Pileus Git - ~andy/linux/blob - fs/ocfs2/super.c

   1 /* -*- mode: c; c-basic-offset: 8; -*-
   2  * vim: noexpandtab sw=8 ts=8 sts=0:
   3  *
   4  * super.c
   5  *
   6  * load/unload driver, mount/dismount volumes
   7  *
   8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
   9  *
  10  * This program is free software; you can redistribute it and/or
  11  * modify it under the terms of the GNU General Public
  12  * License as published by the Free Software Foundation; either
  13  * version 2 of the License, or (at your option) any later version.
  14  *
  15  * This program is distributed in the hope that it will be useful,
  16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  18  * General Public License for more details.
  19  *
  20  * You should have received a copy of the GNU General Public
  21  * License along with this program; if not, write to the
  22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
  23  * Boston, MA 021110-1307, USA.
  24  */
  25
  26 #include <linux/module.h>
  27 #include <linux/fs.h>
  28 #include <linux/types.h>
  29 #include <linux/slab.h>
  30 #include <linux/highmem.h>
  31 #include <linux/init.h>
  32 #include <linux/random.h>
  33 #include <linux/statfs.h>
  34 #include <linux/moduleparam.h>
  35 #include <linux/blkdev.h>
  36 #include <linux/socket.h>
  37 #include <linux/inet.h>
  38 #include <linux/parser.h>
  39 #include <linux/crc32.h>
  40 #include <linux/debugfs.h>
  41 #include <linux/mount.h>
  42 #include <linux/seq_file.h>
  43 #include <linux/quotaops.h>
  44
  45 #define MLOG_MASK_PREFIX ML_SUPER
  46 #include <cluster/masklog.h>
  47
  48 #include "ocfs2.h"
  49
  50 /* this should be the only file to include a version 1 header */
  51 #include "ocfs1_fs_compat.h"
  52
  53 #include "alloc.h"
  54 #include "blockcheck.h"
  55 #include "dlmglue.h"
  56 #include "export.h"
  57 #include "extent_map.h"
  58 #include "heartbeat.h"
  59 #include "inode.h"
  60 #include "journal.h"
  61 #include "localalloc.h"
  62 #include "namei.h"
  63 #include "slot_map.h"
  64 #include "super.h"
  65 #include "sysfile.h"
  66 #include "uptodate.h"
  67 #include "ver.h"
  68 #include "xattr.h"
  69 #include "quota.h"
  70 #include "refcounttree.h"
  71 #include "suballoc.h"
  72
  73 #include "buffer_head_io.h"
  74
  75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  76 struct kmem_cache *ocfs2_dquot_cachep;
  77 struct kmem_cache *ocfs2_qf_chunk_cachep;
  78
  79 /* OCFS2 needs to schedule several differnt types of work which
  80  * require cluster locking, disk I/O, recovery waits, etc. Since these
  81  * types of work tend to be heavy we avoid using the kernel events
  82  * workqueue and schedule on our own. */
  83 struct workqueue_struct *ocfs2_wq = NULL;
  84
  85 static struct dentry *ocfs2_debugfs_root = NULL;
  86
  87 MODULE_AUTHOR("Oracle");
  88 MODULE_LICENSE("GPL");
  89
  90 struct mount_options
  91 {
  92         unsigned long   commit_interval;
  93         unsigned long   mount_opt;
  94         unsigned int    atime_quantum;
  95         signed short    slot;
  96         int             localalloc_opt;
  97         unsigned int    resv_level;
  98         int             dir_resv_level;
  99         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 100 };
 101
 102 static int ocfs2_parse_options(struct super_block *sb, char *options,
 103                                struct mount_options *mopt,
 104                                int is_remount);
 105 static int ocfs2_check_set_options(struct super_block *sb,
 106                                    struct mount_options *options);
 107 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 108 static void ocfs2_put_super(struct super_block *sb);
 109 static int ocfs2_mount_volume(struct super_block *sb);
 110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 112 static int ocfs2_initialize_mem_caches(void);
 113 static void ocfs2_free_mem_caches(void);
 114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 115
 116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 117
 118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 119
 120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 123 static int ocfs2_check_volume(struct ocfs2_super *osb);
 124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 125                                struct buffer_head *bh,
 126                                u32 sectsize,
 127                                struct ocfs2_blockcheck_stats *stats);
 128 static int ocfs2_initialize_super(struct super_block *sb,
 129                                   struct buffer_head *bh,
 130                                   int sector_size,
 131                                   struct ocfs2_blockcheck_stats *stats);
 132 static int ocfs2_get_sector(struct super_block *sb,
 133                             struct buffer_head **bh,
 134                             int block,
 135                             int sect_size);
 136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 137 static void ocfs2_destroy_inode(struct inode *inode);
 138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 141
 142 static const struct super_operations ocfs2_sops = {
 143         .statfs         = ocfs2_statfs,
 144         .alloc_inode    = ocfs2_alloc_inode,
 145         .destroy_inode  = ocfs2_destroy_inode,
 146         .drop_inode     = ocfs2_drop_inode,
 147         .evict_inode    = ocfs2_evict_inode,
 148         .sync_fs        = ocfs2_sync_fs,
 149         .put_super      = ocfs2_put_super,
 150         .remount_fs     = ocfs2_remount,
 151         .show_options   = ocfs2_show_options,
 152         .quota_read     = ocfs2_quota_read,
 153         .quota_write    = ocfs2_quota_write,
 154 };
 155
 156 enum {
 157         Opt_barrier,
 158         Opt_err_panic,
 159         Opt_err_ro,
 160         Opt_intr,
 161         Opt_nointr,
 162         Opt_hb_none,
 163         Opt_hb_local,
 164         Opt_hb_global,
 165         Opt_data_ordered,
 166         Opt_data_writeback,
 167         Opt_atime_quantum,
 168         Opt_slot,
 169         Opt_commit,
 170         Opt_localalloc,
 171         Opt_localflocks,
 172         Opt_stack,
 173         Opt_user_xattr,
 174         Opt_nouser_xattr,
 175         Opt_inode64,
 176         Opt_acl,
 177         Opt_noacl,
 178         Opt_usrquota,
 179         Opt_grpquota,
 180         Opt_coherency_buffered,
 181         Opt_coherency_full,
 182         Opt_resv_level,
 183         Opt_dir_resv_level,
 184         Opt_err,
 185 };
 186
 187 static const match_table_t tokens = {
 188         {Opt_barrier, "barrier=%u"},
 189         {Opt_err_panic, "errors=panic"},
 190         {Opt_err_ro, "errors=remount-ro"},
 191         {Opt_intr, "intr"},
 192         {Opt_nointr, "nointr"},
 193         {Opt_hb_none, OCFS2_HB_NONE},
 194         {Opt_hb_local, OCFS2_HB_LOCAL},
 195         {Opt_hb_global, OCFS2_HB_GLOBAL},
 196         {Opt_data_ordered, "data=ordered"},
 197         {Opt_data_writeback, "data=writeback"},
 198         {Opt_atime_quantum, "atime_quantum=%u"},
 199         {Opt_slot, "preferred_slot=%u"},
 200         {Opt_commit, "commit=%u"},
 201         {Opt_localalloc, "localalloc=%d"},
 202         {Opt_localflocks, "localflocks"},
 203         {Opt_stack, "cluster_stack=%s"},
 204         {Opt_user_xattr, "user_xattr"},
 205         {Opt_nouser_xattr, "nouser_xattr"},
 206         {Opt_inode64, "inode64"},
 207         {Opt_acl, "acl"},
 208         {Opt_noacl, "noacl"},
 209         {Opt_usrquota, "usrquota"},
 210         {Opt_grpquota, "grpquota"},
 211         {Opt_coherency_buffered, "coherency=buffered"},
 212         {Opt_coherency_full, "coherency=full"},
 213         {Opt_resv_level, "resv_level=%u"},
 214         {Opt_dir_resv_level, "dir_resv_level=%u"},
 215         {Opt_err, NULL}
 216 };
 217
 218 #ifdef CONFIG_DEBUG_FS
 219 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 220 {
 221         struct ocfs2_cluster_connection *cconn = osb->cconn;
 222         struct ocfs2_recovery_map *rm = osb->recovery_map;
 223         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 224         int i, out = 0;
 225
 226         out += snprintf(buf + out, len - out,
 227                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 228                         "Device", osb->dev_str, osb->uuid_str,
 229                         osb->fs_generation, osb->vol_label);
 230
 231         out += snprintf(buf + out, len - out,
 232                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 233                         atomic_read(&osb->vol_state), osb->osb_flags);
 234
 235         out += snprintf(buf + out, len - out,
 236                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 237                         osb->sb->s_blocksize, osb->s_clustersize);
 238
 239         out += snprintf(buf + out, len - out,
 240                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 241                         "ROcompat: 0x%X\n",
 242                         "Features", osb->s_feature_compat,
 243                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 244
 245         out += snprintf(buf + out, len - out,
 246                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 247                         osb->s_mount_opt, osb->s_atime_quantum);
 248
 249         if (cconn) {
 250                 out += snprintf(buf + out, len - out,
 251                                 "%10s => Stack: %s  Name: %*s  "
 252                                 "Version: %d.%d\n", "Cluster",
 253                                 (*osb->osb_cluster_stack == '\0' ?
 254                                  "o2cb" : osb->osb_cluster_stack),
 255                                 cconn->cc_namelen, cconn->cc_name,
 256                                 cconn->cc_version.pv_major,
 257                                 cconn->cc_version.pv_minor);
 258         }
 259
 260         spin_lock(&osb->dc_task_lock);
 261         out += snprintf(buf + out, len - out,
 262                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 263                         "WorkSeq: %lu\n", "DownCnvt",
 264                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 265                         osb->blocked_lock_count, osb->dc_wake_sequence,
 266                         osb->dc_work_sequence);
 267         spin_unlock(&osb->dc_task_lock);
 268
 269         spin_lock(&osb->osb_lock);
 270         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 271                         "Recovery",
 272                         (osb->recovery_thread_task ?
