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         .clear_inode    = ocfs2_clear_inode,
 149         .delete_inode   = ocfs2_delete_inode,
 150         .sync_fs        = ocfs2_sync_fs,
 151         .put_super      = ocfs2_put_super,
 152         .remount_fs     = ocfs2_remount,
 153         .show_options   = ocfs2_show_options,
 154         .quota_read     = ocfs2_quota_read,
 155         .quota_write    = ocfs2_quota_write,
 156 };
 157
 158 enum {
 159         Opt_barrier,
 160         Opt_err_panic,
 161         Opt_err_ro,
 162         Opt_intr,
 163         Opt_nointr,
 164         Opt_hb_none,
 165         Opt_hb_local,
 166         Opt_data_ordered,
 167         Opt_data_writeback,
 168         Opt_atime_quantum,
 169         Opt_slot,
 170         Opt_commit,
 171         Opt_localalloc,
 172         Opt_localflocks,
 173         Opt_stack,
 174         Opt_user_xattr,
 175         Opt_nouser_xattr,
 176         Opt_inode64,
 177         Opt_acl,
 178         Opt_noacl,
 179         Opt_usrquota,
 180         Opt_grpquota,
 181         Opt_resv_level,
 182         Opt_dir_resv_level,
 183         Opt_err,
 184 };
 185
 186 static const match_table_t tokens = {
 187         {Opt_barrier, "barrier=%u"},
 188         {Opt_err_panic, "errors=panic"},
 189         {Opt_err_ro, "errors=remount-ro"},
 190         {Opt_intr, "intr"},
 191         {Opt_nointr, "nointr"},
 192         {Opt_hb_none, OCFS2_HB_NONE},
 193         {Opt_hb_local, OCFS2_HB_LOCAL},
 194         {Opt_data_ordered, "data=ordered"},
 195         {Opt_data_writeback, "data=writeback"},
 196         {Opt_atime_quantum, "atime_quantum=%u"},
 197         {Opt_slot, "preferred_slot=%u"},
 198         {Opt_commit, "commit=%u"},
 199         {Opt_localalloc, "localalloc=%d"},
 200         {Opt_localflocks, "localflocks"},
 201         {Opt_stack, "cluster_stack=%s"},
 202         {Opt_user_xattr, "user_xattr"},
 203         {Opt_nouser_xattr, "nouser_xattr"},
 204         {Opt_inode64, "inode64"},
 205         {Opt_acl, "acl"},
 206         {Opt_noacl, "noacl"},
 207         {Opt_usrquota, "usrquota"},
 208         {Opt_grpquota, "grpquota"},
 209         {Opt_resv_level, "resv_level=%u"},
 210         {Opt_dir_resv_level, "dir_resv_level=%u"},
 211         {Opt_err, NULL}
 212 };
 213
 214 #ifdef CONFIG_DEBUG_FS
 215 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
 216 {
 217         struct ocfs2_cluster_connection *cconn = osb->cconn;
 218         struct ocfs2_recovery_map *rm = osb->recovery_map;
 219         struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
 220         int i, out = 0;
 221
 222         out += snprintf(buf + out, len - out,
 223                         "%10s => Id: %-s  Uuid: %-s  Gen: 0x%X  Label: %-s\n",
 224                         "Device", osb->dev_str, osb->uuid_str,
 225                         osb->fs_generation, osb->vol_label);
 226
 227         out += snprintf(buf + out, len - out,
 228                         "%10s => State: %d  Flags: 0x%lX\n", "Volume",
 229                         atomic_read(&osb->vol_state), osb->osb_flags);
 230
 231         out += snprintf(buf + out, len - out,
 232                         "%10s => Block: %lu  Cluster: %d\n", "Sizes",
 233                         osb->sb->s_blocksize, osb->s_clustersize);
 234
 235         out += snprintf(buf + out, len - out,
 236                         "%10s => Compat: 0x%X  Incompat: 0x%X  "
 237                         "ROcompat: 0x%X\n",
 238                         "Features", osb->s_feature_compat,
 239                         osb->s_feature_incompat, osb->s_feature_ro_compat);
 240
 241         out += snprintf(buf + out, len - out,
 242                         "%10s => Opts: 0x%lX  AtimeQuanta: %u\n", "Mount",
 243                         osb->s_mount_opt, osb->s_atime_quantum);
 244
 245         if (cconn) {
 246                 out += snprintf(buf + out, len - out,
 247                                 "%10s => Stack: %s  Name: %*s  "
 248                                 "Version: %d.%d\n", "Cluster",
 249                                 (*osb->osb_cluster_stack == '\0' ?
 250                                  "o2cb" : osb->osb_cluster_stack),
 251                                 cconn->cc_namelen, cconn->cc_name,
 252                                 cconn->cc_version.pv_major,
 253                                 cconn->cc_version.pv_minor);
 254         }
 255
 256         spin_lock(&osb->dc_task_lock);
 257         out += snprintf(buf + out, len - out,
 258                         "%10s => Pid: %d  Count: %lu  WakeSeq: %lu  "
 259                         "WorkSeq: %lu\n", "DownCnvt",
 260                         (osb->dc_task ?  task_pid_nr(osb->dc_task) : -1),
 261                         osb->blocked_lock_count, osb->dc_wake_sequence,
 262                         osb->dc_work_sequence);
 263         spin_unlock(&osb->dc_task_lock);
 264
 265         spin_lock(&osb->osb_lock);
 266         out += snprintf(buf + out, len - out, "%10s => Pid: %d  Nodes:",
 267                         "Recovery",
 268                         (osb->recovery_thread_task ?
 269                          task_pid_nr(osb->recovery_thread_task) : -1));
 270         if (rm->rm_used == 0)
 271                 out += snprintf(buf + out, len - out, " None\n");
 272         else {
 273                 for (i = 0; i < rm->rm_used; i++)
 274                         out += snprintf(buf + out, len - out, " %d",
 275                                         rm->rm_entries[i]);
 276                 out += snprintf(buf + out, len - out, "\n");
 277         }
 278         spin_unlock(&osb->osb_lock);
 279
 280         out += snprintf(buf + out, len - out,
 281                         "%10s => Pid: %d  Interval: %lu  Needs: %d\n", "Commit",
 282                         (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
 283                         osb->osb_commit_interval,
 284                         atomic_read(&osb->needs_checkpoint));
 285
 286         out += snprintf(buf + out, len - out,
 287                         "%10s => State: %d  TxnId: %lu  NumTxns: %d\n",
 288                         "Journal", osb->journal->j_state,
 289                         osb->journal->j_trans_id,
 290                         atomic_read(&osb->journal->j_num_trans));
 291
 292         out += snprintf(buf + out, len - out,
 293                         "%10s => GlobalAllocs: %d  LocalAllocs: %d  "
 294                         "SubAllocs: %d  LAWinMoves: %d  SAExtends: %d\n",
 295                         "Stats",
 296                         atomic_read(&osb->alloc_stats.bitmap_data),
 297                         atomic_read(&osb->alloc_stats.local_data),
 298                         atomic_read(&osb->alloc_stats.bg_allocs),
 299                         atomic_read(&osb->alloc_stats.moves),
 300                         atomic_read(&osb->alloc_stats.bg_extends));
 301
 302         out += snprintf(buf + out, len - out,
 303                         "%10s => State: %u  Descriptor: %llu  Size: %u bits  "
 304                         "Default: %u bits\n",
 305                         "LocalAlloc", osb->local_alloc_state,
 306                         (unsigned long long)osb->la_last_gd,
 307                         osb->local_alloc_bits, osb->local_alloc_default_bits);
 308
 309         spin_lock(&osb->osb_lock);
 310         out += snprintf(buf + out, len - out,
 311                         "%10s => InodeSlot: %d  StolenInodes: %d, "
 312                         "MetaSlot: %d  StolenMeta: %d\n", "Steal",
 313                         osb->s_inode_steal_slot,
 314                         atomic_read(&osb->s_num_inodes_stolen),
 315                         osb->s_meta_steal_slot,
 316                         atomic_read(&osb->s_num_meta_stolen));
 317         spin_unlock(&osb->osb_lock);
 318
 319         out += snprintf(buf + out, len - out, "OrphanScan => ");
 320         out += snprintf(buf + out, len - out, "Local: %u  Global: %u ",
 321                         os->os_count, os->os_seqno);
 322         out += snprintf(buf + out, len - out, " Last Scan: ");
 323         if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
 324                 out += snprintf(buf + out, len - out, "Disabled\n");
 325         else
 326                 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
 327                                 (get_seconds() - os->os_scantime.tv_sec));
 328
 329         out += snprintf(buf + out, len - out, "%10s => %3s  %10s\n",
 330                         "Slots", "Num", "RecoGen");
 331         for (i = 0; i < osb->max_slots; ++i) {
 332                 out += snprintf(buf + out, len - out,
 333                                 "%10s  %c %3d  %10d\n",
 334                                 " ",
 335                                 (i == osb->slot_num ? '*' : ' '),
 336                                 i, osb->slot_recovery_generations[i]);
 337         }
 338
 339         return out;
 340 }
 341
 342 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 343 {
 344         struct ocfs2_super *osb = inode->i_private;
 345         char *buf = NULL;
 346
 347         buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
 348         if (!buf)
 349                 goto bail;
 350
 351         i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
 352
 353         file->private_data = buf;
 354
 355         return 0;
 356 bail:
 357         return -ENOMEM;
 358 }
 359
 360 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 361 {
 362         kfree(file->private_data);
 363         return 0;
 364 }
 365
 366 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 367                                 size_t nbytes, loff_t *ppos)
 368 {
 369         return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
 370                                        i_size_read(file->f_mapping->host));
 371 }
 372 #else
 373 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
 374 {
 375         return 0;
 376 }
 377 static int ocfs2_debug_release(struct inode *inode, struct file *file)
 378 {
 379         return 0;
 380 }
 381 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
 382                                 size_t nbytes, loff_t *ppos)
 383 {
 384         return 0;
 385 }
 386 #endif  /* CONFIG_DEBUG_FS */
 387
 388 static const struct file_operations ocfs2_osb_debug_fops = {
 389         .open =         ocfs2_osb_debug_open,
 390         .release =      ocfs2_debug_release,
 391         .read =         ocfs2_debug_read,
 392         .llseek =       generic_file_llseek,
 393 };
 394
 395 static int ocfs2_sync_fs(struct super_block *sb, int wait)
 396 {
 397         int status;
 398         tid_t target;
 399         struct ocfs2_super *osb = OCFS2_SB(sb);
 400
 401         if (ocfs2_is_hard_readonly(osb))
 402                 return -EROFS;
 403
 404         if (wait) {
 405                 status = ocfs2_flush_truncate_log(osb);
 406                 if (status < 0)
 407                         mlog_errno(status);
 408         } else {
 409                 ocfs2_schedule_truncate_log_flush(osb, 0);
 410         }
 411
 412         if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
 413                                       &target)) {
 414                 if (wait)
 415                         jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
 416                                              target);
 417         }
 418         return 0;
 419 }
 420
 421 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
 422 {
 423         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
 424             && (ino == USER_QUOTA_SYSTEM_INODE
 425                 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
 426                 return 0;
 427         if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
 428             && (ino == GROUP_QUOTA_SYSTEM_INODE
 429                 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
 430                 return 0;
 431         return 1;
 432 }
 433
 434 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
 435 {
 436         struct inode *new = NULL;
 437         int status = 0;
 438         int i;
 439
 440         mlog_entry_void();
 441
 442         new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 443         if (IS_ERR(new)) {
 444                 status = PTR_ERR(new);
 445                 mlog_errno(status);
 446                 goto bail;
 447         }
 448         osb->root_inode = new;
 449
 450         new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
 451         if (IS_ERR(new)) {
 452                 status = PTR_ERR(new);
 453                 mlog_errno(status);
 454                 goto bail;
 455         }
 456         osb->sys_root_inode = new;
 457
 458         for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
 459              i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
