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_i_callback(struct rcu_head *head)
 573 {
 574         struct inode *inode = container_of(head, struct inode, i_rcu);
 575         INIT_LIST_HEAD(&inode->i_dentry);
 576         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 577 }
 578
 579 static void ocfs2_destroy_inode(struct inode *inode)
 580 {
 581         call_rcu(&inode->i_rcu, ocfs2_i_callback);
 582 }
 583
 584 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 585                                                 unsigned int cbits)
 586 {
 587         unsigned int bytes = 1 << cbits;
 588         unsigned int trim = bytes;
 589         unsigned int bitshift = 32;
 590
 591         /*
 592          * i_size and all block offsets in ocfs2 are always 64 bits
 593          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 594          * 64 bit platforms, cluster size will be the limiting factor.
 595          */
 596
 597 #if BITS_PER_LONG == 32
 598 # if defined(CONFIG_LBDAF)
 599         BUILD_BUG_ON(sizeof(sector_t) != 8);
 600         /*
 601          * We might be limited by page cache size.
 602          */
 603         if (bytes > PAGE_CACHE_SIZE) {
 604                 bytes = PAGE_CACHE_SIZE;
 605                 trim = 1;
 606                 /*
 607                  * Shift by 31 here so that we don't get larger than
 608                  * MAX_LFS_FILESIZE
 609                  */
 610                 bitshift = 31;
 611         }
 612 # else
 613         /*
 614          * We are limited by the size of sector_t. Use block size, as
 615          * that's what we expose to the VFS.
 616          */
 617         bytes = 1 << bbits;
 618         trim = 1;
 619         bitshift = 31;
 620 # endif
 621 #endif
 622
 623         /*
 624          * Trim by a whole cluster when we can actually approach the
 625          * on-disk limits. Otherwise we can overflow i_clusters when
 626          * an extent start is at the max offset.
 627          */
 628         return (((unsigned long long)bytes) << bitshift) - trim;
 629 }
 630
 631 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 632 {
 633         int incompat_features;
 634         int ret = 0;
 635         struct mount_options parsed_options;
 636         struct ocfs2_super *osb = OCFS2_SB(sb);
 637         u32 tmp;
 638
 639         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 640             !ocfs2_check_set_options(sb, &parsed_options)) {
 641                 ret = -EINVAL;
 642                 goto out;
 643         }
 644
 645         tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
 646                 OCFS2_MOUNT_HB_NONE;
 647         if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
 648                 ret = -EINVAL;
 649                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 650                 goto out;
 651         }
 652
 653         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 654             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 655                 ret = -EINVAL;
 656                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 657                 goto out;
 658         }
 659
 660         /* Probably don't want this on remount; it might
 661          * mess with other nodes */
 662         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 663             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 664                 ret = -EINVAL;
 665                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 666                 goto out;
 667         }
 668
 669         /* We're going to/from readonly mode. */
 670         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 671                 /* Disable quota accounting before remounting RO */
 672                 if (*flags & MS_RDONLY) {
 673                         ret = ocfs2_susp_quotas(osb, 0);
 674                         if (ret < 0)
 675                                 goto out;
 676                 }
 677                 /* Lock here so the check of HARD_RO and the potential
 678                  * setting of SOFT_RO is atomic. */
 679                 spin_lock(&osb->osb_lock);
 680                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 681                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 682                         ret = -EROFS;
 683                         goto unlock_osb;
 684                 }
 685
 686                 if (*flags & MS_RDONLY) {
 687                         mlog(0, "Going to ro mode.\n");
 688                         sb->s_flags |= MS_RDONLY;
 689                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 690                 } else {
 691                         mlog(0, "Making ro filesystem writeable.\n");
 692
 693                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 694                                 mlog(ML_ERROR, "Cannot remount RDWR "
 695                                      "filesystem due to previous errors.\n");
 696                                 ret = -EROFS;
 697                                 goto unlock_osb;
 698                         }
 699                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 700                         if (incompat_features) {
 701                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 702                                      "of unsupported optional features "
 703                                      "(%x).\n", incompat_features);
 704                                 ret = -EINVAL;
 705                                 goto unlock_osb;
 706                         }
 707                         sb->s_flags &= ~MS_RDONLY;
 708                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 709                 }
 710 unlock_osb:
 711                 spin_unlock(&osb->osb_lock);
 712                 /* Enable quota accounting after remounting RW */
 713                 if (!ret && !(*flags & MS_RDONLY)) {
 714                         if (sb_any_quota_suspended(sb))
 715                                 ret = ocfs2_susp_quotas(osb, 1);
 716                         else
 717                                 ret = ocfs2_enable_quotas(osb);
 718                         if (ret < 0) {
 719                                 /* Return back changes... */
 720                                 spin_lock(&osb->osb_lock);
 721                                 sb->s_flags |= MS_RDONLY;
 722                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 723                                 spin_unlock(&osb->osb_lock);
 724                                 goto out;
 725                         }
 726                 }
 727         }
 728
 729         if (!ret) {
 730                 /* Only save off the new mount options in case of a successful
 731                  * remount. */
 732                 osb->s_mount_opt = parsed_options.mount_opt;
 733                 osb->s_atime_quantum = parsed_options.atime_quantum;
 734                 osb->preferred_slot = parsed_options.slot;
 735                 if (parsed_options.commit_interval)
 736                         osb->osb_commit_interval = parsed_options.commit_interval;
 737
 738                 if (!ocfs2_is_hard_readonly(osb))
 739                         ocfs2_set_journal_params(osb);
 740
 741                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 742                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 743                                                         MS_POSIXACL : 0);
 744         }
 745 out:
 746         return ret;
 747 }
 748
 749 static int ocfs2_sb_probe(struct super_block *sb,
 750                           struct buffer_head **bh,
 751                           int *sector_size,
 752                           struct ocfs2_blockcheck_stats *stats)
 753 {
 754         int status, tmpstat;
 755         struct ocfs1_vol_disk_hdr *hdr;
 756         struct ocfs2_dinode *di;
 757         int blksize;
 758
 759         *bh = NULL;
 760
 761         /* may be > 512 */
 762         *sector_size = bdev_logical_block_size(sb->s_bdev);
 763         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 764                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 765                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 766                 status = -EINVAL;
 767                 goto bail;
 768         }
 769
 770         /* Can this really happen? */
 771         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 772                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 773
 774         /* check block zero for old format */
 775         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 776         if (status < 0) {
 777                 mlog_errno(status);
 778                 goto bail;
 779         }
 780         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 781         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 782                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 783                      hdr->major_version, hdr->minor_version);
 784                 status = -EINVAL;
 785         }
 786         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 787                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 788                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 789                      hdr->signature);
 790                 status = -EINVAL;
 791         }
 792         brelse(*bh);
 793         *bh = NULL;
 794         if (status < 0) {
 795                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 796                      "upgraded before mounting with ocfs v2\n");
 797                 goto bail;
 798         }
 799
 800         /*
 801          * Now check at magic offset for 512, 1024, 2048, 4096
 802          * blocksizes.  4096 is the maximum blocksize because it is
 803          * the minimum clustersize.
