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         mlog_entry_void();
 445
 446         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 447         if (IS_ERR(new)) {
 448                 status = PTR_ERR(new);
 449                 mlog_errno(status);
 450                 goto bail;
 451         }
 452         osb->root_inode = new;
 453
 454         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 455         if (IS_ERR(new)) {
 456                 status = PTR_ERR(new);
 457                 mlog_errno(status);
 458                 goto bail;
 459         }
 460         osb->sys_root_inode = new;
 461
 462         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 463              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 464                 if (!ocfs2_need_system_inode(osb, i))
 465                         continue;
 466                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 467                 if (!new) {
 468                         ocfs2_release_system_inodes(osb);
 469                         status = -EINVAL;
 470                         mlog_errno(status);
 471                         /* FIXME: Should ERROR_RO_FS */
 472                         mlog(ML_ERROR, "Unable to load system inode %d, "
 473                              "possibly corrupt fs?", i);
 474                         goto bail;
 475                 }
 476                 // the array now has one ref, so drop this one
 477                 iput(new);
 478         }
 479
 480 bail:
 481         mlog_exit(status);
 482         return status;
 483 }
 484
 485 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 486 {
 487         struct inode *new = NULL;
 488         int status = 0;
 489         int i;
 490
 491         mlog_entry_void();
 492
 493         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 494              i < NUM_SYSTEM_INODES;
 495              i++) {
 496                 if (!ocfs2_need_system_inode(osb, i))
 497                         continue;
 498                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 499                 if (!new) {
 500                         ocfs2_release_system_inodes(osb);
 501                         status = -EINVAL;
 502                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 503                              status, i, osb->slot_num);
 504                         goto bail;
 505                 }
 506                 /* the array now has one ref, so drop this one */
 507                 iput(new);
 508         }
 509
 510 bail:
 511         mlog_exit(status);
 512         return status;
 513 }
 514
 515 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 516 {
 517         int i;
 518         struct inode *inode;
 519
 520         mlog_entry_void();
 521
 522         for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
 523                 inode = osb->global_system_inodes[i];
 524                 if (inode) {
 525                         iput(inode);
 526                         osb->global_system_inodes[i] = NULL;
 527                 }
 528         }
 529
 530         inode = osb->sys_root_inode;
 531         if (inode) {
 532                 iput(inode);
 533                 osb->sys_root_inode = NULL;
 534         }
 535
 536         inode = osb->root_inode;
 537         if (inode) {
 538                 iput(inode);
 539                 osb->root_inode = NULL;
 540         }
 541
 542         if (!osb->local_system_inodes)
 543                 goto out;
 544
 545         for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
 546                 if (osb->local_system_inodes[i]) {
 547                         iput(osb->local_system_inodes[i]);
 548                         osb->local_system_inodes[i] = NULL;
 549                 }
 550         }
 551
 552         kfree(osb->local_system_inodes);
 553         osb->local_system_inodes = NULL;
 554
 555 out:
 556         mlog_exit(0);
 557 }
 558
 559 /* We're allocating fs objects, use GFP_NOFS */
 560 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 561 {
 562         struct ocfs2_inode_info *oi;
 563
 564         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 565         if (!oi)
 566                 return NULL;
 567
 568         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 569         return &oi->vfs_inode;
 570 }
 571
 572 static void ocfs2_destroy_inode(struct inode *inode)
 573 {
 574         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 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                           char *path)
 991 {
 992         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 993                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 994
 995         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 996                 return -EINVAL;
 997
 998         return dquot_enable(sb_dqopt(sb)->files[type], type,
 999                             format_id, DQUOT_LIMITS_ENABLED);
1000 }
1001
1002 /* Handle quota off quotactl */
1003 static int ocfs2_quota_off(struct super_block *sb, int type)
1004 {
1005         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1006 }
1007
1008 static const struct quotactl_ops ocfs2_quotactl_ops = {
1009         .quota_on       = ocfs2_quota_on,
1010         .quota_off      = ocfs2_quota_off,
1011         .quota_sync     = dquot_quota_sync,
1012         .get_info       = dquot_get_dqinfo,
1013         .set_info       = dquot_set_dqinfo,
1014         .get_dqblk      = dquot_get_dqblk,
1015         .set_dqblk      = dquot_set_dqblk,
1016 };
1017
1018 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1019 {
1020         struct dentry *root;
1021         int status, sector_size;
1022         struct mount_options parsed_options;
1023         struct inode *inode = NULL;
1024         struct ocfs2_super *osb = NULL;
1025         struct buffer_head *bh = NULL;
1026         char nodestr[8];
1027         struct ocfs2_blockcheck_stats stats;
1028
1029         mlog_entry("%p, %p, %i", sb, data, silent);
1030
1031         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1032                 status = -EINVAL;
1033                 goto read_super_error;
1034         }
1035
1036         /* probe for superblock */
1037         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1038         if (status < 0) {
1039                 mlog(ML_ERROR, "superblock probe failed!\n");
1040                 goto read_super_error;
1041         }
1042
1043         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1044         osb = OCFS2_SB(sb);
1045         if (status < 0) {
1046                 mlog_errno(status);
1047                 goto read_super_error;
1048         }
1049         brelse(bh);
1050         bh = NULL;
1051
1052         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1053                 status = -EINVAL;
1054                 goto read_super_error;
1055         }
1056         osb->s_mount_opt = parsed_options.mount_opt;
1057         osb->s_atime_quantum = parsed_options.atime_quantum;
1058         osb->preferred_slot = parsed_options.slot;
1059         osb->osb_commit_interval = parsed_options.commit_interval;
1060
1061         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1062         osb->osb_resv_level = parsed_options.resv_level;
1063         osb->osb_dir_resv_level = parsed_options.resv_level;
1064         if (parsed_options.dir_resv_level == -1)
1065                 osb->osb_dir_resv_level = parsed_options.resv_level;
1066         else
1067                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1068
1069         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1070         if (status)
1071                 goto read_super_error;
1072
1073         sb->s_magic = OCFS2_SUPER_MAGIC;
1074
1075         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1076                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1077
1078         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1079          * heartbeat=none */
1080         if (bdev_read_only(sb->s_bdev)) {
1081                 if (!(sb->s_flags & MS_RDONLY)) {
1082                         status = -EACCES;
1083                         mlog(ML_ERROR, "Readonly device detected but readonly "