 273                          task_pid_nr(osb->recovery_thread_task) : -1));
 274         if (rm->rm_used == 0)
 275                 out += snprintf(buf + out, len - out, " None\n");
 276         else {
 277                 for (i = 0; i < rm->rm_used; i++)
 278                         out += snprintf(buf + out, len - out, " %d",
 279                                         rm->rm_entries[i]);
 280                 out += snprintf(buf + out, len - out, "\n");
 281         }
 282         spin_unlock(&osb->osb_lock);
 283
 284         out += snprintf(buf + out, len - out,
 285                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 286                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 287                         osb->osb_commit_interval,
 288                         atomic_read(&osb->needs_checkpoint));
 289
 290         out += snprintf(buf + out, len - out,
 291                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 292                         "Journal", osb->journal->j_state,
 293                         osb->journal->j_trans_id,
 294                         atomic_read(&osb->journal->j_num_trans));
 295
 296         out += snprintf(buf + out, len - out,
 297                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 298                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 299                         "Stats",
 300                         atomic_read(&osb->alloc_stats.bitmap_data),
 301                         atomic_read(&osb->alloc_stats.local_data),
 302                         atomic_read(&osb->alloc_stats.bg_allocs),
 303                         atomic_read(&osb->alloc_stats.moves),
 304                         atomic_read(&osb->alloc_stats.bg_extends));
 305
 306         out += snprintf(buf + out, len - out,
 307                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 308                         "Default: %u bits\n",
 309                         "LocalAlloc", osb->local_alloc_state,
 310                         (unsigned long long)osb->la_last_gd,
 311                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 312
 313         spin_lock(&osb->osb_lock);
 314         out += snprintf(buf + out, len - out,
 315                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 316                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 317                         osb->s_inode_steal_slot,
 318                         atomic_read(&osb->s_num_inodes_stolen),
 319                         osb->s_meta_steal_slot,
 320                         atomic_read(&osb->s_num_meta_stolen));
 321         spin_unlock(&osb->osb_lock);
 322
 323         out += snprintf(buf + out, len - out, "OrphanScan => ");
 324         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 325                         os->os_count, os->os_seqno);
 326         out += snprintf(buf + out, len - out, " Last Scan: ");
 327         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 328                 out += snprintf(buf + out, len - out, "Disabled\n");
 329         else
 330                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 331                                 (get_seconds() - os->os_scantime.tv_sec));
 332
 333         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 334                         "Slots", "Num", "RecoGen");
 335         for (i = 0; i < osb->max_slots; ++i) {
 336                 out += snprintf(buf + out, len - out,
 337                                 "%10s  %c %3d  %10d\n",
 338                                 " ",
 339                                 (i == osb->slot_num ? '*' : ' '),
 340                                 i, osb->slot_recovery_generations[i]);
 341         }
 342
 343         return out;
 344 }
 345
 346 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 347 {
 348         struct ocfs2_super *osb = inode->i_private;
 349         char *buf = NULL;
 350
 351         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 352         if (!buf)
 353                 goto bail;
 354
 355         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 356
 357         file->private_data = buf;
 358
 359         return 0;
 360 bail:
 361         return -ENOMEM;
 362 }
 363
 364 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 365 {
 366         kfree(file->private_data);
 367         return 0;
 368 }
 369
 370 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 371                                 size_t nbytes, loff_t *ppos)
 372 {
 373         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 374                                        i_size_read(file->f_mapping->host));
 375 }
 376 #else
 377 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 382 {
 383         return 0;
 384 }
 385 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 386                                 size_t nbytes, loff_t *ppos)
 387 {
 388         return 0;
 389 }
 390 #endif  /* CONFIG_DEBUG_FS */
 391
 392 static const struct file_operations ocfs2_osb_debug_fops = {
 393         .open =         ocfs2_osb_debug_open,
 394         .release =      ocfs2_debug_release,
 395         .read =         ocfs2_debug_read,
 396         .llseek =       generic_file_llseek,
 397 };
 398
 399 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 400 {
 401         int status;
 402         tid_t target;
 403         struct ocfs2_super *osb = OCFS2_SB(sb);
 404
 405         if (ocfs2_is_hard_readonly(osb))
 406                 return -EROFS;
 407
 408         if (wait) {
 409                 status = ocfs2_flush_truncate_log(osb);
 410                 if (status < 0)
 411                         mlog_errno(status);
 412         } else {
 413                 ocfs2_schedule_truncate_log_flush(osb, 0);
 414         }
 415
 416         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 417                                       &target)) {
 418                 if (wait)
 419                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 420                                              target);
 421         }
 422         return 0;
 423 }
 424
 425 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 426 {
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 428             && (ino == USER_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 432             && (ino == GROUP_QUOTA_SYSTEM_INODE
 433                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 434                 return 0;
 435         return 1;
 436 }
 437
 438 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 439 {
 440         struct inode *new = NULL;
 441         int status = 0;
 442         int i;
 443
 444         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 445         if (IS_ERR(new)) {
 446                 status = PTR_ERR(new);
 447                 mlog_errno(status);
 448                 goto bail;
 449         }
 450         osb->root_inode = new;
 451
 452         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 453         if (IS_ERR(new)) {
 454                 status = PTR_ERR(new);
 455                 mlog_errno(status);
 456                 goto bail;
 457         }
 458         osb->sys_root_inode = new;
 459
 460         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 461              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 462                 if (!ocfs2_need_system_inode(osb, i))
 463                         continue;
 464                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 465                 if (!new) {
 466                         ocfs2_release_system_inodes(osb);
 467                         status = -EINVAL;
 468                         mlog_errno(status);
 469                         /* FIXME: Should ERROR_RO_FS */
 470                         mlog(ML_ERROR, "Unable to load system inode %d, "
 471                              "possibly corrupt fs?", i);
 472                         goto bail;
 473                 }
 474                 // the array now has one ref, so drop this one
 475                 iput(new);
 476         }
 477
 478 bail:
 479         if (status)
 480                 mlog_errno(status);
 481         return status;
 482 }
 483
 484 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 485 {
 486         struct inode *new = NULL;
 487         int status = 0;
 488         int i;
 489
 490         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 491              i < NUM_SYSTEM_INODES;
 492              i++) {
 493                 if (!ocfs2_need_system_inode(osb, i))
 494                         continue;
 495                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 496                 if (!new) {
 497                         ocfs2_release_system_inodes(osb);
 498                         status = -EINVAL;
 499                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 500                              status, i, osb->slot_num);
 501                         goto bail;
 502                 }
 503                 /* the array now has one ref, so drop this one */
 504                 iput(new);
 505         }
 506
 507 bail:
 508         if (status)
 509                 mlog_errno(status);
 510         return status;
 511 }
 512
 513 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 514 {
 515         int i;
 516         struct inode *inode;
 517
 518         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 519                 inode = osb->global_system_inodes[i];
 520                 if (inode) {
 521                         iput(inode);
 522                         osb->global_system_inodes[i] = NULL;
 523                 }
 524         }
 525
 526         inode = osb->sys_root_inode;
 527         if (inode) {
 528                 iput(inode);
 529                 osb->sys_root_inode = NULL;
 530         }
 531
 532         inode = osb->root_inode;
 533         if (inode) {
 534                 iput(inode);
 535                 osb->root_inode = NULL;
 536         }
 537
 538         if (!osb->local_system_inodes)
 539                 return;
 540
 541         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 542                 if (osb->local_system_inodes[i]) {
 543                         iput(osb->local_system_inodes[i]);
 544                         osb->local_system_inodes[i] = NULL;
 545                 }
 546         }
 547
 548         kfree(osb->local_system_inodes);
 549         osb->local_system_inodes = NULL;
 550 }
 551
 552 /* We're allocating fs objects, use GFP_NOFS */
 553 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 554 {
 555         struct ocfs2_inode_info *oi;
 556
 557         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 558         if (!oi)
 559                 return NULL;
 560
 561         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 562         return &oi->vfs_inode;
 563 }
 564
 565 static void ocfs2_i_callback(struct rcu_head *head)
 566 {
 567         struct inode *inode = container_of(head, struct inode, i_rcu);
 568         INIT_LIST_HEAD(&inode->i_dentry);
 569         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 570 }
 571
 572 static void ocfs2_destroy_inode(struct inode *inode)
 573 {
 574         call_rcu(&inode->i_rcu, ocfs2_i_callback);
 575 }
 576
 577 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 578                                                 unsigned int cbits)
 579 {
 580         unsigned int bytes = 1 << cbits;
 581         unsigned int trim = bytes;
 582         unsigned int bitshift = 32;
 583
 584         /*
 585          * i_size and all block offsets in ocfs2 are always 64 bits
 586          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 587          * 64 bit platforms, cluster size will be the limiting factor.
 588          */
 589
 590 #if BITS_PER_LONG == 32
 591 # if defined(CONFIG_LBDAF)
 592         BUILD_BUG_ON(sizeof(sector_t) != 8);
 593         /*
 594          * We might be limited by page cache size.
 595          */
 596         if (bytes > PAGE_CACHE_SIZE) {
 597                 bytes = PAGE_CACHE_SIZE;
 598                 trim = 1;
 599                 /*
 600                  * Shift by 31 here so that we don't get larger than
 601                  * MAX_LFS_FILESIZE
 602                  */
 603                 bitshift = 31;
 604         }
 605 # else
 606         /*
 607          * We are limited by the size of sector_t. Use block size, as
 608          * that's what we expose to the VFS.
 609          */
 610         bytes = 1 << bbits;
 611         trim = 1;
 612         bitshift = 31;
 613 # endif
 614 #endif
 615
 616         /*
 617          * Trim by a whole cluster when we can actually approach the
 618          * on-disk limits. Otherwise we can overflow i_clusters when
 619          * an extent start is at the max offset.
