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         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 613             !ocfs2_check_set_options(sb, &parsed_options)) {
 614                 ret = -EINVAL;
 615                 goto out;
 616         }
 617
 618         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
 619             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 620                 ret = -EINVAL;
 621                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 622                 goto out;
 623         }
 624
 625         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 626             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 627                 ret = -EINVAL;
 628                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 629                 goto out;
 630         }
 631
 632         /* Probably don't want this on remount; it might
 633          * mess with other nodes */
 634         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 635             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 636                 ret = -EINVAL;
 637                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 638                 goto out;
 639         }
 640
 641         /* We're going to/from readonly mode. */
 642         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 643                 /* Disable quota accounting before remounting RO */
 644                 if (*flags & MS_RDONLY) {
 645                         ret = ocfs2_susp_quotas(osb, 0);
 646                         if (ret < 0)
 647                                 goto out;
 648                 }
 649                 /* Lock here so the check of HARD_RO and the potential
 650                  * setting of SOFT_RO is atomic. */
 651                 spin_lock(&osb->osb_lock);
 652                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 653                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 654                         ret = -EROFS;
 655                         goto unlock_osb;
 656                 }
 657
 658                 if (*flags & MS_RDONLY) {
 659                         mlog(0, "Going to ro mode.\n");
 660                         sb->s_flags |= MS_RDONLY;
 661                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 662                 } else {
 663                         mlog(0, "Making ro filesystem writeable.\n");
 664
 665                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 666                                 mlog(ML_ERROR, "Cannot remount RDWR "
 667                                      "filesystem due to previous errors.\n");
 668                                 ret = -EROFS;
 669                                 goto unlock_osb;
 670                         }
 671                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 672                         if (incompat_features) {
 673                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 674                                      "of unsupported optional features "
 675                                      "(%x).\n", incompat_features);
 676                                 ret = -EINVAL;
 677                                 goto unlock_osb;
 678                         }
 679                         sb->s_flags &= ~MS_RDONLY;
 680                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 681                 }
 682 unlock_osb:
 683                 spin_unlock(&osb->osb_lock);
 684                 /* Enable quota accounting after remounting RW */
 685                 if (!ret && !(*flags & MS_RDONLY)) {
 686                         if (sb_any_quota_suspended(sb))
 687                                 ret = ocfs2_susp_quotas(osb, 1);
 688                         else
 689                                 ret = ocfs2_enable_quotas(osb);
 690                         if (ret < 0) {
 691                                 /* Return back changes... */
 692                                 spin_lock(&osb->osb_lock);
 693                                 sb->s_flags |= MS_RDONLY;
 694                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 695                                 spin_unlock(&osb->osb_lock);
 696                                 goto out;
 697                         }
 698                 }
 699         }
 700
 701         if (!ret) {
 702                 /* Only save off the new mount options in case of a successful
 703                  * remount. */
 704                 osb->s_mount_opt = parsed_options.mount_opt;
 705                 osb->s_atime_quantum = parsed_options.atime_quantum;
 706                 osb->preferred_slot = parsed_options.slot;
 707                 if (parsed_options.commit_interval)
 708                         osb->osb_commit_interval = parsed_options.commit_interval;
 709
 710                 if (!ocfs2_is_hard_readonly(osb))
 711                         ocfs2_set_journal_params(osb);
 712
 713                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 714                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 715                                                         MS_POSIXACL : 0);
 716         }
 717 out:
 718         return ret;
 719 }
 720
 721 static int ocfs2_sb_probe(struct super_block *sb,
 722                           struct buffer_head **bh,
 723                           int *sector_size,
 724                           struct ocfs2_blockcheck_stats *stats)
 725 {
 726         int status, tmpstat;
 727         struct ocfs1_vol_disk_hdr *hdr;
 728         struct ocfs2_dinode *di;
 729         int blksize;
 730
 731         *bh = NULL;
 732
 733         /* may be > 512 */
 734         *sector_size = bdev_logical_block_size(sb->s_bdev);
 735         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 736                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 737                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 738                 status = -EINVAL;
 739                 goto bail;
 740         }
 741
 742         /* Can this really happen? */
 743         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 744                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 745
 746         /* check block zero for old format */
 747         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 748         if (status < 0) {
 749                 mlog_errno(status);
 750                 goto bail;
 751         }
 752         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 753         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 754                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 755                      hdr->major_version, hdr->minor_version);
 756                 status = -EINVAL;
 757         }
 758         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 759                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 760                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 761                      hdr->signature);
 762                 status = -EINVAL;
 763         }
 764         brelse(*bh);
 765         *bh = NULL;
 766         if (status < 0) {
 767                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 768                      "upgraded before mounting with ocfs v2\n");
 769                 goto bail;
 770         }
 771
 772         /*
 773          * Now check at magic offset for 512, 1024, 2048, 4096
 774          * blocksizes.  4096 is the maximum blocksize because it is
 775          * the minimum clustersize.
 776          */
 777         status = -EINVAL;
 778         for (blksize = *sector_size;
 779              blksize <= OCFS2_MAX_BLOCKSIZE;
 780              blksize <<= 1) {
 781                 tmpstat = ocfs2_get_sector(sb, bh,
 782                                            OCFS2_SUPER_BLOCK_BLKNO,
 783                                            blksize);
 784                 if (tmpstat < 0) {
 785                         status = tmpstat;
 786                         mlog_errno(status);
 787                         break;
 788                 }
 789                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 790                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 791                 spin_lock_init(&stats->b_lock);
 792                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 793                 if (tmpstat < 0) {
 794                         brelse(*bh);
 795                         *bh = NULL;
 796                 }
 797                 if (tmpstat != -EAGAIN) {
 798                         status = tmpstat;
 799                         break;
 800                 }
 801         }
 802
 803 bail:
 804         return status;
 805 }
 806
 807 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 808 {
 809         if (ocfs2_mount_local(osb)) {
 810                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 811                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 812                              "mounted device.\n");
 813                         return -EINVAL;
 814                 }
 815         }
 816
 817         if (ocfs2_userspace_stack(osb)) {
 818                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 819                         mlog(ML_ERROR, "Userspace stack expected, but "
 820                              "o2cb heartbeat arguments passed to mount\n");
 821                         return -EINVAL;
 822                 }
 823         }
 824
 825         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 826                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 827                     !ocfs2_userspace_stack(osb)) {
 828                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 829                              "a read-write clustered device.\n");
 830                         return -EINVAL;
 831                 }
 832         }
 833
 834         return 0;
 835 }
 836
 837 /*
 838  * If we're using a userspace stack, mount should have passed
 839  * a name that matches the disk.  If not, mount should not
 840  * have passed a stack.
