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 #include <linux/smp_lock.h>
  45
  46 #define MLOG_MASK_PREFIX ML_SUPER
  47 #include <cluster/masklog.h>
  48
  49 #include "ocfs2.h"
  50
  51 /* this should be the only file to include a version 1 header */
  52 #include "ocfs1_fs_compat.h"
  53
  54 #include "alloc.h"
  55 #include "blockcheck.h"
  56 #include "dlmglue.h"
  57 #include "export.h"
  58 #include "extent_map.h"
  59 #include "heartbeat.h"
  60 #include "inode.h"
  61 #include "journal.h"
  62 #include "localalloc.h"
  63 #include "namei.h"
  64 #include "slot_map.h"
  65 #include "super.h"
  66 #include "sysfile.h"
  67 #include "uptodate.h"
  68 #include "ver.h"
  69 #include "xattr.h"
  70 #include "quota.h"
  71 #include "refcounttree.h"
  72 #include "suballoc.h"
  73
  74 #include "buffer_head_io.h"
  75
  76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
  77 struct kmem_cache *ocfs2_dquot_cachep;
  78 struct kmem_cache *ocfs2_qf_chunk_cachep;
  79
  80 /* OCFS2 needs to schedule several differnt types of work which
  81  * require cluster locking, disk I/O, recovery waits, etc. Since these
  82  * types of work tend to be heavy we avoid using the kernel events
  83  * workqueue and schedule on our own. */
  84 struct workqueue_struct *ocfs2_wq = NULL;
  85
  86 static struct dentry *ocfs2_debugfs_root = NULL;
  87
  88 MODULE_AUTHOR("Oracle");
  89 MODULE_LICENSE("GPL");
  90
  91 struct mount_options
  92 {
  93         unsigned long   commit_interval;
  94         unsigned long   mount_opt;
  95         unsigned int    atime_quantum;
  96         signed short    slot;
  97         int             localalloc_opt;
  98         unsigned int    resv_level;
  99         int             dir_resv_level;
 100         char            cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
 101 };
 102
 103 static int ocfs2_parse_options(struct super_block *sb, char *options,
 104                                struct mount_options *mopt,
 105                                int is_remount);
 106 static int ocfs2_check_set_options(struct super_block *sb,
 107                                    struct mount_options *options);
 108 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
 109 static void ocfs2_put_super(struct super_block *sb);
 110 static int ocfs2_mount_volume(struct super_block *sb);
 111 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
 112 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
 113 static int ocfs2_initialize_mem_caches(void);
 114 static void ocfs2_free_mem_caches(void);
 115 static void ocfs2_delete_osb(struct ocfs2_super *osb);
 116
 117 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
 118
 119 static int ocfs2_sync_fs(struct super_block *sb, int wait);
 120
 121 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
 122 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
 123 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
 124 static int ocfs2_check_volume(struct ocfs2_super *osb);
 125 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
 126                                struct buffer_head *bh,
 127                                u32 sectsize,
 128                                struct ocfs2_blockcheck_stats *stats);
 129 static int ocfs2_initialize_super(struct super_block *sb,
 130                                   struct buffer_head *bh,
 131                                   int sector_size,
 132                                   struct ocfs2_blockcheck_stats *stats);
 133 static int ocfs2_get_sector(struct super_block *sb,
 134                             struct buffer_head **bh,
 135                             int block,
 136                             int sect_size);
 137 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
 138 static void ocfs2_destroy_inode(struct inode *inode);
 139 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
 140 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
 141 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
 142
 143 static const struct super_operations ocfs2_sops = {
 144         .statfs         = ocfs2_statfs,
 145         .alloc_inode    = ocfs2_alloc_inode,
 146         .destroy_inode  = ocfs2_destroy_inode,
 147         .drop_inode     = ocfs2_drop_inode,
 148         .evict_inode    = ocfs2_evict_inode,
 149         .sync_fs        = ocfs2_sync_fs,
 150         .put_super      = ocfs2_put_super,
 151         .remount_fs     = ocfs2_remount,
 152         .show_options   = ocfs2_show_options,
 153         .quota_read     = ocfs2_quota_read,
 154         .quota_write    = ocfs2_quota_write,
 155 };
 156
 157 enum {
 158         Opt_barrier,
 159         Opt_err_panic,
 160         Opt_err_ro,
 161         Opt_intr,
 162         Opt_nointr,
 163         Opt_hb_none,
 164         Opt_hb_local,
 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_resv_level,
 181         Opt_dir_resv_level,
 182         Opt_err,
 183 };
 184
 185 static const match_table_t tokens = {
 186         {Opt_barrier, "barrier=%u"},
 187         {Opt_err_panic, "errors=panic"},
 188         {Opt_err_ro, "errors=remount-ro"},
 189         {Opt_intr, "intr"},
 190         {Opt_nointr, "nointr"},
 191         {Opt_hb_none, OCFS2_HB_NONE},
 192         {Opt_hb_local, OCFS2_HB_LOCAL},
 193         {Opt_data_ordered, "data=ordered"},
 194         {Opt_data_writeback, "data=writeback"},
 195         {Opt_atime_quantum, "atime_quantum=%u"},
 196         {Opt_slot, "preferred_slot=%u"},
 197         {Opt_commit, "commit=%u"},
 198         {Opt_localalloc, "localalloc=%d"},
 199         {Opt_localflocks, "localflocks"},
 200         {Opt_stack, "cluster_stack=%s"},
 201         {Opt_user_xattr, "user_xattr"},
 202         {Opt_nouser_xattr, "nouser_xattr"},
 203         {Opt_inode64, "inode64"},
 204         {Opt_acl, "acl"},
 205         {Opt_noacl, "noacl"},
 206         {Opt_usrquota, "usrquota"},
 207         {Opt_grpquota, "grpquota"},
 208         {Opt_resv_level, "resv_level=%u"},
 209         {Opt_dir_resv_level, "dir_resv_level=%u"},
 210         {Opt_err, NULL}
 211 };
 212
 213 #ifdef CONFIG_DEBUG_FS
 214 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 215 {
 216         struct ocfs2_cluster_connection *cconn = osb->cconn;
 217         struct ocfs2_recovery_map *rm = osb->recovery_map;
 218         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 219         int i, out = 0;
 220
 221         out += snprintf(buf + out, len - out,
 222                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 223                         "Device", osb->dev_str, osb->uuid_str,
 224                         osb->fs_generation, osb->vol_label);
 225
 226         out += snprintf(buf + out, len - out,
 227                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 228                         atomic_read(&osb->vol_state), osb->osb_flags);
 229
 230         out += snprintf(buf + out, len - out,
 231                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 232                         osb->sb->s_blocksize, osb->s_clustersize);
 233
 234         out += snprintf(buf + out, len - out,
 235                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 236                         "ROcompat: 0x%X\n",
 237                         "Features", osb->s_feature_compat,
 238                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 239
 240         out += snprintf(buf + out, len - out,
 241                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 242                         osb->s_mount_opt, osb->s_atime_quantum);
 243
 244         if (cconn) {
 245                 out += snprintf(buf + out, len - out,
 246                                 "%10s => Stack: %s  Name: %*s  "
 247                                 "Version: %d.%d\n", "Cluster",
 248                                 (*osb->osb_cluster_stack == '\0' ?
 249                                  "o2cb" : osb->osb_cluster_stack),
 250                                 cconn->cc_namelen, cconn->cc_name,
 251                                 cconn->cc_version.pv_major,
 252                                 cconn->cc_version.pv_minor);
 253         }
 254
 255         spin_lock(&osb->dc_task_lock);
 256         out += snprintf(buf + out, len - out,
 257                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 258                         "WorkSeq: %lu\n", "DownCnvt",
 259                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 260                         osb->blocked_lock_count, osb->dc_wake_sequence,
 261                         osb->dc_work_sequence);
 262         spin_unlock(&osb->dc_task_lock);
 263
 264         spin_lock(&osb->osb_lock);
 265         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 266                         "Recovery",
 267                         (osb->recovery_thread_task ?