1084                              "mount was not specified.\n");
1085                         goto read_super_error;
1086                 }
1087
1088                 /* You should not be able to start a local heartbeat
1089                  * on a readonly device. */
1090                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1091                         status = -EROFS;
1092                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1093                              "device.\n");
1094                         goto read_super_error;
1095                 }
1096
1097                 status = ocfs2_check_journals_nolocks(osb);
1098                 if (status < 0) {
1099                         if (status == -EROFS)
1100                                 mlog(ML_ERROR, "Recovery required on readonly "
1101                                      "file system, but write access is "
1102                                      "unavailable.\n");
1103                         else
1104                                 mlog_errno(status);
1105                         goto read_super_error;
1106                 }
1107
1108                 ocfs2_set_ro_flag(osb, 1);
1109
1110                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1111                        "services will be utilized for this mount. Recovery "
1112                        "will be skipped.\n");
1113         }
1114
1115         if (!ocfs2_is_hard_readonly(osb)) {
1116                 if (sb->s_flags & MS_RDONLY)
1117                         ocfs2_set_ro_flag(osb, 0);
1118         }
1119
1120         status = ocfs2_verify_heartbeat(osb);
1121         if (status < 0) {
1122                 mlog_errno(status);
1123                 goto read_super_error;
1124         }
1125
1126         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1127                                                  ocfs2_debugfs_root);
1128         if (!osb->osb_debug_root) {
1129                 status = -EINVAL;
1130                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1131                 goto read_super_error;
1132         }
1133
1134         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1135                                             osb->osb_debug_root,
1136                                             osb,
1137                                             &ocfs2_osb_debug_fops);
1138         if (!osb->osb_ctxt) {
1139                 status = -EINVAL;
1140                 mlog_errno(status);
1141                 goto read_super_error;
1142         }
1143
1144         if (ocfs2_meta_ecc(osb)) {
1145                 status = ocfs2_blockcheck_stats_debugfs_install(
1146                                                 &osb->osb_ecc_stats,
1147                                                 osb->osb_debug_root);
1148                 if (status) {
1149                         mlog(ML_ERROR,
1150                              "Unable to create blockcheck statistics "
1151                              "files\n");
1152                         goto read_super_error;
1153                 }
1154         }
1155
1156         status = ocfs2_mount_volume(sb);
1157         if (osb->root_inode)
1158                 inode = igrab(osb->root_inode);
1159
1160         if (status < 0)
1161                 goto read_super_error;
1162
1163         if (!inode) {
1164                 status = -EIO;
1165                 mlog_errno(status);
1166                 goto read_super_error;
1167         }
1168
1169         root = d_alloc_root(inode);
1170         if (!root) {
1171                 status = -ENOMEM;
1172                 mlog_errno(status);
1173                 goto read_super_error;
1174         }
1175
1176         sb->s_root = root;
1177
1178         ocfs2_complete_mount_recovery(osb);
1179
1180         if (ocfs2_mount_local(osb))
1181                 snprintf(nodestr, sizeof(nodestr), "local");
1182         else
1183                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1184
1185         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1186                "with %s data mode.\n",
1187                osb->dev_str, nodestr, osb->slot_num,
1188                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1189                "ordered");
1190
1191         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1192         wake_up(&osb->osb_mount_event);
1193
1194         /* Now we can initialize quotas because we can afford to wait
1195          * for cluster locks recovery now. That also means that truncation
1196          * log recovery can happen but that waits for proper quota setup */
1197         if (!(sb->s_flags & MS_RDONLY)) {
1198                 status = ocfs2_enable_quotas(osb);
1199                 if (status < 0) {
1200                         /* We have to err-out specially here because
1201                          * s_root is already set */
1202                         mlog_errno(status);
1203                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1204                         wake_up(&osb->osb_mount_event);
1205                         mlog_exit(status);
1206                         return status;
1207                 }
1208         }
1209
1210         ocfs2_complete_quota_recovery(osb);
1211
1212         /* Now we wake up again for processes waiting for quotas */
1213         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1214         wake_up(&osb->osb_mount_event);
1215
1216         /* Start this when the mount is almost sure of being successful */
1217         ocfs2_orphan_scan_start(osb);
1218
1219         mlog_exit(status);
1220         return status;
1221
1222 read_super_error:
1223         brelse(bh);
1224
1225         if (inode)
1226                 iput(inode);
1227
1228         if (osb) {
1229                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1230                 wake_up(&osb->osb_mount_event);
1231                 ocfs2_dismount_volume(sb, 1);
1232         }
1233
1234         mlog_exit(status);
1235         return status;
1236 }
1237
1238 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1239                         int flags,
1240                         const char *dev_name,
1241                         void *data)
1242 {
1243         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1244 }
1245
1246 static void ocfs2_kill_sb(struct super_block *sb)
1247 {
1248         struct ocfs2_super *osb = OCFS2_SB(sb);
1249
1250         /* Failed mount? */
1251         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1252                 goto out;
1253
1254         /* Prevent further queueing of inode drop events */
1255         spin_lock(&dentry_list_lock);
1256         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1257         spin_unlock(&dentry_list_lock);
1258         /* Wait for work to finish and/or remove it */
1259         cancel_work_sync(&osb->dentry_lock_work);
1260 out:
1261         kill_block_super(sb);
1262 }
1263
1264 static struct file_system_type ocfs2_fs_type = {
1265         .owner          = THIS_MODULE,
1266         .name           = "ocfs2",
1267         .mount          = ocfs2_mount,
1268         .kill_sb        = ocfs2_kill_sb,
1269
1270         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1271         .next           = NULL
1272 };
1273
1274 static int ocfs2_check_set_options(struct super_block *sb,
1275                                    struct mount_options *options)
1276 {
1277         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1278             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1279                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1280                 mlog(ML_ERROR, "User quotas were requested, but this "
1281                      "filesystem does not have the feature enabled.\n");
1282                 return 0;
1283         }
1284         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1285             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1286                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1287                 mlog(ML_ERROR, "Group quotas were requested, but this "
1288                      "filesystem does not have the feature enabled.\n");
1289                 return 0;
1290         }