 460                 if (!ocfs2_need_system_inode(osb, i))
 461                         continue;
 462                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 463                 if (!new) {
 464                         ocfs2_release_system_inodes(osb);
 465                         status = -EINVAL;
 466                         mlog_errno(status);
 467                         /* FIXME: Should ERROR_RO_FS */
 468                         mlog(ML_ERROR, "Unable to load system inode %d, "
 469                              "possibly corrupt fs?", i);
 470                         goto bail;
 471                 }
 472                 // the array now has one ref, so drop this one
 473                 iput(new);
 474         }
 475
 476 bail:
 477         mlog_exit(status);
 478         return status;
 479 }
 480
 481 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
 482 {
 483         struct inode *new = NULL;
 484         int status = 0;
 485         int i;
 486
 487         mlog_entry_void();
 488
 489         for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
 490              i < NUM_SYSTEM_INODES;
 491              i++) {
 492                 if (!ocfs2_need_system_inode(osb, i))
 493                         continue;
 494                 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
 495                 if (!new) {
 496                         ocfs2_release_system_inodes(osb);
 497                         status = -EINVAL;
 498                         mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
 499                              status, i, osb->slot_num);
 500                         goto bail;
 501                 }
 502                 /* the array now has one ref, so drop this one */
 503                 iput(new);
 504         }
 505
 506 bail:
 507         mlog_exit(status);
 508         return status;
 509 }
 510
 511 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
 512 {
 513         int i;
 514         struct inode *inode;
 515
 516         mlog_entry_void();
 517
 518         for (i = 0; i < NUM_SYSTEM_INODES; i++) {
 519                 inode = osb->system_inodes[i];
 520                 if (inode) {
 521                         iput(inode);
 522                         osb->system_inodes[i] = NULL;
 523                 }
 524         }
 525
 526         inode = osb->sys_root_inode;
 527         if (inode) {
 528                 iput(inode);
 529                 osb->sys_root_inode = NULL;
 530         }
 531
 532         inode = osb->root_inode;
 533         if (inode) {
 534                 iput(inode);
 535                 osb->root_inode = NULL;
 536         }
 537
 538         mlog_exit(0);
 539 }
 540
 541 /* We're allocating fs objects, use GFP_NOFS */
 542 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
 543 {
 544         struct ocfs2_inode_info *oi;
 545
 546         oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
 547         if (!oi)
 548                 return NULL;
 549
 550         jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
 551         return &oi->vfs_inode;
 552 }
 553
 554 static void ocfs2_destroy_inode(struct inode *inode)
 555 {
 556         kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
 557 }
 558
 559 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
 560                                                 unsigned int cbits)
 561 {
 562         unsigned int bytes = 1 << cbits;
 563         unsigned int trim = bytes;
 564         unsigned int bitshift = 32;
 565
 566         /*
 567          * i_size and all block offsets in ocfs2 are always 64 bits
 568          * wide. i_clusters is 32 bits, in cluster-sized units. So on
 569          * 64 bit platforms, cluster size will be the limiting factor.
 570          */
 571
 572 #if BITS_PER_LONG == 32
 573 # if defined(CONFIG_LBDAF)
 574         BUILD_BUG_ON(sizeof(sector_t) != 8);
 575         /*
 576          * We might be limited by page cache size.
 577          */
 578         if (bytes > PAGE_CACHE_SIZE) {
 579                 bytes = PAGE_CACHE_SIZE;
 580                 trim = 1;
 581                 /*
 582                  * Shift by 31 here so that we don't get larger than
 583                  * MAX_LFS_FILESIZE
 584                  */
 585                 bitshift = 31;
 586         }
 587 # else
 588         /*
 589          * We are limited by the size of sector_t. Use block size, as
 590          * that's what we expose to the VFS.
 591          */
 592         bytes = 1 << bbits;
 593         trim = 1;
 594         bitshift = 31;
 595 # endif
 596 #endif
 597
 598         /*
 599          * Trim by a whole cluster when we can actually approach the
 600          * on-disk limits. Otherwise we can overflow i_clusters when
 601          * an extent start is at the max offset.
 602          */
 603         return (((unsigned long long)bytes) << bitshift) - trim;
 604 }
 605
 606 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
 607 {
 608         int incompat_features;
 609         int ret = 0;
 610         struct mount_options parsed_options;
 611         struct ocfs2_super *osb = OCFS2_SB(sb);
 612
 613         lock_kernel();
 614
 615         if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
 616             !ocfs2_check_set_options(sb, &parsed_options)) {
 617                 ret = -EINVAL;
 618                 goto out;
 619         }
 620
 621         if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
 622             (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 623                 ret = -EINVAL;
 624                 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
 625                 goto out;
 626         }
 627
 628         if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
 629             (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
 630                 ret = -EINVAL;
 631                 mlog(ML_ERROR, "Cannot change data mode on remount\n");
 632                 goto out;
 633         }
 634
 635         /* Probably don't want this on remount; it might
 636          * mess with other nodes */
 637         if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
 638             (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
 639                 ret = -EINVAL;
 640                 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
 641                 goto out;
 642         }
 643
 644         /* We're going to/from readonly mode. */
 645         if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
 646                 /* Disable quota accounting before remounting RO */
 647                 if (*flags & MS_RDONLY) {
 648                         ret = ocfs2_susp_quotas(osb, 0);
 649                         if (ret < 0)
 650                                 goto out;
 651                 }
 652                 /* Lock here so the check of HARD_RO and the potential
 653                  * setting of SOFT_RO is atomic. */
 654                 spin_lock(&osb->osb_lock);
 655                 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
 656                         mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
 657                         ret = -EROFS;
 658                         goto unlock_osb;
 659                 }
 660
 661                 if (*flags & MS_RDONLY) {
 662                         mlog(0, "Going to ro mode.\n");
 663                         sb->s_flags |= MS_RDONLY;
 664                         osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 665                 } else {
 666                         mlog(0, "Making ro filesystem writeable.\n");
 667
 668                         if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
 669                                 mlog(ML_ERROR, "Cannot remount RDWR "
 670                                      "filesystem due to previous errors.\n");
 671                                 ret = -EROFS;
 672                                 goto unlock_osb;
 673                         }
 674                         incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
 675                         if (incompat_features) {
 676                                 mlog(ML_ERROR, "Cannot remount RDWR because "
 677                                      "of unsupported optional features "
 678                                      "(%x).\n", incompat_features);
 679                                 ret = -EINVAL;
 680                                 goto unlock_osb;
 681                         }
 682                         sb->s_flags &= ~MS_RDONLY;
 683                         osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
 684                 }
 685 unlock_osb:
 686                 spin_unlock(&osb->osb_lock);
 687                 /* Enable quota accounting after remounting RW */
 688                 if (!ret && !(*flags & MS_RDONLY)) {
 689                         if (sb_any_quota_suspended(sb))
 690                                 ret = ocfs2_susp_quotas(osb, 1);
 691                         else
 692                                 ret = ocfs2_enable_quotas(osb);
 693                         if (ret < 0) {
 694                                 /* Return back changes... */
 695                                 spin_lock(&osb->osb_lock);
 696                                 sb->s_flags |= MS_RDONLY;
 697                                 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
 698                                 spin_unlock(&osb->osb_lock);
 699                                 goto out;
 700                         }
 701                 }
 702         }
 703
 704         if (!ret) {
 705                 /* Only save off the new mount options in case of a successful
 706                  * remount. */
 707                 osb->s_mount_opt = parsed_options.mount_opt;
 708                 osb->s_atime_quantum = parsed_options.atime_quantum;
 709                 osb->preferred_slot = parsed_options.slot;
 710                 if (parsed_options.commit_interval)
 711                         osb->osb_commit_interval = parsed_options.commit_interval;
 712
 713                 if (!ocfs2_is_hard_readonly(osb))
 714                         ocfs2_set_journal_params(osb);
 715
 716                 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
 717                         ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
 718                                                         MS_POSIXACL : 0);
 719         }
 720 out:
 721         unlock_kernel();
 722         return ret;
 723 }
 724
 725 static int ocfs2_sb_probe(struct super_block *sb,
 726                           struct buffer_head **bh,
 727                           int *sector_size,
 728                           struct ocfs2_blockcheck_stats *stats)
 729 {
 730         int status, tmpstat;
 731         struct ocfs1_vol_disk_hdr *hdr;
 732         struct ocfs2_dinode *di;
 733         int blksize;
 734
 735         *bh = NULL;
 736
 737         /* may be > 512 */
 738         *sector_size = bdev_logical_block_size(sb->s_bdev);
 739         if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
 740                 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
 741                      *sector_size, OCFS2_MAX_BLOCKSIZE);
 742                 status = -EINVAL;
 743                 goto bail;
 744         }
 745
 746         /* Can this really happen? */
 747         if (*sector_size < OCFS2_MIN_BLOCKSIZE)
 748                 *sector_size = OCFS2_MIN_BLOCKSIZE;
 749
 750         /* check block zero for old format */
 751         status = ocfs2_get_sector(sb, bh, 0, *sector_size);
 752         if (status < 0) {
 753                 mlog_errno(status);
 754                 goto bail;
 755         }
 756         hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
 757         if (hdr->major_version == OCFS1_MAJOR_VERSION) {
 758                 mlog(ML_ERROR, "incompatible version: %u.%u\n",
 759                      hdr->major_version, hdr->minor_version);
 760                 status = -EINVAL;
 761         }
 762         if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
 763                    strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
 764                 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
 765                      hdr->signature);
 766                 status = -EINVAL;
 767         }
 768         brelse(*bh);
 769         *bh = NULL;
 770         if (status < 0) {
 771                 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
 772                      "upgraded before mounting with ocfs v2\n");
 773                 goto bail;
 774         }
 775
 776         /*
 777          * Now check at magic offset for 512, 1024, 2048, 4096
 778          * blocksizes.  4096 is the maximum blocksize because it is
 779          * the minimum clustersize.