 841  */
 842 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 843                                         struct mount_options *mopt)
 844 {
 845         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 846                 mlog(ML_ERROR,
 847                      "cluster stack passed to mount, but this filesystem "
 848                      "does not support it\n");
 849                 return -EINVAL;
 850         }
 851
 852         if (ocfs2_userspace_stack(osb) &&
 853             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 854                     OCFS2_STACK_LABEL_LEN)) {
 855                 mlog(ML_ERROR,
 856                      "cluster stack passed to mount (\"%s\") does not "
 857                      "match the filesystem (\"%s\")\n",
 858                      mopt->cluster_stack,
 859                      osb->osb_cluster_stack);
 860                 return -EINVAL;
 861         }
 862
 863         return 0;
 864 }
 865
 866 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 867 {
 868         int type;
 869         struct super_block *sb = osb->sb;
 870         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 871                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 872         int status = 0;
 873
 874         for (type = 0; type < MAXQUOTAS; type++) {
 875                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 876                         continue;
 877                 if (unsuspend)
 878                         status = dquot_resume(sb, type);
 879                 else {
 880                         struct ocfs2_mem_dqinfo *oinfo;
 881
 882                         /* Cancel periodic syncing before suspending */
 883                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 884                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 885                         status = dquot_suspend(sb, type);
 886                 }
 887                 if (status < 0)
 888                         break;
 889         }
 890         if (status < 0)
 891                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 892                      "remount (error = %d).\n", status);
 893         return status;
 894 }
 895
 896 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 897 {
 898         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 899         struct super_block *sb = osb->sb;
 900         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 901                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 902         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 903                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 904         int status;
 905         int type;
 906
 907         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 908         for (type = 0; type < MAXQUOTAS; type++) {
 909                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 910                         continue;
 911                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 912                                                         osb->slot_num);
 913                 if (!inode[type]) {
 914                         status = -ENOENT;
 915                         goto out_quota_off;
 916                 }
 917                 status = dquot_enable(inode[type], type, QFMT_OCFS2,
 918                                       DQUOT_USAGE_ENABLED);
 919                 if (status < 0)
 920                         goto out_quota_off;
 921         }
 922
 923         for (type = 0; type < MAXQUOTAS; type++)
 924                 iput(inode[type]);
 925         return 0;
 926 out_quota_off:
 927         ocfs2_disable_quotas(osb);
 928         for (type = 0; type < MAXQUOTAS; type++)
 929                 iput(inode[type]);
 930         mlog_errno(status);
 931         return status;
 932 }
 933
 934 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 935 {
 936         int type;
 937         struct inode *inode;
 938         struct super_block *sb = osb->sb;
 939         struct ocfs2_mem_dqinfo *oinfo;
 940
 941         /* We mostly ignore errors in this function because there's not much
 942          * we can do when we see them */
 943         for (type = 0; type < MAXQUOTAS; type++) {
 944                 if (!sb_has_quota_loaded(sb, type))
 945                         continue;
 946                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 947                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 948                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 949                 inode = igrab(sb->s_dquot.files[type]);
 950                 /* Turn off quotas. This will remove all dquot structures from
 951                  * memory and so they will be automatically synced to global
 952                  * quota files */
 953                 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
 954                                         DQUOT_LIMITS_ENABLED);
 955                 if (!inode)
 956                         continue;
 957                 iput(inode);
 958         }
 959 }
 960
 961 /* Handle quota on quotactl */
 962 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 963                           char *path)
 964 {
 965         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 966                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 967
 968         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 969                 return -EINVAL;
 970
 971         return dquot_enable(sb_dqopt(sb)->files[type], type,
 972                             format_id, DQUOT_LIMITS_ENABLED);
 973 }
 974
 975 /* Handle quota off quotactl */
 976 static int ocfs2_quota_off(struct super_block *sb, int type)
 977 {
 978         return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
 979 }
 980
 981 static const struct quotactl_ops ocfs2_quotactl_ops = {
 982         .quota_on       = ocfs2_quota_on,
 983         .quota_off      = ocfs2_quota_off,
 984         .quota_sync     = dquot_quota_sync,
 985         .get_info       = dquot_get_dqinfo,
 986         .set_info       = dquot_set_dqinfo,
 987         .get_dqblk      = dquot_get_dqblk,
 988         .set_dqblk      = dquot_set_dqblk,
 989 };
 990
 991 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
 992 {
 993         struct dentry *root;
 994         int status, sector_size;
 995         struct mount_options parsed_options;
 996         struct inode *inode = NULL;
 997         struct ocfs2_super *osb = NULL;
 998         struct buffer_head *bh = NULL;
 999         char nodestr[8];
1000         struct ocfs2_blockcheck_stats stats;
1001
1002         mlog_entry("%p, %p, %i", sb, data, silent);
1003
1004         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1005                 status = -EINVAL;
1006                 goto read_super_error;
1007         }
1008
1009         /* probe for superblock */
1010         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1011         if (status < 0) {
1012                 mlog(ML_ERROR, "superblock probe failed!\n");
1013                 goto read_super_error;
1014         }
1015
1016         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1017         osb = OCFS2_SB(sb);
1018         if (status < 0) {
1019                 mlog_errno(status);
1020                 goto read_super_error;
1021         }
1022         brelse(bh);
1023         bh = NULL;
1024
1025         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1026                 status = -EINVAL;
1027                 goto read_super_error;
1028         }
1029         osb->s_mount_opt = parsed_options.mount_opt;
1030         osb->s_atime_quantum = parsed_options.atime_quantum;
1031         osb->preferred_slot = parsed_options.slot;
1032         osb->osb_commit_interval = parsed_options.commit_interval;
1033
1034         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1035         osb->osb_resv_level = parsed_options.resv_level;
1036         osb->osb_dir_resv_level = parsed_options.resv_level;
1037         if (parsed_options.dir_resv_level == -1)
1038                 osb->osb_dir_resv_level = parsed_options.resv_level;
1039         else
1040                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1041
1042         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1043         if (status)
1044                 goto read_super_error;
1045
1046         sb->s_magic = OCFS2_SUPER_MAGIC;
1047
1048         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1049                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1050
1051         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1052          * heartbeat=none */
1053         if (bdev_read_only(sb->s_bdev)) {
1054                 if (!(sb->s_flags & MS_RDONLY)) {
1055                         status = -EACCES;
1056                         mlog(ML_ERROR, "Readonly device detected but readonly "
1057                              "mount was not specified.\n");
1058                         goto read_super_error;
1059                 }
1060
1061                 /* You should not be able to start a local heartbeat
1062                  * on a readonly device. */
1063                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1064                         status = -EROFS;
1065                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1066                              "device.\n");
1067                         goto read_super_error;
1068                 }
1069
1070                 status = ocfs2_check_journals_nolocks(osb);
1071                 if (status < 0) {
1072                         if (status == -EROFS)
1073                                 mlog(ML_ERROR, "Recovery required on readonly "
1074                                      "file system, but write access is "
1075                                      "unavailable.\n");
1076                         else
1077                                 mlog_errno(status);
1078                         goto read_super_error;
1079                 }
1080
1081                 ocfs2_set_ro_flag(osb, 1);
1082
1083                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1084                        "services will be utilized for this mount. Recovery "
1085                        "will be skipped.\n");
1086         }
1087
1088         if (!ocfs2_is_hard_readonly(osb)) {
1089                 if (sb->s_flags & MS_RDONLY)
1090                         ocfs2_set_ro_flag(osb, 0);
1091         }
1092
1093         status = ocfs2_verify_heartbeat(osb);
1094         if (status < 0) {
1095                 mlog_errno(status);
1096                 goto read_super_error;
1097         }
1098
1099         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1100                                                  ocfs2_debugfs_root);
1101         if (!osb->osb_debug_root) {
1102                 status = -EINVAL;
1103                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1104                 goto read_super_error;
1105         }
1106
1107         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1108                                             osb->osb_debug_root,
1109                                             osb,
1110                                             &ocfs2_osb_debug_fops);
1111         if (!osb->osb_ctxt) {
1112                 status = -EINVAL;
1113                 mlog_errno(status);
1114                 goto read_super_error;
1115         }
1116
1117         if (ocfs2_meta_ecc(osb)) {
1118                 status = ocfs2_blockcheck_stats_debugfs_install(
1119                                                 &osb->osb_ecc_stats,
1120                                                 osb->osb_debug_root);
1121                 if (status) {
1122                         mlog(ML_ERROR,
1123                              "Unable to create blockcheck statistics "