1291         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1292             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1293                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1294                      "feature is not enabled\n");
1295                 return 0;
1296         }
1297         /* No ACL setting specified? Use XATTR feature... */
1298         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1299                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1300                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1301                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1302                 else
1303                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1304         }
1305         return 1;
1306 }
1307
1308 static int ocfs2_parse_options(struct super_block *sb,
1309                                char *options,
1310                                struct mount_options *mopt,
1311                                int is_remount)
1312 {
1313         int status;
1314         char *p;
1315         u32 tmp;
1316
1317         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1318                    options ? options : "(none)");
1319
1320         mopt->commit_interval = 0;
1321         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1322         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1323         mopt->slot = OCFS2_INVALID_SLOT;
1324         mopt->localalloc_opt = -1;
1325         mopt->cluster_stack[0] = '\0';
1326         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1327         mopt->dir_resv_level = -1;
1328
1329         if (!options) {
1330                 status = 1;
1331                 goto bail;
1332         }
1333
1334         while ((p = strsep(&options, ",")) != NULL) {
1335                 int token, option;
1336                 substring_t args[MAX_OPT_ARGS];
1337
1338                 if (!*p)
1339                         continue;
1340
1341                 token = match_token(p, tokens, args);
1342                 switch (token) {
1343                 case Opt_hb_local:
1344                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1345                         break;
1346                 case Opt_hb_none:
1347                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1348                         break;
1349                 case Opt_hb_global:
1350                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1351                         break;
1352                 case Opt_barrier:
1353                         if (match_int(&args[0], &option)) {
1354                                 status = 0;
1355                                 goto bail;
1356                         }
1357                         if (option)
1358                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1359                         else
1360                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1361                         break;
1362                 case Opt_intr:
1363                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1364                         break;
1365                 case Opt_nointr:
1366                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1367                         break;
1368                 case Opt_err_panic:
1369                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1370                         break;
1371                 case Opt_err_ro:
1372                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1373                         break;
1374                 case Opt_data_ordered:
1375                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1376                         break;
1377                 case Opt_data_writeback:
1378                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1379                         break;
1380                 case Opt_user_xattr:
1381                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1382                         break;
1383                 case Opt_nouser_xattr:
1384                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1385                         break;
1386                 case Opt_atime_quantum:
1387                         if (match_int(&args[0], &option)) {
1388                                 status = 0;
1389                                 goto bail;
1390                         }
1391                         if (option >= 0)
1392                                 mopt->atime_quantum = option;
1393                         break;
1394                 case Opt_slot:
1395                         option = 0;
1396                         if (match_int(&args[0], &option)) {
1397                                 status = 0;
1398                                 goto bail;
1399                         }
1400                         if (option)
1401                                 mopt->slot = (s16)option;
1402                         break;
1403                 case Opt_commit:
1404                         option = 0;
1405                         if (match_int(&args[0], &option)) {
1406                                 status = 0;
1407                                 goto bail;
1408                         }
1409                         if (option < 0)
1410                                 return 0;
1411                         if (option == 0)
1412                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1413                         mopt->commit_interval = HZ * option;
1414                         break;
1415                 case Opt_localalloc:
1416                         option = 0;
1417                         if (match_int(&args[0], &option)) {
1418                                 status = 0;
1419                                 goto bail;
1420                         }
1421                         if (option >= 0)
1422                                 mopt->localalloc_opt = option;
1423                         break;
1424                 case Opt_localflocks:
1425                         /*
1426                          * Changing this during remount could race
1427                          * flock() requests, or "unbalance" existing
1428                          * ones (e.g., a lock is taken in one mode but
1429                          * dropped in the other). If users care enough
1430                          * to flip locking modes during remount, we
1431                          * could add a "local" flag to individual
1432                          * flock structures for proper tracking of
1433                          * state.
1434                          */
1435                         if (!is_remount)
1436                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1437                         break;
1438                 case Opt_stack:
1439                         /* Check both that the option we were passed
1440                          * is of the right length and that it is a proper
1441                          * string of the right length.
1442                          */
1443                         if (((args[0].to - args[0].from) !=
1444                              OCFS2_STACK_LABEL_LEN) ||
1445                             (strnlen(args[0].from,
1446                                      OCFS2_STACK_LABEL_LEN) !=
1447                              OCFS2_STACK_LABEL_LEN)) {
1448                                 mlog(ML_ERROR,
1449                                      "Invalid cluster_stack option\n");
1450                                 status = 0;
1451                                 goto bail;
1452                         }
1453                         memcpy(mopt->cluster_stack, args[0].from,
1454                                OCFS2_STACK_LABEL_LEN);
1455                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1456                         break;
1457                 case Opt_inode64:
1458                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1459                         break;
1460                 case Opt_usrquota:
1461                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1462                         break;
1463                 case Opt_grpquota:
1464                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1465                         break;