 268                          task_pid_nr(osb->recovery_thread_task) : -1));
 269         if (rm->rm_used == 0)
 270                 out += snprintf(buf + out, len - out, " None\n");
 271         else {
 272                 for (i = 0; i < rm->rm_used; i++)
 273                         out += snprintf(buf + out, len - out, " %d",
 274                                         rm->rm_entries[i]);
 275                 out += snprintf(buf + out, len - out, "\n");
 276         }
 277         spin_unlock(&osb->osb_lock);
 278
 279         out += snprintf(buf + out, len - out,
 280                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 281                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 282                         osb->osb_commit_interval,
 283                         atomic_read(&osb->needs_checkpoint));
 284
 285         out += snprintf(buf + out, len - out,
 286                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 287                         "Journal", osb->journal->j_state,
 288                         osb->journal->j_trans_id,
 289                         atomic_read(&osb->journal->j_num_trans));
 290
 291         out += snprintf(buf + out, len - out,
 292                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 293                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 294                         "Stats",
 295                         atomic_read(&osb->alloc_stats.bitmap_data),
 296                         atomic_read(&osb->alloc_stats.local_data),
 297                         atomic_read(&osb->alloc_stats.bg_allocs),
 298                         atomic_read(&osb->alloc_stats.moves),
 299                         atomic_read(&osb->alloc_stats.bg_extends));
 300
 301         out += snprintf(buf + out, len - out,
 302                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 303                         "Default: %u bits\n",
 304                         "LocalAlloc", osb->local_alloc_state,
 305                         (unsigned long long)osb->la_last_gd,
 306                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 307
 308         spin_lock(&osb->osb_lock);
 309         out += snprintf(buf + out, len - out,
 310                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 311                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 312                         osb->s_inode_steal_slot,
 313                         atomic_read(&osb->s_num_inodes_stolen),
 314                         osb->s_meta_steal_slot,
 315                         atomic_read(&osb->s_num_meta_stolen));
 316         spin_unlock(&osb->osb_lock);
 317
 318         out += snprintf(buf + out, len - out, "OrphanScan => ");
 319         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 320                         os->os_count, os->os_seqno);
 321         out += snprintf(buf + out, len - out, " Last Scan: ");
 322         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 323                 out += snprintf(buf + out, len - out, "Disabled\n");
 324         else
 325                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 326                                 (get_seconds() - os->os_scantime.tv_sec));
 327
 328         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 329                         "Slots", "Num", "RecoGen");
 330         for (i = 0; i < osb->max_slots; ++i) {
 331                 out += snprintf(buf + out, len - out,
 332                                 "%10s  %c %3d  %10d\n",
 333                                 " ",
 334                                 (i == osb->slot_num ? '*' : ' '),
 335                                 i, osb->slot_recovery_generations[i]);
 336         }
 337
 338         return out;
 339 }
 340
 341 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 342 {
 343         struct ocfs2_super *osb = inode->i_private;
 344         char *buf = NULL;
 345
 346         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 347         if (!buf)
 348                 goto bail;
 349
 350         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 351
 352         file->private_data = buf;
 353
 354         return 0;
 355 bail:
 356         return -ENOMEM;
 357 }
 358
 359 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 360 {
 361         kfree(file->private_data);
 362         return 0;
 363 }
 364
 365 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 366                                 size_t nbytes, loff_t *ppos)
 367 {
 368         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 369                                        i_size_read(file->f_mapping->host));
 370 }
 371 #else
 372 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 373 {
 374         return 0;
 375 }
 376 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 377 {
 378         return 0;
 379 }
 380 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 381                                 size_t nbytes, loff_t *ppos)
 382 {
 383         return 0;
 384 }
 385 #endif  /* CONFIG_DEBUG_FS */
 386
 387 static const struct file_operations ocfs2_osb_debug_fops = {
 388         .open =         ocfs2_osb_debug_open,
 389         .release =      ocfs2_debug_release,
 390         .read =         ocfs2_debug_read,
 391         .llseek =       generic_file_llseek,
 392 };
 393
 394 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 395 {
 396         int status;
 397         tid_t target;
 398         struct ocfs2_super *osb = OCFS2_SB(sb);
 399
 400         if (ocfs2_is_hard_readonly(osb))
 401                 return -EROFS;
 402
 403         if (wait) {
 404                 status = ocfs2_flush_truncate_log(osb);
 405                 if (status < 0)
 406                         mlog_errno(status);
 407         } else {
 408                 ocfs2_schedule_truncate_log_flush(osb, 0);
 409         }
 410
 411         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 412                                       &target)) {
 413                 if (wait)
 414                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 415                                              target);
 416         }
 417         return 0;
 418 }
 419
 420 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 421 {
 422         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 423             && (ino == USER_QUOTA_SYSTEM_INODE
 424                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 425                 return 0;
 426         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 427             && (ino == GROUP_QUOTA_SYSTEM_INODE
 428                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 429                 return 0;
 430         return 1;
 431 }
 432
 433 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 434 {
 435         struct inode *new = NULL;
 436         int status = 0;
 437         int i;
 438
 439         mlog_entry_void();
 440
 441         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 442         if (IS_ERR(new)) {
 443                 status = PTR_ERR(new);
 444                 mlog_errno(status);
 445                 goto bail;
 446         }
 447         osb->root_inode = new;
 448
 449         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 450         if (IS_ERR(new)) {
 451                 status = PTR_ERR(new);
 452                 mlog_errno(status);
 453                 goto bail;
 454         }
 455         osb->sys_root_inode = new;
 456
 457         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 458              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 459                 if (!ocfs2_need_system_inode(osb, i))
 460                         continue;
 461                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 462                 if (!new) {
 463                         ocfs2_release_system_inodes(osb);
 464                         status = -EINVAL;
 465                         mlog_errno(status);
 466                         /* FIXME: Should ERROR_RO_FS */
 467                         mlog(ML_ERROR, "Unable to load system inode %d, "
 468                              "possibly corrupt fs?", i);
 469                         goto bail;
 470                 }
 471                 // the array now has one ref, so drop this one
 472                 iput(new);
 473         }
 474
 475 bail:
 476         mlog_exit(status);
 477         return status;
 478 }
 479
 480 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 481 {
 482         struct inode *new = NULL;
 483         int status = 0;
 484         int i;
 485
 486         mlog_entry_void();
 487
 488         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 489              i < NUM_SYSTEM_INODES;
 490              i++) {
 491                 if (!ocfs2_need_system_inode(osb, i))
 492                         continue;
 493                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 494                 if (!new) {
 495                         ocfs2_release_system_inodes(osb);
 496                         status = -EINVAL;
 497                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 498                              status, i, osb->slot_num);
 499                         goto bail;
 500                 }
 501                 /* the array now has one ref, so drop this one */
 502                 iput(new);
 503         }
 504
 505 bail:
 506         mlog_exit(status);
 507         return status;
 508 }
 509
 510 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 511 {
 512         int i;
 513         struct inode *inode;
 514
 515         mlog_entry_void();
 516
 517         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
 518                 inode = osb->system_inodes[i];
 519                 if (inode) {
 520                         iput(inode);
 521                         osb->system_inodes[i] = NULL;
 522                 }
 523         }
 524
 525         inode = osb->sys_root_inode;
 526         if (inode) {
 527                 iput(inode);
 528                 osb->sys_root_inode = NULL;
 529         }
 530
 531         inode = osb->root_inode;
 532         if (inode) {
 533                 iput(inode);
 534                 osb->root_inode = NULL;
 535         }
 536
 537         mlog_exit(0);
 538 }
 539
 540 /* We're allocating fs objects, use GFP_NOFS */
 541 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 542 {
 543         struct ocfs2_inode_info *oi;
 544
 545         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 546         if (!oi)
 547                 return NULL;
 548
 549         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 550         return &oi->vfs_inode;
 551 }
 552
 553 static void ocfs2_destroy_inode(struct inode *inode)
 554 {
 555         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 556 }
 557
 558 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 559                                                 unsigned int cbits)
 560 {
 561         unsigned int bytes = 1 << cbits;
 562         unsigned int trim = bytes;
 563         unsigned int bitshift = 32;
 564
 565         /*
 566          * i_size and all block offsets in ocfs2 are always 64 bits
 567          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 568          * 64 bit platforms, cluster size will be the limiting factor.
 569          */
 570
 571 #if BITS_PER_LONG == 32
 572 # if defined(CONFIG_LBDAF)
 573         BUILD_BUG_ON(sizeof(sector_t) != 8);
 574         /*
 575          * We might be limited by page cache size.
 576          */
 577         if (bytes > PAGE_CACHE_SIZE) {
 578                 bytes = PAGE_CACHE_SIZE;
 579                 trim = 1;
 580                 /*
 581                  * Shift by 31 here so that we don't get larger than
 582                  * MAX_LFS_FILESIZE
 583                  */
 584                 bitshift = 31;
 585         }
 586 # else
 587         /*
 588          * We are limited by the size of sector_t. Use block size, as
 589          * that's what we expose to the VFS.
 590          */
 591         bytes = 1 << bbits;
 592         trim = 1;
 593         bitshift = 31;
 594 # endif
 595 #endif
 596
 597         /*
 598          * Trim by a whole cluster when we can actually approach the
 599          * on-disk limits. Otherwise we can overflow i_clusters when
 600          * an extent start is at the max offset.