 780          */
 781         status = -EINVAL;
 782         for (blksize = *sector_size;
 783              blksize <= OCFS2_MAX_BLOCKSIZE;
 784              blksize <<= 1) {
 785                 tmpstat = ocfs2_get_sector(sb, bh,
 786                                            OCFS2_SUPER_BLOCK_BLKNO,
 787                                            blksize);
 788                 if (tmpstat < 0) {
 789                         status = tmpstat;
 790                         mlog_errno(status);
 791                         break;
 792                 }
 793                 di = (struct ocfs2_dinode *) (*bh)->b_data;
 794                 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
 795                 spin_lock_init(&stats->b_lock);
 796                 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
 797                 if (tmpstat < 0) {
 798                         brelse(*bh);
 799                         *bh = NULL;
 800                 }
 801                 if (tmpstat != -EAGAIN) {
 802                         status = tmpstat;
 803                         break;
 804                 }
 805         }
 806
 807 bail:
 808         return status;
 809 }
 810
 811 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
 812 {
 813         if (ocfs2_mount_local(osb)) {
 814                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 815                         mlog(ML_ERROR, "Cannot heartbeat on a locally "
 816                              "mounted device.\n");
 817                         return -EINVAL;
 818                 }
 819         }
 820
 821         if (ocfs2_userspace_stack(osb)) {
 822                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
 823                         mlog(ML_ERROR, "Userspace stack expected, but "
 824                              "o2cb heartbeat arguments passed to mount\n");
 825                         return -EINVAL;
 826                 }
 827         }
 828
 829         if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
 830                 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
 831                     !ocfs2_userspace_stack(osb)) {
 832                         mlog(ML_ERROR, "Heartbeat has to be started to mount "
 833                              "a read-write clustered device.\n");
 834                         return -EINVAL;
 835                 }
 836         }
 837
 838         return 0;
 839 }
 840
 841 /*
 842  * If we're using a userspace stack, mount should have passed
 843  * a name that matches the disk.  If not, mount should not
 844  * have passed a stack.
 845  */
 846 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
 847                                         struct mount_options *mopt)
 848 {
 849         if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
 850                 mlog(ML_ERROR,
 851                      "cluster stack passed to mount, but this filesystem "
 852                      "does not support it\n");
 853                 return -EINVAL;
 854         }
 855
 856         if (ocfs2_userspace_stack(osb) &&
 857             strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
 858                     OCFS2_STACK_LABEL_LEN)) {
 859                 mlog(ML_ERROR,
 860                      "cluster stack passed to mount (\"%s\") does not "
 861                      "match the filesystem (\"%s\")\n",
 862                      mopt->cluster_stack,
 863                      osb->osb_cluster_stack);
 864                 return -EINVAL;
 865         }
 866
 867         return 0;
 868 }
 869
 870 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
 871 {
 872         int type;
 873         struct super_block *sb = osb->sb;
 874         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 875                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 876         int status = 0;
 877
 878         for (type = 0; type < MAXQUOTAS; type++) {
 879                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 880                         continue;
 881                 if (unsuspend)
 882                         status = vfs_quota_enable(
 883                                         sb_dqopt(sb)->files[type],
 884                                         type, QFMT_OCFS2,
 885                                         DQUOT_SUSPENDED);
 886                 else {
 887                         struct ocfs2_mem_dqinfo *oinfo;
 888
 889                         /* Cancel periodic syncing before suspending */
 890                         oinfo = sb_dqinfo(sb, type)->dqi_priv;
 891                         cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 892                         status = vfs_quota_disable(sb, type,
 893                                                    DQUOT_SUSPENDED);
 894                 }
 895                 if (status < 0)
 896                         break;
 897         }
 898         if (status < 0)
 899                 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
 900                      "remount (error = %d).\n", status);
 901         return status;
 902 }
 903
 904 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
 905 {
 906         struct inode *inode[MAXQUOTAS] = { NULL, NULL };
 907         struct super_block *sb = osb->sb;
 908         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 909                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 910         unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
 911                                         LOCAL_GROUP_QUOTA_SYSTEM_INODE };
 912         int status;
 913         int type;
 914
 915         sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
 916         for (type = 0; type < MAXQUOTAS; type++) {
 917                 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 918                         continue;
 919                 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
 920                                                         osb->slot_num);
 921                 if (!inode[type]) {
 922                         status = -ENOENT;
 923                         goto out_quota_off;
 924                 }
 925                 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
 926                                                 DQUOT_USAGE_ENABLED);
 927                 if (status < 0)
 928                         goto out_quota_off;
 929         }
 930
 931         for (type = 0; type < MAXQUOTAS; type++)
 932                 iput(inode[type]);
 933         return 0;
 934 out_quota_off:
 935         ocfs2_disable_quotas(osb);
 936         for (type = 0; type < MAXQUOTAS; type++)
 937                 iput(inode[type]);
 938         mlog_errno(status);
 939         return status;
 940 }
 941
 942 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
 943 {
 944         int type;
 945         struct inode *inode;
 946         struct super_block *sb = osb->sb;
 947         struct ocfs2_mem_dqinfo *oinfo;
 948
 949         /* We mostly ignore errors in this function because there's not much
 950          * we can do when we see them */
 951         for (type = 0; type < MAXQUOTAS; type++) {
 952                 if (!sb_has_quota_loaded(sb, type))
 953                         continue;
 954                 /* Cancel periodic syncing before we grab dqonoff_mutex */
 955                 oinfo = sb_dqinfo(sb, type)->dqi_priv;
 956                 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
 957                 inode = igrab(sb->s_dquot.files[type]);
 958                 /* Turn off quotas. This will remove all dquot structures from
 959                  * memory and so they will be automatically synced to global
 960                  * quota files */
 961                 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
 962                                             DQUOT_LIMITS_ENABLED);
 963                 if (!inode)
 964                         continue;
 965                 iput(inode);
 966         }
 967 }
 968
 969 /* Handle quota on quotactl */
 970 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
 971                           char *path, int remount)
 972 {
 973         unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
 974                                              OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
 975
 976         if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
 977                 return -EINVAL;
 978
 979         if (remount)
 980                 return 0;       /* Just ignore it has been handled in
 981                                  * ocfs2_remount() */
 982         return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
 983                                     format_id, DQUOT_LIMITS_ENABLED);
 984 }
 985
 986 /* Handle quota off quotactl */
 987 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
 988 {
 989         if (remount)
 990                 return 0;       /* Ignore now and handle later in
 991                                  * ocfs2_remount() */
 992         return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
 993 }
 994
 995 static const struct quotactl_ops ocfs2_quotactl_ops = {
 996         .quota_on       = ocfs2_quota_on,
 997         .quota_off      = ocfs2_quota_off,
 998         .quota_sync     = vfs_quota_sync,
 999         .get_info       = vfs_get_dqinfo,
1000         .set_info       = vfs_set_dqinfo,
1001         .get_dqblk      = vfs_get_dqblk,
1002         .set_dqblk      = vfs_set_dqblk,
1003 };
1004
1005 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1006 {
1007         struct dentry *root;
1008         int status, sector_size;
1009         struct mount_options parsed_options;
1010         struct inode *inode = NULL;
1011         struct ocfs2_super *osb = NULL;
1012         struct buffer_head *bh = NULL;
1013         char nodestr[8];
1014         struct ocfs2_blockcheck_stats stats;
1015
1016         mlog_entry("%p, %p, %i", sb, data, silent);
1017
1018         if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1019                 status = -EINVAL;
1020                 goto read_super_error;
1021         }
1022
1023         /* probe for superblock */
1024         status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1025         if (status < 0) {
1026                 mlog(ML_ERROR, "superblock probe failed!\n");
1027                 goto read_super_error;
1028         }
1029
1030         status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1031         osb = OCFS2_SB(sb);
1032         if (status < 0) {
1033                 mlog_errno(status);
1034                 goto read_super_error;
1035         }
1036         brelse(bh);
1037         bh = NULL;
1038
1039         if (!ocfs2_check_set_options(sb, &parsed_options)) {
1040                 status = -EINVAL;
1041                 goto read_super_error;
1042         }
1043         osb->s_mount_opt = parsed_options.mount_opt;
1044         osb->s_atime_quantum = parsed_options.atime_quantum;
1045         osb->preferred_slot = parsed_options.slot;
1046         osb->osb_commit_interval = parsed_options.commit_interval;
1047
1048         ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1049         osb->osb_resv_level = parsed_options.resv_level;
1050         osb->osb_dir_resv_level = parsed_options.resv_level;
1051         if (parsed_options.dir_resv_level == -1)
1052                 osb->osb_dir_resv_level = parsed_options.resv_level;
1053         else
1054                 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1055
1056         status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1057         if (status)
1058                 goto read_super_error;
1059
1060         sb->s_magic = OCFS2_SUPER_MAGIC;
1061
1062         sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1063                 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1064
1065         /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1066          * heartbeat=none */
1067         if (bdev_read_only(sb->s_bdev)) {
1068                 if (!(sb->s_flags & MS_RDONLY)) {
1069                         status = -EACCES;
1070                         mlog(ML_ERROR, "Readonly device detected but readonly "
1071                              "mount was not specified.\n");
1072                         goto read_super_error;
1073                 }
1074
1075                 /* You should not be able to start a local heartbeat
1076                  * on a readonly device. */
1077                 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1078                         status = -EROFS;
1079                         mlog(ML_ERROR, "Local heartbeat specified on readonly "
1080                              "device.\n");
1081                         goto read_super_error;
1082                 }
1083
1084                 status = ocfs2_check_journals_nolocks(osb);
1085                 if (status < 0) {
1086                         if (status == -EROFS)
1087                                 mlog(ML_ERROR, "Recovery required on readonly "
1088                                      "file system, but write access is "
1089                                      "unavailable.\n");
1090                         else
1091                                 mlog_errno(status);
1092                         goto read_super_error;
1093                 }
1094
1095                 ocfs2_set_ro_flag(osb, 1);
1096
1097                 printk(KERN_NOTICE "Readonly device detected. No cluster "
1098                        "services will be utilized for this mount. Recovery "
1099                        "will be skipped.\n");
1100         }
1101
1102         if (!ocfs2_is_hard_readonly(osb)) {
1103                 if (sb->s_flags & MS_RDONLY)