1124                              "files\n");
1125                         goto read_super_error;
1126                 }
1127         }
1128
1129         status = ocfs2_mount_volume(sb);
1130         if (osb->root_inode)
1131                 inode = igrab(osb->root_inode);
1132
1133         if (status < 0)
1134                 goto read_super_error;
1135
1136         if (!inode) {
1137                 status = -EIO;
1138                 mlog_errno(status);
1139                 goto read_super_error;
1140         }
1141
1142         root = d_alloc_root(inode);
1143         if (!root) {
1144                 status = -ENOMEM;
1145                 mlog_errno(status);
1146                 goto read_super_error;
1147         }
1148
1149         sb->s_root = root;
1150
1151         ocfs2_complete_mount_recovery(osb);
1152
1153         if (ocfs2_mount_local(osb))
1154                 snprintf(nodestr, sizeof(nodestr), "local");
1155         else
1156                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1157
1158         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1159                "with %s data mode.\n",
1160                osb->dev_str, nodestr, osb->slot_num,
1161                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1162                "ordered");
1163
1164         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1165         wake_up(&osb->osb_mount_event);
1166
1167         /* Now we can initialize quotas because we can afford to wait
1168          * for cluster locks recovery now. That also means that truncation
1169          * log recovery can happen but that waits for proper quota setup */
1170         if (!(sb->s_flags & MS_RDONLY)) {
1171                 status = ocfs2_enable_quotas(osb);
1172                 if (status < 0) {
1173                         /* We have to err-out specially here because
1174                          * s_root is already set */
1175                         mlog_errno(status);
1176                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1177                         wake_up(&osb->osb_mount_event);
1178                         mlog_exit(status);
1179                         return status;
1180                 }
1181         }
1182
1183         ocfs2_complete_quota_recovery(osb);
1184
1185         /* Now we wake up again for processes waiting for quotas */
1186         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1187         wake_up(&osb->osb_mount_event);
1188
1189         /* Start this when the mount is almost sure of being successful */
1190         ocfs2_orphan_scan_start(osb);
1191
1192         mlog_exit(status);
1193         return status;
1194
1195 read_super_error:
1196         brelse(bh);
1197
1198         if (inode)
1199                 iput(inode);
1200
1201         if (osb) {
1202                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1203                 wake_up(&osb->osb_mount_event);
1204                 ocfs2_dismount_volume(sb, 1);
1205         }
1206
1207         mlog_exit(status);
1208         return status;
1209 }
1210
1211 static int ocfs2_get_sb(struct file_system_type *fs_type,
1212                         int flags,
1213                         const char *dev_name,
1214                         void *data,
1215                         struct vfsmount *mnt)
1216 {
1217         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1218                            mnt);
1219 }
1220
1221 static void ocfs2_kill_sb(struct super_block *sb)
1222 {
1223         struct ocfs2_super *osb = OCFS2_SB(sb);
1224
1225         /* Failed mount? */
1226         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1227                 goto out;
1228
1229         /* Prevent further queueing of inode drop events */
1230         spin_lock(&dentry_list_lock);
1231         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1232         spin_unlock(&dentry_list_lock);
1233         /* Wait for work to finish and/or remove it */
1234         cancel_work_sync(&osb->dentry_lock_work);
1235 out:
1236         kill_block_super(sb);
1237 }
1238
1239 static struct file_system_type ocfs2_fs_type = {
1240         .owner          = THIS_MODULE,
1241         .name           = "ocfs2",
1242         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1243                                         * the fs? */
1244         .kill_sb        = ocfs2_kill_sb,
1245
1246         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1247         .next           = NULL
1248 };
1249
1250 static int ocfs2_check_set_options(struct super_block *sb,
1251                                    struct mount_options *options)
1252 {
1253         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1254             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1255                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1256                 mlog(ML_ERROR, "User quotas were requested, but this "
1257                      "filesystem does not have the feature enabled.\n");
1258                 return 0;
1259         }
1260         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1261             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1262                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1263                 mlog(ML_ERROR, "Group quotas were requested, but this "
1264                      "filesystem does not have the feature enabled.\n");
1265                 return 0;
1266         }
1267         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1268             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1269                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1270                      "feature is not enabled\n");
1271                 return 0;
1272         }
1273         /* No ACL setting specified? Use XATTR feature... */
1274         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1275                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1276                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1277                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1278                 else
1279                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1280         }
1281         return 1;
1282 }
1283
1284 static int ocfs2_parse_options(struct super_block *sb,
1285                                char *options,
1286                                struct mount_options *mopt,
1287                                int is_remount)
1288 {
1289         int status;
1290         char *p;
1291
1292         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1293                    options ? options : "(none)");
1294
1295         mopt->commit_interval = 0;
1296         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1297         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1298         mopt->slot = OCFS2_INVALID_SLOT;
1299         mopt->localalloc_opt = -1;
1300         mopt->cluster_stack[0] = '\0';
1301         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1302         mopt->dir_resv_level = -1;
1303
1304         if (!options) {
1305                 status = 1;
1306                 goto bail;
1307         }
1308
1309         while ((p = strsep(&options, ",")) != NULL) {
1310                 int token, option;
1311                 substring_t args[MAX_OPT_ARGS];
1312
1313                 if (!*p)
1314                         continue;
1315
1316                 token = match_token(p, tokens, args);
1317                 switch (token) {
1318                 case Opt_hb_local:
1319                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1320                         break;
1321                 case Opt_hb_none:
1322                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1323                         break;
1324                 case Opt_barrier:
1325                         if (match_int(&args[0], &option)) {
1326                                 status = 0;
1327                                 goto bail;
1328                         }
1329                         if (option)
1330                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1331                         else
1332                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1333                         break;
1334                 case Opt_intr:
1335                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1336                         break;
1337                 case Opt_nointr:
1338                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1339                         break;
1340                 case Opt_err_panic:
1341                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1342                         break;
1343                 case Opt_err_ro:
1344                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1345                         break;
1346                 case Opt_data_ordered:
1347                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1348                         break;
1349                 case Opt_data_writeback:
1350                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1351                         break;
1352                 case Opt_user_xattr:
1353                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1354                         break;
1355                 case Opt_nouser_xattr:
1356                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1357                         break;
1358                 case Opt_atime_quantum:
1359                         if (match_int(&args[0], &option)) {
1360                                 status = 0;
1361                                 goto bail;
1362                         }
1363                         if (option >= 0)
1364                                 mopt->atime_quantum = option;
1365                         break;
1366                 case Opt_slot:
1367                         option = 0;
1368                         if (match_int(&args[0], &option)) {
1369                                 status = 0;
1370                                 goto bail;
1371                         }
1372                         if (option)
1373                                 mopt->slot = (s16)option;
1374                         break;
1375                 case Opt_commit:
1376                         option = 0;
1377                         if (match_int(&args[0], &option)) {
1378                                 status = 0;
1379                                 goto bail;
1380                         }
1381                         if (option < 0)
1382                                 return 0;
1383                         if (option == 0)
1384                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1385                         mopt->commit_interval = HZ * option;
1386                         break;
1387                 case Opt_localalloc:
1388                         option = 0;
1389                         if (match_int(&args[0], &option)) {
1390                                 status = 0;
1391                                 goto bail;
1392                         }
1393                         if (option >= 0)
1394                                 mopt->localalloc_opt = option;
1395                         break;
1396                 case Opt_localflocks:
1397                         /*
1398                          * Changing this during remount could race
1399                          * flock() requests, or "unbalance" existing
1400                          * ones (e.g., a lock is taken in one mode but
1401                          * dropped in the other). If users care enough
1402                          * to flip locking modes during remount, we
1403                          * could add a "local" flag to individual
1404                          * flock structures for proper tracking of
1405                          * state.