1466                 case Opt_coherency_buffered:
1467                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1468                         break;
1469                 case Opt_coherency_full:
1470                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1471                         break;
1472                 case Opt_acl:
1473                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1474                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1475                         break;
1476                 case Opt_noacl:
1477                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1478                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1479                         break;
1480                 case Opt_resv_level:
1481                         if (is_remount)
1482                                 break;
1483                         if (match_int(&args[0], &option)) {
1484                                 status = 0;
1485                                 goto bail;
1486                         }
1487                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1488                             option < OCFS2_MAX_RESV_LEVEL)
1489                                 mopt->resv_level = option;
1490                         break;
1491                 case Opt_dir_resv_level:
1492                         if (is_remount)
1493                                 break;
1494                         if (match_int(&args[0], &option)) {
1495                                 status = 0;
1496                                 goto bail;
1497                         }
1498                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1499                             option < OCFS2_MAX_RESV_LEVEL)
1500                                 mopt->dir_resv_level = option;
1501                         break;
1502                 default:
1503                         mlog(ML_ERROR,
1504                              "Unrecognized mount option \"%s\" "
1505                              "or missing value\n", p);
1506                         status = 0;
1507                         goto bail;
1508                 }
1509         }
1510
1511         /* Ensure only one heartbeat mode */
1512         tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1513                                  OCFS2_MOUNT_HB_NONE);
1514         if (hweight32(tmp) != 1) {
1515                 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1516                 status = 0;
1517                 goto bail;
1518         }
1519
1520         status = 1;
1521
1522 bail:
1523         mlog_exit(status);
1524         return status;
1525 }
1526
1527 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1528 {
1529         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1530         unsigned long opts = osb->s_mount_opt;
1531         unsigned int local_alloc_megs;
1532
1533         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1534                 seq_printf(s, ",_netdev");
1535                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1536                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1537                 else
1538                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1539         } else
1540                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1541
1542         if (opts & OCFS2_MOUNT_NOINTR)
1543                 seq_printf(s, ",nointr");
1544
1545         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1546                 seq_printf(s, ",data=writeback");
1547         else
1548                 seq_printf(s, ",data=ordered");
1549
1550         if (opts & OCFS2_MOUNT_BARRIER)
1551                 seq_printf(s, ",barrier=1");
1552
1553         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1554                 seq_printf(s, ",errors=panic");
1555         else
1556                 seq_printf(s, ",errors=remount-ro");
1557
1558         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1559                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1560
1561         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1562                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1563
1564         if (osb->osb_commit_interval)
1565                 seq_printf(s, ",commit=%u",
1566                            (unsigned) (osb->osb_commit_interval / HZ));
1567
1568         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1569         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1570                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1571
1572         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1573                 seq_printf(s, ",localflocks,");
1574
1575         if (osb->osb_cluster_stack[0])
1576                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1577                            osb->osb_cluster_stack);
1578         if (opts & OCFS2_MOUNT_USRQUOTA)
1579                 seq_printf(s, ",usrquota");
1580         if (opts & OCFS2_MOUNT_GRPQUOTA)
1581                 seq_printf(s, ",grpquota");
1582
1583         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1584                 seq_printf(s, ",coherency=buffered");
1585         else
1586                 seq_printf(s, ",coherency=full");
1587
1588         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1589                 seq_printf(s, ",nouser_xattr");
1590         else
1591                 seq_printf(s, ",user_xattr");
1592
1593         if (opts & OCFS2_MOUNT_INODE64)
1594                 seq_printf(s, ",inode64");
1595
1596         if (opts & OCFS2_MOUNT_POSIX_ACL)
1597                 seq_printf(s, ",acl");
1598         else
1599                 seq_printf(s, ",noacl");
1600
1601         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1602                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1603
1604         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1605                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1606
1607         return 0;
1608 }
1609
1610 static int __init ocfs2_init(void)
1611 {
1612         int status;
1613
1614         mlog_entry_void();
1615
1616         ocfs2_print_version();
1617
1618         status = init_ocfs2_uptodate_cache();
1619         if (status < 0) {
1620                 mlog_errno(status);
1621                 goto leave;
1622         }
1623
1624         status = ocfs2_initialize_mem_caches();
1625         if (status < 0) {
1626                 mlog_errno(status);
1627                 goto leave;
1628         }
1629
1630         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1631         if (!ocfs2_wq) {
1632                 status = -ENOMEM;
1633                 goto leave;
1634         }
1635
1636         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1637         if (!ocfs2_debugfs_root) {
1638                 status = -EFAULT;
1639                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1640         }
1641
1642         status = ocfs2_quota_setup();
1643         if (status)
1644                 goto leave;
1645
1646         ocfs2_set_locking_protocol();
1647
1648         status = register_quota_format(&ocfs2_quota_format);
1649 leave:
1650         if (status < 0) {
1651                 ocfs2_quota_shutdown();
1652                 ocfs2_free_mem_caches();
1653                 exit_ocfs2_uptodate_cache();
1654         }
1655
1656         mlog_exit(status);
1657
1658         if (status >= 0) {
1659                 return register_filesystem(&ocfs2_fs_type);
1660         } else
1661                 return -1;
1662 }
1663
1664 static void __exit ocfs2_exit(void)
1665 {
1666         mlog_entry_void();
1667
1668         ocfs2_quota_shutdown();
1669
1670         if (ocfs2_wq) {
1671                 flush_workqueue(ocfs2_wq);
1672                 destroy_workqueue(ocfs2_wq);
1673         }
1674
1675         unregister_quota_format(&ocfs2_quota_format);
1676
1677         debugfs_remove(ocfs2_debugfs_root);
1678
1679         ocfs2_free_mem_caches();
1680
1681         unregister_filesystem(&ocfs2_fs_type);
1682
1683         exit_ocfs2_uptodate_cache();
1684
1685         mlog_exit_void();
1686 }
1687