1104                         ocfs2_set_ro_flag(osb, 0);
1105         }
1106
1107         status = ocfs2_verify_heartbeat(osb);
1108         if (status < 0) {
1109                 mlog_errno(status);
1110                 goto read_super_error;
1111         }
1112
1113         osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1114                                                  ocfs2_debugfs_root);
1115         if (!osb->osb_debug_root) {
1116                 status = -EINVAL;
1117                 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1118                 goto read_super_error;
1119         }
1120
1121         osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1122                                             osb->osb_debug_root,
1123                                             osb,
1124                                             &ocfs2_osb_debug_fops);
1125         if (!osb->osb_ctxt) {
1126                 status = -EINVAL;
1127                 mlog_errno(status);
1128                 goto read_super_error;
1129         }
1130
1131         if (ocfs2_meta_ecc(osb)) {
1132                 status = ocfs2_blockcheck_stats_debugfs_install(
1133                                                 &osb->osb_ecc_stats,
1134                                                 osb->osb_debug_root);
1135                 if (status) {
1136                         mlog(ML_ERROR,
1137                              "Unable to create blockcheck statistics "
1138                              "files\n");
1139                         goto read_super_error;
1140                 }
1141         }
1142
1143         status = ocfs2_mount_volume(sb);
1144         if (osb->root_inode)
1145                 inode = igrab(osb->root_inode);
1146
1147         if (status < 0)
1148                 goto read_super_error;
1149
1150         if (!inode) {
1151                 status = -EIO;
1152                 mlog_errno(status);
1153                 goto read_super_error;
1154         }
1155
1156         root = d_alloc_root(inode);
1157         if (!root) {
1158                 status = -ENOMEM;
1159                 mlog_errno(status);
1160                 goto read_super_error;
1161         }
1162
1163         sb->s_root = root;
1164
1165         ocfs2_complete_mount_recovery(osb);
1166
1167         if (ocfs2_mount_local(osb))
1168                 snprintf(nodestr, sizeof(nodestr), "local");
1169         else
1170                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1171
1172         printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1173                "with %s data mode.\n",
1174                osb->dev_str, nodestr, osb->slot_num,
1175                osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1176                "ordered");
1177
1178         atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1179         wake_up(&osb->osb_mount_event);
1180
1181         /* Now we can initialize quotas because we can afford to wait
1182          * for cluster locks recovery now. That also means that truncation
1183          * log recovery can happen but that waits for proper quota setup */
1184         if (!(sb->s_flags & MS_RDONLY)) {
1185                 status = ocfs2_enable_quotas(osb);
1186                 if (status < 0) {
1187                         /* We have to err-out specially here because
1188                          * s_root is already set */
1189                         mlog_errno(status);
1190                         atomic_set(&osb->vol_state, VOLUME_DISABLED);
1191                         wake_up(&osb->osb_mount_event);
1192                         mlog_exit(status);
1193                         return status;
1194                 }
1195         }
1196
1197         ocfs2_complete_quota_recovery(osb);
1198
1199         /* Now we wake up again for processes waiting for quotas */
1200         atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1201         wake_up(&osb->osb_mount_event);
1202
1203         /* Start this when the mount is almost sure of being successful */
1204         ocfs2_orphan_scan_start(osb);
1205
1206         mlog_exit(status);
1207         return status;
1208
1209 read_super_error:
1210         brelse(bh);
1211
1212         if (inode)
1213                 iput(inode);
1214
1215         if (osb) {
1216                 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1217                 wake_up(&osb->osb_mount_event);
1218                 ocfs2_dismount_volume(sb, 1);
1219         }
1220
1221         mlog_exit(status);
1222         return status;
1223 }
1224
1225 static int ocfs2_get_sb(struct file_system_type *fs_type,
1226                         int flags,
1227                         const char *dev_name,
1228                         void *data,
1229                         struct vfsmount *mnt)
1230 {
1231         return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1232                            mnt);
1233 }
1234
1235 static void ocfs2_kill_sb(struct super_block *sb)
1236 {
1237         struct ocfs2_super *osb = OCFS2_SB(sb);
1238
1239         /* Failed mount? */
1240         if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1241                 goto out;
1242
1243         /* Prevent further queueing of inode drop events */
1244         spin_lock(&dentry_list_lock);
1245         ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1246         spin_unlock(&dentry_list_lock);
1247         /* Wait for work to finish and/or remove it */
1248         cancel_work_sync(&osb->dentry_lock_work);
1249 out:
1250         kill_block_super(sb);
1251 }
1252
1253 static struct file_system_type ocfs2_fs_type = {
1254         .owner          = THIS_MODULE,
1255         .name           = "ocfs2",
1256         .get_sb         = ocfs2_get_sb, /* is this called when we mount
1257                                         * the fs? */
1258         .kill_sb        = ocfs2_kill_sb,
1259
1260         .fs_flags       = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1261         .next           = NULL
1262 };
1263
1264 static int ocfs2_check_set_options(struct super_block *sb,
1265                                    struct mount_options *options)
1266 {
1267         if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1268             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1269                                          OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1270                 mlog(ML_ERROR, "User quotas were requested, but this "
1271                      "filesystem does not have the feature enabled.\n");
1272                 return 0;
1273         }
1274         if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1275             !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1276                                          OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1277                 mlog(ML_ERROR, "Group quotas were requested, but this "
1278                      "filesystem does not have the feature enabled.\n");
1279                 return 0;
1280         }
1281         if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1282             !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1283                 mlog(ML_ERROR, "ACL support requested but extended attributes "
1284                      "feature is not enabled\n");
1285                 return 0;
1286         }
1287         /* No ACL setting specified? Use XATTR feature... */
1288         if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1289                                     OCFS2_MOUNT_NO_POSIX_ACL))) {
1290                 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1291                         options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1292                 else
1293                         options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1294         }
1295         return 1;
1296 }
1297
1298 static int ocfs2_parse_options(struct super_block *sb,
1299                                char *options,
1300                                struct mount_options *mopt,
1301                                int is_remount)
1302 {
1303         int status;
1304         char *p;
1305
1306         mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1307                    options ? options : "(none)");
1308
1309         mopt->commit_interval = 0;
1310         mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1311         mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1312         mopt->slot = OCFS2_INVALID_SLOT;
1313         mopt->localalloc_opt = -1;
1314         mopt->cluster_stack[0] = '\0';
1315         mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1316         mopt->dir_resv_level = -1;
1317
1318         if (!options) {
1319                 status = 1;
1320                 goto bail;
1321         }
1322
1323         while ((p = strsep(&options, ",")) != NULL) {
1324                 int token, option;
1325                 substring_t args[MAX_OPT_ARGS];
1326
1327                 if (!*p)
1328                         continue;
1329
1330                 token = match_token(p, tokens, args);
1331                 switch (token) {
1332                 case Opt_hb_local:
1333                         mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1334                         break;
1335                 case Opt_hb_none:
1336                         mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1337                         break;
1338                 case Opt_barrier:
1339                         if (match_int(&args[0], &option)) {
1340                                 status = 0;
1341                                 goto bail;
1342                         }
1343                         if (option)
1344                                 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1345                         else
1346                                 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1347                         break;
1348                 case Opt_intr:
1349                         mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1350                         break;
1351                 case Opt_nointr:
1352                         mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1353                         break;
1354                 case Opt_err_panic:
1355                         mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1356                         break;
1357                 case Opt_err_ro:
1358                         mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1359                         break;
1360                 case Opt_data_ordered:
1361                         mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1362                         break;
1363                 case Opt_data_writeback:
1364                         mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1365                         break;
1366                 case Opt_user_xattr:
1367                         mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1368                         break;
1369                 case Opt_nouser_xattr:
1370                         mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1371                         break;
1372                 case Opt_atime_quantum:
1373                         if (match_int(&args[0], &option)) {
1374                                 status = 0;
1375                                 goto bail;
1376                         }
1377                         if (option >= 0)
1378                                 mopt->atime_quantum = option;
1379                         break;
1380                 case Opt_slot:
1381                         option = 0;
1382                         if (match_int(&args[0], &option)) {
1383                                 status = 0;
1384                                 goto bail;
1385                         }
1386                         if (option)
1387                                 mopt->slot = (s16)option;
1388                         break;
1389                 case Opt_commit:
1390                         option = 0;
1391                         if (match_int(&args[0], &option)) {
1392                                 status = 0;
1393                                 goto bail;
1394                         }
1395                         if (option < 0)
1396                                 return 0;
1397                         if (option == 0)
1398                                 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1399                         mopt->commit_interval = HZ * option;
1400                         break;
1401                 case Opt_localalloc:
1402                         option = 0;
1403                         if (match_int(&args[0], &option)) {
1404                                 status = 0;
1405                                 goto bail;
1406                         }
1407                         if (option >= 0)
1408                                 mopt->localalloc_opt = option;
1409                         break;
1410                 case Opt_localflocks:
1411                         /*
1412                          * Changing this during remount could race
1413                          * flock() requests, or "unbalance" existing
1414                          * ones (e.g., a lock is taken in one mode but
1415                          * dropped in the other). If users care enough
1416                          * to flip locking modes during remount, we
1417                          * could add a "local" flag to individual
1418                          * flock structures for proper tracking of
1419                          * state.
