfs2_inode_cachep);
1752                 kmem_cache_destroy(ocfs2_dquot_cachep);
1753                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1754                 return -ENOMEM;
1755         }
1756
1757         return 0;
1758 }
1759
1760 static void ocfs2_free_mem_caches(void)
1761 {
1762         /*
1763          * Make sure all delayed rcu free inodes are flushed before we
1764          * destroy cache.
1765          */
1766         rcu_barrier();
1767         kmem_cache_destroy(ocfs2_inode_cachep);
1768         ocfs2_inode_cachep = NULL;
1769
1770         kmem_cache_destroy(ocfs2_dquot_cachep);
1771         ocfs2_dquot_cachep = NULL;
1772
1773         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1774         ocfs2_qf_chunk_cachep = NULL;
1775 }
1776
1777 static int ocfs2_get_sector(struct super_block *sb,
1778                             struct buffer_head **bh,
1779                             int block,
1780                             int sect_size)
1781 {
1782         if (!sb_set_blocksize(sb, sect_size)) {
1783                 mlog(ML_ERROR, "unable to set blocksize\n");
1784                 return -EIO;
1785         }
1786
1787         *bh = sb_getblk(sb, block);
1788         if (!*bh) {
1789                 mlog_errno(-ENOMEM);
1790                 return -ENOMEM;
1791         }
1792         lock_buffer(*bh);
1793         if (!buffer_dirty(*bh))
1794                 clear_buffer_uptodate(*bh);
1795         unlock_buffer(*bh);
1796         ll_rw_block(REQ_OP_READ, 0, 1, bh);
1797         wait_on_buffer(*bh);
1798         if (!buffer_uptodate(*bh)) {
1799                 mlog_errno(-EIO);
1800                 brelse(*bh);
1801                 *bh = NULL;
1802                 return -EIO;
1803         }
1804
1805         return 0;
1806 }
1807
1808 static int ocfs2_mount_volume(struct super_block *sb)
1809 {
1810         int status = 0;
1811         int unlock_super = 0;
1812         struct ocfs2_super *osb = OCFS2_SB(sb);
1813
1814         if (ocfs2_is_hard_readonly(osb))
1815                 goto leave;
1816
1817         mutex_init(&osb->obs_trim_fs_mutex);
1818
1819         status = ocfs2_dlm_init(osb);
1820         if (status < 0) {
1821                 mlog_errno(status);
1822                 if (status == -EBADR && ocfs2_userspace_stack(osb))
1823                         mlog(ML_ERROR, "couldn't mount because cluster name on"
1824                         " disk does not match the running cluster name.\n");
1825                 goto leave;
1826         }
1827
1828         status = ocfs2_super_lock(osb, 1);
1829         if (status < 0) {
1830                 mlog_errno(status);
1831                 goto leave;
1832         }
1833         unlock_super = 1;
1834
1835         /* This will load up the node map and add ourselves to it. */
1836         status = ocfs2_find_slot(osb);
1837         if (status < 0) {
1838                 mlog_errno(status);
1839                 goto leave;
1840         }
1841
1842         /* load all node-local system inodes */
1843         status = ocfs2_init_local_system_inodes(osb);
1844         if (status < 0) {
1845                 mlog_errno(status);
1846                 goto leave;
1847         }
1848
1849         status = ocfs2_check_volume(osb);
1850         if (status < 0) {
1851                 mlog_errno(status);
1852                 goto leave;
1853         }
1854
1855         status = ocfs2_truncate_log_init(osb);
1856         if (status < 0)
1857                 mlog_errno(status);
1858
1859 leave:
1860         if (unlock_super)
1861                 ocfs2_super_unlock(osb, 1);
1862
1863         return status;
1864 }
1865
1866 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1867 {
1868         int tmp, hangup_needed = 0;
1869         struct ocfs2_super *osb = NULL;
1870         char nodestr[12];
1871
1872         trace_ocfs2_dismount_volume(sb);
1873
1874         BUG_ON(!sb);
1875         osb = OCFS2_SB(sb);
1876         BUG_ON(!osb);
1877
1878         /* Remove file check sysfs related directores/files,
1879          * and wait for the pending file check operations */
1880         ocfs2_filecheck_remove_sysfs(osb);
1881
1882         kset_unregister(osb->osb_dev_kset);
1883
1884         /* Orphan scan should be stopped as early as possible */
1885         ocfs2_orphan_scan_stop(osb);
1886
1887         ocfs2_disable_quotas(osb);
1888
1889         /* All dquots should be freed by now */
1890         WARN_ON(!llist_empty(&osb->dquot_drop_list));
1891         /* Wait for worker to be done with the work structure in osb */
1892         cancel_work_sync(&osb->dquot_drop_work);
1893
1894         ocfs2_shutdown_local_alloc(osb);
1895
1896         ocfs2_truncate_log_shutdown(osb);
1897
1898         /* This will disable recovery and flush any recovery work. */
1899         ocfs2_recovery_exit(osb);
1900
1901         ocfs2_journal_shutdown(osb);
1902
1903         ocfs2_sync_blockdev(sb);
1904
1905         ocfs2_purge_refcount_trees(osb);
1906
1907         /* No cluster connection means we've failed during mount, so skip