1688 static void ocfs2_put_super(struct super_block *sb)
1689 {
1690         mlog_entry("(0x%p)\n", sb);
1691
1692         ocfs2_sync_blockdev(sb);
1693         ocfs2_dismount_volume(sb, 0);
1694
1695         mlog_exit_void();
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_entry("(%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         mlog_exit(status);
1755
1756         return status;
1757 }
1758
1759 static void ocfs2_inode_init_once(void *data)
1760 {
1761         struct ocfs2_inode_info *oi = data;
1762
1763         oi->ip_flags = 0;
1764         oi->ip_open_count = 0;
1765         spin_lock_init(&oi->ip_lock);
1766         ocfs2_extent_map_init(&oi->vfs_inode);
1767         INIT_LIST_HEAD(&oi->ip_io_markers);
1768         oi->ip_dir_start_lookup = 0;
1769
1770         init_rwsem(&oi->ip_alloc_sem);
1771         init_rwsem(&oi->ip_xattr_sem);
1772         mutex_init(&oi->ip_io_mutex);
1773
1774         oi->ip_blkno = 0ULL;
1775         oi->ip_clusters = 0;
1776
1777         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1778
1779         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1780         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1781         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1782
1783         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1784                                   &ocfs2_inode_caching_ops);
1785
1786         inode_init_once(&oi->vfs_inode);
1787 }
1788
1789 static int ocfs2_initialize_mem_caches(void)
1790 {
1791         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1792                                        sizeof(struct ocfs2_inode_info),
1793                                        0,
1794                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1795                                                 SLAB_MEM_SPREAD),
1796                                        ocfs2_inode_init_once);
1797         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1798                                         sizeof(struct ocfs2_dquot),
1799                                         0,
1800                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1801                                                 SLAB_MEM_SPREAD),
1802                                         NULL);
1803         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1804                                         sizeof(struct ocfs2_quota_chunk),
1805                                         0,
1806                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1807                                         NULL);
1808         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1809             !ocfs2_qf_chunk_cachep) {
1810                 if (ocfs2_inode_cachep)
1811                         kmem_cache_destroy(ocfs2_inode_cachep);
1812                 if (ocfs2_dquot_cachep)
1813                         kmem_cache_destroy(ocfs2_dquot_cachep);
1814                 if (ocfs2_qf_chunk_cachep)
1815                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1816                 return -ENOMEM;
1817         }
1818
1819         return 0;
1820 }
1821
1822 static void ocfs2_free_mem_caches(void)
1823 {
1824         if (ocfs2_inode_cachep)
1825                 kmem_cache_destroy(ocfs2_inode_cachep);
1826         ocfs2_inode_cachep = NULL;
1827
1828         if (ocfs2_dquot_cachep)
1829                 kmem_cache_destroy(ocfs2_dquot_cachep);
1830         ocfs2_dquot_cachep = NULL;
1831
1832         if (ocfs2_qf_chunk_cachep)
1833                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1834         ocfs2_qf_chunk_cachep = NULL;
1835 }
1836
1837 static int ocfs2_get_sector(struct super_block *sb,
1838                             struct buffer_head **bh,
1839                             int block,
1840                             int sect_size)
1841 {
1842         if (!sb_set_blocksize(sb, sect_size)) {
1843                 mlog(ML_ERROR, "unable to set blocksize\n");
1844                 return -EIO;
1845         }
1846
1847         *bh = sb_getblk(sb, block);
1848         if (!*bh) {
1849                 mlog_errno(-EIO);
1850                 return -EIO;
1851         }
1852         lock_buffer(*bh);
1853         if (!buffer_dirty(*bh))
1854                 clear_buffer_uptodate(*bh);
1855         unlock_buffer(*bh);
1856         ll_rw_block(READ, 1, bh);
1857         wait_on_buffer(*bh);
1858         if (!buffer_uptodate(*bh)) {
1859                 mlog_errno(-EIO);
1860                 brelse(*bh);
1861                 *bh = NULL;
1862                 return -EIO;
1863         }
1864
1865         return 0;
1866 }
1867
1868 static int ocfs2_mount_volume(struct super_block *sb)
1869 {
1870         int status = 0;
1871         int unlock_super = 0;
1872         struct ocfs2_super *osb = OCFS2_SB(sb);
1873
1874         mlog_entry_void();
1875
1876         if (ocfs2_is_hard_readonly(osb))
1877                 goto leave;
1878
1879         status = ocfs2_dlm_init(osb);
1880         if (status < 0) {
1881                 mlog_errno(status);
1882                 goto leave;
1883         }
1884
1885         status = ocfs2_super_lock(osb, 1);
1886         if (status < 0) {
1887                 mlog_errno(status);
1888                 goto leave;
1889         }
1890         unlock_super = 1;
1891
1892         /* This will load up the node map and add ourselves to it. */
1893         status = ocfs2_find_slot(osb);
1894         if (status < 0) {
1895                 mlog_errno(status);
1896                 goto leave;
1897         }
1898
1899         /* load all node-local system inodes */
1900         status = ocfs2_init_local_system_inodes(osb);
1901         if (status < 0) {
1902                 mlog_errno(status);
1903                 goto leave;
1904         }
1905
1906         status = ocfs2_check_volume(osb);
1907         if (status < 0) {
1908                 mlog_errno(status);
1909                 goto leave;
1910         }
1911
1912         status = ocfs2_truncate_log_init(osb);
1913         if (status < 0)
1914                 mlog_errno(status);
1915
1916 leave:
1917         if (unlock_super)
1918                 ocfs2_super_unlock(osb, 1);
1919
1920         mlog_exit(status);
1921         return status;
1922 }
1923
1924 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1925 {
1926         int tmp, hangup_needed = 0;
1927         struct ocfs2_super *osb = NULL;
1928         char nodestr[8];
1929
1930         mlog_entry("(0x%p)\n", sb);
1931
1932         BUG_ON(!sb);
1933         osb = OCFS2_SB(sb);
1934         BUG_ON(!osb);
1935
1936         debugfs_remove(osb->osb_ctxt);
1937
1938         /*
1939          * Flush inode dropping work queue so that deletes are
1940          * performed while the filesystem is still working
1941          */
1942         ocfs2_drop_all_dl_inodes(osb);
1943
1944         /* Orphan scan should be stopped as early as possible */
1945         ocfs2_orphan_scan_stop(osb);
1946
1947         ocfs2_disable_quotas(osb);
1948
1949         ocfs2_shutdown_local_alloc(osb);
1950
1951         ocfs2_truncate_log_shutdown(osb);
1952
1953         /* This will disable recovery and flush any recovery work. */
1954         ocfs2_recovery_exit(osb);
1955
1956         ocfs2_journal_shutdown(osb);
1957
1958         ocfs2_sync_blockdev(sb);
1959
1960         ocfs2_purge_refcount_trees(osb);
1961
1962         /* No cluster connection means we've failed during mount, so skip
1963          * all the steps which depended on that to complete. */
1964         if (osb->cconn) {
1965                 tmp = ocfs2_super_lock(osb, 1);
1966                 if (tmp < 0) {
1967                         mlog_errno(tmp);
1968                         return;
1969                 }
1970         }
1971
1972         if (osb->slot_num != OCFS2_INVALID_SLOT)
1973                 ocfs2_put_slot(osb);
1974
1975         if (osb->cconn)
1976                 ocfs2_super_unlock(osb, 1);
1977
1978         ocfs2_release_system_inodes(osb);
1979
1980         /*
1981          * If we're dismounting due to mount error, mount.ocfs2 will clean
1982          * up heartbeat.  If we're a local mount, there is no heartbeat.
1983          * If we failed before we got a uuid_str yet, we can't stop
1984          * heartbeat.  Otherwise, do it.