1420                          */
1421                         if (!is_remount)
1422                                 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1423                         break;
1424                 case Opt_stack:
1425                         /* Check both that the option we were passed
1426                          * is of the right length and that it is a proper
1427                          * string of the right length.
1428                          */
1429                         if (((args[0].to - args[0].from) !=
1430                              OCFS2_STACK_LABEL_LEN) ||
1431                             (strnlen(args[0].from,
1432                                      OCFS2_STACK_LABEL_LEN) !=
1433                              OCFS2_STACK_LABEL_LEN)) {
1434                                 mlog(ML_ERROR,
1435                                      "Invalid cluster_stack option\n");
1436                                 status = 0;
1437                                 goto bail;
1438                         }
1439                         memcpy(mopt->cluster_stack, args[0].from,
1440                                OCFS2_STACK_LABEL_LEN);
1441                         mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1442                         break;
1443                 case Opt_inode64:
1444                         mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1445                         break;
1446                 case Opt_usrquota:
1447                         mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1448                         break;
1449                 case Opt_grpquota:
1450                         mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1451                         break;
1452                 case Opt_acl:
1453                         mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1454                         mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1455                         break;
1456                 case Opt_noacl:
1457                         mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1458                         mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1459                         break;
1460                 case Opt_resv_level:
1461                         if (is_remount)
1462                                 break;
1463                         if (match_int(&args[0], &option)) {
1464                                 status = 0;
1465                                 goto bail;
1466                         }
1467                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1468                             option < OCFS2_MAX_RESV_LEVEL)
1469                                 mopt->resv_level = option;
1470                         break;
1471                 case Opt_dir_resv_level:
1472                         if (is_remount)
1473                                 break;
1474                         if (match_int(&args[0], &option)) {
1475                                 status = 0;
1476                                 goto bail;
1477                         }
1478                         if (option >= OCFS2_MIN_RESV_LEVEL &&
1479                             option < OCFS2_MAX_RESV_LEVEL)
1480                                 mopt->dir_resv_level = option;
1481                         break;
1482                 default:
1483                         mlog(ML_ERROR,
1484                              "Unrecognized mount option \"%s\" "
1485                              "or missing value\n", p);
1486                         status = 0;
1487                         goto bail;
1488                 }
1489         }
1490
1491         status = 1;
1492
1493 bail:
1494         mlog_exit(status);
1495         return status;
1496 }
1497
1498 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1499 {
1500         struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1501         unsigned long opts = osb->s_mount_opt;
1502         unsigned int local_alloc_megs;
1503
1504         if (opts & OCFS2_MOUNT_HB_LOCAL)
1505                 seq_printf(s, ",_netdev,heartbeat=local");
1506         else
1507                 seq_printf(s, ",heartbeat=none");
1508
1509         if (opts & OCFS2_MOUNT_NOINTR)
1510                 seq_printf(s, ",nointr");
1511
1512         if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1513                 seq_printf(s, ",data=writeback");
1514         else
1515                 seq_printf(s, ",data=ordered");
1516
1517         if (opts & OCFS2_MOUNT_BARRIER)
1518                 seq_printf(s, ",barrier=1");
1519
1520         if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1521                 seq_printf(s, ",errors=panic");
1522         else
1523                 seq_printf(s, ",errors=remount-ro");
1524
1525         if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1526                 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1527
1528         if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1529                 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1530
1531         if (osb->osb_commit_interval)
1532                 seq_printf(s, ",commit=%u",
1533                            (unsigned) (osb->osb_commit_interval / HZ));
1534
1535         local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1536         if (local_alloc_megs != ocfs2_la_default_mb(osb))
1537                 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1538
1539         if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1540                 seq_printf(s, ",localflocks,");
1541
1542         if (osb->osb_cluster_stack[0])
1543                 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1544                            osb->osb_cluster_stack);
1545         if (opts & OCFS2_MOUNT_USRQUOTA)
1546                 seq_printf(s, ",usrquota");
1547         if (opts & OCFS2_MOUNT_GRPQUOTA)
1548                 seq_printf(s, ",grpquota");
1549
1550         if (opts & OCFS2_MOUNT_NOUSERXATTR)
1551                 seq_printf(s, ",nouser_xattr");
1552         else
1553                 seq_printf(s, ",user_xattr");
1554
1555         if (opts & OCFS2_MOUNT_INODE64)
1556                 seq_printf(s, ",inode64");
1557
1558         if (opts & OCFS2_MOUNT_POSIX_ACL)
1559                 seq_printf(s, ",acl");
1560         else
1561                 seq_printf(s, ",noacl");
1562
1563         if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1564                 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1565
1566         if (osb->osb_dir_resv_level != osb->osb_resv_level)
1567                 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1568
1569         return 0;
1570 }
1571
1572 static int __init ocfs2_init(void)
1573 {
1574         int status;
1575
1576         mlog_entry_void();
1577
1578         ocfs2_print_version();
1579
1580         status = init_ocfs2_uptodate_cache();
1581         if (status < 0) {
1582                 mlog_errno(status);
1583                 goto leave;
1584         }
1585
1586         status = ocfs2_initialize_mem_caches();
1587         if (status < 0) {
1588                 mlog_errno(status);
1589                 goto leave;
1590         }
1591
1592         ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1593         if (!ocfs2_wq) {
1594                 status = -ENOMEM;
1595                 goto leave;
1596         }
1597
1598         ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1599         if (!ocfs2_debugfs_root) {
1600                 status = -EFAULT;
1601                 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1602         }
1603
1604         status = ocfs2_quota_setup();
1605         if (status)
1606                 goto leave;
1607
1608         ocfs2_set_locking_protocol();
1609
1610         status = register_quota_format(&ocfs2_quota_format);
1611 leave:
1612         if (status < 0) {
1613                 ocfs2_quota_shutdown();
1614                 ocfs2_free_mem_caches();
1615                 exit_ocfs2_uptodate_cache();
1616         }
1617
1618         mlog_exit(status);
1619
1620         if (status >= 0) {
1621                 return register_filesystem(&ocfs2_fs_type);
1622         } else
1623                 return -1;
1624 }
1625
1626 static void __exit ocfs2_exit(void)
1627 {
1628         mlog_entry_void();
1629
1630         ocfs2_quota_shutdown();
1631
1632         if (ocfs2_wq) {
1633                 flush_workqueue(ocfs2_wq);
1634                 destroy_workqueue(ocfs2_wq);
1635         }
1636
1637         unregister_quota_format(&ocfs2_quota_format);
1638
1639         debugfs_remove(ocfs2_debugfs_root);
1640
1641         ocfs2_free_mem_caches();
1642
1643         unregister_filesystem(&ocfs2_fs_type);
1644
1645         exit_ocfs2_uptodate_cache();
1646
1647         mlog_exit_void();
1648 }
1649
1650 static void ocfs2_put_super(struct super_block *sb)
1651 {
1652         mlog_entry("(0x%p)\n", sb);
1653
1654         lock_kernel();
1655
1656         ocfs2_sync_blockdev(sb);
1657         ocfs2_dismount_volume(sb, 0);
1658
1659         unlock_kernel();
1660
1661         mlog_exit_void();
1662 }
1663
1664 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1665 {
1666         struct ocfs2_super *osb;
1667         u32 numbits, freebits;
1668         int status;
1669         struct ocfs2_dinode *bm_lock;
1670         struct buffer_head *bh = NULL;
1671         struct inode *inode = NULL;
1672
1673         mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1674
1675         osb = OCFS2_SB(dentry->d_sb);
1676
1677         inode = ocfs2_get_system_file_inode(osb,
1678                                             GLOBAL_BITMAP_SYSTEM_INODE,