1908          * all the steps which depended on that to complete. */
1909         if (osb->cconn) {
1910                 tmp = ocfs2_super_lock(osb, 1);
1911                 if (tmp < 0) {
1912                         mlog_errno(tmp);
1913                         return;
1914                 }
1915         }
1916
1917         if (osb->slot_num != OCFS2_INVALID_SLOT)
1918                 ocfs2_put_slot(osb);
1919
1920         if (osb->cconn)
1921                 ocfs2_super_unlock(osb, 1);
1922
1923         ocfs2_release_system_inodes(osb);
1924
1925         /*
1926          * If we're dismounting due to mount error, mount.ocfs2 will clean
1927          * up heartbeat.  If we're a local mount, there is no heartbeat.
1928          * If we failed before we got a uuid_str yet, we can't stop
1929          * heartbeat.  Otherwise, do it.
1930          */
1931         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1932             !ocfs2_is_hard_readonly(osb))
1933                 hangup_needed = 1;
1934
1935         if (osb->cconn)
1936                 ocfs2_dlm_shutdown(osb, hangup_needed);
1937
1938         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1939         debugfs_remove_recursive(osb->osb_debug_root);
1940
1941         if (hangup_needed)
1942                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1943
1944         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1945
1946         if (ocfs2_mount_local(osb))
1947                 snprintf(nodestr, sizeof(nodestr), "local");
1948         else
1949                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1950
1951         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1952                osb->dev_str, nodestr);
1953
1954         ocfs2_delete_osb(osb);
1955         kfree(osb);
1956         sb->s_dev = 0;
1957         sb->s_fs_info = NULL;
1958 }
1959
1960 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1961                                 unsigned uuid_bytes)
1962 {
1963         int i, ret;
1964         char *ptr;
1965
1966         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1967
1968         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1969         if (osb->uuid_str == NULL)
1970                 return -ENOMEM;
1971
1972         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1973                 /* print with null */
1974                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1975                 if (ret != 2) /* drop super cleans up */
1976                         return -EINVAL;
1977                 /* then only advance past the last char */
1978                 ptr += 2;
1979         }
1980
1981         return 0;
1982 }
1983
1984 /* Make sure entire volume is addressable by our journal.  Requires
1985    osb_clusters_at_boot to be valid and for the journal to have been
1986    initialized by ocfs2_journal_init(). */
1987 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1988 {
1989         int status = 0;
1990         u64 max_block =
1991                 ocfs2_clusters_to_blocks(osb->sb,
1992                                          osb->osb_clusters_at_boot) - 1;
1993
1994         /* 32-bit block number is always OK. */
1995         if (max_block <= (u32)~0ULL)
1996                 goto out;
1997
1998         /* Volume is "huge", so see if our journal is new enough to
1999            support it. */
2000         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2001                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2002               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2003                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2004                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2005                      "Enable the 'block64' journal option with tunefs.ocfs2");
2006                 status = -EFBIG;
2007                 goto out;
2008         }
2009
2010  out:
2011         return status;
2012 }
2013
2014 static int ocfs2_initialize_super(struct super_block *sb,
2015                                   struct buffer_head *bh,
2016                                   int sector_size,
2017                                   struct ocfs2_blockcheck_stats *stats)
2018 {
2019         int status;
2020         int i, cbits, bbits;
2021         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2022         struct inode *inode = NULL;
2023         struct ocfs2_journal *journal;
2024         struct ocfs2_super *osb;
2025         u64 total_blocks;
2026
2027         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2028         if (!osb) {
2029                 status = -ENOMEM;
2030                 mlog_errno(status);
2031                 goto bail;
2032         }
2033
2034         sb->s_fs_info = osb;
2035         sb->s_op = &ocfs2_sops;
2036         sb->s_d_op = &ocfs2_dentry_ops;
2037         sb->s_export_op = &ocfs2_export_ops;
2038         sb->s_qcop = &dquot_quotactl_sysfile_ops;
2039         sb->dq_op = &ocfs2_quota_operations;