 601          */
 602         return (((unsigned long long)bytes) << bitshift) - trim;
 603 }
 604
 605 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 606 {
 607         int incompat_features;
 608         int ret = 0;
 609         struct mount_options parsed_options;
 610         struct ocfs2_super *osb = OCFS2_SB(sb);
 611
 612         lock_kernel();
 613
 614         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 615             !ocfs2_check_set_options(sb, &parsed_options)) {
 616                 ret = -EINVAL;
 617                 goto out;
 618         }
 619
 620         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
 621             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 622                 ret = -EINVAL;
 623                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 624                 goto out;
 625         }
 626
 627         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 628             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 629                 ret = -EINVAL;
 630                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 631                 goto out;
 632         }
 633
 634         /* Probably don't want this on remount; it might
 635          * mess with other nodes */
 636         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 637             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 638                 ret = -EINVAL;
 639                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 640                 goto out;
 641         }
 642
 643         /* We're going to/from readonly mode. */
 644         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 645                 /* Disable quota accounting before remounting RO */
 646                 if (*flags & MS_RDONLY) {
 647                         ret = ocfs2_susp_quotas(osb, 0);
 648                         if (ret < 0)
 649                                 goto out;
 650                 }
 651                 /* Lock here so the check of HARD_RO and the potential
 652                  * setting of SOFT_RO is atomic. */
 653                 spin_lock(&osb->osb_lock);
 654                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 655                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 656                         ret = -EROFS;
 657                         goto unlock_osb;
 658                 }
 659
 660                 if (*flags & MS_RDONLY) {
 661                         mlog(0, "Going to ro mode.\n");
 662                         sb->s_flags |= MS_RDONLY;
 663                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 664                 } else {
 665                         mlog(0, "Making ro filesystem writeable.\n");
 666
 667                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 668                                 mlog(ML_ERROR, "Cannot remount RDWR "
 669                                      "filesystem due to previous errors.\n");
 670                                 ret = -EROFS;
 671                                 goto unlock_osb;
 672                         }
 673                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 674                         if (incompat_features) {
 675                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 676                                      "of unsupported optional features "
 677                                      "(%x).\n", incompat_features);
 678                                 ret = -EINVAL;
 679                                 goto unlock_osb;
 680                         }
 681                         sb->s_flags &= ~MS_RDONLY;
 682                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 683                 }
 684 unlock_osb:
 685                 spin_unlock(&osb->osb_lock);
 686                 /* Enable quota accounting after remounting RW */
 687                 if (!ret && !(*flags & MS_RDONLY)) {
 688                         if (sb_any_quota_suspended(sb))
 689                                 ret = ocfs2_susp_quotas(osb, 1);
 690                         else
 691                                 ret = ocfs2_enable_quotas(osb);
 692                         if (ret < 0) {
 693                                 /* Return back changes... */
 694                                 spin_lock(&osb->osb_lock);
 695                                 sb->s_flags |= MS_RDONLY;
 696                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 697                                 spin_unlock(&osb->osb_lock);
 698                                 goto out;
 699                         }
 700                 }
 701         }
 702
 703         if (!ret) {
 704                 /* Only save off the new mount options in case of a successful
 705                  * remount. */
 706                 osb->s_mount_opt = parsed_options.mount_opt;
 707                 osb->s_atime_quantum = parsed_options.atime_quantum;
 708                 osb->preferred_slot = parsed_options.slot;
 709                 if (parsed_options.commit_interval)
 710                         osb->osb_commit_interval = parsed_options.commit_interval;
 711
 712                 if (!ocfs2_is_hard_readonly(osb))
 713                         ocfs2_set_journal_params(osb);
 714
 715                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 716                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 717                                                         MS_POSIXACL : 0);
 718         }
 719 out:
 720         unlock_kernel();
 721         return ret;
 722 }
 723
 724 static int ocfs2_sb_probe(struct super_block *sb,
 725                           struct buffer_head **bh,
 726                           int *sector_size,
 727                           struct ocfs2_blockcheck_stats *stats)
 728 {
 729         int status, tmpstat;
 730         struct ocfs1_vol_disk_hdr *hdr;
 731         struct ocfs2_dinode *di;
 732         int blksize;
 733
 734         *bh = NULL;
 735
 736         /* may be > 512 */
 737         *sector_size = bdev_logical_block_size(sb->s_bdev);
 738         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 739                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 740                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 741                 status = -EINVAL;
 742                 goto bail;
 743         }
 744
 745         /* Can this really happen? */
 746         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 747                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 748
 749         /* check block zero for old format */
 750         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 751         if (status < 0) {
 752                 mlog_errno(status);
 753                 goto bail;
 754         }
 755         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 756         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 757                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 758                      hdr->major_version, hdr->minor_version);
 759                 status = -EINVAL;
 760         }
 761         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 762                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 763                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 764                      hdr->signature);
 765                 status = -EINVAL;
 766         }
 767         brelse(*bh);
 768         *bh = NULL;
 769         if (status < 0) {
 770                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 771                      "upgraded before mounting with ocfs v2\n");
 772                 goto bail;
 773         }
 774
 775         /*
 776          * Now check at magic offset for 512, 1024, 2048, 4096
 777          * blocksizes.  4096 is the maximum blocksize because it is
 778          * the minimum clustersize.
 779          */
 780         status = -EINVAL;
 781         for (blksize = *sector_size;
 782              blksize <= OCFS2_MAX_BLOCKSIZE;
 783              blksize <<= 1) {
 784                 tmpstat = ocfs2_get_sector(sb, bh,
 785                                            OCFS2_SUPER_BLOCK_BLKNO,
 786                                            blksize);
 787                 if (tmpstat < 0) {
 788                         status = tmpstat;
 789                         mlog_errno(status);
 790                         break;
 791                 }
 792                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 793                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 794                 spin_lock_init(&stats->b_lock);
 795                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 796                 if (tmpstat < 0) {
 797                         brelse(*bh);
 798                         *bh = NULL;
 799                 }
 800                 if (tmpstat != -EAGAIN) {
 801                         status = tmpstat;
 802                         break;
 803                 }
 804         }
 805
 806 bail:
 807         return status;
 808 }
 809
 810 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 811 {
 812         if (ocfs2_mount_local(osb)) {
 813                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 814                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 815                              "mounted device.\n");
 816                         return -EINVAL;
 817                 }
 818         }
 819
 820         if (ocfs2_userspace_stack(osb)) {
 821                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 822                         mlog(ML_ERROR, "Userspace stack expected, but "
 823                              "o2cb heartbeat arguments passed to mount\n");
 824                         return -EINVAL;
 825                 }
 826         }
 827
 828         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 829                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 830                     !ocfs2_userspace_stack(osb)) {
 831                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 832                              "a read-write clustered device.\n");
 833                         return -EINVAL;
 834                 }
 835         }
 836
 837         return 0;
 838 }
 839
 840 /*
 841  * If we're using a userspace stack, mount should have passed
 842  * a name that matches the disk.  If not, mount should not
 843  * have passed a stack.
 844  */
 845 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 846                                         struct mount_options *mopt)
 847 {
 848         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 849                 mlog(ML_ERROR,
 850                      "cluster stack passed to mount, but this filesystem "
 851                      "does not support it\n");
 852                 return -EINVAL;
 853         }
 854
 855         if (ocfs2_userspace_stack(osb) &&
 856             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 857                     OCFS2_STACK_LABEL_LEN)) {
 858                 mlog(ML_ERROR,
 859                      "cluster stack passed to mount (\"%s\") does not "
 860                      "match the filesystem (\"%s\")\n",
 861                      mopt->cluster_stack,
 862                      osb->osb_cluster_stack);
 863                 return -EINVAL;
 864         }
 865
 866         return 0;
 867 }
 868
 869 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 870 {
 871         int type;
 872         struct super_block *sb = osb->sb;
 873         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 874                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 875         int status = 0;
 876
 877         for (type = 0; type < MAXQUOTAS; type++) {
 878                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 879                         continue;
 880                 if (unsuspend)
 881                         status = dquot_resume(sb, type);
 882                 else {
 883                         struct ocfs2_mem_dqinfo *oinfo;
 884
 885                         /* Cancel periodic syncing before suspending */
 886                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 887                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 888                         status = dquot_suspend(sb, type);
 889                 }
 890                 if (status < 0)
 891                         break;
 892         }
 893         if (status < 0)
 894                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 895                      "remount (error = %d).\n", status);
 896         return status;
 897 }
 898
 899 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 900 {
 901         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 902         struct super_block *sb = osb->sb;
 903         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 904                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 905         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 906                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 907         int status;
 908         int type;
 909
 910         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 911         for (type = 0; type < MAXQUOTAS; type++) {
 912                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 913                         continue;
 914                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 915                                                         osb->slot_num);
 916                 if (!inode[type]) {
 917                         status = -ENOENT;
 918                         goto out_quota_off;
 919                 }
 920                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 921                                       DQUOT_USAGE_ENABLED);
 922                 if (status < 0)
 923                         goto out_quota_off;
 924         }
 925
 926         for (type = 0; type < MAXQUOTAS; type++)
 927                 iput(inode[type]);
 928         return 0;
 929 out_quota_off:
 930         ocfs2_disable_quotas(osb);
 931         for (type = 0; type < MAXQUOTAS; type++)
 932                 iput(inode[type]);
 933         mlog_errno(status);
 934         return status;
 935 }
 936
 937 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 938 {
 939         int type;
 940         struct inode *inode;
 941         struct super_block *sb = osb->sb;
 942         struct ocfs2_mem_dqinfo *oinfo;
 943
 944         /* We mostly ignore errors in this function because there's not much
 945          * we can do when we see them */
 946         for (type = 0; type < MAXQUOTAS; type++) {
 947                 if (!sb_has_quota_loaded(sb, type))
 948                         continue;
 949                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 950                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 951                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 952                 inode = igrab(sb->s_dquot.files[type]);
 953                 /* Turn off quotas. This will remove all dquot structures from
 954                  * memory and so they will be automatically synced to global
 955                  * quota files */
 956                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 957                                         DQUOT_LIMITS_ENABLED);
 958                 if (!inode)
 959                         continue;
 960                 iput(inode);
 961         }
 962 }
 963
 964 /* Handle quota on quotactl */
 965 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 966                           char *path)
 967 {
 968         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 969                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 970
 971         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 972                 return -EINVAL;
 973
 974         return dquot_enable(sb_dqopt(sb)->files[type], type,
 975                             format_id, DQUOT_LIMITS_ENABLED);
 976 }
 977
 978 /* Handle quota off quotactl */
 979 static int ocfs2_quota_off(struct super_block *sb, int type)
 980 {
 981         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
 982 }
 983
 984 static const struct quotactl_ops ocfs2_quotactl_ops = {
 985         .quota_on       = ocfs2_quota_on,
 986         .quota_off      = ocfs2_quota_off,
 987         .quota_sync     = dquot_quota_sync,
 988         .get_info       = dquot_get_dqinfo,
 989         .set_info       = dquot_set_dqinfo,
 990         .get_dqblk      = dquot_get_dqblk,
 991         .set_dqblk      = dquot_set_dqblk,
 992 };
 993
 994 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 995 {
 996         struct dentry *root;
 997         int status, sector_size;
 998         struct mount_options parsed_options;
 999         struct inode *inode = NULL;
1000         struct ocfs2_super *osb = NULL;
1001         struct buffer_head *bh = NULL;
1002         char nodestr[8];
1003         struct ocfs2_blockcheck_stats stats;
1004