1985          */
1986         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1987                 hangup_needed = 1;
1988
1989         if (osb->cconn)
1990                 ocfs2_dlm_shutdown(osb, hangup_needed);
1991
1992         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1993         debugfs_remove(osb->osb_debug_root);
1994
1995         if (hangup_needed)
1996                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1997
1998         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1999
2000         if (ocfs2_mount_local(osb))
2001                 snprintf(nodestr, sizeof(nodestr), "local");
2002         else
2003                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2004
2005         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2006                osb->dev_str, nodestr);
2007
2008         ocfs2_delete_osb(osb);
2009         kfree(osb);
2010         sb->s_dev = 0;
2011         sb->s_fs_info = NULL;
2012 }
2013
2014 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2015                                 unsigned uuid_bytes)
2016 {
2017         int i, ret;
2018         char *ptr;
2019
2020         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2021
2022         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2023         if (osb->uuid_str == NULL)
2024                 return -ENOMEM;
2025
2026         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2027                 /* print with null */
2028                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2029                 if (ret != 2) /* drop super cleans up */
2030                         return -EINVAL;
2031                 /* then only advance past the last char */
2032                 ptr += 2;
2033         }
2034
2035         return 0;
2036 }
2037
2038 /* Make sure entire volume is addressable by our journal.  Requires
2039    osb_clusters_at_boot to be valid and for the journal to have been
2040    initialized by ocfs2_journal_init(). */
2041 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2042 {
2043         int status = 0;
2044         u64 max_block =
2045                 ocfs2_clusters_to_blocks(osb->sb,
2046                                          osb->osb_clusters_at_boot) - 1;
2047
2048         /* 32-bit block number is always OK. */
2049         if (max_block <= (u32)~0ULL)
2050                 goto out;
2051
2052         /* Volume is "huge", so see if our journal is new enough to
2053            support it. */
2054         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2055                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2056               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2057                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2058                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2059                      "Enable the 'block64' journal option with tunefs.ocfs2");
2060                 status = -EFBIG;
2061                 goto out;
2062         }
2063
2064  out:
2065         return status;
2066 }
2067
2068 static int ocfs2_initialize_super(struct super_block *sb,
2069                                   struct buffer_head *bh,
2070                                   int sector_size,
2071                                   struct ocfs2_blockcheck_stats *stats)
2072 {
2073         int status;
2074         int i, cbits, bbits;
2075         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2076         struct inode *inode = NULL;
2077         struct ocfs2_journal *journal;
2078         __le32 uuid_net_key;
2079         struct ocfs2_super *osb;
2080         u64 total_blocks;
2081
2082         mlog_entry_void();
2083
2084         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2085         if (!osb) {
2086                 status = -ENOMEM;
2087                 mlog_errno(status);
2088                 goto bail;
2089         }
2090
2091         sb->s_fs_info = osb;
2092         sb->s_op = &ocfs2_sops;
2093         sb->s_export_op = &ocfs2_export_ops;
2094         sb->s_qcop = &ocfs2_quotactl_ops;
2095         sb->dq_op = &ocfs2_quota_operations;
2096         sb->s_xattr = ocfs2_xattr_handlers;
2097         sb->s_time_gran = 1;
2098         sb->s_flags |= MS_NOATIME;
2099         /* this is needed to support O_LARGEFILE */
2100         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2101         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2102         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2103
2104         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2105
2106         for (i = 0; i < 3; i++)
2107                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2108         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2109
2110         osb->sb = sb;
2111         /* Save off for ocfs2_rw_direct */
2112         osb->s_sectsize_bits = blksize_bits(sector_size);
2113         BUG_ON(!osb->s_sectsize_bits);
2114
2115         spin_lock_init(&osb->dc_task_lock);
2116         init_waitqueue_head(&osb->dc_event);
2117         osb->dc_work_sequence = 0;
2118         osb->dc_wake_sequence = 0;
2119         INIT_LIST_HEAD(&osb->blocked_lock_list);
2120         osb->blocked_lock_count = 0;
2121         spin_lock_init(&osb->osb_lock);
2122         spin_lock_init(&osb->osb_xattr_lock);
2123         ocfs2_init_steal_slots(osb);
2124
2125         atomic_set(&osb->alloc_stats.moves, 0);
2126         atomic_set(&osb->alloc_stats.local_data, 0);
2127         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2128         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2129         atomic_set(&osb->alloc_stats.bg_extends, 0);
2130
2131         /* Copy the blockcheck stats from the superblock probe */
2132         osb->osb_ecc_stats = *stats;
2133
2134         ocfs2_init_node_maps(osb);
2135
2136         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2137                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2138
2139         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2140         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2141                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2142                      osb->max_slots);
2143                 status = -EINVAL;
2144                 goto bail;
2145         }
2146         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2147
2148         ocfs2_orphan_scan_init(osb);
2149
2150         status = ocfs2_recovery_init(osb);
2151         if (status) {
2152                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2153                 mlog_errno(status);
2154                 goto bail;
2155         }
2156
2157         init_waitqueue_head(&osb->checkpoint_event);
2158         atomic_set(&osb->needs_checkpoint, 0);
2159
2160         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2161
2162         osb->slot_num = OCFS2_INVALID_SLOT;
2163
2164         osb->s_xattr_inline_size = le16_to_cpu(
2165                                         di->id2.i_super.s_xattr_inline_size);
2166
2167         osb->local_alloc_state = OCFS2_LA_UNUSED;
2168         osb->local_alloc_bh = NULL;
2169         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2170
2171         init_waitqueue_head(&osb->osb_mount_event);
2172
2173         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2174         if (status) {
2175                 mlog_errno(status);
2176                 goto bail;
2177         }
2178
2179         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2180         if (!osb->vol_label) {
2181                 mlog(ML_ERROR, "unable to alloc vol label\n");
2182                 status = -ENOMEM;
2183                 goto bail;
2184         }
2185
2186         osb->slot_recovery_generations =
2187                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2188                         GFP_KERNEL);
2189         if (!osb->slot_recovery_generations) {
2190                 status = -ENOMEM;
2191                 mlog_errno(status);
2192                 goto bail;
2193         }
2194
2195         init_waitqueue_head(&osb->osb_wipe_event);
2196         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2197                                         sizeof(*osb->osb_orphan_wipes),
2198                                         GFP_KERNEL);
2199         if (!osb->osb_orphan_wipes) {
2200                 status = -ENOMEM;
2201                 mlog_errno(status);
2202                 goto bail;
2203         }
2204
2205         osb->osb_rf_lock_tree = RB_ROOT;
2206
2207         osb->s_feature_compat =
2208                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2209         osb->s_feature_ro_compat =
2210                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2211         osb->s_feature_incompat =
2212                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2213
2214         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2215                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2216                      "optional features (%x).\n", i);
2217                 status = -EINVAL;
2218                 goto bail;
2219         }
2220         if (!(osb->sb->s_flags & MS_RDONLY) &&
2221             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2222                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2223                      "unsupported optional features (%x).\n", i);
2224                 status = -EINVAL;
2225                 goto bail;
2226         }
2227
2228         if (ocfs2_clusterinfo_valid(osb)) {
2229                 osb->osb_stackflags =
2230                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2231                 memcpy(osb->osb_cluster_stack,
2232                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2233                        OCFS2_STACK_LABEL_LEN);
2234                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2235                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2236                         mlog(ML_ERROR,
2237                              "couldn't mount because of an invalid "
2238                              "cluster stack label (%s) \n",
2239                              osb->osb_cluster_stack);
2240                         status = -EINVAL;
2241                         goto bail;
2242                 }
2243         } else {
2244                 /* The empty string is identical with classic tools that
2245                  * don't know about s_cluster_info. */
2246                 osb->osb_cluster_stack[0] = '\0';
2247         }
2248
2249         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2250
2251         /* FIXME
2252          * This should be done in ocfs2_journal_init(), but unknown
2253          * ordering issues will cause the filesystem to crash.