2040         sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2041         sb->s_xattr = ocfs2_xattr_handlers;
2042         sb->s_time_gran = 1;
2043         sb->s_flags |= SB_NOATIME;
2044         /* this is needed to support O_LARGEFILE */
2045         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2046         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2047         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2048         memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2049                sizeof(di->id2.i_super.s_uuid));
2050
2051         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2052
2053         for (i = 0; i < 3; i++)
2054                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2055         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2056
2057         osb->sb = sb;
2058         osb->s_sectsize_bits = blksize_bits(sector_size);
2059         BUG_ON(!osb->s_sectsize_bits);
2060
2061         spin_lock_init(&osb->dc_task_lock);
2062         init_waitqueue_head(&osb->dc_event);
2063         osb->dc_work_sequence = 0;
2064         osb->dc_wake_sequence = 0;
2065         INIT_LIST_HEAD(&osb->blocked_lock_list);
2066         osb->blocked_lock_count = 0;
2067         spin_lock_init(&osb->osb_lock);
2068         spin_lock_init(&osb->osb_xattr_lock);
2069         ocfs2_init_steal_slots(osb);
2070
2071         mutex_init(&osb->system_file_mutex);
2072
2073         atomic_set(&osb->alloc_stats.moves, 0);
2074         atomic_set(&osb->alloc_stats.local_data, 0);
2075         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2076         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2077         atomic_set(&osb->alloc_stats.bg_extends, 0);
2078
2079         /* Copy the blockcheck stats from the superblock probe */
2080         osb->osb_ecc_stats = *stats;
2081
2082         ocfs2_init_node_maps(osb);
2083
2084         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2085                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2086
2087         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2088         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2089                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2090                      osb->max_slots);
2091                 status = -EINVAL;
2092                 goto bail;
2093         }
2094
2095         ocfs2_orphan_scan_init(osb);
2096
2097         status = ocfs2_recovery_init(osb);
2098         if (status) {
2099                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2100                 mlog_errno(status);
2101                 goto bail;
2102         }
2103
2104         init_waitqueue_head(&osb->checkpoint_event);
2105
2106         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2107
2108         osb->slot_num = OCFS2_INVALID_SLOT;
2109
2110         osb->s_xattr_inline_size = le16_to_cpu(
2111                                         di->id2.i_super.s_xattr_inline_size);
2112
2113         osb->local_alloc_state = OCFS2_LA_UNUSED;
2114         osb->local_alloc_bh = NULL;
2115         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2116
2117         init_waitqueue_head(&osb->osb_mount_event);
2118
2119         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2120         if (status) {
2121                 mlog_errno(status);
2122                 goto bail;
2123         }
2124
2125         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2126         if (!osb->vol_label) {
2127                 mlog(ML_ERROR, "unable to alloc vol label\n");
2128                 status = -ENOMEM;
2129                 goto bail;
2130         }
2131
2132         osb->slot_recovery_generations =
2133                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2134                         GFP_KERNEL);
2135         if (!osb->slot_recovery_generations) {
2136                 status = -ENOMEM;
2137                 mlog_errno(status);
2138                 goto bail;
2139         }
2140
2141         init_waitqueue_head(&osb->osb_wipe_event);
2142         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2143                                         sizeof(*osb->osb_orphan_wipes),
2144                                         GFP_KERNEL);
2145         if (!osb->osb_orphan_wipes) {
2146                 status = -ENOMEM;
2147                 mlog_errno(status);
2148                 goto bail;
2149         }
2150
2151         osb->osb_rf_lock_tree = RB_ROOT;
2152
2153         osb->s_feature_compat =
2154                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2155         osb->s_feature_ro_compat =
2156                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2157         osb->s_feature_incompat =
2158                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2159
2160         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2161                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2162                      "optional features (%x).\n", i);