1005         mlog_entry("%p, %p, %i", sb, data, silent);
1006
1007         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1008                 status = -EINVAL;
1009                 goto read_super_error;
1010         }
1011
1012         /* probe for superblock */
1013         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1014         if (status < 0) {
1015                 mlog(ML_ERROR, "superblock probe failed!\n");
1016                 goto read_super_error;
1017         }
1018
1019         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1020         osb = OCFS2_SB(sb);
1021         if (status < 0) {
1022                 mlog_errno(status);
1023                 goto read_super_error;
1024         }
1025         brelse(bh);
1026         bh = NULL;
1027
1028         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1029                 status = -EINVAL;
1030                 goto read_super_error;
1031         }
1032         osb->s_mount_opt = parsed_options.mount_opt;
1033         osb->s_atime_quantum = parsed_options.atime_quantum;
1034         osb->preferred_slot = parsed_options.slot;
1035         osb->osb_commit_interval = parsed_options.commit_interval;
1036
1037         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1038         osb->osb_resv_level = parsed_options.resv_level;
1039         osb->osb_dir_resv_level = parsed_options.resv_level;
1040         if (parsed_options.dir_resv_level == -1)
1041                 osb->osb_dir_resv_level = parsed_options.resv_level;
1042         else
1043                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1044
1045         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1046         if (status)
1047                 goto read_super_error;
1048
1049         sb->s_magic = OCFS2_SUPER_MAGIC;
1050
1051         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1052                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1053
1054         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1055          * heartbeat=none */
1056         if (bdev_read_only(sb->s_bdev)) {
1057                 if (!(sb->s_flags & MS_RDONLY)) {
1058                         status = -EACCES;
1059                         mlog(ML_ERROR, "Readonly device detected but readonly "
1060                              "mount was not specified.\n");
1061                         goto read_super_error;
1062                 }
1063
1064                 /* You should not be able to start a local heartbeat
1065                  * on a readonly device. */
1066                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1067                         status = -EROFS;
1068                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1069                              "device.\n");
1070                         goto read_super_error;
1071                 }
1072
1073                 status = ocfs2_check_journals_nolocks(osb);
1074                 if (status < 0) {
1075                         if (status == -EROFS)
1076                                 mlog(ML_ERROR, "Recovery required on readonly "
1077                                      "file system, but write access is "
1078                                      "unavailable.\n");
1079                         else
1080                                 mlog_errno(status);
1081                         goto read_super_error;
1082                 }
1083
1084                 ocfs2_set_ro_flag(osb, 1);
1085
1086                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1087                        "services will be utilized for this mount. Recovery "
1088                        "will be skipped.\n");
1089         }
1090
1091         if (!ocfs2_is_hard_readonly(osb)) {
1092                 if (sb->s_flags & MS_RDONLY)
1093                         ocfs2_set_ro_flag(osb, 0);
1094         }
1095
1096         status = ocfs2_verify_heartbeat(osb);
1097         if (status < 0) {
1098                 mlog_errno(status);
1099                 goto read_super_error;
1100         }
1101
1102         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1103                                                  ocfs2_debugfs_root);
1104         if (!osb->osb_debug_root) {
1105                 status = -EINVAL;
1106                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1107                 goto read_super_error;
1108         }
1109
1110         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1111                                             osb->osb_debug_root,
1112                                             osb,
1113                                             &ocfs2_osb_debug_fops);
1114         if (!osb->osb_ctxt) {
1115                 status = -EINVAL;
1116                 mlog_errno(status);
1117                 goto read_super_error;
1118         }
1119
1120         if (ocfs2_meta_ecc(osb)) {
1121                 status = ocfs2_blockcheck_stats_debugfs_install(
1122                                                 &osb->osb_ecc_stats,
1123                                                 osb->osb_debug_root);
1124                 if (status) {
1125                         mlog(ML_ERROR,
1126                              "Unable to create blockcheck statistics "
1127                              "files\n");
1128                         goto read_super_error;
1129                 }
1130         }
1131
1132         status = ocfs2_mount_volume(sb);
1133         if (osb->root_inode)
1134                 inode = igrab(osb->root_inode);
1135
1136         if (status < 0)
1137                 goto read_super_error;
1138
1139         if (!inode) {
1140                 status = -EIO;
1141                 mlog_errno(status);
1142                 goto read_super_error;
1143         }
1144
1145         root = d_alloc_root(inode);
1146         if (!root) {
1147                 status = -ENOMEM;
1148                 mlog_errno(status);
1149                 goto read_super_error;
1150         }
1151
1152         sb->s_root = root;
1153
1154         ocfs2_complete_mount_recovery(osb);
1155
1156         if (ocfs2_mount_local(osb))
1157                 snprintf(nodestr, sizeof(nodestr), "local");
1158         else
1159                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1160
1161         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1162                "with %s data mode.\n",
1163                osb->dev_str, nodestr, osb->slot_num,
1164                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1165                "ordered");
1166
1167         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1168         wake_up(&osb->osb_mount_event);
1169
1170         /* Now we can initialize quotas because we can afford to wait
1171          * for cluster locks recovery now. That also means that truncation
1172          * log recovery can happen but that waits for proper quota setup */
1173         if (!(sb->s_flags & MS_RDONLY)) {
1174                 status = ocfs2_enable_quotas(osb);
1175                 if (status < 0) {
1176                         /* We have to err-out specially here because
1177                          * s_root is already set */
1178                         mlog_errno(status);
1179                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1180                         wake_up(&osb->osb_mount_event);
1181                         mlog_exit(status);
1182                         return status;
1183                 }
1184         }
1185
1186         ocfs2_complete_quota_recovery(osb);
1187
1188         /* Now we wake up again for processes waiting for quotas */
1189         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1190         wake_up(&osb->osb_mount_event);
1191
1192         /* Start this when the mount is almost sure of being successful */
1193         ocfs2_orphan_scan_start(osb);
1194
1195         mlog_exit(status);
1196         return status;
1197
1198 read_super_error:
1199         brelse(bh);
1200
1201         if (inode)
1202                 iput(inode);
1203
1204         if (osb) {
1205                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1206                 wake_up(&osb->osb_mount_event);
1207                 ocfs2_dismount_volume(sb, 1);
1208         }
1209
1210         mlog_exit(status);
1211         return status;
1212 }
1213
1214 static int ocfs2_get_sb(struct file_system_type *fs_type,
1215                         int flags,
1216                         const char *dev_name,
1217                         void *data,
1218                         struct vfsmount *mnt)
1219 {
1220         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1221                            mnt);
1222 }
1223
1224 static void ocfs2_kill_sb(struct super_block *sb)
1225 {
1226         struct ocfs2_super *osb = OCFS2_SB(sb);
1227
1228         /* Failed mount? */
1229         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1230                 goto out;
1231
1232         /* Prevent further queueing of inode drop events */
1233         spin_lock(&dentry_list_lock);
1234         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1235         spin_unlock(&dentry_list_lock);
1236         /* Wait for work to finish and/or remove it */
1237         cancel_work_sync(&osb->dentry_lock_work);
1238 out:
1239         kill_block_super(sb);
1240 }
1241
1242 static struct file_system_type ocfs2_fs_type = {
1243         .owner          = THIS_MODULE,
1244         .name           = "ocfs2",
1245         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1246                                         * the fs? */
1247         .kill_sb        = ocfs2_kill_sb,
1248
1249         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1250         .next           = NULL
1251 };
1252
1253 static int ocfs2_check_set_options(struct super_block *sb,
1254                                    struct mount_options *options)
1255 {
1256         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1257             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1258                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1259                 mlog(ML_ERROR, "User quotas were requested, but this "
1260                      "filesystem does not have the feature enabled.\n");
1261                 return 0;
1262         }
1263         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1264             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1265                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1266                 mlog(ML_ERROR, "Group quotas were requested, but this "
1267                      "filesystem does not have the feature enabled.\n");
1268                 return 0;
1269         }
1270         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1271             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1272                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1273                      "feature is not enabled\n");
1274                 return 0;
1275         }
1276         /* No ACL setting specified? Use XATTR feature... */
1277         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1278                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1279                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1280                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1281                 else
1282                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1283         }
1284         return 1;
1285 }
1286
1287 static int ocfs2_parse_options(struct super_block *sb,
1288                                char *options,
1289                                struct mount_options *mopt,
1290                                int is_remount)
1291 {
1292         int status;
1293         char *p;
1294
1295         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1296                    options ? options : "(none)");
1297
1298         mopt->commit_interval = 0;
1299         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1300         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1301         mopt->slot = OCFS2_INVALID_SLOT;
1302         mopt->localalloc_opt = -1;
1303         mopt->cluster_stack[0] = '\0';
1304         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1305         mopt->dir_resv_level = -1;
1306
1307         if (!options) {
1308                 status = 1;
1309                 goto bail;
1310         }
1311
1312         while ((p = strsep(&options, ",")) != NULL) {
1313                 int token, option;
1314                 substring_t args[MAX_OPT_ARGS];
1315
1316                 if (!*p)
1317                         continue;
1318
1319                 token = match_token(p, tokens, args);
1320                 switch (token) {
1321                 case Opt_hb_local:
1322                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1323                         break;
1324                 case Opt_hb_none:
1325                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1326                         break;
1327                 case Opt_barrier:
1328                         if (match_int(&args[0], &option)) {
1329                                 status = 0;
1330                                 goto bail;
1331                         }
1332                         if (option)
1333                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1334                         else
1335                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1336                         break;
1337                 case Opt_intr:
1338                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1339                         break;
1340                 case Opt_nointr:
1341                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1342                         break;
1343                 case Opt_err_panic:
1344                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1345                         break;
1346                 case Opt_err_ro:
1347                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1348                         break;
1349                 case Opt_data_ordered:
1350                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1351                         break;
1352                 case Opt_data_writeback:
1353                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1354                         break;
1355                 case Opt_user_xattr:
1356                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1357                         break;
1358                 case Opt_nouser_xattr:
1359                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1360                         break;
1361                 case Opt_atime_quantum:
1362                         if (match_int(&args[0], &option)) {
1363                                 status = 0;
1364                                 goto bail;
1365                         }
1366                         if (option >= 0)
1367                                 mopt->atime_quantum = option;
1368                         break;
1369                 case Opt_slot:
1370                         option = 0;
1371                         if (match_int(&args[0], &option)) {
1372                                 status = 0;
1373                                 goto bail;
1374                         }
1375                         if (option)
1376                                 mopt->slot = (s16)option;
1377                         break;
1378                 case Opt_commit:
1379                         option = 0;
1380                         if (match_int(&args[0], &option)) {
1381                                 status = 0;
1382                                 goto bail;
1383                         }
1384                         if (option < 0)
1385                                 return 0;
1386                         if (option == 0)
1387                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1388                         mopt->commit_interval = HZ * option;
1389                         break;
1390                 case Opt_localalloc:
1391                         option = 0;
1392                         if (match_int(&args[0], &option)) {
1393                                 status = 0;
1394                                 goto bail;
1395                         }
1396                         if (option >= 0)
1397                                 mopt->localalloc_opt = option;
1398                         break;
1399                 case Opt_localflocks:
1400                         /*
1401                          * Changing this during remount could race
1402                          * flock() requests, or "unbalance" existing
1403                          * ones (e.g., a lock is taken in one mode but
1404                          * dropped in the other). If users care enough
1405                          * to flip locking modes during remount, we
1406                          * could add a "local" flag to individual
1407                          * flock structures for proper tracking of
1408                          * state.