1406                          */
1407                         if (!is_remount)
1408                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1409                         break;
1410                 case Opt_stack:
1411                         /* Check both that the option we were passed
1412                          * is of the right length and that it is a proper
1413                          * string of the right length.
1414                          */
1415                         if (((args[0].to - args[0].from) !=
1416                              OCFS2_STACK_LABEL_LEN) ||
1417                             (strnlen(args[0].from,
1418                                      OCFS2_STACK_LABEL_LEN) !=
1419                              OCFS2_STACK_LABEL_LEN)) {
1420                                 mlog(ML_ERROR,
1421                                      "Invalid cluster_stack option\n");
1422                                 status = 0;
1423                                 goto bail;
1424                         }
1425                         memcpy(mopt->cluster_stack, args[0].from,
1426                                OCFS2_STACK_LABEL_LEN);
1427                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1428                         break;
1429                 case Opt_inode64:
1430                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1431                         break;
1432                 case Opt_usrquota:
1433                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1434                         break;
1435                 case Opt_grpquota:
1436                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1437                         break;
1438                 case Opt_acl:
1439                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1440                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1441                         break;
1442                 case Opt_noacl:
1443                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1444                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1445                         break;
1446                 case Opt_resv_level:
1447                         if (is_remount)
1448                                 break;
1449                         if (match_int(&args[0], &option)) {
1450                                 status = 0;
1451                                 goto bail;
1452                         }
1453                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1454                             option < OCFS2_MAX_RESV_LEVEL)
1455                                 mopt->resv_level = option;
1456                         break;
1457                 case Opt_dir_resv_level:
1458                         if (is_remount)
1459                                 break;
1460                         if (match_int(&args[0], &option)) {
1461                                 status = 0;
1462                                 goto bail;
1463                         }
1464                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1465                             option < OCFS2_MAX_RESV_LEVEL)
1466                                 mopt->dir_resv_level = option;
1467                         break;
1468                 default:
1469                         mlog(ML_ERROR,
1470                              "Unrecognized mount option \"%s\" "
1471                              "or missing value\n", p);
1472                         status = 0;
1473                         goto bail;
1474                 }
1475         }
1476
1477         status = 1;
1478
1479 bail:
1480         mlog_exit(status);
1481         return status;
1482 }
1483
1484 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1485 {
1486         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1487         unsigned long opts = osb->s_mount_opt;
1488         unsigned int local_alloc_megs;
1489
1490         if (opts & OCFS2_MOUNT_HB_LOCAL)
1491                 seq_printf(s, ",_netdev,heartbeat=local");
1492         else
1493                 seq_printf(s, ",heartbeat=none");
1494
1495         if (opts & OCFS2_MOUNT_NOINTR)
1496                 seq_printf(s, ",nointr");
1497
1498         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1499                 seq_printf(s, ",data=writeback");
1500         else
1501                 seq_printf(s, ",data=ordered");
1502
1503         if (opts & OCFS2_MOUNT_BARRIER)
1504                 seq_printf(s, ",barrier=1");
1505
1506         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1507                 seq_printf(s, ",errors=panic");
1508         else
1509                 seq_printf(s, ",errors=remount-ro");
1510
1511         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1512                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1513
1514         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1515                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1516
1517         if (osb->osb_commit_interval)
1518                 seq_printf(s, ",commit=%u",
1519                            (unsigned) (osb->osb_commit_interval / HZ));
1520
1521         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1522         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1523                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1524
1525         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1526                 seq_printf(s, ",localflocks,");
1527
1528         if (osb->osb_cluster_stack[0])
1529                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1530                            osb->osb_cluster_stack);
1531         if (opts & OCFS2_MOUNT_USRQUOTA)
1532                 seq_printf(s, ",usrquota");
1533         if (opts & OCFS2_MOUNT_GRPQUOTA)
1534                 seq_printf(s, ",grpquota");
1535
1536         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1537                 seq_printf(s, ",nouser_xattr");
1538         else
1539                 seq_printf(s, ",user_xattr");
1540
1541         if (opts & OCFS2_MOUNT_INODE64)
1542                 seq_printf(s, ",inode64");
1543
1544         if (opts & OCFS2_MOUNT_POSIX_ACL)
1545                 seq_printf(s, ",acl");
1546         else
1547                 seq_printf(s, ",noacl");
1548
1549         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1550                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1551
1552         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1553                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1554
1555         return 0;
1556 }
1557
1558 static int __init ocfs2_init(void)
1559 {
1560         int status;
1561
1562         mlog_entry_void();
1563
1564         ocfs2_print_version();
1565
1566         status = init_ocfs2_uptodate_cache();
1567         if (status < 0) {
1568                 mlog_errno(status);
1569                 goto leave;
1570         }
1571
1572         status = ocfs2_initialize_mem_caches();
1573         if (status < 0) {
1574                 mlog_errno(status);
1575                 goto leave;
1576         }
1577
1578         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1579         if (!ocfs2_wq) {
1580                 status = -ENOMEM;
1581                 goto leave;
1582         }
1583
1584         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1585         if (!ocfs2_debugfs_root) {
1586                 status = -EFAULT;
1587                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1588         }
1589
1590         status = ocfs2_quota_setup();
1591         if (status)
1592                 goto leave;
1593
1594         ocfs2_set_locking_protocol();
1595
1596         status = register_quota_format(&ocfs2_quota_format);
1597 leave:
1598         if (status < 0) {
1599                 ocfs2_quota_shutdown();
1600                 ocfs2_free_mem_caches();
1601                 exit_ocfs2_uptodate_cache();
1602         }
1603
1604         mlog_exit(status);
1605
1606         if (status >= 0) {
1607                 return register_filesystem(&ocfs2_fs_type);
1608         } else
1609                 return -1;
1610 }
1611
1612 static void __exit ocfs2_exit(void)
1613 {
1614         mlog_entry_void();
1615
1616         ocfs2_quota_shutdown();
1617
1618         if (ocfs2_wq) {
1619                 flush_workqueue(ocfs2_wq);
1620                 destroy_workqueue(ocfs2_wq);
1621         }
1622
1623         unregister_quota_format(&ocfs2_quota_format);
1624
1625         debugfs_remove(ocfs2_debugfs_root);
1626
1627         ocfs2_free_mem_caches();
1628
1629         unregister_filesystem(&ocfs2_fs_type);
1630
1631         exit_ocfs2_uptodate_cache();
1632
1633         mlog_exit_void();
1634 }
1635
1636 static void ocfs2_put_super(struct super_block *sb)
1637 {
1638         mlog_entry("(0x%p)\n", sb);
1639
1640         ocfs2_sync_blockdev(sb);
1641         ocfs2_dismount_volume(sb, 0);
1642
1643         mlog_exit_void();
1644 }
1645
1646 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1647 {