2163                 status = -EINVAL;
2164                 goto bail;
2165         }
2166         if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2167                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2168                      "unsupported optional features (%x).\n", i);
2169                 status = -EINVAL;
2170                 goto bail;
2171         }
2172
2173         if (ocfs2_clusterinfo_valid(osb)) {
2174                 osb->osb_stackflags =
2175                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2176                 strlcpy(osb->osb_cluster_stack,
2177                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2178                        OCFS2_STACK_LABEL_LEN + 1);
2179                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2180                         mlog(ML_ERROR,
2181                              "couldn't mount because of an invalid "
2182                              "cluster stack label (%s) \n",
2183                              osb->osb_cluster_stack);
2184                         status = -EINVAL;
2185                         goto bail;
2186                 }
2187                 strlcpy(osb->osb_cluster_name,
2188                         OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2189                         OCFS2_CLUSTER_NAME_LEN + 1);
2190         } else {
2191                 /* The empty string is identical with classic tools that
2192                  * don't know about s_cluster_info. */
2193                 osb->osb_cluster_stack[0] = '\0';
2194         }
2195
2196         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2197
2198         /* FIXME
2199          * This should be done in ocfs2_journal_init(), but unknown
2200          * ordering issues will cause the filesystem to crash.
2201          * If anyone wants to figure out what part of the code
2202          * refers to osb->journal before ocfs2_journal_init() is run,
2203          * be my guest.
2204          */
2205         /* initialize our journal structure */
2206
2207         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2208         if (!journal) {
2209                 mlog(ML_ERROR, "unable to alloc journal\n");
2210                 status = -ENOMEM;
2211                 goto bail;
2212         }
2213         osb->journal = journal;
2214         journal->j_osb = osb;
2215
2216         atomic_set(&journal->j_num_trans, 0);
2217         init_rwsem(&journal->j_trans_barrier);
2218         init_waitqueue_head(&journal->j_checkpointed);
2219         spin_lock_init(&journal->j_lock);
2220         journal->j_trans_id = (unsigned long) 1;
2221         INIT_LIST_HEAD(&journal->j_la_cleanups);
2222         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2223         journal->j_state = OCFS2_JOURNAL_FREE;
2224
2225         INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2226         init_llist_head(&osb->dquot_drop_list);
2227
2228         /* get some pseudo constants for clustersize bits */
2229         osb->s_clustersize_bits =
2230                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2231         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2232
2233         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2234             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2235                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2236                      osb->s_clustersize);
2237                 status = -EINVAL;
2238                 goto bail;
2239         }
2240
2241         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2242                                                 le32_to_cpu(di->i_clusters));
2243
2244         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2245                                            total_blocks);
2246         if (status) {
2247                 mlog(ML_ERROR, "Volume too large "
2248                      "to mount safely on this system");
2249                 status = -EFBIG;
2250                 goto bail;
2251         }
2252
2253         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2254                                  sizeof(di->id2.i_super.s_uuid))) {
2255                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2256                 status = -ENOMEM;
2257                 goto bail;
2258         }
2259
2260         strlcpy(osb->vol_label, di->id2.i_super.s_label,
2261                 OCFS2_MAX_VOL_LABEL_LEN);
2262         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2263         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2264         osb->first_cluster_group_blkno =
2265                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2266         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2267         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2268         trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2269                                      (unsigned long long)osb->root_blkno,
2270                                      (unsigned long long)osb->system_dir_blkno,