 804          */
 805         status = -EINVAL;
 806         for (blksize = *sector_size;
 807              blksize <= OCFS2_MAX_BLOCKSIZE;
 808              blksize <<= 1) {
 809                 tmpstat = ocfs2_get_sector(sb, bh,
 810                                            OCFS2_SUPER_BLOCK_BLKNO,
 811                                            blksize);
 812                 if (tmpstat < 0) {
 813                         status = tmpstat;
 814                         mlog_errno(status);
 815                         break;
 816                 }
 817                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 818                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 819                 spin_lock_init(&stats->b_lock);
 820                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 821                 if (tmpstat < 0) {
 822                         brelse(*bh);
 823                         *bh = NULL;
 824                 }
 825                 if (tmpstat != -EAGAIN) {
 826                         status = tmpstat;
 827                         break;
 828                 }
 829         }
 830
 831 bail:
 832         return status;
 833 }
 834
 835 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 836 {
 837         u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
 838
 839         if (osb->s_mount_opt & hb_enabled) {
 840                 if (ocfs2_mount_local(osb)) {
 841                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 842                              "mounted device.\n");
 843                         return -EINVAL;
 844                 }
 845                 if (ocfs2_userspace_stack(osb)) {
 846                         mlog(ML_ERROR, "Userspace stack expected, but "
 847                              "o2cb heartbeat arguments passed to mount\n");
 848                         return -EINVAL;
 849                 }
 850                 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
 851                      !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
 852                     ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
 853                      ocfs2_cluster_o2cb_global_heartbeat(osb))) {
 854                         mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
 855                         return -EINVAL;
 856                 }
 857         }
 858
 859         if (!(osb->s_mount_opt & hb_enabled)) {
 860                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 861                     !ocfs2_userspace_stack(osb)) {
 862                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 863                              "a read-write clustered device.\n");
 864                         return -EINVAL;
 865                 }
 866         }
 867
 868         return 0;
 869 }
 870
 871 /*
 872  * If we're using a userspace stack, mount should have passed
 873  * a name that matches the disk.  If not, mount should not
 874  * have passed a stack.
 875  */
 876 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 877                                         struct mount_options *mopt)
 878 {
 879         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 880                 mlog(ML_ERROR,
 881                      "cluster stack passed to mount, but this filesystem "
 882                      "does not support it\n");
 883                 return -EINVAL;
 884         }
 885
 886         if (ocfs2_userspace_stack(osb) &&
 887             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 888                     OCFS2_STACK_LABEL_LEN)) {
 889                 mlog(ML_ERROR,
 890                      "cluster stack passed to mount (\"%s\") does not "
 891                      "match the filesystem (\"%s\")\n",
 892                      mopt->cluster_stack,
 893                      osb->osb_cluster_stack);
 894                 return -EINVAL;
 895         }
 896
 897         return 0;
 898 }
 899
 900 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 901 {
 902         int type;
 903         struct super_block *sb = osb->sb;
 904         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 905                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 906         int status = 0;
 907
 908         for (type = 0; type < MAXQUOTAS; type++) {
 909                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 910                         continue;
 911                 if (unsuspend)
 912                         status = dquot_resume(sb, type);
 913                 else {
 914                         struct ocfs2_mem_dqinfo *oinfo;
 915
 916                         /* Cancel periodic syncing before suspending */
 917                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 918                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 919                         status = dquot_suspend(sb, type);
 920                 }
 921                 if (status < 0)
 922                         break;
 923         }
 924         if (status < 0)
 925                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 926                      "remount (error = %d).\n", status);
 927         return status;
 928 }
 929
 930 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 931 {
 932         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 933         struct super_block *sb = osb->sb;
 934         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 935                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 936         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 937                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 938         int status;
 939         int type;
 940
 941         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 942         for (type = 0; type < MAXQUOTAS; type++) {
 943                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 944                         continue;
 945                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 946                                                         osb->slot_num);
 947                 if (!inode[type]) {
 948                         status = -ENOENT;
 949                         goto out_quota_off;
 950                 }
 951                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 952                                       DQUOT_USAGE_ENABLED);
 953                 if (status < 0)
 954                         goto out_quota_off;
 955         }
 956
 957         for (type = 0; type < MAXQUOTAS; type++)
 958                 iput(inode[type]);
 959         return 0;
 960 out_quota_off:
 961         ocfs2_disable_quotas(osb);
 962         for (type = 0; type < MAXQUOTAS; type++)
 963                 iput(inode[type]);
 964         mlog_errno(status);
 965         return status;
 966 }
 967
 968 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 969 {
 970         int type;
 971         struct inode *inode;
 972         struct super_block *sb = osb->sb;
 973         struct ocfs2_mem_dqinfo *oinfo;
 974
 975         /* We mostly ignore errors in this function because there's not much
 976          * we can do when we see them */
 977         for (type = 0; type < MAXQUOTAS; type++) {
 978                 if (!sb_has_quota_loaded(sb, type))
 979                         continue;
 980                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 981                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 982                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 983                 inode = igrab(sb->s_dquot.files[type]);
 984                 /* Turn off quotas. This will remove all dquot structures from
 985                  * memory and so they will be automatically synced to global
 986                  * quota files */
 987                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 988                                         DQUOT_LIMITS_ENABLED);
 989                 if (!inode)
 990                         continue;
 991                 iput(inode);
 992         }
 993 }
 994
 995 /* Handle quota on quotactl */
 996 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 997                           char *path)
 998 {
 999         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
1000                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
1001
1002         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1003                 return -EINVAL;
1004
1005         return dquot_enable(sb_dqopt(sb)->files[type], type,
1006                             format_id, DQUOT_LIMITS_ENABLED);
1007 }
1008
1009 /* Handle quota off quotactl */
1010 static int ocfs2_quota_off(struct super_block *sb, int type)
1011 {
1012         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1013 }
1014
1015 static const struct quotactl_ops ocfs2_quotactl_ops = {
1016         .quota_on       = ocfs2_quota_on,
1017         .quota_off      = ocfs2_quota_off,
1018         .quota_sync     = dquot_quota_sync,
1019         .get_info       = dquot_get_dqinfo,
1020         .set_info       = dquot_set_dqinfo,
1021         .get_dqblk      = dquot_get_dqblk,
1022         .set_dqblk      = dquot_set_dqblk,
1023 };
1024
1025 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1026 {
1027         struct dentry *root;
1028         int status, sector_size;
1029         struct mount_options parsed_options;
1030         struct inode *inode = NULL;
1031         struct ocfs2_super *osb = NULL;
1032         struct buffer_head *bh = NULL;
1033         char nodestr[8];
1034         struct ocfs2_blockcheck_stats stats;
1035
1036         mlog_entry("%p, %p, %i", sb, data, silent);
1037
1038         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1039                 status = -EINVAL;
1040                 goto read_super_error;
1041         }
1042
1043         /* probe for superblock */
1044         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1045         if (status < 0) {
1046                 mlog(ML_ERROR, "superblock probe failed!\n");
1047                 goto read_super_error;
1048         }
1049
1050         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1051         osb = OCFS2_SB(sb);
1052         if (status < 0) {
1053                 mlog_errno(status);
1054                 goto read_super_error;
1055         }
1056         brelse(bh);
1057         bh = NULL;
1058
1059         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1060                 status = -EINVAL;
1061                 goto read_super_error;
1062         }
1063         osb->s_mount_opt = parsed_options.mount_opt;
1064         osb->s_atime_quantum = parsed_options.atime_quantum;
1065         osb->preferred_slot = parsed_options.slot;
1066         osb->osb_commit_interval = parsed_options.commit_interval;
1067
1068         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1069         osb->osb_resv_level = parsed_options.resv_level;
1070         osb->osb_dir_resv_level = parsed_options.resv_level;
1071         if (parsed_options.dir_resv_level == -1)
1072                 osb->osb_dir_resv_level = parsed_options.resv_level;
1073         else
1074                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1075
1076         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1077         if (status)
1078                 goto read_super_error;
1079
1080         sb->s_magic = OCFS2_SUPER_MAGIC;
1081
1082         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1083                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1084
1085         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1086          * heartbeat=none */
1087         if (bdev_read_only(sb->s_bdev)) {
1088                 if (!(sb->s_flags & MS_RDONLY)) {
1089                         status = -EACCES;
1090                         mlog(ML_ERROR, "Readonly device detected but readonly "
1091                              "mount was not specified.\n");
1092                         goto read_super_error;
1093                 }
1094
1095                 /* You should not be able to start a local heartbeat
1096                  * on a readonly device. */
1097                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1098                         status = -EROFS;
1099                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1100                              "device.\n");
1101                         goto read_super_error;
1102                 }
1103
1104                 status = ocfs2_check_journals_nolocks(osb);
1105                 if (status < 0) {
1106                         if (status == -EROFS)
1107                                 mlog(ML_ERROR, "Recovery required on readonly "
1108                                      "file system, but write access is "
1109                                      "unavailable.\n");
1110                         else
1111                                 mlog_errno(status);
1112                         goto read_super_error;
1113                 }
1114
1115                 ocfs2_set_ro_flag(osb, 1);
1116
1117                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1118                        "services will be utilized for this mount. Recovery "
1119                        "will be skipped.\n");
1120         }
1121