1679                                             OCFS2_INVALID_SLOT);
1680         if (!inode) {
1681                 mlog(ML_ERROR, "failed to get bitmap inode\n");
1682                 status = -EIO;
1683                 goto bail;
1684         }
1685
1686         status = ocfs2_inode_lock(inode, &bh, 0);
1687         if (status < 0) {
1688                 mlog_errno(status);
1689                 goto bail;
1690         }
1691
1692         bm_lock = (struct ocfs2_dinode *) bh->b_data;
1693
1694         numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1695         freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1696
1697         buf->f_type = OCFS2_SUPER_MAGIC;
1698         buf->f_bsize = dentry->d_sb->s_blocksize;
1699         buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1700         buf->f_blocks = ((sector_t) numbits) *
1701                         (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1702         buf->f_bfree = ((sector_t) freebits) *
1703                        (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1704         buf->f_bavail = buf->f_bfree;
1705         buf->f_files = numbits;
1706         buf->f_ffree = freebits;
1707         buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1708                                 & 0xFFFFFFFFUL;
1709         buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1710                                 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1711
1712         brelse(bh);
1713
1714         ocfs2_inode_unlock(inode, 0);
1715         status = 0;
1716 bail:
1717         if (inode)
1718                 iput(inode);
1719
1720         mlog_exit(status);
1721
1722         return status;
1723 }
1724
1725 static void ocfs2_inode_init_once(void *data)
1726 {
1727         struct ocfs2_inode_info *oi = data;
1728
1729         oi->ip_flags = 0;
1730         oi->ip_open_count = 0;
1731         spin_lock_init(&oi->ip_lock);
1732         ocfs2_extent_map_init(&oi->vfs_inode);
1733         INIT_LIST_HEAD(&oi->ip_io_markers);
1734         oi->ip_dir_start_lookup = 0;
1735
1736         init_rwsem(&oi->ip_alloc_sem);
1737         init_rwsem(&oi->ip_xattr_sem);
1738         mutex_init(&oi->ip_io_mutex);
1739
1740         oi->ip_blkno = 0ULL;
1741         oi->ip_clusters = 0;
1742
1743         ocfs2_resv_init_once(&oi->ip_la_data_resv);
1744
1745         ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1746         ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1747         ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1748
1749         ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1750                                   &ocfs2_inode_caching_ops);
1751
1752         inode_init_once(&oi->vfs_inode);
1753 }
1754
1755 static int ocfs2_initialize_mem_caches(void)
1756 {
1757         ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1758                                        sizeof(struct ocfs2_inode_info),
1759                                        0,
1760                                        (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1761                                                 SLAB_MEM_SPREAD),
1762                                        ocfs2_inode_init_once);
1763         ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1764                                         sizeof(struct ocfs2_dquot),
1765                                         0,
1766                                         (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1767                                                 SLAB_MEM_SPREAD),
1768                                         NULL);
1769         ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1770                                         sizeof(struct ocfs2_quota_chunk),
1771                                         0,
1772                                         (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1773                                         NULL);
1774         if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1775             !ocfs2_qf_chunk_cachep) {
1776                 if (ocfs2_inode_cachep)
1777                         kmem_cache_destroy(ocfs2_inode_cachep);
1778                 if (ocfs2_dquot_cachep)
1779                         kmem_cache_destroy(ocfs2_dquot_cachep);
1780                 if (ocfs2_qf_chunk_cachep)
1781                         kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1782                 return -ENOMEM;
1783         }
1784
1785         return 0;
1786 }
1787
1788 static void ocfs2_free_mem_caches(void)
1789 {
1790         if (ocfs2_inode_cachep)
1791                 kmem_cache_destroy(ocfs2_inode_cachep);
1792         ocfs2_inode_cachep = NULL;
1793
1794         if (ocfs2_dquot_cachep)
1795                 kmem_cache_destroy(ocfs2_dquot_cachep);
1796         ocfs2_dquot_cachep = NULL;
1797
1798         if (ocfs2_qf_chunk_cachep)
1799                 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1800         ocfs2_qf_chunk_cachep = NULL;
1801 }
1802
1803 static int ocfs2_get_sector(struct super_block *sb,
1804                             struct buffer_head **bh,
1805                             int block,
1806                             int sect_size)
1807 {
1808         if (!sb_set_blocksize(sb, sect_size)) {
1809                 mlog(ML_ERROR, "unable to set blocksize\n");
1810                 return -EIO;
1811         }
1812
1813         *bh = sb_getblk(sb, block);
1814         if (!*bh) {
1815                 mlog_errno(-EIO);
1816                 return -EIO;
1817         }
1818         lock_buffer(*bh);
1819         if (!buffer_dirty(*bh))
1820                 clear_buffer_uptodate(*bh);
1821         unlock_buffer(*bh);
1822         ll_rw_block(READ, 1, bh);
1823         wait_on_buffer(*bh);
1824         if (!buffer_uptodate(*bh)) {
1825                 mlog_errno(-EIO);
1826                 brelse(*bh);
1827                 *bh = NULL;
1828                 return -EIO;
1829         }
1830
1831         return 0;
1832 }
1833
1834 static int ocfs2_mount_volume(struct super_block *sb)
1835 {
1836         int status = 0;
1837         int unlock_super = 0;
1838         struct ocfs2_super *osb = OCFS2_SB(sb);
1839
1840         mlog_entry_void();
1841
1842         if (ocfs2_is_hard_readonly(osb))
1843                 goto leave;
1844
1845         status = ocfs2_dlm_init(osb);
1846         if (status < 0) {
1847                 mlog_errno(status);
1848                 goto leave;
1849         }
1850
1851         status = ocfs2_super_lock(osb, 1);
1852         if (status < 0) {
1853                 mlog_errno(status);
1854                 goto leave;
1855         }
1856         unlock_super = 1;
1857
1858         /* This will load up the node map and add ourselves to it. */
1859         status = ocfs2_find_slot(osb);
1860         if (status < 0) {
1861                 mlog_errno(status);
1862                 goto leave;
1863         }
1864
1865         /* load all node-local system inodes */
1866         status = ocfs2_init_local_system_inodes(osb);
1867         if (status < 0) {
1868                 mlog_errno(status);
1869                 goto leave;
1870         }
1871
1872         status = ocfs2_check_volume(osb);
1873         if (status < 0) {
1874                 mlog_errno(status);
1875                 goto leave;
1876         }
1877
1878         status = ocfs2_truncate_log_init(osb);
1879         if (status < 0)
1880                 mlog_errno(status);
1881
1882 leave:
1883         if (unlock_super)
1884                 ocfs2_super_unlock(osb, 1);
1885
1886         mlog_exit(status);
1887         return status;
1888 }
1889
1890 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1891 {
1892         int tmp, hangup_needed = 0;
1893         struct ocfs2_super *osb = NULL;
1894         char nodestr[8];
1895
1896         mlog_entry("(0x%p)\n", sb);
1897
1898         BUG_ON(!sb);
1899         osb = OCFS2_SB(sb);
1900         BUG_ON(!osb);
1901
1902         debugfs_remove(osb->osb_ctxt);
1903
1904         /*
1905          * Flush inode dropping work queue so that deletes are
1906          * performed while the filesystem is still working
1907          */
1908         ocfs2_drop_all_dl_inodes(osb);
1909
1910         /* Orphan scan should be stopped as early as possible */
1911         ocfs2_orphan_scan_stop(osb);
1912
1913         ocfs2_disable_quotas(osb);
1914
1915         ocfs2_shutdown_local_alloc(osb);
1916
1917         ocfs2_truncate_log_shutdown(osb);
1918
1919         /* This will disable recovery and flush any recovery work. */
1920         ocfs2_recovery_exit(osb);
1921
1922         ocfs2_journal_shutdown(osb);
1923
1924         ocfs2_sync_blockdev(sb);
1925
1926         ocfs2_purge_refcount_trees(osb);
1927
1928         /* No cluster connection means we've failed during mount, so skip
1929          * all the steps which depended on that to complete. */
1930         if (osb->cconn) {
1931                 tmp = ocfs2_super_lock(osb, 1);
1932                 if (tmp < 0) {
1933                         mlog_errno(tmp);
1934                         return;
1935                 }
1936         }
1937
1938         if (osb->slot_num != OCFS2_INVALID_SLOT)
1939                 ocfs2_put_slot(osb);
1940
1941         if (osb->cconn)
1942                 ocfs2_super_unlock(osb, 1);
1943
1944         ocfs2_release_system_inodes(osb);
1945
1946         /*
1947          * If we're dismounting due to mount error, mount.ocfs2 will clean
1948          * up heartbeat.  If we're a local mount, there is no heartbeat.
1949          * If we failed before we got a uuid_str yet, we can't stop
1950          * heartbeat.  Otherwise, do it.