2254          * If anyone wants to figure out what part of the code
2255          * refers to osb->journal before ocfs2_journal_init() is run,
2256          * be my guest.
2257          */
2258         /* initialize our journal structure */
2259
2260         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2261         if (!journal) {
2262                 mlog(ML_ERROR, "unable to alloc journal\n");
2263                 status = -ENOMEM;
2264                 goto bail;
2265         }
2266         osb->journal = journal;
2267         journal->j_osb = osb;
2268
2269         atomic_set(&journal->j_num_trans, 0);
2270         init_rwsem(&journal->j_trans_barrier);
2271         init_waitqueue_head(&journal->j_checkpointed);
2272         spin_lock_init(&journal->j_lock);
2273         journal->j_trans_id = (unsigned long) 1;
2274         INIT_LIST_HEAD(&journal->j_la_cleanups);
2275         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2276         journal->j_state = OCFS2_JOURNAL_FREE;
2277
2278         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2279         osb->dentry_lock_list = NULL;
2280
2281         /* get some pseudo constants for clustersize bits */
2282         osb->s_clustersize_bits =
2283                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2284         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2285         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2286
2287         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2288             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2289                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2290                      osb->s_clustersize);
2291                 status = -EINVAL;
2292                 goto bail;
2293         }
2294
2295         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2296                                                 le32_to_cpu(di->i_clusters));
2297
2298         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2299                                            total_blocks);
2300         if (status) {
2301                 mlog(ML_ERROR, "Volume too large "
2302                      "to mount safely on this system");
2303                 status = -EFBIG;
2304                 goto bail;
2305         }
2306
2307         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2308                                  sizeof(di->id2.i_super.s_uuid))) {
2309                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2310                 status = -ENOMEM;
2311                 goto bail;
2312         }
2313
2314         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2315
2316         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2317         osb->vol_label[63] = '\0';
2318         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2319         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2320         osb->first_cluster_group_blkno =
2321                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2322         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2323         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2324         mlog(0, "vol_label: %s\n", osb->vol_label);
2325         mlog(0, "uuid: %s\n", osb->uuid_str);
2326         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2327              (unsigned long long)osb->root_blkno,
2328              (unsigned long long)osb->system_dir_blkno);
2329
2330         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2331         if (!osb->osb_dlm_debug) {
2332                 status = -ENOMEM;
2333                 mlog_errno(status);
2334                 goto bail;
2335         }
2336
2337         atomic_set(&osb->vol_state, VOLUME_INIT);
2338
2339         /* load root, system_dir, and all global system inodes */
2340         status = ocfs2_init_global_system_inodes(osb);
2341         if (status < 0) {
2342                 mlog_errno(status);
2343                 goto bail;
2344         }
2345
2346         /*
2347          * global bitmap
2348          */
2349         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2350                                             OCFS2_INVALID_SLOT);
2351         if (!inode) {
2352                 status = -EINVAL;
2353                 mlog_errno(status);
2354                 goto bail;
2355         }
2356
2357         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2358         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2359         iput(inode);
2360
2361         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2362                                  osb->s_feature_incompat) * 8;
2363
2364         status = ocfs2_init_slot_info(osb);
2365         if (status < 0) {
2366                 mlog_errno(status);
2367                 goto bail;
2368         }
2369
2370 bail:
2371         mlog_exit(status);
2372         return status;
2373 }
2374
2375 /*
2376  * will return: -EAGAIN if it is ok to keep searching for superblocks
2377  *              -EINVAL if there is a bad superblock
2378  *              0 on success
2379  */
2380 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2381                                struct buffer_head *bh,
2382                                u32 blksz,
2383                                struct ocfs2_blockcheck_stats *stats)
2384 {
2385         int status = -EAGAIN;
2386
2387         mlog_entry_void();
2388
2389         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2390                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2391                 /* We have to do a raw check of the feature here */
2392                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2393                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2394                         status = ocfs2_block_check_validate(bh->b_data,
2395                                                             bh->b_size,
2396                                                             &di->i_check,
2397                                                             stats);
2398                         if (status)
2399                                 goto out;
2400                 }
2401                 status = -EINVAL;
2402                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2403                         mlog(ML_ERROR, "found superblock with incorrect block "
2404                              "size: found %u, should be %u\n",
2405                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2406                                blksz);
2407                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2408                            OCFS2_MAJOR_REV_LEVEL ||
2409                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2410                            OCFS2_MINOR_REV_LEVEL) {
2411                         mlog(ML_ERROR, "found superblock with bad version: "
2412                              "found %u.%u, should be %u.%u\n",
2413                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2414                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2415                              OCFS2_MAJOR_REV_LEVEL,
2416                              OCFS2_MINOR_REV_LEVEL);
2417                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2418                         mlog(ML_ERROR, "bad block number on superblock: "
2419                              "found %llu, should be %llu\n",
2420                              (unsigned long long)le64_to_cpu(di->i_blkno),
2421                              (unsigned long long)bh->b_blocknr);
2422                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2423                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2424                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2425                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2426                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2427                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2428                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2429                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2430                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2431                         mlog(ML_ERROR,
2432                              "Superblock slots found greater than file system "
2433                              "maximum: found %u, max %u\n",
2434                              le16_to_cpu(di->id2.i_super.s_max_slots),
2435                              OCFS2_MAX_SLOTS);