1409                          */
1410                         if (!is_remount)
1411                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1412                         break;
1413                 case Opt_stack:
1414                         /* Check both that the option we were passed
1415                          * is of the right length and that it is a proper
1416                          * string of the right length.
1417                          */
1418                         if (((args[0].to - args[0].from) !=
1419                              OCFS2_STACK_LABEL_LEN) ||
1420                             (strnlen(args[0].from,
1421                                      OCFS2_STACK_LABEL_LEN) !=
1422                              OCFS2_STACK_LABEL_LEN)) {
1423                                 mlog(ML_ERROR,
1424                                      "Invalid cluster_stack option\n");
1425                                 status = 0;
1426                                 goto bail;
1427                         }
1428                         memcpy(mopt->cluster_stack, args[0].from,
1429                                OCFS2_STACK_LABEL_LEN);
1430                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1431                         break;
1432                 case Opt_inode64:
1433                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1434                         break;
1435                 case Opt_usrquota:
1436                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1437                         break;
1438                 case Opt_grpquota:
1439                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1440                         break;
1441                 case Opt_acl:
1442                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1443                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1444                         break;
1445                 case Opt_noacl:
1446                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1447                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1448                         break;
1449                 case Opt_resv_level:
1450                         if (is_remount)
1451                                 break;
1452                         if (match_int(&args[0], &option)) {
1453                                 status = 0;
1454                                 goto bail;
1455                         }
1456                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1457                             option < OCFS2_MAX_RESV_LEVEL)
1458                                 mopt->resv_level = option;
1459                         break;
1460                 case Opt_dir_resv_level:
1461                         if (is_remount)
1462                                 break;
1463                         if (match_int(&args[0], &option)) {
1464                                 status = 0;
1465                                 goto bail;
1466                         }
1467                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1468                             option < OCFS2_MAX_RESV_LEVEL)
1469                                 mopt->dir_resv_level = option;
1470                         break;
1471                 default:
1472                         mlog(ML_ERROR,
1473                              "Unrecognized mount option \"%s\" "
1474                              "or missing value\n", p);
1475                         status = 0;
1476                         goto bail;
1477                 }
1478         }
1479
1480         status = 1;
1481
1482 bail:
1483         mlog_exit(status);
1484         return status;
1485 }
1486
1487 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1488 {
1489         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1490         unsigned long opts = osb->s_mount_opt;
1491         unsigned int local_alloc_megs;
1492
1493         if (opts & OCFS2_MOUNT_HB_LOCAL)
1494                 seq_printf(s, ",_netdev,heartbeat=local");
1495         else
1496                 seq_printf(s, ",heartbeat=none");
1497
1498         if (opts & OCFS2_MOUNT_NOINTR)
1499                 seq_printf(s, ",nointr");
1500
1501         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1502                 seq_printf(s, ",data=writeback");
1503         else
1504                 seq_printf(s, ",data=ordered");
1505
1506         if (opts & OCFS2_MOUNT_BARRIER)
1507                 seq_printf(s, ",barrier=1");
1508
1509         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1510                 seq_printf(s, ",errors=panic");
1511         else
1512                 seq_printf(s, ",errors=remount-ro");
1513
1514         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1515                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1516
1517         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1518                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1519
1520         if (osb->osb_commit_interval)
1521                 seq_printf(s, ",commit=%u",
1522                            (unsigned) (osb->osb_commit_interval / HZ));
1523
1524         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1525         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1526                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1527
1528         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1529                 seq_printf(s, ",localflocks,");
1530
1531         if (osb->osb_cluster_stack[0])
1532                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1533                            osb->osb_cluster_stack);
1534         if (opts & OCFS2_MOUNT_USRQUOTA)
1535                 seq_printf(s, ",usrquota");
1536         if (opts & OCFS2_MOUNT_GRPQUOTA)
1537                 seq_printf(s, ",grpquota");
1538
1539         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1540                 seq_printf(s, ",nouser_xattr");
1541         else
1542                 seq_printf(s, ",user_xattr");
1543
1544         if (opts & OCFS2_MOUNT_INODE64)
1545                 seq_printf(s, ",inode64");
1546
1547         if (opts & OCFS2_MOUNT_POSIX_ACL)
1548                 seq_printf(s, ",acl");
1549         else
1550                 seq_printf(s, ",noacl");
1551
1552         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1553                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1554
1555         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1556                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1557
1558         return 0;
1559 }
1560
1561 static int __init ocfs2_init(void)
1562 {
1563         int status;
1564
1565         mlog_entry_void();
1566
1567         ocfs2_print_version();
1568
1569         status = init_ocfs2_uptodate_cache();
1570         if (status < 0) {
1571                 mlog_errno(status);
1572                 goto leave;
1573         }
1574
1575         status = ocfs2_initialize_mem_caches();
1576         if (status < 0) {
1577                 mlog_errno(status);
1578                 goto leave;
1579         }
1580
1581         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1582         if (!ocfs2_wq) {
1583                 status = -ENOMEM;
1584                 goto leave;
1585         }
1586
1587         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1588         if (!ocfs2_debugfs_root) {
1589                 status = -EFAULT;
1590                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1591         }
1592
1593         status = ocfs2_quota_setup();
1594         if (status)
1595                 goto leave;
1596
1597         ocfs2_set_locking_protocol();
1598
1599         status = register_quota_format(&ocfs2_quota_format);
1600 leave:
1601         if (status < 0) {
1602                 ocfs2_quota_shutdown();
1603                 ocfs2_free_mem_caches();
1604                 exit_ocfs2_uptodate_cache();
1605         }
1606
1607         mlog_exit(status);
1608
1609         if (status >= 0) {
1610                 return register_filesystem(&ocfs2_fs_type);
1611         } else
1612                 return -1;
1613 }
1614
1615 static void __exit ocfs2_exit(void)
1616 {
1617         mlog_entry_void();
1618
1619         ocfs2_quota_shutdown();
1620
1621         if (ocfs2_wq) {
1622                 flush_workqueue(ocfs2_wq);
1623                 destroy_workqueue(ocfs2_wq);
1624         }
1625
1626         unregister_quota_format(&ocfs2_quota_format);
1627
1628         debugfs_remove(ocfs2_debugfs_root);
1629
1630         ocfs2_free_mem_caches();
1631
1632         unregister_filesystem(&ocfs2_fs_type);
1633
1634         exit_ocfs2_uptodate_cache();
1635
1636         mlog_exit_void();
1637 }
1638
1639 static void ocfs2_put_super(struct super_block *sb)
1640 {
1641         mlog_entry("(0x%p)\n", sb);
1642
1643         lock_kernel();
1644
1645         ocfs2_sync_blockdev(sb);
1646         ocfs2_dismount_volume(sb, 0);
1647
1648         unlock_kernel();
1649
1650         mlog_exit_void();
1651 }
1652
1653 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1654 {
1655         struct ocfs2_super *osb;
1656         u32 numbits, freebits;
1657         int status;
1658         struct ocfs2_dinode *bm_lock;
1659         struct buffer_head *bh = NULL;
1660         struct inode *inode = NULL;
1661
1662         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1663
1664         osb = OCFS2_SB(dentry->d_sb);
1665
1666         inode = ocfs2_get_system_file_inode(osb,
1667                                             GLOBAL_BITMAP_SYSTEM_INODE,