1951          */
1952         if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1953                 hangup_needed = 1;
1954
1955         if (osb->cconn)
1956                 ocfs2_dlm_shutdown(osb, hangup_needed);
1957
1958         ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1959         debugfs_remove(osb->osb_debug_root);
1960
1961         if (hangup_needed)
1962                 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1963
1964         atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1965
1966         if (ocfs2_mount_local(osb))
1967                 snprintf(nodestr, sizeof(nodestr), "local");
1968         else
1969                 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1970
1971         printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1972                osb->dev_str, nodestr);
1973
1974         ocfs2_delete_osb(osb);
1975         kfree(osb);
1976         sb->s_dev = 0;
1977         sb->s_fs_info = NULL;
1978 }
1979
1980 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1981                                 unsigned uuid_bytes)
1982 {
1983         int i, ret;
1984         char *ptr;
1985
1986         BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1987
1988         osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1989         if (osb->uuid_str == NULL)
1990                 return -ENOMEM;
1991
1992         for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1993                 /* print with null */
1994                 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1995                 if (ret != 2) /* drop super cleans up */
1996                         return -EINVAL;
1997                 /* then only advance past the last char */
1998                 ptr += 2;
1999         }
2000
2001         return 0;
2002 }
2003
2004 static int ocfs2_initialize_super(struct super_block *sb,
2005                                   struct buffer_head *bh,
2006                                   int sector_size,
2007                                   struct ocfs2_blockcheck_stats *stats)
2008 {
2009         int status;
2010         int i, cbits, bbits;
2011         struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2012         struct inode *inode = NULL;
2013         struct ocfs2_journal *journal;
2014         __le32 uuid_net_key;
2015         struct ocfs2_super *osb;
2016
2017         mlog_entry_void();
2018
2019         osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2020         if (!osb) {
2021                 status = -ENOMEM;
2022                 mlog_errno(status);
2023                 goto bail;
2024         }
2025
2026         sb->s_fs_info = osb;
2027         sb->s_op = &ocfs2_sops;
2028         sb->s_export_op = &ocfs2_export_ops;
2029         sb->s_qcop = &ocfs2_quotactl_ops;
2030         sb->dq_op = &ocfs2_quota_operations;
2031         sb->s_xattr = ocfs2_xattr_handlers;
2032         sb->s_time_gran = 1;
2033         sb->s_flags |= MS_NOATIME;
2034         /* this is needed to support O_LARGEFILE */
2035         cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2036         bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2037         sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2038
2039         osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2040
2041         for (i = 0; i < 3; i++)
2042                 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2043         osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2044
2045         osb->sb = sb;
2046         /* Save off for ocfs2_rw_direct */
2047         osb->s_sectsize_bits = blksize_bits(sector_size);
2048         BUG_ON(!osb->s_sectsize_bits);
2049
2050         spin_lock_init(&osb->dc_task_lock);
2051         init_waitqueue_head(&osb->dc_event);
2052         osb->dc_work_sequence = 0;
2053         osb->dc_wake_sequence = 0;
2054         INIT_LIST_HEAD(&osb->blocked_lock_list);
2055         osb->blocked_lock_count = 0;
2056         spin_lock_init(&osb->osb_lock);
2057         spin_lock_init(&osb->osb_xattr_lock);
2058         ocfs2_init_steal_slots(osb);
2059
2060         atomic_set(&osb->alloc_stats.moves, 0);
2061         atomic_set(&osb->alloc_stats.local_data, 0);
2062         atomic_set(&osb->alloc_stats.bitmap_data, 0);
2063         atomic_set(&osb->alloc_stats.bg_allocs, 0);
2064         atomic_set(&osb->alloc_stats.bg_extends, 0);
2065
2066         /* Copy the blockcheck stats from the superblock probe */
2067         osb->osb_ecc_stats = *stats;
2068
2069         ocfs2_init_node_maps(osb);
2070
2071         snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2072                  MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2073
2074         ocfs2_orphan_scan_init(osb);
2075
2076         status = ocfs2_recovery_init(osb);
2077         if (status) {
2078                 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2079                 mlog_errno(status);
2080                 goto bail;
2081         }
2082
2083         init_waitqueue_head(&osb->checkpoint_event);
2084         atomic_set(&osb->needs_checkpoint, 0);
2085
2086         osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2087
2088         osb->slot_num = OCFS2_INVALID_SLOT;
2089
2090         osb->s_xattr_inline_size = le16_to_cpu(
2091                                         di->id2.i_super.s_xattr_inline_size);
2092
2093         osb->local_alloc_state = OCFS2_LA_UNUSED;
2094         osb->local_alloc_bh = NULL;
2095         INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2096
2097         init_waitqueue_head(&osb->osb_mount_event);
2098
2099         status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2100         if (status) {
2101                 mlog_errno(status);
2102                 goto bail;
2103         }
2104
2105         osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2106         if (!osb->vol_label) {
2107                 mlog(ML_ERROR, "unable to alloc vol label\n");
2108                 status = -ENOMEM;
2109                 goto bail;
2110         }
2111
2112         osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2113         if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2114                 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2115                      osb->max_slots);
2116                 status = -EINVAL;
2117                 goto bail;
2118         }
2119         mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2120
2121         osb->slot_recovery_generations =
2122                 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2123                         GFP_KERNEL);
2124         if (!osb->slot_recovery_generations) {
2125                 status = -ENOMEM;
2126                 mlog_errno(status);
2127                 goto bail;
2128         }
2129
2130         init_waitqueue_head(&osb->osb_wipe_event);
2131         osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2132                                         sizeof(*osb->osb_orphan_wipes),
2133                                         GFP_KERNEL);
2134         if (!osb->osb_orphan_wipes) {
2135                 status = -ENOMEM;
2136                 mlog_errno(status);
2137                 goto bail;
2138         }
2139
2140         osb->osb_rf_lock_tree = RB_ROOT;
2141
2142         osb->s_feature_compat =
2143                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2144         osb->s_feature_ro_compat =
2145                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2146         osb->s_feature_incompat =
2147                 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2148
2149         if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2150                 mlog(ML_ERROR, "couldn't mount because of unsupported "
2151                      "optional features (%x).\n", i);
2152                 status = -EINVAL;
2153                 goto bail;
2154         }
2155         if (!(osb->sb->s_flags & MS_RDONLY) &&
2156             (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2157                 mlog(ML_ERROR, "couldn't mount RDWR because of "
2158                      "unsupported optional features (%x).\n", i);
2159                 status = -EINVAL;
2160                 goto bail;
2161         }
2162
2163         if (ocfs2_userspace_stack(osb)) {
2164                 memcpy(osb->osb_cluster_stack,
2165                        OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2166                        OCFS2_STACK_LABEL_LEN);
2167                 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2168                 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2169                         mlog(ML_ERROR,
2170                              "couldn't mount because of an invalid "
2171                              "cluster stack label (%s) \n",
2172                              osb->osb_cluster_stack);
2173                         status = -EINVAL;
2174                         goto bail;
2175                 }
2176         } else {
2177                 /* The empty string is identical with classic tools that
2178                  * don't know about s_cluster_info. */
2179                 osb->osb_cluster_stack[0] = '\0';
2180         }
2181
2182         get_random_bytes(&osb->s_next_generation, sizeof(u32));
2183
2184         /* FIXME
2185          * This should be done in ocfs2_journal_init(), but unknown
2186          * ordering issues will cause the filesystem to crash.
2187          * If anyone wants to figure out what part of the code
2188          * refers to osb->journal before ocfs2_journal_init() is run,
2189          * be my guest.