 620          */
 621         return (((unsigned long long)bytes) << bitshift) - trim;
 622 }
 623
 624 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 625 {
 626         int incompat_features;
 627         int ret = 0;
 628         struct mount_options parsed_options;
 629         struct ocfs2_super *osb = OCFS2_SB(sb);
 630         u32 tmp;
 631
 632         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 633             !ocfs2_check_set_options(sb, &parsed_options)) {
 634                 ret = -EINVAL;
 635                 goto out;
 636         }
 637
 638         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 639                 OCFS2_MOUNT_HB_NONE;
 640         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 641                 ret = -EINVAL;
 642                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 643                 goto out;
 644         }
 645
 646         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 647             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 648                 ret = -EINVAL;
 649                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 650                 goto out;
 651         }
 652
 653         /* Probably don't want this on remount; it might
 654          * mess with other nodes */
 655         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 656             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 657                 ret = -EINVAL;
 658                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 659                 goto out;
 660         }
 661
 662         /* We're going to/from readonly mode. */
 663         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 664                 /* Disable quota accounting before remounting RO */
 665                 if (*flags & MS_RDONLY) {
 666                         ret = ocfs2_susp_quotas(osb, 0);
 667                         if (ret < 0)
 668                                 goto out;
 669                 }
 670                 /* Lock here so the check of HARD_RO and the potential
 671                  * setting of SOFT_RO is atomic. */
 672                 spin_lock(&osb->osb_lock);
 673                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 674                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 675                         ret = -EROFS;
 676                         goto unlock_osb;
 677                 }
 678
 679                 if (*flags & MS_RDONLY) {
 680                         mlog(0, "Going to ro mode.\n");
 681                         sb->s_flags |= MS_RDONLY;
 682                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 683                 } else {
 684                         mlog(0, "Making ro filesystem writeable.\n");
 685
 686                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 687                                 mlog(ML_ERROR, "Cannot remount RDWR "
 688                                      "filesystem due to previous errors.\n");
 689                                 ret = -EROFS;
 690                                 goto unlock_osb;
 691                         }
 692                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 693                         if (incompat_features) {
 694                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 695                                      "of unsupported optional features "
 696                                      "(%x).\n", incompat_features);
 697                                 ret = -EINVAL;
 698                                 goto unlock_osb;
 699                         }
 700                         sb->s_flags &= ~MS_RDONLY;
 701                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 702                 }
 703 unlock_osb:
 704                 spin_unlock(&osb->osb_lock);
 705                 /* Enable quota accounting after remounting RW */
 706                 if (!ret && !(*flags & MS_RDONLY)) {
 707                         if (sb_any_quota_suspended(sb))
 708                                 ret = ocfs2_susp_quotas(osb, 1);
 709                         else
 710                                 ret = ocfs2_enable_quotas(osb);
 711                         if (ret < 0) {
 712                                 /* Return back changes... */
 713                                 spin_lock(&osb->osb_lock);
 714                                 sb->s_flags |= MS_RDONLY;
 715                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 716                                 spin_unlock(&osb->osb_lock);
 717                                 goto out;
 718                         }
 719                 }
 720         }
 721
 722         if (!ret) {
 723                 /* Only save off the new mount options in case of a successful
 724                  * remount. */
 725                 osb->s_mount_opt = parsed_options.mount_opt;
 726                 osb->s_atime_quantum = parsed_options.atime_quantum;
 727                 osb->preferred_slot = parsed_options.slot;
 728                 if (parsed_options.commit_interval)
 729                         osb->osb_commit_interval = parsed_options.commit_interval;
 730
 731                 if (!ocfs2_is_hard_readonly(osb))
 732                         ocfs2_set_journal_params(osb);
 733
 734                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 735                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 736                                                         MS_POSIXACL : 0);
 737         }
 738 out:
 739         return ret;
 740 }
 741
 742 static int ocfs2_sb_probe(struct super_block *sb,
 743                           struct buffer_head **bh,
 744                           int *sector_size,
 745                           struct ocfs2_blockcheck_stats *stats)
 746 {
 747         int status, tmpstat;
 748         struct ocfs1_vol_disk_hdr *hdr;
 749         struct ocfs2_dinode *di;
 750         int blksize;
 751
 752         *bh = NULL;
 753
 754         /* may be > 512 */
 755         *sector_size = bdev_logical_block_size(sb->s_bdev);
 756         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 757                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 758                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 759                 status = -EINVAL;
 760                 goto bail;
 761         }
 762
 763         /* Can this really happen? */
 764         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 765                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 766
 767         /* check block zero for old format */
 768         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 769         if (status < 0) {
 770                 mlog_errno(status);
 771                 goto bail;
 772         }
 773         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 774         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 775                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 776                      hdr->major_version, hdr->minor_version);
 777                 status = -EINVAL;
 778         }
 779         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 780                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 781                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 782                      hdr->signature);
 783                 status = -EINVAL;
 784         }
 785         brelse(*bh);
 786         *bh = NULL;
 787         if (status < 0) {
 788                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 789                      "upgraded before mounting with ocfs v2\n");
 790                 goto bail;
 791         }
 792
 793         /*
 794          * Now check at magic offset for 512, 1024, 2048, 4096
 795          * blocksizes.  4096 is the maximum blocksize because it is
 796          * the minimum clustersize.
 797          */
 798         status = -EINVAL;
 799         for (blksize = *sector_size;
 800              blksize <= OCFS2_MAX_BLOCKSIZE;
 801              blksize <<= 1) {
 802                 tmpstat = ocfs2_get_sector(sb, bh,
 803                                            OCFS2_SUPER_BLOCK_BLKNO,
 804                                            blksize);
 805                 if (tmpstat < 0) {
 806                         status = tmpstat;
 807                         mlog_errno(status);
 808                         break;
 809                 }
 810                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 811                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 812                 spin_lock_init(&stats->b_lock);
 813                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 814                 if (tmpstat < 0) {
 815                         brelse(*bh);
 816                         *bh = NULL;
 817                 }
 818                 if (tmpstat != -EAGAIN) {
 819                         status = tmpstat;
 820                         break;
 821                 }
 822         }
 823
 824 bail:
 825         return status;
 826 }
 827
 828 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 829 {
 830         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 831
 832         if (osb->s_mount_opt & hb_enabled) {
 833                 if (ocfs2_mount_local(osb)) {
 834                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 835                              "mounted device.\n");
 836                         return -EINVAL;
 837                 }
 838                 if (ocfs2_userspace_stack(osb)) {
 839                         mlog(ML_ERROR, "Userspace stack expected, but "
 840                              "o2cb heartbeat arguments passed to mount\n");
 841                         return -EINVAL;
 842                 }
 843                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 844                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 845                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 846                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 847                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 848                         return -EINVAL;
 849                 }
 850         }
 851
 852         if (!(osb->s_mount_opt & hb_enabled)) {
 853                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 854                     !ocfs2_userspace_stack(osb)) {
 855                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 856                              "a read-write clustered device.\n");
 857                         return -EINVAL;
 858                 }
 859         }
 860
 861         return 0;
 862 }
 863
 864 /*
 865  * If we're using a userspace stack, mount should have passed
 866  * a name that matches the disk.  If not, mount should not
 867  * have passed a stack.