1648         struct ocfs2_super *osb;
1649         u32 numbits, freebits;
1650         int status;
1651         struct ocfs2_dinode *bm_lock;
1652         struct buffer_head *bh = NULL;
1653         struct inode *inode = NULL;
1654
1655         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1656
1657         osb = OCFS2_SB(dentry->d_sb);
1658
1659         inode = ocfs2_get_system_file_inode(osb,
1660                                             GLOBAL_BITMAP_SYSTEM_INODE,
1661                                             OCFS2_INVALID_SLOT);
1662         if (!inode) {
1663                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1664                 status = -EIO;
1665                 goto bail;
1666         }
1667
1668         status = ocfs2_inode_lock(inode, &bh, 0);
1669         if (status < 0) {
1670                 mlog_errno(status);
1671                 goto bail;
1672         }
1673
1674         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1675
1676         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1677         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1678
1679         buf->f_type = OCFS2_SUPER_MAGIC;
1680         buf->f_bsize = dentry->d_sb->s_blocksize;
1681         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1682         buf->f_blocks = ((sector_t) numbits) *
1683                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1684         buf->f_bfree = ((sector_t) freebits) *
1685                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1686         buf->f_bavail = buf->f_bfree;
1687         buf->f_files = numbits;
1688         buf->f_ffree = freebits;
1689         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1690                                 & 0xFFFFFFFFUL;
1691         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1692                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1693
1694         brelse(bh);
1695
1696         ocfs2_inode_unlock(inode, 0);
1697         status = 0;
1698 bail:
1699         if (inode)
1700                 iput(inode);
1701
1702         mlog_exit(status);
1703
1704         return status;
1705 }
1706
1707 static void ocfs2_inode_init_once(void *data)
1708 {
1709         struct ocfs2_inode_info *oi = data;
1710
1711         oi->ip_flags = 0;
1712         oi->ip_open_count = 0;
1713         spin_lock_init(&oi->ip_lock);
1714         ocfs2_extent_map_init(&oi->vfs_inode);
1715         INIT_LIST_HEAD(&oi->ip_io_markers);
1716         oi->ip_dir_start_lookup = 0;
1717
1718         init_rwsem(&oi->ip_alloc_sem);
1719         init_rwsem(&oi->ip_xattr_sem);
1720         mutex_init(&oi->ip_io_mutex);
1721
1722         oi->ip_blkno = 0ULL;
1723         oi->ip_clusters = 0;
1724
1725         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1726
1727         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1728         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1729         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1730
1731         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1732                                   &ocfs2_inode_caching_ops);
1733
1734         inode_init_once(&oi->vfs_inode);
1735 }
1736
1737 static int ocfs2_initialize_mem_caches(void)
1738 {
1739         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1740                                        sizeof(struct ocfs2_inode_info),
1741                                        0,
1742                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1743                                                 SLAB_MEM_SPREAD),
1744                                        ocfs2_inode_init_once);
1745         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1746                                         sizeof(struct ocfs2_dquot),
1747                                         0,
1748                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1749                                                 SLAB_MEM_SPREAD),
1750                                         NULL);
1751         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1752                                         sizeof(struct ocfs2_quota_chunk),
1753                                         0,
1754                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1755                                         NULL);
1756         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1757             !ocfs2_qf_chunk_cachep) {
1758                 if (ocfs2_inode_cachep)
1759                         kmem_cache_destroy(ocfs2_inode_cachep);
1760                 if (ocfs2_dquot_cachep)
1761                         kmem_cache_destroy(ocfs2_dquot_cachep);
1762                 if (ocfs2_qf_chunk_cachep)
1763                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1764                 return -ENOMEM;
1765         }
1766
1767         return 0;
1768 }
1769
1770 static void ocfs2_free_mem_caches(void)
1771 {
1772         if (ocfs2_inode_cachep)
1773                 kmem_cache_destroy(ocfs2_inode_cachep);
1774         ocfs2_inode_cachep = NULL;
1775
1776         if (ocfs2_dquot_cachep)
1777                 kmem_cache_destroy(ocfs2_dquot_cachep);
1778         ocfs2_dquot_cachep = NULL;
1779
1780         if (ocfs2_qf_chunk_cachep)
1781                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1782         ocfs2_qf_chunk_cachep = NULL;
1783 }
1784
1785 static int ocfs2_get_sector(struct super_block *sb,
1786                             struct buffer_head **bh,
1787                             int block,
1788                             int sect_size)
1789 {
1790         if (!sb_set_blocksize(sb, sect_size)) {
1791                 mlog(ML_ERROR, "unable to set blocksize\n");
1792                 return -EIO;
1793         }
1794
1795         *bh = sb_getblk(sb, block);
1796         if (!*bh) {
1797                 mlog_errno(-EIO);
1798                 return -EIO;
1799         }
1800         lock_buffer(*bh);
1801         if (!buffer_dirty(*bh))
1802                 clear_buffer_uptodate(*bh);
1803         unlock_buffer(*bh);
1804         ll_rw_block(READ, 1, bh);
1805         wait_on_buffer(*bh);
1806         if (!buffer_uptodate(*bh)) {
1807                 mlog_errno(-EIO);
1808                 brelse(*bh);
1809                 *bh = NULL;
1810                 return -EIO;
1811         }
1812
1813         return 0;
1814 }
1815
1816 static int ocfs2_mount_volume(struct super_block *sb)
1817 {
1818         int status = 0;
1819         int unlock_super = 0;
1820         struct ocfs2_super *osb = OCFS2_SB(sb);
1821
1822         mlog_entry_void();
1823
1824         if (ocfs2_is_hard_readonly(osb))
1825                 goto leave;
1826
1827         status = ocfs2_dlm_init(osb);
1828         if (status < 0) {
1829                 mlog_errno(status);
1830                 goto leave;
1831         }
1832
1833         status = ocfs2_super_lock(osb, 1);
1834         if (status < 0) {
1835                 mlog_errno(status);
1836                 goto leave;
1837         }
1838         unlock_super = 1;
1839
1840         /* This will load up the node map and add ourselves to it. */
1841         status = ocfs2_find_slot(osb);
1842         if (status < 0) {
1843                 mlog_errno(status);
1844                 goto leave;
1845         }
1846
1847         /* load all node-local system inodes */
1848         status = ocfs2_init_local_system_inodes(osb);
1849         if (status < 0) {
1850                 mlog_errno(status);
1851                 goto leave;
1852         }
1853
1854         status = ocfs2_check_volume(osb);
1855         if (status < 0) {
1856                 mlog_errno(status);
1857                 goto leave;
1858         }
1859
1860         status = ocfs2_truncate_log_init(osb);
1861         if (status < 0)
1862                 mlog_errno(status);
1863
1864 leave:
1865         if (unlock_super)
1866                 ocfs2_super_unlock(osb, 1);
1867
1868         mlog_exit(status);
1869         return status;
1870 }
1871
1872 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1873 {
1874         int tmp, hangup_needed = 0;
1875         struct ocfs2_super *osb = NULL;
1876         char nodestr[8];
1877
1878         mlog_entry("(0x%p)\n", sb);
1879
1880         BUG_ON(!sb);
1881         osb = OCFS2_SB(sb);
1882         BUG_ON(!osb);
1883
1884         debugfs_remove(osb->osb_ctxt);
1885
1886         /*
1887          * Flush inode dropping work queue so that deletes are
1888          * performed while the filesystem is still working
1889          */
1890         ocfs2_drop_all_dl_inodes(osb);
1891
1892         /* Orphan scan should be stopped as early as possible */
1893         ocfs2_orphan_scan_stop(osb);
1894
1895         ocfs2_disable_quotas(osb);
1896
1897         ocfs2_shutdown_local_alloc(osb);
1898
1899         ocfs2_truncate_log_shutdown(osb);
1900
1901         /* This will disable recovery and flush any recovery work. */
1902         ocfs2_recovery_exit(osb);
1903
1904         ocfs2_journal_shutdown(osb);