2190          */
2191         /* initialize our journal structure */
2192
2193         journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2194         if (!journal) {
2195                 mlog(ML_ERROR, "unable to alloc journal\n");
2196                 status = -ENOMEM;
2197                 goto bail;
2198         }
2199         osb->journal = journal;
2200         journal->j_osb = osb;
2201
2202         atomic_set(&journal->j_num_trans, 0);
2203         init_rwsem(&journal->j_trans_barrier);
2204         init_waitqueue_head(&journal->j_checkpointed);
2205         spin_lock_init(&journal->j_lock);
2206         journal->j_trans_id = (unsigned long) 1;
2207         INIT_LIST_HEAD(&journal->j_la_cleanups);
2208         INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2209         journal->j_state = OCFS2_JOURNAL_FREE;
2210
2211         INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2212         osb->dentry_lock_list = NULL;
2213
2214         /* get some pseudo constants for clustersize bits */
2215         osb->s_clustersize_bits =
2216                 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2217         osb->s_clustersize = 1 << osb->s_clustersize_bits;
2218         mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2219
2220         if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2221             osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2222                 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2223                      osb->s_clustersize);
2224                 status = -EINVAL;
2225                 goto bail;
2226         }
2227
2228         if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2229             > (u32)~0UL) {
2230                 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2231                      "what jbd can address in 32 bits.\n");
2232                 status = -EINVAL;
2233                 goto bail;
2234         }
2235
2236         if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2237                                  sizeof(di->id2.i_super.s_uuid))) {
2238                 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2239                 status = -ENOMEM;
2240                 goto bail;
2241         }
2242
2243         memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2244
2245         strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2246         osb->vol_label[63] = '\0';
2247         osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2248         osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2249         osb->first_cluster_group_blkno =
2250                 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2251         osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2252         osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2253         mlog(0, "vol_label: %s\n", osb->vol_label);
2254         mlog(0, "uuid: %s\n", osb->uuid_str);
2255         mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2256              (unsigned long long)osb->root_blkno,
2257              (unsigned long long)osb->system_dir_blkno);
2258
2259         osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2260         if (!osb->osb_dlm_debug) {
2261                 status = -ENOMEM;
2262                 mlog_errno(status);
2263                 goto bail;
2264         }
2265
2266         atomic_set(&osb->vol_state, VOLUME_INIT);
2267
2268         /* load root, system_dir, and all global system inodes */
2269         status = ocfs2_init_global_system_inodes(osb);
2270         if (status < 0) {
2271                 mlog_errno(status);
2272                 goto bail;
2273         }
2274
2275         /*
2276          * global bitmap
2277          */
2278         inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2279                                             OCFS2_INVALID_SLOT);
2280         if (!inode) {
2281                 status = -EINVAL;
2282                 mlog_errno(status);
2283                 goto bail;
2284         }
2285
2286         osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2287         osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2288         iput(inode);
2289
2290         osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2291                                  osb->s_feature_incompat) * 8;
2292
2293         status = ocfs2_init_slot_info(osb);
2294         if (status < 0) {
2295                 mlog_errno(status);
2296                 goto bail;
2297         }
2298
2299 bail:
2300         mlog_exit(status);
2301         return status;
2302 }
2303
2304 /*
2305  * will return: -EAGAIN if it is ok to keep searching for superblocks
2306  *              -EINVAL if there is a bad superblock
2307  *              0 on success
2308  */
2309 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2310                                struct buffer_head *bh,
2311                                u32 blksz,
2312                                struct ocfs2_blockcheck_stats *stats)
2313 {
2314         int status = -EAGAIN;
2315
2316         mlog_entry_void();
2317
2318         if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2319                    strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2320                 /* We have to do a raw check of the feature here */
2321                 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2322                     OCFS2_FEATURE_INCOMPAT_META_ECC) {
2323                         status = ocfs2_block_check_validate(bh->b_data,
2324                                                             bh->b_size,
2325                                                             &di->i_check,
2326                                                             stats);
2327                         if (status)
2328                                 goto out;
2329                 }
2330                 status = -EINVAL;
2331                 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2332                         mlog(ML_ERROR, "found superblock with incorrect block "
2333                              "size: found %u, should be %u\n",
2334                              1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2335                                blksz);
2336                 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2337                            OCFS2_MAJOR_REV_LEVEL ||
2338                            le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2339                            OCFS2_MINOR_REV_LEVEL) {
2340                         mlog(ML_ERROR, "found superblock with bad version: "
2341                              "found %u.%u, should be %u.%u\n",
2342                              le16_to_cpu(di->id2.i_super.s_major_rev_level),
2343                              le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2344                              OCFS2_MAJOR_REV_LEVEL,
2345                              OCFS2_MINOR_REV_LEVEL);
2346                 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2347                         mlog(ML_ERROR, "bad block number on superblock: "
2348                              "found %llu, should be %llu\n",
2349                              (unsigned long long)le64_to_cpu(di->i_blkno),
2350                              (unsigned long long)bh->b_blocknr);
2351                 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2352                             le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2353                         mlog(ML_ERROR, "bad cluster size found: %u\n",
2354                              1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2355                 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2356                         mlog(ML_ERROR, "bad root_blkno: 0\n");
2357                 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2358                         mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2359                 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2360                         mlog(ML_ERROR,
2361                              "Superblock slots found greater than file system "
2362                              "maximum: found %u, max %u\n",
2363                              le16_to_cpu(di->id2.i_super.s_max_slots),
2364                              OCFS2_MAX_SLOTS);
2365                 } else {
2366                         /* found it! */
2367                         status = 0;
2368                 }
2369         }
2370
2371 out:
2372         mlog_exit(status);
2373         return status;
2374 }
2375
2376 static int ocfs2_check_volume(struct ocfs2_super *osb)
2377 {
2378         int status;
2379         int dirty;
2380         int local;
2381         struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2382                                                   * recover
2383                                                   * ourselves. */
2384
2385         mlog_entry_void();
2386
2387         /* Init our journal object. */
2388         status = ocfs2_journal_init(osb->journal, &dirty);
2389         if (status < 0) {
2390                 mlog(ML_ERROR, "Could not initialize journal!\n");
2391                 goto finally;
2392         }
2393
2394         /* If the journal was unmounted cleanly then we don't want to
2395          * recover anything. Otherwise, journal_load will do that
2396          * dirty work for us :) */
2397         if (!dirty) {
2398                 status = ocfs2_journal_wipe(osb->journal, 0);
2399                 if (status < 0) {
2400                         mlog_errno(status);
2401                         goto finally;
2402                 }
2403         } else {
2404                 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2405                      "recovering volume.\n");
2406         }
2407
2408         local = ocfs2_mount_local(osb);
2409
2410         /* will play back anything left in the journal. */
2411         status = ocfs2_journal_load(osb->journal, local, dirty);
2412         if (status < 0) {
2413                 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2414                 goto finally;
2415         }
2416
2417         if (dirty) {
2418                 /* recover my local alloc if we didn't unmount cleanly. */
2419                 status = ocfs2_begin_local_alloc_recovery(osb,
2420                                                           osb->slot_num,
2421                                                           &local_alloc);
2422                 if (status < 0) {
2423                         mlog_errno(status);
2424                         goto finally;
2425                 }
2426                 /* we complete the recovery process after we've marked
2427                  * ourselves as mounted. */
2428         }
2429
2430         mlog(0, "Journal loaded.\n");
2431
2432         status = ocfs2_load_local_alloc(osb);
2433         if (status < 0) {
2434                 mlog_errno(status);
2435                 goto finally;
2436         }
2437
2438         if (dirty) {
2439                 /* Recovery will be completed after we've mounted the
2440                  * rest of the volume. */
2441                 osb->dirty = 1;
2442                 osb->local_alloc_copy = local_alloc;
2443                 local_alloc = NULL;
2444         }
2445
2446         /* go through each journal, trylock it and if you get the
2447          * lock, and it's marked as dirty, set the bit in the recover
2448          * map and launch a recovery thread for it. */
2449         status = ocfs2_mark_dead_nodes(osb);
2450         if (status < 0) {
2451                 mlog_errno(status);
2452                 goto finally;
2453         }
2454
2455         status = ocfs2_compute_replay_slots(osb);
2456         if (status < 0)
2457                 mlog_errno(status);
2458
2459 finally:
2460         if (local_alloc)
2461                 kfree(local_alloc);
2462
2463         mlog_exit(status);
2464         return status;
2465 }
2466
2467 /*
2468  * The routine gets called from dismount or close whenever a dismount on
2469  * volume is requested and the osb open count becomes 1.
2470  * It will remove the osb from the global list and also free up all the
2471  * initialized resources and fileobject.
2472  */
2473 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2474 {
2475         mlog_entry_void();
2476
2477         /* This function assumes that the caller has the main osb resource */
2478
2479         ocfs2_free_slot_info(osb);
2480
2481         kfree(osb->osb_orphan_wipes);
2482         kfree(osb->slot_recovery_generations);
2483         /* FIXME
2484          * This belongs in journal shutdown, but because we have to
2485          * allocate osb->journal at the start of ocfs2_initalize_osb(),
2486          * we free it here.
2487          */
2488         kfree(osb->journal);
2489         if (osb->local_alloc_copy)
2490                 kfree(osb->local_alloc_copy);
2491         kfree(osb->uuid_str);
2492         ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2493         memset(osb, 0, sizeof(struct ocfs2_super));
2494
2495         mlog_exit_void();
2496 }
2497
2498 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2499  * panic(). We do not support continue-on-error operation. */
2500 static void ocfs2_handle_error(struct super_block *sb)
2501 {
2502         struct ocfs2_super *osb = OCFS2_SB(sb);
2503
2504         if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2505                 panic("OCFS2: (device %s): panic forced after error\n",
2506                       sb->s_id);
2507
2508         ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2509
2510         if (sb->s_flags & MS_RDONLY &&
2511             (ocfs2_is_soft_readonly(osb) ||
2512              ocfs2_is_hard_readonly(osb)))
2513                 return;
2514
2515         printk(KERN_CRIT "File system is now read-only due to the potential "
2516                "of on-disk corruption. Please run fsck.ocfs2 once the file "
2517                "system is unmounted.\n");
2518         sb->s_flags |= MS_RDONLY;
2519         ocfs2_set_ro_flag(osb, 0);
2520 }
2521
2522 static char error_buf[1024];
2523
2524 void __ocfs2_error(struct super_block *sb,
2525                    const char *function,
2526                    const char *fmt, ...)
2527 {
2528         va_list args;
2529
2530         va_start(args, fmt);
2531         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2532         va_end(args);
2533
2534         /* Not using mlog here because we want to show the actual
2535          * function the error came from. */
2536         printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2537                sb->s_id, function, error_buf);
2538
2539         ocfs2_handle_error(sb);
2540 }
2541
2542 /* Handle critical errors. This is intentionally more drastic than
2543  * ocfs2_handle_error, so we only use for things like journal errors,
2544  * etc. */
2545 void __ocfs2_abort(struct super_block* sb,
2546                    const char *function,
2547                    const char *fmt, ...)
2548 {
2549         va_list args;
2550
2551         va_start(args, fmt);
2552         vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2553         va_end(args);
2554
2555         printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2556                sb->s_id, function, error_buf);
2557
2558         /* We don't have the cluster support yet to go straight to
2559          * hard readonly in here. Until then, we want to keep
2560          * ocfs2_abort() so that we can at least mark critical
2561          * errors.
2562          *
2563          * TODO: This should abort the journal and alert other nodes
2564          * that our slot needs recovery. */
2565
2566         /* Force a panic(). This stinks, but it's better than letting
2567          * things continue without having a proper hard readonly
2568          * here. */
2569         if (!ocfs2_mount_local(OCFS2_SB(sb)))
2570                 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2571         ocfs2_handle_error(sb);
2572 }
2573
2574 /*
2575  * Void signal blockers, because in-kernel sigprocmask() only fails
2576  * when SIG_* is wrong.
2577  */
2578 void ocfs2_block_signals(sigset_t *oldset)
2579 {
2580         int rc;
2581         sigset_t blocked;
2582
2583         sigfillset(&blocked);
2584         rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2585         BUG_ON(rc);
2586 }
2587
2588 void ocfs2_unblock_signals(sigset_t *oldset)
2589 {
2590         int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2591         BUG_ON(rc);
2592 }
2593
2594 module_init(ocfs2_init);
2595 module_exit(ocfs2_exit);