 868  */
 869 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 870                                         struct mount_options *mopt)
 871 {
 872         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 873                 mlog(ML_ERROR,
 874                      "cluster stack passed to mount, but this filesystem "
 875                      "does not support it\n");
 876                 return -EINVAL;
 877         }
 878
 879         if (ocfs2_userspace_stack(osb) &&
 880             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 881                     OCFS2_STACK_LABEL_LEN)) {
 882                 mlog(ML_ERROR,
 883                      "cluster stack passed to mount (\"%s\") does not "
 884                      "match the filesystem (\"%s\")\n",
 885                      mopt->cluster_stack,
 886                      osb->osb_cluster_stack);
 887                 return -EINVAL;
 888         }
 889
 890         return 0;
 891 }
 892
 893 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 894 {
 895         int type;
 896         struct super_block *sb = osb->sb;
 897         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 898                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 899         int status = 0;
 900
 901         for (type = 0; type < MAXQUOTAS; type++) {
 902                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 903                         continue;
 904                 if (unsuspend)
 905                         status = dquot_resume(sb, type);
 906                 else {
 907                         struct ocfs2_mem_dqinfo *oinfo;
 908
 909                         /* Cancel periodic syncing before suspending */
 910                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 911                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 912                         status = dquot_suspend(sb, type);
 913                 }
 914                 if (status < 0)
 915                         break;
 916         }
 917         if (status < 0)
 918                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 919                      "remount (error = %d).\n", status);
 920         return status;
 921 }
 922
 923 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 924 {
 925         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 926         struct super_block *sb = osb->sb;
 927         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 928                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 929         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 930                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 931         int status;
 932         int type;
 933
 934         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 935         for (type = 0; type < MAXQUOTAS; type++) {
 936                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 937                         continue;
 938                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 939                                                         osb->slot_num);
 940                 if (!inode[type]) {
 941                         status = -ENOENT;
 942                         goto out_quota_off;
 943                 }
 944                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 945                                       DQUOT_USAGE_ENABLED);
 946                 if (status < 0)
 947                         goto out_quota_off;
 948         }
 949
 950         for (type = 0; type < MAXQUOTAS; type++)
 951                 iput(inode[type]);
 952         return 0;
 953 out_quota_off:
 954         ocfs2_disable_quotas(osb);
 955         for (type = 0; type < MAXQUOTAS; type++)
 956                 iput(inode[type]);
 957         mlog_errno(status);
 958         return status;
 959 }
 960
 961 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 962 {
 963         int type;
 964         struct inode *inode;
 965         struct super_block *sb = osb->sb;
 966         struct ocfs2_mem_dqinfo *oinfo;
 967
 968         /* We mostly ignore errors in this function because there's not much
 969          * we can do when we see them */
 970         for (type = 0; type < MAXQUOTAS; type++) {
 971                 if (!sb_has_quota_loaded(sb, type))
 972                         continue;
 973                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 974                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 975                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 976                 inode = igrab(sb->s_dquot.files[type]);
 977                 /* Turn off quotas. This will remove all dquot structures from
 978                  * memory and so they will be automatically synced to global
 979                  * quota files */
 980                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 981                                         DQUOT_LIMITS_ENABLED);
 982                 if (!inode)
 983                         continue;
 984                 iput(inode);
 985         }
 986 }
 987
 988 /* Handle quota on quotactl */
 989 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
 990 {
 991         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 992                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 993
 994         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 995                 return -EINVAL;
 996
 997         return dquot_enable(sb_dqopt(sb)->files[type], type,
 998                             format_id, DQUOT_LIMITS_ENABLED);
 999 }
1000
1001 /* Handle quota off quotactl */
1002 static int ocfs2_quota_off(struct super_block *sb, int type)
1003 {
1004         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1005 }
1006
1007 static const struct quotactl_ops ocfs2_quotactl_ops = {
1008         .quota_on_meta  = ocfs2_quota_on,
1009         .quota_off      = ocfs2_quota_off,
1010         .quota_sync     = dquot_quota_sync,
1011         .get_info       = dquot_get_dqinfo,
1012         .set_info       = dquot_set_dqinfo,
1013         .get_dqblk      = dquot_get_dqblk,
1014         .set_dqblk      = dquot_set_dqblk,
1015 };
1016
1017 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1018 {
1019         struct dentry *root;
1020         int status, sector_size;
1021         struct mount_options parsed_options;
1022         struct inode *inode = NULL;
1023         struct ocfs2_super *osb = NULL;
1024         struct buffer_head *bh = NULL;
1025         char nodestr[8];
1026         struct ocfs2_blockcheck_stats stats;
1027
1028         mlog(0, "%p, %p, %i", sb, data, silent);
1029
1030         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1031                 status = -EINVAL;
1032                 goto read_super_error;
1033         }
1034
1035         /* probe for superblock */
1036         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1037         if (status < 0) {
1038                 mlog(ML_ERROR, "superblock probe failed!\n");
1039                 goto read_super_error;
1040         }
1041
1042         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1043         osb = OCFS2_SB(sb);
1044         if (status < 0) {
1045                 mlog_errno(status);
1046                 goto read_super_error;
1047         }
1048         brelse(bh);
1049         bh = NULL;
1050
1051         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1052                 status = -EINVAL;
1053                 goto read_super_error;
1054         }
1055         osb->s_mount_opt = parsed_options.mount_opt;
1056         osb->s_atime_quantum = parsed_options.atime_quantum;
1057         osb->preferred_slot = parsed_options.slot;
1058         osb->osb_commit_interval = parsed_options.commit_interval;
1059
1060         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1061         osb->osb_resv_level = parsed_options.resv_level;
1062         osb->osb_dir_resv_level = parsed_options.resv_level;
1063         if (parsed_options.dir_resv_level == -1)
1064                 osb->osb_dir_resv_level = parsed_options.resv_level;
1065         else
1066                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1067
1068         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1069         if (status)
1070                 goto read_super_error;
1071
1072         sb->s_magic = OCFS2_SUPER_MAGIC;
1073
1074         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1075                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1076
1077         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1078          * heartbeat=none */
1079         if (bdev_read_only(sb->s_bdev)) {
1080                 if (!(sb->s_flags & MS_RDONLY)) {
1081                         status = -EACCES;
1082                         mlog(ML_ERROR, "Readonly device detected but readonly "
1083                              "mount was not specified.\n");
1084                         goto read_super_error;
1085                 }
1086
1087                 /* You should not be able to start a local heartbeat
1088                  * on a readonly device. */
1089                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1090                         status = -EROFS;
1091                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1092                              "device.\n");
1093                         goto read_super_error;
1094                 }
1095
1096                 status = ocfs2_check_journals_nolocks(osb);
1097                 if (status < 0) {
1098                         if (status == -EROFS)
1099                                 mlog(ML_ERROR, "Recovery required on readonly "
1100                                      "file system, but write access is "
1101                                      "unavailable.\n");
1102                         else
1103                                 mlog_errno(status);
1104                         goto read_super_error;
1105                 }
1106
1107                 ocfs2_set_ro_flag(osb, 1);
1108
1109                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1110                        "services will be utilized for this mount. Recovery "
1111                        "will be skipped.\n");
1112         }
1113
1114         if (!ocfs2_is_hard_readonly(osb)) {
1115                 if (sb->s_flags & MS_RDONLY)
1116                         ocfs2_set_ro_flag(osb, 0);
1117         }
1118
1119         status = ocfs2_verify_heartbeat(osb);
1120         if (status < 0) {
1121                 mlog_errno(status);
1122                 goto read_super_error;
1123         }
1124
1125         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1126                                                  ocfs2_debugfs_root);
1127         if (!osb->osb_debug_root) {
1128                 status = -EINVAL;
1129                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1130                 goto read_super_error;
1131         }
1132
1133         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1134                                             osb->osb_debug_root,
1135                                             osb,
1136                                             &ocfs2_osb_debug_fops);
1137         if (!osb->osb_ctxt) {
1138                 status = -EINVAL;
1139                 mlog_errno(status);
1140                 goto read_super_error;
1141         }
1142
1143         if (ocfs2_meta_ecc(osb)) {
1144                 status = ocfs2_blockcheck_stats_debugfs_install(
1145                                                 &osb->osb_ecc_stats,
1146                                                 osb->osb_debug_root);
1147                 if (status) {
1148                         mlog(ML_ERROR,
1149                              "Unable to create blockcheck statistics "
1150                              "files\n");
1151                         goto read_super_error;
1152                 }
1153         }
1154
1155         status = ocfs2_mount_volume(sb);
1156         if (osb->root_inode)
1157                 inode = igrab(osb->root_inode);
1158
1159         if (status < 0)
1160                 goto read_super_error;
1161
1162         if (!inode) {
1163                 status = -EIO;
1164                 mlog_errno(status);
1165                 goto read_super_error;
1166         }
1167
1168         root = d_alloc_root(inode);
1169         if (!root) {
1170                 status = -ENOMEM;
1171                 mlog_errno(status);
1172                 goto read_super_error;
1173         }
1174
1175         sb->s_root = root;
1176
1177         ocfs2_complete_mount_recovery(osb);
1178
1179         if (ocfs2_mount_local(osb))
1180                 snprintf(nodestr, sizeof(nodestr), "local");
1181         else
1182                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1183
1184         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1185                "with %s data mode.\n",
1186                osb->dev_str, nodestr, osb->slot_num,
1187                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1188                "ordered");
1189
1190         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1191         wake_up(&osb->osb_mount_event);
1192
1193         /* Now we can initialize quotas because we can afford to wait
1194          * for cluster locks recovery now. That also means that truncation
1195          * log recovery can happen but that waits for proper quota setup */
1196         if (!(sb->s_flags & MS_RDONLY)) {
1197                 status = ocfs2_enable_quotas(osb);
1198                 if (status < 0) {
1199                         /* We have to err-out specially here because
1200                          * s_root is already set */
1201                         mlog_errno(status);
1202                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1203                         wake_up(&osb->osb_mount_event);
1204                         return status;
1205                 }
1206         }
1207
1208         ocfs2_complete_quota_recovery(osb);
1209
1210         /* Now we wake up again for processes waiting for quotas */
1211         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1212         wake_up(&osb->osb_mount_event);
1213
1214         /* Start this when the mount is almost sure of being successful */
1215         ocfs2_orphan_scan_start(osb);
1216
1217         return status;
1218
1219 read_super_error:
1220         brelse(bh);
1221
1222         if (inode)
1223                 iput(inode);
1224
1225         if (osb) {
1226                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1227                 wake_up(&osb->osb_mount_event);
1228                 ocfs2_dismount_volume(sb, 1);
1229         }
1230
1231         if (status)