2271                                      osb->s_clustersize_bits);
2272
2273         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2274         if (!osb->osb_dlm_debug) {
2275                 status = -ENOMEM;
2276                 mlog_errno(status);
2277                 goto bail;
2278         }
2279
2280         atomic_set(&osb->vol_state, VOLUME_INIT);
2281
2282         /* load root, system_dir, and all global system inodes */
2283         status = ocfs2_init_global_system_inodes(osb);
2284         if (status < 0) {
2285                 mlog_errno(status);
2286                 goto bail;
2287         }
2288
2289         /*
2290          * global bitmap
2291          */
2292         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2293                                             OCFS2_INVALID_SLOT);
2294         if (!inode) {
2295                 status = -EINVAL;
2296                 mlog_errno(status);
2297                 goto bail;
2298         }
2299
2300         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2301         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2302         iput(inode);
2303
2304         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2305                                  osb->s_feature_incompat) * 8;
2306
2307         status = ocfs2_init_slot_info(osb);
2308         if (status < 0) {
2309                 mlog_errno(status);
2310                 goto bail;
2311         }
2312         cleancache_init_shared_fs(sb);
2313
2314         osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2315         if (!osb->ocfs2_wq) {
2316                 status = -ENOMEM;
2317                 mlog_errno(status);
2318         }
2319
2320 bail:
2321         return status;
2322 }
2323
2324 /*
2325  * will return: -EAGAIN if it is ok to keep searching for superblocks
2326  *              -EINVAL if there is a bad superblock
2327  *              0 on success
2328  */
2329 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2330                                struct buffer_head *bh,
2331                                u32 blksz,
2332                                struct ocfs2_blockcheck_stats *stats)
2333 {
2334         int status = -EAGAIN;
2335
2336         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2337                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2338                 /* We have to do a raw check of the feature here */
2339                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2340                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2341                         status = ocfs2_block_check_validate(bh->b_data,
2342                                                             bh->b_size,
2343                                                             &di->i_check,
2344                                                             stats);
2345                         if (status)
2346                                 goto out;
2347                 }
2348                 status = -EINVAL;
2349                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2350                         mlog(ML_ERROR, "found superblock with incorrect block "
2351                              "size: found %u, should be %u\n",
2352                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2353                                blksz);
2354                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2355                            OCFS2_MAJOR_REV_LEVEL ||
2356                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2357                            OCFS2_MINOR_REV_LEVEL) {
2358                         mlog(ML_ERROR, "found superblock with bad version: "
2359                              "found %u.%u, should be %u.%u\n",
2360                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2361                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2362                              OCFS2_MAJOR_REV_LEVEL,
2363                              OCFS2_MINOR_REV_LEVEL);
2364                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2365                         mlog(ML_ERROR, "bad block number on superblock: "
2366                              "found %llu, should be %llu\n",
2367                              (unsigned long long)le64_to_cpu(di->i_blkno),
2368                              (unsigned long long)bh->b_blocknr);
2369                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2370                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2371                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2372                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2373                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2374                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2375                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2376                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2377                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2378                         mlog(ML_ERROR,