1122         if (!ocfs2_is_hard_readonly(osb)) {
1123                 if (sb->s_flags & MS_RDONLY)
1124                         ocfs2_set_ro_flag(osb, 0);
1125         }
1126
1127         status = ocfs2_verify_heartbeat(osb);
1128         if (status < 0) {
1129                 mlog_errno(status);
1130                 goto read_super_error;
1131         }
1132
1133         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1134                                                  ocfs2_debugfs_root);
1135         if (!osb->osb_debug_root) {
1136                 status = -EINVAL;
1137                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1138                 goto read_super_error;
1139         }
1140
1141         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1142                                             osb->osb_debug_root,
1143                                             osb,
1144                                             &ocfs2_osb_debug_fops);
1145         if (!osb->osb_ctxt) {
1146                 status = -EINVAL;
1147                 mlog_errno(status);
1148                 goto read_super_error;
1149         }
1150
1151         if (ocfs2_meta_ecc(osb)) {
1152                 status = ocfs2_blockcheck_stats_debugfs_install(
1153                                                 &osb->osb_ecc_stats,
1154                                                 osb->osb_debug_root);
1155                 if (status) {
1156                         mlog(ML_ERROR,
1157                              "Unable to create blockcheck statistics "
1158                              "files\n");
1159                         goto read_super_error;
1160                 }
1161         }
1162
1163         status = ocfs2_mount_volume(sb);
1164         if (osb->root_inode)
1165                 inode = igrab(osb->root_inode);
1166
1167         if (status < 0)
1168                 goto read_super_error;
1169
1170         if (!inode) {
1171                 status = -EIO;
1172                 mlog_errno(status);
1173                 goto read_super_error;
1174         }
1175
1176         root = d_alloc_root(inode);
1177         if (!root) {
1178                 status = -ENOMEM;
1179                 mlog_errno(status);
1180                 goto read_super_error;
1181         }
1182
1183         sb->s_root = root;
1184
1185         ocfs2_complete_mount_recovery(osb);
1186
1187         if (ocfs2_mount_local(osb))
1188                 snprintf(nodestr, sizeof(nodestr), "local");
1189         else
1190                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1191
1192         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1193                "with %s data mode.\n",
1194                osb->dev_str, nodestr, osb->slot_num,
1195                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1196                "ordered");
1197
1198         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1199         wake_up(&osb->osb_mount_event);
1200
1201         /* Now we can initialize quotas because we can afford to wait
1202          * for cluster locks recovery now. That also means that truncation
1203          * log recovery can happen but that waits for proper quota setup */
1204         if (!(sb->s_flags & MS_RDONLY)) {
1205                 status = ocfs2_enable_quotas(osb);
1206                 if (status < 0) {
1207                         /* We have to err-out specially here because
1208                          * s_root is already set */
1209                         mlog_errno(status);
1210                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1211                         wake_up(&osb->osb_mount_event);
1212                         mlog_exit(status);
1213                         return status;
1214                 }
1215         }
1216
1217         ocfs2_complete_quota_recovery(osb);
1218
1219         /* Now we wake up again for processes waiting for quotas */
1220         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1221         wake_up(&osb->osb_mount_event);
1222
1223         /* Start this when the mount is almost sure of being successful */
1224         ocfs2_orphan_scan_start(osb);
1225
1226         mlog_exit(status);
1227         return status;
1228
1229 read_super_error:
1230         brelse(bh);
1231
1232         if (inode)
1233                 iput(inode);
1234
1235         if (osb) {
1236                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1237                 wake_up(&osb->osb_mount_event);
1238                 ocfs2_dismount_volume(sb, 1);
1239         }
1240
1241         mlog_exit(status);
1242         return status;
1243 }
1244
1245 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1246                         int flags,
1247                         const char *dev_name,
1248                         void *data)
1249 {
1250         return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1251 }
1252
1253 static void ocfs2_kill_sb(struct super_block *sb)
1254 {
1255         struct ocfs2_super *osb = OCFS2_SB(sb);
1256
1257         /* Failed mount? */
1258         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1259                 goto out;
1260
1261         /* Prevent further queueing of inode drop events */
1262         spin_lock(&dentry_list_lock);
1263         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1264         spin_unlock(&dentry_list_lock);
1265         /* Wait for work to finish and/or remove it */
1266         cancel_work_sync(&osb->dentry_lock_work);
1267 out:
1268         kill_block_super(sb);
1269 }
1270
1271 static struct file_system_type ocfs2_fs_type = {
1272         .owner          = THIS_MODULE,
1273         .name           = "ocfs2",
1274         .mount          = ocfs2_mount,
1275         .kill_sb        = ocfs2_kill_sb,
1276
1277         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1278         .next           = NULL
1279 };
1280
1281 static int ocfs2_check_set_options(struct super_block *sb,
1282                                    struct mount_options *options)
1283 {
1284         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1285             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1286                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1287                 mlog(ML_ERROR, "User 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_GRPQUOTA &&
1292             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1293                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1294                 mlog(ML_ERROR, "Group quotas were requested, but this "
1295                      "filesystem does not have the feature enabled.\n");
1296                 return 0;
1297         }
1298         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1299             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1300                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1301                      "feature is not enabled\n");
1302                 return 0;
1303         }
1304         /* No ACL setting specified? Use XATTR feature... */
1305         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1306                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1307                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1308                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1309                 else
1310                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1311         }
1312         return 1;
1313 }
1314
1315 static int ocfs2_parse_options(struct super_block *sb,
1316                                char *options,
1317                                struct mount_options *mopt,
1318                                int is_remount)
1319 {
1320         int status;
1321         char *p;
1322         u32 tmp;
1323
1324         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1325                    options ? options : "(none)");
1326
1327         mopt->commit_interval = 0;
1328         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1329         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1330         mopt->slot = OCFS2_INVALID_SLOT;
1331         mopt->localalloc_opt = -1;
1332         mopt->cluster_stack[0] = '\0';
1333         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1334         mopt->dir_resv_level = -1;
1335
1336         if (!options) {
1337                 status = 1;
1338                 goto bail;
1339         }
1340
1341         while ((p = strsep(&options, ",")) != NULL) {
1342                 int token, option;
1343                 substring_t args[MAX_OPT_ARGS];
1344
1345                 if (!*p)
1346                         continue;
1347
1348                 token = match_token(p, tokens, args);
1349                 switch (token) {
1350                 case Opt_hb_local:
1351                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1352                         break;
1353                 case Opt_hb_none:
1354                         mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1355                         break;
1356                 case Opt_hb_global:
1357                         mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1358                         break;
1359                 case Opt_barrier:
1360                         if (match_int(&args[0], &option)) {
1361                                 status = 0;
1362                                 goto bail;
1363                         }
1364                         if (option)
1365                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1366                         else
1367                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1368                         break;
1369                 case Opt_intr:
1370                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1371                         break;
1372                 case Opt_nointr:
1373                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1374                         break;
1375                 case Opt_err_panic:
1376                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1377                         break;
1378                 case Opt_err_ro:
1379                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1380                         break;
1381                 case Opt_data_ordered:
1382                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1383                         break;
1384                 case Opt_data_writeback:
1385                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1386                         break;
1387                 case Opt_user_xattr:
1388                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1389                         break;
1390                 case Opt_nouser_xattr:
1391                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1392                         break;
1393                 case Opt_atime_quantum:
1394                         if (match_int(&args[0], &option)) {
1395                                 status = 0;
1396                                 goto bail;
1397                         }
1398                         if (option >= 0)
1399                                 mopt->atime_quantum = option;
1400                         break;
1401                 case Opt_slot:
1402                         option = 0;
1403                         if (match_int(&args[0], &option)) {
1404                                 status = 0;
1405                                 goto bail;
1406                         }
1407                         if (option)
1408                                 mopt->slot = (s16)option;
1409                         break;
1410                 case Opt_commit:
1411                         option = 0;
1412                         if (match_int(&args[0], &option)) {
1413                                 status = 0;
1414                                 goto bail;
1415                         }
1416                         if (option < 0)
1417                                 return 0;
1418                         if (option == 0)
1419                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1420                         mopt->commit_interval = HZ * option;
1421                         break;
1422                 case Opt_localalloc:
1423                         option = 0;
1424                         if (match_int(&args[0], &option)) {
1425                                 status = 0;
1426                                 goto bail;
1427                         }
1428                         if (option >= 0)
1429                                 mopt->localalloc_opt = option;
1430                         break;
1431                 case Opt_localflocks:
1432                         /*
1433                          * Changing this during remount could race
1434                          * flock() requests, or "unbalance" existing
1435                          * ones (e.g., a lock is taken in one mode but
1436                          * dropped in the other). If users care enough
1437                          * to flip locking modes during remount, we
1438                          * could add a "local" flag to individual
1439                          * flock structures for proper tracking of
1440                          * state.