1232                 mlog_errno(status);
1233         return status;
1234 }
1235
1236 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1237                         int flags,
1238                         const char *dev_name,
1239                         void *data)
1240 {
1241         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1242 }
1243
1244 static void ocfs2_kill_sb(struct super_block *sb)
1245 {
1246         struct ocfs2_super *osb = OCFS2_SB(sb);
1247
1248         /* Failed mount? */
1249         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1250                 goto out;
1251
1252         /* Prevent further queueing of inode drop events */
1253         spin_lock(&dentry_list_lock);
1254         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1255         spin_unlock(&dentry_list_lock);
1256         /* Wait for work to finish and/or remove it */
1257         cancel_work_sync(&osb->dentry_lock_work);
1258 out:
1259         kill_block_super(sb);
1260 }
1261
1262 static struct file_system_type ocfs2_fs_type = {
1263         .owner          = THIS_MODULE,
1264         .name           = "ocfs2",
1265         .mount          = ocfs2_mount,
1266         .kill_sb        = ocfs2_kill_sb,
1267
1268         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1269         .next           = NULL
1270 };
1271
1272 static int ocfs2_check_set_options(struct super_block *sb,
1273                                    struct mount_options *options)
1274 {
1275         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1276             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1277                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1278                 mlog(ML_ERROR, "User quotas were requested, but this "
1279                      "filesystem does not have the feature enabled.\n");
1280                 return 0;
1281         }
1282         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1283             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1284                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1285                 mlog(ML_ERROR, "Group quotas were requested, but this "
1286                      "filesystem does not have the feature enabled.\n");
1287                 return 0;
1288         }
1289         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1290             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1291                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1292                      "feature is not enabled\n");
1293                 return 0;
1294         }
1295         /* No ACL setting specified? Use XATTR feature... */
1296         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1297                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1298                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1299                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1300                 else
1301                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1302         }
1303         return 1;
1304 }
1305
1306 static int ocfs2_parse_options(struct super_block *sb,
1307                                char *options,
1308                                struct mount_options *mopt,
1309                                int is_remount)
1310 {
1311         int status, user_stack = 0;
1312         char *p;
1313         u32 tmp;
1314
1315         mlog(0, "remount: %d, options: \"%s\"\n", is_remount,
1316              options ? options : "(none)");
1317
1318         mopt->commit_interval = 0;
1319         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1320         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1321         mopt->slot = OCFS2_INVALID_SLOT;
1322         mopt->localalloc_opt = -1;
1323         mopt->cluster_stack[0] = '\0';
1324         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1325         mopt->dir_resv_level = -1;
1326
1327         if (!options) {
1328                 status = 1;
1329                 goto bail;
1330         }
1331
1332         while ((p = strsep(&options, ",")) != NULL) {
1333                 int token, option;
1334                 substring_t args[MAX_OPT_ARGS];
1335
1336                 if (!*p)
1337                         continue;
1338
1339                 token = match_token(p, tokens, args);
1340                 switch (token) {
1341                 case Opt_hb_local:
1342                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1343                         break;
1344                 case Opt_hb_none:
1345                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1346                         break;
1347                 case Opt_hb_global:
1348                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1349                         break;
1350                 case Opt_barrier:
1351                         if (match_int(&args[0], &option)) {
1352                                 status = 0;
1353                                 goto bail;
1354                         }
1355                         if (option)
1356                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1357                         else
1358                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1359                         break;
1360                 case Opt_intr:
1361                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1362                         break;
1363                 case Opt_nointr:
1364                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1365                         break;
1366                 case Opt_err_panic:
1367                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1368                         break;
1369                 case Opt_err_ro:
1370                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1371                         break;
1372                 case Opt_data_ordered:
1373                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1374                         break;
1375                 case Opt_data_writeback:
1376                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1377                         break;
1378                 case Opt_user_xattr:
1379                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1380                         break;
1381                 case Opt_nouser_xattr:
1382                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1383                         break;
1384                 case Opt_atime_quantum:
1385                         if (match_int(&args[0], &option)) {
1386                                 status = 0;
1387                                 goto bail;
1388                         }
1389                         if (option >= 0)
1390                                 mopt->atime_quantum = option;
1391                         break;
1392                 case Opt_slot:
1393                         option = 0;
1394                         if (match_int(&args[0], &option)) {
1395                                 status = 0;
1396                                 goto bail;
1397                         }
1398                         if (option)
1399                                 mopt->slot = (s16)option;
1400                         break;
1401                 case Opt_commit:
1402                         option = 0;
1403                         if (match_int(&args[0], &option)) {
1404                                 status = 0;
1405                                 goto bail;
1406                         }
1407                         if (option < 0)
1408                                 return 0;
1409                         if (option == 0)
1410                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1411                         mopt->commit_interval = HZ * option;
1412                         break;
1413                 case Opt_localalloc:
1414                         option = 0;
1415                         if (match_int(&args[0], &option)) {
1416                                 status = 0;
1417                                 goto bail;
1418                         }
1419                         if (option >= 0)
1420                                 mopt->localalloc_opt = option;
1421                         break;
1422                 case Opt_localflocks:
1423                         /*
1424                          * Changing this during remount could race
1425                          * flock() requests, or "unbalance" existing
1426                          * ones (e.g., a lock is taken in one mode but
1427                          * dropped in the other). If users care enough
1428                          * to flip locking modes during remount, we
1429                          * could add a "local" flag to individual
1430                          * flock structures for proper tracking of
1431                          * state.
1432                          */
1433                         if (!is_remount)
1434                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1435                         break;
1436                 case Opt_stack:
1437                         /* Check both that the option we were passed
1438                          * is of the right length and that it is a proper
1439                          * string of the right length.
1440                          */
1441                         if (((args[0].to - args[0].from) !=
1442                              OCFS2_STACK_LABEL_LEN) ||
1443                             (strnlen(args[0].from,
1444                                      OCFS2_STACK_LABEL_LEN) !=
1445                              OCFS2_STACK_LABEL_LEN)) {
1446                                 mlog(ML_ERROR,
1447                                      "Invalid cluster_stack option\n");
1448                                 status = 0;
1449                                 goto bail;
1450                         }
1451                         memcpy(mopt->cluster_stack, args[0].from,
1452                                OCFS2_STACK_LABEL_LEN);
1453                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1454                         /*
1455                          * Open code the memcmp here as we don't have
1456                          * an osb to pass to
1457                          * ocfs2_userspace_stack().
1458                          */
1459                         if (memcmp(mopt->cluster_stack,
1460                                    OCFS2_CLASSIC_CLUSTER_STACK,
1461                                    OCFS2_STACK_LABEL_LEN))
1462                                 user_stack = 1;
1463                         break;
1464                 case Opt_inode64:
1465                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1466                         break;
1467                 case Opt_usrquota:
1468                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1469                         break;
1470                 case Opt_grpquota:
1471                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1472                         break;
1473                 case Opt_coherency_buffered:
1474                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475                         break;
1476                 case Opt_coherency_full:
1477                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1478                         break;
1479                 case Opt_acl:
1480                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1481                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1482                         break;
1483                 case Opt_noacl:
1484                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1485                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1486                         break;
1487                 case Opt_resv_level:
1488                         if (is_remount)
1489                                 break;
1490                         if (match_int(&args[0], &option)) {
1491                                 status = 0;
1492                                 goto bail;
1493                         }
1494                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1495                             option < OCFS2_MAX_RESV_LEVEL)
1496                                 mopt->resv_level = option;
1497                         break;
1498                 case Opt_dir_resv_level:
1499                         if (is_remount)
1500                                 break;
1501                         if (match_int(&args[0], &option)) {
1502                                 status = 0;
1503                                 goto bail;
1504                         }
1505                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1506                             option < OCFS2_MAX_RESV_LEVEL)
1507                                 mopt->dir_resv_level = option;
1508                         break;
1509                 default:
1510                         mlog(ML_ERROR,
1511                              "Unrecognized mount option \"%s\" "
1512                              "or missing value\n", p);
1513                         status = 0;
1514                         goto bail;
1515                 }
1516         }
1517
1518         if (user_stack == 0) {
1519                 /* Ensure only one heartbeat mode */
1520                 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1521                                          OCFS2_MOUNT_HB_GLOBAL |
1522                                          OCFS2_MOUNT_HB_NONE);
1523                 if (hweight32(tmp) != 1) {
1524                         mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1525                         status = 0;
1526                         goto bail;
1527                 }
1528         }
1529
1530         status = 1;
1531
1532 bail:
1533         return status;
1534 }
1535
1536 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1537 {
1538         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1539         unsigned long opts = osb->s_mount_opt;
1540         unsigned int local_alloc_megs;
1541
1542         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1543                 seq_printf(s, ",_netdev");
1544                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1545                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1546                 else
1547                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1548         } else
1549                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1550
1551         if (opts & OCFS2_MOUNT_NOINTR)
1552                 seq_printf(s, ",nointr");
1553
1554         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1555                 seq_printf(s, ",data=writeback");