1905
1906         ocfs2_sync_blockdev(sb);
1907
1908         ocfs2_purge_refcount_trees(osb);
1909
1910         /* No cluster connection means we've failed during mount, so skip
1911          * all the steps which depended on that to complete. */
1912         if (osb->cconn) {
1913                 tmp = ocfs2_super_lock(osb, 1);
1914                 if (tmp < 0) {
1915                         mlog_errno(tmp);
1916                         return;
1917                 }
1918         }
1919
1920         if (osb->slot_num != OCFS2_INVALID_SLOT)
1921                 ocfs2_put_slot(osb);
1922
1923         if (osb->cconn)
1924                 ocfs2_super_unlock(osb, 1);
1925
1926         ocfs2_release_system_inodes(osb);
1927
1928         /*
1929          * If we're dismounting due to mount error, mount.ocfs2 will clean
1930          * up heartbeat.  If we're a local mount, there is no heartbeat.
1931          * If we failed before we got a uuid_str yet, we can't stop
1932          * heartbeat.  Otherwise, do it.
1933          */
1934         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1935                 hangup_needed = 1;
1936
1937         if (osb->cconn)
1938                 ocfs2_dlm_shutdown(osb, hangup_needed);
1939
1940         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1941         debugfs_remove(osb->osb_debug_root);
1942
1943         if (hangup_needed)
1944                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1945
1946         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1947
1948         if (ocfs2_mount_local(osb))
1949                 snprintf(nodestr, sizeof(nodestr), "local");
1950         else
1951                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1952
1953         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1954                osb->dev_str, nodestr);
1955
1956         ocfs2_delete_osb(osb);
1957         kfree(osb);
1958         sb->s_dev = 0;
1959         sb->s_fs_info = NULL;
1960 }
1961
1962 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1963                                 unsigned uuid_bytes)
1964 {
1965         int i, ret;
1966         char *ptr;
1967
1968         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1969
1970         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1971         if (osb->uuid_str == NULL)
1972                 return -ENOMEM;
1973
1974         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1975                 /* print with null */
1976                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1977                 if (ret != 2) /* drop super cleans up */
1978                         return -EINVAL;
1979                 /* then only advance past the last char */
1980                 ptr += 2;
1981         }
1982
1983         return 0;
1984 }
1985
1986 static int ocfs2_initialize_super(struct super_block *sb,
1987                                   struct buffer_head *bh,
1988                                   int sector_size,
1989                                   struct ocfs2_blockcheck_stats *stats)
1990 {
1991         int status;
1992         int i, cbits, bbits;
1993         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1994         struct inode *inode = NULL;
1995         struct ocfs2_journal *journal;
1996         __le32 uuid_net_key;
1997         struct ocfs2_super *osb;
1998
1999         mlog_entry_void();
2000
2001         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2002         if (!osb) {
2003                 status = -ENOMEM;
2004                 mlog_errno(status);
2005                 goto bail;
2006         }
2007
2008         sb->s_fs_info = osb;
2009         sb->s_op = &ocfs2_sops;
2010         sb->s_export_op = &ocfs2_export_ops;
2011         sb->s_qcop = &ocfs2_quotactl_ops;
2012         sb->dq_op = &ocfs2_quota_operations;
2013         sb->s_xattr = ocfs2_xattr_handlers;
2014         sb->s_time_gran = 1;
2015         sb->s_flags |= MS_NOATIME;
2016         /* this is needed to support O_LARGEFILE */
2017         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2018         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2019         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2020
2021         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2022
2023         for (i = 0; i < 3; i++)
2024                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2025         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2026
2027         osb->sb = sb;
2028         /* Save off for ocfs2_rw_direct */
2029         osb->s_sectsize_bits = blksize_bits(sector_size);
2030         BUG_ON(!osb->s_sectsize_bits);
2031
2032         spin_lock_init(&osb->dc_task_lock);
2033         init_waitqueue_head(&osb->dc_event);
2034         osb->dc_work_sequence = 0;
2035         osb->dc_wake_sequence = 0;
2036         INIT_LIST_HEAD(&osb->blocked_lock_list);
2037         osb->blocked_lock_count = 0;
2038         spin_lock_init(&osb->osb_lock);
2039         spin_lock_init(&osb->osb_xattr_lock);
2040         ocfs2_init_steal_slots(osb);
2041
2042         atomic_set(&osb->alloc_stats.moves, 0);
2043         atomic_set(&osb->alloc_stats.local_data, 0);
2044         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2045         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2046         atomic_set(&osb->alloc_stats.bg_extends, 0);
2047
2048         /* Copy the blockcheck stats from the superblock probe */
2049         osb->osb_ecc_stats = *stats;
2050
2051         ocfs2_init_node_maps(osb);
2052
2053         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2054                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2055
2056         ocfs2_orphan_scan_init(osb);
2057
2058         status = ocfs2_recovery_init(osb);
2059         if (status) {
2060                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2061                 mlog_errno(status);
2062                 goto bail;
2063         }
2064
2065         init_waitqueue_head(&osb->checkpoint_event);
2066         atomic_set(&osb->needs_checkpoint, 0);
2067
2068         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2069
2070         osb->slot_num = OCFS2_INVALID_SLOT;
2071
2072         osb->s_xattr_inline_size = le16_to_cpu(
2073                                         di->id2.i_super.s_xattr_inline_size);
2074
2075         osb->local_alloc_state = OCFS2_LA_UNUSED;
2076         osb->local_alloc_bh = NULL;
2077         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2078
2079         init_waitqueue_head(&osb->osb_mount_event);
2080
2081         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2082         if (status) {
2083                 mlog_errno(status);
2084                 goto bail;
2085         }
2086
2087         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2088         if (!osb->vol_label) {
2089                 mlog(ML_ERROR, "unable to alloc vol label\n");
2090                 status = -ENOMEM;
2091                 goto bail;
2092         }
2093
2094         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2095         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2096                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2097                      osb->max_slots);
2098                 status = -EINVAL;
2099                 goto bail;
2100         }
2101         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2102
2103         osb->slot_recovery_generations =
2104                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2105                         GFP_KERNEL);
2106         if (!osb->slot_recovery_generations) {
2107                 status = -ENOMEM;
2108                 mlog_errno(status);
2109                 goto bail;
2110         }
2111
2112         init_waitqueue_head(&osb->osb_wipe_event);
2113         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2114                                         sizeof(*osb->osb_orphan_wipes),
2115                                         GFP_KERNEL);
2116         if (!osb->osb_orphan_wipes) {
2117                 status = -ENOMEM;
2118                 mlog_errno(status);
2119                 goto bail;
2120         }
2121
2122         osb->osb_rf_lock_tree = RB_ROOT;
2123
2124         osb->s_feature_compat =
2125                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2126         osb->s_feature_ro_compat =
2127                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2128         osb->s_feature_incompat =
2129                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2130
2131         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2132                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2133                      "optional features (%x).\n", i);
2134                 status = -EINVAL;
2135                 goto bail;
2136         }
2137         if (!(osb->sb->s_flags & MS_RDONLY) &&