1441                          */
1442                         if (!is_remount)
1443                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1444                         break;
1445                 case Opt_stack:
1446                         /* Check both that the option we were passed
1447                          * is of the right length and that it is a proper
1448                          * string of the right length.
1449                          */
1450                         if (((args[0].to - args[0].from) !=
1451                              OCFS2_STACK_LABEL_LEN) ||
1452                             (strnlen(args[0].from,
1453                                      OCFS2_STACK_LABEL_LEN) !=
1454                              OCFS2_STACK_LABEL_LEN)) {
1455                                 mlog(ML_ERROR,
1456                                      "Invalid cluster_stack option\n");
1457                                 status = 0;
1458                                 goto bail;
1459                         }
1460                         memcpy(mopt->cluster_stack, args[0].from,
1461                                OCFS2_STACK_LABEL_LEN);
1462                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1463                         break;
1464                 case Opt_inode64:
1465                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1466                         break;
1467                 case Opt_usrquota:
1468                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1469                         break;
1470                 case Opt_grpquota:
1471                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1472                         break;
1473                 case Opt_coherency_buffered:
1474                         mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1475                         break;
1476                 case Opt_coherency_full:
1477                         mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1478                         break;
1479                 case Opt_acl:
1480                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1481                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1482                         break;
1483                 case Opt_noacl:
1484                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1485                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1486                         break;
1487                 case Opt_resv_level:
1488                         if (is_remount)
1489                                 break;
1490                         if (match_int(&args[0], &option)) {
1491                                 status = 0;
1492                                 goto bail;
1493                         }
1494                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1495                             option < OCFS2_MAX_RESV_LEVEL)
1496                                 mopt->resv_level = option;
1497                         break;
1498                 case Opt_dir_resv_level:
1499                         if (is_remount)
1500                                 break;
1501                         if (match_int(&args[0], &option)) {
1502                                 status = 0;
1503                                 goto bail;
1504                         }
1505                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1506                             option < OCFS2_MAX_RESV_LEVEL)
1507                                 mopt->dir_resv_level = option;
1508                         break;
1509                 default:
1510                         mlog(ML_ERROR,
1511                              "Unrecognized mount option \"%s\" "
1512                              "or missing value\n", p);
1513                         status = 0;
1514                         goto bail;
1515                 }
1516         }
1517
1518         /* Ensure only one heartbeat mode */
1519         tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
1520                                  OCFS2_MOUNT_HB_NONE);
1521         if (hweight32(tmp) != 1) {
1522                 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1523                 status = 0;
1524                 goto bail;
1525         }
1526
1527         status = 1;
1528
1529 bail:
1530         mlog_exit(status);
1531         return status;
1532 }
1533
1534 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1535 {
1536         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1537         unsigned long opts = osb->s_mount_opt;
1538         unsigned int local_alloc_megs;
1539
1540         if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1541                 seq_printf(s, ",_netdev");
1542                 if (opts & OCFS2_MOUNT_HB_LOCAL)
1543                         seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1544                 else
1545                         seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1546         } else
1547                 seq_printf(s, ",%s", OCFS2_HB_NONE);
1548
1549         if (opts & OCFS2_MOUNT_NOINTR)
1550                 seq_printf(s, ",nointr");
1551
1552         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1553                 seq_printf(s, ",data=writeback");
1554         else
1555                 seq_printf(s, ",data=ordered");
1556
1557         if (opts & OCFS2_MOUNT_BARRIER)
1558                 seq_printf(s, ",barrier=1");
1559
1560         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1561                 seq_printf(s, ",errors=panic");
1562         else
1563                 seq_printf(s, ",errors=remount-ro");
1564
1565         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1566                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1567
1568         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1569                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1570
1571         if (osb->osb_commit_interval)
1572                 seq_printf(s, ",commit=%u",
1573                            (unsigned) (osb->osb_commit_interval / HZ));
1574
1575         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1576         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1577                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1578
1579         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1580                 seq_printf(s, ",localflocks,");
1581
1582         if (osb->osb_cluster_stack[0])
1583                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1584                            osb->osb_cluster_stack);
1585         if (opts & OCFS2_MOUNT_USRQUOTA)
1586                 seq_printf(s, ",usrquota");
1587         if (opts & OCFS2_MOUNT_GRPQUOTA)
1588                 seq_printf(s, ",grpquota");
1589
1590         if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1591                 seq_printf(s, ",coherency=buffered");
1592         else
1593                 seq_printf(s, ",coherency=full");
1594
1595         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1596                 seq_printf(s, ",nouser_xattr");
1597         else
1598                 seq_printf(s, ",user_xattr");
1599
1600         if (opts & OCFS2_MOUNT_INODE64)
1601                 seq_printf(s, ",inode64");
1602
1603         if (opts & OCFS2_MOUNT_POSIX_ACL)
1604                 seq_printf(s, ",acl");
1605         else
1606                 seq_printf(s, ",noacl");
1607
1608         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1609                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1610
1611         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1612                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1613
1614         return 0;
1615 }
1616
1617 static int __init ocfs2_init(void)
1618 {
1619         int status;
1620
1621         mlog_entry_void();
1622
1623         ocfs2_print_version();
1624
1625         status = init_ocfs2_uptodate_cache();
1626         if (status < 0) {
1627                 mlog_errno(status);
1628                 goto leave;
1629         }
1630
1631         status = ocfs2_initialize_mem_caches();
1632         if (status < 0) {
1633                 mlog_errno(status);
1634                 goto leave;
1635         }
1636
1637         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1638         if (!ocfs2_wq) {
1639                 status = -ENOMEM;
1640                 goto leave;
1641         }
1642
1643         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1644         if (!ocfs2_debugfs_root) {
1645                 status = -EFAULT;
1646                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1647         }
1648
1649         status = ocfs2_quota_setup();
1650         if (status)
1651                 goto leave;
1652
1653         ocfs2_set_locking_protocol();
1654
1655         status = register_quota_format(&ocfs2_quota_format);
1656 leave:
1657         if (status < 0) {
1658                 ocfs2_quota_shutdown();
1659                 ocfs2_free_mem_caches();
1660                 exit_ocfs2_uptodate_cache();
1661         }
1662
1663         mlog_exit(status);
1664
1665         if (status >= 0) {
1666                 return register_filesystem(&ocfs2_fs_type);
1667         } else
1668                 return -1;
1669 }
1670
1671 static void __exit ocfs2_exit(void)
1672 {
1673         mlog_entry_void();
1674
1675         ocfs2_quota_shutdown();
1676
1677         if (ocfs2_wq) {
1678                 flush_workqueue(ocfs2_wq);
1679                 destroy_workqueue(ocfs2_wq);
1680         }
1681