1668                                             OCFS2_INVALID_SLOT);
1669         if (!inode) {
1670                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1671                 status = -EIO;
1672                 goto bail;
1673         }
1674
1675         status = ocfs2_inode_lock(inode, &bh, 0);
1676         if (status < 0) {
1677                 mlog_errno(status);
1678                 goto bail;
1679         }
1680
1681         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1682
1683         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1684         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1685
1686         buf->f_type = OCFS2_SUPER_MAGIC;
1687         buf->f_bsize = dentry->d_sb->s_blocksize;
1688         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1689         buf->f_blocks = ((sector_t) numbits) *
1690                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1691         buf->f_bfree = ((sector_t) freebits) *
1692                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1693         buf->f_bavail = buf->f_bfree;
1694         buf->f_files = numbits;
1695         buf->f_ffree = freebits;
1696         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1697                                 & 0xFFFFFFFFUL;
1698         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1699                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1700
1701         brelse(bh);
1702
1703         ocfs2_inode_unlock(inode, 0);
1704         status = 0;
1705 bail:
1706         if (inode)
1707                 iput(inode);
1708
1709         mlog_exit(status);
1710
1711         return status;
1712 }
1713
1714 static void ocfs2_inode_init_once(void *data)
1715 {
1716         struct ocfs2_inode_info *oi = data;
1717
1718         oi->ip_flags = 0;
1719         oi->ip_open_count = 0;
1720         spin_lock_init(&oi->ip_lock);
1721         ocfs2_extent_map_init(&oi->vfs_inode);
1722         INIT_LIST_HEAD(&oi->ip_io_markers);
1723         oi->ip_dir_start_lookup = 0;
1724
1725         init_rwsem(&oi->ip_alloc_sem);
1726         init_rwsem(&oi->ip_xattr_sem);
1727         mutex_init(&oi->ip_io_mutex);
1728
1729         oi->ip_blkno = 0ULL;
1730         oi->ip_clusters = 0;
1731
1732         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1733
1734         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1735         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1736         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1737
1738         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1739                                   &ocfs2_inode_caching_ops);
1740
1741         inode_init_once(&oi->vfs_inode);
1742 }
1743
1744 static int ocfs2_initialize_mem_caches(void)
1745 {
1746         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1747                                        sizeof(struct ocfs2_inode_info),
1748                                        0,
1749                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1750                                                 SLAB_MEM_SPREAD),
1751                                        ocfs2_inode_init_once);
1752         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1753                                         sizeof(struct ocfs2_dquot),
1754                                         0,
1755                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1756                                                 SLAB_MEM_SPREAD),
1757                                         NULL);
1758         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1759                                         sizeof(struct ocfs2_quota_chunk),
1760                                         0,
1761                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1762                                         NULL);
1763         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1764             !ocfs2_qf_chunk_cachep) {
1765                 if (ocfs2_inode_cachep)
1766                         kmem_cache_destroy(ocfs2_inode_cachep);
1767                 if (ocfs2_dquot_cachep)
1768                         kmem_cache_destroy(ocfs2_dquot_cachep);
1769                 if (ocfs2_qf_chunk_cachep)
1770                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1771                 return -ENOMEM;
1772         }
1773
1774         return 0;
1775 }
1776
1777 static void ocfs2_free_mem_caches(void)
1778 {
1779         if (ocfs2_inode_cachep)
1780                 kmem_cache_destroy(ocfs2_inode_cachep);
1781         ocfs2_inode_cachep = NULL;
1782
1783         if (ocfs2_dquot_cachep)
1784                 kmem_cache_destroy(ocfs2_dquot_cachep);
1785         ocfs2_dquot_cachep = NULL;
1786
1787         if (ocfs2_qf_chunk_cachep)
1788                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1789         ocfs2_qf_chunk_cachep = NULL;
1790 }
1791
1792 static int ocfs2_get_sector(struct super_block *sb,
1793                             struct buffer_head **bh,
1794                             int block,
1795                             int sect_size)
1796 {
1797         if (!sb_set_blocksize(sb, sect_size)) {
1798                 mlog(ML_ERROR, "unable to set blocksize\n");
1799                 return -EIO;
1800         }
1801
1802         *bh = sb_getblk(sb, block);
1803         if (!*bh) {
1804                 mlog_errno(-EIO);
1805                 return -EIO;
1806         }
1807         lock_buffer(*bh);
1808         if (!buffer_dirty(*bh))
1809                 clear_buffer_uptodate(*bh);
1810         unlock_buffer(*bh);
1811         ll_rw_block(READ, 1, bh);
1812         wait_on_buffer(*bh);
1813         if (!buffer_uptodate(*bh)) {
1814                 mlog_errno(-EIO);
1815                 brelse(*bh);
1816                 *bh = NULL;
1817                 return -EIO;
1818         }
1819
1820         return 0;
1821 }
1822
1823 static int ocfs2_mount_volume(struct super_block *sb)
1824 {
1825         int status = 0;
1826         int unlock_super = 0;
1827         struct ocfs2_super *osb = OCFS2_SB(sb);
1828
1829         mlog_entry_void();
1830
1831         if (ocfs2_is_hard_readonly(osb))
1832                 goto leave;
1833
1834         status = ocfs2_dlm_init(osb);
1835         if (status < 0) {
1836                 mlog_errno(status);
1837                 goto leave;
1838         }
1839
1840         status = ocfs2_super_lock(osb, 1);
1841         if (status < 0) {
1842                 mlog_errno(status);
1843                 goto leave;
1844         }
1845         unlock_super = 1;
1846
1847         /* This will load up the node map and add ourselves to it. */
1848         status = ocfs2_find_slot(osb);
1849         if (status < 0) {
1850                 mlog_errno(status);
1851                 goto leave;
1852         }
1853
1854         /* load all node-local system inodes */
1855         status = ocfs2_init_local_system_inodes(osb);
1856         if (status < 0) {
1857                 mlog_errno(status);
1858                 goto leave;
1859         }
1860
1861         status = ocfs2_check_volume(osb);
1862         if (status < 0) {
1863                 mlog_errno(status);
1864                 goto leave;
1865         }
1866
1867         status = ocfs2_truncate_log_init(osb);
1868         if (status < 0)
1869                 mlog_errno(status);
1870
1871 leave:
1872         if (unlock_super)
1873                 ocfs2_super_unlock(osb, 1);
1874
1875         mlog_exit(status);
1876         return status;
1877 }
1878
1879 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1880 {
1881         int tmp, hangup_needed = 0;
1882         struct ocfs2_super *osb = NULL;
1883         char nodestr[8];
1884
1885         mlog_entry("(0x%p)\n", sb);
1886
1887         BUG_ON(!sb);
1888         osb = OCFS2_SB(sb);
1889         BUG_ON(!osb);
1890
1891         debugfs_remove(osb->osb_ctxt);
1892
1893         /*
1894          * Flush inode dropping work queue so that deletes are
1895          * performed while the filesystem is still working
1896          */
1897         ocfs2_drop_all_dl_inodes(osb);
1898
1899         /* Orphan scan should be stopped as early as possible */
1900         ocfs2_orphan_scan_stop(osb);
1901
1902         ocfs2_disable_quotas(osb);
1903
1904         ocfs2_shutdown_local_alloc(osb);
1905
1906         ocfs2_truncate_log_shutdown(osb);
1907
1908         /* This will disable recovery and flush any recovery work. */
1909         ocfs2_recovery_exit(osb);
1910
1911         ocfs2_journal_shutdown(osb);
1912
1913         ocfs2_sync_blockdev(sb);
1914
1915         ocfs2_purge_refcount_trees(osb);
1916
1917         /* No cluster connection means we've failed during mount, so skip
1918          * all the steps which depended on that to complete. */
1919         if (osb->cconn) {
1920                 tmp = ocfs2_super_lock(osb, 1);
1921                 if (tmp < 0) {
1922                         mlog_errno(tmp);
1923                         return;
1924                 }
1925         }
1926
1927         if (osb->slot_num != OCFS2_INVALID_SLOT)
1928                 ocfs2_put_slot(osb);
1929
1930         if (osb->cconn)
1931                 ocfs2_super_unlock(osb, 1);
1932
1933         ocfs2_release_system_inodes(osb);
1934
1935         /*
1936          * If we're dismounting due to mount error, mount.ocfs2 will clean
1937          * up heartbeat.  If we're a local mount, there is no heartbeat.
1938          * If we failed before we got a uuid_str yet, we can't stop
1939          * heartbeat.  Otherwise, do it.