2138             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2139                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2140                      "unsupported optional features (%x).\n", i);
2141                 status = -EINVAL;
2142                 goto bail;
2143         }
2144
2145         if (ocfs2_userspace_stack(osb)) {
2146                 memcpy(osb->osb_cluster_stack,
2147                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2148                        OCFS2_STACK_LABEL_LEN);
2149                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2150                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2151                         mlog(ML_ERROR,
2152                              "couldn't mount because of an invalid "
2153                              "cluster stack label (%s) \n",
2154                              osb->osb_cluster_stack);
2155                         status = -EINVAL;
2156                         goto bail;
2157                 }
2158         } else {
2159                 /* The empty string is identical with classic tools that
2160                  * don't know about s_cluster_info. */
2161                 osb->osb_cluster_stack[0] = '\0';
2162         }
2163
2164         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2165
2166         /* FIXME
2167          * This should be done in ocfs2_journal_init(), but unknown
2168          * ordering issues will cause the filesystem to crash.
2169          * If anyone wants to figure out what part of the code
2170          * refers to osb->journal before ocfs2_journal_init() is run,
2171          * be my guest.
2172          */
2173         /* initialize our journal structure */
2174
2175         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2176         if (!journal) {
2177                 mlog(ML_ERROR, "unable to alloc journal\n");
2178                 status = -ENOMEM;
2179                 goto bail;
2180         }
2181         osb->journal = journal;
2182         journal->j_osb = osb;
2183
2184         atomic_set(&journal->j_num_trans, 0);
2185         init_rwsem(&journal->j_trans_barrier);
2186         init_waitqueue_head(&journal->j_checkpointed);
2187         spin_lock_init(&journal->j_lock);
2188         journal->j_trans_id = (unsigned long) 1;
2189         INIT_LIST_HEAD(&journal->j_la_cleanups);
2190         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2191         journal->j_state = OCFS2_JOURNAL_FREE;
2192
2193         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2194         osb->dentry_lock_list = NULL;
2195
2196         /* get some pseudo constants for clustersize bits */
2197         osb->s_clustersize_bits =
2198                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2199         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2200         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2201
2202         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2203             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2204                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2205                      osb->s_clustersize);
2206                 status = -EINVAL;
2207                 goto bail;
2208         }
2209
2210         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2211             > (u32)~0UL) {
2212                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2213                      "what jbd can address in 32 bits.\n");
2214                 status = -EINVAL;
2215                 goto bail;
2216         }
2217
2218         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2219                                  sizeof(di->id2.i_super.s_uuid))) {
2220                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2221                 status = -ENOMEM;
2222                 goto bail;
2223         }
2224
2225         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2226
2227         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2228         osb->vol_label[63] = '\0';
2229         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2230         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2231         osb->first_cluster_group_blkno =
2232                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2233         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2234         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2235         mlog(0, "vol_label: %s\n", osb->vol_label);
2236         mlog(0, "uuid: %s\n", osb->uuid_str);
2237         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2238              (unsigned long long)osb->root_blkno,
2239              (unsigned long long)osb->system_dir_blkno);
2240
2241         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2242         if (!osb->osb_dlm_debug) {
2243                 status = -ENOMEM;
2244                 mlog_errno(status);
2245                 goto bail;
2246         }
2247
2248         atomic_set(&osb->vol_state, VOLUME_INIT);
2249
2250         /* load root, system_dir, and all global system inodes */
2251         status = ocfs2_init_global_system_inodes(osb);
2252         if (status < 0) {
2253                 mlog_errno(status);
2254                 goto bail;
2255         }
2256
2257         /*
2258          * global bitmap
2259          */
2260         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2261                                             OCFS2_INVALID_SLOT);
2262         if (!inode) {
2263                 status = -EINVAL;
2264                 mlog_errno(status);
2265                 goto bail;
2266         }
2267
2268         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2269         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2270         iput(inode);
2271
2272         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2273                                  osb->s_feature_incompat) * 8;
2274
2275         status = ocfs2_init_slot_info(osb);
2276         if (status < 0) {
2277                 mlog_errno(status);
2278                 goto bail;
2279         }
2280
2281 bail:
2282         mlog_exit(status);
2283         return status;
2284 }
2285
2286 /*
2287  * will return: -EAGAIN if it is ok to keep searching for superblocks
2288  *              -EINVAL if there is a bad superblock
2289  *              0 on success
2290  */
2291 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2292                                struct buffer_head *bh,
2293                                u32 blksz,
2294                                struct ocfs2_blockcheck_stats *stats)
2295 {
2296         int status = -EAGAIN;
2297
2298         mlog_entry_void();
2299
2300         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2301                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2302                 /* We have to do a raw check of the feature here */
2303                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2304                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2305                         status = ocfs2_block_check_validate(bh->b_data,
2306                                                             bh->b_size,
2307                                                             &di->i_check,
2308                                                             stats);
2309                         if (status)
2310                                 goto out;
2311                 }
2312                 status = -EINVAL;
2313                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2314                         mlog(ML_ERROR, "found superblock with incorrect block "
2315                              "size: found %u, should be %u\n",
2316                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2317                                blksz);
2318                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2319                            OCFS2_MAJOR_REV_LEVEL ||
2320                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2321                            OCFS2_MINOR_REV_LEVEL) {
2322                         mlog(ML_ERROR, "found superblock with bad version: "
2323                              "found %u.%u, should be %u.%u\n",
2324                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2325                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2326                              OCFS2_MAJOR_REV_LEVEL,
2327                              OCFS2_MINOR_REV_LEVEL);
2328                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2329                         mlog(ML_ERROR, "bad block number on superblock: "
2330                              "found %llu, should be %llu\n",
2331                              (unsigned long long)le64_to_cpu(di->i_blkno),
2332                              (unsigned long long)bh->b_blocknr);