1682         unregister_quota_format(&ocfs2_quota_format);
1683
1684         debugfs_remove(ocfs2_debugfs_root);
1685
1686         ocfs2_free_mem_caches();
1687
1688         unregister_filesystem(&ocfs2_fs_type);
1689
1690         exit_ocfs2_uptodate_cache();
1691
1692         mlog_exit_void();
1693 }
1694
1695 static void ocfs2_put_super(struct super_block *sb)
1696 {
1697         mlog_entry("(0x%p)\n", sb);
1698
1699         ocfs2_sync_blockdev(sb);
1700         ocfs2_dismount_volume(sb, 0);
1701
1702         mlog_exit_void();
1703 }
1704
1705 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1706 {
1707         struct ocfs2_super *osb;
1708         u32 numbits, freebits;
1709         int status;
1710         struct ocfs2_dinode *bm_lock;
1711         struct buffer_head *bh = NULL;
1712         struct inode *inode = NULL;
1713
1714         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1715
1716         osb = OCFS2_SB(dentry->d_sb);
1717
1718         inode = ocfs2_get_system_file_inode(osb,
1719                                             GLOBAL_BITMAP_SYSTEM_INODE,
1720                                             OCFS2_INVALID_SLOT);
1721         if (!inode) {
1722                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1723                 status = -EIO;
1724                 goto bail;
1725         }
1726
1727         status = ocfs2_inode_lock(inode, &bh, 0);
1728         if (status < 0) {
1729                 mlog_errno(status);
1730                 goto bail;
1731         }
1732
1733         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1734
1735         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1736         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1737
1738         buf->f_type = OCFS2_SUPER_MAGIC;
1739         buf->f_bsize = dentry->d_sb->s_blocksize;
1740         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1741         buf->f_blocks = ((sector_t) numbits) *
1742                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1743         buf->f_bfree = ((sector_t) freebits) *
1744                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1745         buf->f_bavail = buf->f_bfree;
1746         buf->f_files = numbits;
1747         buf->f_ffree = freebits;
1748         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1749                                 & 0xFFFFFFFFUL;
1750         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1751                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1752
1753         brelse(bh);
1754
1755         ocfs2_inode_unlock(inode, 0);
1756         status = 0;
1757 bail:
1758         if (inode)
1759                 iput(inode);
1760
1761         mlog_exit(status);
1762
1763         return status;
1764 }
1765
1766 static void ocfs2_inode_init_once(void *data)
1767 {
1768         struct ocfs2_inode_info *oi = data;
1769
1770         oi->ip_flags = 0;
1771         oi->ip_open_count = 0;
1772         spin_lock_init(&oi->ip_lock);
1773         ocfs2_extent_map_init(&oi->vfs_inode);
1774         INIT_LIST_HEAD(&oi->ip_io_markers);
1775         oi->ip_dir_start_lookup = 0;
1776
1777         init_rwsem(&oi->ip_alloc_sem);
1778         init_rwsem(&oi->ip_xattr_sem);
1779         mutex_init(&oi->ip_io_mutex);
1780
1781         oi->ip_blkno = 0ULL;
1782         oi->ip_clusters = 0;
1783
1784         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1785
1786         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1787         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1788         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1789
1790         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1791                                   &ocfs2_inode_caching_ops);
1792
1793         inode_init_once(&oi->vfs_inode);
1794 }
1795
1796 static int ocfs2_initialize_mem_caches(void)
1797 {
1798         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1799                                        sizeof(struct ocfs2_inode_info),
1800                                        0,
1801                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1802                                                 SLAB_MEM_SPREAD),
1803                                        ocfs2_inode_init_once);
1804         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1805                                         sizeof(struct ocfs2_dquot),
1806                                         0,
1807                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1808                                                 SLAB_MEM_SPREAD),
1809                                         NULL);
1810         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1811                                         sizeof(struct ocfs2_quota_chunk),
1812                                         0,
1813                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1814                                         NULL);
1815         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1816             !ocfs2_qf_chunk_cachep) {
1817                 if (ocfs2_inode_cachep)
1818                         kmem_cache_destroy(ocfs2_inode_cachep);
1819                 if (ocfs2_dquot_cachep)
1820                         kmem_cache_destroy(ocfs2_dquot_cachep);
1821                 if (ocfs2_qf_chunk_cachep)
1822                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1823                 return -ENOMEM;
1824         }
1825
1826         return 0;
1827 }
1828
1829 static void ocfs2_free_mem_caches(void)
1830 {
1831         if (ocfs2_inode_cachep)
1832                 kmem_cache_destroy(ocfs2_inode_cachep);
1833         ocfs2_inode_cachep = NULL;
1834
1835         if (ocfs2_dquot_cachep)
1836                 kmem_cache_destroy(ocfs2_dquot_cachep);
1837         ocfs2_dquot_cachep = NULL;
1838
1839         if (ocfs2_qf_chunk_cachep)
1840                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1841         ocfs2_qf_chunk_cachep = NULL;
1842 }
1843
1844 static int ocfs2_get_sector(struct super_block *sb,
1845                             struct buffer_head **bh,
1846                             int block,
1847                             int sect_size)
1848 {
1849         if (!sb_set_blocksize(sb, sect_size)) {
1850                 mlog(ML_ERROR, "unable to set blocksize\n");
1851                 return -EIO;
1852         }
1853
1854         *bh = sb_getblk(sb, block);
1855         if (!*bh) {
1856                 mlog_errno(-EIO);
1857                 return -EIO;
1858         }
1859         lock_buffer(*bh);
1860         if (!buffer_dirty(*bh))
1861                 clear_buffer_uptodate(*bh);
1862         unlock_buffer(*bh);
1863         ll_rw_block(READ, 1, bh);
1864         wait_on_buffer(*bh);
1865         if (!buffer_uptodate(*bh)) {
1866                 mlog_errno(-EIO);
1867                 brelse(*bh);
1868                 *bh = NULL;
1869                 return -EIO;
1870         }
1871
1872         return 0;
1873 }
1874
1875 static int ocfs2_mount_volume(struct super_block *sb)
1876 {
1877         int status = 0;
1878         int unlock_super = 0;
1879         struct ocfs2_super *osb = OCFS2_SB(sb);
1880
1881         mlog_entry_void();
1882
1883         if (ocfs2_is_hard_readonly(osb))
1884                 goto leave;
1885
1886         status = ocfs2_dlm_init(osb);
1887         if (status < 0) {
1888                 mlog_errno(status);
1889                 goto leave;
1890         }
1891
1892         status = ocfs2_super_lock(osb, 1);
1893         if (status < 0) {
1894                 mlog_errno(status);
1895                 goto leave;
1896         }
1897         unlock_super = 1;
1898
1899         /* This will load up the node map and add ourselves to it. */
1900         status = ocfs2_find_slot(osb);
1901         if (status < 0) {
1902                 mlog_errno(status);
1903                 goto leave;
1904         }
1905
1906         /* load all node-local system inodes */
1907         status = ocfs2_init_local_system_inodes(osb);
1908         if (status < 0) {
1909                 mlog_errno(status);
1910                 goto leave;
1911         }
1912
1913         status = ocfs2_check_volume(osb);
1914         if (status < 0) {
1915                 mlog_errno(status);
1916                 goto leave;
1917         }
1918
1919         status = ocfs2_truncate_log_init(osb);
1920         if (status < 0)
1921                 mlog_errno(status);
1922
1923 leave:
1924         if (unlock_super)
1925                 ocfs2_super_unlock(osb, 1);
1926
1927         mlog_exit(status);
1928         return status;
1929 }
1930
1931 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1932 {
1933         int tmp, hangup_needed = 0;
1934         struct ocfs2_super *osb = NULL;
1935         char nodestr[8];
1936
1937         mlog_entry("(0x%p)\n", sb);
1938
1939         BUG_ON(!sb);
1940         osb = OCFS2_SB(sb);
1941         BUG_ON(!osb);
1942
1943         debugfs_remove(osb->osb_ctxt);
1944
1945         /*
1946          * Flush inode dropping work queue so that deletes are
1947          * performed while the filesystem is still working
1948          */
1949         ocfs2_drop_all_dl_inodes(osb);
1950
1951         /* Orphan scan should be stopped as early as possible */
1952         ocfs2_orphan_scan_stop(osb);
1953
1954         ocfs2_disable_quotas(osb);
1955
1956         ocfs2_shutdown_local_alloc(osb);
1957
1958         ocfs2_truncate_log_shutdown(osb);
1959
1960         /* This will disable recovery and flush any recovery work. */
1961         ocfs2_recovery_exit(osb);
1962
1963         ocfs2_journal_shutdown(osb);
1964
1965         ocfs2_sync_blockdev(sb);