1556         else
1557                 seq_printf(s, ",data=ordered");
1558
1559         if (opts & OCFS2_MOUNT_BARRIER)
1560                 seq_printf(s, ",barrier=1");
1561
1562         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1563                 seq_printf(s, ",errors=panic");
1564         else
1565                 seq_printf(s, ",errors=remount-ro");
1566
1567         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1568                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1569
1570         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1571                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1572
1573         if (osb->osb_commit_interval)
1574                 seq_printf(s, ",commit=%u",
1575                            (unsigned) (osb->osb_commit_interval / HZ));
1576
1577         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1578         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1579                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1580
1581         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1582                 seq_printf(s, ",localflocks,");
1583
1584         if (osb->osb_cluster_stack[0])
1585                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1586                            osb->osb_cluster_stack);
1587         if (opts & OCFS2_MOUNT_USRQUOTA)
1588                 seq_printf(s, ",usrquota");
1589         if (opts & OCFS2_MOUNT_GRPQUOTA)
1590                 seq_printf(s, ",grpquota");
1591
1592         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1593                 seq_printf(s, ",coherency=buffered");
1594         else
1595                 seq_printf(s, ",coherency=full");
1596
1597         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1598                 seq_printf(s, ",nouser_xattr");
1599         else
1600                 seq_printf(s, ",user_xattr");
1601
1602         if (opts & OCFS2_MOUNT_INODE64)
1603                 seq_printf(s, ",inode64");
1604
1605         if (opts & OCFS2_MOUNT_POSIX_ACL)
1606                 seq_printf(s, ",acl");
1607         else
1608                 seq_printf(s, ",noacl");
1609
1610         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1611                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1612
1613         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1614                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1615
1616         return 0;
1617 }
1618
1619 static int __init ocfs2_init(void)
1620 {
1621         int status;
1622
1623         ocfs2_print_version();
1624
1625         status = init_ocfs2_uptodate_cache();
1626         if (status < 0) {
1627                 mlog_errno(status);
1628                 goto leave;
1629         }
1630
1631         status = ocfs2_initialize_mem_caches();
1632         if (status < 0) {
1633                 mlog_errno(status);
1634                 goto leave;
1635         }
1636
1637         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1638         if (!ocfs2_wq) {
1639                 status = -ENOMEM;
1640                 goto leave;
1641         }
1642
1643         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1644         if (!ocfs2_debugfs_root) {
1645                 status = -EFAULT;
1646                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1647         }
1648
1649         status = ocfs2_quota_setup();
1650         if (status)
1651                 goto leave;
1652
1653         ocfs2_set_locking_protocol();
1654
1655         status = register_quota_format(&ocfs2_quota_format);
1656 leave:
1657         if (status < 0) {
1658                 ocfs2_quota_shutdown();
1659                 ocfs2_free_mem_caches();
1660                 exit_ocfs2_uptodate_cache();
1661                 mlog_errno(status);
1662         }
1663
1664         if (status >= 0) {
1665                 return register_filesystem(&ocfs2_fs_type);
1666         } else
1667                 return -1;
1668 }
1669
1670 static void __exit ocfs2_exit(void)
1671 {
1672         ocfs2_quota_shutdown();
1673
1674         if (ocfs2_wq) {
1675                 flush_workqueue(ocfs2_wq);
1676                 destroy_workqueue(ocfs2_wq);
1677         }
1678
1679         unregister_quota_format(&ocfs2_quota_format);
1680
1681         debugfs_remove(ocfs2_debugfs_root);
1682
1683         ocfs2_free_mem_caches();
1684
1685         unregister_filesystem(&ocfs2_fs_type);
1686
1687         exit_ocfs2_uptodate_cache();
1688 }
1689
1690 static void ocfs2_put_super(struct super_block *sb)
1691 {
1692         mlog(0, "(0x%p)\n", sb);
1693
1694         ocfs2_sync_blockdev(sb);
1695         ocfs2_dismount_volume(sb, 0);
1696 }
1697
1698 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1699 {
1700         struct ocfs2_super *osb;
1701         u32 numbits, freebits;
1702         int status;
1703         struct ocfs2_dinode *bm_lock;
1704         struct buffer_head *bh = NULL;
1705         struct inode *inode = NULL;
1706
1707         mlog(0, "(%p, %p)\n", dentry->d_sb, buf);
1708
1709         osb = OCFS2_SB(dentry->d_sb);
1710
1711         inode = ocfs2_get_system_file_inode(osb,
1712                                             GLOBAL_BITMAP_SYSTEM_INODE,
1713                                             OCFS2_INVALID_SLOT);
1714         if (!inode) {
1715                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1716                 status = -EIO;
1717                 goto bail;
1718         }
1719
1720         status = ocfs2_inode_lock(inode, &bh, 0);
1721         if (status < 0) {
1722                 mlog_errno(status);
1723                 goto bail;
1724         }
1725
1726         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1727
1728         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1729         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1730
1731         buf->f_type = OCFS2_SUPER_MAGIC;
1732         buf->f_bsize = dentry->d_sb->s_blocksize;
1733         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1734         buf->f_blocks = ((sector_t) numbits) *
1735                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1736         buf->f_bfree = ((sector_t) freebits) *
1737                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1738         buf->f_bavail = buf->f_bfree;
1739         buf->f_files = numbits;
1740         buf->f_ffree = freebits;
1741         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1742                                 & 0xFFFFFFFFUL;
1743         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1744                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1745
1746         brelse(bh);
1747
1748         ocfs2_inode_unlock(inode, 0);
1749         status = 0;
1750 bail:
1751         if (inode)
1752                 iput(inode);
1753
1754         if (status)
1755                 mlog_errno(status);
1756
1757         return status;
1758 }
1759
1760 static void ocfs2_inode_init_once(void *data)
1761 {
1762         struct ocfs2_inode_info *oi = data;
1763
1764         oi->ip_flags = 0;
1765         oi->ip_open_count = 0;
1766         spin_lock_init(&oi->ip_lock);
1767         ocfs2_extent_map_init(&oi->vfs_inode);
1768         INIT_LIST_HEAD(&oi->ip_io_markers);
1769         oi->ip_dir_start_lookup = 0;
1770
1771         init_rwsem(&oi->ip_alloc_sem);
1772         init_rwsem(&oi->ip_xattr_sem);
1773         mutex_init(&oi->ip_io_mutex);
1774
1775         oi->ip_blkno = 0ULL;
1776         oi->ip_clusters = 0;
1777
1778         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1779
1780         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1781         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1782         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1783
1784         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1785                                   &ocfs2_inode_caching_ops);
1786
1787         inode_init_once(&oi->vfs_inode);
1788 }
1789
1790 static int ocfs2_initialize_mem_caches(void)
1791 {
1792         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1793                                        sizeof(struct ocfs2_inode_info),
1794                                        0,
1795                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1796                                                 SLAB_MEM_SPREAD),
1797                                        ocfs2_inode_init_once);
1798         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1799                                         sizeof(struct ocfs2_dquot),
1800                                         0,
1801                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1802                                                 SLAB_MEM_SPREAD),
1803                                         NULL);
1804         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1805                                         sizeof(struct ocfs2_quota_chunk),
1806                                         0,
1807                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1808                                         NULL);
1809         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1810             !ocfs2_qf_chunk_cachep) {
1811                 if (ocfs2_inode_cachep)
1812                         kmem_cache_destroy(ocfs2_inode_cachep);
1813                 if (ocfs2_dquot_cachep)
1814                         kmem_cache_destroy(ocfs2_dquot_cachep);
1815                 if (ocfs2_qf_chunk_cachep)
1816                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1817                 return -ENOMEM;
1818         }
1819
1820         return 0;
1821 }
1822
1823 static void ocfs2_free_mem_caches(void)
1824 {
1825         if (ocfs2_inode_cachep)
1826                 kmem_cache_destroy(ocfs2_inode_cachep);
1827         ocfs2_inode_cachep = NULL;
1828
1829         if (ocfs2_dquot_cachep)
1830                 kmem_cache_destroy(ocfs2_dquot_cachep);
1831         ocfs2_dquot_cachep = NULL;
1832
1833         if (ocfs2_qf_chunk_cachep)
1834                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1835         ocfs2_qf_chunk_cachep = NULL;
1836 }
1837
1838 static int ocfs2_get_sector(struct super_block *sb,
1839                             struct buffer_head **bh,
1840                             int block,
1841                             int sect_size)
1842 {
1843         if (!sb_set_blocksize(sb, sect_size)) {
1844                 mlog(ML_ERROR, "unable to set blocksize\n");
1845                 return -EIO;
1846         }
1847
1848         *bh = sb_getblk(sb, block);
1849         if (!*bh) {
1850                 mlog_errno(-EIO);
1851                 return -EIO;
1852         }
1853         lock_buffer(*bh);
1854         if (!buffer_dirty(*bh))
1855                 clear_buffer_uptodate(*bh);
1856         unlock_buffer(*bh);
1857         ll_rw_block(READ, 1, bh);
1858         wait_on_buffer(*bh);
1859         if (!buffer_uptodate(*bh)) {
1860                 mlog_errno(-EIO);
1861                 brelse(*bh);
1862                 *bh = NULL;
1863                 return -EIO;
1864         }
1865
1866         return 0;
1867 }
1868
1869 static int ocfs2_mount_volume(struct super_block *sb)
1870 {
1871         int status = 0;
1872         int unlock_super = 0;
1873         struct ocfs2_super *osb = OCFS2_SB(sb);
1874
1875         if (ocfs2_is_hard_readonly(osb))
1876                 goto leave;
1877
1878         status = ocfs2_dlm_init(osb);
1879         if (status < 0) {
1880                 mlog_errno(status);
1881                 goto leave;
1882         }
1883
1884         status = ocfs2_super_lock(osb, 1);
1885         if (status < 0) {
1886                 mlog_errno(status);
1887                 goto leave;
1888         }
1889         unlock_super = 1;
1890
1891         /* This will load up the node map and add ourselves to it. */
1892         status = ocfs2_find_slot(osb);
1893         if (status < 0) {
1894                 mlog_errno(status);
1895                 goto leave;
1896         }
1897
1898         /* load all node-local system inodes */
1899         status = ocfs2_init_local_system_inodes(osb);
1900         if (status < 0) {
1901                 mlog_errno(status);
1902                 goto leave;
1903         }
1904
1905         status = ocfs2_check_volume(osb);
1906         if (status < 0) {
1907                 mlog_errno(status);
1908                 goto leave;
1909         }
1910
1911         status = ocfs2_truncate_log_init(osb);
1912         if (status < 0)
1913                 mlog_errno(status);
1914
1915 leave:
1916         if (unlock_super)
1917                 ocfs2_super_unlock(osb, 1);
1918
1919         return status;
1920 }
1921
1922 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1923 {
1924         int tmp, hangup_needed = 0;
1925         struct ocfs2_super *osb = NULL;
1926         char nodestr[8];
1927
1928         mlog(0, "(0x%p)\n", sb);
1929
1930         BUG_ON(!sb);
1931         osb = OCFS2_SB(sb);
1932         BUG_ON(!osb);
1933
1934         debugfs_remove(osb->osb_ctxt);
1935
1936         /*
1937          * Flush inode dropping work queue so that deletes are
1938          * performed while the filesystem is still working
1939          */
1940         ocfs2_drop_all_dl_inodes(osb);
1941
1942         /* Orphan scan should be stopped as early as possible */
1943         ocfs2_orphan_scan_stop(osb);
1944
1945         ocfs2_disable_quotas(osb);
1946
1947         ocfs2_shutdown_local_alloc(osb);
1948
1949         ocfs2_truncate_log_shutdown(osb);
1950
1951         /* This will disable recovery and flush any recovery work. */
1952         ocfs2_recovery_exit(osb);
1953
1954         ocfs2_journal_shutdown(osb);
1955
1956         ocfs2_sync_blockdev(sb);
1957
1958         ocfs2_purge_refcount_trees(osb);
1959
1960         /* No cluster connection means we've failed during mount, so skip
1961          * all the steps which depended on that to complete. */
1962         if (osb->cconn) {
1963                 tmp = ocfs2_super_lock(osb, 1);
1964                 if (tmp < 0) {
1965                         mlog_errno(tmp);
1966                         return;
1967                 }
1968         }
1969
1970         if (osb->slot_num != OCFS2_INVALID_SLOT)
1971                 ocfs2_put_slot(osb);
1972
1973         if (osb->cconn)
1974                 ocfs2_super_unlock(osb, 1);
1975
1976         ocfs2_release_system_inodes(osb);
1977
1978         /*
1979          * If we're dismounting due to mount error, mount.ocfs2 will clean
1980          * up heartbeat.  If we're a local mount, there is no heartbeat.
1981          * If we failed before we got a uuid_str yet, we can't stop
1982          * heartbeat.  Otherwise, do it.