2379                              "Superblock slots found greater than file system "
2380                              "maximum: found %u, max %u\n",
2381                              le16_to_cpu(di->id2.i_super.s_max_slots),
2382                              OCFS2_MAX_SLOTS);
2383                 } else {
2384                         /* found it! */
2385                         status = 0;
2386                 }
2387         }
2388
2389 out:
2390         if (status && status != -EAGAIN)
2391                 mlog_errno(status);
2392         return status;
2393 }
2394
2395 static int ocfs2_check_volume(struct ocfs2_super *osb)
2396 {
2397         int status;
2398         int dirty;
2399         int local;
2400         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2401                                                   * recover
2402                                                   * ourselves. */
2403
2404         /* Init our journal object. */
2405         status = ocfs2_journal_init(osb->journal, &dirty);
2406         if (status < 0) {
2407                 mlog(ML_ERROR, "Could not initialize journal!\n");
2408                 goto finally;
2409         }
2410
2411         /* Now that journal has been initialized, check to make sure
2412            entire volume is addressable. */
2413         status = ocfs2_journal_addressable(osb);
2414         if (status)
2415                 goto finally;
2416
2417         /* If the journal was unmounted cleanly then we don't want to
2418          * recover anything. Otherwise, journal_load will do that
2419          * dirty work for us :) */
2420         if (!dirty) {
2421                 status = ocfs2_journal_wipe(osb->journal, 0);
2422                 if (status < 0) {
2423                         mlog_errno(status);
2424                         goto finally;
2425                 }
2426         } else {
2427                 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2428                        "unmounted cleanly, recovering it.\n", osb->dev_str);
2429         }
2430
2431         local = ocfs2_mount_local(osb);
2432
2433         /* will play back anything left in the journal. */
2434         status = ocfs2_journal_load(osb->journal, local, dirty);
2435         if (status < 0) {
2436                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2437                 goto finally;
2438         }
2439
2440         if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2441                 jbd2_journal_set_features(osb->journal->j_journal,
2442                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2443                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2444         else
2445                 jbd2_journal_clear_features(osb->journal->j_journal,
2446                                 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2447                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2448
2449         if (dirty) {
2450                 /* recover my local alloc if we didn't unmount cleanly. */
2451                 status = ocfs2_begin_local_alloc_recovery(osb,
2452                                                           osb->slot_num,
2453                                                           &local_alloc);
2454                 if (status < 0) {
2455                         mlog_errno(status);
2456                         goto finally;
2457                 }
2458                 /* we complete the recovery process after we've marked
2459                  * ourselves as mounted. */
2460         }
2461
2462         status = ocfs2_load_local_alloc(osb);
2463         if (status < 0) {
2464                 mlog_errno(status);
2465                 goto finally;
2466         }
2467
2468         if (dirty) {
2469                 /* Recovery will be completed after we've mounted the
2470                  * rest of the volume. */
2471                 osb->local_alloc_copy = local_alloc;
2472                 local_alloc = NULL;
2473         }
2474
2475         /* go through each journal, trylock it and if you get the
2476          * lock, and it's marked as dirty, set the bit in the recover
2477          * map and launch a recovery thread for it. */
2478         status = ocfs2_mark_dead_nodes(osb);
2479         if (status < 0) {
2480                 mlog_errno(status);
2481                 goto finally;
2482         }
2483
2484         status = ocfs2_compute_replay_slots(osb);
2485         if (status < 0)
2486                 mlog_errno(status);
2487
2488 finally:
2489         kfree(local_alloc);
2490
2491         if (status)
2492                 mlog_errno(status);
2493         return status;
2494 }
2495
2496 /*
2497  * The routine gets called from dismount or close whenever a dismount on
2498  * volume is requested and the osb open count becomes 1.
2499  * It will remove the osb from the global list and also free up all the
2500  * initialized resources and fileobject.