1966
1967         ocfs2_purge_refcount_trees(osb);
1968
1969         /* No cluster connection means we've failed during mount, so skip
1970          * all the steps which depended on that to complete. */
1971         if (osb->cconn) {
1972                 tmp = ocfs2_super_lock(osb, 1);
1973                 if (tmp < 0) {
1974                         mlog_errno(tmp);
1975                         return;
1976                 }
1977         }
1978
1979         if (osb->slot_num != OCFS2_INVALID_SLOT)
1980                 ocfs2_put_slot(osb);
1981
1982         if (osb->cconn)
1983                 ocfs2_super_unlock(osb, 1);
1984
1985         ocfs2_release_system_inodes(osb);
1986
1987         /*
1988          * If we're dismounting due to mount error, mount.ocfs2 will clean
1989          * up heartbeat.  If we're a local mount, there is no heartbeat.
1990          * If we failed before we got a uuid_str yet, we can't stop
1991          * heartbeat.  Otherwise, do it.
1992          */
1993         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1994                 hangup_needed = 1;
1995
1996         if (osb->cconn)
1997                 ocfs2_dlm_shutdown(osb, hangup_needed);
1998
1999         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
2000         debugfs_remove(osb->osb_debug_root);
2001
2002         if (hangup_needed)
2003                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
2004
2005         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2006
2007         if (ocfs2_mount_local(osb))
2008                 snprintf(nodestr, sizeof(nodestr), "local");
2009         else
2010                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2011
2012         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2013                osb->dev_str, nodestr);
2014
2015         ocfs2_delete_osb(osb);
2016         kfree(osb);
2017         sb->s_dev = 0;
2018         sb->s_fs_info = NULL;
2019 }
2020
2021 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2022                                 unsigned uuid_bytes)
2023 {
2024         int i, ret;
2025         char *ptr;
2026
2027         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2028
2029         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2030         if (osb->uuid_str == NULL)
2031                 return -ENOMEM;
2032
2033         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2034                 /* print with null */
2035                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2036                 if (ret != 2) /* drop super cleans up */
2037                         return -EINVAL;
2038                 /* then only advance past the last char */
2039                 ptr += 2;
2040         }
2041
2042         return 0;
2043 }
2044
2045 /* Make sure entire volume is addressable by our journal.  Requires
2046    osb_clusters_at_boot to be valid and for the journal to have been
2047    initialized by ocfs2_journal_init(). */
2048 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2049 {
2050         int status = 0;
2051         u64 max_block =
2052                 ocfs2_clusters_to_blocks(osb->sb,
2053                                          osb->osb_clusters_at_boot) - 1;
2054
2055         /* 32-bit block number is always OK. */
2056         if (max_block <= (u32)~0ULL)
2057                 goto out;
2058
2059         /* Volume is "huge", so see if our journal is new enough to
2060            support it. */
2061         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2062                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2063               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2064                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2065                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2066                      "Enable the 'block64' journal option with tunefs.ocfs2");
2067                 status = -EFBIG;
2068                 goto out;
2069         }
2070
2071  out:
2072         return status;
2073 }
2074
2075 static int ocfs2_initialize_super(struct super_block *sb,
2076                                   struct buffer_head *bh,
2077                                   int sector_size,
2078                                   struct ocfs2_blockcheck_stats *stats)
2079 {
2080         int status;
2081         int i, cbits, bbits;
2082         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2083         struct inode *inode = NULL;
2084         struct ocfs2_journal *journal;
2085         __le32 uuid_net_key;
2086         struct ocfs2_super *osb;
2087         u64 total_blocks;
2088
2089         mlog_entry_void();
2090
2091         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2092         if (!osb) {
2093                 status = -ENOMEM;
2094                 mlog_errno(status);
2095                 goto bail;
2096         }
2097
2098         sb->s_fs_info = osb;
2099         sb->s_op = &ocfs2_sops;
2100         sb->s_d_op = &ocfs2_dentry_ops;
2101         sb->s_export_op = &ocfs2_export_ops;
2102         sb->s_qcop = &ocfs2_quotactl_ops;
2103         sb->dq_op = &ocfs2_quota_operations;
2104         sb->s_xattr = ocfs2_xattr_handlers;
2105         sb->s_time_gran = 1;
2106         sb->s_flags |= MS_NOATIME;
2107         /* this is needed to support O_LARGEFILE */
2108         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2109         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2110         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2111
2112         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2113
2114         for (i = 0; i < 3; i++)
2115                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2116         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2117
2118         osb->sb = sb;
2119         /* Save off for ocfs2_rw_direct */
2120         osb->s_sectsize_bits = blksize_bits(sector_size);
2121         BUG_ON(!osb->s_sectsize_bits);
2122
2123         spin_lock_init(&osb->dc_task_lock);
2124         init_waitqueue_head(&osb->dc_event);
2125         osb->dc_work_sequence = 0;
2126         osb->dc_wake_sequence = 0;
2127         INIT_LIST_HEAD(&osb->blocked_lock_list);
2128         osb->blocked_lock_count = 0;
2129         spin_lock_init(&osb->osb_lock);
2130         spin_lock_init(&osb->osb_xattr_lock);
2131         ocfs2_init_steal_slots(osb);
2132
2133         atomic_set(&osb->alloc_stats.moves, 0);
2134         atomic_set(&osb->alloc_stats.local_data, 0);
2135         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2136         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2137         atomic_set(&osb->alloc_stats.bg_extends, 0);
2138
2139         /* Copy the blockcheck stats from the superblock probe */
2140         osb->osb_ecc_stats = *stats;
2141
2142         ocfs2_init_node_maps(osb);
2143
2144         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2145                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2146
2147         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2148         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2149                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2150                      osb->max_slots);
2151                 status = -EINVAL;
2152                 goto bail;
2153         }
2154         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2155
2156         ocfs2_orphan_scan_init(osb);
2157
2158         status = ocfs2_recovery_init(osb);
2159         if (status) {
2160                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2161                 mlog_errno(status);
2162                 goto bail;
2163         }
2164
2165         init_waitqueue_head(&osb->checkpoint_event);
2166         atomic_set(&osb->needs_checkpoint, 0);
2167
2168         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2169
2170         osb->slot_num = OCFS2_INVALID_SLOT;
2171
2172         osb->s_xattr_inline_size = le16_to_cpu(
2173                                         di->id2.i_super.s_xattr_inline_size);
2174
2175         osb->local_alloc_state = OCFS2_LA_UNUSED;
2176         osb->local_alloc_bh = NULL;
2177         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2178
2179         init_waitqueue_head(&osb->osb_mount_event);
2180
2181         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2182         if (status) {
2183                 mlog_errno(status);
2184                 goto bail;
2185         }
2186
2187         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2188         if (!osb->vol_label) {
2189                 mlog(ML_ERROR, "unable to alloc vol label\n");
2190                 status = -ENOMEM;
2191                 goto bail;
2192         }
2193
2194         osb->slot_recovery_generations =
2195                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2196                         GFP_KERNEL);
2197         if (!osb->slot_recovery_generations) {
2198                 status = -ENOMEM;
2199                 mlog_errno(status);
2200                 goto bail;
2201         }
2202
2203         init_waitqueue_head(&osb->osb_wipe_event);
2204         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2205                                         sizeof(*osb->osb_orphan_wipes),
2206                                         GFP_KERNEL);
2207         if (!osb->osb_orphan_wipes) {
2208                 status = -ENOMEM;
2209                 mlog_errno(status);
2210                 goto bail;
2211         }
2212
2213         osb->osb_rf_lock_tree = RB_ROOT;
2214
2215         osb->s_feature_compat =
2216                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2217         osb->s_feature_ro_compat =