2333                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2334                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2335                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2336                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2337                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2338                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2339                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2340                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2341                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2342                         mlog(ML_ERROR,
2343                              "Superblock slots found greater than file system "
2344                              "maximum: found %u, max %u\n",
2345                              le16_to_cpu(di->id2.i_super.s_max_slots),
2346                              OCFS2_MAX_SLOTS);
2347                 } else {
2348                         /* found it! */
2349                         status = 0;
2350                 }
2351         }
2352
2353 out:
2354         mlog_exit(status);
2355         return status;
2356 }
2357
2358 static int ocfs2_check_volume(struct ocfs2_super *osb)
2359 {
2360         int status;
2361         int dirty;
2362         int local;
2363         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2364                                                   * recover
2365                                                   * ourselves. */
2366
2367         mlog_entry_void();
2368
2369         /* Init our journal object. */
2370         status = ocfs2_journal_init(osb->journal, &dirty);
2371         if (status < 0) {
2372                 mlog(ML_ERROR, "Could not initialize journal!\n");
2373                 goto finally;
2374         }
2375
2376         /* If the journal was unmounted cleanly then we don't want to
2377          * recover anything. Otherwise, journal_load will do that
2378          * dirty work for us :) */
2379         if (!dirty) {
2380                 status = ocfs2_journal_wipe(osb->journal, 0);
2381                 if (status < 0) {
2382                         mlog_errno(status);
2383                         goto finally;
2384                 }
2385         } else {
2386                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2387                      "recovering volume.\n");
2388         }
2389
2390         local = ocfs2_mount_local(osb);
2391
2392         /* will play back anything left in the journal. */
2393         status = ocfs2_journal_load(osb->journal, local, dirty);
2394         if (status < 0) {
2395                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2396                 goto finally;
2397         }
2398
2399         if (dirty) {
2400                 /* recover my local alloc if we didn't unmount cleanly. */
2401                 status = ocfs2_begin_local_alloc_recovery(osb,
2402                                                           osb->slot_num,
2403                                                           &local_alloc);
2404                 if (status < 0) {
2405                         mlog_errno(status);
2406                         goto finally;
2407                 }
2408                 /* we complete the recovery process after we've marked
2409                  * ourselves as mounted. */
2410         }
2411
2412         mlog(0, "Journal loaded.\n");
2413
2414         status = ocfs2_load_local_alloc(osb);
2415         if (status < 0) {
2416                 mlog_errno(status);
2417                 goto finally;
2418         }
2419
2420         if (dirty) {
2421                 /* Recovery will be completed after we've mounted the
2422                  * rest of the volume. */
2423                 osb->dirty = 1;
2424                 osb->local_alloc_copy = local_alloc;
2425                 local_alloc = NULL;
2426         }
2427
2428         /* go through each journal, trylock it and if you get the
2429          * lock, and it's marked as dirty, set the bit in the recover
2430          * map and launch a recovery thread for it. */
2431         status = ocfs2_mark_dead_nodes(osb);
2432         if (status < 0) {
2433                 mlog_errno(status);
2434                 goto finally;
2435         }
2436
2437         status = ocfs2_compute_replay_slots(osb);
2438         if (status < 0)
2439                 mlog_errno(status);
2440
2441 finally:
2442         if (local_alloc)
2443                 kfree(local_alloc);
2444
2445         mlog_exit(status);
2446         return status;
2447 }
2448
2449 /*
2450  * The routine gets called from dismount or close whenever a dismount on
2451  * volume is requested and the osb open count becomes 1.
2452  * It will remove the osb from the global list and also free up all the
2453  * initialized resources and fileobject.
2454  */
2455 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2456 {
2457         mlog_entry_void();
2458
2459         /* This function assumes that the caller has the main osb resource */
2460
2461         ocfs2_free_slot_info(osb);
2462
2463         kfree(osb->osb_orphan_wipes);
2464         kfree(osb->slot_recovery_generations);
2465         /* FIXME
2466          * This belongs in journal shutdown, but because we have to
2467          * allocate osb->journal at the start of ocfs2_initialize_osb(),
2468          * we free it here.
2469          */
2470         kfree(osb->journal);
2471         if (osb->local_alloc_copy)
2472                 kfree(osb->local_alloc_copy);
2473         kfree(osb->uuid_str);
2474         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2475         memset(osb, 0, sizeof(struct ocfs2_super));
2476
2477         mlog_exit_void();
2478 }
2479
2480 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2481  * panic(). We do not support continue-on-error operation. */
2482 static void ocfs2_handle_error(struct super_block *sb)
2483 {
2484         struct ocfs2_super *osb = OCFS2_SB(sb);
2485
2486         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2487                 panic("OCFS2: (device %s): panic forced after error\n",
2488                       sb->s_id);
2489
2490         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2491
2492         if (sb->s_flags & MS_RDONLY &&
2493             (ocfs2_is_soft_readonly(osb) ||
2494              ocfs2_is_hard_readonly(osb)))
2495                 return;
2496
2497         printk(KERN_CRIT "File system is now read-only due to the potential "
2498                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2499                "system is unmounted.\n");
2500         sb->s_flags |= MS_RDONLY;
2501         ocfs2_set_ro_flag(osb, 0);
2502 }
2503
2504 static char error_buf[1024];
2505
2506 void __ocfs2_error(struct super_block *sb,
2507                    const char *function,
2508                    const char *fmt, ...)
2509 {
2510         va_list args;
2511
2512         va_start(args, fmt);
2513         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2514         va_end(args);
2515
2516         /* Not using mlog here because we want to show the actual
2517          * function the error came from. */
2518         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2519                sb->s_id, function, error_buf);
2520
2521         ocfs2_handle_error(sb);
2522 }
2523
2524 /* Handle critical errors. This is intentionally more drastic than
2525  * ocfs2_handle_error, so we only use for things like journal errors,
2526  * etc. */
2527 void __ocfs2_abort(struct super_block* sb,
2528                    const char *function,
2529                    const char *fmt, ...)
2530 {
2531         va_list args;
2532
2533         va_start(args, fmt);
2534         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2535         va_end(args);
2536
2537         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2538                sb->s_id, function, error_buf);
2539
2540         /* We don't have the cluster support yet to go straight to
2541          * hard readonly in here. Until then, we want to keep
2542          * ocfs2_abort() so that we can at least mark critical
2543          * errors.
2544          *
2545          * TODO: This should abort the journal and alert other nodes
2546          * that our slot needs recovery. */
2547
2548         /* Force a panic(). This stinks, but it's better than letting
2549          * things continue without having a proper hard readonly
2550          * here. */
2551         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2552                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2553         ocfs2_handle_error(sb);
2554 }
2555
2556 /*
2557  * Void signal blockers, because in-kernel sigprocmask() only fails
2558  * when SIG_* is wrong.
2559  */
2560 void ocfs2_block_signals(sigset_t *oldset)
2561 {
2562         int rc;
2563         sigset_t blocked;
2564
2565         sigfillset(&blocked);
2566         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2567         BUG_ON(rc);
2568 }
2569
2570 void ocfs2_unblock_signals(sigset_t *oldset)
2571 {
2572         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2573         BUG_ON(rc);
2574 }
2575
2576 module_init(ocfs2_init);
2577 module_exit(ocfs2_exit);