2501  */
2502 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2503 {
2504         /* This function assumes that the caller has the main osb resource */
2505
2506         /* ocfs2_initializer_super have already created this workqueue */
2507         if (osb->ocfs2_wq)
2508                 destroy_workqueue(osb->ocfs2_wq);
2509
2510         ocfs2_free_slot_info(osb);
2511
2512         kfree(osb->osb_orphan_wipes);
2513         kfree(osb->slot_recovery_generations);
2514         /* FIXME
2515          * This belongs in journal shutdown, but because we have to
2516          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2517          * we free it here.
2518          */
2519         kfree(osb->journal);
2520         kfree(osb->local_alloc_copy);
2521         kfree(osb->uuid_str);
2522         kfree(osb->vol_label);
2523         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2524         memset(osb, 0, sizeof(struct ocfs2_super));
2525 }
2526
2527 /* Depending on the mount option passed, perform one of the following:
2528  * Put OCFS2 into a readonly state (default)
2529  * Return EIO so that only the process errs
2530  * Fix the error as if fsck.ocfs2 -y
2531  * panic
2532  */
2533 static int ocfs2_handle_error(struct super_block *sb)
2534 {
2535         struct ocfs2_super *osb = OCFS2_SB(sb);
2536         int rv = 0;
2537
2538         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2539         pr_crit("On-disk corruption discovered. "
2540                 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2541
2542         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2543                 panic("OCFS2: (device %s): panic forced after error\n",
2544                       sb->s_id);
2545         } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2546                 pr_crit("OCFS2: Returning error to the calling process.\n");
2547                 rv = -EIO;
2548         } else { /* default option */
2549                 rv = -EROFS;
2550                 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2551                         return rv;
2552
2553                 pr_crit("OCFS2: File system is now read-only.\n");
2554                 sb->s_flags |= SB_RDONLY;
2555                 ocfs2_set_ro_flag(osb, 0);
2556         }
2557
2558         return rv;
2559 }
2560
2561 int __ocfs2_error(struct super_block *sb, const char *function,
2562                   const char *fmt, ...)
2563 {
2564         struct va_format vaf;
2565         va_list args;
2566
2567         va_start(args, fmt);
2568         vaf.fmt = fmt;
2569         vaf.va = &args;
2570
2571         /* Not using mlog here because we want to show the actual
2572          * function the error came from. */
2573         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2574                sb->s_id, function, &vaf);
2575
2576         va_end(args);
2577
2578         return ocfs2_handle_error(sb);
2579 }
2580
2581 /* Handle critical errors. This is intentionally more drastic than
2582  * ocfs2_handle_error, so we only use for things like journal errors,
2583  * etc. */
2584 void __ocfs2_abort(struct super_block *sb, const char *function,
2585                    const char *fmt, ...)
2586 {
2587         struct va_format vaf;
2588         va_list args;
2589
2590         va_start(args, fmt);
2591
2592         vaf.fmt = fmt;
2593         vaf.va = &args;
2594
2595         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2596                sb->s_id, function, &vaf);
2597
2598         va_end(args);
2599
2600         /* We don't have the cluster support yet to go straight to
2601          * hard readonly in here. Until then, we want to keep
2602          * ocfs2_abort() so that we can at least mark critical
2603          * errors.
2604          *
2605          * TODO: This should abort the journal and alert other nodes
2606          * that our slot needs recovery. */
2607
2608         /* Force a panic(). This stinks, but it's better than letting
2609          * things continue without having a proper hard readonly
2610          * here. */
2611         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2612                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2613         ocfs2_handle_error(sb);
2614 }
2615
2616 /*
2617  * Void signal blockers, because in-kernel sigprocmask() only fails
2618  * when SIG_* is wrong.
2619  */
2620 void ocfs2_block_signals(sigset_t *oldset)
2621 {
2622         int rc;
2623         sigset_t blocked;
2624
2625         sigfillset(&blocked);
2626         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2627         BUG_ON(rc);
2628 }
2629
2630 void ocfs2_unblock_signals(sigset_t *oldset)
2631 {
2632         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2633         BUG_ON(rc);
2634 }
2635
2636 module_init(ocfs2_init);
2637 module_exit(ocfs2_exit);