2218                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2219         osb->s_feature_incompat =
2220                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2221
2222         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2223                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2224                      "optional features (%x).\n", i);
2225                 status = -EINVAL;
2226                 goto bail;
2227         }
2228         if (!(osb->sb->s_flags & MS_RDONLY) &&
2229             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2230                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2231                      "unsupported optional features (%x).\n", i);
2232                 status = -EINVAL;
2233                 goto bail;
2234         }
2235
2236         if (ocfs2_clusterinfo_valid(osb)) {
2237                 osb->osb_stackflags =
2238                         OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2239                 memcpy(osb->osb_cluster_stack,
2240                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2241                        OCFS2_STACK_LABEL_LEN);
2242                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2243                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2244                         mlog(ML_ERROR,
2245                              "couldn't mount because of an invalid "
2246                              "cluster stack label (%s) \n",
2247                              osb->osb_cluster_stack);
2248                         status = -EINVAL;
2249                         goto bail;
2250                 }
2251         } else {
2252                 /* The empty string is identical with classic tools that
2253                  * don't know about s_cluster_info. */
2254                 osb->osb_cluster_stack[0] = '\0';
2255         }
2256
2257         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2258
2259         /* FIXME
2260          * This should be done in ocfs2_journal_init(), but unknown
2261          * ordering issues will cause the filesystem to crash.
2262          * If anyone wants to figure out what part of the code
2263          * refers to osb->journal before ocfs2_journal_init() is run,
2264          * be my guest.
2265          */
2266         /* initialize our journal structure */
2267
2268         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2269         if (!journal) {
2270                 mlog(ML_ERROR, "unable to alloc journal\n");
2271                 status = -ENOMEM;
2272                 goto bail;
2273         }
2274         osb->journal = journal;
2275         journal->j_osb = osb;
2276
2277         atomic_set(&journal->j_num_trans, 0);
2278         init_rwsem(&journal->j_trans_barrier);
2279         init_waitqueue_head(&journal->j_checkpointed);
2280         spin_lock_init(&journal->j_lock);
2281         journal->j_trans_id = (unsigned long) 1;
2282         INIT_LIST_HEAD(&journal->j_la_cleanups);
2283         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2284         journal->j_state = OCFS2_JOURNAL_FREE;
2285
2286         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2287         osb->dentry_lock_list = NULL;
2288
2289         /* get some pseudo constants for clustersize bits */
2290         osb->s_clustersize_bits =
2291                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2292         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2293         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2294
2295         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2296             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2297                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2298                      osb->s_clustersize);
2299                 status = -EINVAL;
2300                 goto bail;
2301         }
2302
2303         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2304                                                 le32_to_cpu(di->i_clusters));
2305
2306         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2307                                            total_blocks);
2308         if (status) {
2309                 mlog(ML_ERROR, "Volume too large "
2310                      "to mount safely on this system");
2311                 status = -EFBIG;
2312                 goto bail;
2313         }
2314
2315         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2316                                  sizeof(di->id2.i_super.s_uuid))) {
2317                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2318                 status = -ENOMEM;
2319                 goto bail;
2320         }
2321
2322         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2323
2324         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2325         osb->vol_label[63] = '\0';
2326         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2327         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2328         osb->first_cluster_group_blkno =
2329                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2330         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2331         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2332         mlog(0, "vol_label: %s\n", osb->vol_label);
2333         mlog(0, "uuid: %s\n", osb->uuid_str);
2334         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2335              (unsigned long long)osb->root_blkno,
2336              (unsigned long long)osb->system_dir_blkno);
2337
2338         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2339         if (!osb->osb_dlm_debug) {
2340                 status = -ENOMEM;
2341                 mlog_errno(status);
2342                 goto bail;
2343         }
2344
2345         atomic_set(&osb->vol_state, VOLUME_INIT);
2346
2347         /* load root, system_dir, and all global system inodes */
2348         status = ocfs2_init_global_system_inodes(osb);
2349         if (status < 0) {
2350                 mlog_errno(status);
2351                 goto bail;
2352         }
2353
2354         /*
2355          * global bitmap
2356          */
2357         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2358                                             OCFS2_INVALID_SLOT);
2359         if (!inode) {
2360                 status = -EINVAL;
2361                 mlog_errno(status);
2362                 goto bail;
2363         }
2364
2365         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2366         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2367         iput(inode);
2368
2369         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2370                                  osb->s_feature_incompat) * 8;
2371
2372         status = ocfs2_init_slot_info(osb);
2373         if (status < 0) {
2374                 mlog_errno(status);
2375                 goto bail;
2376         }
2377
2378 bail:
2379         mlog_exit(status);
2380         return status;
2381 }
2382
2383 /*
2384  * will return: -EAGAIN if it is ok to keep searching for superblocks
2385  *              -EINVAL if there is a bad superblock
2386  *              0 on success
2387  */
2388 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2389                                struct buffer_head *bh,
2390                                u32 blksz,
2391                                struct ocfs2_blockcheck_stats *stats)
2392 {
2393         int status = -EAGAIN;
2394
2395         mlog_entry_void();
2396
2397         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2398                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2399                 /* We have to do a raw check of the feature here */
2400                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2401                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2402                         status = ocfs2_block_check_validate(bh->b_data,
2403                                                             bh->b_size,
2404                                                             &di->i_check,
2405                                                             stats);
2406                         if (status)
2407                                 goto out;
2408                 }
2409                 status = -EINVAL;
2410                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2411                         mlog(ML_ERROR, "found superblock with incorrect block "
2412                              "size: found %u, should be %u\n",
2413                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2414                                blksz);
2415                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2416                            OCFS2_MAJOR_REV_LEVEL ||
2417                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2418                            OCFS2_MINOR_REV_LEVEL) {
2419                         mlog(ML_ERROR, "found superblock with bad version: "
2420                              "found %u.%u, should be %u.%u\n",
2421                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2422                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2423                              OCFS2_MAJOR_REV_LEVEL,
2424                              OCFS2_MINOR_REV_LEVEL);
2425                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2426                         mlog(ML_ERROR, "bad block number on superblock: "
2427                              "found %llu, should be %llu\n",