1983          */
1984         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1985                 hangup_needed = 1;
1986
1987         if (osb->cconn)
1988                 ocfs2_dlm_shutdown(osb, hangup_needed);
1989
1990         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1991         debugfs_remove(osb->osb_debug_root);
1992
1993         if (hangup_needed)
1994                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1995
1996         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1997
1998         if (ocfs2_mount_local(osb))
1999                 snprintf(nodestr, sizeof(nodestr), "local");
2000         else
2001                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2002
2003         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2004                osb->dev_str, nodestr);
2005
2006         ocfs2_delete_osb(osb);
2007         kfree(osb);
2008         sb->s_dev = 0;
2009         sb->s_fs_info = NULL;
2010 }
2011
2012 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2013                                 unsigned uuid_bytes)
2014 {
2015         int i, ret;
2016         char *ptr;
2017
2018         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2019
2020         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2021         if (osb->uuid_str == NULL)
2022                 return -ENOMEM;
2023
2024         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2025                 /* print with null */
2026                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2027                 if (ret != 2) /* drop super cleans up */
2028                         return -EINVAL;
2029                 /* then only advance past the last char */
2030                 ptr += 2;
2031         }
2032
2033         return 0;
2034 }
2035
2036 /* Make sure entire volume is addressable by our journal.  Requires
2037    osb_clusters_at_boot to be valid and for the journal to have been
2038    initialized by ocfs2_journal_init(). */
2039 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2040 {
2041         int status = 0;
2042         u64 max_block =
2043                 ocfs2_clusters_to_blocks(osb->sb,
2044                                          osb->osb_clusters_at_boot) - 1;
2045
2046         /* 32-bit block number is always OK. */
2047         if (max_block <= (u32)~0ULL)
2048                 goto out;
2049
2050         /* Volume is "huge", so see if our journal is new enough to
2051            support it. */
2052         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2053                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2054               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2055                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2056                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2057                      "Enable the 'block64' journal option with tunefs.ocfs2");
2058                 status = -EFBIG;
2059                 goto out;
2060         }
2061
2062  out:
2063         return status;
2064 }
2065
2066 static int ocfs2_initialize_super(struct super_block *sb,
2067                                   struct buffer_head *bh,
2068                                   int sector_size,
2069                                   struct ocfs2_blockcheck_stats *stats)
2070 {
2071         int status;
2072         int i, cbits, bbits;
2073         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2074         struct inode *inode = NULL;
2075         struct ocfs2_journal *journal;
2076         __le32 uuid_net_key;
2077         struct ocfs2_super *osb;
2078         u64 total_blocks;
2079
2080         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2081         if (!osb) {
2082                 status = -ENOMEM;
2083                 mlog_errno(status);
2084                 goto bail;
2085         }
2086
2087         sb->s_fs_info = osb;
2088         sb->s_op = &ocfs2_sops;
2089         sb->s_d_op = &ocfs2_dentry_ops;
2090         sb->s_export_op = &ocfs2_export_ops;
2091         sb->s_qcop = &ocfs2_quotactl_ops;
2092         sb->dq_op = &ocfs2_quota_operations;
2093         sb->s_xattr = ocfs2_xattr_handlers;
2094         sb->s_time_gran = 1;
2095         sb->s_flags |= MS_NOATIME;
2096         /* this is needed to support O_LARGEFILE */
2097         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2098         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2099         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2100
2101         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2102
2103         for (i = 0; i < 3; i++)
2104                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2105         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2106
2107         osb->sb = sb;
2108         /* Save off for ocfs2_rw_direct */
2109         osb->s_sectsize_bits = blksize_bits(sector_size);
2110         BUG_ON(!osb->s_sectsize_bits);
2111
2112         spin_lock_init(&osb->dc_task_lock);
2113         init_waitqueue_head(&osb->dc_event);
2114         osb->dc_work_sequence = 0;
2115         osb->dc_wake_sequence = 0;
2116         INIT_LIST_HEAD(&osb->blocked_lock_list);
2117         osb->blocked_lock_count = 0;
2118         spin_lock_init(&osb->osb_lock);
2119         spin_lock_init(&osb->osb_xattr_lock);
2120         ocfs2_init_steal_slots(osb);
2121
2122         atomic_set(&osb->alloc_stats.moves, 0);
2123         atomic_set(&osb->alloc_stats.local_data, 0);
2124         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2125         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2126         atomic_set(&osb->alloc_stats.bg_extends, 0);
2127
2128         /* Copy the blockcheck stats from the superblock probe */
2129         osb->osb_ecc_stats = *stats;
2130
2131         ocfs2_init_node_maps(osb);
2132
2133         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2134                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2135
2136         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2137         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2138                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2139                      osb->max_slots);
2140                 status = -EINVAL;
2141                 goto bail;
2142         }
2143         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2144
2145         ocfs2_orphan_scan_init(osb);
2146
2147         status = ocfs2_recovery_init(osb);
2148         if (status) {
2149                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2150                 mlog_errno(status);
2151                 goto bail;
2152         }
2153
2154         init_waitqueue_head(&osb->checkpoint_event);
2155         atomic_set(&osb->needs_checkpoint, 0);
2156
2157         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2158
2159         osb->slot_num = OCFS2_INVALID_SLOT;
2160
2161         osb->s_xattr_inline_size = le16_to_cpu(
2162                                         di->id2.i_super.s_xattr_inline_size);
2163
2164         osb->local_alloc_state = OCFS2_LA_UNUSED;
2165         osb->local_alloc_bh = NULL;
2166         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2167
2168         init_waitqueue_head(&osb->osb_mount_event);
2169
2170         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2171         if (status) {
2172                 mlog_errno(status);
2173                 goto bail;
2174         }
2175
2176         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2177         if (!osb->vol_label) {
2178                 mlog(ML_ERROR, "unable to alloc vol label\n");
2179                 status = -ENOMEM;
2180                 goto bail;
2181         }
2182
2183         osb->slot_recovery_generations =
2184                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2185                         GFP_KERNEL);
2186         if (!osb->slot_recovery_generations) {
2187                 status = -ENOMEM;
2188                 mlog_errno(status);
2189                 goto bail;
2190         }
2191
2192         init_waitqueue_head(&osb->osb_wipe_event);
2193         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2194                                         sizeof(*osb->osb_orphan_wipes),
2195                                         GFP_KERNEL);
2196         if (!osb->osb_orphan_wipes) {
2197                 status = -ENOMEM;
2198                 mlog_errno(status);
2199                 goto bail;
2200         }
2201
2202         osb->osb_rf_lock_tree = RB_ROOT;
2203
2204         osb->s_feature_compat =
2205                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2206         osb->s_feature_ro_compat =
2207                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2208         osb->s_feature_incompat =
2209                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2210
2211         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2212                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2213                      "optional features (%x).\n", i);
2214                 status = -EINVAL;
2215                 goto bail;
2216         }
2217         if (!(osb->sb->s_flags & MS_RDONLY) &&
2218             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2219                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2220                      "unsupported optional features (%x).\n", i);
2221                 status = -EINVAL;
2222                 goto bail;
2223         }
2224
2225         if (ocfs2_clusterinfo_valid(osb)) {
2226                 osb->osb_stackflags =
2227                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2228                 memcpy(osb->osb_cluster_stack,
2229                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2230                        OCFS2_STACK_LABEL_LEN);
2231                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2232                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2233                         mlog(ML_ERROR,
2234                              "couldn't mount because of an invalid "
2235                              "cluster stack label (%s) \n",
2236                              osb->osb_cluster_stack);
2237                         status = -EINVAL;
2238                         goto bail;
2239                 }
2240         } else {
2241                 /* The empty string is identical with classic tools that
2242                  * don't know about s_cluster_info. */
2243                 osb->osb_cluster_stack[0] = '\0';
2244         }
2245
2246         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2247
2248         /* FIXME
2249          * This should be done in ocfs2_journal_init(), but unknown
2250          * ordering issues will cause the filesystem to crash.
2251          * If anyone wants to figure out what part of the code
2252          * refers to osb->journal before ocfs2_journal_init() is run,
2253          * be my guest.