1940          */
1941         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1942                 hangup_needed = 1;
1943
1944         if (osb->cconn)
1945                 ocfs2_dlm_shutdown(osb, hangup_needed);
1946
1947         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1948         debugfs_remove(osb->osb_debug_root);
1949
1950         if (hangup_needed)
1951                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1952
1953         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1954
1955         if (ocfs2_mount_local(osb))
1956                 snprintf(nodestr, sizeof(nodestr), "local");
1957         else
1958                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1959
1960         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1961                osb->dev_str, nodestr);
1962
1963         ocfs2_delete_osb(osb);
1964         kfree(osb);
1965         sb->s_dev = 0;
1966         sb->s_fs_info = NULL;
1967 }
1968
1969 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1970                                 unsigned uuid_bytes)
1971 {
1972         int i, ret;
1973         char *ptr;
1974
1975         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1976
1977         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1978         if (osb->uuid_str == NULL)
1979                 return -ENOMEM;
1980
1981         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1982                 /* print with null */
1983                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1984                 if (ret != 2) /* drop super cleans up */
1985                         return -EINVAL;
1986                 /* then only advance past the last char */
1987                 ptr += 2;
1988         }
1989
1990         return 0;
1991 }
1992
1993 /* Make sure entire volume is addressable by our journal.  Requires
1994    osb_clusters_at_boot to be valid and for the journal to have been
1995    initialized by ocfs2_journal_init(). */
1996 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1997 {
1998         int status = 0;
1999         u64 max_block =
2000                 ocfs2_clusters_to_blocks(osb->sb,
2001                                          osb->osb_clusters_at_boot) - 1;
2002
2003         /* 32-bit block number is always OK. */
2004         if (max_block <= (u32)~0ULL)
2005                 goto out;
2006
2007         /* Volume is "huge", so see if our journal is new enough to
2008            support it. */
2009         if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2010                                        OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2011               jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2012                                                JBD2_FEATURE_INCOMPAT_64BIT))) {
2013                 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2014                      "Enable the 'block64' journal option with tunefs.ocfs2");
2015                 status = -EFBIG;
2016                 goto out;
2017         }
2018
2019  out:
2020         return status;
2021 }
2022
2023 static int ocfs2_initialize_super(struct super_block *sb,
2024                                   struct buffer_head *bh,
2025                                   int sector_size,
2026                                   struct ocfs2_blockcheck_stats *stats)
2027 {
2028         int status;
2029         int i, cbits, bbits;
2030         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2031         struct inode *inode = NULL;
2032         struct ocfs2_journal *journal;
2033         __le32 uuid_net_key;
2034         struct ocfs2_super *osb;
2035         u64 total_blocks;
2036
2037         mlog_entry_void();
2038
2039         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2040         if (!osb) {
2041                 status = -ENOMEM;
2042                 mlog_errno(status);
2043                 goto bail;
2044         }
2045
2046         sb->s_fs_info = osb;
2047         sb->s_op = &ocfs2_sops;
2048         sb->s_export_op = &ocfs2_export_ops;
2049         sb->s_qcop = &ocfs2_quotactl_ops;
2050         sb->dq_op = &ocfs2_quota_operations;
2051         sb->s_xattr = ocfs2_xattr_handlers;
2052         sb->s_time_gran = 1;
2053         sb->s_flags |= MS_NOATIME;
2054         /* this is needed to support O_LARGEFILE */
2055         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2056         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2057         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2058
2059         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2060
2061         for (i = 0; i < 3; i++)
2062                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2063         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2064
2065         osb->sb = sb;
2066         /* Save off for ocfs2_rw_direct */
2067         osb->s_sectsize_bits = blksize_bits(sector_size);
2068         BUG_ON(!osb->s_sectsize_bits);
2069
2070         spin_lock_init(&osb->dc_task_lock);
2071         init_waitqueue_head(&osb->dc_event);
2072         osb->dc_work_sequence = 0;
2073         osb->dc_wake_sequence = 0;
2074         INIT_LIST_HEAD(&osb->blocked_lock_list);
2075         osb->blocked_lock_count = 0;
2076         spin_lock_init(&osb->osb_lock);
2077         spin_lock_init(&osb->osb_xattr_lock);
2078         ocfs2_init_steal_slots(osb);
2079
2080         atomic_set(&osb->alloc_stats.moves, 0);
2081         atomic_set(&osb->alloc_stats.local_data, 0);
2082         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2083         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2084         atomic_set(&osb->alloc_stats.bg_extends, 0);
2085
2086         /* Copy the blockcheck stats from the superblock probe */
2087         osb->osb_ecc_stats = *stats;
2088
2089         ocfs2_init_node_maps(osb);
2090
2091         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2092                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2093
2094         ocfs2_orphan_scan_init(osb);
2095
2096         status = ocfs2_recovery_init(osb);
2097         if (status) {
2098                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2099                 mlog_errno(status);
2100                 goto bail;
2101         }
2102
2103         init_waitqueue_head(&osb->checkpoint_event);
2104         atomic_set(&osb->needs_checkpoint, 0);
2105
2106         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2107
2108         osb->slot_num = OCFS2_INVALID_SLOT;
2109
2110         osb->s_xattr_inline_size = le16_to_cpu(
2111                                         di->id2.i_super.s_xattr_inline_size);
2112
2113         osb->local_alloc_state = OCFS2_LA_UNUSED;
2114         osb->local_alloc_bh = NULL;
2115         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2116
2117         init_waitqueue_head(&osb->osb_mount_event);
2118
2119         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2120         if (status) {
2121                 mlog_errno(status);
2122                 goto bail;
2123         }
2124
2125         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2126         if (!osb->vol_label) {
2127                 mlog(ML_ERROR, "unable to alloc vol label\n");
2128                 status = -ENOMEM;
2129                 goto bail;
2130         }
2131
2132         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2133         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2134                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2135                      osb->max_slots);
2136                 status = -EINVAL;
2137                 goto bail;
2138         }
2139         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2140
2141         osb->slot_recovery_generations =
2142                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2143                         GFP_KERNEL);
2144         if (!osb->slot_recovery_generations) {
2145                 status = -ENOMEM;
2146                 mlog_errno(status);
2147                 goto bail;
2148         }
2149
2150         init_waitqueue_head(&osb->osb_wipe_event);
2151         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2152                                         sizeof(*osb->osb_orphan_wipes),
2153                                         GFP_KERNEL);
2154         if (!osb->osb_orphan_wipes) {
2155                 status = -ENOMEM;
2156                 mlog_errno(status);
2157                 goto bail;
2158         }
2159
2160         osb->osb_rf_lock_tree = RB_ROOT;
2161
2162         osb->s_feature_compat =
2163                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2164         osb->s_feature_ro_compat =
2165                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2166         osb->s_feature_incompat =
2167                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2168
2169         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2170                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2171                      "optional features (%x).\n", i);
2172                 status = -EINVAL;
2173                 goto bail;
2174         }
2175         if (!(osb->sb->s_flags & MS_RDONLY) &&
2176             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2177                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2178                      "unsupported optional features (%x).\n", i);
2179                 status = -EINVAL;
2180                 goto bail;
2181         }
2182
2183         if (ocfs2_userspace_stack(osb)) {
2184                 memcpy(osb->osb_cluster_stack,
2185                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2186                        OCFS2_STACK_LABEL_LEN);
2187                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2188                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2189                         mlog(ML_ERROR,
2190                              "couldn't mount because of an invalid "
2191                              "cluster stack label (%s) \n",
2192                              osb->osb_cluster_stack);
2193                         status = -EINVAL;
2194                         goto bail;
2195                 }
2196         } else {
2197                 /* The empty string is identical with classic tools that
2198                  * don't know about s_cluster_info. */
2199                 osb->osb_cluster_stack[0] = '\0';
2200         }
2201
2202         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2203
2204         /* FIXME
2205          * This should be done in ocfs2_journal_init(), but unknown
2206          * ordering issues will cause the filesystem to crash.
2207          * If anyone wants to figure out what part of the code
2208          * refers to osb->journal before ocfs2_journal_init() is run,
2209          * be my guest.