2436                 } else {
2437                         /* found it! */
2438                         status = 0;
2439                 }
2440         }
2441
2442 out:
2443         mlog_exit(status);
2444         return status;
2445 }
2446
2447 static int ocfs2_check_volume(struct ocfs2_super *osb)
2448 {
2449         int status;
2450         int dirty;
2451         int local;
2452         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2453                                                   * recover
2454                                                   * ourselves. */
2455
2456         mlog_entry_void();
2457
2458         /* Init our journal object. */
2459         status = ocfs2_journal_init(osb->journal, &dirty);
2460         if (status < 0) {
2461                 mlog(ML_ERROR, "Could not initialize journal!\n");
2462                 goto finally;
2463         }
2464
2465         /* Now that journal has been initialized, check to make sure
2466            entire volume is addressable. */
2467         status = ocfs2_journal_addressable(osb);
2468         if (status)
2469                 goto finally;
2470
2471         /* If the journal was unmounted cleanly then we don't want to
2472          * recover anything. Otherwise, journal_load will do that
2473          * dirty work for us :) */
2474         if (!dirty) {
2475                 status = ocfs2_journal_wipe(osb->journal, 0);
2476                 if (status < 0) {
2477                         mlog_errno(status);
2478                         goto finally;
2479                 }
2480         } else {
2481                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2482                      "recovering volume.\n");
2483         }
2484
2485         local = ocfs2_mount_local(osb);
2486
2487         /* will play back anything left in the journal. */
2488         status = ocfs2_journal_load(osb->journal, local, dirty);
2489         if (status < 0) {
2490                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2491                 goto finally;
2492         }
2493
2494         if (dirty) {
2495                 /* recover my local alloc if we didn't unmount cleanly. */
2496                 status = ocfs2_begin_local_alloc_recovery(osb,
2497                                                           osb->slot_num,
2498                                                           &local_alloc);
2499                 if (status < 0) {
2500                         mlog_errno(status);
2501                         goto finally;
2502                 }
2503                 /* we complete the recovery process after we've marked
2504                  * ourselves as mounted. */
2505         }
2506
2507         mlog(0, "Journal loaded.\n");
2508
2509         status = ocfs2_load_local_alloc(osb);
2510         if (status < 0) {
2511                 mlog_errno(status);
2512                 goto finally;
2513         }
2514
2515         if (dirty) {
2516                 /* Recovery will be completed after we've mounted the
2517                  * rest of the volume. */
2518                 osb->dirty = 1;
2519                 osb->local_alloc_copy = local_alloc;
2520                 local_alloc = NULL;
2521         }
2522
2523         /* go through each journal, trylock it and if you get the
2524          * lock, and it's marked as dirty, set the bit in the recover
2525          * map and launch a recovery thread for it. */
2526         status = ocfs2_mark_dead_nodes(osb);
2527         if (status < 0) {
2528                 mlog_errno(status);
2529                 goto finally;
2530         }
2531
2532         status = ocfs2_compute_replay_slots(osb);
2533         if (status < 0)
2534                 mlog_errno(status);
2535
2536 finally:
2537         if (local_alloc)
2538                 kfree(local_alloc);
2539
2540         mlog_exit(status);
2541         return status;
2542 }
2543
2544 /*
2545  * The routine gets called from dismount or close whenever a dismount on
2546  * volume is requested and the osb open count becomes 1.
2547  * It will remove the osb from the global list and also free up all the
2548  * initialized resources and fileobject.
2549  */
2550 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2551 {
2552         mlog_entry_void();
2553
2554         /* This function assumes that the caller has the main osb resource */
2555
2556         ocfs2_free_slot_info(osb);
2557
2558         kfree(osb->osb_orphan_wipes);
2559         kfree(osb->slot_recovery_generations);
2560         /* FIXME
2561          * This belongs in journal shutdown, but because we have to
2562          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2563          * we free it here.
2564          */
2565         kfree(osb->journal);
2566         if (osb->local_alloc_copy)
2567                 kfree(osb->local_alloc_copy);
2568         kfree(osb->uuid_str);
2569         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2570         memset(osb, 0, sizeof(struct ocfs2_super));
2571
2572         mlog_exit_void();
2573 }
2574
2575 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2576  * panic(). We do not support continue-on-error operation. */
2577 static void ocfs2_handle_error(struct super_block *sb)
2578 {
2579         struct ocfs2_super *osb = OCFS2_SB(sb);
2580
2581         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2582                 panic("OCFS2: (device %s): panic forced after error\n",
2583                       sb->s_id);
2584
2585         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2586
2587         if (sb->s_flags & MS_RDONLY &&
2588             (ocfs2_is_soft_readonly(osb) ||
2589              ocfs2_is_hard_readonly(osb)))
2590                 return;
2591
2592         printk(KERN_CRIT "File system is now read-only due to the potential "
2593                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2594                "system is unmounted.\n");
2595         sb->s_flags |= MS_RDONLY;
2596         ocfs2_set_ro_flag(osb, 0);
2597 }
2598
2599 static char error_buf[1024];
2600
2601 void __ocfs2_error(struct super_block *sb,
2602                    const char *function,
2603                    const char *fmt, ...)
2604 {
2605         va_list args;
2606
2607         va_start(args, fmt);
2608         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2609         va_end(args);
2610
2611         /* Not using mlog here because we want to show the actual
2612          * function the error came from. */
2613         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2614                sb->s_id, function, error_buf);
2615
2616         ocfs2_handle_error(sb);
2617 }
2618
2619 /* Handle critical errors. This is intentionally more drastic than
2620  * ocfs2_handle_error, so we only use for things like journal errors,
2621  * etc. */
2622 void __ocfs2_abort(struct super_block* sb,
2623                    const char *function,
2624                    const char *fmt, ...)
2625 {
2626         va_list args;
2627
2628         va_start(args, fmt);
2629         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2630         va_end(args);
2631
2632         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2633                sb->s_id, function, error_buf);
2634
2635         /* We don't have the cluster support yet to go straight to
2636          * hard readonly in here. Until then, we want to keep
2637          * ocfs2_abort() so that we can at least mark critical
2638          * errors.
2639          *
2640          * TODO: This should abort the journal and alert other nodes
2641          * that our slot needs recovery. */
2642
2643         /* Force a panic(). This stinks, but it's better than letting
2644          * things continue without having a proper hard readonly
2645          * here. */
2646         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2647                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2648         ocfs2_handle_error(sb);
2649 }
2650
2651 /*
2652  * Void signal blockers, because in-kernel sigprocmask() only fails
2653  * when SIG_* is wrong.
2654  */
2655 void ocfs2_block_signals(sigset_t *oldset)
2656 {
2657         int rc;
2658         sigset_t blocked;
2659
2660         sigfillset(&blocked);
2661         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2662         BUG_ON(rc);
2663 }
2664
2665 void ocfs2_unblock_signals(sigset_t *oldset)
2666 {
2667         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2668         BUG_ON(rc);
2669 }
2670
2671 module_init(ocfs2_init);
2672 module_exit(ocfs2_exit);