2254          */
2255         /* initialize our journal structure */
2256
2257         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2258         if (!journal) {
2259                 mlog(ML_ERROR, "unable to alloc journal\n");
2260                 status = -ENOMEM;
2261                 goto bail;
2262         }
2263         osb->journal = journal;
2264         journal->j_osb = osb;
2265
2266         atomic_set(&journal->j_num_trans, 0);
2267         init_rwsem(&journal->j_trans_barrier);
2268         init_waitqueue_head(&journal->j_checkpointed);
2269         spin_lock_init(&journal->j_lock);
2270         journal->j_trans_id = (unsigned long) 1;
2271         INIT_LIST_HEAD(&journal->j_la_cleanups);
2272         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2273         journal->j_state = OCFS2_JOURNAL_FREE;
2274
2275         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2276         osb->dentry_lock_list = NULL;
2277
2278         /* get some pseudo constants for clustersize bits */
2279         osb->s_clustersize_bits =
2280                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2281         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2282         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2283
2284         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2285             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2286                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2287                      osb->s_clustersize);
2288                 status = -EINVAL;
2289                 goto bail;
2290         }
2291
2292         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2293                                                 le32_to_cpu(di->i_clusters));
2294
2295         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2296                                            total_blocks);
2297         if (status) {
2298                 mlog(ML_ERROR, "Volume too large "
2299                      "to mount safely on this system");
2300                 status = -EFBIG;
2301                 goto bail;
2302         }
2303
2304         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2305                                  sizeof(di->id2.i_super.s_uuid))) {
2306                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2307                 status = -ENOMEM;
2308                 goto bail;
2309         }
2310
2311         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2312
2313         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2314         osb->vol_label[63] = '\0';
2315         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2316         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2317         osb->first_cluster_group_blkno =
2318                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2319         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2320         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2321         mlog(0, "vol_label: %s\n", osb->vol_label);
2322         mlog(0, "uuid: %s\n", osb->uuid_str);
2323         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2324              (unsigned long long)osb->root_blkno,
2325              (unsigned long long)osb->system_dir_blkno);
2326
2327         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2328         if (!osb->osb_dlm_debug) {
2329                 status = -ENOMEM;
2330                 mlog_errno(status);
2331                 goto bail;
2332         }
2333
2334         atomic_set(&osb->vol_state, VOLUME_INIT);
2335
2336         /* load root, system_dir, and all global system inodes */
2337         status = ocfs2_init_global_system_inodes(osb);
2338         if (status < 0) {
2339                 mlog_errno(status);
2340                 goto bail;
2341         }
2342
2343         /*
2344          * global bitmap
2345          */
2346         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2347                                             OCFS2_INVALID_SLOT);
2348         if (!inode) {
2349                 status = -EINVAL;
2350                 mlog_errno(status);
2351                 goto bail;
2352         }
2353
2354         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2355         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2356         iput(inode);
2357
2358         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2359                                  osb->s_feature_incompat) * 8;
2360
2361         status = ocfs2_init_slot_info(osb);
2362         if (status < 0) {
2363                 mlog_errno(status);
2364                 goto bail;
2365         }
2366
2367 bail:
2368         return status;
2369 }
2370
2371 /*
2372  * will return: -EAGAIN if it is ok to keep searching for superblocks
2373  *              -EINVAL if there is a bad superblock
2374  *              0 on success
2375  */
2376 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2377                                struct buffer_head *bh,
2378                                u32 blksz,
2379                                struct ocfs2_blockcheck_stats *stats)
2380 {
2381         int status = -EAGAIN;
2382
2383         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2384                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2385                 /* We have to do a raw check of the feature here */
2386                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2387                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2388                         status = ocfs2_block_check_validate(bh->b_data,
2389                                                             bh->b_size,
2390                                                             &di->i_check,
2391                                                             stats);
2392                         if (status)
2393                                 goto out;
2394                 }
2395                 status = -EINVAL;
2396                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2397                         mlog(ML_ERROR, "found superblock with incorrect block "
2398                              "size: found %u, should be %u\n",
2399                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2400                                blksz);
2401                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2402                            OCFS2_MAJOR_REV_LEVEL ||
2403                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2404                            OCFS2_MINOR_REV_LEVEL) {
2405                         mlog(ML_ERROR, "found superblock with bad version: "
2406                              "found %u.%u, should be %u.%u\n",
2407                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2408                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2409                              OCFS2_MAJOR_REV_LEVEL,
2410                              OCFS2_MINOR_REV_LEVEL);
2411                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2412                         mlog(ML_ERROR, "bad block number on superblock: "
2413                              "found %llu, should be %llu\n",
2414                              (unsigned long long)le64_to_cpu(di->i_blkno),
2415                              (unsigned long long)bh->b_blocknr);
2416                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2417                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2418                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2419                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2420                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2421                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2422                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2423                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2424                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2425                         mlog(ML_ERROR,
2426                              "Superblock slots found greater than file system "
2427                              "maximum: found %u, max %u\n",
2428                              le16_to_cpu(di->id2.i_super.s_max_slots),
2429                              OCFS2_MAX_SLOTS);
2430                 } else {
2431                         /* found it! */
2432                         status = 0;
2433                 }
2434         }
2435
2436 out:
2437         if (status && status != -EAGAIN)
2438                 mlog_errno(status);
2439         return status;
2440 }
2441
2442 static int ocfs2_check_volume(struct ocfs2_super *osb)
2443 {
2444         int status;
2445         int dirty;
2446         int local;
2447         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2448                                                   * recover
2449                                                   * ourselves. */
2450
2451         /* Init our journal object. */
2452         status = ocfs2_journal_init(osb->journal, &dirty);
2453         if (status < 0) {
2454                 mlog(ML_ERROR, "Could not initialize journal!\n");
2455                 goto finally;
2456         }
2457
2458         /* Now that journal has been initialized, check to make sure
2459            entire volume is addressable. */
2460         status = ocfs2_journal_addressable(osb);
2461         if (status)
2462                 goto finally;
2463
2464         /* If the journal was unmounted cleanly then we don't want to
2465          * recover anything. Otherwise, journal_load will do that
2466          * dirty work for us :) */
2467         if (!dirty) {
2468                 status = ocfs2_journal_wipe(osb->journal, 0);
2469                 if (status < 0) {
2470                         mlog_errno(status);
2471                         goto finally;
2472                 }
2473         } else {
2474                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2475                      "recovering volume.\n");
2476         }
2477
2478         local = ocfs2_mount_local(osb);
2479
2480         /* will play back anything left in the journal. */
2481         status = ocfs2_journal_load(osb->journal, local, dirty);
2482         if (status < 0) {
2483                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2484                 goto finally;
2485         }
2486
2487         if (dirty) {
2488                 /* recover my local alloc if we didn't unmount cleanly. */
2489                 status = ocfs2_begin_local_alloc_recovery(osb,
2490                                                           osb->slot_num,
2491                                                           &local_alloc);
2492                 if (status < 0) {
2493                         mlog_errno(status);
2494                         goto finally;
2495                 }
2496                 /* we complete the recovery process after we've marked
2497                  * ourselves as mounted. */
2498         }
2499
2500         mlog(0, "Journal loaded.\n");
2501
2502         status = ocfs2_load_local_alloc(osb);
2503         if (status < 0) {
2504                 mlog_errno(status);
2505                 goto finally;
2506         }
2507
2508         if (dirty) {
2509                 /* Recovery will be completed after we've mounted the
2510                  * rest of the volume. */
2511                 osb->dirty = 1;
2512                 osb->local_alloc_copy = local_alloc;
2513                 local_alloc = NULL;
2514         }
2515
2516         /* go through each journal, trylock it and if you get the
2517          * lock, and it's marked as dirty, set the bit in the recover
2518          * map and launch a recovery thread for it. */
2519         status = ocfs2_mark_dead_nodes(osb);
2520         if (status < 0) {
2521                 mlog_errno(status);
2522                 goto finally;
2523         }
2524
2525         status = ocfs2_compute_replay_slots(osb);
2526         if (status < 0)
2527                 mlog_errno(status);
2528
2529 finally:
2530         if (local_alloc)
2531                 kfree(local_alloc);
2532
2533         if (status)
2534                 mlog_errno(status);
2535         return status;
2536 }
2537
2538 /*
2539  * The routine gets called from dismount or close whenever a dismount on
2540  * volume is requested and the osb open count becomes 1.
2541  * It will remove the osb from the global list and also free up all the
2542  * initialized resources and fileobject.
2543  */
2544 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2545 {
2546         /* This function assumes that the caller has the main osb resource */
2547
2548         ocfs2_free_slot_info(osb);
2549
2550         kfree(osb->osb_orphan_wipes);
2551         kfree(osb->slot_recovery_generations);
2552         /* FIXME
2553          * This belongs in journal shutdown, but because we have to
2554          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2555          * we free it here.
2556          */
2557         kfree(osb->journal);
2558         if (osb->local_alloc_copy)
2559                 kfree(osb->local_alloc_copy);
2560         kfree(osb->uuid_str);
2561         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2562         memset(osb, 0, sizeof(struct ocfs2_super));
2563 }
2564
2565 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2566  * panic(). We do not support continue-on-error operation. */
2567 static void ocfs2_handle_error(struct super_block *sb)
2568 {
2569         struct ocfs2_super *osb = OCFS2_SB(sb);
2570
2571         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2572                 panic("OCFS2: (device %s): panic forced after error\n",
2573                       sb->s_id);
2574
2575         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2576
2577         if (sb->s_flags & MS_RDONLY &&
2578             (ocfs2_is_soft_readonly(osb) ||
2579              ocfs2_is_hard_readonly(osb)))
2580                 return;
2581
2582         printk(KERN_CRIT "File system is now read-only due to the potential "
2583                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2584                "system is unmounted.\n");
2585         sb->s_flags |= MS_RDONLY;
2586         ocfs2_set_ro_flag(osb, 0);
2587 }
2588
2589 static char error_buf[1024];
2590
2591 void __ocfs2_error(struct super_block *sb,
2592                    const char *function,
2593                    const char *fmt, ...)
2594 {
2595         va_list args;
2596
2597         va_start(args, fmt);
2598         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2599         va_end(args);
2600
2601         /* Not using mlog here because we want to show the actual
2602          * function the error came from. */
2603         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2604                sb->s_id, function, error_buf);
2605
2606         ocfs2_handle_error(sb);
2607 }
2608
2609 /* Handle critical errors. This is intentionally more drastic than
2610  * ocfs2_handle_error, so we only use for things like journal errors,
2611  * etc. */
2612 void __ocfs2_abort(struct super_block* sb,
2613                    const char *function,
2614                    const char *fmt, ...)
2615 {
2616         va_list args;
2617
2618         va_start(args, fmt);
2619         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2620         va_end(args);
2621
2622         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2623                sb->s_id, function, error_buf);
2624
2625         /* We don't have the cluster support yet to go straight to
2626          * hard readonly in here. Until then, we want to keep
2627          * ocfs2_abort() so that we can at least mark critical
2628          * errors.
2629          *
2630          * TODO: This should abort the journal and alert other nodes
2631          * that our slot needs recovery. */
2632
2633         /* Force a panic(). This stinks, but it's better than letting
2634          * things continue without having a proper hard readonly
2635          * here. */
2636         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2637                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2638         ocfs2_handle_error(sb);
2639 }
2640
2641 /*
2642  * Void signal blockers, because in-kernel sigprocmask() only fails
2643  * when SIG_* is wrong.
2644  */
2645 void ocfs2_block_signals(sigset_t *oldset)
2646 {
2647         int rc;
2648         sigset_t blocked;
2649
2650         sigfillset(&blocked);
2651         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2652         BUG_ON(rc);
2653 }
2654
2655 void ocfs2_unblock_signals(sigset_t *oldset)
2656 {
2657         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2658         BUG_ON(rc);
2659 }
2660
2661 module_init(ocfs2_init);
2662 module_exit(ocfs2_exit);