2210          */
2211         /* initialize our journal structure */
2212
2213         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2214         if (!journal) {
2215                 mlog(ML_ERROR, "unable to alloc journal\n");
2216                 status = -ENOMEM;
2217                 goto bail;
2218         }
2219         osb->journal = journal;
2220         journal->j_osb = osb;
2221
2222         atomic_set(&journal->j_num_trans, 0);
2223         init_rwsem(&journal->j_trans_barrier);
2224         init_waitqueue_head(&journal->j_checkpointed);
2225         spin_lock_init(&journal->j_lock);
2226         journal->j_trans_id = (unsigned long) 1;
2227         INIT_LIST_HEAD(&journal->j_la_cleanups);
2228         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2229         journal->j_state = OCFS2_JOURNAL_FREE;
2230
2231         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2232         osb->dentry_lock_list = NULL;
2233
2234         /* get some pseudo constants for clustersize bits */
2235         osb->s_clustersize_bits =
2236                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2237         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2238         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2239
2240         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2241             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2242                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2243                      osb->s_clustersize);
2244                 status = -EINVAL;
2245                 goto bail;
2246         }
2247
2248         total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2249                                                 le32_to_cpu(di->i_clusters));
2250
2251         status = generic_check_addressable(osb->sb->s_blocksize_bits,
2252                                            total_blocks);
2253         if (status) {
2254                 mlog(ML_ERROR, "Volume too large "
2255                      "to mount safely on this system");
2256                 status = -EFBIG;
2257                 goto bail;
2258         }
2259
2260         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2261                                  sizeof(di->id2.i_super.s_uuid))) {
2262                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2263                 status = -ENOMEM;
2264                 goto bail;
2265         }
2266
2267         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2268
2269         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2270         osb->vol_label[63] = '\0';
2271         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2272         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2273         osb->first_cluster_group_blkno =
2274                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2275         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2276         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2277         mlog(0, "vol_label: %s\n", osb->vol_label);
2278         mlog(0, "uuid: %s\n", osb->uuid_str);
2279         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2280              (unsigned long long)osb->root_blkno,
2281              (unsigned long long)osb->system_dir_blkno);
2282
2283         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2284         if (!osb->osb_dlm_debug) {
2285                 status = -ENOMEM;
2286                 mlog_errno(status);
2287                 goto bail;
2288         }
2289
2290         atomic_set(&osb->vol_state, VOLUME_INIT);
2291
2292         /* load root, system_dir, and all global system inodes */
2293         status = ocfs2_init_global_system_inodes(osb);
2294         if (status < 0) {
2295                 mlog_errno(status);
2296                 goto bail;
2297         }
2298
2299         /*
2300          * global bitmap
2301          */
2302         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2303                                             OCFS2_INVALID_SLOT);
2304         if (!inode) {
2305                 status = -EINVAL;
2306                 mlog_errno(status);
2307                 goto bail;
2308         }
2309
2310         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2311         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2312         iput(inode);
2313
2314         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2315                                  osb->s_feature_incompat) * 8;
2316
2317         status = ocfs2_init_slot_info(osb);
2318         if (status < 0) {
2319                 mlog_errno(status);
2320                 goto bail;
2321         }
2322
2323 bail:
2324         mlog_exit(status);
2325         return status;
2326 }
2327
2328 /*
2329  * will return: -EAGAIN if it is ok to keep searching for superblocks
2330  *              -EINVAL if there is a bad superblock
2331  *              0 on success
2332  */
2333 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2334                                struct buffer_head *bh,
2335                                u32 blksz,
2336                                struct ocfs2_blockcheck_stats *stats)
2337 {
2338         int status = -EAGAIN;
2339
2340         mlog_entry_void();
2341
2342         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2343                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2344                 /* We have to do a raw check of the feature here */
2345                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2346                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2347                         status = ocfs2_block_check_validate(bh->b_data,
2348                                                             bh->b_size,
2349                                                             &di->i_check,
2350                                                             stats);
2351                         if (status)
2352                                 goto out;
2353                 }
2354                 status = -EINVAL;
2355                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2356                         mlog(ML_ERROR, "found superblock with incorrect block "
2357                              "size: found %u, should be %u\n",
2358                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2359                                blksz);
2360                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2361                            OCFS2_MAJOR_REV_LEVEL ||
2362                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2363                            OCFS2_MINOR_REV_LEVEL) {
2364                         mlog(ML_ERROR, "found superblock with bad version: "
2365                              "found %u.%u, should be %u.%u\n",
2366                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2367                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2368                              OCFS2_MAJOR_REV_LEVEL,
2369                              OCFS2_MINOR_REV_LEVEL);
2370                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2371                         mlog(ML_ERROR, "bad block number on superblock: "
2372                              "found %llu, should be %llu\n",
2373                              (unsigned long long)le64_to_cpu(di->i_blkno),
2374                              (unsigned long long)bh->b_blocknr);
2375                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2376                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2377                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2378                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2379                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2380                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2381                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2382                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2383                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2384                         mlog(ML_ERROR,
2385                              "Superblock slots found greater than file system "
2386                              "maximum: found %u, max %u\n",
2387                              le16_to_cpu(di->id2.i_super.s_max_slots),
2388                              OCFS2_MAX_SLOTS);
2389                 } else {
2390                         /* found it! */
2391                         status = 0;
2392                 }
2393         }
2394
2395 out:
2396         mlog_exit(status);
2397         return status;
2398 }
2399
2400 static int ocfs2_check_volume(struct ocfs2_super *osb)
2401 {
2402         int status;
2403         int dirty;
2404         int local;
2405         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2406                                                   * recover
2407                                                   * ourselves. */
2408
2409         mlog_entry_void();
2410
2411         /* Init our journal object. */
2412         status = ocfs2_journal_init(osb->journal, &dirty);
2413         if (status < 0) {
2414                 mlog(ML_ERROR, "Could not initialize journal!\n");
2415                 goto finally;
2416         }
2417
2418         /* Now that journal has been initialized, check to make sure
2419            entire volume is addressable. */
2420         status = ocfs2_journal_addressable(osb);
2421         if (status)
2422                 goto finally;
2423
2424         /* If the journal was unmounted cleanly then we don't want to
2425          * recover anything. Otherwise, journal_load will do that
2426          * dirty work for us :) */
2427         if (!dirty) {
2428                 status = ocfs2_journal_wipe(osb->journal, 0);
2429                 if (status < 0) {
2430                         mlog_errno(status);
2431                         goto finally;
2432                 }
2433         } else {
2434                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2435                      "recovering volume.\n");
2436         }
2437
2438         local = ocfs2_mount_local(osb);
2439
2440         /* will play back anything left in the journal. */
2441         status = ocfs2_journal_load(osb->journal, local, dirty);
2442         if (status < 0) {
2443                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2444                 goto finally;
2445         }
2446
2447         if (dirty) {
2448                 /* recover my local alloc if we didn't unmount cleanly. */
2449                 status = ocfs2_begin_local_alloc_recovery(osb,
2450                                                           osb->slot_num,
2451                                                           &local_alloc);
2452                 if (status < 0) {
2453                         mlog_errno(status);
2454                         goto finally;
2455                 }
2456                 /* we complete the recovery process after we've marked
2457                  * ourselves as mounted. */
2458         }
2459
2460         mlog(0, "Journal loaded.\n");
2461
2462         status = ocfs2_load_local_alloc(osb);
2463         if (status < 0) {
2464                 mlog_errno(status);
2465                 goto finally;
2466         }
2467
2468         if (dirty) {
2469                 /* Recovery will be completed after we've mounted the
2470                  * rest of the volume. */
2471                 osb->dirty = 1;
2472                 osb->local_alloc_copy = local_alloc;
2473                 local_alloc = NULL;
2474         }
2475
2476         /* go through each journal, trylock it and if you get the
2477          * lock, and it's marked as dirty, set the bit in the recover
2478          * map and launch a recovery thread for it. */
2479         status = ocfs2_mark_dead_nodes(osb);
2480         if (status < 0) {
2481                 mlog_errno(status);
2482                 goto finally;
2483         }
2484
2485         status = ocfs2_compute_replay_slots(osb);
2486         if (status < 0)
2487                 mlog_errno(status);
2488
2489 finally:
2490         if (local_alloc)
2491                 kfree(local_alloc);
2492
2493         mlog_exit(status);
2494         return status;
2495 }
2496
2497 /*
2498  * The routine gets called from dismount or close whenever a dismount on
2499  * volume is requested and the osb open count becomes 1.
2500  * It will remove the osb from the global list and also free up all the
2501  * initialized resources and fileobject.
2502  */
2503 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2504 {
2505         mlog_entry_void();
2506
2507         /* This function assumes that the caller has the main osb resource */
2508
2509         ocfs2_free_slot_info(osb);
2510
2511         kfree(osb->osb_orphan_wipes);
2512         kfree(osb->slot_recovery_generations);
2513         /* FIXME
2514          * This belongs in journal shutdown, but because we have to
2515          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2516          * we free it here.
2517          */
2518         kfree(osb->journal);
2519         if (osb->local_alloc_copy)
2520                 kfree(osb->local_alloc_copy);
2521         kfree(osb->uuid_str);
2522         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2523         memset(osb, 0, sizeof(struct ocfs2_super));
2524
2525         mlog_exit_void();
2526 }
2527
2528 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2529  * panic(). We do not support continue-on-error operation. */
2530 static void ocfs2_handle_error(struct super_block *sb)
2531 {
2532         struct ocfs2_super *osb = OCFS2_SB(sb);
2533
2534         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2535                 panic("OCFS2: (device %s): panic forced after error\n",
2536                       sb->s_id);
2537
2538         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2539
2540         if (sb->s_flags & MS_RDONLY &&
2541             (ocfs2_is_soft_readonly(osb) ||
2542              ocfs2_is_hard_readonly(osb)))
2543                 return;
2544
2545         printk(KERN_CRIT "File system is now read-only due to the potential "
2546                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2547                "system is unmounted.\n");
2548         sb->s_flags |= MS_RDONLY;
2549         ocfs2_set_ro_flag(osb, 0);
2550 }
2551
2552 static char error_buf[1024];
2553
2554 void __ocfs2_error(struct super_block *sb,
2555                    const char *function,
2556                    const char *fmt, ...)
2557 {
2558         va_list args;
2559
2560         va_start(args, fmt);
2561         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2562         va_end(args);
2563
2564         /* Not using mlog here because we want to show the actual
2565          * function the error came from. */
2566         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2567                sb->s_id, function, error_buf);
2568
2569         ocfs2_handle_error(sb);
2570 }
2571
2572 /* Handle critical errors. This is intentionally more drastic than
2573  * ocfs2_handle_error, so we only use for things like journal errors,
2574  * etc. */
2575 void __ocfs2_abort(struct super_block* sb,
2576                    const char *function,
2577                    const char *fmt, ...)
2578 {
2579         va_list args;
2580
2581         va_start(args, fmt);
2582         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2583         va_end(args);
2584
2585         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2586                sb->s_id, function, error_buf);
2587
2588         /* We don't have the cluster support yet to go straight to
2589          * hard readonly in here. Until then, we want to keep
2590          * ocfs2_abort() so that we can at least mark critical
2591          * errors.
2592          *
2593          * TODO: This should abort the journal and alert other nodes
2594          * that our slot needs recovery. */
2595
2596         /* Force a panic(). This stinks, but it's better than letting
2597          * things continue without having a proper hard readonly
2598          * here. */
2599         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2600                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2601         ocfs2_handle_error(sb);
2602 }
2603
2604 /*
2605  * Void signal blockers, because in-kernel sigprocmask() only fails
2606  * when SIG_* is wrong.
2607  */
2608 void ocfs2_block_signals(sigset_t *oldset)
2609 {
2610         int rc;
2611         sigset_t blocked;
2612
2613         sigfillset(&blocked);
2614         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2615         BUG_ON(rc);
2616 }
2617
2618 void ocfs2_unblock_signals(sigset_t *oldset)
2619 {
2620         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2621         BUG_ON(rc);
2622 }
2623
2624 module_init(ocfs2_init);
2625 module_exit(ocfs2_exit);