2428                              (unsigned long long)le64_to_cpu(di->i_blkno),
2429                              (unsigned long long)bh->b_blocknr);
2430                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2431                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2432                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2433                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2434                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2435                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2436                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2437                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2438                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2439                         mlog(ML_ERROR,
2440                              "Superblock slots found greater than file system "
2441                              "maximum: found %u, max %u\n",
2442                              le16_to_cpu(di->id2.i_super.s_max_slots),
2443                              OCFS2_MAX_SLOTS);
2444                 } else {
2445                         /* found it! */
2446                         status = 0;
2447                 }
2448         }
2449
2450 out:
2451         mlog_exit(status);
2452         return status;
2453 }
2454
2455 static int ocfs2_check_volume(struct ocfs2_super *osb)
2456 {
2457         int status;
2458         int dirty;
2459         int local;
2460         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2461                                                   * recover
2462                                                   * ourselves. */
2463
2464         mlog_entry_void();
2465
2466         /* Init our journal object. */
2467         status = ocfs2_journal_init(osb->journal, &dirty);
2468         if (status < 0) {
2469                 mlog(ML_ERROR, "Could not initialize journal!\n");
2470                 goto finally;
2471         }
2472
2473         /* Now that journal has been initialized, check to make sure
2474            entire volume is addressable. */
2475         status = ocfs2_journal_addressable(osb);
2476         if (status)
2477                 goto finally;
2478
2479         /* If the journal was unmounted cleanly then we don't want to
2480          * recover anything. Otherwise, journal_load will do that
2481          * dirty work for us :) */
2482         if (!dirty) {
2483                 status = ocfs2_journal_wipe(osb->journal, 0);
2484                 if (status < 0) {
2485                         mlog_errno(status);
2486                         goto finally;
2487                 }
2488         } else {
2489                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2490                      "recovering volume.\n");
2491         }
2492
2493         local = ocfs2_mount_local(osb);
2494
2495         /* will play back anything left in the journal. */
2496         status = ocfs2_journal_load(osb->journal, local, dirty);
2497         if (status < 0) {
2498                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2499                 goto finally;
2500         }
2501
2502         if (dirty) {
2503                 /* recover my local alloc if we didn't unmount cleanly. */
2504                 status = ocfs2_begin_local_alloc_recovery(osb,
2505                                                           osb->slot_num,
2506                                                           &local_alloc);
2507                 if (status < 0) {
2508                         mlog_errno(status);
2509                         goto finally;
2510                 }
2511                 /* we complete the recovery process after we've marked
2512                  * ourselves as mounted. */
2513         }
2514
2515         mlog(0, "Journal loaded.\n");
2516
2517         status = ocfs2_load_local_alloc(osb);
2518         if (status < 0) {
2519                 mlog_errno(status);
2520                 goto finally;
2521         }
2522
2523         if (dirty) {
2524                 /* Recovery will be completed after we've mounted the
2525                  * rest of the volume. */
2526                 osb->dirty = 1;
2527                 osb->local_alloc_copy = local_alloc;
2528                 local_alloc = NULL;
2529         }
2530
2531         /* go through each journal, trylock it and if you get the
2532          * lock, and it's marked as dirty, set the bit in the recover
2533          * map and launch a recovery thread for it. */
2534         status = ocfs2_mark_dead_nodes(osb);
2535         if (status < 0) {
2536                 mlog_errno(status);
2537                 goto finally;
2538         }
2539
2540         status = ocfs2_compute_replay_slots(osb);
2541         if (status < 0)
2542                 mlog_errno(status);
2543
2544 finally:
2545         if (local_alloc)
2546                 kfree(local_alloc);
2547
2548         mlog_exit(status);
2549         return status;
2550 }
2551
2552 /*
2553  * The routine gets called from dismount or close whenever a dismount on
2554  * volume is requested and the osb open count becomes 1.
2555  * It will remove the osb from the global list and also free up all the
2556  * initialized resources and fileobject.
2557  */
2558 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2559 {
2560         mlog_entry_void();
2561
2562         /* This function assumes that the caller has the main osb resource */
2563
2564         ocfs2_free_slot_info(osb);
2565
2566         kfree(osb->osb_orphan_wipes);
2567         kfree(osb->slot_recovery_generations);
2568         /* FIXME
2569          * This belongs in journal shutdown, but because we have to
2570          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2571          * we free it here.
2572          */
2573         kfree(osb->journal);
2574         if (osb->local_alloc_copy)
2575                 kfree(osb->local_alloc_copy);
2576         kfree(osb->uuid_str);
2577         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2578         memset(osb, 0, sizeof(struct ocfs2_super));
2579
2580         mlog_exit_void();
2581 }
2582
2583 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2584  * panic(). We do not support continue-on-error operation. */
2585 static void ocfs2_handle_error(struct super_block *sb)
2586 {
2587         struct ocfs2_super *osb = OCFS2_SB(sb);
2588
2589         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2590                 panic("OCFS2: (device %s): panic forced after error\n",
2591                       sb->s_id);
2592
2593         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2594
2595         if (sb->s_flags & MS_RDONLY &&
2596             (ocfs2_is_soft_readonly(osb) ||
2597              ocfs2_is_hard_readonly(osb)))
2598                 return;
2599
2600         printk(KERN_CRIT "File system is now read-only due to the potential "
2601                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2602                "system is unmounted.\n");
2603         sb->s_flags |= MS_RDONLY;
2604         ocfs2_set_ro_flag(osb, 0);
2605 }
2606
2607 static char error_buf[1024];
2608
2609 void __ocfs2_error(struct super_block *sb,
2610                    const char *function,
2611                    const char *fmt, ...)
2612 {
2613         va_list args;
2614
2615         va_start(args, fmt);
2616         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2617         va_end(args);
2618
2619         /* Not using mlog here because we want to show the actual
2620          * function the error came from. */
2621         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2622                sb->s_id, function, error_buf);
2623
2624         ocfs2_handle_error(sb);
2625 }
2626
2627 /* Handle critical errors. This is intentionally more drastic than
2628  * ocfs2_handle_error, so we only use for things like journal errors,
2629  * etc. */
2630 void __ocfs2_abort(struct super_block* sb,
2631                    const char *function,
2632                    const char *fmt, ...)
2633 {
2634         va_list args;
2635
2636         va_start(args, fmt);
2637         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2638         va_end(args);
2639
2640         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2641                sb->s_id, function, error_buf);
2642
2643         /* We don't have the cluster support yet to go straight to
2644          * hard readonly in here. Until then, we want to keep
2645          * ocfs2_abort() so that we can at least mark critical
2646          * errors.
2647          *
2648          * TODO: This should abort the journal and alert other nodes
2649          * that our slot needs recovery. */
2650
2651         /* Force a panic(). This stinks, but it's better than letting
2652          * things continue without having a proper hard readonly
2653          * here. */
2654         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2655                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2656         ocfs2_handle_error(sb);
2657 }
2658
2659 /*
2660  * Void signal blockers, because in-kernel sigprocmask() only fails
2661  * when SIG_* is wrong.
2662  */
2663 void ocfs2_block_signals(sigset_t *oldset)
2664 {
2665         int rc;
2666         sigset_t blocked;
2667
2668         sigfillset(&blocked);
2669         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2670         BUG_ON(rc);
2671 }
2672
2673 void ocfs2_unblock_signals(sigset_t *oldset)
2674 {
2675         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2676         BUG_ON(rc);
2677 }
2678
2679 module_init(ocfs2_init);
2680 module_exit(ocfs2_exit);